Sasakian Geometry

Charles P. Boyer and Krzysztof Galicki

Department of Mathematics and Statistics, University of New Mex-
ico, Albuquerque, N.M. 87131
Preface
This book is a result of a fteen year long collaboration which has led us to
understand and appreciate the importance of Sasakian manifolds as an integral
part of Riemannian Geometry. In the early nineties neither of us was aware or
particularly interested in Sasakian structures. This rapidly changed in 1992 when,
together with Ben Mann, we realized that smooth 3-Sasakian manifolds, which are
automatically Einstein, are orbibundles over positive quaternionic Kahler orbifolds.
In the smooth case this bundle was rst described by Konishi. As quaternionic
Kahler orbifolds are easily manufactured via quaternionic reduction, likewise we
were able to construct large families of smooth compact 3-Sasakian spaces with
relative ease. Searching through the literature afterwards we were surprised that a
folklore conjecture attributed to Tanno stated that any 3-Sasakian manifold should
be a spherical space form.
Soon after we began to understand that the 3-Sasakian geometry, though very
interesting, was too specialized a case. To a large extent our main motivation
had to do with Einstein metrics. We realized that the 3-Sasakian manifolds were
merely a special case of the more general theory of Sasaki-Einstein manifolds and
shifted our focus to that more general case. With time it became clearer and clearer
that Sasakian geometry is one of the richest sources of complete Einstein metrics of
positive scalar curvature. We began to appreciate that Sasakian geometry, naturally
sandwiched between two dierent Kahler geometries, was at least as interesting
and important as the latter.
It was Nigel Hitchin who rst suggested to us that perhaps it is time to write
a modern book on the subject. We started thinking about the project in 2002. It
was a challenging endeavor: we aimed at writing a monograph which should give a
fairly complete account of our own contributions to the subject combined with an
advanced graduate level textbook describing the foundational material in a modern
language. Unlike the Kahler case, there are very few books of this sort discussing
Sasakian manifolds.
Another diculty which we were to discover later was the rapid development of
the subject. Some new results, such as the construction of Sasaki-Einstein metrics
on exotic spheres with Janos Kollar, had to do with the dynamics of our own
research program. But many other important results were obtained by others.
To complicate things even more, Sasaki-Einstein manifolds appear to play a very
special role in the so-called AdS/CFT duality conjecture and they have received an
enormous amount of attention among physicists working in this area. Some of the
intriguing new results which we decided to include in the book, albeit very briey,
indeed came quite unexpectedly from considerations in Superstring Theory.
In the end we know our book will not be as complete as we would have wanted
or hoped it to be. We nally had to stop writing while knowing all too well that the
iii
iv PREFACE
book is neither complete nor perfect. Few books of this sort are. We are pleased to
see so much renewed interest in the subject and hope that our work will be of help
to a number of mathematicians and physicists, researchers and graduate students
alike.
Over the years we have beneted immensely from many discussions with many
colleagues and collaborators about the mathematics contained in this book. Here
we are happy to take this opportunity to thank: B. Acharya, I. Agricola, D. Alek-
seevsky, V. Apostolov, F. Battaglia, C. Bar, H. Baum, F. Belgun, L. Berard Bergery,
R. Bielawski, C. Bohm, A. Buium, D. Blair, O. Biquard, J.-P. Bourguignon, R.
Bryant, D. Calderbank, A.

Cap, J. Cheeger, T. Colding, A. Dancer, O. Dearricott,
T. Draghici, Y. Eliashberg, J. Figueroa OFarrill, M. Fernandez, T. Friedrich, E.
Gasparim, P. Gauduchon, H. Geiges, W. Goldman, G. Grantcharov, K. Grove, M.
Harada, T. Hausel, G. Hernandez, L. Hernandez, R. Herrera, O. Hijazi, N. Hitchin,
H. Hofer, T. Holm, J. Hurtubise, J. Isenberg, G. Jensen, J. Johnson, A. de Jong,
D. Joyce, L. Katzarkov, J. Konderak, M. Kontsevich, J. Kollar, H. B. Lawson,
C. LeBrun, B. Mann, P. Matzeu, S. Marchiafava, R. Mazzeo, J. Milgram, A. Mo-
roianu, P.-A. Nagy, M. Nakamaye, T. Nitta, P. Nurowski, L. Ornea, H. Pedersen,
P. Piccinni, M. Pilca, Y.-S. Poon, E. Rees, P. Rukimbira, S. Salamon, R. Schoen,
L. Schwachhofer, U. Semmelmann, S. Simanca, M. Singer, J. Sparks, J. Starr, S.
Stolz, A. Swann, Ch. Thomas, G. Tian, M. Verbitsky, M. Wang, J. Wisniewski, R.
Wolak, D. Wraith, S.-T. Yau, D. Zagier, and W. Ziller. It is probably inevitable
that we have missed the names of some friends and colleagues to whom we deeply
apologize.
We owe our deep gratitude to Janos Kollar. He is not just a collaborator on
one of the more important papers we wrote. His continuous help and involvement
throughout writing of this book was invaluable. Much of the material presented
in chapters 10 and 11 was written with his expert help and advice. We also wish
to thank Ilka Agricola, Elizabeth Gasparim, Eugene Lerman, Michael Nakamaye,
and Santiago Simanca and Thomas Friedrich for carefully reading certain parts
of our book and providing invaluable comments and corrections. We thank Evan
Thomas for helping with computations used to compile the various tables appearing
in the appendices. We also thank our graduate students: J. Cuadros, R. Gomez,
D. Grandini, J. Kania, and R. Sanchez-Silva for weeding out various mistakes while
sitting in several courses we taught at UNM using early versions of the book.
The second author would like to express special thanks to the Max-Planck-
Institut f ur Mathematik in Bonn for the generous support and hospitality. Several
chapters of this book were written during K.G.s sabbatical visit at the MPIM
during the calendar year 2004 and also later during two shorter visits in 2005 and
2006. Both of us thank the National Science Foundation for continuous support of
our many projects, including this one.
Finally, we thank Jessica Churchman and Alison Jones from the Oxford Uni-
versity Press for their patience, continuing interest in our work on the project and
much help in the latter stages of preparing the manuscript for publication.
Last but not least we would like to thank our families, especially Margaret and
Rowan, for support and patience and for putting up with us while we were working
on The Book day and night.
Albuquerque, February 2007.
Contents
Preface iii
Introduction 1
Chapter 1. Structures on Manifolds 9
1.1. Sheaves and Sheaf Cohomology 9
1.2. Principal and Associated Bundles 14
1.3. Connections in Principal and Associated Vector Bundles 18
1.4. G-Structures 23
1.5. Pseudogroup Structures 36
1.6. Group Actions on Manifolds 38
Chapter 2. Foliations 51
2.1. Examples of Foliations 51
2.2. Haeiger Structures 52
2.3. Leaf Holonomy and the Holonomy Groupoid 54
2.4. Basic Cohomology 59
2.5. Transverse Geometry 60
2.6. Riemannian Flows 69
Chapter 3. Kahler Manifolds 75
3.1. Complex Manifolds and Kahler Metrics 76
3.2. Curvature of Kahler Manifolds 82
3.3. Hodge Theory on Kahler Manifolds 88
3.4. Complex Vector Bundles and Chern Classes 92
3.5. Line Bundles and Divisors 94
3.6. The Calabi Conjecture and the Calabi-Yau Theorem 102
Chapter 4. Fundamentals of Orbifolds 105
4.1. Basic Denitions 105
4.2. Orbisheaves and orbibundles 108
4.3. Groupoids, Orbifold Invariants and Classifying Spaces 115
4.4. Complex Orbifolds 123
4.5. Weighted Projective Spaces 133
4.6. Hypersurfaces in Weighted Projective Spaces 138
4.7. Seifert Bundles 144
Chapter 5. Kahler-Einstein Metrics 151
5.1. Some Elementary Considerations 152
5.2. The Monge-Amp`ere Problem and the Continuity Method 153
5.3. Obstructions in the Positive Case 160
v
vi CONTENTS
5.4. Kahler-Einstein Metrics on Hypersurfaces in CP(w) 162
5.5. Automorphisms and the Moduli Problem 173
Chapter 6. Almost Contact and Contact Geometry 179
6.1. Contact Structures 180
6.2. Almost Contact Structures 190
6.3. Almost Contact Metric Structures 195
6.4. Contact Metric Structures 198
6.5. Structures on Cones 201
Chapter 7. K-Contact and Sasakian Structures 207
7.1. Quasi-regularity and the Structure Theorems 207
7.2. The Transverse Geometry of the Characteristic Foliation 214
7.3. Curvature Properties of K-Contact and Sasakian Structures 219
7.4. Topology of K-Contact and Sasakian Manifolds 229
7.5. Sasakian Geometry and Algebraic Geometry 236
7.6. New Sasakian Structures from Old 250
Chapter 8. Symmetries and Sasakian Structures 257
8.1. Automorphisms of Sasakian Structures and Isometries 257
8.2. Deformation Classes of Sasakian Structures 265
8.3. Homogeneous Sasakian Manifolds 272
8.4. Symmetry Reduction and Moment Maps 276
8.5. Contact and Sasakian Reduction 290
Chapter 9. Links as Sasakian Manifolds 299
9.1. Preliminaries 299
9.2. Sasakian Structures and Weighted Homogeneous Polynomials 300
9.3. The Milnor Fibration and the Topology of Links 302
9.4. The Dierential Topology of Links 312
9.5. Positive Sasakian Structures on Links 319
9.6. Links of Complete Intersections 326
Chapter 10. Sasakian Geometry in Dimensions Three and Five 329
10.1. Sasakian Geometry in Dimension Three 329
10.2. Sasakian Structures and the Topology of 5-Manifolds 335
10.3. Sasakian Links in Dimension Five 352
10.4. Regular Sasakian Structures on 5-Manifolds 360
Chapter 11. Sasaki-Einstein Geometry 369
11.1. Foundations of Sasaki-Einstein Geometry 370
11.2. Extremal Sasakian Metrics 378
11.3. Further Obstructions to Sasaki-Einstein Structures 382
11.4. Sasaki-Einstein Metrics in Dimensions Five 388
11.5. Sasaki-Einstein Metrics on Homotopy Spheres 403
11.6. The Sasaki-Einstein semi-group 407
11.7. Sasaki-Einstein Metrics in Dimensions Seven and Higher 409
11.8. Sasakian -Einstein Metrics 417
Chapter 12. Quaternionic Kahler and Hyperkahler Manifolds 421
12.1. Quaternionic Geometry of H
n
and HP
n
422
CONTENTS vii
12.2. Quaternionic Kahler Metrics 428
12.3. Positive Quaternionic Kahler Manifolds and Symmetries 434
12.4. Quaternionic Kahler Reduction 438
12.5. Compact Quaternionic Kahler Orbifolds 443
12.6. Hypercomplex and Hyperhermitian Structures 453
12.7. Hyperkahler Manifolds 455
12.8. Hyperkahler Quotients 458
12.9. Toric Hyperkahler Metrics 461
12.10. ALE Spaces and Other Hyperkahler Quotients 465
Chapter 13. 3-Sasakian Manifolds 473
13.1. Almost Hypercontact Manifolds and 3-Sasakian Structures 474
13.2. Basic Properties 478
13.3. The Fundamental Foliations T

and T
Q
480
13.4. Homogeneous 3-Sasakian Manifolds 491
13.5. 3-Sasakian Cohomology 494
13.6. Symmetry Reduction 499
13.7. Toric 3-Sasakian Manifolds 506
13.8. Cohomogeneity One 3-Sasakian 7-Manifolds 522
13.9. Non-Toric 3-Sasakian Manifolds in Dimension 11 and 15 525
Chapter 14. Sasakian Structures, Killing Spinors, and Supersymmetry 529
14.1. The Dirac Operator and Killing Spinors 529
14.2. Real Killing Spinors, Holonomy and Bars Correspondence 533
14.3. Geometries Associated with 3-Sasakian 7-manifolds 535
14.4. Geometries Associated with Sasaki-Einstein 5-manifolds 542
14.5. Geometric Structures on Manifolds and Supersymmetry 545
Appendix A 551
A.1. Preliminaries on Groupoids 551
A.2. The Classifying Space of a Topological Groupoid 555
Appendix B 559
B.1. Reids List of K3 Surfaces as hypersurfaces in CP
4
(w) 559
B.2. Dierential topology of 2/(o
3
o
4
) and 2/(o
5
o
6
) 560
B.3. Tables of Kahler-Einstein metrics on hypersurfaces CP(w) 561
B.4. Positive Breiskorn-Pham Links in Dimension 5 564
B.5. The Yau-Yu Links in Dimensions 5 567
Bibliography 569
Index 607
Introduction
In 1960 Shigeo Sasaki [Sas60] began the study of almost contact structures
in terms of certain tensor elds, but it wasnt until [SH62] that what are now
called Sasakian manifolds rst appeared under the name of normal contact metric
structure. By 1965 the terms Sasakian structure and Sasakian manifold be-
gan to be used more frequently replacing the original expressions. For a number of
years these manifolds were intensively studied by a group of Japanese geometers.
The subject did get some attention in the United States mainly due to the papers
of Goldberg and Blair. Nevertheless, the main interest in the eld remained in-
side Japan, nding it hard to spread out and attract broader attention beyond its
birthplace either in the United States or in Europe.
Over a period of four years between 1965 and 1968 Sasaki wrote a three part
set of lecture notes which appeared as an internal publication of the Mathematical
Institute of the Tohoku University under the title Almost contact manifolds, Part
I-III [Sas65, Sas67, Sas68]. Put together, the work amounts to almost 500 pages.
Even today, after 40 years, the breath, depth and the relative completeness of the
Sasaki lectures is truly quite remarkable. It is hard to understand why they did not
make it as a monograph in some prestigious Western book series; it is a pity. As it
is, the notes are not easily available and, consequently, not well-known
1
. Outside
Japan the rst and important attempt to give a broader account of the subject was
given eight years later by Blair [Bla76a].
After 1968 Sasaki himself was less active although he continued to publish
until 1980. Yet he had already created a new subeld of Riemannian geometry
which slowly started to attract attention worldwide, not just in Japan. In 1966
Brieskorn wrote his famous paper describing a beautiful geometric model for all
homotopy spheres which bound parallelizable manifolds [Bri66]. In 1976 Sasaki
[ST76, SH76] realized that Brieskorn manifolds admit almost contact and con-
tact structures. (This very important fact was independently observed by several
other mathematicians: Abe-Erbacher [AE75], Lutz-Meckert [LM76], and Thomas
[Tho76].) Thirty years later the Brieskorn-Pham links as well as more general links
of weighted homogeneous polynomials, are the key players in several chapters of
our book. Yet again, Sasaki seemed to have had both the necessary intuition and
a broad vision in understanding what is and what is not of true importance
2
.
1
We became aware of the Sasaki notes mainly because of our work on this book. We obtained
a copy of the lectures in 2003 form David Blair and we would like to thank him for sharing them
with us.
2
We know little about Sasakis non-mathematical life. He was born in 1912 and what we do
know is from the volume of his selected papers edited in 1985 by Tachibana [Sas85]. There one
nds a short introduction by S. S. Chern and an essay by Sasaki in which he mostly discusses
his life as a working mathematician. He apparently died almost 20 years ago on August 12th,
1987. Sadly, his death passed without any notice, strangely forgotten. We could not nd an
1
2 INTRODUCTION
Over the years Sasakian geometry has taken a back seat to other areas of
Riemannian geometry, most prominently to the study of Riemannian geometry
with reduced holonomy groups. Nevertheless, as we shall see, Sasakian geome-
try is closely related to all these other geometries. Although generally a Sasakian
manifold has the generic holonomy oO(n), the Riemannian cone over a Sasakian
manifold does have reduced holonomy making the study of Sasakian geometry quite
tractable. A quick perusal of Bergers list of possible irreducible Riemannian holo-
nomy groups (see Table 1.4.1 below) shows that there are ve innite series of these
holonomy groups. All ve are related to Sasakian geometry: generally a Sasakian
metric itself has generic holonomy oO(n). and its Riemannian cone has holonomy
l(n); the Riemannian cone of a general Sasaki-Einstein structure has holonomy
ol(n); whereas, the Riemannian cone of a general 3-Sasakian structure has holo-
nomy oj(n). The remaining innite series in Table 1.4.1 is the group oj(n)oj(1)
of quaternionic Kahler geometry which is closely related to 3-Sasakian geometry as
we discuss shortly. The two remaining irreducible Riemannian holonomy groups
G
2
and Spin(7) are related to Sasakian geometry in a less direct way as we discuss
in Chapter 14. In this same spirit our book attempts to show Sasakian geometry
not as a separate subeld of Riemannian geometry but rather through its inter-
relation to other geometries. This is perhaps the most important feature of the
subject. The study of Sasakian manifolds brings together several dierent elds
of mathematics from dierential and algebraic topology through complex algebraic
geometry to Riemannian manifolds with special holonomy.
The closest relative of Sasakian geometry is Kahlerian geometry, the importance
of which is dicult to overestimate mainly because of its role in algebraic geometry.
But Sasakian geometry also has a very algebro-geometric avor. In fact, there is an
algebraic structure on every Sasakian manifold; whereas, Voisin has shown recently
that there are Kahler manifolds that admit no algebraic structure whatsoever. To
better understand the relation between Sasakian and Kahlerian geometries we begin
with the more familiar relation between contact and symplectic geometries. Let
(`. . ) be a contact manifold where is a contact form on ` and is its Reeb
vector eld. It is easy to see that the cone (C(`) = R
+
`. = d(t)) is
symplectic. Likewise, the Reeb eld denes a foliation of ` and the transverse
space Z is also symplectic. When the foliation is regular the transverse space is a
smooth symplectic manifold giving a projection called Boothby-Wang bration,
and

= d relates the contact and the symplectic structures as indicated by

(((`). ) (`. . )

(Z. ).
English language mathematical obituary honoring his life and work and commemorating his death;
apparently, his passing was noted just in a short obituary of a local newspaper. After his death
Tokyo Science University where he briey worked after retiring from Tohoku took care of his
mathematical heritage with some of his manuscripts placed at the university library. We are
grateful to Professor Yoshinobu Kamishima for this information.
INTRODUCTION 3
We do not have any Riemannian structure yet. It is quite reasonable to ask
if there is a Riemannian metric o on ` which best ts into the above diagram.
As the preferred metrics adapted to symplectic forms are Kahler metrics one could
ask for the Riemannian structure which would make the cone with the warped
product metric o = dt
2
+ t
2
o together with the symplectic form into a Kahler
manifold? Then o and dene a complex structure

. Alternatively, one could ask
for a Riemannian metric o on ` which would dene a Kahler metric / on Z via a
Riemannian submersion. Surprisingly, in both cases the answer to these questions
leads naturally and uniquely to Sasakian Geometry. Our diagram becomes
(((`). . o.

) (`. . . o. )

(Z. . /. J).
From this point of view it is quite clear that Kahlerian and Sasakian geome-
tries are inseparable, Sasakian Geometry being naturally sandwiched between two
dierent types of Kahlerian Geometry. Yet the two fared dierently over the years.
Since Erich Kahlers seminal article several dominant gures of the mathematical
scene of the XXth century have, step after step along a 50 year period, transformed
the subject into a major area of Mathematics that has inuenced the evolution of
the discipline much further than could have conceivably been anticipated by any-
one writes Jean-Pierre Bourguignon in his tributary article The unabated vitality
of Kahlerian geometry published in [Kah03]. Sasakian Geometry has not been as
lucky. There always has been interesting work in the area, but for unclear reasons
it has never attracted people with the same broad vision, people who would set
out to formulate and then work on fundamental problems. Yet, arguably Sasakian
manifolds are at least as interesting as Kahlerian manifolds.
Our own research of the last decade has been an attempt to bring Sasakian
geometry back into the main stream. We believe that a modern book on Sasakian
manifolds is long overdue. Most of the early results are scattered, often buried
in old and hard to get journals. They are typically written in an old fashioned
language. Worse than that, articles with interesting results are drowned in a vast
sea of papers of little importance. There are very few graduate level texts on the
subject. There is the book Structures on manifolds written by Yano and Kon in 1984
[YK84], but this book is over 20 years old and treats both Sasakian and Kahlerian
geometries as a subeld of Riemannian geometry. Recently, Blair substantially
updated his well-known Contact manifolds in Riemannian geometry in the Springer
Lecture Notes series [Bla76a] with Riemannian geometry of contact and symplectic
manifolds [Bla02]. Again the major emphasis as well as the techniques used are
Riemannian in nature. Our monograph naturally complements Blairs as it employs
an entirely dierent philosophy and follows a dierent approach. First we develop
the important relations between Sasakian geometry and the algebraic geometry of
Kahler (actually projective algebraic) orbifolds. Secondly, our major motivation
to begin with was in proving the existence of Einstein metrics. So we have the
understanding of Sasaki-Einstein metrics as a main goal toward which to work, but
4 INTRODUCTION
we have also come to appreciate the beauty and richness of Sasakian geometry in
its own right.
Our book breaks more or less naturally into two parts. Chapters 1 through
9 provide an introduction to the modern study of Sasakian geometry. Starting
with Chapter 10 the book becomes more of a research monograph describing many
of our own results in the subject. However, the extensive introduction should
make it accessible to graduate students as well as non-expert researchers in related
elds. We have used parts or our monograph as a textbook for advanced graduate
courses. For example, assuming that the students have some basic knowledge of
Riemannian geometry, algebraic geometry, and some algebraic topology, a course
treating Sasakian geometry starting with Chapter 4 through Chapter 11 is possible,
drawing on the rst three chapters as review. Many of the results in Chapters 4-11
are given with full proofs bringing the student to the forefront of research in the area.
As a guide we use the principle that proofs, or at least an outline of proofs, are given
for important results that are not found in another book. Our book also contains
many examples, for we believe that the learning process is substantially enhanced
by working through examples. We also have exercises scattered throughout the
text for the reader to sharpen her/his skills. Open problems of varying or unknown
diculty are listed, some of which could be the basis of a dissertation. The text is
aimed at mathematicians, but we hope it will nd many readers among physicists,
particularly those working in Superstring Theory.
We begin in Chapter 1 by introducing various geometries that play more or
less important roles in the way they relate to Sasakian structures. We espouse the
point of view that a geometric structure is best described as a G-structure which,
in addition, may or may not be (partially) integrable. As Sasakian manifolds are all
examples of Riemannian foliations with one-dimensional leaves, Chapter 2 takes the
reader into the world of foliations with a particular focus on the Riemannian case.
The literature is full of excellent books on the subject so we just select the topics
most relevant to us. Chapter 3 reviews some basic facts about Kahler manifolds.
Again, we are very selective choosing only what is needed later in describing the
two Kahler geometries of the Kahler-Sasaki sandwich. Of particular interest is
Yaus famous proof of the Calabi conjecture.
A key tool that allows for connecting Sasakian structures to other geometric
structures is the theory of Riemannian orbifolds and orbifold bundles or orbibun-
dles. For that reason Chapter 4 is crucial in setting the stage for an in depth study
of Sasakian manifolds which begins later in Chapter 7. Orbifolds just as manifolds
have become a household name to the well trained geometer. Nevertheless, a lot
of important results are scattered throughout the literature, and orbifolds typically
appear within a specic context. There is a forthcoming book on orbifolds, Orb-
ifolds and Stringy Topology [ALR06] by Adem, Leida, and Ruan, but this has a
particularly topological bent. Hence, we take some time and eort to prepare the
reader introducing all the basic concepts from the point of view needed in subse-
quent chapters. By Chapter 5 we are ready for a second trip into the realm of
Kahler geometry. However, now the focus is on Kahler-Einstein metrics, in partic-
ular positive scalar curvature Kahler-Einstein metrics on compact Fano orbifolds.
We introduce some basic techniques that allow for proving various existence results.
We also briey discuss obstructions. Chapter 6 presents the necessary foundational
INTRODUCTION 5
material on almost contact and contact geometry. This leads directly to the de-
nition of a Sasakian structure introduced at the very end.
The study of Sasakian geometry nally begins with Chapter 7. We rst present
the important structure theorems, and then gather all the results concerning the
geometry, topology, and curvature properties of both K-contact and Sasakian man-
ifolds. Most of the curvature results are standard and can be found in Blairs book
[Bla02], but our main focus is dierent: we stress the relation between Sasakian
and algebraic geometry, as well as the basic cohomology associated with a Sasakian
structure. A main tool used in the text is the transverse Yau Theorem due to El
Kacimi-Alaoui. In the companion Chapter 8 we present known results concerning
symmetries of Sasakian structures. We introduce the Sasakian analogue of the bet-
ter known symplectic/contact reduction. Then we study toric contact and toric
Sasakian manifolds and prove several Delzant-type results.
Chapter 9 is devoted to the geometry of links of isolated hypersurface singulari-
ties as well as a review of the dierential topology of homotopy spheres a la Kervaire
and Milnor [KM63]. A main reference for the study of such links is Milnors classic
text Singular points of complex hypersurfaces [Mil68], but also Dimcas Singular-
ities and topology of hypersurfacs [Dim92] is used. The dierential topology of
links is a beautiful piece of mathematics, and this chapter oers a hands-on users
guide approach with much emphasis on the famous work of Brieskorn [Bri66]. Of
importance for us is that when the singularities arise from weighted homogeneous
polynomials the links have a natural Sasakian structure with either denite (posi-
tive or negative) or null basic rst Chern class. Emphasis is given to the positive
case which corresponds to having positive Ricci curvature. In Chapter 10 we dis-
cuss the Sasakian geometry in low dimensions. In dimension 3 there is a complete
classication. Dimension 5 is large enough to be interesting, yet small enough to
hope for some partial classication. We concentrate on the simply connected case
as there we can rely on the Smale-Barden classication. In terms of Sasakian struc-
tures our main focus is on the case of positive Sasakian structures. In considerable
detail we describe several remarkable theorems of Kollar which show how positivity
severely restricts the topology of a manifold which is to admit a positive Sasakian
structure.
Chapter 11 is central to the whole book and perhaps the main reason and jus-
tication for it. Much of this chapter is based on a new method for proving the ex-
istence of Einstein metrics on odd dimensional manifolds introduced by the authors
in 2001 [BG01b]. We realized there that links of isolated hypersurface singulari-
ties obtained from weighted homogeneous polynomials admit Sasakian structures.
Moreover, by using an orbifold version [BG00b] of an old result of Kobayashi, we
successfully tied the problem to making use of the continuity method for proving
the existence of Kahler-Einstein metrics on compact Kahler orbifolds. For the au-
thors this was the original raison detre for Chapters 4, 5, and 9 of the book.
In a series of papers the authors and their collaborators have successfully applied
this method to prove the existence of Sasaki-Einstein metrics on many 5-manifolds,
on odd dimensional homotopy spheres that bound parallelizible manifolds, as well
as on odd dimensional rational homology spheres. Furthermore, our method has
been substantially generalized by Kollar who has pushed our understanding much
further especially in dimension ve. Although a complete classication is perhaps
6 INTRODUCTION
not within reach we now begin to have a really good grasp of Sasaki-Einstein ge-
ometry in dimension 5. In addition, we discuss toric Sasaki-Einstein geometry in
dimension 5 which began with the work of Gauntlett, Martelli, Sparks, and Wal-
dram [GMSW04b], and culminates with the very recent work of Cho, Futaki,
Ono, and Wang [FOW06, CFO07] which shows in arbitrary odd dimension that
any toric Sasakian manifold with positive anticanonical Sasakian structure admits
a compatible Sasaki-Einstein metric. We also discuss extremal Sasaki metrics de-
ned in analogy with the extremal Kahler metrics, and introduce the Sasaki-Futaki
invariant [BGS06]. In addition to lifting the well-known obstructions of positive
Kahler-Einstein metrics we also present some new results due to Gauntlett, Martelli,
Sparks, and Yau [GMSY06] involving two well known estimates, one due to Lich-
nerowicz, and the other to Bishop. We also present Sasaki-Einstein metrics obtained
via the join construction described earlier in Chapter 7. We end this long chapter
with a brief discussion of Sasakian--Einstein metrics.
Chapter 12 gives an extensive overview of various quaternionic geometries. The
main focus is on the positive quaternionic Kahler (QK) manifolds (orbifolds) and
on the hyperkehler manifolds (orbifolds). The reason for such an extensive treat-
ment has to do with Chapter 13. The 3-Sasakian manifolds studied there cannot
be introduced without a deeper understanding of these two quaternionic geome-
tries, just as Sasakian and Sasaki-Einstein manifolds cannot be studied without
Kahlerian and Kahler-Einstein geometry. The Sasaki-Einstein manifolds of Chap-
ter 13 have a completely dierent avor than the ones that appeared in Chapter
11. It is not only that these occur only in dimensions 4: + 3, but also that they
have a somewhat richer geometric structure. In addition the method in which
the metrics are obtained is completely dierent. In Chapter 11, with some excep-
tions, we mostly get our existence results via the continuity method applied to the
Monge-Amp`ere equation. Very few metrics are known explicitly, though there are
exceptions. Most of the 3-Sasakian metrics we consider are obtained via symmetry
reduction similar to the hyperkahler and quaternionic Kahler reduction. Indeed the
three quotients are all related. So the manifolds and the metrics we get are quite
often explicit and can be studied as quotients. Again, there are some exceptions.
Finally, Chapter 14 gives a very brief overview of the rich theory of Killing spinors.
There we describe some other geometries and show how they relate to 3-Sasakian
7-manifolds and Sasaki-Einstein 5-manifolds. At the end we very briey comment
on how Sasakian geometry naturally appears in various supersymmetric physical
theories. Both Sasaki-Einstein geometry and geometries with exceptional holonomy
have appeared in various models of supersymmetric String Theory fuelling vigorous
interest in them by mathematicians and physicists alike.
We also have added two appendices. The rst appendix gives a very brief in-
troduction to groupoids and their classifying spaces which are employed in Chapter
4, while the second gives many tables listing links of hypersurface singularities that
are used throughout the book.
We have compiled a very extensive bibliography. There are various reasons for
its size. We should remark that in this day and age of easy internet access, with
MathSciNet and Google, it would make no sense to simply compile a bibliography
of every paper with the words Sasaki or Sasakian in the title. Anyone with an
access to MathSciNet can easily compile such a list of 809 papers
3
so it would serve
3
As we checked on February 19, 2007.
INTRODUCTION 7
no purpose to do it for this book. In a way, in spite of its size, we were very careful
and selective in choosing all bibliography items. Our book brings together so many
dierent areas of mathematics, that having good references becomes essential. In
some cases the proofs we give are only sketches and in such instances we wanted
to refer the reader to the best place he/she could nd more details. That is quite
often the original source but not always. We refer to various books, monographs,
and lecture notes. We have tried to be both selective and accurate in attributing
various results with care. This can be at times a hard task. We suspect that we
did not always get the references and proper credits exactly right. This is almost
inevitable considering we are not experts in many of the areas of mathematics that
substantially enter as part of the book. In any case, we apologize for any omissions
and errors; these are certainly not intentional
4
.
Finishing this book was a challenging task. We had to deal with an increasing
number of new and interesting results appearing every few months. It is always a bit
dangerous to include material based on articles that have not yet been published.
To make the book as up-to-date as possible we took the risk to include some of
these new results without giving proofs. On the other hand we are happy to see
that the eld is active and we very much hope that our book will become somewhat
outdated in a few years.
4
There are many examples in the literature where incorrect attributions are made. A recent
case that we just uncovered is that of nearly Kahler manifolds which are usually attributed to
Gray [Gra69b, Gra70], yet they were discovered 10 years earlier by Tachibana [Tac59]. The
fact that any nearly Kahler 6-manifold is Einstein is also attributed to Gray in [Gra76], yet it
was proven earlier by Matsumoto [Mat72]. See Section 14.3.2.
CHAPTER 1
Structures on Manifolds
A unifying viewpoint for doing dierential geometry involving dierent struc-
tures is that of a G-structure, where G can stand for geometric or more appro-
priately a Lie group. A (rst order) G-structure is just a reduction of the bundle
of frames of a manifold from the general linear group to a subgroup. For example,
from the point of view of G-structures a Riemannian metric on a manifold corre-
sponds to a reduction of the frame bundle with group G1(n. R) to the orthonormal
frame bundle with subgroup, the orthogonal group O(n. R). Many other examples
will be given below. It will be important to have at our disposal the general theory
of connections in principal and associated bre bundles of which the G-structures
mentioned above are special cases. However, before embarking on the study of such
structures we give a short review of sheaves and their cohomology groups. Sheaves
are more general than bundles, but they are a bit too general for describing geomet-
ric structures on manifolds. They are mainly used to pass from local information
to global information.
1.1. Sheaves and Sheaf Cohomology
Sheaves, which were invented by Jean Leray as a prisoner of war during World
War II, have become an important tool in geometry. Here we give a very brief tour
of sheaf theory referring to the literature [Bre97, GR65, Hir66, Wel80] for more
detail and proofs. Our presentation follows [Wel80] fairly closely.
Denition 1.1.1: Let A be a topological space. A presheaf on A is an assignment
to each nonempty open set l A a set T(l) together with maps, called restriction
maps,

U
V
: T(l)T(\ )
for each pair of open sets l and \ with \ l that satisfy the conditions
U
U
= id
U
and
U
W
=
U
V

V
W
whenever, \ \ l.
Very often the sets T(l) have some additional algebraic structure, such as a
group, a ring, or a module structure. In this case we assume that the restriction
maps preserve the algebraic structure. So a presheaf is a contravariant functor
from the category of open sets of a xed topological space with inclusion maps as
morphisms to the category of groups, rings, or modules whose morphisms are the
homomorphisms of that category.
9
10 1. STRUCTURES ON MANIFOLDS
A morphism of presheaves 1 : T( on A is a family of homomorphisms
1
U
: T(l)((l) such that the diagram
(1.1.1)
T(l)
h
U
((l)

U
V

U
V
T(\ )
h
V
((\ )
commutes, where \ l A. If the homomorphisms 1
U
: T(l)((l) are
monomorphisms then T is a subpresheaf of (.
Given a presheaf T over A we can consider the direct limit
T
x
= lim

xU
T(l)
with respect to the restriction maps
U
V
. Clearly, T
x
inherits whatever algebraic
structure the sets T(l) have. T
x
is called the stalk of T at r. and for r l there
is a natural projection
U
x
: T(l)T
x
by sending an element : T(l) to its
equivalence class :
x
in the direct limit. :
x
is called the germ of : at r.
Denition 1.1.2: A presheaf T is called a sheaf if for every collection l
i
of open
subsets of A with l = l
i
. the presheaf T satises the following two conditions
(i) If :. t T(l) and
U
Ui
(:) =
U
Ui
(t) for all i. then : = t.
(ii) If :
i
T(l
i
) and if for any , with l
i
l
j
= the equality
U
i
UiUj
(:
i
) =

U
j
U
i
U
j
(:
j
) holds for all i. then there is an : T(l) such that
U
U
i
(:) = :
i
for all i.
A morphism of sheaves is a morphism of the underlying presheaves. In par-
ticular, an isomorphism of sheaves is a morphism of sheaves such that the all the
maps 1
U
: T(l)((l) are isomorphisms. Not every presheaf is a sheaf. For
example consider C and the presheaf that assigns to every open set l C the
algebra of bounded holomorphic functions B(l) on l. Since there are no bounded
holomorphic functions on all of C this presheaf violates condition (ii) of denition
1.1.2 above. Nevertheless, as we shall see shortly we can associate a sheaf to any
presheaf in a fairly natural way. First we give an important example of a presheaf
that is a sheaf.
Example 1.1.3: Let A and Y be topological spaces and consider the presheaf (
X,Y
on A which associates to every open subset l of A the set (
X,Y
(l) of all continuous
maps from l to Y. The restriction map
U
V
is just the natural restriction, i.e. if
1 (
X,Y
(l) and \ l. then
U
V
(1) = 1[
V
. It is easy to see that (
X,Y
satises
the two conditions of Denition 1.1.2 and thus denes a sheaf on A. An important
special case is obtained by taking Y to be either of the continuous elds F = R or
C. In this case we note that (
X,F
has the structure of an F-algebra. Two particular
cases of interest are when A is either a smooth real manifold or a complex manifold.
In these cases we are more interested in the subsheaf c
X
(
X,R
of smooth functions
and the subsheaf O
X
(
X,C
of holomorphic functions, respectively. Such sheaves
are called the structure sheaf of A.
Denition 1.1.4: An etale space over a topological space A is a topological space
Y together with a continuous map : Y A that is a local homeomorphism.
1.1. SHEAVES AND SHEAF COHOMOLOGY 11
For each open set l A we can consider the set (l. Y ) of continuous sections
of . that is, the subset of 1 (
X,Y
(l) that satisfy 1 = id
U
. Then the assignment
that assigns to each open subset l of A the set of continuous sections (l. Y ) forms
a subsheaf of (
X,Y
which we denote by o
X,Y
. We shall now associate to any presheaf
T an etale space

T. Then by taking sections of

T we shall get a sheaf. Dene

T by
(1.1.2)

T = .
xX
T
x
.
Let :

TA denote the natural projection that sends any :
x
T
x
to r. We
dene a topology on

T as follows: for each : T(l) dene a map : : l

T by
:(r) = :
x
. This satises : = id
U
. Then we let the sets
: [ l is open in A. : T(l)
be a basis for the topology on

T. Then and all functions : are continuous, and it
is easy to check that is a local homeomorphism, so that

T is an etale space over
A. But we have already seen that the set of sections of an etale space form a sheaf.
We denote the sheaf associated to the etale space

T by

o
X,Y
. Thus, beginning
with a presheaf T we have associated a sheaf

o
X,Y
. This sheaf is called the sheaf
associated to or generated by the presheaf T.
Exercise 1.1: Show that if one starts with a sheaf T then the sheaf associated to
T is isomorphic to T.
Denition 1.1.5: A ringed space is a pair (A. /) consisting of a topological
space A together with a sheaf of rings / on A. called the structure sheaf. (A. /)
is a locally ringed space if for each point r A, the stalk /
x
is a local ring.
All ringed spaces considered in this book will be locally ringed spaces, so we
often omit the word locally and just refer to a ringed space. We now consider the
structure sheaves c
M
and O
M
on a real or complex manifold `. respectively. We
denote by / the sheaf c on a real manifold or the sheaf O on a complex manifold.
Denition 1.1.6: Let (A. /) be a ringed space with structure sheaf / given by c
(or O), respectively. A sheaf T of /-modules is said to be locally free of rank
r if A can be covered by open sets l such that there is an isomorphism of sheaves
T[
U
/[
U

r times
/[
U
. A locally free rank 1 sheaf is called an invertible sheaf.
Let : 1` be an F-vector bundle over a smooth manifold `. As in the
case of an etale space the subset of 1 (
M,E
satisfying 1 = id
M
denes a
subsheaf of (
M,E
, called the sheaf of germs of continuous sections of 1. Actually,
we are more interested in the sheaf of germs of smooth sections of 1 which we
denote by c(1). On a complex manifold ` a C-vector bundle can have a special
type of structure.
Denition 1.1.7: Let ` be a complex manifold. A complex vector bundle :
1` on ` is said to be holomorphic if 1 is a complex manifold with
holomorphic, and the transition functions for 1 can be taken to be holomorphic
functions.
Of course, the transition functions being holomorphic is equivalent to the local
trivializations being holomorphic. Then we have
Proposition 1.1.8: Let ` be a real or complex manifold. There is a one-to-one
correspondence between smooth vector bundles on ` and locally free c
M
-sheaves on
12 1. STRUCTURES ON MANIFOLDS
`. Similarly, when ` is a complex manifold, there is a one-to-one correspondence
between holomorphic vector bundles on ` and locally free O
M
-sheaves on `.
The correspondence is given by associating to a holomorphic vector bundle
1 the sheaf of germs of holomorphic sections of 1. Because of Proposition 1.1.8
we shall often use interchangeably the concepts of vector bundles and locally free
sheaves. Of particular interest is the rank one case in the holomorphic category
which gives a one-to-one correspondence between holomorphic line bundles and
invertible O
M
-sheaves.
As mentioned previously the power of sheaf theory is as a tool in passing from
the local to the global. This is accomplished by the way of sheaf cohomology theory
which we now briey describe.
1.1.1. Sheaf Cohomology. There are several approaches to sheaf cohomol-
ogy theory. There is the axiomatic approach used in [GR65], the derived functor
approach using the resolution by discontinuous sections due to Godement [God58]
and espoused in [Bre97, Wel80], and nally the

Cech theoretic approach described
in [GH78b] and [Kod86]. We prefer to begin with the latter and then present the
so-called Abstract de Rham Theorem using resolutions as in [Wel80].
For simplicity we assume that A is paracompact and Hausdor. Let T be a
sheaf A, and let U = l
i

i
be a locally nite cover by open sets. Dene the set
C
p
(U. T) of p-chains of T by
(1.1.3) C
p
(U. T) =

i0,...,ip
T(l
i
0
l
i
p
).
where we assume that the indices i
j
are distinct. We denote elements of C
p
(U. T)
by
i
0
,...,i
p
. The coboundary operator
p
: C
p
(U. T)C
p+1
(U. T) is dened by
(1.1.4) (
p
)
i0,...,ip+1
=
p+1

j=0
(1)
j

i0,...,

ij,...,ip+1
[
Ui
0
Ui
p
.
where as usual

i means remove that index. This gives rise to a cochain complex
C
0
(U. T)

0
C
1
(U. T)

1
C
p
(U. T)

p
C
p+1
(U. T)
and we dene the cohomology groups H
p
(U. T) by
H
p
(U. T) =
ker
p
Im
p1
.
If Vis a cover of A rening U. there are homomorphisms : H
p
(U. T)H
p
(V. T).
Thus, we can take the direct limit of the cohomology groups H
p
(U. T) as the cover
becomes ner and ner, so we dene the cohomology group with coecients in the
sheaf T by
(1.1.5) H
p
(A. T) = lim

U
H
p
(U. T).
Note that generally the groups H
p
(U. T) depend on the cover, but there are certain
special covers known as acyclic covers for which we have the following theorem of
Leray
Theorem 1.1.9: If the cover U is acyclic in the sense that H
q
(l
i0
l
ip
. T)
vanishes for all c 0 and all i
0
. . . . . i
p
, then H

(`. T) H

(U. T).
1.1. SHEAVES AND SHEAF COHOMOLOGY 13
In practice, one always computes H

(U. T) for an acyclic cover and then uses

the above theorem.
Example 1.1.10: Consider the sphere o
2
=(r
1
. r
2
. r
3
) R
3
[

3
i=1
(r
i
)
2
= 1 and
the constant sheaf T = R. together with two covers. The rst is the stereographic
cover dened by two open sets l

= o
2
(0. 0. 1). This cover is not acyclic
since l
+
l

has the homotopy type of a circle. The second cover is a cover

consisting of a cover of the lower hemisphere by 3 open sets together with the
upper hemisphere. This gives an acyclic cover from which one can compute the
cohomology groups H
i
(o
2
. R) of o
2
. See Example 9.3 of [BT82].
The sheaf cohomology groups have some nice properties. For any sheaf T the
cohomology group H
0
(A. T) is the space of global sections T(A) of T. and a sheaf
morphism 1 : /B induces a group homomorphism 1
q
: H
q
(A. /)H
q
(A. B).
Moreover, a short exact sequence of sheaves,
0/B(0
induces a long exact sequence in cohomology,
(1.1.6)
0H
0
(`. /)H
0
(`. B)H
0
(`. ()

H
1
(`. /)

H
q
(`. /) .
where is the well-known connecting homomorphism.
To proceed further we need some denitions.
Denition 1.1.11: A sheaf T is soft if for any closed subset o A, the restriction
map
X
U
: T(A)T(o) is surjective. T is ne if for any locally nite open cover
l
i
of A there is a partition of unity of T. i.e. there is a family of sheaf morphisms

i
: TT such that

i
= 1. and
i
(T
x
) = 0 for all r in some neighborhood of
A l
i
.
The importance of soft sheaves is the following
Proposition 1.1.12: If T is a soft sheaf, then H
q
(A. T) = 0 for all c 0.
Fine sheaves are special cases of soft sheaves, viz.
Proposition 1.1.13: Fine sheaves are soft.
Example 1.1.14: The structure sheaf c
M
of a real smooth manifold is ne, thus
soft. The structure sheaf O
M
of a complex manifold is neither ne nor soft, nor are
the constant sheaves.
Denition 1.1.15: Let T be a sheaf on A. A resolution of T is an exact sequence
of sheaves T
i
of the form
0TT
0
T
1
T
k
.
The resolution is acyclic if H
q
(A. T
i
) = 0 for all c 0 and all i 0.
A resolution of a sheaf T is conveniently written in the shorthand notation
0TT

. By Proposition 1.1.12 if the sheaves T

i
of a resolution are soft
sheaves, then the resolution is acyclic. Thus, a resolution by soft or ne sheaves is
acyclic. A resolution of sheaves gives rise to a cochain complex C

= C

(T

) on
global sections
(1.1.7) 0T(A)T
0
(A)
d
0
T
1
(A)
d
1

d
k1
T
k
(A)
d
k
.
14 1. STRUCTURES ON MANIFOLDS
In general this sequence is exact only at T(A). but we do have d
k+1
d
k
= 0. The
cohomology group
H
k
(C

) =
ker d
k
im d
k1
of this complex is called the /
th
derived group of C

. We are now ready for the

Abstract de Rham Theorem:
Theorem 1.1.16: Let T be a sheaf over a paracompact Hausdor space A, and
let 0TT

be an acyclic resolution of T. Then there is a natural isomorphism

H
q
(A. T) H
q
(C

) =
ker d
k
im d
k1
.
This is a very powerful theorem of which the usual de Rham theorem is a special
case. We give this as an example below. The Dolbeault Theorem is another special
case which will be given in Chapter 3.
Example 1.1.17: de Rhams Theorem. Let ` be a smooth manifold and let
c
p
denote the sheaf of germs of sections of the exterior bundles
p
` with c
0
= c.
Let T = R the constant sheaf on `. Then by the well-known Poincare Lemma we
get a resolution of the constant sheaf R. viz
0Rc
0
c
1
c
n1
c
n
0.
Since the sheaves c
p
are ne this resolution is acyclic, so by the abstract de Rham
Theorem 1.1.16 we have an isomorphism
H
k
(`. R) H
k

ker d
k
im d
k1

= H
k
deRh
(`. R).
Moreover, H
k
(`. R) can be identied with the /
th
singular cohomology group by
taking an acyclic resolution of the constant sheaf R by sheaves of singular cochains.
See [Wel80].
1.2. Principal and Associated Bundles
We begin by considering principal bre bundles. The main reference here is
the classic text of Kobayashi and Nomizu[KN63, KN69].
Denition 1.2.1: A principal bundle 1 over ` with Lie group G consists of:
(1) A pair of smooth manifolds 1 and ` together with a smooth surjection : 1
`.
(2) A smooth free action of the Lie group G on 1, 1 : 1 G 1 given by the
right action (n. o) 1
a
n = no such that the quotient is `, i.e., ` = 1G.
(3) 1 is locally trivial. More precisely, every point j ` has a neighborhood l
and a dieomorphism :
1
(l) l G dened by (n) = ((n). (n)). where
:
1
(l) G satises the compatibility condition (no) = (n)o.
As usual the manifold 1 is called the total space, the manifold ` is called the base
space, and for j `,
1
(j) is called the bre at j. If we x a point n
0

1
(j)
the map sending o G to n
0
o
1
(j) identies
1
(j) with G. Moreover, one
easily sees that this map is a dieomorphism from G to
1
(j) whose inverse is .
When we want to emphasize the base space and the group we write 1(`. G) for 1,
or we refer to 1 as a principal G-bundle over `. Notice that for any open subset
l `.
1
(l) is a principal G-bundle over l. Another example is the trivial
G-bundle when 1 is just the product ` G.
1.2. PRINCIPAL AND ASSOCIATED BUNDLES 15
Now let l

be an open cover of ` such that for each the map

1
(l

)l

G is a dieomorphism, and satises the compatibility condition

of Denition 1.2.1. Then if n
1
(l

) we see that

(no)(

(no))
1
=

(n)(

(n))
1
for all o G. Hence, we can dene a smooth map

: l

G by
(1.2.1)

((n)) =

(n)(

(n))
1
.
The smooth maps

are called transition functions for the principal bundle

1(`. G). and one easily sees that for each j l

they satisfy the

cocycle condition
(1.2.2)

(j) =

(j)

(j).
The appellation cocycle condition comes from the fact that equivalence classes of
transition functions are elements of the sheaf cohomology set H
1
(`. (). where (
denotes the sheaf of germs of maps from opens sets of ` to G. Here two transition
functions

and
t

are equivalent if there are maps

: l

G such that

(j) =

(j)

(j)(

(j))
1
. and the inverse element in G is given by

(j) =
(

(j))
1
. The reader is referred to Section 1.1 below for more details on sheaf
theory and its cohomology. We then have the well-known
Theorem 1.2.2: Let l

be an open cover on the smooth manifold ` and let

G be a Lie group. Suppose further that on each nonempty intersection l

there exist smooth maps

: l

G satisfying (1.2.2). Then there is a

principal G-bundle 1(`. G) with transition functions

. Moreover, there is a
bijective correspondence between isomorphism classes of principal G-bundles and
elements of H
1
(`. ().
The rst statement in the theorem says that the transition functions determine
the principal bundle, and the last statement says that the set H
1
(`. () classies
principal G-bundles over `. The denition of an isomorphism of principal bundles
will be given below. It is only in the case that G is Abelian, for example, G = o
1
,
the circle, that H
1
(`. () has itself the structure of an Abelian group, and more
standard techniques such as the exponential sequence can be used to relate this to
the integral cohomology of `. We will discuss this later in more detail.
Perhaps the most important example of a principal bundle is:
Example 1.2.3: The linear frame bundle 1(`). Recall that a frame n =
(A
1
. . . . . A
n
) at a point j ` is a basis of the tangent space T
p
`. We let 1(`)
denote the set of all frames at all points of `. Then : 1(`)` is a principal
bundle on ` with group G1(n. R). The action of G1(n. R) on 1(`) is just given
by matrix multiplication from the right, that is (n. ) n for n 1(`) and
G1(n. R). The local triviality can be seen by choosing a local coordinate chart
(l; r
1
. . . . . r
n
) on ` and writing the vectors A
i
of the frame n in local coordinates
as
A
i
=

j
A
j
i

r
j
.
where the A
j
i
are the components of a non-singular matrix (A
j
i
) of smooth functions
on l. Then the dieomorphism :
1
(l)l G1(n. R) is given by () =
((). (A
j
i
())). We remark that a point n 1(`) can be viewed as a linear map
16 1. STRUCTURES ON MANIFOLDS
n : R
n
T
p
` with j = (n) by n(c
i
) = A
i
. where c
1
. . . . . c
n
denotes the
standard basis of R
n
.
Exercise 1.2: Let G be a Lie group and H a closed Lie subgroup. Show that the
Lie group G can be viewed as a principal H-bundle G(GH. H) over the homoge-
neous manifold GH.
Now suppose we are given a principal G-bundle 1(`. G) over `, and a prin-
cipal H-bundle Q(. H) on . A homomorphism of principal bundles consists of a
smooth map

1 : Q1 together with a Lie group homomorphism / : HG such
that for all n Q and o H.

1 satises

1(no) =

1(n)/(o). This condition implies
that

1 maps bres to bres; hence, there is a smooth map 1 : ` such that
the diagram
(1.2.3)
Q

f
1

f
`
commutes. If in addition

1 is an embedding of smooth manifolds and / is a group
monomorphism then we say that

1 : Q1 is an embedding of principal bundles.
This implies that the map 1 : ` is also an embedding. In the case that
= ` and 1 is the identity map on `. then Q(`. H) is called a subbundle
or a reduction of 1(`. G) to the group H. We also say that Q(`. H) is a reduced
subbundle of 1(`. G). It may not be possible to reduce a principal bundle 1(`. G)
to a given subgroup H. In general, there are topological obstructions to doing so. We
shall see many examples of this below. If also / is an isomorphism then Q(`. H)
and 1(`. G) are isomorphic as principal bundles. We also say that 1(`. G) is
trivial if it is isomorphic to the product ` G. Observe that the local triviality
conditions says that every point j ` has a neighborhood l such that
1
(l) is
trivial. Concerning the transition functions we have
Proposition 1.2.4: A principal G-bundle 1(`. G) can be reduced to a Lie sub-
group H if and only if there is an open cover l

of ` with transition functions

taking their values in H.

Proof. Suppose that 1(`. G) can be reduced to a Lie subgroup H. Then there
is a principal H-bundle Q(`. H) together with a smooth bundle map 1 : Q1
such that the diagram (1.2.3) commutes and 1(no) = 1(n)o for all o H G. On
the open set
1
Q
(l

) Q(`. H) the maps

Q

and
P

are related by
Q

=
P

1.
It follows that for any 1 there exists an o G and n Q such that = 1(n)o
and
P

() =
Q

(n)o. Hence, the transition functions

P

satisfy
(1.2.4)
P

() =
P

()(
P

())
1
=
Q

(n)o o
1
(
Q

(n))
1
=
Q

(n)(
Q

(n))
1
.
and thus, have their values in H.
Conversely, given transition functions
P

: l

G which take their

values in the Lie subgroup H. a standard result says that
P

is smooth as a
map into H. Thus, by Proposition 1.2.2 there is a principal H-bundle Q(`. H)
with transition functions
P

. To construct the bundle map 1 we dene maps

1

:
1
Q
(l

)
1
P
(l

) by putting 1

=
1
P

Q
. One easily sees that 1

= 1

1.2. PRINCIPAL AND ASSOCIATED BUNDLES 17

on l

. and so denes a global bundle map 1 : Q1 with the requisite

properties.
Let 1(`. G) be a principal bundle and 1 a G-manifold, that is, a manifold
together with a smooth action of the Lie group G. We denote this action by left
multiplication as r o
1
r. where r 1 and o G. Then we have the product
action on 1 1 dened by (n. r) (no. o
1
r). The quotient space (1 1)G
by this action is a smooth bre bundle 1(`. 1. G. 1) called the bundle associated
to 1(`. G) with bre 1. Note that the projection map
E
is dened as follows: let
[n. r] denote equivalence class of the pair (n. r) 1 1. where (n. r) is equivalent
to (n
t
. r
t
) if there is an o G such that (n
t
. r
t
) = (no. o
1
r). Then dene the
projection map
E
: 1` by
E
([n. r]) =
P
(n). One easily checks that this
is well dened. In the case that 1 is a vector space, say R
k
, and G = G1(/. R)
the associated bundle 1 = 1(`. R
k
. G1(/. R). 1) is called a vector bundle (of rank
/) associated to the principal bundle 1(`. G1(/. R)). or just a real vector bundle.
Similarly, if the bre 1 is C
k
with group G = G1(/. C). then 1 is called a complex
vector bundle. We shall often combine these an refer to a real or complex vector
bundle as an F-bundle, where F = R or C. A vector bundle of rank one is called a
line bundle.
Exercise 1.3: Use
E
to dene a dierential structure on 1(`. 1. G. 1) that
makes it a smooth bre bundle with projection map
E
and bre 1.
Exercise 1.4: Show that the tangent bundle T` of ` is a vector bundle associ-
ated to the principal bundle 1(`). More generally show that the tensor bundles
T
r
s
`. exterior bundles
p
`. and symmetric bundles o
p
` are vector bundles as-
sociated to the principal bundle 1(`).
One often describes structures on manifolds by tensor elds. Examples are
complex structures, symplectic structures, Riemannian structures. These arise as
smooth sections of certain bundles over ` which are associated to the frame bundle
1(`) or its subbundles. Recall that a section of a bre bundle
E
: 1`
is a (smooth) map : : `1 such that
E
: = 1l
M
. Unless otherwise stated
sections will be smooth. Vector bundles always have sections (e.g., the zero section);
whereas, a principal bundle has a section if and only if it is trivial. This is easy to
verify directly, but also follows from Theorem 1.2.5 below. By a local section of 1
we shall mean a section of the bundle
1
E
(l) for some open set l in `.
Theorem 1.2.5: A principal G-bundle 1(`. G) is reducible to a closed subgroup H
if and only if the associated bundle 1(`. GH. G. 1) admits a section : : `1.
Moreover, there is a bijective correspondence between such sections : and subbundles
Q(`. H) of 1(`. G).
Proof. Suppose 1(`. G) is reducible to a closed subgroup H and let 1 :
Q1 be the subbundle. The associated bundle 1 can be identied with the
quotient space 1H by mapping the equivalence class [n. oH] 1 to the equivalence
class [no] 1H. Let : 11 = 1H denote the natural projection. Then it is
easy to see that 1 is constant on the bres of Q. Thus, we can dene a section
: : `1 by setting :(r) = (1(n)). where
P
(1(n)) = r.
Conversely, let : : `1 = 1H be a section, and consider 1 as a principal
H-bundle over 1 = 1H with transition functions /

: \

H. where \

is
an open cover of 1. Then we can dene transition functions for 1(`. G) with values
18 1. STRUCTURES ON MANIFOLDS
in H by setting

= /

:. Thus, by Proposition 1.2.2 1(`. G) is reducible to

the subgroup H. The correspondence between sections and H-subbundles can be
seen to be 1-1.
Let 1(`. G) be a principal bundle on a smooth manifold ` with G = G1(/. R),
and let 1 be an associated real vector bundle of rank /. Reductions of 1(`. G) to
subgroups G G1(/. R) correspond to adding certain structures to its associated
vector bundle 1. In general, there are obstructions to be able to do this. For exam-
ple, a reduction to the group G1(/. R)
+
corresponds to choosing an orientation on
1. which can be done if and only if the bundle 1 is orientable. Such obstructions are
often given in terms of so-called characteristic classes [MS74]. Obstruction theory
for general bre bundles is expounded in the classic text of Steenrod [Ste51] and
also in [Hus66]. 1 is orientable if and only if its rst Stiefel-Whitney class n
1
(1)
vanishes. See Appendix A. Many more examples of such obstructions are discussed
for G-structures in Section 1.4. Another example of such a reduction is the reduc-
tion of 1(`. G1(/. R)) to its maximal compact subgroup O(/. R). Since G1(/. R)
is homotopy equivalent to the orthogonal group O(/. R) there are no obstructions
to performing this reduction. This corresponds to choosing a Riemannian metric
on 1. If 1 is also orientable then we can reduce further to the special orthogonal
group oO(/. R). The corresponding principal bundle is now 1(`. oO(/. R)).
Another type of structure comes from lifting instead of reduction. Thus, since

1
(oO(/. R)) = Z
2
the group oO(/. R) has a two-sheeted universal covering group
called Spin(/). So given a 1(`. oO(/. R)) or one of its associated vector bundles 1
one can ask whether the bundle 1(`. oO(/. R)) can be lifted to a covering bundle
1(`. Spin(/)) with group Spin(/)? When such a covering exists the bundle 1 is
said to have a spin structure. We refer to [LM89] for a full treatment of spin
structures. The exact sequence of Lie groups 0Z
2
Spin(/)oO(/)0
induces the coboundary map n
2
: H
1
(`. oO(/))H
2
(`. Z
2
) whose image is
just the second Stiefel-Whitney class n
2
(1) described briey in Appendix A. So
an oriented vector bundle 1 admits a spin structure if and only if n
2
(1) = 0.
Moreover, the distinct spin structures on 1 are in one-to-one correspondence with
the elements of H
1
(`. Z
2
). We say that a smooth oriented manifold ` is a spin
manifold if T` admits a spin structure. So ` is spin if and only if n
1
(`) =
n
2
(`) = 0. Spin manifolds admit certain vector bundles that do not exist on non-
spin manifolds, namely, those whose bres are representations of Spin(n) that are
not representations of oO(n) or G1(n. R). Such vector bundles are called spinor
bundles and its sections are called spinor elds. Evaluation at a point gives a
spinor. Notice that spinor bundles depend on a choice of Riemannian metric. A
study of these representations involves Cliord algebras which will not be treated
in any detail in this text, and will not appear until Chapter 14. See [LM89] for a
thorough treatment.
1.3. Connections in Principal and Associated Vector Bundles
In this section we briey review the fundamentals of the theory of connections.
As the reader undoubtedly knows there are various settings and denitions for a
connection. We begin with the most general as well as most abstract, namely that
of connections in a principal bre bundle. We then discuss connections in associated
vector bundles, and the relationship to the former. Both formulations have their
advantages and will be used in the sequel. This brief discussion will also allow us to
1.3. CONNECTIONS IN PRINCIPAL AND ASSOCIATED VECTOR BUNDLES 19
set our notation and terminology for the remainder of the text. Again the standard
reference for much of this material is the classic text of Kobayashi and Nomizu
[KN63, KN69].
Let 1 be a principal bundle over ` with group G. Let n 1 and consider the
tangent space T
u
1 at n. Let G
u
be the vertical subspace of T
u
1 consisting of all
vectors that are tangent to the bre
1
((n)). The dierential of the restriction
of the map to the bre through n identies G
u
with the tangent space T
e
G at
the identity c G, and thus with the Lie algebra g of G.
Denition 1.3.1: A connection in 1 is an assignment to each n 1 a com-
plimentary subspace H
u
to G
u
T
u
1, called the horizontal subspace, that for all
n 1 satises:
(i) T
u
1 = G
u
H
u
.
(ii) H
ua
= (1
a
)

H
u
for all o G.
(iii) H
u
depends smoothly on n.
In words a connection is a G-equivariant choice of compliment to the vertical
that varies smoothly with n. If A T
u
1 then A and /A denotes the vertical
and horizontal components of A, respectively. We can describe a connection in
fancierterminology as follows. We dene the vertical subbundle of T1 by 11 =

uP
G
u
. and the horizontal subbundle by H1 =

uP
H
u
. Then we have an exact
sequence of G-modules
0 11 T1 O 0.
Then a connection in 1 is a splitting of this exact sequence as G-modules, and thus
gives a decomposition of the tangent bundle T1 as G-modules, viz.
T1 = 11 H1.
For a Lie group G acting smoothly on a manifold ` there is well-known homo-
morphism from the Lie algebra g of G to the Lie algebra A(`) of smooth vector
elds on `. and if G acts eectively (which we assume) this is a monomorphism.
Given g we let

denote its image in A(`). If we specialize to the case of the

right action of G on a principal G-bundle 1, the vector eld

is tangent to the
bres
1
((n)) at each point n 1, and is called the fundamental vertical vector
eld on 1 associated to g. Evaluation of

at a point n 1 gives a vector

space isomorphism of the Lie algebra g with the vertical tangent space G
u
at n.
Denition 1.3.2: Given a connection on 1, we dene the connection form
associated to to be the 1-form on 1 with values in the Lie algebra g by setting
(A) equal to the unique g such that (

)
u
is the vertical component of A.
Clearly, satises (A) = 0 if A is horizontal. In the sequel we are particularly
interested in the case of a circle bundle. In this case the Lie algebra g R. and
it is common to take 1 as a generator of g. Actually a connection 1-form denes a
connection as is seen by the following
Proposition 1.3.3: The connection form satises
(i) (

) = for all g.
(ii) 1

a
= od
a
1 for all g.
Conversely, given a g valued 1-form on 1 which satises conditions (i) and (ii),
there is a unique connection in 1 whose connection form is .
20 1. STRUCTURES ON MANIFOLDS
We shall often refer to a connection 1-form as simply a connection. By
localizing and using partitions of unity one has
Theorem 1.3.4: Every principal bundle 1(`. G) admits a connection.
Let 1(`. G) be a principal bundle and \ be a nite dimensional vector space
over F = R. C or the quaternions
1
Q. Let
r
(1. \ ) denote the set of smooth :-forms
on 1 with values in \. that is the set of smooth sections of the bundle
r
(1)\. This
set has some natural algebraic structures. It is an innite dimensional vector space
over F as well as a C

(1)
R
F module. Consider a representation : G nt \
on \ . A pseudotensorial form of degree : and of type (. \ ) is :-form on 1 with
values in \ such that
1

a
= (o
1
) .
is called tensorial if is pseudotensorial and satises (A
1
. . . . . A
r
) = 0 if at
least one of the vector elds A
1
. . . . . A
r
is vertical. In particular, a pseudotenso-
rial 0-form, which is automatically tensorial, is just a smooth G-equivariant map
from 1 to \. We denote by T
r
(1. \ ) the closed subspace of tensorial r-forms. Now
suppose that is a connection in 1(`. G). Then the connection form of is
pseudotensorial of type (od. g), but not tensorial. Notice, however, Proposition
1.3.3 implies that the dierence
t
of any two connections is tensorial. Given
any pseudotensorial form we can dene a tensorial form by taking its horizontal
projection, i.e., /(A
1
. . . . . A
r
) = (/A
1
. . . . . /A
r
). Generally d is only pseu-
dotensorial, even if is tensorial. However, the exterior covariant derivative 1
dened as the horizontal projection
(1.3.1) 1(A
1
. . . . . A
r+1
) = d(/A
1
. . . . . /A
r+1
)
is always tensorial. In particular, if is a connection 1-form, 1 is a tensorial 2-
form of type (od. g) called the curvature 2-form of and usually denoted by . It is
a smooth section of
2
(1) g. and satises the famous Cartan structure equations
(1.3.2) = 1 = d +
1
2
[. ].
as well as the well-known Bianchi identities
(1.3.3) 1 = 0.
The meaning of the bracket in the Equation (1.3.2) is [. ](A. Y ) = [(A). (Y )].
Generally, the exterior covariant derivative of any tensorial form is
1 = d + [. ].
A connection or is said to be at if its curvature vanishes.
Tensorial forms can be described alternatively in terms of associated bundles.
Let 1 = 1
G
\ be the F-vector bundle on ` with standard bre \ and associated
to the F-representation of G. Then there is a 1-1 correspondence between tensorial
:-forms T
r
(1. \ ) of type (. \ ) and smooth sections of the bundle
r
(`) 1 as
follows: By xing a point n 1 the natural projection 1 \ 1 gives an F-
linear isomorphism from the vector space \ to the bre 1
(u)
. Then if A
1
. . . . A
r
are tangent vectors at (n) ` we dene the section

of
r
(`) 1 by

(A
1
. . . . . A
r
) = n(A

1
. . . . . A

r
).
1
We shall need all three types of vector spaces in the sequel. Some care must be taken
when working with quaternionic vector spaces, due to the noncommutativity of Q. For example,
quaternionic vector spaces are considered by multiplication from the left.
1.3. CONNECTIONS IN PRINCIPAL AND ASSOCIATED VECTOR BUNDLES 21
where A

denotes any vector on 1 that projects to A on `. One checks that

is independent of the choices made. In particular, smooth G-equivariant functions

1\ correspond to smooth sections of 1.
Now a connection in 1 induces a connection in the vector bundle 1 and,
more generally, in the vector bundles
r
(`) 1. We denote the C

(`)-module
of smooth sections of
r
(`) 1 by
r
(`. 1).
Denition 1.3.5: A (Koszul) connection on an F-vector bundle 1 is an F-
linear map on sections: : (1)
1
(`. 1) satisfying the Leibnitz rule (1:) =
1: +d1 :. where 1 C

(`) and : is a smooth section of 1.

Given a connection on 1(`. G) and an associated vector bundle 1 = 1
G
\.
we can dene a connection on 1 by setting
= n1n
1
.
Conversely, suppose we have an F-vector bundle 1 with associated principal bundle
1(`. G) and F-representation of G on the standard bre \ so that 1 = 1
G
\.
Then given a connection on 1. we obtain a connection on 1 as follows: Let
s = (:
1
. . . . . :
k
) be a G-frame of local sections of 1. If A T
x
`. then the subspace
H
u
of 1
u
with (n) = r dened by
H
u
= s

A [
X
s = 0
is a G-equivariant complement to the vertical subspace G
u
. This denes the con-
nection in 1. The Koszul connection in the associated bundles is often referred
to as the covariant derivative, and a section : of 1 is said to be covariantly constant
if : = 0.
As in Proposition 1.3.3 it is convenient to express the connection in terms of
a 1-form. However, this can only be done locally in terms of a local trivialization
of the vector bundle 1. Given a local trivialization 1[
U
l \. we can write
(1.3.4) [
U
= d +
U
.
where the exterior derivative d represents the at connection on l and
U
is a g-
valued 1-form on l. Given another such trivialization on the open set \ `. and
transition functions o : l \G relating the two trivializations in the overlap,
one obtains the relation
(1.3.5)
W
= o
1

U
o +o
1
do.
Conversely, given a cover of ` by open sets l

with g-valued 1-forms

U
on each
open set in the cover satisfying Equation (1.3.5) in the overlaps, one can reconstruct
the connection on 1. This formulation is essentially Cartans denition of con-
nection, whereas, the formulation given in Proposition 1.3.3 is due to his student
Ehresmann.
There is a natural extension of Equation (1.3.4) to the bundles
r
(`) 1. We
shall use the notation in [DK90] and write the exterior covariant derivative as
d
A
= d + :
r
(`. 1)
r+1
(`. 1).
Here the symbol denotes a family of g-valued 1-forms on open sets of ` satisfying
Equation (1.3.5) which act linearly via the representation

: g1nd \ on the
local sections of 1 obtained from a local trivialization of 1. The curvature form
22 1. STRUCTURES ON MANIFOLDS
= 1 on 1 corresponds by the isomorphism n to the smooth section 1
A
= d
A

of
2
(`) g. The Cartan structure Equations (1.3.2) then take the form
(1.3.6) 1
A
= d
A
= d+ .
where here again we follow the convention in [DK90] using the wedge product in
lieu of the brackets to emphasize as an endomorphism of \ via local trivializa-
tions.
Exercise 1.5: Show that if we write the g-valued connection 1-form as =

i
dr
i
and its curvature 2-form 1
A
=

1
ij
dr
i
dr
j
in a local coordinate chart
(l; r). Equation (1.3.6) can be written as
1
ij
=

j
r
i

i
r
j
+ [
i
.
j
].
A connection in 1(`. G) allows us to dene the notion of parallel translation
of a bre of 1 along any piecewise smooth curve in `. This is done as follows: Let
r
0
. r
1
` and let : [0. 1]` be a curve in ` with (0) = r
0
and (1) = r
1
.
Now at each point n 1 there is a vector space isomorphism : T
(u)
`H
u
.
So xing n
0

1
(r
0
) we can lift horizontally to a unique piecewise smooth
curve

in 1 such that

(0) = n
0
and

(t) = ( (t)) for all t [0. 1]. This gives

an isomorphism of bres
1
((t))
1
(r
0
). called parallel translation along .
Now suppose that is a loop at r
0
, then parallel translation gives automorphisms
of the bre
1
(r
0
). By composing loops and running the loop backwards, we see
that the set of such automorphisms form a group. Moreover, if we x a point
n
1
(r
0
) this group can be identied with a connected Lie subgroup of G,
called the holonomy group through n and denoted by Hol
u
. If we restrict ourselves
to loops at (n) that are null homotopic, then we obtain a normal subgroup Hol
0
u
of Hol
u
known as the restricted holonomy group through n. The groups Hol
0
u
and
Hol
u
enjoy the following properties:
(i) If n. 1 can be joined by a horizontal curve then Hol
u
= Hol
v
and
Hol
0
u
= Hol
0
v
.
(ii) If = no for o G. then Hol
u
= Ad
a
Hol
v
and H
0
(n) = Ad
a
Hol
0
u
.
A fundamental result in the theory of holonomy groups is the so-called Reduc-
tion Theorem which we now state. Of course, we refer to [KN63] for its proof:
Theorem 1.3.6: Let ` be a smooth connected manifold and 1(`. G) be a prin-
cipal G-bundle with a connection . Let n 1 be an arbitrary point and let 1(n)
denote the subset of points in 1 that can be joined to n by a horizontal curve. Then
1(n) is a reduced subbundle of 1 with structure group Hol
u
, and the connection
restricts to a connection on 1(n).
The subbundle 1(n) is called the holonomy bundle through n. and we call such
a connection reducible. It has its values in the Lie algebra hol
u
of Hol
u
. A well-
known theorem of Ambrose and Singer [AS53] characterizes the Lie algebra hol
u
as precisely the Lie algebra spanned by the curvature
v
(A. Y ). where 1(n)
and A. Y are arbitrary horizontal vectors at . Conversely, if Q(`. H) 1(`G)
is principal subbundle corresponding to a Lie subgroup H G, a connection in
a principal bundle Q(`. H) can be extended to a connection in 1(`. G).
1.4. G-STRUCTURES 23
1.4. G-Structures
In this section we describe structures on manifolds from the unifying viewpoint
of G-structures. There are several texts where this point of view is expounded
[Kob72, Mol77, Sal89, Ste83]. Here we consider only rst order G-structures,
that is, reductions of the bundle of linear frames on `. Here is the precise denition.
Denition 1.4.1: Let G G1(n. R) be a subgroup, then a G-structure on ` is
a reduction of the frame bundle 1(`) to the subgroup G. The G-structure is said
to be integrable if every point of ` has a local coordinate chart (l; r) such that
the local section

r
1
. . . . .

r
n

of 1(`) is a local section of the reduced bundle 1(`. G). Such a coordinate chart
is called admissible.
Let us return to the bundle of linear frames 1(`) of a smooth manifold `.
On 1(`) there is a canonical R
n
-valued 1-form dened as follows: as seen in
Example 1.2.3 we can view any point n 1(`) as a vector space isomorphism
n : RT
(u)
`. So can be dened by
(1.4.1) '. A` = n
1

A.
where A T
u
1(`) and '. ` denotes the natural pairing between the tangent bun-
dle to the bundle of linear frames T1(`) and its dual cotangent bundle T

1(`).
If 1(`. G) is a G-structure on `. i.e., a subbundle of 1(`). then we can restrict
to 1(`. G) giving a canonical 1-form on the G-structure. The canonical 1-form on
a G-structure 1(`. G) behaves functorially under the action of the general linear
group G1(n. R) on 1(`). Indeed we have
Lemma 1.4.2: For any o G G1(n. R) the transformation rule holds:
1

a
= o
1
.
Proof. For any vector eld A on 1(`. G), we have
'1

a
. A` = '. 1
a
A` = (no)
1

1
a
A = o
1
n
1

A = 'o
1
. A`.
Generally, there are topological obstructions to the existence of G-structures.
To see that such a reduction does not always exist let G = c the identity group.
Then an c-structure on ` is nothing but a global frame or parallelism of `. But
it is well-known that global frames do not always exist, that is that ` may not
be parallelizable. For example, the 2-sphere o
2
does not even have one nowhere
vanishing vector eld let alone a global frame. Even if there is a G-structure
on ` there may not be an integrable G-structure. For example, as seen below
every manifold ` admits many O(n. R)-structures, but a compact ` admits an
integrable O(n. R)-structure only if ` is covered by a torus (see Example 1.4.7
below). On the other hand every G1(n. R) structure on ` is integrable. The
following proposition is evident.
Proposition 1.4.3: A G-structure 1(`. G) is integrable if and only if there is an
atlas of coordinate charts (l

; r
()
)
I
on ` whose Jacobian matrices

x
()
i
x
()
j

i,j
lie in G at all points of l

.
24 1. STRUCTURES ON MANIFOLDS
Recall from Section 1.3 that associated to any connection in a principal bundle
there is the fundamental curvature 2-form. So for any connection in 1(`), usually
called a linear connection, we have its curvature = 1. However, since the linear
frame bundle 1(`) has a canonical 1-from associated to it, we have another
2-form associated to any linear connection . namely = 1. called the torsion
2-form. In the case of linear connections we can add to Cartans structure Equation
(1.3.2) the so-called First Structure Equation
(1.4.2) d + = .
We can also add to the Bianchi identities (1.3.3), the First Bianchi identities
(1.4.3) 1 = .
Exercise 1.6: Show that for linear connections the usual expressions for the torsion
and curvature tensors
T(A. Y ) =
X
Y
Y
A[A. Y ]. 1(A. Y )7 =
X

Y
7
Y

X
7
[X,Y ]
7
are related to the corresponding Cartan expressions by
T(A. Y ) = n(2(A

. Y

)). 1(A. Y )7 = n(2(A

. Y

))(n
1
7).
where A

. Y

are the horizontal lifts of the vectors A. Y T

(u)
`, respectively,
and n 1(`) is any point.
This entire discussion holds for any G-structure 1(`. G) 1(`) with a con-
nection . We shall often refer to a linear connection in a G-structure 1(`. G) as
a G-connection. A linear connection with = 0 is said to be torsion-free. Clearly
the notions of parallel translation and holonomy apply to the case of G-structures.
We now want to put Theorem 1.2.5 to work by seeing how certain natural
tensor elds dene G-structures on `. Suppose that T
0
is an element of the tensor
algebra T (R
n
) over R
n
and that G is the largest closed Lie subgroup of G1(n. R)
that leaves T
0
invariant. Viewing a point n 1(`) as a vector space isomorphism
n : R
n
T
(u)
`. we obtain an induced isomorphism n

: T (R
n
)T (T
(u)
`) of
the tensor algebras. Then the image T = n

T
0
is a section of the tensor algebra
bundle T (`). If T
0
is a tensor of type (:. :) then T is a section of the tensor bundle
T
r
s
`. In any case because of the invariance of T
0
under G, the tensor eld T denes
a section of the associated bundle 1(`)G. In this case we say that 1(`. G) is a
G-structure dened by the tensor T
0
.
Proposition 1.4.4: Let 1(`. G) be a G-structure dened by the tensor T
0
. Then
1(`. G) is integrable if and only if there exists an atlas of charts (l

; r
()
)
I
on ` such that the corresponding tensor eld T = n

T
0
has constant components
on l.
Proof. () Let 1(`. G) be integrable and let (l; r) be a coordinate chart,
then the frame n = (

x1
. . . . .

xn
) belongs to 1(`. G). Let c
i
denote the standard
basis for R
n
and c
j
its dual basis. Then

xi
= n(c
i
). and dr
j
= n(c
j
). So if
T
0
= t
i
1
i
r
j1js
c
i
1
c
i
r
c
j1
c
js
is a tensor of type (:. :), we have
T = n

T
0
= t
i
1
i
r
j1js

r
i1


r
ir
dr
j
1
dr
j
s
.
Hence, T has constant components on l.
1.4. G-STRUCTURES 25
() Conversely, suppose that (l; r) is a coordinate chart such that T = n

T
0
has constant components. So T is a constant section of the associated bundle
1(l)G. Then there is a linear transformation of coordinates on l to new coor-
dinates (n
1
. . . . . n
n
) such that the frame n = (

y
1
. . . . .

y
n
) belongs to 1(`. G).
Thus, the G-structure 1(`. G) is integrable.
One easily sees that if a G-structure 1(`. G) is dened by a tensor eld T and
is a G-connection, then T is covariantly constant with respect to . or equivalently,
T = 0. Generally, the integrability condition of Denition 1.4.1 can be very
restrictive, and we will discuss various levels of integrability. First, it is easy to
show that
Proposition 1.4.5: An integrable G structure admits a torsion-free connection.
It is now convenient to consider a much less restrictive denition of integrability.
Denition 1.4.6: A G-structure 1(`. G) is said to be 1-integrable if it admits
a torsion-free G-connection.
So an integrable G-structure is automatically 1-integrable. The failure of the
existence of torsion-free G-connection can be seen as the rst order obstruction to
integrability, hence, the name 1-integrable. The question of uniqueness of a torsion-
free connection, assuming one exists, is related to prolongations of G-structures
which we shall briey treat below (See Denition 1.6.8). We now wish to consider
many examples of G-structures. Our rst example is a good example where non-
integrable G-structures are of more interest than the integrable ones.
Example 1.4.7: Riemannian metrics. We consider a reduction of the frame
bundle 1(`) to the orthogonal group
O(n. R) = G1(n. R) [
t
= 1l
n
.
The reduced bundle O(`) 1(`) is called the orthonormal frame bundle of `.
Theorem 1.2.5 says that such a reduction is equivalent to a choice of section of the
associated bundle 1(`. G1(n. R)O(n. R). 1(`)) = 1(`)O(n. R). We show that
such a section is just a Riemannian metric on `. As mentioned previously each
point n 1(`) gives an isomorphism of the standard vector space R
n
with the
tangent space T
(u)
`. Let '. ` denote the Euclidean inner product on R
n
. and let
A. Y T
(u)
`. Then
o
u
(A. Y ) = 'n
1
A. n
1
Y `
denes an inner product on T
(u)
`. Furthermore, if o O(n. R) we have
o
ua
(A. Y ) = '(no)
1
A. (no)
1
Y ` = 'o
1
n
1
A. o
1
n
1
Y `
= 'n
1
A. n
1
Y ` = o
u
(A. Y ).
where the second to the last equality holds by the invariance of '. ` under O(n. R).
This shows that o
u
is constant along the bres of O(`). and thus, is a section of
1(`)O(n. R). Thus, the associated bundle 1(`)O(n. R) can be identied with
a subbundle of the vector bundle Sym
2
T

` of symmetric covariant 2-tensors on

`. The choice of n modulo O(n. R) corresponds to a choice of Riemannian metric
on `. Since manifolds are paracompact the standard partition of unity argument
shows that such sections and hence, such reductions always exist.
When are O(n)-structures integrable? According to Proposition 1.4.4 this oc-
curs when the metric tensor o has constant components in some coordinate chart
26 1. STRUCTURES ON MANIFOLDS
(l; r). and then by a change of coordinate, say (\ ; n) the metric can be brought
to the form
o =

i
(dn
i
)
2
in \. By the Second Fundamental Theorem of Riemannian geometry this happens
precisely when the Riemann curvature tensor vanishes. For example, if ` is a com-
pact 2-dimensional manifold, then it is either a torus or a Klein bottle depending
on whether it is orientable or not. This is quite restrictive. On the other hand
the First Fundamental Theorem of Riemannian geometry says that there exists
a unique torsion-free Riemannian connection, denoted
g
, called the Levi-
Civita connection. So all Riemannian G-structures are 1-integrable. Similarly, a
reduction to the group oO(n. R) = O(n. R) G1
+
(n. R) corresponds to oriented
Riemannian geometry. In particular, one can consider the parallel translation de-
ned by the Levi-Civita connection and its associated holonomy group which is a
subgroup of the structure group O(n. R) (oO(n. R) in the oriented case). Since this
connection
g
is uniquely associated to the metric o, we denote it by Hol(o), and
refer to it as the Riemannian holonomy group or just the holonomy group when
the context is clear. Indeed, it is precisely this Riemannian holonomy that plays
an important role in this book. Now on a Riemannian manifold (`. o) there is a
canonical epimorphism
1
(`)Hol(o)Hol
0
(o). in particular, if
1
(`) = 0 then
Hol(o) = Hol
0
(o). In 1955 Berger proved the following theorem [Ber55] concerning
Riemannian holonomy:
Theorem 1.4.8: Let (`. o) be an oriented Riemannian manifold which is neither
locally a Riemannian product nor locally symmetric. Then the restricted holonomy
group Hol
0
(o) is one of the following groups listed in Table 1.4.1.
Table 1.4.1: Bergers Riemannian Holonomy Groups
Hol
0
(o) dim(`) Geometry of ` Comments
oO(n) n orientable Riemannian generic Riemannian
l(n) 2n Kahler generic Kahler
ol(n) 2n Calabi-Yau Ricci-at Kahler
oj(n) oj(1) 4n quaternionic Kahler Einstein
oj(n) 4n hyperkahler Ricci-at
G
2
7 G
2
-manifold Ricci-at
ojin(7) 8 ojin(7)-manifold Ricci-at
We will encounter all the geometries listed in this table throughout this book.
Most of them will already be introduced in this chapter as G-structures. Orig-
inally, Bergers list included ojin(9) but Alekseevsky proved that any manifold
with such holonomy group must be symmetric [Ale68]. In the same paper Berger
also claimed a classication of all holonomy groups of torsion-free ane (linear)
connections that act irreducibly. He produced a list of possible holonomy rep-
resentations up to what he claimed was a nite number of exceptions. But his
classication had some gaps discovered 35 years later by Bryant [Bry91]. An in-
nite series of exotic holonomies was found in [CMS96] and nally the classication
in the non-Riemannian ane case was completed by Merkulov and Schwachhofer
[MS99]. We refer the reader to [MS99] for the proof, references and the history of
the general ane case. In the Riemannian case a new geometric proof of Bergers
1.4. G-STRUCTURES 27
Theorem is now available [Olm05]. An excellent review of the subject just prior
to the Merkulov and Schwachhofers classication can be found in [Bry96]. We
should add that one of the rst non-trivial results concerning manifolds with the
exceptional holonomy groups of the last two rows of Table 1.4.1 is due to Bonan
[Bon66] who established Ricci-atness of manifolds with parallel spinors.
It is clear in the Riemannian case that non-integrable structures are much
more interesting than integrable ones, and there is a big gap between the integrable
structures and the 1-integrable structures. The obstructions to integrability occur
at order two, namely the Riemannian curvature. In order to study these further one
needs to study certain invariants of the G-structure which involves a detailed study
of the Riemannian curvature. For a general G-structure these invariants involve
Spencer cohomology to which we refer the reader to the literature [Fuj72, KS72,
SS65, Ste83]. In the case of O(n. R)-structures the relevant Spencer cohomology
group consists of the Riemann curvature tensor. In our next example the integrable
case is very rich.
Example 1.4.9: Almost complex structures. A complex structure on a real
vector space \ is an endomorphism J of \ that satises J
2
= 1l. where 1l denotes
the identity endomorphism of \. We can give \ the structure of a complex vector
space by dening scalar multiplication with scalars in C by
(o +i/) = o +/J.
Conversely, if \ is a complex vector space we dene the endomorphism J on \ by
J = i.
Let
1
. . . . .
n
be basis for \ as a complex vector space, then it is easy to check
that
1
. . . . .
n
. J
1
. . . . . J
n
is a basis for \ as a real vector space. In particular,
this implies that any real vector space with a complex structure has even real
dimension. For example, by writing .
j
= r
j
+ in
j
for , = 1. . . . . n, the complex
vector (.
1
. . . . . .
n
) in the complex vector space C
n
is identied with the real vector
(r
1
. . . . . r
n
. n
1
. . . . . n
n
) in the real vector space R
2n
. The canonical complex structure
J
0
on \ is given in block form with respect to the standard basis of R
2n
by
J
0
=

0 1l
n
1l
n
0

.
where 1l
n
denotes the n n identity matrix. We can identify the complex linear
group G1(n. C) with the subgroup of matrices in G1(2n. R) that commute with J
0
.
Thus, the complex n n matrix +i1 is identied with the real 2n 2n matrix

1
1

.
On a manifold ` an almost complex structure is an endomorphism J
x
of the
tangent space T
x
` that varies smoothly with r and satises J
2
x
= 1l
x
at each
point r `. or in other words, a smooth section J of the endomorphism bundle
End T` satisfying J
2
= 1l. As in the Riemannian case such a section can be
obtained from the canonical complex structure J
0
on R
2n
by identifying n 1(`)
with a linear map n : R
2n
T
(u)
`. Thus, we dene
J
u
= nJ
0
n
1
.
28 1. STRUCTURES ON MANIFOLDS
The invariance of J
0
under the subgroup G1(n. C) G1(2n. R) shows that J
u
is
constant along the bres of the bundle C1(`) of complex linear frames on ` and
thus, gives a section of the associated bundle 1(`. G1(2n. R)G1(n. C). 1(`)) =
1(`)G1(n. C) which is the subbundle of End T` consisting of endomorphisms
that satisfy J
2
= 1l. Unlike in the Riemannian case it is not always possible to nd
such a section, or equivalently to obtain a reduction of the real frame bundle 1(`)
to the complex linear group G1(n. C). In general, this is a fairly deep topological
question. The existence of an almost complex structure J on ` is equivalent to
the splitting of the complexied tangent bundle T` C = T
1,0
` T
0,1
`. If
one has such a splitting, J can be dened to be multiplication by i on T
1,0
`
and multiplication by 1 on T
0,1
`. Conversely, given J one easily sees that it is
diagnonalizable over T` C with eigenvalues i.
By Proposition 1.4.3 a G1(n. C)-structure is integrable when there is an atlas
of coordinate charts l

; r
()
1
. . . . . r
()
n
. n
()
1
. . . . . n
()
n

I
whose Jacobian matrix
lies in G1(n. C) G1(2n. R). This corresponds to the complex coordinates .
()
j
=
r
()
j
+in
()
j
being independent of the complex conjugate coordinates .
()
i
. that is the
transition functions are holomorphic or equivalently satisfy the Cauchy-Riemann
equations. An integrable almost complex structure is called a complex structure,
and a smooth manifold ` with a complex structure is called a complex manifold.
The structure tensor of the G-structure that measures the failure of integrability is
the Nijenhuis torsion tensor dened by
(1.4.4)
J
(A. Y ) = [JA. JY ] [A. Y ] J[A. JY ] J[JA. Y ].
The vanishing of this tensor eld is equivalent to the integrability of the almost com-
plex structure. If one assumes that ` is real analytic this follows from a Frobenius
integrability argument (cf. [KN69], Appendix 8). However, in the C

case this is
a much deeper result due to Newlander and Nirenberg [NN57]. Summarizing an
almost complex structure is integrable if and only if the Nijenhuis torsion tensor

J
vanishes. This is indeed the torsion of an almost complex connection, so the
only obstruction to integrability occurs at level one, that is, in this case 1-integrable
implies integrable.
Following up on the remark at the end of the second paragraph of Example
1.4.9, one can ask why a partition of unity argument doesnt work in the case of
almost complex structures. The point is that one cannot guarantee the condition
J
2
= 1l at every point of the manifold by using partitions of unity. A better under-
standing is obtained from the point of view of G-structures. An alternative proof
of the existence of Riemannian metrics on any manifold is as follows: by Theorem
1.2.5 giving a Riemannian metric is equivalent to giving a reduction of the frame
bundle to the orthogonal group O(n. R). This reduction can be accomplished (gen-
erally over any paracompact space) since as topological spaces the general linear
group G1(n. R) decomposes as O(n. R) R
n(n+1)
2
. the so-called Iwasawa decompo-
sition. Thus, G1(n. R) is homotopy equivalent to O(n. R). Then paracompactness
allows one to construct this homotopy equivalence at each point of ` giving the
reduction. However, in the case of reducing 1(`) to the complex linear group
G1(n. C) there is no such homotopy equivalence. Indeed, G1(2n. R) is homotopy
equivalent to O(2n. R). whereas, again the Iwasawa decomposition for G1(n. C)
says that G1(n. C) is homotopy equivalent to the unitary group l(n). that is, the
subgroup of G1(n. C) that leaves invariant the standard Hermitian form '. ` on
1.4. G-STRUCTURES 29
C
n
. It is well-known that O(2n. R) and l(n) are not homotopy equivalent, e.g.
they have dierent homology groups. So, generally, there are obstructions to such
reductions. A general theory of obstructions is expounded in the classic text of
Steenrod [Ste51]. A well-known example of an even dimensional manifold which
admits no almost complex structure is the 4-sphere o
4
. This can be shown either by
obstruction theory or using the Atiyah-Singer index theorem. Another interesting
example is o
6
which is known to admit an almost complex structure, but it is still
an open question whether o
6
has an integrable almost complex structure, i.e., a
complex structure.
Exercise 1.7: Let G = G1
+
(n. R) be the subgroup of matrices in G1(n. R) whose
determinant is positive. Show that ` admits such a G-structure if and only if `
is orientable. Show that a G1
+
(n. R)-structure on ` coincides with a choice of
orientation of `. What about integrability?
Exercise 1.8: Show that an almost complex structure on ` determines an orien-
tation on `. hence, any almost complex manifold is orientable.
Notice that an oriented Riemannian manifold gives a reduction of 1(`) to the
special orthogonal group oO(n) = O(n) [ det() = 1. and that oO(n) =
G1
+
(n. R) O(n). We will frequently use G-structures that occur as a combination
of other G-structures or, alternatively, a G-structure dened by more than one
tensor. Another important example of this is:
Example 1.4.10: Almost Hermitian structures. Let G = l(n) the unitary
group dened by
l(n) = G1(n. C) [

t
= 1l
n
.
This is precisely the group that leaves the standard Hermitian form 'n. ` = n
1

1
+
+ n
n

n
in C
n
invariant. Since l(n) = G1(n. C) O(2n), a l(n)-structure or
almost Hermitian structure consists of an almost complex structure J together with
a Riemannian metric o on ` satisfying the compatibility condition o(JA. JY ) =
o(A. Y ). The two tensor elds o and J give rise to a 2-form called the Kahler
form dened by (A. Y ) = o(A. JY ). The structure is called almost Kahler if is
closed, Hermitian if J is integrable, and it is called Kahler if both J is integrable
and is closed. A l(n)-structure is integrable if both the almost complex structure
and the Riemannian structures are integrable which is very restrictive. The only
compact examples are covered by a complex torus. A less restrictive notion is that
of being 1-integrable. A l(n)-structure is 1-integrable if and only if it is Kahler.
This is equivalent to the l(n) bundle coinciding with the holonomy bundle. We
shall discuss these structures in much more detail in Chapter 3.
These ideas can also be applied to indenite or pseudo-Riemannian and pseudo-
Hermitian metrics
2
. For example,
Example 1.4.11: Pseudo-Riemannian structures. Let O(j. c) denote the sub-
group of G1(n. R) that leaves the quadratic form n
2
1
+ +n
2
p
n
2
p+1
n
2
p+q
invariant. Then an O(j. c)-structure on ` is the same as a choice of pseudo-
Riemannian metric o on ` of signature (j. c). If j or c equals 1, then this is called
a Lorentzian structure and o a Lorentzian metric. Similarly, one has almost pseudo-
Hermitian structures by considering the pseudo-unitary groups l(j. c) which is the
2
One should be cautioned that the terminology pseudo-Hermitian structure is used in an
entirely dierent sense in [Web78]
30 1. STRUCTURES ON MANIFOLDS
group that leaves invariant an indenite Hermitian form of signature (j. c). Clearly,
l(j. c) = G1(j + c. C) O(2j. 2c). The integrability question is similar to that of
Example 1.4.7.
Unlike Riemannian metrics (structures), pseudo-Riemannian metrics do not
always exist on `. See for example [HE73, ON83] for the existence of Lorentzian
metrics on `. In fact, it is easy to see that ` admits a Lorentzian metric if and
only if it admits a nowhere vanishing vector eld.
Exercise 1.9: Dene the group
CO(n) = G1(n. R) [
t
= c1l
n
. c 0 .
Show that a CO(n)-structure on ` coincides with a conformal class of Riemannian
metrics on `. Recall that a conformal class is an equivalence class of Riemannian
metrics, where the Riemannian metrics o and o
t
are equivalent if there exists a
positive function on ` such that o
t
= o.
A CO(n)-structure on ` is called a conformal structure on `; however, CO(n)
is usually not called the conformal group, a name that is usually reserved for the
automorphism group of a conformal structure, and this group is generally larger
than CO(n). This phenomenon is related to prolongations of G-structures which is
briey touched upon below. The reader is referred to [Kob72, Ste83] for complete
discussions of prolongations of G-structures. Conformal structures can be dened
similarly in the case of pseudo-Riemannian structures. Our next example will also
be of fundamental importance to us.
Example 1.4.12: Almost symplectic structures. Consider the non-degenerate
antisymmetric bilinear form

0
(u. v) = n
1

n+1
+ +n
n

2n
n
n+1

1
n
2n

n
on the vector space R
2n
. Notice that
0
can also be written in terms of the standard
Euclidean metric '. ` and canonical complex structure J
0
as

0
(u. v) = 'u. J
0
v`.
The subgroup of G1(2n. R) that leaves
0
invariant is called the real symplectic
group and denoted
3
by oj(n. R). It is easy to see that this group can be dened by
oj(n. R) = G1(2n. R) [
t
J
0
= J
0
.
An almost symplectic structure on a manifold ` is then given by transporting
the antisymmetric form
0
to the manifold as before by the linear isomorphism
n : R
2n
T
p
`. viz. for A. Y T
p
` we dene
(A. Y ) =
0
(n
1
A. n
1
Y ).
is a section of the bundle 1(`. G1(2n. R)oj(n. R). 1(`)) = 1(`)oj(n. R)
which can be identied with the subbundle of the exterior bundle
2
` bundle con-
sisting of non-degenerate 2-forms. By Theorem 1.2.5 this corresponds to a reduction
of 1(`) to the real symplectic group oj(n. R). The appearance of the tensor J
0
sug-
gests that there may be a connection between almost complex structures and almost
symplectic structures. This is indeed the case for oj(n. R) O(2n. R) = l(n) =
G1(n. C) O(2n. R) Since, an O(2n. R)-structure exists on any 2n-dimensional
3
The notation for this group is by no means standard. It is sometimes written as Sp(2n, R),
and is written as Sp(2n) in [MS98]. We shall employ the notation Sp(n) for the compact
symplectic group which is dierent.
1.4. G-STRUCTURES 31
manifold, we see that `
2n
admits an almost symplectic structure if and only if it
admits an almost complex structure. Of course, the two structures are dierent.
By Proposition 1.4.4 an oj(n. R)-structure is integrable if and only if the tensor
eld has constant components with respect to some local coordinates. In this
case there are local coordinates (r
1
. . . . . r
n
. n
1
. . . . . n
n
) on ` such that takes the
form
=
n

i=1
dr
i
dn
i
.
A well-known theorem of Darboux says that an oj(n. R)-structure is integrable if
and only if the tensor eld is closed, that is d = 0. An integrable oj(n. R)-
structure on ` is called a symplectic structure. In this case the pair (`. ) is
called a symplectic manifold.
Let (`. ) be a symplectic manifold and consider the group S(`. ) of sym-
plectomorphisms on `. i.e., the subgroup of Di ` that leaves invariant. The
Lie algebra s(`. ) of S(`. ) is the subalgebra of vector elds A on ` such that

X
= 0. This equation implies that the 1-form A is closed. If it is also exact
then the there is a smooth function H on ` such that A = dH. In this case
the function H is called a Hamiltonian function and the A a Hamiltonian vector
eld. Notice that H is only dened up to a constant and if H
1
(`. R) = 0 then
every A s(`. ) is Hamiltonian. We refer to [LM87] for further development.
From the discussion above about Lie groups, it should be clear that an almost
Hermitian structure gives naturally an almost symplectic structure. So we recover
Example 1.4.10 as a particular case of an almost symplectic structure with an added
compatible Riemannian structure.
Exercise 1.10: Dene the conformal symplectic group by
Coj(n. R) = G1(2n. R) [
t
J
0
= cJ
0
. c R

.
where R

denotes R 0. Show that a Coj(n. R)-structure on ` coincides with

a choice of equivalence class of non-degenerate 2-forms, where the 2-forms and

t
are equivalent if there is a nowhere zero function on ` such that
t
= .
Such a structure can be called a conformal almost symplectic structure. Show that
if the Coj(n. R)-structure is integrable, then the non-degenerate 2-forms are closed
and is a constant. The resulting structure is known as a conformal symplectic
structure.
Example 1.4.13: j-dimensional distribution A j-dimensional distribution on
an n-dimensional manifold ` is a choice of j-dimensional subbundle 1 of the
tangent bundle T`. The group G that stabilizes 1 is given by
G1(j. c; R) =

0
C 1

[ G1(j. R). 1 G1(c. R). C

q,p

.
where
p,q
denotes the j by c real matrices, and j + c = n. An integrable
G1(j. c; R)-structure on ` is just a foliation T of `. that is, every point of `
has a foliated coordinate chart (l; ) with coordinates (r
1
. . . . . r
p
; n
1
. . . . . n
q
) such
that (

x1
. . . . .

xp
) spans 1[l. and the Jacobian matrix of the change of foliated
charts lies in G1(j. c; R). An atlas U = (l

)
I
of such foliated charts is
called a foliated atlas for `. Two foliated atlases U and V of the same codimen-
sion c correspond to the same foliation T if and only if they satisfy the coherence
property that the disjoint union U . V is a foliated atlas for T.
32 1. STRUCTURES ON MANIFOLDS
Globally, the foliation T gives a partition of ` into disjoint j-dimensional
immersed submanifolds L

. called the leaves of T. By the classical Frobenius The-

orem, a p-dimensional distribution is integrable if and only if it involutive, that is
the Lie bracket of any sections of the subbundle 1 is also a section of 1. This has a
dual formulation as follows: let 1
0
denote the subbundle of T

` that annihilates
1. Then the Frobenius Theorem says that 1 is integrable if and only if the ideal
1(1
0
) generated by the sections of 1
0
is closed under exterior dierentiation, i.e.,
1(1
0
) is a dierential ideal. Of particular interest in foliation theory is a study
of the transverse geometry and this is studied in more detail in Chapter 2. The
transverse geometry is the geometry of the space of leaves of the foliation dened
to be the quotient space dened by the equivalence relation that r n if r and
n lie on the same leaf. However, this can be very crudely dened as a topological
space, so one studies the transverse geometry directly on the foliated manifold.
Alternatively, a foliation can be described by patching together local submer-
sions. This is the point of view advanced by Haeiger [Hae70] and will be discussed
in Section 2.2. A (global) submersion (hence, a bre bundle) is a special case of a
foliation. It is often called also a simple foliation. More development of foliation
theory is given below in Chapter 2.
There are various special cases of Example 1.4.13 of special interest.
Example 1.4.14: Almost product structures. This is Example 1.4.13 with
C = 0. In this case the tangent bundle to ` splits as a direct sum of two subbun-
dles T` = T
+
T

. By Proposition 1.4.3 an integrable almost product structure

implies the integrability of the complementary subbundles. In particular, the uni-
versal covering

` is actually a product of two manifolds, one of dimension j and
the other of dimension c. One can easily see that an almost product structure is
equivalent to the existence of an endomorphism 1 End(T`) such that 1
2
= 1l.
Namely, given such 1 we can introduce the projection operators
1

=
1
2
(1l 1). 1
+
1

= 1

1
+
= 0. 1
+
+1

= 1l.
We get the desired spitting T` = T
+
T

by taking T

= 1

(T`). As in
the almost complex case the Nijenhuis tensor
P
(A. Y ) is the obstruction to the
integrability of the two distributions T

. A Riemannian almost product structure

is a further reduction of the structure group to O(j. R) O(c. R) which is not
obstructed, i.e., if a manifold admits an almost product structure it also admits an
almost product Riemannian metric structure. It follows that a Riemannian almost
product structure is a triple (`. . o), where is an almost product structure and
o is a Riemannian metric with the property
o(1A. 1Y ) = o(A. Y ).
A Riemannian almost product structure is said to be a locally Riemannian product
structure if 1 = 0, i.e., is parallel with respect to the Levi-Civita connection of
o. The condition 1 = 0 implies the integrability
P
(A. Y ) = 0. Note also that if
a manifold ` admits an almost product structure with group G1(j. R) G1(c. R)
then it also admits an O(j. c) pseudo-Riemannian structure.
Exercise 1.11: Let (`
n
. 1. o) be an almost product structure. Show that 1 is a
local O(n. R) matrix with all eigenvalues 1 and Tr(1) = 2c n, where c = rk(T
+
)
is the number of positive eigenvalues. Furthermore, show that one can choose a
local chart and a basis so that 1. 1

take the following simple forms

1.4. G-STRUCTURES 33
1 =

1l
q
0
0 1l
nq

. 1
+
=

1l
q
0
0 0

. 1

0 0
0 1l
nq

.
We are especially interested in almost product structures when j = 2: and
c = 1 which is related to almost contact structures treated in detail in Chapter 6.
Unfortunately, the terminology with regard to contact and almost contact struc-
tures is very misleading. For example, a contact structure is generally not an almost
contact structure in contrast to the usual use of this terminology; moreover both
of these structures are dened in [Gra59]! This was noted and xed in [Sto74], at
least for topological purposes, but we would like consistency with geometric struc-
tures as well. We shall refer to our structure as an almost contact G-structure to
distinguish it from an almost contact structure whose denition is by now too well
ensconced in the literature.
Example 1.4.15: Almost contact G-structure. Consider the group G dened
as follows:
G =

0
c /

[ Coj(:. R). / G1(1. R). c

1,2m

.
This G-structure picks out a codimension 1 subbundle with a conformal almost
symplectic structure on it. Here an integrable G-structure is related to what could
be called a conformal cosymplectic structure.
In the case of Example 1.4.15 we shall be more interested in the non-integrable
case. In Chapter 6 we shall describe its relation with almost contact and contact
structures.
Another structure that is related to both Examples 1.4.9 and 1.4.13 is a CR-
structure. Here the concept of an almost CR-structure is not standard. Indeed
the denition of a CR-structure varies as well. Some authors require that a CR-
structure be dened by a codimension one subbundle (cf. [Jac90]), the idea being
that of modelling a real hypersurface in C
n+1
. We shall be more general.
Example 1.4.16: Almost CR-structures. An almost CR-structure of codimen-
sion c is a G-structure with G given by
G =

0
C 1

[ G1(:. C). 1 G1(c. R). C

q,2m

.
where G1(:. C) is realized as a subgroup of G1(2:. R) as described in Example
1.4.9. This G picks out a 2:-dimensional subbundle 1 of T` with an almost
complex structure, that is a smooth section J of the endomorphism bundle End 1
satisfying J
2
= 1l
E
. We say that (1. J) denes a CR-structure on ` if for any
smooth sections A. Y of 1 the vector eld [JA. Y ] + [A. JY ] is a section of 1 and
the Nijenhuis tensor
J
, as dened in Example 1.4.9, vanishes. We are interested
in the case of codimension one, i.e., c = 1. If ` is orientable then there exists a
1-form on ` such that 1 = ker . The symmetric bilinear form 1(A. Y ) on 1
dened by 1(A. Y ) = d(A. JY ) is called the Levi form of the CR-structure. The
CR-structure is said to be strictly pseudoconvex if 1 is either positive or negative
denite. For an extensive treatment of CR-structures we refer the reader to the
recent book [DT06].
34 1. STRUCTURES ON MANIFOLDS
Example 1.4.17: 1-structures. An 1-structure of corank c is a special kind of
an almost product structure. It is a G-structure with G given by
G =

0
0 1

[ G1(n. C). 1 G1(c. R)

.
1-structures were rst introduced and studied by Yano [Yan63]. One can easily see
that the reduction implies an existence of a non-vanishing section 1 of the bundle
End(T`) with constant rank such that
1
3
+1 = 0.
Given 1 we dene two commuting projections 1
+
= 1
2
and 1

= 1
2
+ 1l, which,
in turn, give the splitting T` = T
+
T

. Clearly, 1
2
(1A) = 1(A) so that 1
denes and almost complex structure on the subbundle T
+
= Im(1) T`. In
particular, further reduction to the subgroup l(n) O(c. R) can be achieved by
choosing a locally product Riemannian metric o on ` which is compatible with 1
in the following natural sense
o(1A. Y ) = o(A. 1Y ).
Clearly, such a metric always exists and the triple (`. 1. o) is said to be a metric 1-
structure. When c = 0, an 1-structure is simply an almost complex structure, and
with a compatible metric it is an almost Hermitian structure of two earlier examples.
When c = 1 an 1-structure is a special case of an almost contact G-structure. In
particular, when c = 1 and the structure group reduces further to G1(n. C) 1 an
1-structure is a (strict) almost contact structure discussed in Chapter 6. It follows
that there exist a nowhere vanishing vector eld and its dual 1-form such that
[Bla76a]
(1.4.5) 1 = 1l + .
Note that in the latter case picking a compatible Riemannian metric, i.e., the
metric o which satises o(1A. 1Y ) = o(A. Y ) (A)(Y ) is always possible and
gives further reduction of the structure group to l(n) 1 which is the standard
denition of an almost contact (metric) structure due to Gray [Gra59]. When c 2
one can consider the so-called globally framed (metric) 1-structure as the further
reduction if the structure group to l(n) 1l
q
. The denition implies that the
bundle T

= Ker(1) is parallelizable, i.e., there exist global frame eld

1
. . . . .
s

spanning T

= Ker(1). Such structures were introduced and studied by Goldberg

and Yano [GY71, GY70] and Blair [Bla70]. In some literature they are also
called f.pk-structure (see the recent book [FIP04] for more details and references).
Example 1.4.18: Quaternionic structures. Quaternionic G-structures ap-
peared early on in the work of Libermann [Lib54], Berger [Ber55], Obata [Oba56],
and Martinelli [Mar59], and continued in [Ale68, Bon67, Gra69a, Wol65] to
cite a few. Our brief treatment here follows that of Salamon [Sal86] with further
development in Chapter 12. Let H
n
be a quaternionic vector space with n 1.
By writing each quaternionic coordinate as n
i
= n
0
i
+ n
1
i
i + n
2
i
j + n
3
i
k we natu-
rally identify H
n
R
4n
. Let G1(n. H) G1(4n. R) be a group of non-singular
quaternionic n n matrices and consider the product G1(n. H) G1(1. H). For
any (A. ) we can dene G1(n. H) G1(1. H)-action on H
n
by (u; (A. )) A u.
The central R

acts trivially and we denote the quotient group by G1(n. H)oj(1),

1.4. G-STRUCTURES 35
where oj(1) G1(1. H) is the subgroup of unit quaternions. In particular, it fol-
lows that G = G1(n. H)oj(1) is a subgroup of G1(4n. R), and one easily sees
that G1(n. H)oj(1) =

G1(n. H) oj(1)

Z
2
. An almost quaternionic struc-
ture on a manifold ` is a reduction of the frame bundle 1(`) to the subgroup
G1(n. H)oj(1). Theorem 1.2.5 gives a one-to-one correspondence between sections
of the associated vector bundle 1(`. G1(4n. R)G and reductions Q of the frame
bundle 1(`) to the group G. Moreover, in analogy with Example 1.4.9 we see that
sections of 1(`. G1(4n. R)G) correspond to the existence of a triple of local sec-
tions 1
1
. 1
2
. 1
3
of a rank 3 vector subbundle O End(T`) satisfying the algebra
of the quaternions. Hence, one can consider either a reduction to the bundle Q
or the existence of the subbundle O as an almost quaternionic manifold. In the
world of quaternionic geometry the usual notion of integrability is much too strong.
For example, the only simply connected compact manifold which admits a quater-
nionic atlas whose transition functions are in G1(n. H)oj(1) is the quaternionic
projective space HP(n) [Mar70, Som75]. More interesting quaternionic geome-
tries emerge by considering 1-integrable (recall Denition 1.4.6) almost quaternionic
G-structures. A 1-integrable almost quaternionic structure is called a quaternionic
structure. In the case of quaternionic structures torsion-free connections are not
unique. Such a connection [Opr77, Opr84] is called an Oproiu connection in
[AM96b] or a quaternionic connection.
One can always equip ` with a Riemannian metric o which preserves O, i.e.,
o(JY. JY ) = o(A. Y ) for any local section J of O. The triple (`. O. o) is then
called almost quaternionic Hermitian manifold. This corresponds to a further re-
duction of the bundle Q to its maximal compact subgroup oj(n)oj(1). An almost
quaternionic (Hermitian) manifold with a torsion-free connection preserving O is
called a quaternionic (Hermitian) manifold [Sal86]. In dimension greater than 4 an
almost quaternionic Hermitian structure is 1-integrable if and only if it is quater-
nionic Kahler [AM96b]. The usual denition of a quaternionic Kahler structure
(in dimension greater than 4) is that the Riemannian holonomy lies in oj(n)oj(1).
Quaternionic structures will be described in more detail in Chapter 12. In dimen-
sion 4 the denitions of quaternionic and quaternionic Kahler involve duality of
Weyls conformal curvature and will be dened in Chapter 12.
Example 1.4.19: Hypercomplex structure. A G1(n. H)-structure on a mani-
fold ` is called an almost hypercomplex structure. Thus, an almost hypercomplex
manifold is the special case of an almost quaternionic manifold, where the sub-
bundle O End(T`) admits a trivialization by global sections I 1
1
. 1
2
. 1
3

satisfying quaternion algebra. Notice that the quaternionic relations on O allows

one to dene a two-spheres worth of almost complex structures
(1.4.6) 1() =
1
1
1
+
2
1
2
+
3
1
3
.
where = (
1
.
2
.
3
) o
2
. So ` together with 1() is usually called an almost
hypercomplex manifold. A G1(n. H)-connection satises 1() = 0 for all
o
2
. An almost hypercomplex structure that is 1-integrable is called hypercomplex.
In this case torsion-free G1(n. H)-connections are unique. Such a connection is
called the Obata connection . Compact four dimensional manifolds admitting a
hypercomplex structure were classied in [Boy88a].
One can always equip ` with a Riemannian metric o which preserves 1(),
i.e., o(1()Y. 1()Y ) = o(A. Y ) for any o
2
. The triple (`. I. /) is then called
36 1. STRUCTURES ON MANIFOLDS
an almost hyperhermitian manifold and provides further reduction of the structure
group to oj(n). A hypercomplex structure with a compatible Riemannian metric
is called a hyperhermitian structure. An almost hyperhermitian structure is 1-
integrable if and only if it is hyperkahler. This is equivalent to the usual denition
of a hyperkahler structure as one in which the Riemannian holonomy is contained
in oj(n). Both hypercomplex and hyperkahler structures are treated in more detail
in Chapter 12.
Example 1.4.20: Quaternionic 1-structures. A G-structure with the group
G = G1(n. H)oj(1) 1l
m
G1(4n + :. R) is called an almost quaternionic 1-
structure of corank :. It can be seen that such a reduction is equivalent to the
existence of two subbundles Q End(T`) and 1 T` of rank 3 and : respec-
tively, such that Q admits local sections J
1
. J
2
. J
3
satisfying quaternionic algebra
relations, and 1 is parallelizable. One can choose global sections
1
. . . . .
m
of 1
so that
J
i
(

) = 0.

J
i
= 0. J
2
i
= 1l +
m

=1

.
where = 1. . . . . :. i 1. 2. 3 and
1
. . . . .
m
is the dual frame in 1

. When
the sections J
1
. J
2
. J
3
are global, i.e., the bundle Q is trivial we call (`. Q. 1) a
hyper 1-structure of corank :. Such structure is a G-structure with G = oj(n)1l
m
.
When : = 0 these denitions reduce to the usual notions of almost quaternionic
and hypercomplex geometry in the two previous examples. When : = 3 a hyper 1-
structure is called an almost hypercontact structure introduced by Kuo [Kuo70] as
an almost contact (metric) 3-structures. The geometry of quaternionic and hyper
1-structures of corank 3 was studied by [Her96] and we will come back to some of
these examples in Chapter 13.
1.5. Pseudogroup Structures
We have seen that the notion of G-structure is a unifying concept for study-
ing geometric structures. Another unifying concept is the idea of a pseudogroup
of transformations. Roughly speaking a pseudogroup is a group of transforma-
tions, where there may be some problems with domains of denition. The concept
goes back to Lie and Cartan, but the precise notion was dened more recently by
Ehresmann [Ehr53]. The ideas were then developed much further by Kodaira and
Spencer among others, cf. [GS66, Kod60b, KS58, Spe62]. Here is the denition
following [Kob61]
Denition 1.5.1: A pseudogroup of transformations on R
n
is a set of local
dieomorphisms which satisfy
(i) Each 1 is a dieomorphism of an open subset of R
n
. called the
domain of 1, onto an open subset of R
n
, called the range of 1.
(ii) If l =
i
l
i
. where each l
i
is an open subset of R
n
. then a dieomor-
phism 1 with domain l belongs to if and only if its restriction to each
l
i
belongs to .
(iii) For every open subset l of R
n
, the identity transformation on l belongs
to .
(iv) If 1 then so is 1
1
.
(v) If 1 : l\ and 1
t
: l
t
\
t
are elements of and \ l
t
= . then
the dieomorphism 1
t
1 : 1
1
(\ l
t
)1
t
(\ l
t
) belongs to .
1.5. PSEUDOGROUP STRUCTURES 37
is said to be transitive if for every pair of points j. c R
n
there is an 1
such that 1(j) = c.
More generally, one can replace R
n
by a dierent model space [Kob61], for
example a smooth manifold `. The importance of pseudogroups for our purpose
lies in its ability to describe dierent types of atlases of coordinate charts.
Denition 1.5.2: Let be a transitive pseudogroup, and ` a second countable,
Hausdor topological space. A -atlas for ` is a family of pairs (l

)
I
,
called charts, such that
(i) The collection l

I
of open sets covers `. i.e., each l

is an open
subset of `. and

= `.
(ii) Each

is a homeomorphism of l

onto an open subset

(l

) R
n
such that whenever l

= . the map

(l

) :

(l

)
belongs to .
A -atlas on ` is said to be maximal if it is not properly contained in any
other -atlas on `.
Clearly, if is the pseudogroup of all local dieomorphisms of R
n
, this denition
coincides with the usual denition of an atlas of coordinate charts for a dierentiable
manifold. As in the usual case, it is easily seen that every -atlas is contained in
a unique maximal -atlas. Thus, a given -atlas denes uniquely a -structure on
`. The relation between transitive pseudogroups and G-structure should now be
evident from Proposition 1.4.3.
Proposition 1.5.3: Let G be a Lie subgroup of G1(n. R). and let
G
be the pseu-
dogroup of local dieomorphisms of R
n
whose Jacobian matrix lies in G at each
point of its domain. Then the
G
-structures on ` are in 11 correspondence with
the integrable G-structures on `.
It should be clear that the pseudogroup
G
is transitive since it contains the
translations. Notice that not every -structure is a
G
-structure for some G
G1(n. R). We will see an important example of this later. Of course, as with
integrable G-structures, or just G-structures, for a given there may not exist any
-structure at all on `. That is, generally, there are topological obstructions to
the existence of a -structure. Two important
G
-structures are for G = G1(n. C)
and G = oj(
n
2
. R). It is clear from the two examples 1.4.9 and 1.4.12 that the

GL(n,C)
-structures are just the complex structures, while the
Sp(
n
2
,R)
-structures
are the symplectic structures. The example of the six sphere o
6
is interesting in
this context. As mentioned in the paragraph after Example 1.4.9 o
6
is known to
admit an almost complex structure, i.e., a G1(3. C)-structure, but it is not known
whether it has a complex structure, i.e., a
GL(3,C)
-structure. From the symplectic
point of view, o
6
also admits an almost symplectic structure, i.e., an oj(3. R)-
structure, since it admits an almost complex structure; however, it cannot admit
a symplectic structure, i.e., a
Sp(3,R)
-structure, since the closed 2-form would
provide a non-vanishing element of H
2
(o
6
. R) by Stokes Theorem.
Following [SS65] we relate pseudogroups to collections of vector elds. Let
A

be a collection of smooth vector elds each dened on an open set l

`.
Each A

generates a local1-parameter group

(t) of local dieomorphisms dened

for small t and for some open subset of l

. The collection

denes a family
38 1. STRUCTURES ON MANIFOLDS
of local dieomorphisms and thus generates a pseudogroup { on `. Fix a point
r ` and consider the subpseudogroup {
0
of { generated by those vector elds
in the collection that vanish at r. Then any local 1-parameter group

(t) {
0
satises

(t)(r) = r for all t in its domain. The dierential

(t)

of any such
local 1-parameter group is a linear map on the vector space T
x
`. Running through
all such local 1-parameter groups in {
0
. determines a Lie subalgebra g
0
of o|(n. R).
We call g
0
the linear isotropy algebra of the pseudogroup {. This denition diers a
bit from the one in [SS65], but they should be equivalent. The connected subgroup
G
0
of G1(n. R) determined by the Lie algebra g
0
is called the linear isotropy group
of {.
Let be a pseudogroup of local transformations
4
on a smooth manifold `. We
can consider germs of elements 1 . that is two maps 1. o : l\ in sending
r ` to n ` are equivalent if there is an open set \ l such that 1[
W
= o[
W
.
The equivalence class 1
x
is the germ of the map 1 at r. Note that 1
x
can only be
evaluated at r in which case 1
x
(r) = n. Thus, we have a source map :(1
x
) = r.
a target map t(1
x
) = 1
x
(r) = n. and a composition dened as follows: if 1
x
is a
germ of a map from r ` to n `. and o
y
is a germ of a map from n to .. then
the composition o
y
1
x
is well dened and is the germ at r of the map o 1 from
r to .. One easily sees that the set (() of germs of maps of the pseudogroup
forms a groupoid (cf. Appendix A), called a groupoid of germs of the pseudogroup
. Furthermore, we can give (() a topology that makes the source map : a local
homeomorphism. So : : (()` becomes an etale space over `. Thus, (()
is an etale groupoid constructed from germs, or in the terminology of Haeiger
[Hae58], a faisceau de groupoids (sheaf of groupoids). Conversely given a groupoid
of germs ( on a manifold `. we can reconstruct a pseudogroup as follows: let
: l( be a local section of the source map of (. Composing with the target
projection t : (` gives a local dieomorphism of `. and it is easy to check that
the set of all such dieomorphisms forms a pseudogroup of local transformations of
`. Summarizing we have
Proposition 1.5.4: Let ` be a smooth manifold and a pseudogroup of local
transformations on `. Then the set (() can be given the structure of a etale
groupoid. Conversely given a groupoid of germs ( over `. the set
t [ t is the target projection. : l( is a local section of the source map :
is a pseudogroup of local transformations on `.
This proposition says that the notion of pseudogroups of transformation on `
and that of groupoids of germs on ` are equivalent.
Exercise 1.12: Check the details of Proposition 1.5.4. That is, given a pseu-
dogroup of transformations on `. show that passing to germs one obtains an etale
groupoid, and conversely, given a groupoid of germs on `. the set dened above
forms a pseudogroup.
1.6. Group Actions on Manifolds
In this section we will review some basic properties of group actions on mani-
folds. This is a classical subject treated in several texts [MZ55, Bre72, Kob72,
4
By a pseudogroup of local transformations on M we mean any subpseudogroup of the
pseudogroup of local dieomorphisms of M. More generally one could take M to be a topological
space and a subpseudogroup of the pseudogroup of local homeomorphisms of M.
1.6. GROUP ACTIONS ON MANIFOLDS 39
DK00] to name a few. We mainly follow the approach and the notation of Duis-
termaat and Kolk and refer the reader to their modern textbook for more details
[DK00]. Recall that a topological group is a group G endowed with a topology
such that the multiplication map j : ```, j(o
1
. o
2
) = o
1
o
2
and the inverse
map : ``, (o) = o
1
are continuous. A discrete group is a topological group
with the discrete topology. For a topological manifold ` we denote by Hom(`)
the group of homeomorphisms of `. If ` is a smooth manifold the group of dieo-
morphisms is denoted by Diff(`). We give Diff(`) the compact-open C

topology.
This is a good topology when ` is compact, but in the non-compact case it does
not control the behavior at innity. Since we deal mainly with compact manifolds
we do not concern ourselves with this failure. We refer the reader to [Ban97] for
further discussion of dieomorphism groups.
Denition 1.6.1: An action of a topological group G on a topological manifold
` is a homomorphism / : G Hom(`) of G into the group of homeomorphisms
of `. We say that the action / is continuous if the mapping
(1.6.1) (o. r) o r = /(o)(r) : G` `.
also denoted by / : G ` `. is continuous. We say that / is a proper
action if the associated map G` ` ` given by
(1.6.2) (o. r) (r. o r)
is proper, i.e, the inverse image of any compact set is compact. If ` is a smooth
manifold and G is only a topological group we say that / is smooth if /(o)
Diff(`) for each o G. However, if G is a Lie group a smooth action / means
that the map (1.6.1) is smooth.
Let / be an action of G on ` and let c G be the identity element. For any
r ` the subgroup G
x
= o G [ o r = r is called the isotropy subgroup of the
action at the point r `, or the stabilizer of r ` under the action. Then / is
called eective if Ker(/) = 1l. The action / is free if G
x
= c for all r `
and locally free if G
x
is a nite group for all r `.
For each r ` the orbit through r is dened as
(1.6.3) G r = n ` [n = o r. o G `.
The orbit G r through r is an immersed submanifold of `. and there is a natural
identication of G r with the coset space GG
x
.
It is clear that the relation r n if n G r is an equivalence relation
partitioning ` into orbits. We shall denote the quotient space ` by `G,
and call it the space of orbits. The map : ``G sending r G r is the
canonical projection. We equip `G with the quotient topology, i.e., \ is open in
`G if and only if
1
(\ ) is open in `.
Exercise 1.13: Let ` be a manifold with group action / : G` `. Show
that `G is Hausdor if and only if the set (r. n) ` ` [ n G r is a
closed subset of ` `.
In this book we will mainly be concerned with actions which yield Hausdor
quotients. The importance of proper group actions is realized through the following
Proposition 1.6.2: If the action / of G on ` is proper then the quotient `G
is Hausdor.
40 1. STRUCTURES ON MANIFOLDS
Example 1.6.3: A typical example of a non-proper group action is the so-called
irrational ow on a torus. Let ` = T
2
= oO(2)oO(2) which is an Abelian group.
So for any R we can consider a subgroup G() = ((t). (t). t R, where
(t) =

cos t sin t
sin t cos t

.
Then G() acts on ` by left (or right) multiplication. For Q the orbits are
all closed in `. In this case G() is homeomorphic to o
1
and so is the quotient
`G(). However, for R`Q the orbits are not closed. They are immersed but
not embedded submanifolds homeomorphic to R.
The closure of any orbit equals to ` and the quotient `G() is non-Hausdor.
There are also examples with non-Hausdor quotients with all orbits closed.
Consider, for instance, G = R acting on R
2
via

t. (r
1
. r
2
)

(r
1
+tr
2
. r
2
). t R. (r
1
. r
2
) R
2
.
Here the orbits are indeed closed: they are either lines parallel to the r
1
-axis, or
points on the r
1
-axis. It is easy to see that the quotient is non-Hausdor.
Denition 1.6.4: A G-manifold is a smooth manifold ` together with a smooth
action / of a group G.
The importance of proper and free actions is immediately realized by the fol-
lowing classical result.
Theorem 1.6.5: Let G be a Lie group acting smoothly, freely, and properly on
`. Then the orbit space `G is a smooth manifold of dimension dim(`)
dim(G) with a unique dierentiable structure having the following property: if
: ` `G is the canonical projection then for every o `G there exists an
open neighborhood | o and a dieomorphism
= (. :) :
1
(|) G|
such that, for each r
1
(|). o G we have
(1.6.4) (r) = :(r). (o r) = (o(r). :(r)).
In fact, it follows that the map is a smooth submersion and ` = 1(`G. G)
is a principal G-bundle over `G, and the map is a local trivialization. If we add
a bit more structure the hypothesis on the action of Theorem 1.6.5 can be weakened
to be only locally free. This is of great importance to us in this book, and will be
discussed further in the forthcoming chapters. See, for example, Theorem 2.5.11
below.
A case of particular interest to us is when a group G leaves invariant some
geometric structure. Following [Kob72] we have
Denition 1.6.6: Let 1 be a G-structure over `. A dieomorphism 1 : ``
is an automorphism of the G-structure if the induced map 1

: 11 is an
isomorphism.
The automorphisms of 1 form a subgroup of Diff(`) and is denoted by Aut(1).
In particular, if the G-structure 1 is dened by a tensor eld T then 1 : ``
is an automorphism of 1 if and only if it leaves T invariant. There are many
such examples of interest to us throughout this book. We shall often denote the
automorphism group by Aut(T), but in some cases there is more common standard
terminology. For example, if (`. o) is a Riemannian manifold the automorphism
1.6. GROUP ACTIONS ON MANIFOLDS 41
group Aut(o) is called the isometry group and denoted by Isom(`. o). We shall
come across other examples as we proceed. Since Aut(1) a subgroup of Diff(`) it
is endowed with the compact-open topology, and we are interested in when it is a
Lie group. Recall [Kob72] that a Lie subalgebra g gl(n. R) is said to be elliptic
if g contains no matrix of rank one. Then we have
Proposition 1.6.7: Let 1 be a G-structure on a compact manifold ` with group
G and Lie algebra g. If g is elliptic then Aut(1) is a Lie group with respect to the
compact-open C

topology.
Examples of elliptic Lie algebras are the orthogonal Lie algebras o(n. R). the
complex Lie algebra gl(n. C) gl(2n. R). and any of their subalgebras. Thus,
the automorphism groups of a compact Riemannian manifold, or of a compact
almost complex manifold are Lie groups. In general compactness is a necessary
hypothesis for Aut(1) of an elliptic G-structure to be a Lie group. For example the
complex automorphism group Aut(C
n
) is not a Lie group. However, for Riemannian
and certain other G-structures compactness is not necessary. In order to develop
more we need the notion of prolongation. Since this is amply treated in several
books [Ste83, Kob72] we give only a very brief treatment here. Let \ be an
n-dimensional vector space and g a Lie subalgebra of gl(n. R) = End(\ ).
Denition 1.6.8: The rst prolongation g
(1)
of g is dened by
g
(1)
= T Hom(\. g) [ T(n) = T()n for all n. \ .
The /
th
prolongation g
(k)
is dened inductively by g
(k+1)
= g
(k)
(1)
. A Lie algebra
g is said to be of nite type of order / if for some /. g
(k)
= 0, but g
(k1)
= 0
otherwise g has innite type.
Similarly one can dene the prolongations of a Lie group G G1(n. R). Such
prolongations give rise to higher order G-structures for which we have no real need.
We say the a G-structure is of nite type of order / if its Lie algebra g is of nite
type of order /. It is easy to see that if h g and g
(k)
= 0 then h
(k)
= 0. It
is equally easy to show that o(n. R)
(1)
= 0. and that gl(n. R) and gl(n. C) are of
innite type. However, gl(n. H) is of nite type, in fact gl(n. H)
(1)
= 0. Many of the
G-structures that are of importance in this book are of innite type, in particular
those of Examples 1.4.9, 1.4.12 1.4.13 1.4.15, and 1.4.16. G-structures that are of
nite type, but with g
(1)
= 0 are also of interest. This occurs in two important
cases for us, the conformal G-structures with G = CO(n) and the quaternionic G-
structures with G = G1(n. H)oj(1). These both have g
(1)
= 0, but g
(2)
= 0. When
a G-structure is 1-integrable, the vector space g
(1)
parameterizes the torsion-free
connections, and if g
(1)
= 0 the torsion-free connection is unique. We now have
Theorem 1.6.9: Let 1(`. G) be a G-structure of nite type of order /. Then
Aut(1) is a Lie group of dimension at most dim ` + dim g + dim g
(1)
+ +
dim g
(k1)
.
The proof of this general theorem is given in [Ste83, Kob72]. It, of course,
has many precursors the most famous of which is the somewhat stronger result
essentially due to Myers and Steenrod Theorem [MS39]:
Theorem 1.6.10: Let (`. o) be a Riemannian manifold with nitely many con-
nected components. Then the group of isometries Isom(`. o) is a nite-dimensional
Lie group. The action of Isom(`. o) on ` is proper and its Lie group topology
coincides with the compact-open C

topology as a subgroup of Diff(`).

42 1. STRUCTURES ON MANIFOLDS
In particular, this theorem says that the isometry group Isom(`. o) of a com-
pact manifold is compact. We discuss many other examples of G-structures of
nite type throughout the book. The G-structures dealt with in this book for
which Aut(1) is not a Lie group include the symplectic and contact structures.
Denition 1.6.11: Let G be a group acting on a manifold `. We say that the
action is transitive if for any r. n ` there exists o G such that n = o r. In
such a case we say that ` is G-homogeneous. If Aut(1) is the automorphism
group of a G-structure 1(`. G) we say that ` is a homogeneous G-structure
if Aut(1) acts transitively on `. If G is a Lie group which acts transitively on `
we say that ` is a homogeneous manifold.
When a manifold admits a Lie group G acting transitively and is a subgroup
of the Aut(1) for some G-structure 1(`. G), we say that ` is a homogeneous (G-
structure) manifold, for example a homogeneous complex or homogeneous contact
manifold, etc.
Example 1.6.12: Let G be a Lie group and H any subgroup. The H acts on G
from the left and from the right:
(1.6.5) /
L
(/)o = /o.
(1.6.6) /
R
(/)o = o/
1
.
The corresponding quotient spaces are denoted by H`G and GH. Both of these
actions are free but they may or may not be proper. If H is a closed subgroup of
G then the H-action is proper and the quotient GH is a homogeneous smooth
G-manifold under the natural left G-action.
More generally, we will be interested in proper actions which are not necessarily
free. In such case the space of orbits is not always a manifold; nevertheless, it has
a very tractable structure, namely that of an orbifold which we discuss in Chapter
4.
Consider the dierential (/
x
)

: g T
x
`, where g T
e
G is identied with
the tangent space of G at the identity o = c.
Denition 1.6.13: Let / : G` ` be a smooth action of a Lie group on a
manifold `. A smooth slice at r
0
` for the action / is a smooth submanifold
o `, r
0
o such that
(i) T
x
0
` = (/
x
0
)

(g) T
x
0
o and T
x
0
` = (/
x
0
)

(g) +T
x
0
o for all r o,
(ii) o is G
x0
-invariant,
(iii) if r o, o G, and /(o)(r) o, then o G
x0
.
It follows that the inclusion o ` induces a bijective map G
x0
r G r
from the space oG
x
0
of G
x
0
orbits in o onto an open neighborhood of G r
0
in
`G.
Denition 1.6.14: The action / : G ` ` is said to be proper at r
0
if
for any convergent sequence r
i
r
0
in ` and a sequence o
i

iN
in G such that
o
i
r
i
r
0
, there exists a convergent subsequence of o
i

iN
.
Naturally, if / : G` ` is proper then it is proper at every point r `.
But the converse is in general not true. However, if / is proper at every point of
` and `G is Hausdor then is proper.
1.6. GROUP ACTIONS ON MANIFOLDS 43
Lemma 1.6.15: The action / : G ` ` is proper at r
0
` if and only
if there exists a neighborhood | r
0
such that o G [ /(o)(|) | = has a
compact closure in G. In particular G
x0
is a compact subgroup of G.
The following important theorem was proved by Palais [Pal61].
Theorem 1.6.16: Let / : G` ` be a smooth action of a Lie group G on
a manifold `. Suppose that the action is proper at r
0
`. Then there exists a
smooth slice o at r
0
for the action /.
If ` is a smooth manifold and G acts properly, the quotient is both Hausdor
and paracompact. One can dene a function on an open subset of `G to be
smooth if and only if

1 = 1 is a smooth function on `. Then a standard

partition of unity and averaging argument can be used to construct G-invariant
Riemannian metric on `.
Theorem 1.6.17: Let / : G` ` be a proper action of G on `. Then `
has a G-invariant Riemannian metric o.
Theorems 1.6.10 and 1.6.17 show that in the case of smooth manifolds with
a nite number of connected components the category of smooth proper eective
Lie group actions coincides with the category of closed subgroups of isometries for
Riemannian structures on `.
Denition 1.6.18: A smooth stratication of a manifold ` is a partition of
` into locally closed sets such that
(i) Any ` is a smooth, connected, nite-dimensional manifold without
boundary.
(ii) is locally nite.
(iii) If `
1
. `
2
and `
1
`
2
= then `
1
`
2
.
(iv) If `
1
. `
2
, `
1
= `
2
and `
1
`
2
then dim(`
1
) < dim(`
2
).
The third condition is called the frontier condition. Furthermore, is said to be
Whitney regular if the following two conditions are met.
(a) For each `
1
. `
2
, `
2
`
1
, `
1
= `
2
and each sequence r
i

iN
such that
lim
i
r
i
= r `
2
. lim
i
T
xi
`
1
= 1
we have T
x
`
1
1.
(b) If r
i

iN
is a sequence as in (a) and n
i

iN
is a sequence in the limit
stratum `
2
such that n
i
r and n
i
= r
i
for all i N then each limit of
one-dimensional subspaces R (r
i
. n
i
) of T
x
i
`, for i is contained
in 1. Here is any dieomorphism from an open neighborhood of the
diagonal in ` ` to an open neighborhood of the zero section of the
tangent bundle T` and the set of limit lines does not depend on the
choice of .
We will now consider the a stratication induced by a proper eective action
/ : G` `. As before we will follow closely the notation of [DK00].
Denition 1.6.19: We say that points r. n ` and the orbits G r. G n `G
are of the same type, with notation r n, G r G n respectively, if there
exists a G-equivariant bijection from G r to G n. We say that r dominates n and
G r dominates G n (n r, G n G r) if there exists a G-equivariant mapping
from G r to G n. The equivalence classes under are called orbit types in `
44 1. STRUCTURES ON MANIFOLDS
and `G respectively and will be denoted by
`

x
= n ` [ r n. `

Gx
G = G n `G [ G n G r.
The relation is a pre-order in ` and we dene
`

x
= n ` [ r n. `

Gx
G = G n `G [ G n G r.
The key to this denition is the following lemma describing the properties of
the above relations.
Lemma 1.6.20: We have the following
(i) r n G
y
= o
1
G
x
o for some o G, i.e., G
x
and G
y
are conjugated
in G.
(ii) r n o
1
G
x
o G
y
for some o G, i.e., G
x
is conjugated in G to
a subgroup of G
y
.
(iii) For any G-equivariant mapping : G r G n, G
x
is a subgroup of
G
(x)
and is a smooth bration with ber G
(x)
G
x
.
(iv) r n and n r r n.
Denition 1.6.21: Let H be any subgroup of G. We dene the set of xed points
for H as
(1.6.7) `
H
= n ` [ / n = n. for all / H.
We denote by (H) the conjugacy class of groups conjugated to H in G and by `
(H)
the set of all points in ` with stabilizers conjugated to H.
One can show that each connected component of `
H
is a closed submanifold
of ` but dierent connected components can have dierent dimensions. Naturally,
the orbit types give a stratication of ` and this stratication is typically called
stratication by orbit types. However, an orbit type can be a union of connected
components of dierent dimensions. That is why we consider a ner stratication
of each `

x
= `
(G
x
)
by the so-called local action type.
Denition 1.6.22: The points r. n ` and the orbits G r, G n, respectively,
are said to be of the same local type (with the notation r n and G r G n) if
there exists a G-equivariant dieomorphism from an open G-invariant neighborhood
| r onto an open G-invariant neighborhood 1 n in `.
The relation partitions further partitions `
x
and `

Gx
G into subclasses
called local action types and denoted by
(1.6.8) `

x
= n ` [ n r. `

Gx
G = G n `G [ G n G n.
The following theorem describes fundamental properties of the stratication by
local action types.
Theorem 1.6.23: Let / : G` ` be a smooth and proper action.
(i) Each local action type is an open and closed subset of the corresponding
orbit type.
(ii) The set `

x
`
G
x
is open in `
G
x
and a locally closed smooth submani-
fold of ` (that is all its connected components have the same dimension).
In addition it is (G
x
)-invariant, when (G
x
) is the normalizer of G
x
in G.
(iii) The canonical projection : ` `G maps `

x
`
G
x
onto `

Gx
G,
its bers in `

x
`
G
x
are the orbits for the (G
x
)-action on `

x
`
G
x
.
1.6. GROUP ACTIONS ON MANIFOLDS 45
The action of (G
x
)G
x
is proper and free. This gives `

Gx
G a unique
structure of a smooth manifold for which : `

x
`
G
x
`

Gx
G is
the principal ber bundle with structure group (G
x
)G
x
.
(iv) `

x
is a locally closed smooth submanifold in `. Furthermore, the pro-
jection : `

x
`

Gx
is a smooth bre bundle with bre GG
x
.
(v) We have dim(`

x
) = dim(G) dim(G
x
) +dim(`

x
`
G
x
). and, hence,
dim(`

x
G) = dim((G
x
)) dim(G
x
) + dim(`

x
`
Gx
).
The local niteness of the stratication follows now from the fact that in the
case of proper actions, locally there can only be nitely many orbit types. More
precisely
Proposition 1.6.24: For a proper smooth Lie group action G on ` there are
locally only nitely many distinct orbit types with locally only nitely many con-
nected components. If ` is compact there are only nitely many distinct connected
components of orbit types and the same is true if ` is a nite-dimensional vector
space and G acts linearly.
Theorem 1.6.25: The connected components of the orbit types in ` from a Whit-
ney stratication of `.
A part of the information about the orbit type stratication for the action
/ : G ` ` can be encoded into the directed graph = (/. G. `). The
vertices of are the connected components of the orbit types in `G. If
1
.
2

then we draw an arrow from
1

2
if
1

2
. It follows that if
1

2
then

2

1
and dim(
1
) dim(
2
) as long as
1
and
2
are distinct.
1

2
denes
a partial ordering in in which any chain has at most 1 + dim(`G) elements.
Example 1.6.26: [Complex weighted projective plane] Let ` = o
5
C
3
, be
the unit sphere in C
3
, G = o
1
and consider the action /
p
: o
1
o
5
o
5
given
by
/
p
(. z) = (
p
1
.
1
.
p
2
.
2
.
p
3
.
3
).
where z = (.
1
. .
2
. .
3
) are the linear complex coordinates on C
3
, o
1
is a unit com-
plex number, and p = (j
1
. j
2
. j
3
) are positive integers such that gcd(j
1
. j
2
. j
3
) = 1,
d
ij
= gcd(j
i
. j
j
) 1 for all i = ,. Then the directed graph has 7 vertices which
can be described as follows: Let be any non-empty subset of 1. 2. 3 and denote

the o
1
-invariant subset in o
5
o
1
corresponding to the non-vanishing of all com-
plex coordinates .
j
on o
5
C
3
with , . With this notation can be described
as follows.
46 1. STRUCTURES ON MANIFOLDS

1

12

31

123

3

23

In the case any d

ij
= gcd(j
i
. j
j
) = 1 or/and some j
t
j
: equal to 1, the directed
graph is obtained by deleting the corresponding vertices and the arrows that orig-
inate there and appropriately re-dening the set
123
. In the case p = (1. 1. 1) we
get just one vertex as the o
1
-action is free and there is only one orbit type.
Denition 1.6.27: Let / : G` ` be a smooth proper action of a Lie group
G on `. Then the orbit G r, r ` is said to be a principal orbit if its local
action type `

x
is open in `, i.e., `

x
belongs to the maximal element of . We
write `
princ
= r ` [ G r is a principal orbit, and `
princ
G for the set of
principal obits.
Denition 1.6.28: The orbit G r is said to be regular if the dimension of the
orbits G n is constant for all n suciently close to r. Points lying on regular
orbits are said to be regular points and their collection forms `
reg
. The points in
`
reg
``
princ
are called exceptional and any orbit through an exceptional point is
called an exceptional orbit. Finally, the points in ` ` `
reg
are called singular
and any orbit through a singular point is called a singular orbit.
It follows that in every connected component of G` a nearby non-principal
orbit is an exceptional orbit if and only if it has the same dimension as the principal
orbit. A nearby non-principal orbit is singular if and only if its dimension is strictly
smaller than the dimension of the a principal orbit. Note that in the Example
1.6.26 all orbits are regular so that `
reg
= `. Hence, non-principal orbits are all
exceptional.
Clearly, `
princ
`
reg
` and both are dense open subsets in `. Both of
these sets are also G-invariant so that the inclusions project to open dense subsets
`
princ
G `
reg
G `G. The following theorem is typically called principal
orbit theorem.
Theorem 1.6.29: Let / : G` ` be a smooth proper action of a Lie group
G on `. Them ` ` `
reg
is equal to the union of local orbit types of codimension
2 in `. For every connected component `

of `, the subset `
reg
`

is
connected, open and dense in `

. Each connected component of G` contains

only one principal orbit type which is a connected, open, and dense subset of it. In
particular, if ` is connected there is only one principal orbit type.
Without loss of generality, by restricting to subsets of ` of the from G `

,
one can assume that G` is actually connected. In particular, this is always the
1.6. GROUP ACTIONS ON MANIFOLDS 47
case when ` itself is connected. In this book such will most often be the case.
When both `G and ` are connected the general stratication picture simplies
a lot. We have
Corollary 1.6.30: Let / : G` ` be a smooth proper action of a Lie group
G on ` such that `G is connected. Let r `
princ
and let C be the connected
component of r in `
princ
`
G
x
. Then
(i) `
princ
= `

x
= `

x
.
(ii) The union of connected components of `
Gx
that meet `
princ
is a closed
smooth submanifold of ` which contains (G
x
) C = `
princ
`
G
x
.
(iii) G
(C)
= o G [ /(o)(C) = C induces a G-equivariant dieomorphism
GG
x

G
(C)
/Gx
C `
princ
.
We next consider some special cases in some more detail. One important class
of actions is when G is a compact Lie group acting locally freely on a compact
manifold `, in which case the quotient `G can be identied with a compact
smooth orbifold. These will be discussed in chapter 4. In fact, most, though not
all compact orbifolds arise this way.
Denition 1.6.31: A smooth, proper, locally free action / : o
1
` ` is called
a Seifert bered o
1
-structure and ` is called Seifert bered o
1
-manifold.
The actions /
p
of the Example 1.6.26 give innitely many distinct Seifert bred
o
1
-structures on o
5
. In Chapter 4 we will give a more general denition of a Seifert
bration of a manifold ` in the language of orbifolds and orbibundles.
Denition 1.6.32: Let (`. o) be a Riemannian manifold and let G = Isom(`. o)
be its isometry group. We say that (`. o) is homogeneous if G acts transitively
on `. We say that (`. o) is of cohomogeneity 0 < / dim(`) when the
principal orbit of the G-action has dimension dim(`) /.
The most tractable cases amenable to further analysis are the homogeneous, co-
homogeneity one, and cohomogeneity two, cases. Homogeneous Riemannian mani-
folds will be of special interest to us throughout this book, but they are trivial from
the point of view of the stratication discussed in this section. The next case is that
of cohomogeneity one. In this case the stratication is simple as there are not too
many candidates for `G. We have the following theorem of Mostert [Mos57]:
Theorem 1.6.33: Let G be a compact connected Lie group acting on a connected
manifold ` of dimension n such that there is at least one orbit of dimension n1.
Then the quotient space `G must be one of the following
(i) `G = R,
(ii) `G = [1. ),
(iii) `G = o
1
,
(iv) `G = [1. +1].
Furthermore, the non-principal orbits correspond to the boundary points.
The general description can be found in Section 8 of Chapter IV of [Bre72], in
particular, Theorem IV.8.2 and its proof where it is assumed that ` is compact.
Compactness eliminates the rst two cases. In the third case ` (GH)
Z
R is a
bundle over o
1
and has innite fundamental group. We follow the presentation in
[GZ02]. However, we focus our attention on the last case where there are exactly
two non-principal orbits.
48 1. STRUCTURES ON MANIFOLDS
Consider any G-invariant metric o and a geodesic c : [1. +1] ` perpendic-
ular to the orbits with endpoints r

= c(1). The isotropy group at every point

c(t), 1 < t < 1 equals to G
c(t)
= H. Let us denote by 1

= G
x
the isotropy
groups at the end points of the geodesic. All orbits GH through c(t), t = 1
are principal. The two non-principal orbits are G1

= G r

. Let 1
l+1

be
the normal disc to the orbit G1

at r

, where |

= dim(G) dim(1

) are the
dimensions of non-principal orbits. By the slice theorem ` decomposes as
(1.6.9) ` = G
K
1
l
+
+1

G/H
G
K
1
l

+1
and we have the following group diagram
(1.6.10) G
1

1
+
j+

H
h

h+

which is also written as H 1

. 1
+
G. The non-principal orbits are denoted
by 1

= G1

and note that 1

H o
l

must be spheres. The non-principal

orbit, say 1

= G1

, is exceptional if |

= 0. Otherwise, 1

is singular. Hence,
a cohomogeneity one manifold ` with two non-principal orbits determines the
diagram 1.6.10. Conversely, any such group diagram determines ` via (1.6.9). We
remark that in the case there is only one non-principal orbit ` = `
K
1
l

+1
.
Remark 1.6.1: Observe that a group diagram as in (1.6.10), where we assume
that /

are embeddings, but ,

are only homomorphisms with nite kernel and

,

= ,
+
/
+
= ,
0
with 1

H = o
l

, denes a cohomogeneity one orbifold

O: The regular orbits, being hypersurfaces, have no orbifold singularities, and we
can therefore assume that ,
0
is an embedding, although we still allow the action
of G to be ineective otherwise. A neighborhood of a singular orbit is given by
1(1

) = G
K
1
l

+1
, where 1

acts on G via right multiplication: o/ = o,

(/)
and on 1
l

+1
via the natural linear extension of the action of 1

on o
l

. This
then can be written as 1(1

) = G
(K/kerj)
(1
l

+1
ker,

) and the singularity

normal to the smooth singular orbit G,

(1

) is o
l

ker,

.
Parker [Par86] began a classication of the smooth cohomogeneity one actions
of compact Lie groups on compact connected 4-manifolds although there are some
missing cases. In Section 12.5, we shall be interested in certain cohomogeneity one
oO(3)-actions. Here we present an example, shown to us by Wolfgang Ziller, for
the group oj(1) that is relevant to the discussion in Section 12.5, in particular to
diagram 12.5.9.
Example 1.6.34: Let ` be a compact 4-manifold, G = oj(1) = ol(2), with quo-
tient of type (iv) above. We consider four cases where the non-principal orbits are
singular of dimension one. The rst case we consider is ` = o
4
. Here we represent
o
4
as the unit sphere in the ve dimensional representation of oO(3) consisting
of trace-free symmetric matrices. (See Equations 12.5.7 and 12.5.8 below). This
action can be lifted to an non-eective action of the group oj(1) of unit quaternions
which we treat here in order to facilitate the forthcoming discussion. We have the
1.6. GROUP ACTIONS ON MANIFOLDS 49
group diagram
oj(1)
1

1
+
.
j+

Q
h

h
+

where Q is the 8-element quaternion group Q = 1. i

1
. i
2
. i
3
, 1

= 'c
i1
. i
2
`
and 1

= 'c
i
2

. i
1
`. Each group in this diagram can be factored by a central Z
2
giving diagram 12.5.9. This is action 62 in [Par86].
All other possible diagrams are obtained from this one by the removal of certain
nite subgroups from 1

which gives a new manifold together with a branch cover.

For example, consider
oj(1)
1

'c
i2
`.
j
+
_

Z
4
= 'i
2
`
h

h+

with 1

as before. This gives ` = CP

2
as a well-known 2-fold branch cover
: CP
2
o
4
. This action is missing in [Par86].
Further branching is obtained by taking
oj(1)
'c
i1
`
j

'c
i1
`.
j+
_

Z
2
= 1
h

h
+

This gives ` = CP
1
CP
1
as a 2-fold branched cover : CP
1
CP
1
CP
2
branched over a conic which is the image of the diagonal. In [Par86] this is action
50 with n = 2.
The nal example corresponds to the group diagram
oj(1)
'c
i1
`
j

'c
i1
`.
j
+

1
h

h
+

50 1. STRUCTURES ON MANIFOLDS
which corresponds to ` = CP
2
#CP
2
: a blow-up of CP
2
at one point. This is
action 45 in [Par86].
Exercise 1.14: Fill in the details for Example 1.6.34.
We nish this section with a short discussion of the cohomogeneity 2 case. In
general, when `G is 2-dimensional the situation can be quite complicated even
when it is connected. Nevertheless, under some mild assumptions, the structure is
actually quite manageable in which case `G is a disc with boundary, the boundary
corresponding to non-principal orbits. The following theorem, due to Montgomery,
Samelson, and Yang [MSY56], is Theorem 8.6 in Chapter IV of [Bre72] to which
we refer for the proof.
Theorem 1.6.35: Let ` be a compact connected manifold with a cohomogeneity 2
action of a compact connected Lie group G. Suppose H
1
(`. Z) = 0 and that there
exist a singular orbit in `G. Then there are no exceptional orbits and `G is a
2-disc with the boundary corresponding to the non-principal orbits.
By denition, any circle action on a 3-dimensional manifold is of cohomogene-
ity 2. The following example shows that the existence of a singular orbit in the
hypothesis of Theorem 1.6.35 is indeed necessary.
Example 1.6.36: [Football orbifold] Let ` = o
3
C
2
, be the unit sphere in
C
2
, G = o
1
and consider the action /
p
: o
1
o
3
o
3
given by

p
(. z) = (
p
1
.
1
.
p
2
.
2
).
where z = (.
1
. .
2
) are the linear complex coordinates on C
2
, o
1
is a unit
complex number, and p = (j
1
. j
2
) are positive integers such that gcd(j
1
. j
2
) = 1.
There are three orbit types:
(a) principal orbits through (.
1
. .
2
) o
2
when neither complex coordinate
vanishes and with G
(z
1
,z
2
)
= c,
(b) two exceptional orbits: through (.
1
. 0) o
3
with G
(z
1
,0)
= Z
p
2
, and
through (0. .
2
) o
3
with G
(0,z2)
= Z
p
1
.
The set `
princ
o
1
can be identied with a punctured disc (cylinder) with the
boundary as two circles corresponding to the exceptional orbits. The quotient
space is the orbifold CP
1
(j
1
. j
2
) which is topologically a sphere with two isolated
orbifold points.
In Chapter 13 we will consider many examples of 3-Sasakian manifolds with
cohomogeneity 2 isometric actions for which Theorem 1.6.35 applies.
CHAPTER 2
Foliations
Foliations were dened in Example 1.4.13. It is the purpose of this section to
gather together those well-known facts concerning foliations that we shall need in
the sequel. The theory of foliations is very well developed, there being many books
on the subject [CC00, CC03, CL85, God91, HH86, Mol88, Rei83, Tam92,
Ton97]. The modern theory began with the work of Ehresmann and Reeb in the
1940s (cf. [ER44]), but as with much in mathematics its roots go back to the 19th
century. The type of foliations that we are mainly concerned with in this book are
the so-called Riemannian foliations. This type of foliation has been developed in
the books of Reinhart [Rei83], Molino [Mol88] and Tondeur [Ton97].
2.1. Examples of Foliations
We give here several examples of foliations which should provide insight later
on. Only those examples which illustrate concepts pertinent to our development
are given.
Example 2.1.1: Submersions. As mentioned previously a submersion 1 : `.
where ` and are smooth manifolds, is a very special case of a foliation, called
a simple foliation. A bre bundle is a special case of a submersion; however, not
every submersion satises the local triviality condition. If one removes a point from
a bre of a bre bundle, one still has a submersion, but local triviality fails. For
a more interesting examples of submersions that are not locally trivial see [CC00]
pages 6-7. The Ehresmann Fibration Theorem states that if 1 is proper then a
submersion is a bre bundle.
The next example illustrates phenomena that will re-occur in our work.
Example 2.1.2: Linear ows on tori. For simplicity we consider the two dimen-
sional torus T
2
dened as the quotient space by the integer lattice, viz. R
2
(ZZ).
A linear ow is the ow generated by the vector elds
A = o
1

r
1
+o
2

r
2
where o
i
R for i = 1. 2 are not both zero. For each such pair (o
1
. o
2
) R
2
(0. 0)
the vector eld A generates a subbundle 1 of TT
2
. and thus denes a 1-dimensional
foliation T on T
2
. (Frobenius integrability is automatic for 1-dimensional distribu-
tions). The leaf passing through the point (r
1
(0). r
2
(0)) T
2
is given by the image
of the ow : RR
2
dened by
(t) = (r
1
(0) +o
1
t. r
2
(0) +o
2
t).
We assume that o
1
= 0. There are two cases to consider:
51
52 2. FOLIATIONS
(i) The ratio
a
2
a
1
is rational. In this case the vector eld A is periodic and
the leaves of the foliation are circles. The foliation T is simple, and in
fact describes T
2
as an o
1
bundle over o
1
.
(ii) The ratio
a
2
a1
is irrational. In this case the leaves are dieomorphic to
R. and each leaf is dense in T
2
. Hence, the topology of the quotient
space T
2
T is not even a T
1
space (referring to the separation axioms of
point-set topology).
Exercise 2.1: Prove the assertions (i) and (ii) of Example 2.1.2.
Our next example generalizes a well-known result in Lie theory.
Example 2.1.3: A locally free action of a Lie group. Let G be a Lie group
which acts locally freely on a smooth manifold `. that is there is a smooth map
G `` such that the isotropy subgroup at any point is a discrete subgroup
of G. Let g denote the Lie algebra of G. The action induces a monomorphism
: g(T`) of Lie algebras, and since this action is locally free the image (g)
generates a subbundle 1 of T` of dimension (dim g). Furthermore, since (g) is
subalgebra of (T`). the subbundle 1 is integrable and denes a foliation T on
`. In the next chapter we show that if the action of G is also proper and there
is a Riemannian metric on ` with isometry group G. then the space of leaves
`G is an orbifold such that the natural projection : ``G is an orbifold
Riemannian submersion. This generalizes to the case of locally free action the well-
known theorem of Lie theory that says that if G acts properly and freely on `.
then the space of leaves of T. i.e., the orbit space `G of the action is a smooth
manifold such that ` is the total space of a principal G-bundle over `G.
2.2. Haeiger Structures
Let (l

) be a foliated atlas with coordinates

= (r
()
1
. . . . . r
()
p
; n
()
1
. . . . . n
()
q
).
Then there are local submersions 1

: l

R
q
dened to be the composition of

: l

R
p+q
= R
p
R
q
with projection onto the second factor. The inverse
image 1
1
(n) of a point n R
q
denes a plaque of T. that is a component of
l

. In the overlap of foliated charts l

the local submersions 1

. 1

are
related by
(2.2.1) 1

.
where the transition functions

: 1

(l

)1

(l

) are dieomorphisms
that satisfy Haeigers cocycle conditions
(2.2.2)

=
1

on l

on l

.
This point of view has particular interest in regard to transverse G-structures which
are dened below in Denition 2.5.2.
Haeiger [Hae62] has generalized this notion in order to construct a correct
homotopy theory for foliations. It has also been generalized by using other pseu-
dogroups, cf. [Hae71].
Denition 2.2.1: Let be a subpseudogroup of the pseudogroup
GL(q,R)
of local
dieomorphisms of R
q
, and let A be a topological space. A Haeiger cocycle on
A is given by the following data:
(i) an open cover U = l

I
of A.
2.2. HAEFLIGER STRUCTURES 53
(ii) continuous maps 1

: l

R
q
. called local projections,
(iii) for each . 1 and r l

. a dieomorphism

q
from
a neighborhood of 1

(r) onto a neighborhood of 1

(r) where the germ

of the map 1

at the point 1

(r) varies continuously with r. such that

1

. and for all r l

one has

.
We remark that if the local projections 1

are open maps, then the second

condition in (iii) follows from the rst. It is also clear that if A is a smooth
manifold and the local projections 1

are local submersions we recover Haeigers

description of a foliation given above. In this case we say that the Haeiger cocycle
is a Haeiger cocycle of a foliation, and in the case of a proper subpseudogroup

q

GL(q,R)
, it is a Haeiger cocycle of a foliation with an additional transverse
structure. We shall consider this case in more detail in section 2.5. Denition 2.2.1
also makes perfect sense if we replace R
q
by an appropriate c-dimensional model
space . and
q
by any pseudogroup of transformations on .
Next we give an alternative denition of a Haeiger cocycle in terms of groupoids.
We refer to Appendix A for a brief discussion of groupoids. Actually this second
denition is a bit more general since it holds for any topological groupoid (. When
( is a groupoid of germs this denition is completely equivalent to Denition 2.2.1
by Proposition 1.5.4.
Denition 2.2.2: Let ( be a topological groupoid with A as its space of units. A
Haeiger cocycle on A consists of an open cover l

together with a family

of continuous maps o

: l

( such that for every r l

the
condition o

(r) = o

(r)o

(r) holds. Two such cocycles are equivalent if they

can be extended to a cocycle on the union of the covers.
Notice that local projection maps are implicit in this denition. They are
obtained by the composition 1

= : o

: l

A. We now take the groupoid (

to be a groupoid of germs ((
q
) of a pseudogroup
q
. Haeigers cocycle condition
2.2.2 is a cocycle for the cohomology set H
1
(U. ((
q
)). In this set two cocycles o

and o
t

are equivalent if there exist germs of dieomorphisms

q
such that
on 1

(l

) the relation o
t

holds. The equivalence stated in

Denition 2.2.2 is the following: we say that the Haeiger cocycle o

corresponding
to the cover U is equivalent to the Haeiger cocycle o
t

corresponding to the cover

U
t
if there is a Haeiger cocycle o
tt

on the disjoint union U . U

t
that restricts
to o

on U and to o
t

on U
t
. As usual by taking the inductive limit over ner
covers we obtain the cohomology set H
1
(A. ((
q
)) whose elements we refer to as

q
-structures or Haeiger structures. Unfortunately, the classes in H
1
(A. ((
q
))
are not invariant under homotopy, so we need to dene the set
q
(A) of homotopy
classes of elements in H
1
(A. ((
q
)). We say the two cocycles o
1
. o
2
H
1
(A. ((
q
))
are homotopic if there exists an element H
1
(A1. ((
q
)) such that

j
= o
j
.
where 1 = [0. 1]. and
j
: AA 1 are the natural face maps.
Exercise 2.2: Show that the property of being homotopic denes an equivalence
relation on H
1
(A. ((
q
)).
Now we are interested in the case when the Haeiger structure comes from a
foliation T on a smooth manifold `. In this case we have
54 2. FOLIATIONS
Proposition 2.2.3: Let ` be a smooth manifold with a foliation T of codimension
c. Then T determines a unique Haeiger structure (o) H
1
(`. ((
q
)). Further-
more, any two foliations that determine the same Haeiger structure are isomor-
phic.
Proof. Given a foliated atlas U = (l

) we obtain local submersions

1

: l

R
q
as above. This uniquely determines a cocycle o

by Equation
2.2.1 which determines a unique germ o
x

. and this germ determines a unique

Haeiger structure (o) H
1
(U. ((
q
)). Suppose that V = (\

) is another
foliated atlas for T. and o

: \

R
q
are the corresponding local submersions.
This determines a unique element () H
1
(V. ((
q
)). Let W = (\

) be a
common renement of U and V. By abuse of notation we let 1

. o

denote their
corresponding restrictions to \

. Then for each there is a dieomorphism

q
such that o

. So by the uniqueness of the dieomorphism in 2.2.1 and we

have

. Then by passing to germs this gives a unique class in

H
1
(W. ((
q
)). and hence, in H
1
(`. ((
q
)).
Suppose two foliations T and T
t
determine the same Haeiger structure, then
they determine the same element (o) H
1
(U. ((
q
)) for some common foliated
atlas U.
2.3. Leaf Holonomy and the Holonomy Groupoid
In this section we describe an important invariant of foliations, the holonomy
groupoid. We begin by describing the leaf holonomy of a given leaf, a concept
which generalizes the classical rst return map of Poincare from dynamical systems
theory. Let (`. T) be a foliated manifold with dimension of ` equal to n. and
the dimension of T equal to j. A submanifold

` is called a transversal or
transverse submanifold if at each point r we have
T
(x)
` = TL
x
+T
x
.
In order to construct leaf holonomy, we now describe, following [Mol88], a proce-
dure known as sliding along the leaves. Consider two points r
0
and r
t
0
lying on the
same leaf L
0
. Choose transversals and
t
through r
0
and r
t
0
, respectively, and
a piecewise smooth path : [0. 1]L
0
such that (0) = r
0
and (1) = r
t
0
. Subdi-
vide the interval [0. 1] into subintervals [t
1
. t

] with 0 = t
0
< t
1
< < t
k
= 1
such that each ([t
1
. t

]) lies in a simple open set l

of a foliated atlas. More-

over, since ([t
1
. t

]) is connected it lies on a unique plaque of L

0
l

. Let

be transversals at (t

). Then since (t
1
) lies on a single plaque in l
1
l

.
for each = 1. . . . . / there is a dieomorphism

from an open neighborhood

\
i1

1
about (t
1
) onto an open neighborhood \

about (t

)
which is constructed from the local submersions 1

: l

R
q
by the commutative
diagram
(2.3.1)
\
1

.
`
R
q
where the diagonal arrows are 1
1
[\
1
and 1

[\

. respectively. By looking at the

local foliated coordinate charts (l
1
; r
(1)
1
. . . . . r
(1)
p
. n
(1)
1
. . . . . n
(1)
q
) and
2.3. LEAF HOLONOMY AND THE HOLONOMY GROUPOID 55
(l

; r
()
1
. . . . . r
()
p
. n
()
1
. . . . . n
()
q
) one can easily identify the coordinate represen-
tative of

with the Haeiger cocycle o

(1)
. It follows that the composition

k1
=
k

1
: \
0
\
k
is a dieomorphism that depends on the subdivision
and the intermediate transversals only through their domains. It does, however,
depend on the choice of foliated atlas, and the transversals at r
0
and r
t
0
. The set of
all local dieomorphisms obtained in this way form a pseudogroup
/,N,N
. called
the holonomy pseudogroup of T. This is a bit of a misnomer; however, a dierent
choice of foliated atlas, and dierent choices of transversals give pseudogroups that
are locally conjugate, so it makes sense to speak of the holonomy pseudogroup.
Since the dieomorphism
k1
described above depends on the path, the subdi-
vision, and intermediate transversals through their domains, it makes sense to pass
to germs. We denote the germ of the dieomorphism
k1
at r
0
by /

. This is now
independent of the subdivision and transversals. Notice also that if
t
is another
path with the same endpoints and : [0. 1] [0. 1]

is a homotopy between
and
t
. then /

only depends on the homotopy class [] of (cf. [CC00]). Now

suppose that r
t
0
= r
0
in which case we can take
t
= . Then /

denes the
germ of a local dieomorphism of which leaves r
0
xed. We let (
x0
x
0
denote the
set of all such germs obtained by sliding along leaves. (The notation will be clear
shortly). This set forms a group under composition. Notice that if and
t
are
two loops at r
0
. then /

= /

. so one gets a group epimorphism

(2.3.2) /
x
0
:
1
(L
0
. r
0
)(
x0
x
0
Diff
x
0
.
where Diff
x
0
denotes the group of germs of dieomorphisms of that x r
0
. It
can be identied with the group of germs of local dieomorphisms of R
q
which x
the origin.
Denition 2.3.1: The group (
x0
x0
is called the holonomy group of the leaf L
0
at
r
0
.
If a dierent transversal
t
is chosen at r
0
, then one can identify the germ of
the transversal
t
with the germ of the transversal . So the leaf holonomy group
is well dened up to this identication. Furthermore, if r
1
L
0
is another point of
the same leaf, any path : [0. 1]L
0
satisfying (0) = r
0
and (1) = r
1
induces
an isomorphism

: (
x0
x
0
(
x1
x
1
.
where

(/
x0
) = /

/
x0
/

1. Thus, it makes sense to talk about the holonomy

group Hol(L) of a leaf L.
Proposition 2.3.2: Let T be a simple foliation. Then Hol(L) is trivial for all
leaves L of T.
Proof. Since the foliation is simple the local submersions in diagram 2.3.1 are
the restrictions of a global submersion 1. Thus, after identifying \
1
with \

by
translation along the leaves, the map

in 2.3.1 is the identity, so sliding along the

leaves produces the germ of the identity dieomorphism.
The following proposition says that for any foliation the generic leaf has
trivial holonomy.
Proposition 2.3.3: For any foliated manifold the leaves having a trivial holonomy
group form a residual (i.e., a countable intersection of dense open sets) set.
Proof. Cf. [CC00].
56 2. FOLIATIONS
The kernel of the homomorphism /
x
of Equation 2.3.2 is a normal subgroup
of the fundamental group of the leaf called the holonomy kernel, and denote by
Holker(L). A space (leaf) whose fundamental group is Holker(L) is called the ho-
lonomy covering. We consider several examples regarding holonomy. The rst is a
continuation of Example 2.1.2.
Example 2.3.4: Linear ows on tori. Again for simplicity we consider only T
2
.
In both the irrational and rational cases the holonomy groups are all trivial, but
for dierent reasons. In the rational case the foliation is simple, so the holonomy is
trivial. In the irrational case the leaves are simply connected, so the holonomy is
again trivial. However, in this latter case the holonomy pseudogroup is nontrivial.
The next example is more interesting and plays an important role in the sequel.
Example 2.3.5: Linear ows on odd dimensional spheres. Consider the unit
sphere in C
n+1
dened by
o
2n+1
= z = (.
0
. . . . . .
n
) C
n+1
[ [.
0
[
2
+ +[.
n
[
2
= 1.
By linear ow we mean linear in the complex Cartesian coordinates z = (.
0
. . .
n
).
Again for simplicity we consider the case n = 1. Let n
0
and n
1
be two non-zero
real numbers, and consider the real vector eld on C
2
= i

n
0
.
0

.
0
+n
1
.
1

.
1
c.c

.
Restricted to o
3
this vector eld is everywhere tangent to o
3
. and denes a nowhere
vanishing vector eld on o
3
. Thus, generates a 1-dimensional foliation on o
3
whose
ow is given by
(.
0
. .
1
) (c
2iw
0
t
.
0
. c
2iw
1
t
.
1
).
Again there are two cases to consider depending on whether the ratio
w1
w
0
is rational
or irrational. If the ratio is rational by a reparameterization and considering the
complex conjugate coordinates if necessary, we can take the pair (n
0
. n
1
) to be
coprime positive integers. In this case all the leaves are circles. The leaf through
r
0
= (1. 0) is a circle which returns to r
0
at t =
1
w
0
. whereas the nearby leaves with
.
1
= 0 take n
0
times as long to return. So the holonomy group Hol(L
x
0
) Z
w
0
.
Similarly we see for r
1
= (0. 1) the holonomy through the leaf L
x
1
is Hol(L
x
1
)
Z
w
1
. We also see that Holker(L
i
) n
i
Z. All other leaves have trivial holonomy.
Now consider the case when the ratio
w1
w0
is irrational. Consider the leaf through
a point with .
0
.
1
= 0. say the point (
1

2
.
1

2
). This leaf is non-compact; in fact,
it is the irrational ow on the torus T
2
in o
3
dened by [.
0
[
2
= [.
1
[
2
=
1
2
. So this
leaf is not closed in o
3
. Its closure is just the 2-torus T
2
. The same phenomenon
occurs for any point (.
0
. .
1
) o
3
with .
0
.
1
= 0. However, the two leaves through
r
0
= (1. 0) and r
1
= (0. 1) are circles for they return after a time t =
1
w0
and t =
1
w1
.
respectively. Nevertheless, all leaves have trivial holonomy which can be seen from
(v) of Theorem 2.5.10 below, but as in Example 2.3.4 the holonomy pseudogroup
is non-trivial.
The holonomy groups attached to a foliation can be studied together by form-
ing a more global object, the holonomy groupoid. Although the ideas go back to
Ehresmann, Reeb, and Thom, it appears that its rst precise formulation was given
by Winkelnkemper [Win83]. See also [Rei83].
2.3. LEAF HOLONOMY AND THE HOLONOMY GROUPOID 57
Proposition 2.3.6: Let (`. T) be a foliated manifold. The collection of all triples
(r. n. []), where r. n ` lie on the same leaf L, is a piecewise smooth path
from r to n, and [] denotes the holonomy equivalence class of . i.e.,
t
if
t

1
= c (
y
y
. is a Lie groupoid of dimension n + j called the holonomy
groupoid
1
of (`. T) and denoted by ( = ((`. T). Furthermore, if every holonomy
germ is conjugate to an analytic germ, ((`. T) is Hausdor.
Proof. In Appendix A.1 we give a brief review of groupoids. Clearly, the set
of objects G
0
is ` and this is a smooth Hausdor manifold, while G
1
is the set
of triples (r. n. []). We show that G
1
is a smooth manifold by following Connes
[Rei83] and exhibiting a system of coordinates on it. We then check that the
structure maps are smooth. Write the coordinates as (. n) and (
t
. n) having the
same transverse coordinate n at the beginning and end of a path, respectively. Here
and
t
are coordinates along the leaf. Then (
t
. . n) is a coordinate system on
G
1
. and a change of coordinates is of the form
1(
t
. . n) = (1
t
(
t
. n). 1(. n). o(n)).
where (. n) (1(. n). o(n)) is a change of coordinates on `. This gives G
1
a
locally Euclidean topology of dimension n+j that is not necessarily Hausdor. This
topology is also 2nd countable (see [Rei83], pg 137). Moreover, G
1
is Hausdor if
and only if each holonomy germ at r ` which is the identity on some open set
whose closure contains r is the identity germ at r [Rei83, Win83]. In particular,
if every holonomy germ is conjugate to an analytic germ, the groupoid ((`. T)
will be Hausdor.
The canonical source and target maps : : (` and t : (`. are dened
by
(2.3.3) :(r. n. []) = r. t(r. n. []) = n.
while the unit is n(r) = (r. r. 0), where 0 denotes the constant loop. Further-
more, the associative multiplication on ( is dened when the range of one element
coincides with the source of the other, specically in our case we have
(2.3.4) (r. n. []) (r
t
. n
t
. [
t
]) = (r. n
t
. [
t
]) if n = r
t
The inverse map is dened by (r. n. []) = (n. r. [
1
]). To see that the structure
maps are smooth, we write in coordinates:
:(
t
. . n) = (. n). t(
t
. . n) = (
t
. n). n(. n) = (. . n).
(
t
. . n) = (.
t
. n). (
tt
.
t
. n) (
t
. . n) = (
tt
. . n).
So the structure maps are smooth. Furthermore, both : and t are submersions.
This completes the proof.
The fact that the holonomy groupoid ((`. T) of a foliation is, in general, not
Hausdor is of very little consequence for us, since it known to be Hausdor when
the foliation T is Riemannian (see Denition 2.5.4 and Theorem 2.5.10 below)
which is the case of interest to us. Notice that T induces a foliation
2
of dimension
2j on ((`. T) by pulling back the foliation T to ((`. T) by the source map :. We
denote this foliation on ((`. T) by T
(
. If 1 denotes the integrable subbundle of
T` corresponding to the foliation T, then the integrable subbundle of T((`. T)
1
This is called the graph of the foliation in [Win83].
2
A dierent 2p-dimensional foliation having trivial holonomy groups is described in [Win83].
58 2. FOLIATIONS
corresponding to T
(
is :

1 1. where 1 is the vertical bundle consisting of the

tangent vectors to the bres of the submersion :. So the leaf L
(x,y,[])
of T
(
through
(r. n. []) ((`. T) projects under : and t to the leaf L
x
of T. Furthermore,
these projections induce an isomorphism of the corresponding holonomy groups,
see [Rei83] Proposition IV.2.44. Notice that the both subbundles :

1 and 1 of
T((`. T) are separately integrable. Summarizing we have
Proposition 2.3.7: Let T be a j-dimensional foliation on a manifold `. Then
the foliation ((`. T) T pullsback under either : or t to a 2j-dimensional foliation
T
(
on the holonomy groupoid ((`. T) that is invariant under the involution . If
L
(x,y,[])
denotes the leaf of T
(
through (r. n. []) ((`. T) and L
x
denotes the
leaf of T through r `. then there is an isomorphism Hol(L
(x,y,[])
) Hol(L
x
).
More information about the holonomy groupoid ((`. T) can be ascertained
when T is a Riemannian foliation. This will be described later in this section. Now
we consider another important groupoid introduced by Haeiger [Hae84] which he
called the transverse holonomy groupoid. It is closely related to, but dierent from
((`. T). Here is the denition.
Denition 2.3.8: Let be a complete transversal. The transverse holonomy
groupoid associated to is the full subgroupoid (
N
N
of ((`. T).
(
N
N
is dened in the Appendix by Equations A.1.2. Here complete means that
intersects each leaf at least once. Complete transversals always exist since we
do not require them to be connected. Of course, (
N
N
depends on . but Haeiger
[Hae84] shows that (
N
N
is independent of up to Morita equivalence. In fact
Theorem 2.3.9: Let T be a j-dimensional foliation on a manifold of dimension
j+c. and let be any complete transversal. Then the transverse holonomy groupoid
(
N
N
is an etale Lie groupoid of dimension c that is Morita equivalent to the holonomy
groupoid ((`. T).
Proof. (
N
= t
1
() is a submanifold of G
1
such that
codim((
N
G
1
) = codim( G
0
) = j.
Since G
1
has dimension 2j+c we see that dim (
N
= j+c = dim G
0
. Now consider
the source map :
N
= :[
(
N restricted to (
N
. If we show that this map :
N
: (
N
G
0
is an immersion, it will be a local dieomorphism. Then (
N
N
= :
1
N
() will be
a submanifold of (
N
of dimension c. So the restriction of :
N
to (
N
N
will be a
local dieomorphism. Thus, (
N
N
will be etale. To see that the dierential d:
n
is
injective at a morphism (o : rn) (
N
G
1
. we consider a tangent vector
T
g
G
1
. It will be tangent to (
N
if (dt)
g
() T
y
. Since : : G
1
G
0
is
a submersion, it denes a j-dimensional foliation

T on G
1
whose leaves are the
connected components of the bres of the submersion. Thus, if d:
N
() = 0 with
tangent to (
N
. then is both tangent to the bres of the source submersion and
dt
g
() is tangent to . But (dt)
g
maps the tangent space to the leaf

L
g
of

T at o to
the tangent space to the leaf L
y
of T at n. Thus, = 0 which proves the injectivity
of d:
N
. or equivalently, that :
N
is a local dieomorphism.
To prove the last statement we consider the smooth functor 1 : (
N
N
((`. T)
dened by 1(r. n. [o]) = (
N
(r).
N
(n). [o]). We see from the fact that the transversal
is complete that :
N
is surjective. This shows that the map :
2
:
G0
G
1
G
0
on pairs (r. o) satisfying
N
(r) = t(o) is surjective. Moreover, one easily checks
2.4. BASIC COHOMOLOGY 59
that corresponding square is a pullback. Thus, 1 : (
N
N
((`. T) is an essential
equivalence, and hence (
N
N
and ((`. T) are Morita equivalent.
In particular, Theorem 2.3.9 says that the etale groupoid (
N
N
is independent of
up to Morita equivalence. We shall see later that Morita equivalent groupoids
have homotopy equivalent classifying spaces. We will consider (
N
N
further in the
Chapter 4.
We are most interested in the case where the foliation arises from a locally free
action of a Lie group (2.1.3) with nite isotropy groups. So we recall the action
groupoid from the Appendix (Denition A.1.5). For G a Lie group and ` a smooth
manifold, we consider the action groupoid `G. According to Reinhart ([Rei83],
pg. 138) the space of morphisms of the holonomy groupoid of a foliation coming
from a locally free action of a Lie group G is just ` G. So in this case these two
groupoids are isomorphic.
2.4. Basic Cohomology
Basic cohomology was introduced by Reinhart in [Rei59b]. We begin with a
brief review following [Ton97]. Let T be a foliation of a smooth manifold `. A
dierential :-form on ` is said to be basic if for all vector elds \ on ` that
are tangent to the leaves of T the following conditions hold:
(2.4.1) \ = 0.
V
= 0.
In a local foliated coordinate chart (l. ) with = (r
1
. . . . . r
p
; n
1
. . . . . n
q
) a basic
:-form takes the form
(2.4.2) =

i
1
...i
r
(n
1
. . . . . n
q
)dn
i
1
dn
i
r
.
where the sum is taken over all repeated indices.
Let
r
B
(T) denote sheaf of germs of basic :-forms on `. and
r
B
(T) the set of
its global sections. The direct sum
B
(T) =
r

r
B
(T) is closed under addition and
exterior multiplication, and so is a subalgebra of the algebra of exterior dierential
forms on `. Furthermore, exterior derivation takes basic forms to basic forms, that
is,
(2.4.3)
V
d = d
V
= 0. and \ d =
V
d(\ ) = 0.
Thus, the subalgebra
B
(T) forms a subcomplex of the de Rham complex, and
its cohomology ring H

B
(T) = H

B
(T. d
B
) is called the basic cohomology ring of
T. Here d
B
denotes the restriction of the exterior derivative d to the basic forms

B
(T).
The groups H
r
B
(T) are dened for all 0 : c, where c is the codimension
of the foliation, but generally they may be innite dimensional for : 2. However,
they are nite dimensional for Riemannian foliations on compact manifolds which
is mainly the case of interest to us. Let us look at the groups H
r
B
(T) for : = 0. 1.
The set
0
B
(T) is just the set of functions that are constant along the leaves of
T. so the cocycles with respect to d
B
are just the constant functions. Thus, if
` is connected H
0
B
(T) R. Let us now consider H
1
B
(T). We claim that H
1
B
(T)
injects into H
1
(`. R) induced by the natural inclusion
B
(T)(`). Let be
a 1-cocycle, i.e., d
B
= 0. where is a smooth section of
1
B
(T). So if = d1 for
some smooth function on `. then we have \ 1 = \ d1 = \ d = 0 for any
60 2. FOLIATIONS
vector eld \ tangent to the leaves of T. But this says that 1
1
B
(T). so is a
coboundary with respect to d
B
. This proves the claim. We have arrived at
Proposition 2.4.1: For any foliation T on a connected manifold `.
(i) H
0
B
(T) H
0
(`. R) R.
(ii) H
1
B
(T) is a subgroup of H
1
(`. R).
The appropriate notion of basic for vector elds is
Denition 2.4.2: A vector eld A on a foliated manifold (`. T) is said to be
foliate with respect to T if for every vector eld \ tangent to the leaves of T. the
Lie bracket [A. \ ] is tangent to the leaves of T.
Let us set some notation. The Lie algebra of all smooth vector elds on ` is
denoted by (`). We denote by
J
(`) the Lie subalgebra of (`) consisting of
those vector elds that are tangent to the leaves of the foliation T. These are just
the smooth sections of the integrable vector bundle 1 dening T. The set of foliate
vector elds also forms a Lie subalgebra of (`). It is denoted by fol(`. T). From
Denition 2.4.2 one sees that it is precisely the normalizer of
J
(`) in (`).
Equivalently, its local one parameter group preserves the foliation. We denote the
subgroup of the dieomorphism group Diff(`) that preserve the foliation T by
Fol(`. T). that is
(2.4.4) Fol(`. T) = Diff(`) [

1 1.
Exercise 2.3: Show that in local foliated coordinates (r
1
. . . . . r
p
; n
1
. . . . . n
q
) a fo-
liate vector eld A takes the form
(2.4.5) A =
q

a=1

a
(n
1
. . . . . n
q
)

n
a
+
p

i=1
1
i
(r
1
. . . . . r
p
; n
1
. . . . . n
q
)

r
i
.
where
a
and 1
i
are smooth functions of the indicated variables.
2.5. Transverse Geometry
We are interested in foliations with a transverse structure. These were intro-
duced by Conlon [Con74] and further developed by Molino [Mol75, Mol88] in
a slightly dierent guise. Here we follow [Mol88], but our notation is somewhat
dierent. Let T be a j-dimensional foliation on the manifold ` with integrable
subbundle 1. There is an exact sequence of vector bundles, viz.
(2.5.1) 01T`

Q
Q0.
The quotient bundle Q is called the normal bundle of the foliation, and it is often
denoted by (T) to emphasize the foliation. Notice that any foliate vector eld A
projects to a section

A of (T) that is independent of the coordinates along the
leaves. This follows easily from the coordinate form 2.4.5 together with a change in
foliated coordinate charts. The set of all such sections forms a Lie algebra, denoted
by trans(`. T). and we have an exact sequence of Lie algebras
(2.5.2) 0
J
(`)fol(`. T)trans(`. T)0.
Here the Lie bracket on trans(`. T) is dened by [

A.

Y ] = [A. Y ]. We call elements
of trans(`. T) transverse vector elds. Note also that the Lie derivative of basic
tensor elds with respect to transverse vector elds is well dened [Mol88].
2.5. TRANSVERSE GEOMETRY 61
A transverse frame at the point r ` is a basis (Y
1
. . . . . Y
q
) for the bre Q
x
of
Q at r. (Y
i
is not a tangent vector at r, but an equivalence class of tangent vectors
modulo vectors in 1
x
.) The set 1
T
(`. T) of all transverse frames at all points of
` forms a principal G1(c. R) bundle on `. called the transverse frame bundle on
(`. T). Let G G1(c. R) be a Lie subgroup and let
T
: 1
T
(`. G. T)` denote
the corresponding principal G subbundle of 1
T
(`. T). As with the ordinary frame
bundle a point . 1
T
(`. T) can be viewed as the linear map . : R
q
Q

T
(z)
that assigns to the standard basis of R
q
the frame . = (Y
1
. . . . . Y
q
). Also as in the
usual case 1
T
(`. T) and its subbundles 1
T
(`. G. T) have a canonical R
q
-valued
1-form
T
dened by
(2.5.3) '
T
. A` = .
1

T
A.
The foliation T on ` lifts to a foliation T
T
on 1
T
(`. T) as follows: we dene a
foliation on 1
T
(`. T) by the j-dimensional subbundle 1
T
T1
T
(`. T) generated
by the vectors in T
z
1
T
(`. T) that satisfy
(2.5.4) 1
Tz
= A
z
T
z
1
T
(`. T) [ A
z

T
= A
z
d
T
= 0 for all . 1
T
(`. T).
We now have
Proposition 2.5.1: Let (`. T) be a foliated manifold and let 1
T
(`. T) be its
transverse frame bundle. The subbundle 1
T
dened by equation 2.5.4 is integrable,
and thus denes a foliation T
T
on 1
T
(`. T). The leaves of this foliation project
under
T
to the leaves of the foliation T.
Proof. Let A. Y be vector elds on 1
T
(`. T) that are sections of the sub-
bundle 1
T
. Then
0 = d
T
(A. Y ) = A
T
(Y ) Y
T
(A)
T
([A. Y ]) =
T
([A. Y ]).
so [A. Y ]
T
= 0. But also we have for any section A of 1
T
that

T
= d(A
T
) +A d
T
= 0.
Thus,
0 = [
X
.
Y
]
T
=
[X,Y ]

T
= d([A. Y ]
T
) + [A. Y ] d
T
= [A. Y ] d
T
which implies that 1
T
is integrable by Frobenius Theorem.
Now suppose that L
T
is a leaf of T
T
. Let l be a simple open set of ` such
that the quotient map of the foliation restricted to l is a submersion 1
U
: lR
q
.
Then on 1
T
[l =
1
T
(l), the canonical 1-form
T
is just the pullback 1

U
. where
is the canonical 1-form on the frame bundle 1(R
q
) on R
q
. So in local coordinates
(r
1
. . . . . r
p
; n
1
. . . . . n
q
;
1
1
. . . . .
q
1
. . . . .
q
q
) on 1
T
. equation 2.5.4 implies that the
tangent vectors to 1
T
that are tangent to the leaves of the foliation T
T
are spanned
by the

x
j
and

A
i
j
. Thus, these vectors project under
T
to vectors that are tangent
to the leaves of the foliation T on `. But since
T
is surjective this implies that
the leaf L
T
of T
T
projects to a leaf of T.
The foliation T
T
is called the lifted foliation. We are now ready for
Denition 2.5.2: Let G be a Lie subgroup of G1(c. R). and let 1
T
(`. G. T) be
a principal G subbundle of 1
T
(`. T). We say that 1
T
(`. G. T) is a transverse
G-structure if at each point . 1
T
(`. G. T) the tangent space T
z
1
T
(`. G. T)
contains the subspace 1
Tz
tangent to the leaves of the lifted foliation T
T
.
62 2. FOLIATIONS
Important examples for us will be when G = O(c. R). G = G1(:. C). and
G = l(:). where c = 2:. These correspond to transverse Riemannian structures,
transverse almost complex structures, and transverse almost Hermitian structures,
respectively. The following result [Mol88] gives a characterization of transverse
G-structures in terms of the local submersions.
Proposition 2.5.3: Let 1
T
(`. G. T) be a principal G subbundle of 1
T
(`. T).
Then 1
T
(`. G. T) is a transverse G-structure if and only if for every simple open
set l. the bundle 1
T
(`. G. T)[
U
is the pullback by
T
: 1
T
(l. T[
U
)1(R
q
) of a
G-structure 1(R
q
. G) on R
q
.
Proof. The argument is purely local, and the details are left to the reader.
2.5.1. Transverse Riemannian Geometry. The foliations that interest us
most are those with a transverse Riemannian structure and actually more, a trans-
verse Kahler structure. It turns out that allowing for a transverse Riemannian
structure adds quite a bit of tractability to the subject which was begun by Rein-
hart [Rei59a], and subsequently developed by Molino and others. See [Mol88]
and references therein.
Denition 2.5.4: A Riemannian foliation is a foliation with a transverse Rie-
mannian structure, i.e., a transverse G-structure with G = O(c. R).
A Riemannian foliation (`. T) denes a transverse Riemannian metric o
T
on
` by
(2.5.5) o
T
(A. Y ) = '.
1

Q
A. .
1

Q
Y `.
where A. Y T`

T
(z)
and
Q
is the projection in the exact sequence 2.5.1. Notice
that o
T
denes a nonnegative symmetric bilinear form on T`. and viewed as a
section of Sym
2
T` has kernel 1T` +T`1. and thus denes a Riemannian
metric on the quotient bundle Q.
Proposition 2.5.5: Let T be a codimension c foliation on `. There is a 1-1 corre-
spondence between transverse O(c. R)-structures on ` and transverse Riemannian
metrics on `.
The second denition is that of a bundle-like metric which is due to Reinhart
[Rei59a].
Denition 2.5.6: A Riemannian metric o is said to be bundle-like with respect
to a foliation T if for any foliate horizontal vector elds A. Y , the function o(A. Y )
is basic, i.e., for any vector eld \ along the leaves of T the equation \ o(A. Y ) = 0.
The following is well-known [Mol88]
Proposition 2.5.7: If o is a bundle-like metric with respect to a foliation T, then
T is a Riemannian foliation. Conversely, if T is a Riemannian foliation with trans-
verse metric o
T
. then there exist bundle-like metrics o whose associated transverse
metric is o
T
.
Proof. Let o be a bundle-like metric on `. This metric splits the exact se-
quence (2.5.1) as
(2.5.6) T` = 1 1

.
2.5. TRANSVERSE GEOMETRY 63
where 1

denotes the subbundle of T` consisting of all vectors perpendicular to

the vectors of 1 with respect to the Riemannian metric o. Thus, any vector eld
A on ` can be decomposed as
A = A

+A

.
where A

is a section of 1 and A

is a section of 1

. Dene a transverse met-

ric on (`. T) by o
T
(A. Y ) = o(A

. Y

). Since o is bundle-like, this makes T a

Riemannian foliation.
Conversely, given a Riemannian foliation, we have a transverse metric o
T
de-
ned by Equation 2.5.5, and we can choose a Riemannian metric in the vector
bundle 1 to obtain a bundle-like metric on `.
In terms of Haeiger structures of Section 2.2, the transverse metric o
T
can be
obtained by pulling back Riemannian metrics on R
q
by the local submersions 1

.
In this case the transition functions

local isometries between local Riemannian

metrics. In fact one can give an equivalent denition of a Riemannian foliation in
terms Haeiger cocycles:
Proposition 2.5.8: Let (`. T) be a foliated manifold, and let be the associated
Haeiger cocyle. Then (`. T) is a Riemannian foliation if and only if there are
Riemannian metrics o

on each 1

(l

) such that o

.
Proof. Given a foliation T whose associated Haeiger structure satises o

with Riemannian metrics o

on each open set 1

(l

) R
q
. we obtain a
transverse Riemannian metric o
T,
on each open set l

and the cocycle conditions

imply that these metrics agree on each overlap l

giving a global transverse

Riemannian metric o
T
. Hence, the foliation is Riemannian. Conversely, if T is
Riemannian we have a transverse Riemannian metric o
T
which extends to a bundle-
like metric o on `. Since for any foliate vector elds A. Y the function o(A. Y ) is
basic, we can dene Riemannian metrics o

on each 1

(l

) by o

(1

A. 1

Y ) =
o[
U
(A. Y ). It is then easy to check that o

.
For a Riemannian foliation on a compact manifold we have [Mol88, Rei83]
Proposition 2.5.9: Let (`. T) be a Riemannian foliation on a compact connected
manifold `. Then the leaves of T all have the same universal cover.
In fact a common covering of the leaves appears naturally in the holonomy
groupoid ((`. T) of a Riemannian foliation. Generally, for a Riemannian foliation
((`. T) has some nice special properties. We say that a Riemannian submersion
is horizontally complete if any horizontal geodesic can be extended for all values of
its ane parameter.
Theorem 2.5.10: Let T be a j-dimensional Riemannian foliation on the manifold
`. Then
(i) The holonomy groupoid ((`. T) of T is a Hausdor Lie groupoid.
(ii) There is a unique Riemannian metric o
(
on ((`. T) such that both :
and t are Riemannian submersions.
(iii) The inverse map is an isometry with respect to o
(
.
(iv) The 2j dimensional foliation T
(
on ((`. T) is a Riemannian foliation
such that each leaf is locally the Riemannian product of leaves of the
j-dimensional foliations.
64 2. FOLIATIONS
(v) If the source submersion : : ((`. T)` is horizontally complete, then
it is a locally trivial bre bundle whose bres are precisely the holonomy
covers of the leaves of T.
(vi) If ` is compact and all leaves are compact, then ((`. T) is compact.
Proof. (i): The holonomy groupoid is a Lie groupoid by Proposition 2.3.6.
Since the transverse geometry of T is orthogonal, so are the leaf holonomy groups.
Thus, the holonomy is linear, and hence, analytic. So the result follows by Propo-
sition 2.3.6. For the remainder see Proposition IV.4.23 of [Rei83].
Next we give a theorem of Molino [Mol88] that is of great importance to us in
the sequel. It says that a Riemannian foliation with compact leaves has a tractable
space of leaves, namely an orbifold which we discuss in detail in Chapter 4.
Theorem 2.5.11: Let T be a j-dimensional Riemannian foliation with compact
leaves on the manifold `. Then the space of leaves `T admits the structure of
a j-dimensional orbifold such that the canonical projection : ``T is an
orbifold Riemannian submersion.
Proof. The proof can be found in [Mol88], Proposition 3.7.
Now we discuss the relation between the curvature of (`. o) and the curvature
of the transverse metric o
T
. where o is a bundle-like metric whose transverse com-
ponent is o
T
. Accordingly we split T` as in 2.5.6. This relationship is essentially
that of ONeill for Riemannian submersions [ON66, Rei83, Ton97]. We rst
dene the induced connection
T
in the bundle 1

.
(2.5.7)
T
X
Y =

E
(
X
Y ) if A is a smooth section of 1

E
[\. Y ] if A = \ is a smooth section of 1,
where Y is a smooth section of 1. It is left as an exercise to mimic the standard
proof of the rst fundamental theorem of Riemannian geometry that
T
is the
unique torsion free connection such that the transverse metric o
T
is parallel, i.e.,

T
X
o
T
= 0.
Exercise 2.4: Show that the connection
T
is the unique torsion-free metric con-
nection with respect to the transverse metric o
T
.
We let 1
T
denote the Riemannian curvature tensor eld with respect to o
T
.
Then if 7. \ are any vector elds on `. and A is a vector eld in 1

. we have
(2.5.8) 1
T
(7. \)A =
T
Z

T
W
A
T
W

T
Z
A
T
[Z,W]
.
Proposition 2.5.12: For any smooth section \ of 1 the following hold:
(i)
V

T
= 0.
(ii) \ 1
T
= 0.
(iii)
V
1
T
= 0.
Proof. (i) follows as usual from the fact that
T
is determined uniquely in
term the transverse metric o
T
and
V
o
T
= 0. (ii): If \ is a vector eld on ` and
A is a smooth section of 1

we have
1
T
(\. \)A =
T
V

T
W
A
T
W

T
V
A
T
[V,W]
A
=
V

T
W
A
T
W

V
A
T

V
W
A = (
V

T
)
W
(A) = 0.
We leave the proof of (iii) as an exercise.
2.5. TRANSVERSE GEOMETRY 65
Exercise 2.5: Prove (iii) of Proposition 2.5.12.
Let us x some common notation [Bes87, Ton97]: we let 1
i
denote any vector
elds on `. while we let l. \. \ denote vector elds in the integrable subbundle 1.
and A. Y. 7 be vector elds in 1

. We also let : T`1. and

: T`1

denote the corresponding natural projections. Following ONeill [ON83] (see also
Gray [Gra67]) we introduce the type (2. 1) tensor eld
(2.5.9) T
E
1
1
2
= (
(E1)

(1
2
)) +

(
(E1)
(1
2
)).
Exercise 2.6: Show that T dened by 2.5.9 is a tensor eld of type (2. 1).
Clearly, T only depends of projection of 1
1
along the leaves, so
(2.5.10) T
X
l = T
X
Y = 0.
We also have
T
U
\ =

(
U
\ ). T
U
A = (
U
A). (2.5.11)
T
U
\ = T
V
l. (2.5.12)
T is anti-symmetric, i.e., o(T
U
\. A) = o(\. T
U
A). (2.5.13)
The symmetry property Equation 2.5.12 follows by taking l and \ as coordinate
vector elds along the leaves of the foliation T (valid since T is a tensor eld),
and using the fact that is torsion-free, while the anti-symmetry property 2.5.13
follows since is a metric connection. Notice that by Equation 2.5.13 T vanishes
if and only if T
U
\ = 0 for all vector elds tangent to the leaves of T. Since any leaf
1 is an immersed submanifold of `. we see from Equation 2.5.11 that 1 is totally
geodesic if and only if T vanishes along 1. If T
U
\ = 0 for all l and \ then all
leaves of T are totally geodesic. In this case we say that the foliation T is a totally
geodesic foliation. Summarizing we have
Proposition 2.5.13: A foliation T is totally geodesic if and only if T = 0.
Reversing the roles of the subbundles 1 and 1

in Equation 2.5.9 we obtain

(2.5.14)
E
1
1
2
= (

(E1)

(1
2
)) +

(E1)
(1
2
)).
Exercise 2.7: Show that dened by 2.5.14 is a tensor eld of type (2. 1).
For ONeills tensor eld, we have the following

U
\ = 0.
U
A = 0. (2.5.15)

X
l =

(
X
l).
X
Y = (
X
Y ). (2.5.16)

X
Y =
Y
A. (2.5.17)
is anti-symmetric, i.e., o(
X
l. Y ) = o(l.
X
Y ). (2.5.18)
Equations 2.5.15 and 2.5.16 are obvious from the denition. Equation 2.5.17 will
be proved in Proposition 2.5.14 below, while again Equation 2.5.18 follows since
o = 0.
Proposition 2.5.14: For any horizontal vector elds A. Y we have

X
Y =
1
2
([A. Y ]).
Proof. We compute
([A. Y ]) = (
X
Y
Y
A) =
X
Y
Y
A.
66 2. FOLIATIONS
The result will now follow if we establish Equation 2.5.17. It suces to show that

X
A = 0. Without loss of generality we may take the horizontal vector eld A to
be foliate. We have for arbitrary vertical l
o(l.
X
A) = o(l. (
X
A)) = o(l.
X
A) = o(
X
l. A)
= o(
U
A. A) o([A. l]. A) =
1
2
lo(A. A) = 0.
The last expression vanishes since o(A. A) is basic when A is foliate.
Notice that Equation 2.5.18 implies that vanishes precisely when
X
Y van-
ishes for all horizontal vector elds A and Y. and Proposition 2.5.14 says that

X
Y vanishes for all A. Y if and only if the subbundle 1

is integrable. Next let

us see how the covariant derivative on ` decomposes. We let

be the induced
connection along the leaves. Then we have Gauss formula for the leaves
(2.5.19)
U
\ =

U
\ +T
U
\.
We also have

U
A = T
U
A +

(
U
A). (2.5.20)

X
l = (
X
l) +
X
l. (2.5.21)

X
Y =
X
Y +
T
X
Y. (2.5.22)
Concerning the covariant derivatives of and T we have the following:
Lemma 2.5.15: The following hold:
(i) (
U
)
V
=
T
U
V
. (
X
T)
Y
= T
A
X
Y
.
(ii) (
X
)
V
=
A
X
V
. (
V
T)
Y
= T
T
V
Y
.
(iii) o((
U
)
X
\. \) = o(T
U
\.
X
\) o(T
U
\.
X
\ ).
(iv) o((
E1
)
X
Y. \ ) is anti-symmetric in A. Y.
(v) o((
E
1
T)
U
\. A) is anti-symmetric in l. \.
(vi) So((
Z
)
X
Y. \ ) = So(
X
Y. T
V
7). where S denotes the cyclic sum
taken over A. Y. 7.
Proof. We establish the rst identity of (ii), and leave the rest as an exercise.
For any vector eld 1 we have
(
X
)
V
1 =
X
(
V
1)

X
V
1
V
(
X
1) =
A
X
V
1.
Exercise 2.8: Finish the proof of Lemma 2.5.15.
Theorem 2.5.16: Let 1.

1, and 1
T
denote the Riemann curvature associated to
the connections .

, and
T
. respectively Then the following curvature identities
2.5. TRANSVERSE GEOMETRY 67
hold:
o(1(l. \ )\. \
t
) = o(

1(l. \ )\. \
t
) +o(T
U
\. T
V
\
t
) o(T
V
\. T
U
\
t
);
o(1(l. \ )\. A) = o((
U
T
V
)\. A) o((
V
T
U
)\. A);
o(1(A. l)Y. \ ) = o(T
U
A. T
V
Y ) o((
X
T
U
)\. Y )
o(
X
l.
Y
\ ) o((
U
)
X
Y. \ );
o(1(l. \ )A. Y ) = o((
V
)
X
Y. l) o((
U
)
X
Y. \ )
+o(
X
\.
Y
l) o(
X
l.
Y
\ )
+o(T
U
A. T
V
Y ) o(T
V
A. T
U
Y );
o(1(A. Y )7. l) = o(
Y
7. T
U
A) +o(
Z
A. T
U
Y )
+o((
Z
)
X
Y. l) o(
X
Y. T
U
7);
o(1(A. Y )7. 7
t
) = o(1
T
(A. Y )7. 7
t
) + 2o(
X
Y.
Z
7
t
)
+o(
X
7.
Y
7
t
) o(
X
7
t
.
Y
7)
Proof. The rst equation is just Gauss Equation along the leaves viewed as
immersed submanifolds, while the second equation is the Codazzi equation along
the leaves. We leave the proof for the other equations to the reader.
We now have an easy corollary for the various sectional curvatures.
Corollary 2.5.17: Let A. Y. l. \ be linearly independent unit vector elds with
respect to the metric o. and let 1. 1
T
.

1 denote the sectional curvatures with respect
to the metrics o. o
T
. o[
1
. respectively. Then one has
1(l. \ ) =

1(l. \ ) +[T
U
\ [
2
o(T
U
l. T
V
\ ).
1(A. l) = o((
X
T)
U
l. A) [T
U
A[
2
+[
X
l[
2
.
1(A. Y ) = 1
T
(A. Y ) 3[
X
Y [
2
.
We would like expressions for the Ricci and scalar curvatures similar to those
of Theorem 2.5.16. First we need some preliminary development which we lift from
Besse [Bes87]. In what follows we let A
i

i
and l
a

a
denote local orthonormal
frames for the horizontal subspace 1

and vertical subspace 1. respectively. The

following identities hold:
o(
X
.
Y
) =

i
o(
X
A
i
.
Y
A
i
) =

a
o(
X
l
a
.
Y
l
a
). (2.5.23)
o(
X
. T
U
) =

i
o(
X
A
i
. T
U
A
i
) =

a
o(
X
l
a
. T
U
l
a
). (2.5.24)
o(l. \ ) =

i
o(
Xi
l.
Xi
\ ). (2.5.25)
o(TA. TY ) =

a
o(T
Ua
A. T
Ua
Y ). (2.5.26)
Next for any tensor eld T we dene the projected divergence operators,

T
T =

i
(
X
i
T )
X
i
. (2.5.27)

T =

a
(
Ua
T )
Ua
. (2.5.28)
68 2. FOLIATIONS
which satisfy =

+
T
with respect to the ordinary divergence operator . We
dene the mean curvature vector eld along the leaves of T by
(2.5.29) =

a
T
U
a
l
a
which vanishes if and only if each leaf is a minimal submanifold of `. We are now
ready for the relations between the Ricci curvatures. We let Ric. Ric
T
. and

Ric
denote the Ricci curvatures of o. o
T
. and the family of metrics o[
1
. respectively.
Theorem 2.5.18: The Ricci curvatures Ric. Ric
T
. and

Ric satisfy
Ric(l. \ ) =

Ric(l. \ ) o(. T
U
\ ) +o(l. \ ) +

i
o((
Xi
T)
U
\. A
i
).
Ric(A. l) = o((

T)l. A) +o(
U
. A) o((
T
)A. l) 2o(
X
. T
U
).
Ric(A. Y ) = Ric
T
(A. Y ) 2o(
X
.
Y
) o(TA. TY ) +
1
2
(o(
X
. Y ) +o(
Y
. A)).
Dening the quantities
[[
2
=

i
o(
X
i
.
X
i
) =

a
o(l
a
. l
a
). (2.5.30)
[T[
2
=

i
o(TA
i
. TA
i
) =

a
o(T
U
a
. T
U
a
). (2.5.31)
we can give the relation between the various scalar curvatures, viz.
Corollary 2.5.19: If :. :
T
. : denote the scalar curvatures of the metrics o. o
T
. o[
1
.
respectively, then
: = :
T
+ : [[
2
[T[
2
[[
2
2
T
.
2.5.2. Transverse Complex Geometry. Here we describe only the basics
of transverse complex and almost complex geometry. We present much more detail
in the special case of interest to us, namely transverse almost Kahler and Kahler
geometry, in Sections 7.2 and 7.5 below. For related work on transverse Kahler
foliations see [NT88]. Let (`. T) be a foliated manifold, then
Denition 2.5.20: A transverse almost complex structure is a smooth sec-
tion

J of End (T) that satises

J
2
= 1l
(J)
.
This implies that the normal bundle (T) is even dimensional, say 2c. Then
alternatively, a transverse almost complex structure on a foliated manifold (`. T)
is a transverse G-structure with G = G1(c. C). We are particularly interested in
the case when the transverse almost complex structure is integrable. This is done
in terms of a Nijenhuis torsion tensor in the case of almost contact structures
in subsection 6.5.4 of Chapter 6. Here it seems both more conceptual and more
convenient to describe the integrable case in terms of Haeiger structures. It is
more convenient from the point of view of deformation theory adopted in Section
8.2. Let (`. T) be a foliated manifold of codimension 2c with local submersions
1

: l

R
2q
= C
q
related by Equation 2.2.1 with transition functions

that
satisfy Haeigers cocycle conditions 2.2.2. Then
Denition 2.5.21: The foliated manifold (`. T) is said to have a transverse
holomorphic structure if the local dieomorphisms

: 1

(l

)1

(l

) are biholomorphisms for all . .

2.6. RIEMANNIAN FLOWS 69
It is easy to see that a foliated manifold (`. T) with a transverse holomorphic
structure denes a transverse almost complex structure

J on the normal bundle
(T). We can then decompose the complexication (T)
C
= (T) C into the
eigenspaces of

J.
(2.5.32) (T)
C
= (T)
1,0
(T)
0,1
.
Similarly, we get a splitting of the co-normal bundle
(2.5.33)

(T)
C
=

(T)
1,0

(T)
0,1
.
and this induces a splitting of the exterior dierential algebra over

(T)
C

(T)
C
=

p+q=r

p,q

(T)
C
.
where
p,q

(T)
C
=
p

(T)
1,0

(T)
0,1
. We let
p,q
B
denote the set of basic
forms of type (j. c). that is the set of smooth basic sections of
p,q

(T)
C
. As in
the usual case there is a splitting
(2.5.34)
r
B
(T) C =

p+q=r

p,q
B
.
as well as the basic Dolbeault complex
(2.5.35) 0
p,0
B

p,1
B

p,n
B
0.
together with its basic Dolbeault cohomology groups H
p,q
B
(T

). This will be devel-

oped much further in the context of K-contact and Sasakian geometry in Section
7.2.
We end this section with a discussion of transformation groups that preserve
the transverse holomorphic structure. Notice that any Fol(`. T) induces
a map

: (T) (T). So we dene the group of transversely holomorphic

transformations H
T
(T.

J) by
(2.5.36) H
T
(T.

J) = Fol(`. T

) [

J =

J

.
Since a 1-parameter subgroup of any element of
J
(`) induces the identity on
(T

), the group H
T
(T.

J) is innite dimensional. We are mainly interested in the
innitesimal version, that is in the Lie algebra of H
T
(T.

J).
Denition 2.5.22: Let (`. T) be a foliated manifold with a transverse holomor-
phic structure

J. We say that a vector eld A fol(`. T) is transversally holo-
morphic if given any section

Y of (T), we have
[

A.

J

Y ] =

J [A. Y ] .
The set of all such vector elds will be denoted by h
T
(T.

J).
Exercise 2.9: Show that h
T
(T.

J) is a Lie subalgebra of trans(`. T).
2.6. Riemannian Flows
We shall be interested in 1-dimensional Riemannian foliations.
Denition 2.6.1: A one dimensional oriented Riemannian foliation is called a
Riemannian ow.
70 2. FOLIATIONS
Before we begin our study of Riemannian ows, we briey mention one dimen-
sional foliations that are never Riemannian, the so-called Anosov ows [Ano69].
Any oriented 1-dimensional foliation on a smooth manifold ` is determined by a
nowhere vanishing vector eld \. The ow
t
determined by \ is said to be Anosov
if the tangent bundle T` splits as T` = 1
V
1
s
1
u
. where 1
V
is the trivial
line bundle generated by \ and 1
s
(1
u
) denote the stable (unstable) subbundles
dened as follows: there exist constants c 1 and o 0 such that for all t 0 the
estimates hold:
(i) [(
t
)

A[ cc
at
[A[ for every smooth section A of 1
s
,
(ii) [(
t
)

A[ c
1
c
at
[A[ for every smooth section A of 1
u
.
Here the norm [ [ is taken with respect to some Riemannian metric on `. but
when ` is compact the condition of being Anosov is actually independent of the
metric. An important example of an Anosov ow is the geodesic ow on the unit
tangent bundle over a compact manifold of constant negative curvature [Ano69].
See also [Ton97].
For a given Riemannian foliation (`. T) one is interested in whether the leaves
are closed in the topology of `. This is of particular interest to us in the case of
Riemannian ows. In order to study this in more detail we consider certain types
of singular foliations. Here we follow Molino [Mol88].
Denition 2.6.2: Let ` be a smooth manifold. A singular Riemannian folia-
tion oT is a partition of ` into connected immersed submanifolds, called leaves
that satisfy:
(i) The Lie subalgebra of vector elds tangent to the leaves of oT is transitive
on each leaf.
(ii) There exists a Riemannian metric o on ` such that every geodesic that
is perpendicular at one point to a leaf remains perpendicular to every leaf
it meets.
It is not dicult to see that any ordinary Riemannian foliation satises these
conditions, and thus is a special case of a singular Riemannian foliation. Further-
more, Molino shows a converse in the following sense. If T is a foliation on ` (in
the usual sense) and the conditions of Denition 2.6.2 hold, then T is a Riemannian
foliation with respect to o. For us the importance of singular Riemannian foliations
is given by
Proposition 2.6.3 (Molino): Let (`. T) be a Riemannian foliation on a compact
connected manifold `. Then the closures

T of the leaves of T form a singular
Riemannian foliation for which the leaves are embedded submanifolds of `.
Proof. See [Mol88] Proposition 6.2 and Theorem 5.1.
We are interested in the case when the Riemannian ow is an isometry, and then
many simplications occur. Indeed the following theorem of Carri`ere [Car84b] (see
also Appendix A of [Mol88] by Carri`ere) becomes much easier to prove when the
ow is an isometry.
Theorem 2.6.4: Let T be Riemannian ow on a compact manifold `. Then the
leaf closure

L of a leaf L of T is dieomorphic to a torus T
k
. and T restricted to

L is conjugate to a linear ow.

2.6. RIEMANNIAN FLOWS 71
Proof. Here we give the proof assuming that the ow is an isometry, but the
theorem holds more generally. See [Car84b] for the proof of the general case. Let o
be a bundle-like Riemannian metric for (`. T) and let denote the one parameter
subgroup of the isometry group I(`. o) generated by T. Since ` is compact, so is
I(`. o). and the closure

is a torus T I(`. o). For any r ` the closure r
of the orbit r coincides with the orbit

r of the closure

. So if
x
denotes the
isotropy subgroup at r, we see that

x
is a torus T
k
. Hence, the leaf closure

L
is a torus. Furthermore, is a subgroup of the torus, so it is conjugate to a linear
ow.
Note that the dimension / = /(L) of the torus in general depends on the leaf L
and is clearly bounded by the dimension of the manifold. This leads to
Denition 2.6.5: The toral rank or just rank rk(`. T) of a Riemannian ow
T on a compact manifold ` is dened by
rk(`. T) = max
1M
/(L).
It is clear that 1 rk(`. T) dim `. The rk(`. T) = 1 case will be of great
interest to us. In this case the leaves are all circles, and the space of leaves will be
a compact orbifold. See Chapter 4 for details.
Denition 2.6.6: A one dimensional foliation T on ` is said to be isometric if
there is a Riemannian metric o on ` such that the ow is an isometry of o.
We have
Proposition 2.6.7: An isometric one dimensional foliation is Riemannian.
Proof. Let \ be a nowhere vanishing vector eld on ` generating a one
dimensional foliation T that is isometric with respect to a Riemannian metric o.
and let A. Y be any horizontal foliate vector elds on `. Then
(2.6.1) \ o(A. Y ) = (
V
o)(A. Y ) +o([\. A]. Y ) +o(A. [\. Y ]).
The rst term vanishes since \ is an innitesimal isometry (Killing vector eld),
and the last two terms vanish since A. Y are both horizontal and foliate. Thus, o
is a bundle-like metric for (`. T). So T is Riemannian by Proposition 2.5.7.
The converse is not true in general; however, see Proposition 2.6.11 below.
Carri`ere has related the isometry of a one dimensional foliation to the orbits being
geodesics. The following result is well-known [Ton97].
Lemma 2.6.8: Let T be a one dimensional foliation on a manifold `. and let \
be a nowhere vanishing vector eld generating T. Then the following are equivalent:
(i) There exists a Riemannian metric o on ` such that \ has unit length
and the orbits of \ are geodesics.
(ii) There exists a 1-form on ` such that (\ ) = 1 and
V
= 0.
(iii) There exists a 1-form on ` such that (\ ) = 1 and \ d = 0.
Proof. Choose a metric o on ` such that \ has unit length and dene to
be the dual of \ , that is = o(\. ). Then the equivalence of (ii) and (iii) follows
immediately from the well-known formula
X
=
X
d +d
X
. Now let denote
any ane connection on `. Then, the orbits of \ are geodesics with respect to the
connection if and only if the mean curvature vector eld (
V
\ )

vanishes, where
indicates the component perpendicular to \. Equivalently, this means that the
72 2. FOLIATIONS
dual mean curvature 1-form = o((
V
\ )

. ) vanishes. But for any vector eld

A perpendicular to \ , we have
(A) = o(
V
\. A) = \ o(\. A)
V
o(\. A) o(\.
V
A)
= (
V
)(A) +([\. A])
V
o(\. A) (
V
A)
= (
V
)(A) +(
V
A) (
X
\ ) +((\. A))
V
o(\. A) (
V
A)
= (
V
)(A) +
1
2
(
X
o)(\. \ ) (
V
o)(\. A) +((\. A)).
where denotes the torsion tensor of . In particular, choosing to be the Levi-
Civita connection proves the result.
Now we have the following result due to Carri`ere.
Proposition 2.6.9: A Riemannian ow whose orbits are geodesics is isometric.
Conversely, the orbits of a one dimensional isometric foliation are geodesics.
Proof. Let T be a Riemannian ow generated by the nowhere vanishing vector
eld \ whose orbits are geodesics, and let o be a bundle-like metric such that \
has unit length. Consider Equation 2.6.1 for horizontal vector elds A. Y. Since
this equation is tensorial we can take A. Y to be foliate as well. So the last two
terms on the right hand side vanish as does the left hand side since o is bundle-like.
Thus, (
V
o)(A. Y ) = 0 for A. Y horizontal. On the other hand if is the1-form
dual to \ , then by (iii) of Lemma 2.6.8 we have for A horizontal
0 = d(\. A) = \ (A) A(\ ) ([\. A]) = ([\. A]).
which implies that (
V
o)(\. A) = 0. Also (
V
o)(\. \ ) = 0 since \ has constant
length. Thus, \ is a Killing vector eld and the ow is isometric.
For the converse let T be a one dimensional isometric foliation. Then T is
generated by a nowhere vanishing vector eld \. and there is Riemannian metric
o on ` such that \ is a Killing vector eld. Let be the 1-form dual to \ with
respect to o. Then condition (ii) of Lemma 2.6.8 is satised. Thus, the orbits of \
are geodesics.
The following is an immediate consequence of the proof of Theorem 2.6.4.
Proposition 2.6.10: Let (`. T) be an isometric Riemannian ow with toral rank
rk(`. T) = /. Then the isometry group I(`. o) contains a torus subgroup T
k
of
dimension /. Moreover, the closure of any leaf is a subtorus T
j
T
k
of dimension
1 , /.
Molino (See [Mol88] and references therein) has studied the leaf closures for
general Riemannian foliations T and has shown the existence of a sheaf of local
transverse Killing vector elds C(`. T) that he calls the commuting sheaf of the
foliation. Proposition 2.6.10 says that for isometric Riemannian ows the sheaf
C(`. T) admits precisely / global sections. In fact on a compact manifold ` a
result of Molino and Sergiescu [MS85] says that the commuting sheaf admits a
global trivialization if and only if T is isometric. Furthermore, Proposition 5.5
of [Mol88] says that for a Riemannian foliation on a compact simply connected
manifold, C(`. T) admits a global trivialization. Thus,
Proposition 2.6.11: On a simply connected compact manifold every Riemannian
ow is isometric.
2.6. RIEMANNIAN FLOWS 73
We are now ready for a Theorem of Wadsley [Wad75] that will be of great
importance to us in the sequel.
Theorem 2.6.12: Let T be a one dimensional foliation on ` such that all the
leaves of T are circles. Then there is a smooth action : o
1
`` whose
orbits are precisely the leaves of T if and only if there exists a Riemannian metric
on ` such that the leaves of T are geodesics.
Next we discuss an obstruction to the existence of Riemannian ows noticed by
Carri`ere. This comes from a seminal paper of Gromov [Gro81] in which he denes
two important invariants of smooth manifolds. The rst, called the minimal volume
is dened by
(2.6.2) MinVol(`) = inf
]K(g)]1
Vol(`. o).
where Vol(`. o) denotes the volume of ` with respect to the Riemannian metric o.
and the inmum is taken over all complete metrics whose sectional curvatures 1(o)
are bounded in absolute value by 1. One easily sees that, for example, for compact
oriented 2-manifolds `
2
one has MinVol(`
2
) = 2[(`
2
)[. where denotes the
Euler characteristic. On the other hand Gromov noticed that if a smooth manifold
` admits a locally free o
1
action, then by scaling the metrics along the orbits of
the o
1
. one sees that MinVol(`) = 0. and Carri`ere [Car84b] noticed that this
argument generalizes to arbitrary Riemannian ows.
The second invariant of Gromov is intuitively less geometrical, but computa-
tionally more useful. It is the so-called simplicial volume of ` dened
3
as follows:
consider the singular chain complex C

of all nite combinations c =

i
:
i

i
of
singular simplices
i
in ` with real coecients :
i
. Dene the simplicial norm of
c by [[c[[ =

i
[:
i
[. This gives a pseudo-norm on the singular homology H

(`. R)
by [[[.][[ = inf
z
[[[.][[. where the inmum is taken over all singular homology cycles
representing [.]. If ` is an orientable manifold, its simplicial volume [[`[[ (also
known as the Gromov invariant of `) is dened to be the simplicial norm of the
fundamental class of `. If ` is not orientable, then [[`[[ =
1
2
[[

`[[. where

` is the
2-fold cover of `. The simplicial volume enjoys the following important properties:
(i) If 1 : ``
t
is a proper map of degree d. then [[`[[ d[[`
t
[[. and if
1 is a d-sheeted covering map, [[`[[ = d[[`
t
[[.
(ii) If `
1
is compact and `
2
is arbitrary, then there is a positive constant
C depending only on the dimension of `
1
`
2
such that
C[[`
1
[[ [[`
2
[[ [[`
1
`
2
[[ C
1
[[`
1
[[ [[`
2
[[.
(iii) If dim `
1
= dim `
1
3. then [[`
1
#`
2
[[ = [[`
1
[[ +[[`
2
[[.
Furthermore, Gromov proves the following estimate for the simplicial and minimal
volumes
(2.6.3) [[`[[ (n 1)
n
n!MinVol(`).
We mention that if ` is closed (compact without boundary), then [[`[[ is com-
pletely determined by the fundamental group. In particular, compact simply con-
nected manifolds have vanishing Gromov invariant. (More generally [[`[[ vanishes
for any compact manifold whose fundamental group is amenable).
3
Here M can be any topological space.
74 2. FOLIATIONS
Summarizing we have the following theorem of Carri`ere.
Theorem 2.6.13: If a compact manifold ` admits a Riemannian ow, then
MinVol(`) = 0. In particular, [[`[[ = 0 and all the Pontrjagin numbers of `
vanish.
We are interested in seeing which manifolds ` have non-vanishing Gromov in-
variant [[`[[. In [Gro81] Gromov proves a theorem which he attributes to Thurston
which says that a complete Riemannian manifold of nite volume whose sectional
curvatures are pinched between / and 1 for some / 1 has non-vanishing
Gromov invariant [[`[[. At about the same time Inoue and Yano [IY82] proved
that any compact Riemannian manifold whose sectional curvatures is bounded from
above by for 0 has non-vanishing Gromov invariant. We state the precise
results as:
Theorem 2.6.14: Let (`. o) be a n-dimensional Riemannian manifold.
(i) Suppose that (`. o) is complete with nite volume 0 < Vol(`. o) < .
and that the sectional curvatures of o are pinched between / and 1 for
some real number / 1. Then the estimate
Vol(`. o) C
n
[[`[[
holds, where C
n
is a positive constant depending only on n.
(ii) Let (`. o) be compact without boundary whose sectional curvatures are
bounded from above by for some 0. Then there exists a positive
constant C
n
such that
[[`[[ C
n

n/2
Vol(`. o).
Hence, in both cases [[`[[ = 0.
It follows immediately that no manifold satisfying either of the two conditions
of this theorem can admit a Riemannian ow.
CHAPTER 3
Kahler Manifolds
Kahler metrics
1
were introduced and studied by Erich Kahler in 1933 [Kah33]
(recently reprinted in [Kah03]). But it was not until late the 40ties when the
importance of Kahler manifolds in both Riemannian and algebraic geometry was
nally realized. This was largely due to the fundamental work of Chern. Already
in 1946 Chern introduces the notion of his Chern classes for Hermitian manifolds
[Che46]. As a special case, it follows that the cohomology class of the Ricci 2-form

on a Kahler manifold (`. J. o.

g
) does not depend on the metric o but only on
the complex structure and it is a xed multiple of the rst Chern class of `
[
g
] = 2c
1
(`) .
(cf. Proposition 3.6.1 of the last section of this chapter). All of a sudden Kahler
metrics dramatically gained in their importance. Throughout the 50ties they were
studied by such giants as Borel, Chern, Hodge, Kobayashi, Kodaira, Lichnerowicz,
Spencer, among many others, and culminating in Weils well known book [Wei58].
However, it was not until 10 years after the Cherns article when Calabi grasped
the real signicance of this remarkable relation and what it means in the context
of Kahler-Einstein metrics [Cal56, Cal57]. Calabis work directly or indirectly re-
sulted in several mesmerizing conjectures which captivated the mathematical world
for years to come and stimulated incredible interest in Kahler geometry that has
remained until today
2
. The relation
[
g
] = [
g
] = 2c
1
(`)
will be the focal point of Chapter 3 and even more so in Chapter 5. Since Kahler
geometry and algebraic geometry from a dierential viewpoint play a crucial role
in this book, we take some care in setting things up in spite of the fact that there
are several comprehensive treatments of this subject in a book form. We mention
here the following books where excellent presentations of Kahler geometry can be
found: [KN69, LB70, GH78b, Wel80, Zhe00, Voi02, Bal06] to name a few.
We thus recall the basic properties of Kahler manifolds and Kahler metrics for the
most part without proofs. We focus on the curvature properties, Chern classes and
end with a discussion of the famous Calabi Conjecture. In Chapter 4 we describe
1
Schouten considered Kahler manifolds 4 years earlier in the article

Uber unitare Geometrie
but never got much credit for his work [Sch29, SvD30]. Curiously, neither Schouten nor Kahler
appeared very interested in their own inventions. The term Kahler manifolds became standard
after the WWII in the late 40ties. Interestingly, Schouten and von Dantzig arrive at the Kahler
condition while investigating parallel transport of the Levi-Civita connection associated to the a
Hermitian metric.
2
We refer the reader interested in the historical development of Kahler geometry from its
birth to the volume devoted to the mathematical works of Kahler [Kah03] and to Jean-Pierre
Bourguignons excellent article in that volume.
75
76 3. K

AHLER MANIFOLDS
how Kahler and algebraic geometry generalize to the category of Kahler orbifolds.
Finally, in Chapter 5 we focus primarily on Kahler-Einstein metrics of positive
scalar curvature.
3.1. Complex Manifolds and Kahler Metrics
Let us start with a denition of a complex manifold. There are several fre-
quently used approaches to dene such a structure. Perhaps the most natural one
is the one that imitates the usual denition of a smooth structure on a real manifold.
Denition 3.1.1: Let ` be a real manifold of dimension 2n. A complex chart
on ` is a pair (l. ) such that l ` is open and : l C
n
is dieomorphism
between l and an open set (l) C
n
. We say that ` is a complex manifold if
it admits and atlas / = l

of complex charts whose transition functions

(l

(l

)
are all biholomorphic.
Recall from Example 1.4.9 that if ` is a real manifold than a smooth section J
of the bundle of endomorphisms End(T`) such that J
2
= 1l is called an almost
complex structure. Note that we must have dim(`) = 2n. We can extend J to
act on the complexied tangent bundle T`
R
C by C-linearity. Then J induces
a splitting T`
R
C = T
1,0
` T
0,1
`. where T
1,0
` and T
0,1
` are eigenspaces
with eigenvalues i. respectively. Note that T` is naturally isomorphic to T
1,0
`
by the map A
1
2
(A iJA).
An almost complex structure is said to be integrable if ` admits an atlas of
complex charts with holomorphic transition functions such that J corresponds to
the induced complex multiplication on T`
R
C. Hence, a manifold with an inte-
grable almost complex structure is (by denition) a complex manifold. Conversely,
given a complex manifold ` we can dene J by
J

.
j
= i

.
j
.
where (.
1
. . . . . .
n
) is a local holomorphic coordinate system.
The theorem of Newlander and Nirenberg [NN57] asserts that integrability of
J is equivalent of the vanishing of the Nijenhuis tensor Equation 1.4.4 In view of
this we have the following
Theorem/Denition 3.1.2: Let ` be a real manifold and J an almost complex
structure on `. The Nijenhuis tensor
J
0 on ` if and only if J is integrable
in which case we call J a complex structure and the pair (`. J) a complex
manifold.
The second part of this statement is very often taken as an alternative denition
of a complex manifold. As we will see it is perhaps the most geometric and indeed
we will use it most often. A third way of describing a complex manifold as a real
manifold with torsion-free G1(n. C)-structure was described earlier in Example
1.4.9. For maps between complex manifolds we have
Denition 3.1.3: Let (`
1
. J
1
), (`
2
. J
2
) be two complex manifolds. A map 1 :
`
1
`
2
is called holomorphic if J
2
(d1()) = d1(J
1
()) for all (T`). If,
in addition, 1
1
exists and is holomorphic than 1 is called a biholomorphism
between `
1
and `
2
.
3.1. COMPLEX MANIFOLDS AND K

AHLER METRICS 77
If `
2
= `
1
= ` and J
2
= J
1
= J the set of biholomorphisms from ` to itself
form a group Aut(`. J). Generally it is not a Lie group; however, it is a Lie group
when ` is compact by Proposition 1.6.7. In fact if ` is compact it is a complex
Lie group. Explicitly, [Kob72] we have
Proposition 3.1.4: Let ` be a compact complex manifold. Then Aut(`. J) is
a complex Lie transformation group and its Lie algebra aut(`. J) consists of the
holomorphic vector elds on `.
For a much more extended discussion of the complex automorphism group on
Kahler manifolds we refer the reader to Chapter III of [Kob72].
One can extend J to act on the complexied cotangent bundle T

`
R
C which
then splits as T
(1,0)
` T
(0,1)
`. This in turn denes the splitting of the bundle
of complex j-forms on `

p
T

`
R
C =
p

j=0

j
T
(1,0)
`
pj
T
(0,1)
` =
p

j=0

j,pj
` .
where the last equality denes the space of forms of type (,. j ,). Note that such
decomposition holds even if J is only almost complex.
Proposition 3.1.5: Let (`. J) be an almost complex manifold. The following
conditions are equivalent:
(i) [\. \] (T
1,0
`) for all \. \ (T
1,0
`).
(ii) [\. \] (T
0,1
`) for all \. \ (T
0,1
`).
(iii) Im(d)[
(
1,0
M)
(
2,0
`) (
1,1
`).
Im(d)[
(
0,1
M)
(
0,2
`) (
1,1
`).
(iv) Im(d)[
(
p,q
M)
(
p+1,q
`) (
p,q+1
`).
(v)
J
0.
Proof. We will only show the equivalence of (i), (iii), and (v) as the rest is
evident. To establish equivalence of (i) and (iii) we use
2d(A. Y ) = A((Y )) Y ((A)) ([A. Y ]) .
which, if is a 1-form of type (1. 0) and A. Y are vector elds of type (0. 1), clearly
vanishes. Hence, d does not have any (0. 2)-part. To establish equivalence of (i)
and (v) we consider 1 = [AiJA. Y iJY ] for any two real vector elds A. Y on
`. It is easy to see that 1 +iJ1 = (A. Y ) iJ((A. Y )).. But 1 +iJ1 = 0
if and only if 1 is a vector eld of type (1. 0) if and only if
J
0.
Thus, if (`. J) is a complex manifold, composing the exterior derivative acting
on (/. |)-forms with projections on (
k+1,l
`) and (
k,l+1
`). denes natural
splitting of d into two parts. They are denoted by and

. where d = +

. Since
d
2
= 0 we get
2
=

2
= 0 and

. We also denote by c
k,l
the sheaf of
germs of smooth sections of the bundles
k,l
`. and use the same notation .

. etc.
for these operators acting on local sections of c
k,l
. We also let
p
denote the sheaf
of germs of holomorphic sections of the bundle
p,0
`. Then, by the well-known

-Poincare lemma [GR65], we have a resolution of this sheaf, namely

0
p
c
p,0

c
p,1

c
p,n1

c
p,n
0 .
78 3. K

AHLER MANIFOLDS
Moreover, since the sheaves c
k,l
are ne sheaves, this resolution is acyclic. So if we
dene the Dolbeault cohomology groups [Dol53] of (`. J)
H
k,l

(`) =
Ker

: (
k,l
) (
k,l+1
)

Im

: (
k,l1
) (
k,l
)

to be the derived groups of the cochain complex on global sections (1.1.7), the Ab-
stract de Rham Theorem 1.1.16 gives the well-known Dolbeault Theorem [Dol53]
Theorem 3.1.6: Let ` be a complex manifold. Then there is an isomorphism
H
q
(`.
p
) H
p,q

(`) .
In Section 3.4 we shall give a more general version of this theorem by twisting
with a vector bundle.
Denition 3.1.7: Let (`. J) be an almost complex manifold and let o be a Rie-
mannian metric on ` such that
o(JA. JY ) = o(A. Y ). A. Y (T`) .
Then o is called an Hermitian metric and (`. J. o) is called an almost Hermit-
ian manifold. Furthermore, given a Hermitian metric we dene its fundamental
2-form or Kahler form
g
of o by

g
(A. Y ) = o(A. JY ). A. Y (T`) .
The triple (J. o.
g
) is called an almost Hermitian structure on `.
Note that an almost complex manifold always admits a Hermitian metric.
An almost Hermitian structure (J. o.
g
) with an integrable J is called Hermit-
ian. Given an Hermitian metric o we can extend it to a unique Hermitian scalar
product / on the complexied tangent bundle T`
R
C satisfying
(i) /(

7.

\) = /(7. \) for all 7. \ (T`
R
C).
(ii) /(7.

7) 0 fora all 7 = 0 T`
R
C.
(iii) /(7.

\) = 0 for 7 (T
1,0
`) and \ (T
0,1
`).
Locally, in a complex coordinate chart (l; .
1
. . . . . .
n
) we can write the metric / as
(3.1.1) / =
n

i,j=1
/
i

j
d.
i
d .
j
.
where /
i

j
(j) is a Hermitian nn matrix. Then the Hermitian form / can be written
in terms of the Riemannian metric o and fundamental 2-form
3
as / = o 2i
g
and

g
can be written as
(3.1.2)
g
=
i
2
n

i,j=1
/
i

j
d.
i
d .
j
.
Also, note that the volume form satises

n
g
= n!

i
2

n
det(/
i

j
)d.
1
d .
1
d.
n
d .
n
and, hence,
n
g
= n!dvol
g
. where dvol
g
is the Riemannian volume form of (`. o).
3
The factor -2 multiplying
g
is chosen so that the volume form satises
n
g
= n!dvol
g
.
3.1. COMPLEX MANIFOLDS AND K

AHLER METRICS 79
Denition 3.1.8: A Hermitian manifold (`. J. o.
g
) is said to be Kahler if
g
is a closed 2-form. We call o a Kahler metric,
g
its Kahler form, and the
triple (J. o.
g
) a Kahler structure on `.
When we want to emphasize a Kahler form, we shall use the notation (`. )
to denote a Kahler manifold. A manifold with a Kahler metric is often called a
manifold of Kahler type.
From the previous discussion we see that a Kahler form is a real (1. 1)-form
which is both -closed and

-closed. Hence [
g
] denes a Dolbeault cohomology
class in H
1,1

(`). Next we state several propositions and lemmas which describes

some basic local and global properties of Kahler manifolds.
Proposition 3.1.9: Let (`. J. o.
g
) be an almost Hermitian manifold of real di-
mension 2n and let
g
be the fundamental 2-form associated to o. Let be the
Levi-Civita connection of o. Then the following conditions are equivalent:
(i) J = 0,
(ii)
g
= 0,
(iii) the holonomy group Hol(`. o) is contained in l(n),
(iv) (`. J. o.
g
) is 1-integrable,
(v) (`. J. o.
g
) is Kahler.
Yet another characterization of the Kahler condition is
Proposition 3.1.10: A Hermitian metric o on a complex manifold (`. J) is
Kahler if and only if for any point j ` there exists a local holomorphic chart
(l; .
1
. . . . . .
n
) such that /
i

j
(j) =
ij
. d/
i

j
(j) = 0.
Such a metric is said to osculate to order 2 to the Euclidean metric, and the
coordinate chart is said to be normal at j.
We consider some important example of complex Kahler manifolds [KN69].
Example 3.1.11: Let ` = R
2n
and let (r
1
. . . . . r
n
. n
1
. . . . . n
n
) be the global coor-
dinate chart on `. We can dene
J

r
j
=

n
j
. J

n
j
=

r
j
.
Clearly, J
2
= 1l and
J
0 as J is a constant matrix on `. We can dene
.
1
. . . . . .
n
by .
j
= r
j
+in
j
for , = 1. . . . . n. We have
J

.
j
= i

.
j
. J

.
j
= i

.
j
so that (.
1
. . . . . .
n
) C
n
is a (global) holomorphic chart compatible with J. Now let
us write the Hermitian metric / and the fundamental 2-form in complex coordinates:
Let (.
1
. . . . . .
n
) be the global holomorphic chart with .
i
= r
i
+ in
i
. Since d.
i
=
dr
i
+idn
i
. d .
i
= dr
i
idn
i
i, i = 1. . . . . n and the Hermitian metric / reads
/ =
n

i,j=1
/
i

j
d.
i
d .
j
=
n

i=1
d.
i
d .
i
.
As the fundamental 2-form
g
is simply minus the imaginary part of / we can write

g
=
i
2
n

i,j=1
/
i

j
d.
i
d .
j
=
i
2
n

i=1
d.
i
d .
i
which is clearly closed.
80 3. K

AHLER MANIFOLDS
The ducial example of a compact complex manifold is given by our next
example.
Example 3.1.12: Complex Projective Space CP
n
is dened as the set of lines
through the origin in C
n+1
. Let z = (.
0
. . . . . .
n
) be a point (vector) in C
n+1
` 0.
We say that two non-zero vectors z and z
t
are equivalent if there is a C

such that z
t
= z. Then CP
n
is the quotient space (C
n+1
` 0) . We let
: C
n+1
` 0CP
n
denote the natural projection. Local coordinate charts
are dened as follows. Let

l
i
be the open subspace of C
n+1
such that .
i
= 0. Then
CP
n
is covered by open sets l
i
= (

l
i
)
n
i=0
. together with homeomorphisms

i
: l
i
C
n
= R
2n
dened by

i
([.
0
. . . . . .
n
]) =

.
0
.
i
. . . . .
.
i1
.
i
.
.
i+1
.
i
. . . . .
.
n
.
i

.
The coordinates on the right hand side of this equation are called ane coordinates,
and it is convenient to set .
i
= 1 and write [z[
2
= [.
1
[
2
+ +

[.
i
[
2
+ + [.
n
[
2
.
where as usual the hat means remove that element. Then the Fubini-Study metric
in l
i
CP
n
is dened by
o
k

j
=

2
.
k
.
j
ln(1 +[z[
2
) .
and the Kahler form is
(3.1.3)
g
=
i
2

ln(1 +[z[
2
) =
i
2

k
d.
k
d .
k
1 +[z[
2

(

k
.
k
d.
k
) (

j
.
j
d .
j
)
(1 +[z[
2
)
2

.
Clearly,
g
is closed and one can check that both the Riemannian metric o and
Kahler form
g
are globally dened on CP
n
.
Notice that by restricting the map to the unit sphere in C
n+1
gives the well-
known Hopf bration o
1
o
2n+1
CP
n
which as should become apparent will
be considered as the ducial example for this book.
This example has an interesting generalization, namely:
Example 3.1.13: Let G:
p
(C
q+p
) be the set of j-dimensional complex subspaces of
the complex vector space C
q+p
. We will dene complex structure on the complex
Grassmannian G:
p
(C
q+p
) by construction an atlas of holomorphic charts. Let
(.
1
. . . . . .
p+q
) be the natural coordinates on C
q+p
. We can think of .

as the complex
linear mapping .

: C
q+p
C. Now, consider the partition of the set 1. . . . . j +
c =
c
=
1
. . . . .
p

p+1
. . . . .
p
. where both sets are ordered in the
increasing order. Let let l

G:
p
(C
q+p
) be the subset of j-planes \ for which
.

1
[
W
. . . . . .

p
[
W
are linearly independent. Since, for each \ l

, the mappings
.

1
[
W
. . . . . .

p
[
W
form a basis for the dual space of \, we can write
.

p+k
[
W
=
p

j=1
n
j
k
(.
j
[
W
) . / = 1. . . . . c.
The coecients n
j
k
uniquely dene a j c complex matrix which maybe thought
of as an element of C
pq
. Setting

(\) = (n
j
k
)
for each partition denes maps

: l

C
pq
. It is a simple exercise to check that
these maps are injective and onto. Hence, the family o = l

of

p+1
p

charts
3.1. COMPLEX MANIFOLDS AND K

AHLER METRICS 81
form a holomorphic atlas on G:
p
(C
q+p
). There is a generalization of Equation 3.1.3
which shows that G:
p
(C
q+p
) is Kahler [KN69]. When j = 1. c = n we recover
the standard atlas of the (n + 1) holomorphic charts on complex projective space
described in Example 3.1.12. In particular, we have G:
1
(C
n+1
) = CP
n
.
One can describe further interesting generalizations by considering the set of
all nested subspaces C
k
1
C
k
2
C
k
l
C
n
with 0 < /
1
< /
2
< < /
l
< n.
This gives rise to the so-called ag manifolds F|
k
1
,...,k
l
(C
n
), cf. [Akh90, Akh95].
In the case | = 1 we recover the Grassmannian. Even further generalizations are
obtained by considering the generalized ag manifolds G1. where G is a complex
semi-simple Lie group and 1 is a parabolic subgroup (the ordinary ag manifolds
correspond to taking G = o1(n. C)). These are all homogeneous Kahler manifolds.
In fact, a well-known theorem of Borel and Remmert [BR62] (See also [Akh95])
says that any compact homogeneous Kahler manifold ` is of the form (G1)
(`). where (`) is a complex torus known as the Albanese of `.
Now, let 1 C

(`). As the i

1 =
1
2
( +

)(

)1 is real we conclude that

i

1 is a real closed 2-form of type (1. 1). We have the following two lemmas
Lemma 3.1.14: [local i

-lemma] Let (
1,1
l) be a smooth, closed, real
(1. 1)-form on a unit disc l = 1
n
C
(1) C
n
. Then there exists 1 C

(l) such
that = i

1.
Lemma 3.1.15: [global i

-lemma] Let (`. J) be a compact complex manifold

and let be a real (1. 1)-form on ` satisfying = d for some real 1-form . Then
there exists a smooth real function 1 such that = i

1. In particular, if o
1
. o
2
are
two dierent Kahler metrics on ` such that [
g1
] = [
g2
] H
2
(`. R). Then

g
2
=
g
1
+i

1 .
and 1 is unique up to a constant.
Lemma 3.1.14 says that on a Kahler manifold locally one can always nd a
function which generates the Kahler form (and, hence, the Kahler metric) via
the simple formula
g
= i

1. Such a function is called a Kahler potential. How-

ever, if ` is compact a Kahler potential cannot exists globally. This is a simple
consequence of the following
Lemma 3.1.16: Let (`. J. o.
g
) be a compact Kahler manifold with the Kahler
form
g
. Then [
k
g
] H
2k
(`. R) is non-trivial for all / = 0. . . . . n.
Proof. This is a simple consequence of the Stokes theorem and the fact that

n
g
= n!dvol
g
. For let
k
g
= d be exact. Then
0 < n!vol(`) =

n
g
=

k
g

nk
g
=

M
d
nk
g
=

M
d(
nk
g
) = 0 .
so we get a contradiction. In particular, when / = 1 the lemma implies that there
is no globally dened function on ` such that
g
= i

1.
Let (`. J) be a complex manifold and let be a submanifold of `. Then, in
particular at each point j , the tangent space T
p
T
p
` is a vector subspace.
We say that is a complex submanifold of ` if T
p
is a complex subspace for
each j . i.e., J(T
p
) = T
p
. Now, if is a complex submanifold then the
restriction J
N
of J to T is a complex structure on and the inclusion map
82 3. K

AHLER MANIFOLDS
: ` has the property
J(d
p
(n)) = (d
p
)(J
N
(n)) . n T
p
.
that is is a holomorphic map between two complex manifolds (. J
N
) and (`. J).
Since we can pullback Kahler forms, we have a simple but important result.
Proposition 3.1.17: A complex submanifold of a Kahler manifold is Kahler.
Example 3.1.18: Consider the previous example with ` = CP
n
and let the inclu-
sion map : CP
n
be dened by setting
=
f
= : CP
n
[f (:) = 0 .
where f = (1
1
. . . . . 1
k
) and 1
k
C[.
0
. . . . . .
n
] are complex homogeneous polynomials
in (n + 1) variables for each / = 0. 1. . . . . n. Any such set is called a projective
algebraic variety. If, in addition
J
is smooth then it is a complex submanifold in
the way described above; hence, it is a smooth projective algebraic variety which
by Proposition 3.1.17 is always Kahler. For instance, if / = 1 then is called a
hypersurface. For example, let f = 1. where 1 is
1([.
0
. . . . . .
n
]) = .
p
0
+ +.
p
n+1
.
Then
f
is called a Fermat hypersurface. When j = 2 we get a complex quadric.
The concept of variety can vary somewhat depending on the context. We now
formalize this. An important point is that varieties allow for a certain type of
singular behavior.
Denition 3.1.19: We have
(i) An ane algebraic variety is the common zero locus of a collection of
polynomials in C
n
.
(ii) A projective algebraic variety is the common zero locus in CP
n
of a
collection of homogeneous polynomials in C
n+1
.
(iii) An analytic (sub)variety \ of a complex manifold ` is a closed subset
that can written locally as the common zero locus of a nite collection of
local holomorphic functions in `. In particular, A closed subset `
of a complex manifold ` is said to be a hypersurface if every point
j has an open neighborhood l and a nonzero holomorphic function
1 : lC such that l = j l [ 1(n) = 0.
By a variety we mean any of these three, and by an algebraic variety either of
the rst two. When working with algebraic varieties it is common to use the Zariski
topology which is dened by the condition that its closed subsets are precisely the
common zero loci of polynomials. So by a subvariety of an algebraic variety \ we
shall mean any Zariski closed subset of \.
Denition 3.1.20: A variety \ is said to be irreducible if it cannot be written
as the union of two subvarieties \
1
and \
2
with \
i
= \.
3.2. Curvature of Kahler Manifolds
Let (`. J. o.
g
) be a Kahler manifold and the Levi-Civita connection. We
can extend in a C-linear way to (T`
R
C). In a local chart (l; .
1
. . . . . .
n
) we
have

z
1
. . . . .

z
n
and

z
1
. . . . .

z
n
as bases for T
1,0
` and T
0,1
`. respectively.
3.2. CURVATURE OF K

AHLER MANIFOLDS 83
Dene the Christoel symbols
C
AB
as follows

/z
A

.
B
=
n

C=1

C
AB

.
C
. . 1. C 1. . . . . n.

1. . . . . n .
It follows by C-linearity that

C
AB
=

A

B
.
C
AB
=
C
BA
.
Lemma 3.2.1: The only non-vanishing Christoel symbols are
k
ij
and

j
. where

n
ij
=

k
o
n

k
o
j

k
.
i
.
Proof. We have

/z
i

.
j
=

k
ij

.
k
+

k
ij

.
k
.

/zi

.
j
=

k
i

.
k
+

k
i

.
k
.
Since, J = 0 and J acts by multiplication by i on the basis of T
1,0
` and i on
the basis of T
0,1
`, we get

/z
i

J

.
j

= J
/z
i

.
j
.
By comparing two sides of this equation we get
i

k
ij

.
k
+

k
ij

.
k

= i

k
ij

.
k

k
ij

.
k

.
which means that

k
ij
= 0. Similar argument applied to the second equation yields

k
i

j
= 0. while

k
ij
= 0 follows by the symmetry two lower indices. Hence, the
only non-zero Christoel symbols are
k
ij
and their complex conjugate

j
. Since
Ao(Y. 7) = o(
X
Y. 7) +o(Y.
X
7) we obtain

.
i
o
j

k
=

.
i
o

.
j
.

.
k

= o

l
ij

.
l
.

.
k

l
ij
o
l

k
.
Contracting both sides of this equation with the inverse of the metric gives the
formula.
In a local chart (l; .
1
. . . . . .
n
) we can dene the following matrix-valued (1. 0)-
form
= o
1
o =

i=1
o
1
o
.
i
d.
i
.
where o = (o
i

j
). The above lemma implies that
Proposition 3.2.2: Let (`. J. o.
g
) be a Kahler manifold and let be the Levi-
Civita connection of o extended to T`
R
C T
1,0
` T
0,1
`. Accordingly,
decomposes =
1,0
+
0,1
. is the connection 1-form of
1,0
.
84 3. K

AHLER MANIFOLDS
Note that we have the following simple expression

j
ij
=

j,

k
o
j

k
o
j

k
.
i
=

.
i
ln G.
where G = det(o) = det(o
i

j
). The determinant G C

(|) is a smooth real

function dened locally on |. The last equality follows from the well-known formula
det() = expTr ln as we have

.
i
ln G = Tr

o
1
o
.
i

n,k
o
nk
o
k n
.
i
.
These simple formulas for the Christoel symbols simplify the expression for the
full Riemann curvature tensor. Recall that the Riemannian curvature tensor 1 of
the metric o one ` is dened by 1(A. Y ) : (T`)(T`). where
1(A. Y )7 =
X

Y
7
Y

X
7
[X,Y ]
7 . A. Y. 7 (T`) .
Point-wise 1(A. Y ) is a skew-endomorphism of T
p
` at any j |. Since also
1(A. Y ) = 1(Y. A) one can think of 1 as a 2-from on ` with values in skew-
endomorphisms of T`. It is sometimes convenient to think of 1 as the curvature
operator, that is a sections of
2

2
T` dened by
{(A Y. \ 7)) = o(1(A. Y )7. \) . A. Y. 7. \ (T`) .
Alternatively, one can introduce 1 as the section of
4
(T

`), i.e., the four-linear

map 1 : (T`) (T`) (T`) (T`)C

(`). dened by
1(A. Y. \. 7) = o(1(A. Y )7. \ ) . A. Y. \. 7 (T`) .
All these dierent pictures are useful.
If (`. J. o) is complex than one can extend all these curvature tensors by C-
linearity to T`
R
C. In addition, on a Kahler manifold we have J = 0 so that
1(\. \)J7 = J1(\. \)7 . \. \. 7 (T`
R
C) .
The above property has many important consequences. Since o is Hermitian we
get 1(l. \. J\. J7) = 1(l. \. \. 7). Hence, 1(l. \. \. 7) = 0 unless \. 7 are
of dierent type. In particular, in the Kahler case one can dene the following
curvature tensors:
Denition 3.2.3: Let (`. J. o.
g
) be a Kahler manifold and let 1 be the curvature
tensor of (`. o). Extending 1 to T`
R
C we dene the following:
(i) Riemann curvature tensor 1 as section of
4
(T
1,0
`). i.e., the map
1 : (T
1,0
`) (T
0,1
`) (T
1,0
`) (T
0,1
`)C

(`. C). given

by
(A. Y. 7. \) 1(A.

Y . 7.

\) .
(ii) Riemann curvature tensor 1
X,

Y
as a section of End(T
1,0
`), i.e.,
the map 1
X,

Y
: ((T
1,0
`))(T
0,1
`). A. Y (T
1,0
`) given by
1
X,

Y
(7) = 1(A.

Y )

7 .
(iii) A real (1. 1)-form with values in (End(T
1,0
`)). i.e., a skew-Hermitian
map : (T
1,0
`) (T
0,1
`)(End(T
1,0
`))
(A.

Y ) = 1
X,

Y
. A. Y (T
1,0
`) .
called the Kahler-Riemann curvature form.
3.2. CURVATURE OF K

AHLER MANIFOLDS 85
(iv) Riemann curvature operator { as a section of
2
(
1,1
`). i.e., the
map { : (
1,1
`) (
1,1
`)C

(`. C) given by
{(A

Y . \

\) = 1(A.

Y . 7.

\) .
Proposition 3.2.4: Let (l; .
1
. . . . . .
n
) be a local coordinate chart on a Kahler
manifold (`. J. o.
g
). Let 1 (
4
(T
1,0
`)) be the Riemann curvature tensor.
Then
1
i

jk

l
=

2
o
k

l
.
i
.
j

p, r,m
o
p r
o
k r
.
i
o
p

l
.
j
.
Proof. It remains to do the local computations. We have
1
i

jk

l
= 1

.
i
.

.
j
.

.
k
.

.
l

= o

.
i
.

.
j

.
l
.

.
k

.
Since
1

.
i
.

.
j

.
l
=
/z
i

/ z
j

.
l

/ z
j

/z
i

.
l
=
/z
i

.
n

.
Hence
1
i

jk

l
= o

/z
i

.
n

.

.
k

=

.
i
o

.
n
.

.
k

.
n
.

m
ik

.
m

=

.
i

l
o
nk

n,p

m
ik
o
nm
=
=

.
i

n,p
o
np
o
lp
.
j
o
nk

n,p, r,m
o
np
o
lp
.
j
o
m r
o
k r
.
i
o
nm
=

2
o
k

l
.
i
.
j

p, r,m
o
p r
o
k r
.
i
o
p

l
.
j
.

Denition 3.2.5: Let 1

X,

Y
: (T
1,0
`)(T
1,0
`). A. Y (T
1,0
`) be the
Riemann curvature tensor on a Kahler manifold (`. J. o.
g
). The Ricci curva-
ture tensor is the map Ric

: (T
1,0
`) (T
1,0
)C

(`. C) dened by the

trace
Ric

(A.

Y ) = Tr(1
X,

Y
) .
A priori it is not clear that Ric

(A.

Y ) is the Ricci curvature tensor in the usual
sense. However, it is clear that Ric(A.

Y ) denes a Hermitian symmetric form on
T
1,0
p
`. For with respect to any unitary basis c
1
. . . . . c
n
of T
1,0
p
` we have
Ric

(A.

Y ) =
n

i=1
1(A.

Y . c
i
. c
i
) .
so that
Ric

(A.

Y ) =
n

i=1
1(

A. Y. c
i
. c
i
) =
n

i=1
1(Y.

A. c
i
. c
i
) = Ric

(Y.

A) .
86 3. K

AHLER MANIFOLDS
Proposition 3.2.6: The Ricci curvature tensor Ric

: T
1,0
` T
1,0
` C is a
Hermitian symmetric form on T
1,0
p
` at each j `. In a local chart (l; .
1
. . . . . .
n
)
this form can be written as
Ric

=
n

i,

j=1
1
i

j
d.
i
d .
j
.
where
1
i

j
=

2
.
i
.
j
(ln G) .
The real part of this Hermitian symmetric form is the Ricci curvature tensor of the
metric o while the imaginary part is a real (1. 1)-form on ` called the Ricci form
and denoted by

g
. With respect to the local chart (l; .
1
. . . . . .
n
) we have

=
i
2

i,

j
1
i

j
d.
i
d .
j
=

(ln G) .
Proof. By denition we have
1
i

j
=

l
o
k

l
1
i

jk

l
=

2
.
i
.
j
(ln G) .
A priori it is not obvious that Ric(A.

Y ) coincides with the usual denition of the
Ricci curvature. A simple calculation shows that this is indeed the case.
Note that the Kahler Riemann curvature 2-form = d + is simple the
curvature 2-form of the
1,0
part of the Levi-Civita connection. In local coordinates
we can write

l
k
=

i,

j, p
1
i

jk p
o
l p
d.
i
d .
j
and, hence, the Ricci form

is a closed 2-form which is an invariant under the

complex linear group G1(n. C). namely,
(3.2.1)

= Tr() .
There are several more curvatures typically considered in the context of complex
and Kahler manifolds. The usual notion of sectional curvature is one of them.
Recall that for any A. Y (T`) we dene the sectional curvature of the 2-plane
T
p
` spanned at point j ` by A
p
and Y
p
as
1(. j) = 1(A. Y ) =
1(A. Y. Y. A)
[A Y [
2
.
where [A Y [
2
= [A[
2
[Y [
2
[o(A. Y )]
2
= o(A. A)o(Y. Y ) [o(A. Y )]
2
is the area
square of the parallelogram spanned by A. Y. Now, if (`. o) is complete, simply
connected and of constant sectional curvature then (depending on the sign) (`. o) is
isometric to o
n
. R
n
. or the real hyperbolic ball 1
n
R
(1). Now, in complex dimension
greater than 1 the only Kahler manifolds of constant sectional curvature are at.
To obtain a Kahler analogue of the real space forms we introduce the following
Denition 3.2.7: Let (`. J. o.
g
) be a Kahler manifold and let A. Y (T
1,0
`).
Then
1(A. Y ) =
1(A.

A. Y.

Y )
o(A.

A)o(Y.

Y )
3.2. CURVATURE OF K

AHLER MANIFOLDS 87
is called the bisectional curvature of o in the direction of A. Y while
H(A) = 1(A. A) =
1(A.

A. A.

A)
[o(A.

A)]
2
is called the holomorphic sectional curvature of o in the direction of A.
If we write A =
1

2
(l + iJl). Y =
1

2
(\ + iJ\ ) for some real vector elds
l. \ (T`) we have
1(A. Y ) =
[Jl \ [
2
[l[
2
[\ [
2
1(Jl. \ ) +
[l \ [
2
[l[
2
[\ [
2
1(l. \ ) .
and H(A) = 1(l. Jl). where 1 the sectional curvature of o. In particular, if o
is of positive (negative, non-positive, nonnegative) sectional curvature then o is of
positive (negative, non-positive, nonnegative) bisectional curvature. Clearly, the
holomorphic sectional curvature is the curvature of all J-invariant planes. It is not
hard to show that on a Kahler manifold the holomorphic sectional curvature H
determines the Riemann curvature tensor completely.
Example 3.2.8: This is a continuation of the discussion of complex projective space
as described in Example 3.1.12 An easy computation gives

n
g
= (i2)
2
n!
(1 +[z[
2
)
n+1
d.
1
d .
1
d.
n
d .
n
.
which means that G = det(o
i

j
) = (1 + [z[
2
)
(n+1)
. Hence, The Ricci curvature
tensor
(3.2.2) 1
i

j
=

2
.
i
.
j
ln G = (n + 1)o
i

j
.
Exercise 3.1: Show that the Fubini-Study metric is l(n + 1)-invariant and that
l(n + 1) acts on ` transitively.
It now follows that in order to compute the full Riemann curvature tensor 1
i

jk

l
it is sucient to compute it at one point [1. 0. . . . . 0]. Using local expressions for
1
i

jk

l
and o
i

j
one can easily see that at z = 0 we have 1
i

jk

l
=
i

l
+
i

j
. In
particular, as o
i

j
[
z=0
=
i

j
at any other point
1
i

jk

l
= o
i

j
o
k

l
+o
i

l
o
k

j
and ` is of constant holomorphic sectional (and bisectional) curvature c = +1.
Example 3.2.9: [Bergman Metric on the Complex Ball.] Consider ` =
1
n
C
(1) = z C
n
[ [z[
2
< 1 and let

g
=
i
2

ln(1 [z[
2
) .
Likewise, one can show that Bergman metric o is l(n. 1)-invariant and that l(n. 1)
acts on ` = 1
n
C
(1) transitively. Repeating the calculations of the previous example
we can easily see that 1
i

j
= (n + 1)o
i

j
and (`. o) is of constant holomorphic
sectional curvature c = 1.
It can be shown that any two simply-connected complete Kahler manifolds with
constant sectional curvature c are holomorphically isometric. Hence, the above
examples together with the at metric on C
n
show that
88 3. K

AHLER MANIFOLDS
Theorem 3.2.10: Let (`
2n
. J. o.
g
) be a complete Kahler manifold of constant
holomorphic sectional curvature c. Depending on the sign of c and up to scaling,
the universal cover

` is holomorphically isometric to CP
n
. C
n
. or 1
n
C
(1).
3.3. Hodge Theory on Kahler Manifolds
Let (`
2n
. J. o.
g
) be a compact Hermitian manifold. We dene the Hodge star
operator to be the map : (
k,l
`)(
nk,nl
)` dened by
(3.3.1) = o(.

)dvol
g
= o(.

)

n
g
n!
. . (
k,l
`) .
Note that this map is linear over R but conjugate-linear over C. In addition, we
have
2
= (1)
k+1
. The action of the Hodge star operator on any complex /-
form (
k
` C) is dened by the splitting of as a sum of forms of type
(/ ,. ,). Locally, in a holomorphic chart (l; .
1
. . . . . .
n
). n = dim
C
(`) one can
write
(
k,l
`) =

]I]=k,]J]=l

I

J
d.
I
d .
J
.
I

J
C

(l. C)

.
where
1 = (i
1
. . . . . i
k
). 1 i
1
< < i
k
n. d.
I
= d.
i
1
d.
i
k
.
J = (,
1
. . . . . ,
l
). 1 ,
1
< < ,
k
n. d .
J
= d .
k
1
d .
j
l
.
Let us assume that (l; .
1
. . . . . .
n
) are normal coordinates, i.e., d.
1
. . . . . d.
n
is
a unitary frame of (
1,0
`). Denote by

1.

J the likewise ordered complements of
1. J in 1. . . . . n and dene (1) by
d.
I
d.

I
= (1)
(I)
d.
1
d.
n
.
Then we have an explicit formula
(
I

J
d.
I
d .
J
) = 2
k+lm
i
n
c
IJ

I

J
d.

I
d .

J
.
where the sign c
IJ
= (1)
n(n1)/2+(nk)l+(I)+(J)
. For compact ` we can dene
an 1
2
Hermitian inner product on each (
k,l
`) by
(3.3.2) '. ` =

M
. . (
k,l
`) .
One can check that is an isometry with respect to this inner product, i.e.,
'. ` = '. `. We proceed to dene
(3.3.3)

=

: (
k,l
`)(
k,l1
`) .
It follows that this is a formal adjoint of the

operator as for any (
k,l1
`)
and any (
k,l
`) we have

=

+ (1)
k+l1


=

( ) = d( ) .
so that, by Stokes theorem, '

. ` = '.

` . Similarly, once can dene

= : (
k,l
`)(
k1,l
`)
which is an adjoint of the operator. Hence, we have 6 dierent operators acting
on forms d. d

. .

.

.

and 3 natural Laplacians

d
= dd

+d

d .

.
3.3. HODGE THEORY ON K

AHLER MANIFOLDS 89
The Laplacian

is a self-adjoint operator and

'

. ` = '

.

` +'

` 0 .
Denition 3.3.1: The space H
k,l
(`) (
k,l
`)
H
k,l
(`) = (
p,q
`) [

= 0 .
is called the space of harmonic (/. |)-forms. Furthermore, dim(H
k,l
(`)) = /
k,l
(`)
are called the Hodge numbers of `.
We now have the famous Hodge Theorem.
Theorem 3.3.2: On a compact Hermitian manifold (`
2n
. J. o.
g
) the dimensions
dim(H
k,l
(`)) < for all 0 /. | n. Furthermore, we have the decompositions
(
k,l
`) = H
k,l
(`)

((
k,l1
`))

((
k,l1
`)) .
In particular, any class in H
k,l

(`) has a unique harmonic representative, i.e., we

have a complex vector space isomorphism
H
k,l
(`) H
k,l

(`) .
Let us introduce the projections

k,l
: (

T` C)(
k,l
`) .

r
=
k+l=r

k,l
: (

T` C)(
r
` C) .
Since the metric o is Hermitian we have
k,l
=
k,l
. In general, given a complex

d
-harmonic :-form the projection
k,l
need not be

-harmonic. However,
on a Kahler manifold we have some additional structure.
To show this let us introduce the operator
1 : (
k,l
`)(
k+1,l+1
`)
dened by wedging with the Kahler form
g
. i.e., 1() =
g
. The adjoint of 1
in the 1
2
-norm is given by
1

= (1)
p+q
1 : (
k,l
`)(
k1,l1
`) .
We have the following
Lemma 3.3.3: [Kahler Identities] Let (`. J. o.
g
) be a Kahler manifold. Then
[1.

] = i

. [1.

] = i .
[1

. ] = i

. [1

.

] = i

.
Proof. Since both sides of all these inequalities are rst order dierential
operators it is enough to check the for the Euclidean space C
n
with the standard
Hermitian metric. The general statement follows then from the fact that Kahler
condition is equivalent to the existence of normal coordinates.
Exercise 3.2: Verify the identities of Lemma 3.3.3 in normal coordinates.
Proposition 3.3.4: Let (`. J. o.
g
) be a compact Kahler manifold. The
(i) [1

.
d
] = [1

.
d
] = 0 .
(ii)
d
= 2

= 2

.
(iii) [
d
.
p,q
] = 0 .
90 3. K

AHLER MANIFOLDS
Proof. Now, the proof of the proposition. We have

d
= ( +

)(

) +(

)( +

) =

+(

) +(

) .
Using Kahler identities we can replace

= i[1

. ] and
(

) = i([1

. ] + [1

. ]) = 0 .
On the other hand, using [1

.

] = i

. we have
i

= (i

) + (i

) = (1

) + (1

) =
= 1

+

1

= [. 1

+

[. 1

] = i(

) = i

There are two important consequences of the above propositions. The rst
is the so-called Hodge decomposition. The second is the so-called hard Lefschetz
theorem. Let us rst discuss the Hodge structure. We dene
H
k,l
(`. C) =
7
k,l
d
(`)
d(

` C) 7
k,l
d
(`)
.
to be the space of d-closed forms of type (/. |) modulo d-exact forms and let H
r
d
(`)
be the space of
d
-harmonic :-forms. Since the usual Laplacian
d
is real, com-
mutes with
k,l
and equals 12 of the

-Laplacian we have
H
r
(`) =
k+l=r
H
k,l
(`) .
H
k,l

(`) = H
k,l
(`) = H
k,l
(`) = H
k,l
(`. C) .
Using the usual Hodge theorem H

DR
(`) = H

(`) for the Laplacian

d
we get
Theorem 3.3.5: For a compact Kahler manifold we have
H
r
(`. C) =
k+l=r
H
k,l
(`. C) . H
k,l
(`. C) = H
l,k
(`. C) .
Since [.

] = 0, by Kodaira-Serre duality
4
the map
: H
k,l
(`)H
nk,nl
(`).
is an isomorphism. Summarizing, we have the following properties of the Hodge
numbers
Corollary 3.3.6: 1ct(`
2n
. J. o.
g
) be a compact Kahler manifold, /
k,l
(`) be the
Hodge numbers, and /
r
(`) =

k+l=r
/
k,l
(`) the Betti numbers of `. Then for
all 0 /. | n we have
(i) /
k,l
(`) < ,
(ii) /
k,k
(`) 1 and /
n,n
= /
0,0
= 1,
(iii) /
k,l
(`) = /
l,k
(`) = /
nk,nl
(`).
(iv) /
2r+1
(`) are even and /
1,0
(`) = /
0,1
(`) =
1
2
/
1
(`) is a topological
invariant.
4
The general version of the Kodaira-Serre Duality Theorem is given in Theorem 3.4.9 below.
3.3. HODGE THEORY ON K

AHLER MANIFOLDS 91
For any compact Kahler manifold once can consider the Hodge diamond, that
is the arrangement of the Hodge numbers in a diamond-shape array. For instance,
when n = 3 we can have
/
3,3
/
3,2
/
2,3
/
3,1
/
2,2
/
1,3
/
3,0
/
2,1
/
1,2
/
0,3
/
2,0
/
1,1
/
0,2
/
1,0
/
0,1
/
0,0
The above corollary implies that every Hodge diamond has two symmetries: con-
jugation gives the symmetry through the central vertical axis while Hodge star
yields the symmetry through the center of the diamond. There are more relations
between Hodge numbers. They are due to the Lefschetz decomposition which we
will discuss below.
Exercise 3.3: Let us dene an operator / =

2n
r=0
(n :)
r
. Show that we have
the following relations
[1

. 1] = /. [/. 1] = 21. [/. 1

] = 21

and, hence, 1. 1

. / are generators of the Lie algebra sl(2. R).

Denition 3.3.7: Let (`. J. o.
g
) be a compact Kahler manifold. We dene the
primitive cohomology groups of ` as kernels of the 1

-operator, i.e.,
1
r
(`. C) = ker

: H
r
(`. C)H
r2
(`. C)).
1
k,l
(`. C) = ker

: H
k,l
(`. C)H
k1,l1
(`. C)).
The dimension of 1
r
(`. C) is called primitive :
th
Betti number of `.
Since 1. 1

. / all commute with the Kahler Laplacian we have

1
r
(`. C) =
k+l=r
1
k,l
(`. C).
The generators 1. 1

. / give a nite-dimensional representation of the Lie algebra

sl(2. R) acting on H

(`. C). where H

r
(`. C) is the eigenspace of the operator /
with eigenvalue (n :). The next theorem follows directly from the theory of
nite-dimensional representations of sl(2. R):
Theorem 3.3.8: (Hard Lefschetz) On a compact Kahler manifold ` the map
1
k
: H
nk
(`)H
n+k
(`) is an isomorphism for 1 / n. Furthermore,
H
r
(`) =

k0
1
k
1
r2k
(`) .
92 3. K

AHLER MANIFOLDS
3.4. Complex Vector Bundles and Chern Classes
Let (`. J) be a smooth manifold and let 1` be a complex vector bundle
of complex rank : over `. Let be a complex Koszul connection (cf. Denition
1.3.5) on 1 with curvature form . Consider the space
rr
(C) of complex : :
matrices. For any A
rr
(C) we dene
det(A +I) = 1
r
(A) +1
r1
(A) + +
r1
1
1
(A) +
r
.
Clearly, for any i = 1. . . . . : the function 1
i
:
rr
(C)C is a G1(:. C)-invariant,
complex homogeneous polynomial of deg(1
i
) = i. Note that 1
i
is the i
th
elementary
symmetric function of the eigenvalues of A. In particular, 1
r
(A) = det(A) and
1
1
(A) = Tr(A).
Denition 3.4.1: Let 1` be rank : complex vector bundle over `. and let
be a complex connection on 1 with curvature 2-form . For each i = 1. . . . . : we
dene the 2i-form
1
i

i
2

= c
i
(1. )
and call it the i
th
Chern form of 1.
We have the following
Proposition/Denition 3.4.2: Given (1. ) and any 1 i :, the i
th
Chern
form c
i
(1. ) is closed. Furthermore, if

is another complex connection on 1
the dierence c
i
(1. ) c
i
(1.

) is exact, i.e., the cohomology class [c

i
(1. )]
H
2i
DR
(`) is independent of . The resulting cohomology class is called the i
th
Chern class of 1 and is denoted by c
i
(1).
When working with Chern classes it is convenient to consider the total Chern
class c(1) =

i=0
c
i
(1). where c
0
(1) = 1 and the sum is always nite for a nite
rank vector bundle, so c(1) H

(`. Z). For Whitney sums the total Chern class

satises c(1 1
t
) = c(1)c(1
t
). If

1 denotes the complex conjugate bundle to 1.
then c
i
(

1) = (1)
i
c
i
(1) and since

1 is isomorphic to the dual complex vector
bundle c
i
(1

) = (1)
i
c
i
(1). There are also certain multiplicative sequences of
interest [Hir66, MS74]. We dene the Chern roots r
i
of a rank / complex vector
bundle by c(1) = 1+c
1
(1)+ +c
k
(1) =

k
i=1
(1+r
i
). Then the Chern character
ch(1) and the Todd class Td(1) are dened by
(3.4.1) ch(1) =
k

i=1
c
xi
. Td(1) =
k

i=1
r
i
1 c
x
i
.
Generally, these are elements in the rational cohomology ring H

(`. Q). Here is

the celebrated Hirzebruch-Riemann-Roch Theorem:
Theorem 3.4.3: Let 1 be a holomorphic vector bundle on a compact complex
manifold `
n
. Then
(`. 1) = t
2n
(Td(`)ch(1)) .
where (`. 1) =

n
i=1
/
i
(`. O(1)) is the holomorphic Euler characteristic of 1.
When 1 is the trivial bundle we get an invariant of the complex structure,
namely the holomorphic Euler characteristic (`. O) and Theorem 3.4.3 gives the
Todd-Hirzebruch formula (`. O) = t
2n
(Td(`)). the later being known as the
Todd genus.
3.4. COMPLEX VECTOR BUNDLES AND CHERN CLASSES 93
Example 3.4.4: The complexied tangent bundle. Consider the case where 1 =
T` C is the complexied tangent bundle of a complex manifold which splits as
T`C = T
1,0
`T
0,1
`. Now T
1,0
` is a holomorphic vector bundle, called the
holomorphic tangent bundle, which is isomorphic as a real vector bundle to T`.
Now T
1,0
` has a complex connection
1,0
which can be used to compute its Chern
classes, c
i
(`) = c
i
(T
1,0
`). called the i
th
Chern class of the complex manifold `.
and denoted simply by c
i
when the complex manifold ` is understood. These
Chern classes depend only on the homotopy class of the complex structure J on `.
Note that the top class c
n
(`) = c(`) is the Euler class. Given any partition 1 =
i
1
. . . . . i
r
of the integer n. we can dene the 1
th
Chern number c
I
[`] by evaluation
on the fundamental homology class [`]. that is, c
I
[`] = 'c
i
1
c
i
r
. [`]`. Note
that the top Chern number c
n
[`] is just the Euler-Poincare characteristic c
n
[`] =
(`) =

i
(1)
i
/
i
(`).
In low dimensions the Todd-Hirzebruch formula reduces to well-known classical
formulae. Recall that /
0,1
= c is called the irregularity. For n = 1 (compact
Riemann surfaces) we get /
0,1
= o. where o is the genus of the Riemann surface.
For n = 2 we get Noethers formula for compact complex surfaces, (`. O) =
1 c +j
g
=
1
12
(c
2
1
[`] +c
2
[`]).
Now suppose that (`. J) is a complex manifold and 1 is a complex vector
bundle on `. We can consider the tensor product bundle 1
k,l
`. and we let
c
k,l
(1) denote the sheaf of germs of smooth sections of 1
k,l
`. Smooth sections
of this sheaf are (/. |)-forms with coecients in 1. the set of which we denote by

k,l
(1). The connection in 1 induces a connection, also written as , in
k,l
(1).
This connection splits as =
1,0
+
0,1
giving maps

1,0
:
k,l
(1)
k+1,l
(1).
0,1
:
k,l
(1)
k,l+1
(1).
We are interested in the case when 1 admits a holomorphic structure. Then simi-
larly, we let
p
(1) denote the sheaf of germs of holomorphic sections of the bundle

p,0
1. So we shall analyze the structure of connections in holomorphic vector
bundles following [Kob87].
Theorem 3.4.5: A smooth complex vector bundle 1 over a complex manifold ad-
mits a holomorphic structure if and only if there is a connection in 1 such that

0,1
=

.
The holomorphic structure in 1 is uniquely determined by the condition
0,1
=

. and this condition says that the (0. 2) component

0,1

0,1
of the curvature
of vanishes. It is straightforward to generalize Denition 3.1.7 to an arbitrary
complex vector bundle.
Denition 3.4.6: An Hermitian metric / on a complex vector bundle 1 is an
assignment of an Hermitian inner product to each bre 1
x
of 1 that varies smoothly
with r. A connection in 1 is called an Hermitian connection if / = 0.
A vector bundle equipped with a Hermitian metric is often called an Hermitian
vector bundle. Using partitions of unity one easily sees that Hermitian metrics
exists on complex vector bundles.
Proposition 3.4.7: Let 1 be a holomorphic vector bundle with an Hermitian met-
ric /. Then there exists a unique Hermitian connection such that
0,1
=

.
Again for a proof see [Kob87]. The unique connection of Proposition 3.4.7 is
called the Hermitian connection. There is a version of Hodge theory tensored with
94 3. K

AHLER MANIFOLDS
holomorphic vector bundles with an Hermitian metric. Such a choice of Hermitian
metric gives an isomorphism between a holomorphic vector bundle 1 and its dual
1

. Then we can dene the Hodge star isomorphism

E
= :
k,l
(1)
nk,nl
(1

)
or, alternatively, on the sheaf level

E
= : c
k,l
(1)c
nk,nl
(1

).
We can now extend the Hodge inner product 3.3.2 to an inner product on
k,l
(1)
by
(3.4.2) '. ` =

M

E
. .
k,l
(1) .
where we write .
k,l
(1) = (
k,l
` 1) as = c. and = 1.
respectively, in which case we have
E
= ()'c. 1`
E
. where 'c. 1`
E
is the
pairing between 1 and 1

.
Now consider the cochain complex

k,l
(1)

k,l+1
(1)
and let H
p,q

(`. 1) denote the derived cohomology groups of this complex. The

operator

has a (formal) adjoint

E
=
E

E
with respect to the Hodge inner
product 3.4.2, and we have complex Laplacian

E
=

E
+

acting on
,
(1).
As in Denition 3.3.1 we dene the space H
k,l
(`. 1) of harmonic 1-valued (/. |)-
forms by
(3.4.3) H
k,l
(`. 1) =
k,l
(1) [

E
= 0 .
Combining the resulting vector bundle version of the Hodge isomorphism The-
orem 3.3.2 with the Abstract de Rham Theorem 1.1.16, and using the acyclic

-
resolution of the sheaf
p
(1) and one gets Serres generalized Dolbeault Theorem
for holomorphic vector bundles [Ser55]:
Theorem 3.4.8: Let ` be a complex manifold, and 1 a holomorphic vector bundle
on `. Then there are isomorphisms
H
q
(`.
p
(1)) H
p,q

(`. 1) H
p,q
(`. 1) .
This theorem can be used to prove the celebrated Kodaira-Serre Duality The-
orem [Ser55]:
Theorem 3.4.9: Let ` be a compact complex manifold of complex dimension n.
and let 1 be a holomorphic vector bundle over A. There there is a conjugate-linear
isomorphism
: H
q
(`.
p
(1))H
nq
(`.
np
(1

)) .
3.5. Line Bundles and Divisors
The purpose of this section is to give a brief review of the fundamental concepts
employed in complex manifold theory, namely, line bundles and divisors as well as to
discuss their interrelationship. We refer to the literature [GH78b, Wel80, Voi02,
Laz04a] for complete treatments of these important subjects.
3.5. LINE BUNDLES AND DIVISORS 95
3.5.1. Line Bundles. We begin by considering the set of smooth complex
line bundles on a complex manifold `. These are determined by their transition
functions which have values on in the group G1(1. C) = C

. The isomorphism
classes of complex line bundles form a group under tensor product, and by Theorem
1.2.2 isomorphism classes of C

-bundles over ` are in one-to-one correspondence

with elements of the sheaf cohomology group H
1
(`. c

). Thus, we consider the

exponential short exact sequence of sheaves on `
(3.5.1) 0Z

c
exp
c

0 .
where the map is (/) = 2i/ and the exponential map sends the germ 1 of any
holomorphic function to exp(1). Since c is a ne sheaf, Propositions 1.1.12 and
1.1.13 imply H
q
(`. c) = 0 for all c 0. So the induced long exact cohomology
sequence (1.1.6) gives an isomorphism H
1
(`. c

) H
2
(`. Z) which says that
the topological invariant H
2
(`. Z) can be thought of as the group of complex line
bundles on `. This isomorphism is realized by associating to a complex line bundle
L its rst Chern class c
1
(L).
To study the holomorphic line bundles on ` we consider the exact sequence
(3.5.2) 0Z

O
exp
O

0 .
This induces a long exact sequence in cohomology,
(3.5.3) H
1
(`. O)H
1
(`. O

)

H
2
(`. Z)H
2
(`. O) .
The group H
1
(`. O

) represents the group of holomorphic line bundles on ` with

group multiplication being the tensor product, and the inverse bundle being the
dual bundle. This group is called the Picard group of ` and often denoted by
Pic(`). As seen above the connecting homomorphism takes a holomorphic line
bundle L to its rst Chern class c
1
(L). and the group H
2
(`. Z) is isomorphic to the
group of topological complex line bundles on `. So if H
2
(`. O) = 0 we see that
not every complex line bundle gives rise to a holomorphic line bundle. Similarly,
if H
1
(`. O) = 0. there can be inequivalent holomorphic bundles associated to the
same complex line bundle. The kernel of the map is denoted by Pic
0
(`) and
represents the subgroup of holomorphic line bundles that are trivial topologically.
The quotient group Pic(`)Pic
0
(`) is known as the Neron-Severi group denoted
by o(`). The rank of o(`) is called the Picard number of ` and denoted
by (`). When ` is a smooth projective algebraic variety, we have 1 (`)
/
2
(`). and it follows from the well-known Lefschetz Theorem on (1. 1) classes that
there is an isomorphism o(`) H
1,1
(`. C) H
2
(`. Z). So o(`) is a free
Abelian group of rank (`).
The notion of positivity is fundamental in the study of holomorphic line bun-
dles.
Denition 3.5.1: Let ` be a complex manifold and 1 a holomorphic line bundle
on `. We say that 1 is positive (negative) if it carries an Hermitian met-
ric whose curvature form with respect to the Hermitian connection is a positive
(negative) (1. 1)-form on `.
Since Chern classes can be computed with respect to any connection, we see
that the rst Chern class c
1
(1) can be represented by
i
2
. Notice that when 1 is
positive, the form
i
2
denes a Kahler metric on ` with integral cohomology
class. This leads to
96 3. K

AHLER MANIFOLDS
Denition 3.5.2: The pair (`. 1) consisting of a complex manifold ` with a
positive line bundle 1 is called a polarized Kahler manifold.
Of course, the complex manifold ` is necessarily a manifold of Kahler type, and
the Kahler structure is that determined by the curvature of 1. namely, =
i
2
. So
we often refer to the pair (`. []) as a polarized Kahler manifold. If ` is a Kahler
manifold with an integral Kahler class [] or equivalently a positive line bundle 1.
we also say that ` admits a polarization or is polarized by [] or 1.
There is a complex line bundle canonically associated with every complex man-
ifold, called the canonical bundle.
Denition 3.5.3: Let ` be a complex manifold of complex dimension n. The
n
th
exterior power
n
T
(1,0)
` =
n,0
` is a holomorphic line bundle, called the
canonical bundle and denoted by /
M
. The dual or inverse line bundle /
1
M
is
called the anticanonical bundle.
When the underlying manifold ` is understood we often write just / for /
M
.
It is easy to see that
Proposition 3.5.4: The rst Chern class of /
M
satises
c
1
(/
M
) = c
1
(`) .
There are important discrete invariants associated to /
M
and hence, to `. On
any complex manifold we dene the plurigenera by
(3.5.4) 1
m
(`) = dim H
0
(`. O(/
m
)) = /
0
(`. O(/
m
)) .
Recall that 1
1
(`) is called the geometric genus and is usually denoted by j
g
. Note
that by the well-known Dolbeault Theorem 3.4.8, j
g
= /
n,0
. Another important
invariant is the holomorphic Euler characteristic (`. O) =

i
/
i
(`. O).
Consider the commutative ring
(3.5.5) 1(`) = C

m1
H
0
(`. O(/
m
)) .
R(M), called the canonical ring, has nite transcendence degree tr(1(`)) over C.
Denition 3.5.5: Let ` be a compact complex manifold. The Kodaira dimension
of ` is dened by
Kod(`) =

if 1(`) = C
tr(1(`)) 1 otherwise,
where tr(1) denotes the transcendence degree of the ring 1.
More generally one can dene (cf. [Laz04a]) the Iataka dimension (`. L) of
any line bundle L by simply replacing / by L in Equation 3.5.5. Then Kod(`) =
(`. /). The Iataka dimension and, therefore, the Kodaira dimension can be de-
ned on any normal projective algebraic variety. We shall make use of the Iataka
dimension when discussing the orbifold Kodaira dimension in Chapter 4. Recall
that the transcendence degree of the eld of meromorphic functions on a compact
complex manifold is called the algebraic dimension and is denoted by o(`). Re-
garding the Kodaira dimension we have Kod(`) o(`) n = dim `. The
Kodaira dimension gives a measure of the asymptotic growth of the plurigenera.
Theorem 3.5.6: Let ` be compact complex manifold. Then
(i) Kod(`) = if and only if 1
m
(`) = 0 for all :;
3.5. LINE BUNDLES AND DIVISORS 97
(ii) Kod(`) = 0 if and only if 1
m
(`) = 0 or 1. but not 0 for all :;
(iii) Kod(`) = / if and only if there exist constants . such that :
k

1
m
(`) :
k
.
The importance of the Kodaira dimension is that it is a birational invariant.
Recall [Har77] that a rational map is dened as follows: (it is not a map in the usual
sense) let A. Y be varieties (either both ane or both projective) and consider pairs
(l.
U
). where l is a nonempty open subset of A. and
U
: lY is a holomorphic
map. Two such pairs (l.
U
) and (\.
V
) are equivalent if
U
[
UV
=
V
[
UV
. Then
a rational map : AY is an equivalence class of such pairs. is said to be
dominant of if the image
U
is dense in Y. A birational map is a rational map that
admits an inverse, namely, a rational map : Y A such that = 1l
X
and
= id
Y
. For a thorough treatment of rational maps we recommend the recent
book [Har92].
The most important example of a rational map is the blowing-up map which
we now describe.
Example 3.5.7: [Blowing-up] Let 1
r
(0) be a ball of radius : centered at 0 in
C
n
. where n 2. and let z = (.
1
. . . . . .
n
) be the standard coordinates in C
n
. Then
blowing-up 1
r
(0) at 0 is the complex manifold dened by

1 = (z. y) 1 CP
n1
[ .
i
n
j
= .
j
n
i
. 1 i < , n .
where y = (n
1
. . . . . n
n
) are homogeneous coordinates for CP
n1
. There is a natural
surjective holomorphic map :

11 dened as the restriction to

1 of the
projection map 1 CP
n1
onto the rst factor. For z = 0 the bre
1
(z) is the
single point (z. [z]); whereas, at z = 0 we have
1
(0) CP
n1
. So

1 is 1 with the
origin 0 replaced by a projective space CP
n1
. The bre
1
(0) = CP
n1
is called
the exceptional divisor, and often denoted by 1. Away from the exceptional divisor
is a biholomorphism. Replacing

1 by 1 is called blowing-down. One can show
that

1 is dieomorphic to 1#CP
n
. where # denotes the connected sum operation
and CP
n
is CP
n
with the reverse orientation.
It is easy to transfer the blowing-up process to an arbitrary complex manifold
`. Let j ` and \ a neighborhood containing j that is biholomorphic to 1 with
j mapping to 0. Identifying \ with 1, we let

` be the manifold obtained from `
by replacing \ = 1 by

1. So

` is a complex manifold that is dieomorphic to
`#CP
n
. Analytically, is a birational map, so we have the following birational
invariants: o(

`) = o(`). 1
m
(

`) = 1
m
(`). and Kod(

`) = Kod(`). We also
have an isomorphism H
i
(`. O
M
) H
i
(

`. O

M
) for all i 0. and /

M
=

/
M

O((n1)1). where O(1) is the line bundle associated to the divisor 1 as described
in subsection 3.5.2 below.
Blowing-up can be applied to singular points of algebraic varieties to obtain
smooth manifolds. This procedure has led to the celebrated Hironaka Resolution
of Singularities Theorem which says that a singular algebraic variety can be desin-
gularized after a nite sequence of blowing-ups.
Now for certain complex manifolds there is a vanishing theorem H
q
(`. O) =
0 for all c 0. When this happens we get an isomorphism between the Picard
group H
1
(`. O

) of holomorphic line bundles on ` and the topological invariant

H
2
(`. Z). Such vanishing theorems are of great importance in complex geometry.
98 3. K

AHLER MANIFOLDS
Here we only give the well-known Kodaira-Nakano vanishing Theorem, and we refer
to Chapter 4 of [Laz04a] and [SS85] for further development.
Theorem 3.5.8: Let ` be a compact complex manifold of complex dimension n.
and let L be a holomorphic line bundle on `.
(i) If L/
1
is positive, then H
q
(`. O(L)) = 0 for all c 0. In particular,
if the anticanonical bundle /
1
is positive, H
q
(`. O) = 0 for all c 0.
(ii) If L is negative, then H
q
(`.
p
(L)) = 0 for all j +c < n.
Compact manifolds with positive anticanonical bundles are called Fano man-
ifolds, and it follows immediately from the long exact sequence 3.5.3 and (i) of
Theorem 3.5.8 that
Corollary 3.5.9: For any Fano manifold ` there are isomorphisms Pic(`)
o(`) H
2
(`. Z). So for Fano manifolds (`) = /
2
(`).
Example 3.5.10: A simple example of a Fano manifold is the complex projective
space CP
n
as discussed in Examples 3.1.12 and 3.2.8. It follows from Equation 3.2.2
and Proposition 3.6.1 that c
1
(CP
n
) is positive. It is known that H
2
(CP
n
. Z) =
Z. and it follows from Proposition 3.5.4 and Theorem 3.5.8 that Pic(CP
n
)
o(CP
n
) Z. The positive generator of Pic(CP
n
) is known as the hyperplane
bundle and denoted by H or as we do using the invertible sheaf notation O(1). So
every holomorphic line bundle on CP
n
is O(n) for some n Z. The line bundle
O(1) is called the tautological bundle since its total space is the C
n+1
from which
CP
n
is constructed. It is easy to compute the canonical bundle of CP
n
nding
/
CP
n = O(n 1).
Example 3.5.11: Low dimensional Fano manifolds. The smooth Fano varieties
have been classied for n = 1. 2. 3. There is a unique one dimensional smooth Fano
variety up to isomorphism; it is the complex projective plane CP
1
. For n = 2 they
are classical and known as del Pezzo surfaces. A smooth del Pezzo surface is, up to
isomorphism, one of the following:
(i) CP
2
.
(ii) a smooth quadric Q = CP
1
CP
1
.
(iii) a double cover 1
1
of a quadric cone Q
t
CP
3
ramied along a smooth curve
of degree 6 not passing through the vertex of the cone,
(iv) a double cover 1
2
of CP
2
ramied along a smooth curve of degree 4.
(v) a surface 1
d
CP
d
. d = 1
2
F
d
. where 3 d 7.
(vi) a geometrically ruled surface F
1
with the exceptional section :. :
2
= 1.
One can show that surfaces 1
d
, d = 1. . . . . 7 can be obtained by blowing up of 9d
points on CP
2
which are in general position, i.e., no two of these points lie on a
line nor any three lie on a conic. Also, F
1
can be obtained by blowing CP
2
at 1
point. Thus, as a smooth manifold a del Pezzo surface must be dieomorphic to
CP
2
. Q = CP
1
CP
1
. F
1
= CP
2
#CP
2
. or 1
d
= CP
2
#(9 d)CP
2
. where 1 d 7.
The classication for n = 3 was begun by Fano and almost completed by
Iskovskikh in [Isk77, Isk78, Isk79]. However, an additional 3-fold was found
a bit later by Mukai and Umemura [MU83], and it was shown by Prokhorov
[Pro90] to complete the classication. We refer to [IP99] for a survey including
the complete list of Fano 3-folds. Generally, Fano manifolds and orbifolds are of
much importance for us in this book. For recent treatments of Fano varieties see
[Kol96, IP99].
3.5. LINE BUNDLES AND DIVISORS 99
More generally, any generalized ag manifold G1 is Fano. Fano manifolds
have the important property that they can always be embedded into a complex
projective space CP
N
for some . i.e., they are examples of projective algebraic
varieties as discussed in Example 3.1.18. The beautiful Kodaira embedding theorem
gives precise conditions to have a projective algebraic variety.
Theorem 3.5.12: Let ` be a compact complex manifold and 1 a holomorphic line
bundle on `. Then 1 is positive if and only if there is a holomorphic embedding
: `CP
N
for some such that

O(1) = 1
m
for some : 0.
The condition of the holomorphic embedding gives rise to
Denition 3.5.13: A holomorphic line bundle 1 on a complex manifold ` is
said to be ample if there is an embedding : `CP
N
for some such that

O(1) = 1
m
for some : 0. If we can take : = 1 then 1 is said to be very
ample.
The Kodaira Embedding Theorem can now be reformulated in two more equiv-
alent ways.
Theorem 3.5.14: Let ` be a compact complex manifold. Then
(i) A holomorphic line bundle 1 on ` is positive if and only if it is ample
in which case ` is a projective algebraic variety.
(ii) ` admits polarization if and only if it is projective algebraic.
So a Fano manifold could be dened by the condition that the anticanonical
line bundle /
1
M
is ample. In this case the corresponding embedding is said to be
an anticanonical embedding. Similarly, if the canonical line bundle /
M
is ample,
then the manifold ` is projective algebraic and the corresponding embedding is
called a canonical embedding.
A very important criterion for the ampleness of a line bundle was discovered
by Nakai in the case of complex surfaces and generalized to arbitrary smooth al-
gebraic varieties by Moishezon. It was generalized further to complete schemes by
Kleiman. Here we simply state the version for smooth algebraic varieties and refer
to Lazarsfelds recent book [Laz04a] for further discussion and proof of the scheme
theoretic version.
Theorem 3.5.15: Let ` be a compact manifold of complex dimension n. and 1
a holomorphic line bundle on `. Then 1 is ample if and only if

V
c
1
(1)
k
0
for every irreducible subvariety \ ` of dimension /.
In the case of surfaces Nakais criterion can be stated as 1 is ample if and only
if c
2
1
(1) 0 and

D
c
1
(1) 0
for every eective divisor 1 on `.
If the Kahler class [] is a rational class, that is it lies in H
2
(`. Q). then /[] is
an integral class for some positive integer /. So /[] denes a positive holomorphic
line bundle 1 on `. and Theorem 3.5.14 implies that (`. ) is a projective algebraic
variety. A Kahler manifold (`. ) with [] H
2
(`. Q) is called a Hodge manifold.
100 3. K

AHLER MANIFOLDS
Example 3.5.16: A complex torus is the quotient manifold T
n
= C
n
. where
is a lattice, that is a discrete subgroup of C
n
of rank 2n. A at Kahler structure on
C
n
induces a Kahler structure on T
n
. However, for a generic complex structure,
the Kahler class [] H
1,1
(`. C) H
2
(`. R) is not a rational class. So generally
a complex torus T
n
is not a projective algebraic variety. When one can nd a
compatible Kahler form such that [] is a rational class, T
n
is projective algebraic
by the Kodaira Embedding Theorem 3.5.12 in which case it is called an Abelian
variety. Alternatively, a complex torus T
n
which admits a positive line bundle 1
is an Abelian variety. So the pair (T
n
. 1) is called a polarized Abelian variety. For
a complete treatment of Abelian varieties we refer the reader to [LB92].
3.5.2. Divisors. There are two equivalent way to describe divisors on smooth
complex manifolds. Since they are not equivalent for singular algebraic varieties or
more generally for complex spaces, we discuss both of these here. The singular case
is treated in Section 4.4 below.
Denition 3.5.17: A Weil divisor 1 on a complex manifold ` is a locally nite
formal linear combination of irreducible analytic hypersurfaces \
i
1 =

i
o
i
\
i
. o
i
Z.
where locally nite means that every point j ` has a neighborhood intersecting
only nitely many of the \
i
s. 1 is said to be eective if o
i
0 for all i with not
all o
i
equal to zero.
Under the formal sum operation Weil divisors form a group, called the divisor
group and denoted by Div(`). Now from Denition 3.1.19 a divisor is described
locally by the zero set of holomorphic functions.
Recall that a meromorphic function on an open set l ` is a ratio 1o of
relatively prime holomorphic 1. o on l. We let denote the sheaf of holomor-
phic functions and

the subsheaf of not identically zero meromorphic functions.

Consider the short exact sequence
(3.5.6) 0O

0 .
We have
Denition 3.5.18: Let A be a complex manifold or an algebraic variety. Elements
of the group
H
0
(A.

)
are called Cartier divisors on A. A Cartier divisor is principal if it is the divisor
of a global meromorphic function, that is, it is in the image of the natural quotient
map H
0
(A.

)H
0
(A.

).
Since on a smooth manifold the local rings O
x
are unique factorization domains
(UFD) Weil divisors and Cartier divisors coincide.
Theorem 3.5.19: Let ` be a smooth complex manifold. Then there is an isomor-
phism
Div(`) H
0
(`.

) .
This isomorphism does not hold on singular complex spaces. For example the
complex spaces discussed in Section 4.4.1 below are not locally factorial. On
smooth complex manifolds we shall often identify Weil divisors and Cartier divisors
3.5. LINE BUNDLES AND DIVISORS 101
by just referring to a divisor. Two divisors 1 and 1
t
on ` are said to be linearly
equivalent, written 1
t
1. if 1
t
= 1 +(1). where (1) denotes the divisor dened
by the global meromorphic function 1 as follows: write 1 locally as 1 = o/ as
(1) = ord(o)7
g
ord(/)7
h
. Here 7
g
denotes the zero set of the holomorphic function
o and ord(o) denotes the its order of vanishing. We denote by [1[ the set of all
divisors on ` that are linearly equivalent to 1. It is called the linear system of
divisors dened by 1. The common intersection

]D]
1
t
is called the base locus
of linear system [1[.
We now describe the relationship between line bundles and divisors. From the
short exact sequence (3.5.6) one has
(3.5.7) 0H
0
(`.

)H
0
(`.

)H
1
(`. O

)H
1
(`.

) .
This says that every divisor 1 on ` determines a holomorphic line bundle O(1).
and the line bundle O(1) is holomorphically trivial if and only if 1 is the di-
visor of a meromorphic function. The quotient H
0
(`.

)H
0
(`.

) is
called the Cartier divisor class group and is denoted by CaCl(`). Furthermore,
H
1
(`.

) = 0 if and only if every holomorphic line bundle on ` has a global

meromorphic section. In this case we get an isomorphism between the divisor class
group and the Picard group. This happens, for example, for smooth projective
algebraic varieties, that is
Proposition 3.5.20: Let A be a smooth projective algebraic variety. Then Cl(A)
Pic(A).
We now briey describe the intersection theory for divisors. The general theory
is laid out in [Ful84], but our treatment follows closely the abbreviated version in
[Laz04a]. Here we work in the category of varieties (either ane or projective).
Let A be an irreducible variety. To each Cartier divisor 1 H
0
(A.

) we
associate a line bundle O(1). and to this line bundle we can associate a Chern
class c
1
(O(1)) H
2
(A. Z). Now let C A be an irreducible curve, that is a one
dimensional irreducible complex subvariety of A. and denote by [C] its homology
class in H
2
(A. Z). Then
Denition 3.5.21: The intersection number 1 C Z of a Cartier divisor 1
with an irreducible curve C is dened by
1 C =

C
1 = 'c
1
(O(1)). [C]` Z.
where '. ` denotes the Kronecker pairing.
This denition can easily be generalized as follows: let \ be a k-dimensional
subvariety of A. and let 1
1
. . . . . 1
k
be Cartier divisors on A. Then we dene the
intersection number by
(3.5.8) 1
1
1
k
\ = 'c
1
(O(1
1
)) c
1
(O(1
k
)) . [\ ]`
Denition 3.5.22: We say that two Cartier divisors 1. 1
t
are numerically equiv-
alent, denoted by 1
t

n
1 if 1
t
C = 1C for all irreducible curves C. A divisor 1
is numerically trivial if it is numerically equivalent to 0. We denote by Num(A)
the subgroup of H
0
(A.

) consisting of numerically trivial Cartier divisors.

We now have another characterization of the Neron-Severi group, namely as
the quotient group of numerical classes of Cartier divisors,
o(A) = H
0
(A.

)Num(A) .
102 3. K

AHLER MANIFOLDS
We shall also make use of the following terminology [Laz04a].
Denition 3.5.23: Let A be a complete variety of dimension n. A (Cartier) divi-
sor 1 is said to be numerically eective or nef if for every irreducible curve C
we have 1 C 0. The divisor 1 is big if the Iataka dimension (A. O(1)) = n.
An important criterion for the bigness of a nef divisor is: a nef divisor 1 is big
if and only if 1
n
0.
3.6. The Calabi Conjecture and the Calabi-Yau Theorem
The Ricci form

dened in Proposition 3.2.6 plays a special role in Kahler

geometry. Indeed, from Equation 3.2.1 and Denition/Proposition 3.4.2 we have
Proposition 3.6.1: Let (`. J. o.
g
) be a Kahler manifold and

the Ricci form

of o. Then
[

] = 2c
1
(`) H
2
DR
(`) H
1,1
(`. C).
This proposition says that on a Kahler manifold the Ricci curvature 2-form

of any Kahler metric represents the cohomology class 2c

1
(`). The well-known
Calabi Conjecture [Cal56, Cal57] is the question whether or not the converse is
also true, that is, the assertion that on a compact complex manifold any real closed
(1. 1)-form that represents 2c
1
(`) is the Ricci form of a unique Kahler metric o on
`. This conjecture was proven in the celebrated work of Yau which was announced
in [Yau77] and the details of which appeared in [Yau78]. (Uniqueness was actually
proved by Calabi [Cal57] before the conjecture was settled). By now there are
several excellent treatments in book form [Aub82, Aub98, Joy00, Siu87, Tia00].
While these proofs all assume one is working with compact Kahler manifolds, the
proofs readily extend to the case of compact Kahler orbifolds since the computations
are local in nature and one needs only to consider local functions and sections on the
local uniformizing neighborhoods that are invariant under the local uniformizing
groups.
To be more specic we begin with a couple of denitions
Denition 3.6.2: Let (`. J. o.
g
) be a compact Kahler manifold. The Kahler
cone of `
1(`) = [] H
1,1
(`. C) H
2
(`. R) [ =
h
for some Kahler metric / .
i.e., it is the set of all possible Kahler classes on `.
It is easy to show that 1(`) is a convex open set in H
1,1
(`. C) H
2
(`. R).
Of particular interest to us is the Kahler lattice 1
L
(`) which is the intersection
of 1(`) with the Neron-Severi group, i.e., 1
L
(`) = 1(`) o(`). Beyond
surfaces, not much is known about the Kahler cone. However, recent progress has
been made by Demailly and Paun [DP04] who characterize 1(`) as a connected
component of the set of (1. 1) cohomology classes which are numerically positive on
all analytic cycles. This generalizes the well-known Nakai-Moishezon criterion for
ample line bundles.
Denition 3.6.3: Let (`. J. o.
g
) be a compact Kahler manifold and 1(`) its
Kahler cone. For any xed Kahler class [] 1(`) we dene
1
[]
= / (
2
(T`)) [ / is a Kahler metric and [] = [
h
]
to be the space of all Kahler metrics in a given cohomology class.
3.6. THE CALABI CONJECTURE AND THE CALABI-YAU THEOREM 103
The global i

-lemma provides for a very simple description of the space of

Kahler metrics 1
[]
. Suppose we have a Kahler metric o in a with Kahler class
[
g
] = [] 1(`). If / 1
[]
is another Kahler metric then, up to a constant,
there exists a global function C

(`. R) such that

h

g
= i

. We could
x constant by requiring, for example, that

M
dvol
g
= 0. Hence, we have
Corollary 3.6.4: Let (`. J. o.
g
) be a compact Kahler manifold with [
g
] = []
1(`). Then, relative to the metric o the space of all Kahler metric is 1
[]
can be
described as
1
[]
= C

(`. R) [
h
=
g
+i

0.

M
dvol
g
= 0 .
where the 2-form
h
0 means that
h
(A. JY ) is a Hermitian metric on `.
The following theorem is the statement of the famous conjecture made by Calabi
and proved by Yau in 1977-78 [Yau77, Yau78].
Theorem 3.6.5: Let (`. J. o.
g
) be a compact Kahler manifold. Then any real
(1. 1)-form on ` which represents the cohomology class 2c
1
(`) is the Ricci
form of a unique Kahler metric / such that [
h
] = [
g
].
Let us reformulate the problem using the global i

-lemma. We start with a

given Kahler metric o on ` in Kahler class [
g
] = []. Since
g
also represents
2c
1
(`) there exists a globally dened function 1 C

(`. R) such that

g
= i

1 .
Appropriately, 1 could be called a discrepancy potential function for the Calabi
problem and we could x the constant by asking that

M
(c
f
1)dvol
g
= 0.
Now suppose that the desired solution of the problem is a metric / 1
[]
. We
know that the Kahler form of / can be written as

h
=
g
+i

.
for some smooth function C

(`. R). We normalize as in previous corollary.

Combining these two equations we see that

g
= i

1 .
If we dene a smooth function 1 C

(`. R) relating the volume forms of the two

metrics dvol
h
= c
F
dvol
g
then the left-hand side of the above equation takes the
following form
i

1 =
h

g
= i

.
and, hence, simply i

(1 1) = 0. Hence, 1 = 1 + c. But since we normalized

M
(c
f
1)dvol
g
= 0 we must have c = 0. Hence, 1 = 1. or dvol
h
= c
f
dvol
g
. We
can now give two more equivalent formulations of the Calabi Problem.
Theorem 3.6.6: Let (`. J. o.
g
) be a compact Kahler manifold, [
g
] = 1(`)
the corresponding Kahler class and
g
the Ricci form. Consider any positive (1. 1)-
form on ` such that [] = 2c
1
(`). Let
g
= i

1. with

M
(c
f
1)dvol
g
= 0.
(i) There exists a unique Kahler metric / 1
[]
whose volume form satises
dvol
h
= c
f
dvol
g
.
(ii) Let (l; .
1
. . . . . .
n
) be a local complex chart on ` with respect to which the
metric o = (o
i

j
). Then, up to a constant, there exists a unique smooth
104 3. K

AHLER MANIFOLDS
function in 1
[]
, which satises the following equation
det

o
i

j
+

2

zi zj

det(o
i

j
)
= c
f
.
The equation in (ii) is called the Monge-Amp`ere equation. Part (i) gives a
very simple geometric characterization of the Calabi-Yau theorem. On a compact
Kahler manifold one can always nd a metric with arbitrarily prescribed volume
form. The uniqueness part of this theorem was already proved by Calabi. This
part involves only the Maximum Principle. The existence proof uses continuity
methods and it involves several dicult a priori estimates. These were found by
Yau [Yau78] in 1978. For the details we refer the reader to the books mentioned
previously. Here we only list some immediate consequences of this theorem:
Corollary 3.6.7: Let (`. J. o.
g
) be a compact Kahler manifold with c
1
(`) = 0.
Then ` admits a unique Kahler Ricci-at metric.
Such a metric has holonomy group inside ol(n). Kahler manifolds with this
property are called Calabi-Yau manifolds.
Corollary 3.6.8: Let (`. J. o.
g
) be a compact Kahler manifold with c
1
(`) 0.
Then ` admits a Kahler metric of positive Ricci curvature.
We can combine this corollary with an older result of Kobayashi [Kob61] to
give a beautiful proof of
Theorem 3.6.9: Any Fano manifold is simply connected.
Proof. Let ` be a Fano manifold. Then its anticanonical line bundle /
1
M
is
positive, so c
1
(`) = c
1
(/
1
M
) 0. Thus, by Corollary 3.6.8 ` has a Kahler metric
of positive Ricci curvature. By Myers Theorem ` has nite fundamental group.
Let

` be the universal cover of `. It is compact, Fano and a nite, say d-fold,
cover of `. By the Kodaira-Nakano Vanishing Theorem 3.5.8 H
q
(

`. O) = 0 for
all c 0 which implies that the holomorphic Euler characteristic (

`. O) = 1. On
the other hand, (

`. O) = d(`. O) = d implying that d = 1. So ` is simply

connected.
We shall return to important consequences of Yaus Theorem in Chapter 5.
CHAPTER 4
Fundamentals of Orbifolds
The notion of orbifold was introduced under the name V-manifold by Sa-
take [Sat56] in 1956, and subsequently he developed Riemannian geometry on
V-manifolds [Sat57] ending with a proof of the Gauss-Bonnet theorem for V-
manifolds. Contemporaneously, Baily introduced complex V-manifolds and gen-
eralized both the Hodge decomposition theorem [Bai56], and Kodairas projective
embedding theorem [Bai57] to V-manifolds. Somewhat later in the late 1970s
and early 1980s Kawasaki generalized various index theorems [Kaw79, Kaw78,
Kaw81] to the category of V-manifolds. It was about this time that Thurston
[Thu79] rediscovered the concept of V-manifold, under the name of orbifold, in his
study of the geometry of 3-manifolds, and dened the orbifold fundamental group

orb
1
. By now orbifold has become the accepted term for these objects and we shall
follow suit. However, we do use the name V-bundle interchangeably with orbibun-
dle for bre bundles in this category. At this stage we can also mention the recent
book [ALR06] which we refer to for certain foundational results, especially from
the groupoid point of view. Although we start with a very general setting, the real
focus of this chapter is on complex and specically on Kahler orbifolds as they are
be of crucial importance in understanding the quasi-regular Sasakian structures of
Chapter 7. However, more general orbifold structures, such as quaternionic Kahler
orbifolds of Chapters 12 and 13, will appear and will be equally important.
4.1. Basic Denitions
Orbifolds arise naturally as spaces of leaves of Riemannian foliations with com-
pact leaves, and we are particularly interested in this point of view. Conversely,
every orbifold can be realized in this way. In fact, given an orbifold O. we can
construct on it the V-bundle of orthonormal frames whose total space 1 is a
smooth manifold with a locally free action of the orthogonal group O(n) such that
O = 1O(n). Thus, every orbifold can be realized as the quotient space by a lo-
cally free action of a compact Lie group. The denition given here is close to
Satakes original denition [Sat56, Sat57] (cf. [Bai56]), but with more modern
terminology as in Moerdijk and Pronk [MP97]. Here F denotes either R or C.
Denition 4.1.1: Let A be a paracompact Hausdor space. An orbifold chart
or local uniformizing system on A is a triple (

l. . ). where

l is connected
open subset of F
n
containing the origin, is a nite group acting eectively as
dieomorphisms of

l. and :

ll is a continuous map onto an open set l A
such that = for all and the induced natural map of

l onto l is a
homeomorphism. An injection or embedding between two such charts

l. .
and

l
t
.
t
.
t
is a smooth embedding :

l

l
t
such that
t
= . An orbifold
atlas on A is a family | =

l
i
.
i
.
i
of orbifold charts such that
105
106 4. FUNDAMENTALS OF ORBIFOLDS
(i) A =
i

i
(

l
i
).
(ii) given two charts (

l
i
.
i
.
i
) and (

l
j
.
j
.
j
) with l
i
=
i
(

l
i
) and l
j
=

j
(

l
j
) and a point r l
i
l
j
, there exist an open neighborhood l
k
of r,
and a chart (

l
k
.
k
.
k
) such that there are injections
ik
: (

l
k
.
k
.
k
)
(

l
i
.
i
.
i
) and
jk
: (

l
k
.
k
.
k
)(

l
j
.
j
.
j
).
An atlas | is said to be a renement of an atlas 1 if there exists an injection of every
chart of | into some chart of 1. Two orbifold atlases are said to be equivalent
if they have a common renement. A smooth orbifold (or V-manifold) is a
paracompact Hausdor space A with an equivalence class of orbifold atlases. We
denote the orbifold by A = (A. |). If every nite group consists of orientation
preserving dieomorphisms and there is an atlas such that all the injections are
orientation preserving then the orbifold is orientable.
Remarks 4.1.1: 1) The nite groups
i
are called local uniformizing groups, and it
follows from the usual slice theorem that we can always take them to be nite sub-
groups O(n. F) G1(n. F). where O(n. R) = O(n) is the usual orthogonal group,
and O(n. C) = l(n). the unitary group. Of course, in the orientable case the local
uniformizing groups lie in oO(n). The condition that the local uniformizing groups
act eectively is not always imposed in the literature, and there are occasions when
this is too restrictive. However, in our situation we are almost always concerned
with eective orbifolds
1
so it is convenient to incorporate this as part of our de-
nition. We shall refer to the more general situation when at least one of the local
uniformizing groups does not act eectively as a non-eective orbifold [HM04]. 2)
Every orbifold atlas is contained in a unique maximal atlas satisfying the required
properties. Thus, we shall often think of the orbifold A = (A. |). where | is the
unique maximal atlas. This justies the notation. 3) Notice that any satises
( r) = (r) for any r

l. Thus, every denes an injection by r r.
The following lemma was proved by Satake in [Sat57] under the added as-
sumption that the xed point set has codimension at least two
2
. However, this
assumption was removed recently by Moerdijk an Pronk [MP97]. We refer to
Proposition 2.12 and Lemma 2.11 of [MM03] for the proof.
Lemma 4.1.2: Let
1
.
2
: (

l. . )(

l
t
.
t
.
t
) be two injections. Then there
exists a unique
t

t
such that
2
=
t

1
.
Let (A. |) be an orbifold and choose a local uniformizing system l. . .
Let r A be any point, and let j
1
(r). then up to conjugacy the isotropy
subgroup
p
depends only on r. and accordingly we shall denote this isotropy
subgroup by
x
. We now have
Denition 4.1.3: A point of A whose isotropy subgroup
x
= id is called a singu-
lar point. Those points with
x
= id are called regular points. The set of singular
points is called the orbifold singular locus or orbifold singular set, and is
denoted by
orb
(A).
The subset of regular points is an open dense subset of A. The isotropy groups
give a natural stratication of A by saying that two points lie in the same stratum
if their isotropy subgroups are conjugate. Thus, we get a decomposition of A as
(4.1.1) A = A
orb
reg
.
j

orb
j
(A) .
orb
(A) = .
j

orb
j
(A) .
1
Eective orbifolds are sometimes called reduced orbifolds, cf. [CR04].
2
The codimension two condition occurred as part of Satakes original denition of V-manifold.
4.1. BASIC DEFINITIONS 107
where the union is taken over all conjugacy classes. The dense open subset of regular
points A
orb
reg
forms the principal stratum corresponding to the trivial conjugacy
class. It is well-known that an orbifold singular point may or may not be an honest
singularity. For example as a variety A may have both smooth points and singular
points. The set of smooth or regular points A
reg
is generally larger than the set
of orbifold regular points, i.e., A
orb
reg
A
reg
. Likewise, the orbifold singular locus

orb
(A) is larger than the singular locus (A) as a variety, (A)
orb
(A). This
is discussed in Section 4.4 below.
One way to construct an orbifold is from a group acting properly discontin-
uously on a manifold `. The quotient then has the structure of an orbifold in a
natural way [Sat57]. To see this consider a point j `. The isotropy subgroup

p
is nite since the action is properly discontinuous. Moreover, there is a
neighborhood

l
p
of each point j such that

l
p


l
p
= if
p
and

l
p
=

l
p
if
p
. Then the natural projection :

l
p

l
p

p
= l
p
is a local homeo-
morphism satisfying = for
p
. Construct a cover l
p
for ` and
local uniformizing systems of the form

l. . by adding when necessary nite

intersections of the form l
p1
l
p
k
to guarantee that condition (ii) of Denition
4.1.1 holds. The quotient `, which is always Hausdor, is then an orbifold.
Denition 4.1.4: An orbifold A = (A. |) is said to be developable
3
if A = `
for some discrete group acting properly discontinuously on `. and the orbifold
atlas | is constructed as above.
Although not every orbifold is developable, it will be proven later in Proposition
4.2.17 that every orbifold can be realized as the quotient of a manifold by a compact
Lie group
4
G. In this case G acts locally freely so that A is the space of leaves of
a foliation T. and the isotropy subgroup
x
is precisely the leaf holonomy group of
the leaf r. (See Example 2.1.3 below.) This identies the local uniformizing groups
with the leaf holonomy groups of T. An orbifold A is a smooth manifold or in the
complex analytic category a complex manifold if and only if
x
= id for all r A.
In this case we can take = id and = id. and the denition of an orbifold reduces
to the usual denition of a smooth manifold.
Denition 4.1.5: The order (A) of an orbifold A = (A. |) is the least common
multiple of the orders of the isotropy groups when it exists. If there is no such least
common multiple, we say A has innite order.
Example 4.1.6: Orbifold structures on o
2
. This example is taken from [BH99].
Consider o
2
= C as the Riemann sphere with two marked points, the north
pole and south pole 0, respectively. We construct an orbifold structure on o
2
using two uniformizing charts
0
: C\
0
= o
2
` and

: C\

=
o
2
` 0 dened by
0
(.) = .
m
and

(n) = n
n
. The injection satises
n = (.) = (1.)
m
n
so that
0
(.) =

(.) holds. The uniformizing groups

xing and 0. respectively, are cyclic groups C
m
. C
n
of order : and n. respectively.
3
The name good was used by Thurston in [Thu79], but we do not like this terminology,
since from our viewpoint it is precisely Thurstons bad orbifolds that are good in the sense that
they are important and useful to us. On the other hand Thurstons good orbifolds are of less
importance to us. We prefer the appellation developable that was introduced by Bridson and
Haeiger in [BH99].
4
It is not known whether this holds for noneective orbifolds. However, see [HM04] for
recent progress on this issue.
108 4. FUNDAMENTALS OF ORBIFOLDS
This orbifold is developable if and only if : = n in which case it is the quotient
space o
2
C
m
. Note that in all cases the order of this orbifold is lcm(:. n).
For orbifold structures on Riemann surfaces of genus greater than 1 we refer
to the constructions in Chapter III( of [BH99]. In this very interesting chapter
Bridson and Haeiger discuss orbifolds from the point of view of etale groupoids
(cf. Proposition 4.3.1 below). In particular, they generalize the following result
attributed to Gromov in [BH99].
Proposition 4.1.7: A complete Riemannian orbifold
5
of non-positive sectional
curvature is developable.
The notions of structure preserving maps and bundle theory, or more gener-
ally sheaf theory, are fundamental to doing geometry on any object. The stan-
dard notions of smooth maps between orbifolds as well as V-bundles on orb-
ifolds (orbibundles) have been given in an analogous manner to manifolds (see
[Sat56, Sat57, Bai56, Bai57]). However, it wasnt realized until recently (see
for example [CR04, MP97]) that certain problems arise, namely it does not in
general induce morphisms of sheaves or V-bundles. This led to the notion of good
map in [CR04]; however, we follow the development in [MP97] and [Moe02].
Denition 4.1.8: Let A = (A. |) and \ = (Y. 1) be orbifolds. A map 1 : AY
is said to be smooth if for every point r A there are orbifold charts (

l. . )
about r and (

\ . . ) about 1(r) such that 1 maps l = (

l) into \ = (

\ )
and there exist local lifts

1

U
:

l

\ satisfying

1

U
= 1 . Furthermore,
1 is said to be good if it is compatible with the injections in the sense that if

ij
: (

l
j
.
j
.
j
)(

l
i
.
i
.
i
) is an injection on A, then there is an injection
j(
ij
) : (

\
j
.
j
.
j
)(

\
i
.
i
.
i
) on \, such that
(i)

1

U
i

ij
= j(
ij
)

1

U
j
.
(ii) j(
ij

jk
) = j(
ij
) j(
jk
).
We shall often write 1 : A\ for a smooth map of orbifolds and refer to
this as an orbifold map. It is easy to see that the composition of smooth maps is
smooth, so the collection of orbifolds with smooth maps forms a category. Thus,
we also have
Denition 4.1.9: Two orbifolds A = (A. |) and \ = (Y. 1) are said to be dif-
feomorphic or equivalent if there exist smooth maps 1 : AY and o : Y A
such that 1 o = 1l
Y
and o 1 = 1l
X
. where 1l
X
. 1l
Y
are the respective identity maps.
Notice that in particular a dieomorphism between orbifolds gives a homeo-
morphism of the underlying topological spaces.
4.2. Orbisheaves and orbibundles
We now wish to describe sheaves and bundles over orbifolds. Following [MP97]
we have
Denition 4.2.1: Let A = (A. |) be an orbifold. A sheaf T on A or orbisheaf
on A consists of a sheaf T

U
over

l for each orbifold chart (

l. . ) such that
for each injection : (

l. . )(

\ . . ) there is an isomorphism of sheaves

T() : T

V
which is functorial in .
5
For the denition of a Riemannian orbifold, see Denition 4.2.11 below.
4.2. ORBISHEAVES AND ORBIBUNDLES 109
Let T be an orbisheaf on A. and | = (

l. . ) an orbifold chart. By 3)
of Remarks 4.1.1 each denes an injection

, so there is a sheaf map

T() : T

U
. and if : T

U,x
is a point in the stalk over r

l. then
T()(:) lies in the stalk over
1
r. This denes an action of on the sheaf T

U
,
which says that T

U
is a -equivariant sheaf on

l. So every orbisheaf T is equivariant
under the local uniformizing groups
i
. We now have
Denition 4.2.2: The structure orbisheaf O

of an orbifold A is the orbisheaf

dened by the structure sheaf O

U
on each orbifold chart (

l. . ).
O

is well dened since each injection : (

l. . )(

l
t
.
t
.
t
) induces an
isomorphism O

by sending 1 O

U,x
to 1
1
(

)
x
. The
structure sheaves that are of most interest to us are the smooth structure sheaf
when O

U
is the sheaf of C

functions on

l for real orbifolds, the holomorphic
structure sheaf when O

U
is the sheaf of holomorphic functions on

l for complex
orbifolds, and the regular structure when O

U
is the sheaf of regular functions on

l.
Denition 4.2.3: A morphism of orbisheaves : TT
t
is a family of sheaf
maps

U
: T

U
T
t

U
. one for each orbifold chart, that is compatible with the injec-
tions in the sense that for each injection the diagram
T

U
T
t

J()

()

T
t

V
commutes.
So the collection of all sheaves of sets, Abelian groups, rings, etc. forms a cate-
gory
6

A that we call the category of orbisheaves on A. It is important to realize that

A is quite dierent from the category of sheaves on A unless the orbifold structure
is trivial, i.e. all the local uniformizing groups
i
are the identity. Nevertheless,
one can identify sheaves with orbisheaves.
Lemma 4.2.4: Let A = (A. |) be an orbifold. Then a sheaf T
X
on A induces
an orbisheaf, denoted by T

. on A by pulling back the restrictions T

U
i
to the local
uniformizing neighborhoods

l
i
. Conversely, an orbisheaf T

on A induces a sheaf
T
X
on A by locally taking the direct image sheaves of the subsheaf of -invariant
sections on the local uniformizing neighborhoods.
Proof. This follows directly from the denitions. Given an orbifold atlas
| =

l
i
.
i
.
i
and an injection
ji
:

l
i
.
i
.
i

l
j
.
j
.
j
. there are sheaves
T

U
i
=

i
T
Ui
and T

U
j
=

j
T
Uj
on A. We have l
i
l
j
. so

ji
T

Uj
=

ji

j
T
Uj
(
j

ji
)

(T
Uj
[
Ui
) =

i
T
Ui
= T

Ui
.
Similarly, it is easy to check that this is functorial in .
To describe the converse consider an orbisheaf T on A. This gives a sheaf T

U
on each local uniformizing neighborhood (

l. . ). Then we get a sheaf T on A

6
Actually this is a special type of topos called a smooth etendue by Grothendieck and Verdier
[AGV72], but we shall not use this terminology.
110 4. FUNDAMENTALS OF ORBIFOLDS
by taking the direct image of the -invariant sections, that is, T is the sheaf on A
associated to the presheaf l H
0
(
1
(l). T

U
)

.
Any map

1 :

A

\ of orbisheaves sends O

to O
`
. and thus induces a mor-
phism 1 : A\. However, the converse is not true. We need a good orbifold map
as dened by Denition 4.1.8 to induce a morphism of orbisheaves. Nevertheless,
the following does hold [MP97]:
Proposition 4.2.5: A dieomorphism of orbifolds A \ induces an equivalence
of orbisheaves

A

\.
Proof. See [MP97] for details.
The structure sheaf O

of an orbifold A is a sheaf of commutative rings on each

local uniformizing neighborhood

l
i
. and we can construct an orbisheaf of O

modules on A by giving a sheaf

U
of O

U
modules on each orbifold chart (

l. . ).
As usual for a sheaf T

U
on

l we denote by (T

U
)
r
the direct sumT

U

r times
T

U
.
Denition 4.2.6: An orbisheaf of O

modules on A is said to be locally

V-free or just locally free if each point r A has an orbifold chart such that

U
(O

U
)
r
for some positive integer :. called the rank of . An orbisheaf of
rank one is called invertible orbisheaf or a V-invertible sheaf.
We use the terms locally free and locally V-free interchangeably when referring
to orbisheaf. The terms V-free and V-invertible are used mainly for emphasis since
a sheaf induces an orbisheaf which is generally a dierent structure.
Many of the usual dierential geometric concepts that hold for smooth or com-
plex analytic manifolds also hold in the orbifold category, in particular the impor-
tant notion of a ber bundle.
Denition 4.2.7: A V-bundle or orbibundle over an orbifold A = (A. |) con-
sists of a ber bundle 1

U
over

l for each orbifold chart (

l
i
.
i
.
i
) | with Lie
group G and ber 1 a smooth G-manifold which is independent of

l
i
together with
a homomorphism /

Ui
:
i
G satisfying:
(i) If / lies in the ber over r
i


l
i
then for each
i
. //

U
i
() lies in the
ber over
1
r
i
.
(ii) If the map
ji
:

l
i

l
j
is an injection, then there is a bundle map

ji
: 1

Uj
[
ji(

Ui)
1

Ui
satisfying the condition that if
i
. and

t

j
is the unique element such that
ji
=
t

ji
. then /

U
i
()

ji
=

ji
/

Uj
(
t
). and if
kj
:

l
j

l
k
is another such injection then
(
kj

ji
)

ji

kj
.
If the ber 1 is a vector space of dimension : and G acts on 1 as linear transfor-
mations of 1. then the V-bundle is called a vector V-bundle of rank :. Similarly,
if 1 is the Lie group G with its right action, then the V-bundle is called a principal
V-bundle.
Remark 4.2.1: Both the denitions of orbisheaves and orbibundles (V-bundles)
consist of a sequence sheaves or bundles dened on the disjoint union

l = .
i

l
i
of
the local uniformizing neighborhoods that satisfy certain compatibility conditions
with respect to the action of the local uniformizing groups and the injections.
The space

l is the space of objects of a certain etale groupoid, and this plays
4.2. ORBISHEAVES AND ORBIBUNDLES 111
an important role when describing invariants of an orbifold. We discuss this in
detail in the next section.
We now describe the total space 1 of a V-bundle over A. and see that it inherits
a (generally) non-eective orbifold structure. We construct an orbifold structure
on 1 with local uniformizing systems 1

U
i
.

i
.

i
by gluing together the sets
1

Uj

i
. By choosing the local uniformizing neighborhoods of A small enough, we
can always take 1

Ui
to be the product

l
i
1 which we shall heretofore assume.
The action of the local uniformizing group
i
on

l
i
extends to an action on

l
i
1
by sending ( r
i
. /)

l
i
1 to (
1
r
i
. //

U
i
()). so the local uniformizing groups

i
are dened to be subgroups of
i
that stabilize the point (r
i
. /). We are particularly
interested in the case of a principal bundle in which case the bre is a Lie group G.
so the image /

U
i
(

i
) acts freely on 1. Then we have
Lemma 4.2.8: Let G be a Lie group and 1 the total space of a principal V-bundle
over an orbifold A = (A. |). Then 1 is a smooth manifold if and only if the maps
/

U
i
are injective for all i.
Remark 4.2.2: We shall often denote a V-bundle by the standard notation :
1A and think of this as an orbifold bration. It must be understood, however,
that an orbifold bration is not a bration in the usual sense. Shortly, we shall
show that it is a bration rationally. As in the case of orbisheaves a smooth map
between orbifolds does not in general induce a bundle map except in the case
of dieomorphisms as in Proposition 4.2.5. For more general maps we need to
assume that the map is good as in Denition 4.1.8. However, isomorphism classes
of V-bundles can be described in the usual manner. We leave this description to
the reader. An absolute V-bundle resembles a bundle in the ordinary sense, and
corresponds to being able to take /

U
= id. for all local uniformizing neighborhoods

l. In particular, the trivial V-bundle A1 is absolute. Another important notion

introduced by Kawasaki [Kaw78] is that of proper. A V-bundle 1 is said to be
proper if the local uniformizing groups

i
of 1 act eectively on A when viewed
as subgroups of the local uniformizing groups
i
on A. Any V-bundle with smooth
total space is clearly proper. The Kawasaki index theorems such as his Riemann-
Roch Theorem used in section 3.2 require the V-bundles to be proper.
The notion of sections of bundles works just as well in the orbifold category.
Denition 4.2.9: Let c = (1. |

) be a V-bundle over an orbifold A. Then a

section of c over an open set \ A is a section

U
of the bundle 1

U
for each
orbifold chart (

l. . ) such that l \ and for any r

l we have
(i) For each .

U
(
1
r) = /

U
()

U
( r).
(ii) If : (

l. . )(

l
t
.
t
.
t
) is an injection, then

(( r)) =

U
( r).
If each of the local sections

U
is continuous, smooth, holomorphic, etc., we say
that is continuous, smooth, holomorphic, etc., respectively. Given local sections

U
of a vector V-bundle or orbisheaf we can always construct -invariant local
sections by averaging over the group, i.e., we dene an invariant local section by
(4.2.1)
I

U
=
1
[[

U
.
Notice that this determines a well dened map from the underlying space, namely

U
: l = (

l)1

U
. Using invariant local sections over each orbifold chart we
112 4. FUNDAMENTALS OF ORBIFOLDS
obtain global invariant sections. We often work with these invariant sections, and
view objects on an orbifold interchangeably as invariant objects on

l or objects on
l. Conversely, given a section of a sheaf T
X
over the regular subspace A
reg
we
can extend it to a
i
invariant section over each local uniformizing neighborhood

l
i
. This gives a section of the corresponding orbisheaf T

.
The standard notions of tangent bundle, cotangent bundle, and all the associ-
ated tensor bundles all have V-bundle analogues [Bai56, Sat56, Sat57].
Example 4.2.10: [Tangent V-bundle of A] Given a uniformizing chart (

l
i
.
i
.
i
)
in |. we take 1

Ui
= T

l
i
the tangent bundle of the open submanifold

l
i

R
n
. Let (n
1
i
. . . . . n
n
i
) denote the local coordinates of

l
i
. The bre 1 is R
n
with
its coordinate basis n
k
i
at each point j

l
i
. and for each injection
ji
:
(

l
i
.
i
.
i
)(

l
j
.
j
.
j
) we have the Jacobian matrix
J
ji
=

y
k
j

ji
y
l
i

which satises J
ji
( j) G1(n. R) for each j

l
i
. and its inverse is the transition
function for the bundle 1

U
i
. So G = G1(n. R). Since, as mentioned previously,
each local uniformizing group
i
denes an injection, the homomorphisms /

U
i
are
injective, and satisfy condition i) of Denition 4.2.7. Furthermore, the injections
dene a bundle map satisfying condition ii) of the denition. This denes the
tangent V-bundle TA = (TA. |

). where TA is the total space of the V-bundle

and the orbifold atlas |

has orbifold charts of the form (

l
i
R
n
.

i
.

i
). where
the local uniformizing group

i
is
i
acting on

l
i
R
n
by ( r. ) (
1
r. /

U
i
()).
and

i
:

l
i
R
n
(

l
i
R
n
)

i
is the natural quotient projection. Thus, the
tangent V-bundle of an orbifold is always proper. A smooth invariant section of
TA is called a vector eld on A.
The cotangent V-bundle and the tensor bundles are constructed similarly.
Thus, one easily constructs Riemannian metrics, symplectic 2-forms, connections,
etc. such that the injections are maps of the corresponding G-structure. In partic-
ular, the objects are invariant under the local uniformizing groups. For example,
Denition 4.2.11: A Riemannian metric o on an orbifold A = (A. |) is a Rie-
mannian metric o
i
on each local uniformizing neighborhood

l
i
that is invariant un-
der the local uniformizing group
i
. and the injections
ji
: (

l
i
.
i
.
i
)(

l
j
.
j
.
j
)
are isometries, i.e.,

ji
(o
j
[

ji
(U
i
)
) = o
i
. Similarly, if A = (A. |) is a complex orb-
ifold, then an Hermitian metric / is a
i
-invariant Hermitian metric on each
neighborhood

l
i
such that the injection maps are Hermitian isometries. An orb-
ifold with a Riemannian (Hermitian) metric is called a Riemannian (Hermitian)
orbifold.
Then a slight modication of usual partition of unity argument gives (cf.
[MM03] for details of the proof):
Proposition 4.2.12: Every orbifold admits a Riemannian metric, and every com-
plex orbifold admits an Hermitian metric.
Remark 4.2.3: If an orbifold structure is given on a smooth manifold `. then an
orbifold Riemannian metric is not necessarily a Riemannian metric on `.
4.2. ORBISHEAVES AND ORBIBUNDLES 113
Integration theory also goes through. In particular, if \ is an open subset of
(

l) then the integral of an n-form (measurable) is dened by

(4.2.2)

V
=
1
[[

1
(V )

U
.
So all of the standard integration techniques, such as Stokes theorem, hold on
orbifolds.
Example 4.2.13: [V-bundle or orbibundle 1(A) of linear frames of A] Con-
tinuing with the notation of Example 4.2.10, we now put 1 = G1(n. R). The
local bundle 1

U
is now just the frame bundle on the local uniformizing neighbor-
hood

l. The linear frame V-bundle 1(A) = (1(A). |

) has orbifold charts of the

form (

l
i
G1(n. R).

i
.

i
). where the local uniformizing group

i
is
i
acting
on

l
i
G1(n. R) by ( r. ) (
1
r. /

Ui
()). and

i
:

l
i
G1(n. R)(

l
i

G1(n. R))

i
is the natural quotient projection. In this case

i
acts freely on the
second factor, so the action of the local uniformizing groups

i
of the linear frame
V-bundle is always free. Thus, the total space 1(A) of the linear frame V-bundle
1(A) of any orbifold A is a smooth manifold. The right action of G1(n. R) on
1(A) is only locally free with the uniformizing groups
i
as the isotropy group.
We wish to describe reduction of principal orbibundles, but rst we need the
notion of an orbibundle map.
Denition 4.2.14: A V-bundle (orbibundle) map : 11
t
consists of a
family of bundle maps

U
: 1

U
1
t

U
. one for each orbifold chart, that is compatible
with the injections : (

l. . )(

\ . . ) and
t
: (

l
t
.
t
.
t
)(

\
t
.
t
.
t
) in
the sense that the diagram
1

V
1
t

U
1
t

commutes.
By rening the orbifold charts if necessary we can take

l such that 1

U


l1.
Then by identifying 1

U
with

l 1 and letting
1

U
denote the map

U
on the rst
factor, we see that Denition 4.2.14 implies
1

V
=
t

1

U
. To dene a reduction
of a principal orbibundle to a subgroup, we require that the diagram (1.2.3) on
the uniformizing neighborhoods patch together to give an orbibundle map. Let us
describe this on the level of the principal frame orbibundle, as the generalization
to arbitrary principal orbibundles is straightforward. Let : GG1(n. R) be a
subgroup and consider the corresponding subbundle Q

U
of the linear frame bundle
1

U
on the uniformizing neighborhood

l of an orbifold A. Then diagram (1.2.3)
becomes
(4.2.3)
Q

U
1

l
id

l .
114 4. FUNDAMENTALS OF ORBIFOLDS
where

U
satises

U
(no) =

U
(o). Combining this diagram with the one of
Denition 4.2.14 gives the commutative diagram:
(4.2.4) Q

U
1

l

l
Q

V
1

\

\

id

id

This then together with Example 4.2.13 allows us to dene the notion of a G-
structure on an orbifold.
Denition 4.2.15: Let A be an orbifold and G be a subgroup of G1(n. R). Then a
G-structure on an orbifold A is a reduction of the linear frame orbibundle 1(A)
to the subgroup G such that the natural homomorphism is an orbibundle map.
The notion of integrable G-structure works equally well on orbifolds A. It simply
means that the G-structure on each local uniformizing neighborhood

l is integrable.
Exercise 4.1: Formulate the reduction process for general principal orbibundles
and state and prove the orbifold analog of Theorem 1.2.5.
Example 4.2.16: [The orthonormal frame orbibundle] The discussion in Ex-
ample 1.4.7 carries over to the orbifold case with the obvious changes. As in that
case a reduction of the linear frame orbibundle to the orthogonal group is equiva-
lent to a choice of an orbifold Riemannian metric. That this can always be done
follows from a standard partition of unity argument. Since the local uniformizing
groups
i
can be taken to be subgroups of the orthogonal group O(n). the total
space 1O(A) of the orthonormal frame V-bundle is also a smooth manifold.
This example immediately implies the following result of Satake.
Proposition 4.2.17: Every orbifold A = (A. |) can be presented as a quotient
space of a locally free action of a compact Lie group G on a manifold `.
This proposition implies that every orbifold can be realized as the space of leaves
of a foliation. Of course, generally the converse is not true; Molinos Theorem 2.5.11
gives conditions under which a converse holds.
The usual relation between locally free sheaves and vector bundles also holds
for orbifolds. We consider both real and complex vector V-bundles and refer to
them as F-vector V-bundles, where F = R or C.
Proposition 4.2.18: There is a one-to-one correspondence between isomorphism
classes of F-vector V-bundles on A and isomorphism classes of locally V-free or-
bisheaves on A. In particular, F line V-bundles correspond to invertible orbisheaves.
4.3. GROUPOIDS, ORBIFOLD INVARIANTS AND CLASSIFYING SPACES 115
Proof. Given a rank : vector V-bundle c over an orbifold A we construct
the sheaf T

U
of local section of the local bundle 1

U
. By the usual result T

U
is a
locally free sheaf of rank :. Then the bundle map

associated with each injection

: (

l. . )(

\ . . ) gives an isomorphismT() : T

V
that is functorial
in . This gives a locally V-free orbisheaf T.
Conversely, if T is a locally V-free orbisheaf then on each local uniformizing
neighborhood

l, the locally free sheaf T

U
corresponds to the sheaf of sections of
some vector bundle 1

U
on

l. The transition functions o
ij
:

l
i

l
j
G1(:. R) are
obtained from the isomorphisms o
i
: T

Ui
O
r

Ui
by setting o
ij
= o
i
o
1
j
.
4.3. Groupoids, Orbifold Invariants and Classifying Spaces
In this section we describe certain orbifold invariants which are important to us
since they give rise to certain invariants of deformation classes of Sasakian struc-
tures. For this description it is convenient to use the language of groupoids. It
turns out that orbifolds can be conveniently represented by etale groupoids. The
relation between orbifolds and groupoids was rst elucidated by Haeiger in his
seminal paper [Hae84], and more recently developed by Moerdijk and collabora-
tors, [MP97, Moe02, MM03]. Again we refer the reader to the Appendix A.1
for basic facts about groupoids. Haeigers point of departure is his groupoid of
germs discussed in Proposition 1.5.4.
Consider an orbifold A = (A. |) with a given orbifold atlas |. Let

l = .
i

l
i
denote the disjoint union of the

l
i
. and consider the pseudogroup {

of local
dieomorphisms of

l generated by the injections
ji
and their inverses. The action
of {

on

l denes an equivalence relation on

l by saying that r n if they lie on
the same orbit, that is, if there exists an 1 {

such that n = 1(r). The quotient

space

l{

is identied with the underlying space A of the orbifold. To see this

we dene the projection map :

lA to be the union of the maps
i
. and
notice that r
i


l
i
. and r
j


l
j
satises (r
i
) = (r
j
) if and only if there exists
an injection
ji
: (

l
i
.
i
.
i
)(

l
j
.
j
.
j
) such that
ji
= . So we let ((|)
denote the groupoid of germs of dieomorphisms associated to {

as in Proposition
1.5.4. ((|) is a proper eective etale Lie groupoid with space of objects (
0
=

l
and space of arrows ( the set of germs of the injections. Thus, we have identied
A with the orbit space (
0
(. In this sense we say that the orbifold A = (A. |) is
represented by the groupoid ((|). In fact we have [MM03]
Proposition 4.3.1: Let A = (A. |) be an orbifold with a xed atlas |. then ((|)
is a proper eective etale groupoid. Furthermore, if A
t
= (A
t
. |
t
) is another such
orbifold, then ((|) and ((|
t
) are Morita equivalent if and only if the orbifolds A
and A
t
are isomorphic.
Conversely, given a proper eective etale Lie groupoid (, the isotropy subgroup
(
x
x
at any point r (
0
is nite, and r has an open neighborhood l
x
with an eective
action of (
x
x
which induces an isomorphism of etale groupoids [MM03]
([
U
l
x
(
x
x
.
where the right hand side is the action groupoid of Denition A.1.5. Now we can
construct a canonical orbifold structure on the space of orbits (
0
( as follows: we
can choose the neighborhood l
x
of r so that it is dieomorphic to an open subset

l
x
of R
n
via a dieomorphism
x
: l
x

l
x
. Let : (
0
(
0
( be the natural
116 4. FUNDAMENTALS OF ORBIFOLDS
projection. Then we obtain an atlas | of orbifold charts of the form (

l
x
.
x
(
x
x

1
x
.
1
x
). Now suppose that \
y
is contained in l
x
with
y
the natural inclusion.
Then the injection

xy
: (

\
y
.
y
(
y
y

1
y
.
1
y
)(

l
x
.
x
(
x
x

1
x
.
1
x
)
is given by
xy
=
x

y

1
y
. Thus, ((
0
(. |) is an orbifold. We have arrived at
Proposition 4.3.2: Let ( be a proper eective etale groupoid. Then (
0
( has a
canonical orbifold structure.
Thus, an orbifold structure on a paracompact Hausdor space is completely
equivalent to a proper etale Lie groupoid. This gives credence to the name orbifold
groupoid of Denition A.1.6. The groupoid ( also denes a Haeiger cocyle by
Denition 2.2.2 which in this case is just the cocycle dened by the cover | together
with the injection maps from double intersections viewed as maps from

l
i

l
j
to
(. This cocycle denes an element
ji
H
1
(

l. (). These cocycles are related to

principal ( bundles. We refer to Section 2.2 and [Hae84, GH06] for more detail.
Actually it is only the Morita equivalence class of a groupoid that is pertinent
to our discussion of orbifold invariants. The following theorem due to Moerdijk and
Pronk [MP97, Moe02, MM03] plays an important role relating the holonomy
groupoid of a foliation to orbifolds.
Theorem 4.3.3: Let ( be a Lie groupoid. Then the following are equivalent:
(i) ( is Morita equivalent to an eective proper etale groupoid, i.e., an orb-
ifold groupoid.
(ii) ( is Morita equivalent to the holonomy groupoid of a foliation with com-
pact leaves and nite holonomy groups.
(iii) ( is Morita equivalent to an action groupoid `1. where 1 is a compact
Lie group acting on ` such that each isotropy group 1
x
is nite and acts
eectively on the normal bundle.
The reader is invited to compare Theorem 4.3.3 dealing with categorical equiv-
alence of groupoids with Molinos Theorem 2.5.11 about the space of leaves of a
foliation while keeping in mind the relation between proper etale groupoids and
orbifolds. From the sheaf theoretic viewpoint we have, along similar lines, the
following theorem of Moerdijk and Pronk [MP97]:
Theorem 4.3.4: The following are equivalent as categories (ringed topoi):
(i) The topos of orbisheaves on orbifolds.
(ii) The topos of equivariant sheaves on a G-manifold A with nite isotropy
groups and faithful slice representations.
(iii) The topos of sheaves on some eective proper etale groupoid.
We are now ready to describe some important orbifold invariants. As discussed
in Appendix A.2, Haeiger constructed a classifying space 1( for any topological
groupoid (. On any orbifold we can choose a Riemannian metric by Proposition
4.2.12. In this case the space of arrows of the etale Lie groupoid ((|) is generated
by the pseudogroup of local isometries. Haeiger [Hae84] gives an explicit Borel-
type construction of the classifying space of an orbifold. Let us look at the linear
frame bundle of Example 4.2.13 of an orbifold A from the point of view of the etale
groupoid ((|) representing it. We dene the orthonormal frame bundle of the
space of units

l of ((|) by 1O(

l) = .
i
1O(

l
i
). It is a principal O(n)-bundle over
4.3. GROUPOIDS, ORBIFOLD INVARIANTS AND CLASSIFYING SPACES 117

l. The dierential of the groupoid action on

l induces an action of the groupoid
( on 1O(

l). and as in Example 4.2.13 this action is free. From the denitions one
easily sees that the quotient space 1O(

l)( is the orthonormal frame V-bundle

1O(A) of the orbifold which is a smooth manifold as discussed in Example 4.2.13.
Furthermore, 1O(A) depends only on A. that is, on the equivalence class of orbifold
atlases, and not on the particular atlas |. Thus, we may also consider 1O(

l) to
be a principal (-bundle over 1O(A).
Recall that the universal principal O(n)-bundle 1O(n)1O(n) whose total
space 1O(n) is a contractible space on which O(n) acts freely [Hus66, Mil56a].
So we can construct the associated O(n)-bundle 1O(

l)
O(n)
1O(n). It is an O(n)-
bundle over

l with contractible bres 1O(n). Since the action of the groupoid ( is
trivial on 1O(n) and commutes with the O(n) action on 1O(

l), the groupoid ( acts

on both the total space and base space of the bundle j : 1O(

l)
O(n)
1O(n)

l.
From the denitions and construction in Appendix 2.5.16, we see that this is just
the universal space 1(. i.e.,
(4.3.1) 1( = 1O(

l)
O(n)
1O(n).
Taking the quotient by ( one gets a V-bundle j : 1O(A)
O(n)
1O(n)A
so the classifying space 1( is 1O(A)
O(n)
1O(n). We thus have a commutative
diagram
(4.3.2)
1(

1(

l = (
0

(
0
( = A .
Note that since 1O(A) is independent of a particular atlas representing the orbifold
A. so is 1(. In fact from Theorem A.2.3, up to weak homotopy equivalence 1(
depends only on the Morita equivalence class of (. Thus, it makes sense to dene
Denition 4.3.5: We call 1( the classifying space of the orbifold A. and denote
it by 1A.
It follows also from Theorem 4.3.3 that the weak homotopy type of 1( is an
invariant of any foliation whose holonomy groupoid is Morita equivalent to the
orbifold groupoid (. Note also that the map j is not a bration, since the bres are
not all homotopy equivalent. A generic bre of j. that is a bre over a regular point
r A is 1O(n). a contractible space, whereas, a singular bre is an Eilenberg-
MacLane space 1(. 1). where is the local uniformizing group.
The above discussion allows us, following [Hae84, HS91] to make the impor-
tant
Denition 4.3.6: We dene the orbifold cohomology, homology, and homo-
topy groups by
H
i
orb
(A. Z) = H
i
(1A. Z) . H
orb
i
(A. Z) = H
i
(1A. Z) .
i
(A)
orb
=
i
(1A) .
We caution that the orbifold cohomology groups dened here are not the orb-
ifold cohomology groups dened by Chen and Ruan [CR04] which are not used in
this book. In [Thu79] Thurston introduced the orbifold fundamental group
orb
1
as a group of deck transformations. It was shown in [HD84] that the orbifold
118 4. FUNDAMENTALS OF ORBIFOLDS
fundamental group
orb
1
(A) dened in Denition 4.3.6 is equivalent to Thurstons
better known denition. See Theorem 4.3.19 below. We should mention if the orb-
ifold A is a smooth manifold these orbifold groups coincide with the usual groups.
It should be noted that generally these orbifold groups are not topological invari-
ants of A. but invariants of the orbifold structure. In order to better understand
the orbifold cohomology groups, we study the Leray spectral sequence of the map
j : 1AA. For any sheaf T on 1A we let 1
s
j(T) denote the derived functor
sheaves, that is the sheaves associated to the presheaves l H
s
(j
1
(l). T). Then
Lerays theorem [Bre97] says that there is a spectral sequence 1
r,s
r
with 1
2
term
given by
1
r,s
2
= H
r
(A. 1
s
j(T))
converging to H
r+s
(A. T).
Lemma 4.3.7: Let A = (A. |) be a compact orbifold and an Abelian group.
Then the derived functor sheaves 1
s
j() have their support contained in the sin-
gular set
orb
of A with stalks given by the group cohomology H
s
(
i
. ) at r
i

l
i

orb
.
Proof. Let (

l
i
.
i
.
i
) be a local uniformizing system for the open set l
i
=

i
(

l
i
) A. On

l
i
the action of the groupoid ( is the action of
i
. So j
1
(l
i
) is
identied with

l
i

i
1O(n) which is an Eilenberg-MacLane space 1(
i
. 1) = 1
i
.
Thus, H
q
(j
1
(l
i
). ) is the group cohomology H
q
(
i
. ). So for c 0 the stalks
vanish at regular points and is H
q
(
i
. ) at the singular point r
i
.
Since the group cohomology H
q
(
i
. ) vanishes if the order [
i
[ is invertible in
. the spectral sequence collapses and we have [Hae84]
Corollary 4.3.8: Let A = (A. |) be an orbifold of order (A) < . Then if
(A) is invertible in , the groups H
q
orb
(A. ) and H
q
(A. ) are isomorphic. In
particular, H
q
orb
(A. ) H
q
(A. ) if = Q. R. C. or Z
p
. where gcd(j. (A)) = 1.
There is a similar spectral sequence for the homology groups [CM00] such that
H
r
(A. 1
s
j(T)) = H
orb
r+s
(A. T).
so Corollary 4.3.8 holds for the homology groups as well. We should also warn the
reader that the groups H
p
orb
(A. T) or H
orb
p
(A. T) do not in general vanish above
the dimension of A.
We are especially interested in orbifolds whose local uniformizing groups are
nite subgroups of the circle group o
1
. so we give these a special name.
Denition 4.3.9: An orbifold A = (A. |) is said to be locally cyclic if it has an
orbifold atlas all of whose local uniformizing groups are cyclic groups.
In the case of cyclic groups the cohomology groups H
q
(
i
. ) are well under-
stood. Indeed, for Z
m
[Bro82]
H
q
(Z
m
. Z) =

Z for c = 0,
Z
m
for c 0 even,
0 for c odd.
We want to relate orbibundles on A to ordinary bundles on 1A. To do so
we need the analogue of an atlas of charts on 1A. This is done by passing to
the nite models of 1O(n) as discussed in Appendix A. So a smooth map on 1G
means smooth maps on the nite models 1G(/) which are smooth manifolds. Let
4.3. GROUPOIDS, ORBIFOLD INVARIANTS AND CLASSIFYING SPACES 119

l
i
.
i
.
i
be a cover of uniformizing charts for the orbifold A. and let

1
i
= 1O(

l
i
)
denote the restriction of 1O(

l) to

l
i
. The group O(n) acts locally freely on

1
i
with isotropy group
i
xing the orthonormal frames over the center o
i
l
i
. So
we have homeomorphisms

1
i
O(n)

l
i

i
l
i
. Thus, we can cover 1A by
neighborhoods of the form

l
i

i
1O(n). where
i
. Now by rening the cover if
necessary we have injection maps
ji
:

l
i

l
j
and these induce the change of
charts maps
G
ji
:

l
i

i
1O(n)

l
j

j
1O(n)
given by G
ji
([ r
i
. c]) = [
ji
( r
i
). c]. This is well-dened since the unique element
j
is
identied with
i
under the identication of
i
as a subgroup of
j
. This will allow
us to construct local data on

l
i

i
1O(n) by considering smooth (or holomorphic)
data on the

l
i
and continuous data on 1O(n) which is smooth on the nite models
1O(n)(/) and invariant under the
i
action. Since the
ji
are dieomorphisms (or
biholomorphisms) this local data will then patch to give global data on 1A. For
example we denote by c the sheaf of germs of complex-valued functions on 1A that
are smooth in

l
i
and continuous in 1O(n). We shall call global (local) sections of
the sheaf c smooth functions on 1A. Similarly we refer to smooth maps from
1A. For example, if 1 : A
1
= (A
1
. |
1
)A
2
= (A
2
. |
2
) is a good map of orbifolds,
then this induces a smooth map 1 : 1A
1
1A
2
. However, by Proposition 4.2.5
we have an equivalence of orbisheaves, and hence an equivalence of orbibundles.
Thus,
Proposition 4.3.10: An orbifold dieomorphism 1 : A
1
A
2
induces a dieo-
morphism 1 : 1A
1
1A
2
.
Similarly, because of the commutative diagram

l
i
1O(n)

l
i

l
i

i
1O(n)
pi
l
i
.
the map j : 1AA is smooth, since its local covering maps j
i
are smooth. Now
we are ready for our description of V-bundles in terms of ordinary bundles.
Theorem 4.3.11: Let A = (A. |) be an orbifold. There is a one-to-one correspon-
dence between isomorphism classes of orbibundles on A with group G and generic
ber 1 and isomorphism classes of bundles on 1A with group G and ber 1.
Proof. A V-bundle on A is a bundle on

l
i
for each local uniformizing neigh-
borhood

l
i
together with a group homomorphism /

Ui
Ho:(
i
. G) that satisfy
the compatibility conditions of Denition 4.2.7. This gives an action of
i
on 1

Ui
.
Now cover 1A by neighborhoods of the form

l
i

i
1O(n). where we make use of
the fact that the local uniformizing groups
i
can be taken as subgroups of O(n).
There is an action of
i
on

l
i
11O(n) given by ( r
i
. n. c) (
1
r
i
. n/
U
i
(). c).
and this gives a G-bundle over

l
i

i
1O(n) with ber 1 for each i. Moreover, the
compatibility condition (ii) of Denition 4.2.7 guarantees that these bundles patch
together to give a G-bundle on 1A with ber 1.
Conversely, given a G-bundle on 1A with ber 1. restricting to

l
i

i
1O(n)
gives a G-bundle there. Since for each i

l
i

i
1O(n) is the Eilenberg-MacLane
space 1(
i
. 1). there is a one-to-one correspondence between isomorphism classes
120 4. FUNDAMENTALS OF ORBIFOLDS
of G-bundles on

l
i

i
1O(n) and conjugacy classes of group homomorphisms
Hom(
i
. G). The fact that these G-bundles come from a global G-bundle on 1A
implies that the compatibility conditions (ii) of Denition 4.2.7 are satised. The
correspondence can be seen to be bijective.
It is well-known that isomorphism classes of principal G-bundles on a CW
complex A are in one-to-one correspondence with the homotopy classes [A. 1G] of
maps from A to 1G. Thus, we have
Corollary 4.3.12: Let G be a Lie group. There is a one-to-one correspondence be-
tween principal orbibundles on the orbifold A = (A. |) with group G and homotopy
classes of maps [1A. 1G].
Another corollary of Theorem 4.3.11 whose proof is straightforward is:
Corollary 4.3.13: The isomorphism classes of orbibundles on A with group G
are in one-to-one correspondence with elements of the non-Abelian cohomology set
H
1
(1A. G) = H
1
orb
(A. G). where G is the sheaf of germs of maps to the group G.
Recall the cohomology ring H

(1G) for the classical groups.

H

(1O(n). Z
2
) Z
2
[n
1
. n
2
. . . . . n
n
] .
H

(1l(n). Z) Z[c
1
. c
2
. . . . . c
n
] .
H

(1oj(n). Z) Z[j
1
. j
2
. . . . . j
n
] . (4.3.3)
where n
i
are the Stiefel-Whitney classes, c
i
are the Chern classes, and j
i
the
Pontryagin classes. By Theorem 4.3.11 a principal G V-bundle 1 on A can be
represented uniquely up to homotopy by 1

1G. where : 1A1G

is a classifying map. So for any ring the V-bundle 1 denes a map

:
H

(1G. )H

orb
(A. ). Elements in the image of this map are the orbifold
characteristic classes of the orbibundles on A. For example the Chern classes of a
principal l(n)-orbibundle, or the Pontryagin classes of a principal oj(n) V-bundle
are integral classes in H

orb
(A. Z).
Now Chern-Weil theory proceeds on orbifolds as on manifolds. As in the mani-
fold case every principal V-bundle 1 admits a connection, and this can be described
by an invariant connection 1-form on 1 with values in the Lie algebra g of G. that
is a smooth section of
1
1 g that is invariant under the local uniformizing groups.
Let 1(G) denote the algebra of polynomials on g that are invariant under the adjoint
action of G on g. The algebra 1(G) is graded by degree,
1(G) =

k
1
k
(G).
and to each polynomial 1 1
k
(G) one denes a 2/-form on the orbifold 1 by
1()(A
1
. . . . . A
2k
) =
1
2/!

(A
(1)
. A
(2)
). . . . . ((A
(2k1)
.
(2k)
)

.
where is the curvature 2-form of . A
1
. . . . . A
2k
are smooth vector elds on 1.
the sum is taken over all permutations , and c

denotes the sign of . This 2/-form

is basic and projects to a closed 2-form on the underlying space A of the orbifold.
Moreover, its de Rham cohomology class is independent of the choice of connection
. This gives the Weil homomorphism
n : 1(G)H

dRh
(A. R) H

(A. R) .
4.3. GROUPOIDS, ORBIFOLD INVARIANTS AND CLASSIFYING SPACES 121
When A is a smooth manifold then n(1(G)) lies in the image of the inclusion
H

(A. Z)H

(A. R). This is not true for orbifolds. However, we shall see that
n(1(G)) does lie in the image of H

(A. Q)H

(A. R). Moreover, there is a uni-

versal homomorphism 1(G)H

(1G. R) which is an isomorphism when G is com-

pact. So in this case we get a homomorphism

: H

(1G. R)H

(A. R). El-

ements of H

(A. R) that are in the image of are known as real characteristic

classes. We now relate this construction of real characteristic classes in H

(A. R) to
the integral characteristic classes described above. First we see that Corollary 4.3.8
implies that j : 1AA gives a ring isomorphism j

: H

(A. Q)H

orb
(A. Q).
So the integral classes in H

orb
(A. Z) H

orb
(A. Q) map to rational classes in
H

(A. Q) under (j

)
1
. Furthermore, since both constructions are universal the
characteristic classes constructed by the Chern-Weil theory pullback under j

to
the integral characteristic classes in H

orb
(A. Z). Of course, we have worked with
the principal bundles here, but everything works for the corresponding associated
vector bundles as well. We denote the orbifold rational Chern (Pontrjagin) classes
in H

(A. Q) of a complex (real) V-bundle 1 on an orbifold A by c

orb
i
(1)(j
orb
i
(1)).
respectively. Summarizing we have
Theorem 4.3.14: Let G = l(n) or oj(n) and A an orbifold. Then the orbifold
Chern classes c
orb
i
(1) and orbifold Pontrjagin classes j
orb
i
(1) are dened for any
V-bundle 1 on A with group G. respectively. Moreover, j

c
orb
i
(1) and j

j
orb
i
(1)
are integral classes in H

orb
(A. Z).
Similarly for G = oO(2n) we can dene the orbifold Euler class c
orb
H
2n
(A. Q)
such that j

c
orb
H
2n
(A. Z). and c
orb
vanishes for the odd special orthogonal
groups. Satake [Sat57] denes the orbifold Euler characteristic
orb
(A) as
(4.3.4)
orb
(A) =

Ci
(1)
dim(C
i
)
1
[(C
i
)[
.
where the sum is taken over a cell decomposition of A that is compatible with the
stratication 4.1.1 in the sense that every cell lies in a single stratum. This can be
rewritten in the more invariant form as
(4.3.5)
orb
(A) =

S
(1)
dim(S)
(o)
[(o)[
.
where the sum is taken overall strata o of the stratication (4.1.1), and (o) is the
ordinary Euler class of the stratum o. It is emphasized that generally
orb
(A) is a
rational number. Satake [Sat57] proved the orbifold version of the Gauss-Bonnet
Theorem. The Euler class c
orb
H
2n
(A. Q) can be represented by the top invariant
curvature form of the Riemannian curvature, so the Gauss-Bonnet formula takes
the form
(4.3.6)
orb
(A) = 'c
orb
. [A]` =

X
.
where [A] H
2n
(A. Q) denotes the fundamental class of the compact oriented
2n-dimensional orbifold A.
We are particularly interested in the case of circle V-bundles on an orbifold. In
this case the group G is the circle group l(1). and the relevant characteristic class
is c
orb
1
H
2
(A. Q). or equivalently its image j

c
orb
1
H
2
orb
(A. Z). We have
122 4. FUNDAMENTALS OF ORBIFOLDS
Theorem 4.3.15: The isomorphism classes of circle V-bundles 1 over an orbifold
A = (A. |) are in one-to-one correspondence with elements of H
2
orb
(A. Z). and the
bijection is given by the orbifold rst Chern class j

c
orb
1
. Furthermore, the total
space of 1 is a smooth manifold if the local homomorphisms /

Ui
of the local uni-
formizing groups
i
mapping into the group l(1) of the bundle are monomorphisms
for all local uniformizing charts (

l
i
.
i
.
i
).
Proof. Let c denote sheaf of germs of smooth complex-valued functions on
1A and c

the sub-sheaf of nowhere vanishing complex valued smooth functions.

The isomorphism classes of complex line bundles on 1A. hence complex line V-
bundles on A. are in one-to-one correspondence with the elements of the cohomology
group H
1
orb
(A. c

). Now 1A is an innite paracompact CW complex, and c is a

ne sheaf, so the exponential exact sequence
0Zcc

0
gives an isomorphism
H
1
orb
(A. c

)

H
2
orb
(A. Z) .
Now let S denote the sheaf of germs of maps from 1A to l(1). We can identify S
as a subsheaf of c

by identifying l(1) as the unit circle in C. There is a deformation

retraction of c

onto S as etale spaces, so there is an isomorphism H

1
orb
(A. c

)
H
1
orb
(A. S

). Now Corollary 4.3.13 implies the rst statement. In the case of circle
bundles 1 over a CW complex it is well-known that the connecting homomorphism
can be represented by the rst Chern class of 1. The last statement follows from
the discussion after Denition 4.2.7. This completes the proof.
It is straightforward to generalize Theorem 4.3.15 in analogy with the manifold
case to
Theorem 4.3.16: The isomorphism classes of principal T
n
V-bundles 1 over an
orbifold A = (A. |) are in one-to-one correspondence with elements of H
2
orb
(A. Z
n
).
In [HS91] Haeiger and Salem study torus actions on orbifolds. There it is
shown that H
2
orb
(A. Z
n
) classies T
n
V-bundles up to local equivalence. This gives
a monomorphism H
2
(A. Z
n
)H
2
orb
(A. Z
n
) which, of course, is an isomorphism
rationally. [HS91] also study the exact homotopy sequence associated with a locally
free action of T
n
on an orbifold. We shall need
Denition 4.3.17: A (smooth) action of a Lie group G on an orbifold A =
(A. |) is a continuous action / : G AA such that for each o G and
r A there are local uniformizing charts (

l. . ) and (

l
t
.
t
.
t
) with r (

l).
/(o. r)
t
(

l
t
). together with an open neighborhood \ of o G and a smooth
map

/ : \

l

l
t
such that
t
(

/(o. r)) = /(o. ( r)) and for each o \ the
map r

/(o. r) is a dieomorphism.
The Haeiger-Salem theory applies to the following situation. Let G be a
compact Lie group acting locally freely on an orbifold \ with quotient orbifold
A. This gives rise to a bration 1O(n) G1\1A. which induces the long
exact homotopy sequence. Thus, we have
Theorem 4.3.18: Let G be a compact Lie group that acts locally freely on an
orbifold \ = (Y. 1) with quotient orbifold A = (A. |). Then the sequence

i
(G)
orb
i
(\)
orb
i
(A)
i1
(G)
4.4. COMPLEX ORBIFOLDS 123
of homotopy groups is exact.
We are particularly interested in the orbifold fundamental group
orb
1
(A) =

1
(1A). Thurston [Thu79] has given an alternative denition of
orb
1
(A) as the
group of deck transformations. Here we state this as a theorem.
Theorem 4.3.19: Let A = (A. |) be an orbifold. Then there exists a universal
covering orbifold

A = (

A.

|). The orbifold fundamental group
orb
1
(A) is the group
of deck transformations of the universal covering map :

AA.
Proof. The existence of a universal covering orbifold was given in Chapter
13 of [Thu79]. The proof that Haeigers denition and Thurstons denition of

orb
1
(A) coincide was given in [HD84] to which we refer.
4.4. Complex Orbifolds
On a complex orbifold there is a -invariant tensor eld J of type (1. 1) which
describes the complex structure on the tangent V-bundle TA. The almost complex
structure J gives rise in the usual way to the V-bundles
r,s
of dierential forms
of type (:. :). The standard concepts of Hermitian and Kahler metrics hold equally
well on orbifolds, and all the special identities involving Kahler, Einstein, or Kahler-
Einstein geometry hold. In particular, the standard Weizenbock formulas hold.
This is described in more detail in Chapter 3.
4.4.1. The Underlying Space and Divisors. Let A = (A. |) be a complex
orbifold. Then the underlying complex space A is a paracompact Hausdor space
with at worst quotient singularities. We can give A the structure of a locally C-
ringed space (A. O) as follows: The structure sheaf O
X
has stalks O
x
for r l A
that are isomorphic to the local ring O

C
n of germs of -invariant holomorphic func-
tions, where is the local uniformizing group of the uniformizing system (

l. . ).
Such rings are reduced, that is, they contain no nilpotent elements. (The reader is
referred to [GR65, GR84] for basic facts about locally ringed spaces.) Since the
local uniformizing groups are the identity on the dense set A
orb
reg
of regular orbifold
points, the ring O
x
is isomorphic to the ring O
C
n of convergent power series in
C
n
when r A is regular. In particular, O
C
n is a unique factorization domain,
or UFD for short. However, at a singular orbifold point r
orb
(A). the ring
O
x
is isomorphic to the ring O

x
C
n. which for
x
= id is not necessarily a UFD
[Mum61, Bri68]. Nevertheless it follows from a standard result (cf. [Bou89], pg
323) that the rings O

x
C
n are always integrally closed.
Recall that an element G1(n. C) is called a reection if has eigenvalue
1 with multiplicity n 1. Thus, a reection xes a hyperplane in C
n
. hence, its
name. A nite subgroup G1(n. C) is a reection group if it is generated by
reections. A nite subgroup G1(n. C) is small if it contains no reections.
We have the following well-known result [Pri67] (See also [Dim92]):
Theorem 4.4.1: Let G1(n. C) be a nite subgroup. Then
(i) The quotient space C
n
is smooth if and only if is a reection group.
(ii) There is a small group such that C
n
and C
n
are biholomorphic.
In the case n = 1 every nite subgroup of G1(1. C) = C

is a reection group
and has the form Z
n
for some integer n 2. so we have
Corollary 4.4.2: For every nite subgroup C

, the quotient C is smooth.

124 4. FUNDAMENTALS OF ORBIFOLDS
It follows from Theorem 4.4.1 that if a local uniformizing group of an orbifold
A = (A. |) contains a reection, the orbifold singular set will be bigger than the
singular set of the underlying complex space. Thus we have (A)
orb
(A). and
it follows from Theorem 4.4.1 that (A) =
orb
(A) if and only if none of the local
uniformizing groups
i
of the orbifold A = (A. |) contain a reection. If some
i
contains a reection, then the reection xes a hyperplane in

l
i
giving rise to a
ramication divisor on

l
i
and a branch divisor on A. We discuss this more in detail
below.
Now combining the discussion of the last several paragraphs we arrive at
Proposition 4.4.3: Let O
X
be the structure sheaf of a complex orbifold A =
(A. |). Then
(i) If r A
reg
then O
x
is isomorphic to the ring O
C
n of convergent power
series.
(ii) If r (A) with local uniformizing group . then there is a nontrivial
small subgroup of such that O
x
is isomorphic to the subring O

C
n of
invariant convergent power series, so O
x
is not a UFD.
(iii) For any r A the local ring O
x
is integrally closed.
This Proposition together with well-known results (cf. [GR84]) have the fol-
lowing consequence:
Proposition 4.4.4: The locally ringed space (A. O
X
) associated to a complex orb-
ifold has the following properties:
(i) (A. O
X
) is a reduced normal complex space,
(ii) the singular locus (A) is a closed reduced complex subspace of A and
has complex codimension at least two in A.
(iii) the smooth locus A
reg
is a complex manifold and a dense open subset of
A.
If (A) is compact it follows from Proposition 4.4.4 and Hartogs Theorem
that any holomorphic function on A
reg
extends to a holomorphic function on all
of A. Furthermore, Weil divisors can be dened on (A. O
X
). although there may
be none in the purely analytical setting. Similarly Cartier divisors can also be de-
ned. However, generally in contrast to Theorem 3.5.19, they are not the same as
we discuss below. But A
reg
is a complex manifold, so Theorem 3.5.19 applies and
Weil divisors and Cartier divisors coincide on A
reg
. Let us set some notation as
in subsection 3.5.2. For any normal variety A we let Div(A) denote the group of
Weil divisors on A. and H
0
(A.

) the group of Cartier divisors. Similarly,

we denote by Cl(A) the divisor class group of Weil divisors modulo linear equiva-
lence, and by CaCl(A) the group of Cartier divisor classes (Cartier divisors modulo
principal divisors). Moreover, by (ii) of Proposition 1.1.5 we have
Lemma 4.4.5: Let A be the underlying complex space of a complex orbifold. Then
Cl(A) Cl(A
reg
) CaCl(A
reg
).
However, not every Cartier divisor on A
reg
extends to a Cartier divisor on A.
If 1 is a Weil divisor with r 1 (A) = . then O
x
is not a UFD, so 1 is
not locally principal near r. The obstruction for being locally principal near r lies
in the local divisor class group Cl(A. r) = Cl(

l). Thus, the Cartier divisors are

identied with the subgroup of Div(A) consisting of locally principal divisors. So
the group CaCl(A) is generally a subgroup of the divisor class group Cl(A). In our
4.4. COMPLEX ORBIFOLDS 125
singular situation in lieu of Theorem 3.5.19 we have only an exact sequence
(4.4.1) 0CaCl(A)Cl(A)Cl(A)CaCl(A)0 .
If A is a projective algebraic variety (or scheme) [Har77], which is what is of
interest to us in this book, then there is an isomorphism CaCl(A) PicA. where
PicA denotes the Picard group of holomorphic line bundles (equivalently, invertible
sheaves) on A. Because of this we shall often use PicA to denote the group CaCl(A)
as well as the group of holomorphic line bundles or invertible sheaves. If 1
Div(A) the sheaf O(1) is not necessarily invertible. It is, however, reexive of rank
1 [Kol04], and one can characterize Cl(A) as group of rank 1 reexive sheaves. The
underlying space A of a complex orbifold enjoys another important property.
Denition 4.4.6: A normal reduced complex space is said to be Q-factorial if
every Weil divisor is Q-Cartier, i.e., some multiple of it is a Cartier divisor.
From the denition of orbifolds it is easy to see that
Proposition 4.4.7: The underlying space A of a complex orbifold is Q-factorial.
In order to work eciently with complex orbifolds, it is convenient to consider
Q-divisors on A. These are just formal nite linear combinations of analytic hyper-
surfaces with rational coecients, that is, Weil divisors with coecients in Q. Then
generalizing Denition 4.4.6 we say that a Q-divisor is Q-Cartier if some multiple
of it is a Cartier divisor. The concepts of eective, nef, big, etc. all apply equally
well to Q-divisors, but now for example the intersection number is only a rational
number instead of an integer. We refer to the literature [Laz04a, Laz04b, Mat02]
for more details. It will be convenient for us to work in the so-called log category,
that is one considers pairs (A. ). where A is the underlying space of a complex
orbifold and is a Q-divisor on A.
Denition 4.4.8: The branch divisor of an orbifold A = (A. |) is a Q-divisor
on A of the form
=

1
1
:

.
where the sum is taken over all Weil divisors 1

that lie in the orbifold singular

locus
orb
(A), and :

is the gcd of the orders of the local uniformizing groups

taken over all points of 1

and is called the ramication index of 1

.
Notice that if a Weil divisor 1 lies in
orb
(A). the local uniformizing group
of every point is nontrivial, and the gcd of the orders of all the local uniformizing
groups on 1 is its ramication index. As emphasized by Kollar [Kol04], pg. 15,
a complex orbifold A = (A. |) is uniquely determined by the pair (A. ). The
local branched covers
i
:

l
i
l
i
are uniquely determined by the conditions
that
i
is unramied over l
i
` (
orb
(A) l
i
). and the ramication index :

of 1

is the largest integer dividing the ramication index of

i
for all i with
1

l
i
= . So one can think of the orbifold A as the pair (A. ). This point
of view is also espoused in [Cam04]. Then by applying Theorem 4.4.1 we see
that one can associate to every complex orbifold (A. ) another complex orbifold
(A. ) which we call the unramied orbifold associated with (A. ). Notice that if
(A) = then (A. ) is a smooth complex manifold. We denote the order of the
unramied orbifold (A. ) by = (A). and more generally the order of the local
uniformizing group
x
at r (A) by (r). If the unramied orbifold (A. ) is of
nite type, multiplication by denes a map Cl(A)CaCl(A). More generally
126 4. FUNDAMENTALS OF ORBIFOLDS
if 1 Div(A) and (1) = lcm
xD
(r) then (1)1 is Cartier. We illustrate the
discussion of this subsection with an example.
Example 4.4.9: Consider the weighted projective spaces CP(w) described in Sec-
tion 4.5 below. For simplicity we take n = 2 and w = (1. 4. 6). The orbifold struc-
ture on CP(1. 4. 6) is the standard orbifold structure described in Section 4.5. In
this case the Weil divisor 1
0
= (.
0
= 0) gives a branch divisor
1
2
1
0
with ramica-
tion index 2. The order of the orbifold (CP(1. 4. 6).
1
2
1
0
) is = 12. The unramied
orbifold is CP(1. 2. 3) with its standard orbifold structure whose order is = 6.
Now CP(1. 2. 3) has two singular points, (0. 1. 0) with local uniformizing group the
cyclic group Z
2
, and (0. 0. 1) with local uniformizing group Z
3
. Both of these points
are contained in the Weil divisor 1
0
. and 1
0
is not a Cartier divisor. However, 61
0
is Cartier. Notice also that the divisors 1
1
= (.
1
= 0) and 1
2
= (.
2
= 0) are not
Cartier, but 31
1
and 21
2
are Cartier. Both CaCl(CP(1. 2. 3)) and Cl(CP(1. 2. 3))
are isomorphic to Z. and we have an exact sequence
0CaCl(CP(1. 2. 3))Cl(CP(1. 2. 3))Z
6
0 .
So Pic(CP(1. 2. 3)) CaCl(CP(1. 2. 3)) 6Cl(CP(1. 2. 3)).
We end this subsection by recalling
Denition 4.4.10: A canonical divisor 1
X
is any divisor on A such that its
restriction to A
reg
is associated to the canonical bundle
n
X
reg
.
Notice that 1
X
is only dened up to linear equivalence.
4.4.2. Orbidivisors and Orbisheaves. As with bundles and sheaves the
notion of a divisor in the category of orbifolds consists of a sequence of divisors,
one on each local uniformizing neighborhood that patch together in the right way.
Denition 4.4.11: An orbidivisor or Baily divisor on a complex orbifold A is
a Cartier divisor 1

U
on each local uniformizing system

l. . that satises the

two conditions
(i) If for each r A and . 1 T
x
then 1 T
x
.
(ii) If :

l. .

l
t
.
t
.
t
is an injection and 1 T
t
(x)
then 1
T
x
. Here T denotes the divisor sheaf on

l.
A Baily divisor is called absolute if on each local uniformizing system

l. .
the divisor 1

U
can be written as 1

U
= (1). where 1 is the quotient of -invariant
holomorphic functions on

l.
Thus, a Baily divisor is viewed as a divisor on the space of objects (
0
=

l
of the associated etale groupoid that satisfy certain compatibility conditions with
respect to the morphisms of the groupoid.
There are several important orbisheaves on complex orbifolds that we shall
work with. First, there is the structure orbisheaf O

dened by Denition 4.2.2,

where now O

U
is the sheaf of holomorphic functions on

l. Similarly there is the
meromorphic orbisheaf

consisting of meromorphic functions on each local uni-

formizing neighborhood

l. Finally, there is the canonical orbisheaf of a complex
orbifold. On a complex orbifold A of complex dimension n. we denote by
p
the
orbisheaf of holomorphic dierential j-forms on A. This is the orbisheaf constructed
from the usual sheaf
p

U
i
on each orbifold chart.
p
is a locally free orbisheaf of
rank

n
p

.
4.4. COMPLEX ORBIFOLDS 127
Denition 4.4.12: The canonical orbisheaf of a complex orbifold A of complex
dimension n is the orbisheaf
n
.
Given a Weil divisor 1 on A we can lift its restriction 1
U
i
to l
i
to a divisor

Ui
on the uniformizing neighborhood

l
i
by

1

Ui
=
1
i
(1l
i
). It is easy to check
that this denes a Baily divisor

1 on the orbifold A = (A. |). In fact, if : Z
divides the order of all the local uniformizing groups at all points of a Q-divisor
1
m
1. then it lifts to a Baily divisor. In particular, since the ramication index of a
branch divisor is the gcd of the orders of the uniformizing groups at all point of
we have arrived at
Proposition 4.4.13: The branch divisor or more generally any Q-divisor on
A of the form

. where :

is a ramication index and /

Z. lifts to a
Baily divisor 1 on A = (A. |).
A Baily divisor 1 obtained as the lift of a branch divisor is called a ramication
divisor. The following is straightforward
Proposition 4.4.14: To each Baily divisor 1 on A there corresponds a complex
line V-bundle [1] or equivalently an invertible orbisheaf O(1). Absolute Baily di-
visors on a complex orbifold A correspond to Cartier divisors on the underlying
complex space A.
Remark 4.4.1: Let A = (A. |) be a complex orbifold. Its canonical orbisheaf
by Proposition 4.2.18 determines an orbibundle which will be called the canonical
orbibundle of A and denoted by /

. We will sometimes write /

orb

, especially
in the presence of a non-trivial branch divisor . The anticanonical orbibundle
will always be denoted by /
1

. One can interchangeably think of 1

orb

as the
canonical orbibundle, the orbifold canonical divisor (see below), as well as the
orbifold canonical class (rational orbifold rst Chern class) in H
2
(A. Q). On the
other hand, when A = (A. ) one may consider a simpler complex orbifold: namely
(A. ). This orbifolds canonical divisor coincides with the canonical divisor 1
X
dened in (4.4.10). When A is non-singular, 1
X
traditionally also denotes the
canonical line bundle. When A has quotient singularities this is not dened. On
the other hand, there is a well-dened canonical orbibundle of the complex orbifold
(A. ) and this orbibundle will be denoted by /
X
with its dual written as /
1
X
.
The most important Baily divisor on a complex orbifold A = (A. |) is the
orbifold canonical divisor 1
orb

which is any Baily divisor associated to the canonical

orbibundle 1

. In the presence of branch divisors 1

an orbifold canonical divisor

1
orb

is not the same (meaning not linearly equivalent) as the canonical divisor 1
X
of the underlying complex space A. In fact we have
Proposition 4.4.15: The orbifold canonical divisor 1
orb

and canonical divisor

1
X
are related by
1
orb

1
X
+

1
1
:

.
In particular, 1
orb

1
X
if and only if there are no branch divisors.
Proof. Let (1
1
m
)1 be an irreducible branch divisor. Assume that 1 in-
tersects l
i
nontrivially, where

l
i
is a local uniformizing neighborhood. By deni-
tion the ramication index : divides the order of
i
. We can choose coordinates
(r
1
. . . . . r
n
) on

l
i
so that at a general point of the ramication divisor 1
i
the map
128 4. FUNDAMENTALS OF ORBIFOLDS
takes the form

i
(r
1
. . . . . r
n
) = (r
m
1
. r
2
. . . . . r
n
) = (.
1
. . . . . .
n
) .
A local section dr
1
dr
n
of the canonical bundle satises

i
(d.
1
d.
n
) = :r
m1
1
dr
1
dr
2
dr
n
.
so that the ramication divisor 1
i
is represented by r
m1
1
= 0. This expresses a
local section of the canonical bundle 1

Ui
as the sum of a local section of the pullback
of the canonical bundle 1
U
i
and a pole of order :1 along the ramication divisor
which translates to a pole of order
m1
m
along the pullback of the branch divisor
1. that is
1

Ui

i
1
U
i
+ (:1)1
i
=

i
1
U
i
+
:1
:

i
1.
This holds in all uniformizing neighborhoods, and for all irreducible branch divisors.
This proves the result.
In terms of the orbifold rational Chern class c
orb
1
(A) dened in the paragraph
above Theorem 4.3.14, Proposition 4.4.15 implies
(4.4.2) c
orb
1
(A) = c
1
(1

) = c
1
(A)

1
1
:

c
1
(O(1

)) H
2
(A. Q) .
Here c
1
(A) = c
1
(1
X
) H
2
(A. Q) is the rst Chern class of A. We will sometimes
write c
orb
1
(A) = c
orb
1
(A. ) to indicate the branch divisor.
The orbifold Kodaira dimension can be dened analogously with the manifold
case if one replaces the canonical bundle /
X
by the canonical orbibundle /
orb

.
Denition 4.4.16: Let A = (A. |) be a compact complex orbifold. The orbifold
Kodaira dimension Kod
orb
(A) is dened
Kod
orb
(A) =

if 1(A) = C
tr(1(A)) 1 otherwise,
where tr(1) denotes the transcendence degree of the ring 1.
Here 1(A) is the orbifold canonical ring dened by
(4.4.3) 1(A) = C

m1
H
0
(A. O(:1
orb

)).
Note that by passing to invariant sections Kod
orb
(A) is just the Itaka dimension of
the line bundle
O((1
orb

)) = O

1
X
+

1
1
:

.
Of special interest to us is the case Kod
orb
(A) = .
Denition 4.4.17: We say that a compact complex orbifold A is a (log) Fano
orbifold if c
orb
1
(A) 0. Alternatively, a Fano orbifold is a complex orbifold whose
anti-canonical orbibundle 1
1

is ample. A Fano orbifold surface is also called a

del Pezzo orbifold.
Actually this is a special case of a somewhat more general denition used in
Mori theory which we will make use of in Chapter 10.
Denition 4.4.18: A log del Pezzo surface is a pair (A. ). where
4.4. COMPLEX ORBIFOLDS 129
(i) o is a normal, projective surface,
(ii) =

o
i
1
i
is a linear combination of distinct irreducible divisors with
0 o
i
1. o
i
Q. and
(iii) (1
S
+ ) is ample. (That is, a suitable multiple of it is an ample
Cartier divisor.)
The Q-divisor is called the boundary.
Thus a 2-dimensional log Fano orbifold is a log del Pezzo surface. In the orbifold
case the boundary satises o
i
=
mi1
mi
. where the :
i
are natural numbers. Such a
boundary is sometimes called standard.
In analogy with the manifold case we also say that a complex orbifold A =
(A. |) is a Calabi-Yau orbifold if c
orb
1
(A) = 0. However, an important result of
Kollar [Kol05] given in Proposition 4.7.11 below says that under mild conditions
this is the same as c
1
(A) = 0.
Remark 4.4.2: It is clear from (4.4.2) that in the presence of a non-trivial branch
divisor , the complex space A may be a Fano variety in the usual algebraic sense
of the anti-canonical bundle of A being ample, while, depending on the divisor
, the complex orbifold A = (A. |) may or may not be Fano in the orbifold
sense. We would like to highlight this important dierence: our denition of a
Fano orbifold is not equivalent to A being Fano as an algebraic variety. We will
soon consider many examples of this situation. We warn the reader that when A
is not smooth other inequivalent notions of Fano or Q-Fano can be found in the
literature. Our extension of the Fano condition to the orbifold case is quite natural
in several dierent contexts. Most importantly, as we shall see later, a compact
Kahler Fano orbifold admits an orbifold Kahler metric of positive Ricci curvature
in the same Kahler class, this being one of the consequences of the orbifold version
of the Calabi-Yau theorem.
Example 4.4.19: Let (; .
1
. . . . . .
k
) be a Riemann surface with / marked points.
We can give (; .
1
. . . . . .
k
) an orbifold structure by dening local uniformizing
systems (

l
i
. C
mi
.
i
) centered at the point .
i
. where C
mi
is the cyclic group of
order :
i
and
i
:

l
i
l
i
=

l
i
C
m
i
is the branched covering map
i
(.) = .
m
i
.
An orbifold canonical divisor is given by
1
orb
(;z
1
,...,z
k
)

1
1
:
i

(.
i
) .
where 1

is an ordinary canonical divisor on the Riemann surface . Thus, if o

denotes the genus of the orbifold Chern number is
c
orb
1
(; .
1
. . . . . .
k
) = 2 2o / +
k

i=1
1
:
i
which equals the orbifold Euler characteristic
orb
(; .
1
. . . . . .
k
). In order that
(; .
1
. . . . . .
k
) be a Fano orbifold we need o = 0. and / < 4. When o = 0 we
have = o
2
with / = 1. 2. 3 marked points, and the Chern number becomes
c
orb
1
(o
2
; .
1
. . . . . .
k
) = 2 / +
k

i=1
1
:
i
.
For / = 1 we have c
orb
1
(o
2
; .
1
) = 1+(1:
1
). while for / = 2. we recapture Example
4.1.6 with c
orb
1
(o
2
; .
1
. .
2
) = (1:
1
) + (1:
2
). For / = 3 we obtain a Fano orbifold
130 4. FUNDAMENTALS OF ORBIFOLDS
only for (:
1
. :
2
. :
3
) = (j. 2. 2). (3. 3. 2). (4. 3. 2). and (5. 3. 2). These can all be
realized as hypersurfaces of Brieskorn-Pham polynomials in a weighted projective
space, see Example 4.6.15 below and subsection 10.1.1 of Chapter 10. Fano orbifold
structures on Riemann surfaces are usually called spherical [Mil75]. As in the
usual case there is a uniformization theorem for Riemannian orbifold metrics with
Gaussian curvature equal to 1. 0. 1. This is no longer a topological invariant, but an
invariant of the orbifold structure according to the orbifold Gauss-Bonnet formula
4.3.6.
Exercise 4.2: Determine for which / and :
i
one obtains Euclidean orbifolds
(i.e., c
orb
1
= 0) and hyperbolic orbifolds (c
orb
1
< 0) on o
2
.
4.4.3. The Orbifold Picard and Neron-Severi Groups. In [BG97b] the
authors introduced the set Pic
orb
(A) of equivalence classes holomorphic line V-
bundles over a complex orbifold A = (A. |). First we notice that complex line
V-bundles on A correspond to circle bundles on A which form a group by Corollary
4.3.13 namely H
1
orb
(A. c

). There are several approaches to deal with holomorphic

objects such as Pic
orb
(A). One may: 1) work directly on the V-bundle using holo-
morphic data in Denition 4.2.7, or 2) work on 1A using Theorem 4.3.11 in the
holomorphic context. These two approaches are actually very closely related since
holomorphic data on 1A really means holomorphic data on the set of objects of the
etale groupoid (, and this is the same as holomorphic data on the V-bundles since
the charts of 1A are

l
i

i
1l(n) with holomorphic data on

l
i
and smooth
data on 1l(n). In this case the group G is a complex Lie group and the bre 1 is
a complex manifold. In fact, we are most interested in the case of holomorphic line
bundles or holomorphic C

bundles
7
. We have arrived at the holomorphic version
of Theorem 4.3.11:
Theorem 4.4.20: Let A = (A. |) be an complex orbifold. There is a one-to-one
correspondence between isomorphism classes of holomorphic V-bundles on A with
group G and generic ber 1 and isomorphism classes of holomorphic bundles on
1A with group G and ber 1. In particular, holomorphic line V-bundles on A
correspond to holomorphic line bundles on 1A.
This leads to
Denition 4.4.21: The Abelian group of holomorphic line V-bundles on an orb-
ifold A is denoted by Pic
orb
(A).
The relationship between the orbifold Picard group Pic
orb
(A) and the ordinary
Picard group Pic(A) is given by:
Proposition 4.4.22: Let A = (A. |) be a complex orbifold of order . Then the
following hold:
(i) Pic(A) is naturally a subgroup of Pic
orb
(A).
(ii) Pic(A)
Z
Q Pic
orb
(A)
Z
Q.
(iii) If is nite, in particular if A is compact, the map 1 1

Pic
orb
(A)Pic(A)
is a homomorphism of Abelian groups.
7
A third approach using Grothendieck topologies was taken in [SW99].
4.4. COMPLEX ORBIFOLDS 131
Proof. (i): Ordinary holomorphic line bundles on A corresponds to absolute
line orbibundles on A. so there is a monomorphism Pic(A)Pic
orb
(A). (ii):
This follows from Corollary 4.3.8 and Theorem 4.4.20. (iii): If is nite, the map
1 1

, is a group homomorphism sending Pic

orb
(A) to Pic(A). since the absolute
orbibundle 1

is identied with an element of Pic(A).

We now study holomorphic line V-bundles on A by considering ordinary line
bundles on 1A that are holomorphic in the complex coordinates on

l. Let A be
a compact complex orbifold. We consider the classifying space 1A as covered by
charts of the form

l
i

i
1l(n). In the complex category the local uniformizing
neighborhoods

l
i
are identied with C
n
with the origin as the xed point of
i
. and
the injections
ji
of Denition 4.1.1 are biholomorphisms on their image. We now
dene the sheaf / of germs of holomorphic functions on 1A to be the sub-sheaf
of c consisting of functions that are holomorphic in the

l
i
. Similarly, /

denotes
the subsheaf of germs of nowhere vanishing functions. So the last statement of
Theorem 4.4.20 can be rephrased as
Theorem 4.4.23: On a complex orbifold there is an isomorphism
Pic
orb
(A) H
1
orb
(A. /

) .
The exponential exact sequence 0Z//

0 on 1A yields the long

exact sequence
H
1
orb
(A. Z)H
1
orb
(A. /)Pic
orb
(A)
p

c
orb
1
H
2
orb
(A. Z)H
2
orb
(A. /)
which, in turn, gives rise to an exact sequence
0H
1
orb
(A. /)H
1
orb
(A. Z)Pic
orb
(A)
p

c
orb
1
H
2
orb
(A. Z)
and the image of the map j

c
orb
1
is the orbifold Neron-Severi group
o
orb
(A) = Pic
orb
(A)H
1
orb
(A. /) H
1,1
(A. C) H
2
orb
(A. Z) .
Following the usual convention the rank
orb
(A) of o
orb
(A) is called the orbifold
Picard number.
Next we give slight generalization of the Fano index. We generalized the notion
of Fano index to orbifolds in [BG00b]. However, it is convenient to be more general
and consider a divisibility index as in [BGO07].
Denition 4.4.24: Let A be a complex orbifold such that c
orb
1
(A) is denite. The
divisibility index or just index of A is the largest positive integer : such that
p

c
orb
1
()
m
is an element of the free part of H
2
orb
(A. Z). The index of A is denoted by
1

.
Now we have
Lemma 4.4.25: Let A be a complex orbifold with 1

= :. Then there is a holo-

morphic line V-bundle L Pic
orb
(A) such that L
m
= /
1

.
Proof. The idea of the proof is simple, but since we are working on 1A we
write out the details. First on 1A we dene the following sheaf c
p,q
of dierential
forms: Let (.
1
. . . . . .
n
) be complex coordinates on

l
i
. Then using the standard
multi-index notation, we construct the sheaf c
p,q
whose stalks are spanned by
elements of the form 1
IJ
(.. c)d.
I
d .
J
. where 1 = i
1
i
p
and J = ,
1
,
q
are
the usual multi-indices, and 1 is a smooth function on

l
i
1O(2n) satisfying
132 4. FUNDAMENTALS OF ORBIFOLDS
1(
1
.. c) = 1(.. c) for
i
. We have, as usual, dierential operators

and the
Dolbeault complex

c
p,q

c
p,q+1

.
Notice that c
0,0
= c and ker(

: c
0,0
c
0,1
) = /.
By Theorem 4.3.15 there is a complex line V-bundle L such that L
m
= /
1

.
The transition function /
ij
for L are nowhere vanishing local sections of c over

l
i

i
1O(2n)

l
j

j
1O(2n) and the transition functions for L
m
= /
1

are
/
m
ij
. But since /
1

is holomorphic

/
m
ij
= 0; hence,

/
ij
= 0 implying that L is
holomorphic.
4.4.4. Kahler and Algebraic Orbifolds. It is usually straightforward to
generalize the geometry of Kahler manifolds as reviewed in Chapter 3 to the case
of orbifolds. However, as seen previously subtleties can and do occur. Here we
study the geometry of putting Kahler metrics on a complex orbifold. As discussed
in Section 4.2 a section of an orbibundle on an orbifold is just a sequence of sections
on the local uniformizing covers which patch together in the right way. We then
work with invariant sections by averaging over the local uniformizing groups as
in Equation 4.2.1. In this way the usual notions carry over easily to orbifolds.
For example, if A = (A. |) is a complex orbifold with an Hermitian metric o
and corresponding Hermitian 2-form
g
dened by
g
(A. Y ) = o(A. JY ). Then
A is a Kahler orbifold if
g
is a closed 2-form. We shall often denote a Kahler
orbifold by (A. ) or by (A. o.
g
) when we want to emphasize the orbifold metric
o. Of course, this denition is essentially the same as the manifold case, only the
interpretation of the objects is dierent. Here o. J and
g
are sequences of
i
-
invariant Hermitian metrics, almost complex structures, and a 2-forms, respectively,
on each local uniformizing neighborhood that are compatible with the injections
as discussed previously. With this in mind the results of Chapter 3 carry over
to orbifolds in a straightforward way. Now we are mainly interested in polarized
Kahler orbifolds whose denition is the orbifold analog of Denition 3.5.2, viz.
Denition 4.4.26: The pair (A. L) consisting of a complex orbifold A = (A. |)
with a positive line orbibundle L is called a polarized Kahler orbifold.
Exactly as in Denition 3.5.1, a positive line orbibundle is an orbibundle that
carries an Hermitian metric whose curvature form with respect to the Hermitian
connection is positive. Analogous to the manifold case polarized Kahler orbifold are
algebraic. This is the content of the Kodaira-Baily Embedding Theorem [Bai57]:
Theorem 4.4.27: Let A = (A. |) be a compact complex orbifold and suppose that
A has a positive orbibundle L. Then A is a projective algebraic variety.
This theorem is proved in much that same way as the usual Kodaira embedding
theorem by noticing that if is the order of the orbifold then L

is a positive
line bundle on A in the usual sense. See [Bai57] for details. In the process one
also makes use of a vanishing theorem, that we call the Kodaira-Baily Vanishing
Theorem.
Theorem 4.4.28: Let A be a compact complex orbifold, /

the canonical orbi-

bundle and L any holomorphic line orbibundle on A. If L /
1

is a positive line
orbibundle, then
H
q
(A. O(L)) = 0 for c 1 .
4.5. WEIGHTED PROJECTIVE SPACES 133
This theorem is of much interest in its own right. In analogy with the manifold
case we have
Denition 4.4.29: A compact Kahler orbifold A = (A. ) is called a Hodge
orbifold if [j

] lies in H
2
orb
(Z. Z). where j : 1ZZ is the natural projection
dened in section 4.3.
With this denition a restatement of the Kodaira-Baily embedding theorem
[Bai57] is:
Theorem 4.4.30 (Baily): A Hodge orbifold is a polarized projective algebraic va-
riety.
Now Yaus theorem and its corollaries have a direct translation to the orbifold
case.
Theorem 4.4.31: Let (A. o.
g
) be a compact Kahler orbifold. Then any real
(1. 1)-form on A which represents the cohomology class 2c
orb
1
(A) is the Ricci
form of a unique Kahler orbifold metric / such that [
h
] = [
g
].
It follows from Theorem 4.3.4 that this theorem is a special case of the more
general foliation version of El Kacimi-Alaoui [EKA90] given in Theorem 7.5.19
below. The orbifold version of Theorem 3.6.6 and its corollaries follows directly
from Theorem 4.4.31. These will be used to much advantage in Chapter 5.
4.5. Weighted Projective Spaces
In this section we discuss weighted projective spaces both as singular (or non-
singular) algebraic varieties, and as complex orbifolds. It is important to regard
orbifolds as added structure, and distinguish the orbifold from its underlying
algebraic variety. In this sense an orbifold is a type of Deligne-Mumford stack.
We begin with the algebraic geometric treatment of weighted projective spaces.
References here are [Del75, Dol82, IF00].
Given a sequence w = (n
0
. . . . . n
n
) of positive integers one can form the
graded polynomial ring o(w) = C[.
0
. . . . . .
n
]. where .
i
has weight n
i
. As an alge-
braic variety [Dol82] the weighted projective space with weights w = (n
0
. . . . . n
n
)
is dened to be the projective scheme Proj(o(w)). and is denoted by CP(w) =
CP(n
0
. . . . . n
n
). Weighted projective spaces are examples of normal projective al-
gebraic varieties with only mild quotient singularities. As in the case of ordinary
projective space the coordinates (.
0
. . . . . .
n
) are called homogeneous coordinates.
It is convenient to view the weights as the components of a vector w (Z
+
)
n+1
.
and we shall usually assume that they are ordered n
0
n
1
n
n
and
that gcd(n
0
. . . . . n
n
) = 1
8
. Notice that these two conditions are not additional
assumptions since the ordering can be performed by a permutation of the ane
coordinates, and the gcd condition can be obtained by redening the coordinate on
C

. It will at times be necessary to perform these operations.

In this book we are very interested in a canonical orbifold structure that comes
naturally by dening CP(w) as a certain orbit space. So our ocial denition of
a weighted projective should reect this.
8
We will sometimes use the notation gcd(w
0
, . . . , w
n
) (w
0
, . . . , w
n
). Since (w
0
, . . . , w
n
)
also denotes a vector w, the latter notation will only be used when dropping the gcd in front of
the brackets causes no ambiguity.
134 4. FUNDAMENTALS OF ORBIFOLDS
Denition 4.5.1: The weighted C

-action on the ane space C

n+1
. denoted by
C

(w). is dened by
(.
0
. . . . . .
n
) (
w
0
.
0
. . . . .
w
n
.
n
) .
where w = (n
0
. . . . . n
n
) is a sequence of positive integers.
Then we obtain the weighted projective space with a canonical orbifold struc-
ture, viz.
Denition 4.5.2: The weighted projective space CP(w) is dened as the quo-
tient space (C
n+1
` 0)C

(w).
Let us describe the canonically induced orbifold structure on CP(w). Set

l
i
=
(.
1
. . . . . .
m
) [ .
i
= 1, and let
i
C

(w) be the subgroup of n

th
i
roots of unity.
Note that

l
i
is invariant under the action of
i
. Set l
i
=

l
i

i
. Note that the
C

(w)-orbits on (C
m
` 0) ` (z
i
= 0) are in one-to-one correspondence with the
points of l
i
. Let us explicitly construct an orbifold atlas on CP(w). We cover
CP(w) by open sets of the form
(4.5.1) l
i
= [.
0
. . . . . .
n
] CP(w) [ .
i
= 0 .
For the local uniformizing system

l
i
.
i
.
i
we have open sets

l
i
C
n
with ane
coordinates (n
i0
. . . . . n
ii
. . . . . n
in
) which satisfy
(4.5.2) n
w
i
ij
=
.
wi
j
.
wj
i
.
Here n means that element is removed. Let us use some vector notation and
write y
i
= (n
i0
. . . . . n
ii
. . . . . n
in
) and y
wi
i
= (n
wi
i0
. . . . . n
wi
ii
. . . . . n
wi
in
). The maps
i
:

l
i
l
i
are
i
(y
i
) = y
wi
i
. The local uniformizing groups
i
are the cyclic groups
Z
w
i
which act on

l
i
by sending y
i
to cy
i
. where c Z
w
i
. The origin is clearly
a xed point of this action. The atlas | of uniformizing charts consists of the

l
i
.
i
.
i
together with their nontrivial intersections which are charts of the form

l
i
0


l
i
k
.
i
0
i
k
.
i
0
i
k
. where
i
0
i
k
is the cyclic group Z
gcd(w
i
0
,...,w
i
k
)
.
The map
i
0
i
k
:

l
i
0


l
i
k
l
i
0
l
i
k
is given by
i
0
i
k
(y
i
0
i
k
) =
y
gcd(w
i
0
,...,w
i
k
)
i
0
i
k
. Here the notation y
i0i
k
designates an n-vector lying in the open
set

l
i0


l
i
k
C
n
. We illustrate the injection maps in double overlaps. There
is an open set

\
ij


l
i

l
j
so that the injection map
ij,i
:

\
ij

l
i
is written as

ij,i
(n
ij
) = n
t
ij
. where t =
gcd(w
i
,w
j
)
w
i
and we choose the principal branch. This sat-
ises the condition
i

ij,i
=
ij
. and any other such injection map satisfying this
condition diers by a dierent choice of branch, that is, by an element of
i
= Z
w
i
.
Keeping with our notation we denote this orbifold by {(w) = (CP(w). |). and refer
to it as the standard orbifold structure on CP(w). Unless otherwise stated when we
refer to an orbifold structure on CP(w) we shall always mean the standard orbifold
structure. Clearly, from Denition 4.3.9 and the above description we have
Proposition 4.5.3: The orbifolds {(w) = (CP(w). |) are locally cyclic.
Weighted projective spaces are also important examples of toric varieties [Ful93]
since they admit the action of a complex torus (C

)
n
. This torus action can be de-
scribed as follows. Let the torus (C

)
n+1
act on C
n+1
by
(.
0
. . . . . .
n
) (
1
.
0
. . . . .
n
.
n
) .
4.5. WEIGHTED PROJECTIVE SPACES 135
Then the quotient group (C

)
n+1
C

(w) acts on CP(w) preserving the standard

orbifold structure. So {(w) = (CP(w). |) is a toric orbifold as well.
The orbifold singular locus
orb
(CP(w)) is determined by setting certain of the
homogeneous coordinates equal to zero. For any proper subset J n
0
. . . . . n
n

let o
J
denote the subset of CP(w) such that .
j
= 0 for all , J. Let J
c
denote
the compliment of J in n
0
. . . . . n
n
. Then o
J
lies in
orb
(CP(w)) if and only if
gcd(n
i
1
. . . . . n
i
k
) 1. where J
c
= n
i
1
. . . . . n
i
k
. The local uniformizing group
at generic points in o
J
is then the cyclic group Z
gcd(wi
1
,...,wi
k
)
. It is clear that

orb
(CP(w)) is nonempty as long as w = (1. . . . . 1). and that
orb
(CP(w)) has a
natural stratication according to the subsets o
J
.
We now consider some important sheaves and orbisheaves on CP(w). For each
integer n we have the sheaves O
CP(w)
(n) dened as the sheaf associated to the
graded o(w)-module o(w)(:). where o(w)(:)
j
= o(w)
m+j
. But by Lemma 4.2.4
O
CP(w)
(n) also gives rise to an orbisheaf which we denote O
1(w)
(n). We are in-
terested in Pic
orb
({(w)). Here we make use of Proposition 4.2.18 and identify
holomorphic line V-bundles on {(w) with invertible orbisheaves. We have
Theorem 4.5.4: The Abelian group Pic
orb
({(w)) Z is generated by the invert-
ible orbisheaves O
1(w)
(1). That is, every invertible orbisheaf (holomorphic line
V-bundle) on {(w) is of the form O
1(w)
(n) for some n Z.
Proof. First the isomorphism Pic
orb
({(w)) Z follows from Proposition
4.4.22. On the uniformizing neighborhood

l
i
the orbisheaf O
1(w)
(1) is dened by
the sheaf

i
O
Ui
(1) =
1
i
O
Ui
(1)

1
i
c
U
i
O

U
i
. As usual we let (.
0
. . . . . .
n
) denote
the homogeneous coordinates on CP(w) with .
i
of weight n
i
. Then the n
th
i
root
w
i

.
i
is a well dened section of
1
i
O
Ui
(1)

1
i
c
U
i
O

U
i
for each i = 0. . . . . n. Now
any section : of
1
i
O
U
i
(1)

1
i
c
U
i
O

U
i
is a nite linear combination of sections
of the form
.
m
i
0
i
0
.
m
in
i
n
.
mi
i
.
where the i
k
run over the n indices dierent than ,, and the :
i
k
and :
i
are
nonnegative integers satisfying

i
k
,=i
:
i
k
n
i
k
= :
i
n
i
+ 1. Moreover, by Equation
(4.5.2), .
j
= n
ij
w
i

.
w
j
i
for all , = i. Thus, : = j(y
i
)
w
i

.
i
. where j is a polynomial
in the components of y
i
. Hence,
w
i

.
i
is a basis for

i
O
U
i
(1) for each i = 0. . . . . n
which implies that the orbisheaf O
1(w)
(1) is invertible. Similarly,
w
i

.
m
i
is a basis
for

i
O
Ui
(:). so O
1(w)
(:) is a V-invertible orbisheaf. Next we show that there is
an isomorphism of orbisheaves
(4.5.3) O
1(w)
(/) O
1(w)
(|) O
1(w)
(/ +|)
for all /. | Z. We know that for each / Z.
w
i

.
k
i
is a basis for the orbisheaf

i
O
U
i
(/). This gives an isomorphism of bases
w
i

.
k
i

w
i

.
l
i

w
i

.
k+l
i
. and
proves the isomorphism of orbisheaves (4.5.3). This shows that O
1(w)
(1) gener-
ates Pic
orb
({(w)). Clearly a similar analysis holds for O
1(w)
(1).
Lemma 4.5.5: The ordinary sheaves O
CP(w)
(n) are not generally invertible.
136 4. FUNDAMENTALS OF ORBIFOLDS
Proof. To see that O
CP(w)
(1) is not generally invertible we consider CP(1. 1. 2)
([BR86], pg 130). Consider the sections of O
U2
(1). It is minimally generated by
.
0
and .
1
. but it is not free, since for example
.
0
.
1
.
2
.
0

.
2
0
.
2
.
1
= 0 .
We now consider the properties of weighted projective spaces as algebraic va-
rieties. First we give an alternative description of weighted projective spaces as
complex spaces. Consider the nite group G
w
= Z
w
0
Z
w
n
. Then CP
n
G
w
is a developable compact complex orbifold. We have [Dol82]
Lemma 4.5.6: As algebraic varieties there is an isomorphism CP(w) CP
n
G
w
.
But no such isomorphism of orbifolds exists unless w = (1. . . . . 1).
Proof. One easily sees that the isomorphism is induced by the map .
i
.
w
i
i
.
However, there can be no isomorphism of orbifolds which can be seen by considering
the orbifold invariant
orb
1
. By the long exact homotopy sequence 4.3.18 one sees
that
orb
1
(CP(w)) is the identity; whereas,
orb
1
(CP
n
G
w
) G
w
.
In the future by a weighted projective space we shall mean CP(w) with its
induced orbifold structure as a quotient of o
2n+1
by a locally free circle action.
There is another isomorphism of algebraic varieties between weighted projective
spaces that is more important to us in the sequel. This was rst noticed by Delorme
[Del75]. First we give some notation.
Denition 4.5.7: For any weight vector w we dene for each i = 0. . . . . n the
integers:
(i) d
i
= gcd(n
0
. . . . . n
i1
. n
i+1
. . . . . n
n
).
(ii) c
i
= lcm(d
0
. . . . . d
i1
. d
i+1
. . . . . d
n
).
(iii) o
w
= lcm(d
0
. . . . . d
n
).
and a new weight vector w = (
w0
e0
. . . . .
wn
en
).
Note that (n
0
. . . . . n
n
) = 1 implies that (d
i
. d
j
) = 1 for all i. ,. so in this case
the c
i
and o
w
are just the corresponding products. More generally the following is
easily veried:
Lemma 4.5.8: The following hold:
(i) d
i
[n
j
for all , = i.
(ii) c
i
[n
i
for all i.
(iii) n
i
[(CP(w)) for all i.
(iv) d
i
[(CP(w)) for all i.
(v) o
w
[(CP(w)).
Note that (ii) implies that the weight vector w is integer-valued, and that
w = w if and only if d
i
= 1 for all i = 0. . . . . n. This leads to the notion of
well-formedness.
Denition 4.5.9: We say that the weight vector w = (n
0
. . . . . n
n
) or the weighted
projective space CP(w) with its standard orbifold structure {(w) is well-formed
if d
i
= 1 for all i = 0. . . . . n.
Thus, w is well-formed if and only if w = w. We now have [Del75]
Proposition 4.5.10: There is an isomorphism CP(w) CP( w) of algebraic va-
rieties. But there is no orbifold isomorphism unless w = w. In particular, as
4.5. WEIGHTED PROJECTIVE SPACES 137
algebraic varieties every weighted projective line CP(n
0
. n
1
) is isomorphic to the
projective line CP
1
.
Proof. Following [Dol82] dene the graded subring o
t
of the graded ring
o(w) by
o
t
=

kZ
o(w)
ka
w
.
Then writing o
t
= C[n
0
. . . . . n
n
] we see that n
i
= .
d
i
i
has weight
a
w
w
i
e
i
since c
i
d
i
=
o
w
. Thus, by 2.4.7 of [Gro61] we have Proj(o( w)) Proj(o
t
) Proj(o(w)) which
proves the rst statement. For the second statement it is clear that unless w = w
the weighted projective spaces CP( w) and CP(w) have dierent local uniformizing
groups. The last statement is clear from Denition 4.5.7.
Remark 4.5.1: The isomorphism of Proposition 4.5.10 is due to the existence of
the reection groups that leave hyperplanes invariant (See Theorem 4.4.1 and the
preceding paragraph). Elements of such groups are also called pseudoreections
[Dol82] or quasireections [IF00]. In the case of CP(w) these nite groups are the
cyclic groups Z
d
i
.
Example 4.5.11: Here we give an explicit example of the isomorphism of Proposi-
tion 4.5.10 which relates to the famous Poincare homology sphere as well as other
homology 3-spheres (cf. Example 10.1.10 below). The weighted projective space is
CP(6. 2(6/ 1). 3(6/ 1)). where / 1. Here we have w = (6. 2(6/ 1). 3(6/ 1))
so d
0
= 6/ 1. d
1
= 3. d
2
= 2. o
w
= 6(6/ 1). Let us go through the isomorphism
in steps, rst using d
0
then d
1
and nally d
2
. We have
CP(6. 2(6/ 1). 3(6/ 1)) CP(6. 2. 3) CP(2. 2. 1) CP(1. 1. 1) = CP
2
.
Thus, as algebraic varieties CP(6. 2(6/ 1). 3(6/ 1)) is isomorphic to the complex
projective plane CP
2
. and similarly with the projective spaces occurring at the
intermediate stages.
The following is a result of Delorme [Del75].
Lemma 4.5.12: For any , Z there is a unique J Z and sheaf isomorphisms
O
CP(w)
(,) O
CP(w)
(Jo
w
) O
CP( w)
(J) .
Proof. The isomorphisms can be described as follows (cf. [BR86]): recall
that O
CP(w)
(,) is the sheaf associated to the graded o(w)-module o(w)(,) de-
ned by o(w)(,)
l
= o(w)
j+l
. Now there is an equality of schemes Proj(o(w)) =
Proj(o
t
(w)). where o
t
(w) is the subring of o(w) dened by
o
t
(w) =

kZ
o(w)
kaw
.
The isomorphisms are then induced by the equality of graded o
t
(w)-modules

kZ
o(w)(,)
kaw
=

kZ
.
b
0
(j)
0
.
b
n
(j)
n

o(w)(,

i
/
i
(,)n
i
)

kaw
.
where /
i
(,) are the unique integers 0 /
i
(,) < d
i
satisfying , = /
i
(,)n
i
+ c
i
d
i
for
some c
i
Z. So the integer J is

,

i
/
i
(,)n
i

o
w
.
Exercise 4.3: Show that if d
i
1 for all i = 0. . . . . n, then there is a sheaf
isomorphism O
CP(w)
([w[) O
CP( w)
.
138 4. FUNDAMENTALS OF ORBIFOLDS
It should be clear from the explicit construction of these isomorphisms that
they cannot be isomorphisms of orbisheaves. Here is an example showing that the
isomorphism of orbisheaves given by Equation 4.5.3 is not generally an isomorphism
of sheaves, that is the canonical homomorphism of sheaves
(4.5.4) O
CP(w)
(/) O
CP(w)
(|)O
CP(w)
(/ +|)
is not generally an isomorphism.
Example 4.5.13: [BR86] Consider the weighted projective space CP(2. 3). As al-
gebraic varieties CP(2. 3) CP
1
. So by Lemma 4.5.12 and its proof there are
isomorphisms
O
CP(2,3)
(2) O
CP
1. O
CP(2,3)
(4) O
CP
1. O
CP(2,3)
(6) O
CP
1(1).
It follow that the canonical homomorphism 4.5.4 cannot be an isomorphism.
Nevertheless, we do have
Lemma 4.5.14: For any integers /. | there is an isomorphism of sheaves
O
CP(w)
(/o
w
) O
CP(w)
(|o
w
)O
CP(w)
((/ +|)o
w
).
We now turn to the description of orbifold canonical divisors 1
orb
1(w)
of the
orbifolds {(w). and their rational Chern classes.
Lemma 4.5.15: The following hold:
(i) 1
orb
1(w)

1
CP( w)
+

i
(1
1
d
i
)[(.
i
)].
(ii) o
w
c
orb
1
(/
orb
1(w)
) = c
1
(O
1(w)
([w[)).
In particular, any weighted projective space with its standard orbifold structure is
log Fano.
Proof. (i) follows easily from Proposition 4.4.15. To prove (ii) we notice
that the divisor (.
i
) has degree n
i
. and we can clear the denominator in (i) by
multiplying by o
w
. This avoids working with fractional sheaves. So we see that
(i) and Equation 4.4.2 give
o
w
c
orb
1
(/
orb
1(w)
) = c
1

O
1(w)

o
w
([ w[ +

(1
1
d
i
) n
i
)

= c
1

O
1(w)
(o
w

(1
1
d
i
1) n
i
)

= c
1

O
1(w)
(o
w

n
i
d
i
)

= c
1

O
1(w)
(o
w

n
i
c
i
d
i
)

= c
1
(O
1(w)
([w[)) .
Here we have made use of Denition 4.5.7, Lemma 4.5.8, and o
w
= c
i
d
i
. This proves
(ii).
Remark 4.5.2: Lemma 4.5.15 is far from true generally. Example 4.6.15 below
illustrates non-standard orbifold structures on CP
n
that are not Fano.
4.6. Hypersurfaces in Weighted Projective Spaces
Next we are interested in certain hypersurfaces, or more generally, complete
intersections in CP(w). As we shall see throughout this text, these hypersurfaces
provide a wide range of interesting examples of Kahler orbifolds, and ultimately
Sasakian structures on odd dimensional manifolds. Although the end product al-
most exclusively involves the hypersurface case, the foundations are laid for the
4.6. HYPERSURFACES IN WEIGHTED PROJECTIVE SPACES 139
more general case of weighted homogeneous varieties. We are interested in both
their algebraic geometric as well as their orbifold properties.
Denition 4.6.1: A polynomial 1 C[.
0
. . . . . .
n
] is said to be a weighted ho-
mogeneous polynomial of degree d and weight w = (n
0
. . . . . n
n
) if for any
C

1(
w
0
.
0
. . . . .
w
n
.
n
) =
d
1(.
0
. . . . . .
n
) .
The space of weighted homogeneous polynomials of degree d is a
d
-eigenspace
of the induced C

(w)-action on C[.
0
. . . . . .
n
] and in analogy with the classical case
will be denoted by H
0
(CP(w). O(d)). If \ C
n+1
is a variety dened by a collection
of weighted homogeneous polynomials 1
1
. . . . . 1
p
with weights (n
0
. . . . . n
n
), then \
is invariant under the C

-action 4.5.1. This statement has a converse due to Orlik

and Wagreich [OW71b].
Proposition 4.6.2: Let \ C
n+1
be an irreducible analytic variety which is in-
variant under the C

-action dened by Equation 4.5.1. Then \ is an algebraic

variety and the ideal of polynomials vanishing on \ is generated by weighted homo-
geneous polynomials of Denition 4.6.1.
Proof. Let C.
1
. . . . . .
n
denote the ring of convergent power series. Then
for 1 C.
0
. . . . . .
n
there is a unique polynomial 1
i
for each i such that
(4.6.1) 1( z) =

i=0

i
1
i
(z) .
This series converges for small enough. If 1 vanishes on \ near the origin 0
in C
n+1
. then z \ implies that

i

i
1
i
(z) = 0 for small enough. Hence,
1
i
(z) = 0 for all i and all z \ near o. Let 1(\ ) denote the ideal of functions in
C.
0
. . . . . .
n
vanishing on \. and let . be the ideal generated by 1
(j)
i
. where
the 1
(j)
s are generators of 1(\ ). Clearly, . 1(\ ). Now the locus of zeroes of
. is \. since if z \ and is within the radius of convergence of 1
(j)
for all ,,
there must be some 1
(j)
i
with 1
(j)
i
(z) = 0. Thus, by Hilberts Nullstellensatz the
radical of . is 1(\ ). Let .
t
be the ideal generated by 1
(j)
i
in the polynomial ring
C[.
0
. . . . . .
n
]. and let 1
t
=

.
t
. Then 1
t
C[.
0
. . . . . .
n
] =

. = 1(\ ). So 1(\ ) is
generated by polynomials, and the algebraic variety dened by 1(\ ) is \. Now if 1
is polynomial in 1(\ ) then there is only a nite number of is in Equation (4.6.1)
with 1
i
dierent from zero. Since the 1
(j)
s are generators of 1(\ ). the weighted
homogeneous polynomials 1
(j)
i
also generate 1(\ ).
Remark 4.6.1: Notice that for a given (n + 1)-tuple of weights w = (n
0
. . . . . n
n
)
and degree d there may be no such polynomial 1. So the existence of a weighted
homogeneous polynomial 1 C[.
0
. . . . . .
n
] puts constraints on the set of the pair
(w; d).
In analogy with Denition 3.1.19 we have
Denition 4.6.3: A weighted (homogeneous) variety A is the common zero
locus in CP(w) of a collection of weighted homogeneous polynomials in C[.
0
. . . . . .
n
].
A weighted (homogeneous) hypersurface A
f
is the zero locus in CP(w) of
a single weighted homogeneous polynomial 1. A weighted variety A is called a
weighted complete intersection if the number of polynomials in the collection
equals the codimension of A.
140 4. FUNDAMENTALS OF ORBIFOLDS
Since each weighted homogeneous polynomial has a degree, we can associate
precisely c = codim(A) degrees (d
1
. . . . . d
c
) to a weighted complete intersection A.
We call d = (d
1
. . . . . d
k
) the multidegree of the complete intersection.
Notice that a weighted variety denes another ane variety as the zero locus in
C
n+1
of a collection of weighted homogeneous polynomials which is clearly invariant
under the C

(w)-action. This leads to

Denition 4.6.4: Let A CP(w) be a weighted variety in CP(w) and : C
n+1
`
0 CP(w) the canonical projection. The punctured ane cone (

(A) over
A is dened by (

(A) =
1
(A) and the ane cone ((A) is the completion of
(

(A) in C
n+1
.
Notice that (

(A) cannot have isolated singularities. We shall assume that

when a weighted homogeneous polynomial 1 does exist that it is non-degenerate in
the sense that each variable occurs. In this treatise we shall only be interested in
the case where either 1 is regular at the origin o or when o is an isolated singular
point. Notice that 1 will be regular at o if and only if some variable, say .
n
. occurs
as a linear monomial in 1. and this occurs if and only if n
n
= d. In this case we
have a linear cone and we can remove the variable .
n
and consider the reduced
weighted homogeneous polynomial o C[.
0
. . . . . .
n1
]. Below we give conditions
on (w; d) such that the hypersurface A
f
has an isolated singularity at the origin.
Denition 4.6.5: We say that a weighted variety A CP
n
(w) is quasi-smooth
of dimension : if its ane cone ((A) is smooth of dimension : + 1 outside the
origin 0. that is, (

(A) is smooth.
When A CP(w) is quasi-smooth all its singularities are quotient singularities.
In fact since a weighted homogeneous variety is invariant under the C

(w)-action
we have
Proposition 4.6.6: Let : ACP(w) be a quasi-smooth weighted homogeneous
variety. Then the orbifold structure {(w) = (CP(w). |) on CP(w) naturally in-
duces a locally cyclic orbifold structure A on A.
Proof. Since A is quasi-smooth and invariant under the C

-action, the local

uniformizing groups of the orbifold A are just the isotropy subgroups of the weighted
C

(w)-action on the punctured ane cone (

(A) C
n+1
. If r A has local
uniformizing group
x
, then this is also the local uniformizing group of (r)
CP(w). So the set of local uniformizing groups of the orbifold A is precisely the
subset of local uniformizing groups of {(w) taken over all image points of . Then
since {(w) is a cyclic orbifold, so is A.
We will sometimes emphasize the degree d of a weighted variety and write A =
A
d
CP(w). Note that the pair (d. w) determines not just one 1 satisfying the non-
degeneracy requirement, but rather a family of polynomials in H
0
(CP(w). O(d))
and by abuse of notation we shall write interchangeably A
d
for a specic member
or a suciently general member of this deformation family. The context should
make clear which is meant. Let us consider an important example which will play
a signicant role in this book.
Example 4.6.7: Consider weighted homogeneous polynomials 1 of the form
(4.6.2) 1(.
0
. . . . . .
n
) = .
a0
0
+ +.
a
n
n
H
0
(CP(w). O(d)) .
4.6. HYPERSURFACES IN WEIGHTED PROJECTIVE SPACES 141
where o
i
1 and n
i
o
i
= d for all i = 0. . . . . n. Such polynomials were introduced
by Pham [Pha65] and used by Brieskorn [Bri66] in his famous study of exotic
spheres. They have become known as Brieskorn-Pham polynomials, and we shall
often employ the abbreviation BP polynomials. Note that since we always assume
gcd(n
0
. . . . . n
n
) = 1. we have d = lcm(o
0
. . . . . o
n
). It is easy to see that the zero
locus A
f
CP(w) is quasi-smooth, so A
f
is an orbifold with orbifold structure in-
duced by the standard orbifold structure of CP(w). This orbifold structure depends
on the pair (w; d) or equivalently on the exponent vector a = (o
0
. . . . . o
n
). We denote
the induced orbifold structure by A(a). Now we can perturb the BP polynomial 1
as follows. Let j H
0
(CP(w). O(d)) be any weighted homogeneous polynomial of
degree d and weight vector w. Let A
a,p
= A
f+p
denote the corresponding weighted
homogeneous hypersurface together with its ane cone (
a,p
= (
f+p
. We say that
A
a,p
and/or (
a,p
is a (weighted homogeneous) perturbation of the BP polynomial 1
if the following genericity condition holds:
(GC) The intersections of (
a,p
with any number of hyperplanes (.
i
= 0)
are all smooth outside the origin.
We shall often be interested in a generic member of this family.
We can naturally extend Denition 4.6.3 to weighted complete intersections,
and as in the hypersurface case we will write A
d
1
,...,d
k
CP(w) to denote a su-
ciently general member of the family dened by the pair (w; d).
We now extend the notion of well-formedness given in Denition 4.5.9 to sub-
varieties of weighted projective spaces. This seems to have rst appeared in [IF00].
Denition 4.6.8: A subvariety A CP(w) of codimension / is well-formed if
CP(w) is well-formed and A contains no codimension / + 1 singular stratum of
CP(w). It follows that codim
X
(A CP(w)
sing
) 2.
It follows from Denition 4.6.8 and Theorem 4.4.1 that a quasi-smooth well-
formed variety A CP(w) is a locally cyclic complex orbifold with no branch
divisors. Thus, we have
Proposition 4.6.9: Let A be a quasi-smooth subvariety of CP(w). Then A is
well-formed if and only if its orbifold singular locus and algebro-geometric singular
locus coincide, equivalently A has no branch divisors.
In the case of complete intersections the well-formedness condition can be ex-
pressed as a divisibility condition on the pair (w; d).
Corollary 4.6.10: Let A
d
CP(w) be a weighted complete intersection of multi-
degree d = (d
1
. . . . . d
k
). Then A
d
is well-formed if and only if CP(w) is well-formed
and for all 1 : / the gcd of any (n 1 / +:) of n
0
. . . . . n
n
must divide
at least : of d
1
. . . . . d
k
. If A
d
CP(w) is a weighted homogeneous hypersurface,
then A
d
is well-formed if and only if CP(w) is well-formed and for all 0 i < , n
we have gcd(n
0
. . . . . n
i
. . . . . n
j
. . . . . n
n
)[d. In particular, if A
d
= A
f
. where 1 is a
Brieskorn-Pham polynomial, then A
d
is well-formed if and only if CP(w) is well-
formed.
We continue the discussion of Example 4.6.7.
Example 4.6.11: For a BP polynomial 1 (or a perturbation thereof) of Equation
4.6.2, we see that the branch divisor of A(a) is
=

1
1
d
i

1
i
.
142 4. FUNDAMENTALS OF ORBIFOLDS
where d
i
is dened in Denition 4.5.7 and 1
i
are the hyperplane divisors (.
i
= 0).
Notice that if d
i
= 1 there is no contribution of 1
i
to the branch divisor, or
equivalently 1
i
is not contained in the orbifold singular locus
orb
(A). Of course,
in the well-formed case d
i
= 1 for all i, and there is no branch divisor. When
we want to emphasize the appearance of branch divisors we shall use the notation
A(a) = (A. ) with the a or equivalently the BP polynomial 1 understood. We can
obtain other orbifold structures on the same variety A by applying the procedure
of Section 4.5. Ultimately, we end up with a well-formed weight vector w with the
orbifold structure (A. ). This will give rise to a new BP polynomial with a new
exponent vector b which we describe in more detail below.
It follows from Proposition 4.4.15 that in the non-well-formed case the orbifold
canonical divisor and canonical divisor are not the same. In this case it is the
orbifold canonical bundle and corresponding orbisheaf that is of interest to us. In
either case the adjunction formula holds as usual as long as one replaces the canon-
ical bundle /
X
by the orbifold canonical bundle /
orb

. Somewhat more generally

we have [BGK05]
Proposition 4.6.12: Let 1 H
0
(CP
n
(w). O(d)) and let A
f
and C
f
be as in
Denition 4.6.3. Assume also that A
f
is quasi-smooth, n 2 and :(d [w[) 0.
Then the following three spaces are naturally isomorphic:
(i) Global sections of
s
/
orb

.
(ii) C

(w)-invariant global sections of

s
/
C
f
.
(iii) The space of weighted homogeneous polynomials of weight :(d [w[),
modulo multiples of 1.
Proof. Dierential forms on A
f
correspond to C

-invariant dierential forms

on C
f
` 0. so global sections of
s
/
orb

can be identied with C

(w)-invariant
global sections of
s
/
C
f
\0
. If n 2 then C
f
is a hypersurface of dimension
2 with an isolated singularity at the origin, thus normal. Hence, by Hartogs
global sections of
s
/
C
f
agree with global sections of
s
/
C
f
\0
. This shows the
equivalence of (i) and (ii).
The C

(w)-action on has weight [w[ d. thus

s
/
C
f
is the trivial bundle on
C
f
. where the C

(w)-action has weight :([w[ d). Its invariant global sections are
thus given by homogeneous polynomials of weight :(d [w[) times the generator
.
Putting : = 1 in Proposition 4.6.12 gives the adjunction formula
(4.6.3) /
orb

= O

(d [w[) .
The general adjunction formula can be proved as usual as long as one replaces the
canonical divisor by the orbifold canonical divisor.
Proposition 4.6.13 (Adjunction formula): Let 1 be a quasi-smooth orbidivisor
in the weighted projective orbifold {(w). Then the adjunction formula
/
orb
D
= O(1
orb
D
) = O(1
orb
1(w)
+1)[
D
holds.
It follows that
dc
orb
1
(/
orb
(a)
) = c
1
(O
1(w)
(d [w[))
and we get
4.6. HYPERSURFACES IN WEIGHTED PROJECTIVE SPACES 143
Corollary 4.6.14: Let A
d
CP(w) be a quasi-smooth weighted hypersurface of
degree d. The orbifold A
f
= (A
f
. ) is log Fano if and only if d < [w[.
One can easily generalize this statement to the case of complete intersections.
Let us consider an example of a family of quasi-smooth but not well-formed hy-
persurfaces which illustrates various concepts discussed earlier as well as being of
interest to us later.
Example 4.6.15: Consider the Brieskorn-Pham polynomials 1 of Examples 4.6.7
and 4.6.11 where now the (n+1)-tuple a = (o
0
. . . . . o
n
) consists of pairwise relatively
prime integers for all i = 0. . . . . n. Then d = o
0
o
n
and n
i
= o
0
o
i
o
n
. so
d
i
= o
i
. Clearly, CP(w) is not well-formed and as an algebraic variety we have
CP(w) = CP
n
. Likewise, it is easy to see that A
d
CP(w) is simply CP
n1
.
However, the orbifold structure very much depends on a with
A(a) = (A
d
. ) =

CP
n1
.

1
1
o
i
1
i

.
where 1
i
CP
n1
are the coordinate hyperplanes .
i
= 0. According to the Corol-
lary 4.6.14 the orbifold A(a) is log Fano if and only if
(4.6.4) 1 <
n

i=0
1
o
i
.
Thus A
d
= CP
n1
is smooth and certainly Fano as an algebraic variety while the
orbifold A(a) = (CP
n1
. ) is log Fano only when the inequality (4.6.4) holds.
This can be seen by computing the orbifold rst Chern class of A(a). Let [H] be
the hyperplane class. Then c
1
(CP
n1
) = n[H], [1
i
] = [H] and, hence,
c
orb
1
(A(a)) = n[H]
n

i=0
(1 1o
i
)[H] =

i=0
1
o
i
1

[H] .
When n = 2 the orbifold A(a) is topologically just CP
1
o
2
. but the orbifold
structure and the geometry of A(a) depend on a in a rather complicated way
[Mil75]. In particular, A(a) admits an orbifold Riemannian metric of constant
positive curvature if and only if it is log Fano. As we are considering the case that
o
0
. o
1
. o
2
are pairwise relatively prime, this occurs in only one case, a = (2. 3. 5). If
1

2
i=0
1
a
i
the orbifold geometry is hyperbolic, and one can show that A(a) is
actually a quotient of a smooth Riemann surface
g
of genus o = o(a) 0 by a
discrete group of isometries. Hence, in such a case A(a) admits a negative constant
curvature orbifold metric. See Example 4.4.19 as well as Proposition 4.1.7. We
shall revisit this and similar examples in Chapters 5, 9, 10 and 11.
More generally, for the case of a Riemann surface
w
realized as hypersurfaces
of a weighted homogeneous polynomial in CP(n
0
. n
1
. n
2
). one has the genus formula
[OW71a, Orl72b]
(4.6.5) o(
(w;d)
) =
1
2

d
2
n
0
n
1
n
2
d

i<j
gcd(n
i
. n
j
)
n
i
n
j
+

i
gcd(d. n
i
)
n
i
1

.
Notice that this equation reduces to the well-known (cf. [GH78b]) genus formula
o =
1
2
(d 1)(d 2) for smooth curves of degree d in CP
2
when all the weights are
1.
144 4. FUNDAMENTALS OF ORBIFOLDS
We nish this section with the quasi-smoothness criterion for hypersurfaces
A
d
CP(w). We rst note that all linear cones are quasi-smooth. If a weighted
hypersurface is not a linear cone we get [IF00]
Theorem 4.6.16: Let A
d
CP
n
(w) be a general hypersurface which is not a
linear cone. Then A
d
is quasi-smooth if and only if for every non-empty subset
1 = i
0
. . . . . i
k1
0. 1. . . . . n either one of the two following conditions holds
(i) There exists a monomial .
m
0
i
0
.
m
k1
i
k1
H
0
(CP(w). O(d)).
(ii) For all j = 1. . . . . / there exist .
m0,
i
0
.
m
k1,
i
k1
.
e
H
0
(CP(w). O(d)).
where c

are / distinct elements.

Exercise 4.4: [Quasi-smoothness criterion for surfaces] Show that a general
hypersurface A
d
CP(n
0
. n
1
. n
2
. n
3
) is quasi-smooth if and only if all of the
following three conditions hold [IF00, JK01b]
(i) For each i = 0. . . . . 3 there is a , and a monomial .
m
i
i
.
j
H
0
(CP(w). O(d)).
Here , = i is possible.
(ii) If gcd(n
i
. n
j
) 1 then there is a monomial .
b
i
i
.
b
j
j
H
0
(CP(w). O(d)).
(iii) For every i. , either there is a monomial .
bi
i
.
bj
j
H
0
(CP(w). O(d)).
or there are monomials .
c
i
i
.
c
j
j
.
k
and .
d
i
i
.
d
j
j
.
l
H
0
(CP(w). O(d)) with
/. | = i. ,.
4.7. Seifert Bundles
Seifert bred 3-manifolds were originally introduced and studied by Seifert in
1932 [Sei32, ST80]. Roughly speaking, these are 3-manifolds which are unions
of circles in a very particular way. More precisely, one could start with a trivial
bred solid torus o
1
1
2
. Then a bred solid torus is a solid torus which is nitely
covered by a trivial solid torus. This can be obtained by cutting the trivial solid
torus along a disc, rotating the disc through the angle 2cj and gluing it back
together. A bred Klein bottle is a solid Klein bottle which is doubly covered by a
trivial solid torus. Up to bred homeomorphism there is only one such bred solid
Klein bottle.
Denition 4.7.1: A Seifert bred 3-space is a 3-manifold ` with a decomposition
of ` into circles called bres such that each circle in ` has a neighborhood in `
which is a union of bers and it is isomorphic to a bred solid torus or Klein bottle.
Seifert original denition [Sei32] excludes the Klein bottle case. The Denition
4.7.1 is due to Scott [Sco83] and in the world of 3-manifolds it has some obvious
advantages. Epstein proved that a compact 3-manifolds is a Seifert bred space
(in the above sense) if and only if it is bred by circles [Eps72]. Hence, one
can introduce the transverse space A with the canonical projection : ` A.
Clearly, A is an orbifold surface and `, in a certain sense, is an orbifold bundle
over A. However, this denition is perhaps too general from our point of view.
For example, it can be seen that if ` is a Seifert bred space then ` admits an
o
1
-action leaving the foliation invariant if and only if one can coherently orient all
the bers. Hence, in general a Seifert bred 3-space is not a principal o
1
-orbifold
bundle over a compact 2-dimensional orbifold surface. We will be slightly less
general in dening Seifert bred n-spaces for which both `
n
and A are (naturally)
oriented. On the other hand we want to be able to work with Seifert brations in
4.7. SEIFERT BUNDLES 145
arbitrary dimension. For the extensive discussion of the general theory of Seifert
bred 3-spaces we refer the interested reader to [Sco83].
Higher dimensional Seifert bred manifolds were investigated rst Orlik and
Weigreich [OW75] who observed that in many cases of interest Seifert bred mani-
folds correspond to holomorphic Seifert C

-bundles, and started to develop a general

theory of holomorphic Seifert G-bundles for any complex Lie group G. More re-
cently Kollar has generalized this construction and we closely follow his approach
[Kol04, Kol05]. First we give some preliminaries. Recall [GR84] that a complex
space A (all our complex spaces are reduced and second countable unless other-
wise stated) is holomorphically convex if for every innite discrete closed subset o
of A there is a holomorphic function whose set of values on o is unbounded. A
holomorphically convex space A is Stein if every compact complex subspace of A
is nite.
Denition 4.7.2: Let A be a normal complex space. A Seifert C

-bundle over
A is a normal complex space Y together with a holomorphic map 1 : Y A and
a C

-action on Y satisfying the following two conditions.

(i) The pre-image 1
1
(l) of any open Stein subset l of A is Stein and
C

-equivariant (with respect to the trivial action on A).

(ii) For every j A, the C

-action on the ber Y

p
= 1
1
(j). C

Y
p
Y
p
is C

-equivariantly biholomorphic to the natural C

-action on C

j
m
for
some : = :(j. YA). where j
m
C

denotes the group of :

th
roots of
unity.
The number :(j. YA) is called the multiplicity of the Seifert ber over j.
One can always assume that the C

-action is eective, that is, :(j. YA) =

1 for generic j A. Note that even when Y is smooth, A can have quotient
singularities.
A classication of Seifert C

-bundles for A a smooth manifold with H

1
(A. Z)
torsion free is given in [OW75]. The most general case in arbitrary characteristic
is discussed in detail in [Kol04]. We begin with the following description of Seifert
C

-bundles Let 1 : Y A be a Seifert C

-bundle. The set of points j A :

:(j. YA) 1 is a closed analytic subset of A. It can be written as the union
of Weil divisors 1
i
A and of a subset of codimension at least 2 contained in
Sing(A). The multiplicity :(j. YA) is constant on a dense open subset of each
1
i
. this common value is called the multiplicity of the Seifert C

-bundle over 1
i
;
denote it by :
i
= :(1
i
). We call the Q-divisor =

(1
1
mi
)1
i
the branch
divisor of 1 : Y A. We will write 1 : Y (A. ) to indicate the branch divisor.
Kollar shows that there is the following correspondence [Kol04]
Theorem 4.7.3: Let A be a normal reduced complex space with at worst quotient
singularities and =

(1
1
mi
)1
i
a Q-divisor. There is a one-to-one correspon-
dence between Seifert C

-bundles 1 : Y (A. ) and the following data:

(i) For each 1
i
an integer 0 /
i
< :
i
. relatively prime to :
i
. and
(ii) a linear equivalence class of Weil divisors 1 Div(A).
We now wish to relate Seifert bundles to orbibundles on orbifolds. First follow-
ing [Kol05] we consider the case when A is smooth as in [OW75]. First let [[
denote the support of the Q-divisor . and set :() = lcm(:
i
). Then Y `1
1
([[)
is an :()-sheeted covering space of the C

-bundle over A`[[ with Chern class

:()

[1] +

i
b
i
m
i
[1
i
]

. (Here for convenience we adopt the common convention of

146 4. FUNDAMENTALS OF ORBIFOLDS
identifying a line bundle [1] associated with a divisor with its Chern class c
1
([1]).)
This is a well-dened integral cohomology class since the :
i
divide :() for all i.
More generally if A is singular but satises the hypothesis of Theorem 4.7.3, then
the Seifert bundle 1 : Y (A. ) is determined uniquely by its restriction to the
smooth locus A
reg
. Now we dene the Chern class of the Seifert bundle as a rational
cohomology class of A which denes an integral class j

c
1
(YA) H
2
orb
(A. Z).
Denition 4.7.4: The Chern class of the Seifert bundle 1 : Y (A. ) is
dened as
c
1
(YA) = [1] +

b
i
m
i
[1
i
] H
2
(A. Q) .
where 1 is a Weil divisor as in Theorem 4.7.3.
The triple (1
i
. :
i
. /
i
) are often called orbit invariants of the Seifert bundle.
We shall see in Theorem 4.7.10 below that under fairly mild conditions a Seifert
bundle is uniquely determined by its Chern class. We are now ready for
Theorem 4.7.5: Every C

-Seifert bundle 1 : Y (A. ) with Y smooth is a

C

-orbibundle over the locally cyclic orbifold A = (A. ). Conversely, if (A. ) is

locally cyclic orbifold and 1 : Y (A. ) is a C

-orbibundle over A whose local

uniformizing groups inject into the group C

of the orbibundle, then 1 : Y (A. )

is a C

-Seifert bundle with Y smooth.

Proof. Let 1 : Y (A. ) be a Seifert C

-bundle with Y smooth. For j A

pick any n 1
1
(j) and a j
m
-invariant smooth hypersurface \
p
Y transversal
to 1
1
(j) for : = :(j. YA). Then
p
: \
p
l
p
= \
p
j
m
gives an orbifold
structure on A = (A. |) on A. The orbifold branch divisor coincides with the
branch divisor of the Seifert bundle. Since Y is smooth each l
p
is a quotient by the
cyclic group j
m
. Hence A = (A. ) is locally cyclic. Now consider the Weil divisor
1 Div(A) and the Q-divisor

b
i
mi
1
i
Div(A) Q. The Weil divisor 1 lifts
to Baily divisor

1 on A and since the :
i
are the ramication indices of a branch
divisor, the Q-divisor

bi
mi
1
i
also lifts to a Baily divisor 1 by Proposition 4.4.13.
So the Q-divisor 1 +

b
i
mi
1
i
lifts to a unique Baily divisor which by Proposition
4.4.14 corresponds to an orbibundle over A. For the converse one can simply retrace
the steps and use Lemma 4.2.8.
Our main interest in Seifert bundles involves the circle subgroup of C

. As a
real Lie group, C

= o
1
R. Since a Seifert bundle involves subgroups of o
1
only,
we can write Y = ` R. Then restricting 1 to ` we have
Denition 4.7.6: We call 1 : ` (A. ) a Seifert o
1
-bundle or simply a
Seifert bundle.
Clearly, Y and ` have isomorphic homology and homotopy groups, and Y is
smooth if and only if ` is smooth. So it follows that 1 : ` (A. ) is a principal
o
1
-orbibundle over the orbifold A = (A. ), and Theorem 4.7.5 becomes
Theorem 4.7.7: Every Seifert o
1
-bundle 1 : ` (A. ) with ` smooth is a o
1
-
orbibundle over the locally cyclic orbifold A = (A. ). Conversely, if (A. ) is lo-
cally cyclic orbifold and 1 : `(A. ) is a o
1
-orbibundle over A whose local uni-
formizing groups inject into the group o
1
of the orbibundle, then 1 : `(A. )
is a Seifert o
1
-bundle with ` smooth.
4.7. SEIFERT BUNDLES 147
In general, the total space of an orbibundle or Seifert bundle over an orbifold
is an again an orbifold. So we want to characterize the smoothness of ` or equiv-
alently Y in terms of algebro-geometric properties. For this we describe the divisor
class group of a quotient singularity. Let Dim
C
A = n, and pick a point r A such
that Y is smooth along 1
1
(r). Let

l. . be a local uniformizing system for A

with r l = (

l). The local uniformizing group is the cyclic group j

m
. We can
take

l to be the polydisc 1
n
C
n
in which case l = (

l) is biholomorphic to
1
n
j
m
. By diagonalizing the cyclic group action, we see that locally the irreducible
components of passing through r are the quotients of (some of) the coordinate
hyperplanes in 1
n
. By an analysis similar to that in the proof of Lemma 4.5.12
we can obtain equivalent varieties by removing the reection elements of j
m
which
amounts to writing : = : :
1
:
n
. where the :
1
. . . . . :
n
are pairwise relatively
prime and the multiplicities of the irreducible components of passing through r.
(We add the necessary number of 1s when there are fewer than n such components.)
The quotient

1
n
= 1
n
j
m
1
m
n
is smooth, so as varieties 1
n
j
m

=

1
n
j
r
and
the j
r
-action is xed point free outside a codimension 2 subset. So the local divisor
class group Cl(A. r) at r is the cyclic group j
r(x)
Z
r
. Given a Weil divisor 1
containing r, we can restrict it to the neighborhood l

= 1
n
j
r(x)
of r. This gives
a well dened map 1
x
: Cl(A)Cl(A. r) Z
r(x)
whose kernel is precisely the
subgroup of classes of locally principal divisors at r. Following Kollar we introduce
some useful notation:
1. :(r. ) = lcm(:
i
[ r 1
i
).
2. :() = lcm(:(r. ) [ r A) = lcm(:
i
).
3. :(r. A) = :(r),
4. :(A) = lcm(:(r. A)) [ r A).
5. :(r. A. ) = :(r).
6. :(A. ) = lcm(:(r. A. ) [ r A) = lcm(:(r) [ r A) = (A).
Note that :(r. A. ) = :(r. ):(r. A) and :(A. )[:():(A) but the latter
can be dierent. Also since :
i
divides :() for each i. the class :()c
1
(YA)
gives a well dened element of Cl(A). Similarly, :(A. )c
1
(YA) gives a well
dened element of PicA. Now we dene
(4.7.1) :(r. ) c
1
(YA) = :(r. )[1] +

i]xDi
:(r. )/
i
:
i
[1
i
]
and note that it is a well dened element of Cl(A). Kollar then gives the following
simple smoothness criterion [Kol05]:
Proposition 4.7.8: Let Y = Y (1.

bi
mi
1
i
) be a Seifert bundle over the orbifold
(A.

(1
1
mi
)1
i
). Then Y is smooth along 1
1
(r) if and only if 1
x
(:(r. )
c
1
(YA)) is a generator of the local class group Cl(A. r).
Let ` (A.

(1
1
m
i
)1
i
) be the associated Seifert o
1
-bundle. This, as we
remarked is a principal circle orbibundle over the orbifold (A.

(1
1
m
i
)1
i
). It is
easy to see that this is equivalent to the requirement that the orbifold uniformizing
groups inject into the structure group o
1
of the principal orbibundle.
Some fairly general information can be obtained for a general Seifert bundle
` (A.

(1
1
m
i
)1
i
) with a smooth total space. By Lerays Theorem there is
148 4. FUNDAMENTALS OF ORBIFOLDS
a Leray spectral sequence (in cohomology) whose 1
2
term is
(4.7.2) 1
i,j
2
= H
i
(A. 1
j
1

Z
M
) = H
i+j
(`. Z) .
Since every bre is an o
1
. the only non-vanishing derived functor sheaf is 1
1
1

Z
M
.
As with Corollary 4.3.8 it is straightforward to show
Proposition 4.7.9: Let ` (A.

(1
1
m
i
)1
i
) be a Seifert o
1
-bundle, and let
is any group in which (A) is invertible. Then
(i) there is an isomorphism 1
1
1

M

X
: hence, there is a Gysin sequence
H
p
(A.
X
)

H
p
(`.
M
)
j
p
H
p1
(A.
X
)

H
p+1
(A.
X
) .
(ii) there is a natural injection : 1
1
1

Z
M
Z
X
which is an isomorphism
on points where :(r) = 1.
(iii) the dierential d
2
: 1
0,1
= H
0
(A. 1
1
1

M
)H
2
(A.
X
) is identied
with cupping with c
1
(YA).
Even so, in general, it is not easy to compute the cohomology ring of the bundle,
or even understand the fundamental group
1
(`). However, assuming that ` is
smooth, from Theorem 4.3.18 we get an exact sequence
(4.7.3)
orb
2
(A. )Z
1
(`)
orb
1
(A. )1
which often allows for an explicit computation of H
1
(`. Z). In principle, the orb-
ifold invariants of (A. ) and the rst Chern class of the Seifert bundle should be
sucient to determine the cohomology of ` at least when H
orb
1
(A. Z) = 0. Indeed
using results in [OW75] Kollar [Kol04] proves
Theorem 4.7.10: Let A = (A. ) be a complex orbifold such that H
orb
1
(A. Z) = 0.
Then a Seifert C

-bundle (hence, a Seifert o

1
-bundle) 1 : Y (A. ) is uniquely
determined by its Chern class c
1
(YA) H
2
(A. Q).
Another important result of [Kol05] says that when H
orb
1
(A. Z) = 0 there are
no branch divisors on a Calabi-Yau orbifold.
Proposition 4.7.11: Let 1 : ` (A.

(1
1
mi
)1
i
) be a Seifert bundle, `
smooth and H
orb
1
(A. Z) = 0. Suppose that A is a Calabi-Yau orbifold, that is,
1
X
+

(1
1
m
i
)1
i
is numerically trivial. Then

(1
1
m
i
)1
i
= 0 and 1
orb

1
X
.
So it is important to have conditions for the vanishing of H
orb
1
(A. Z) or even
better for
orb
1
(A). Necessary and sucient conditions were worked out by Kollar
[Kol05] in the case of complex dimension 2, but before describing these we give a
result of importance to us.
Proposition 4.7.12: Let A = (A. |) be a quasi-smooth weighted complete inter-
section in CP(w) with dimA 2. Then
orb
1
(A) = 0.
Proof. This is a direct consequence of Proposition 9.6.1 below and the long
exact homotopy sequence 4.3.18 applied to the Seifert o
1
-bundle.
4.7.1. Seifert Bundles of Dimension Five. We now specialize to the case
of complex surfaces where much more is known, but rst we need a general denition
concerning singularities. We recall [KM98, Laz04a] that a normal variety \ has
rational singularities if there exists a resolution of singularities : \
t
\ such
that the derived functor sheaves 1
i

O
V
vanish for i 0.
4.7. SEIFERT BUNDLES 149
Proposition 4.7.13: Let A be a normal projective surface with rational singu-
larities, and let A = (A. ) be an orbifold structure with branch divisor =

i
(1
1
mi
)1
i
. Then H
orb
1
(A. Z) = 0 if and only if the following conditions hold:
(i) H
1
(A
reg
. Z) = 0. and
(ii) the map : H
2
(A. Z)

j
Z
m
j
dened by () = (1
j
) mod :
j
is
surjective.
If in addition H
1
(A. Z) = 0 and H
2
(A. Q) = Q. then H
orb
1
(A. Z) = 0 if
and only if
(iii) Cl(A) Z.
(iv) the :
j
are pairwise relatively prime, and
(v) for every , the degree of 1
j
is relatively prime to :
j
.
Again one can say more for projective surfaces; Kollar [Kol05] shows that in
this case the spectral sequence degenerates at the 1
3
term and one obtains:
Theorem 4.7.14: Let 1 : `
5
(A. =

i
(1
1
m
i
)1
i
) be a smooth Seifert
bundle over a projective surface with rational singularities. Set : = /
2
(A) and
assume that H
orb
1
(A. Z) = 0. Then the cohomology groups H
i
(`
5
. Z) are
i 0 1 2 3 4 5
H
i
(`
5
. Z) Z 0 Z
s1
Z
d
Z
s1

i
(Z
m
i
)
2g(D
i
)
Z
d
Z
where d is the largest natural number such that :()c
1
(YA) Cl(A) is divisi-
ble by d, and o(1) is the genus of the Riemann surface 1. (Here we employ the
convention that Z
1
= 0).
For details we refer to the literature [OW75, Kol05]. Theorem 4.7.14 says
that the torsion in H
3
(`
5
. Z) (or equivalently in H
2
(`
5
. Z) when H
1
(`
5
. Z) = 0)
is determined by the nonrational curves in the branch divisor . Indeed one has
an immediate corollary of Theorem 4.7.14 and Proposition 4.6.9 which was rst
obtained in [BG01b] by the authors using a result of Randell [Ran75].
Corollary 4.7.15: Let 1
f
be the link of an isolated hypersurface singularity de-
ned by a weighted homogeneous polynomial 1 in four complex variables, and let A
denoted the corresponding hypersurface in CP(n
0
. n
1
. n
2
. n
3
). Suppose further that
A is well-formed, then Tor (H
2
(1
f
. Z)) = 0.
We end this section by considering an example that illustrates some of the ideas
presented in this chapter.
Example 4.7.16: We build on our discussion of Example 4.4.9 where A = CP(1. 2. 3).
but with a dierent branch divisor as in Example 6.6 of [Kol05]. Here the branch
divisor has two irreducible components, namely =
1
2
1
0
+
10
11
1
t
. where 1
t
=
(.
6
0
+.
3
1
+.
2
2
= 0) and 1
0
is as in Example 4.4.9. Recall also that 1
0
is a generator
of Cl(CP(1. 2. 3)) Z. Let us compute 1
orb

of A = (A. ). First, the divisor 1

t
is
a member of the linear system [O(6)[, so 1
t
61
0
. Second, the canonical bundle
of CP(1. 2. 3) is O(6). so from Proposition 4.4.15 we have
1
orb

61
0
+
1
2
1
0
+
10
11
61
0
(6 +
1
2
+
60
11
)1
0

1
22
1
0
.
So A = (CP(1. 2. 3).
1
2
1
0
+
10
11
1
t
) is a Fano orbifold.
We consider the orbibundle over (A. ) with Chern class c
1
(YA) =
1
22
[1
0
].
To check the smoothness of Y or equivalently `
5
we need to determine the orbit
invariants.
150 4. FUNDAMENTALS OF ORBIFOLDS
Since Cl(A) Z. we can write 1 /1
0
for some integer /. Then using
Denition 4.7.4 we write
1
22
1
0
/1
0
+
1
2
1
0
+
/
11
(61
0
)
for 0 < / < 11. We get / = 1 and / = 1. Now A = CP(1. 2. 3) has precisely
two singular points, r
2
= [0. 1. 0]. and r
3
= [0. 0. 1] both of which lie in 1
0
but
not in 1
t
. So :(r
2
. ) = :(r
3
. ) = 2. and it follows from Equation 4.7.1 that
:(r
i
. )c
1
(YA)) = 21
0
+
2
2
1
0
= 1
0
. But 1
x
i
(1
0
) is a generator of Cl(A. r
i
)
for both i = 2. 3. and by Proposition 4.7.8 the total spaces `
5
and Y are smooth.
Using Theorem 4.7.14 we show that `
5
is a rational homology sphere with
H
1
(`
5
. Z) = 0 and H
2
(`
5
. Z) = Z
11
Z
11
. Now :() = lcm
i
(:
i
) = 22. so
:()c
1
(YA) = [1
0
] implying that d = 1. Furthermore, from the genus formula
4.6.5 we see that o(1
t
) = 1. whereas, 1
0
has o(1
0
) = 0. So H
2
(`
5
. Z) = Z
11
Z
11
.
CHAPTER 5
Kahler-Einstein Metrics
Recall that a Riemannian metric o on a manifold ` is Einstein if Ric
g
= o
for some constant . So there are three cases, positive, negative, and null, depend-
ing on whether is positive, negative, or zero, respectively. The reader is referred
to the Bible of Einstein geometry, namely Besse [Bes87], and to [LW99] for
more recent developments. It is worth mentioning that although there has been
an enormous amount of progress in producing new examples of Einstein metrics,
most of the main problems in Section I of the Introduction of [Bes87] still re-
main. The one notable exception is the ten year old result of Catanese and LeBrun
[CL97] which shows that in dimension 4n with n 2 there are smooth compact
manifolds that admit both positive and negative Einstein metrics. However, an
important problem that still remains is that there is still no known topological
obstruction to the existence of Einstein metrics in dimensions greater than four
where one has the well-known Hitchin-Thorpe inequality and its generalizations
[LeB96, Kot98b, LeB99, LeB01]. Indeed there has been much progress in our
understanding of Einstein geometry on four manifolds. The reader is referred to
the excellent expository articles of LeBrun [LeB99, LeB03] for a discussion about
what was known about Einstein metrics in dimension four circa 2003. Of particular
interest is the relationship between the existence of Einstein metrics and dierential
topology. This is developed further in Chapter 11 in the context of Sasaki-Einstein
geometry. However, in dimension greater than four very little is known about both
the existence and the moduli problem concerning Einstein metrics. At this stage
the best we can do is to produce examples. An important theme in Einstein ge-
ometry that is not touched on at all in this book is the notion of collapse and its
relation to the moduli space of Einstein metrics developed mainly by Anderson,
Cheeger, Colding, and Tian. See [Che03, CT06] and references therein.
The Einstein equations if Kahler geometry have immediate cohomological con-
sequences. We must have
[
g
] = [
g
] = 2c
1
(`)
so that the rst Chern class of ` cannot be indenite. From the outset that
excludes most compact Kahler manifolds from considerations. As discussed in the
last section of Chapter 3 Yaus proof of the Calabi Conjecture [Yau77, Yau78]
implies each compact complex manifold with c
1
(`) = 0 must have a Kahler Ricci-
at metric. However, in the two other cases when c
1
(`) < 0 and c
1
(`) 0 one
needs to consider a modied Monge-Amp`ere problem. The existence of a Kahler-
Einstein metric of negative scalar curvature on a compact manifold with c
1
(`) < 0
was proved contemporaneously and independently by and Aubin [Aub76, Aub78].
In the case of c
1
(`) 0 there are known obstructions to solving the Monge-Amp`ere
problem. In the 50ties the only known obstructions had to do with the theorem of
151
152 5. K

AHLER-EINSTEIN METRICS
Matsushima which says that if (`. J. o.
g
) is Kahler-Einstein then the Lie algebra
h(`) of the holomorphic vector elds on ` must be reductive [Mat57b]. This
led Calabi to conjecture that if h(`) = 0 there should be no obstructions. If true
that would have indeed provided a very eective way of deciding on whether or not
a given Fano manifold admits a Kahler-Einstein metric. It took years of research
and truly deep insights of Tian to show that, on one hand, this conjecture is true
for surfaces but, on the other hand, it fails in higher dimensions.
So the problem is to give necessary and sucient conditions for the class c
1
(`)
to admit a Kahler-Einstein metric. For compact manifolds this was done in 1997
by Tian [Tia97] who gave necessary and sucient analytic conditions in terms of
the properness of a certain functional. This is also related to a certain stability
criterion. Along the way Tian also exhibited the rst example of a smooth Fano
3-fold with no holomorphic vector elds which admits no Kahler-Einstein metric, as
well as proving that del Pezzo surfaces admit positive Kahler-Einstein metrics if and
only if the Lie algebra of innitesimal automorphisms is reductive [Tia90]. In spite
of all this, the problem of formulating necessary and sucient algebraic conditions
under which a compact Fano manifold (orbifold) admits an Einstein metric still
remains open. It is precisely nding algebraic conditions that is of upmost interest
to us. We refer the reader to Tians book [Tia00] and review article [Tia99] as well
as his original articles for a complete treatment (see also Chapter 7 of [Aub98]).
More generally, today one is interested in the larger class of the so-called ex-
tremal Kahler metrics (cf. [Tia00, Sim04] and references therein) which general-
ize both Kahler-Einstein metrics and Kahler metrics of constant scalar curvature.
We will discuss extremal metrics briey in Section 5.3. The subject of extremal
Kahler metrics and how they relate to various notions of geometric and algebraic
stability goes far beyond the scope of this book. Here we had to settle for a careful
selection of topics most relevant for the material of Chapter 11 with the focus on
the existence of Kahler-Einstein metrics on compact Fano manifolds and orbifolds.
5.1. Some Elementary Considerations
Here we concentrate on the case that o is both Kahler and Einstein in which
case we say that o is Kahler-Einstein. It follows from Proposition 3.6.1 that if
(`. J. o.
g
) is Kahler-Einstein then the rst Chern class must be proportional to
the cohomology class of the Kahler form, that is, c
1
(`) = [
g
]. so we have
Proposition 5.1.1: A necessary condition that a Kahler manifold admits a Kahler-
Einstein metric is that the rst Chern class c
1
(`) is either denite or null.
One often refers to such metrics as the canonical Kahler metrics (cf. [Tia00]).
There is an analogous statement for Kahler orbifolds A = (A. |). but here we must
use the orbifold rst Chern class c
orb
1
as dened in Theorem 4.3.14 which in general
is dierent than c
1
(A). Thus, we have
Proposition 5.1.2: A necessary condition that a Kahler orbifold A admits an
orbifold Kahler-Einstein metric is that the orbifold rst Chern class c
orb
1
(A) is
either denite or null.
There can exist Kahler-Einstein orbifold metrics on algebraic varieties A whose
rst Chern class c
1
(A) is indenite. Of course, in this case c
orb
1
(A) must be denite
or null.
5.2. THE MONGE-AMP
`
ERE PROBLEM AND THE CONTINUITY METHOD 153
We can consider the existence of Kahler-Einstein metrics with arbitrary Ein-
stein constant . By scaling we can assume that = 0. 1. Specically, let
(`. J. o.
g
) be a compact Kaher manifold. We would like to know if one can al-
ways nd a Kahler-Einstein metric / /
[
g
]
. Recall that on a Kahler-Einstein
manifold
g
=
g
. This implies that 2c
1
(`) = [
g
]. Now, if c
1
(`) 0 we
must have = +1 because [
g
] is the Kahler class, so that when ` is compact
[
g
] = 0. Similarly, when c
1
(`) < 0 the only allowable sign of a Kahler-Einstein
metric on ` is = 1. Clearly, when c
1
(`) = 0 we must have = 0 as [
g
] = 0.
As we have already pointed out the = 0 case follows from Yaus solution to the
Calabi conjecture.
Remark 5.1.1: Note that when ` is not compact one can have Kahler-Einstein
metrics on ` of any sign. The simplest example here is ` = C
n
. First we can
take the at metric which is Einstein with = 0. Then we can identify ` with a
complex hyperbolic ball and put the Bergman metric on it with = 1. Finally,
we can take ` = C
n
CP(n) with the Fubini-Study metric and = +1. The rst
two metrics are complete, the third is not. One cannot have a complete metric of
positive Ricci curvature on a non-compact manifold.
5.2. The Monge-Amp`ere Problem and the Continuity Method
This section is a continuation of Sections 3.6 and 4.4.4. Let (A. o.
g
) be
a Kahler orbifold with underlying complex space A and Kahler class [j

g
]
H
2
orb
(A. Z). Let us reformulate the existence problem using the global i

-lemma.
Suppose there exists an orbifold Einstein metric / /
[g]
. Starting with the original
Kahler metric o on A we have a globally dened function 1 C

(A. R)

that is
invariant under the local uniformizing groups such that

g

g
= i

1 .
As before we will call 1 a discrepancy potential function. We also x the constant
by asking that

X
(c
f
1)dvol
g
= 0. Let / /
[]
be an Einstein metric for which

h
=
h
. Using global i

-lemma once again we have a globally dened invariant

function C

(A. R)

such that
h

g
= i

. We shall x the constant in

later. Using these two equations we easily get

g

h
= i

(1 ) .
Dening 1 so that
n
h
= c
F

n
g
we can write this equation as
i

1 = i

(1 ) .
This implies that 1 = 1 + c. We have already xed the constant in 1 so
c depends only on the choice of . We can make c = 0 by choosing such that

X
(c
f
1)dvol
g
= 0. Hence, using Theorem 4.4.31, Theorem 3.6.6 becomes
Proposition 5.2.1: Let (A. ) be a compact Kahler orbifold with c
orb
1
0. where
= 1. or with c
orb
1
= 0 and = 0. Let [] H
2
orb
(A. Z) be a Kahler class and
o. / two Kahler metrics in /
[]
with Ricci forms
g
.
h
. Let 1. C

(A. R)

be
dened by
g

g
= i

1.
h

g
= i

1. Fix the relative constant of 1

by setting

X
(c
f
1)dvol
g
= 0. Then the metric / is Einstein with Einstein
constant if and only if satises the following Monge-Amp`ere equation

n
h
= c
f

n
g
.
154 5. K

AHLER-EINSTEIN METRICS
which in a local uniformizing complex chart (l; .
1
. . . . . .
n
) is written as
(5.2.1)
det

o
i

j
+

2

zi zj

det(o
i

j
)
= c
f
.
Note that by setting = 0 we get the Monge-Amp`ere equation for the original
Calabi problem described in Theorems 3.6.5 and 3.6.6. The character of the Monge-
Amp`ere equation above very much depends on the choice of . The case of < 0 is
actually the simplest, as the necessary a priori C
0
-estimates can be derived using
the Maximum Principle. This was done by Aubin [Aub76] and independently by
Yau [Yau77, Yau78]. As discussed previously the extension to the case of compact
orbifolds is straightforward, so we have
Theorem 5.2.2: Let (A. J. o.
g
) be a compact Kahler orbifold with c
orb
1
< 0 or
c
orb
1
= 0. Then there exists a unique orbifold Kahler-Einstein metric / with [
h
]
such that
h
=
h
.
The case of 0 is more dicult and it has been known for quite some time
that there are actually non-trivial obstructions to the existence of Kahler-Einstein
metrics which we discuss in the next section. For ease of discussion we rst treat the
positive case for smooth manifolds. The rst important step toward understanding
the c
1
(`) positive case was taken by Aubin [Aub84] who showed how to apply
the continuity method in this case. Rewrite equation (5.2.1) with a parameter
t [0. 1] :
(5.2.2) det

o
i

j
+

2

.
i
.
j

= det(o
i

j
)c
ft
subject to the condition
(5.2.3) o
t
i

j
= o
i

j
+

2

.
i
.
j
0
which denes a Kahler metric on `. Assuming that a solution (t) of Equation
(5.2.2) exists at t. we get a Kahler metric o
t
t
and its 2-form
t
t
. We set o
t
0
= o. and
let
t
t
be the Ricci form of o
t
t
. then a calculation shows that
(5.2.4)
t
t
= (1 t) +t
t
t
.
Now we dene the set
(5.2.5) o = t [0. 1] [ Equation (5.2.2) has a solution (t) .
By Theorem 3.6.6 (0) is a solution, so o is non-empty. If o = [0. 1] then Equation
(5.2.4) gives us a Kahler-Einstein metric at t = 1. So to prove the existence of a
Kahler-Einstein metric it suces to show that the set o is both open and closed in
[0. 1]. That o is open is an inverse function theorem argument [Aub84, Aub98].
The problem then is to nd necessary and sucient conditions that guarantee that
o is closed. Such necessary and sucient analytic conditions were found by Tian
in 1997 [Tia97] in terms of the properness of a certain functional; however, such
conditions are often hard to implement. The applications we have in mind in this
book require algebraic conditions that are easy to give.
The rst steps toward nding sucient conditions for guaranteeing the exis-
tence of a Kahler-Einstein metric were taken by Tian [Tia87] and independently
5.2. THE MONGE-AMP
`
ERE PROBLEM AND THE CONTINUITY METHOD 155
by Siu [Siu88]. Let us dene
1
1(`. o) =

C
2
(`) [ positivity (5.2.3) holds.

M
dvol
g
= 0

.
Then Tian [Tia87] shows that there exist positive constants and C depending
only on ` and o such that
(5.2.6)

M
c

dvol
g
C
for all functions 1(`. o). This gives a uniform bound
(5.2.7) (`) = sup
>0

C 0 such that (5.2.6) holds for all 1(`. o)

.
It is easy to see that (`) depends only on the cohomology class [
g
]. and is
invariant under biholomorphisms. Then we have [Tia87] (also [Siu88]):
Theorem 5.2.3: Let (`. J. o.
g
) be a compact Kahler manifold of complex di-
mension n with c
1
(`) 0. If
g
c
1
(`) and (`)
n
n+1
. then ` admits a
Kahler-Einstein metric in the same Kahler class as
g
.
In certain cases Tian is able to obtain a lower bound for (`) as follows.
Proposition 5.2.4: Let (`. J. o.
g
) be a compact Kahler manifold having fami-
lies of curves C
1

. . . . . C
N

. where CP
n
is the parameter, and subvarieties
o
1
. . . . . o
N
such that
(i) o
1
o
N
= .
(ii) ``o
j
=

C
j

(``o
j
)

. C
j

C
j

(``o
j
) = . and C
j

(`` o
j
)
is smooth for each .
(iii) for all . ` `
i
o
i
. T
z
C
j
i
[ C
j
j
. spans T
z
`. and for all . o
i
,
either T
z
C
j
j
[ . C
j
j
(``o
j
) spans T
z
`, or there exists C
j
j
such
that . C
j
j
(` ` o
j
) with C
j
j
o
i
= nite number of points.
(iv) for all i = 1. . . . . and all CP
n
. 4vol
g
(C
i

) .
Then (`)
4

.
It is easy to generalize these results to orbifolds, and we give orbifold examples
later where this inequality holds. In order to apply Proposition 5.2.4 successfully
to prove the existence of Kahler-Einstein metrics one needs to invoke some symme-
tries. So one considers a subgroup G of the group of automorphisms of the Kahler
structure, and denes the subset 1
G
(`. o) = 1(`. o) [ is G invariant.
Then we dene
(5.2.8)
G
(`) = sup
>0

C 0 such that (5.2.6) holds for all 1

G
(`. o)

.
Clearly,
G
(`) (`). Then Theorem 5.2.3 has a G-invariant version, namely
Theorem 5.2.5: Let (`. J. o.
g
) be a compact Kahler manifold of complex di-
mension n with c
1
(`) 0. and let G be a compact subgroup of the group of Kahler
automorphisms. If
g
c
1
(`) and
G
(`)
n
n+1
. then ` admits a Kahler-
Einstein metric in the same Kahler class as
g
.
It is this G-invariant version that is most eective for proving the existence of
Kahler-Einstein metrics on smooth manifolds.
1
Our normalization

M
dvol
g
= 0 which follows [Aub98] is dierent from Tians.
156 5. K

AHLER-EINSTEIN METRICS
Example 5.2.6: [Fermat hypersurfaces] Both Tian and Siu showed how to prove
the existence of Kahler-Einstein metrics on Fermat hypersurfaces. Here we follow
Tian [Tia87]. Consider the Fermat hypersurface 1
n,p
dened by
.
p
0
+ +.
p
n
= 0
in CP
n
for n 3. First, by adjunction theory 1
n,p
is Fano if and only if n+1j 0.
The hypersurface 1
n,p
is invariant under the group G generated by the sym-
metric group
n+1
acting as permutations of the n + 1 homogeneous coordinates
.
0
. . . . . .
n
and n + 1 copies of the cyclic group C
p
Z
p
acting by .
i

i
.
i
. where

i
C
p
. Now in the presence of symmetries we can obtain an improved lower bound.
We need to replace item (iv) in Proposition 5.2.4 by
(iv) Let G
j
G be the subgroup that preserves the bration of ``o
j
by the
curves C
j

(` ` o
j
). then o
j
is invariant under G
j
and the inequality
4vol
g
(C
i

)
ord(G
j
)
.
where ord(G
j
) = min
z(M\S
j
)
]G
j
]
]StabzGj]
. where Stab
z
is the isotropy
subgroup at . in G
j
.
Then
G
(1
n,p
)
4

. This makes smaller giving a better bound for

G
(1
n,p
).
Tian (see [Tia87] for details) then obtains the estimate
4(n+1p)
2
. So using
Theorem 5.2.5, we get the existence of Kahler-Einstein metrics if
n 1
n

G
(`)
4

2
(n + 1 j)
holds. That is, we get Kahler-Einstein metrics for j = n 1. n. Nadel [Nad90]
has improved this result by proving the existence of Kahler-Einstein metrics on
1
n,p
for
n
2
j n using multiplier ideal sheaves. We should also mention that the
complex quadric 1
n,2
is an Hermitian symmetric space (the oriented Grassmannian
G
+
2
(R
n+1
)) that is known to admit a Kahler-Einstein metric [KN69].
Let us briey describe what happens if we fail to obtain the uniform bound
(5.2.6). Then there exists an increasing sequence t
k
with t
k
< 1 such that
(5.2.9)

k
dvol
g
as /
for all (
n
n+1
. 1). We now introduce the concept of a multiplier ideal sheaf
[Nad89, Nad90].
Denition 5.2.7: Let 1() denote the sheaf associated to the presheaf dened by
associating to each open set l ` the set of sections

1 O
M
(l) [

U
[1[
2
c

dvol
g
<

.
We can now state Nadels Theorem [Nad90]:
Theorem 5.2.8: Let (`. J. o.
g
) be a compact Kahler manifold of complex dimen-
sion n with c
1
(`) 0. and let G be a compact group of automorphisms. If ` does
not admit a G-invariant multiplier ideal sheaf 1() = O
M
for all (
n
n+1
. 1).
then ` admits a Kahler-Einstein metric.
5.2. THE MONGE-AMP
`
ERE PROBLEM AND THE CONTINUITY METHOD 157
The point is that if one can nd a sequence satisfying Equation 5.2.9 then there
is a non-trivial multiplier ideal sheaf 1() = O
M
. otherwise 1() = O
M
. and the
set o of Equation 5.2.5 becomes the whole interval [0. 1] giving a Kahler-Einstein
metric at t = 1. We refer to the literature [Aub98, DK01, Nad90] for further
discussion and proofs.
Once one has some familiarity working with orbifolds, the generalization of
these ideas to orbifolds is straightforward. Here computations are done on the
local uniformizing covers (

l. ) working with -invariant objects. So one works

with orbisheaves as discussed in Chapter 4. Let O

denote the structure orbisheaf

of Denition 4.2.2. One now writes down Equations (5.2.2)-(5.2.5) on the local
uniformizing neighborhoods with all objects being -invariant. Then a multiplier
ideal orbisheaf is:
Denition 5.2.9: Let 1() denote the orbisheaf associated to the presheaf dened
by associating to each local uniformizing open set (

l. ) the set of sections

1 O

l)

U
[1[
2
c

dvol
g
<

.
The orbifold version of Theorem 5.2.8 was given by Demailly and Kollar [DK01].
Theorem 5.2.10: Let A be a compact Kahler orbifold of complex dimension n
with c
orb
1
(A) 0. and let G be a compact group of automorphisms. If A does not
admit a G-invariant multiplier ideal orbisheaf 1() = O

for all (
n
n+1
. 1) .
then A admits a Kahler-Einstein orbifold metric.
Remark 5.2.1: Recall from Equation 4.4.2 that c
orb
1
(A) and c
1
(A) are dierent in
the presence of branch divisors. So positive Kahler-Einstein orbifold metrics can
exist on algebraic varieties that are not Fano. They are, however, orbifold Fano.
Multiplier ideal sheaves can also be described in terms of log resolutions of
singularities as we now explain [Dem96, Dem01, Kol97, KM98, Laz04b]. The
idea is that the uniform boundedness of the integrals (5.2.9) can be expressed in
terms of the mildness of the singularities of a pair (A. 1). where A is a normal
variety and 1 is a Q-divisor on A. However, in order to dene log resolutions, we
need to make sense out of pulling back divisors by a map. So we assume that A
is Q-factorial. Thus, Weil Q-divisors are Q-Cartier, so their pullback by a map is
well dened. This causes no problem since we are interested in algebraic varieties
A that are the underlying complex space of a complex orbifold and Propositions
1.1.5 and 4.4.7 guarantee that such an A is both normal and Q-factorial.
Denition 5.2.11: A divisor 1 =

i
1
i
is said to have simple normal cross-
ings if each 1
i
is smooth and 1 is dened in a local neighborhood chart (l; .)
by equations of the form .
1
.
k
= 0 for some / n. A Q-divisor

i
o
i
1
i
has
simple normal crossings if the divisor

i
1
i
has simple normal crossings.
Denition 5.2.12: Let 1 be a Q-divisor on A. A log resolution of the pair
(A. 1) is a projective birational map j : A
t
A with A
t
smooth such that the
divisor j

1 +

i
1
i
has simple normal crossings. Here 1
i
are the exceptional
divisors.
We rst describe the multiplier ideal sheaves associated with the underlying
variety. Given a log resolution j : A
t
A of (A. 1) we can dene a multiplier
ideal sheaf 1(1) by
(5.2.10) 1(1) = j

O
X

1
X
j

1
X
+1|

O
X
.
158 5. K

AHLER-EINSTEIN METRICS
where 1| denotes the integral part of 1. It follows from a theorem of Esnault and
Viehweg (cf. [Laz04b], pg 156) that 1(1) is independent of the log resolution.
In the case of an orbifold A = (A. |) we replace the variety A by a local
uniformizing neighborhood :

ll. and if the orbifold has a non-trivial branch
divisor then we must replace the canonical divisor

1
U
with the orbifold canon-
ical divisor 1
orb

U
=

(1
U
+). This gives rise to a multiplier ideal orbisheaf 1(1)
dened as the sequence of sheaves associated to the presheaves which associates to
each local uniformizing neighborhood

l the sheaf
(5.2.11) 1(1)(

l) = j

O
X

1
X
j

(1
orb

U
+

1)

U
.
We now make the connection between the orbisheaf 1() of Denition 5.2.9
and the orbisheaf dened by Equation (5.2.11). Let 1 =

i
o
i
1
i
be an eective
-invariant Q-divisor on a local uniformizing neighborhood

l. and assume that
1
i
is principal, that is, determined by a -invariant holomorphic dening function
o
i
. Then the function
D
=

i
o
i
log [o
i
[ is plurisubharmonic and locally denes a
multiplier ideal sheaf
1(
D
)

U
=

1 O
X
(

l)

U
[1[
2

[o
i
[
2a
i
dvol
g
<

on

l. Now the o
i
can be taken as the coordinate functions of a local coordinate
chart. So the condition that the integral be locally nite is that the holomorphic
function 1 be divisible by

[o
i
[
m
i
. where :
i
is an integer satisfying :
i
o
i
1
for each i. Note that since o
i
Q this means that :
i
o
i
|. By Theorem 9.3.42
of [Laz04b] the multiplier ideal sheaf 1(
D
)

U
is equal to the analytic extension
1(1)
an
(

l) of the algebraic multiplier ideal sheaf 1(1)(

l) on each uniformizing
neighborhood

l. Thus, the corresponding multiplier ideal orbisheaves 1(
D
) and
1(1)
an
coincide. Hereafter, we shall use these two orbisheaves interchangeably and
denote the orbisheaf by 1(A. 1).
Denition 5.2.13: Let 1 be an eective Q-divisor on an orbifold A = (A. |).
Then
(i) The pair (A. 1) is Kawamata log terminal or klt if 1(A. 1) = O

.
(ii) The pair (A. 1) is log canonical if 1

A. (1 c)1

= O

for all 0 <

c < 1.
An alternative formulation of Denition 5.2.13 can be obtained by noticing that
the condition 1(A. 1)(

l) = O

U
is equivalent to the condition
(5.2.12) ord
E

1
X
j

1
U
+ +1|

1
for every exceptional divisor in the log resolution j : A
t

l. This gives rise to the

denition used in [BGN03b] (and also in [KM98]), namely
Denition 5.2.14: The pair (A. 1) is klt or Kawamata log-terminal) if for
each local uniformizing neighborhood

l there exists a log resolution of singularities
j : A
t


l such that
1
X

n
j

(1
orb

U
+

1) +

o
i
1
i
with o
i
1 for all j-exceptional 1
i
. If o
i
1 then the pair (A. ) is said to
have log-canonical singularities.
5.2. THE MONGE-AMP
`
ERE PROBLEM AND THE CONTINUITY METHOD 159
Clearly, a klt pair (A. 1) is log-canonical. It is now a simple matter to refor-
mulate Theorems 5.2.8 and 5.2.10 in terms of the klt condition. For example, we
have
Theorem 5.2.15: Let (A. J. o. ) be a compact Kahler orbifold of complex di-
mension n with c
orb
1
(A) 0. and let G be a compact group of automorphisms
on (A. J. o. ). Assume that there is an c 0 such that (A.
n+
n+1
1) is klt for
every G-invariant eective Q-divisor 1
n
1
orb

. Then A has an orbifold Kahler-

Einstein metric.
This theorem has given some rather startling results which we describe in Chap-
ter 11 [JK01b, BGN03b, BGK05] even for the case G = id. Let us rephrase
this more explicitly in terms of holomorphic sections for the case G = id [DK01].
Theorem 5.2.16: Let (A. J. o. ) be a compact Kahler orbifold of complex dimen-
sional n such that the orbifold anticanonical divisor 1
1

is ample. The continuity

method produces a Kahler-Einstein orbifold metric / on A if there is a
n
n+1
such that for every : 1 and for every holomorphic section
s
H
0
(A
orb
. O(1
s

))
we have [
s
[

s
1
2
(A). that is

[
s
[

2
s
dvol
g
< +.
When applying this theorem we will need to check the 1
2
condition along
certain divisors H
i
. This is accomplished by reducing the problem to an analogous
problem on H
i
and using induction. In algebraic geometry, this method is called
inversion of adjunction. Conjectured by Shokurov, the following version is due
to Kollar [Kol92, 17.6]. It was observed by [Man93] that it can also be derived
from the 1
2
-extension theorem of Ohsawa and Takegoshi [OT87]. See [Kol97] or
[KM98] for more detailed expositions.
Theorem 5.2.17 (Inversion of adjunction): Let ` be a smooth manifold, H `
a smooth divisor with equation (/ = 0) and o a holomorphic function on `. Let
o
H
denote the restriction of o to H and assume that it is not identically zero. The
following are equivalent:
(i) [o[
c
[/[
1
is locally 1
2
near H for every 0.
(ii) [o
H
[
c
is locally 1
2
on H.
We end this section by briey describing the situation for complex surfaces. In
this case Tian [Tia90] has given a classication of all smooth compact complex
surfaces that admit a positive Kahler-Einstein metric. Recall from Example 3.5.11
that the del Pezzo surfaces are precisely CP
2
. CP
1
CP
1
. or CP
2
#/CP
2
. where
1 / 8. Now there are well-known obstructions to the existence of positive
Kahler-Einstein metrics as discussed in Section 5.3 below. In particular, Example
5.3.2 says that CP
2
#/CP
2
with / = 1 or 2 cannot admit a Kahler-Einstein metric.
However, Tians theorem says that these are the only smooth compact complex
del Pezzo surfaces which do not admit a Kahler-Einstein metric. Specically, Tian
[Tia90] proved
Theorem 5.2.18: A smooth compact complex surface ` with c
1
(`) positive ad-
mits a Kahler-Einstein metric if and only if the Lie algebra of the automorphism
group of the complex structure is reductive.
160 5. K

AHLER-EINSTEIN METRICS
Somewhat earlier Tian and Yau [TY87] had shown that for each smooth man-
ifold CP
2
#/CP
2
with 3 / 8 there is a complex structure admitting a Kahler-
Einstein metric, and independently Siu [Siu88] proved the existence of a Kahler-
Einstein metric on CP
2
#3CP
2
. We mention that for 5 / 8 there are non-trivial
moduli of Kahler-Einstein metrics. See Exercise 5.3 below.
5.3. Obstructions in the Positive Case
The rst types of obstructions that were found involved the Lie algebra of
holomorphic vector elds. Let h(`) be the complex Lie algebra of all holo-
morphic vector elds on `. Matsushima [Mat57a] proved that on a compact
Kahler-Einstein manifold with c
1
(`) 0. h(`) must be reductive, i.e., h(`) =
7(h(`)) [h(`). h(`)]. This result was generalized by Lichnerowicz [Lic57] and
then later by Calabi [Cal85]. For convenience here we give the Lichnerowicz ver-
sion. The more general result of Calabi will be generalized further to certain trans-
verse Kahler structures in Theorem 11.3.1 below. A detailed proof of Calabis
theorem appears in [Sim04].
Theorem 5.3.1: Let (`. J. o.
g
) be a compact Kahler manifold of constant scalar
curvature. Then h(`) = a(`) h
0
(`). where a(`) is the subalgebra of h(`)
consisting of all parallel vector elds and h
0
(`) is the complexication of the Lie
algebra of Killing vector elds. In particular, if (`. J) admits a Kahler-Einstein
metric, then h(`) is reductive.
For further discussion and proof we refer the reader to [Kob72] and [Bes87].
Example 5.3.2: For an easy application of this theorem consider `
k
= CP
n
blown
up at / points in general position, where 1 / n. It is easy to see that the
Lie algebra h(`
k
) is not reductive; hence, `
k
cannot admit a Kahler-Einstein
metric. For details in the n = 2 case see, for example, [Fut88] or [Tia00]. The
generalization to n dimensions is straightforward.
Theorem 5.3.1 was generalized by Calabi [Cal85] to the case of extremal met-
rics. Extremal Kahler metrics are critical points of the energy functional
c(o) =

M
:
2
g
dvol
g
.
where :
g
is the scalar curvature of the Kahler metric o. Solving the Euler-Lagrange
equations for this functional shows that a Kahler metric o is extremal if and only
if the vector eld o:od
g
:
g
is holomorphic. Thus, Kahler metrics of constant scalar
curvature are extremal. In particular, Kahler-Einstein metrics are extremal. How-
ever, in the more general case of non constant scalar curvature extremal metrics,
the Lie algebra h
0
(`) is more complicated. See, for example, [Fut88] or [Sim04]
for details and proofs.
The next type of obstruction again involves the Lie algebra h(`) and was
found by Futaki [Fut83]. We follow the presentation in [Sim04]. Given a Kahler
metric on ` or equivalently 2-form . its Ricci form represents the rst Chern
class c
1
(`). up to a factor of 2. But by Hodge theory has a unique harmonic
representative
H
. and by Lemma 3.1.15 we can write
(5.3.1) =
H
+i

5.3. OBSTRUCTIONS IN THE POSITIVE CASE 161

for some real function

. called the Ricci potential. Then the Futaki invariant or

character is a map F : h 1(`)C dened by
(5.3.2) F(A. []) =

M
A(

)dvol

.
One can check that F only depends on the cohomology class []. and not on the
specic Kahler metric or form . Furthermore, it is easy to see that for any two
holomorphic vector elds A. Y. we have F([A. Y ]. []) = 0. The main result on
Futaki invariant is
Theorem 5.3.3: Let (`. J. o.
g
) be a compact Kahler manifold, and suppose that
the o is an extremal metric. Then o has constant scalar curvature if and only if
F(A. []) = 0 for all holomorphic vector elds A.
We are particularly interested in the case that c
1
(`) is positive and the coho-
mology class [] the Kahler metric is proportional to c
1
(`). Then after an appro-
priate scaling, Equation (5.3.1) becomes
(5.3.3) = +i

.
In this case an extremal Kahler metric with F(A. []) = 0 is a Kahler-Einstein
metric.
Example 5.3.4: [Fut88] Consider CP
m
1
CP
m
2
and let H
i
denote the hyperplane
bundle over CP
mi
. Let
i
: CP
m1
CP
m2
CP
mi
be the natural projection, and
let `
m1,m2
be the total space of the projectivization CP(

1
H
1

2
H
2
). Then
c
1
(`
m1,m2
) is positive, and the connected component of the automorphism group
of `
m
1
,m
2
is the projectivized group P(G1(:
1
. C)G1(:
2
. C)). So the Lie algebra
of holomorphic vector elds is reductive. Let A be an innitesimal generator of
the kernel of the projection map P(G1(:
1
. C) G1(:
2
. C))P(G1(:
1
. C))
P(G1(:
2
. C)). Then generally F(A. []) = 0. For example Futaki [Fut88] shows
that for `
1,2
. F(A. []) = 90. Thus, `
1,2
does not admit a Kahler-Einstein
metric.
In particular, the non-vanishing of F is an obstruction to the existence of Kahler-
Einstein metrics. Just as Matsushima theorem the Futaki invariant provides ob-
structions only in the presence of holomorphic vector elds. Another conjecture
of Calabi (formulated long before Futakis papers) speculated that Kahler-Einstein
metrics should exist on a Kahler manifold with c
1
(`) 0 and no holomorphic
vector elds. This was shown, however, to be false, rst in the category of orbifolds
by Ding and Tian [DT92], and nally Tian [Tia97] exhibited a smooth Fano 3-fold
with no holomorphic vector elds and no Kahler-Einstein metric (see also [Tia00]).
In order to understand the existence and obstructions in the absence of any holo-
morphic vector elds Tian constructs a generalized Futaki character [Tia97]. He
then formulates an alternative to the false Calabi conjecture: vanishing of his
generalized obstruction should now be necessary and sucient. All that quite nat-
urally leads to the study of the various notions of geometric/algebraic stability.
Simplifying a bit, in the case of the canonical Kahler metrics, in essence one deals
with the following fundamental questions:
Open Problem 5.3.1: Let (`. J. o.
g
) be a compact Kahler manifold (orbifold)
with positive (orbifold) rst Chern class. What are the necessary and sucient
conditions for ` to admit an (orbifold) Kahler-Einstein metric? In case ` is a
162 5. K

AHLER-EINSTEIN METRICS
complex algebraic variety, can such conditions be understood/formulated exclu-
sively in terms of the algebraic-geometric properties of `? Finally, given such `
can one eectively determine if ` can admit a Kahler-Einstein metric or not?
Obstructions to the existence of Kahler-Einstein orbifold metrics on log Fano
orbifolds will be discussed later within the context of Sasakian geometry in Sections
11.3.2 and 11.3.3.
5.4. Kahler-Einstein Metrics on Hypersurfaces in CP(w)
The algebraic geometry of hypersurfaces and complete intersections in weighted
projective spaces CP(w) was discussed in Section 4.6 where they were rst intro-
duced. In particular, they are all examples of compact Kahler orbifolds (manifolds).
In this section we shall examine some applications of the general existence results
discussed in Section 5.2 when applied to this specic case.
5.4.1. Smooth Hypersurfaces in CP(w). Let us begin by considering ex-
amples of smooth embedded hypersurfaces in CP(w). For convenience we mainly
treat the case dimension two, that is, complex surfaces. First we recall the well-
known work of Freedman (cf. [GS99]) which says that simply connected smooth
compact 4-manifolds ` are determined up to homeomorphism by their intersection
form Q : H
2
(`. Z) H
2
(`. Z)Z. In fact, Freedman proved that for every uni-
modular symmetric bilinear form Q there exists a compact simply connected topo-
logical 4-manifold whose intersection form is Q. Unfortunately, in the case of even
intersection form it is not known exactly which correspond to smooth 4-manifolds.
There is an outstanding conjecture, known as the
11
8
-conjecture, that if true, would
solve the problem. The case of odd intersection form is relatively easy. All sim-
ply connected smooth 4-manifolds with odd intersection form are homeomorphic
to /
+
2
CP
2
#/

2
CP
2
. where /

2
is the dimension of the maximal positive (negative)
subspace of Q on H
2
(`. Z). Since the pioneering work of Donaldson and Freedman
in the early eighties, the study of the dierential topology of 4-manifolds has been
a major theme in modern mathematics. We refer the reader to the monographs
[DK90, FM94, GS99, Sco05] for a detailed treatment of this very interesting
subject.
Example 5.4.1: We continue the discussion of the Fermat hypersurfaces 1
n,p

CP
n+1
of Example 5.2.6. As mentioned there 1
n,p
are Fano if n + 1 j and are
known to admit a Kahler-Einstein metric if in addition j
n
2
. Furthermore, when
j n+1 the hypersurfaces 1
n,p
also admit a Kahler-Einstein metric by the Aubin-
Yau Theorem 5.2.2. For simplicity we consider now the case of complex surfaces,
i.e., n = 3. Here there are two del Pezzo surfaces, j = 2 which is a complex quadric
biholomorphic to CP
1
CP
1
and it is well-known that this admits a Kahler-Einstein
metric. Of course, this is dieomorphic to o
2
o
2
and its intersection form Q is
given by the matrix H =

0 1
1 0

. The case j = 3 is known to be CP

2
blown up at
6 points, that is, 1
3,3
= CP
2
#6CP
2
which is also known to admit Kahler-Einstein
metrics by Tian and Yau [TY87]. Its intersection form Q = '1` 6'1`. The case
j = 4 gives the well-known 13 surface which admits a Ricci at metric by Yaus
result 5.2.2. Its intersection form is 2(1
8
) 3H. where 1
8
is the Cartan matrix
for the simple Lie algebra 1
8
. Generally for the Fermat surface 1
3,p
the intersection
5.4. K

AHLER-EINSTEIN METRICS ON HYPERSURFACES IN CP(w) 163

form Q is |
p
'1` :
p
'1` for j odd and |
p
H n
p
(1
8
) for j even, where
|
p
=
1
3
(j
3
6j
2
+11j3). :
p
=
1
3
(j1)(2j
2
4j+3). n
p
=
1
24
j(j
2
4) .
So, for example, the Fermat surface 1
3,5
is homeomorphic to 9CP
2
#44CP
2
. whereas,
1
3,6
is homeomorphic to 413#9(o
2
o
2
). For j 5 the Fermat surfaces 1
3,p
are
simply connected complex surfaces of general type all of which admit negative
Kahler-Einstein metrics. The general method will be discussed in more detail in
Example 5.4.2 below. See also [DK90, GS99].
Next we consider some complex algebraic manifolds that are hypersurfaces in
the more general weighted projective spaces CP(w). According to Proposition 9.3.22
any Brieskorn-Pham polynomial with pairwise relatively prime weights describes a
smooth projective algebraic variety in CP(w). Again for simplicity we consider the
surface case. The Table below gives the examples of all del Pezzo surfaces A
w
that
can be written as smooth hypersurfaces in weighted projective spaces. All of these
are known to admit positive Kahler-Einstein metrics [TY87]. We only indicate a
particular polynomial, as the most general such hypersurface involves many such
monomials which for reasons of space we do not include. We shall discuss this
further in Section 5.5.
Table 1: del Pezzo surfaces as smooth hypersurfaces in CP
3
(w)
|w| d w f
w
d X
w
1 (1,1,1,1) z
3
0
+ z
3
1
+ z
3
2
+ z
3
3
3 CP
2
#6CP
2
1 (1,1,1,2) z
4
0
+ z
4
1
+ z
4
2
+ z
2
3
4 CP
2
#7CP
2
1 (1,1,2,3) z
6
0
+ z
6
1
+ z
3
2
+ z
2
3
6 CP
2
#8CP
2
2 (1,1,1,1) z
2
0
+ z
2
1
+ z
2
2
+ z
2
3
2 CP
1
CP
1
3 (1,1,1,d) z
d
0
+ z
d
1
+ z
d
2
+ z
3
d CP
2
There are only two K3 surfaces that can be written as smooth hypersurfaces
in weighted projective spaces. Of course, these admit Ricci at Kahler-Einstein
metrics. They are given in Table 2.
Table 2: K3 surfaces as smooth hypersurfaces in CP
3
(w)
|w| d w f
w
d X
w
0 (1,1,1,1) z
4
0
+ z
4
1
+ z
4
2
+ z
4
3
4 K3
0 (1,1,1,3) z
6
0
+ z
6
1
+ z
6
2
+ z
2
3
6 K3
For all other Brieskorn-Pham polynomials with relatively prime weights, the
index [w[d is negative. There are innitely many of these and they are all surfaces
of general type and admit negative Kahler-Einstein metrics. It is easy to see that
for any such surface the geometric genus is at least 2. Suce it to present a few
examples.
164 5. K

AHLER-EINSTEIN METRICS
Example 5.4.2: These examples are complex surfaces A
w
of general type with
c
1
(A) negative, and unimodular indenite intersection form Q that are smoothly
embedded in complex weighted projective space CP(w). By Proposition 9.3.22 these
can be obtained from Brieskorn polynomials with relatively prime weights.
By Freedman smooth simply connected compact 4-manifolds with indenite
intersection form Q are determined up to homeomorphism by the parity, rank and
signature of Q. We have the following two cases according to the parity of Q.
(i) If Q is odd, A
w
is homeomorphic to /
+
2
CP
2
#/

2
CP
2
.
(ii) If Q is even, A
w
is homeomorphic to :13#n(o
2
o
2
). where : and n
are given by
: =

16
. n =
/
2
[[
2
+
3
16
.
We mention here the well-known theorem of Rohlin (cf. [GS99]) which states
that for a smooth 4-manifold with Q even the signature is divisible by 16. To
determine the homeomorphism type of A
w
it suces to compute the numbers
/
+
2
= /

2
(A
w
). or alternatively /
2
(A
w
) and /
+
2
(A
w
). The former is easily computed
using the techniques discussed in Chapter 9. While for the latter we use the well-
known formula for complex surfaces with even second Betti number, /
+
2
= 1 +2j
g
.
and j
g
= /
0,2
= /
2,0
. So by Proposition 4.6.12 j
g
equals the number of monomials
of degree d [w[. Then /

2
and can easily be determined. We consider the case
of odd and even intersection form separately.
We give two tables of innite sequences of smooth hypersurfaces in the weighted
projective space CP(w). Table 3 lists two innite sequences with Q odd; whereas,
Table 4 gives one innite sequence for the case Q even. Here both / and | are
integers greater than 1. Other sequences can be worked out in a straightforward
way.
Table 3: Hypersurfaces in CP
3
(w) homeomorphic to b
+
2
CP
2
#b

2
CP
2
b

2
w f
w
d b
+
2
14k
2
12k + 4 (1, 1, 1, k) z
3k
0
+ z
3k
1
+ z
3k
2
+ z
3
3
3k 4k
2
6k + 3
12l
2
6l + 1 (1,1,1,2l) z
4l
0
+ z
4l
1
+ z
4l
2
+ z
2
3
4l 4l
2
6l + 3
The rst two members of each of these sequences are 7CP
2
#36CP
2
. 21CP
2
#94CP
2
.
and 7CP
2
#37CP
2
. 21CP
2
#91CP
2
. Let us analyze the rst member of each sequence
a bit further illustrating an important phenomenon discovered by LeBrun [LeB96,
LeB01] and Kotschick [Kot98a].
Example 5.4.3: This example is taken from [LeB03]. The rst member of each
sequence are the surfaces
A
0
= .
6
0
+.
6
1
+.
6
2
+.
3
3
= 0 . A
2
= .
8
0
+.
8
1
+.
8
2
+.
2
3
= 0 .
and as mentioned they are complex surfaces of general type that are homeomorphic
to 7CP
2
#36CP
2
and 7CP
2
#37CP
2
. respectively. Now blow-up the rst one A
0
at
one point to obtain another complex surface A
1
homeomorphic to 7CP
2
#37CP
2
.
Using Seiberg-Witten theory, LeBrun [LeB01] gives the estimate
(2 + 3)(A
1
)
2
3
c
2
1
(A
0
)
5.4. K

AHLER-EINSTEIN METRICS ON HYPERSURFACES IN CP(w) 165

for A
1
= A
0
#CP
2
to admit any Einstein metric whatsoever. Now c
2
1
(A
0
) = 3
so (2 + 3)(A
1
) = c
2
1
(A
1
) = 2. Thus, A
1
and A
2
are both homeomorphic to
7CP
2
#37CP
2
with A
2
admitting a Kahler-Einstein metric, but A
1
admitting no
Einstein metric at all.
Exercise 5.1: Perform an analysis similar to Example 5.4.3 for the general se-
quences in Table 3.
A series giving surfaces of general type with even intersection form is:
Table 4: Hypersurfaces in CP
3
(w) homeomorphic to mK3#n(S
2
S
2
)
m w f
w
d n
l(l1)
2
(1, 1, 1, 2l + 1) z
4l+2
0
+ z
4l+2
1
+ z
4l+2
2
+ z
2
3
4l+2
5l
2
9l+2
2
The rst two members of this sequence are 13#2(o
2
o
2
) and 313#10(o
2
o
2
).
As indicated in the previous examples, Kahler-Einstein metrics abound in the
negative case. So we concentrate on the positive and null cases. Before passing on
to the case of Fano hypersurfaces which is our main interest, we briey discuss the
Calabi-Yau case.
5.4.2. Quasi-smooth Calabi-Yau Hypersurfaces in CP(w). Recall that
an orbifold A = (A. |) is Calabi-Yau if c
orb
1
(A) = 0. and that this is equivalent to
the triviality of the ordinary canonical class when H
orb
1
(A. Z) = 0 by Proposition
4.7.11. Combining this with Propositions 4.7.12, 4.6.13, and Theorem 4.4.31 gives
Theorem 5.4.4: Let A = (A. |) be a quasi-smooth weighted hypersurface in
CP(w) satisfying [w[ = d. Then w is well-formed, and A admits a unique Ricci
at orbifold metric.
When dim
C
A = 2 one can say more. Indeed, Kollar [Kol05] shows that
the minimal resolution of A is a smooth K3 surface as long as H
orb
1
(A. Z) = 0.
Furthermore, if /
2
(A) = 22 then A = 13. Thus using Theorem 4.7.14 we can
sharpen the results for the surface case (See Theorem 10.3.8 below for details).
Theorem 5.4.5: Let A = (A. |) be an orbifold K3 surface with H
orb
1
(A. Z) = 0.
Then H
2
(A. Z) = Z
s
. where 3 : 22 and : = 22 if and only if A = 13.
In 1979 Miles Reid [Rei80] produced a list of 95 weighted K3 surfaces as well-
formed hypersurfaces in the weighted projective space CP(n
0
. n
1
. n
2
. n
3
). This list
is reproduced in [IF00] and includes the two cases in Table 2 above. In Appendix
B.1 we give Reids list together with the second Betti number of each surface A
which can be computed using the Milnor-Orlik procedure described below in Chap-
ter 9, Subsection 9.3.1. Explicitly, we have a computer program which computes
the second Betti number of the link (Seifert bundle over A) by a means of the
formula of Corollary 9.3.13 below. Interestingly, Reids list produces all possible /
2
except /
2
= 3 and 18.
Theorem 5.4.4 holds equally well for quasi-smooth weighted complete intersec-
tions as long as one replaces the degree d by [d[ =

i
d
i
. A list of 84 weighted K3
surfaces as well-formed weighted complete intersections was also given in [IF00].
In higher dimensions the physicists have generated many examples. In partic-
ular, Candelas et.al. [CLS90] produced a list of over 6000 Calabi-Yau orbifolds in
166 5. K

AHLER-EINSTEIN METRICS
CP(n
0
. n
1
. n
2
. n
3
. n
4
). This was one of the rst papers that illustrated the inter-
esting new phenomenon known as mirror symmetry which has since taken on a life
of its own with important consequences for enumerative geometry. We refer the
reader to the recent books [CK99, HKK
+
03] and references therein.
5.4.3. Quasi-smooth Fano Hypersurfaces in CP(w). Suppose A CP(w)
is a quasi-smooth irreducible hypersurface in weighted projective space with its in-
duced orbifold structure A. We also assume that A is Fano, that is the orbifold
anticanonical bundle /
1

is ample. In the case that A is given as the zero set of

a weighted homogeneous polynomial, the Fano condition is 1

= [w[ d 0. by
Corollary 4.6.14. In terms of divisors we want to describe under what conditions
the log pair (A
d
. 1) is klt whenever 1 1
orb

. where
n1
n
and 1 is ef-
fective. As discussed in Section 5.2 this amounts to 1
2
conditions on the sections
of powers of the orbifold anticanonical bundle O(1
1

) = O(1

). We emphasize
once more that the klt condition is only a sucient condition for the existence of
a Kahler-Einstein orbifold metric, and that we give only sucient conditions to be
klt. We have [BGK05] the following direct consequence of Theorem 5.2.16 applied
to weighted hypersurfaces.
Corollary 5.4.6: Let 1 H
0
(CP(w). O(d)) be quasi-smooth weighted homoge-
neous polynomial of degree d
f
and weight vector w. and let A
d
f
denote the cor-
responding weighted homogeneous hypersurface with its induced orbifold structure
A. Assume further that A is a Fano orbifold, i.e., that 1

= [w[ d
f
0.
The continuity method produces a Kahler-Einstein orbifold metric on A
d
f
if there
is a
n1
n
such that for every nontrivial weighted homogeneous polynomial
o H
0
(A
d
f
. O(:1

)) of degree :1

, the function [o[

/s
is locally 1
2
on the punc-
tured ane cone (

f
.
Generally it is not easy to decide whether a given function [o[
c
is locally 1
2
or not. However, for o dened as in Corollary 5.4.6, it is not dicult, though it
can be somewhat tedious, to estimate the multiplicities of its zeroes by intersection
theory. The following theorem was given in [BGK05].
Theorem 5.4.7: Let A
d
CP(w) be a weighted homogeneous hypersurface as in
Corollary 5.4.6, and assume further that the intersections of (
f
with any number of
hyperplanes (.
i
= 0) are smooth outside the origin. If / is a weighted homogeneous
polynomial of degree d
h
and for any c R the estimate cd
f
d
h
< min
i,j
n
i
n
j

holds, then [/[

c

i
[.
i
[

i
1
is locally 1
2
on (

f
for any
i
0.
This theorem is a generalization of Proposition 11 of [JK01b] as well as results
in [JK01b, BGK05] and in the Appendix of the preprint version of [BGN02a]
(math.DG/0108113). These all correspond to the case
i
= 1 in Theorem 5.4.7,
and in this case one easily obtains the following
Corollary 5.4.8: Let A
d
CP(w) be a weighted homogeneous hypersurface as in
Corollary 5.4.6. Then A
d
admits a Kahler-Einstein orbifold metric if the following
estimate holds:
d1

<
n
(n 1)
min
i,j
n
i
n
j
.
outline of proof. The proof of Theorem 5.4.7 is somewhat technical, so we
provide only an outline, and refer to the literature for details. The proof proceeds
by induction on dimension, reducing to the surface case. Suppose for a moment that
5.4. K

AHLER-EINSTEIN METRICS ON HYPERSURFACES IN CP(w) 167

A is a surface with at worst quotient singularities. Johnson and Kollar [JK01b]
gave sucient conditions for the pair (A. 1) to be klt, namely
Lemma 5.4.9: Let A be a normal surface with at worst quotient singularities and
1 a Weil divisor on A. Then (A. 1) is klt if the following three conditions hold:
(i) If 1 =

r
i=1

i
1
i
then
i
< 1 for all i,
(ii) for all smooth points r A, mult
x
(1) 1,
(iii) if 1 A is a singular point and : Y A a local nite cover re-
solving the singularity at 1 then

1 has multiplicity at most one at

Q =
1
(1).
Condition (i) is also necessary, while neither (ii) nor (iii) is necessary. When
passing to the case of general dimension we cut A with the appropriate number of
hypersurfaces, and the condition of the hypothesis on the intersection of hyperplanes
allows us to apply inversion of adjunction (Theorem 5.2.17) to conclude that (A. 1)
is klt. In general dimension the conditions (i) and (ii) remain essentially the same,
whereas, we must replace condition (iii) by
(iii) Suppose r A is a singular point of A with local uniformizing group
x
and suppose H
1
. H
2
. . . . . H
dimX
are hypersurfaces through r (including
the 1
i
), then we require the intersection number to satisfy
A H
1
H
dimX

1
[
x
[
.
The details can be found in varying degree of detail in several places [JK01b,
JK01a, BGK05], but perhaps the most detail appears in the Appendix of the
preprint version (math.DG/0108113) of [BGN02a].
Let us now specialize to the case of a Brieskorn-Pham polynomial of Examples
4.6.7 and 4.6.11, then the Fano condition [w[ d 0 becomes 1 <

n
i=0
1
ai
. So if
A
d
CP(w) is a Fano hypersurface arising from a BP polynomial, the estimate of
Corollary 5.4.8 takes the form
(5.4.1) 1 <
n

i=0
1
o
i
< 1 +
n
n 1
min
i,j

1
o
i
o
j

.
These conditions are far from optimal, and we shall see how Theorem 5.4.7 gives
better estimates in the non-well-formed case. This comes about by reducing, via
the procedure of Section 4.5, to a well-formed orbifold. Let 1 be a BP polynomial
or a weighted homogeneous perturbation thereof.
Lemma 5.4.10: Let w = (n
0
. . . . . n
n
) be the weight vector and a = (o
0
. . . . . o
n
)
the exponent vector of a BP polynomial 1 of degree d. and let d
j
. c
j
and o
w
be given
by Denition 4.5.7. Then there is a BP polynomial o of degree

d = lcm(/
i
) with a
well-formed weight vector w = ( n
0
. . . . . n
n
) and exponent vector b = (/
0
. . /
n
)
such that there is an isomorphism of algebraic varieties A
f
A
g
. Furthermore, the
following conditions hold
(i) d = o
w

d.
(ii) d = C
j
d
j
. where C
j
= lcm(o
i
[i = ,).
(iii) /
j
= gcd(o
j
. C
j
).
for all , = 0. . . . . n.
168 5. K

AHLER-EINSTEIN METRICS
Proof. We apply the procedure discussed in Section 4.5 (cf. Example 4.5.11).
We begin with the BP polynomial 1 = .
a
0
0
+ + .
a
n
n
such that A
f
CP(w).
By Proposition 4.5.10 there is an isomorphism : CP(w)CP( w) of algebraic
varieties dened in local coordinates by

r
j
= .
dj
j
. Now has degree o
w
= lcm(d
i
)
and

r
j
has degree n
j
d
j
= d
j
c
j
n
j
= o
w
n
j
. Thus, r
j
has weight n
j
. Now d = n
i
o
i
for all i = 0. . . . . n. and by Lemma 4.5.8 d
j
divides n
i
for all i = ,. and is relatively
prime to n
j
. Thus, d
j
must divide o
j
for each ,. So we dene /
i
=
ai
di
and consider
the BP polynomial o =

n
i=0
r
b
i
i
together with its corresponding weighted variety
A
g
CP( w). Then we have
1(.
0
. . . . . .
n
) =
n

i=0
.
ai
i
=
n

i=0
(.
di
i
)
b
i
=
n

i=0
(

r
i
)
b
i
=

o(r
0
. . . . . r
n
)

.
Then A
g
and A
f
are isomorphic as algebraic varieties (but not of course as weighted
varieties or orbifolds). It follows from Lemma 4.5.8 that w is well-formed. This
proves the rst part.
For the three conditions, we rst see that (i) follows from the fact that the de-
gree of is o
w
. (ii) follows from the denition of C
j
and d
j
together with the equality
o
i
n
i
= d for all i. Then (iii) follows from d/
j
= C
j
d
j
/
j
= C
j
o
j
= d gcd(o
j
. C
j
).
We are now ready to prove our main estimate for Brieskorn-Pham polynomials
which was given in [BGK05]. Once again this estimate is not optimal at least in
certain cases as we shall see shortly.
Theorem 5.4.11: Let A
d
be a weighted homogeneous hypersurface arising from a
Brieskorn-Pham polynomial or a weighted homogeneous perturbation thereof (i.e.,
condition (GC) of Example 4.6.7 holds). Then the orbifold A
d
is Fano if and only
if
1 <
n

i=0
1
o
i
and admits a Kahler-Einstein orbifold metric if
1 <
n

i=0
1
o
i
< 1 +
n
n 1
min
i,j

1
o
i
.
1
/
i
/
j

.
In Theorem 5.4.11 the klt condition is the right hand inequality, and since
/
i
o
i
for all i it clearly is at least as good as the estimate of Equation (5.4.1).
We shall see that in a sense the less well-formed w is the better the estimate
becomes, whereas, in the well-formed case the estimate coincides with the estimate
of Equation (5.4.1). Notice also from Lemma 5.4.10 that if the o
i
s are pairwise
relatively prime the /
i
s are all 1. so the estimate is fairly good in this case, although
it can be improved upon as seen by Theorem 5.4.12 below.
Proof of Theorem 5.4.11. By Corollary 5.4.6 we need to show that for ev-
ery : 0 and for every weighted homogeneous polynomial o of weighted degree
:1
X
d
= :d(

o
1
i
1), the function [o[
/s
is locally 1
2
on the punctured ane
cone (

f+p
of the perturbed BP polynomial 1 + j. where j is any weighted homo-
geneous polynomial of degree d
f
such that the ane cone (
f+p
satises condition
(GC) of Example 4.6.7. By Lemma 5.4.10 we reduce this to a problem on a per-
turbation of the simpler Brieskorn-Pham polynomial o to which we can apply the
sharper estimate given by Theorem 5.4.7. As in the proof of Lemma 5.4.10 we
5.4. K

AHLER-EINSTEIN METRICS ON HYPERSURFACES IN CP(w) 169

easily arrive at a perturbation c of o such that j(.
0
. . . . . .
n
) = c(r
0
. . . . . r
n
). This
then gives a branched covering map : (

f+p
(

g+p
of the punctured ane cones
of the perturbed polynomials.
More generally we consider an arbitrary weighted homogeneous polynomial
/ with degree d
h
. If the variable .
i
has exponent c
i
in / then there is a unique
nonnegative integer :
i
such that .
ci
i
= .
ri
i
.
dibi
i
= .
ri
i
r
bi
i
. So there is a polynomial
H(r
0
. . . . . r
n
) of degree d
H
such that
(5.4.2) /(.
0
. . . . . .
n
) =

i
.
ri
i
H(r
0
. . . . . r
n
).
Since the Jacobian of has a (d
i
1)-fold zero along (.
i
= 0), we see that
() [/[
/s
is locally 1
2
on (

f+p
if and only if [H[
/s

[r
i
[

e
i
sd
i
+
1
d
i
1
is
locally 1
2
on (

g+q
.
We now apply Corollary 5.4.6 and Theorem 5.4.7 to the latter estimate with
/ H
0
(CP(w). O(:1

)). First, each variable r

i
must have an exponent greater
than 1. This is equivalent to c
i
<
1
:. But we know from above that c
i
n
i

d
h
= :1

= :d(

o
1
i
1). So if we assume that (

o
1
i
1) <
n
n1
min
i

1
a
i
we
have
c
i
o
i
:(

o
1
i
1) < o
i
:
n
n 1
min
i

1
o
i

<
1
o
i
: min
i

1
o
i

<
1
:.
Next if we can satisfy the condition in the hypothesis of Theorem 5.4.7 for the
well-formed perturbed BP polynomial o + c. then we can use () together with
Corollary 5.4.6 to conclude the proof. Consider a weighted homogeneous polynomial
/ H
0
(A. O(:1

)). We have using Equation (5.4.2)

d
H
d
g

d
2
g
d
f
d
h
= d
2
g
:

i
1
o
i
1

:
n
n 1
min
i,j

d
2
g
/
i
/
j

= :
n
n 1
min
i,j
n
i
n
j
.
For the special case of A = CP
n1
Ghigi and Kollar [GK05] have improved
the klt estimate of Theorem 5.4.11 by roughly a factor of n. Their idea is based
on the paper [AGP06] where the existence of Kahler-Einstein metrics on certain
manifolds of the form `G. where G is a nite group, implies the existence on the
Galois cover `. In [GK05] this method is applied to the identity map on CP
n1
.
Let us consider the orbifold A(a) = (A
d
. ) =

CP
n1
.

1
1
a
i

1
i

. of Example
4.6.15, where the o
i
s are pairwise relatively prime, 1
i
= .
i
= 0 for i = 1. . . . . n
and 1
0
= .
1
+ +.
n
= 0. Under these conditions the Ghigi-Kollar theorem is
Theorem 5.4.12: Let a = (o
0
. . . . . o
n
) be a sequence of relatively prime integers.
Then the orbifold A(a) = (A
d
. ) =

CP
n1
.

1
1
a
i

1
i

admits a Kahler-
Einstein metric if the following estimate holds:
0 <
n

i=0
1
o
i
1 < nmin
i

1
o
i

.
Surprisingly, it turns out that the Ghigi-Kollar estimate in this case is sharp,
i.e., the converse of this theorem is also true. We will discuss this in the proof
of Theorem 11.5.1 in Chapter 11 after we introduce obstructions to the existence
of Sasaki-Einstein metrics on links. In particular, we shall see that these new
obstructions give non-existence result for Kahler-Einstein metrics on many log del
Pezzo surfaces.
170 5. K

AHLER-EINSTEIN METRICS
We are interested in nding sequences of integers a = (o
0
. . . . . o
n
) with n 3
that satisfy the inequalities of Theorem 5.4.11. For low values of n these inequalities
are easily solved on a computer. However, it is not only instructive to obtain some
solutions by hand, but also we can prove some interesting results by judicious
choices of sequences. We consider a very important sequence that does not satisfy
the inequalities, but marks the borderline of the left hand inequality. It is the
so-called extremal sequence or Sylvesters sequence [GKP89] determined by the
recursion relation
(5.4.3) c
k+1
= 1 +c
0
c
k
= c
2
k
c
k
+ 1
beginning with c
0
= 2. It starts as
2. 3. 7. 43. 1807. 3263443. 10650056950807. ...
The importance of this sequence is that it satises
(5.4.4)
n

i=0
1
c
i
= 1
1
c
0
c
n
.
So we see that sequences of the form a = (o
0
= c
0
. . . . . o
n1
= c
n1
. o
n
) satisfy the
left hand inequality as long as o
n
< c
0
c
n1
. Furthermore, by construction the
rst n elements of such sequences are all pairwise relatively prime. It remains to
analyze the right hand inequality.
Example 5.4.13: We continue the discussion of Example 4.6.15. Here recall that
a = (o
0
. . . . . o
n
) consists of pairwise relatively prime integers, so d = o
0
o
n
and n
i
= o
0
o
i
o
n
. Thus, /
j
= gcd(o
j
. C
j
) = 1 for all , in which case we
can use the sharper estimate of Theorem 5.4.12. There are many sequences that
satisfy this condition. Take a = (o
0
= c
0
. . . . . o
n1
= c
n1
. o
n
) with c
n1
<
o
n
< c
0
c
n1
with o
n
relatively prime to c
0
c
n1
. then both inequalities are
satised. For example for n = 3 this gives the sequence (2. 3. 7. o
3
) with o
3
=
11. 13. 17. 19. 23. 25. 29. 31. 37. 41. But we can also relax the condition o
n1
= c
n1
;
for example, we can take the sequence (2. 3. 5. o
3
) giving a larger range for o
3
,
namely, o
3
= 7. 11. 13. 17. 19. 23. 29. 31. 37. 41. 43. 47. 49. 53. 59. Next, for n = 4 we
take o
4
to be any integer between 43 and 1806 that is relatively prime to 1806. etc.
This example is just a special case and with the aid of a computer one can take
full advantage of the sharper estimates of both Theorem 5.4.11 and Theorem 5.4.12.
Our program, which is described in [BGKT05], generates a list of 68 sequences
for n = 3 and 8494 for n = 4. We shall make important use of these results later
in Chapter 11. In particular, we show that the number of sequences grow double
exponentially with dimension.
Now we discuss the more general case of weighted homogeneous polynomials,
which is based on the papers [JK01b, JK01a, Ara02, BGN03b, BGN02b].
These papers all assume that the weights are well-formed. This was done mainly for
convenience and can be dropped. However, for our discussion here we shall assume
that the weight vector w is well-formed, and restrict ourselves to the surface case,
i.e., n = 3. We rst need to impose the conditions guaranteeing quasi-smoothness,
Exercise 4.4. From condition (i) of Exercise 4.4 we see that we need
(5.4.5) :
i
n
i
+n
j(i)
= d = n
0
+n
1
+n
2
+n
3
1

.
5.4. K

AHLER-EINSTEIN METRICS ON HYPERSURFACES IN CP(w) 171

where :
i
is a positive integer and both i and ,(i) are integers ranging over 0. 1. 2. 3.
Note we can have ,(i) = i. Since here we are only interested in Fano orbifold, we
impose the condition that the index be positive. In what follows we assume without
loss of generality that the weights are ordered n
0
n
1
n
2
n
3
. Then we have
Lemma 5.4.14: Assuming the quasi-smoothness conditions (i) and (iii) of Exer-
cise 4.4, and that 1

0., the following bounds hold

1. 1 :
3
2.
2. Either 2 :
2
4. or 2n
0
1

.
3. Either 2 :
1
10. or 21

= n
0
+ n
1
. or 21

3n
0
. or quasi-
smoothness condition(ii) of Exercise 4.4 is violated.
Proof. The proof is straightforward, though a bit tedious, so we refer the
reader to [BGN03b] for details.
The solutions of the linear system (5.4.5) satisfying Lemma 5.4.14 split into two
cases depending on whether :
0
is bounded or not. The case that :
0
is unbounded,
gives series solutions depending on an integer /, and the case when :
0
is bounded,
give the so-called sporadic solutions. The series solutions can be worked out by
hand, but the sporadic solutions require the aid of a computer. In [BGN03b]
we modied the computer program of [JK01b] to solve the system (5.4.5) for any
index 1 as well as to discard some solutions that are not quasi-smooth. Notice that
there are solutions with arbitrary :
1
that are quasi-smooth, but our lemma implies
that these must satisfy 21 3n
0
or 21 = n
0
+n
1
. As we shall see, in both cases the
sucient condition for the existence of a Kahler-Einstein metric fails. Examples of
such a series are: (1. 1. /. /) of degree 2/ and index 2, (1. 2. 2/ +1. 2/ +1) of degree
2(2/ +1) and index 3, and (2. 2. 2/ +1. 2/ +1) of degree 2(2/ +1) and index 4, as
well as the general series (1 n. 1 +n. n. n +n) of degree 2n +n +1 and index 1.
There are also double series such as (1. 1. :. :+/) of degree 2:+/ and index 2.
We now make use of the inequalities in Lemma 5.4.14 involving 1

.
Lemma 5.4.15: If 21

3n
0
or 21

= n
0
+n
1
then for every c 0 there exists
a divisor 1 [ 1

[ such that the pair

A.
2+
3
1

is not klt.
Proof. The rst inequality is easy for it implies that the divisor 1
0
= (.
0
= 0)
satises :1
0
[O

(1

)[ for :
3
2
. But this says that
2+
3
1
0
has multiplicity
greater than 1 at a smooth point, so (A.
2+
3
1) cannot be klt by (ii) of Lemma
5.4.9.
Next assume the second equality holds. We notice in the proof (Lemma 5.1 in
[BGN03b]) of Lemma 5.4.14 that 21 = n
0
+ n
1
occurs in precisely one case and
in this case we have n
3
= n
1
+n
2
1. But then w must have the form
w = (1 n. 1 +n. n. n +n)
for some n Z
+
with n 1 + n. and some non-negative integer n < 1 and in
this case the degree d = 2n+n +1. Now suppose that .
a
0
0
.
a
1
1
.
a
2
2
.
a
3
3
is a monomial
occurring with nonzero coecient in the polynomial dening A. We claim that if
o
0
= 0 then o
1
= 0. Indeed, if o
0
= o
1
= 0 then
o
2
n +o
3
n = 2n + 21 +n.
But n 1+n so the only possible solutions would require o
2
+o
3
= 3 and n = 21+n
and one readily checks that these hypersurfaces are never quasi-smooth. Thus the
divisor 1 = .
0
= 0 A has at least two components, 1 and 1. where 1 is the
172 5. K

AHLER-EINSTEIN METRICS
line .
0
= .
1
= 0 and 1 is dened, inside the weighted projective plane .
0
= 0,
by a polynomial 1(.
1
. .
2
. .
3
). Moreover, 1(.
1
. .
2
. .
3
) = .
2
2
+ .
2
1
(o(.
1
. .
2
. .
3
)). Note
that the point 1 = (0. 0. 0. 1) A since n + n does not divide the degree of A.
Thus if : C
2
A is a local cover of the quotient singularity at 1 = (0. 0. 0. 1)
then

1 =

1 +

1 has multiplicity at least 1 + mult

0
(

1) at the origin. To
compute the multiplicity of

1 at the origin, let Y = 7(1(.

1
. .
2
. .
3
)), the zero set
of 1. Then Y A = 1 G. where G does not contain the point (0. 0. 0. 1). Thus
mult
0
(

1) = mult
0
(

7(1)) 2 .
Consequently, (A.
2+
3
1) is never klt as it always has multiplicity 2 at 0. But
I
In
1 [ 1

[ so this completes the proof of the lemma.

Now with the aid of our computer program we have
Theorem 5.4.16: Let A = (A. |) be a well-formed quasi-smooth log del Pezzo
surface with weight vector w. index 0 < 1

10 and degree d embedded in the

weighted projective space CP(w) = CP(n
0
. n
1
. n
2
. n
3
). Suppose further that A sat-
ises the klt condition on the pair (A.
2+
3
1). where 1 [ 1

[. then A must
belong to one of the cases listed in Tables B.2.1 and B.2.2 in Appendix B.
Outline of Proof. With 1

0 the orbifold A is Fano, and the quasi-

smoothness and klt assumptions together with Lemmas 5.4.14 and 5.4.15 imply the
range for :
i
indicated in Lemma 5.4.14 for the computer program as well as the
inequalities 21 < 3n
0
and 21 = n
0
+ n
1
. The computer program then produces
tables B.2.1 and B.2.2 of Appendix B assuming the indicated range for the index
1

. The last column of these tables indicates whether the klt condition (which
implies that A admits a Kahler-Einstein metric) holds (Y) or is unknown (?). To
verify the klt condition for the sporadic case we use the following corollary which is
a sharpening of Corollary 5.4.8 for the surface case and whose proof can be found
in [BGN03b]:
Corollary 5.4.17: Let w = (n
0
. n
1
. n
2
. n
3
) and A CP(w) be a quasi-smooth
surface of degree d = n
0
+ n
1
+ n
2
+ n
3
1

. Then A admits a Kahler-Einstein

orbifold metric if 21

d < 3n
0
n
1
. If the line .
0
= .
1
= 0 A then 21

d < 3n
0
n
2
is also sucient. If the point (0. 0. 0. 1) A then 21

d < 3n
0
n
3
is also sucient.
The proof of the klt condition in the series case is more delicate. Again full
proofs can be found in [BGN03b, BGN02b].
Remark 5.4.1: The computer program indicates that there are neither series solu-
tions nor sporadic solutions satisfying the hypothesis of Theorem 5.4.16 for 1

10.
In fact, an easy argument shows that there are no such solutions for suciently large
1

.
2
A more systematic study of all (not just well-formed) log del Pezzo hypersur-
faces in CP(n
0
. n
1
. n
2
. n
3
) can be carried out in a similar way. Here the starting
point has recently been established by Yau and Yu [YY05] who, based on the much
earlier work of Kouchnirenko [Kou76] and Orlik and Randell [OR77], presented a
table of 19 distinct types of quasi-smooth weighted homogeneous hypersurfaces A
f
in CP(n
0
. n
1
. n
2
. n
3
). More precisely we have
2
The code for the C program used to generate the tables of the Theorem 5.4.16 are available
at: http://www.math.unm.edu/galicki/papers/publications.html.
5.5. AUTOMORPHISMS AND MODULI 173
Theorem 5.4.18: Let 1(z) C[.
0
. .
1
. .
2
. .
3
] be such that Y
f
= z C
4
[ 1(z) = 0
has only an isolated singularity at the origin. Then 1(z) = /(z) + o(z). where
/(z) is one of the 19 classes of weighted homogeneous polynomials with isolated
singularities at the origin listed in Table B.5.1 of Appendix B and /(z) and o(z) have
no monomial in common. If 1(z) is weighted homogeneous with rational weights
(dn
0
. dn
1
. dn
2
. dn
3
) then so are /(z) and o(z). Furthermore, let Y
h
= z
C
4
[ /(z) = 0 and let 1
f
and 1
h
be the associated links. Then 1
f
is equivariantly
dieomorphic to 1
h
.
Open Problem 5.4.1: All A
h
CP(n
0
. n
1
. n
2
. n
3
). where /(z) is one of the
weighted homogeneous polynomials mentioned in Theorem 5.4.18, have denite or
zero rst Chern class. One can partition the list depending on the sign of c
orb
1
(A
h
)
and try to determine which of the log del Pezzo surfaces can possibly admit a
Kahler-Einstein metric using the methods of this chapter.
We end this section with a brief discussion of the higher dimensional case. The
ideas described here can clearly be applied to any dimension, although computer
searches become much more time consuming as the dimension increases. In this
regard the only further results known are in dimension 7. where Johnson and Kollar
[JK01a] have produced a list of 4442 Fano 3-folds that embed into weighted projec-
tive 4-spaces CP(n
0
. n
1
. n
2
. n
3
. n
4
). Out of this list they showed that 1936 of these
admit Kahler-Einstein metrics using the techniques described above. However, so
far little is known about the topology of the orbifolds on this list. To date the only
known results are those of [BGN02a] where 184 3-folds with /
3
= 0 were isolated.
This was used to give Sasaki-Einstein metrics on rational homology 7-spheres, and
will be discussed further in Chapter 11.
Example 5.4.19: [Iano-Fletcher/Reid Fano 3-folds] Consider any orbifold 13
surface of the Appendix B.1. Suppose we take any A
d
CP(n
0
. n
1
. n
2
. n
3
) repre-
sented by some A
f
with some 1 = 1(.
0
. .
1
. .
2
. .
3
). Then o = n
d
+ 1(.
0
. .
1
. .
2
. .
3
)
is a weighted homogeneous polynomial of the same degree and with weights w =
(1. n
0
. n
1
. n
2
. n
3
). The 3-fold A
g
is Fano of index one as 1 = 1 + n
0
+ n
1
+ n
2
+
n
3
d = 1. In [IF00] Iano-Fletcher shows that
Lemma 5.4.20: There is a one-to-one correspondence between the following two
classes
(i) the set of families of quasismooth, well-formed weighted hypersurfaces
A
2
d
CP(n
0
. n
1
. n
2
. n
3
) with only canonical singularities and trivial orb-
ifold canonical class 1
orb

= 0.
(ii) the set of families A
3
d
CP(1. n
0
. n
1
. n
2
. n
3
) of quasismooth weighted
3-folds of index 1 = 1.
Hence, one gets 95 examples of anti-canonically embedded Fano 3-folds which
correspond to the Reids K3 orbifold surfaces. It was recently shown by Chelstov
that all but 5 of these admit Kahler-Einstein orbifold metrics [Che07].
5.5. Automorphisms and the Moduli Problem
In this section we are mainly concerned with describing the moduli space of
K ahler-Einstein metrics on weighted homogeneous hypersurfaces in CP(w). This
will then give moduli for Sasaki-Einstein and Sasaki--Einstein metrics which we
discuss in detail in Chapter 11.
174 5. K

AHLER-EINSTEIN METRICS
5.5.1. Automorphisms of Weighted Projective Spaces. We begin by
discussing the group Aut(CP(w)) of complex automorphisms of a weighted projec-
tive space CP(w). The group C

(w) of Denition 4.5.1 is a subgroup of the group

Aut(C
n+1
) of biholomorphic maps from C
n+1
to itself, and we let 7(C

(w)) denote
the centralizer of C

(w) in Aut(C
n+1
). Clearly C

(w) 7(C

(w)). We have
Lemma 5.5.1: The automorphism group Aut(CP(w)) is isomorphic to the quo-
tient group 7(C

(w))C

(w) and 7(C

(w)) can be characterized as the subgroup

of Aut(C
n+1
) consisting of biholomorphic maps dened by sending .
i
to a linear
combination of monomials of degree n
i
for all i = 0. . . . . n. Thus, Aut(CP(w)) is a
Lie group of dimension

i
/
0
(CP(w). O(n
i
)) 1.
Proof. First, Aut(CP(w)) is a Lie group by Proposition 1.6.7 which holds
equally well for compact orbifolds. Furthermore, since CP(w) = (C
n+1
`0)C

(w)
the elements of Aut(CP(w)) all lift to biholomorphic maps of C
n+1
` 0 which
commute with C

(w). It follows by Hartogs Theorem that Aut(CP(w)) is a closed

subgroup of Aut(C
n+1
). Moreover, if Aut(C
n+1
) commutes with C

(w) then it
descends to element in Aut(CP(w)). So there is a group epimorphism
7(C

(w))Aut(CP(w))
whose kernel is C

(w).
Now let 7(C

(w)). Then (.
i
) = o
i
(.
0
. . . . . .
n
). where o
i
is a holomorphic
function of the n+1 variables .
0
. . . . . .
n
. Since commutes with C

(w). o
i
satises
o
i
(
w
0
.
0
. . . . .
w
n
.
n
) =
w
i
o
i
(.
0
. . . . . .
n
) .
that is, o
i
is a weighted homogeneous polynomial of degree n
i
. This means that
o
i
consists of a linear combination of monomials of degree n
i
. Such monomials
span the nite dimensional vector space H
0
(CP(w). O(n
i
)) whose dimension is
/
0
(CP(w). O(n
i
)).
Now let A
f
be a quasi-smooth weighted hypersurface in CP(w). and let Aut(A
f
)
denote its complex automorphism group. We are interested in knowing when
Aut(A
f
) is a closed subgroup of Aut(CP(w)). Then
Lemma 5.5.2: If n 4 or if n = 3 and [w[ = d. then Aut(A
f
) is the closed
subgroup of Aut(CP(w)) that leaves 1 invariant, that is,

1 = 1. Hence, Aut(A
f
)
is an algebraic group.
Proof. Let
f
: A
f
CP(w) denote the inclusion map. If n 4 then
Pic
orb
(A
f
) = Z and the line orbibundle

f
O(1) generates Pic
orb
(A
f
) = Z by
the Grothendieck-Lefschetz theorem [Gro68]. Thus,

f
O(1) is invariant under
Aut(A
f
). so it is a closed algebraic subgroup of Aut(CP(w)). This also occurs when
n = 3 and [w[ = d since in this case either /
orb
or (/
orb
)
1
is ample, invari-
ant under Aut(A
f
). and the line orbibundle

f
O(1) is a unique root of /
orb
or
(/
orb
)
1
.
Our next result says that Aut(A
f
) is a discrete group, or alternatively that the
Lie algebra aut(A
f
) vanishes, in most cases.
Lemma 5.5.3: If 2n
i
< d for all but at most one i then the group Aut(A
f
) is
discrete. If in addition n 4 or n = 3 and [w[ = d. then the group Aut(A
f
) is
nite.
5.5. AUTOMORPHISMS AND MODULI 175
Proof. For the rst statement it is enough to show that there are no one
parameter subgroups of Aut(A
f
) under the hypothesis of the lemma. Let
t
be a
one parameter subgroup of Aut(A
f
) Aut(CP(w)). and write
.
t
i
=
t
(.
i
) = .
i
+

j1
t
j
o
ij
(.
0
. . . . . .
n
) .
The condition of invariance is
1(
t
(.
0
). . . . .
t
(.
n
)) = 1(.
0
. . . . . .
n
) .
Dierentiating with respect to .
i
. expanding in a Taylor series in t and equating
the coecients of t
j
for each , gives

k
o
kj
1
.
k
= 0 .
Since the degree of o
kj
is n
k
and the degree of
f
z
k
is d n
k
. and not all
f
z
k
s
vanish, this implies that either all o
kj
vanish or there is a cancellation for two
dierent values of /. say / and |. But the only way this can happen is that n
k
= n
l
and that both
f
z
k
and
f
z
l
are linear in .
k
or .
l
. But this implies that n
k
= n
l
=
d
2
.
This proves the rst statement. The second statement follows from Lemma 5.5.2
since an algebraic group has nitely many components.
This result was stated incorrectly in [BGK05]. The hypothesis n 3 or n = 2
and [w[ = d was left o. It was, however, used correctly since all cases treated in
[BGK05] are Fano. That the extra hypothesis is necessary in general is illustrated
by the classical example of Fano and Severi who exhibited a quartic surface in CP
3
.
where Aut(A
f
) is an innite discrete group. This is discussed in detail in [MM64].
We mention also the related result of Kobayashi [Kob59a, Kob72] which says
that Aut(A
f
) is nite for complex manifolds with negative rst Chern class.
5.5.2. The Moduli Problem for Kahler-Einstein metrics. It is not our
intention here to discuss moduli theory in any generality. This is a very important
part of modern mathematics which is treated in detail elsewhere, cf. [Kod86,
MFK94, SU02]. Rather, we are interested in the moduli problem for Kahler-
Einstein metrics, and more specically mainly those arising from hypersurfaces
embedded in weighted projective spaces. Later in Section 8.2.1 we review Kodaira-
Spencer theory in the context of the transverse holomorphic geometry of a foliation,
but here we proceed in a much more pedestrian fashion by contenting ourselves with
a count of eective parameters for Kahler-Einstein orbifold metrics. For now the
only bit of information from complex deformation theory that we need explicitly
is the well-known fact [Kod86] that H
1
(A. ) is the tangent space to the space
of deformations of complex structures on A. where is the sheaf of germs of
holomorphic vector elds on A. First, we mention a general result
Theorem 5.5.4: Let A = (A. |) be a compact orbifold with a Kahler-Einstein
orbifold metric o. Then
(i) If c
1
(A)
orb
< 0 then o is unique up to homothety.
(ii) If c
1
(A)
orb
= 0 then o is unique up to homothety in each positive (1. 1)
cohomology class.
(iii) If c
1
(A)
orb
0 then o is unique up to a complex automorphism.
176 5. K

AHLER-EINSTEIN METRICS
In the manifold case uniqueness in (i) and (ii) was proved originally by Calabi
and follows from Theorem 5.2.2. Case (iii) is due to Bando and Mabuchi [BM87].
A proof can also be found in [Aub98]. The proof for orbifolds is essentially the
same.
We mention that in case (iii) of Theorem 5.5.4, the Lie algebra of the complex
automorphism group is the complexication of the Lie algebra of the isometry group
by Theorem 5.3.1. Theorem 5.5.4 says that the moduli space of Kahler-Einstein
metrics is determined by the moduli space of complex structures. In case (ii) one
should also take into account the contribution from all the admissible (1. 1)-classes.
Thus, one needs to consider the moduli space of marked pairs consisting of complex
manifold (or orbifold) A and Kahler class on it (cf. [BHPVdV04], pg 334-336,
and Chapter 12 of [Bes87]). For example, in the case of K3 surfaces this moduli
space has real dimension 60; therefore, the moduli space of Kahler-Einstein metrics
(up to homothety) on K3 has real dimension 59. However, the moduli space of
marked pairs is of much less interest to us in this book since we are more concerned
with the moduli space of a polarized Kahler orbifold.
The study of Einstein moduli spaces on complex manifolds is due to Koiso
[Koi83]. Here we present a very brief summary of his work which has been treated
in much more depth in Chapter 12 of [Bes87]. Following Besse we let o be an
orbifold Einstein metric on an orbifold
3
`, and denote the space of innitesimal
Einstein deformations by (o). This space is characterized by symmetric 2-tensors
/ which satisfy certain equations due to Berger and Ebin [BE69] who show that
(o) is nite dimensional and should be thought of as the tangent space to the
premoduli space of Einstein structures near o. Here we are interested in the case
that A is a complex orbifold and o is a Kahler-Einstein metric on it. In this case
one gets a splitting (o) = (o)
H
(o)
A
into Hermitian and skew-Hermitian parts.
Then for manifolds Koiso [Koi83] proves the following result (which also holds for
complex orbifolds as long as c
1
(A) is replaced by c
orb
1
(A)).
Theorem 5.5.5: Let (J. o) be a Kahler-Einstein structure on A with xed volume
say, one, and assume that J belongs to a complete
4
family of complex structures.
Then
(i) if c
orb
1
(A) < 0 or c
orb
1
(A) = 0. then every Einstein metric o
t
in the
premoduli space is Kahler with respect to a deformed complex structure,
(ii) if c
orb
1
(A) 0. (o)
H
= 0. and there are no nonzero holomorphic vector
elds on A. then every Einstein metric o
t
in the premoduli space is Kahler
with respect to a deformed complex structure.
Moreover, the following hold
(a) if c
orb
1
(A) < 0. then dim (o) = 2dim
C
H
1
(A. ).
(b) if c
orb
1
(A) = 0. then
dim (o) = 2dim
C
H
1
(A. ) 2dim
C
H
2,0
(A. C) + dim H
1,1
(A. R) 1 .
(c) if c
orb
1
(A) 0. then dim (o) 2dim
C
H
1
(A. ).
This implies that the subset of Kahler-Einstein metrics is open in the premoduli
space of innitesimal Einstein deformations.
3
The treatment in [Koi83] as well as [Bes87] assumes that M is a manifold, but the discus-
sion carries through for orbifolds by working on the local uniformizing covers.
4
See Koiso or Denition 1.7 of [KS58] for the denition.
5.5. AUTOMORPHISMS AND MODULI 177
Exercise 5.2: Show that Koisos proof of Theorem 5.5.5 holds for orbifolds when
c
1
(A) is replaced by c
orb
1
(A).
5.5.3. The Moduli Problem for Weighted Hypersurfaces. We now turn
to the problem at hand, namely the description of the moduli space of complex
embeddings of a projective algebraic orbifold A as a weighted hypersurface into a
xed weighted projective space CP(w).
Proposition 5.5.6: Let A
f
be a quasi-smooth weighted hypersurface in CP(w)
corresponding to the weighted homogenous polynomial 1 of degree d and weight
vector w = (n
0
. . . . . n
n
). Let Aut(A
f
) denote the group of automorphisms of A
f
.
Assume also that n 3. Then the complex orbifolds A = (A
f
. |) form a continuous
family of complex dimension
/
0

CP(w). O(d)

i
/
0

CP(w). O(n
i
)

+ dim Aut(A
f
) .
Proof. The space of weighted homogeneous polynomials of degree d is the
complex vector space H
0

CP(w). O(d)

of dimension /
0

CP(w). O(d)

. The action
of Aut(CP(w)) on CP(w) induces a representation of Aut(CP(w)) on the vector
space H
0

CP(w). O(d)

. The hypersurfaces cut out by two polynomials

1. o H
0

CP(w). O(d)

are isomorphic if they dier by a transformation in Aut(CP(w)). So we get a complex

family of orbifolds of complex dimension at least
/
0

CP(w). O(d)

i
/
0

CP(w). O(n
i
)

.
However, the isotropy subgroup of 1 H
0

CP(w). O(d)

is Aut(A
f
). so any one
parameter subgroup of Aut(A
f
) cannot contribute to the equivalence between 1
and o implying the result.
We now combine Proposition 5.5.6 with Theorem 5.5.4 to obtain
Theorem 5.5.7: Let A
f
be a quasi-smooth weighted hypersurface of degree d in
CP(w) and let 1 = [w[ d < 0. Then
(i) If 1 < 0 then A
f
admits a 2

/
0

CP(w). O(d)

i
/
0

CP(w). O(n
i
)

dimensional family of Kahler-Einstein metrics up to homothety,

(ii) If 1 = 0 A
f
admits a 2

/
0

CP(w). O(d)

i
/
0

CP(w). O(n
i
)

di-
mensional family of Kahler-Einstein metrics up to homothety in each
positive (1. 1) cohomology class.
(iii) If 1 0 and A
f
admits a Kahler-Einstein metric for a generic 1 then
it admits a 2

/
0

CP(w). O(d)

i
/
0

CP(w). O(n
i
)

+ dim Aut(A
f
)

dimensional family of Kahler-Einstein metrics up to homothety.

We give an application of this theorem.
Example 5.5.8: Consider the log del Pezzo orbifold surface A
210
of degree 210
obtained from the sequence a = (35. 7. 3. 2). This is represented by the zero set
of the weighted homogeneous polynomial 1 = 1
7
(.
5
0
. .
1
) + .
3
2
+ .
2
3
in the weighted
projective space CP(w) with weight vector w = (6. 30. 70. 105). where 1
7
is any
suciently general homogenous septic polynomial. The conditions of Theorem
178 5. K

AHLER-EINSTEIN METRICS
5.4.11 are easily seen to be satised. So A
210
admits Kahler-Einstein metrics.
Aut(CP(w)) is given modulo the C

(w) action by the transformations

(.
0
. .
1
. .
2
. .
3
) (
0
.
0
.
1
.
1
+
10
.
5
0
.
2
.
2
.
3
.
3
) .
So we have

i
/
0
(CP(w). O(n
i
)) = 5. and by Lemma 5.5.3 dim Aut(A
f
)) = 0.
Now it is easy to see that /
0
(CP(w). O(210)) = 10. So we have a 2(10 5) = 10
parameter family of Kahler-Einstein metrics on A
210
. In Chapter 11 we show that
this gives a 10 parameter family of Sasaki-Einstein metrics on the 5-sphere o
5
.
An important result of Kodaira and Spencer ([KS58], Theorem 14.1) says that
for smooth hypersurfaces H
n,d
of degree d in the complex projective space CP
n+1
the complex analytic family given by Proposition 5.5.6 is actually complete except
when n = 2 and d = 4. It would be interesting to obtain a similar result for weighted
hypersurfaces.
Open Problem 5.5.1: Discuss the completeness problem for weighted homoge-
neous hypersurfaces in CP(w).
It is also easy to work out the dimension of the moduli space of Kahler-Einstein
metrics on A
k
equal to CP
n
blown-up at / points for certain /. For simplicity we
consider n = 2 and the del Pezzo surfaces A
k
for 0 < / 8. As discussed previously
there are no Kahler-Einstein metrics on A
1
or A
2
.
Exercise 5.3: Assuming Tians Theorem 5.2.18 show that the moduli space of
Kahler-Einstein metrics on A
k
for 4 / 8 has real dimension 4(/ 4).
Exercise 5.4: Compute the dimension of the parameter space of Kahler-Einstein
metrics for the rst three entries of Table 1 in Subsection 5.4.1 using Theorem 5.5.7.
See that it agrees with the answer in Exercise 5.3.
Exercise 5.5: Compute the dimension of the parameter space of Kahler-Einstein
metrics for the surfaces in Tables 2, 3, and 4 of Subsection 5.4.1
CHAPTER 6
Almost Contact and Contact Geometry
In this chapter we describe the basics of contact and almost contact geometry.
Contact and symplectic geometries both lie at the very heart of the mathematical
foundations of Hamiltonian and Lagrangian mechanics. As such they have been
studied extensively for over three centuries. We refer the reader interested in the
geometric description of mechanics to the classic texts of Arnold [Arn78] or Abra-
ham and Marsden [AM78] on the subject. In this chapter we rather focus on a
slightly dierent aspect of almost contact and contact structures in the way they
naturally lead towards the development of Sasakian geometry. We also do not touch
on the interesting advances in contact topology, but refer the reader to the recent
review [Gei06].
Our goal is to show how Riemannian metrics can be naturally incorporated
into the fabric of contact geometry. Starting with a quasi-regular (strict) contact
manifold (`. ) we will consider it simultaneously with the associated symplectic
cone on one hand and the symplectic orbifold transverse space on the other. At
the beginning of this chapter we shall develop all the necessary non-Riemannian,
i.e., contact and symplectic ingredients. At the end we will arrive at the denition
of the Sasakian geometry together with the full description of how it relates to the
two Kahler geometries of the cone and the transverse space, respectively. We can
summarize the context and the structure of Chapter 6 by putting together the two
diagrams of the Introduction.
(((`). ) (`. . )

(Z. )
=
(((`). .

. o) (`. . . . o)

(Z. . J. o)
The geometric structures described in the second box will be the main subject of
Chapter 7. We remark yet again that, unfortunately, the terminology is not com-
pletely standard although preference has been given to dening a contact structure
as a codimension one subbundle of the tangent bundle that is as far as possible from
being integrable (see Proposition 6.1.9 below). However, what is worse, is that pref-
erence has also been given to dening an almost contact manifold [Gra59] in terms
of an equivalence class of reductions of the orthonormal frame bundle to the group
l(n) 1. and with this convention a contact manifold is not necessarily an almost
contact manifold! Although this, in the words of Stong [Sto74], is terrible, it
seems to be too well ensconced in the literature to be rectiable. Fortunately, for
179
180 6. ALMOST CONTACT AND CONTACT GEOMETRY
our purposes this problem of terminology is mute, since the confusion only arises
when the contact line bundle is not trivial, and the contact structures of interest to
us always have trivial contact line bundle. Nevertheless, we take some pains to set
things up correctly in the beginning. We shall then revert to the usual convention.
6.1. Contact Structures
6.1.1. Contact Transformations in R
2n+1
. Contact transformations arose
in the theory of Analytical Mechanics developed in the 19th century by Hamilton,
Jacobi, Lagrange, and Legendre. But its rst systematic treatment was given by
Sophus Lie [Lie77] in 1896 under the German name Ber uhrungstransformation.
Consider R
2n+1
with Cartesian coordinates (r
1
. . . . . r
n
; n
1
. . . . . n
n
; .). together with
a 1-form given by
(6.1.1) = d.

i
n
i
dr
i
.
It is easy to see that satises
(6.1.2) (d)
n
= 0
on R
2n+1
. A 1-form that satises this equation is called a contact form.
Remark 6.1.1: The choice of in equation (6.1.1) is not unique. Other choices
are probably just as reasonable, for example,

t
= d. +
1
2
n

i=0
(r
i
dn
i
n
i
dr
i
) .
This clearly satises (6.1.2), and diers from by an exact form, namely,
1
2
d(x y).
We shall mainly use the form of equation (6.1.1) which we call the standard
1-form on R
2n+1
. There is local converse to this, namely, the contact version of
Darbouxs theorem.
Theorem 6.1.1: Let be a 1-form on R
2n+1
that satises equation (6.1.2). Then
there is an open set l R
2n+1
and local coordinates (r
1
. . . . . r
n
; n
1
. . . . . n
n
; .) such
that has the form (6.1.1) in l.
Darbouxs theorem says that locally all contact forms look the same. A proof
can be found in most books on Analytical Mechanics, e.g., [AM78, LM87]. The
modern proof is due to Moser [Mos65] and Weinstein [Wei77a]
Denition 6.1.2: A dieomorphism : l\ where l. \ are open sets of R
2n+1
is called a contact transformation if there is a nowhere vanishing smooth func-
tion 1 such that

= 1.
If 1 1 on l, then is called a strict contact transformation.
The collection
Con
of all such contact transformations forms a pseudogroup,
called the contact pseudogroup; it is one of the transitive innite, simple pseu-
dogroups of Lie and Cartan. In fact
Con
is an example of a transitive pseudogroup
which does not come from an integrable G-structure. It is, however, related to a
non-integrable G-structure as we shall see.
Exercise 6.1: Show that the subset of strict contact transformations forms a sub-
pseudogroup
sCon
of
Con
.
6.1. CONTACT STRUCTURES 181
Next we describe the Lie algebra of innitesimal contact transformations. By
taking
t
to be a local 1-parameter group of transformations generated by a vector
eld A on l, it is easy to see that the innitesimal version of Denition 6.1.2 is:
Denition 6.1.3: An innitesimal contact transformation is a local vector
eld A dened on an open set l R
2n+1
that satises

X
= 1.
where 1 is a smooth function on l. If 1 vanishes on l. then A is called a innites-
imal strict contact transformation. Let scon(l) and con(l) denote the subsets
of all vector elds on l consisting of innitesimal strict contact transformations
and innitesimal contact transformations, respectively.
Exercise 6.2: Show that con(l) forms a Lie subalgebra of the Lie algebra of all
vector elds on l. Show that scon(l) is a Lie subalgebra of con(l).
It is easy to see that the vector space con(l) of all innitesimal contact trans-
formations on l forms a Lie algebra under the Lie bracket of vector elds. In order
to see that the contact pseudogroup
Con
is transitive, we shall compute explic-
itly the Lie algebra of innitesimal contact transformations. Using the well-known
identity
X
= d(A ) +A d, one sees that A must take the form
(6.1.3) A =

1
n

i=1
n
i
1
n
i

.

n

i=1
1
n
i

r
i
+
n

i=1

n
i
1
.
+
1
r
i

n
i
.
where 1 is an arbitrary smooth function on l known as a Hamiltonian function
for the innitesimal contact transformation A.
Example 6.1.4: Consider the subalgebra h
2n+1
of scon(R
2n+1
) spanned by the
vector elds
(6.1.4) 7 =

.
. A
i
=

r
i
. Y
i
=

n
i
+r
i

.
which satisfy the Lie bracket identities
[A
i
. Y
j
] =
ij
7 . [A
i
. 7] = [Y
i
. 7] = 0 .
h
2n+1
is known as the Heisenberg algebra, and the Lie group H
2n+1
that it generates
is the Heisenberg group. Here we have realized H
2n+1
as a Lie subgroup of the
pseudogroup
sCon
of strict contact transformations on R
2n+1
.
Exercise 6.3: Show that the Heisenberg group H
2n+1
is realized as a subgroup of

sCon
by the transformations (using vector notation)
(x. y. .) (x +a. y +b. . +c +x b) .
Notice that as manifolds H
2n+1
= R
2n+1
whose group law is given by (a. b. c)
(x. y. .) = the right hand side of the above equation. It is clear that H
2n+1
acts
transitively on R
2n+1
; hence, we have arrived at
Proposition 6.1.5:
sCon
and
Con
are transitive pseudogroups.
6.1.2. Basic Denitions. We begin with the most general denition of a
contact manifold. We give a more restricted denition below, which is the one we
concentrate on in this monograph.
Denition 6.1.6: A 2n+1 dimensional manifold ` with a
Con
-structure is called
a contact manifold. If ` has a
sCon
-structure, then it is called a strict contact
manifold.
182 6. ALMOST CONTACT AND CONTACT GEOMETRY
The denition of contact manifold appears to be due to Gray [Gra59]. This
has also been given the somewhat cumbersome name contact structure in the wider
sense, but we shall not use this terminology. A contact structure on ` gives rise to
a real line bundle L called the contact line bundle as follows. In the chart (l
i
;
i
)
we have the contact form

i
. and in the overlap l
i
l
j
there exists a nowhere
vanishing function
ij
such that
(6.1.5)

i
=
ij

j
.
The
ij
dene a cocycle for the sheaf cohomology group H
1
(`. (/(1. R)) where
(/(1. R) is the sheaf of nowhere vanishing real functions on `. and thus a real
line bundle L. The cocycle
ij
is trivial or a coboundary if for all i. , the
ij
can be taken to be positive, that is, local sections of the sheaf (/(1. R)
+
of strictly
positive functions. In this case, there are positive functions 1
i
such that for all i. ,
we have 1
i

i
= 1
j

j
. Thus, we have a nowhere vanishing global section of L
dened on each l
i
by [
U
i
= 1
i

i
. This trivializes L in which case ` admits a
global contact 1-form. In the sequel we shall deal almost exclusively with contact
manifolds with trivial contact line bundle L. In fact it is quite common to take this
as the denition of a contact manifold. We shall often consider a contact structure

Con
on a strict contact manifold `. In this case the maximal atlas dening the
strict contact structure is a subatlas of the corresponding maximal atlas dening

Con
. In this case the contact structure can be dened as an equivalence class of
globally dened contact 1-forms. Thus, we are led to our second slightly more
restrictive denition of a contact manifold.
Denition 6.1.7: A 2n + 1 dimensional manifold ` is a (strict) contact man-
ifold if there exists a 1-form , called a contact 1-form, on ` such that
(d)
n
= 0
everywhere on `. A contact structure on ` is an equivalence class of such
1-forms where
t
if there is a nowhere vanishing function 1 on ` such that

t
= 1.
For purposes of this section only, we shall refer to a contact structure in the
sense of Denition 6.1.7 as a strict contact structure. Clearly for any real line
bundle L the tensor product with itself L
2
is trivial. This gives rise to the following
result, attributed to Sasaki [Bla02]:
Proposition 6.1.8: Let `
2n+1
be a contact manifold which is not strict, then
its 2-fold cover is a strict contact manifold. In particular, every simply connected
contact manifold is a strict contact manifold.
A contact structure also gives rise to a codimension one subbundle of T`.
On l
i
we dene T
i
= ker

i
. and on l
i
l
j
equation (6.1.5) holds implying that
T
i
= T
j
. So we get a well dened subbundle T of T`. called the contact subbundle,
or contact distribution which is as far from being integrable as possible. This
terminology stems from the denition of integrability in the sense of the classical
Frobenius Theorem. Let T be a codimension one subbundle of T` so there is an
exact sequence of vector bundles
(6.1.6) 0TT`Q0 .
6.1. CONTACT STRUCTURES 183
Dualizing we see that Q

is just the contact line bundle L. so we have the exact

sequence
0LT

`T

0 .
Now T is integrable if and only if there are local sections
i
of L that satisfy

i
d
i
= 0 for all i. Hence, we say that a subbundle T of T` dened as the
kernel of the local 1-forms
i
is as far from being integrable as possible if the local
1-forms satisfy
i
(d
i
)
n
= 0. Thus, the contact subbundle actually characterizes
the contact structure, that is, we have
Proposition 6.1.9: Let ` be a (2n +1)-dimensional manifold. Then a codimen-
sion one subbundle T of T` denes a contact structure on ` if and only if T is
as far from being integrable as possible.
Denition 6.1.10: If the vector bundle T is oriented we say that the contact
manifold ` is co-oriented.
The following is a result of Stong [Sto74] which corrects an error in Proposition
2.2.1 and Theorem 2.2.2 of [Gra59]. This error was propagated in [Kob72].
Proposition 6.1.11: Let ` be (2n+1)-dimensional contact manifold with contact
bundle T. Then
(i) If n is odd, then ` is oriented.
(ii) If n is even, then ` is co-oriented. Thus, in this case ` has a strict
contact structure if and only if ` is oriented.
Proof. There is a cover l
i
of ` together with 1-forms
i
on l
i
dening
the contact structure. In the overlap l
i
l
j
= there are non-vanishing functions
1
i
. 1
j
such that 1
i

i
= 1
j

j
. Then on l
i
l
j
= we have
(d
i
)
n
[
1
=

1
j
1
i

n
(d
j
)
n
[
1
.
i
(d
i
)
n
=

1
j
1
i

n+1

j
(d
j
)
n
.
Thus, if n is odd, we get a well-dened orientation; whereas, if n is even we get an
orientation on T. In the latter case ` is orientable if and only if we can choose the
1
i
all positive, that is, if and only if the contact line bundle is trivial.
Corollary 6.1.12: Let ` be a strict contact manifold. Then ` is both oriented
and co-oriented.
Remark 6.1.2: Recall that if (`
2n
. J) is a complex manifold and

` denotes the
same smooth manifold but with reversed orientation than it is a non-trivial problem
to determine if

` admits any complex structure

J compatible with that orientation.
The simplest example is CP
2
which is not a complex manifold. For surfaces the
complete answer can be found in [Kot97]. One can ask a similar question about
any strict contact manifold (`
2n+1
. ). That is, if we reverse the orientation on
` xed by , is there a compatible contact form on

`? The answer is yes when
n = 1 due to Maritnets theorem (cf. Theorem 6.1.22). When n = 2 one can
always take . Hence, the rst dimension when this question becomes of interest
is dimension 7. Unfortunately, very little is known in general about existence and
non-existence of contact structures in higher dimensions. In particular, other than
orientability there are very few known obstructions (cf. Theorem 6.2.7).
In this book almost all contact structures that we discuss are both oriented and
co-oriented. Notice that if (`. ) is a co-oriented contact manifold then (T. d)
denes a symplectic subbundle of T`.
184 6. ALMOST CONTACT AND CONTACT GEOMETRY
Contact structures have been dened in the complex setting by Kobayashi
[Kob59b], and will play and important role in Chapters 12 and 13. Complex
contact transformations can be dened in analogy with Denition 6.1.2 by simply
replacing R
2n+1
with C
2n+1
. Then the coordinates and 1-form of equation (6.1.1)
are holomorphic. One then denes the pseudogroup
con
C
of biholomorphic maps
from open sets C
2n+1
to open sets of C
2n+1
leaving invariant. Then in complete
analogy with Denition 6.1.6 we have
Denition 6.1.13: A complex manifold ` with a
Con
C
-structure is called a com-
plex contact manifold.
Such a manifold must have odd complex dimension, and in analogy with the real
case a complex contact structure is given by locally dened holomorphic 1-forms
(l
i
.
i
) satisfying
(i)
i
(d
i
)
n
= 0 in l
i
for each i.
(ii) if l
i
l
j
= then there exists a nowhere vanishing holomorphic function
1
ij
on l
i
l
j
which satises
i
= 1
ij

j
for all distinct i. ,.
The locally dened 1-forms patch together to dene a holomorphic line bundle L,
so that the
i
are 1-forms with values in L. Then the equation

i
(d
i
)
n
= 1
n+1
ij

j
(d
j
)
n
gives a trivialization of the bundle /
M
L
n+1
. hence, an isomorphism L /

1
n+1
M
.
This entire discussion works equally well for orbifolds. We only need to take the
open sets l
i
to be the local uniformizing neighborhoods and the 1-forms
i
and
transition functions 1
ij
to be invariant under the local uniformizing groups. Of
course, /
M
must be replaced by the canonical orbibundle /
orb

. Summarizing we
have
Proposition 6.1.14: Let (A. L) be a complex contact orbifold of dimension 2n+1.
Then L (/
orb

1
n+1
. In particular, the index 1

is divisible by n+1. Conversely, if

A = (A. |) orbifold with a holomorphic line orbibundle L satisfying L
n+1
/
orb

1l, then (A. L) is a complex contact orbifold.
6.1.3. Examples. Our rst example is the most obvious.
Example 6.1.15: ` = R
2n+1
. The global 1-form given by equation (6.1.1) denes
a strict contact structure on R
2n+1
. The contact subbundle T is spanned by the
vector elds
x
i
+ n
i

z
and
y
i
. This is called the standard contact structure
on R
2n+1
.
The next example will be fundamental in the sequel.
Example 6.1.16: ` = o
2n+1
. the unit (2n + 1)-sphere. Consider the 1-form
on R
2n+2
dened in the standard Cartesian coordinates (r
0
. . . . . r
n
. n
0
. . . . . n
n
) by

n
i=0
(r
i
dn
i
n
i
dr
i
). Let denote its restriction to the unit sphere. We claim that
(d)
n
= 0 everywhere on o
2n+1
. A straightforward computation shows that
(d)
n
= 2
n
n!

i=0
r
i
dr
0
dn
0


dr
i
dn
i
dr
n
dn
n

i=0
n
i
dr
0
dn
0
dr
i

dn
i
dr
n
dn
n

S
2n+1
.
6.1. CONTACT STRUCTURES 185
where as usual

dr means the dr is omitted. Now the 1-form =

n
i=0
(r
i
dr
i
+n
i
dn
i
)
is normal to the sphere o
2n+1
and an easy computation gives
(d)
n
= 2
n
n!
n

i=0

(r
i
)
2
+ (n
i
)
2

dr
0
dn
0
dr
n
dn
n
.
This shows that (d)
n
vanishes nowhere on o
2n+1
. This denes the standard
contact structure on o
2n+1
. Notice also that is invariant under the reection
(r
i
. n
i
) (r
i
. n
i
) for all i = 0. . . . . n. Thus, we get an induced contact structure
on the real projective space RP
2n+1
as well.
There is a straightforward generalization of the example above due to Gray
[Gra59] (see also [Bla02, YK84]) that shows that any immersed hypersurface in
R
2n+2
such that no point of the tangent space to the hypersurface contains the
origin 0 R
2n+2
admits a strict contact structure. Here is the construction:
Proposition 6.1.17: Let ` be an immersed hypersurface in R
2n+2
such that no
tangent space of ` contains the origin of R
2n+2
. Then ` has a strict contact
structure.
Proof. Again consider the 1-form in Cartesian coordinates (r
0
. . . . . r
2n+1
) of
R
2n+2
dened by
=
n

i=0
(r
i
dr
n+1+i
r
n+1+i
dr
i
) .
Choose a point x
0
` together with a basis (
1
. . . . .
2n+1
) of T
x
0`. Consider
the vector w whose components are n
i
= (dr
i
)(
1
. . . . .
2n+1
). where denotes
the Hodge star operator with respect to the Euclidean metric on R
2n+2
. Now w is
perpendicular to T
x
0` viewed as a subspace of R
2n+2
. Regarding x
0
as a vector,
a simple computation gives
(d)
n
(
1
. . . . .
2n+1
) = x
0
w.
Thus, the right hand side of this equation vanishes if and only if the tangent space
T
x
0` contains the origin. So under the hypothesis of the proposition, the restriction
of to ` denes a strict contact structure on `.
Another variation on the same theme, again due to Gray, uses the cotangent
bundle T

of a manifold . On T

there is a tautological or canonical 1-form

dened as follows: if T

at the covector T

. then is dened by
(6.1.7) () = '.

` .
where : T

is the natural projection and '. ` denotes the natural pairing

between T
x
and T

x
. It is not dicult to show that (d)
n+1
= 0. Thus, the
cotangent bundle of any manifold has a natural symplectic structure.
Proposition 6.1.18: Let ` be an immersed submanifold of T

with its canonical

1-form . Then the restriction of to ` denes a contact structure on ` if and
only if the following conditions hold:
(i) dim (`
1
(r)) = n for all r .
(ii) ` = . where is identied with the zero section of T

.
(iii) No tangent space to `
1
(r) in
1
(r) contains the origin in
1
(r).
Furthermore, if the conditions are satised the contact structure is strict.
186 6. ALMOST CONTACT AND CONTACT GEOMETRY
Proof. The proof is similar to that of Proposition 6.1.17. If (r
0
. . . . . r
n
) are
local coordinates on an open set l . and (n
0
. . . . . n
n
) are coordinates in the
bres, then (r
0
. . . . . r
n
. n
0
. . . . . n
n
) are coordinates in
1
(l). In these coordinates
takes the form =

n
i=0
n
i
dr
i
. Clearly, vanishes along the zero section of
T

, so condition (ii) is necessary. As in the previous case let (

1
. . . . .
2n+1
) be a
local frame for `
1
(l). Dene a tangent vector n to the bres
1
(r) whose
components with respect to the coordinate frame in
1
(r) are
n
i
= dn
0
dr
0


dn
i
dr
i
dn
n
dr
n
(
1
. . . . .
2n+1
) .
where only the dns are removed. Then a simple computation gives
(6.1.8) (d)
n
(
1
. . . . .
2n+1
) =
n

i=0
n
i
n
i
.
Now the proposition follows as before.
As an application of this proposition we have immediately
Example 6.1.19: The unit sphere bundle T

1
of a manifold with respect to
some Riemannian metric. One easily checks that T

1
satises the three conditions
of Proposition 6.1.18. Thus, T

1
has a (strict) contact structure.
Proposition 6.1.18 suggests the following
Denition 6.1.20: Let (. ) be a symplectic manifold. A hypersurface `

is said to be of contact type if there is a 1-form on ` such that

= d.
Next is an example of a contact structure due to Gray [Gra59] which shows
(ironically) that (ii) of Proposition 2.2.1 as well as Theorem 2.2.2 of [Gra59] and
Theorem 7.3 of Chapter I of [Kob72] are incorrect. This was rst noticed in
[Sto74].
Example 6.1.21: ` = R
n+1
RP
n
. Let r
i
for i = 0. . . . . n denote the Cartesian
coordinates on R
n+1
and t
i
for i = 0. . . . . n denote the homogeneous coordinates
on the real projective space RP
n
. Let l
i
R
n+1
RP
n
be the ane neighborhood
dened by t
i
= 0. Clearly l
i

n
i=0
cover `. We dene the contact structure by a
sequence of 1-forms
i
dened in l
i
by

i
= dr
i
+

j,=i
t
j
t
i
dr
j
=

j
t
j
t
i
dr
j
.
In the overlap l
i
l
j
we see that

j
=
t
i
t
j

i
.
and this denes the contact line bundle L which is non-trivial since it is induced
by the tautological line bundle on RP
n
. Hence, there is no globally dened contact
1-form on ` which denes the contact structure, that is, the contact structure is
not strict. An easy calculation also shows that

j
(d
j
)
n
=

t
i
t
j

n+1

i
(d
i
)
n
.
so ` is orientable if and only if n is odd, and in this case ` is not co-orientable.
Stong [Sto74] noticed the following generalization of this example:
6.1. CONTACT STRUCTURES 187
Exercise 6.4: Consider the projectivized cotangent bundle : ` = P(T

).
where has dimension n+1. Let l be the open set of the chart dened by the local
coordinates (r
0
. . . . . r
n
). and let (n
0
. . . . . n
n
) denote the homogeneous coordinates
in the bres of
1
(l) T

. Let l
i
` be the open set of
1
(l) dened by
setting n
i
= 0. On l
i
dene the 1-form
i
= (n
i
)
1

n
j=0
n
j
dr
j
. Show that the
i
s
patch together to dene a contact structure on `.
For further examples see Blairs recent book [Bla02]. Of particular interest is
the fundamental theorem of Martinet in dimension three [Mar71].
Theorem 6.1.22: Every orientable 3-manifold admits a contact structure.
Following Blair [Bla02] we give an explicit contact structure on the 3-torus T
3
.
Example 6.1.23: ` = T
3
= R
3
Z
3
. Let (r. n. .) denote standard Cartesian coor-
dinates in R
3
. and consider the 1-form
= sin ndr + cos nd. .
Then we nd d = dr dn d.. The contact subbundle T is spanned by the
vector elds

n
. cos n

r
sin n

.

.
In particular, Blair discusses the so-called overtwisted contact structures in
dimension 3 of Eliashberg. These seem to have no analogues in higher dimensions.
Lutz [Lut79] proved the existence of contact structures on principal T
2
-bundles
over 3-manifolds. In particular, the 5-torus T
5
admits a contact structure. It has
recently been proven that every odd dimensional torus admits a contact structure
[Bou02].
Exercise 6.5: Let (r
1
. . . . . r
5
) denote the standard coordinates on T
5
= R
5
Z
5
.
Show that the 1-form
= (sin r
1
cos r
3
sin r
2
sin r
3
)dr
4
+ (sin r
1
sinr
3
+ sin r
2
cos r
3
)dr
5
+sin r
2
cos r
2
dr
1
sin r
1
cos r
1
dr
2
+ cos r
1
cos r
2
dr
3
denes a contact structure on T
5
.
A strict contact structure has an associated canonical vector eld introduced
by Reeb [Ree52] which is of great importance in the sequel.
Lemma 6.1.24: On a strict contact manifold (`. ) there is a unique vector eld
, called the Reeb vector eld, satisfying the two conditions
= 1. d = 0 .
Proof. The volume element (d)
n
gives an isomorphism between the
C

(`)-module of vector elds on ` and the C

(`)-module of 2n-forms on
`. Thus, there is a unique vector eld that satises the equation ( (d)
n
) =
(d)
n
. Contracting again with respect to gives (d)
n
= 0. and expanding gives
(d)
n
= ()(d)
n
( (d)
n
) = ()(d)
n
which implies () = 1. But then
since d is a 2-form 0 = (d)
n
= n( d) (d)
n1
. But since the rank of the
2-form d is n. this implies that d = 0.
The Reeb vector eld is often called the characteristic vector eld and it
uniquely determines a 1-dimensional foliation T

on (`. ) called the character-

istic foliation. We let 1

denote the trivial line bundle consisting of tangent vectors

188 6. ALMOST CONTACT AND CONTACT GEOMETRY
that are tangent to the leaves of T

. This splits the exact sequence 6.1.6 as

(6.1.9) T` = T 1

.
Denition 6.1.25: The characteristic foliation T

of a strict contact structure is

said to be quasi-regular if there is a positive integer / such that each point has a
foliated coordinate chart (l. r) such that each leaf of T

passes through l at most /

times. If / = 1 then the foliation is called regular. We use the terminology non-
regular to mean quasi-regular, but not regular; whereas, if T

is not quasi-regular,
it is said to be irregular.
The notion of quasi-regularity in Denition 6.1.25 is due to Thomas [Tho76]
who used the terminology almost regular instead of quasi-regular. He then proved
that the ow of the Reeb vector eld for any almost regular contact structure is
equivalent to a locally free circle action. However, this follows from the earlier
theorem of Wadsley [Wad75] 2.6.12.
Remark 6.1.3: We will also say that the Reeb vector eld is regular (quasi-
regular, non-regular, irregular). Note that since the Reeb vector eld is uniquely
determined by the contact form, it is perfectly acceptable to talk about a regular
(quasi-regular, non-regular, irregular) contact form on ` as well. However, in a
given contact structure dierent choices of the 1-form produce dierent character-
istic foliations which can have drastically dierent behavior. We shall see explicit
examples of this in Chapter 7 (cf. Example 7.1.12). Hence, it makes no sense to
say, for example, that a contact structure on ` is regular. Nevertheless, we can
ask
Question 6.1.1: Given a contact manifold (`. T) does there exist a regular 1-form
in the contact structure T?
When the answer is yes, by abuse of language, we will sometimes say that a contact
manifold ` is regular, meaning that ` is a strict contact manifold (`. ) which
has been quipped with a regular contact from .
In 1958 Boothby and Wang [BW58] (cf. Theorem 8.3.5) studied homogeneous
contact manifolds which they proved must be regular in the sense of Question 6.1.1.
This will be discussed in Chapter 8. In particular, they show that a regular strict
compact contact manifold has a canonical structure of a principal circle bundle over
a compact symplectic manifold.
Theorem 6.1.26: Let (`. ) be a regular compact strict contact manifold. Then
` is the total space of a principal circle bundle : `Z over the space of leaves
Z = `T

. Furthermore, Z is a compact symplectic manifold with symplectic

form , [] H
2
(Z. Z), and is a connection form on the bundle with curvature
d =

.
Remark 6.1.4: This theorem does, in a sense, serve as the backbone of the material
of the next several chapters. It is quite clear the it can easily be extended to the
quasi-regular case where instead of circle bundle and a smooth symplectic manifold
we can talk about orbibundles and symplectic orbifolds. We will discuss and prove
various generalizations of this result in this and the next chapter (cf. Theorem 6.3.8
and Theorem 7.1.3).
Let (`. ) be a strict contact manifold. An immersed submanifold : `
is called an integral submanifold of the contact structure if

= 0. An integral
submanifold is said to be maximal if it is not contained as a proper subset of any
6.1. CONTACT STRUCTURES 189
other integral submanifold. Maximal integral submanifolds of a contact manifold
are often called Legendrian submanifolds.
Proposition 6.1.27: On a contact manifold (`. ) of dimension 2n+1 the Legren-
drian submanifolds have dimension n.
Proof. It is easy to see that integral submanifolds of dimension n exist. By
Darbouxs theorem there is a coordinate chart (l; r
1
. . . . . r
n
. n
1
. . . . . n
n
. .) such
that takes the form of equation (6.1.1). Then dene the submanifold locally by
r
i
= o
i
. . = c for some constants o
i
. c. This shows that submanifolds of dimension
n exist. On the other hand suppose that is an integral submanifold of dimension
/. and let A and Y be vector elds that are tangent to . Then the equation
o(A. Y ) = d(A. Y ) = 0
says that A is normal to . Since is clearly normal to there can be at most
n linearly independent local vector elds tangent to .
6.1.4. Global Contact Transformations. We begin by describing some el-
ementary properties of the group of contact transformations. A good book for the
further study of this important group is that of Banyaga [Ban97].
Denition 6.1.28: Let ` be a strict contact manifold, and let Con(`. T) denote
the group of contact transformations, i.e., the subgroup of the group Diff(`) of
dieomorphisms of ` that leaves the contact distribution T invariant. Alternatively
xing a contact form such that T = ker . then Con(`. T) can be characterized
as
Con(`. T) = Diff(`) [

= 1 for 1 C

(`) nowhere vanishing .

With the 1-form xed we are also interested in the subgroup Con(`. ) of strict
contact transformations dened by the condition

= .
The uniqueness of the Reeb vector eld implies
Lemma 6.1.29: If Con(`. ) then

= .
We denote the Lie algebras
1
of Con(`. T) and Con(`. ) by con(`. T) and
con(`. ). respectively. They can be characterized as follows:
con(`. T) = A (`) [
X
= o for some o C

(`) .
(6.1.10) con(`. ) = A (`) [
X
= 0 .
where (`) denotes the Lie algebra of smooth vector elds on `. Clearly, con(`. )
is a Lie subalgebra of con(`. T). The innitesimal analog of Lemma 6.1.29 is
Lemma 6.1.30: If A con(`. ) then [A. ] = 0. so lies in the center of
con(`. ). In particular, every vector eld in con(`. ) is foliate with respect to the
characteristic foliation T

. i.e., con(`. ) is a subalgebra of fol(`. T

).
In contrast to the symplectic case, every innitesimal contact transformation
is Hamiltonian. More explicitly,
Proposition 6.1.31: The contact 1-form induces a Lie algebra isomorphism
between the Lie algebra con(`. T) of innitesimal contact transformations and
1
The quotes occur since such Lie algebras are actually associated with the corresponding
pseudogroups, not groups of global transformations.
190 6. ALMOST CONTACT AND CONTACT GEOMETRY
the Lie algebra C

(`) of smooth functions with the Jacobi bracket . . The

isomorphism is dened by
(A) = (A).
Furthermore, under this isomorphism the subalgebra con(`. ) is isomorphic to
the subalgebra C

(`)

of functions in C

(`) that are invariant under the ow

generated by the Reeb vector eld . that is the subalgebra of functions that are basic
with respect to the characteristic foliation T

.
Recall [LM87] that the Jacobi bracket is dened by 1. o = ([A
f
. A
g
]). The
function (A) is known as the contact Hamiltonian function of A. Conversely, for
any smooth function 1 the vector eld A
f
=
1
(1) is an innitesimal contact
transformation. In particular, =
1
(1).
6.1.5. Deformation Theory of Contact Structures. The deformation the-
ory of contact structures was done by Gray [Gra59], the main result being Grays
Stability Theorem which essentially says that there are no local invariants in contact
geometry. It is the analog of Mosers Stability Theorem of symplectic geometry,
cf. [MS98, Ban97]. Grays idea was to apply the Kodaira-Spencer deformation
theory to contact geometry; however, it is possible to give a proof that is similar to
the proof of Mosers Theorem [MS98]. For us it suces here to give a statement of
the result. Recall that an isotopy between two dieomorphisms
0
.
1
Diff(`)
is a smooth map : ` [0. 1]` such that
(i) (r. 0) =
0
(r).
(ii) (r. 1) =
1
(r).
(iii)
t
(r) = (r. t) Diff(`) for all t [0. 1].
Here is Grays Stability Theorem:
Theorem 6.1.32: For t [0. 1] let
t
be a smooth family of contact forms on
a compact manifold `. Then there exists an isotopy
t
and a family of smooth
nowhere vanishing functions 1
t
such that

t
= 1
t

0
for each t [0. 1].
Let T
t
= ker
t
with T
0
= T. Then this theorem says that every contact
structure in the smooth family T
t
of contact structures is contactomorphic to T.
so there are no local invariants for contact structures.
6.2. Almost Contact Structures
6.2.1. Denitions and Topological Properties. Arguably the most e-
cient way to dene almost structures is through G-structures which we shall employ.
Our rst denition of an almost contact structure is given below in 6.2.2 is com-
pletely non-standard, but at least it is terminologically consistent. As mentioned
previously the standard denition of almost contact structures [Bla76a, Bla02,
YK84] is too restrictive. For example, the contact manifold ` = R
n+1
RP
n
of
Example 6.1.21 is not an almost contact manifold in the usual sense. Recall the
almost contact G-structure of Example 1.4.15, then we have
Theorem 6.2.1: A contact structure on a manifold ` is a non-integrable almost
contact G-structure.
Proof. Let us look at the contact pseudogroup
Con
at the level of the tangent
bundle. Let
Con
, then its dierential

leaves the contact subbundle T

invariant. Choose a local 1-form
i
representing the contact structure. Then d
i
[
1
denes a symplectic structure on T over l
i
. Fix a point r l
i
. and let
Con
x
denote
6.2. ALMOST CONTACT STRUCTURES 191
the subpseudogroup generated by the vector elds in con(l
i
) that vanish at r. The
dierentials of the elements of
Con
x
span the linear isotropy algebra of con which
determines the G-structure. Let
Con
x
. then

leaves the symplectic form d

i
[
1
invariant up to a multiplicative factor. So the linear isotropy group is precisely the
group G of Example 1.4.15. Furthermore, since
i
and the symplectic form d
i
[
1
satisfy equation (6.1.2), this G-structure is not integrable.
Hence, we choose to dene an almost contact structure in terms of G-structures
as follows:
Denition 6.2.2: An almost contact structure on ` is a reduction of the
frame bundle to the group G dened in Example 1.4.15, that is, an almost contact
G-structure.
Thus, an almost contact structure picks out a codimension one subbundle T of
T` with a conformal symplectic structure on it. By reducing to maximal compact
subgroups we have
Theorem 6.2.3: A manifold ` of dimension 2n + 1 admits an almost contact
structure if and only if its frame bundle can be reduced to the group l(n) Z
2
Z
2
.
where the rst Z
2
corresponds to the sign of the conformal factor in Coj(n. R)
and the second Z
2
corresponds to the sign of the determinant. Moreover, if ` is
orientable, there is a reduction to the group l(n) Z
2
. which further reduces to the
group l(n) 1 if and only if the subbundle T of T` is orientable. Furthermore, if
` is non-orientable, but T is orientable, then there is a reduction to the subgroup
l(n) Z
2
.
Proof. The group Coj(n. R) is homotopy equivalent to Z
2
oj(n. R). where
the group Z
2
is identied with the sign of the conformal factor, and oj(n. R) is
homotopy equivalent to its maximal compact subgroup l(n). So the group G of
Example 1.4.15 is homotopy equivalent to the block matrices of the form
G =

C 0
0 /

[ C H. / Z
2

.
where H l(n)Z
2
is realized as the subgroup of O(2n. R) generated by matrices
of the form

1
1

1 0 0
0 1 0
.
.
.
.
.
. 0
0 0 1

.
where and 1 are n by n block matrices that satisfy

t
+

1
t
1 = 1l
n
. (

t
1)
t
=

t
1.
If ` is orientable we can take det ` = /. so the group reduces to l(n) Z
2
.
Moreover, the subbundle T is orientable if and only if we can take det C = 1. in
which case the group reduces to l(n) 1.
Combining Theorem 6.2.3 and Proposition 6.1.11 we have
Corollary 6.2.4: Let ` be a 2n+1-dimensional contact manifold, then the frame
bundle can be reduced to the subgroup l(n) Z
2
.
(i) If n is odd, then ` can be reduced to the group l(n) 1 if and only if
` is co-orientable.
192 6. ALMOST CONTACT AND CONTACT GEOMETRY
(ii) If n is even, then ` can be reduced to the group l(n) 1 if and only if
` is orientable.
(iii) ` has a strict contact structure if and only if ` can be reduced to the
group l(n) 1.
We shall now discuss the standard denition of an almost contact structure
[Bla76a, Bla02, YK84], which is more closely related to strict contact structures
than contact structures. However, even in this regard, the denition includes ad-
ditional structure, namely a choice of vector eld. Also from the point of view of
G-structures, it would be more consistent to dene almost contact structures using
a transverse symplectic form instead of a transverse almost complex structure.
Of course, in the presence of a metric the two are equivalent, so we shall stick with
what is commonly done. At least for the purposes of this section we shall often add
the word strict to the denition for emphasis.
Denition 6.2.5: A (strict) almost contact structure on a dierentiable man-
ifolds ` is a triple (. . ). where is a tensor eld of type (1. 1) (i.e., an endo-
morphism of T`), is a vector eld, and is a 1-form which satisfy
() = 1 and = 1l + .
where 1l is the identity endomorphism on T`. A smooth manifold with such a
structure is called an almost contact manifold.
Such an almost contact structure (. . ) has three other almost contact struc-
tures canonically associated to it, namely (. . ). (. . ). and (. . ).
Notice that an almost contact manifold is necessarily odd dimensional, and one eas-
ily veries the identities [Bla76a, Bla02, YK84]
(6.2.1) = 0. = 0 .
Thus, an almost contact manifold ` comes canonically equipped with the following
structures: Since the vector eld is nowhere vanishing, it generates a 1-dimensional
subbundle 1

of the tangent bundle T`; hence, an almost contact manifold ` has

associated to it a 1-dimensional foliation T

. called the characteristic foliation. The

1-form is called the characteristic 1-form and it plays the role of the characteristic
form in foliation theory [Ton97], even though requires a Riemannian metric for
its denition. The codimension one subbundle T = ker has an almost complex
structure dened by J = [
1
. Thus, as in the case of a contact structure there is a
canonical splitting of the tangent bundle
(6.2.2) T` T 1

.
and from the point of view of G-structures, a choice of a strict almost contact
structure on a 2n + 1 dimensional manifold ` corresponds to a reduction of the
frame bundle to the subgroup G1(n. C) 1 of G1(2n + 1. R). Thus, in particular,
an almost contact structure is naturally an almost product structure of type (2n. 1)
with a transverse almost complex structure. Hence, the subbundle T. which we
refer to as the horizontal subbundle of T`. together with the endomorphism J
provides ` with an almost CR structure (T. J) of codimension one. Combining
this discussion with Theorem 1.2.5 we have arrived at
Theorem 6.2.6: Let ` be a smooth manifold of dimension 2n + 1. There is a
one-to-one correspondence between the strict almost contact structures (. . ) on
` and the reductions of the frame bundle 1(`) to the group G1(n. C) 1.
6.2. ALMOST CONTACT STRUCTURES 193
Notice that a choice of strict almost contact structure (. . ) on ` chooses
both an orientation and a transverse orientation on (`. T

). The almost contact

structure (. . ) chooses the opposite orientation, and (. . ) chooses the
opposite orientation on T if and only if n is odd.
Corollary 6.2.4 together with standard topological obstruction theory give nec-
essary conditions for the existence of a strict almost contact structure. See Appen-
dix A for a brief discussion of characteristic classes and [MS74, LM89] for more
details. The characteristic classes on ` are obtained by pulling back the universal
characteristic classes on the classifying space 1l(n) by the classifying map. But it
is well-known that 1l(n) has vanishing odd Stiefel-Whitney classes n
2i+1
as well
as vanishing integral Stiefel-Whitney classes \
i
. The former implies the vanishing
of all Stiefel-Whitney numbers, and by a famous theorem of Thom that ` is the
boundary of a compact manifold. Summarizing, we have
Theorem 6.2.7: Let (`. . . ) be a strict almost contact manifold of dimension
2n + 1. then the odd Stiefel-Whitney classes n
2i+1
(`) vanish and all the integral
Stiefel-Whitney classes \
i
(`) vanish. Thus, all Stiefel-Whitney numbers vanish,
and ` is the boundary of a compact manifold.
Actually, this result holds a bit more generally for all three cases of Corollary
6.2.4. Theorem 6.2.7 is essentially due to Gray [Gra59] who also proved that the
vanishing of \
3
(`) is sucient in dimension 5.
Corollary 6.2.8: [Gra59] A 5-dimensional orientable manifold ` admits an al-
most contact structure (. . ) if and only if the integral Stiefel-Whitney class
\
3
(`) vanishes.
6.2.2. Equivalence and Automorphisms. Two strict almost contact man-
ifolds (`. . . ) and (`
t
.
t
.
t
.
t
) are isomorphic if there is a dieomorphism
1 : ``
t
such that
(6.2.3)
t
= 1

.
t
= (1
1
)

.
t
1

= 1

.
When (`. . . ) is a contact manifold such an isomorphism induces an isomor-
phism of the corresponding symplectic vector bundles (T. d) and (T
t
. d
t
). We
shall also be interested in certain automorphism groups. For any vector subbun-
dle 1 T` we let Aut(1) denote the subgroup of dieomorphisms that leave 1
invariant, that is,
(6.2.4) Aut(1) = Diff(`) [

1 1 .
In the case that 1 = ker for some 1-form . this is equivalent to

= 1 for some
nowhere vanishing function 1. Thus, when is a contact form Aut(T) = Con(`. T).
Similarly, if 1 = 1

the line bundle generated by a nowhere vanishing vector

eld . then Aut(1

) = Fol(T

). We also have subgroups Aut() Aut(T) and

Aut() Aut(T

) dened by
(6.2.5) Aut() = Diff(`) [

=
and
(6.2.6) Aut() = Diff(`) [

= .
194 6. ALMOST CONTACT AND CONTACT GEOMETRY
respectively. These groups are all innite dimensional and have corresponding Lie
algebras dened by
aut(T) = vector elds A [
X
= 1 for some 1 C

(`) . (6.2.7)
aut() = vector elds A [
X
= 0 . (6.2.8)
aut() = vector elds A [ [A. ] = 0 . (6.2.9)
Notice that when generates a foliation T

on ` the Lie algebra aut() is a Lie

subalgebra of the Lie algebra fol(`. T

) of foliate vector elds.

Let us now consider the case of an almost contact structure (. . ). In this
case we dene the group leaving the tensor eld invariant by
(6.2.10) Aut(`. ) = Diff(`) [

.
Its Lie algebra of innitesimal transformations is given by
(6.2.11) aut(`. ) = vector elds A [
X
= 0 .
It follows from Equations 6.2.1 that
(6.2.12) Aut(`. ) Aut(T) Fol(`. T

) .
and similarly for their Lie algebras. It is also convenient to dene the group of
CR-transformations of an almost contact manifold (`. . . ) by
(6.2.13) CR(`. T. J) = Aut(T) [

J = J

.
In the case that the almost CR structure is integrable and ` is compact, this group
was shown to be a Lie group in [CM74]. Its Lie algebra cr(`. T. J) is dened by
(6.2.14) cr(`. T. J) = A aut(T) [
X
J = 0 .
Notice that
X
J makes sense even though J is not a tensor eld on `. This
follows from the fact that the vector eld A leaves T invariant. In [BGS06] the
following lemma is proved under a more restrictive hypothesis, but it is easy to see
that the proof given there holds in the present case.
Lemma 6.2.9: Let (`. . . ) be an almost contact manifold with underlying al-
most CR structure (T. J) with T = ker and J = [
1
. Then
Aut(`. ) = CR(`. T. J) Fol(`. T

) .
We now dene the automorphism group Aut(. . ) of an almost contact struc-
ture as
(6.2.15) Aut(. . ) = 1 Diff ` [ 1

= . 1

= . 1

= 1

.
Clearly, we have Aut(. . ) = Aut() Aut() Aut(`. ). We also dene the Lie
algebra of Aut(. . ) by
(6.2.16) aut(. . ) = vector elds A [
X
=
X
= [A. ] = 0 .
However, unlike the strict contact case the vector eld does not necessarily lie in
aut(. . ). See for example Lemmas 6.3.3 and 6.3.4 below.
Notice that in the case that denes a contact structure, Lemma 6.1.29 implies
that the group Aut(. . ) can be characterized as
(6.2.17) Aut(. . ) = 1 Con(`. ) [ 1

= 1

= Con(`. ) Aut(`. ) .
6.3. ALMOST CONTACT METRIC STRUCTURES 195
Exercise 6.6: Let (`. . . ) be an almost contact manifold. Show that the Lie
algebra cr(`. T. J) can be characterized as the set of all vector elds A on ` such
that
X
= [A. ] .
The main theorem on automorphisms of strict almost contact structures is due
to Sasaki [Sas67].
Theorem 6.2.10: Let ` be a compact manifold with a strict almost contact struc-
ture (. . ). Then the group of automorphisms Aut(. . ) is a Lie group.
Proof. By Theorem 6.2.6 the Lie algebra of the structure group is gl(n. C)
gl(2n. R). which is elliptic since it has no matrix of rank one. So the theorem follows
from Proposition 1.6.7.
6.3. Almost Contact Metric Structures
In order to investigate further we need a compatible metric structure.
Denition 6.3.1: Let ` be a (strict) almost contact manifold. A Riemannian
metric o on ` is said to be compatible with the almost contact structure if for
any vector elds A. Y on ` we have
o(A. Y ) = o(A. Y ) (A)(Y ) .
An almost contact structure with a compatible metric is called an almost contact
metric structure.
One immediately sees by setting Y = in the equation above that o(. A) =
(A). It is easy to construct compatible Riemannian metrics on an almost contact
manifold. Compatible metrics are often called associated metrics [Bla02].
Exercise 6.7: Let o
t
be any Riemannian metric on a strict almost contact manifold
(`. . . ) and dene
o =
1
2

o
t
(
2

2
) +o
t
( )

+ .
Show that o denes a Riemannian metric on ` that is compatible with the almost
contact structure. Conversely, show that any compatible metric must take the form
o = o
1
+ . where o
1
is a metric on the vector bundle T.
Almost contact metric structures correspond to almost product structures of
type (2n. 1) with a transverse almost Hermitian structure. As with Theorem 6.2.6
we have
Proposition 6.3.2: There is a one-to-one correspondence between strict almost
contact metric structures (. . . o) on ` and reductions of the frame bundle to
the group l(n) 1.
If ` is a smooth manifold of dimension 2n + 1 with an almost contact metric
structure, then there is a natural 2-form on ` of rank 2n dened by
(6.3.1) (A. Y ) = o(A. Y ) .
In this case it is easy to see that the 2n + 1 form ()
n
= 0 everywhere on `.
and that is skew-symmetric. The following lemma is a reformulation of Lemma
2.6.8 in terms of almost contact manifolds.
Lemma 6.3.3: Let (`. . . ) be an almost contact manifold. Then the following
are equivalent:
196 6. ALMOST CONTACT AND CONTACT GEOMETRY
(i) There exists a compatible Riemannian metric on ` such that the orbits
of are geodesics.
(ii)

= 0.
(iii) d = 0.
We remark that for 1-dimensional foliations condition (i) in Lemma 6.3.3 is
equivalent to tautness [Ton97]. Thus, we can rephrase Lemma 6.3.3 as:
Lemma 6.3.4: Let (`. . . ) be an almost contact manifold. The associated
characteristic foliation T

is taut if and only the characteristic 1-form is invariant

under the ow of .
Next we wish to characterize those almost contact structures that have Rie-
mannian metrics that are compatible with the characteristic foliation, that is that
the characteristic foliation be a Riemannian foliation 2.5.4. As shown in Proposition
2.5.7 this is equivalent to the condition that the compatible metric be bundle-like,
cf. 2.5.6. We show that this happens precisely when the transverse metric o
1
= o[T
is invariant under the ow of , that is when

o
1
= 0.
Proposition 6.3.5: Let (`. . . ) be an almost contact manifold and o be a
Riemannian metric that is compatible with the almost contact structure (. . ).
Then o is bundle-like and T

is taut if and only if is a Killing eld of o.

Proof. By Exercise 6.7 any compatible metric o takes the form
o = o
1
. o
1
=
1
2

o
t
(
2

2
) +o
t
( )

.
where o
t
is an arbitrary Riemannian metric. The computation is local and there
is a local frame of foliate vector elds. Computing the Lie derivatives we nd for
foliate horizontal vector elds A. Y
(

o)(A. Y ) = (

o
1
)(A. Y ) . (

o)(A. ) = d(. A). (

o)(. ) = 0
from which the result easily follows by Lemma 6.3.3.
It would be interesting to nd examples of almost contact structures whose
characteristic foliation is not taut, but nevertheless admits a compatible bundle-
like metric. Next we give the orbifold version of a Theorem of Hatakeyama [Hat63]
who shows that the total space of a principal G
1
bundle over an almost complex
manifold where G
1
is a 1-dimensional Lie group admits the structure of an almost
contact manifold in a natural way. Now let (. J) be an almost complex orbifold,
and : ` a principal G
1
V-bundle over . Generally, the total space ` is
an orbifold with bres dieomorphic to G
1
. There are two cases (we assume G
1
is
connected): (1) G
1
= R and (2) G
1
= o
1
. But according to Denition 4.2.7 there
is a group homomorphism from the local uniformizing groups
i
to G
1
. But for
case (1) G
1
= R the only homomorphism from a nite group to the additive reals
is trivial. Hence, : ` is an absolute V-bundle, i.e. an ordinary real line
bundle, and these are classied by homotopy classes of maps [. 1Z
2
] H
1
(. Z
2
).
In this case the singularities of are not desingularized by the V-bundle, so if
is a non-trivial orbifold so is `. Henceforth, for case (1) we restrict ourselves to
the smooth category where Hatakeyamas theorem holds. Case (2) with G
1
= o
1
is
more interesting since ` can desingularize . By Lemma 4.2.8 this occurs precisely
when the homomorphisms /
i
:
i
o
1
are injective.
6.3. ALMOST CONTACT METRIC STRUCTURES 197
In either case Hatakeyamas construction goes through with only slight modi-
cation. Let denote the vector eld on ` generating the G
1
action, and let be
a connection 1-form on the V-bundle : `. Then with respect to the basis 1
of the Lie algebra R of G
1
. we have () = 1. and the horizontal bundle T = ker .
Now any point n ` can be written as n = ( r. o)

l o
1
. where

l is local uni-
formizing neighborhood, and there is a vector space isomorphism

: T
u
T
x

l.
Let c
u
: T
x

lT
u
denote the inverse map. Then we have a commutative diagram
(6.3.2)
T
x

l
R
h()
T

1
x

l

q
u

q
uh()
T
u
R
h()
T
uh()
.
where (

l. . ) is a local uniformizing system with . and / is the homomor-

phism? We then dene the tensor eld by
(6.3.3)
u
= c
u
J
x

.
This tensor eld is invariant under the group of the bundle, namely,
ua
= (1
a
)

u
.
Theorem 6.3.6: Let (. J) be an almost complex orbifold and let : ` be
a principal G
1
V-bundle over . Suppose that is a generator of the G
1
action
that corresponds to 1 in the Lie algebra g
1
R. and that is a connection 1-form
on `. Then (. . ) with dened by equation (6.3.3) denes an almost contact
structure on `. Furthermore, if / is a Hermitian metric on compatible with J
then
o =

/ +
is a bundle-like Riemannian metric on ` compatible with the almost contact struc-
ture, and is a Killing vector eld for o.
Our next task is to formulate a converse to Theorem 6.3.6. Notice that the
ingredients are a bundle-like metric with a Killing vector eld, and a -invariant
tensor eld . On an almost contact manifold (`. . . ) these conditions are
related mainly through Proposition 6.3.5; however, the invariance of is not pinned
down completely.
Lemma 6.3.7: On an almost contact manifold (`. . . ) the following relations
hold:

() = 0. (

) =

. (

)
h
+ (

)
h
= 0 .
where the superscript / means horizontal component. So if the tensor eld is
invariant under the ow of then T

is taut.
Proof. The rst equation follows from the rst of equations (6.2.1). To prove
the remaining two equations, take the Lie derivative with respect to of the second
equation in Denition 6.2.5 giving
(6.3.4)

.
Taking of this implies that the second equation holds on T while the second
equation of (6.2.1) implies that it hold also on . Taking the horizontal component
equation (6.3.4) gives the third equation. The last statement follows from the
second equation and Lemma 6.3.4.
198 6. ALMOST CONTACT AND CONTACT GEOMETRY
First we shall make the simplifying assumption that the leaves of the charac-
teristic foliation T

are compact. In this case the leaves are all dieomorphic to a

circle, however, the foliation need not be simple.
Theorem 6.3.8: Let (`. . . ) be an almost contact manifold such that the leaves
of the characteristic foliation are all compact. Suppose also that (`. . . ) admits
a compatible Riemannian metric o such that is a Killing vector eld which leaves
the tensor eld invariant. Then the space of leaves `T

has the structure of

an almost Hermitian orbifold such that the canonical projection : ``T

is
an orbifold Riemannian submersion, and a principal o
1
V-bundle over `T

with
connection 1-form .
Proof. Since is a Killing vector eld for the compatible Riemannian metric
o, Proposition 6.3.5 says that o is bundle-like and the leaves of T

are geodesics.
Thus, T

is a taut Riemannian foliation. Hence, since the leaves are all compact
Molinos theorem 2.5.11 implies that the space of leaves is an orbifold, and the
projection : ``T

is an orbifold Riemannian submersion. Moreover, by

Lemma 6.3.3 the leaves are geodesics, so Wadsleys Theorem 2.6.12 implies that
the leaves are the orbits of an o
1
action on `. Since the holonomy groups of T

are
nite cyclic, this action is locally free, and thus : ``T

is a principal o
1
V-bundle over `T

. Again by Lemma 6.3.3 the horizontal subbundle T of T`

is invariant under the o
1
action and thus denes a connection on : ``T

.
and it is clear from Denition 1.3.2 that is a connection 1-form of T. Finally,
for any vector eld A on `T

we can dene an endomorphism 1 on `T

by
setting (1A)
h
= A
h
where the superscript / denotes the horizontal lift. This is
well dened by the invariance of the tensor eld under the circle action. So one
easily sees that 1 denes an almost complex structure on the orbifold `T

.
We end this section by mentioning in passing the classication of types of almost
contact metric structures, although we do not make use of them in this text. This
is somewhat involved, being accomplished by studying the representations of the
structure group l(n) 1 on certain natural vector spaces [AG86, CG90, CM92].
6.4. Contact Metric Structures
The case of most interest to us is when the almost contact structure is a strict
contact structure. We now investigate this case further.
We now wish to relate strict contact structures to strict almost contact struc-
tures. While a strict contact structure uniquely determines the characteristic vector
eld , we need to add some more structure, namely the endomorphism . in order
to obtain an underlying almost contact structure! Recall that on a contact man-
ifold a choice of 1-form gives the contact bundle the structure of a symplectic
vector bundle, namely (T. d). An almost complex structure J in T is said to be
compatible (cf. [MS98]
2
) with the symplectic form d if d(JA. JY ) = d(A. Y )
for all vector elds A. Y , and d(JA. A) 0 for all A = 0 hold. We can extend
the endomorphism J to an endomorphism on all of T` by setting = 0. The
conditions then become
(6.4.1) d(A. Y ) = d(A. Y ) for all A. Y . d(A. A) 0 for all A = 0.
2
We use the opposite convention to McDu and Salamon [MS98] who use (X, JX) > 0.
6.4. CONTACT METRIC STRUCTURES 199
and one easily checks that (. . ) denes an almost contact structure on `. Ac-
cordingly, we have
Denition 6.4.1: Let (`. ) be a contact manifold. Then an almost contact struc-
ture (.
t
. ) is said to be compatible with the contact structure if
t
= . is its
Reeb vector eld, and there is a smooth section of the endomorphism bundle of
T` such that
2
= 1l + and conditions (6.4.1) hold. We denote by /(()
the set of compatible almost contact structures on (`. ).
Every compatible almost contact structure is determined uniquely by the endo-
morphism which in turn is determined uniquely by the almost complex structure
on T. So we can give the set /(() the subspace topology of the space of sections
(End T`) with the C

compact open topology. Then we have

Proposition 6.4.2: Let (`. ) be a strict contact manifold. Then there is a
one-to-one correspondence between compatible almost complex structures J on the
symplectic vector bundle (T. d) and compatible almost contact structures (. . ).
Moreover, the space /(() is contractible.
Proof. The rst statement is clear from what was said above, and the second
statement follows from Proposition 2.61 of [MS98].
Notice that d(A. Y ) is symmetric in A and Y and is essentially the Levi form
1

of the almost CR-structure dened by (T. [T). So the last of conditions (6.4.1)
implies that the Levi form 1

is strictly pseudoconvex; hence, 1

= d ( 1l)
denes a Riemannian metric o
1
on the vector bundle T. We can then dene a
metric o on ` by
(6.4.2) o = o
1
+ = d ( 1l) +
which, as one can easily check, is compatible with the almost contact structure
(. . ). i.e., (. . . o) is an almost contact metric structure. Exercise 6.7 shows
that compatible Riemannian metrics always exist on a strict contact manifold.
Riemannian metrics of the form (6.4.2) are called associated Riemannian metrics,
and there is precisely one for every compatible almost contact structure. Thus,
Proposition 6.4.3: The set of associated Riemannian metrics are in one-to-one
correspondence with elements of /(().
Denition 6.4.4: A (strict) contact manifold (`. ) with a compatible (strict) al-
most contact metric structure (. . . o) such that o(A. Y ) = d(A. Y ) is called a
contact metric structure, and (`. . . . o) is called a contact metric man-
ifold.
Remark 6.4.1: It is possible to have a contact manifold (`. ) with a compatible
almost contact structure (. . ) and a Riemannian metric o which is compatible
with the almost contact structure, but which is not the associated metric, i.e.,
o = d (1l). We refer to such a structure as a (strict) contact structure with a
compatible Riemannian metric. Of course, a contact metric structure is a contact
structure with a compatible Riemannian metric. Unless stated to the contrary, the
metric of a contact metric structure will mean an associated Riemannian metric.
An immediate corollary of Lemmas 6.3.3 and 6.1.24 is
Corollary 6.4.5: On a strict contact manifold (`. ) the characteristic foliation
T

is taut with respect to any compatible Riemannian metric.

200 6. ALMOST CONTACT AND CONTACT GEOMETRY
Lemma 6.4.6: On a strict contact metric manifold (`. . o) the 1-form is co-
closed, i.e., = 0.
Proof. Recall that = d. where is the Hodge star operator with respect
to o. The Riemannian volume dvol
g
satises
dvol
g
=
1
n!
(d)
n
.
so we have
= d =
1
n!
d(d)
n
= 0 .
We are interested in the case when the characteristic foliation T

of a contact
metric structure is a Riemannian foliation. Although a Riemannian ow is not
necessarily isometric in general, it turns out that if the characteristic ow T

of a
contact metric structure is Riemannian if and only if it is isometric. First we have
Denition 6.4.7: A contact metric structure (. . . o) is called K-contact if
is a Killing vector eld of o. i.e., if

o = 0. A manifold with such a structure is

called a K-contact manifold.
Proposition 6.4.8: On a contact metric manifold (`. . o) the following condi-
tions are equivalent:
(i) The characteristic foliation T

is a Riemannian foliation.
(ii) o is bundle-like.
(iii) The Reeb ow is an isometry.
(iv) The Reeb ow is a CR transformation, that is, it belongs to CR(`. T. J).
(v) The Reeb ow leaves the (1. 1) tensor eld invariant, that it belongs to
Aut(`. ).
(vi) The contact metric structure (`. . o) is K-contact.
Proof. The equivalence of (i) and (ii) is Proposition 2.5.7. The two condi-
tions of Lemma 6.1.24 that characterize the Reeb vector eld imply that both the
transverse symplectic form d and the contact form are invariant under the Reeb
ow. From Denition 2.5.6 o is bundle-like if and only if the transverse metric o
1
is basic, that is, if and only if the Reeb ow leaves o
1
invariant as well. Hence,
the equivalence of (ii) and (iii). Since d is invariant under the Reeb ow, o
1
is
invariant if and only if J is invariant, or equivalently if and only if is invariant.
This implies the equivalence of (iii) with both (iv) and (v). Also (iii) and (vi) are
clearly equivalent.
We are particularly interested in a certain type of CR-structure, namely when
the Levi form 1

= d ([
1
1l) is strictly pseudoconvex. The following denition
will prove convenient.
Denition 6.4.9: Let (T. J) be a strictly pseudoconvex almost CR structure on
`. We say that (T. J) is of Sasaki type if there exists a K-contact structure
o = (. . . o) such that T = ker and [
1
= J.
Clearly, if the almost CR structure on ` is integrable, then the K-contact
structure o will be Sasakian.
6.5. STRUCTURES ON CONES 201
6.5. Structures on Cones
We wish to describe certain types of conical geometry. Let ` be a smooth
manifold, and consider the cone on ` as C(`) = ` R
+
. We shall identify `
with ` 1. We rst describe the Riemannian situation.
6.5.1. Riemannian Cones. In this case the natural Riemannian structure
on C(`) is not the product metric, but the so-called warped product [ON83]
`
r
2 R
+
. where : denotes the coordinate on R
+
. More generally, we have
Denition 6.5.1: For any Riemannian metric o
M
on `. the warped product
metric on C(`) = R
+
` is the Riemannian metric dened by
o = d:
2
+
2
(:)o
M
.
where : R
+
and = (:) is a smooth function, called the warping function. If
(:) = : then (C(`). o) is simply called the Riemannian cone or metric cone
on `.
In this and the following chapters we will almost exclusively deal with the
metric cones. However, a more general cone metrics will also be discussed in Chap-
ter 14. It is clear that there is a one-to-one correspondence between Riemannian
metrics on ` and cone metrics on C(`). The cone metric admits a group of ho-
mothety transformations dened by (r. :) (r. :). The innitesimal generator of
the homothety group is the Liouville (or Euler) vector eld dened by
(6.5.1) = :

:
.
It is clear that the metric cone of Denition 6.5.1 is homogeneous of degree 2 and
satises

o = 2o .
6.5.2. Almost Complex and Almost Hermitian Cones. We are inter-
ested in the geometry on C(`) that corresponds to an almost contact structure on
`. Given an almost contact structure (. . ) on ` we dene a section 1 of the
endomorphism bundle of the tangent bundle TC(`) of the cone by
(6.5.2) 1Y = Y +(Y ). 1 = .
It is easy to see that 1 denes an almost complex structure on C(`) that is ho-
mogeneous of degree 0 in :. Alternatively, 1 is invariant under the ow of and
we
(6.5.3)

1 = 0 .
Conversely, we begin with an almost complex structure 1 on C(`) such that
(i) I is invariant under . i.e., equation (6.5.3) holds.
(ii) The vector eld 1 is tangent to `.
From this one easily sees that = 1 denes a nowhere vanishing vector eld on `.
Letting 1

denote the trivial line bundle generated by . we have an exact sequence

01

T`Q0 .
and there is a one-to-one correspondence between the splittings of this exact se-
quence and 1-forms on ` that satisfy () = 1. This correspondence is given by
202 6. ALMOST CONTACT AND CONTACT GEOMETRY
ker . One can then easily dene an endomorphism of T` by
(6.5.4) =

1. on ker
0 on 1

.
It is easy to check that
2
= 1l + . so (. . ) denes an almost contact
structure on `.
The choice of in the paragraph above can be determined by giving a compat-
ible Riemannian structure. Explicitly,
Proposition 6.5.2: There is a one-to-one correspondence between almost contact
structures (. . ) on ` and almost complex structures 1 on C(`) satisfying the
two conditions above and equation (6.5.4). Furthermore, if o is a Riemannian
metric on ` that is compatible with the almost contact structure, then the cone
metric / = d:
2
+:
2
o is Hermitian with respect to the almost complex structure 1. and
the one-to-one correspondence above induces a one-to-one correspondence between
almost contact metric structures on ` and almost Hermitian cone structures on
C(`).
Proof. As described in the paragraph above an almost contact structure
(. . ) on ` determines a unique almost complex structure 1 on C(`) that
is invariant under . and vice-versa. Furthermore, if o is a metric on ` satisfying
o(A. Y ) = o(A. Y ) (A)(Y ). then one checks (exercise) that the cone metric
/ = d:
2
+ :
2
o is Hermitian, that is that / satises /(1A. 1Y ) = /(A. Y ). where 1
is given by equation (6.5.2). This correspondence is clearly injective. Finally, given
an almost Hermitian structure (1. /) where / is a cone metric and 1 is invariant
under . we dene by (A) =
1
r
2
/(1. A) and by equation (6.5.4) and check
that o is compatible with the almost contact structure.
Exercise 6.8: Fill in the details of the proof of Proposition 6.5.2 by showing that
the cone metric / = d:
2
+ :
2
o is Hermitian if and only if o is compatible with the
almost contact structure.
6.5.3. Symplectic Cones. We begin with the following
Denition 6.5.3: A symplectic cone is a cone C(`) = ` R
+
with a sym-
plectic structure which has a one parameter group
t
of homotheties whose inn-
itesimal generator is a vector eld on R
+
.
Denition 6.5.3 is referred to as a symplectic Liouville structure in [LM87].
Notice that there is a freedom regarding the degree of homogeneity of the homoth-
eties. As we shall want the symplectic structure on C(`) to be compatible with
the standard cone metric given in Denition 6.5.1, we ask that the symplectic form
be homogeneous
3
of degree 2. that is the equation
(6.5.5) 2 =

= d( )
holds. Notice that it follows from this equation that is an exact 2-form, so we
dene the 1-form on C(`) by = giving = d . We can dene a 1-form
on ` by = [
M1
. Then we see that = :
2
. and this gives as
(6.5.6) = 2:d: +:
2
d .
3
Note that in [BG00a] we took to be homogeneous of degree one.
6.5. STRUCTURES ON CONES 203
The condition that
n+1
= 0 implies that gives ` the structure of a strict
contact manifold with a xed 1-form. Conversely, one can easily reverse this proce-
dure: given a strict contact manifold (`. ). dening on C(`) by equation 6.5.6
gives the cone C(`) a symplectic structure with homotheties. In [Eli98, MS98]
(C(`). ) is called a symplectization of ` while it is called a symplectication of
` in [LM87]. The terminology is apparently due to Arnold [Arn78]. We have
arrived at the well-known result
Proposition 6.5.4: Let be a 1-form on the manifold `. Then denes a strict
contact structure on ` if and only if the 2-form d(:
2
) denes a symplectic struc-
ture on the cone C(`) = ` R
+
.
Let (. o) be a contact metric structure on ` with o an associated metric. Then
by Proposition 6.4.3 there is a unique almost contact structure (. . ) satisfying
o(A. Y ) = d(A. Y ) and d = 0. The metric o = d ( 1l) + on `
corresponds to the metric d:
2
+:
2
(d (1l) + ) on C(`). We have arrived
at
Proposition 6.5.5: There is a one-to-one correspondence between the contact met-
ric structures (. . . o) on ` and almost Kahler structures (d:
2
+ :
2
o. d(:
2
). 1)
on C(`).
So what does the K-contact condition on ` correspond to? We have
Proposition 6.5.6: A contact metric structure (. . . o) is K-contact if and only
if i pseudo-holomorphic with respect to the almost Kahler structure (d:
2
+
:
2
o. d(:
2
). 1) on C(`).
Proof. By the properties of the Reeb vector eld one easily sees that i is
pseudo-holomorphic, i.e., an innitesimal automorphism of 1 if and only if

1 =

= 0. But by Proposition 6.4.8 this holds if and only if

o
1
= 0.
Remark 6.5.1: In the case that the complex vector eld i on C(`) is not
pseudo-holomorphic, a quotient formed by dividing by the resulting C

action, or
equivalently, by the symplectic reduction of the o
1
action will lose both the almost
complex structure and the Riemannian structure.
6.5.4. Integrability and Normality. The main thrust of this monograph
will deal with the case in which the complex structure on the C(`) is integrable.
This leads to [SH62]
Denition 6.5.7: An almost contact structure (. . ) is said to be normal if the
corresponding almost complex structure 1 on C(`) is integrable.
Combining this denition with Proposition 6.5.2 immediately gives
Proposition 6.5.8: There is a one-to-one correspondence between the normal al-
most contact structures (. . ) on ` and almost complex structures 1 on C(`)
that are integrable and satisfy the conditions of Proposition 6.5.2. Furthermore, if
o is a Riemannian metric on ` that is compatible with the almost contact struc-
ture, then the cone metric / = d:
2
+ :
2
o is Hermitian with respect to the complex
structure 1.
As is well-known and discussed in Example 1.4.9 the integrability of 1 is equiv-
alent to the vanishing of its Nijenhuis torsion tensor
I
given by equation (1.4.4).
More generally the Nijenhuis torsion tensor

can be dened for any (1. 1)-tensor

204 6. ALMOST CONTACT AND CONTACT GEOMETRY
eld . viz.
(6.5.7)

(A. Y ) = [A. Y ] +
2
[A. Y ] [A. Y ] [A. Y ] .
Note that the Nijenhuis torsion tensor of a (1. 1)-tensor eld is a tensor eld of type
(1. 2). and that

(Y. A) =

(A. Y ). In particular we want to determine the

normality condition in terms of the Nijenhuis torsion tensor for the tensor eld
of an almost contact structure.
Theorem 6.5.9: An almost contact structure (. . ) on ` is normal if and only
if

= 2 d.
Proof. Since
I
is a tensor eld of type (1. 2) on C(`). it suces to compute
it on pairs of vector elds of the form A. Y and A. . where A and Y are vector
elds on `. Using equation (6.5.2)

I
(A. Y ) = [1A. 1Y ] [A. Y ] 1[A. 1Y ] 1[1A. Y ]
= [A +(A). Y +(Y )] [A. Y ]
1[A. Y +(Y )] 1[A +(A). Y ]
= [A. Y ] +
2
[A. Y ] [A. Y ] [A. Y ] + [(A). Y ]
+ [A. (Y )] ([A. Y ]) A(Y )1 +Y (A)1
([A. Y ] + [A. Y ])
=

(A. Y ) (Y (A) A(Y ) ([A. Y ] + [A. Y ]))

+ (A(Y ) Y (A) ([A. Y ]))
=

(A. Y ) + 2d(A. Y ) + 2(d(A. Y ) +d(A. Y ))

=

(A. Y ) + 2d(A. Y ) 2

(
Y
)(A) (
X
)(Y )

.
Next we compute

I
(A. ) = [1A. 1] [A. ] 1[A. 1] 1[1A. ]
= [A +(A). ] + [A. ] +([A. ])
[A +(A). ] ([A +(A). ])
= [A. ] + [A. ] +

([A. ]) +(A)

= (

)(A) + (

)(A).
At this stage it is traditional to dene certain tensor elds, and we shall follow suit.

(1)
(A. Y ) =

(A. Y ) + 2d(A. Y ) . (6.5.8)

(2)
(A. Y ) = (
X
)(Y ) (
Y
)(A) . (6.5.9)

(3)
(A) = (

)(A).
(4)
(A) = (

)(A) . (6.5.10)
Equating to zero separately the components of
I
(A. Y ) that are tangent to `
and those that are proportional to , shows that (. . ) is normal if and only if

(i)
(A. Y ) = 0 for all i = 1. . . . . 4. However, the following lemma shows that the
vanishing of
(1)
is both necessary and sucient for normality.
Lemma 6.5.10: If
(1)
vanishes so does
(i)
for i = 2. 3. 4.
6.5. STRUCTURES ON CONES 205
Proof. For any vector eld A on `. consider

(1)
(A. ) =

(A. ) + 2d(A. )
=
2
[A. ] [A. ] (A) ([A. ])
= [A. ] +([A. ]) [A. ] (A) ([A. ])
= [A. ] [A. ] (A) .
Applying to this equation shows that the vanishing of
(1)
implies ([A. ]) +
(A) = 0 which is equivalent to

= d = 0. This shows that

(4)
= 0.
Next we notice that replacing A by A in the equation for
(1)
(A. ). gives
(6.5.11)
(1)
(A. ) = [. A] [. A] = (

)(A) =
(3)
(A) .
So the vanishing of
(1)
implies the vanishing of
(3)
.
Finally, we apply to
(1)
(A. Y ) which gives
(
(1)
(A. Y )) = (

(A. Y )) + 2d(A. Y )
= ([
2
A. Y ]) + 2d(A. Y )
= ([A. Y ]) +([(A). Y ]) + 2d(A. Y )
= ([A. Y ] Y (A) +(A)([. Y ]) + 2d(A. Y )
= 2d(A. Y ) 2d(Y. A) 2(A)d(. Y )
= 2

(
X
)(Y ) (
Y
)(A)

2(A)d(. Y )
= 2
(2)
(A. Y ) 2(A)d(. Y ) .
But we saw in the argument for the vanishing of
(4)
that the vanishing of
(1)
implies d = 0. hence, the last term on the right vanishes whenever
(1)
vanishes,
so the vanishing of
(1)
implies the vanishing of
(2)
.
This completes the proof of the theorem.
Let us compare normality with the integrability of the underlying almost CR-
structure.
Corollary 6.5.11: An almost contact metric structure (. . . o) on ` is normal
if and only if
(i) The almost CR-structure (T. [
1
) is integrable, and
(ii) is invariant under . i.e.,
(3)
= 0.
Proof. We set J = [
1
and compute the Nijenhuis tensor for A. Y sections
of T.

J
(A. Y ) = [JA. JY ] [A. Y ] J([A. JY ] +[JA. Y ]) =

(A. Y ) +2d([A. Y ]) .
Moreover, from equations (6.5.8) and (6.5.11) we have
(

)(A) =
(1)
(A. ) =

(A. ) + 2d(A. ) .
Using these two equations the result follows from Theorem 6.5.9.
In the case of a contact metric structure the four tensor elds dened above
simplify a bit.
Proposition 6.5.12: Let (. . . o) be a contact metric structure. Then
(2)
=

(4)
= 0. Furthermore,
(3)
= 0 if and only if (. . . o) is K-contact.
206 6. ALMOST CONTACT AND CONTACT GEOMETRY
Proof. First we have

(4)
= (

) = d(()) + d = 0
when (. . . o) is contact. Next we have

(2)
(A. Y ) = (
X
)(Y ) (
Y
)(A)
= A()(Y )) ([A. Y ]) Y ()(A)) +([Y. A])
= 2d(A. Y ) 2d(Y. A) = 2o(Y. A) 2o(A. Y ) = 0 .
For
(3)
we notice rst that since (. . . o) is contact,

d = 0. Since o
is the associated metric 6.4.2, we have

o = d (

1l) .
So the vanishing of
(3)
implies that is Killing. Conversely, if is Killing, the
non-degeneracy of the bilinear form d on T implies that
(3)
(A) = 0 when A is
a section of T. But since = 0.
(3)
vanishes identically on .
Our nal denition of this subsection is the major object of study in this mono-
graph.
Denition 6.5.13: A normal contact metric structure o = (. . . o) on ` is
called a Sasakian structure. A pair (`. o) called a Sasakian manifold.
It follows immediately from Corollary 6.5.11 and Proposition 6.5.12 that
Proposition 6.5.14: A Sasakian structure o = (. . . o) is a K-contact structure
whose almost CR-structure is integrable.
Actually as seen in Lemma 6.3.3
(4)
vanishes as soon as there is a compatible
metric of the almost contact structure such that integral curves of are geodesics.
From the denition of normality 6.5.7 and Proposition 6.5.5, Denition 6.5.13 is
equivalent to:
Denition 6.5.15: A contact metric manifold (`. . . . o) is Sasakian if its
metric cone (C(`). d:
2
+:
2
o. d(:
2
). 1) is Kahler.
CHAPTER 7
K-Contact and Sasakian Structures
In this chapter we begin a detailed study of our main subject: Sasakian Geome-
try. Since there are many properties that also hold for K-contact structures, we will
often consider both. The reason for imposing the K-contact condition and not work-
ing with the more general contact metric structures comes from Proposition 6.4.8
which says, among other things, that a contact metric structure is K-contact if and
only if the characteristic foliation T

is Riemannian. The latter condition is neces-

sary for much of the geometry that we develop. Indeed, early (circa 1960) in the de-
velopment of global contact geometry the nice situation was that described by the
well-known Boothby-Wang bration (Theorem 6.1.26 of Chapter 6). As mentioned
in Chapter 6 Sasakian structures were rst studied by Sasaki and Hatakeyama in
[SH62] under the name normal contact metric structures (cf. Denition 6.5.13).
In 1963 Hatakeyama, Ogawa, and Tanno [HOT63] introduced the name K-contact
manifold, and proved the basic Theorem 7.3.7 found below. There was a systematic
development of the curvature properties of both K-contact and Sasakian geometries
throughout the sixties and seventies, see the books [Bla76a, Bla02, YK84]. The
present chapter summarizes all the basic geometric properties of such manifolds
(`. . . . o) as well as adding some more recent developments of a topological
nature due mainly to Rukimbira and those of a algebro-geometric nature due to
the authors and their collaborators. Throughout this chapter we will learn how
most of these properties, starting with the key structure Theorem 7.1.3, closely
evolve around the geometric objects depicted in the following diagram:
(((`). .

. o) (`. . . . o)

(Z. . J. /).
7.1. Quasi-regularity and the Structure Theorems
In this section we begin with a more detailed discussion of the regularity prop-
erties introduced in Denition 6.1.25. To best illustrate this denition consider
the linear ows on the sphere described by Example 2.3.5. It is not dicult to
show that these ows are the characteristic ows belonging to K-contact, in fact,
Sasakian structures on o
2n+1
. We shall see this explicitly shortly. Note that the
case in Example 2.3.5 when the ratio n
0
n
1
is rational (irrational) corresponds to
a quasi-regular (irregular) contact form, respectively. For quasi-regular K-contact
or Sasakian structures we have:
207
208 7. K-CONTACT AND SASAKIAN STRUCTURES
Denition 7.1.1: The order of a quasi-regular K-contact structure o is the least
common multiple (assuming it exists) of the orders of the leaf holonomy groups of
the characteristic foliation, and is denoted by (o).
A K-contact manifold with < is said to be of nite order. In the quasi-
regular case the order (o) coincides with the order of the orbifold Z of Denition
4.1.5. Of course, any compact quasi-regular K-contact manifold is of nite or-
der, and the order of a regular K-contact structure is one. In the compact case
quasi-regularity implies K-contact. In fact, compactness of the leaves is all that is
necessary. We see that from the existence of the Reeb vector eld 6.1.24, Lemma
6.3.3, and Wadsleys Theorem 2.6.12 we obtain
Proposition 7.1.2: Let (. ) be a quasi-regular strict contact structure on ` such
that the leaves of the characteristic foliation T

are compact. Then the Reeb ow

generates a locally free circle action on `. Furthermore, there exists a Riemannian
metric o such that the induced structure (. . o) is K-contact.
Proof. Since the Reeb vector eld satises

= 0. Lemma 6.3.3 implies

that there exists a compatible Riemannian metric such that the leaves of T

are
geodesics. Thus, Wadsleys Theorem implies that the leaves are the orbits of a
locally free circle action.
To prove the second statement we let
t
denote the Reeb ow generating the
locally free circle action. Let o
0
be any compatible metric. The orbits
t
(r) are
geodesics with respect to this metric, but o
0
is not necessarily invariant under the
circle action generated by
t
. However, the metric o dened by
o =

S
1

t
o
0
dt
is invariant under the o
1
-action, and the Reeb ow
t
is an isometry. Furthermore,
Lemma 6.3.3 implies that the orbits
t
(r) are also geodesics of o. Hence, (`. . o)
is K-contact by Proposition 6.4.8.
The second statement of this proposition is due to Rukimbira [Ruk95a]. We
now have a special case of Theorem 6.3.8 for K-contact manifolds, namely our rst
structure theorem:
Theorem 7.1.3: Let (`. . . . o) be a quasi-regular K-contact manifold with
compact leaves. Then
(i) The space of leaves `T

is an almost Kahler orbifold Z such that the

canonical projection : ``T

is an orbifold Riemannian submer-

sion.
(ii) ` is the total space of a principal o
1
orbibundle (V-bundle) over `T

with connection 1-form whose curvature d is the pullback by of a

symplectic form on `T

.
(iii) The symplectic form denes a non-trivial integral orbifold cohomology
class, that is, [j

] H
2
orb
(`T

. Z) where j is the natural projection

dened by equation (4.3.5).
(iv) The leaves of T

are all geodesics.

(v) If the characteristic foliation T

is regular, then the circle action is free

and ` is the total space of a principal o
1
bundle over an almost Kahler,
hence, symplectic manifold dening an integral class [] H
2
(`T

. Z).
(vi) (`. . . . o) is Sasakian if and only if (`T

. ) is Kahler.
7.1. QUASI-REGULARITY AND THE STRUCTURE THEOREMS 209
Remark 7.1.1: Part (v) of this theorem is essentially a restatement of the Boothby-
Wang Theorem 6.1.26. In the regular K-contact (Sasakian) case it was proved
by Hatakeyama (cf. [Hat63], Theorem 4). Hence, Theorem 7.1.3 is the orb-
ifold/orbibundle generalization of the Boothby-Wang bration. Actually the K-
contact condition can be dropped as seen by Proposition 7.1.2.
The reader should recall the relation between quasi-regular Sasakian manifolds
and Seifert o
1
-bundles as discussed in Section 4.7. Indeed, Theorem 4.7.7 says
that o
1
orbibundles over complex orbifolds and Seifert o
1
-bundles are one and
the same. Then Theorem 7.5.2 below says that a compact quasi-regular Sasakian
manifold is nothing but an o
1
-Seifert bundle over a normal projective algebraic
variety. Of course, dimension 3 is the classical case (cf. [Orl72b] and references
therein). This was generalized to n dimensions when the projective algebraic variety
is smooth in [OW75]. As presented in Section 4.7, Kollar has recently begun a
more general study of Seifert bundles [Kol07, Kol05, Kol04, Kol06a] which has
already produced substantial results in Sasakian geometry that we discuss later in
Chapter 10. Although Seifert bundles were originally dened in the case that `o
1
is complex, it is straightforward to generalize this to the case of a locally free circle
action [Kol06a]. Here we are only interested in the case of quasi-regular K-contact
structures, and we think of these as orbibundles over symplectic orbifolds Z. As in
the complex case [OW75], if ` is a compact quasi-regular K-contact manifold of
order . there is a commutative diagram
(7.1.1) `

.
.

Z
where `

is the quotient `Z

. Generally, `

is a developable orbifold with

orb
1
(`

) Z

. Now there is an induced action of o

1
Z

= o
1
on `

which is
free, so the southwest arrow is equivalent to the projection map of a principal o
1
bundle over 7. We can also consider the cone C(`) which is a C

orbibundle over
Z in which case the commutative diagram 7.1.1 becomes
(7.1.2) C(`)

C(`

) .
.

Z
When ` is Sasakian, so is `
u
. and the orbifold is projective algebraic. In the
case the southeast arrow is a branched cover. This shows that the natural orbifold
bration associated to any compact quasi-regular Sasakian manifold factors in terms
of a branched cover and an o
1
-bundle map.
Theorems 7.1.3 and 4.3.18 give some immediate information about quasi-regular
K-contact manifolds.
210 7. K-CONTACT AND SASAKIAN STRUCTURES
Proposition 7.1.4: Let ` be a quasi-regular K-contact manifold with orbifold
bration : `Z. Then we have
(i)
i
(`) =
orb
i
(Z) for i 3.
(ii) The sequence
0
2
(`)
orb
2
(Z)Z
1
(`)
orb
1
(Z)1
is exact. In particular, if ` is simply connected, then
orb
1
(Z) = 1 and

orb
2
(Z)
2
(`) Z.
More information relating the homology and cohomology groups of ` and Z
can be obtained by studying harmonic theory as well as the Leray spectral sequence
of the map . This is treated in more detail in Section 7.4.
Example 7.1.5 (Sasakian structure on o
2n+1
[SH62]): We begin with the
standard contact structure
1
on o
2n+1
. namely the restriction of
0
=

n
i=0
(n
i
dr
i

r
i
dn
i
) to the unit sphere
o
2n+1
= z = (x. y) C
n+1
[
n

i=0
[.
i
[
2
=
n

i=0

(r
i
)
2
+ (n
i
)
2

= 1 .
where r
i
. n
i
denote the real and imaginary components of .
i
C. In complex
coordinates the 1-form
0
takes the form

0
=
i
2
n

j=0
(.
j
d .
j
.
j
d.
j
) .
The Reeb vector eld is

0
=
n

i=0

n
i

r
i
r
i

n
i

= i
n

j=0

.
j

.
j
.
j

.
j

while the (1. 1)-tensor eld

0
is obtained by restricting the standard complex
structure is
1 =
n

i=0

n
i
dr
i

r
i
dn
i

= i
n

j=0

.
j
d.
j


.
j
d .
j

on R
2n+2
to T To
2n+1
and extends trivially to the real line bundle generated by

0
. Explicitly, we nd

0
=

i,j

(r
i
r
j

ij
)

r
i
dn
j
(n
i
n
j

ij
)

n
i
dr
j
+r
j
n
i

n
i
dn
j
r
i
n
j

r
i
dr
j

.
The canonical metric o
0
on o
2n+1
is the at metric on R
2n+1
restricted to the
o
2n+1
. It has constant sectional curvature 1 and satises
o
0
= d
0
(
0
1l) +
0

0
.
It is easy to check that
0
o
0
= 0. so (
0
.
0
.
0
. o
0
) is a K-contact structure.
Moreover, since the 1 describes the standard at complex structure on C
n+1
, this
1
Our choice of 1-form
0
here is the negative of the one used in Example 6.1.16. The reason
for this is to keep the standard conventions of [Bla02, YK84] for Sasakian geometry, while at
the same time using the standard complex structure on C
n+1
and not the complex conjugate. Of
course there is no change in the underlying contact structure, but only a change of orientation or
co-orientation.
7.1. QUASI-REGULARITY AND THE STRUCTURE THEOREMS 211
structure is Sasakian. We refer to the structure o
0
= (
0
.
0
.
0
. o
0
) as the stan-
dard Sasakian structure on o
2n+1
. We shall also refer to the conjugate struc-
ture o
c
0
= (
0
.
0
.
0
. o
0
) as a standard Sasakian structure when no confusion
should arise. (See Denition 7.5.11 below).
These are both regular Sasakian structures, and are nothing but the Hopf
brations as principal o
1
bundles over the complex projective space CP
n
. Now
C
n+1
with the standard complex structure and the positive oriented C

-action
(i.e., z :c
it
z) is the total space of the tautological line bundle O(1). Since
for the Sasakian structure o
0
the circle action has the negative orientation, o
0
is
associated with the positive line bundle O(1). as it must by Theorem 7.5.1. The
conjugate Sasakian structure o
c
0
is associated with complex line bundle O(1) but
with the complex conjugate structure, and as such it is also a positive line bundle.
Before turning to the irregular case, we give an inversion theorem to Theorem
7.1.3. In the regular K-contact (and also Sasakian) case the result was observed by
Hatakeyama [Hat63].
Theorem 7.1.6: Let (Z. . J) be an almost Kahler orbifold with [j

] H
2
orb
(Z. Z).
and let ` denote the total space of the circle V-bundle dened by the class []. Then
the orbifold ` admits a K-contact structure (. . . o) such that d =

where
: `Z is the natural orbifold projection map. Furthermore, if all the local
uniformizing groups of Z inject into the structure group o
1
, then ` is a smooth
K-contact manifold.
Remark 7.1.2: Notice that the Sasakian structure on ` is not uniquely deter-
mined, but is determined only up to a gauge transformation of the o
1
V-bundle.
Once a gauge is chosen the Sasakian structure on ` is then uniquely determined by
the Kahler structure and o
1
V-bundle on Z. We discuss such gauge transformations
further in Section 7.5.
The following is immediate:
Corollary 7.1.7: Let ` be a smooth manifold that admits a quasi-regular K-
contact structure (. . . o), then the leaf holonomy groups, or equivalently, the
local uniformizing groups of the orbifold `T

are all cyclic.

To study the irregular case we recall the toral rank of a Riemannian ow 2.6.5.
Here we specialize to a K-contact structure (. . . o) on a smooth manifold `.
We denote the isometry group of (`. o) by Isom(`. o). and the group of automor-
phisms of the K-contact structure (`. . o). by Aut(. . . o). We let isom(`. o)
and aut(. . . o) denote the corresponding Lie algebras. We study these Lie groups
and Lie algebras more thoroughly in Chapter 8. When ` is compact the well-known
theorem of Myers and Steenrod says that Isom(`. o) is a compact Lie group.
Moreover, Aut(. . . o) is a closed Lie subgroup of Isom(`. o) [Tan69a]. In the
K-contact case the Reeb ow belongs to the automorphism group Aut(. . . o)
which is a compact Lie group. Thus, the closure T of the Reeb ow is a compact
commutative Lie group which lies in Aut(. . . o). Now the Reeb ow is a strict
contact transformation lying in the center of the group of strict contact transforma-
tions [LM87]; hence, it lies in the center of Aut(. . . o). It follows that T also lies
in the center of Aut(. . . o). Combining this discussion with Carri`eres Theorem
2.6.4, we have
Proposition 7.1.8: Let (`. . . . o) be a compact K-contact metric manifold.
Then the leaf closures of the Reeb ow are the orbits of a torus T lying in the
212 7. K-CONTACT AND SASAKIAN STRUCTURES
center of the Lie group Aut(. . . o) of automorphisms of (. . . o) and the Reeb
ow is conjugate to a linear ow on T.
The dimension of the maximal torus in Proposition 2.6.3 is the toral rank
rk(`. T

) of (. . . o). and the contact geometry places a smaller upper bound on

the toral rank [Ruk94].
Lemma 7.1.9: Let (`. . . . o) be a compact K-contact metric manifold of di-
mension 2n+1. Then the rank rk(`. T

) is independent of the metric o and satises

1 rk(`. T

) n+1. Moreover, rk(`. T

) rk Aut(. . . o) dim a(. . . o).

Proof. Consider the Lie algebra t of T. It consists of the Reeb vector eld
together with the innitesimal generators of the leaf closures. The projections of
these generators onto T are global sections of Molinos commuting sheaf [Mol79,
Mol88] C(`. T

). Thus, they give integral submanifolds of the subbundle T. But

from Proposition 6.1.27 the integral submanifolds of T of maximal dimension, that
is the Legendrian submanifolds of the contact structure, have dimension n. Hence,
these together with the Reeb vector eld generate a torus of dimension at most
n + 1. Furthermore, Molino shows that the commuting sheaf is independent of the
transverse metric, so rk(`. T

) is independent of o.
Remark 7.1.3: The case rk(`. T

) = 1 is the quasi-regular case, while the other

extreme rk(`. T

) = n+1 is the toric case studied in [BM93a, BM96, BG00a].

It might seem that the irregular case could prove to be intractable, but this
is not at all the case at least if ` is compact thanks to the recent approximation
theorem of Rukimbira [Ruk95a]. This result is based on Carri`eres Theorem 2.6.4
on Riemannian ows. The theorem proves that any irregular K-contact structure
can be approximated by quasi-regular ones. We thus arrive at our second structure
theorem:
Theorem 7.1.10: Let ` be compact. Any K-contact structure (. . . o) on `
is either quasi-regular or there is a sequence of quasi-regular K-contact structures
(
j
.
j
.
j
. o
j
)
j
converging in the compact-open C

-topology to (. . . o). In par-

ticular, if ` admits an irregular K-contact structure, it admits many locally free
circle actions.
Conversely, any quasi-regular strict contact structure on a compact manifold
` admits a compatible K-contact structure (. . . o).
Proof. Assume the K-contact structure is irregular. Since ` is compact we
can apply Proposition 7.1.8 which says that the closure

L of any leaf L of the
characteristic foliation T

is a torus. Furthermore, by Proposition 2.6.3

L is the
leaf of a singular Riemannian foliation

T

which is the closure of T

. The vector
elds dening

T

span a commutative Lie subalgebra t(. . . o) of the Lie algebra

a(. . . o) of innitesimal automorphisms of the K-contact structure. Since by
Proposition 7.1.8 the ow of is conjugate to a linear ow on

L. there exist a
sequence of vector elds
j
t(. . . o) such that
j
= +
j
is quasi-regular
for each , and lim
j

j
= 0. Now dene a 1-form

j
=

1 +(
j
)
.
We can choose the vector elds
j
small enough in the compact-open C

-topology
so that
j
is a contact 1-form for each ,.
7.1. QUASI-REGULARITY AND THE STRUCTURE THEOREMS 213
Next we dene the (1. 1)-tensor elds and metrics by
(7.1.3)
j
=
1
1 +(
j
)

j
.
and
(7.1.4) o
j
= d
j
(
j
1l) +
j

j
.
where T = ker = ker
j
. Again one can choose the vector elds
j
small enough so
that (
j
.
j
.
j
. o
j
) are K-contact structures converging to (. . . o). This proves
the rst statement. Since ` is compact and the ow is quasi-regular, the leaves of
T

are compact. The second statement then follows from the second statement of
Proposition 7.1.2.
An immediate corollary of Theorem 7.1.10 is
Corollary 7.1.11: Every K-contact manifold admits a xed point free circle action.
Exercise 7.1: Fill in the details of the proof of Theorem 7.1.10 by showing that
the tensor elds (
j
.
j
.
j
. o
j
) satisfy the conditions of Denition 6.4.1.
To consider non-regular, both quasi-regular and irregular examples we con-
struct deformation of the round sphere of Example 7.1.5. Such deformations will
be formalized later in section 8.2.
Example 7.1.12: Consider the standard Sasakian structure o
0
= (
0
.
0
.
0
. o
0
) de-
scribed in Example 7.1.5. Now the isometry group I(o
2n+1
. o
0
) of the round sphere
is O(2n + 2. R). and the automorphism group Aut(
0
.
0
.
0
. o
0
) of the Sasakian
structure is the subgroup of O(2n+2. R) that commutes with
0
. namely, l(n+1).
The maximal torus T
n+1
of l(n+1) has n+1 dimensions. Let c
i

n
i=0
denote the
standard basis for its Lie algebra t
n+1
= R
n+1
. Under the action of T
n+1
on o
2n+1
.
the standard basis vectors c
i

n
i=0
correspond to the vector elds H
i

n
i=0
. where
H
i
= n
i

r
i
r
i

n
i
.
and the Reeb vector eld
0
can be written as
0
=

i
H
i
.
We now deform the standard Sasakian structure of Example 7.1.5 by con-
sidering any positive vector w = (n
0
. . . . . n
n
) t
n+1
= R
n+1
with n
i
0 for all
i = 0. . . . . n. The vector eld
w
corresponding to w is the one described in Exam-
ple 2.3.5 and the corresponding 1-form is denoted by
w
. In this way we obtain a
deformed Sasakian structure o
w
= (
w
.
w
.
w
. o
w
) dened by
(7.1.5)
w
=
n

i=0
n
i
H
i
.
w
=

n
i=0
n
i
((r
i
)
2
+ (n
i
)
2
)
.
where
w
and o
w
are determined by equations (7.1.3) and (7.1.4), respectively with
, replaced by w. Notice that
(
w
) = 1 +(
w
) =
n

i=0
n
i
((r
i
)
2
+ (n
i
)
2
) 0 .
This deformation procedure was rst introduced by Takahashi [Tak78] and will be
formalized in Section 8.2. It follows (cf. Section 8.2.3) that o
w
is indeed a Sasakian
structure as long as all the components of w are positive. We shall refer to o
w
as
the weighted Sasakian structure on o
2n+1
. and we denote the unit sphere with
214 7. K-CONTACT AND SASAKIAN STRUCTURES
this Sasakian structure by o
2n+1
w
. In this case we have the linear ow as described
in Example 2.3.5.
The deformed structures o
w
, however, are not all distinct. The Weyl group
J(ol(n + 1)) = Nor(T
n+1
)T
n+1
acts as outer automorphisms on the maximal
torus, and it is well-known that J(ol(n + 1)) =
n+1
, the permutation group
on n + 1 letters. So for any
n+1
the Sasakian structures o
w
and o
(w)
are
isomorphic. Thus, we shall frequently take the weights to be ordered such that
n
0
n
1
n
n
. It is important to note that all the Sasakian structures o
w
belong to the same underlying contact structure on o
2n+1
. namely the standard
one. Later we shall discover Sasakian structures on o
2n+1
belonging to distinct,
non-standard contact structures.
If all the components of w are rational numbers, then the leaves characteristic
foliation T

are all circles and T

is equivalent to a locally free circle action by

Wadsley Theorem 2.6.12. In this case we can clear the denominators by multiplying
by their least common multiple and then redene the circle action. So without loss
of generality we can assume that the n
i
are all integers satisfying gcd(n
0
. . . . . n
n
) =
1. Now the Sasakian structure o
w
on o
2n+1
is quasi-regular and we have a natural
orbifold bration : o
2n+1
{(w) where {(w) denotes the weighted projective
space CP(w) with its natural orbifold structure as discussed in Section 4.5.
Exercise 7.2: Show that for the structures o
w
of Example 7.1.12 the conjugation
map o
w
o
c
w
is induced by an isometry of o
w
.
7.2. The Transverse Geometry of the Characteristic Foliation
7.2.1. Basic Cohomology of K-Contact Structures. Recall the basic de
Rham complex
B
(T) of a foliation and its basic cohomology ring H

B
(T) from
Section 2.4. As seen previously a K-contact structure has a taut Riemannian ow,
and in this case the spectral sequence of the complex simplies. Our treatment
follows [Ton97]. As seen in Proposition 2.6.3 the closure of the leaves of the
characteristic foliation T

is a torus T Aut(. . . o) Isom(`. o). Let (`)

T
denote the subalgebra of (`) consisting of T-invariant forms. Since the orbit of
is dense in T any basic form is T-invariant, so we have
B
(T) (`)
T
. We have
Proposition 7.2.1: There is an exact sequence of complexes
0

B
(T

(`)
T

1
B
(T

)0.
Proof. First we show that is basic for any T-invariant form . This
means the two conditions ( ) = 0 and d( ) are satised. The rst is
an identity while the second follows from

( ) = 0. since T. Furthermore,
up to sign is a map of complexes. Next we show that is surjective. Let

r1
B
(T

). then ( ) = . and if T is an element of the Lie algebra t of

T. we have

T
( ) = (
T
) +
T
= 0.
The rst term vanishes since T t a(`. . . . o). and the second term vanishes
since basic forms are T-invariant. This proves the surjectivity. Finally, exactness
of the middle term follows from the denitions of
B
(T

) and (`)
T
and the fact
that t.
Now the cohomology of the complex (`)
G
for any compact connected Lie
group is isomorphic to the usual de Rham cohomology [GHV73], so we get a long
7.2. THE TRANSVERSE GEOMETRY OF THE CHARACTERISTIC FOLIATION 215
exact cohomology sequence
(7.2.1)
H
p
B
(T

)

H
p
(`. R)
jp
H
p1
B
(T

)

H
p+1
B
(T

) .
where is the connecting homomorphism given by []
B
= [d ]
B
= [d]
B

[]
B
. and ,
p
is the composition of the map induced by with the isomorphism
H
r
((`)
T
) H
r
(`. R) mentioned above. This sequence is of fundamental im-
portance in the study of Sasakian geometry. It is the foliation analogue of the
rational or real Gysin sequence for the quasi-regular case such as in Proposition
4.7.9. In fact, we have
Proposition 7.2.2: Let (`. . . . o) be a quasi-regular K-contact manifold with
: ``T

an orbifold Riemannian submersion. Then induces a ring iso-

morphism

: H

(`T

. R)

H

B
(T

) .
Exercise 7.3: Prove Proposition 7.2.2.
On compact K-contact manifolds the basic cohomology groups enjoy some
special properties. In particular, there is a transverse Hodge theory [EKAH86,
KT87, Ton97] which we shall briey review. We begin by dening the transverse
Hodge star operator . It is dened in terms of the usual Hodge star by
(7.2.2) = ( ) = (1)
r
.
The adjoint
B
:
r
B
(T

)
r1
B
(T

) of d
B
is dened by
(7.2.3)
B
= d
B
.
The basic Laplacian
B
is dened in terms of d
B
and its adjoint
B
by
(7.2.4)
B
= d
B

B
+
B
d
B
.
The space H
r
B
(T

) of basic harmonic r-forms is then dened to be the kernel of

B
:
r
B
(T

)
r
B
(T

). or alternatively H
r
B
(T

) = ker d
B
ker
B
. The transverse
Hodge theorem [EKAH86] then says that each basic cohomology class has a unique
harmonic representative, i.e., H
r
B
(T

) = H
r
B
(T

). From transverse Hodge theory

we get
Proposition 7.2.3: Let (`. . . . o) be a compact K-contact manifold of dimen-
sion 2n + 1. Then we have
(i) The groups H
r
B
(T

) are nite dimensional and H

r
B
(T

) = 0 for : 2n.
(ii) H
2n
B
(T

) R.
(iii) The class [d]
B
H
2
B
(T

) is non-trivial.
(iv) H
2r
B
(T

) 0 for 0 : n.
(v) H
1
(`. R) H
1
B
(T

).
(vi) (transverse Poincare duality) There is a non-degenerate pairing
: H
r
(T

) H
2nr
(T

)R.
Proof. (i) follows from transverse Hodge theory in the usual way. (ii): With
j = 2n + 1 the exact sequence 7.2.1 gives an isomorphism
(7.2.5) H
2n+1
(`)
j
2n+1
H
2n
B
(T

).
proving (ii). To prove (iii) we notice that the 2-form d satises (d)
n
= 0.
and that the isomorphism ,
2n+1
in (7.2.5) (which is essentially ) implies that d
216 7. K-CONTACT AND SASAKIAN STRUCTURES
denes a non-trivial class in H
2
B
(T

). Clearly, (iv) holds since (d)

n
H
2n
B
(T

) is
non-trivial. To prove (v) we notice that the beginning of the exact sequence 7.2.1
is
0H
1
B
(T

)H
1
(`. R)
j
1
R

0
H
2
B
(T

).
By (iii)
0
is injective, so 0 = ker
0
= im ,
1
; hence, ,
1
is the zero map. (vi)
follows by transverse Hodge theory [KT87, Ton97]. The pairing is that induced
from the pairing

:
r
B
(T

)
2nr
B
(T

)R
dened by
(7.2.6)

(. ) =

M
( ) .
This induces an isomorphism on the transverse harmonic forms, and denes a
non-degenerate inner product '. `
B
=

(. ). See [KT87, Ton97] for details.

Remark 7.2.1: The class [d]

B
H
2
B
(T

) is called the fundamental basic class of

the K-contact structure (. . . o). and if ` is compact [d]
B
= 0.
Next we consider some fundamental invariants of K-contact structures [BGN03a].
We dene the basic Betti numbers
(7.2.7) /
B
r
(T

) = dim H
r
B
(T

).
and the basic Euler characteristic of T

given by
(7.2.8) (T

) =
2n

r=0
(1)
r
/
B
r
(T

).
Notice that by (v) of Proposition 7.2.3 /
1
(T

) = /
1
(`), a topological invariant of
`. Using transverse Poincare duality we have
Proposition 7.2.4: On a compact K-contact manifold `
2n+1
we have
/
B
2nr
(T

) = /
B
r
(T

). (T

) = 2 2/
B
1
(T

) + 2/
B
2
(T

) + (1)
n
/
B
n
.
We end this section with a result from [EKAN93] which implies that the basic
cohomology ring is an invariant of a K-contact structure.
Theorem 7.2.5: Let `. `
t
be manifolds of dimension 2n + 1 with complete K-
contact structures o = (. . . o) and o
t
= (
t
.
t
.
t
. o
t
). respectively. Suppose
further that there exists a dieomorphism 1 : ``
t
sending o to o
t
. i.e., such
that

t
= 1

.
t
= (1
1
)

.
t
1

= 1

. o
t
= (1
1
)

o .
Then, 1 induces a ring isomorphism 1

: H

B
(T

)H

B
(T

).
Actually in Theorem 7.4.14 below we shall see that for any Sasakian structures
on dieomorphic manifolds, the basic cohomology groups are isomorphic. However,
it says nothing about the ring structure.
7.2. THE TRANSVERSE GEOMETRY OF THE CHARACTERISTIC FOLIATION 217
7.2.2. Basic Cohomology of Sasakian Structures. The basic cohomology
groups H
p,q
B
(T

) are fundamental invariants of a Sasakian structure which enjoy

many of the same properties as the ordinary Dolbeault cohomology of a Kahler
structure. Here we explore some of the elementary properties.
Theorem 7.2.6: Let (`. o) be a compact Sasakian manifold of dimension 2n+1.
Then we have
(i) H
n,n
B
(T

) R.
(ii) The class [d]
B
= 0 lies in H
1,1
B
(T

).
(iii) H
p,p
B
(T

) 0.
(iv) H
2p+1
B
(T

) has even dimension for j < [n2].

(v) (Transverse Hodge Decomposition) H
r
B
(T

) =

p+q=r
H
p,q
B
(T

).
(vi) Complex conjugation induces an anti-linear isomorphism H
p,q
B
(T

)
H
q,p
B
(T

).
(vii) (Transverse Serre Duality) There is an isomorphism
H
p,q
(T

) H
np,nq
(T

) .
Proof. The proof is very similar to the standard case [EKA90]. As usual we
dene the operator 1 :
r
(T

)
r+2
(T

) by
(7.2.9) 1 = d .
and its adjoint :
r
(T

)
r2
(T

) by
(7.2.10) = 1 .
We also have the adjoint operators

:
p,q
(T

)
p,q1
(T

)
and

:
p,q
(T

)
p1,q
(T

)
of and

. respectively, dened by
(7.2.11)

= .

.
As usual, we get the following
Lemma 7.2.7: On a Sasakian manifold the following hold:
(i) [. ] = i

.
(ii) [.

] = i

.
(iii)

= 0.

+

= 0.
(iv)

.
If we dene
(7.2.12)
B

.
B

.
then
Lemma 7.2.8: On a Sasakian manifold one has

B
= 2
B

.
The theorem easily follows from Lemmas 7.2.7 and 7.2.8.
Remark 7.2.2: Item (vii) of Theorem 7.2.6 holds without the Kahlerian hypothesis.
Thus, it holds for normal almost contact structures such that
(3)
= 0.
218 7. K-CONTACT AND SASAKIAN STRUCTURES
Theorem 7.2.6 gives rise to fundamental invariants of the Sasakian structure.
Now we have the basic Hodge numbers
(7.2.13) /
p,q
B
(T

) = dim H
p,q
B
(T

)
which are related to the basic Betti numbers by
(7.2.14) /
B
r
(T

) =

p+q=r
/
p,q
(T

) .
and satisfy
(7.2.15) /
p,q
B
(T

) = /
q,p
B
(T

) . /
p,q
B
(T

) = /
np,nq
B
(T

) .
In addition to the basic Euler characteristic on a Sasakian manifold we dene the
basic holomorphic Euler characteristic by
(7.2.16)
hol
(T

) =
n

q=0
(1)
q
/
0,q
B
(T

) .
There is also a transverse Lefschetz decomposition [EKA90]. Consider the
operator 1 :
r
B
(T

)
r+2
B
(T

) of equation (7.2.9), and let 1

r
(T

) denote the
so-called basic primitive cohomology group, that is, the kernel of the map :
H
r
B
(T

)H
r2
B
(T

) dened by Equation 7.2.10. Notice that as a subset of H

r
B
(T

).
each element of 1
r
(T

) has a unique harmonic representative. Thus, the transverse

Hodge Theorem [EKAH86] allows us to dene the set {
r
(T

) of primitive basic
harmonic forms. Now the Sasakian version of the transverse Hard Lefschetz The-
orem [EKA90] is:
Theorem 7.2.9: On a compact Sasakian manifold (`. o) of dimension 2n+1 the
map 1
k
: H
nk
B
(T

)H
n+k
B
(T

) is an isomorphism for 1 / n. Furthermore,

H
r
B
(T

) =

k0
1
k
1
r2k
(T

) .
We have the following useful corollary of Theorem 7.2.9:
Corollary 7.2.10: On a compact Sasakian manifold (`. o) of dimension 2n + 1
we have an isomorphism
H
r
B
(T

) 1
r
(T

) 1H
r2
B
(T

)
for 0 : n. Moreover, since 1 is injective for this range of : we also have
/
B
r
= j
B
r
+/
B
r2
. where j
B
r
= dim 1
r
(T

).
If the dimension of ` is 4n+1 we can also dene the basic Hirzebruch signature

B
(T

) to be the signature of the bilinear form dened by the cup product on the
middle basic cohomology group H
2n
B
(T

). Using the transverse Lefschetz Theorem

7.2.9 [EKA90] one can obtain the usual formula:
(7.2.17)
B
(T

) =

p,q
(1)
p
/
p,q
B
(T

) =

pq(2)
(1)
p
/
p,q
B
(T

) .
As is usual in complex geometry we introduce the basic geometric genus j
g
(T

) =
/
0,n
B
(T

). the basic arithmetic genus j

a
(T

) = (1)
n
(
hol
(T

) 1). and the basic

irregularity c = c(o) = /
0,1
B
. We remark that c =
1
2
/
B
1
(T

) =
1
2
/
1
(`) is actually a
topological invariant by (v) of Proposition 7.2.3. In the case n = 2, things simplify
7.3. CURVATURE PROPERTIES OF K-CONTACT AND SASAKIAN STRUCTURES 219
much more. All the basic Hodge numbers are given in terms of the three invariants
c. j
g
(T

) and /
1,1
B
(T

). and we have the relations

B
(T

) = 2 + 2j
g
(T

) 4c +/
1,1
B
(T

).
B
(T

) = 2 + 2j
g
(T

) /
1,1
B
(T

).

B
(T

) +
B
(T

) = 4
hol
(T

).
The fact that c is actually a topological invariant has some nice consequences for
n = 2.
7.3. Curvature Properties of K-Contact and Sasakian Structures
For any geometry it is important to study the curvature properties of a preferred
connection. In our case this connection is almost always the Levi-Civita connection,
but other important connections can and do occur. Nevertheless, our development
here concerns the curvature of the Levi-Civita connection.
7.3.1. Some Elementary Properties of Contact Metric Structures.
Here we develop some useful formulas for K-contact and Sasakian structures. We
begin by proving a general expression for the covariant derivative of for an almost
contact metric structure.
Lemma 7.3.1: Let (. . . o) be an almost contact metric structure on ` and
dene the 2-form by (A. Y ) = o(A. Y ). Then the following formula holds:
2o((
X
)Y. 7) = 3d(A. Y. 7) 3d(A. Y. 7) +o(
(1)
(Y. 7). A)
+
(2)
(Y. 7)(A) + 2d(Y. A)(7) 2d(7. A)(Y )
for any vector elds A. Y. 7 on `.
Proof. The formula follows by a straightforward albeit tedious computation
from the well-known formulas for the covariant and exterior derivative, namely
2o(
X
Y. 7) = Ao(Y. 7) +Y o(A. 7) 7o(A. Y )
+o([A. Y ]. 7) +o([7. A]. Y ) o([Y. 7]. A) .
and
3d(A. Y. 7) = A(Y. 7) +Y (7. A) +7(A. Y )
([A. Y ]. 7) + ([7. A]. Y ) ([Y. 7]. A) .
The details are left as an exercise.
Lemma 7.3.1 easily simplies in the case of a contact metric structure with its
associated metric.
Lemma 7.3.2: Let (. . . o) be a contact metric structure on `. Then we have
2o((
X
)Y. 7) = o(
(1)
(Y. 7). A) + 2d(Y. A)(7) 2d(7. A)(Y ) .
In particular,

= 0.
Lemma 7.3.3: For a contact metric structure we have
(i)
(3)
is a symmetric endomorphism in the sense that
o(
(3)
(A). Y ) = o(
(3)
(Y ). A).
(ii) =
1
2

(3)
.
(iii)
(3)
+
(3)
= 0.
(iv)
(3)
() = 0.
(3)
= 0.
220 7. K-CONTACT AND SASAKIAN STRUCTURES
(v) tr
(3)
= 0.
Proof. To prove (i) we rst compute

(3)
(A) = (

)(A) = [. A] [. A] =

(A)
X
(

A
X
)
= (

)(A) +

A
X

A +
X
=
X

X
.
where we have used the last statement of Lemma 7.3.2. Thus,
(3)
() =

which vanishes by Lemma 2.6.8. Next we claim that o(

(3)
(A). ) = (
(3)
(A)) =
0. From above we have (
(3)
(A)) = (
X
) which vanishes since is a metric
connection. This proves (ii) when A or Y is . Now assume that A and Y are
orthogonal to . Continuing our computation above
o(
(3)
(A). Y ) = o(
X
. Y ) o(
X
. Y ) = o(
X
. Y ) +o(.
X
Y )
= (
X
Y ) +(
X
Y ) .
and hence,
o(
(3)
(Y ). A) = (
Y
A) +(
Y
A) .
But it is not dicult to see that the equation
(2)
(A. Y ) = 0 implies that these
two expressions are equal. This proves (i).
To prove (ii) we consider the equation of Lemma 7.3.2 with Y = . and A and
7 orthogonal to . We have
2o((
X
). 7) = o(
(1)
(. 7). A) 2d(7. A)
= o(
2
[. 7] [. 7]. A) 2o(7. A)
= o((

)7. A) 2o(7. A)
= o((

)7. A) +((

)7)(A) 2o(7. A) .
Therefore,
2o(
X
. 7) = o(
(3)
(7). A) 2o(A. 7) = o(
(3)
(A). 7) 2o(A. 7) .
Since this holds for arbitrary 7 we have 2
X
=
(3)
(A) + 2A. Applying to
this equation proves (ii).
Both (iii) and (iv) are a result of Lemma 6.3.7. Finally, (v) holds since by (iii)
if is an eigenvalue of the symmetric operator
(3)
so is .
Combining Lemma 7.3.3 and Proposition 6.5.12 we have
Proposition 7.3.4: A contact metric structure (. . . o) is K-contact if and only
if = .
For a K-contact structure (. . . o) the characteristic foliation T

is Riemann-
ian, so we have the ONeill tensors T. at our disposal. First, one easily sees that
T = 0. and
(7.3.1)
X
Y =
1
2
([A. Y ]) =
1
2
([A. Y ]) = d(A. Y ) .
Indeed,
Lemma 7.3.5: For a K-contact manifold the following identities hold:
(i) T = 0.
(ii)
X
Y = d(A. Y ).
(iii)
X
= A.
(iv) [[
2
= 2n.
7.3. CURVATURE PROPERTIES OF K-CONTACT AND SASAKIAN STRUCTURES 221
Proof. It only remains to prove (iii) and (iv). We have that
o(
X
. Y ) = o(.
X
Y ) = (
X
Y ) = d(A. Y ) = o(A. Y )
holds for all horizontal Y. This proves (iii) since
X
is horizontal. To prove (iv)
we let 1
i

2n
i=1
denote a local orthonormal frame for T. and using (iii) we compute
[[
2
=
2n

i=1
o(
E
i
.
E
i
) =
2n

i=1
o(1
i
. 1
i
) =
2n

i=1
o(1
i
. 1
i
) = 2n.
We can write in the succinct form
(7.3.2)
X
= A A d .
Combining Lemmas 2.5.15 and 7.3.5 we have
Lemma 7.3.6: For a K-contact manifold the following identities hold:
(i) (

= 0 .
(ii) (
X
)

=
X
.
(iii) ((

)
X
) = 0 .
We can obtain more information about the behavior of the covariant derivatives
of from Theorem 2.5.16, but rst we need to investigate curvature properties of
K-contact manifolds.
7.3.2. Curvature and K-Contact Structures. Here we collect some useful
results concerning the Riemannian curvature 1(A. Y ) of K-contact manifolds. We
begin with the fundamental result of [HOT63]:
Theorem 7.3.7: A Riemannian manifold (`. o) is K-contact if and only if (`. o)
admits a Killing vector eld of unit length such that the sectional curvature of all
plane sections containing equals one.
First we obtain a general expression for 1(A. )Y for a metric contact structure.
Lemma 7.3.8: For a contact metric structure (. . . o) we have for any A
1(A. ) = A (A)
1
4
(
(3)
)
2
(A) +
1
2
(

(3)
)(A).
Moreover, if (. . . o) is K-contact manifold then for any vector eld A and for
Y is a section of T the following identities hold:
(i) 1(A. ) = A (A) .
(ii) 1(A. )Y = (
X
)(Y ) .
Proof. We compute
1(A. )Y =
X

X
Y
[X,]
Y
=
X
(
Y
+ [. Y ])

X
Y
[.
X
Y ]
[X,]
Y
=
X

X
Y
(

)
X
Y . (7.3.3)
222 7. K-CONTACT AND SASAKIAN STRUCTURES
Putting Y = in this equation and using Lemmas 7.3.2 and 7.3.3 we have
1(A. ) =

[.
X
]
[,X]

=
X
1
2
N
(3)
(X)
[. A
1
2

(3)
(A)] [. A]

1
2

(3)
([. A]))
= (A +
1
2

(3)
(A))
1
2

(3)
(A +
1
2

(3)
(A)) +
(3)
(A)
+
1
2
(
(3)
)
2
(A) +
1
2
(

(3)
)(A)
= A (A) +
1
4
(
(3)
)
2
(A) +
(3)
(A) +
1
2
(

(3)
)(A) .
Now it is easy to verify the formula
(7.3.4)

(3)
=

(3)
2
(3)
(
(3)
)
2
.
By applying to this formula and substituting it into the expression for 1(A. )
above proves the rst statement. (i) follows immediately from the rst statement
by Proposition 6.5.12. To prove (ii) we have from above
1(A. )Y =
X

X
Y
(

)
X
Y
which for a K-contact structure becomes
1(A. )Y =
X
Y +
X
Y = (
X
)(Y )
and proves (ii).
Exercise 7.4: Prove formula 7.3.4 above.
Proof of Theorem 7.3.7. Suppose rst that (`. . . . o) is K-contact. Then
is a unit length Killing vector eld, and using Lemma 7.3.8 we have for A or-
thogonal to .
1(A. ) =
o(1(A. ). A)
o(A. A)
=
o(A. A)
o(A. A)
= 1.
Conversely, suppose that is a unit length Killing vector eld. We dene
and by
(A) = o(. A). A =
X
.
Then as in the proof of Lemma 2.6.8 we have =

= 0. Now since is a
Killing vector eld it leaves invariant, so we have from above
1(A. ) =

=
X
=
2
A .
This gives for any vector eld A orthogonal to .
1 = 1(A. ) =
o(1(A. ). A)
o(A. A)
=
o(
2
A. A)
o(A. A)
.
Since this must hold for all vector elds A orthogonal to . we must have
2
A = A
for all vector elds A orthogonal to . Thus,
2
A = A +(A) for arbitrary A.
Moreover, since is a Killing and is metric and torsion-free we have
2d(A. Y ) = A(Y ) Y (A) ([A. Y ]) = Ao(. Y ) Y o(. A) o(. [A. Y ])
= o(
X
. Y ) +o(.
X
Y ) o(
Y
. A) o(.
Y
A) o(. [A. Y ])
= o(
X
. Y ) o(
Y
. A) = 2o(
X
. Y ) = 2o(A. Y ) = 2o(A. Y ) .
7.3. CURVATURE PROPERTIES OF K-CONTACT AND SASAKIAN STRUCTURES 223
This proves that the structure (. . . o) is K-contact.
Next we consider the Ricci curvature
(7.3.5) Ric(A. Y ) =

i
o

1(A. 1
i
)1
i
. Y

i
o

1(1
i
. A)Y. 1
i

.
where 1
i
is a local orthonormal frame. We have a result of Blair [Bla77]:
Proposition 7.3.9: A contact metric structure (. . . o) on a manifold of dimen-
sion 2n + 1 is K-contact if and only if Ric(. ) = 2n.
Proof. Let 1
0
= . 1
1
. . . . . 1
2n
be a local orthonormal frame, then we have
from Lemma 7.3.8
Ric(. ) =
2n

i=1
o

1(1
i
. ). 1
i

=
2n

i=1
o(1
i
. 1
i
)
2n

i=1

1
4
(
(3)
)
2
(1
i
). 1
i

1
2
o

(3)
)(1
i
). 1
i

= 2n
1
4
2n

i=1

(3)
(1
i
).
(3)
(1
i
)

2o

(3)
)(1
i
). 1
i

.
Now since on a contact metric manifold A and A are orthogonal, we can take the
local orthonormal frame to be of the form . 1
1
. . . . . 1
n
. 1
1
. . . . . 1
n
. Then we
have
Ric(. ) = 2n
1
4
2n

i=1
o

(3)
(1
i
).
(3)
(1
i
)

1
2
n

i=1
o

(3)
)(1
i
). 1
i

1
2
n

i=1
o

(3)
)(1
i
). 1
i

.
We claim that the last two terms cancel so that we have
(7.3.6) Ric(. ) = 2n
1
4
2n

i=1
(o(
(3)
(1
i
).
(3)
(1
i
)) = 2n
1
4
tr (
(3)
)
2
.
Since each term under the sum is positive, this proves the result.
Exercise 7.5: Prove the following formula for contact metric structures
1(A. ) = 1(A. ) + 2
2
A +
1
2
(
(3)
)
2
(A) .
and use it to prove the claim above verifying equation (7.3.6).
Next we relate the curvature for K-contact manifolds to the ONeill tensor eld
. First we notice that the rst two equations and the fourth equation of Theorem
2.5.16 are vacuous in the K-contact case. In what follows A. Y. 7. 7
t
are always
sections of T.
Lemma 7.3.10: On a K-contact manifold (`. . . . o) the following identities
hold:
(i)

)
X
Y

= 0.
(ii)

(
Z
)
X
Y

= o

(
Z
)(A). Y

.
224 7. K-CONTACT AND SASAKIAN STRUCTURES
(iii) o

1(A. Y )7. 7
t

= o

1
T
(A. Y )7. 7
t

+2d(A. Y )d(7. 7
t
)
+d(A. 7)d(Y. 7
t
) d(A. 7
t
)d(Y. 7) .
Proof. (i): The third equation of Theorem 2.5.16 gives the identity
o

1(A. )Y.

= o

X
.
Y

)
X
Y.

= o

A. A

)
X
Y.

= o(A. Y ) o

)
X
Y.

.
Using the well-known Riemann curvature identities and Lemma 7.3.8 the left hand
side of this equation reduces as
o

1(A. )Y.

= o

1(A. ). Y

= o(A. Y ) .
This proves (i). To prove (ii) we notice that the fth equation of Theorem 2.5.16
becomes
o

1(A. Y )7.

= o

(
Z
)
X
Y.

(
Z
)
X
Y

.
Again by the Riemann curvature identities and Lemma 7.3.8, the left hand side is
o

1(A. Y )7.

= o

1(7. )A. Y

= o

(
Z
)(A). Y

which proves (ii). Then (iii) follows immediately by (vi) of Theorem 2.5.16 and (ii)
of Lemma 7.3.5.
From Theorem 2.5.18 one obtains relations between the Ricci curvature of o
and the transverse Ricci curvature for K-contact manifolds. But rst we need a
lemma.
Lemma 7.3.11: Let (. . . o) be a contact metric structure on `. and consider
the local orthonormal frame . 1
1
. . . . . 1
n
. 1
1
. . . . . 1
n
. Then the following re-
lation holds:
2n

i=1
o

(
E
i
)(1
i
). A

= 2n(A) .
where 1
n+i
= 1
i
.
Proof. From Lemma 7.3.2 we get
(7.3.7) 2o

(
Ei
)1
i
. A

= o

(1
i
. A). 1
i

+ 2o(1
i
. 1
i
)(A) .
Now it is easy to see that

(A. 1
i+n
) =

(A. 1
i
) ([1
i
. A]) .
Using this equation and summing on i in Equation 7.3.7 gives the result.
We now have
Theorem 7.3.12: Let (`. . . . o) be a K-contact manifold and let Ric
T
denote
the Ricci curvature of the transverse metric o
T
. Then the following identities hold:
(i) Ric(A. ) = 2n(A) for any vector eld A,
(ii) Ric(A. Y ) = Ric
T
(A. Y ) 2o(A. Y ). where A. Y are sections of T.
Proof. To prove (i) we see that the second equation of Theorem 2.5.18 gives
Ric(A. ) = o

(
T
)A.

.
So choosing a local orthonormal basis 1
i

i
of T as above, we compute using (ii)
of Lemma 7.3.10
o(
T
)A. ) =

i
o

(
Ei
)
Ei
)(A).

i
o

(
Ei
)(1
i
). A

.
7.3. CURVATURE PROPERTIES OF K-CONTACT AND SASAKIAN STRUCTURES 225
So applying Lemma 7.3.11 implies (i). For (ii) we only need to compute o(
X
.
Y
)
using Equation 2.5.23
o(
X
.
Y
) = o(
X
.
Y
) = o(A. Y ) = o(A. Y ) .
For sectional curvatures of a K-contact structure we have
Lemma 7.3.13: Let A. Y be smooth sections of T of unit length. Then the sec-
tional curvatures satisfy
(i) 1(A. ) = 1.
(ii) 1(A. Y ) = 1
T
(A. Y ) 3(d(A. Y ))
2
.
Proof. This follows immediately from Theorem 7.3.7, Corollary 2.5.17, and
Lemma 7.3.5.
We are also interested in the so-called -sectional curvatures which play a
role in Sasakian geometry analogous to the holomorphic sectional curvatures in
Kahlerian geometry.
Denition 7.3.14: The -sectional curvature H(A) of a K-contact manifold
is dened by H(A) = 1(A. A). where 1 is the ordinary sectional curvature.
Then an immediate consequence of Lemma 7.3.13 is:
Lemma 7.3.15: Let A be a section of T of unit length. Then
H(A) = 1
T
(A. A) 3 .
In the next section we shall see that in the case of Sasakian manifolds, the
-sectional curvature determines the full curvature tensor 1.
7.3.3. Curvature Properties of Sasakian Structures. Now we give nec-
essary and sucient condition on the covariant derivative of for an almost contact
structure to be Sasakian.
Theorem 7.3.16: An almost contact metric manifold (`. . . . o) is Sasakian if
and only if

X
= A o A .
Proof. Let (`. o) be embedded in its cone (C(`). d:
2
+:
2
o) as the hypersur-
face : = 1. Then the almost contact metric structure (. . . o) is Sasakian if and
only if the cone metric o = d:
2
+ :
2
o is Kahler with respect to complex structure
given by equation (6.5.2). Furthermore, the complex structure 1 is Kahler with
respect to the metric o if and only if it is parallel with respect to the Levi-Civita
connection

of o. Now for any vector elds A. Y on ` we have, using Gauss
formula

X
Y =
X
Y +o(:A. Y ) and Weingartens equation

X
= :A.
0 = (

X
1)(Y ) =

X
1Y 1

X
Y =

X
(Y +(Y )) ( + )

X
Y
=
X
Y +o(:A. Y ) +A(Y )(Y ):A

X
Y o(:A. Y )1(
X
Y )
= (
X
)(Y ) o(A. Y ) +(Y )A +

A(Y ) o(A. Y ) (
X
Y )

.
Noting that the normal terms give an identity for any almost contact metric struc-
ture, one sees that the tangential term gives the result. Notice that we have used
the easily veried fact that the shape operator : equals 1l for a hypersurface `
embedded metrically in the cone C(`) as : = 1.
226 7. K-CONTACT AND SASAKIAN STRUCTURES
Exercise 7.6: Show that the normal terms in the above equation vanish identically
on any almost contact metric manifold.
Next we have a characterization of Sasakian structures in terms of curvature.
Proposition 7.3.17: Let (. . . o) be a contact metric structure. Then the fol-
lowing are equivalent:
(i) (. . . o) is Sasakian.
(ii) 1(A. Y ) = (Y )A (A)Y.
(iii) 1(A. )Y = (Y )A o(A. Y ).
Proof. (i) (ii) and (iii): We have
1(A. Y ) =
X

Y

Y

X

[X,Y ]

=
X

Y +
1
2

(3)
(Y )

+
Y

A +
1
2

(3)
(A)

+ [A. Y ] +
1
2

(3)
([A. Y ])
= (
X
)(Y ) + (
Y
)(A)
1
2

(
X

(3)
)((Y ) (
Y

(3)
)(A)

.
If (. . . o) is Sasakian then
(3)
= 0. and Theorem 7.3.16 implies (ii). That (i)
implies (iii) is an immediate consequence of Lemma 7.3.8 and Theorem 7.3.16.
Now assume that either (ii) or (iii) holds. Then we have 1(A. ) = A for any
A orthogonal to . So using Exercise 7.5 we have
1
2
(
(3)
)
2
(A) = 1(A. ) 1(A. ) + 2A =
2
A +A = 0.
But by Lemma 7.3.3
(3)
is symmetric and annihilates , so this implies that
(3)
=
0. Thus, (. . . o) is K-contact. But (ii) of Lemma 7.3.8 says that 1(A. )Y =
(
X
)(Y ), so (iii) implies (i) by Theorem 7.3.16. To see that (ii) also implies (i)
we use the symmetries of the Riemann curvature tensor to give
o

(
X
)(Y ). 7

= o

1(A. )Y. 7

= o

1(Y. 7). A

= o

(7)Y (Y )7. A

= o

o(A. Y ) (Y )A. 7

.
This holds for arbitrary 7 so (ii) implies (i) again by Theorem 7.3.16.
The proof of Theorem 7.3.16 suggests that we should look for conditions that
a real hypersurface in a Kahler manifold be Sasakian. We have
Theorem 7.3.18: A real hypersurface ` in a Kahler manifold (. 1. o) is Sasakian
if the shape operator : satises : = 1l + for some smooth function . with
= 1\ , \ a unit normal vector eld of ` and the dual 1-form of .
Conversely, if (`. o) is a Sasakian manifold embedded as a real hypersurface in
then the shape operator : satises the above condition.
Proof. Let ` be a (2n + 1)-dimensional hypersurface in a Kahler manifold
(. 1. o) of complex dimensional n+1 with complex structure tensor 1. Let o denote
the restriction of o to ` and

the Levi-Civita connection on . Let be the
Levi-Civita connection on `. Then the Gauss and Weingarten equations are
(7.3.8)

X
Y =
X
Y +o(:A. Y )\.

X
\ = :A .
7.3. CURVATURE PROPERTIES OF K-CONTACT AND SASAKIAN STRUCTURES 227
where A. Y are vector elds on `. and \ is a unit normal eld to `. Dene
= 1\ and by 1A = A + (A)\ where is the dual 1-form to . It is easy
to check that (. . . o) denes an almost contact metric structure on `.
Now since 1 is parallel with respect to

. we have

X
1Y = 1

X
Y. The left
hand side gives using equations (7.3.8)

X
1Y =

X
(Y +(Y )\ ) =
X
Y +o(:A. Y )\ +A(Y )\ (Y ):A .
while the right hand sign equals
1

X
Y = 1(
X
Y +o(:A. Y )\ ) =
X
Y +(
X
Y )\ o(:A. Y ) .
Equating the components tangent to ` gives
(7.3.9) (
X
)(Y ) = (Y ):A o(:A. Y ) .
By Theorem 7.3.16 ` is Sasakian if and only if
(
X
)(Y ) = o(A. Y ) (Y )A .
Equating this with Equation 7.3.9 and using the symmetry of the second funda-
mental form gives the solution
:A = A +(A)
for some smooth function . which proves the result.
Theorem 7.3.18 says that the embedding of a Sasakian hypersurface into a
Kahler manifold is umbilic on the contact subbundle. The principal curvatures are
all equal there. Another approach to the problem of embedded hypersurfaces in
Kahler manifolds is given by
Exercise 7.7: Let ` be a real hypersurface embedded in a Kahler manifold .
Show that

(A. Y ) + 2d(A. Y ) = (A)[. :]Y (Y )[. :]A .

Thus, (`. . . . o) is Sasakian if [. :] = 0.
Next we give a brief discussion of -sectional curvatures which were dened in
Denition 7.3.14. Only the main results are given here. The proofs as well as more
detail are given in several books [Bla76a, Bla02, YK84]. All of these results have
strong analogs in Kahlerian geometry.
Theorem 7.3.19: On a Sasakian manifold ` the -sectional curvature H deter-
mines the curvature. Moreover, assuming dim ` 5. if the -sectional curvature
H at any point is independent of the choice of -section at the point, then it is
constant, say c, on ` and the Riemannian curvature is given by
1(A. Y )7 =
c + 3
4

o(Y. 7)A o(A. 7)Y

+
c 1
4

(A)(7)Y (Y )(7)A +o(A. 7)(Y ) o(Y. 7)(A)

+d(7. Y )A d(7. A)Y + 2d(A. Y )7

.
Notice that the case c = 1 greatly simplies giving the formula for a Riemannian
metric of constant sectional curvature equal to 1. Thus, c = 1 gives the standard
round sphere.
Unlike Kahler geometry Sasakian geometry is not scale invariant. That is, if
o is a Sasakian metric and is a non-zero constant, o is not Sasakian. There is,
228 7. K-CONTACT AND SASAKIAN STRUCTURES
however, a transverse homothety deformation or T-homothety in [YK84] which we
now describe. The earliest reference seems to be the thesis of Moskal [Mos66]. It
was used subsequently by Sasaki in his lecture notes [Sas68] and by Tanno [Tan68].
Consider the deformation of a contact metric structure o = (. . . o) dened by
(7.3.10)
t
= o
1
.
t
= o.
t
= . o
t
= oo + (o
2
o)
for o R
+
. We call this deformation a transverse homothety deformation. We shall
see later that this also occurs as part of what we call o-homology classes described
in Denition 7.5.13 below. It is easy to see that
Proposition 7.3.20: If o = (. . . o) is a contact metric, K-contact, or Sasakian
structure, then so is o
t
= (
t
.
t
.
t
. o
t
) determined by equations (7.3.10). Moreover,
o
t
belongs to the same underlying CR structure as o.
Exercise 7.8: First prove Proposition 7.3.20. Then show that if o = (. . . o) is
the standard round sphere o
2n+1
with sectional curvature 1. the Sasakian structure
o
t
= (
t
.
t
.
t
. o
t
) obtained by a transverse homothety deformation has constant
-sectional curvature c =
4
a
3.
We denote the sphere o
2n+1
with constant -sectional curvature equal to c by
o
2n+1
(c). Note that in this case the range of c is 3 < c < .
Exercise 7.9: Consider the transverse homothety of a Sasakian structure dened
in equation (7.3.10). Show that if
t
and denote the Levi-Civita connections of
o
t
and o, respectively, then the dierence tensor =
t
is given by
= (o 1)( + ).
Show that the Riemann curvature tensor 1
t
for o
t
is related to that for o by
1
t
(A. Y )7 = 1(A. Y )7 + (
X
)(Y. 7) (
Y
)(A. 7)
+

A. (Y. 7)

Y. (A. 7)

.
and determine an explicit expression in terms of the Sasakian structure (. . . o).
Show directly that the Ricci tensor of o
t
is given by
(7.3.11) Ric
g
= Ric
g
2(o 1)o + 2(o 1)(on +n + 1) .
Exercise 7.10: Prove relation (7.3.11) using Theorem 7.3.12 as well as the trans-
formation formula for the scalar curvature,
(7.3.12) :
g
= o
1
(:
g
+ 2n) 2n.
Next we describe another natural Sasakian structure.
Example 7.3.21: Recall the standard contact structure on R
2n+1
dened by equa-
tion (6.1.1), namely = d.

i
n
i
dr
i
. The Reeb vector eld is given by =

z
.
Next we dene
=

r
i
+n
i

dn
i

n
i
dr
i

.
Then the associated metric 6.4.2 becomes
(7.3.13) o =

(dr
i
)
2
+ (dn
i
)
2

+ .
Notice from (7.3.13) that the transverse metric o
T
is at, so it follows from Lemma
7.3.15 that the -sectional curvature H(A) = 3 for all sections A of T of unit
7.4. TOPOLOGY OF K-CONTACT AND SASAKIAN MANIFOLDS 229
length. We denote this Riemannian structure by R
2n+1
(3). This example is related
to the (2n+1)-dimensional Heisenberg group as we discuss in Example 8.1.12 below.
Exercise 7.11: Prove that o = (. . . o) is a Sasakian structure on R
2n+1
. What
happens to (R
2n+1
. o) under a transverse homothety?
Example 7.3.22: Recall the complex ball 1
n
C
(1) of Example 3.2.9 with the Bergman
metric o
0
of constant sectional curvature 1. We can scale this metric to have con-
stant holomorphic sectional curvature / < 0. Let o
T
denote this metric, and
T
the corresponding Kahler form. Since 1
n
C
(1) is simply connected there is a 1-form
such that
T
= d. We construct a Sasakian structure on 1
n
C
(1) R by setting
= dt+

and =

t
. where : 1
n
C
(1)R1
n
C
(1) is the natural projection and
t is the coordinate on R. We dene to be the horizontal lift of the almost complex
structure tensor J and zero in the vertical direction. The metric is o = o
T
+
as usual. Then o = (. . . o) denes a Sasakian structure on 1
n
C
(1) R which by
Lemma 7.3.15 has constant -sectional curvature H = c = / 3. We denote this
Riemannian structure by 1
n
C
(/) R.
Exercise 7.12: Show that 1
n
C
(/) R maps to 1
n
C
(1) R under a transverse ho-
mothety.
The examples discussed above are known as Sasakian space forms and the
following theorem of Tanno [Tan69a] gives the classication of simply connected
Sasakian space forms. So there are precisely three Sasakian space forms up to
transverse homothety.
Theorem 7.3.23: Let `(c)be a complete simply connected Sasakian manifold of
constant -sectional curvature c. Then `(c) is one of the following:
(i) If c 3. `(c) is isomorphic to o
2n+1
(c).
(ii) If c = 3, `(c) is isomorphic to R
2n+1
(3).
(iii) If c < 3, `(c) is isomorphic to 1
n
C
(/) R. where c = / 3.
For Sasakian -symmetric spaces the reader is referred to [Bla02] and refer-
ences therein.
7.4. Topology of K-Contact and Sasakian Manifolds
We had seen previously in Section 6.2 that a necessary condition for a manifold
to admit a strict almost contact structure is the vanishing of certain characteristic
classes, namely the odd Stiefel-Whitney classes. Here we discuss some additional
topological obstructions for a manifold to admit a K-contact or Sasakian struc-
ture. The rst such obstruction was noticed by Rukimbira [Ruk93] in 1993, and
somewhat later by Itoh [Ito97]. Rukimbiras result applies to what he called R-
contact structures, that is, contact metric structures whose Reeb vector eld denes
a Riemannian ow. This condition is ostensively weaker than the K-contact con-
dition. (Clearly, any K-contact structure is R-contact.) However, later Rukimbira
[Ruk95a] proved that every R-contact structure admits a compatible K-contact
metric. Hence, we restrict ourselves to K-contact structures.
Recall that the cuplength of a topological space `. denoted by cup(`) is the
largest integer / such that there exists classes
1
. . . . .
k


H

(`. R) satisfying

1

k
= 0. Clearly if ` is an orbifold or manifold, we have 0 cup(`)
dim(`), and if in addition ` is compact cup(`) 0. For example, the (2n + 1)-
torus T
2n+1
has maximal cuplength, namely cup(T
2n+1
) = 2n + 1. Now we have
230 7. K-CONTACT AND SASAKIAN STRUCTURES
Theorem 7.4.1: Let ` be a compact 1-contact manifold of dimension 2n + 1.
Then,
1 cup(`) 2n.
We have an immediate corollary of this theorem:
Corollary 7.4.2: The manifolds
g1

g
l
o
1
do not admit a K-contact
structure, where
g
denotes a Riemann surface of genus o 1. In particular, a
torus cannot admit a K-contact structure.
On the other hand odd dimensional tori do admit contact structures [Bou02],
as evidenced by Example 6.1.23 and Exercise 6.5. The proof of Theorem 7.4.1 is
based on the following well-known result of Tachibana [Tac65].
Lemma 7.4.3: Let (`. . . o) be a compact K-contact manifold. Then every har-
monic 1-form satises () = 0.
Proof. Any closed 1-form can be decomposed into = + 1 where
satises () = 0. By Hodge theory harmonic forms are invariant under isometries,
and is an innitesimal isometry. Thus,
0 =

= d(()) = d1.
so 1 is constant. This implies that d +1d = 0. So by Stokes theorem
0 =

M
d( (d)
n1
=

M
1 (d)
n
.
which, since (d)
n
is a volume form, implies that 1 = 0. So = proving the
result.
Proof. (of Theorem 7.4.1). By Hodge theory every cohomology class can be
represented uniquely by a harmonic form, and the wedge product on forms induces
the cup product on de Rham cohomology classes. So suppose that cup(`) = 2n+1,
then the Riemannian volume dvol form can be written uniquely up to order as
dvol =
1

2n+1
.
where
i
are harmonic 1-forms. Since ` admits a K-contact structure, the har-
monic forms
i
can be given with respect to the K-contact metric o. Moreover, the
characteristic vector eld is a Killing vector eld, and we can extend to a local or-
thonormal basis . c
1
. . . . . c
2n
. But then Lemma 7.4.3 gives dvol(. c
1
. . . . . c
2n
) =
0. which implies the vanishing of the cohomology class [dvol] giving a contradic-
tion.
We mention that Example 3.5.16 below gives Sasakian, hence, K-contact man-
ifolds, with cup(`) = 2n. Rukimbira [Ruk94] noticed that one can do better than
the upper bound given by Lemma 7.1.9, namely
Theorem 7.4.4: Let (`. . . . o) be a compact K-contact manifold of dimension
2n + 1 with rst Betti number /
1
= /
1
(`). Then
1 rk(`. T

) min(n + 1. 2n + 1 /
1
) .
Proof. By Hodge theory there are precisely /
1
independent harmonic 1-forms

1
. . . . .
b
1
. Let 7
1
. . . . . 7
b
1
denote their dual harmonic vector elds, and let / =
rk(`. T

). Then by Proposition 2.6.3 the leaf closures of the characteristic foliation

T

are tori T whose maximal dimension is /. So on a dense open set the leaf
closures are tori T
k
of dimension precisely /. Furthermore, T
k
lies in the center of
7.4. TOPOLOGY OF K-CONTACT AND SASAKIAN MANIFOLDS 231
the automorphism group Aut(. . . o). So if \ t
k
. the Lie algebra of T
k
. it is a
Killing vector eld. Thus, by Tachibanas Lemma 7.4.3 we have
0 =
i
(\ ) = o(7
i
. \ )
for all i = 1. . . . . /
1
and any \ t
k
. It follows that / +/
1
2n + 1 which together
with Lemma 7.1.9 proves the result.
Together with Theorem 7.1.3, Rukimbiras Theorem 7.4.4 has an interesting
corollary.
Corollary 7.4.5: Let (`. . . . o) be a compact K-contact metric manifold of
dimension 2n +1 whose rst Betti number /
1
(`) = 2n. Then ` is the total space
of a principal o
1
V-bundle over a compact almost Kahler orbifold Z whose rst
Betti number /
1
(Z) = 2n.
Another consequence of Tachibanas Lemma 7.4.3 is
Proposition 7.4.6: There are no parallel vector elds on a compact K-contact
manifold.
Proof. Parallel vector elds are harmonic so Lemma 7.4.3 implies that a par-
allel vector eld is orthogonal to . Thus, for parallel and for A an arbitrary
vector eld we have
0 =
X
o(. ) = o(
X
. ) +o(.
X
) = o(A. )
by Proposition 7.3.4. But being parallel is a section of T. so it must vanish by
the non-degeneracy of on T.
A quasi-regular K-contact manifold clearly has an innite number of closed
characteristics, that is, an innite number of closed integral curves of the charac-
teristic vector eld . The famous Weinstein conjecture[Wei79] states that for any
compact hypersurface of contact type with vanishing rst Betti number, the Reeb
vector eld has a periodic orbit, i.e., a closed integral curve. It is also widely
believed that the conjecture holds when the condition on the rst Betti number is
dropped. Much progress has been made recently on this conjecture, but it still re-
mains open in full generality. We refer the reader to [Bla02, MS98] and references
therein for further discussion and results. For K-contact manifolds the conjecture
was proven by Banyaga [Ban90], and more recently Rukimbira [Ruk95b, Ruk99]
proved the following renement:
Theorem 7.4.7: Let (`. . . . o) be a compact K-contact manifold of dimension
2n + 1. Then has at least n + 1 closed integral curves. Furthermore, if ` is
simply connected and has precisely n+1 closed characteristic curves. Then ` is
homeomorphic to the sphere o
2n+1
.
Proof. Let (`. . . . o) be a compact irregular K-contact manifold of di-
mension 2n + 1. Then by Theorem 7.1.10 there exists a quasi-regular K-contact
structure (
1
.
1
.
1
. o
1
) on ` close to (. . . o). Furthermore, by Theorem 2.6.4,
the vector elds and
1
are equivalent to linear ows on a torus; hence, they com-
mute and
1
is an innitesimal automorphism of the K-contact structure (. . . o).
Consider the function (
1
). It is invariant under the circle action generated by
1
.
If j ` is a critical point of the function (
1
). then we have
0 = d((
1
))(j) = (
1
)(j) (
1
d)(j) = (
1
d)(j).
232 7. K-CONTACT AND SASAKIAN STRUCTURES
It follows that the vectors
1
(j) and (j) are proportional. Furthermore, since (
1
)
is invariant under the circle action generated by
1
. it follows that
1
(j) and (j)
are proportional at all points of the leaf L
1
of T
1
through j. So the leaves of T

and the leaves of T

1
coincide at the critical points of (
1
). Since the leaves of T
1
are all circles, the leaves of T

are circles at these critical points.

Now we can use Lusternik-Schnirelman theory to get a lower bound on the
number of critical curves. This was done by Weinstein [Wei77b] by showing that
the Lusternik-Schnirelman Theorem generalizes to compact orbifolds (Notice that
this proves Weinsteins conjecture for K-contact manifolds assuming that critical
points exist). In fact Weinstein shows that there are at least Cat(`T
1
) criti-
cal curves of any circle invariant function on `. where Cat(`T

1
) denotes the
Lusternik-Schnirelman category
2
of the orbifold `T
1
. But on any topological
space o we have the inequality Cat(o) cup(o) +1. Clearly, on a compact almost
Kahler orbifold of dimension 2n, we have cup n. Putting these estimates together
proves the rst part.
The proof of the second statement uses Morse-Bott theory. We only give an
outline of the proof here. First using curvature computations Rukimbira shows that
the Hessian is non-degenerate, so Botts generalization [Bot54] of Morse theory
holds. Moreover, the fact that there is a transverse almost complex structure implies
that the Morse indices are all even. Then since there are precisely n + 1 closed
characteristics, one for each index 0. 2. . . . . 2n, the Morse inequalities imply that all
the Betti numbers are either 0 or 1. Then using the gradient ow for the function
(
1
) one obtains a stratication of `
2n+1
as follows:
`
1
`
3
`
2n1
`
2n+1
.
where each `
2k+1
a is closed embedded submanifold with / + 1 non-degenerate
critical circles of the restriction of (
1
) to `
2k+1
. Rukimbira then shows using
the Mayer-Vietoris sequence that if H
1
(`
2k+1
. Z) = 0 then H
1
(`
2k1
. Z) = 0.
and if in addition `
2k1
is a homology sphere then so is `
2k+1
. By hypothesis
H
1
(`
2n+1
. Z) = 0. and using Carri`eres classication [Car84a] of Riemannian
ows on 3-manifolds one sees that `
3
is dieomorphic to o
3
. This then together
with Smales generalized Poincare Theorem [Sma61] proves the result.
Another related theorem of Rukimbira [Ruk95b] is:
Theorem 7.4.8: Let (`. . . . o) be a compact K-contact manifold, and suppose
that has only a nite number of closed integral curves. Then /
1
(`) = 0.
A corollary of this result is that any toric K-contact manifold ` must have
/
1
(`) = 0. There is a stronger result for toric K-contact manifolds due to Lerman
[Ler04b] which says that the fundamental group of any K-contact toric manifold
is nite and determined combinatorially by the image of the contact moment map.
We will treat toric contact manifolds in some detail later.
Our next example of an obstruction to K-contact structures involves Gromovs
invariants MinVol and [[`[[ discussed at the end of Chapter 2. Since a K-contact
structure is in particular a Riemannian ow, we have immediately from Theorem
2.6.13
2
Recall that for a topological space Cat(S) is the minimal number of open sets needed to
cover S. This denition is not standard though. For example, the minimal number of such open
sets minus one is frequently used [CLOT03].
7.4. TOPOLOGY OF K-CONTACT AND SASAKIAN MANIFOLDS 233
Corollary 7.4.9: Let ` admit a K-contact structure. Then both MinVol(`) and
Gromov invariant [[`[[ vanish.
From Theorem 2.6.14, the estimate 2.6.3, and the properties of the simplicial
volume discussed in Chapter 1, we have
Corollary 7.4.10: If ` decomposes as ` = `
1
# #`
l
where some `
i
is
compact and admits a metric whose sectional curvature is bounded from above by
for some 0. then ` does not admit a K-contact structure.
For obstructions to Sasakian manifolds we have the following theorem which ap-
pears to have been discovered independently by various geometers [BG67, Fuj66,
Tan67]. It is the analogue of a well-known theorem in Kahler geometry.
Theorem 7.4.11: Let `
2n+1
be a compact manifold admitting a Sasakian struc-
ture o. Then the j
th
Betti number /
p
(`) is even for j odd with 1 j n. and
for j even with n < j 2n.
Proof. For j = 1 this follows immediately from (v) of Proposition 7.2.3, (vi)
of Theorem 7.2.6 and equation (7.2.14). For j 1 it is much more involved. We
present the main technical result as a lemma and leave the proof as an exercise.
The rst part is a generalization of Lemma 7.4.3 and the lemma itself is again due
to Tachibana [Tac65]:
Lemma 7.4.12: Let (`
2n+1
. o) be a compact Sasakian manifold. Then for 1
j n any harmonic j-form n is orthogonal to . i.e., n = 0. Moreover, for the
same range of j, the j-form n dened by
n(A
1
. . . . . A
p
) =
p

i=1
n(A
1
. . . . . A
i
. . . . . A
p
)
is harmonic.
Assuming the lemma, we proceed with the proof of Theorem 7.4.11. Dene the
j-form 1n by
1n(A
1
. . . . . A
p
) = n(A
1
. . . . . A
p
).
and for each / = 0. . . . . j consider the iterates
k
n. By Lemma 7.4.12 each
k
n
is harmonic. Furthermore, it should be clear that 1n can be written as a linear
combination of the harmonic forms
k
n
p
i=0
. Hence, 1n is itself harmonic. So 1
is a map 1 : H
p
H
p
on harmonic j-forms. Moreover, it satises 1
2
= (1)
p
1l
on H
p
. Thus, for j n odd, the real harmonic forms n and 1n are independent,
so the result follows for this range of j. The other range of j follows by Poincare
duality.
Exercise 7.13: Prove Lemma 7.4.12.
More can be said about harmonic theory on compact Sasakian manifolds.
Proposition 7.4.13: Let (`
2n+1
. o) be a compact Sasakian manifold, and let j
be an integer satisfying 1 j n. Then a j-form n is harmonic if and only if it is
primitive and basic harmonic.
Proof. Let n be a harmonic j-form on `. Then since n is closed and horizontal
by Lemma 7.4.12, n is basic. By the denition of , equation (7.2.2), we see that
n = n. Thus,
d(n) = d n d(n) = 1n d
B
n.
234 7. K-CONTACT AND SASAKIAN STRUCTURES
The two terms on the right hand side of this equation are independent so the left
hand side vanishes if and only if both terms on the right vanish. The vanishing
of the rst term on the right says that n is primitive, while the vanishing of the
second says that n is basic harmonic.
Proposition 7.4.13 says that the basic primitive cohomology groups 1
r
(T

) of
a Sasakian manifold are actually topological invariants. This implies then that the
basic cohomology groups H
r
B
(T

) are also topological invariants. More precisely,

Theorem 7.4.14: Let ` and `
t
be homeomorphic compact Sasakian manifolds
with Sasakian structures o = (. . . o) and o
t
= (
t
.
t
.
t
. o
t
). respectively. Then
their basic cohomology groups are isomorphic, i.e., H
r
B
(T

) H
r
B
(T

). Further-
more, the Betti numbers /
r
and basic Betti numbers /
B
r
are related by /
r
= /
B
r
/
B
r2
for 0 : n. In particular, if : is odd, /
r
= /
B
r
.
Proof. By Proposition 7.4.13 the basic primitive cohomology groups 1
r
(T

)
are isomorphic to the de Rham cohomology of ` which is a topological invariant.
So 1
r
(T

) 1
r
(T

). But the transverse hard Lefschetz Theorem 7.2.9 computes

H
r
B
(T

) in terms of 1
r
(T

) by wedging with d. and computes H

r
B
(T

) in terms
of 1
r
(T

) by wedging with d
t
. and these are isomorphic operations. This proves
the rst statement. The second statement then follows from Corollary 7.2.10.
Although the basic cohomology groups H
r
B
(T

) cannot distinguish Sasakian

structures, the basic Hodge groups H
p,q
B
(T

) can as we show by explicit exam-

ple later. Notice that Theorem 7.4.14 and (iv) of Theorem 7.2.6 yield Theorem
7.4.11 as an immediate corollary. Both proofs of Theorem 7.4.11, however, rely on
Tachibanas Lemma 7.4.12. In the case of a quasi-regular K-contact or Sasakian
manifold `, Proposition 7.2.2 together with Theorem 7.4.14 relate the real coho-
mology of ` with the real cohomology of the space of leaves 7 = `T

. However,
they do not give integral information. For integral information one needs to analyze
the Leray spectral sequence of the corresponding Seifert o
1
-bundle. In this regard
see Proposition 4.7.9.
Next we discuss without proof some rigidity results regarding Sasakian man-
ifolds with positive sectional curvature. This work appears to have begun with
Goldberg [Gol65, Gol67] under the added assumption of regularity, but reached
its fruition in the work of Tanno [Tan68] and the unpublished thesis of Moskal
[Mos66] which is described in detail in Sasaki [Sas68].
Theorem 7.4.15: Let (`. o) be a compact Sasakian manifold with positive sec-
tional curvature. Then /
2
(`) = 0. If in addition the scalar curvature is constant,
then (`. o) is isometric to the sphere o
2n+1
(c) of constant -sectional curvature c
for some 3 < c < .
The rst statement is due to Tanno [Tan68] with a weaker version established a
few years earlier in [TO66], while the second statement is due to Moskal. Somewhat
stronger results are available in the low dimensions of three and ve, as well as under
the added assumption of being Einstein. These are treated in detail in Chapters 10
and 11.
The rst known example of a symplectic manifold `
4
that does not admit a
Kahler structure is the well-known primary Kodaira surface from complex surface
7.4. TOPOLOGY OF K-CONTACT AND SASAKIAN MANIFOLDS 235
theory [BPVdV84]. This was pointed out by Thurston
3
[Thu76] who constructed
it as a non-trivial T
2
-bundle over T
2
with rst homology group equal to Z
3
. This
gives rise to a K-contact manifold `
5
that admits no Sasakian structure. Later `
4
was shown to admit an indenite Kahler structure in [FGG88]. Further examples
of compact almost Kahler manifolds admitting no Kahler structure are given in
[CFdL85, CFG86]. Many of these examples, including Thurstons, involve the
Heisenberg group. The geometry of Thurstons example is discussed more fully in
[Abb84].
Example 7.4.16: A non-trivial T
3
-bundle over T
2
. Consider the real three-step
nilpotent group G
5
(R) which as a set is R
5
. Using vector notation the group action
is
(7.4.1) (c. a. b) (.. x. y) = (x +a. y +b. . +c +o
2
r
1
+/
2
n
1
+o
2
/
2
n
2
+o
2
n
2
2
) .
where
=

1 o
2
0 1

.
Let G
5
(Z) denote G
5
with integer values for its parameters, and let `
5
denote the
quotient G(R)G(Z). Then
1
(`
5
) G
5
(Z) which as a set is Z
5
. The commutator
subgroup is [G
5
(Z). G
5
(Z)] = Z
2
. So H
1
(`
5
. Z) Z
3
. Moreover, by construction
`
5
is a o
1
-bundle over Thurstons non-trivial T
2
-bundle over T
2
which is an almost
Kahler manifold. Thus, `
5
is K-contact but not Sasakian by Theorems 7.1.6 and
7.4.11. It does, however, admit a pseudo-Sasakian structure owing to the indenite
Kahler structure on the base. This is a type of normal almost contact structure.
As with Kahler manifolds it is natural to ask to what extend K-contact mani-
folds are more general than Sasakian manifolds. In the symplectic case, the rst sim-
ply connected examples of non-Kahler symplectic manifolds was given by McDu
[McD84]. Later Gompf [Gom95] gave a systematic construction which produced
innitely many examples of simply connected symplectic non-Kahler manifolds.
It seems likely that a similar result should be true in the K-contact case, but so
far there are no known simply connected K-contact manifolds which do not admit
Sasakian structures. Thus, we have
Open Problem 7.4.1: Find examples of simply connected K-contact manifolds
which cannot admit Sasakian structures.
The work of Kollar [Kol05, Kol06a] should prove important in attacking this
problem. Indeed Kollar has discovered torsion obstructions to the existence of
Sasakian structures in dimension ve. We discuss this work in detail in Chapter
10. Perhaps it is worth mentioning here that the techniques of Gompf and Kollar
appear to be quite dierent. Another known obstruction to the existence of Kahler
structures is the existence of non-vanishing Massey products [DGMS75]. This
technique has been utilized in some of the references mentioned above, cf. [CFG93].
However, non-vanishing Massey products do not obstruct Sasakian structures as the
three dimensional Heisenberg group shows. Nevertheless, certain Massey products
do obstruct Sasakian structures, but they depend on the basic cohomology class
3
Thurston gave his example as a counterexample to a result of Guggenheimer who claimed
that the odd Betti numbers of a symplectic manifold were even after P. Libermann had noticed
that Guggenheimers proof was incomplete.
236 7. K-CONTACT AND SASAKIAN STRUCTURES
of a K-contact structure. It would be very interesting to obtain a topological
characterization of such obstructions.
As with Kahler manifolds another interesting question concerns the possible
fundamental groups of compact Sasakian manifolds. In the case of compact Kahler
manifolds quite a bit is known concerning which groups can be the fundamental
group of a compact Kahler manifold [ABC
+
96], whereas, almost nothing is known
for compact Sasakian manifolds. As with Kahler manifolds it is clear that there are
restrictions on the fundamental group of a compact Sasakian manifold, for example
Theorem 7.4.11 says that the rst Betti numbers of any compact Sasakian manifold
must be even. Moving away from the Kahler category to symplectic geometry, there
is again a result of Gompf [Gom95] saying that any nitely presented group can
be realized as the fundamental group of a compact symplectic manifold. Similarly
ACampo and Kotschick [AK94] have shown that any nitely presented group
is the fundamental group of some compact contact manifold. It is not known
whether there are any restrictions on the fundamental group of a compact K-contact
manifold. As in the Kahler case we have
Denition 7.4.17: A group is called a Sasaki group if it is the fundamental
group of some compact Sasakian manifold. The rst Betti number of /
1
() is
the rank of the Abelianization [. ].
As with Kahler metrics Sasakian metrics can be lifted to coverings, so subgroups
of nite index of a Sasaki group is Sasaki. Thus, we have the following negative
result:
Proposition 7.4.18: A group which contains a nite index subgroup
t
with
/
1
(
t
) odd is not Sasaki. In particular any non-trivial free group cannot be Sasaki.
It appears as though the analogy with Kahler groups stops with this proposi-
tion. For example we have seen above that Sasaki groups unlike Kahler groups can
have non-vanishing Massey products.
7.5. Sasakian Geometry and Algebraic Geometry
We begin the study of Sasakian structures per se. First we translate Theorem
7.3.7 to the Sasakian situation. In the case of a compact quasi-regular Sasakian
manifold ` we know that the space of leaves `T

is a compact Riemannian
orbifold Z. But since the transverse geometry on ` is Kahler, the orbifold must
be Kahler. But actually in the quasi-regular case more is true. It follows from
Theorem 7.3.7 that ` is the total space of a V-bundle over Z. and the curvature
of the connection form is precisely the pullback of the Kahler form on Z. Thus,
Z satises an orbifold integrality condition which we now elaborate. This integral-
ity condition ties Sasakian geometry on compact manifolds to projective algebraic
geometry.
We can now combine the results of this section with Theorem 6.3.8 or Theorem
7.1.3 to give one of the foundational theorems of our subject.
Theorem 7.5.1: Let o = (. . . o) be a compact quasi-regular Sasakian structure
on a smooth manifold of dimension 2n + 1, and let 7 denote the space of leaves of
the characteristic foliation (just as topological space). Then
(i) 7 carries the structure of a Hodge orbifold Z = (7. ) with an orbifold
Kahler metric / and Kahler form which denes an integral class [j

]
7.5. SASAKIAN GEOMETRY AND ALGEBRAIC GEOMETRY 237
in H
2
orb
(Z. Z) in such a way that : (`. o)(Z. /) is an orbifold Rie-
mannian submersion, and a principal o
1
orbibundle over Z. Furthermore,
satises

= d.
(ii) The bers of are geodesics.
(iii) 7 is also a Q-factorial, polarized, normal projective algebraic variety.
(iv) If o is regular then the orbifold structure is trivial and : `7 is a
principal circle bundle over a smooth projective algebraic variety.
Proof. (i) and (ii) follow from Theorem 7.1.3 and the fact that the transverse
almost contact structure is integrable. That Z is a projective algebraic variety
follows from the Kodaira-Baily embedding theorem 4.4.30, and since the only sin-
gularities are quotient singularities it is both normal and Q-factorial. Now denes
a rational class in H
2
(Z. Q) which pullback to the integral class [j

] H
2
orb
(Z. Z)
and thus denes a holomorphic line V-bundle whose unit circle bundle is the prin-
cipal o
1
V-bundle. Thus, (Z. []) is a polarization.
Notice that a quasi-regular Sasakian structure on a compact manifold can be
thought of as a transverse projective algebraic structure. For compact Kahler mani-
folds there was a conjecture attributed to Kodaira that says that any Kahler struc-
ture can be approximated or equivalently is a deformation of an algebraic structure.
It was shown to be true in case of compact Kahler surfaces by Kodaira [Kod60a].
However, it has been recently shown by Voisin [Voi04] that this does not hold in
complex dimension four and higher. More explicitly Voisin has given examples of
simply connected compact Kahler manifolds whose cohomology ring is not isomor-
phic to the cohomology ring of a projective algebraic variety. Perhaps surprisingly
the analog in Sasakian geometry does hold. A corollary of the Structure Theorem
7.1.10 which will be stated precisely in the next section says that every Sasakian
structure is a deformation of a quasi-regular Sasakian structure. Thus, in a real
sense (no pun intended), Sasakian geometry is more algebraic than Kahlerian ge-
ometry.
The Inversion Theorem 7.1.6 now becomes
Theorem 7.5.2: Let (Z. . J) be a Hodge orbifold with [j

] H
2
orb
(Z. Z). and
let ` denote the total space of the circle V-bundle dened by the class []. Then
the orbifold ` admits a quasi-regular Sasakian structure (. . . o) such that d =

where : `Z is the natural orbifold projection map, or alternatively :

`Z is a Seifert o
1
-bundle. Furthermore, if all the local uniformizing groups of
Z inject into the structure group o
1
, then Z is a cyclic orbifold and ` is a smooth
Sasakian manifold.
Example 7.5.3: Consider the weighted sphere o
2n+1
w
as described in Example
7.1.12 with integral weights ordered as 0 < n
0
n
n
. The Sasakian structure
o
w
= (
w
.
w
.
w
. o
w
) is quasi-regular for every such weight vector w. We shall
assume that gcd(n
0
. . . . . n
n
) = 1 for otherwise we can perform a transverse homo-
thety 7.3.10 to obtain another Sasakian structure. This doesnt change the orbifold
structure of the base space Z. but only resales the Kahler metric /. In our case
Z
w
= P(w) the weighted projective space of Denition 4.5.2. This is a Kahler orb-
ifold as well as a Q-factorial normal projective algebraic variety. The polarization is
dened by the unique Kahler class [
w
] satisfying

[
w
] = [d
w
]
B
. On the orbifold
Z
w
the canonical orbisheaf of Denition 4.4.12 is the orbisheaf O
P(w)
([w[). The
Sasakian structure o
w
is determined by the orbifold structure of Z
w
. For example,
238 7. K-CONTACT AND SASAKIAN STRUCTURES
consider the weighted Sasakian structure o
w
with w = (6. 2(6/1). 3(6/1)) with
order (o) = 6(6/ 1). The orbifold Z
w
is that given by Example 4.5.11, namely
P(6. 2(6/ 1). 3(6/ 1)). As an algebraic variety P(6. 2(6/ 1). 3(6/ 1)) is just
P
2
. but not as orbifolds. The Sasakian structure with Z = P
2
is that of the round
sphere of Example 7.1.5 and has order = 1.
Another similar example is:
Example 7.5.4: Consider as above the weighted projective space CP
2
(n
0
. n
1
. n
2
)
dened by the usual weighted C

action on C
3
`0. where now n
0
. n
1
. n
2
are pair-
wise relatively prime integers. As an algebraic variety CP
2
(n
0
. n
1
. n
2
) is equivalent
[Kol96] to CP
2
Z
w
0
Z
w
1
Z
w
2
. But as orbifolds these are distinct, since the for-
mer has
orb
1
= 0. whereas the latter has
orb
1
= Z
w0
Z
w1
Z
w2
. The induced
metrics are also dierent. In the latter case the metric is just the Fubini-Study
metric pushed to the quotient which is Kahler-Einstein. The Sasakian structure on
the corresponding o
1
V-bundle is just the standard Sasaki-Einstein structure on the
lens space o
5
Z
w
0
Z
w
1
Z
w
2
; whereas, in the former case the Sasakian structure
is just the weighted sphere as described in Example 7.1.12 and the induced metric
is ellipsoidal and not Einstein. Likewise, the metric / on CP
2
(n
0
. n
1
. n
2
) is not
Kahler-Einstein. In both cases the order is the same, namely = n
0
n
1
n
2
.
Let Z be a Kahler orbifold and recall the Kahler cone 1(Z) from Denition
3.6.2. In 1(Z) we dene the Kahler lattice 1
L
(Z) of the complex orbifold Z to be
the lattice
(7.5.1) 1
L
(Z) = [] 1(Z) [ [j

] H
2
orb
(Z. Z) .
From the Inversion Theorem 7.5.2 we have
Theorem 7.5.5: Let (Z. ) be a Hodge orbifold. Every [] 1
L
(Z) determines
a principal o
1
V-bundle : `
[]
Z and choosing a connection in `
[]
whose
curvature is

determines a Sasakian structure on `

[]
.
Recall that the class [] H
2
(7. Q) is just the rational Chern class of the o
1
orbibundle, and generally [] does not uniquely determine a complex line orbibun-
dle. However, Kollar (Proposition 53 in [Kol04]) has recently proven that [] does
uniquely determine a complex orbibundle if H
orb
1
(Z. Z) = 0 in which case the o
1
V-bundle : `
[]
Z with its CR structure is uniquely determined.
7.5.1. The Space of Sasakian Structures. Here we consider those Sasakian
structures that have the same Reeb vector eld. Fix a Sasakian structure o =
(. . . o) and dene the set
(7.5.2) F() = Sasakian structures (
t
.
t
.
t
. o
t
) [
t
= .
We give the set F() the C

compact-open topology as sections of vector bundles,

and call it the space of Sasakian structures compatible with or just the space of
Sasakian structures when the Reeb vector eld is understood to be . Dene the real
1-form =
t
. Then is basic, so [d
t
]
B
= [d]
B
. Hence, all Sasakian structures
in F() correspond to the same basic cohomology class in H
2
B
(T

). Furthermore, it
is clear that the order (o) does not depend on the Sasakian structure o F(),
and thus is an invariant of the space F().
Let

J denote the underlying almost complex structure on (T

). and

:
T`(T

) the natural projection. We dene the subspace F(.

J) F() to be
7.5. SASAKIAN GEOMETRY AND ALGEBRAIC GEOMETRY 239
the subset of all Sasakian structures (
t
.
t
.
t
. o
t
) F() such that the diagram
(7.5.3)
T`

T`

(T

J
(T

).
commutes. The elements of F(.

J) consist of all Sasakian structures o
t
F() with
the same transverse holomorphic structure.
We can now give an alternative description of F(.

J). but rst we need the
transverse

lemma due to El Kacimi-Alaoui [EKA90].

Lemma 7.5.6: Let (`. o) be a compact Sasakian manifold, and let .
t
be basic
real closed (1. 1)-forms such that []
B
= [
t
]
B
. then there exists a smooth basic
function such that
t
= +i

.
Proof. The proof is similar to the usual case [EKA90].
As in the Kahler case the basic 2-form d can be written locally in terms of a
basic potential function , viz. d = i

. So as with Kahler geometry, Sasakian

geometry is locally determined by one potential function [GKN].
A real basic (1. 1)-form on a Sasakian manifold (`. . . . o) is said to be
positive, written 0. if (A. A) 0 for all smooth sections A of T. We can
obtain a characterization of the space of Sasakian metrics on ` whose Reeb vector
eld is and whose transverse holomorphic structure is

J. We now dene certain
subspaces of the space of all smooth basic functions on (`. T

).
(
1
0
(o) = C

(`)
B
[ d +i

0.

M
dvol
g
.
(
0
0
(o) = C

(`)
B
[

M
dvol
g
= 0 .
Proposition 7.5.7: Let (`. . . . o) be a compact Sasakian manifold with under-
lying transverse holomorphic structure

J. Then F(.

J) = (
1
0
(o)(
0
0
(o)H
1
(`. Z).
Proof. Let o
t
= (.
t
.
t
. o
t
) F(.

J). then as mentioned above d and d
t
represent the same basic cohomology class in H
1,1
B
(T

). Thus, by Lemma 7.5.6

(7.5.4) d
t
= d +i

for some smooth basic function . Since o

t
is Sasakian we must have d+i

0.
However, the function does not uniquely determine the Sasakian structure. Dene
d
c
=
i
2
(

) so that i

= dd
c
. Then the 1-form =
t
d
c
is closed and
basic; hence, it denes an element of H
1
(`. R) H
1
(T

) which lies in the image of

the natural inclusion H
1
(`. Z)H
1
(`. R). Let denote the class in H
1
(`. Z)
dened by . Then any other closed basic 1-form
t
belonging to the same class
diers from by d where is a smooth basic function on `. Since both and
are only dened up to a constant, they can be chosen to satisfy the normalization
condition

M
1 dvol
g
= 0. This then determines the pair (. ) uniquely. It is now
easy to see that the map o
t
(. . ) is one-to-one.
Conversely, with a xed Sasakian structure o = (. . . o) let (. . )
(
1
0
(o) (
0
0
(o) H
1
(`. Z) be given. The class H
1
(`. Z) denes a closed
240 7. K-CONTACT AND SASAKIAN STRUCTURES
1-form up to a the dierential of a smooth basic function. So the pair (. ) de-
nes the 1-form uniquely, the constant being xed by the normalization condition
on . Dene
t
by
(7.5.5)
t
= +d
c
+ .
Then
t
is a contact form on ` since if
t
(d + dd
c
)
n
were to vanish at some
point, taking the interior product with shows that (d+dd
c
)
n
would vanish which
violates the positivity condition. Now putting
t
= d
c
+ we have
t
= +
t
. We
dene the endomorphism
t
and metric o
t
by

t
=
t
.
o
t
= d
t
(
t
1l) +
t

t
. (7.5.6)
Clearly,
t
and project to the same transverse holomorphic structure

J. To see
that
t
is uniquely determined by the triple (. . ) we notice that if
t
and
both project to

J then there is a smooth 1-form such that
t
= + .
But then the condition
t

t
= 0 implies that =
t
. Finally, we need to
check the positivity condition d
t
(
t
1l) 0. as well as the condition (
t
)
2
=
1l +
t
. These are straightforward and left as an exercise. Furthermore, under
the above identication the topology of F(.

J) coincides with the relative topology
as a subspace of C

(`)
B
C

(`)
B
H
1
(`. Z).
The group H
1
(`. Z) labels the components F(.

J)

of F(.

J). We have
Corollary 7.5.8: Each component F(.

J)

is contractible.
If we x the component F(.

J)

, it is convenient to think of a transformation

of the form given by equation (7.5.5) as a deformation within a give component.
This point of view was adobted in [Bel01] and we formalize this as:
Denition 7.5.9: A transformation within F(.

J)

of the form + for

some 1-form is called a deformation of type II.
Deformations of type I will be dened in Denition 8.2.10 below.
Remark 7.5.1: It is interesting to note that in the quasi-regular case F(.

J) can
be viewed as a subspace of the space of connections in a l(1) V-bundle, and indeed
even in the irregular case as a subspace of the space of connections in a 1-dimensional
foliation. Explicitly, F(.

J) is the subspace of connections with strictly positive
curvature satisfying the normalization condition. The function is then identied
with a null-homotopic gauge transformation.
Our next result says that of all the Sasakian structures in F() have the same
volume. (We always take the volume element dvol
g
to have positive orientation.)
Explicitly,
Proposition 7.5.10: Let `
2n+1
be a compact Sasakian manifold. Then for o. o
t

F() arbitrary, we have Vol(`. o) = Vol(`. o
t
).
Proof. The innitesimal volume element of a Sasakian metric is related to
the contact volume element, up to orientation, by
dvol
g
=
1
n!
(d)
n
.
7.5. SASAKIAN GEOMETRY AND ALGEBRAIC GEOMETRY 241
Now by our discussion above there is a smooth basic function and closed basic
1-form such that
dvol
g
=
1
n!

t
(d
t
)
n
= ( +d
c
+) (d +dd
c
)
n
= (d +dd
c
)
n
.
where the last equality holds since d
c
+ is basic. Now expand
(d +dd
c
)
n
= (d)
n
+
n

i=1

n
i

(d)
ni
(dd
c
)
i
.
so that
Vol(`. o
t
) =

M
dvol
g
=
1
n!

M
(d +dd
c
)
n
=
1
n!

(d)
n
+
n

i=1

n
i

(d)
ni
(dd
c
)
i

=
1
n!

M
(d)
n
+
1
n!
n

i=1

n
i

M
d

(d)
ni
d
c
(dd
c
)
i1

=
1
n!

M
(d)
n

1
n!
n

i=1

n
i

M
d

(d)
ni
d
c
(dd
c
)
i1

M
dvol
g
= Vol(`. o) .
where the last line follows by Stokes theorem.
Henceforth, we shall often write F() for F(.

J) when it is understood that the
transverse complex structure is xed. Later in Section 8.2.1 we discuss deformations
of the transverse holomorphic structure. It will be convenient to consider a set
of Sasakian structures that is slightly larger than F(). namely those Sasakian
structures that have the same characteristic foliation. First we have
Denition 7.5.11: For any Sasakian structure o = (. . . o) there is the con-
jugate Sasakian structure dened by o
c
= (
c
.
c
.
c
. o) = (. . . o)
F(T

) .
Clearly, this denition holds equally well for K-contact structures. We dene
F(T

) to be the set of all Sasakian structures whose characteristic foliation is T

.
Clearly, we have F() F(T

). Recall the transverse homothety 7.3.10. Then xing

o we dene
(7.5.7) F
+
(T

) =

aR
+
F(o
1
) .
and F

(T

) to be the image of F
+
(T

) under conjugation. We now have

Lemma 7.5.12: The decomposition holds
F(T

) = F
+
(T

) . F

(T

) .
Proof. The one inclusion is obvious. Let o
t
= (
t
.
t
.
t
. o
t
) F(T

). Then
there is a nowhere vanishing smooth function 1 such that
t
= 1
1
. and
t
= 1.
Then the condition
t
d
t
= 0 implies that
(7.5.8) d1 = (1).
242 7. K-CONTACT AND SASAKIAN STRUCTURES
and this implies that A1 = 0 for any horizontal vector eld A. So 1 only depends
on the leaf coordinates. Taking d of Equation 7.5.8 gives
0 = d(1) + (1)d .
If we can show that the rst term vanishes, then 1 = 0 so 1 will be constant
which will prove the lemma. Now the rst term d(1) = d
h
(1) where
d
h
is the exterior derivative in the horizontal direction. So we compute A1 for
A horizontal. We have, using the fact that 1 is independent of the horizontal
coordinates, A1 = [A. ]1 = 0. since [A. ] is horizontal.
This discussion shows that if we x a foliation T

and a class H
1
(`. Z),
the subset of homology classes represented by compatible Sasakian structures forms
a line minus the origin in H
2
B
(T

). and that conjugation interchanges the positive

and negative rays. We shall refer to F

(T

as an o-homology class of Sasakian

structures, or simply a Sasakian ray, and we shall always mean F
+
(T

unless
otherwise stated. We need
Denition 7.5.13: Two Sasakian structures o = (. . . o) and o
t
= (
t
.
t
.
t
. o
t
)
in F(T

on a smooth manifold ` are said to be o-homologous if there is an

o R
+
such that
t
= o
1
and [d
t
]
B
= o[d]
B
.
The o-homology classes form a set of two elements that can be identied with
positive and negative rays in H
2
B
(T

). So every Sasakian structure in F(T

is o-
homologous to o or its conjugate o
c
. For example, the standard Sasakian structure
o
0
on o
2n+1
discussed in Example 7.1.5 is a representative in the standard o-
deformation class F
+
(T

0
). while o
c
0
= (
0
.
0
.
0
. o
0
) is a representative in
F

(T
0
). Of course, the other representatives of F(T
0
) do not, in general, have
constant sectional curvature. The dieomorphism C : o
2n+1
o
2n+1
induced by
complex conjugation on C
n+1
sends o
0
to o
c
0
. Note that C is an element of the
contactomorphism group Con(o
2n+1
. T
0
= ker
0
) (as well as the isometry group
Isom(`. o
0
)) which preserves orientation if n is odd, and reverses orientation for
n even. For n odd C reverses the co-orientation. A natural question to ask is:
Question 7.5.1: Given a Sasakian manifold (`. o) when does there exist a C
Con(`. T) Isom(`. o
0
) such that C(o) = o
c
?
This question is discussed further in Chapter 8. Summarizing we have
Theorem 7.5.14: The space of Sasakian structures F(T

corresponding to a
class H
1
(`. Z) consists of two contractible components F
+
(T

and F

(T

that are isomorphic under conjugation. Moreover, any two Sasakian structures in
F(T

are contactomorphic.
We are specially interested in the case that the Sasakian structure o is quasi-
regular. Then according to Theorem 7.5.2 o determines a Seifert o
1
-bundle. Since
quasi-regularity is a property of the foliation T

, every o F(T

)
+

determines
the same Seifert bundle, whereas, elements of F(T

correspond to the conjugate

Seifert o
1
-bundle with the orientation of the bres reversed and with the complex
conjugate structure on Z. Thus, we can think of F(T

)
+
, or by abuse of terminology
F(T

), as a certain type of Seifert o

1
-bundle. This leads us to
Denition 7.5.15: If T

is quasi-regular we call F(T

)
+

(or even F(T

) a
Sasaki-Seifert structure.
7.5. SASAKIAN GEOMETRY AND ALGEBRAIC GEOMETRY 243
Thus, a Sasaki-Seifert structure is just an o-deformation class of Sasakian struc-
tures under the additional assumption of quasi-regularity. Strictly speaking the
transverse complex structure

J is xed for a given Sasaki-Seifert structure since
the complex structure of the orbifold A = (A. ) is xed. However, as we shall
see Sasaki-Seifert structures often occur in continuous families (moduli), and we
will sometimes use the term Sasaki-Seifert structure to refer to an entire family of
such structures. The context should make this clear. Note that the Approximation
Theorem 7.1.10 says that a manifold which admits a Sasakian structure necessarily
admits a Sasaki-Seifert structure.
Open Problem 7.5.1: The following two problems are of paramount interest to
us in this book.
(1) Determine which manifolds admit Sasakian structures.
(2) Determine all Sasaki-Seifert structures on a given manifold M.
In full generality these problems seem quite hopeless to answer. However, we
shall see in Chapter 10 that under some added conditions, such as positivity to be
discussed in the next section, one can go quite far in certain cases.
7.5.2. Chern Classes and Basic Chern Classes. Consider the complex
vector bundle T on a Sasakian manifold (`. . . . o). As such T has Chern
classes c
1
(T). . . . . c
n
(T) which can be computed by choosing a connection
1
in T
[Kob87]. We choose
1
to be the transverse connection
T
of 2.5.7 in which case
the curvature is just the transverse curvature 1
T
given by equation (2.5.8). Let us
choose a local foliate unitary transverse frame (A
1
. . . . . A
n
). and denote by
T
the
transverse curvature 2-form with respect to this frame. We have
Proposition 7.5.16:
T
is a basic (1. 1)-form.
Proof. The transition functions on the transverse coordinates for the foliation
T

are holomorphic and only involve the transverse coordinates. So the usual proof
carries over to this case.
Since the curvature 2-form
T
has type (1. 1) it follows as in ordinary Chern-
Weil theory that
Theorem/Denition 7.5.17: The /th Chern class c
k
(T) of the complex vector
bundle T is represented by the basic (/. /)-form
k
determined by the formula
det

1l
n

1
2i

= 1 +
1
+ +
k
.
Since
k
is a closed basic (/. /)-form it represents an element in H
k,k
B
(T

)
H
2k
B
(T

) that is called the basic /

th
Chern class and denoted by c
k
(T

).
Notice that the Chern classes c
k
(T) are invariants of the underlying contact
structure.
Recall the transverse Ricci tensor Ric
T
of o
T
. It is related to the Ricci tensor
Ric
g
of o by the second equation in Theorem 7.3.12. Now as usual dene the Ricci
form
g
and transverse Ricci form
T
g
by
(7.5.9)
g
(A. Y ) = Ric
g
(A. Y ).
T
g
(A. Y ) = Ric
T
(A. Y )
for smooth sections A. Y of T. It is easy to check that these are anti-symmetric of
type (1. 1) and Theorem 7.3.12 implies that they are related by
(7.5.10)
T
g
=
g
+ 2d .
244 7. K-CONTACT AND SASAKIAN STRUCTURES
Thus, as in the usual case we have
Lemma 7.5.18: The basic class 2c
1
(T

) H
1,1
B
(T

) is represented by the trans-

verse Ricci form
T
g
.
More importantly there is a transverse version of Yaus famous Theorem given
by El Kacimi-Alaoui [EKA90]:
Theorem 7.5.19: If 2c
1
(T

) is represented by a real basic (1. 1) form

T
. then
it is the Ricci curvature form of a unique transverse Kahler form
T
in the same
basic cohomology class as d.
This is translated into Sasakian language as
Theorem 7.5.20: Let (`. . . . o) be a Sasakian manifold whose basic rst Chern
class c
1
(T

) is represented by the real basic (1. 1) form

1
2
. then there is a unique
Sasakian structure (.
1
.
1
. o
1
) F() homologous to (. . . o) such that
g
1
=
2d
1
is the Ricci form of o
1
. and
1
= +
1
. with
1
=
1
2
d
c
. The metric o
1
and endomorphism
1
are then given by equations (7.5.6).
We also have
Proposition 7.5.21: The basic Chern classes c
k
(T

) are independent of the choice

of a Sasakian structure in F(T

). In particular, if T

is quasi-regular c
k
(T

) only
depends on the Sasaki-Seifert structure.
Proof. Since the
k
is invariant under a G1(n. C) change of local frame, it
is also invariant under a homothety of the transverse metric. So the basic Chern
classes c
k
(T

) are invariant under a transverse homothety, as well as invariant under

the addition of a (1. 1)-form in the image of

.
We denote by c
k
(T)
R
the image of the class c
k
(T) under the natural inclusion
H
2k
(`. Z)H
2k
(`. R).
Lemma 7.5.22: The basic Chern class c
k
(T

) maps to c
k
(T)
R
under the natural
inclusion

. Moreover, an element []
B
H
2k
B
(T

) maps to c
k
(T)
R
if and only if
there exists a closed basic (2/ 2)-form such that
[]
B
= c
k
(T

) + [1]
B
.
In particular, c
k
(T) is a torsion class if and only if there exists a closed basic
(2/ 2)-form such that c
k
(T

) = [1]
B
.
Proof. Since the connection in T is the transverse connection
T
and its
curvature form is basic by Proposition 7.5.16, the Chern classes c
k
(T) are computed
by the
k
s of Theorem/Denition 7.5.17. This proves the rst statement.
From the exact sequence (7.2.1) we have the diagram of exact sequences
(7.5.11)
0

H
2k
(`. Z)

H
2k2
B
(T

)

H
2k
B
(T

)

H
2k
(`. R) .
So the second statement follows since ([]
B
) = [d]
B
[]
B
= [d ]
B
= [1]
B
.
and the last statement is clear from the diagram (7.5.11).
7.5. SASAKIAN GEOMETRY AND ALGEBRAIC GEOMETRY 245
In case of a quasi-regular Sasakian structure, the basic Chern classes are related
to the real orbifold Chern classes in a rather obvious way, namely:
Proposition 7.5.23: Let (`. . . . o) be a quasi-regular Sasakian manifold with
o
1
orbibundle : `Z. Then as real cohomology classes c
k
(T

) =

c
orb
k
(Z).
Proof. This follows immediately from the fact that the transverse Ricci form

T
is the pullback of the Ricci form on the orbifold Z.
We now concentrate on the rst Chern classes c
1
(T) and c
1
(T

). We give an
immediate corollary of Lemma 7.5.22, but rst a denition.
Denition 7.5.24: A Sasakian structure o = (. . . o) F(T

) or a Sasaki-
Seifert structure F(T

) is said to be of positive (negative) type if c

1
(T

) can be
represented by a positive (negative) denite (1. 1)-form. If either of these two con-
ditions is satised F(T

) is said to be of denite type. F(T

) is said to be of null
type if c
1
(T

) = 0. A Sasakian structure (. . . o) which is neither denite nor

null is said to be of indenite type. A Sasakian structure o = (. . . o) F(T

)
is called anticanonical (canonical) if c
orb
1
(Z) is a positive (negative) multiple of
[d]
B
.
We apply the denition to F(T

) whether T

is quasi-regular or not. We usually

drop the terminology type by saying for example, a positive Sasakian structure
instead of a Sasakian structure of positive type. The following self-evident result
says that being anticanonical or canonical is a property of the deformation class
F(T

).
Proposition 7.5.25: If o = (. . . o) F(T

) is anticanonical (or canonical)

then so is any other Sasakian structure o
t
= (. . . o) F(T

). An anticanonical
(canonical) Sasakian structure is positive (negative), respectively.
The converse of he last statement is certainly not true as seen from the Wang-
Ziller manifolds of Example 7.6.10 below.
Now we consider the case / = 1 in Lemma 7.5.22. Then diagram (7.5.11)
becomes
0

H
2
(`. Z)

0H
0
B
(T

)

H
2
B
(T

H
2
(`. R) .
and Lemma 7.5.22 implies
Corollary 7.5.26: Let o = (. . . o) be a Sasakian structure with underlying
contact bundle T. Then c
1
(T) is a torsion class if and only if there exists a real
number o such that c
1
(T

) = o[d]
B
. So
(i) if c
1
(T

) = 0 then c
1
(T) is a torsion class if and only if the Sasakian
structure o is either anticanonical or canonical,
(ii) if o is a null Sasakian structure then c
1
(T) is a torsion class.
Remark 7.5.2: Let ` be a quasi-regular Sasakian manifold with /
2
(`) 0 and
space of leaves 7 = `T

with its induced orbifold structure Z = (7. |). Assume

also that c
orb
1
(Z) is either positive or negative. The condition of being anticanonical
246 7. K-CONTACT AND SASAKIAN STRUCTURES
or canonical chooses a ray in the Kahler cone 1(Z). Thus, there are many positive
(negative) Sasakian structures that are not anticanonical or canonical corresponding
to choices of Seifert o
1
-bundle (orbibundle) whose rst Chern class c
1
(`7) is not
a multiple of c
1
(T

). The situation is dierent in the null case. If c

orb
1
(Z) = 0.
every (1. 1) class in 1(Z) satisfying the correct positivity and integrality conditions
gives rise to a null Sasakian structure which stand more or less on equal footing.
For example as discussed later in Chapter 11, all null Sasakian structures admit
a null -Einstein metric, whereas, only relatively few positive (negative) Sasakian
structures are anticanonical (canonical) which is a necessary condition to admit a
non-null -Einstein metric.
In analogy with common terminology of smooth algebraic varieties we see that
a positive Sasakian structure is a transverse Fano structure, while a null Sasakian
structure is a transverse Calabi-Yau structure. The negative Sasakian case cor-
responds to the canonical bundle being ample, and could be called a transverse
canonical structure; however, this conicts with the terminology introduced in Def-
inition 7.5.24, so we do not use it.
Note that c
1
(T) is an invariant of the complex vector bundle T. but it is
related to a topological invariant, namely the second Stiefel-Whitney class n
2
(`).
It is well-known (cf. [LM89]) that a vector bundle 1 admits a spin structure if
and only if n
2
(1) = 0. and that the distinct spin structures on 1 are in one-to-one
correspondence with the elements of H
1
(`. Z
2
). Furthermore, if 1 is a complex
vector bundle then n
2
(1) is the mod 2 reduction of c
1
(1). Indeed, let (`. ) be a
strict contact manifold with contact bundle T. Let be the Reeb vector eld, then
we have a natural splitting
T` = T 1

.
where 1

is the trivial real line bundle generated by . Thus,

n
2
(`) = n
2
(T`) = n
2
(T 1

) = n
2
(T) .
So if ` admits a compatible Sasakian structure, the contact bundle T is a complex
vector bundle, so n
2
(`) is the mod 2 reduction of c
1
(T) H
2
(`. Z). Hence,
Corollary 7.5.26 implies
Theorem 7.5.27: Let (`. . . . o) be a Sasakian manifold with H
1
(`. Z)
tor
= 0.
and suppose that c
1
(T

) = o[d]
B
for some real number o. Then ` is spin. In
particular, if ` satises H
1
(`. Z)
tor
= 0 and admits an anticanonical, canonical,
or null Sasakian structure, then ` is spin.
Theorem 7.5.27 says that n
2
(`) is an obstruction for a simply connected man-
ifold ` to admit a Sasakian structure o = (. . . o) satisfying c
1
(T

) = o[d]
B
.
and the next example shows that some condition on H
1
(`. Z) is necessary. (Note
that Proposition 2.6 of [BGN03a] is incorrect. This is illustrated by Example
7.6.10 below where a counterexample is given. The correct statement concerning
Sasakian manifolds and spin is Theorem 7.5.27).
Example 7.5.28: Consider the real projective spaces RP
2n+1
which clearly admit
Sasakian structures (the induced metric from round sphere metric on o
2n+1
is
Sasakian). For any Sasakian structure H
2
B
(T

) is one dimensional, so c
1
(T

) =
o[d]
B
is automatically satised. Moreover [LM89],
n
2
(RP
2n+1
) (n + 1) mod 2 .
7.5. SASAKIAN GEOMETRY AND ALGEBRAIC GEOMETRY 247
So RP
2n+1
is not spin if n is even. This shows that a condition on H
1
(`. Z) is
needed.
The type of Sasakian structures that can occur on rational homology spheres
are quite restrictive. We can give a characterization of type of Sasakian structures
that can occur on rational homology spheres in terms of the basic cohomology ring
H

B
(T

).
Proposition 7.5.29: Let `
2n+1
be a manifold with a Sasakian structure o =
(. . . o). Then `
2n+1
is a rational homology sphere if and only if the basic coho-
mology ring H

B
(T

) is isomorphic to the real cohomology ring of complex projective

space CP
n
. Moreover, any Sasakian structure o on a rational homology sphere sat-
ises c
1
(T

) = o[d]
B
for some non-zero constant o. or equivalently, the Sasakian
structure o is either anticanonical or canonical. Hence, o is either positive or
negative.
Proof. To prove the rst statement we analyze the basic long exact sequence
(7.2.1). If ` is a rational homology sphere this implies the isomorphisms
H
p1
B
(T

) H
p+1
B
(T

)
for j = 1. . . . . 2n 1. and since H
1
B
(T

) = 0 and H
2
B
(T

) is generated by [d]
B
.
this implies the ring isomorphism
(7.5.12) H

B
(T

) R

[d]
B

([d]
B
)
n+1
.
Conversely, the ring isomorphism (7.5.12) implies that the connecting homomor-
phism in (7.2.1) is an isomorphism. Putting j = 2:+1 shows that H
2r+1
(`. R) =
H
2r+2
(`. R) = 0 for : = 0. . . . . n 1 which implies that `
2n+1
is a rational ho-
mology sphere. This proves the rst statement.
For the second statement we rst notice that the ring isomorphism 7.5.12 im-
plies that c
1
(T

) = o[d]
B
for some o R. Moreover, o cannot vanish since this
would imply that the basic geometric genus j
g
(T

) = 0 implying /
n
(`) = 0 by the
transverse Hodge decomposition 7.2.6 which would give a contradiction.
Notice that Proposition 7.5.29 and Theorem 7.5.27 have the following
Corollary 7.5.30: A rational homology sphere ` with H
1
(`. Z)
tor
= 0 which
admits a Sasakian structure is a spin manifold.
An example of a rational homology sphere satisfying H
1
(`. Z)
tor
= 0 is the
simply connected homogeneous 5-manifold ol(3)oO(3). This cannot admit a
Sasakian structure. In fact as discussed in Section 10.2 below ol(3)oO(3) does
not admit a contact structure nor an almost contact structure. Other examples of
rational homology 5-spheres, due to Kollar [Kol05], which cannot admit a Sasakian
structure are discussed in Chapter 10. We shall see some non-spin Sasakian mani-
folds in Example 7.6.10 below.
Although not every Sasakian manifold is spin, every Sasakian manifold (in
fact, any K-contact manifold) admits a Spin
c
-structure since n
2
(`) is the mod
2 reduction of an integral class, namely c
1
(T). In fact since any Sasakian mani-
fold `
2n+1
admits a reduction of the frame bundle to the group l(n) 1. there
is a canonical Spin
c
-structure on `
2n+1
determined by the canonical homomor-
phism l(n)Spin
c
2n
. We refer to Appendix D of [LM89] for the details of Spin
c
-
structures. The observation that any Sasakian manifold is Spin
c
was rst made in
[Mor97].
248 7. K-CONTACT AND SASAKIAN STRUCTURES
7.5.3. Positive Sasakian Structures. In this book a special role is played
by Sasakian structures with positive basic rst Chern class c
1
(T

). Our main results

in this area appear in [BGN03a]. However, before embarking on this discussion we
give a very brief discussion of the well-known obstructions to the existence of pos-
itive scalar curvature metrics. For a more thorough treatment we refer the reader
to [LM89, RS01]. The rst of these obstructions was discovered in 1963 by Lich-
nerowicz [Lic63a] in his study of harmonic spinors in dimension 4n. Later Hitchin
[Hit74] generalized this to an invariant (`) which lives in the real K-theory
group 1O
n
which is non-vanishing only in dimensions 0 mod 4. or 1. 2 mod 8.
Then the Hitchin-Lichnerowicz Theorem says that if ` is a compact spin manifold
with (`) = 0. then ` does not admit a metric of positive scalar curvature. A
converse to this theorem was attacked by Gromov and Lawson [GL80] with much
progress toward its solution, but the full converse was left as a conjecture, known
as the Gromov-Lawson conjecture. The simply connected case was settled in 1992
by Stolz [Sto92], but the full conjecture still remains open. Stolz proved that a
simply connected manifold of dimension n 5 admits a metric of positive scalar
curvature if and only if (`) = 0. This is known to be false in dimension 4 [RS01].
For a complete statement of the Gromov-Lawson conjecture and its variants the
reader is referred to [RS01]. In the context of Sasakian geometry in dimension 1
mod 8 one can ask
Question 7.5.2: Are there Sasakian manifolds that do not admit metrics of posi-
tive scalar curvature?
From Stolzs Theorem or the earlier result of Gromov and Lawson [GL80] any
simply connected manifold of dimension 5 or 7 admits a metric of positive scalar
curvature, so for an armative answer to the question in the simply connected case
one should start with dimension 9 and the Hitchin exotic spheres which we briey
discuss later in Section 9.4.
We now turn to a general discussion of positive Sasakian structures. Before
doing so we mention that Kollar [Kol05] has recently found strong conditions on
the torsion subgroup of H
2
(`. Z) for a simply connected ve manifold ` to admit
a positive Sasakian structure. We describe this in detail in Chapter 10.
Theorem 7.5.31: Let o = (. . . o) be a positive Sasakian structure on a com-
pact manifold ` of dimension 2n + 1. Then ` admits a Sasakian structure o
t
=
(
t
.
t
.
t
. o
t
) with positive Ricci curvature o-homologous to o.
Proof. c
1
(T

) can be represented by a positive denite basic (1. 1)-form .

so by Theorem 7.5.20 there is a Sasakian structure o
1
= (.
1
.
1
. o
1
) F()
homologous to (. . . o) such that
g1
= 2d
1
is the Ricci form of o
1
. Let o
T
1
denote the transverse Kahler metric of this Sasakian structure. Then by Theorem
7.3.12 the Ricci curvatures of o
T
1
and o
1
are related by
Ric
g
1
[
1
1
1
1
= Ric
g
T
1
2o
T
1
.
Next for any real number o 0 we can perform a transverse homothety (cf. equa-
tion (7.3.10)) of the Sasakian structure by dening
o
T
2
=
1
o
o
T
1
.
2
=
1
o

1
.
2
= o.
2
=
1
.
in which case o
2
= (
2
.
2
.
2
. o
2
) is a Sasakian structure where o
2
= o
T
2
+
2

2
by Proposition 7.3.20. Notice also that o
1
and o
2
both have the same contact
7.5. SASAKIAN GEOMETRY AND ALGEBRAIC GEOMETRY 249
subbundles with the same underlying transverse complex structures, and the same
characteristic foliations. Now since the Ricci tensor is invariant under homothety
we nd
Ric
g
2
[
1
2
1
2
= Ric
g
T
2
2o
T
2
= Ric
g
T
1

2
o
o
T
1
.
But since 1ic
g
T
1
0 and ` is compact there exists o
0
R
+
such that for all o o
0
we have Ric
g
2
[
1
2
1
2
0. But also for any Sasakian metric we have Ric
g
2
(A.
2
) =
2n
2
(A) which implies that the Ricci curvature of o
2
is positive. But clearly o
2
is
o-homologous to o which proves the result.
Next we have a variation of Myers Theorem based on a result due to Hasegawa
and Seino [HS82].
Theorem 7.5.32: Let ` be a manifold which admits a positive Sasakian structure
o = (. . . o) with o complete and c
1
(T

) for some 0. Then ` is compact

with nite fundamental group.
Proof. As in the Proof of Theorem 7.5.31 we have an o-homologous Sasakian
structure (o.
2
.
2
. o
2
) such that
Ric
g2
[
1212
= Ric
g
T
2
2o
T
2
= Ric
g
T
1

2
o
o
T
1

2
o
which is positive for o large enough. So the result follows by (i) of Theorem 7.3.12
and Myers Theorem (cf. [Pet98]).
As in algebraic geometry positivity implies strong restrictions on the type of
transverse holomorphic structure allowed. So we can think about positive Sasakian
structures as transverse Fano structures as the following result [BGN03a] suggests.
Proposition 7.5.33: Let o = (. . . o) be a quasi-regular positive Sasakian struc-
ture on a compact manifold ` with o
1
orbibundle : `Z. then
(i) Ric
g
2 and the orbifold Z is Fano, i.e., c
orb
1
(Z) 0.
(ii) /
p,0
(T

) = 0 for all j 0. so
hol
(T

) = 1. and j
g
(T

) = j
a
(T

) = 0.
(iii) The underlying complex space 7 of the orbifold Z is simply connected.
(iv) As an algebraic variety 7 is uniruled and its Kodaira dimension is .
Proof. (i) follows from equation (7.5.10). (ii): Any non-trivial element of
H
p,0
B
(T

) is represented by a basic (j. 0)-form in the kernel of

. i.e., by a trans-
versely holomorphic j-form . Since o = (. . . o) is quasi-regular, ` is the total
space of an orbifold o
1
V-bundle over a compact Kahler orbifold Z. Moreover, since
is basic, it descends to a non-trivial element H
p,0
(Z). Now since the Sasakian
structure o is positive so is the Kahler structure on Z. i.e. c
orb
1
(Z) 0. and this
implies /
p,0
(Z) = 0 by the Kodaira-Baily vanishing Theorem 4.4.28.
The proof of simple connectivity is essentially a translation of Kobayashis proof
in [Kob61] for the case of manifolds to the case of complex orbifolds. By positivity
of the transverse Ricci form
T
=

. the orbifold Z has at most a nite cover.

By (ii) the transverse holomorphic Euler number (Z. O

) = 1. and this must

also hold on any nite cover. But from the above Z is projective algebraic, hence
complete. So the version of the Hirzebruch-Riemann-Roch Theorem for singular
algebraic varieties due to Baum, Fulton, and Macpherson (cf. [Ful84],pg 354)
applies and we have
1 = (Z. O

) = 'Td(Z). [Z]` .
250 7. K-CONTACT AND SASAKIAN STRUCTURES
where Td(Z) is the Todd class. Moreover, this equation holds on any /-fold cover

Z as well, and 'Td(

Z). [

Z]` = /'Td(Z). [Z]` . This forces / = 1.
Similar results proving simple connectivity in a purely algebro-geometric set-
ting, that is without the orbifold hypothesis appear in [Tsu88] and [Tak00]. If we
restrict ourselves to the regular case we obtain a niteness result as a corollary of the
well-known rational connectedness of smooth Fano varieties [Cam92, KMM92],
namely
Theorem 7.5.34: There is at most a nite number of deformation classes F(T

)
of regular positive Sasakian structures in any dimension.
7.6. New Sasakian Structures from Old
In this section we describe methods for obtaining new Sasakian structures from
old ones. The rst is an invariance condition for submanifolds analogous to the
situation in Kahler geometry. The second is a join construction which applies
to quasi-regular Sasakian structures. This was originally constructed for Sasaki-
Einstein structures in [BG00b] and developed further recently in [BGO07].
7.6.1. Immersions and Invariant Submanifolds. We begin with a deni-
tion about submanifolds of Sasakian manifolds.
Denition 7.6.1: Let (`
2n+1
. . . . o) be a K-contact manifold. An immersed
submanifold of ` is said to be an invariant submanifold if the two conditions
hold:
(i) is tangent to at all points of .
(ii) T
p
T
p
for all points j .
This leads to the simple but important result of Okumura [Oku68] (see also
[YK84]).
Theorem 7.6.2: An invariant submanifold of a Sasakian (K-contact) manifold
(`
2n+1
. . . . o) is Sasakian (K-contact) whose structure is given by the restric-
tion of the tensor elds (. . . o) to .
Proof. The two conditions of Denition 7.6.1 guarantee that and re-
stricted to well-dened tensor elds on . Of course, the restriction of the tensor
elds and o to are always well-dened. Hence, the tensor elds (. . . o)
satisfy all the relations on that they satisfy on `. in particular the normality
condition 6.5.9,

= 2 d. is satised. Thus, with the structure (. . . o)

is Sasakian.
Theorem 7.6.2 will be of much interest to us in Chapter 9. Many further
results about submanifolds of Sasakian manifolds can be found in [YK84] and
references therein. In view of Theorem 7.6.2 it is natural to ask whether there is a
CR or Sasakian version of the famous Kodaira Embedding Theorem 3.5.12. Such
a theorem was claimed by Ornea and Verbitsky in [OV05] as an oshoot of their
study of Vaisman immersion theorems. But they have recently shown [OV06b]
that their claim was mistaken and have provided a x. We shall not delve into
the details of Vaisman manifolds. Suce it to say that they are a special type
of locally conformally Kahler manifold and that if ` is a Sasakian manifold then
` o
1
is a Vaisman manifold. Thus, a Vaisman manifold is a kind of generalized
Hopf manifold which is actually the name that Vaisman gave to them [Vai82]. See
7.6. NEW SASAKIAN STRUCTURES FROM OLD 251
the book [DO98] of Dragomir and Ornea for details and references. Here is the
embedding theorem of [OV06b]:
Theorem 7.6.3: Let be a compact Sasakian manifold. Then there is a positive
integer and CR-embedding of into the sphere o
2N+1
such that cone C(o
2N+1
)
is biholomorphic to C
N
` 0.
It is important to note that the induced CR structure on o
2N+1
is not generally
the standard CR structure on o
2N+1
. Another related result [MY06] gives a
CR embedding of not necessarily compact Sasakian manifolds into C
N
. but not
necessarily into a sphere.
7.6.2. The Join Construction. In this section we apply a construction due
to Wang and Ziller [WZ90] to dene a multiplication on the set of quasi-regular
Sasakian orbifolds. This was done originally in [BG00b] in the case of Sasaki-
Einstein orbifolds which is perhaps of more interest, but there is an easy general-
ization to the strict Sasakian case [BGO07]. The idea is quite simple and is based
on the fact that product of Kahler orbifolds can be given a Kahler orbifold structure
in many natural ways.
Denition 7.6.4: We denote by oO the set of compact quasi-regular Sasakian
orbifolds, by o the subset of oO that are smooth manifolds, and by { o the
subset of compact, simply connected, regular Sasakian manifolds. The set oO is
topologized with the C
m,
topology, and the subsets are given the subspace topology.
The set oO is graded by dimension, that is,
(7.6.1) oO =

n=1
oO
2n+1
.
and similarly for o and {. For each pair of relatively prime positive integers
(/
1
. /
2
) we dene a graded multiplication
(7.6.2)
k1,k2
: oO
2n1+1
oO
2n2+1
oO
2(n
1
+n
2
)+1
as follows: Let
1
.
2
oO of dimension 2n
1
+1 and 2n
2
+1. respectively. Since
each orbifold (
i
. o
i
) has a quasi-regular Sasakian structure o
i
, its Reeb vector
eld generates a locally free circle action, and the quotient space by this action has
a natural orbifold structure Z
i
[Mol88]. Thus, there is a locally free action of the 2-
torus T
2
on the product orbifold
1

2
. and the quotient orbifold is the product
of the orbifolds Z
i
. (Locally free torus actions on orbifolds have been studied in
[HS91]). Now the Sasakian structure on
i
determines a Kahler structure
i
on the orbifold Z
i
, but in order to obtain an integral orbifold cohomology class
[
i
] H
2
orb
(Z
i
. Z) we need to assure that the period of a generic orbit is one. By a
result of Wadsley [Wad75] the period function on a quasi-regular Sasakian orbifold
is lower semi-continuous and constant on the dense open set of regular orbits. This
is because on a Sasakian orbifold all Reeb orbits are geodesics. Thus, by a transverse
homothety we can normalize the period function to be the constant 1 on the dense
open set of regular orbits. In this case the Kahler forms
i
dene integer orbifold
cohomology classes [
i
] H
2
orb
(Z
i
. Z). If 7
i
denotes the underlying complex space
associated with the orbifold Z
i
, one should not confuse H

orb
(Z
i
. Z) with H

(7
i
. Z).
However, they are isomorphic rationally. Now each pair of positive integers /
1
. /
2
give a Kahler form /
1

1
+ /
2

2
on the product. Furthermore, [/
1

1
+ /
2

2
]
252 7. K-CONTACT AND SASAKIAN STRUCTURES
H
2
orb
(Z
1
Z
2
. Z). and thus denes an o
1
V-bundle over the orbifold Z
1
Z
2
whose
total space is an orbifold that we denote by
1

k
1
,k
2

2
. We have
Denition 7.6.5: The orbifold
1

k
1
,k
2

2
constructed above is called the (/
1
. /
2
)-
join of (
1
. o
1
) and (
2
. o
2
).
By construction
1

k
1
,k
2

2
admits a quasi-regular Sasakian structure by
choosing a connection 1-form on
1

k1,k2

2
whose curvature is

(/
1

1
+/
2

2
).
This Sasakian structure is unique up to a gauge transformation of the form
+ d where is a smooth basic function and will be denoted by o
1

k
1
,k
2
o
2
.
This denes the maps in (7.6.2). If o
i
are quasi-regular Sasakian structures on
the compact manifolds `
i
. respectively, we shall use the notation o
1

k
1
,k
2
o
2
and
`
1

k1,k2
`
2
interchangeably depending on whether we want to emphasize the
Sasakian or manifold nature of the join. Notice also that if gcd(/
1
. /
2
) = : and
(/
t
1
. /
t
2
) = (
k
1
m
.
k
2
m
). then gcd(/
t
1
. /
t
2
) = 1 and `
1

k1,k2
`
2
(`
1

1
,k

2
`
2
)Z
m
. In
this case the cohomology class /
t
1

1
+/
t
2

2
is indivisible in H
2
orb
(Z
1
Z
2
. Z). Note
that `
1

k
1
,k
2
`
2
can be realized as the quotient (`
1
`
2
)o
1
(/
1
. /
2
). where the
o
1
action is given by the map
(7.6.3) (r. n) (c
ik2
r. c
ik1
n) .
Alternatively, we can think of the (/
1
. /
2
)-join as an operation on the set of Sasaki-
Seifert structures endowed with the quotient topology.
We are interested in restricting the map
k
1
,k
2
of (7.6.2) to the subset of smooth
Sasakian manifolds, that is in the map
(7.6.4)
k1,k2
: o
2n1+1
o
2n2+1
oO
2(n
1
+n
2
)+1
.
If `
1
and `
2
are quasi-regular Sasakian manifolds, we are interested under what
conditions the orbifold `
1

k
1
,k
2
`
2
is a smooth manifold. Recall from Denition
7.1.1 the order of a quasi-regular Sasakian manifold. We have
Proposition 7.6.6: Let (`
1
. o
1
) and (`
2
. o
2
) be compact quasi-regular Sasakian
manifolds of orders
1
and
2
. respectively. Then for each pair of relatively prime
positive integers /
1
. /
2
, the orbifold (`
1

k
1
,k
2
`
2
. o
1

k
1
,k
2
o
2
) is a smooth quasi-
regular Sasakian manifold if and only if gcd(
1
/
2
.
2
/
1
) = 1. Furthermore, if `
1
and `
2
are simply connected, so is `
1

k
1
,k
2
`
2
.
Proof. Since `
1
`
2
is a 2-torus V-bundle over Z
1
Z
2
. the o
1
V-bundle
`
1

k
1
,k
2
`
2
can be realized as a quotient (`
1
`
2
)o
1
by choosing a certain ho-
momorphism o
1
o
1
o
1
. For the V-bundle on Z
1
Z
2
with rst Chern class given
by [/
1

1
+/
2

2
] we have the action on `
1
`
2
dened by (r. n) (
k2
r.
k1
n).
The condition that `
1

k1,k2
`
2
is in o is that there are no xed points (r. n)
under the above action. If
x
.
y
denote the orders of the leaf holonomy groups
of r. n. respectively, then the condition that (r. n) be a xed point under the o
1
action generated by the Reeb vector eld is that gcd(
x
/
2
.
y
/
1
) = o 1. This con-
dition never holds if and only if gcd(
x
/
2
.
y
/
1
) = 1 for all pairs (r. n). This proves
the rst statement. The second statement follows from the long exact homotopy
sequence.
We make note of some special cases. If `
1
is regular, then `
1

k
1
,k
2
`
2
is
smooth if and only if /
2
is chosen relatively prime to the order
2
. If `
i
are both
regular Sasakian manifolds, then so is `
1

k1,k2
`
2
. The following proposition is
essentially given in [WZ90].
7.6. NEW SASAKIAN STRUCTURES FROM OLD 253
Proposition 7.6.7: Let (`
1
. o
1
) and (`
2
. o
2
) be compact quasi-regular Sasakian
manifolds and assume gcd(
1
/
2
.
2
/
1
) = 1. Then `
1

k
1
,k
2
`
2
is the associated o
1
orbibundle over Z
1
with bre `
2
Z
k2
. In particular, if `
1
is regular and /
2
= 1 then
for each positive integer /. `
1

k,1
`
2
is an `
2
-bundle over the Kahler manifold
Z
1
.
Proof. Following [WZ90] we break up the o
1
action on `
1
`
2
into stages.
First divide by the subgroup Z
k
2
of the circle group o
1
(/
1
. /
2
) dened by equation
(7.6.3) giving `
1
(`
2
Z
k
2
). Letting [n] `
2
Z
k
2
denote the equivalence class of
n `
2
. we see that the quotient group o
1
Z
k
2
acts on `
1
`
2
Z
k
2
by (r. [n])
(c
i
r. [c
i
k
1
k
2

n]) which identies `

1

k
1
,k
2
`
2
as the orbibundle over Z
1
with bre
`
2
Z
k
2
associated to the principal o
1
orbibundle
1
: `
1
Z
1
. This proves the
result.
Notice that the type of the two Sasakian manifolds is compatible with the join
operation in the following sense.
Proposition 7.6.8: Let (`
1
. o
1
) and (`
2
. o
2
) be compact quasi-regular Sasakian
manifolds. Then the join `
1

k1,k2
`
2
is positive, negative, or null if and only if
both Sasakian manifolds `
1
and `
2
are positive, negative, or null, respectively.
Proof. It follows from Theorem 7.1.3 and Proposition 7.5.23 that for any
quasi-regular Sasakian structure on a compact manifold ` we have an orbifold
submersion : `Z satisfying c
1
(T

) =

c
orb
1
(Z) as real cohomology classes.
So the sign or vanishing of c
1
(T

) and c
orb
1
(Z) coincide. Furthermore, for any pair
of integers (/
1
. /
2
) we have
c
1
(T
1

k
1
,k
2
T
2
) =

c
orb
1
(Z
1
Z
2
) =

c
orb
1
(Z
1
) +

c
orb
1
(Z
2
) = c
1
(T
1
) +c
1
(T
2
) .
where T
1

k1,k2
T
2
is the foliation dened by the structure o
1

k1,k2
o
2
on `
1

k1,k2
`
2
and T
1
. T
2
are the characteristic foliations of o
1
. o
2
, respectively. Now suppose
that the Sasakian structures o
1
and o
2
are both denite of the same type. Then
c
1
(T
1
) +c
1
(T
2
) can be represented by either a positive denite or negative denite
basic (1. 1)-form. The null case is clear.
We are interested in computing the topology of the join `
1

k
1
,k
2
`
2
. We rst
obtain some general information about the low Betti numbers of the (/
1
. /
2
)-join
of two orbifolds which follows easily from Theorem 7.4.14
Lemma 7.6.9: Let
i
oO
2ni+1
and let
1

2
be any (/
1
. /
2
)-join. Then
(i) /
1
(
1

2
) = /
1
(
1
) +/
1
(
2
).
(ii) /
2
(
1

2
) = /
2
(
1
) +/
2
(
2
) + 1 if n
i
1.
(iii) /
3
(
1

2
) = /
3
(
1
) +/
3
(
2
) if n
i
3.
(iv) /
4
(
1

2
) = /
4
(
1
)+/
4
(
2
)+/
2
(
1
)/
2
(
2
)+/
2
(
1
)+/
2
(
2
)+1
if n
i
4.
Exercise 7.14: Prove Lemma 7.6.9
When n
i
is outside the indicated range, the formulae are slightly dierent, but
are easily worked out. For general /
r
the formulae are increasingly more compli-
cated and are dierent depending on whether : is even or odd, or whether the
range conditions are satised or not. In order to determine the cohomology of
the (/
1
. /
2
)-join `
1

k1,k2
`
2
of two Sasakian manifolds one needs to analyze the
Leray-Serre spectral sequences in specic cases. Following [WZ90] we describe
254 7. K-CONTACT AND SASAKIAN STRUCTURES
the general setup. First we apply the classifying functor to an orbifold bration
o
1
`Z giving the bration `1Z1o
1
to which we can apply a
Leray-Serre spectral sequence. Next for a pair of Sasakian manifolds `
i
o,
we consider the commutative diagram of orbifold brations
(7.6.5)
`
1
`
2
1(`
1

k
1
,k
2
`
2
) 1o
1

`
1
`
2
1Z
1
1Z
2
1o
1
1o
1
.
The maps are all the obvious ones. In particular, is determined by the o
1
action
() = (
k
2
.
k
1
). Of course, if `
1

k
1
,k
2
`
2
is a smooth manifold then the 1
does not occur in the middle term of the top row. The important point in the
analysis of this diagram is that the dierentials in the spectral sequence of the top
bration are determined through naturality by the dierentials in the sequence of
the bottom bration. Wang and Ziller [WZ90] apply this method to compute the
integral cohomology ring of torus bundles over products of projective spaces.
Example 7.6.10: [The Wang-Ziller manifolds] The Wang-Ziller manifolds `
p,q
k,l
studied in [WZ90] are dened to be o
1
-bundle over CP
p
CP
q
whose rst Chern
class is / + | where and are the positive generators of H

(CP
p
. Z) and
H

(CP
q
. Z). respectively and /. | Z
+
. Thus, they are just the (/. |)-join o
2p+1

k,l
o
2q+1
= `
p,q
k,l
. For simplicity we assume that gcd(/. |) = 1. so by Proposition
7.6.6 `
p,q
k,l
is a simply connected manifold admitting natural Sasakian structures.
Moreover, since the base of the circle bundle is CP
p
CP
q
with its product complex
structure, these Sasakian structures are all positive by Proposition 7.6.8. To analyze
the topology of the manifolds `
p,q
k,l
we follow Wang and Ziller [WZ90] and consider
the free T
2
action on o
2p+1
o
2q+1
dened by (x. y) (c
i
1
x. c
i
2
y) where (x. y)
C
p+1
C
q+1
. The quotient space is CP
p
CP
q
. and `
p,q
k,l
can be identied with
the quotient of o
2p+1
o
2q+1
by the circle dened by (x. y) (c
il
x. c
ik
y).
Applying the spectral sequence analysis described above Wang and Ziller determine
the integral cohomology ring of the manifolds `
p,q
k,l
. Assuming that 1 j c they
show that
(7.6.6) H

(`
p,q
k,l
. Z) = Z[r. n]

(|r)
p+1
. r
q+1
. r
p+1
n. n
2

.
where r is a 2-dimensional class and n is a 2c + 1-dimensional class. Letting :
`
p,q
k,l
CP
p
CP
q
denote the natural bundle projection, we see that the classes .
pull back as

= |r.

= /r. Here, by abuse of notation, we let

also denote the basic classes in H

2
B
(T

). Furthermore, the basic rst Chern class is

c
1
(T

) = (j + 1)

+ (c + 1)

, so we get
c
1
(T) =

c
1
(T

) = (j + 1)

+ (c + 1)

= (|(j + 1) /(c + 1))r.

Thus, we have
(7.6.7) n
2
(`
p,q
k,l
) = (|(j + 1) +/(c + 1))r mod 2.
In certain cases one can determine the manifold completely [WZ90]. For example,
consider j = | = 1 in which case `
1,q
k,1
is an o
2q+1
-bundle over o
2
. The o
k
-bundles
over o
2
are classied by
1
(oO(/ + 1)) Z
2
[Ste51]. So there are precisely two
o
2q+1
-bundles over o
2
. and they are distinguished by n
2
. From Eequation (7.6.7)
7.6. NEW SASAKIAN STRUCTURES FROM OLD 255
we get n
2
(`
1,q
k,1
) = /(c +1)r mod 2. Thus, if / is even or c is odd we get the trivial
bundle o
2
o
2q+1
; whereas, if c is even and / is odd, we get the unique non-trivial
o
2q+1
-bundle over o
2
. This gives an innite number of distinct deformation classes
F(T

) of positive Sasakian structures on these manifolds. Furthermore, Wang and

Ziller show using the foundational work of Sullivan [Sul77] that for xed c and
| the manifolds `
1,q
k,l
the manifolds have only a nite number of dieomorphism
types. In dimension ve (c = 1) we can do somewhat better. In fact for any pair of
relatively prime positive integers (/. |) `
1,1
k,l
is dieomorphic to o
2
o
3
; whereas
later in Chapter 10 we construct a positive Sasakian structure on the non-trivial
o
3
-bundle over o
2
as well as a family of indenite Sasakian structures. Notice for
`
1,1
k,l
, c
1
(T) = 2(/ |)r. Other examples can be found in [WZ90].
Generally, given two known quasi-regular Sasakian manifolds `
1
and `
2
. it can
be quite dicult to compute the dieomorphism type or even the homeomorphism
type of `
1

k1,k2
`
2
. However, see Theorem 11.7.8 below. For further discussion
of Sasakian structures it is convenient to introduce a denition.
Denition 7.6.11: We say that (. o) oO is o-reducible if there are positive
integers (/
1
. /
2
) such that can be written as =
1

k
1
,k
2

2
for some
(
1
. o
1
). (
2
. o
2
) oO. is o-irreducible if it is not o-reducible.
It is clear from the construction that o-irreducibility corresponds to Riemann-
ian irreducibility on the space of leaves. o-reducibility rst occurs in dimension
5. Every simply connected o-reducible regular positive Sasakian 5-manifold is a
Wang-Ziller manifold `
1,1
k,l
= o
3

k,l
o
3
which is dieomorphic to o
2
o
3
for every
pair of relatively prime integers (/. |). Clearly (ii) of Lemma 7.6.9 implies
Corollary 7.6.12: Any Sasakian structure on a rational homology sphere is o-
irreducible.
CHAPTER 8
Symmetries and Sasakian Structures
In this chapter we will undertake a detailed study of the symmetries of con-
tact and Sasakian structures. Automorphisms of almost contact and Sasakian
structures have been investigated early on in series the papers of Tanno [Tan62,
Tan63a, Tan63b, Tan64, Tan69a, Tan70, Tan71] (cf. Vol. II of Sasakis
lecture notes [Sas68]). Here, in a systematic way, we begin by exploring the
relation between the isometry group and its subgroup of transformations which
also preserve the Sasakian structure. We then introduce a deformation theory for
Sasakian structures. The rst specic application of the background material of
the rst two section is the theory of compact homogeneous Sasakian manifolds.
This is a Sasakian version of a similar eort to understand homogeneous com-
pact complex and contact manifolds undertaken in the 50ties by Boothby and
Wang [Wan54, BW58, Boo80]. In fact, as we show, every compact homoge-
neous contact manifold admits a compatible homogeneous Sasakian structure (cf.
Theorem 8.3.6). The second half of this chapter is devoted to the study of contact
(Sasakian) moment maps, convexity, and the Delzant type theorems for toric con-
tact (Sasakian) manifolds. Contact/Sasakian reduction is indeed a natural analogue
of better known symplectic/Kahler quotient of Marsden-Weinstein
1
[MW74].
8.1. Automorphisms of Sasakian Structures and Isometries
Given a Riemannian manifold (`. o) we let Isom(`. o) denote the isometry
group of o. that is, Isom(`. o) is the subgroup of Diff(`) that leaves o invariant.
The well-known theorem of Myers and Steenrod (cf. [Kob72]) says that Isom(`. o)
is a nite dimensional Lie group and if ` is compact so is Isom(`. o). We let
isom(`. o) denote the Lie algebra of Isom(`. o). We now wish to consider contact
metric structures with the metric given by equation (6.4.2). In Section 6.2.2 we
discussed several important groups associated with an almost contact structure,
namely, the automorphism group Aut(. . ), the group of CR transformations
CR(`. T. J), the automorphism group Aut() of the tensor eld . as well as the
group Con(`. T) of contact transformations in the case of a contact structure. Here
we relate these groups to the isometry group Isom(`. o) in the case of a contact
metric structure (`. . . . o). First we notice that for contact structures we have
a natural string of subgroup inclusions
(8.1.1) Aut(. . ) Aut(`. ) CR(`. T. J) Con(`. T) .
1
Sir Michael Atiyah once remarked that the correct original reference regarding the sym-
plecting momentum mapping ought to be the works of Archimedes, c. 287-212 B.C. Indeed, the
height function on the 2-sphere is the symplectic moment map and quite clearly Archimedes was
investigating this function in his work on areas and volumes (cf. famous treatises On Conoids
and Spheroids and On the Sphere and the Cylinder).
257
258 8. SYMMETRIES AND SASAKIAN STRUCTURES
The relation with isometries can be seen by restricting the contact transformations
to the subgroup Con(`. ) of strict contact transformations.
Proposition 8.1.1: Let (`. . . . o) be a contact metric manifold. Then
Aut(. . ) = Con(`. ) Aut(`. ) = Con(`. ) CR(`. T. J)
= Con(`. ) Isom(`. o) .
Furthermore, Aut(. . ) is a closed Lie subgroup of Isom(`. o). hence, it is a Lie
group. If ` is compact so is Aut(. . ).
Proof. By equation (8.1.1) to prove the equalities on the rst line we need
only show that if Con(`. ) CR(`. T. J) then Aut(. . ). But this
follows easily from Lemma 6.1.29 and equation (6.2.1). Furthermore, it is clear
from equation (6.4.2) that this implies 1

o = o. This proves the containment

Aut(. . ) Con(`. ) Isom(`. o). For the converse, suppose 1

= and
1

o = o. then using equation (6.4.2) we have

d(A. Y ) = 1

(d 1l)(A. Y ) = d(1

A. 1

Y ) = d(1

A. 1

Y )
= d(1

1
1

A. 1

Y ) = 1

d(1
1

A. Y ) = d(1
1

A. Y )
which holds for all vector elds A. Y. But then 1
1

= by the non-degeneracy
of d on T. This proves the equality as sets. But since all the groups involved
are dened as subgroups of Diff(`) this is an equality as groups. This equality
characterizes Aut(. . ) as a subgroup of the Lie group Isom(`. o) that leaves
the 1-form invariant. Thus, it is a closed Lie subgroup which is compact if ` is
compact.
Denition 8.1.2: Let (`. . . . o) be a contact metric manifold. We dene
the automorphism group of (`. . . . o) to be the automorphism group of
the underlying almost contact structure and denote it by Aut(`. . . . o). We let
aut(`. . . . o) denote the Lie algebra of Aut(`. . . . o).
We know by Proposition 6.4.8 that
Proposition 8.1.3: A contact metric structure o = (. . . o) with underlying
almost CR structure (T. J) is K-contact if and only if cr(T. J) in which case
actually lies in the center of aut(. . . o).
We now describe the relationship between Aut(`. . . . o) and CR(`. T. J).
The study of the group CR(`. T. J) has a long and varied history [Web77, Lee96,
Sch95]. We state the most general result given in [Sch95]. This result holds for
a general strictly pseudoconvex CR manifold `, that is an integrable almost CR
manifold with a positive denite Levi form.
Theorem 8.1.4: Let ` be a 2n + 1 dimensional manifold with a strictly pseudo-
convex CR structure (T. J), and CR automorphism group CR(T. J). If CR(T. J)
does not act properly on `, then:
(i) If ` is non-compact, then ` is CR dieomorphic to the Heisenberg
group with its standard CR structure.
(ii) If ` is compact, then ` is CR dieomorphic to the sphere o
2n+1
with
its standard CR structure.
In particular, if ` is a closed manifold not CR dieomorphic to the sphere, the
automorphism group of its CR structure is compact.
8.1. AUTOMORPHISMS OF SASAKIAN STRUCTURES AND ISOMETRIES 259
As an immediate corollary we have
Corollary 8.1.5: Let ` be a 2n +1 dimensional manifold with a strictly pseudo-
convex CR structure (T. J) on which CR(T. J) does not act properly. Then (T. J)
is of Sasaki type.
Theorem 8.1.6: Let (T. J) be a strictly pseudoconvex CR structure of Sasaki type
on a compact manifold `. Then there exists a Sasakian structure o = (. . . o).
with underlying CR structure (T. J). whose automorphism group Aut(o) is a max-
imal compact subgroup of CR(T. J). In fact, except for the case when (`. T. J)
is CR dieomorphic to the sphere o
2n+1
with its standard CR structure, the auto-
morphisms group Aut(o) of o is equal to CR(T. J). For the standard CR structure
(T. J) on the sphere, CR(o
2n+1
. T. J) = ol(n + 1. 1).
Proof. This is accomplished by averaging over the maximal compact subgroup
of CR(T. J). See [BGS06] for details. The last statement was apparently given
rst in [Web77].
Now we focus on the characteristic foliation T

rather than the CR structure.

For a K-contact manifold we have an exact sequence
(8.1.2) 0aut(. . . o)q(. . . o)0
and we wish to study the quotient algebra q(. . . o).
Now on a K-contact manifold we have a transverse almost Hermitian structure
dened by a transverse almost complex structure

J and a compatible Hermitian
metric o
T
. Recall from Section 2.5 the Lie algebra trans(`. T

) of transverse vector
elds which are actually equivalences classes of vector elds on `. Since taking Lie
derivatives of basic geometric objects with respect to transverse vector elds is well
dened, we let aut(

J. o
T
) denote the Lie algebra of innitesimal automorphisms of
the transverse almost Hermitian structure, that is,
(8.1.3) aut(

J. o
T
) =

A trans(`. T

) [
X

J = 0.
X
o
T
= 0 .
We would like to identify the Lie algebras q(. . . o) and aut(

J. o
T
). however, as
the following shows this is not always possible. First we have
Lemma 8.1.7: Let A aut(. . . o) and let

A denote its projection onto q(. . . o).
Then

A aut(

J. o
T
).
Proof. This is easy. For example,
[

A.

J

Y ] = [

A.

JY ] = [A. Y ] = [A. Y ] =

J[

A.

Y ]
with a similar computation to show that
X
o
T
= 0.

Conversely, given an element

A aut(

J. o
T
). we would like to extend

A to a
vector eld in A aut(. . . o) by choosing as a representative of

A the section

A
of T. writing A =

A+o and demanding that
X
= 0. This implies the existence
of a smooth basic function o such that do =

A d. The obstruction for solving
this equation lies in H
1
B
(T

). In fact we have an exact sequence

(8.1.4) 0aut(. . . o)aut(

J. o
T
)

H
1
B
(T

) .
where (

A) = [

A d]
B
. We have arrived at
260 8. SYMMETRIES AND SASAKIAN STRUCTURES
Theorem 8.1.8: Let o = (. . . o) be a K-contact structure on a smooth manifold
` with characteristic foliation T

. Then a transverse vector eld

A aut(

J. o
T
)
extends to an innitesimal automorphism A aut(. . . o) if and only the basic
cohomology class [

A d]
B
H
1
B
(T

) vanishes.
Exercise 8.1: Prove that if the obstructions [

A d]
B
vanish for any transverse
Killing eld

A aut(

J. o
T
). then q(. . . o) and aut(

J. o
T
) are isomorphic as Lie
algebras.
In the quasi-regular case the vanishing of [

A d]
B
H
1
B
(T

) on ` translates
to the vanishing of the class [

A ] H
1
(Z. R) on Z. where

= d. This is just
the condition that the innitesimal automorphism

A aut(J. ) be Hamiltonian.
Thus, we can rephrase Theorem 8.1.8 as
Corollary 8.1.9: Let (Z. J. ) be an almost Kahler orbifold with polarization de-
ned by [] H
2
orb
(Z. Z). Then an innitesimal automorphism

A aut(J. ) lifts
to an innitesimal automorphism A aut(. . . o) of the induced K-contact struc-
ture on the total space ` of the circle V-bundle whose orbifold rst Chern class is
[] (cf. Theorem 7.5.2) if and only if

A is Hamiltonian.
Generally in the quasi-regular case the Lie algebra aut(

J. o
T
) is isomorphic to
the Lie algebra of innitesimal automorphisms of the almost Kahler orbifold `T

.
Exercise 8.2: Prove this statement.
Notice that the minimal possible dimension of aut(. . . o) is one. In [Tan69b]
Tanno proved that the maximal dimension of the group of automorphisms of an
almost contact metric structure is (n + 1)
2
. This maximum is realized in Example
8.1.12.
Proposition 8.1.10: Let `
2n+1
be a 2n+1-dimensional manifold with a K-contact
structure (. . . o) of rank /. Then
/ dim aut(. . . o) (n + 1)
2
.
Proof. The lower bound follows from Proposition 7.1.8. By the exact se-
quence (8.1.2), it suces to work locally on a local transversal C
n
with its almost
Hermitian structure. Now any transverse Killing vector eld at a point j C
n
is determined by A
i
p
and (
j
A
i
)
p
. From the invariance of o
T
and

J above, the
later lies in the Lie algebra u(n) of the unitary group, and the former is an arbi-
trary complex vector in C
n
. Thus, the maximal dimension is n
2
+ 2n. Adding the
characteristic vector eld on ` to this gives the result.
Next we give some examples. The rst two examples show that the maximal
dimension in Proposition 8.1.10 is realized.
Example 8.1.11: We consider the standard Sasakian structure (
0
.
0
.
0
. o
0
) on
the round sphere o
2n+1
discussed in Example 7.1.5. Since this is just the usual
Hopf bration over CP
n
with its standard Fubini-Study metric, the automorphism
group Aut(
0
.
0
.
0
. o
0
) is l(n + 1) with the central o
1
being generated by the
Reeb vector eld. Here Aut(
0
.
0
.
0
. o
0
) acts transitively on o
2n+1
, and realizes
the maximal dimension of Proposition 8.1.10.
Our next example shows that the exact sequence of Lie algebras (8.1.2) does
not always split, and that on a non-compact manifold aut(. . . o) need not be
reductive.
8.1. AUTOMORPHISMS OF SASAKIAN STRUCTURES AND ISOMETRIES 261
Example 8.1.12: This is a continuation of Example 7.3.21. One easily sees that
the Sasakian structure o = (. . . o) in 7.3.21 has a transitive group of Sasakian
automorphisms. In fact R
2n+1
is the total space of the 2n +1-dimensional Heisen-
berg group H
2n+1
dened in Exercise 6.3. The Heisenberg Lie algebra h
2n+1
is
generated by the left invariant vector elds
=

.
. o
i
=

n
i
+r
i

. 1
i
=

r
i
which satisfy the Lie bracket relations
(8.1.5) [1
i
. o
j
] =
ij
. [1
i
. ] = [o
i
. ] = 0 .
Thus, h
2n+1
is a non-trivial central extension of the 2n-dimensional Abelian Lie
algebra. Moreover, it is easy to check that h
2n+1
aut(. . . o). however, there
are many more symmetries. The automorphism group of the at Kahler structure
on R
2n
= C
n
is the semidirect product of the unitary group l(n) with the 2n-
dimensional translation group, or in terms of Lie algebras we have a non-trivial
extension
0t
2n
aut(

J. o
T
)u(n)0 .
We can lift elements of aut(

J. o
T
) to vector elds on H by using formula (6.1.3).
From all of this we get a commutative diagram of exact sequences
(8.1.6)
0 0

0 h
2n+1
aut(. . . o) u(n) 0

:
0 t
2n
aut(

J. o
T
) u(n) 0

0 0
Thus, we obtain the Lie algebra of innitesimal automorphisms aut(. . . o) of
the Sasakian manifold (H
2n+1
. . . . o) as a non-trivial extension of u(n) by the
Heisenberg algebra h
2n+1
. In dimension 3 (n = 1) this is one of Thurstons eight
model geometries [Thu97].
The following is an example where none of the innitesimal automorphisms of
(Z. . J) lift to innitesimal automorphisms of (`. . . . o).
Example 8.1.13: Continuing from Example 3.5.16, we let (\. ) be an Abelian
variety. Consider a polarization dened by the nonzero class [] H
2
(\. Z). For
simplicity we consider the principal polarization dened by the standard Kahler
structure dened by
=
i
2

j
d.
j
d .
j
with local coordinates .
j
= r
j
+in
j
. The vector elds
xj
and
yj
are in aut(J. ).
but none are Hamiltonian since
xj
is not exact, with the same for
yj
. Thus,
262 8. SYMMETRIES AND SASAKIAN STRUCTURES
the nontrivial circle bundle o
1
`T
2n
whose rst Chern class is [] admits a
Sasakian structure o = (. . . o) by Theorem 7.5.2. In this case aut(. . . o) is
the one dimensional Lie algebra generated by . This is consistent with the well-
known fact [LB92] that the automorphism group of a polarized Abelian variety
is nite. Notice that here we have cup(`) = 2n. and that c
1
(T

) = 0. In this
case the manifold ` is the nil-manifold ^
2n+1
= H
2n+1
(R)H
2n+1
(Z) which is a
homogeneous manifold for the Heisenberg group H
2n+1
(R) and a Sasakian manifold,
but it is not homogeneous Sasakian.
This example is indicative of a more general result.
Theorem 8.1.14: Let ` be a compact manifold with a K-contact structure o =
(. . . o).
(i) If c
1
(T

) < 0. then aut(. . . o) = . Hence, any K-contact structure

with c
1
(T

) < 0 is quasi-regular.
(ii) If o is a quasi-regular K-contact structure with c
1
(T

) 0 and /
1
(`) =
0. then aut(. . . o) = . Hence, when /
1
(`) = 0 any K-contact
structure with c
1
(T

) = 0 is quasi-regular.
(iii) If o is a Sasakian structure with c
1
(T

) 0. then aut(. . . o)
= . Hence, any Sasakian structure with c
1
(T

) 0 is quasi-regular.
Proof. Let A aut(. . . o) . We claim that A must vanish. Assume
not, then A and span a 2-dimensional commutative subalgebra t
2
of aut(. . . o).
First assume that o is quasi-regular and that c
1
(T

) 0. Now by the Trans-

verse Yau Theorem 7.5.19, or equivalently the Sasakian version, Theorem 7.5.20,
we can assume that the transverse Ricci curvature Ric
g
T
0. Then we have an
orbifold Riemannian submersion : `Z with almost Kahler orbifold struc-
ture (Z. /) whose Ricci curvature satises Ric
h
0. Furthermore, by Lemma
8.1.7 any innitesimal automorphism A of o projects to an innitesimal auto-
morphism of the almost Kahler structure on Z. So to prove (ii) we notice that
/
1
(Z) = /
B
1
(`) = /
1
(`) = 0. so the orbifold (Z. /) has no parallel vector elds.
But 0 c
1
(Z)
orb
implies Ric
h
0. so Bochners well-known theorem (cf. [Pet98]),
which holds equally well for compact orbifolds, implies that any Killing eld on
(Z. /) is parallel. This implies that A = 0 and proves (ii).
For (iii) we rst assume that o is quasi-regular and Sasakian. Again A projects
to a parallel vector eld

A. But now J

A is also parallel since / is Kahler. So

A = J

A / is harmonic, and thus represents a non-trivial element of H
1
(Z. R).
So by Corollary 8.1.9

A cannot be the projection of a vector eld A aut(. . . o).
This is a contradiction unless A = 0 proving (iii) under the added assumption that
o is quasi-regular. But since every quasi-regular Sasakian structure with c
1
(T

) 0
has a one dimensional automorphism group, it follows from Proposition 7.1.8 that
no irregular Sasakian structure can exist in this case. This proves (iii).
To prove (i) we rst assume that o is quasi-regular, so as above ` is an orbifold
Riemannian submersion over an almost Kahler orbifold (Z. /) with Ric
h
< 0. Thus,
by Bochner there are no Killing vector elds on (Z. /). But by Lemma 8.1.7 every
A aut(. . . o) projects to an automorphism of (Z. /) implying that A .
This proves (i) for the quasi-regular case. But then as in the proof of (iii) no
irregular K-contact structure can exist.
The maximal dimension of Isom(`
2n+1
. o) is (2n + 1)(n + 1). We want to
understand the dierence between isom(`. o) and aut(. . . o) or more globally
8.1. AUTOMORPHISMS OF SASAKIAN STRUCTURES AND ISOMETRIES 263
between Isom(`. o) and Aut(. . . o) on a K-contact or Sasakian manifold. First
we notice that any nonzero Killing vector eld in isom(`. o)aut(. . . o) cannot
commute with . Explicitly,
Lemma 8.1.15: Let o = (. . . o) be a K-contact structure, and let be a Killing
vector eld that commutes with . Then aut(. . . o). Moreover, the following
hold:
(()) = 0. (()) = 0 .
Proof. For the rst statement we need to show that

and

both
vanish. Since,

o = 0. the vector eld is ane, that is,

= 0. where is
the Levi-Civita connection. Thus,

() = (

)() +[. ] = 0 .
To show that

vanishes, we have

( o) = [. ] o +

o = 0 .
To prove the two equations we notice that
(()) = (

)() +([. ]) = 0 .
which proves the rst equation, while for the second we have
(()) = (

)() = 0 .
To understand better the dierence between isometries and automorphisms, we
develop two more lemmas the rst of which is due to Tachibana and Yu [TY70],
and the second essentially due to Tanno [Tan70].
Lemma 8.1.16: Let (. . . o) and (
t
.
t
.
t
. o) be two Sasakian structures sharing
the same Riemannian metric o. Then either o(.
t
) is constant or (`. o) is a space
form with sectional curvature equal to 1.
Proof. Put 1 = o(.
t
) and consider the vector eld = grad 1. It is easiest
to work in local coordinates (l; r
i
). Let
i
=
f
dx
i
. Then using Proposition 7.3.4 and
Theorem 7.3.16 one obtains the dierential equation

j
+ 2
i
o
kj
+
k
o
ij
+
j
o
ik
= 0.
Then a Theorem of Obata [Oba66] says that the only nonzero solution for
i
occurs
when o is a metric of constant sectional curvature equal to 1.
Our second lemma addresses the case when 1 = o(.
t
) is constant.
Lemma 8.1.17: Let o = (. . . o) and o
t
= (
t
.
t
.
t
. o) be two Sasakian struc-
tures sharing the same Riemannian metric o. and suppose that (`. o) is not a space
form with sectional curvature equal to 1. Then either
(i) o
t
= o.
(ii) o
t
= o
c
the conjugate Sasakian structure, or
(iii) o and o
t
are part of a 3-Sasakian structure (cf. Proposition 13.2.1 below).
Proof. Since (`. o) is not a space form with sectional curvature equal to 1.
Lemma 8.1.16 says that o(.
t
) is a constant, say o. By the Schwarz inequality
264 8. SYMMETRIES AND SASAKIAN STRUCTURES
[o[ 1. and o = 1 gives case (i) while o = 1 gives case (ii). Thus, we can assume
that [o[ < 1. Following Tanno [Tan70] we dene

=

t
o

1 o
2
.
and it is easy to see that
o(.

) = 0. o(

) = 1 .
Now we dene

= o(

. ). and

. So by linearity Proposition 7.3.17

implies that o

= (

. o) s a Sasakian structure orthogonal to o. For con-

venience we set
1
= .
2
=

. etc, and dene

3
=
1

2
. Then using
0 = A(o(
1
.
2
)) = o(
1
A.
2
) +o(
2
A.
1
)
together with Denition 6.2.5 we see that
1

2
=
2

1
. So from the torsion
freeness of one gets [
1
.
2
] = 2
1

2
= 2
3
. Similar relations are obtained by
cyclic permutation of the superscripts and Proposition 13.2.1 implies the result.
We now have
Theorem 8.1.18: Let (`. . . . o) be a complete Sasakian manifold that is not
isometrically covered by the round sphere, nor part of a 3-Sasakian structure. Then
either Aut(. . . o) = Isom(`. o) or Isom(`. o) is a Z
2
extension of Aut(. . . o).
In particular, if (`. . . . o) is a complete non-compact Sasakian manifold, then
Aut(. . . o) = Isom(`. o) or Isom(`. o) is a Z
2
extension of Aut(. . . o).
Proof. Assuming the hypothesis of Theorem 8.1.18, let Isom(`. o). and
consider the Sasakian structure o

= (

.
1

. o). By Lemma 8.1.17

the only possibilities are o

= o. or o

= o
c
. In the former case Aut(. . . o) =
Isom(`. o). whereas, the latter gives Isom(`. o) as a Z
2
extension of Aut(. . . o).
The last statement follows since both spheres and complete 3-Sasakian manifolds
are compact, cf. Corollary 13.2.3 below.
Corollary 8.1.19: Under the hypothesis of Theorem 8.1.18, we have
(i) isom(`. o) = aut(. . . o).
(ii) 1 dim Isom(`. o) (n + 1)
2
.
Since every complete Sasakian structure has at least a one dimensional group of
automorphisms, in the case of quasi-regular structures we can always form nontrivial
smooth quotients. Evidently we have
Proposition 8.1.20: Let (`. . . . o) be a compact quasi-regular Sasakian man-
ifold, and let G be a nite subgroup of Aut(. . . o). Then the quotient space `G
is an orbifold which inherits the Sasakian structure o = (. . . o) from `. More-
over, if G acts freely on `. then `G is a Sasakian manifold. In particular, every
compact quasi-regular Sasakian manifold admits quotients by cyclic subgroups C
m
of
the circle group generated by . Moreover, if gcd(:. (o)) = 1. the quotient `C
m
is a Sasakian manifold.
One can use this proposition to construct Sasakian structures on lens spaces
from the Sasakian structure on the weighted sphere.
Example 8.1.21: [Lens spaces] Let o
w
denote the weighted Sasakian structure
of weight vector w (R
+
)
n+1
and o
0
the standard Sasakian structure on o
2n+1
.
8.2. DEFORMATION CLASSES OF SASAKIAN STRUCTURES 265
Also let C
p
denote the cyclic subgroup of the maximal torus T
n+1
generated by the
action
(.
0
. . . . . .
n
) (.
0
.
q
1
.
1
. . . . .
q
n
.
n
) .
where is a primitive pth root of unity, and is relatively prime to the c
i
s. The
quotient of o
2n+1
by this action is a smooth manifold known as the lens space
1(j; c
1
. . . . . c
n
). Since T
n+1
Aut(o
w
). the Sasakian structures o
w
as well as o
0
all pass to the quotient manifold 1(j; c
1
. . . . . c
n
). Notice that if we choose w =
(1. c
1
. . . . . c
n
) the cyclic subgroup C
p
T
n+1
is actually a subgroup of the circle
group o
1
w
generated by the Reeb vector eld
w
. Thus, in this case the Sasakian
structure o
w
on 1(j; c
1
. . . . . c
n
) is the one obtained from Proposition 8.1.20. As
in Example 7.1.12 all the Sasakian structures o
w
belong to the same underlying
contact structure on 1(j; c
1
. . . . . c
n
).
Much is known about the topology of lens spaces especially in dimension 3
(i.e., n = 1). For example it is well-known (cf. [Bre93]) that 1(j. c) is homotopy
equivalent to 1(j. c
t
) if and only if either cc
t
or cc
t
is a quadratic residue mod j.
There is a theorem of Brody (cf. [Bre93]) that is perhaps less well-known. It says
that 1(j. c) and 1(j. c
t
) are homeomorphic if and only if either c
t
c mod j or
cc
t
1 mod j.
Notice that one can dene quotient manifolds that are locally Sasakian as
follows. Let ` be a Sasakian manifold and Isom(`. o)`Aut(. . . o). Then
by Theorem 8.1.18 is an involution. Suppose further that this involution acts
freely on `. Then ` is not Sasakian, but it has a locally Sasakian structure
in the sense that its 2-fold cover is Sasakian and the Riemannian metric is just the
Sasakian metric of ` pushed to the quotient. However, generally such manifolds
may not even be orientable. We shall give explicit examples of this phenomenon in
Chapters 9 and 14. See Examples 9.3.23 and 14.2.5 below.
8.2. Deformation Classes of Sasakian Structures
8.2.1. Deformations of Transverse Holomorphic Structures. Here we
briey sketch the deformation theory of transverse holomorphic structures due to
Duchamp and Kalka [DK79, DK80, DK84], Girbau, Haeiger, and Sundarara-
man [GHS83], and Gomez-Mont [GM80]. In [DK79, DK80] an extra assump-
tion was made that is not convenient for our purposes. So we shall mainly follow
[GHS83], but adapted to our purpose. For the basics of deformation theory we re-
fer to [Kod86] and [Kur71]. Here we want to think of ` as a Sasakian manifold of
dimension 2n+1, but we shall be more general in the beginning. We are interested
for now only in the transverse holomorphic structure. So in this subsection T will
denote a transverse holomorphic foliation on `. Let (o. 0) denote a germ at 0 of
an analytic space, that is, a space o with a local structure ring O
0
(o) O
0
(C
k
)J.
where O
0
(C
k
) denotes the algebra of germs of holomorphic on C
k
at the origin, and
J is some ideal. A smooth function o on (o. 0) is an equivalence class of smooth
functions on a neighborhood of 0 C
k
. where the two functions are equivalent if
their dierence lies in the ideal J.
A germ of a deformation of a transversely holomorphic foliation T on ` pa-
rameterized by (o. 0) is given by an open cover l

I
of ` and a family of local
submersions 1
s

: l

C
n
parameterized by (o. 0) that are holomorphic in : o
266 8. SYMMETRIES AND SASAKIAN STRUCTURES
for each r l

. and satisfy the condition

(8.2.1) 1
s

=
s

1
s

on l

.
where
s

: 1
s

(l

)1
s

(l

) is a holomorphic family of biholomorphisms.

Thus, for each : o the map
s

belongs to the pseudogroup

Autut(I)
consisting
of local automorphisms of the standard complex structure 1 on C
n
. Here we denote
the germ of the transverse holomorphic foliation by T
s
. and the original foliation
T = T
0
is dened by the local submersions 1
0

. This induces a foliation T

S
on
(o. 0) ` given by the local submersions (:. r) (:. 1
s

(r)) of (o. 0) l

to
(o. 0)C
n
. The transverse structure of this foliation is locally isomorphic to (o. 0)
C
n
. Consider two deformations of T parameterized by (o. 0). These can be given
by two families of local submersions 1
s

: l

C
n

and 1
ts

: l
t

C
n

.
Two such deformations are isomorphic if there is a smooth family /
s
: ``
of dieomorphisms parameterized by (o. 0) such that on l

l
t

there exists a
holomorphic family
s

of biholomorphisms such that

1
s

=
s

1
ts

/
s
.
Innitesimal deformations are obtained by dierentiating equation (8.2.1) with
respect to : and evaluating at : = 0. This gives rise to the Kodaira-Spencer map
which we now describe. Let

J
denote the sheaf of germs of vector elds on `
that are innitesimal automorphisms of the transversely holomorphic foliation T,
and let
J
be the quotient sheaf by the sheaf of smooth vector elds tangent to
the leaves of T. Let C
n
have its standard complex structure 1 and let
C
n denote
the sheaf of germs of vector elds preserving 1. Then
J
[
U

C
n. and upon
restricting further to l

we have
J
[
UU

C
n[
UU

. Now writing
the Haeiger cocycle
s

in the transverse complex coordinates (.

()
1
. . . . . .
()
n
), we
dene a section of the sheaf
J
[
UU

by

=
n

i=1

(
s

)
i
:
(.
()
1
. . . . . .
()
n
)

s=0

.
()
i
.
From Haeigers cocycle condition (2.2.2) we see that

satises the cocycle con-

dition
(8.2.2)

= 0
in the triple overlap l

= . and thus denes an element [

]
H
1
(`.
J
). The Kodaira-Spencer map is the linear map : T
0
oH
1
(`.
J
)
that assigns to the tangent vector (
d
ds
)
s=0
T
0
o the cohomology class [

] de-
ned by the vector eld

. More generally, one can consider the full cohomology

ring H

(`.
J
) which has the structure of graded Lie algebra induced by the
Lie bracket on
J
. The basic fact about these cohomology groups is that they
are nite dimensional for compact manifolds. This follows from the ellipticity of
the transverse structure and was proven independently in [DK79] and [GM80].
Summarizing we have
Proposition 8.2.1: Let ` be a compact manifold with a transverse holomorphic
foliation T. Then the cohomology groups H
r
(`.
J
) are nite dimensional.
The major result of any deformation theory is that the deformation space (o. 0)
be versal, that is in the case at hand, that any germ of a transverse holomorphic
8.2. DEFORMATION CLASSES OF SASAKIAN STRUCTURES 267
deformation that is close to T is isomorphic to a germ of a transverse holomorphic
foliation T
s
for some : (o. 0). Here is the result [GHS83]:
Theorem 8.2.2: Let T be a transverse holomorphic foliation on a compact man-
ifold `. and let
J
denote the sheaf of germs of transversely holomorphic vector
elds dened above. Then
(i) There is a germ (o. 0) of an analytic space parameterizing a germ of
a deformation T
s
of T such that if T
s

is any germ of a deformation

parameterizing T by the germ (o
t
. 0), there is a holomorphic map :
(o
t
. 0)(o. 0) so that the deformation T
(s)
is isomorphic to T
s

.
(ii) The Kodaira-Spencer map : T
0
oH
1
(`.
J
) is an isomorphism.
(iii) There is an open neighborhood l H
1
(`.
J
) and a holomorphic map
: lH
2
(`.
J
) such that (o. 0) is the germ at 0 of
1
(0). The
2-jet of satises ,
2
(n) =
1
2
[n. n].
The element [n. n] H
2
(`.
J
) is the primary obstruction to performing the
deformation, and (i) of Theorem 8.2.2 says that the analytic space (o. 0) is versal.
It is known as the Kuranishi space of the deformation. Theorem 8.2.2 has the
following important corollaries.
Corollary 8.2.3: If H
1
(`.
J
) = 0. then any innitesimal deformation of T is
trivial, that is, every T
s
close enough to T is isomorphic to T.
Corollary 8.2.4: If H
2
(`.
J
) = 0. then the Kuranishi space of deformations of
T is isomorphic to an open neighborhood of 0 in H
1
(`.
J
).
Notice that (ii) and (iii) of the Theorem 8.2.2 imply that if H
2
(`.
J
) does
not vanish then the Kuranishi space o is singular.
8.2.2. Deformations of Transverse Holomorphic Sasakian Structures.
We now apply the ideas of the previous subsection to the case where ` is a compact
Sasakian manifold. First, we have the evident
Proposition 8.2.5: The characteristic foliation T

of a Sasakian structure o is a
transverse holomorphic foliation.
Since by Theorem 7.1.10 any Sasakian structure on ` can be approximated by
quasi-regular Sasakian structures, we restrict our discussion of deformation theory
to the latter. Here we denote by
J

the sheaf of transversely holomorphic vector

elds on `. and by

the sheaf of germs of holomorphic vector elds on the

complex orbifold Z.
Proposition 8.2.6: Let o be a quasi-regular Sasakian structure on a compact man-
ifold ` with characteristic foliation T

. and let : `Z denote the quotient

projection. Then there is an exact sequence
0H
1
(Z.

)H
1
(`.
J

)H
0
(Z.

)H
2
(Z.

) .
Proof. The Leray sheaf H

(.
J

) of the map is the sheaf associated to

the presheaf l H

(
1
(l).
J

1
(U)
). By Lerays Theorem there is a spectral
sequence for which
1
p,q
2
= H
p
(Z. H
q
(.
J

)) = H
p+q
(`.
J

) .
268 8. SYMMETRIES AND SASAKIAN STRUCTURES
Furthermore, the stalk H
q
(.
J

)
y
at a point n Z is isomorphic to
H
q
(
1
(n).
J

1
(y)
) H
q
(o
1
. (

)
y
)

)
y
. if c = 0. 1;
0. otherwise.
Thus we have
1
p,q
2
=

H
p
(Z.

). if c = 0. 1;
0. otherwise.
So the spectral sequence collapses at the third term and the exact sequence follows
as for the Gysin sequence.
There are two immediate corollaries to this proposition.
Corollary 8.2.7: If H
i
(Z.

) = 0 for i = 1. 2. then H
1
(`.
J

) H
0
(Z.

).
Corollary 8.2.8: If H
0
(Z.

) = 0. then H
1
(`.
J

) H
1
(Z.

).
These are the two extreme cases for which one can think of a deformation
of a transverse holomorphic structure as arising form either a deformation of the
complex structure on Z or a deformation of the vector eld . The rst corollary
applies when there are no nontrivial deformations of the complex structure on Z.
in which case the nontrivial deformations of the transverse holomorphic structure
comes from deformations of the vector eld which arise from the innitesimal
complex automorphisms of Z. An example of this type is ` = o
2n+1
with its
standard Sasakian structure. In this case the base space Z = CP
n
which has no
deformation space. So H
1
(o
2n+1
.
J

) H
0
(CP
n
.

). Deformations of this type

include the weighted spheres of Example 7.1.12 which correspond to elements lying
in the maximal torus of H
0
(CP
n
.

).
On the other hand, the second corollary says that if there are no innitesimal
complex automorphisms, then the all deformations of the transverse holomorphic
structure comes from the deformations of the complex structure on Z. This is the
case in Example 8.1.13 for the nil-manifold ^
2n+1
. where the base space Z is a
polarized smooth Abelian variety
n
of complex dimension n. In this case we have
H
1
(^
2n+1
.
J

) H
1
(
n
.

).
In order to treat deformations through Sasakian structures more constraints
are needed. This should be contrasted with the Kahlerian case where the well-
known Kodaira-Spencer stability theorem [KS60] implies that locally innitesimal
complex deformations of Kahlerian structures remain Kahlerian. The point is that
transversely holomorphic deformations which deform a Riemannian foliation need
not remain Riemannian. Thus, we wish to deform through Riemannian, or more
precisely Kahlerian foliations in which case the Haeiger cocyles
s

are taken from

the subpseudogroup
Aut(I,g)

Aut(I)
of isometries of some locally dened Kahler
metric o on C
n
. Thus, innitesimally the corresponding vector elds on ` are also
local Killing vector elds for the transverse metric o
T
. where o
T
= 1

o for all .
We denote the subsheaf of
J

of such vector elds by

J

,a
. Now let T

be the
characteristic foliation of a quasi-regular Sasakian structure o = (. . . o) on `.
Then by Theorem 7.5.1 there is an orbifold submersion : `Z onto a compact
Kahler orbifold Z whose Kahler metric / satises o
T
=

/. So we let
,a
denote
the subsheaf of

that leaves invariant the Kahler metric / on Z. Then exactly

as in Proposition 8.2.6 we have
8.2. DEFORMATION CLASSES OF SASAKIAN STRUCTURES 269
Proposition 8.2.9: Let o be a quasi-regular Sasakian structure on a compact man-
ifold ` with characteristic foliation T

. and let : `Z denote the quotient

projection. Then there is an exact sequence:
0H
1
(Z.
,a
)H
1
(`.
J

,a
)H
0
(Z.
,a
)H
2
(Z.
,a
).
We mention that in the case of certain deformations through Hermitian folia-
tions, Duchamp and Kalka [DK80]have given a splitting theorem.
8.2.3. Deformations of the Characteristic Foliation. We consider de-
formations of the characteristic foliation T

with a xed transverse holomorphic

structure, that is innitesimal deformations such that the Kodaira-Spencer map
lies in H
1
(`.
J

,a
)H
0
(Z.
,a
). Rather than studying the Kodaira-Spencer
map, we proceed directly.
This type of deformation was rst considered by Takahashi [Tak78] who con-
sidered deformations by adding an innitesimal automorphism aut(. . . o).
However, here we consider somewhat more general deformations rst introduced
by Gauduchon and Ornea [GO98] in the 3-dimensional case, and then used by
Belgun in his classication of 3-dimensional Sasakian structures [Bel00, Bel01].
Explicitly, we consider deformations which preserve the underlying contact struc-
ture as well as the underlying almost CR structure, but deform the contact 1-form
by sending 1 for some positive function 1. We then write
(8.2.3) = 1.

= + .
where

is the Reeb vector eld of . The conditions
(

) = 1.

d = 0
then imply the relation
(8.2.4) 1 =
1
(

)
=
1
1 +()
.
It is clear that is a contact 1-form as long as (

) 0. and that the contact

subbundle T = ker = ker remains unchanged. The deformations also preserve
the CR structure, that is,

[
1
= [
1
. which extends to an endomorphism by setting

. Moreover, we can easily check that

= 0

2
= 1l +


hold. Thus, (

. .

) is an underlying almost contact structure of the contact struc-
ture. Next we dene a Riemannian metric o by
o = d (

1l) .
This metric is compatible with the contact structure, so (

. .

. o) is a contact
metric structure. But according to Corollary 6.5.11 the structure (

. .

. o) is
normal if and only if the almost CR structure

[
1
is integrable and

is invariant
under

. The former condition holds since the CR structure hasnt changed. So if
we begin with a Sasakian structure a deformation through K-contact structures is
automatically a deformation through Sasakian structures.
Denition 8.2.10: Let (T. J) be a strictly pseudoconvex almost CR structure of
Sasaki type on a manifold `. We denote by F(T. J) the set of all K-contact struc-
tures having (T. J) as its underlying CR structure. A deformation within F(T. J)
270 8. SYMMETRIES AND SASAKIAN STRUCTURES
given by a transformation of the form of equation (8.2.3) is called a deformation
of type I.
We now investigate the properties of the space F(T. J). We shall assume that
(T. J) is of Sasaki type so that F(T. J) is nonempty. We have immediately from
Denition 6.4.9 and Proposition 6.4.8 that
Proposition 8.2.11: A contact metric structure o = (. . . o) is in F(T. J) if
and only if cr(T. J).
Next we identify F(T. J) with a convex subset of cr(T. J). Let us x a K-contact
structure o
0
= (
0
.
0
.
0
. o
0
) F(T. J). and let cr
+
(T. J) be the subset of cr(T. J)
such that
0
() 0. There is a natural map
(8.2.5) : F(T. J)cr
+
(T. J)
dened by (o) = . and one easily sees that this map is injective. We claim that
it is also surjective. Let cr
+
(T. J) and dene by
=

0

0
()
.
Clearly () = 1. and since cr
+
(T. J), it leaves T invariant. This implies that
d =

= 0. So is the Reeb vector eld of . We then dene =

0

0
.
and o = d ( 1l) + . So o = (. . . o) F(T. J) by Proposition 8.2.11.
Hereafter, we shall identify F(T. J) with cr
+
(T. J) without further ado.
Proposition 8.2.12: The subset cr
+
(T. J) is an open convex cone in cr(T. J) that
is invariant under the adjoint action of CR(T. J).
Proof. That cr
+
(T. J) is an open subset of cr(T. J) follows immediately from
its denition. Convexity is clear since if
1
.
2
cr
+
(T. J), then
0
((1t)
1
+t
2
) =
(1 t)
0
(
1
) + t
0
(
2
) 0 for all 0 t 1. The conic property is also clear. For
groups of transformations the adjoint action is that induced by the dierential. So
for CR(T. J). we have
0
(

) = (

0
)() = 1

0
() 0 for some positive
function 1

. implying that cr
+
(T. J) is invariant.
The tangent space T
S
(cr
+
(T. J)) at a point o cr
+
(T. J) is identied with the
Lie algebra cr(T. J). Let us now assume that ` is compact and that the almost CR
structure is integrable, so by Theorem 8.1.6 the Lie algebra cr(T. J) is reductive.
We have [BGS06]
Theorem 8.2.13: Let ` be a compact manifold of dimension 2n + 1 with a
CR structure (T. J) of Sasaki type. Then the Lie algebra cr(T. J) decomposes as
cr(T. J) = t
k
+p. where t
k
is the Lie algebra of a maximal torus T
k
of dimension /
with 1 / n + 1 and p is a completely reducible T
k
module. Furthermore, every
A cr
+
(T. J) is conjugate to a positive element in the Lie algebra t
k
.
Proof. First assume that (T. J) is not the standard CR structure on the
sphere. Then according to Proposition 8.1.6 there is a Sasakian structure o
0

o(T. J) such that CR(T. J) = Aut(o
0
) which is a compact Lie group. Then by a
well-known result of Lie theory every element in the Lie algebra aut(o
0
) is conjugate
under the adjoint action of the group Aut(o
0
) to one on t
k
. Moreover, by Proposition
8.2.12 positivity is preserved under this action. The range for the dimension of the
maximal torus follows by a well-known result of Sasakian geometry.
8.2. DEFORMATION CLASSES OF SASAKIAN STRUCTURES 271
In the case that (T. J) is the standard CR structure on the sphere, we know
[Web77] that CR(T. J) = ol(n + 1. 1). in which case there are several maximal
Abelian subalgebras of cr(T. J) = su(n+1. 1). However, it can be seen from a case
by case analysis that the only Abelian subgroup where the positivity condition can
be satised is in the case of a maximal torus. This can be ascertained, for example,
by looking at Theorem 6 of [Dav05].
We now dene t
+
k
= t
+
k
(T. J) = t
k
cr
+
(T. J). As with cr
+
(T. J) this is a
convex open cone in the Lie algebra t
k
. and it can be viewed as the moduli space of
Sasakian structures that are compatible with the CR structure (T. J). The space
t
+
k
(T. J) was introduced and studied in [BGS06]. It is called the Sasaki cone.
Explicitly,
Corollary 8.2.14: The following identications hold:
t
+
k
= t
+
k
(T. J) = cr
+
(T. J)CR(T. J) = F(T. J)CR(T. J) .
Let us consider the case / = 1 which is somewhat special. Since the Reeb
vector eld is central / = 1 implies that dim aut(o) = dim cr(T. J) = 1. In this
case F(T. J) = cr
+
(T. J) = t
+
1
= R
+
and F(T. J) consists of a 1-parameter family
of Sasakian structures given by o
a
= (
a
.
a
.
a
. o
a
). where o 0 and

a
= o
1
.
a
= o .
a
= . o
a
= oo + (o
2
o) .
So the deformation is a transverse homothety. By Theorem 8.1.14 this case arises
when c
1
(T

) 0. that is:
Proposition 8.2.15: Let o = (. . . o) be a Sasakian structure with c
1
(T

) 0.
Then the Sasaki cone has dimension one.
From another point of view one can formulate the condition of deforming
through Sasakian structure as a condition on the function 1. This was done in
[GO98] in terms of the Hessian. We dene the Hessian H
f
(A. Y ) of a function 1
by H
f
(A. Y ) = (
X
d1)(Y ).
Proposition 8.2.16: The deformed contact metric structure (

. .

. o) is Sasakian
if and only if
H
f
1(A. Y ) = H
f
1(A. Y )
for all sections A. Y of T.
The proof of Proposition 8.2.16 is left as a series of exercises the rst of which
is to prove the following
Lemma 8.2.17: Let (. . . o) be a contact metric structure on `. and consider
a deformation of the contact 1-form dened by equations (8.2.3) and (8.2.4). Then
the vector eld takes the form
=

1 1
1

1
2
grad 1
1
.
Exercise 8.3: Prove Lemma 8.2.17.
Exercise 8.4: Prove that
o(
X
Q. Y ) =
1
2
H
f
1(A. Y ) +
1
2
d1
1
()o(A. Y ) .
where 2Q = grad 1
1
. and use this to prove Proposition 8.2.16.
272 8. SYMMETRIES AND SASAKIAN STRUCTURES
Exercise 8.5: Show that if is a Killing vector eld such that = / for some
function /, then / must be constant.
An example of this type of deformation where is a Killing vector eld is the
deformation away from the standard sphere to the weighted sphere as described
in Example 7.1.12. These are obtained by taking to lie on a maximal torus
in the automorphism group of the standard Sasakian round sphere. More gener-
ally, such deformations have an important application, namely in the proof of the
Sasakian analogue of Kodairas conjecture mentioned earlier. The Sasakian case,
however, manifests itself somewhat dierently than the complex case in that type
I deformations do not change the underlying CR structure. This is a special case
of Rukimbiras Theorem 7.1.10, namely
Theorem 8.2.18: On a compact manifold every Sasakian structure o = (. . . o)
is a deformation of a quasi-regular (algebraic) Sasakian structure.
8.3. Homogeneous Sasakian Manifolds
Recall from Denition 1.6.11 that a homogeneous manifold is a smooth man-
ifold ` on which there is a smooth transitive action of a Lie group G. Given a
homogeneous manifold ` one is often interested in homogeneous structures, that
is, geometric structures that are invariant under the action of G. For example, a
homogeneous Riemannian structure is a homogeneous manifold ` together with a
G-invariant Riemannian metric. This can be phrased by saying that the Lie group
Isom(`. o) of isometries of (`. o) acts transitively on `. However, the group of
automorphisms of a G-structure is not generally a Lie group. When it is, the G-
structure is said to be of nite type. Contact and symplectic structures are not of
nite type. So we have
Denition 8.3.1: A contact structure T on a manifold ` is a homogeneous
contact structure if there is a Lie subgroup G Con(`. T) that acts transitively
on `. Similarly, a strict contact structure dened by a global 1-form is homoge-
neous is if there is a Lie subgroup G Con(`. ) that acts transitively on `. In
this case we refer to the triple (`. . G) as a homogeneous contact manifold.
Since this book mainly describes certain contact metric structures whose group
of automorphisms is a nite dimensional Lie group by we focus our attention on
nite dimensional Lie subgroups of Con(`. ). However, before embarking on the
description of the automorphism group of K-contact structures, we discuss briey
a more general result of Lichnerowicz [Lic63b] that we make of use of later. Recall
that a Lie algebra g is reductive if its maximal solvable ideal (radical) coincides with
its center, alternatively, the derived Lie algebra [g. g] is semi-simple. Then in the
compact case, we have the following important theorem of Lichnerowicz [Lic63b]:
Theorem 8.3.2: Let (`. ) be a compact strict contact manifold, and let g be a
nite dimensional subalgebra of con(`. ). Then g is reductive.
This theorem will be important in Section 8.3 where we prove a general state-
ment about compact homogeneous contact manifolds due to Boothby and Wang
[BW58]. In the case of a K-contact structure much of the structure of its auto-
morphism group Aut(. . . o) can be deduced directly from the well-known Myers-
Steenrod Theorem.
8.3. HOMOGENEOUS SASAKIAN MANIFOLDS 273
Corollary 8.3.3: Let ` be a smooth manifold with a contact metric structure
o = (. . . o). Then its automorphism group Aut(. . . o) is a closed nite di-
mensional Lie subgroup of the isometry group Isom(`. o). Furthermore, if ` is
compact the Lie algebra of innitesimal automorphisms aut(. . . o) is reductive
with non-trivial center. In this case the corresponding Lie group Aut(. . . o) is
up to covering isomorphic to G T. where G is a compact semi-simple Lie group
and T is a torus.
Without loss of generality we assume that the Lie group G is connected. A
particular case of interest to us concerns the addition of a compatible Riemannian
metric in which case we do obtain a G-structure of nite type. Hence, we make the
denition below. However, as we shall show the fundamental results of Boothby and
Wang [BW58, Boo80] go a long way toward showing that they are all equivalent.
Denition 8.3.4: A metric contact structure (. o) on a manifold ` is said to be
a homogeneous metric contact structure if the group Aut(`. . o) acts transi-
tively on `. Similarly, a K-contact structure or Sasakian structure o = (. . . o)
on a manifold ` is said to be a homogeneous K-contact or Sasakian struc-
ture if the group Aut(o) acts transitively on `.
We have the following important theorem of Boothby and Wang [BW58].
Theorem 8.3.5: Let (`. . G) be a homogeneous contact manifold. Then the con-
tact form is regular. Moreover, ` = G1 is a ber bundle over GH
0
1 with
ber H
0
11. where H
0
is the connected component of a one dimensional Lie group
H. and H
0
11 is either dieomorphic to o
1
or to R.
Proof. Standard Lie theory together with the foundational work of Palais
[Pal57] applied to the ow of the Reeb vector eld proves regularity. Since the
bres are one dimensional, the result follows.
We mention that in [Ale90] homogeneous contact manifolds are related to co-
adjoint orbits. Let us recall the following well-known terminology. Let G
c
be a
complex semi-simple Lie group. A maximal solvable complex subgroup 1 is called
a Borel subgroup, and 1 is unique up to conjugacy. Any complex subgroup 1 that
contains 1 is called a parabolic subgroup. Then the homogeneous space G
c
1 is
called a generalized ag manifold. A well-known result of Wang [Wan54] (see also
[Akh90]) says that every simply-connected compact homogeneous Kahler manifold
is a generalized ag manifold G
c
1. Thus, we can easily construct homogeneous
Sasakian manifolds by considering circle bundles over generalized ag manifolds.
An additional result of Borel and Remmert [BR62] says that a compact homoge-
neous Kahler manifold A is the product of its Albaneese (A) and a generalized
ag manifold; however, the next theorem shows that the Albaneese factor cannot
contribute to compact homogeneous contact manifolds. We now have the main
result of homogeneous contact manifolds.
Theorem 8.3.6: Let (`. . G) be a compact homogeneous contact manifold. Then
(i) ` admits a homogeneous Sasakian structure with contact form .
(ii) ` is a non-trivial circle bundle over a generalized ag manifold.
(iii) ` has nite fundamental group, and the universal cover

` of ` is
compact with a homogeneous Sasakian structure.
274 8. SYMMETRIES AND SASAKIAN STRUCTURES
Proof. We sketch the proof. Let G be a Lie group acting transitively on
` preserving the contact form . Lichnerowiczs Theorem 8.3.2 says that the Lie
algebra g of G is reductive. Furthermore, since the action is transitive a corollary
[Lic63b] of Theorem 8.3.2 says that g is either semi-simple or has a one dimensional
center. But we know that the group of contactomorphisms of ` has the one
dimensional subgroup generated by the Reeb vector eld in its center. So we can
assume that G is simply connected with a one dimensional center 7. We write ` =
G1 for some closed subgroup 1 of G. Then the Boothby-Wang Theorem 8.3.5
implies that ` = G1 is a circle bundle over the compact homogeneous symplectic
manifold 1 = GH. The group G does not act eectively on 1 since H contains the
center 7. But 1 = GH = (G7)(H7). so writing G = G
1
7 we see that G
1
acts
transitively on 1 with isotropy subgroup H
1
= HG
1
. Now G
1
is simply connected
and semi-simple, and acts eectively up to a discrete normal subgroup. Thus, a
theorem of Borel [Bor60] implies that G
1
is compact and 1 is simply connected.
So G
1
is compact semi-simple, then according to Borel [Bor54] and Lichnerowicz
[Lic53] H is a centralizer of a torus, and GH is homogeneous Kahler
2
. But the
Kahler form denes a positive integral cohomology class [BW58], so 1 = GH
is projective algebraic by the Kodaira embedding theorem. So lifting the Kahler
structure to ` gives a homogeneous Sasakian structure with the same contact form
. This proves (i).
By a result of Wang [Wan54] 1 = GH is a generalized ag manifold, that is
it can be written in the form 1 = G
c
1. where G
c
is a complex semi-simple Lie
group and 1 a parabolic subgroup. (In fact, G is a maximal compact subgroup of
G
c
.) Since the Kahler form on 1 transgresses to d on `. the circle bundle is
non-trivial, proving (ii). The rst part of (iii) now follows since the last statement
implies that /
1
(`) = 0. Furthermore, the group of the deck transformations of
the cover j :

`` is a cyclic subgroup of the central o
1
. so the homogeneous
Sasakian structure on ` lifts to

`.
A weaker version of this theorem was given originally by Boothby and Wang
[BW58] under the more restrictive assumption that ` = GH is not only com-
pact, but also simply connected. However, somewhat later Boothby [Boo80] (in a
memorial paper dedicated to the memory of H.C. Wang) was able to remove sim-
ply connectedness from the hypothesis (see also [DMR82]). In Boothbys paper
[Boo80] he pulls back the contact form to a 1-form on G. and argues that the
semi-simple factor G
1
of G must act transitively. (Actually he incorrectly concludes
that the central element does not occur).
Theorem 8.3.6 says that every homogeneous contact structure on a compact
manifold is homogeneous Sasakian, and that every homogeneous Sasakian structure
on a compact manifold can be realized as a circle bundle over a generalized ag
manifold. However, we do not know whether the homogeneous Sasakian structure
is determined uniquely by the homogeneous contact 1-form. Borel and Hirzebruch
[BH58] give an example of two inequivalent homogeneous complex structures on
the ag manifold 1 = l(4)(l(2)l(1)l(1)). If both of these complex structures
were compatible with the same Kahler form . one could construct two inequivalent
2
There seems to be some folklore around this issue. Boothby and Wang [BW58] attribute
the result to Lichnerowicz citing mimeograph notes from Princeton.
8.3. HOMOGENEOUS SASAKIAN MANIFOLDS 275
homogeneous Sasakian structures with the same contact form on the circle bundle
over 1 determined by [].
Since a compact simply connected Lie group is semi-simple, we have
Corollary 8.3.7: Let (`. . G) be a compact homogeneous contact manifold, then
G can be taken to be a compact semi-simple Lie group.
Remark 8.3.1: Notice that if a K-contact or Sasakian structure o = (. . . o) is
homogeneous (under the Lie group G), then o
t
obtained from o by a transverse
homothety 7.3.10 is also a homogeneous K-contact or Sasakian structure. More
generally, if o
t
F(T

) and =
t
is a G invariant basic 1-form, then o
t
is also
homogeneous.
Example 8.3.8: The Wang-Ziller manifolds `
p,q
k,l
described in Example 7.6.10 are
examples of homogeneous Sasakian manifolds, since the base manifold CP
p
CP
q
is
a generalized ag manifold G
c
1. Furthermore, they are the (/. |)-join of the homo-
geneous Sasakian manifolds o
2p+1
and o
2p+1
with G
c
= o(G1(j. C) G1(c. C)).
This example exhibits some more general properties of the join construction
applied to homogeneous Sasakian manifolds. First we note that (/
1
. /
2
)-join of
two simply connected homogeneous Sasakian manifolds is also a simply connected
homogeneous Sasakian manifold as long as /
1
and /
2
are relatively prime, which
we assume. Thus, the set of simply connected homogeneous Sasakian structures H
forms a submonoid Moreover, a well-known theorem of Borel [Bor54] states that a
simply connected homogeneous Kahler manifold corresponding to a complex semi-
simple Lie group G is Riemannian irreducible if and only if G is simple. Thus, we
have
Theorem 8.3.9: The subset H of compact simply connected homogeneous Sasakian
manifolds is closed under the (/
1
. /
2
)-join operation (with gcd(/
1
. /
2
) = 1). Fur-
thermore, a simply connected homogeneous Sasakian manifold corresponding to a
semi-simple Lie group G is o-irreducible if and only if G is simple.
In Example 8.1.12 we have studied the Heisenberg group H
2n+1
as an example
of a Lie group that admits not only a contact structure, but a Sasakian structure.
In fact, it is easy to see that H
2n+1
admits a homogeneous Sasakian structure.
However, there are very few semi-simple Lie groups that admit a homogeneous
contact structure as the following theorem Boothby and Wang [BW58] shows.
Theorem 8.3.10: Let (` = G. . G) be a homogeneous contact manifold with G
semi-simple, then G is locally isomorphic to either oO(3) or o1(2. R).
Of course, the locally isomorphic compact Lie groups ol(2) and oO(3) admit
homogeneous Sasakian structures, but according to Corollary 8.3.7 and Theorem
8.3.10 these are the only compact Lie groups that admit a homogeneous contact
structure.
Example 8.3.11: The real Heisenberg group H
2n+1
= H
2n+1
(R) is a nilpotent Lie
group, and as mentioned above it carries a homogeneous Sasakian structure. H
2n+1
276 8. SYMMETRIES AND SASAKIAN STRUCTURES
can be realized in terms of certain n + 2 by n + 2 nilpotent matrices, namely
(8.3.1)

1 r
1
r
n
.
0 1 0 n
1
.
.
.
.
.
.

.
.
.
0 0 1 n
n
0 0 0 1

In terms of these matrices the group multiplication on H

2n+1
dened in Exercise 6.3
is matrix multiplication from the right. As discussed previously in Example 7.3.21
the contact 1-form of equation (6.1.1) is invariant under this action, and has a
compatible Sasakian structure with constant -sectional curvature 3. Now con-
sider the nilmanifold N
2n+1
= H
2n+1
(R)H
2n+1
(Z) of left cosets, where H
2n+1
(Z)
is the subgroup of matrices of the form of equation (8.3.1) with integer entrees.
This is the nontrivial circle bundle over an Abelian variety discussed in Example
3.5.16. The contact structure T dened by is invariant under the right action,
and denes a contact structure, also denoted by T. on the quotient N
2n+1
. In fact
the entire Sasakian structure on H
2n+1
passes to the quotient. However, T is not
invariant under the group action from the left. Thus, N
2n+1
is a homogeneous
manifold and a contact (actually Sasakian) manifold, but it is not a homogeneous
contact manifold.
8.4. Symmetry Reduction and Moment Maps
In this section we discuss moment maps and symmetry reduction in various
contexts. We begin with the well-known case of symplectic reduction. Since this
has been treated extensively in many books and review articles (cf. [AM78, GS84,
MS98, CdS01, CdS03, OR04]), we only give a brief description. Historically
Souriaus book [Sou70] is very important. Apparently it is where the term moment
map had its origin. Furthermore, it has recently been translated into English
[Sou97].
8.4.1. A Brief Review of the Moment Map and Symplectic Reduc-
tion. Let (`. ) be a symplectic manifold and let G be a Lie group acting eec-
tively as symplectomorphisms of `. i.e., there is a group monomorphism / : G
S(`. ). We call such an action a symplectic action. Recall from Example 1.4.12
the denition of a Hamiltonian vector eld.
Denition 8.4.1: A symplectic action / : G S(`. ) is said to be Hamilton-
ian if for each element g the associated vector eld A

is Hamiltonian, i.e.,
A

= dH for some function H C

(`).
So the obstruction to being Hamiltonian lies in the de Rham cohomology
group H
1
(`. R). In particular, if this group vanishes then every vector eld A
corresponding to an element in the Lie algebra g is Hamiltonian. The denition of
a Hamiltonian action given by Denition 8.4.1 is that of [LM87], but it is not stan-
dard; for example this is called weakly Hamiltonian in [MS98]. We let ham(`. )
denote the Lie algebra of Hamiltonian vector elds on `. Then a Hamiltonian
action / induces a monomorphism of Lie algebras, viz. /

: gham(`. ). As
mentioned in Example 1.4.12 the Hamiltonian function is dened only up to a
constant, so we have an exact sequence of Lie algebras
0RC

(`)ham(`. )0 .
8.4. SYMMETRY REDUCTION AND MOMENT MAPS 277
Recall that on a symplectic manifold C

(`) has a natural Lie algebra structure

given by the Poisson bracket . which is dened by
(8.4.1) 1. o = (A
f
. A
g
) .
where A
f
denotes the Hamiltonian vector eld dened by the smooth function 1. So
a natural question is: when does the Lie algebra homomorphism/

: gham(`. )
lift to a Lie algebra homomorphism : gC

(`)? This is determined by the

vanishing of a certain cocycle in the Lie algebra cohomology group H
2
(g. R). and
there is always a choice of constants that makes this cocycle vanish. From now
on we assume that the action of G on ` is Hamiltonian and this choice has been
made. This stronger notion is what is called a Hamiltonian action in [MS98] and
it always gives rise to an equivariant moment map or momentum map j : `g

dened by
(8.4.2) d'j. ` = A

.
where g and A

= /

. Furthermore, the map j is Ad

-equivariant in the
sense that
j(/
g
(r)) = Ad

g
j(r) .
The moment map allows one to construct new symplectic manifolds from old ones
by the procedure of symmetry reduction. The main result is the famous Marsden-
Weinstein-Meyer reduction theorem [Mey73, MW74]. The version given here is
from [Ler03a].
Theorem 8.4.2: Let (`. ) be a symplectic manifold with a Hamiltonian action
of a Lie group G. and let j : `g

denote the corresponding moment map.

Suppose further that g

is a regular value of j. and that the action of the

isotropy subgroup G

G is proper on j
1
(). Then G

acts locally freely on

j
1
(). and the quotient `

= j
1
()G

is naturally a symplectic orbifold. If

in addition the action of G

is free on j
1
(). then the quotient `

is a smooth
symplectic manifold. Furthermore, if (`. ) has a compatible Kahler structure and
G acts as Kahler automorphisms, then the quotient has a natural Kahler structure.
For a proof of this and related results see [GS84]. There has been much work
on adapting the reduction procedure to various settings, cf. [OR04]. The case
of most interest is when G is Abelian. In this case the isotropy subgroup G

for
any g

is the whole group G. and any choice of constant gives an equivariant

moment map. Moreover, we can employ the so-called shift trick. If j : `g

is
a moment map for an Abelian group G. and g

is a regular value, then 0 is a

regular value of the shifted moment map j
t
= j .
If is not a regular value of j the situation is more complicated. Neverthe-
less, it was shown by Sjamaar and Lerman [SL91] that in this general case the
reduced space is a stratied symplectic space satisfying Goretsky-MacPherson ax-
ioms [GM88]. More precisely,
Theorem 8.4.3: Let (`. ) be a Hamiltonian G-manifold with the moment map
j : `g

. Let `
(H)
denote the stratum consisting of orbits of type H < G. Then

H
= j
1
(0) `
(H)
is a manifold and the orbit space

`
H
=
j
1
(0) `
(H)
G
278 8. SYMMETRIES AND SASAKIAN STRUCTURES
has a natural symplectic structure
H
whose pullback to
H
coincides with the
restriction of to
H
`. Consequently, the stratication of ` by orbit types
induces the decomposition of the reduced space

` = j
1
(0)G into a disjoint union
of symplectic manifolds

` =

H<G

`
H
.
An important tool in understanding the relation between the symplectic (Kahler)
and the algebraic quotient is the norm of the moment map 1 = [[j[[
2
dened rel-
ative to any G-invariant inner product on the Lie algebra g. Kirwan proved that
the following [Kir84]
Theorem 8.4.4: Let G be a compact Lie group and (`. ) a G-Hamiltonian space
with the moment map j : `g

. For any G-invariant inner product on g the

function 1 = [[j[[
2
is an equivariantly perfect Morse function over Q.
Suppose (`. ) is a compact G-Hamiltonian space such that the 0 g

is
a regular value of the moment map j : `g

. Equivalently, the compact Lie

group G acts on j
1
(0) only with nite isotropy points and the quotient (

`. ) is
a compact symplectic orbifold. We have a natural isomorphism
(8.4.3) H

(

`. Q) H

G
(j
1
(0). Q) .
where H

G
(j
1
(0). Q) is the G-equivariant cohomology of the zero level set. The
inclusion j
1
(0) ` induces a map
(8.4.4) : H

G
(`. Q)H

G
(j
1
(0). Q) H

(

`. Q)
called the Kirwan map. Under the above assumption we have [Kir84]
Theorem 8.4.5: The Kirwan map : H

G
(`. Q)H

(

`. Q) is surjective.
Kirwans surjectivity is the key tool in computing the rational cohomology of
the reduced space. For example, it can be eectively used to determine H

(

`. Q)
when G is a torus in which case the kernel of the Kirwan map can be computed
[Gol02]. Tolman and Weitsman give an explicit formula for the cohomology ring
of the symplectic reduction

` in terms of the cohomology ring of ` and xed
point data. Under certain conditions, their formula also holds for the integral
cohomology ring, and can be used to show that the cohomology of the reduced
space is torsion-free [TW03].
Recall that a complex Lie group 1 is reductive if and only if 1 = G
C
is the
complexication of any maximal compact subgroup G. Consider any Kahler man-
ifold (`. 1. . o) with a holomorphic action of 1 = G
C
. where G Aut(`. 1. . o)
and ` is G-Hamiltonian. Let j : `g

be the associated moment map.

Denition 8.4.6: The point r ` is analytically semi-stable if the closure of
G
C
-orbit through r intersects j
1
(0). The set of semi-stable points is denoted by
`
ss
j
1
(0). The point r ` is called analytically stable if the closure of
G
C
-orbit through r intersects j
1
(0) at a point where dj is surjective.
Note that the notion of analytic (semi-)stability depends on the Kahler metric
and the momentum map. Sjamaar proves [Sja95]
Theorem 8.4.7: The inclusion j
1
(0) `
ss
induces a homeomorphism

` = j
1
(0)G `
ss
G
C
.
as a map of stratied spaces.
8.4. SYMMETRY REDUCTION AND MOMENT MAPS 279
The connection between the two quotients in the above theorem is established
via the Morse function 1 = [[j[[
2
. The set of semi-stable points `
ss
can be seen as
the minimum Morse startum `
min
. i.e., set of all points whose paths of steepest
descent under 1 have limit points in j
1
(0) [Kir84, Sja95].
Theorem 8.4.8: Let (`. J. . o) be a connected Kahler manifold and let
G Aut(`. J. . o) a compact Lie group. Let complexication G
C
act holomorphi-
cally on ` and let ` be a Hamiltonian G-space with the moment map j : `g

.
Then the set `
ss
coincides with the minimum Morse stratum `
min
and induced
by j
1
(0) `
min
map

` = j
1
(0)G `
min
G
C
is a homeomorphism of stratied Kahler spaces.
Example 8.4.9: One important example is when ` is a projective algebraic va-
riety. Suppose 1 acts on a nonsingular complex projective variety ` CP
n
via
a representation : 1G1(n + 1. C). We can choose coordinates so that re-
stricts to the unitary representation of G l(n + 1) acting on `. Suppose `
is G-Hamiltonian and let j : `g

be the associated moment map. Mumfords

geometric invariant theory associates to such an action a projective quotient variety
denoted by `1. The quotient is obtained by restricting the so-called semi-stable
points `
ss
`. Mumfords notion of (semi-)stability is purely algebraic. How-
ever, one can show that in the integral case the analytic stability is the same as
the algebraic stability so that we get the connection between the geometric and the
algebraic quotients [MFK94, Sja95].
Theorem 8.4.10: Let ` CP
n
and let 1 = G
C
be a reductive complex Lie
group acting on ` via a representation : 1G1(n + 1. C). Suppose ` is a
G-Hamiltonian space. Then the notions of analytic and algebraic semi-stability
coincide. Hence, the inclusion j
1
(0) `
ss
induces a homeomorphism

` =
j
1
(0)G `1. as a map of stratied spaces.
We end this brief discussion with a remark that many of these results still hold
in the more general case when ` and G are not compact as long as the moment
map j is assumed to be proper.
Next we give the well-known convexity theorem of Atiyah, Guillemin, and
Sternberg [Ati82, GS82a].
Theorem 8.4.11: Let (`. ) be a compact 2n-dimensional symplectic manifold
with an eective Hamiltonian action of a torus T of dimension /. Let j : `t

be the corresponding Ad

-equivariant moment map. Then the level sets of j are

connected and the image j(`) is a convex polytope in t

.
By denition a convex polytope is the convex hull of a nite set of points j
i
.
called vertices, in t

. It can be represented as
j(`) =

i=1
:
i
j
i
[

i
:
i
= 1. :
i
0

.
The proof of Theorem 8.4.11 uses Bott-Morse theory and is given for example in
[MS98]. The inverse image under j of the vertices is precisely the xed point set
of the T-action. More generally, the inverse image of the /-dimensional faces is
the subset xed by an (n /)-dimensional subtorus of T. We also mention that
codimension one faces are called facets.
280 8. SYMMETRIES AND SASAKIAN STRUCTURES
We are particularly interested in the case / = n. We have
Denition 8.4.12: A toric symplectic manifold is a quadruple (`. . /. j).
where ` is a 2n dimensional manifold, is a symplectic 2-form on `. / is an
eective Hamiltonian action of an n-torus T with moment map j : `t

.
Notice that choosing a particular component H of the moment map gives in
this case what is often called a completely integrable Hamiltonian system. A re-
markable theorem of Delzant [Del88] says that compact toric symplectic manifolds
are completely described by the image of their corresponding moment maps.
Denition 8.4.13: A Delzant polytope R
n
is a polytope satisfying the
following three conditions:
(i) There are n edges meeting at each vertex (simplicity).
(ii) If j is a vertex then each edge meeting at j has the form j + tn
i
. where
t 0 and n
i
Z
n
(rationality).
(iii) For each vertex the corresponding n
1
. . . . . n
n
can be chosen as a Z-basis
for Z
n
(smoothness).
Now Delzants theorem is:
Theorem 8.4.14: There is a one-to-one correspondence between compact toric
symplectic manifolds and Delzant polytopes.
The proof of Delzants theorem has an important corollary, viz.
Corollary 8.4.15: A compact toric symplectic manifold admits a compatible Kahler
structure.
There is an orbifold version of all of this due to Lerman and Tolman [LT97]
which we now describe. First we see that by using the Denition 4.3.17 of a Lie
group action on an orbifold that Denition 8.4.12 easily generalizes to orbifolds.
Denition 8.4.16: A toric symplectic orbifold is a quadruple (A. . /. j).
where A is a 2n dimensional orbifold, is a symplectic 2-form on A. / is an
eective Hamiltonian action of an n-torus T with moment map j : At

.
The Convexity Theorem 8.4.11 holds for symplectic orbifolds with an eective
torus action. However, one obtains an orbifold Delzant polytope given by De-
nition 8.4.13 with the smoothness condition (iii) dropped. In addition to classify
orbifolds one needs to assign labels to the facets of the polytope. We have
Denition 8.4.17: A Lerman-Tolman polytope or lt-polytope for short is a
convex rational simple polytope t

with dim = dim t

and a positive integer

attached to each open facet.
Let A = (A. |) be a symplectic orbifold with a closed non-degenerate 2-form
. Since the local uniformizing groups preserve . the dimensions of the orbifold
strata dier by an even integer. In particular, the orbifold singular set
orb
(A)
has codimension at least two. The inverse image j
1
(1) may or may not lie in

orb
(A). but if it j
1
(1)
orb
(A) = then j
1
(1)
orb
(A). This follows from
Theorem 6.4 of [LT97]. Moreover, if 1 is an open facet with label :
F
Z
+
. the
inverse image j
1
(1) is xed by an o
1
and lies in the stratum of A dened by
Z
m
F
as a local uniformizing group. So for any open facet 1 with label :
F
= 1 the
inverse image j
1
(1) lies in the regular stratum of the orbifold. Thus, we use the
convention that a facet with label 1 is equivalent to no label. We see that facets
with :
F
1 correspond to orbifold branch divisors, cf. Denition 4.4.8. Generally,
8.4. SYMMETRY REDUCTION AND MOMENT MAPS 281
the local uniformizing group at a point r A can be described as follows: let T(r)
denote the set of open facets whose closure contains j(r). and let
F
denote the
primitive outward normal to 1 T(r). Let T
x
denote the isotropy subgroup of
T at r. and let / t
x
denote its integral lattice. The set :
F

F
generates a
sublattice

/ of /; the local uniformizing group at r is then isomorphic to /

/.
We now have the Delzant-Lerman-Tolman theorem:
Theorem 8.4.18: There is a one-to-one correspondence between compact toric
symplectic orbifolds and lt-polytopes. Moreover, the toric symplectic orbifold is a
symplectic reduction of C
n+1
by a closed Abelian subgroup of l(n + 1).
As before the proof of convexity uses Bott-Morse theory, while the construction
of a toric symplectic orbifold from an lt-polytope is similar to Delzants original
proof, see Theorem 8.1 in [LT97]. As in the smooth manifold case we have
Corollary 8.4.19: A compact toric symplectic orbifold A admits a compatible in-
variant complex structure, hence, a Kahler structure. In fact its underlying complex
space A is a projective algebraic variety.
Next we give an example similar to Example 6.8 in [LT97].
Example 8.4.20: Weighted projective space CP(2. 3. 6) with its standard Kahler
structure. This is a quotient of o
5
C
3
by the weighted o
1
w
-action with weights
w = (2. 3. 6). or equivalently the weighted C

-quotient of C
3
` 0. It can also
be constructed by symplectic reduction of the standard symplectic structure on
C
3
. However, in this case 0 is not a regular value of the unshifted moment map.
We take the shifted moment map for this circle action to be j
1
(z) = 2[.
0
[
2
+
3[.
1
[
2
+ 6[.
2
[
2
6. where z = (.
0
. .
1
. .
2
) are the standard coordinates for C
3
. The
choice of the shift vector 6 is a matter of convenience. Generally, for any weighted
projective space CP(w). a convenient choice is lcm(n
0
. . . . . n
n
). Now we have the
identication CP(2. 3. 6) = j
1
1
(0)o
1
w
. The circle o
1
w
can be viewed as the kernel
of the map

: T
3
T
2
induced by the matrix

3 0 1
0 2 1

viewed as a map : Z
3
Z
2
.
We want to compute the T
2
-moment map j
2
: CP(2. 3. 6)t

2
. This can be
viewed as a quotient moment map. The full T
3
-moment map j
3
: C
3
t

3
is
j
3
(z) = ([.
0
[
2
. [.
1
[
2
. [.
2
[
2
). Now the restriction j
3
[

1
1
(0)
descends to a map j
3
:
CP(2. 3. 6)t

3
. To nd j
2
we compose this with a map t

3
t

2
obtained by
identifying t

2
with the hyperplane dened by y
T
y + (0. 0. 1). The image
j
3
(CP(2. 3. 6)) lying in this hyperplane is the polytope dened by
(8.4.5) j
3
(CP(2. 3. 6)) = (r
0
. r
1
. r
2
) t

3
[ r
i
0. 2r
0
+ 3r
1
+ 6r
2
= 6 .
To obtain the corresponding lt-polytope in t

2
corresponding to this polytope, we
notice that the vertices map as follows: (0. 0. 1) (0. 0). (0. 2. 0) (0. 1). and
(3. 0. 0) (1. 0). Thus, lt-polytope is the standard simplex in t

2
= R
2
; however,
there are labels 2 on the horizonal line from (0. 0) to (1. 0). and 3 on the vertical line
from (0. 0) to (0. 1). The diagonal line from (0. 1) to (1. 0) has a 1 or equivalently
no label.
282 8. SYMMETRIES AND SASAKIAN STRUCTURES
Recall from Proposition 4.5.10 that as algebraic varieties there are isomor-
phisms
CP(2. 3. 6) CP(1. 3. 3) CP(2. 1. 2) CP(1. 1. 1) = CP
2
.
All four of these orbifolds correspond to the same simplex, namely the standard
one. They are distinguished as orbifolds by their labels. Of course, the standard
simplex with no labels corresponds to CP
2
. The simplex of the orbifold CP(2. 1. 2)
has horizontal and vertical labels, 2 and 1, respectively; while the simplex of the
orbifold CP(1. 3. 3) has horizontal and vertical labels, 1 and 3, respectively.
Exercise 8.6: Consider the toric symplectic orbifold CP
2
with the symplectic
form descending from the standard Fubini-Study symplectic form on CP
2
. and the
action of Z
2
Z
3
Z
6
on CP
2
is given by [.
0
. .
1
. .
2
] [
0
.
0
.
1
.
1
.
2
.
2
] with

2
0
= 1.
3
1
= 1.
6
2
= 1. Show that the lt-polytope for this toric symplectic orbifold
is the standard simplex as in Example 8.4.20 but now with label 6 on each face.
While polytopes are the combinatorial data involved in symplectic geometry,
the combinatorial data of algebraic geometry are known as fans. We refer to the
books [Oda88] and [Ful93] and reference therein for details. Example 8.4.20 illus-
trates a more general phenomenon.
Proposition 8.4.21: Let A
1
= (A
1
. |
1
.
1
) and A
1
= (A
2
. |
2
.
2
) be compact
toric symplectic orbifolds.
(i) Then their lt-polytopes
1
and
2
are isomorphic as labelled polytopes if
and only if A
1
and A
2
are isomorphic as toric symplectic orbifolds.
(ii)
1
and
2
are isomorphic polytopes, but with possibly dierent labels,
if and only if A
1
and A
2
are isomorphic as toric projective algebraic
varieties.
Proof. Part (i) is proven in [LT97] to which we refer the reader. To prove (ii)
we recall that a convex rational polytope determines a fan, and the fan completely
determines [Oda88] the underlying space as a toric algebraic variety. So isomorphic
polytopes give isomorphic toric algebraic varieties and vice-versa.
It follows from the Morse theory on orbifolds developed in [LT97] that the odd
dimensional cohomology groups on toric symplectic orbifolds vanish. See Proposi-
tion III.47 [Ler03a] of for a proof.
Proposition 8.4.22: Let (Z. ) be a compact toric symplectic orbifold with un-
derlying topological space 7. Then H
2k+1
(7. R) = 0.
8.4.2. The Contact Moment Map and Convexity. In the modern liter-
ature the relation between contact structures and symplectic cones seems to have
rst appeared in Arnolds book [Arn78]. This connection was also used in the
context of group representation theory in [GS82b].
Let (`. ) be a strict contact manifold of dimension 2n + 1 with Reeb vector
eld . and C(`) its symplectication. On C(`) we let S(C(`). ) denote the
group of symplectomorphisms of (C(`). ). and S
0
(C(`). ) the subgroup of
S(C(`). ) that commutes with homotheties, i.e., the automorphism group of
the symplectic Liouville structure. The corresponding Lie algebras are denoted by
s(C(`). ) and s
0
(C(`). ). respectively. Given a vector eld A s
0
(C(`). ).
the fact that A commutes with the Liouville vector eld implies that A = A
M
.
8.4. SYMMETRY REDUCTION AND MOMENT MAPS 283
where A
M
is a vector eld on ` lying in con(`. ). One easily sees that ( cf.
[LM87] pg. 314)
Proposition 8.4.23: There is an isomorphism S
0
(C(`). ) Con(`. ) of topo-
logical groups induced by the natural inclusion `C(`). Innitesimally, there
are Lie algebra isomorphisms s
0
(C(`). ) con(`. ) C

(`)

induced by the
maps A A
M
(A
M
). Furthermore, is in the center of con(`. ).
We now consider the moment map construction for C(`) and `. Let G be a Lie
group acting eectively on the symplectic cone (C(`). ) which leaves invariant the
symplectic form and which commutes with the homothety group. In particular,
G gives rise to a group homomorphism / : GS
0
(C(`). ). and since the action
is eective /(G) is a subgroup of S
0
(C(`). ). Recall from equation (6.5.5) that
on the symplectic cone C(`). the 2-form is exact. Hence, every action / :
GS
0
(C(`). ) is Hamiltonian. As usual from this we obtain a moment map
j : C(`)g

.
and as before we shall assume that it is equivariant. Explicitly j is dened by
d' j. ` = A

.
where A

denotes the vector eld on C(`) corresponding to g. For simplicity

we denote the function ' j. ` by j

. As in Section 6.5.3 we take the symplectic

2-form to be homogeneous of degree 2 and write = d = d(:
2
). Here is the
contact 1-form on `. where we identify ` with the subset ` 1 C(`). An
easy computation then shows that up to an additive constant
(8.4.6) j

= (A

).
With this choice we see that under homotheties : c
t
: we have
(8.4.7) c
2t
. c
2t
. j c
2t
j.
We now specialize to the case when the Lie group G is a torus T with Lie
algebra t. In this case the kernel of the exponential map exp : tT forms a lattice
called the integral lattice of T and denoted by Z
T
. This lattice represents the set of
circle subgroups of T. There is a convexity theorem for symplectic cones [dMT97],
but rst we have
Denition 8.4.24: A subset ( t

is called a polyhedral cone if it can be

represented by
( =

n t

[ 'n.
i
` 0
for some nite set of vectors
i
. It is rational if in addition all the vectors
i
lie
in the integral lattice Z
T
.
The vectors
i
are the inward normal vectors of the polyhedral cone. Without
loss of generality we can assume that each
i
is primitive in the sense that :
i
Z
T
for 0 < : < 1. and that the set
i
is minimal, that is, for any index ,. ( =

i,=j
n t

[ 'n.
i
` 0.
Theorem 8.4.25: Let (C(`). ) be a symplectic cone with ` compact, and let
/(T) be a torus subgroup of S
0
(C(`). ). Assume further that there is an element
t such that ' j. ` 0. Then the image j(C(`)) is a convex polyhedral cone.
The positivity condition ' j. ` 0 is quite strong. For example, it is easy to
see that it implies that 0 is not in the image of the moment map j.
284 8. SYMMETRIES AND SASAKIAN STRUCTURES
We wish to relate the symplectic moment map on C(`) to a contact moment
map on `. This is easy to do since G commutes with homotheties, and we get
an induced action of G on `. By Proposition 8.4.23 we can identify G with a Lie
subgroup of Con(`. ) which by abuse of notation we continue to denote by G.
This gives a moment map by restriction, viz.
j

: `g

.
(8.4.8) j

= j[
M1
. j

= (A

) .
Exercise 8.7: Show that the contact moment map j

of equation (8.4.8) can be

dened intrinsically without reference to the symplectication C(`).
Notice that if 1 is a G-invariant function on `. and
t
= 1 is another contact 1-
form in the same contact structure, then we get a dierent symplectic form
t
which
is again G-invariant, and this gives rise to a dierent moment maps, namely j
t
= 1 j.
and j
t

= 1j

. It is convenient to have a moment map that only depends on the

underlying contact structure and not on the choice of 1-form. In order to dene such
a moment map, we follow Lerman [Ler02a, Ler03a] and consider the annihilator
T
o
of T in T

`. This is just the contact line bundle L dened by equation (6.1.5),

which in the current case is trivial. Choosing a contact 1-form (nowhere vanishing
section of T
o
) trivializes the bundle T
o
` R. as well as chooses an orientation
of T
o
. so we can split T
o
` 0 as T
o
+
T
o

in which case sections of T

o
+
are
identied with 1 with 1 0. Now the idea is to represent the cone C(`) by
T
o
+
`R
+
. Consider the group Con(`. T) of contactomorphisms of (`. T). By
pullback Con(`. T) lifts to an action on T

` which leaves T
o
invariant. Con(`. T)
has a subgroup Con(`. T)
+
dened to be the subset of elements in Con(`. T) that
leave T
o
+
invariant. Notice that for any choice of contact 1-form there is a sequence
of subgroups Con(`. ) Con(`. T)
+
Con(`. T).
Denition 8.4.26: Let G be a Lie group acting eectively as contactomorphisms
that preserves the co-orientation, that is /(G) Con(`. T)
+
. We dene a moment
map : T
o
+
g

by
'(r. ). ` = '. A

(r)`.
If the action of G preserves a contact 1-form of the contact structure (i.e.,
/(G) Con(`. )), then we have a commutative diagram
(8.4.9) T
o
+

or in symbols = j

. An observation of Lerman [Ler02a] says that for proper

group actions the condition that G preserves a contact 1-form is not a restriction.
Lemma 8.4.27: Let G be a Lie group acting properly and eectively on a co-
oriented contact manifold ` such that /(G) Con(`. T)
+
Then there exists a
G-invariant section of T
o
+
, that is a G-invariant contact form such that T = ker .
8.4. SYMMETRY REDUCTION AND MOMENT MAPS 285
Proof. Choose a contact 1-form
t
. If G is compact we dene an invariant
contact form by averaging, viz.
=
1
Vol(G)

G
o

t
do .
If G is non-compact we can use a slice theorem [Pal61] as done in Proposition 2.5.2
of [DK00], or in [Loo01].
Thus, for proper eective actions we can assume that /(G) Con(`. ) for
some contact 1-form . In this case the identication of T
o
+
with the cone C(`)
holds as G-manifolds, and the moment map coincides with j. The advantage of
working with over j

is obvious; the latter depends on the choice of 1-form .

whereas, the former depends only on the contact structure T up to a choice of co-
orientation. Nevertheless, since this book mainly deals with Sasakian or K-contact
geometry where a 1-form is xed working with the moment map j

will often be
sucient.
However, when working with G-invariant contact structures, it is convenient to
dene the moment cone
(8.4.10) C(j

) = t g

[ j

(`). t [0. )
which is an invariant of the contact structure with its G action. Clearly, we have
C(j

) = (T
o
+
) 0. and C(j

) is independent of the choice of 1-form in T

o
+
.
Accordingly, we denote C(j

) by C().
Denition 8.4.28: We say that a Lie group action / : GCon(`. T)
+
is of
Reeb type if there are a contact 1-form of the contact structure T and an
element g such that A

is the Reeb vector eld of .

Notice that a vector g such that A

is the Reeb vector eld satises

'. ` = 1. We call the characteristic vector of g. and the equation '. ` = 1 the
characteristic hyperplane in g

. (This hyperplane is called the Reeb hyperplane in

[BM93a, BM96]). Now choosing the standard Cartesian coordinates (:
0
. . . . . :
n
)
on g

and expanding the characteristic vector in the standard basis c

i

i
for g. this
hyperplane can be written as
(8.4.11)
n

i=0
:
i
n
i
= 1 .
When the coecients n
i
are integers it will often be convenient to rescale the
coordinates :
i
so that the characteristic hyperplane takes the form
(8.4.12)
n

i=0
:
i
n
i
= lcm(n
0
. . . . . n
n
) .
Then the hyperplane arising in Example 8.4.20 is just the characteristic hyperplane
of the associated contact structure. Concerning actions of Reeb type we have
Proposition 8.4.29: The action / : GCon(`. T)
+
is of Reeb type if and only
if there is an element in the center z of g such that '. ` 0.
Proof. The only if part is clear. So if g satises '. ` 0. we have
(A

) 0 for any section : `T

o
+
. We then dene the contact form

t
=
1
(A

)
.
286 8. SYMMETRIES AND SASAKIAN STRUCTURES
Since is in the center of g, the 1-form
t
is G-invariant. But then A

is the Reeb
vector eld of
t
.
Proper actions of Reeb type have a particularly nice property, namely
Proposition 8.4.30: Let (`. T) be a co-oriented contact manifold, and let / :
GCon(`. T)
+
be a proper eective action of Reeb type of a Lie group G. Then
there exists a G-invariant K-contact structure o = (. . . o) belonging to the con-
tact structure dened by T.
Proof. By Lemma 8.4.27 there exists a G invariant contact form with T =
ker . Choose a compatible Riemannian metric. Then as in the proof of Proposition
2.5.2 of [DK00] we obtain a G-invariant Riemannian metric o compatible with the
contact structure. Then the Reeb vector eld of is an innitesimal isometry
of o. Moreover, is dened to be the endomorphism obtained from d using the
metric o. The structure (. . . o) is then K-contact.
Proposition 8.4.30 together with Theorem 7.1.10 have an important corollary.
Corollary 8.4.31: Let (`. T) be a co-oriented compact contact manifold with an
eective /-torus action / : T
k
Con(`. T)
+
. Suppose also that this action is of
Reeb type. Then there is a compatible quasi-regular T
k
-invariant K-contact struc-
ture o = (. . . o) so that ` is an orbifold bration : `Z over a compact
symplectic orbifold (Z. ) with

= d. Furthermore, the quotient To

1
of T
by the o
1
generated by the Reeb vector eld acts eectively on Z preserving the
symplectic structure.
Actually the base orbifold Z has a T
k1
invariant almost Kahler structure.
Indeed we are mainly concerned with the case when G is a torus T in which case
one has a convexity theorem for certain contact manifolds. It was rst proved
by Banyaga and Molino [BM93a, BM96] in the case of compact toric contact
manifolds, that is when the torus has maximal dimension. It was then generalized
by Lerman [Ler02b] for any torus of dimension bigger than two under the condition
that the image of the moment map does not contain the origin. Before describing
this we have
Denition 8.4.32: A toric contact manifold is a quadruple (`. T. T. ). where
(`. T) is an oriented and co-oriented contact manifold of dimension 2n + 1 with
an eective action / : TCon(`. T)
+
of a (n + 1)-torus T and moment map
: T
+
0
t

n+1
.
Lerman [Ler02a] classied all compact toric contact manifolds. In particular,
he shows that most of these are toric contact manifolds of Reeb type. The ones
that are not of Reeb type are completely described and are of less interest to us
in this book. We state Lermans theorem below without proof. Another result in
[Ler02a] which is obvious for toric contact manifolds of Reeb type is:
Proposition 8.4.33: For toric contact manifolds 0 is not in the image of the
moment map .
The convexity theorem is:
Theorem 8.4.34: Let (`. T) be a compact co-oriented contact manifold of dimen-
sion 2n + 1 and let / : TCon(`. T)
+
be an eective action of a torus on `
with moment map . Assume also that dim T 2. and that 0 is not in the image
8.4. SYMMETRY REDUCTION AND MOMENT MAPS 287
of . Then the bres of are connected and the moment cone C() is a convex
rational polyhedral cone.
When the action / is of Reeb type a slightly weaker convexity theorem follows
from Theorem 8.4.25 and Proposition 8.4.29. Since / of Reeb type implies that 0 is
not in the image of the moment map, the convexity result is subsumed by Theorem
8.4.34 at least in the case that dim T 2. We now choose a contact 1-form . i.e.,
a smooth section of T
o
+
. together with its Reeb vector eld. This determines the
characteristic hyperplane '. ` = 1 which intersects the polyhedral cone C() in
a polytope. By applying Corollary 8.4.31 and using the Lerman-Tolman Theorem
8.4.18 it is easy to see [BG00a] that this polytope is rational if and only if the
characteristic vector lies in the integral lattice Z
T
. Summarizing we have
Theorem 8.4.35: Let / : TCon(`. T)
+
be an eective action of a torus of
Reeb type. Then the moment cone C() t

is a convex rational polyhedral cone.

Furthermore, a choice of T-invariant contact form in T
o
+
together with its Reeb
vector eld A

intersects C() t

with the characteristic hyperplane '. ` =

1 giving the image j

(`) as a convex polytope in the characteristic hyperplane.

Moreover, the polytope is rational if and only if the characteristic vector lies in
the integral lattice Z
T
.
Actually the convex rational polyhedral cones of Theorem 8.4.35 must obey
other conditions. The following denition is due to Lerman [Ler02a].
Denition 8.4.36: A rational polyhedral cone C is said to be good if it satises
the following condition: if
k

j=1

n t

[ 'n.
i
j
` 0

is a /-face of C then the

ij
are independent over Z and the equality

j=1
o
j

i
j
[ .
j
R

Z
T
=

j=1
:
j

i
j
[ :
j
Z

holds.
Identifying t with its double dual t

. this condition says that the annihilator

of a / face of C is precisely the Lie algebra of a subtorus H of T.
We are now ready for Lermans classication theorem [Ler02a].
Theorem 8.4.37: Let (`. T. T. ) be a compact toric contact manifold. Then
(i) If dim(`) = 3 and the action of T
2
is free, then ` is dieomorphic to
the 3-torus T
3
. and there is a section of T
+
0
having the form cos nt d
1
+
sin nt d
2
for some positive integer n.
(ii) If dim(`) = 3 and the action of T
2
is not free, then ` is dieomorphic
to a lens space or o
1
o
2
and (`. T. T
2
. ) is classied by two rational
numbers :. c with 0 : < 1. : < c.
(iii) If dim(`) = 2n+1 3 and the action of T
n+1
is free, then ` is a prin-
cipal T
n+1
-bundle over o
n
. Moreover, each principal T
n+1
-bundle over
o
n
has a unique T
n+1
-invariant contact structure making it a contact
toric manifold.
(iv) If dim(`) = 2n + 1 3 and the action of T
n+1
is not free, then the
moment cone C() is a good rational polyhedral cone. Conversely, if
288 8. SYMMETRIES AND SASAKIAN STRUCTURES
C t

is a good rational polyhedral cone, there is a unique toric contact

manifold (`. T. T
n+1
. ) whose moment cone C() is C.
We are mainly interested in case (iv) of this theorem. We have
Proposition 8.4.38: Let (`. T. T. ) be a compact toric contact manifold with
dim(`) = 2n + 1 3. Then the action of T on ` is of Reeb type if and only if it
is not free and the moment cone C() contains no nonzero linear subspace.
Proof. Suppose that the action of T
n+1
is of Reeb type, then by Proposition
8.4.30 there is a compatible K-contact structure, and by perturbing the Reeb vector
eld if necessary it is quasi-regular. Thus, the space of leaves of the characteristic
foliation is a compact symplectic orbifold A. Moreover, the quotient n-torus T
n
=
T
n+1
o
1
. where o
1
is the subgroup of T
n+1
generated by the Reeb vector eld,
acts on the orbifold A preserving its symplectic structure. Such an action is never
free if n 1 [LT97].
Conversely suppose n 1 and the action of T
n+1
is not free, then by (iv) of
Lermans Theorem 8.4.37 C() is a good rational polyhedral cone containing no
nonzero linear subspace. This implies that there is an element t such that
'. ` 0. Thus, by Proposition 8.4.29 the action is of Reeb type.
There is a counterexample to this proposition when dim(`) = 3. namely o
1

o
2
. where the action is neither free nor of Reeb type. We give some examples of
toric contact manifolds of Reeb type.
Example 8.4.39: We consider the standard sphere o
2n+1
with the standard contact
structure 7.1.5. This is the contact structure induced from the standard symplectic
structure on R
2n+2
. In this case the standard contact form
0
and its Reeb vector
eld
0
are those given in Example 7.1.5. The maximal torus T
n+1
is generated
by the vector elds H
i
of that example, and this provides the sphere o
2n+1
with
a toric contact structure. Let c
i

n
i=0
denote the standard basis for t
n+1
= R
n+1
corresponding to the vector elds H
i
. that is, A
e
i
= H
i
. and let c

i
denote the
dual basis of t

n+1
. From equation (8.4.8) the moment map becomes
j
0
(x. y) =
n

i=0

0
(H
i
)c

i
=
n

i=0
[.
i
[
2
c

i
.
So the moment cone C() = C(j

0
) is the union of 0 with the positive 2
n+1
-
tant. If we let (q. p) denote the bre coordinates of T

` corresponding to (x. y)
respectively, we see that
(x. y; q. p) =
n

i=0
(c
i
n
i
j
i
r
i
)c

i
.
Letting :
0
. . . . . :
n
denote the coordinates for t

n+1
we see that the characteristic
hyperplane is just
:
0
+ +:
n
= 1 .
Thus, the image j

0
(o
2n+1
) is just the standard n-simplex in the characteristic
hyperplane with :
i
s as barycentric coordinates. We want to identify this with a
Delzant polytope. For simplicity we put n = 2 and consider the exact sequence
0t
1

t
3

t
2
0 .
8.4. SYMMETRY REDUCTION AND MOMENT MAPS 289
where t
1
is the one dimensional Lie algebra dened by the characteristic vector .
The map c can be represented by the matrix
=

1 0 1
0 1 1

.
The characteristic hyperplane is then represented as
T
y + (0. 0. 1). Thus we
see that the vertices (0. 0. 1). (1. 0. 0). (0. 1. 0) are mapped by c to the vertices
(0. 0). (1. 0). (0. 1). respectively, giving the Delzant polytope as the standard sim-
plex in t

2
= R
2
. This corresponds to the toric symplectic manifold (CP
2
. ). where
is the standard symplectic form given by the Fubini-Study metric.
Choosing a weighted contact form
w
and equivalently a weighted Reeb vector
eld
w
as in Example 7.1.12, determines a dierent characteristic hyperplane,
namely that of equation (8.4.12). The toric CR contact structure is represented
by the moment cone C(). whereas, choosing a particular Reeb vector eld in the
Sasaki cone t
+
n+1
cuts C() by the corresponding characteristic hyperplane.
Example 8.4.40: We revisit Example 8.4.20 in the context of contact geometry.
There we described the weighted projective space as a shifted symplectic reduction
of the cone C
3
0 by the weighted circle action o
1
w
with weights w = (2. 3. 6). The
0 level set j
1
(0) of the shifted moment map is the weighted sphere o
5
w
represented
as 2[.
0
[
2
+ 3[.
1
[
2
+ 6[.
2
[
2
= 6 which diers from the standard representation as in
Example 8.4.39. In this case the contact 1-form on o
5
w
is just a rescaled standard
contact form, namely
1
6
with Reeb vector eld
w
= 2H
0
+3H
1
+6H
2
. The rescaling
of the contact form corresponds to the rescaling of the characteristic hyperplane,
viz. 2:
0
+3:
1
+6:
2
= 6 which coincides with the hyperplane described in Example
8.4.20. Now the rational polyhedral cone corresponding to the standard contact
structure is just the positive octant with the inward normal vectors (1. 0. 0). (0. 1. 0)
and (0. 0. 1). Intersecting the rescaled characteristic hyperplane with this cone gives
the rational polytope 8.4.5. Notice that on the level of contact geometry there is no
need for labels, the branch divisors are encoded in the equation for the characteristic
hyperplane. This information is lost, however, when transforming to polytopes in
t

2
. hence, the need for labels at this level.
Example 8.4.41: Consider the lens spaces 1(j. c
1
. c
2
) with j relatively prime to
c
i
as discussed in Example 8.1.21. The contact structure on 1(j. c
1
. c
2
) is that
induced from the standard contact structure on o
5
. We follow [GL02] (Note that
1(j. c) should be 1(c. j) in [GL02]) and consider the surjective map : T
3
T
3
dened by (
0
.
1
.
2
) = (
p
0
.
q
1
0

1
.
q
2
0

2
). The kernel of is the subgroup
of T
3
dened by
p
0
= 1.
1
=
q
1
0
.
2
=
q
2
0
which is isomorphic to Z
p
. The lens
space 1(j. c
1
. c
2
) is the quotient of o
5
by the action (.
0
. .
1
. .
2
) (.
0
.
q1
.
1
.
q2
.
2
).
where is a j
th
root of unity. There is an action of T
3
on 1(j. c
1
. c
2
) induced
by the standard action of T
3
on o
5
. The image of the induced moment map is the
polyhedral cone
C
p,q1,q2
= :
0
(1. 0. 0) +:
1
(c
1
. j. 0) +:
2
(c
2
. 0. j) [ :
0
. :
1
. :
2
0 .
Consider the sublattice / of the integral lattice T
3
Z
generated by the inward nor-
mal vectors to the facets of C
p,q
1
,q
2
. It is easy to see that / is generated by
(0. 0. 1). (0. 1. 0) and (j. 0. 0). Thus, we verify item (ii) of Theorem 8.4.42 below
that
1
(1(j. c
1
. c
2
)) = T
3
Z
/ Z
p
.
290 8. SYMMETRIES AND SASAKIAN STRUCTURES
Let us choose the standard contact form on o
5
and Reeb vector eld
w
=
H
0
+ c
1
H
1
+ c
2
H
2
with w = (1. c
1
. c
2
). We represent the weighted o
5
w
as [.
0
[
2
+
c
1
[.
1
[
2
+c
2
[.
2
[
2
= c
1
c
2
. where for simplicity we take c
1
and c
2
to be relatively prime.
The rescaled characteristic hyperplane is :
0
+ c
1
:
1
+ c
2
:
2
= c
1
c
2
and intersecting
this with the cone C
p,q1,q2
gives the polytope in the r = c
1
c
2
plane with vertices
(c
1
c
2
. 0. 0). (c
1
c
2
. c
2
j. 0). (c
1
c
2
. 0. c
1
j). Identifying this plane with t

2
= R
2
gives the
lt-polytope with vertices (0. 0). (c
2
j. 0). (0. c
1
j) with no labels. This is the poly-
tope corresponding to the orbifold weighted projective space {(1. c
1
. c
2
) with the
symplectic form j. where [] H
2
orb
({(1. c
1
. c
2
). Z) is a generator. In particular,
the special case c
1
= c
2
= 1 gives CP
2
with symplectic form j. where is the
Fubini-Study 2-form.
Exercise 8.8: Discuss the case in Example 8.4.41 when c
1
and c
2
are not relatively
prime.
Concerning the topology of compact toric contact manifolds we have the fol-
lowing theorem due to Lerman. Items (ii) and (iii) below are given in [Ler04b]
and (i) follows from Proposition 8.4.22 and Corollary 8.4.31 by a Gysin or spectral
sequence argument.
Theorem 8.4.42: Let (`. T. T. ) be a compact toric contact manifold of Reeb
type of dimension 2n+1 3. and let / denote the sublattice of the integral lattice Z
T
generated by the normal vectors to the facets of the polyhedral cone C() associated
to `. Then
(i) The Betti numbers /
2i+1
(`) vanish for i = 0. . . . n 2.
(ii)
1
(`) is isomorphic to the nite Abelian group Z
T
/.
(iii)
2
(`) is a free Abelian group of rank n 1. where is the number
of facets of C().
8.5. Contact and Sasakian Reduction
The contact analogue of the Marsden-Weinstein-Meyer reduction theorem was
rst treated systematically by Albert [Alb89]. However, it appeared earlier in
[GS82a] in the guise of co-isotropic reduction of symplectic cones which is closely
related to a well-known generalization of reduction due to Kazhdan, Kostant, and
Sternberg [KKS78]. A related 3-Sasakian version of contact reduction appeared
later in [BGM94a]. At this stage we were unaware of both [Alb89] and the
connection with [GS82a]. This also appears to be the case of Geiges [Gei97a],
and even more recently in [Loo01] where contact reduction in the more general
case when the contact structure is not necessarily co-oriented is considered. In this
book we are only concerned with contact reduction at 0, and Theorem 8.5.1 below
assumes both that the contact structure is co-oriented and that 0 is a regular value
of the moment map. The singular case when 0 is not a regular value of the moment
map is treated in [LW01]. As in the symplectic case this gives rise to stratied
quotients. A version of reduction at nonzero values of the contact moment map
was described in Albert [Alb89] and is implicit in [GS82a]. A somewhat dierent
approach to reduction at non-zero values of the contact moment map has recently
been developed in [Wil02], and a Sasakian version in [DO06]. Contact reduction
of cosphere bundles both at 0 and not are treated in [DOR02]. The important
consideration about reduction at 0 is that it is independent of the choice of contact
form.
8.5. CONTACT AND SASAKIAN REDUCTION 291
Theorem 8.5.1: Let (`. T) be an oriented and co-oriented contact manifold of
dimension 2n+1. and let G be a compact Lie group with an eective action /(G)
Con(`. T)
+
. Let be a G-invariant contact form with ker = T. and let j

:
`g

denote its moment map. Then if 0 is a regular value of j

. the quotient
`
0
= j
1

(0)G is a contact orbifold of dimension 2(n dim g) +1 whose contact

1-form
0
descends from . Furthermore, the quotient `
0
depends only on the
underlying contact structure T and not on the 1-form. If the action of G is free on
j
1

(0). then the quotient (`

0
. T
0
) is a smooth contact manifold with T
0
= ker
0
.
Proof. Since 0 is a regular value of j

. the dierential dj

has rank equal

to the dimension of G at all points of the zero level set j
1

(0) which is a smooth

embedded G-invariant submanifold of ` of codimension dim G. Moreover, the
dierential dj

(r) : T
x
`T

(x)
g

is the transpose of the evaluation map

A

(r). So G acts locally freely on j

1

(0). Since G is compact there exists

a G-invariant Riemannian metric compatible with the contact form such that
the locally free action of G denes a Riemannian foliation of ` with compact
leaves. Thus, by Molinos Theorem 2.5.11, the space of leaves `
0
= j
1

(0)G is
an orbifold, and the natural map : j
1

(0)j
1

(0)G is an orbifold Riemannian

submersion. Furthermore, if the action is free, then all the local uniformizing groups
are trivial so the quotient `
0
is a smooth manifold.
Now from Denition 8.4.26
(8.5.1) j
1

(0) =
g
(A

) = 0 .
so for all g the vector elds A

restrict to j
1

(0) and lie in the restriction of

the contact subbundle T to j
1

(0). Then it follows from invariance that [

(0)
is
basic. So it descends to a contact 1-form
0
on `
0
. Dening T
0
= ker
0
implies
that (`
0
. T
0
) is a contact orbifold. Moreover, from invariance (`
0
. T
0
) depends
only on the contact structure T and not on the choice of 1-form.
Remark 8.5.1: The hypothesis of Theorem 8.5.1 can be weakened to a proper
eective action of any Lie group preserving the contact structure. In this case
Molinos Theorem does not apply. However, one can use a slice theorem such as
Theorem 2.4.1 of [DK00] to construct an orbifold chart on the quotient. This
applies equally well to Theorems 8.5.2 and 8.5.3 below.
Now let denote the Reeb vector eld of a contact manifold (`. ). Let /(G)
Con(`. ) be a proper eective action of a Lie group G. Notice that if corresponds
to an element in the Lie algebra g. then 0 in not in the image of the moment map
j

. So we cannot perform reduction at 0 by G in this case. Thus when doing contact

reduction by a Lie group G we always assume that the action /(G) Con(`. ) is
not of Reeb type. However, by Proposition 6.1.31 lies in the center of con(`. ).
so commutes with /

g. It follows that for all g we have (A

) = 0. Thus,
from equation (8.5.1), at points of j
1

(0) the Reeb vector eld is tangent to

j
1

(0). and since commutes with /

g, it descends to a vector eld

0
on the
quotient `
0
. It is easy to check that
0
is the Reeb vector eld of
0
.
Given the contact manifold (`. ) we choose a compatible almost contact struc-
ture (. . ) satisfying equations (6.4.1). We then choose the associated Riemann-
ian metric, i.e., o = d ( 1l) + to obtain a contact metric structure
(. . . o). Now let /(G) Con(`. T)
+
be an eective action of a compact Lie
group as in Theorem 8.5.1, but suppose also that /(G) preserves the contact metric
292 8. SYMMETRIES AND SASAKIAN STRUCTURES
structure, that is, /(G) Aut(. . . o). Then for any Aut(. . . o) we have

= .

= .

= .

o = o .
Theorem 8.5.2: Assuming the hypothesis of Theorem 8.5.1, we let (. . . o) be
a contact metric structure associated to the underlying contact structure. Suppose
in addition that /(G) is a subgroup of Aut(. . . o). Then there exists a unique
contact metric structure (
0
.
0
.
0
. o
0
) on the quotient `
0
such that

o =

o
0
and

0
. where : j
1

(0)` denotes the natural inclusion and :

j
1

(0)`
0
is an orbifold submersion. Furthermore, if (. . . o) is K-contact
so is (
0
.
0
.
0
. o
0
).
Proof. We already have constructed (
0
.
0
) on the quotient space `
0
which
under the hypothesis of Theorem 8.5.1 is a contact orbifold. We need to construct

0
and o
0
. The latter is easy since by invariance the induced metric

o descends
to a unique metric o
0
on the complement to the trivial line orbibundle on `
0
generated by
0
. To construct
0
we notice that for any g the vector eld A

restricted to the zero set j

1

(0) is tangent to j
1

(0). whereas, the vector elds

A

are sections of the normal bundle ^j

1

(0). To see this we notice that for any

g. and for any vector Y tangent to j
1

(0) this implies

0 = (
X
)(Y ) = d(A

. Y ) +d((A

))(Y ) = d(A

. Y ) .
So for all g and all Y tangent to j
1

(0). we have o(A

. Y ) = d(A

. Y ) = 0.
Now since the action / of G is locally free, the vector elds A

g
gener-
ate a trivial integrable subbundle 1 of T`. Furthermore, the non-degeneracy of
on the complement to the trivial line bundle 1

gives an isomorphism [
E
:
1^j
1

(0). But we also have the orthogonal decomposition

T`[

(0)
= 1 1 ^j
1

(0) .
where 1 is the orthogonal complement to 1 in Tj
1

(0). So the non-degeneracy

of implies that [
F
is an endomorphism of 1 satisfying ([
F
)
2
= 1l
F
+ .
Now 1 is isomorphic to the tangent bundle of `
0
. and [
F
is invariant under the
group action, so
F
descends to a unique endomorphism
0
on T`
0
satisfying

2
0
= 1l +
0

0
. This constructs the contact metric structure on `
0
.
If

o = 0 it is easy to trace through the identications to prove that

0
o
0
= 0
as well. Thus, (
0
.
0
.
0
. o
0
) is a K-contact structure if (. . . o) is K-contact.
The reduction which is of most interest to us is given by the next theorem
due to Grantcharov and Ornea [GO01]. There are several ways to prove this
theorem. The proof in [GO01] uses the curvature conditions of Proposition 7.3.17
for Sasakian manifolds. Our proof uses Theorem 8.5.2 and Theorem 7.3.16.
Theorem 8.5.3: Let (`. . . . o) be a Sasakian manifold of dimension 2n+1 and
let G be a compact Lie acting as automorphisms of the Sasakian structure. Assume
that 0 is a regular value of the moment map j

: `g

. Then the quotient

orbifold `
0
with its induced structure o
0
= (
0
.
0
.
0
. o
0
) is Sasakian.
Proof. Consider the (2. 1)-tensor eld on `. Since (`. . . . o) is Sasakian
satises = o 1l by Theorem 7.3.16. But from the proof of Theorem
8.5.2 this equation descends to
0

0
=
0
o
0
1l
0
. where
0
is the Levi-Civita
connection of o
0
. Thus, (
0
.
0
.
0
. o
0
) is Sasakian by Theorem 7.3.16.
8.5. CONTACT AND SASAKIAN REDUCTION 293
There is a simple relationship between contact reduction at 0 and symplectic
reduction of a symplectic cone. Let us apply Theorem 8.4.2 with = 0 to a
symplectic cone C(`). First notice that the symplectic 2-form given by equation
(6.5.5) is exact, so every symplectic action is Hamiltonian. By Proposition 8.4.23
the groups S
0
(C(`). ) and Con(`. ) are identied, and their corresponding
moment maps are related by equation (8.4.8). We are ready for
Theorem 8.5.4: Let (`. ) be a contact manifold and (C(`). ) its symplectic
cone with = d(:
2
). Let G be a compact Lie group with an eective action /(G)
Con(`. ) = S
0
(C(`). ). and let j

and j be the moment maps on ` and C(`).

respectively. Suppose that 0 g

is a regular value of the moment map j

. Then
the symplectic reduction (C(`)
0
.
0
) of (C(`). ) is the symplectic cone on the
contact reduction (`
0
.
0
) of (`. ).
Proof. The reduced space C(`)
0
is j
1
(0)/(G). where by abuse of notation
/denotes the action of G on both ` and C(`). It follows from the transformations
(8.4.7) and the fact that /(G) commutes with homotheties that C(`)
0
is a cone.
Moreover, equations (8.4.8) imply that j
1
(0) = j
1

(0) R
+
. So C(`)
0
is a cone
over `
0
. Letting : denote the coordinate on R
+
one easily sees by invariance under
/(G) that
0
= d(:
2

0
). so C(`)
0
is the symplectic cone on `
0
.
Now suppose that ` has a contact metric structure (. . . o). then by Propo-
sition 6.5.5 this determines a unique almost Kahler structure (d:
2
+:
2
o. d(:
2
). 1)
on the symplectic cone C(`). where these structures are related by equations
(6.5.2). Since every element o Aut(. . . o) lifts to an element in o Aut(d:
2
+
:
2
o. d(:
2
). 1). Theorems 8.5.2 and 8.5.4 imply that the corresponding reduced
structures (
0
.
0
.
0
. o
0
) and (d:
2
+ :
2
o
0
. d(:
2

0
). 1
0
) on `
0
and C(`
0
). respec-
tively are also related by equations (6.5.2). By Proposition 6.5.6 ` is K-contact
if and only if i is pseudo-holomorphic, so again by Theorems 8.5.2 and 8.5.4
the reduced structure (
0
.
0
.
0
. o
0
) will be K-contact with i
0
is pseudo-
holomorphic. The reductions are also clearly compatible having Sasakian structures
on ` and `
0
and the corresponding Kahler structures on C(`) and C(`
0
). Let
us denote the reduction procedure, either symplectic or contact, by a double arrow
. Then we can summarize our results of reduction by a compact Lie group G with
the follow commutative diagram
(8.5.2)
` C(`)

`
0
C(`
0
).
where the two-sided arrow can be interpreted as either a quotient projection
` C(`) by R
+
. or an inclusion ` C(`) putting : = 1. In the case
that the contact metric structure (. . . o) on ` is quasi-regular, it is naturally
K-contact by Theorem 7.1.10 and we have a natural orbifold bration : `7.
Thus, we can extend our diagram to the commutative diagram
(8.5.3)
7 ` C(`)

7
0
`
0
C(`
0
).
One can think of the reduction of diagram 8.5.3 as a reduction of Morita equiv-
alent Lie groupoids. The left hand reduction is that of proper etale Lie groupoids
while the reduction of ` and its cone C(`) corresponds to the holonomy groupoids
294 8. SYMMETRIES AND SASAKIAN STRUCTURES
associated to the respective foliation. For reduction of groupoids see [MW88]. Let
us consider an example of the Sasakian reduction procedure given in [GO01] ap-
plied to the diagram 8.5.3.
Example 8.5.5: We begin by considering a particular case of the brations of the
top row of diagram 8.5.3, namely CP
3
o
7
(C
4
`0). Writing C
4
= C
2
C
2
with coordinates x and y on the C
2
factors. Consider the circle action o
1
k,l
of
Example 7.6.10 given by (x. y) (c
il
x. c
ik
y). where / and | are relatively
prime integers. The moment map j : (C
4
` 0)R corresponding to this action
is
j = |[x[
2
+/[y[
2
and the zero set is the cone C(o
3
o
3
). Restricting to o
7
(or equivalently quoti-
enting by R
+
) gives j
1
S
7
(0) = o
3
o
3
and the quotient j
1
S
7
(0)o
1
k,l
is identied
with the Wang-Ziller manifolds `
1,1
k,l
which as discussed in Example 7.6.10 are dif-
feomorphic to o
2
o
3
for all relatively prime integers (/. |). Moreover, the quotient
of j
1
S
7
(0) by the 2-torus dened by (x. y) (c
i1
x. c
i2
y) is CP
1
CP
1
. so our
reduction diagram 8.5.3 becomes
(8.5.4)
CP
3
o
7
C
4
` 0

CP
1
CP
1
`
1,1
k,l
C(`
1,1
k,l
).
Here the manifolds `
1,1
k,l
are inequivalent toric contact manifolds for each pair of
relatively prime integers (/. |). This shows that o
2
o
3
admits a countably innite
number of inequivalent toric contact structures. In fact we shall see that there are
many more toric contact structures on o
2
o
3
.
Exercise 8.9: Construct the polyhedral cones of `
1,1
k,l
and the Delzant polytopes
of the toric symplectic manifolds (o
2
o
2
.
k,l
). and discuss their relationship.
In [BG00a] the authors proved
Theorem 8.5.6: Let (`. T. T. ) be a compact toric contact manifold of Reeb type
with a xed quasi-regular contact form . Then (`. T. T. ) is isomorphic to the
reduction of a sphere o
2N1
with its standard contact structure and with a xed
1-form
a
by a subtorus of T.
Proof. Now ` is compact of dimension, say, 2n + 1. and since is quasi-
regular, the space of leaves Z of the characteristic foliation T is a compact sym-
plectic orbifold of dimension 2n. Furthermore, since ` is toric, so is Z. that is,
there is an n dimensional torus T
n
preserving the symplectic structure on Z.
Furthermore, by Theorem 7.1.3 represents an integral class in H
2
orb
(Z. Z) and `
is the total space of the principal o
1
V-bundle `

Z whose rst Chern class is

represented by d =

.
Now by the Lerman and Tolman Theorem 8.4.18 (Z. ) is isomorphic to the
symplectic reduction (C
N
.
0
) with the standard symplectic structure by a torus
T
Nn
of dimension n. If j
Nn
: C
N
t

Nn
denotes the moment map for the
T
Nn
action, then (Z. ) is isomorphic to (j
1
Nn
()T
Nn
. ). where is a regular
value of j
Nn
and is the unique symplectic 2-form induced by reduction. Let

denote the above isomorphism. It follows that the cohomology class of = (

1
)

is integral in H
2
orb
(j
1
Nn
()T
Nn
. Z). By the orbifold version of the Boothby-Wang
8.5. CONTACT AND SASAKIAN REDUCTION 295
theorem there is an o
1
V-bundle : 1j
1
Nn
()T
Nn
. a connection form on
1 such that d =

. Thus there is an o
1
-equivariant V-bundle map
`

j
1
Nn
()T
Nn
such that

d = d. Thus,

and dier by a closed 1-form. But the space

of closed 1-forms on ` is path-connected, so one can nd a one parameter family
of connections having the same curvature that connect

to . So by Grays
Stability Theorem 6.1.32

and dene the same contact structure. Thus, we

can choose

= . Moreover, by equivariance the characteristic vector of the

contact manifold (1. ) lies in the Lie algebra t
Nn
. so we can split o the circle
that it generates and write T
Nn
= o
1

T
Nn1
, where T
Nn1
is an (n1)-
dimensional torus. It follows that 1 = j
1
Nn
()T
Nn1
. Hereafter, we identify
(`. ) with (j
1
Nn
()T
Nn1
. ). Now j
1
Nn
() is a torus bundle over a compact
manifold, so it is a compact manifold which by construction is an intersection of
n real quadrics in C
N
. It follows that there is a component of the moment map
j
Nn
which takes the form

i
o
i
[.
i
[
2
with o
i
0 for all i. Let a denote the vector in
R
N
whose i
th
component is o
i
. and consider the ellipsoid
a
=

i
o
i
[.
i
[
2
= 1

=
o
2N1
. Then there is a T
Nn
-moment map
a
:
a
t

Nn
such that
1
a
(0) =
j
1
Nn
(). Furthermore, letting
0
=

i
(r
i
dn
i
n
i
dr
i
) we see that d
0
=
0
on C
N
and that
0
[

a
=
a
[

a
. where
a
is the deformed 1-form of Example 8.4.39. Thus,
letting j : j
1
Nn
()` denote the natural submersion, and : j
1
Nn
()
a
the natural inclusion, we see that j

a
. so (`. ) is obtained from (
a
.
a
)
by contact reduction.
Now Corollary 8.4.19 says that every symplectic toric orbifold possesses an
invariant complex structure which is compatible with its symplectic form giving it
a compatible Kahler structure. Combining this with Theorem 8.5.6 gives
Corollary 8.5.7: Every compact toric contact manifold of Reeb type admits a
compatible invariant Sasakian structure.
We now give a brief discussion of a construction due to Lerman [Ler04a] that
allows one to construct K-contact structures on the total space of a bre bundle
whose bres are K-contact. This is a generalization of the join construction of
Section 7.6.2 as well as a certain bre join construction due to Yamazaki [Yam99].
We then apply Lermans construction to toric Sasakian geometry. Actually one
can work within the pure contact setting, and it is the contact analog of symplectic
bre bundles described in [GLS96]. Here is the denition.
Denition 8.5.8: A bre bundle 1`

1 is called a contact bre bundle

if
(i) 1 is a co-oriented contact manifold with contact bundle T.
296 8. SYMMETRIES AND SASAKIAN STRUCTURES
(ii) There exist an open cover l
i

iI
of 1 and local trivializations

i
:
1
(l
i
)l
i
1 such that for every point j l
i
l
j
the transition
functions
j

1
i
[
pF
are elements of Con(1. T)
+
.
We need the notion of fatness of a bundle due to Weinstein [Wei80]. Let
be a connection 1-form in a principal bundle 1(`. G) with Lie group G. and let
= 1 denote its curvature 2-form. Let o g

be any subset in the dual g

of
the Lie algebra g of G. We say that the connection is fat on o if the bilinear map
(8.5.5) j : H1 H1 : R
is non-degenerate for all j o.
We rst recall the main lines of the construction, not in full generality, but
adapted to our needs. Let : 1 1 be a principal G-bundle endowed with a
connection (we dont distinguish between the connection and its 1-form). Let 1
be a K-contact manifold, with xed contact form
F
and Reeb eld
F
. Suppose
G Aut(1. ), i.e., it acts (from the left) on 1 by strong contactomorphisms and
denote by : 1 g

the associated momentum map. Then Lerman [Ler04a]

proves
Theorem 8.5.9: In the above setting, if the connection is fat at all the points
of the image of the momentum map , then the total space ` of the associated
bundle 1
G
1 admits a 1-contact structure.
We recall that fatness at a point g

means that the bilinear map

A
:
H
A
H
A
R is non-degenerate, where H
A
and
A
are, respectively, the horizontal
distribution and the curvature form dened by the connection . In particular, if
G is a torus, the bundle : 1 1 is identied, up to a gauge transformation,
by a connection form such that d =

with [] H
2
(1. Z). Then, if is a
symplectic form, is certainly fat.
Now combining Theorem 8.5.9 with Corollary 8.5.7 one can prove [BGO07]
Theorem 8.5.10: Let 1
2n+1
be a compact toric contact manifold of Reeb type and
with torus T
n+1
Con(1. ). Let : 1 1 be a principal T
n+1
bundle over a
toric compact symplectic manifold 1. Then 1
T
n+1 1 is a toric Sasakian manifold.
It is easy to see [BGO07] how this generalizes the join construction. Suppose
that G = o
1
. 1 is a regular Sasakian manifold over a polarized Kahler base man-
ifold 1
F
. and 1 is a polarized Kahler manifold, then taking the product complex
structure on 1 1
F
reproduces the join construction.
In the paper [Gui94a] and the book [Gui94b], Guillemin constructs canoni-
cal Kahler metrics on compact toric symplectic manifolds from the combinatorial
data of the Delzant polytopes. See also [CDG02, BG04] for further development.
Furthermore, this construction has been generalized to compact toric symplectic
orbifolds by Abreu [Abr01] using [LT97]. This gives rise to special Sasakian
metrics on compact toric contact manifolds of Reeb type. However, such special
structures are far from unique. First, one can deform the characteristic foliation
within a xed contact and CR structure as described in Section 8.2.3. When apply-
ing these deformations to toric geometry we also need to preserve the toric structure.
Thus, this type of deformation is performed by perturbing the contact 1-form as in
equation (8.2.3) in a T
n+1
invariant way, that is, the function 1 is invariant under
T
n+1
. We also take the vector eld in equation (8.2.3) to lie in the image /

t
n+1
.
Such deformations are equivalent to deforming the characteristic hyperplane in t

.
8.5. CONTACT AND SASAKIAN REDUCTION 297
Second, one can add to the contact form a T
n+1
-invariant basic 1-form . This
gives another toric Sasakian structure. In the quasi-regular case this is equivalent
to adding i

1 to the symplectic form on the orbifold base, where 1 is a smooth

T
n
-invariant function.
We now give a brief review of the Guillemin-Abreu construction. Let t

be an lt-polytope and let

denote its interior. Now can be represented by

inequalities of the form
(8.5.6) 'r. n
i
`
i
. i = 1. . . . . d = # of facets of .
where n
i
is the inward normal to the i
th
facet and is a primitive element in the
lattice Z
T
. We dene ane functions |
i
: t

R by
(8.5.7) |
i
(r) = 'r. :
i
n
i
` :
i

i
.
where :
i
is the label of the i
th
facet 1
i
. so |
i
(r) 0 if and only if r

and
|
i
(r) = 0 if and only if r 1
i
. Now according to Theorem 8.4.18 to each lt-
polytope there corresponds a compact toric symplectic orbifold (Z. . T. j) such
that j(7) = . The idea is to construct the Kahler structure on the orbifold Z
from a potential function on the polytope . Dene the Guillemin potential j

by
(8.5.8) j

=
1
2
d

i=1
|
i
(r) log |
i
(r)
which is a smooth function on the interior

. The dense open subset j

1
(

) 7
is dieomorphic to

T
n
on which we have the so-called action-angle coordinates
(r
1
. . . . . r
n
;
1
. . . . .
n
). The potential j

gives rise to Guillemin metrics that are

constructed from the Hessian of a function G. Dene
o
ij
=

2
G
r
i
r
j
.
Then the Guillemin metric is dened by
(8.5.9) o

o
ij
dr
i
dr
j
+o
ij
d
i
d
j

.
where o
ij
are the components of the inverse of o
ij
and G = j

. This Kahler
structure coincides up to isometry with the Kahler structure on the dense open set
j
1
(

) induced from the reduction by Corollary 8.4.19. This can be checked by

choosing a local slice of the T
n
action [Abr01]. Thus, one obtains a smooth Kahler
orbifold metric on all of 7 = j
1
(). Furthermore, one has the freedom to add to
j

the Hessian of a smooth function / on the whole polytope giving more general
toric Kahler orbifold metrics. In fact, while the canonical Guillemin metric enjoys a
certain combinatorial simplicity, it does not enjoy other important properties. For
example, it is not generally extremal
3
. See [Abr01] where the extremal metrics for
weighted projective spaces CP(w) are computed in terms of the combinatorial data.
We refer to the metric of equation (8.5.9) with G = j

as the canonical Guillemin

metric. This construction gives
Theorem 8.5.11: Let (`. T. T. ) be a compact toric contact manifold of Reeb
type with a xed quasi-regular contact form . and let : `Z denote the
3
Nevertheless, a potential similar to the Guillemin potential had been introduced earlier by
Simanca [Sim91, Sim92] to construct extremal Kahler metrics.
298 8. SYMMETRIES AND SASAKIAN STRUCTURES
induced orbifold bundle over the compact Kahler orbifold (Z. ) satisfying d =

.
Then the Riemannian metric
o =

+ .
where o

is the canonical Guillemin metric on the orbifold Z. is a toric Sasakian

metric.
We refer to the metric o of Theorem 8.5.11 as the Guillemin-Sasakian metric.
Of course, as with the canonical Guillemin metric we can add a T
n+1
invariant basic
1-form to to obtain other toric Sasakian metrics. Unlike the Guillemin metric on
Z. the toric Guillemin-Sasakian metrics in Theorem 8.5.11 are unique only up to a
gauge transformation in the choice of satisfying d =

.
CHAPTER 9
Links as Sasakian Manifolds
Links arose early on in the study of knots in three dimensions (cf. [Rol76]),
but it is the higher-dimensional case that is of interest to us. We are interested
in links arising from isolated hypersurface singularities, although we also treat
briey the case of complete intersection singularities. So good references here are
[Mil68, Dim92]. As we shall see in Section 9.2 links of isolated hypersurface sin-
gularities dened by weighted homogeneous polynomials admit Sasakian structures
in a natural way [Abe77, Tak78]. In fact it was realized in the mid seventies that
both Brieskorn manifolds as well as the more general links of isolated hypersurface
singularities arising from weighted homogeneous polynomials admitted both contact
[AE75, LM76, SH76] and almost contact [Abe76, ST76] structures. However,
it was Abe [Abe77] who rst developed the connection with Sasakian geometry
proceeded by Takahashi [Tak78] who viewed the links as invariant submanifolds of
the weighted spheres of Example 7.1.5. See Proposition 9.2.2 below.
9.1. Preliminaries
We begin with the main denitions following Milnors beautiful classical text
[Mil68]. Let (.
0
. . . . . .
n
) denote the standard complex coordinates in C
n+1
. and
let 1 C[.
0
. . . . . .
n
] be a polynomial that is non-constant with respect to each of
the variables .
i
. i.e., the partial derivative
f
z
i
is not everywhere zero for any i =
0. . . . . n. We are interested in the zero locus \
f
of 1 which is a complex hypersurface
in C
n+1
. Recall that a point z \
f
is called a singular point or critical point of \
f
if its dierential d1 vanishes at z. Otherwise, z is called a regular or a simple point.
A critical point z is said to be isolated if there is a neighborhood l of z such that
all the points of l ` z are regular.
Denition 9.1.1: Let z
0
be any point z
0
\
f
. We consider a sphere o
2n+1

(z
0
) of
radius c centered about z
0
. Then the intersection 1
f
(c) = \
f
o
2n+1

(z
0
) is called
a link of \
f
.
If z
0
is a regular point it follows from the well-known Morse Lemma [Mil63]
that for c suciently small, 1
f
(c) is dieomorphic to the standard unknotted big
sphere inside the round sphere o
2n+1

. For that reason we are interested in the case

when z
0
is singular. It should be clear that if z
0
is an isolated singular point then for
c suciently small 1
f
(c) is independent of c in the sense that 1
f
(c) and 1
f
(c
t
) are
dieomorphic for 0 < c
t
< c. Thus, we shall write 1
f
(c) as 1
f
. When z
0
is singular
1
f
is often referred to as the link of a hypersurface singularity. We shall assume
that all variables .
0
. . . . . .
n
occur in 1 for if not we would have a polynomial in one
less variable. We also assume that there is no constant term so that the origin lies
in \
f
. and we want the origin to be a singular point of 1 which means that there
are no linear terms in 1. Consider d1 : C
n+1
C
n+1
the dierential of 1. and let
299
300 9. LINKS AS SASAKIAN MANIFOLDS
(1) be the Jacobi ideal in C[.
0
. . . . . .
n
] generated by d1. The multiplicity j
f
(z
0
)
of 1 at an isolated singular point z
0
\
f
is dened to be the degree of the Gauss
map G : o
2n+1

o
2n+1
dened by
(9.1.1) G(z) =
d1(z)
[[d1(z)[[
.
If z
0
is a regular point then j
f
(z
0
) is dened to be 0. Let z
0
be the origin in C
n+1
and dene the Milnor algebra as the quotient algebra
(9.1.2) `(1) =
C[.
0
. . . . . .
n
]
(1)
.
Then it can be shown [Mil68] that
(9.1.3) j
f
= j
f
(0) = dim
C
`(1) .
j
f
is called the Milnor number of the isolated hypersurface singularity. We shall
see later how `(1) encodes much of the geometry of the link 1
f
.
9.2. Sasakian Structures and Weighted Homogeneous Polynomials
In this section we consider the links of a hypersurface singularity dened by
the weighted homogeneous polynomials of Denition 4.6.1, and show that such
links have a natural Sasakian structure associated with them. According to the
Structure Theorem 7.1.10, in order that a link 1
f
admits a Sasakian structure it
is necessary that it also admits a locally free circle action. Thus, at least locally
the hypersurface \
f
admits a locally free C

action. This leads us naturally to the

weighted homogeneous polynomials of Denition 4.6.1. Consider the ane space
C
n+1
together with the weighted C

(w)-action
1
as in Denition 4.5.1.
The zero set \
f
of a non-degenerate weighted homogeneous polynomial 1. is
just the weighted ane cone C
f
of Denition 4.6.3. If the origin is an isolated
singularity in C
f
. then it is the only singularity of C
f
. In this case we can take
c = 1 so that o
2n+1

(z
0
) becomes the unit sphere o
2n+1
centered at the origin with
its weighted Sasakian structure as in Example 7.1.12. In this case our link
(9.2.1) 1
f
= C
f
o
2n+1
.
is a smooth manifold of dimension 2n 1 which by the Milnor Fibration Theorem
9.3.1 is (n 2)-connected. Compact (n 2)-connected manifolds ` of dimension
2n 1 have been called highly connected by Wall [Wal67b] since their only non-
vanishing homology groups, H
i
(`. Z) occur for i = 0. n1. n. 2n1. Wall [Wal67b]
and his students [Bar65, Wil72] have given a rough classication of these highly
connected manifolds which will be of much interest to us later. In particular, for
n = 3 there is a complete classication [Bar65] which we discuss in detail in
Chapter 10.
When the origin is an isolated singularity of a weighted homogeneous polyno-
mial, the link 1
f
is a smooth manifold. This leads to
1
Our convention of calling the positive integers w
j
weights is not entirely standard. In the
literature on links it is common to refer to the rational numbers d/w
i
as weights (cf. [MO70,
OW71b]). Our convention is clearly more natural in this book, since it relates to the C

(w)-action
dened earlier. However, rational weights will be introduced later in Denition 9.3.8.
9.2. SASAKIAN STRUCTURES AND WEIGHTED HOMOGENEOUS POLYNOMIALS 301
Denition 9.2.1: We denote by \H1
2n1
the set of dieomorphism classes of
smooth closed (2n 1)-dimensional manifolds 1
f
that can be realized as the link
of an isolated hypersurface singularity of a weighted homogeneous polynomial 1 in
n + 1 complex variables, and we let \H1
2n1
(d) denote the subset of \H1
2n1
that are realized by weighted homogeneous polynomials of degree d.
We exclude the case d = 1 as this does not have a singularity. So we have the
equality
(9.2.2) \H1
2n1
=

d2
\H1
2n1
(d) .
There may be polynomials of degree d with dierent weight vectors which give rise to
dierent elements of \H1
2n1
(d). and there may be smooth manifolds represented
by links of dierent degrees. When we x a weight vector w and a degree d we are
interested in all possible monomials such that Theorem 4.6.16 holds.
The circle subgroup o
1
C

acts on the ane space C

n+1
by taking
o
1
in Denition 4.5.1. Recall from Examples 7.1.5, and 7.1.12 that the weighted
Sasakian structure (
w
.
w
.
w
. o
w
) on the unit sphere o
2n+1
comes from the circle
action with the negative orientation generated by the vector eld
w
of equation
(7.1.5). Thus, our convention here is = c
i
with [0. 2]. We now have
[BG01b, Tak78]
Proposition 9.2.2: Let 1
f
be the link of a non-degenerate weighted homogeneous
polynomial of weight w and degree d. The Sasakian structure o
w
= (
w
.
w
.
w
. o
w
)
on o
2n+1
induces by restriction a Sasakian structure, also denoted by o
w
. on the
link 1
f
.
Proof. According to Theorem 7.6.2 we only need to check that for all points
j 1
f
the two conditions hold:
(i)
w
is tangent to 1
f
.
(ii)
w
T
p
1
f
T
p
1
f
.
To see that (i) holds, we dierentiate the dening equation in Denition 4.6.1 with
respect to .

w
1 =
d1
d
= d
d
1 .
So
w
1 vanishes at points of C
f
. hence, at points of 1
f
. This proves (i).
To prove (ii) we rst notice that since C
f
is dened by a holomorphic equation
we have 1TC
f
TC
f
. where 1 is the standard almost complex structure on C
n+1
.
Moreover, we have

w
[
1
L
f
= [
1
L
f
= 1[
1
L
f
.
where T
L
f
= ker
L
f
and to avoid confusion we let
L
f
denote the restriction of

w
to 1
f
. Now at a point j 1
f
we have (T
L
f
)
p
= T
p
T
p
C
f
. Since
w

w
= 0.
we need only check that
w
T
L
f
T
L
f
. But
w
leaves T invariant, and equals 1
on T
L
f
which leaves TC
f
invariant, so it leaves T
L
f
invariant.
Recall from Example 7.1.12 the unit sphere o
2n+1
w
with the weighted Sasakian
structure o
w
= (
w
.
w
.
w
. o
w
). but now with w (Z
+
)
n+1
(R
+
)
n+1
. Thus,
Proposition 9.2.2 says that the natural inclusion : 1
f
o
2n+1
w
is a Sasakian em-
bedding. It is also clear since the vector w lies in the positive integer lattice in R
n+1
that the Sasakian structures o
w
on both 1
f
and o
2n+1
w
are quasi-regular. Thus,
by Theorem 7.5.1 the spaces of leaves in both cases are compact polarized Hodge
302 9. LINKS AS SASAKIAN MANIFOLDS
orbifolds. We denote the space of leaves of 1
f
by Z
f
. and almost by denition the
space of leaves of o
2n+1
w
is the weighted projective space CP(w) = CP(n
0
. . . . . n
n
).
Thus, we have a commutative diagram
(9.2.3)
1
f
o
2n+1
w

Z
f
CP(w) .
where the horizontal arrows are Sasakian and Kahlerian embeddings, respectively,
and the vertical arrows are principal o
1
-orbibundles and orbifold Riemannian sub-
mersions. Of course, the complex orbifold Z
f
can also be represented as the quo-
tient (C
f
` 0)C

w
. Note that a choice of 1 determines (w; d). However, given
(w; d) typically there is no quasi-smooth polynomial 1 matching such data, and
if there is one, there can be a nite-dimensional family as indicated by Theorem
5.5.7. Moreover, for each xed values of the parameters, there is an o-deformation
class of Sasakian structures adapted to the link. We often think of all quasi-smooth
members of such a family. It is convenient to x some notation.
Denition 9.2.3: We denote by 1

(w; d) any quasi-smooth link 1

f
(or the whole
family of such links) given by a weighted homogeneous polynomial 1 with weight
vector w and degree d. We let F(T

(w; d)) denote the o-deformation class of the

family.
We use the star to distinguish such general links from the Brieskorn-Pham type
links discussed later. Alternatively, one could think of it as the quasi-smoothness
restriction. We have the following immediate consequence of the orbifold adjunction
formula equation (4.6.3):
Proposition 9.2.4: Let 1
f
1

(w; d) and let o

w
denote its induced Sasakian
structure. Then o
w
is quasi-regular, satises c
1
(T

w
) = o[d
w
]
B
. and is
(i) positive if and only if [w[ d 0 .
(ii) null if and only if [w[ d = 0 .
(iii) negative if and only if [w[ d < 0 .
In particular, the induced Sasakian structure of all such links 1
f
is either anti-
canonical, canonical, or null.
9.3. The Milnor Fibration and the Topology of Links
The topology of links of isolated hypersurface singularities is best understood
using the well-known Fibration Theorem of Milnor [Mil68] which we now discuss.
Theorem 9.3.1: Let z
0
\
f
be any point of the complex hypersurface \
f
. and let
c 0 be suciently small. Then the map : o
2n+1

(z
0
) 1
f
o
1
dened by
(z) =
1(z)
[1(z)[
is the projection map of a smooth bre bundle with a smooth parallelizable bre.
If z
0
is an isolated singular point of 1. then each bre 1 has the homotopy type
of a bouquet of n-spheres o
n
o
n
. and is homotopy equivalent to its closure

1 which is a compact manifold with boundary where the common boundary

1 is
precisely 1
f
. Furthermore, 1
f
is a smooth (n2)-connected manifold of dimension
2n 1.
9.3. THE MILNOR FIBRATION AND THE TOPOLOGY OF LINKS 303
The proof of this theorem uses Morse Theory [Mil63] and can be found in
[Mil68]. The number of o
n
s in the bouquet described in Theorem 9.3.1 is an
important invariant of the link and it is the Milnor number j. If the point z
0
is
regular, it follows from the Morse Lemma [Mil63] that j = 0. and the link 1
f
is
dieomorphic to the big (2n1)-sphere in o
2n+1
. In view of Theorem 9.3.1 we have
Theorem 9.3.2: Let ` \H1
2n1
. Then
(i) ` is highly connected, that is, it is (n 2)-connected.
(ii) ` is the boundary of a compact (n1)-connected parallelizable manifold
\ of dimension 2n.
(iii) If n is even the Betti number /
n1
(`) is even.
(iv) If n 3 then ` is a spin manifold.
Proof. Parts (i) and (ii) follow from the Milnor Fibration Theorem 9.3.1. By
Proposition 9.2.2 ` admits a Sasakian structure, so its odd Betti numbers are even
up to the middle dimension by Theorem 7.4.11. This proves (iii), and (iv) follows
from Proposition 9.2.4 and Theorem 7.5.27 as soon as n 3.
Using the Wang sequence of the Milnor bration together with Alexander-
Poincare duality gives the exact sequence [Mil68]
(9.3.1) 0H
n
(1
f
. Z)H
n
(1. Z)
1lh
H
n
(1. Z)H
n1
(1
f
. Z)0 .
where /

. known as the monodromy map (or characteristic map), comes from the
action of a generator of
1
(o
1
) on the homology of the bre. From this we see that
H
n
(1
f
. Z) = ker(1l /

) is a free Abelian group, and H

n1
(1
f
. Z) = coker(1l /

)
which in general has torsion, but whose free part equals ker(1l/

). So the topology
of 1
f
is encoded in the monodromy map /

. The characteristic polynomial (t) of

the monodromy map is another important link invariant. It is a generalization of
the Alexander polynomial of knot theory, and has also been called the Alexander
polynomial [Hir71, HZ74] of 1
f
. We can get some immediate information from
the exact sequence (9.3.1). If (1) = 0 then 1l /

is non-singular which implies

H
n
(1
f
. Z) = 0. that is 1
f
is a rational homology sphere. Furthermore, if [(1)[ = 1.
then 1l /

is an isomorphism, so 1
f
is a homology sphere, and if n 2 it must
be a homotopy sphere by the Hurewicz Theorem. Thus, we have arrived at:
Proposition 9.3.3: The following hold:
(i) 1
f
is a rational homology sphere if and only if (1) = 0.
(ii) 1
f
is a homology sphere if and only if [(1)[ = 1.
(iii) If 1
f
is a rational homology sphere, then the order of H
n1
(1
f
. Z) equals
[(1)[.
We now briey consider the case n = 2. Let us take z
0
to be the origin in
C
3
. If the origin is at worst an isolated singularity of 1, all that we know is that
1
f
is a smooth connected 3-manifold. But the well-known Theorem of Mumford
[Mum61, Dim92] implies more.
Theorem 9.3.4: If the origin is at worst a normal singularity of 1 then 1
f
is
simply connected if and only if 1
f
is dieomorphic to o
3
.
In fact in the simply connected case the origin is a regular point of 1. Thus,
hypersurface singularities can say nothing about the Poincare conjecture. Never-
theless, they play an important role in understanding homology spheres as we shall
see in Section 9.5.
304 9. LINKS AS SASAKIAN MANIFOLDS
9.3.1. Computing the Alexander Polynomial of Weighted Homoge-
neous Polynomials. In the case that 1 is a weighted homogeneous polynomial
(see Section 9.2 below), there is a well-known algorithm due to Milnor and Orlik
[MO70] for computing the free part of H
n1
(1
f
. Z) in terms of the characteris-
tic polynomial (t) = det(t1l /

). namely the Betti number /

n
(1
f
) = /
n1
(1
f
)
equals the number of factors of (t 1) in (t).
Now we shall assume that 1 is a quasi-smooth weighted homogeneous polyno-
mial. In this case the Milnor number j
f
can be calculated explicitly in terms of
the degree d and weights w = (n
0
. . . . . n
n
). This result is due to Milnor and Orlik
[MO70].
Proposition 9.3.5: Let 1
f
1

(w; d). Then

(9.3.2) j = j(1
f
) =
n

i=0

d
n
i
1

.
Proof. Consider the map o : C
n+1
C
n+1
dened by
o(.
0
. . . . . .
n
) = (.
w
0
0
. . . . . .
w
n
n
) .
Then 1 o is homogeneous of degree d and the i
th
component of the dierential
d1 o is homogeneous of degree d n
i
. So the degree of d1 o at the origin is
(d n
0
) (d n
n
). But the degree of d1 at the origin, which by denition is the
multiplicity j
f
. is obtained by dividing by the degree of o at the origin, giving the
formula.
Remark 9.3.1: It is remarkable that for isolated hypersurface singularities the
product in formula (9.3.2) is always an integer although the ratios dn
i
may not
be integers. This is not necessarily the case if the singularity is not isolated.
The closure

1 of 1 is a manifold with boundary that is homotopy equivalent to
1. and whose boundary is precisely the link 1
f
. The topology of 1
f
is determined
by the monodromy map induced by the o
1
w
-action. In the case of isolated hyper-
surface singularities determined by weighted homogeneous polynomials this has a
particularly nice form.
Proposition 9.3.6: Let 1 be a weighted homogeneous polynomial of an isolated
hypersurface singularity. Then the Milnor bre 1 is dieomorphic to the non-
singular hypersurface
z C
n+1
[ 1(z) = 1 .
and the monodromy map /

is induced by the map / : C

n+1
C
n+1
dened by
/(.
0
. . . . . .
n
) =

c
2w
0
i
d
.
0
. . . . . c
2wni
d
.
n

.
Notice that / is periodic of period d. Thus, the monodromy map /

is diago-
nalizable over C with eigenvalues d roots of unity. Identifying 1 with 1
1
(1) we
let /
j
: 11 denote the ,-fold iterate of /. and 1
j
denote the xed point subset
of 1 under /
j
. The 1
j
s are determined by
(9.3.3) 1
j
=

z 1
1
(1) [ c
2ijw
l
d
.
l
= .
l
for all | = 0. . . . . n

.
By computing the Euler number
j
of the xed point sets 1
j
, Milnor [Mil68]
was able to compute the Alexander polynomial (t) of the link 1
f
using the Weil
9.3. THE MILNOR FIBRATION AND THE TOPOLOGY OF LINKS 305
zeta function
(9.3.4) (t) = exp

j=1

j
,
t
j

.
The zeta function (9.3.4) and the Euler numbers
j
are related by integers :
i
through
(9.3.5)
j
=

l]j
|:
l
. (t) =

j]d
(1 t
j
)
r
j
.
Then we have
Theorem 9.3.7: The Alexander polynomial (t) of the link 1
f
of an isolated
singularity of a weighted homogeneous polynomial 1 is determined by
(t)
(1)
n+1
= (t 1)

j]d
(t
j
1)
r
j
.
See [Mil68] for the proof of this theorem. Milnor and Orlik [MO70] use these
facts to give an algorithm for computing the Alexander polynomial (t) of the
monodromy map in terms of the degree d and weight vector w. The procedure is
this. Associate to any monic polynomial 1 with roots
1
. . . . .
k
C

its divisor
(9.3.6) div 1 = '
1
` + +'
k
`
as an element of the integral ring Z[C

] and let
n
= div(t
n
1). Then the s
satisfy the multiplication rule
(9.3.7)
a

b
= gcd(o. /)
lcm(a,b)
.
Denition 9.3.8: Given 1
f
1

(w; d) we dene rational weights

(9.3.8) (dn
0
. . . . . dn
n
) (n
0

0
. . . . . n
n

n
).
where (u; v) = (n
0
. . . . . n
n
;
0
. . . . .
n
) is given by
(9.3.9) n
i
=
d
gcd(d. n
i
)
.
i
=
n
i
gcd(d. n
i
)
.
As already remarked the rational weights were introduced in [MO70]. In the
literature on links they are quite often denoted by n
i
= n
i

i
so, once again, we
alert the reader that our weights w are dierent. We prefer to think of rational
weights as a (u; v)-data on a link 1
f
1

(w; d). In particular, one can always

translate between the (w; d)-data and the (u; v)-data. In a certain sense they are
dual to one another. We are now ready for
Theorem 9.3.9: The divisor of the Alexander polynomial (t) is given by
div =

u
i

i
1

= (1)
n+1

:
j

.
Proof. From Theorem 9.3.7 we have
div = (1)
n+1

:
j

.
Since the :
j
s are obtained from the Euler numbers
j
of the xed point sets 1
j
. we
need to express
j
in terms of d and the n
i
s, or alternatively in terms of the n
i
s
306 9. LINKS AS SASAKIAN MANIFOLDS
and
i
s. By equation (9.3.3) 1
j
is given by the vanishing of certain coordinates .
i
.
Then by permuting the variables .
i
if necessary we see that if
2
(9.3.10) , 0(n
i
) for i /. , 0(n
i
) for i /
then
1
j
= z C
n+1
[ 1(.
0
. . . . . .
k
. 0. . . . . 0) = 1 .
The polynomial 1
k
(.
0
. . . . . .
k
) = 1(.
0
. . . . . .
k
. 0. . . . . 0) is weighted homogeneous
with an isolated singularity at the origin in C
k+1
. Thus, 1
j
is the Milnor bre of
1
k
. So its only non-vanishing reduced homology group is

H
k
(1
j
. Z) 7

j
. where
by Proposition 9.3.5
j
j
=
k

i=0

d
n
i
1

.
Thus, the Euler number of 1
j
is

j
= 1 + (1)
k
j
j
= 1
k

i=0

1
d
n
i

= 1
k

i=0

1
n
i

.
Notice from equation (9.3.10) that we can rewrite this equation as
(9.3.11) 1
j
=

ui]j

1
n
i

.
This determines the
j
, and hence, the :
j
in terms of the n
i
s. To prove the second
equality we dene for any integer : between 0 and n.
(9.3.12) (n
0
. . . . . n
m
) = (1)
m+1

:
j
(n
0
. . . . . n
m
)
j

.
Viewing this as an element of the rational group ring Q[C

]. we have
Lemma 9.3.10: The identity holds:
(n
0
. . . . . n
p
) = (n
0
. . . . . n
m
)(n
m+1
. . . . . n
p
) .
Proof. Set :
l
= |:
l
. and let
l
. :
l
be the sequences associated to the weights
n
0
. . . . . n
m
through equations (9.3.5) and (9.3.11). Likewise, let
t
l
. :
t
l
be associated
to the weights n
m+1
. . . . . n
p
. and
tt
l
. :
tt
l
be associated to the composite j-tuple
n
0
. . . . . n
p
. Then from equation (9.3.11) we nd

tt
j
=
j
+
t
j

j

t
j
.
Substituting
j
=

i]j
:
i
and similar expresses for the primed and double primed
variables gives

l]j
:
tt
l
=

l]j
:
l
+

l]j
:
t
l

k]j

l]j
:
k
:
t
l
.
Using this equation an induction argument gives
(9.3.13) :
tt
j
= :
j
+:
t
j

lcm(k,l)=j
:
k
:
t
l
.
Now we write the :
j
s in terms of the :
j
s, and multiply equation (9.3.13) by
j
and sum over ,. This gives
(9.3.14)

:
tt
j

j
=

:
j

j
+

:
t
j

:
k

:
t
l

.
2
We use the notation n m(k) or n = m mod k for arithmetics mod k.
9.3. THE MILNOR FIBRATION AND THE TOPOLOGY OF LINKS 307
Here we have used equation (9.3.7). The lemma now follows easily from equation
(9.3.14).
We need one more lemma to nish the proof of the theorem. Namely,
Lemma 9.3.11:
(n
0
. . . . . n
n
) =
n

i=0

u
i

i
1

.
Proof. The proof is by induction on n. For n = 0 we notice from equation
(9.3.11) that

j
(n
0
) =

u
0
v
0
. if n
0
[,;
0. otherwise.
Thus, :
j
satises
:
j
(n
0
) =

u
0
v
0
. if , = n
0
;
0. if , = n
0
.
This shows that
(n
0
) = (1 :
j
(n
0
)
j
) =

1

u0

=

u0

0
1 .
which proves the result for n = 0. Now an easy induction using Lemma 9.3.10
proves the result.
This completes the proof of Theorem 9.3.9.
The Milnor-Orlik algorithm computes the numbers :
j
in terms of (w; d)-data
or, alternatively, in terms of (u; v)-data. In fact, we have
Corollary 9.3.12:
div =

(1)
n+1s
n
i1
n
is

i1

is
lcm(n
i1
. . . . . n
is
)

lcm(u
i
1
,...,u
is
)
.
where the sum is taken over all the 2
n+1
subsets i
1
. . . . . i
s
of 0. . . . . n.
According to the exact sequence (9.3.1) the (n 1)
st
and n
th
Betti numbers
/
n
(1
f
) = /
n1
(1
f
) of the link equals the number of factors of (t1) in the Alexander
polynomial (t). or equivalently the order of vanishing of (t) at t = 1. Thus, using
Corollary 9.3.12 we get
Corollary 9.3.13: Let 1
f
1

(w; d) and let (t) be its Alexander polynomial.

Then
(9.3.15) /
n1
(1
f
) =

(1)
n+1s
n
i1
n
is

i1

is
lcm(n
i1
. . . . . n
is
)
.
where the sum is taken over all the 2
n+1
subsets i
1
. . . . . i
s
of 0. . . . . n.
Working with the Alexander polynomial, the characteristic polynomial of the
monodromy map, entails working over C. that is studying the map
(t1l /

) : H
n
(1. C)H
n
(1. C) .
An alternative procedure to the one taken above would be to diagonalize (t1l /

),
that is nd unimodular matrices l(t) and \ (t) such that
(9.3.16) l(t)(t1l /

)\ (t) = diag

:
1
(t). . . . . :

(t)

.
308 9. LINKS AS SASAKIAN MANIFOLDS
The minimal polynomial :

(t) contains the same information as the characteristic

polynomial (t). In order to handle torsion, however, one needs to work with the
integral polynomial ring Z[t] which is not a principal ideal domain. This poses
several diculties which have still not been entirely overcome. Nevertheless, there
is a conjecture due to Orlik [Orl72a] saying that the matrices l(t) and \ (t) can
be taken with values in the ring Z[t]. This implies
Conjecture 9.3.14: H
n1
(1
f
. Z) = Z
m
1
(1)
Z
m

(1)
. where Z
1
and Z
0
are
the trivial and innite cyclic groups, respectively.
Given an index set i
1
. . . . . i
s
we will denote by 1 all its 2
s
subsets and by
J all its proper subsets. For each ordered subset i
1
. . . . . i
s
0. 1. . . . . n; i
1
<
i
2
< < i
s
one denes inductively the set of 2
s
positive integers, starting with
c

= gcd(n
0
. . . . . n
n
) :
(9.3.17) c
i
1
,...,i
s
=
gcd(n
0
. . . . . n
i1
. . . . . n
is
. . . . . n
n
)

J
c
j1,...jt
.
In addition, starting with /

= c
n+1
. one denes
(9.3.18)
/
i
1
,...,i
s
c
ns+1

i
1
,...,i
s
= c
ns+1

I
(1)
st
n
j
1
n
j
t

j
1

j
t
lcm(n
j
1
. . . . . n
j
t
)
.
where
c
ns+1
=

0. if n : + 1 is even;
1. if n : + 1 is odd.
For any 1 , : = max/
i1,...,is
| we set
(9.3.19) d
j
=

ki
1
,...,i
s
j
c
i
1
,...,i
s
.
The Conjecture 9.3.14 can now be stated more precisely
Conjecture 9.3.15: H
n1
(1
f
. Z)
tor
= Z
d
1
Z
d
r
.
We refer to the above computation of H
n1
(1
f
. Z)
tor
as the Orlik Algorithm.
Remark 9.3.2: It is useful to think of the Orlik Algorithm as follows: Consider
the indexing set = 0. . . . . n. For each subset (including the empty
set and itself) Orlik denes 2
]]
= 2
n+1
pairs of numbers: (c
A
.
A
). The
numbers c
A

A
are all positive integers and they are dened inductively from
c

= gcd(n
0
. . . . . n
n
). (We set c

= 1 though this does not matter.) The numbers

A
are, in general, fractions. Starting with

= /
2
(1
f
) one sets some
A
s
equal to zero depending on the cardinality [[ dening
/
A
=

1 (1)
]\A]

A
.
The numbers c
A
are then used to dene the orders of the cyclic groups that enter
as torsion and the numbers /
A
determine the powers in the prime number de-
composition of the torsion. It is equally important to observe that for the torsion
computation only the pairs (c
A
. /
A
) with /
A
1 matter, so in practice starting
with 2
n+2
numbers we end up with very few that are needed for the calculation of
the relevant homology group.
9.3. THE MILNOR FIBRATION AND THE TOPOLOGY OF LINKS 309
The formula (9.3.19) gives a simple way of computing the relevant torsion group
in all cases that Orliks conjecture is known to be valid. For example, we have the
following [Ran75, OR77]
Proposition 9.3.16: Conjecture 9.3.15 is true in the following cases:
(i) In dimensions 3, i.e., for n = 2.
(ii) For 1(z) = .
a0
0
+ +.
a
n
n
. i.e., the so-called Brieskorn-Pham polynomials.
(iii) For 1(z) = .
a
0
0
+.
0
.
a
1
1
+ +.
n1
.
a
n
n
.
(iv) For certain complete intersections of Brieskorn varieties given by, what
Randell calls, generalized Brieskorn polynomials [Ran75].
Exercise 9.1: [See Section 9.3.2 for the discussion of Brieskorn-Pham links and
notation conventions.] Use Orliks algorithm to show that if 1
f
is the Brieskorn-
Pham link 1
f
= 1(21. 52. 55. 900) then H
2
(1
f
. Z) = Z
24
(Z
7
)
12
(Z
11
)
6
(Z
13
)
8
.
Even more interesting is the BP link 1
g
= 1(78. 170. 627. 2261). Show that Orliks
algorithm gives H
2
(1
g
. Z) = Z
1558
(Z
5
)
36
(Z
7
)
2
(Z
11
)
16
(Z
13
)
288
. In spite of
what one might think we did not come up with these links by a random computer
search. Can you explain how we chose them and come up with your own examples?
You will need to consider the Brieskorn graph G(a) for these two links (which will
shortly be dened in Section 9.3.2) to answer this question which eventually leads
to Theorem 10.3.5. Can one get links (BP or otherwise) which have more than the
of four dierent cyclic groups to some even powers as torsion in H
2
(1
f
. Z)?
Exercise 9.2: Suppose that n = 2 and 1 1

(n
0
. n
1
. n
2
; d). Show that Or-
lik Algorithm reduces to the genus formula (4.6.5), i.e., we have o(
(w;d)
) =
/
1
(1
f
) =
012
. Consider next the n = 3 case and 1 1

(n
0
. n
1
. n
2
. n
3
; d).
Show that Orliks formula (9.3.19) for the torsion in H
2
(1
f
. Z) depends only on
the numbers /
012
. /
013
. /
023
. /
123
and c
012
. c
013
. c
023
. c
123
. Further show that
/
ijk
= 2o(
( w;

d)
). where
w =
(n
i
. n
j
. n
k
)
gcd(n
j
. n
j
. n
k
)
.

d =
d
gcd(n
j
. n
j
. n
k
)
.
0 i < , < / 3. and o(
( w;

d)
). is calculated with the formula (4.6.5). Could

d be fractional? Show by example that it does not follow that all /

ijk
s in Orlik
Algorithm are integers. Explain how (4.6.5) can give a non-integer result.
Open Problem 9.3.1: Prove or disprove the Orlik Conjecture 9.3.15. For n = 3
the conjecture basically follows from Kollars Theorem 4.7.14 and Exercise 9.2 (cf.
Theorem 10.3.5).
To end this section we consider branched covers constructed as the link 1
g
of
the polynomial
(9.3.20) o = .
k
0
+1(.
1
. . . . . .
m
) .
where 1 is a weighted homogeneous polynomial in n variables with an isolated
singularity at the origin. Then 1
g
is a /-fold branched cover of o
2m+1
branched
over the link 1
f
. The degree of 1
g
is d
g
= lcm(/. d
f
). and the weight vector is
w
g
=

d
f
gcd(k,d
f
)
.
k
gcd(k,d
f
)
w
f

. We shall always assume that / 2, since the linear

case / = 1 is a hyperplane in a weighted projective space. The following theorem,
which is Theorem 7.1 of [BGN03c], is a reformulation of results of Savelev [Sav79].
It is a generalization of part (i) of the well-known Brieskorn Graph Theorem 9.3.18
below.
310 9. LINKS AS SASAKIAN MANIFOLDS
Theorem 9.3.17: Let 1(.
1
. . . . . .
m
) be a weighted homogeneous polynomial of de-
gree d and weights w = (n
1
. . . . . n
m
) in C
m
with an isolated singularity at the
origin. Let / Z
+
and consider the link 1
g
of the equation
o = .
k
0
+1(.
1
. . . . . .
m
) = 0 .
Write the numbers
d
w
i
in irreducible form
u
i
v
i
. and suppose that gcd(/. n
i
) = 1
for each i = 1. . . . . :. Then the link 1
g
has weights
(d,kw)
gcd(k,d)
and degree lcm(/. d).
Furthermore, 1
g
is a rational homology sphere with the order [H
m1
(1
g
. Z)[ =
/
b
m2
(L
f
)
. where /
m2
(1
f
) is the (:2)
nd
Betti number of 1
f
.
9.3.2. Links of Brieskorn-Pham Type. Here we discuss in more detail
links of polynomials of Brieskorn-Pham type (9.3.16) (abbreviated BP type) which
are of particular interest in the case of homology spheres. We also discuss when BP
type links admit regular Sasakian structures. In the case of BP polynomials it is
more convenient to describe the links in terms of the Brieskorn exponents written
as the components of vector, viz. a = (o
0
. . . . . o
n
) Z
n+1
+
. We then write 1(z) =
.
a
0
0
+ +.
a
n
n
and denote the associated link by 1
f
= 1(a)
3
. The relation between
the exponents and the weights is now simply that d = o
i
n
i
for all i = 0. . . . . n.
where d is the degree of 1. Hence, on a BP link 1
f
1

(w; d) the exponents a are

the non-trivial part of the (u; v)-data, with u = a and v = (1. . . . . 1).
Let us recall the famous Brieskorn Graph Theorem
4
[Bri66, Dim92]. To
the vector a one associates a graph G(a) whose n + 1 vertices are labelled by
o
0
. . . . . o
n
. Two vertices o
i
and o
j
are connected if and only if gcd(o
i
. o
j
) 1. Let
G
ev
(a) G(a) denote the connected component of G(a) determined by the even
integers. Note that all even vertices belong to G
ev
(a). but G
ev
(a) may contain odd
vertices as well. Then we have
Theorem 9.3.18: The following hold:
(i) The link 1(a) is a rational homology sphere if and only if either G(a)
contains at least one isolated point, or G
ev
(a) has an odd number of
vertices and for any distinct o
i
. o
j
G
ev
(a). gcd(o
i
. o
j
) = 2.
(ii) The link 1(a) is an integral homology sphere if and only if either G(a)
contains at least two isolated points, or G(a) contains one isolated point
and G
ev
(a) has an odd number of vertices and for any distinct o
i
. o
j

G
ev
(a). gcd(o
i
. o
j
) = 2.
Proof. The proof of (i) is given in detail in [Dim92], and that of (ii) in
[Bri66].
In either case a very simple property of the graph G(a) conveys a lot of infor-
mation about the topology of 1(a). According to the Proposition 9.2.4 the sign of
the Sasakian structure on any BP link 1(a) is determined as follows
Corollary 9.3.19: The link 1(a) admits a Sasakian structure that is
(i) positive if and only if

1
ai
1 .
(ii) null if and only if

1
a
i
= 1 .
3
Note that each L(a) is a member of a family L

(w; d), where the pair (w; d) is uniquely

determined by a. That is why we use the asterisk to avoid the confusion.
4
The fascinating history of this theorem has recently come to light in an interesting historical
account by Brieskorn [Bri00]. The graph theorem for homotopy spheres was conjectured by Milnor
in a letter (published in [Bri00]) to John Nash who showed the letter to Brieskorn.
9.3. THE MILNOR FIBRATION AND THE TOPOLOGY OF LINKS 311
(iii) negative if and only if

1
a
i
< 1 .
In particular, BP links provide a rich source of all three types of Sasakian
geometry. Let us consider several important examples of BP links.
Example 9.3.20: Let a = (o
0
. . . . . o
n
) = (2. . . . . 2). Here we have the degree d of
1(z) is 2. and weights are all 1. i.e., n
0
= = n
n
= 1. Thus, by Proposition 9.3.5
the Milnor number j = 1. According to Theorem 9.3.18 1(a) is a rational homology
sphere if and only if n is even. Let us compute the Alexander polynomial (t) for
this link from the Milnor-Orlik procedure described above. First we note that
(
2
1)
2
= (2 2)
2
+ 1 = 1 .
Thus, we have
div =

2
1. if n is even;
1. if n is odd.
So the Alexander polynomial is
(t) =

t + 1. if n is even;
t 1. if n is odd.
This gives
H
n1
(1(a). Z) =

Z
2
. if n is even;
Z. if n is odd.
In this case the link 1(a) = oO(n + 1)oO(n 1) is the Stiefel manifold
\
2
(R
n+1
) of oriented 2-frames in R
n+1
. It is a homogeneous space and it carries
oO(n + 1)-homogeneous Sasaki-Einstein structure discussed in Theorem 11.4.1.
Example 9.3.21: Consider a = (o
0
. . . . . o
n
) = (/. 2. . . . . 2) with n 2. The cor-
responding BP link 1(a) is a /-fold cyclic branched cover of o
2n1
branched over
the link of the quadric .
2
1
+ + .
2
n
. Now we have j = / 1. The divisor of the
Alexander polynomial is
div =

k
1. if n is even;

2k

k

2
+ 1. if n and / are odd;

k

2
+ 1. if n is odd and / is even.
Hence,
(t) =

t
k1
+t
k2
+ +t + 1. if n is even;
t
k1
t
k2
+ t + 1. if n and / are odd;
(t 1)(t
k2
+t
k4
+ +t
2
+ 1). if n is odd and / is even.
An additional calculation of the torsion group with 9.3.15 yields the following an-
swer
H
n1
(1(a). Z) =

Z
k
. if n is even;
0. if n and / are odd;
Z. if n is odd and / is even.
In the rst case 1(a) is a rational homology sphere whose homotopy type changes
with /. The second case is more interesting. Here 1(a) is always a homology and,
hence, a homotopy sphere. But as we shall soon see the smooth structure does
depend on /. The third case is perhaps the most interesting as both the topology
312 9. LINKS AS SASAKIAN MANIFOLDS
and the smooth structure depend on /. We shall discuss the dierential topology
of these examples at the end of Section 9.4.
Another case of interest for Brieskorn-Pham links is when the weights are pair-
wise relatively prime. There are many such examples. For instance, for any rela-
tively prime pair of positive integers j. c one can take a = (j. c. jc. . . . . jc). Then
the weight vector is w = (c. j. 1. . . . . 1). Clearly, Z
f
for this BP link is a smooth
manifold. More generally, we have
Proposition 9.3.22: Let 1(a) be a BP link with pairwise relatively prime weights
w = (n
0
. . . . . n
n
). Then 1(a) is the total space of a circle bundle over a smooth
projective algebraic variety Z
a
and the induced Sasakian structure is regular.
We end this section with example of a construction of locally Sasakian struc-
tures.
Example 9.3.23: Let 1(a) be a link of a BP polynomial 1(z) = .
a
0
0
+ + .
an
n
.
Consider the involution on C
n+1
dened by complex conjugation z z. This
involution restricts to an involution on 1(a) which sends the Sasakian structure
o dened by Proposition 9.2.2 to its conjugate o
c
. Moreover, if all the o
i
s are
even then since all the coecients of 1(z) are positive, the involution acts freely
on 1(a). So the quotient 1(a) is a manifold which is not Sasakian, but it is
locally Sasakian. Notice that 1(a) is orientable if and only if n is even. The
simplest examples are the k-fold branched covers of quadrics of Example 9.3.20
with / even. In particular, considering / = 2 we see that 1(a) is just the Stiefel
manifold \
2
(R
n+1
) with its homogeneous Sasaki-Einstein metric. So \
2
(R
n+1
)
admits an Einstein metric that is locally Sasakian, but not Sasakian.
9.4. The Dierential Topology of Links
The Milnor Fibration Theorem 9.3.1 implies that links of isolated hypersurface
singularities are highly connected manifolds which are the boundaries of a paral-
lelizable manifold. Hence, they are (2n 1)-dimensional closed oriented manifolds
whose only non-vanishing homology groups occur in dimensions 0. n 1. n. 2n 1.
The most obvious examples of such manifolds are the homotopy spheres, so we
shall begin there. In 1956 Milnor [Mil56b] created the eld of dierential topol-
ogy by constructing exotic dierential structures on o
7
. Later the seminal work of
Kervaire and Milnor [KM63] (see also [Hir67]) showed that associated with each
sphere o
n
with n 5 there is an Abelian group
n
consisting of equivalence classes
of oriented homotopy spheres o
n
that are equivalent under oriented h-cobordism
which by Smales h-cobordism theorem [Sma62b, Mil65] implies equivalence un-
der oriented dieomorphism. More explicitly Kervaire and Milnor proved
Theorem 9.4.1: For n 5 the set
n
of equivalence classes of oriented homotopy
spheres of dimension n-form a semi-group under the connected sum operation whose
Grothendieck group completion forms a nite Abelian group, also denoted
n
.
Kervaire and Milnor also dened the subgroup /1
n+1
of
n
which is of great
importance to us.
Theorem/Denition 9.4.2: The subset /1
n+1
of
n
consisting of those homo-
topy spheres which bound parallelizable manifolds form a subgroup of
n
of nite
index, and the short exact sequence
0/1
n+1

n
/1
n+1
0
9.4. THE DIFFERENTIAL TOPOLOGY OF LINKS 313
splits. Furthermore, the quotient group
n
/1
n+1
injects into
n
(o)J(oO). and
is an isomorphism if n = 2
j
2 for some ,.
Here
n
(o) is the n
th
stable homotopy group of spheres and J is the well-known
Hopf-Whitehead J homomorphism, cf. [KM63, Rav86].
Kervaire and Milnor [KM63] also were able to determine /1
n+1
in most cases.
They proved that /1
2m+1
= 0 for : 1. that /1
4m+2
is either 0 or Z
2
, and that
for : 2. /1
4m
is cyclic of order
[/1
4m
[ = 2
2m2
(2
2m1
1) numerator

41
m
:

.
where 1
m
is the :
th
Bernoulli number dened by the power series
.
c
z
1
= 1
.
2
+
1
1
2!
.
2

1
2
4!
.
4
+ .
The numerator of the rational number
4Bm
m
is tabulated in [MS74] up to : = 12.
For /1
4m+2
the situation is still not entirely understood. It entails computing
the Kervaire invariant, which is hard. It is known (see the recent review paper
[Lan00] and references therein) that /1
4m+2
= Z
2
if 4:+2 = 2
i
2 for any i 3.
Furthermore, /1
4m+2
vanishes for : = 1. 3. 7. and 15. The table below lists what
is known about the two groups in rst few dimensions.
n
n
/1
n+1
5 0 0
6 0 0
7 Z
28
Z
28
8 Z
2
0
9 Z
2
Z
2
Z
2
Z
2
10 Z
6
0
11 Z
992
Z
992
12 0 0
13 Z
3
0
14 Z
2
0
15 Z
2
Z
8128
Z
8128
16 Z
2
0
17 Z
2
Z
2
Z
2
Z
2
Z
2
18 Z
2
Z
8
0
19 ? Z
130816
Let us make several remarks about exotic structures on spheres and the above
table. First, homeomorphism implies dieomorphism in dimensions 1. 2. 3. and so
Perelmans recently conrmed proof of the Poincare conjecture implies
3
= 0.
However, in four dimensions the problem remains completely open. The so-called
smooth Poincare conjecture asserts that there are no exotic 4-spheres
5
. In higher
5
Apparently there is no consensus whether the conjecture is true or false. Four is the only
dimension where innitely many exotic structures can and do exist on the same topological man-
ifold. Uncountably many are known to exist on R
4
! See the recent book [Sco05] for a discussion
of what is known about the wild world of 4-manifolds.
314 9. LINKS AS SASAKIAN MANIFOLDS
dimensions the computation of
n
depends on the higher homotopy groups of
spheres and those, in general, are not known.
Orientation is typically not a part of the denition of a smooth structure on a
manifold. Taking the group inverse in
n
corresponds to reversing the orientation.
Hence, to a dierential geometer, in dimension 7 there are only 15 smooth structures
on spheres. The standard o
7
and one exotic 7-sphere admit orientation reversing
dieomorphisms. The other 13 do not admit such a dieomorphism and, hence, they
are counted twice in
7
. In dimension 9, however, there are indeed 8 dierent
smooth structures. Let
n
and
n
be two distinct elements in
n
. Whether or
not one needs to distinguish the two depends on the context. For instance, if one
wants to establish existence (or non-existence) of some special Riemannian metric
on
n
any such result would automatically extends to
n
. Then it is natural to
forget the orientation. On the other hand, on a Sasakian manifold orientation is
xed, hence a part of the Sasakian structure. A construction of a Sasakian structure
on
n
does not automatically give a Sasakian structure on
n
(cf. Remark 6.1.2).
In dimensions n = 8/+1 and n = 8/+2 there is another interesting subgroup of

n
: the group of homotopy n-spheres which bound spin manifolds
spin
n

n
. The
subgroup
spin
n
has index 2. The coset of this group, i.e., all homotopy spheres which
are not boundaries of spin manifolds is quite remarkable. In 1974 Hitchin [Hit74]
introduced the so-called -invariant, showing that (`) = 0 is an obstruction to
` admitting any metric of positive scalar curvature. He observed that his invariant
does not vanish for half of the exotic spheres in dimensions 1 and 2 mod 8. These
exotic spheres are exactly the spheres which do not bound a spin manifold. For
example, in dimension 9 we get
spin
9
= Z
2
Z
2
so that 4 exotic 9-spheres do not
admit any metric of positive scalar curvature. In dimension 10 one has three such
exotic spheres; two forgetting the orientation.
The computation of
9
and
17
is based on a paper of Brumel [Bru70], where
he shows that
4m+1
= Z
2
(
S
4m+1
imJ) if 4:+2 = 2
j
2. Here J :
n
(oO)
S
n
is the Adams J-homomorphism. As pointed out by Linus Kramer the relevant
stable homotopy groups of spheres are known and it follows that
S
9
imJ = Z
2
Z
2
and
S
17
imJ = Z
2
Z
2
Z
2
.
Ten years after Milnors landmark paper, Brieskorn [Bri66] showed how all
odd-dimensional exotic spheres in the /1
n+1
subgroup can be obtained as links of
isolated hypersurface singularities. In fact, all of such spheres can be realized as
very simple BP links. Let us recall Brieskorns result. We begin with an arbitrary
BP link 1(a) of dimension at least 5. The Brieskorn Graph Theorem 9.3.18 tells
us when 1(a) is a topological sphere. Suppose 1(a) is homeomorphic to o
2n1
. It
follows that 1(a) /1
2n
is a boundary of a parallelizable manifold which is the
Milnor ber 1 = \
2n
a
. The determination of the smooth structure of 1(a) = \
2n
a
depends on the parity of n. We being with the simpler case of n = 2: + 1 odd.
Then is /1
4m+2
either 0 or Z
2
. When /1
4m+2
= 0. without any further eort, one
concludes that all such BP links 1(a) /1
4m+2
are standard spheres. Such is the
case, for example, when n = 3. 7 (: = 1. 3). Brieskorn provides a complete answer
when /1
4m+2
= Z
2
and a partial answer when /1
4m+2
is not known.
Theorem 9.4.3: Let 1(a) be a topological sphere and n = 2: + 1 5. The link
1(a) is the so-called Kervaire sphere
4m+1
if and only if the graph G(a) consists
of one isolated point o
k
and another connected component and o
k
3(8). In all
9.4. THE DIFFERENTIAL TOPOLOGY OF LINKS 315
other cases 1(a) is dieomorphic to the standard sphere. Furthermore, the Kervaire
sphere
4m+1
is exotic when /1
4m+2
= Z
2
.
The above theorem separates topological spheres among BP links 1(a)
/1
2m+2
into the standard and exotic every time /1
4m+2
= Z
2
. But even when
/1
4m+2
is not known Theorem 9.4.3 still separates standard spheres from the po-
tentially exotic Kervaire spheres.
Denition 9.4.4: Consider the polynomials
(9.4.1) 1(z) = .
6k1
0
+.
3
1
+.
2
2
+ +.
2
2m
. o(z) = .
2k+1
0
+.
2
1
+.
2
2
+ +.
2
2m+1
.
with / 0. We denote the BP links associated with the above polynomials by
1(6/ 1. 3. 2. . . . . 2) =
4m1
k
. 1(2/ + 1. 2. . . . . 2) =
4m+1
k
.
Example 9.4.5: 1(3. 2. 2. 2) =
5
3
is a 5-sphere and it is a Kervaire sphere. But
there are no exotic 5-spheres so that 1(3. 2. 2. 2) o
5
as is any other topological
5-sphere. 1(3. 2. 2. 2. 2. 2) is Kervaire and Theorem 9.4.3 implies that it is the non-
trivial element in /1
10
= Z
2
. In particular, we get

4m+1
k
= 1(2/ + 1. 2. . . . . 2)

o
4m+1
. if 2/ + 1 = 1 mod8 ;

4m+1
. if 2/ + 1 = 3 mod8 .
and the Kervaire spheres
4m+1
are always exotic when /1
4m+2
= 0.
To analyze
4m1
k
we need more information. Suppose 1(a) is a topological
(4:1)-sphere and an element in /1
4m
.
Theorem 9.4.6: Consider two BP links 1(a) = \
4m
a
. 1(b) = \
4m
b
. There
exists an orientation preserving dieomorphism 1(a) 1(b) if and only if
(\
4m
a
) = (\
4m
b
) mod [/1
4m
[ .
where (\
4m
a
) is the Hirzebruch signature. In particular, 1(a) is dieomorphic to
the standard sphere when (\
4m
a
) = 0 mod [/1
4m
[.
Thus we only need to compute the Hirzebruch signature of the Milnor ber
1 = \
4m
a
. Brieskorn derives a simple combinatorial index formula for
(9.4.2) (\
4m
a
) =
+
(a)

(a).
where

+
(a) = #

x Z
2m+1
[ 0 < r
i
< o
i
and 0 <
2m

i=0
r
i
o
i
< 1 mod 2

(a) = #

x Z
2m+1
[ 0 < r
i
< o
i
and 1 <
2m

i=0
r
i
o
i
< 2 mod 2

.
In principle, calculating the signature amounts to counting points in a nite subset
of an integral lattice whose size depends on a. For arbitrary very large exponents
this can be a challenging task. We remark that using a formula of Eisenstein, Zagier
(cf. [Hir71]) has rewritten this as:
(9.4.3) (\
4m
(a)) =
(1)
m

N1

j=0
cot
(2, + 1)
2
cot
(2, + 1)
2o
0
cot
(2, + 1)
2o
2m
.
where is any common multiple of the o
i
s.
316 9. LINKS AS SASAKIAN MANIFOLDS
Example 9.4.7: Consider the BP links
4m1
k
= 1(6/ 1. 3. 2. . . . . 2) /1
4m
with
: 1. For any / 1 the Brieskorn Graph Theorem 9.3.18 says this is a topological
sphere as G(a) has two isolated points. Let 1(6/ 1. 3. 2. . . . . 2) = \
4m
k
An
elementary calculation with (9.4.2) shows that
(\
4m
k
) = (1)
m
8/ .
Both 1(5. 3. 2. . . . . 2) and 1(7. 3. 2. . . . . 2) represent the so-called Milnor generator
in /1
4m
. In particular, 1(6/ 1. 3. 2. . . . . 2) #
k

4m1
1
realizes all elements in
/1
4m
by simply taking 1 / [/1
4m
[.
Summarizing Examples 9.4.5 and 9.4.7 gives the following result Brieskorn
[Bri66]:
Theorem 9.4.8: For n 3 every homotopy sphere /1
2n
can be realized as
(i)
4m1
k
. / = 1. . . . . [/1
4m
[, when n = 2:
(ii)
4m+1
k
, / = 1. 3, when n = 2:+ 1.
In particular,
4m1
]bP4m]
and
4m+1
3
are dieomorphic to the standard spheres.
In analogy with the Kervaire-Milnor group /1
2n
. Durfee [Dur77] dened the
group 11
2n
. Actually, he rst dened this as a semigroup in [Dur71], and later in
[Dur77] with the same notation denoted the corresponding Grothendieck group.
Thus, we have
Denition 9.4.9: For n 3 let o11
2n
denote the semigroup of dieomorphism
classes of closed oriented (n2)-connected (2n1)-manifolds that bound paralleliz-
able manifolds, and let 11
2n
denote its Grothendieck completion.
As with /1
2n
multiplication in o11
2n
is the connected sum operation, and the
standard sphere is a two-sided identity. Thus, o11
2n
is a monoid. Furthermore,
for n 3 the Kervaire and Milnor group /1
2n
of Denition/Theorem 9.4.2 becomes
a subgroup of 11
2n
by Smales famous h-cobordism theorem [Sma62b, Mil65].
Unless otherwise stated we shall heretofore assume that n 3. We are mainly
interested in those highly connected manifolds that can be realized as links of
isolated hypersurface singularities dened by weighted homogeneous polynomials,
that is by the set \H1
2n1
of Denition 9.2.1. By (i) and (ii) of Theorem 9.3.2
one sees that there is a map
: \H1
2n1
11
2n
which is the composition of the inclusion \H1
2n1
o11
2n
with the natural
semigroup homomorphism o11
2n
11
2n
. The image (\H1
2n1
) is a subset
of 11
2n
, and by (iii) of Theorem 9.3.2 it is a proper subset at least when n is
even. This can be contrasted with /1
2n
which according to Theorem 9.4.8 satises
/1
2n
(\H1
2n1
) = /1
2n
if n 3. Notice, however, that (\H1
2n1
) is not
generally a submonoid.
We now discuss invariants that distinguish elements of 11
2n
. First, by Poincare
duality the only non-vanishing homology groups occur in dimension 0. n 1. n and
2n 1. Moreover, H
n
(1. Z) is free and rank H
n
(1. Z) = rank H
n1
(1. Z). Thus,
our rst invariant is the rank of H
n1
(1. Z). so we dene
(9.4.4) 11
2n
(/) = 1 11
2n
[ rank H
n1
(1. Z) = / .
9.4. THE DIFFERENTIAL TOPOLOGY OF LINKS 317
Now 11
2n
is graded by its rank, so we have
(9.4.5) 11
2n
=

k
11
2n
(/) .
and the monoid multiplication denes a map
(9.4.6) 11
2n
(/
1
) 11
2n
(/
2
)11
2n
(/
1
+/
2
) .
Note that 11
2n
(0) is the submonoid of highly connected rational homology spheres.
The remaining known invariants [Wal67b, Dur71, Dur77] are a linking form
on the torsion subgroup of H
n1
(1) and a quadratic invariant on the 2n-manifold
whose boundary is 1. The precise nature of these invariants depends on whether
n is even or odd. For the case n even Durfee [Dur77] shows that for n 3 and
n = 4. 8 there is an exact sequence
(9.4.7) 0/1
2n
11
2n

Z 1Q(Z)0 .
where 1Q(Z) denotes the Grothendieck group of regular bilinear form modules over
Z. Let us describe the map . The projection onto the rst factor is just the rank of
H
n1
(1) while the projection onto the second factor is Walls quadratic form [Wa2]
which is essentially the classical linking form / on the torsion subgroup of H
n1
(1).
Any two manifolds 1
1
. 1
2
11
2n
such that (1
1
) = (1
2
) dier by a homotopy
sphere, i.e., there is /1
2n
such that 1
2
1
1
#. where means dieomorphic.
But by Theorem 9.4.6 the elements /1
2n
for n even are determined by the
signature of \. This completes the dieomorphism classication for n = 4. 8 even.
The cases n = 4. 8 are more complicated [Wil72, Cro01]. Now, in addition to
the group H
n1
(1) and the linking form /. there is an obstruction cocycle


H
n
(1.
n1
(oO)) H
n
(1. Z). The tangent bundle of 1 restricted to the (n 1)-
skeleton is trivial and

gives the obstruction to triviality on the n-skeleton. If the
torsion subgroup of H
n1
(1) has odd order, then up to decomposability these are
all the invariants. However, if the torsion subgroup of H
n1
(1) has even order,
things are even more complicated, and the analysis in [Wil72] was not complete. It
was recently completed in [Cro01]. The important point for us is that if the torsion
subgroup of H
n1
(1) vanishes, 1 is determined completely up to homeomorphism
by the rank of H
n1
(1). Summarizing we have
Theorem 9.4.10: Let ` be a highly connected manifold in 11
4n
such that
H
2n1
(`. Z) = Z
k
. Then ` is dieomorphic to /#(o
2n1
o
2n
)#
4n1
for some

4n1
/1
4n
.
Notice that by Theorem 7.4.11 /#(o
2n1
o
2n
)#
4n1
can admit a Sasakian
structure only if / is even. So by (iii) of Theorem 9.3.2 11
4n
(2/+1)\H1
4n1
=
. whereas we shall see that 11
4n
(2/) \H1
4n1
= for all /.
For the case n odd the dieomorphism classication was obtained by Wall
[Wal67b], but for our purposes, the presentation in [Dur71] is more convenient.
Let 1 11
2n
with 1 = \. where \ can be taken as (n 1)-connected and
parallelizable. In this case the key invariant is a Z
2
-quadratic form
: H
n
(\. Z)2H
n
(\. Z)Z
2
dened by Wall [Wal62] as follows: Let A be an embedded n-sphere in \ that
represents a non-trivial homology class in H
n
(\. Z). and let [A] denote its image
318 9. LINKS AS SASAKIAN MANIFOLDS
in H
n
(\. Z)2H
n
(\. Z). Then ([A]) is the characteristic class in
ker

n1
(oO(n))
n1
(oO)

Z
2
of the normal bundle of A. Let rad denote the radical of . i.e., the subspace of
the Z
2
-vector space H
n
(\. Z)2H
n
(\. Z). where is singular. Then Durfee [Dur71]
(see also [DK75]) proves
Theorem 9.4.11: Let 1
i
11
2n
for i = 1. 2 with n 3 odd be boundaries of
parallelizable (n1)-connected 2n manifolds \
i
with Z
2
quadratic forms
i
. Suppose
that H
n1
(1
1
. Z) H
n1
(1
2
. Z). then
(i) if n = 3 or 7, then 1
1
and 1
2
are dieomorphic;
(ii) if the torsion subgroups of H
n1
(1
i
. Z) have odd order and
i
[rad
i
0
for i = 1. 2. then 1
1
1
2
#(c(
1
)+c(
2
)). where c is the Arf invariant
and is the Kervaire sphere, i.e., the generator of /1
2n
;
(iii) if the torsion subgroups of H
n1
(1
i
. Z) have odd order and
i
[rad
i
0
for i = 1. 2. then 1
1
1
2
1
2
#.
We end this section by discussing some results of Durfee and Kaufmann [DK75]
concerning the periodicity of branched covers. Let 1 o
2n+1
be a simple bered
knot or link (n 1), by which we mean an (n 2)-connected (2n 1)-embedded
submanifold of o
2n+1
for which the Milnor bration theorem holds. If 1 is the
Milnor ber of the bration : o
2n+1
1o
1
then the monodromy map / :
H
n
(1)H
n
(1) is a fundamental invariant of the link 1. Let 1
k
be a /-fold cyclic
branched cover of o
2n+1
branched along 1. Then Durfee and Kauman [DK75]
show that there is an exact sequence
(9.4.8) H
n
(1)
1l+h++h
k1
H
n
(1)H
n
(1
k
)0 .
So homologically 1
k
is determined by the cokernel of the map 1l +/ + +/
k1
.
Now suppose that 1 is a rational homology sphere and that the monodromy map
/ of 1 has period d. Then since 1l / is invertible, 1l + / + + /
d1
is the zero
map in (9.4.8), and this determines the homology of 1
d
. The following lemma is
due to Durfee and Kauman [DK75]
Lemma 9.4.12: Let 1 be a bered knot in o
2n+1
which is a rational homology
sphere such that the monodromy map has period d. Suppose further that 1
k
is a
/-fold cyclic cover of o
2n+1
branched along 1. Then
(i) H
n
(1
d
) H
n
(1) Z

. where j is the Milnor number of 1.

(ii) H

(1
k+d
) H

(1
k
) for all / 0.
(iii) H

(1
dk
) H

(1
k
) for all 0 < / < d.
Notice that (i) determines a large class of (n1)-connected (2n+1)-manifolds
whose middle homology group H
n
is free, and in certain cases this determines the
manifold up to homeomorphism. Items (ii) and (iii) give a homological periodicity.
Durfee and Kauman also show that there are both homeomorphism and dieo-
morphism periodicities in the case that n is odd and n = 1. 3. 7. In particular in
this case, when the link 1 is a rational homology sphere, 1
k+d
is homeomorphic
to 1
k
. To obtain the dieomorphism periodicity let
k
denote the signature of
the intersection form on the Milnor ber 1
k
. Again assuming that 1 is a rational
homology sphere and / has periodicity d. one nds that 1
k+d
is dieomorphic to

d+1
8
#1
k
. where

d+1
8
denotes

d+1
8
copies of the Milnor sphere . Here we state
the slightly more general Theorem 6.4 of Durfee [Dur77]:
9.5. POSITIVE SASAKIAN STRUCTURES ON LINKS 319
Theorem 9.4.13: For even n = 2. 4. 8 let 1
i
be (n 2)-manifolds that bound
parallelizable manifolds \
i
. with i = 1. 2. Suppose that the quadratic forms of 1
i
are isomorphic and H
n1
(1
1
. Z) H
n1
(1
2
. Z). Then (\
2
)(\
1
) is divisible by
8, and 1
2
is dieomorphic to 1
1
#
1
8
((\
2
)(\
1
)). where (\ ) is the Hirzebruch
signature of \.
Remark 9.4.1: Theorem 6.4 of [Dur77] as well as Theorem 5.3 of [DK75] ex-
clude the cases n = 4 and 8. However, it follows from [Wil72] and [Cro01] that
the dieomorphism classication still holds in these cases since the links we are
considering here have no element of even order in the torsion subgroup of H
n1
(In
fact the torsion subgroup vanishes in the case above). This remark also pertains to
the discussion for Theorem 3 below.
9.5. Positive Sasakian Structures on Links
The importance of positive Sasakian structures comes from Theorem 7.5.31
which says that they give rise to Sasakian metrics with positive Ricci curvature.
The existence of metrics of positive Ricci curvature on certain manifolds has been
an area of some interest. Most of the results obtained in this direction have used
the techniques of surgery theory. We mention here two such results. First there is
the work of Sha and Yang [SY91] who proved using surgery theory the existence
of metrics of positive Ricci curvature on the connected sum of products of spheres.
There is also the work of Wraith [Wra97] who proved that all homotopy spheres
that are boundaries of parallelizable manifolds admit a metric of positive Ricci
curvature. A proof of this result using the techniques elaborated in this book
was given in [BGN03c] and presented as Theorem 9.5.6 below. In [BGN03a]
a special case of the Sha-Yang result was established by showing the existence of
Sasakian metrics with positive Ricci curvature on the connected sums /(o
2
o
3
).
This and other examples on simply connected 5-manifolds will be discussed further
in Chapter 10. Of course, the existence of Sasaki-Einstein metrics as presented in
Chapter 11 also provides examples of Sasakian metrics of positive Ricci curvature.
We start with some applications of Theorem 7.5.31. First we can combine
Proposition 9.2.4 with Theorem 7.5.31 to obtain
Theorem 9.5.1: Let 1
f
be the link of an isolated hypersurface singularity of a
weighted homogeneous polynomial 1. and suppose that [w[ d 0. Then 1
f
admits
a Sasakian metric with positive Ricci curvature.
Next we have
Theorem 9.5.2: Let 1
f
be the link of a w.h.p. 1
t
(.
2
. . . . . .
n
) 1

(w
t
. d
t
). As-
sume that the origin in C
n1
is the only singularity so that 1
f
is smooth. Consider
the weighted homogeneous polynomial
1 = .
2
0
+.
2
1
+1
t
of degree d = lcm(2. d
t
). Then the link 1
f
admits a Sasakian structure with positive
Ricci curvature and /
n1
(1
f
) = /
n3
(1
f
).
Proof. There are two cases. If d
t
is odd then the weight vector of 1 is w =
(d
t
. d
t
. 2w
t
). whereas, if d
t
is even, then w = (
d

2
.
d

2
. w
t
). In the rst case we have
[w[ d = d
t
+ d
t
+ 2[w
t
[ 2d
t
= 2[w
t
[ 0. while in the second case [w[ d =
d

2
+
d

2
+[w
t
[d
t
= [w
t
[ 0. In either case 1
f
admits a Sasakian metric with positive
Ricci curvature by Theorem 9.5.1. The Betti number equality is actually well-known
320 9. LINKS AS SASAKIAN MANIFOLDS
and follows from the Sebastiani-Thom Theorem [ST71, KN77, Dim92], but for
completeness we give an independent derivation suited to our context. We consider
the divisor div
f
of the Alexander polynomial
f
of 1. By Theorem 9.3.9 we have
div
f
=

ui

i
1

= (
2
1)
2

t
i
1

t
i
1

= div
f
.
and for i = 2. . . . . n we have
d
w
i
=
d

i
. So n
i
= n
t
i
and
i
=
t
i
for the same range of
i. Thus, the Betti number equality follows from Corollary 9.3.12.
9.5.1. Positive Sasakian Structures on Homology Spheres. We begin
by studying homology spheres, both rational homology spheres and homotopy
spheres. An immediate corollary of Theorem 9.5.2 is
Corollary 9.5.3: Let 1
f
be the link of a weighted homogeneous polynomial that
is a rational homology sphere. Then
1 = .
2
0
+.
2
1
+1
t
is the link 1
f
of a rational homology sphere that admits a Sasakian structure with
positive Ricci curvature.
Example 9.5.4: Consider a BP link 1(a) with a = (/. j. . . . . j) for any relatively
prime positive integers /. j Z
+
. Brieskorn Graph Theorem 9.3.18 implies that this
is always a rational homology sphere as the graph G(a) has one isolated vertex. In
addition, it is never a homology sphere unless j = 2 and n odd which was already
discussed in the example 9.3.21 of the previous section. Using corollary 9.3.19 we
see that 1(a) is positive if and only if j(/1) < n/ and negative when j(/1) n/.
In either case one gets innitely many examples. As expected, we cannot get null
links as j(/ 1) = n/ when gcd(/. j) = 1. To see that one indeed gets innitely
many, one can take the simplest case of rational homology 5-spheres and compute
the torsion group
H
2
(1(a). Z) = (Z
k
)
(p1)(p2)
.
The only innite series of positive links we get is 1(/. 3. 3. 3). with / 3 prime to
3. 1(/. 3. 3. 3) is dieomorphic to the Barden prime manifold `
5
k
. These links will
be discussed further in Chapters 10 and 11 together with many other examples of
ve- and seven-dimensional links. Note that any 1(/. j. j. j) with j and / 1
prime to j is negative with the exception of 1(3. 4. 4. 4) whose torsion group equals
(Z
3
)
6
. Furthermore, 1(/. j. j. j) is a
1
2
(j 1)(j 2)-fold connected sum of `
5
k
.
Theorem 9.5.5: Every (2n 2)-connected (4n 1)-manifold that is the boundary
of a parallelizable manifold whose homology group H
2n1
(1. Z) is isomorphic to Z
3
admits Sasakian metrics with positive Ricci curvature. There are precisely 2[/1
4n
[
such smooth oriented manifolds.
Proof. This is essentially a corollary of Proposition 7.2 of [Dur77] where
Durfee considers the link 1
k
of the Brieskorn-Pham polynomial .
k
0
+ .
3
1
+ .
2
2
+
+.
2
n
for even n 4. He shows that H
n
(1
2
. Z) H
n
(1
4
. Z) Z
3
. but that 1
2
and 1
4
have inequivalent linking forms. Furthermore, 1
6l+2
is dieomorphic to
1
2
#(1)
n
2
|
4n1
and 1
6l+4
is dieomorphic to 1
4
#(1)
n
2
|
4n1
. where
4n1
=
1
5
is the Milnor generator.
9.5. POSITIVE SASAKIAN STRUCTURES ON LINKS 321
There are two distinct non-homeomorphic topological manifolds in Theorem
9.5.5 and they are distinguished by their linking form in H
2n1
(1. Z) Z
3
. Each
topological manifold is comprised of [/1
4n
[ distinct dieomorphism types.
For homotopy spheres we have the following result [BGN03c]:
Theorem 9.5.6: For n 3 let
2n1
be a homotopy sphere which can be realized
as the boundary of a parallelizable manifold. Then
2n1
admits Sasakian metrics
with positive Ricci curvature.
Proof. By Theorem 9.4.8 every homotopy sphere that is the boundary of a
parallelizable manifold can be realized as the link 1
f
of the Example 9.4.5. By
Theorem 9.5.2 any all such links admits Sasakian metrics with positive Ricci cur-
vature.
For more examples of positive Sasakian structures on homotopy spheres see
[BGN03c].
Lopez de Medrano [LdM71] has shown that the quotient space of any xed
point free involution on a homotopy sphere is homotopy equivalent to a real pro-
jective space RP
n
. Here we are interested in xed point free involutions on
2n+1
that also preserve the Sasakian structure. In particular, we consider the well-known
involution T on
4m+1
p
dened by (.
0
. .
1
. . . . . .
2m+1
) (.
0
. .
1
. . . . . .
2m+1
). One
easily sees that the group 'T` generated by T acts freely on
4m+1
. so by Lopez
de Medranos theorem the quotient manifold
4m+1
p
'T` is homotopy equivalent
to RP
4m+1
. Furthermore, it is clear that 'T` is a subgroup of the weighted circle
group o
1
w
with weights w = (2. 2j + 1. . . . . 2j + 1). Thus, the deformation class of
Sasakian structures passes to the quotient
4m+1
p
'T` to give a deformation class
of Sasakian structures together with their characteristic foliation T

. Furthermore,
by Theorem 9.5.2 the Sasakian structures on
4m+1
p
'T` are positive and admit
Sasakian metrics of positive Ricci curvature. We have
Theorem 9.5.7: On each of the known 2
2m
oriented dieomorphism types of ho-
motopy projective spaces RP
4m+1
there exist deformation types of positive Sasakian
structures, and each deformation class contains Sasakian metrics of positive Ricci
curvature.
Proof. It is known [AB68, Bro73, Gif69a, Gif69b] that there are at least
2
2m
dieomorphism types on a homotopy real projective space of dimension 4:+1.
Atiyah and Bott [AB68] and Browder [Bro73] obtained the bound 2
2m1
which
was then extended to 2
2m
by Gien [Gif69a, Gif69b]. Furthermore, for 0 j <
j
t
< 2
2m
the homotopy real projective spaces
4m+1
p
'T` and
4m+1
p
'T` are not
dieomorphic. Thus, the
4m+1
p
'T` realize the 2
2m
dieomorphism types distin-
guished by Gien and from the discussion above each of these admit deformation
classes of positive Sasakian structures. In dimension 4 +1 = 5. the bound 2
21
= 4
is sharp. This proves the theorem.
9.5.2. Further Examples of Positive Sasakian Structures. Here we de-
scribe some results concerning the existence of positive Sasakian structures on man-
ifolds other than rational homology spheres. The next two theorems were described
in [BG06c].
Theorem 9.5.8: Let n 2 be an integer, then for each positive integer / there
exist Sasakian metrics with positive Ricci curvature in 1
n
(/) of the [/1
4n
[ oriented
322 9. LINKS AS SASAKIAN MANIFOLDS
dieomorphism classes of the (4n 1)-manifolds 2/#(o
2n1
o
2n
). where 1
n
(/)
is determined by an explicit formula given below.
Theorem 9.5.9: For each pair of positive integers integer (n. /) there exists a
(2n1)-connected (4n+1)-manifold 1 with H
2n
(1. Z) free of rank / which admits
a Sasakian metric of positive Ricci curvature. Furthermore, 1 is dieomorphic to
one of manifolds
#/(o
2n
o
2n+1
) . #(/ 1)(o
2n
o
2n+1
)#T . #/(o
2n
o
2n+1
)#
4n+1
.
where T = T
1
(o
2n+1
) is the unit tangent bundle of o
2n+1
. and
4n+1
is the Kervaire
sphere. For / = 1 the manifolds
o
2n
o
2n+1
. (o
2n
o
2n+1
)#
4n+1
. T
all admit Sasakian metrics with positive Ricci curvature. If n = 1 or n = 3 then
#/(o
2n
o
2n+1
) admits a Sasakian metric with positive Ricci curvature for all /.
Proof. The links that we need to prove Theorems 9.5.8 and 9.5.9 involve
Brieskorn polynomials of the form
(9.5.1) 1
p,q
= .
p
0
+.
q
1
+.
2
2
+ +.
2
n
.
The link associated with 1
p,q
is
1
p,q
= 1
1
p,q
(0) o
2n+1
.
By Theorem 9.5.2 all such links admit Sasakian metrics with positive Ricci curva-
ture. One can view 1
p,q
as a j-fold branched cover of o
2n1
branched over the link
1
q
dened by the polynomial
1
q
= .
q
1
+.
2
2
+ +.
2
n
.
For Theorem 9.5.8 we need the link 1
2(2k+1),2k+1
. i.e., j = 2(2/+1). c = 2/+1.
with n even. In this case the degree of 1
2(2k+1),2k+1
is d = 2(2/ + 1) which is the
period of the monodromy map of the link 1
2k+1
. Furthermore, 1
2k+1
is a homology
sphere which can be seen by computing its Alexander polynomial,
(t) =
t
2k+1
+ 1
t + 1
.
Now the link 1
2(2k+1),2k+1
is a 2(2/ + 1) branched cover of o
2n1
branched over
1
2k+1
. so by (i) of Lemma 9.4.12, we have
(9.5.2) H
n1
(1
2(2k+1),2k+1
. Z) H
n
(1
2(2k+1),2k+1
. Z) Z

= Z
2k
.
Here j is the Milnor number of the link 1
2k+1
which is easily computed by the
formula for Brieskorn polynomials, namely
j =
n

i=0
(o
i
1) = (2/ + 1 1) 1 1 = 2/ .
It now follows from Theorem 9.4.10 that the link 1
2(2k+1),2k+1
is dieomorphic to
2/#(o
n1
o
n
)#
4n1
for some
4n1
/1
4n
. We now use the periodicity results
of Durfee and Kauman to determine the dieomorphism type. First we notice that
Theorem 9.4.10 together with Theorem 4.5 of [DK75] imply that for every positive
integer i and every positive integer /. the link 1
2i(2k+1),2k+1
is homeomorphic to
the connected sum 2/#(o
n1
o
n
). The dieomorphism types are determined by
Theorem 9.4.13 (Theorem 6.4 of [Dur77], see also Theorem 5.3 of [DK75]) together
9.5. POSITIVE SASAKIAN STRUCTURES ON LINKS 323
with Remark 9.4.1. Let 1
i,k
denote the Milnor bre of the link 1
2i(2k+1),2k+1
and
(1
i,k
) its Hirzebruch signature. Then Theorem 9.4.13 says that for each pair of
positive integers i. , there is a dieomorphism
(9.5.3) 1
2i(2k+1),2k+1

(1
i,k
) (1
j,k
)
8

#1
2j(2k+1),2k+1
.
Here | denotes the connected sum of | copies of the Milnor sphere, and a minus
sign corresponds to reversing orientation. It follows from Durfees theorem that
the dierence in signatures is always divisible by 8, so this expression makes sense.
Equation (9.5.3) can be iterated; so it is enough to consider the case i = 2 and
, = 1. In order to determine how many distinct dieomorphism types occur in
(9.5.3), we need to compute the signature of the Milnor bres. To do so we adapt
formula (9.4.3) to the case of equation (9.5.3) with = 2(2/ + 1). namely, a =
(2(2/+1). 2/+1. 2 . . . . 2). Notice that we can always take the in Zagiers formula
9.4.3 to be this in which we shall denote (\
4m
(a)) by t
d
since the degree d =
2(2/ + 1) is the periodicity as well. Likewise, we denote by t
2d
the signature t(a)
with a = (4(2/ + 1). 2/ + 1. 2 . . . . 2). We nd
t
d
=
(1)
n
2
4/ + 2
4k+1

j=0
(1)
j
cot
2
(2, + 1)
8/ + 4
cot
(2, + 1)
4/ + 2
.
and
t
2d
=
(1)
n
2
8/ + 4
8k+3

j=0
(1)
j
cot
2
(2, + 1)
16/ + 8
cot
(2, + 1)
4/ + 2
.
We want to compute
k
=
]t
2d
t
d
]
8
. After some algebra we nd

k
=
8k+3

j=0
(1)
j
cot
(2, + 1)
16/ + 8

cot
(2, + 1)
16/ + 8
cot
(2, + 1)
8/ + 4

cot
(2, + 1)
4/ + 2
.
where =
1
64k+32
. Now
k
is always an integer that is independent of n. We now
dene
1
n
(/) =
[/1
4n
[
gcd(
k
. [/1
4n
[)
.
which by equation (9.5.3) represents the number of distinct dieomorphism types
that can be represented by our construction. Using MAPLE we give two tables
consisting of a list of
k
and 1
n
(/) together with the ratio
1
2
(/)
[/1
8
[
=
1
gcd(
k
. [/1
4n
[)
for both the 7-manifolds #2/(o
3
o
4
) and the 11-manifolds #2/(o
5
o
6
) for
various values of /. These two tables are given in Appendix B.2. This proves
Theorem 9.5.8.
Remark 9.5.1: Notice that the prime factorization of [/1
4n
[ consists of high powers
of two together with odd primes coming from the Bernoulli numbers. Since
k
is
independent of n. this gives rise to a bit of a pattern for the ratios
D
n
(k)
]bP4n]
. It is
obvious that for / = 1 all possible dieomorphism types occur, but this seems also
to hold for / = 2. It is of course true whenever [/1
4n
[ is relatively prime to 3. If we
look at the next case namely, /1
16
, we see that [/1
16
[ = 8128 = 2
6
127. Comparing
324 9. LINKS AS SASAKIAN MANIFOLDS
this with [/1
12
[ = 992 = 2
5
31. we see that the same ratios will occur for the case
#2/(o
7
o
8
) as for #2/(o
5
o
6
) for / = 1. . . . . 30.
To prove Theorem 9.5.9 we have n odd and there are several cases. First we
take j = 2(2/ +1). c = 2/ +1 as in the proof of Theorem 9.5.8. Again this leads to
the link 1
2(2k+1),2k+1
with free homology satisfying equation (9.5.2) except now n
is odd. Next we consider c = 2/ in equation (9.5.1). The link 1
2k
of the Brieskorn-
Pham polynomial 1
2k
= .
2k
1
+ .
2
2
+ + .
2
n
is a rational homology sphere by the
Brieskorn Graph Theorem. Furthermore, its monodromy map has period 2/. Then
choosing j = 2/ in equation (9.5.1) the link 1
2k,2k
is 2/-fold branched cover over
o
2n+1
branched over the rational homology sphere 1
2k
. so by item (i) of Lemma
9.4.12, we have
H
n1
(1
2k,2k
. Z) H
n
(1
2k,2k
. Z) Z

= Z
2k1
.
These two cases now give links whose middle homology groups are free of arbitrary
positive rank. However, unlike the case for n even this does not determine the
homeomorphism type unless n = 3. 7 in which case there is a unique dieomorphism
class. Indeed Theorem 9.4.11 implies we need to compute the quadratic form .
and this appears to be quite dicult in all but the simplest case. From Theorem
9.4.11 one can conclude [Dur71] that if ` 11
4n+2
with H
2n
(`. Z) free of
rank one, then it is homeomorphic to o
2n
o
2n+1
or the unit tangent bundle
T = T
1
(o
2n+1
). So the dieomorphism types at most dier by an exotic Kervaire
sphere
4n+1
. Furthermore, o
2n
o
2n+1
. T and (o
2n
o
2n+1
)#
4n+1
generate
the torsion-free submonoid of 11
4n+2
. there being relations in the monoid, namely,
T#T = #2(o
2n
o
2n+1
) and T#
4n+1
= T (Some further relations may exist
depending on n such as T
1
(o
3
) o
2
o
3
). This proves the rst statement in
Theorem 9.5.9.
To prove the second statement we follow Durfee and Kauman and consider
a slightly dierent Brieskorn-Pham polynomial, namely .
2k
0
+ .
2
1
+ + .
2
n
. For
/ = 1 we get as before a link 1
2,2
whose middle homology group is free of rank one.
Thus, it is dieomorphic to one of the three generators above by (i) of Lemma 9.4.12.
Now as / varies we have a homological periodicity by (ii) and (iii) of Lemma 9.4.12.
Durfee and Kauman show that there is an 8-fold dieomorphism periodicity, and
they compute the invariant to show that
1
2,2
T . 1
4,2
(o
2n
o
2n+1
)#
4n+1
.
1
6,2
T#
4n+1
T . 1
8,2
o
2n
o
2n+1
.
This proves Theorem 9.5.9.
9.5.3. Exotic Contact Structures. The material discussed here is well-
known and taken from Sato [Sat77] and Morita [Mor75]. Given an oriented almost
contact structure on a compact manifold `
2n1
, we let (`) be the set of homo-
topy classes of almost contact structures on `. This is in one-to-one correspondence
with the set of homotopy classes of almost complex structures on ` R [Sat77],
and when ` =
2n1
is a homotopy sphere, the latter is known [Mor75]:
(9.5.4) (
2n1
) =
2n1
(oO(2n)l(n))

=

Z Z
2
if n 0 mod 4;
Z
(n1)!
if n 1 mod 4;
Z if n 2 mod 4;
Z(n1)!
2
if n 3 mod 4.
9.5. POSITIVE SASAKIAN STRUCTURES ON LINKS 325
Theorem 9.5.10: On each odd homotopy sphere
2n1
/1
2n
. n 2. there
exist countably innitely many deformation classes of positive Sasakian structures
belonging to non-isomorphic underlying contact structures. Hence, the moduli space
of deformation classes of Sasakian metrics with positive Ricci curvature on
2n1
has innitely many positive components.
Proof. The proof for the homotopy spheres
4m1
and
4m+1
is somewhat
dierent. The case
4m1
uses the proof of Theorem 4.1(i) of Morita [Mor75]. We
represent each homotopy sphere
4m1
/1
4m
by the link of a BP polynomial.
The underlying almost contact structures can be distinguished by an invariant .
Moritas invariant, which is dened in terms of the Hirzebruch signature of the
Milnor bre, the Milnor number j of the link, and the Bernoulli numbers. Explicitly,
(9.5.5) (1(a)) =
(1)
m
(\
4m
)
4o
m

1
2
j(1(a)) .
where
o
m
=
2
2m
(2
2m1
1)
:!
1
m
.
It is more convenient to work with the link 1(6/ + 1. 3. 2. . . . . 2) rather than the
standard Brieskorn link with leading term 6/ 1. This is not a problem since the
two series both represent all homotopy spheres in /1
4m
. A simple computation gives
(1(6/ + 1. 3. 2. . . . . 2)) =

2
o
m
6

/ .
Moreover, there are an innite number of / that correspond to the same homotopy
sphere
4m1
/1
4m
. Thus, we will have a countably innite number of distinct
underlying almost contact structures on the same homotopy sphere as long as o
m
=
1
3
. We show that this always holds. Recall the well-known estimate for the Bernoulli
numbers 1
m

2(2m)!
(2)
2m
(cf. Appendix B of [MS74]). We then have
o
m
=
2
2m
(2
2m1
1)
:!
1
m

2(2
2m1
1)(2:)!

2m
:!
2(2
2m1
1)

:+ 1

m
which is clearly greater than 1 for all : 9. One can easily check, for example
using MAPLE
6
, that o
m
1 for all 2 : < 9. This completes the proof for this
case.
Now every homotopy sphere of the form
4m+1
can be represented as a BP link

4m+1
k
of Denition 9.4.4. Moreover, we know that
4m+1
k
is the standard sphere
if 2/ + 1 1(8). and the Kervaire sphere if 2/ + 1 3(8) which is exotic when
4: + 2 = 2
i
2. In this case the Morita invariant is simply given in terms of the
Milnor number j of the link, viz.
(
4m+1
k
) =
1
2
j(
4m+1
k
) = / .
Thus, as 2/ + 1 varies through 1(8) or 3(8). the invariant (
4m+1
k
) varies
through 0(4) and 3(4). or 1(4) and 2(4). respectively. Let o(:) denote the order of
the group (
4m+1
k
) dened by Eequation (9.5.4). Since for : 1 o(:) 0(4) we
can choose / such that (
4m+1
k
) is any number mod o(:). Thus, there are o(:)
distinct underlying almost contact structures which is nite in this case. To obtain
6
The notation for Bernoulli numbers is not standard. We use the notation used by topologists,
whereas, Maple uses the notation used by applied mathematicians. If we denote this Bernoulli
number by b
m
, then our B
m
equals (1)
m1
b
2m
.
326 9. LINKS AS SASAKIAN MANIFOLDS
an innite number of inequivalent contact structures we use the recent results of
Ustilovsky [Ust99] who showed by explicitly computing the contact homology of
Eliashberg, Givental, and Hofer (cf. [Eli98]) that for the standard sphere o
4m+1
distinct /s satisfying 2/ + 1 1(8) give inequivalent contact structures. How-
ever, Ustilovskys proof works equally well for any homotopy sphere
4m+1
. This
completes the proof.
Remarks 9.5.1: (1) It is interesting to note that for the 5-sphere o
5
(: = 1
above), Ustilovskys results distinguish innitely many inequivalent contact struc-
tures, whereas, there is a unique underlying almost contact structure. (2) It is
tempting to adapt the computations in [Ust99] to the case of the homotopy pro-
jective spaces RP
4m+1
. However, in this case complications arise, not the least of
which is the fact that the simplication of the contact homology arising from the
fact that all the Conley-Zehnder indices are even no longer holds. (3) An algorithm
for computing the contact homology for general Brieskorn manifolds has been given
recently in [vK04].
9.6. Links of Complete Intersections
Here we consider complete intersection singularities dened by j weighted ho-
mogeneous polynomials 1
1
. . . . . 1
p
in C
n+1
of degrees d
1
. . . . . d
p
. respectively. Our
main references are [Dim92, Dol82, IF00, Loo84]. We require that each of the
polynomials 1
i
for i = 1. . . . . j is a weighted homogeneous polynomial of degree
d
i
with weight vector w. We use vector notation and write f = (1
1
. . . . . 1
p
). and
d = (d
1
. . . . . d
p
); the latter is called the multidegree of f . We are interested in the
weighted ane cone C
f
dened by
C
f
= (.
0
. . . . . .
n
) C
n+1
[ f (.
0
. . . . . .
n
) = 0 .
We shall assume that dim
C
C
f
= n+1j and that the origin in C
n+1
is an isolated
singularity, in fact the only singularity, of f . Then the link 1
f
dened by
(9.6.1) 1
f
= C
f
o
2n+1
.
is a smooth manifold of dimension 2(n j) + 1. Furthermore, it is well-known
[Loo84] that
Proposition 9.6.1: Let f = (1
1
. . . . . 1
p
) be j weighted homogeneous polynomials of
multidegree d = (d
1
. . . . . d
p
) with weight vector w = (n
0
. . . . . n
n
). Let 1
f
be the link
of an isolated complete intersection dened by F. Then 1
f
is (nj 1)-connected.
The C

(w) action on C
n+1
restricts to an action on C
f
. and the associated o
1
action restricts to an action on both o
2n+1
and 1
f
. It follows that
w
is tangent to
the submanifold 1
f
and, by abuse of notation, we shall denote by
w
.
w
.
w
. o
w
the corresponding tensor elds on both o
2n+1
w
and 1
f
. Now
w
coincides with
on the contact subbundle T on o
2n+1
which denes an integrable almost complex
structure on T. Moreover, since 1 is a holomorphic function on C
n+1
the Cauchy-
Riemann equations imply that for any smooth section A of T we have
w
A(1) = 0.
Thus, 1
f
is an invariant submanifold of o
2n+1
with its weighted Sasakian structure,
or alternatively the inclusion 1
f
o
2n+1
w
is a Sasakian embedding. This gives an
easy generalization of a previous result of the rst two authors [BG01b], viz.
Theorem 9.6.2: The quadruple (
w
.
w
.
w
. o
w
) gives 1
f
a quasi-regular Sasakian
structure such that 1
f
o
2n+1
w
is a Sasakian embedding.
9.6. LINKS OF COMPLETE INTERSECTIONS 327
Since the o
1
(w)-action is locally free on both o
2n+1
w
and 1
f
. we have well-
dened quotient spaces which by a standard theorem are compact Kahler orbifolds.
The quotient of o
2n+1
w
is the well-known weighted projective space CP(w). and the
quotient of 1
f
is an algebraic variety Z
f
. It is also a suborbifold of CP(w) whose
singularities are quotient singularities induced by those of CP(w). So there is a
commutative diagram
(9.6.2)
1
f
o
2n+1
w

Z
f
CP(w) .
where the horizontal arrows are Sasakian and Kahlerian embeddings, respectively,
and the vertical arrows are orbifold Riemannian submersions. The right vertical
map is nothing but the weighted version of the well-known Hopf bration. The
algebraic variety Z
f
is a complete intersection in the weighted projective space
CP(w). and the condition that the ane cone C
f
be smooth away from the origin
means that Z
f
is quasi-smooth [Dol82]. As with the weight vector w we dene
[d[ =

d
i
. Then we have the following obvious generalization of Proposition 9.2.4:
Proposition 9.6.3: The link 1
f
is a
(i) canonical Sasakian structure if and only if [d[ [w[ 0.
(ii) null Sasakian structure if and only if [d[ [w[ = 0.
(iii) anticanonical Sasakian structure if and only if [d[ [w[ < 0.
In particular all links of complete intersections are either positive denite, negative
denite, or null.
CHAPTER 10
Sasakian Geometry in Dimensions Three and Five
In this chapter we shall discuss the geometry of Sasakian manifolds in rst two
non-trivial dimensions. The topology and geometry of compact three-dimensional
manifolds is extremely rich and varied. There has been enormous progress in the last
twenty years beginning with foundational work of Thurston [Thu97] and Hamilton
[Ham82], and culminating in the amazing papers of Perelman [Per02, Per03b,
Per03a]. Both the Poincare Conjecture and the Thurston Uniformization Con-
jecture have nally succumbed to the powers of the modern Geometric Analysis.
See [KL06, MT06, CZ06] for full details of Perelmans work. In contrast, the
Sasakian geometry of three-dimensional manifolds is much simpler owing to the fact
that any such manifold must be positive, negative or null which essentially reduces
their classication to the uniformization of the compact two-dimensional surfaces.
Furthermore, up to a cover there exists only one compact Sasaki-Einstein manifold
and it is the unit 3-sphere. In dimension ve the situation is far more compli-
cated. There is, however, a classication (see Theorem 10.2.3 below) of closed
simply connected ve manifolds up to dieomorphism due to Smale [Sma62a] and
Barden [Bar65]. As we shall see in this chapter, in many ways dimension ve is
large enough to accommodate many interesting Sasakian structures and yet small
enough to hope for some kind of classication, at least in the simply-connected
case.
10.1. Sasakian Geometry in Dimension Three
By Martinets Theorem 6.1.22 every orientable 3-manifold admits a contact
structure. The study of contact structures on 3-manifolds is quite rich and leads
to many yet unsolved problems. They occur in two types: tight contact structures,
and overtwisted contact structures. The latter have been classied by Eliashberg
[Eli89], leaving the tight contact structures as an important area of investigation.
See [Eli92, Bla02, Hon06, Gei06] for details and references. In contrast, as shown
by Belgun, the Sasakian structures are much more rigid and can be completely
classied [Bel00, Bel01]. Before discussing Belguns theorem, we briey consider
regular Sasakian structures in dimension 3. In the compact case this follows easily
from the classication of compact Riemann surfaces.
Proposition 10.1.1: Every regular compact Sasakian 3-manifold ` is a circle
bundle over a compact Riemann surface
g
of genus o.
Geiges [Gei97b] showed that a compact 3-manifold admits a Sasakian structure
if and only if it is dieomorphic to one of the following:
(i) o
3
with I
0
(o
3
) = oO(4).
(ii)

o1(2. R). where is universal cover of o1(2. R) and I
0
(

o1(2. R)).
(iii) Nil
3
with I
0
(Nil
3
).
329
330 10. SASAKIAN GEOMETRY IN DIMENSIONS THREE AND FIVE
Here is a discrete subgroup of the connected component I
0
of the corresponding
isometry group with respect to a natural metric, and Nil
3
denotes the 3 by 3
nilpotent real matrices, otherwise known as the Heisenberg group. These are three
of the eight model geometries of Thurston [Thu97], and correspond precisely to
the compact Seifert bundles with non-zero Euler characteristic [Sco83]. For further
discussion of the isometry groups we refer to [Sco83]. We refer to the three model
geometries above as spherical,

o1
2
-type, and nil geometry, respectively.
The purpose of this section then is to describe some examples represented as
links of isolated hypersurface singularities. First note that in dimension 3 the basic
cohomology group H
2
(T

) is 1-dimensional so we cannot get indenite forms other

than the zero class. Hence
Proposition 10.1.2: Let ` be a compact 3-dimensional Sasakian manifold. Then
` is either canonical (negative), anticanonical (positive) or null.
These three types of Sasakian structures correspond precisely to the three model
geometries above. Belguns Theorem uniformizes these three cases. We prefer to
rephrase Belguns theorem in terms of Tannos Theorem 7.3.23 on constant -
sectional curvature. In dimension three the -sectional curvature is determined by
one function, namely H(A) = 1(A. A) (cf. [Bel01], Theorem 1).
Theorem 10.1.3: Let ` be a 3-dimensional compact manifold admitting a Sasakian
structure o = (. . . o). Then
(i) If o is positive, ` is spherical, and there is a Sasakian metric of constant
-sectional curvature 1 in the same deformation class as o.
(ii) If o is negative, ` is of

o1
2
type, and there is a Sasakian metric of
constant -sectional curvature 4 in the same deformation class as o.
(iii) If o is null, ` is nil, and there is a Sasakian metric of constant -
sectional curvature 3 in the same deformation class as o.
Actually, when o is either null or negative than ` is up to nite quotient a
regular Sasakian 3-manifold, i.e., a circle bundle over a Riemann surface of positive
genus. In this case the Sasaki cone is one-dimensional (cf. Proposition 8.2.15)
so the only non-trivial deformations are deformations of the Kahler structure on
the underlying (orbifold) Riemann surface. In the positive case, ` is covered by
o
3
and its Sasakian structure is a deformation of a standard Sasakian structure
o
0
. In particular, Theorem 10.1.3 states that any Sasakian structure on o
3
is a
deformation of the standard one. As we have seen in the previous chapter (cf.
Theorem 9.5.10) this is far from true for higher-dimensional spheres. Generally
speaking not much is known about the Sasakian structures on o
2n+1
, n 1.
Exercise 10.1: Prove Theorem 10.1.3.
It is convenient when possible to represent Sasakian 3-manifolds as links of
quasi-smooth weighted homogeneous polynomials. In the case of positive Sasakian
structures this is always possible at least up to deformations, while in the null and
negative cases it is not.
10.1.1. Three-Dimensional Links. Since by Proposition 9.2.2 any link of a
quasi-smooth weighted homogeneous polynomial admits a class of Sasakian struc-
tures, one can obtain fairly explicit constructions of Sasakian structures in dimen-
sion three. Furthermore, it is easy to classify weighted homogeneous polynomials
in 3 variables [OW71a].
10.1. SASAKIAN GEOMETRY IN DIMENSION THREE 331
Proposition 10.1.4: Let /(.
0
. .
1
. .
2
) be a weighted homogeneous polynomial in
three complex variables having an isolated singularity at the origin. Then there
are a permutation on the set 0. 1. 2, non-zero complex numbers
0
.
1
.
2
, and
weighted homogeneous polynomials 1 and o with no monomials in common such
that / can be written as / = 1 +o. where 1(
0
.
(0)
.
1
.
(1)
.
2
.
(2)
) is equal to
1
:
(I) .
a0
0
+.
a1
1
+.
a2
2
.
(II) .
a0
0
+.
a1
1
+.
a2
2
.
1
with o
1
1.
(III) .
a
0
0
+.
a
1
1
.
2
+.
a
2
2
.
1
with o
1
1. o
2
1.
(IV) .
a
0
0
+.
a
1
1
.
0
+.
a
2
2
.
1
with o
0
1.
(V) .
a
0
0
.
1
+.
a
1
1
.
2
+.
a
2
2
.
0
.
(VI) .
a
0
0
+.
0
(.
a
2
2
+.
a
1
1
) +.
b
2
2
.
b
1
1
with (o
0
1)(o
2
/
1
+o
1
/
2
) = o
0
o
1
o
2
.
(VII) .
a
0
0
.
1
+.
0
(.
a
2
2
+.
a
1
1
) +.
b
2
2
.
b
1
1
with (o
0
1)(o
2
/
1
+o
1
/
2
) = (o
0
o
1
1)o
2
.
We shall use Orliks notation 1(o
0
. o
1
. o
2
; I). . . . . 1(o
0
. o
1
. o
2
. /
1
. /
2
; VII) when
referring to any link associated to these classes of polynomials. Note that the link
1(o
0
. o
1
. o
2
; I) = 1(o
0
. o
1
. o
2
) is of Brieskorn-Pham type. All the other types (II-
VII) can be easily written as 1

(w; d) links in the notation of Chapter 9. For

example, we have
1(o
0
. o
1
. o
2
; II) 1

d
o
0
.
d
o
1
.
d(o
1
1)
o
1
o
2
; d

.
1(o
0
. o
1
. o
2
; III) 1

d
o
0
.
d(1 o
2
)
1 o
1
o
2
.
d(1 o
1
)
1 o
1
o
2
; d

.
The 5-dimensional analogue of the link 1(o
0
. o
1
. o
2
; II) will be revisited later in
this chapter (cf. Example 10.3.7).
Exercise 10.2: Compute 1

(w; d) form of each link of Proposition 10.1.4. In each

case nd 1 = [w[ d and determine all the links with a xed sign of 1, i.e., separate
out positive, null, and negative links. Note that there is some duplication in the
list of Proposition 10.1.4. For instance, since 1(2. 2. 2; III) 1

(3. 2. 2; 6) we easily
conclude that 1(2. 2. 2; III) is equivariantly equivalent to the BP link 1(2. 3. 3. I)
1

(3. 2. 2; 6). Using the rst part of this exercise nd all such equivalences in the
list of Proposition 10.1.4.
The following conjecture of Milnor [Mil68] was proven by Orlik [Orl70]
Theorem 10.1.5: Let 1
f
1

(n
0
. n)1. n
2
; d). Then
1. The fundamental group
1
(1) is innite and the universal cover of 1 is
an open 3-cell if and only if d [w[.
2.
1
(1) is innite nilpotent if and only if d = [w[.
Proof. To prove the rst statement Orlik shows that
1
(1) is nite if and
only if [w[ d. If
1
(1) is nite, then the geometry of 1 is elliptic and 1 is covered
by o
3
. In this case it is well-known that
1
(1) is one of the nite subgroups of
ol(2). namely:
(i) = Z
p
the cyclic group of order j.
(ii) = D

m
a binary dihedral group of order 4(:2), where : 3,
(iii) = T

the binary tetrahedral group,

(iv) = O

the binary octahedral group,

1
Actually in [OW71a] Orlik and Wagreich missed the last two cases. The correction appears
in [OW77].
332 10. SASAKIAN GEOMETRY IN DIMENSIONS THREE AND FIVE
(v) = I

the binary icosahedral group.

Orlik [Orl70] shows that all of these geometries are realized by the links occurring
in Proposition 10.1.4 and that this exhausts all cases with [w[ d (cf. Exercise
10.2 for the 1 0 links). These are:
(i) = Z
p
gives the lens spaces 1(j. 1) = 1(2. 2. j; I).
(ii) The binary dihedral group = D

m
gives 1(2. :1. 2; II).
(iii) The binary tetrahedral group = T

gives 1(2. 3. 4; I).

(iv) The binary octahedral group = O

gives 1(2. 3. 3; II).

(v) The binary icosahedral group = I

gives 1(2. 3. 5; I).

This proves the rst part of (1), and the second statement of (1) follows from the
work of Waldhausen [Orl70, Wal67a]. It should be noted that only the lens spaces
of the form 1(j. 1) occur. Proof of (2) is equally simple. The condition [w[ = d
is very restrictive and any such link turns out to be equivariantly dieomorphic to
the three BP links 1(3. 3. 3; I). 1(2. 3. 6; I), 1(2. 4. 4; I) (cf. Exercise 10.2 for the
null links). One can easily calculate the fundamental groups for these three BP
links to check it is nilpotent [Orl70]. Actually, a few years later Milnor calculated
the fundamental group of all BP links 1(o
0
. o
1
. o
2
) [Mil75].
Remark 10.1.1: It is well-known that the ve classes given in the proof are the
ve classes of nite subgroups ol(2) the universal cover of the rotation group
oO(3) and that they are related to the hypersurface singularities as given. The
links are all dieomorphic to o
3
and they are not simply Sasakian but they
exhaust all possible compact 3-Sasakian manifolds in dimension 3 (see Chapter 13
for the denition and the classication). In particular, the cones on these links can
be identied with C
2
C
3
, where the embedding is given by the vanishing of
the associated polynomial. These are precisely the classical Kleinian singularities.
Even more remarkable is that these are related to the Dynkin diagrams of certain
simple Lie algebras. The Kleinian singularities, their resolutions, their relation to
the Dynkin diagrams of type A-D-E and the representation theory of the discrete
subgroups of ol(2) will be discussed in Section 12.10. Here, we summarize all this
in the following table:
Diagram Subgroup [[ Polynomial

p1
Z
p
j .
p
0
+.
2
1
+.
2
2
1
m
D

m
4(:2) .
2
0
.
1
+.
m1
1
+.
2
2
1
6
T

24 .
4
0
+.
3
1
+.
2
2
1
7
O

48 .
3
0
+.
3
1
.
0
+.
2
2
1
8
I

120 .
5
0
+.
3
1
+.
2
2
Let us now specialize to the case of links 1(a) = 1(o
0
. o
1
. o
2
) of Brieskorn-Pham
polynomials which, after Milnor [Mil75], are usually called Brieskorn manifolds.
There are precisely four classes of elliptical (i.e., [w[ d) Brieskorn 3-manifolds,
namely
1(j. 2. 2). 1(3. 3. 2). 1(4. 3. 2). 1(5. 3. 2).
Looking at the above table we see that only three of these explicitly appear; the
fourth one 1(3. 3. 2) 1(2. 2. 2; III) is equivalent to one on the list of Proposi-
tion 10.1.4 as pointed out in Exercise 10.2. There are precisely three Brieskorn 3-
manifolds in the Euclidean case ([w[ = d), namely 1(6. 3. 2). 1(4. 4. 2). and 1(3. 3. 3).
10.1. SASAKIAN GEOMETRY IN DIMENSION THREE 333
These are realized by arbitrary weighted homogeneous polynomials of degrees 6. 4
and 3 respectively, and they exhaust all possibilities that satisfy [w[ = d. The table
below gives the weights, degrees, and number of monomials of the three polynomi-
als.
manifold weight vector degree # of monomials
`
1
(1. 2. 3) 6 7
`
2
(1. 1. 2) 4 9
`
3
(1. 1. 1) 3 10
The remaining Brieskorn 3-manifolds are hyperbolic. Which links of an isolated
hypersurface singularities dened by weighted homogeneous polynomials in three
variables are homology 3-spheres have been characterized by Saeki [Sae87]. The
following easily follows from the Brieskorns graph Theorem 9.3.18.
Proposition 10.1.6: Let 1
f
denote the link of an isolated hypersurface singularity
dened by a weighted homogeneous polynomial 1 in three variables. Suppose further
that 1
f
represents a homology sphere, then 1
f
has the same knot type as the link
of a Brieskorn-Pham polynomial.
The problem of nding which Brieskorn manifolds are homology spheres can
be determined from Brieskorns Graph Theorem 9.3.18.
Proposition 10.1.7: The link 1(a) is a homology sphere if and only if gcd(o
i
. o
j
) =
1 for all i = , = 0. 1. 2.
It is clear from Theorem 10.1.5 and Propositions 10.1.6 and 10.1.7 that the only
non-trivial homology 3-sphere with nite fundamental group that can be realized as
a hypersurface link of a weighted homogeneous polynomial is the case 1(2. 3. 5; I)
which is the famous Poincare homology sphere. Furthermore, it is known that I

is the only non-trivial nite group that can occur as the fundamental group of an
integral homology 3-sphere [Sav02].
Example 10.1.8: [Poincare Homology Sphere] This example has a very inter-
esting history. When Poincare rst made his famous conjecture he rst conjectured
that any homology 3-sphere is homeomorphic to o
3
. He then constructed a coun-
terexample o
3
I

using Felix Kleins binary icosahedral group I

which is a perfect
group (its commutator subgroup [I

. I

] is all of I

) of order 120. Thus,

1
(o
3
I

) =
I

. whereas its Abelianization H

1
(o
3
I

. Z)
1
(o
3
I

)[
1
(o
3
I

).
1
(o
3
I

)] =
0. He then rephrased his conjecture as any homotopy 3-sphere is homeomorphic
to o
3
. The icosahedral group I is the group of isometries of a regular icosahedron
(dodecahedron), and the binary icosahedral group I

is the lift of I oO(3) to its

universal cover ol(2). It has a presentation
I

= (o. /) [ (o/)
2
= o
3
= /
5
.
It is well-known that there are many 3-dimensional homology spheres, and all
but one, the Poincare homology sphere have innite fundamental group. According
Saekis theorem, Proposition 10.1.6, any link of an isolated hypersurface singularity
in three variables which represents a homology sphere is a Brieskorn link 1(a).
Furthermore, Proposition 10.1.7 says that 1(a) is a homology sphere if and only if
the components of a are relatively prime. Let us consider a simple example.
334 10. SASAKIAN GEOMETRY IN DIMENSIONS THREE AND FIVE
Example 10.1.9: Consider the Brieskorn polynomial
1 = .
7
0
+.
3
1
+.
2
2
whose zero locus Z
f
is a hypersurface of degree d = 42 in the weighted projective
space CP(6. 14. 21). We wish to study the Brieskorn manifold 1(7. 3. 2). It has in-
nite fundamental group by Orliks Theorem 10.1.5, and is a homology 3-sphere
which is an orbifold bration over CP
1
with a nontrivial orbifold structure. 1(7. 3. 2)
is the quotient of the universal cover

o1(2. R) of o1(2. R) by a co-compact discrete
subgroup

o1(2. R). There is a covering of 1(7. 3. 2) by the total space of a
non-trivial circle bundle over a Riemann surface of genus o = 3.
Example 10.1.10: Consider the Brieskorn polynomial
1 = .
6k1
0
+.
3
1
+.
2
2
whose zero locus Z
f
is a hypersurface of degree d = 6(6/ 1) in the weighted
projective space CP(6. 2(6/1). 3(6/1)). We wish to study the link 1(6/1. 3. 2).
By Orliks Theorem 10.1.5
1
(1(6/ 1. 3. 2)) is innite if and only if d [w[ =
6/ 7 0. So there is one case with nite fundamental group, namely the link
1(5. 3. 2) = 1(2. 3. 5; I) which represents the Poincare homology sphere o
3
1

.
Thus, by Proposition 10.1.7, for all / 1 the link 1(6/ 1. 3. 2) represents a
homology 3-sphere with innite
1
. This homology 3-sphere is an orbifold bration
over CP
1
with a non-trivial orbifold structure. By Milnors Theorem [Mil75] the
link 1(6/ 1. 3. 2). for (/ 1). is the quotient of the universal cover

o1(2. R) of
o1(2. R) by a co-compact discrete subgroup

o1(2. R). Furthermore, 1(6/
1. 3. 2) has a nite covering by a manifold that is dieomorphic to a non-trivial
circle bundle over a Riemann surface of some genus o 1.
Next we give a brief discussion of how one can distinguish 3-dimensional inte-
gral homology spheres. This can be done by the Casson invariant which roughly
speaking counts the number of irreducible representations of
1
in ol(2). More
explicitly for a Brieskorn homology sphere 1(o
0
. o
1
. o
2
) Fintushel and Stern proved
that (1(o
0
. o
1
. o
2
)) equals
1
2
times the number of conjugacy classes of irre-
ducible representations of
1
(1(o
0
. o
1
. o
2
)) into ol(2). Furthermore, they showed
how (1(o
0
. o
1
. o
2
)) can be computed from the Hirzebruch signature (\ ) of the
parallelizable manifold \ (o
0
. o
1
. o
2
) whose boundary is the link 1(o
0
. o
1
. o
2
). Ex-
plicitly,
(1(o
0
. o
1
. o
2
)) =
(\ (o
0
. o
1
. o
2
))
8
.
One can easily compute the signature of \ (6/ 1. 3. 2) using formula (9.4.2) to
nd that it equals 8/. giving (1(6/ 1. 3. 2)) = /. Thus, 1(6/ 1. 3. 2) give
an innite sequence of inequivalent homology spheres. Moreover, it is known that
(1(7. 3. 2)) = 1, and since / = 1 from the previous sequence is the Poincare
homology sphere, which is clearly distinct, we see that 1(7. 3. 2) is also distinct
from the others. For further discussion and references we refer the reader to the
recent book [Sav02].
Exercise 10.3: One can easily see that 1(6/ + 1. 3. 2) is also an innite sequence
of homology spheres. Note that from the Hirzebruch signature computations the
Casson invariants (1(6/ 1. 3. 2)) = (1(6/ + 1. 3. 2)) = /. Yet, as remarked
1(5. 3. 2) and 1(7. 3. 2) are completely dierent though they are both homology
spheres. What can you say about 1(6/ 1. 3. 2) and 1(6/ + 1. 3. 2), for / 1?
10.2. SASAKIAN STRUCTURES AND THE TOPOLOGY OF 5-MANIFOLDS 335
All regular null Sasakian structures on a compact manifold can be determined
from the work of Milnor [Mil75]. These are realized as quotients of the three-
dimensional Heisenberg group H
3
(R) over R by the three-dimensional integral
Heisenberg group H
3
(Z). cf. Example 8.1.12. Up to covering these are all non-
trivial circle bundles over a complex torus, and it is known [Mil75] that all such
bundles can be constructed from the three-dimensional Heisenberg group H
3
(R)
which be represented as a subgroup of G1(3. R) by considering matrices of the
form
(10.1.1) A =

1 r .
0 1 n
0 0 1

; r. n. . R.
We also dene subgroups H
3
(/. Z) H
3
(Z) of matrices for which r. n. . are integers
divisible by some positive integer /. It is easy to see that the quotient manifold
`
k
= HH
k
is a circle bundle over a torus with Chern number /. The bration
: `
k
T
2
is explicitly given by
(10.1.2) (A) = (r mod /. . mod /) .
Now from example 7.3.21 the Sasakian structure o = (. . . o) on H
3
(R) is dened
by the formulas
(10.1.3)

= d. n dr. =

z
. o = dr
2
+dn
2
+ (d. n dr)
2
= (

x
+n

z
) dn

y
dr.
Now one can easily check that the tensor elds dening o are invariant under the
action of H
3
(R) on itself. Hence, o denes a Sasakian structure on the compact
quotient manifolds `
k
. Milnor [Mil75] observed that the integral homology can
easily be computed, viz.
H
1
(`
k
. Z) = H
k
[H
k
. H
k
] = Z Z Z
k
.
However, as indicated by the table 2 above only the cases / = 1. 2. 3 can be realized
as links of isolated hypersurface singularities of weighted homogeneous polynomials.
10.2. Sasakian Structures and the Topology of 5-Manifolds
To begin our study of Sasakian structures on 5-manifolds, we would like to
review basic results concerning smooth simply connected 5-manifolds. Dimension
5 provides a quite useful venue to study and understand the dierent classes of
Sasakian manifolds. For the most part, although not entirely, we restrict our con-
siderations to simply connected 5-manifolds. The main reason for doing so is an
important theorem of Markov [Mar58, Mar60] which states
Theorem 10.2.1: The class of compact smooth orientable n-manifolds, n 4, is
not classiable under dieomorphism, combinatorial equivalence, homeomorphism
or homotopy type.
Every such manifold ` is given by a triangulation. A classication is under-
stood to mean nding all classes of pairwise inequivalent manifolds and then giving
a nite algorithm which, from the triangulation of `, would determine the class
` belongs to. It can be shown that even the question of simple connectivity of `
cannot be decided by a nite algorithm. Surprisingly perhaps, if we restrict our
attention to simply connected 5-manifolds the full classication is not only possible,
336 10. SASAKIAN GEOMETRY IN DIMENSIONS THREE AND FIVE
but also it is very simple: the complete set of invariants consist of H
2
(`. Z) and
one other invariant i(`) introduced in [Bar65] and dened below.
Theorem/Denition 10.2.2: Let ` be a compact, smooth, oriented, 1-connected
5-manifold (or Smale-Barden manifold throughout this section). Let us write
H
2
(`. Z) as a direct sum of cyclic groups of prime power order
(10.2.1) H
2
(`. Z) = Z
k

p,i

Z
p
i

c(p
i
)
.
where / = /
2
(`), and c(j
i
) = c(j
i
. `). The non-negative integers /. c(j
i
) are
determined by H
2
(`. Z) but the subgroups Z
p
i H
2
(`. Z) are not unique. One
can choose the decomposition (10.2.1) such that the second Stiefel-Whitney class
map
n
2
: H
2
(`. Z) Z
2
is zero on all but one summand Z
2
j . The value , is unique, denoted by i(`),
and called the Barden invariant of `. It can take on any value , for which
c(2
j
) = 0, besides 0 and . Alternatively, i(`) is the smallest , such that there
is an H
2
(`. Z) such that n
2
() = 0 and has order 2
j
.
The following theorem was proved by Smale [Sma62a] in the spin case and later
generalized by Barden [Bar65] without the spin assumption. We shall formulate
it using Bardens notation and refer to it as the Smale-Barden Theorem.
Theorem 10.2.3: The class B of simply connected, closed, oriented, smooth, 5-
manifolds is classiable under dieomorphism. Furthermore, any such ` is dieo-
morphic to one of the spaces
(10.2.2) `
j;k
1
,...,k
s
= A
j
#`
k
1
# #`
k
s
.
where 1 , . : 0, 1 < /
1
and /
i
divides /
i+1
or /
i+1
= . A
complete set of invariants is provided by H
2
(`. Z) and i(`) and the manifolds
A
1
. A
0
. A
j
. A

. `
j
. `

are characterized as follows

` H
2
(`. Z) i(`)
A
1
= ol(3)oO(3) Z
2
1
`
1
= A
0
= o
5
0 0
A
j
, 0 < , < Z
2
j Z
2
j ,
A

Z
`
k
, 1 < / < Z
k
Z
k
0
`

= o
2
o
3
Z 0
We refer to B as the class of Smale-Barden manifolds or B-manifolds for short,
and dene the subclass B
0
to be the subclass of ` B such that i(`) = 0.
Elements of B
0
are called Smale manifolds, and this is precisely the subclass in B
that admit a spin structure, i.e., when n
2
(`) = 0. It is understood that : = 0
means that no `
ki
occurs. Note that A

is the non-trivial o
3
bundle over o
2
. With
the exception of A
1
= A
1
#A
1
. the manifolds A
j
s are indecomposable under
the connected sum operation. On the other hand `
0,k
1
,...,k
s
in the decomposition
10.2.2 is given in terms of invariant factors /
1
[/
2
[ [/
s
; equivalently, this can be
written in terms of elementary divisors j
i
s such as in the decomposition 10.2.1. For
example, there are four inequivalent Smale manifolds with ord(H
2
(`. Z)) = 36
2
.
They can be written in terms of either invariant factors or elementary divisors.
10.2. SASAKIAN STRUCTURES AND THE TOPOLOGY OF 5-MANIFOLDS 337
The equalities of the two types are `
0,36
= `
36
= `
4
#`
9
with : = 1 and as
`
0,2,18
= `
2
#`
18
= 2`
2
#`
9
. `
0,3,12
= `
3
#`
12
= 2`
3
#`
4
. and `
0,6,6
=
`
6
#`
6
= 2`
2
#2`
3
with : = 2. As usual, we write n` to denote the n-fold
connected sum of ` with itself for n 1. It is also convenient to use the convention
0` = o
5
. Note that for 1 < , < . the rational homology 5-spheres `
2
j and A
j
have the same H
2
(`. Z) but i(`
2
j ) = 0 while i(A
j
) = ,. Likewise, `

and A

(as o
3
-bundles over o
2
) are distinguished by their Barden invariants. We will be
interested in the following further special classes of Smale-Barden manifolds:
Denition 10.2.4: Let B be the class of all Smale-Barden manifolds. We dene
the following subclasses of B with obvious further inclusions
(i) oT
R
B
oT
B
B: B-manifolds admitting a free smooth o
1
-action; (a
Seifert bered o
1
-structure),
(ii) /(
QR
B
/(
B
B: B-manifolds admitting a quasi-regular almost contact
structure (an almost contact structure),
(iii) (
R
B
(
QR
B
/(
QR
B
(
B
B: B-manifolds admitting a regular contact
structure (a quasi-regular contact structure, a contact structure),
(iv) o
R
B
o
QR
B
o
B
/
B
(
QR
B
B: B-manifolds admitting regular
Sasakian structure; (a quasi-regular Sasakian structure, a Sasakian struc-
ture, a 1-contact structure),
(v) o
+R
B
o
+QR
B
o
+
B
o
B
B: B-manifolds admitting a (regular, quasi-
regular) positive Sasakian structure; ( o
R
B
o
QR
B
o

B
o
B
B,
o
NR
B
o
NQR
B
o
N
B
o
B
B for negative and null Sasakian struc-
tures),
(vi) oc
R
oc
QR
oc
B
o
+
B
B: B-manifolds admitting a (regular,
quasi-regular) Sasaki-Einstein structure.
For all of these subclasses of B we denote the corresponding subclass of Smale
manifolds by a superscript 0, for example the class of Seifert bered Smale manifolds
is denoted by oT
0
B
. and the regular Seifert bered Smale manifolds by (oT
R
B
)
0
.
We begin by mentioning some equalities that follow directly from previous
results.
Proposition 10.2.5: The following equalities hold:
/
B
= (
QR
B
. o
QR
B
= o
B
. o
+QR
B
= o
+
B
. o
QR
B
= o

B
. o
NQR
B
= o
N
B
.
Proof. The rst equality follows from Proposition 7.1.2 while the remaining
4 are a direct consequence of the Approximation Theorem 7.1.10.
Recall from Theorem 8.1.14 that in the last two cases a much stronger result
holds, namely that any such Sasakian structure is quasi-regular.
The inclusions oT
R
B
oT
B
B which are completely understood. First
concerning the inclusion oT
R
B
oT
B
. it is easy to see that a Smale-Barden manifold
` admits a free circle action if and only if H
2
(`. Z) is torsion free (cf. [Gei91,
Prop.10]), and we discuss the regular case further in Section 10.4 below.
The inclusion oT
B
B is much more subtle. Recall from the previous section
that an existence of a Seifert bration on a compact oriented 3-manifold imposes
a very strong topological restriction: Only six models of the eight Thurston Ge-
ometries are Seifert bered spaces and, as we have seen, three are Sasakian. Like-
wise, in the case of the Smale-Barden manifolds, the existence of a Seifert bered
338 10. SASAKIAN GEOMETRY IN DIMENSIONS THREE AND FIVE
o
1
-structure must impose severe restrictions on the pair H
2
(`. Z). i(`). The
Barden invariant i(`) is forced to assume only three possible values. In addition,
the size of / = /
2
(`) imposes restriction on the structure of the torsion group, the
lower the /
2
(`) the more restrictive the condition. More precisely, Kollar [Kol06a]
proves
Theorem 10.2.6: Let /
2
(`) = / and write H
2
(`. Z) as in equation (10.2.1).
Then ` oT
B
if and only if
(i) i(`) 0. 1. ,
(ii) for every prime j #i : c(j
i
) 0 / + 1,
(iii) i(`) = #i : c(2
i
) 0 /.
We will sometimes refer to conditions (i-iii) of Theorem 10.2.6 as Condition K.
Actually, Kollar observes that the decomposition 10.2.1 and the Barden invariant
i(`) can be considered also for certain non-simply connected 5-manifolds. He
obtains a slightly more general result: a compact oriented 5-manifold ` with
H
1
(`. Z) = 0 admits a Seifert bered o
1
-structure only if the conditions of the
Theorem 10.2.6 are satised. Under the
1
(`) = 0 assumption this is also a
sucient condition. Theorem 10.2.6 is not a classication of all Seifert bered o
1
-
structures on Smale-Barden manifold. For rational homology spheres / = /
2
(`) =
0 and we get
Corollary 10.2.7: Let /
2
(`) = 0. Then ` oT
B
if and only if i(`) = 0. 1 and
for every prime j #i : c(j
i
) 0 1.
Thus, for example, both `
3
and `
9
admit Seifert bered o
1
-structures while
`
3
#`
9
does not. It should be noted that Kollars proof does not give a classi-
cation of all Seifert bered o
1
-structures on any given Smale-Barden 5-manifold.
In fact, one can exhibit innitely many topologically distinct xed point free o
1
-
actions on every ` as in (10.2.6). In principle, the classication of all o
1
-actions
on 5-manifolds is reduced to a question on 4-dimensional cyclic orbifolds, but the
4-dimensional problem is rather delicate.
The proof of Theorem 10.2.6 is quite involved and it can be found in [Kol06a].
The main ideas follow [Sei32, OW75] as one considers any compact smooth ori-
ented 5-manifold with a xed point free circle action and takes the quotient space
Z = `o
1
. Z is a compact orbifold with an orbifold structure (Z. ), where
=

(1
1
m
i
)1
i
is a formal sum of codimension 2 closed subspaces 1
i
Z.
When (Z. ) is an algebraic orbifold surface, is the branch divisor dened in
Denition 4.4.8, but here one must work without any complex analytic assump-
tions on Z. The main technical aspect of the proof is relating the invariants of
` to the invariants of the topological Seifert bundle `(Z. ). In particular,
Kollar rst shows that the conditions (i-iii) of Theorem 10.2.6 are necessary. After
that, for each i(`) = 0. 1. separately he gives an explicit construction of a Seifert
bundle `(CP
2
.

(1
1
mi
)1
i
) with the prescribed allowable homology group.
Remark 10.2.1: The classication of circle actions on 4-manifolds was carried out
by Fintushel [Fin77, Fin78]. In this case the quotient is a 3-manifold (with bound-
ary corresponding to the xed points) endowed with additional data involving links
and certain weights. Foundational questions concerning circle and torus actions on
orbifolds were also considered in [HS91].
We next turn our attention to the question when a Seifert bered o
1
-structure
on a Smale-Barden manifold is associated to an (almost) contact structure. First
10.2. SASAKIAN STRUCTURES AND THE TOPOLOGY OF 5-MANIFOLDS 339
note, that in the realm of Smale-Barden manifolds Grays result, Corollary 6.2.8
translates as follows
Theorem 10.2.8: ` /(
B
if and only if i(`) = 0. .
This theorem completely explains the inclusion /(
B
B. The simplest exam-
ple of a manifold which has a locally free o
1
-action but no almost contact structure
is the symmetric space A
1
= ol(3)oO(3). also known as the Wu manifold
[Bar65]. As we are ultimately interested in those Smale-Barden manifolds ad-
mitting various Sasakian structures, we need to know, in particular, which such
manifolds admit a contact structure. Geiges showed that there are no further ob-
structions beyond the l(2)-reduction obstruction of Gray [Gei91], i.e., /(
B
= (
B
.
More precisely
Theorem 10.2.9: ` /(
B
if and only if `
5
= A
j
#`
k1
# #`
ks
, where
, = 0. . : 0, 1 < /
1
and /
i
divides /
i+1
or /
i+1
= . Furthermore, each such
` /(
B
admits a contact structure in every homotopy class of almost contact
structures.
Comparing the theorems of Geiges and Kollar we note that innitely many
Smale-Barden manifolds cannot support any quasi-regular contact structure. The
question of which Seifert bered o
1
-structures are compatible with some quasi-
regular (almost) contact structure is more dicult. In such cases the transverse
geometry is that of a compact symplectic 4-orbifold.
Denition 10.2.10: Let ` B with /
2
(`) = / and write H
2
(`. Z) as in 10.2.1.
We say that ` satises condition G-K if the pair (H
2
(`. Z. i(`)) fullls all of
the following
(i) i(`) 0. .
(ii) for every prime j #i : c(j
i
) 0 / + 1,
(iii) i(`) = #i : c(2
i
) 0 /.
Alternatively, ` B satises condition G-K if and only if ` (
B
oT
B
.
Note that condition G-K is necessary for ` /(
QR
B
. Unfortunately, we do not
know whether it is also sucient. Hence, we arrive at the following two fundamental
questions
Question 10.2.1: Suppose ` B satises the condition G-K.
(i) Does ` admit a quasi-regular almost contact structure, i.e., what is the
inclusion /(
QR
B
(
B
oT
B
?
(ii) Does ` admit a quasi-regular contact structure, i.e., what is the inclusion
(
QR
B
/(
QR
B
(
B
oT
B
?
Note that these are indeed two separate problems: if ` admits a quasi-regular
almost contact structure the deformed contact structure (which must exist) does
not itself need be quasi-regular. Alternatively, we are interested in those elements of
(
B
oT
B
such that the contact structure and Seifert bered structure are compatible
in the sense that the o
1
action of the Seifert bration is a Reeb vector eld for a
1-form representing the contact structure.
The importance of the above question has to do with the following simple
observation: Any quasi-regular contact manifold admits a compatible 1-contact
structure by Proposition 7.1.2. Also, any 1-contact structure can be approximated
by a sequence of quasi-regular 1-contact structures by Theorem 7.1.10. Restricting
to Smale-Barden manifolds gave us the rst equality in Proposition 10.2.5, /
B
=
340 10. SASAKIAN GEOMETRY IN DIMENSIONS THREE AND FIVE
(
QR
B
. To sum up, we get the following sequence of inclusions
(10.2.3) o
B
/
B
= (
QR
B
/(
QR
B
(
B
oT
B
which highlights the importance of condition G-K as a necessary condition for the
existence of any Sasakian structures.
Corollary 10.2.11: Let ` B. The condition G-K is necessary for the existence
of any 1-contact and, hence, a Sasakian structure on `.
It is disappointing that very little is known about all the inclusions in (10.2.3).
In particular, we do not know if o
B
/
B
is proper. (However, recall Example
7.4.16 which gives a non-simply connected K-contact 5-manifold that admits no
Sasakian structure). This gives rise to
Open Problem 10.2.1: Find a Smale-Barden manifold that admits a K-contact
structure but does not admit a Sasakian structure.
Kollar oers an important glimpse into the question of which Smale-Barden
manifolds admit Sasakian structures. He shows that the existence of a Sasakian
structure can be further obstructed [Kol06a], i.e., the inclusion o
B
(
B
oT
B
is
proper. Perhaps this should not be surprising for in the quasi-regular Sasakian case
the transverse geometry is algebraic, i.e., the space of leaves Z is a Hodge orbifold.
In particular, 1
i
Z must themselves be complex algebraic curves. The extra
condition leads to ner holomorphic obstructions as illustrated by the following
example [Kol06a].
Proposition 10.2.12: Let ` (Z.

(1
1
mi
)1
i
) be a Seifert bundle over an
algebraic orbifold Z. Assume that H
1
(`. Z) = 0 and
H
2
(`. Z) =

p,i

Z
p
i

c(p
i
)
.
i.e., ` is a rational homology sphere. Then there is a degree 2 polynomial c with
integer coecients such that the set c(Z) contains all but 10 elements of the set
c(j
i
). In particular, the set c(j
i
) contains at most 12 + 2

elements of any
interval of length .
Proof. If Z is an algebraic surface with quotient singularities, the Bogomolov-
Miyaoka-Yau-type inequalities of [Kol92, 10.8, 10.14] and [KM99, 9.2] imply that

1
1
r
x

< c(Z).
where :
x
is the order of the fundamental group of the link of a singular point r Z
and c(Z) is the topological Euler number. In our case c(Z) = 3 and so Z has at
most 5 singular points. Kollar shows that there are at most 10 complex algebraic
curves 1
i
which pass through a singular point. Once 1
i
is contained in the smooth
locus of Z, its genus is computed by the adjunction formula
2o(1
i
) = (1
i
(1
i
+1

)) + 2 .
Since Z has Picard number 1, we conclude that there is a degree 2 polynomial with
integer coecients c(t) such that 2o(1
i
) c(Z). The statement follows now from
the following lemma:
Lemma 10.2.13: Let c(t) = ot
2
+ /t + c with o 0. Then the set c(Z) intersects
every interval of length in at most 2 + 2

o elements.

10.2. SASAKIAN STRUCTURES AND THE TOPOLOGY OF 5-MANIFOLDS 341

The conditions of Proposition 10.2.12 are very restrictive as can be seen by the
following example.
Example 10.2.14: Let the j
i
be distinct primes and ` (Z.

(1
1
m
i
)1
i
) a
Seifert bundle over an algebraic orbifold such that H
1
(`. Z) = 0 and
H
2
(`. Z) =
k

i=1

Z
pi

2i
.
Then (10.2.12) implies that / 23. On the other hand, the conditions of (10.2.6)
are satised for any /, and so there are such Seifert bundles for any / and j
i
(even
over Z = CP
2
). The surfaces 1
i
, however, can not be chosen complex algebraic for
/ 23.
Corollary 10.2.15: There exists innitely many distinct rational homology spheres
in (
B
oT
B
which do not admit any Sasakian structure.
Other than the above example, generally we cannot say much about the inclu-
sions o
B
/
B
(
B
oT
B
. On the other hand it follows that either the rst, or
the second, or both inclusions are proper.
10.2.1. Positive Sasakian Structures. We now focus our attention on pos-
itive Sasakian manifolds o
+
B
(
B
oT
B
. the importance of which stems from
Theorem 9.5.1 which implies that any ` o
+
B
admits a Sasakian metric with
positive Ricci curvature. In the positive case one can actually get much more infor-
mation. The point is that, by denition, when ` o
+
B
we get Seifert o
1
-structures
`(Z. ) with c
orb
1
(Z. ) 0. Hence, in such cases the base (Z. ) is a compact
del Pezzo orbifold surface with only cyclic orbifold singularities. The positivity as-
sumption, although not as strong as in the smooth case, still imposes restrictions on
the type of Seifert brations that can occur. These results are again due to Kollar
[Kol05, Kol06b] and the techniques used to prove them are ensconced deeply in
the minimal model program. Since the authors are not experts in this area, we
shall state the relevant results, and in many cases only sketch some ideas of the
proofs, referring to the literature for details.
We begin by sketching the main idea in applying the minimal model program.
One begins with a log del Pezzo surface (o. ) and obtains a minimal resolution o
t
by a sequence of blow ups. Then one obtains a minimal model o
min
by a sequence of
blow downs. o
min
is a minimal rational surface, and there are a limited number of
possibilities, namely it is either CP
2
. CP
1
CP
1
. or one of the Hirzebruch surfaces S
n
with n 2. At each intermediate step one obtains a surface T whose anticanonical
divisor satises 1
T

T
+ H
T
. where H
T
is big and nef (see Denition 3.5.23)
and
T
is the proper transform of . From Theorem 4.7.14 we know that the
torsion in H
2
is determined by the non-rational curves in the branch divisor of
o, and it is not dicult to see that there is at most one such curve. So we can
write = (1
1
m
)1
0
+
t
. where 1
0
is non-rational, and the remainder
t
of the
branch divisor consists of at most rational curves. A case by case analysis shows
that there are constraints between the ramication index : and the genus o of the
non-rational curve 1
0
. Let 1
t
0
denote the birational transform of 1
0
. 1 the section
of S
n
of self-intersection n. and 1 the bre of S
n
. Then the following possibilities
occur [Kol06b]:
342 10. SASAKIAN GEOMETRY IN DIMENSIONS THREE AND FIVE
(1) If : 6 and o 1 then o
min
is CP
2
. CP
1
CP
1
. or S
2
and o = 1 with
1
t
0
[ 1
S
min[.
(2) If : 5 and o 2 then : = 5. o
min
is S
3
and o = 2 with 1
t
0
[21+61[.
(3) If : 3 and o 4 then : = 3. o
min
is S
5
and o = 4 with 1
t
0

[21 + 101[.
(4) If : = 4 and o 2 then either o
min
is S
3
and o = 2 with 1
t
0
[21+61[
or o
min
is S
2
and o = 2 with 1
t
0
[21 + 51[.
Using this analysis it is straightforward to prove the following lemma which we
shall make use of in several places below. Details can be found in (6.8) and (6.13)
of [Kol05].
Lemma 10.2.16: Let (o. =

n
i=0
o
i
1
i
) be a log del Pezzo surface. Assume that
1
0
has genus 1 and o
0
12. Then 1
i
is rational for i 1 and
(i) o(1
0
) = 1 if o
0
56,
(ii) o(1
0
) 2 if o
0
34,
(iii) o(1
0
) 4 if o
0
23.
Assume in addition that o
0
1112 and o
i
12 for every i. Then (o. ) is a log
del Pezzo surface with Du Val singularities and = o
0
1
0
. The Picard number of
o is at most 9.
We recall the following denition from [KM98]. Let (0. A) be a normal surface
singularity with a minimal resolution 1 : Y A. Then (0. A) is a Du Val Singu-
larity if for every exceptional curve 1
i
Y we have 1
Y
1
i
= 0. A normal surface
A with Du Val singularities is characterized by the fact that 1
X
is Cartier.
The rst remarkable theorem of Kollar [Kol05] says that only very few groups
can occur as torsion subgroups of H
2
(`. Z) of a simply connected 5-manifold that
admits a positive Sasakian structure. Recall from Theorem 4.7.14 that torsion
in H
2
(`. Z) is a consequence of the existence of a non-rational component of the
branch divisor. Then the point of Lemma 10.2.16 is that the positivity of c
orb
1
forces
the existence of at most one non-rational component such that both the genus and
ramication index cannot be large, severely restricting the possibilities. We call
such torsion groups admissible. Explicitly,
Theorem 10.2.17: Suppose ` o
+
B
. Then the torsion subgroup of H
2
(`. Z) is
one of the following:
(i) (Z
m
)
2
for any : Z
+
.
(ii) (Z
5
)
4
.
(iii) (Z
4
)
4
,
(iv) (Z
3
)
4
. (Z
3
)
6
or (Z
3
)
8
.
(v) (Z
2
)
2n
for any n Z
+
.
Conversely, for each nite group G in the list above, there is an ` (o
+
B
)
0
with
H
2
(`. Z)
tors
= G. (Note that : = 1 corresponds to zero torsion).
Proof. By the Approximation Theorem 7.1.10, ` admits a Seifert bundle
structure over a log del Pezzo orbifold (o. ). where =

1
1
mi

1
i
. and by
Theorem 4.7.14 if the torsion of H
2
(1. Z) is non-trivial at least one of the curves
1
i
must be non-rational. Then the rst part of Lemma 10.2.16 says that all the
other curves 1
i
for i = 0 must be rational, and the list of possible torsion groups
follows from (i)-(iii) of the lemma.
10.2. SASAKIAN STRUCTURES AND THE TOPOLOGY OF 5-MANIFOLDS 343
The converse can be proved easily by example. For instance, the Brieskorn-
Pham link 1(3. 3. 3. 3/) has homology H
2
(1. Z) = Z
6
Z
k
Z
k
for any / which
gives examples for (i) (including the case of no torsion which is / = 1). All other
torsion groups in (ii)-(v) can be realized as H
2
(`. Z) of a rational homology sphere
in (o
+
B
)
0
and they are given in the discussion of Theorem 10.2.19 below.
Remark 10.2.2: It is quite remarkable that all possible torsion groups for simply
connected 5-manifolds admitting positive Sasakian structures can be realized by
links of weighted homogeneous polynomials. Notice that we do not claim that
every simply connected 5-manifold admitting a positive Sasakian structure can be
realized as the link of weighted homogeneous polynomials. This is clearly false as
A

shows. See also the table in Theorem 10.2.25 below.

One can be more precise about possible torsion when ` B is a rational ho-
mology sphere. When ` o
+
B
. in addition we have that
1
(`) =
orb
1
(Z) = 0.
and c
orb
1
(Z) 0. The rst results concerning the existence of positive Sasakian
structures on rational homology 5-spheres were obtained by the authors in [BG02]
(with erratum [BG06b]) where it was shown that innitely many rational ho-
mology 5-spheres admit positive Sasakian structures. However, since then Kollar
[Kol05] has obtained a complete classication of all rational homology spheres in
o
+
B
together with a complete classication of Seifert bered structures in certain
special cases [Kol05]. First we need
Theorem 10.2.18: There is a one-to-one correspondence between
(i) Seifert bundle structures 1 : ` Z = (o. ) on 5-dimensional, compact
rational homology spheres with H
1
(`. Z) = 0. and
(ii) compact, complex, 2-dimensional, cyclic orbifolds Z = (o. ) with H
2
(o. Q) =
Q and H
orb
1
(Z) = 0.
Under this correspondence,
1
(`) =
orb
1
(Z).
The proof of this theorem is fairly straightforward and left to the reader, who
may consult [Kol05] for details. We are now ready for our main classication
theorem for rational homology spheres.
Theorem 10.2.19: Let ` o
+
B
and /
2
(`) = 0. Then ` (o
+
B
)
0
. and
(i) if H
2
(`. Z) contains a torsion element of order at least 12 then H
2
(`. Z) =
(Z
m
)
2
for some : not divisible by 30. Furthermore, the number of Seifert
bundle structures varies between 1 and 4. depending on : mod 30.
(ii) 2`
5
has a unique family of positive Sasaki-Seifert structures,
(iii) 4`
3
has a unique family of positive Sasakian Seifert structures,
(iv) 2`
4
has exactly 2 families of positive Sasaki-Seifert structures,
(v) n`
2
admits innitely many families of positive Sasaki-Seifert structures
for each n 0.
(vi) the manifolds 2`
3
. 3`
3
. and `
m
. with 2 < : < 12 all admit families of
positive Sasaki-Seifert structure. Moreover, `
3
. `
5
and 2`
3
all admit
innitely many positive Sasaki-Seifert structures.
In particular, with the exception of (Z
30n
)
2
, all torsion groups of Theorem 10.2.17
appear as H
2
(`. Z) for some ` (o
+
B
)
0
with /
2
(`) = 0.
Proof. First, we notice that ` (o
+
B
)
0
by Theorem 10.2.9, and second we
notice that H
2
(`. Z) must be one of the groups listed in Theorem 10.2.17. Again
from the Approximation Theorem 7.1.10 we know that ` must have a Seifert
344 10. SASAKIAN GEOMETRY IN DIMENSIONS THREE AND FIVE
o
1
-bundle structure `(o. ). and by Theorem 10.2.18 there is a one-to-one
correspondence between these and their cyclic orbifolds. Using this we now proceed
to check each of the six assertions. For (i) we apply the second part of Lemma
10.2.16 to conclude that o is a log del Pezzo surface with Du Val singularities.
But such surfaces with Picard number 1 have been classied [Fur86, MZ88], and
described in Proposition 6.5 in [Kol05]. Under the : 12 restriction there are
basically only four possibilities for (Z. ):
1. Z = (o. ) = (CP
2
. (1
1
m
)1). where 1 is a smooth cubic and gcd(:. 3) =
1.
2. Z = (o. ) = (Q. (1
1
m
)1). where Q CP
3
is a quadric cone and
1 is its intersection with a quadric, and gcd(:. 2) = 1. As an orbifold,
Q = CP(1. 1. 2).
3. Z = (o. ) = (CP(1. 2. 3). (1
1
m
)1). where 1 [O(6)[ is smooth. [1]
is 6 times the generator in Cl(Z) and gcd(:. 6) = 1.
4. Z = (o. ) = (o
5
. (1
1
m
)1) with an embedding o
5
CP
5
as a degree
5 surface and 1 is a smooth hyperplane section. [1] is 5 times the
generator in Cl(Z) and gcd(:. 5) = 1.
In all four cases the curve 1 is a smooth member, of genus one, of the linear
system [ 1

[. and the Seifert bundles can be represented by links of weighted

homogeneous polynomials. For each such polynomial one needs to check that the
branch divisor has the correct ramication index, and then compute the genus of
the corresponding curve using the genus formula 4.6.5. In the rst three cases
they are BP Links listed in the tables of Example 10.3.3 below: 1(3. 3. 3. /) with
(/. 3) = 1, 1(2. 4. 4. /) with (/. 2) = 1. and 1(2. 3. 6. /) with (/. 6) = 1. In the last
case the link is given in Example 10.3.7 and is the link of 1(z) = .
6
0
+.
3
1
+.
3
2
+.
0
.
k
3
with (/. 5) = 1. This is a weighted homogeneous polynomial of degree d = 6/ with
weights w = (/. 3/. 2/. 5) and index 1 = 5. So the associated link 1
f
is positive
and one can show it is `
k
. The list of four cases above shows that : cannot be
divisible by 30 and proves the statement in (i).
The manifold 2`
5
is the Brieskorn-Pham link 1(2. 5. 6. 6) and the uniqueness
of the Seifert bration is shown in [Kol05]. This proves (ii). To establish (iii)
we observe that the manifold 4`
3
is the Brieskorn-Pham link 1(2. 3. 10. 10) and
the uniqueness of the Seifert bration can be argued similarly [Kol06b]. The
case (o. ) = (o
5
. (1
1
6
)1) gives `
6
and it can be realized by the link of the
weighted homogeneous polynomial 1(z) = .
6
0
+.
2
1
+.
3
2
+.
0
.
6
3
. The manifolds `
m
for 2 < : < 12 with : = 6 can be all obtained as Brieskorn-Pham links (see
the discussion of positive BP links in Section 10.3 and Table B.4.1 in Appendix
B). Likewise, 2`
3
= 1(2. 3. 7. 14). 3`
3
= 1(2. 3. 8. 8). This proves (vi). To show
(v) consider 1(z) = .
2
0
.
1
+ .
2n+1
1
+ .
2n+1
2
+ .
2
3
. This is a weighted homogeneous
polynomial of degree d = 4n + 4 with weights w = (2n. 2. 2. 2n + 1) and of index
1 = 1. Hence, Z
4n+2
CP(2n. 2. 2. 2n + 1) is log del Pezzo with the associated
positive link 1
f
. One can show that H
2
(1
f
. Z) = (Z
2
)
2n
. Lastly, to establish (iv),
take 1(z) = .
5
0
+.
5
1
+.
0
.
2
2
+.
4
3
. This is a homogeneous polynomial of degree d = 20
with weights w = (4. 4. 8. 5) and index 1 = 1. Thus the associated link 1
f
is positive
and one can show that H
2
(1
f
. Z) = (Z
4
)
4
. The second Seifert bration comes
from the polynomial 1(z) = .
10
0
+.
5
1
+.
2
2
+.
0
.
4
3
. This is a weighted homogeneous
polynomial of degree d = 40 with weights w = (4. 8. 9. 20) with index 1 = 1. Kollar
shows there are no other positive Seifert bered structures in this case [Kol06b].
10.2. SASAKIAN STRUCTURES AND THE TOPOLOGY OF 5-MANIFOLDS 345
On the other hand innitely many positive Seifert brations on `
2
. `
3
. `
5
and 2`
2
can be realized by BP links of Example 10.3.3. In the case of n`
2
for n 0 one can take the link of 1(z) = .
k(2n+1)
0
+ .
2n+1
1
+ .
1
.
2
2
+ .
2
3
with
gcd(/. 2n(2n +1)) = 1 to obtain an innite number of positive Sasakian structures
on n`
2
. For the n = 0 case, i.e., o
5
, the link 1(2. 2. j. c) gives innitely many
positive Sasakian structures.
Theorem 10.2.19 says that the problem of determining which simply connected
rational homology 5-spheres admit positive Sasakian structures is completely un-
derstood. We succinctly summarize this as
Corollary 10.2.20: Let ` be a simply connected rational homology 5-sphere.
Then ` admits a positive Sasakian structure if and only if ` is one of the following
` = `
m
. n`
2
. 2`
3
. 3`
3
. 4`
3
. 2`
4
. 2`
5
.
where n. : are positive integers and : = 30/ for some positive integer /.
Here it is understood that `
1
= o
5
. The problem of determining all positive
Sasaki-Seifert structures on these manifolds is still very much open. This seems
somewhat intractable for ` = o
5
. but in other cases it is quite manageable.
Open Problem 10.2.2: Classify all positive Sasaki-Seifert structures on `
4
, 3`
3
and `
m
for 6 : 11.
With the exception of `
4
it is probably true that the number of such structures
is nite in each case. These sporadic rational homology spheres already appear
in the classication of positive Brieskorn links 10.3.3. Certainly, in each case we
have at least one example. In some cases there are many.
We now consider the case of a general Smale-Barden manifold. The rank of
H
2
(`. Z) and its torsion subgroup are not independent of each other, but the
classication problem becomes much more complicated. One observation that can
be made by looking at Theorems 10.2.17 and 10.2.19 is that the possible torsion
groups occur in two series each depending on a positive integer plus a handful of
exceptional cases.
Kollar [Kol05, Kol06b] obtains only partial results when H
2
(`. Z)
tor
= Z
m

Z
m
. but he is able to get a complete classication in certain exceptional cases. In
particular, there are denitive results in three of the ve exceptional cases listed in
Theorem 10.2.17.
Theorem 10.2.21: Let ` o
+
B
.
(i) If H
2
(`. Z) = Z
k
(Z
5
)
4
. then / = 0 and there is a unique family of
positive Sasaki-Seifert structures;
(ii) If H
2
(`. Z) = Z
k
(Z
3
)
8
. then / = 0 and there is a unique family of
positive Sasaki-Seifert structures;
(iii) If H
2
(`. Z) = Z
k
(Z
4
)
4
then / = 0. 1 and there are exactly 3 families
of positive Sasaki-Seifert structures: 2 for / = 0 and 1 for / = 1.
Proof. This follows by a detailed analysis of the four conditions above Lemma
10.2.16. We refer to [Kol05, Kol06b] for details.
Remark 10.2.3: It is quite remarkable that all the 5-manifolds ` o
+
B
of Theorem
10.2.21 can be realized as links of weighted hypersurface singularities. Cases (i) and
(ii) are realized by the BP links 1(2. 5. 6. 6) and 1(2. 3. 10. 10), respectively, which
are listed in Table B.4.1 of appendix B. For case (iii) two Sasaki-Seifert structures
346 10. SASAKIAN GEOMETRY IN DIMENSIONS THREE AND FIVE
occur on 2`
4
and one on `

#2`
4
. and these are realized as links of the non-BP
weighted homogeneous polynomials
1
1
(z) = .
5
0
+.
5
1
+.
4
2
+.
0
.
2
3
. 1
2
(z) = .
10
0
+.
5
1
+.
0
.
4
2
+.
2
3
.
for / = 0 (the corresponding hypersurfaces are 7
20
CP(4. 4. 5. 8) and 7
40

CP(4. 8. 9. 20). respectively) or
1
3
(z) = .
6
0
+.
6
1
+.
0
.
4
2
+.
2
3
.
which is 7
24
CP(4. 4. 5. 12) for / = 1 [Kol06b]. We give a list of weighted
homogeneous hypersurfaces of this type in Example 10.3.7 below.
In [Kol06b] Kollar also obtains some results for non-simply connected 5-
manifolds. First, as in Proposition 8.1.20, there are the obvious quotients obtained
from the cyclic subgroups of the circle group generated by the Reeb vector eld.
But there are other more interesting quotients. For example Kollar shows (See The-
orem 7 of [Kol06b]) that there is a free involution on the manifolds 5`

#2`
2
and
4`

#2`
2
giving 5-manifolds ` with
1
(`) = Z
2
and H
2
(`. Z) = Z
2
(Z
4
)
4
.
In order to proceed further we need to discuss the notion of a weighted blow up.
The general denition is Denition 4.56 in [KM98]. However, as in [Kol05] only a
special case is needed so we give the denition employed there. Let j be a point on
a smooth curve C o and blow up j repeatedly n times. This gives one (1)-curve
and an (n 1)-chain of (2)-curves. We can contract the chain of (2)-curves to
obtain a singular neighborhood of type
n1
which can be represented locally by
an equation of the form rn .
n
= 0. The resulting surface T is called the blow up
of o of type n at j. If o is a del Pezzo surface with cyclic Du Val singularities and
C o is a smooth elliptic curve, then we denote by 1
n
1
,...,n
k
o any surface which is
obtained as above by performing blow ups of type n
i
at j
i
C. The anticanonical
class of 1
n
1
,...,n
k
o will be nef and big if and only if

n
i
< (1
2
S
), and if this holds
then 1
n
1
,...,n
k
o is a del Pezzo surface with cyclic Du Val singularities for a general
choice of the points j
i
. Applying the minimal program we obtain for o
min
either
one of the 4 log del Pezzo surfaces with Picard number one CP
2
. CP(1. 2. 3). Q. o
5
listed in the proof of Theorem 10.2.19 or CP
1
CP
1
. Then Kollar proves [Kol05]
Proposition 10.2.22: There are 93 deformation types of del Pezzo surfaces with
cyclic Du Val singularities satisfying H
orb
1
(Z) = 0. These are
(i) 1
n1,...,n
k
CP(1. 2. 3) for

n
i
< 6. n
i
2.
(ii) 1
n1,...,n
k
Q for

n
i
< 8. n
i
2.
(iii) 1
n
1
,...,n
k
CP
2
for

n
i
< 9. n
i
2.
(iv) 1
n
1
,...,n
k
CP
1
CP
1
for

n
i
< 8.
(v) o
5
. 1
3
o
5
. 1
4
o
5
and 1
1
CP
2
.
All these satisfy
orb
1
(Z) = 0.
Remark 10.2.4: The count of the 93 deformation types and the ranges for the n
i
in Proposition 10.2.22 takes into account the following isomorphisms
1
1
CP(1. 2. 3) 1
3
Q. 1
1
Q 1
2
CP
2
. 1
m,1
CP
2
1
m
(CP
1
CP
1
). 1
1
o
5
1
5
CP
2
as well as isomorphisms derived from them.
In [Kol06b] Kollar has loosened the restriction that
orb
1
(Z) = 0. In all he
has found 132 families of log del Pezzo surfaces with cyclic Du Val singularities
up to deformations giving precisely the 93 deformation types listed in Proposi-
tion 10.2.22 when
orb
1
(Z) = 0. We are interested in constructing orbibundles or
10.2. SASAKIAN STRUCTURES AND THE TOPOLOGY OF 5-MANIFOLDS 347
equivalently Seifert o
1
-bundles over Z with smooth total space
2
. Let us set some
notation. We let Z = (7. ) denote one of 93 log del Pezzo surfaces (orbifolds)
given by Proposition 10.2.22 with 7 = 1
n1,...,n
k
o. where o is one of the 5 surfaces
CP
2
. CP(1. 2. 3). Q. o
5
. CP
1
CP
1
. Let H be a positive generator of Cl(o). and let
1
i
with i = 1. . . . . / denote the exceptional divisors. Then the divisor class group
Cl(7) is freely generated by

H. 1
1
. . . . . 1
k
except in the case o = CP
1
CP
1
.
where we add a second positive generator 1
0
. The anticanonical divisor of 7 is
1
Z
1
S

i
n
i
1
i
. where 1
S
is the Fano index of o. We choose a Seifert
o
1
-bundle 1 : `(7. ) with rst Chern class
(10.2.4) c
1
(`7) = o[

H] +

i
c
i
[1
i
] +

i
/
i
:
i
[1
i
] .
where 0 /
i
< :
i
. gcd(/
i
. :
i
) = 1. In order to have a Sasakian structure on `
we need to choose c
1
(`7) 0. This can always be done by choosing o large
enough. The condition for assuring the smoothness of `, given by Proposition
4.7.8, is that the local restriction of :(r. ) c
1
(`7) generates the local class
group Cl(7. r) at each point r 7. But the singular points j
i
of the weighted blow
ups lie on their respective exceptional divisor 1
i
which generates the local class
group Cl(7. j
i
) = Z
n
i
. So ` will be smooth if gcd(n
i
. c
i
) = 1 for all i = 1. . . . . /.
and when o is singular gcd(1
S
. o) = 1. Furthermore, ` will be simply connected if
in addition :()c
1
(`7) is a generator of Cl(7).
Now the orbifold rst Chern class of Z = (7. ) is
(10.2.5)
c
orb
1
(Z) = c
1

(1
Z
+

1
1
:
i

1
i
)

= 1
S
[

H]

i
n
i
[1
i
]

1
1
:
i

[1
i
] .
In order that ` admit a positive Sasakian structure it is necessary that c
orb
1
(Z)
0. Let us now specialize to the case at hand, namely that 7 is one of the 93
log del Pezzo surfaces with cyclic Du Val singularities described by Proposition
10.2.22. In this case by Lemma 10.2.16 there is only one component, say 1, in the
branch divisor . and we let : denote its ramication index. Let us dene d(7) =
gcd(n
1
. . . . . n
k
. 1
S
). Now the condition that
orb
1
(Z) = 0 implies by Proposition
4.7.13 that (1) mod : surjects onto Z
m
. Moreover, it is easy to see that
the restriction map

: H
2
(7. Z)H
2
(1. Z) is multiplication by d(7). It follows
that we must have gcd(d(7). :) = 1. Of course, in many cases d(7) = 1 so there
is no condition. Summarizing we have
Proposition 10.2.23: Let Z = (7. = (1
1
m
)1) denote one of 93 log del Pezzo
orbifold surfaces given by Proposition 10.2.22 with c
orb
1
(Z) 0. and let `Z be
the Seifert o
1
-bundle whose rst Chern class c
1
(`7) is given by equation (10.2.4).
Then ` is a smooth simply connected 5-manifold with a positive Sasakian structure,
that is ` o
+
B
if and only if the following conditions hold:
(i) c
1
(`7) 0 and gcd(n
i
. c
i
) = 1 for all i = 1. . . . . /. and when o is
singular gcd(1
S
. o) = 1.
(ii) gcd(d(7). :) = 1.
(iii) :()c
1
(`7) is a generator of Cl(7).
2
Although here we are discussing the case of del Pezzo surfaces with cyclic Du Val singular-
ities, the analysis that follows is slightly more general.
348 10. SASAKIAN GEOMETRY IN DIMENSIONS THREE AND FIVE
We can now combine Propositions 10.2.22 and 10.2.23 with Theorems 10.2.17
and 10.2.21.
Theorem 10.2.24: Let `
5
o
+
B
with / = /
2
(`). Then `
5
is /`

# or
A

#(/ 1)`

#. where is `
m
. n`
2
. 2`
3
. 3`
3
. 4`
3
. 2`
4
. 2`
5
. Further-
more,
(i) If = 4`
3
or 2`
5
then / = 0. i.e., `
5
= = 4`
3
or 2`
5
.
(ii) If = 2`
4
then / = 0. 1 so `
5
= 2`
4
. `

#2`
4
. or A

#2`
4
.
(iii) If = `
m
with : 12 then / 8, so `
5
= `
m
. /`

#`
m
. or
A

#(/ 1)`

#`
m
with 1 / 8. and the number of families of
Sasaki-Seifert structures is nite.
Proof. The only part of this theorem that is not a consequence of the results
mentioned is (iii) and this follows directly from the second statement of Lemma
10.2.16. The niteness is a consequence of the fact that the branch divisor
contains only one irreducible component again by the second statement of Lemma
10.2.16.
Concerning the existence of positive Sasaki-Seifert structures much more is
known in the case of Smale manifolds. Corollary 10.2.20 says that this problem is
completely understood in the case of simply connected rational homology 5-spheres.
Furthermore, Table B.4.1 lists all rational homology 5-spheres admitting a positive
Sasakian structure that can be realized as the link of a BP polynomial. Remarkably
these together with the rational homology spheres in the table of Example 10.3.7
show that all admissible rational homology 5-spheres with a positive Sasakian struc-
ture can be realized by the link of a weighted homogeneous polynomial. However, it
is probably not true that all positive Sasaki-Seifert structures on admissible simply
connected rational homology 5-spheres can be realized by links of weighted homo-
geneous polynomials. As indicated by Open Problem 10.2.2 and as discussed earlier
the classication of all deformation types of positive Sasakian Seifert structures on
admissible rational homology 5-spheres is still open and perhaps even untractable.
Let us now turn to the more general case of Smale manifolds /`

#. where
is a rational homology 5-sphere with an admissible torsion group and / = 0. As
remarked earlier even though `
30n
does not admit a positive Sasakian structure,
/`

#`
30n
does for certain / as evidenced by Table B.4.2. Nevertheless, the
problem as to which ` o
0
B
admit positive Sasakian structures is still very much
open. Much less is known about simply connected 5-manifolds in o
+
B
`(o
+
B
)
0
. Some
regular examples will be given in Section 10.4.
Open Problem 10.2.3: Determine the sets (o
+
B
)
0
and o
+
B
precisely.
A special case of particular interest in the general case is when c
orb
1
(Z) is a ra-
tional multiple of c
1
(`7). or equivalently when the Sasaki-Seifert structure is anti-
canonical, recall Denition 7.5.24. This condition automatically holds when ` is a
rational homology sphere by Proposition 7.5.29, or for any link of a weighted homo-
geneous polynomial by (or more generally for a weighted complete intersection) by
Proposition 9.2.4. The terminology preSE was introduced by Kollar in [Kol06b]
for the anticanonical case. The next theorem due to Kollar [Kol05, Kol06b] re-
duces the 93 cases of Proposition 10.2.23 to 19 in the case of anticanonical Sasakian
Seifert structures.
10.2. SASAKIAN STRUCTURES AND THE TOPOLOGY OF 5-MANIFOLDS 349
Theorem 10.2.25: Let ` o
+
B
with H
2
(`. Z) = Z
k
(Z
m
)
2
for some :
12 and / = /
2
(`) 0, and suppose that the Sasaki-Seifert structure F(T

) is
anticanonical. Then F(T

) is one of the cases listed in the table below. More

precisely, we list all such anticanonical Sasaki-Seifert structures together with the
corresponding 5-manifold and its representation as a link of weighted homogeneous
polynomial when it exists.
` Z Link
8`

#`
m
1
1,1,1,1,1,1,1,1
(CP
2
) 1(2. 3. 6. 6:)
7`

#`
m
1
1,1,1,1,1,1,1
(CP
2
) 1(2. 4. 4. 4:)
6`

#`
m
1
1,1,1,1,1,1
(CP
2
) 1(3. 3. 3. 3:)
5`

#`
m
1
1,1,1,1,1
(CP
2
)
4`

#`
m
1
1,1,1,1
(CP
2
)
4`

#`
m
1
2,2,2
(CP
1
CP
1
). gcd(:. 2) = 1 1(2. 3. 6. 3:)
3`

#`
m
1
1,1,1
(CP
2
)
3`

#`
m
1
2,2
(CP
1
CP
1
). gcd(:. 2) = 1 1(2. 4. 4. 2:)
2`

#`
m
1
1,1
(CP
2
)
2`

#`
m
1
3,3
(CP
2
). gcd(:. 3) = 1 1(2. 3. 6. 2:)
2`

#`
m
1
2
(CP
1
CP
1
). gcd(:. 2) = 1 1

(:. :. :. 2; 3:)
`

#`
m
1
1
(CP
2
)
`

#`
m
CP
1
CP
1
. gcd(:. 2) = 1
`

#`
m
1
4
(Q). gcd(:. 2) = 1 1

(:. 2:. 3:. 4; 6:)

`

#`
m
1
3
(CP
2
). gcd(:. 3) = 1 1

(:. :. 2/. 3; 4:)

Proof. Since : 12 (iii) of Theorem 10.2.24 says that any log del Pezzo sur-
face which occurs must have /
2
9 and be on the list given in Proposition 10.2.22.
We now add the condition that the Sasaki-Seifert structure be anticanonical. This
imposes stringent conditions on the Seifert o
1
-bundle. Since there is only one com-
ponent 1 in the branch divisor and 1 [ 1
Z
[ we nd using equations (10.2.5)
and (10.2.4) that our condition becomes
o[

H] +

k
c
i
1
i
=
:
d(7)
(1
S
[

H]

i
n
i
1
i
)
for some integer :. But from Proposition 10.2.23 smoothness implies gcd(c
i
. n
i
) = 1
for each i. and we also have gcd(d(7). :) = 1. Thus, in the anticanonical case we
must have n
1
= n
2
= = n
k
= d(7). and gcd(o. 1
S
) = 1. Furthermore, when o
is singular n
1
= = n
k
= d(7) = 1
S
. These conditions cut down considerably
the possibilities from the list in Proposition 10.2.22. A straightforward argument
then shows that we obtain the 15 cases in the list above when /
2
(`) 0. (The
4 remaining cases CP
2
. Q. o
5
. and CP(1. 2. 3) have /
2
(`) = 0 giving 19 cases in
all). One can then easily compute the corresponding second Betti number by the
methods of Chapter 9.
The association with the corresponding link comes from our analysis in Section
10.3 below. The cases in the table where no link is given means that one cannot
realize ` as the link of a weighted homogeneous polynomial when : 12 and
the Sasakian structure is positive. This was shown by Kollar in [Kol06b], and we
reproduce his proof: Assume the contrary, that is Z = (7. ) is a hypersurface of
degree d in a weighted projective space CP(w). Then by the adjunction formula
350 10. SASAKIAN GEOMETRY IN DIMENSIONS THREE AND FIVE
(Proposition 4.6.13) we have
1
orb

1
Z
(1
1
:
)1
[w[ d
:
H[
Z
.
where H is the hyperplane class of the weighted projective space. Since 1
[1
Z
[, we conclude that 1
Z
is proportional to H[
Z
. By the Grothendieck-Lefschetz
theorem [Gro68], this implies that 1
Z
= :H[
Z
for some : Z. In the cases
7 = 1
1
CP
2
. . . . . 1
11111
CP
2
we have : = 1, so 1
Z
H[
Z
. Now
/
0
(7. O
Z
(1
Z
)) = /
0
(CP(w). O
CP(w)
(1)) 4
since this dimension is the number of times that 1 occurs among w = (n
0
. n
1
. n
2
. n
3
).
But for the four cases for 7 above we have /
0
(7. O
Z
(1
Z
) 5. giving a contra-
diction. A similar argument works for CP
1
CP
1
.
We make note, however, that 5`

#`
m
can be written as a rather simple
link, namely 1

(2:. 3:. 4:. 7; 14:) with gcd(:. 7) = 1. except that it is negative

for : 1.
Remark 10.2.5: Note that from the table of Theorem 10.2.25 we can make a count
of the number of families of positive Sasakian Seifert structures in each case. In
all cases : 11. We see that the positive Sasaki-Seifert structure on /`

#`
m
for / = 5. 6. 7. 8 is unique up to deformations. There are precisely two families of
positive Sasaki-Seifert structures on 4`

#`
m
and 3`

#`
m
when : is odd,
and one such family when : is even. There are three families of positive Sasakian
Seifert structures on 2`

#`
m
when : is relatively prime to 6, two such families
when : is even and relatively prime to 3. two such families when : is an odd and
a multiple of 3. and only one such family when : is a multiple of 6. There are
four families of positive Sasaki-Seifert structures on `

#`
m
when : is relatively
prime to 6, three such families when : is odd and a multiple of 3. two such families
when : is even and relatively prime to 3. and only one such family when : is a
multiple of 6.
Finally we briey discuss the Smale manifolds with no torsion. These are
o
5
. /`

and they will be treated in much more detail in Chapter 11. Positive
Sasaki-Seifert structures are ubiquitous on o
5
and we discuss them further below
and in Chapter 11. For example in Section 10.3.1 we show that innitely many
such structures arise already in the classication of positive BP links (cf. Table
B.4.3 of Appendix B).The rst examples of positive Sasakian structures on the /-
fold connected sum /`

= /(o
2
o
3
) for all / were given in [BGN03a]. This is
given by the non-BP link appearing as the rst entry of the table in Example 10.3.7
below. Many other examples for particular values of / were given in [BGN03b,
BGN02b] and will be discussed in the context of Sasaki-Einstein geometry in
Chapter 11. Again, just as in the sphere case, each /`

has innitely many BP

link representatives (cf. Table B.4.4).
10.2.2. Smale-Barden Manifolds with Symmetries. So far we have fo-
cused on oT
B
B as the existence of a locally free circle action is the necessary
pre condition for the Sasakian geometry. It is interesting to examine what happens
when ` B is a G-manifold with dim(G) 1. Not surprisingly, the larger the
symmetry group the more restrictions on the topology of `. Let us summarize all
of the known results here. Some of them were already mentioned and others will
10.2. SASAKIAN STRUCTURES AND THE TOPOLOGY OF 5-MANIFOLDS 351
be discussed in the following section. However, we feel it is interesting to put all
them together.
Proposition 10.2.26: Let ` B.
(i) If ` oT
B
then H
2
(`. Z) satises condition K. Conversely, any such
manifold admits innitely many inequivalent locally free circle actions.
(ii) If ` admits a free circle action then H
2
(`. Z)
tor
= 0. Conversely, any
such ` admits a free circle action.
(iii) If ` is homogeneous than ` is A
1
. o
5
or `

.
(iv) If ` is a cohomogeneity one manifold than ` is A
1
. o
5
. `

, or A

and there are innitely many cohomogeneity one structures in each case.
(v) If ` admits an eective T
3
-action then H
2
(`. Z)
tor
= 0. Conversely,
any such ` admits innitely many inequivalent T
3
-actions.
As discussed previously case (i) is due to Kollar [Kol06a]. Cases (ii) and
(v) are due to Geiges [Gei91] and Oh [Oh83], respectively and will be discussed
in the Section 10.4. All compact, homogeneous, 1-connected, 5-manifolds were
classied in [Gor80]. It is clear that A
1
. o
5
. `

and A

have innitely many

cohomogeneity one structures. In the case of o
5
. `

and A

there are even

innitely many cohomogeneity one Einstein metrics [Boh98, GHY04]. But the
fact that there are no other such manifolds and a complete classication of all
inequivalent cohomogeneity one diagrams has been carried out only recently by
Hoelcher [Hoe07]. We nish this chapter with a couple of open problems.
Open Problem 10.2.4: What can one say about ` B admitting an eective
T
2
-action which does not extend to an eective T
3
-action? In particular, what can
one say about H
2
(`. Z)
tor
in this case?
First, note that A
1
= ol(3)oO(3) admits an aective T
2
-action and there-
fore such a condition no longer implies that H
2
(`. Z)
tor
= 0. Secondly, we do get
many examples of Sasakian manifolds with T
2
-actions which are not toric. One
such example is the BP link 1(2. 2. 2. /) with the naturally induced Sasakian struc-
ture which is oO(3) l(1)-invariant and it is of cohomogeneity 1. More generally
all the links 1(2. 2. j. c) (gcd(j. c) 1)(o
2
o
3
) have Sasakian structures with
T
2
-symmetry.
These examples suggest the partial answer to the problem which was recently
communicated to us by Kollar. viz.
Proposition 10.2.27: Let ` o
+
B
. and assume that the dimension of the isome-
try group of the Sasakian structure is greater than one. Then ` = /(o
2
o
3
) for
some non-negative integer / (again / = 0 means ` = o
5
).
Proof. Let ` have Sasakian structure o = (. . . o). If the dimension of
Isom(`. o) 1. then by Theorem 8.1.18 either ` is o
5
with its round sphere
metric or dim Aut(o) = dim Isom(`. o). In both cases we have dim Aut(o) 2.
By the Approximation Theorem 7.1.10 we can assume that o is qusi-regular, and
that the additional element in aut(o) generates a circle. Since
1
(`) = 0 the
exact sequence 8.1.4 implies that this circle action projects to a circle action on
the algebraic variety A = `T

preserving the induced Kahler structure. Thus, it

extends to a C

action on the orbifold (A. ). Now since o is positive, (1

X
+)
is also positive, so A is a rational surface. Writing =

(1
1
mi
)1
i
it suces to
show by Theorem 4.7.14 that all the 1
i
are rational curves.
352 10. SASAKIAN GEOMETRY IN DIMENSIONS THREE AND FIVE
So suppose that o(1
j
) 0 for some ,. Since each 1
i
is invariant under the
C

action and a curve of genus o 0 has no non-trivial C

actions, 1
i
must lie
in the xed point set of the action. Thus, we can assume that a general point of
1
j
is weakly attracting xed point. Then Theorem 4.3 of [BB73] implies that A
contains a Zariski open subset l
j
containing 1
j
such that the C

action retracts
l
j
to 1
j
. That is, there is a C

-equivariant morphism l
j
1
j
whose bres are
isomorphic to C. Thus, there is a non-constant rational map A 1
j
. implying
that 1
j
is rational which gives a contradiction.
Open Problem 10.2.5: Out of the cohomogeneity one Barden-Smale manifolds
we know that A
1
cannot be Sasakian. After a complete classication of all coho-
mogeneity one diagrams for o
5
. `

and A

is accomplished it is natural to ask

which of them admit an invariant Sasakian structure? Are there cohomogeneity
one diagrams with no compatible Sasakian structure at all?
10.3. Sasakian Links in Dimension Five
In this section we present a more systematic discussion of the Smale manifolds
that can be realized as links of weighted homogeneous polynomials in C
4
. The set
of such manifolds was denoted by \H1
5
in Chapter 9, and as seen in Propositions
9.2.2 and 9.2.4 such links admit a natural anticanonical, canonical, or null Sasakian
structure, and they are spin by (iv) of Theorem 9.3.2. Summarizing we have
Proposition 10.3.1: \H1
5
o
0
B
. Furthermore, every Sasakian structure realized
as a link of a weighted homogeneous polynomial is either anticanonical, canonical,
or null.
Just as in dimension three (see Proposition 10.1.4) there is a 5-dimensional
version of the classication problem of links of quasi-smooth weighted homogeneous
polynomials. As mentioned in Chapter 5 such a list of 19 types of polynomials was
given recently by Yau and Yu [YY05] and their table is reproduced as Table B.5.1
in Appendix B. In fact, we have a corollary of Theorem 5.4.18:
Theorem 10.3.2: Let 1 C[.
0
. .
1
. .
2
. .
3
] be a weighted homogeneous polynomial
with weights and degree (w; d) and with only an isolated singularity at 0. Then
the associated link 1
f
1

(w; d) is equivariantly dieomorphic to one of the links

given by the polynomials listed in Table B.5.1. of the Appendix B.
Note that there is a lot of duplication in Table B.5.1. On the other hand, it
follows from Theorem 10.3.2 that Table B.5.1. properly subsumes other tables of
Appendix B: namely, Reids examples in Table B.1.1, the well-formed examples
of Tables B.3.1-2 and the positive BP links classied in Tables B.4.1-4. The rst
entry in Table B.5.1 is the just the set of Brieskorn-Pham polynomials and as seen
above it is quite remarkable how much can be obtained by working with these
alone. Some important non-BP Sasakian links are presented in the context of
Sasaki-Einstein geometry in Chapter 11.
10.3.1. Positive Sasakian Links. It is a straightforward albeit tedious task
to list all the positive links that occur in Table B.5.1. This is precisely what was
done in the preprint version of [YY05] (cf. math.AG/0303302). Indeed the authors
give quite an extensive list of all rational isolated hypersurface singularities dened
by weighted homogenous polynomials in four complex variables. The positivity of
the Sasakian structure on the link is equivalent to the rationality of the singularity
10.3. SASAKIAN LINKS IN DIMENSION FIVE 353
at the origin. They do not, generally, seem to be interested in the topology of the
links or the orbifold structure of the underlying log del Pezzo surfaces. However,
in an earlier paper Yau and Yu classify all links of Table B.5.1 which are positive
and dieomorphic to the standard 5-sphere [YY02].
Next we determine those manifolds in o
+
B
which can be realized as Brieskorn-
Pham links 1(a). It is easy to get a complete classication. Let us assume that the
exponents in a = (o
0
. o
1
. o
2
. o
3
) are in non-decreasing order. We have to consider
two separate situations: First let

2
0
1
a
i
1. Then the last exponent o
3
is arbitrary
so that in each instance we get an innite series of positive links. Apart from
(2. 2. j. ), there are only six such cases: (2. 3. 3. /). (2. 3. 4. /). (2. 3. 5. /). (2. 3. 6. /).
(2. 4. 4. /), (3. 3. 3. /). One can show that 1(2. 2. j. /) has no torsion in H
2
(`. Z). It
follows that 1(2. 2. j. c) (n 1)`

. where n = gcd(j. c). For the six remaining

cases we consider subsequences depending on the modality of the fourth exponent
relative to the rst three. Secondly, when

2
0
1
a
i
< 1 we get only a nite list of 112
sporadic examples. These are 1(3. 4. 4. 4). 1(3. 4. 4. 5) and 1(a). where a is one
of the following
(2. 3. 7. /) 6 < / < 42 (2. 3. 8. /) 7 < / < 24 (2. 3. 9. /) 8 < / < 18
(2. 3. 10. /) 9 < / < 15 (2. 3. 11. /) 10 < / < 14 (2. 4. 5. /) 4 < / < 20
(2. 4. 6. /) 5 < / < 12 (2. 4. 7. /) 6 < / < 10 (2. 5. 5. /) 4 < / < 10
(2. 5. 6. /) 5 < / < 8 (3. 3. 4. /) 3 < / < 12 (3. 3. 5. /) 4 < / < 8
In the four classication tables in Appendix B.4, we separate links according to the
following properties of the second homology group:
(i) H
2
(1(a). Z) = 0 is pure torsion,
(ii) H
2
(1(a). Z) has both non-trivial free part and non-trivial torsion piece,
(iii) H
2
(1(a). Z) = 0.
(iv) H
2
(1(a). Z) = 0 has no torsion subgroup.
In addition, in the case of vanishing torsion we separate out the sporadic links.
Out of the 112 sporadic examples most are actually spheres, more precisely, in 79
cases H
2
(1(a). Z)
tor
= 0 of which 63 are dieomorphic to o
5
and 16 are dieomor-
phic to some n`

. These are split between in Tables B.4.3-4. The 33 sporadic links

with non-trivial torsion are rational homology spheres in 25 cases and are included
in Table B.4.1. The remaining 8 sporadic links with /
2
0 and non-trivial torsion
can be found in Table B.4.2. The computation of H
2
(1(a). Z) gives the following
classication:
Theorem 10.3.3: Let 1(a) be a Brieskorn-Pham link and suppose 1(a) o
+
B
. If
H
2
(1(a). Z)
tor
= 0 then 1(a) is either one of the rational homology spheres listed
in Table B.4.1 of Appendix B, or /
2
(1) 0 and 1(a) is one of the Smale manifolds
listed in Table B.4.2. If H
2
(1(a). Z)
tor
= 0 then 1(a) is either o
5
with its positive
Seifert bered structure listed in Table B.4.3 or /
2
(1) 0 and 1(a) is one of the
Smale manifolds listed in Table B.4.4. In each case when H
2
(1(a). Z)
tor
= 0 we
give the number of distinct positive Sasakian Seifert structures
SJ
(if nite, or
else
SJ
= ).
As mentioned above all positive links of isolated hypersurface singularities from
weighted homogeneous polynomials was given in [YY02]. So our Table B.4.3 con-
tains only the subclass of type I links (i.e., BP links) given in the 10 page list in
354 10. SASAKIAN GEOMETRY IN DIMENSIONS THREE AND FIVE
[YY02]. The computations for rational homology 5-spheres realized by Brieskorn-
Pham links 1(a) are very easy due to a simplication of Orliks formulae (9.3.17)-
(9.3.19) in which case Orliks Torsion Conjecture 9.3.15 is known to hold.
Theorem 10.3.4: Let a = (o
0
. o
1
. o
2
. o
3
) and 1(a) be is a rational homology 5-
sphere. Without loss of generality assume that o
0
= / is the only isolated vertex of
G(a). Then
H
2
(1(a). Z) = (Z
k
)
2g
.
where
2o = 2

i<j
gcd(o
i
. o
j
) +
gcd(o
1
. o
2
) gcd(o
1
. o
3
) gcd(o
2
. o
3
)
gcd(o
1
. o
2
. o
3
)
.
Proof. Let 1(a) be a BP link which is a rational homology 5-sphere. By the
Brieskorn Graph Theorem 9.3.18 we must have at least one isolated point in G(a),
say o
0
= /. If there is more than one isolated vertex the torsion vanishes so that
we can assume that there is exactly one such vertex. The statement follows by
Proposition 9.3.16 from the calculation of the torsion group with Orliks algorithm.
We compute
c

= 1. c
0
= gcd(o
1
. o
2
. o
3
). c
1
= c
2
= c
3
= 1 .
c
0,i
=
gcd(o
j
. o
k
)
gcd(o
1
. o
2
. o
3
)
. c
i,j
= 1 .
c
1,2,3
= /. c
0,j,k
=
gcd(o
1
. o
2
. o
3
)o
i
gcd(o
j
. o
i
)gcd(o
i
. o
k
)
.
However, the only non-vanishing /
i
1
,...,i
s
is /
1,2,3
= 2o.
Note that any / can occur and which o occur depends on the choice of /.
The Theorem 10.3.4 can be conveniently restated as follows: let 1(a) be a rational
homology 5-sphere with non-trivial torsion. The graph G(a) consists of an isolated
point and a triangle. Let us label the triangle with numbers: the edges with
gcd(o
i
. o
j
) and the face with gcd(o
1
. o
2
. o
3
). Then 2o 2 equals the product of
the edge numbers divided by the face number minus the sum of the edge
numbers. The general case when /
2
(1(a)) = 0 is only slightly more complicated.
Exercise 10.4: Let 1
f
= 1(o
1
. o
2
. o
3
) be a 3-dimensional sublink of 1(o
0
. o
1
. o
2
. o
3
).
Show that 2o of Theorem 10.3.4 equals to the rst Betti number of 1
f
which, in
turn, is twice the genus of the orbifold Riemann surface A
f
which is the transverse
space of 1
f
.
Computing the homology of more general links is only slightly more involved.
Let 1(o
0
. o
1
. o
2
. o
3
) be a BP link and let G(a) be its Brieskorn graph. We can
easlily construct a labelled polytope, also denoted by G(a). where in addition all
six edges and all four faces should be labelled with the corresponding gcds. This
give a tetrahedron
/.-, ()*+
a
0
G(a)=

/.-, ()*+
a
1

/.-, ()*+
a2
/.-, ()*+
a3
where we also label the interior of the tetrahedron with t =
a0a1a2a3
lcm(a0,a1,a2,a3)
. The re-
lation with the usual Brieskorn graph introduced in Chapter 9 is simple enough: we
just need to remove all edges of the tetrahedron labelled by 1 to give the Brieskorn
10.3. SASAKIAN LINKS IN DIMENSION FIVE 355
graph. Whether one uses a labelled polytope or a Brieskorn graph is a choice dic-
tated by convenience. In the situation of Theorem 10.3.4, for instance, it is more
convenient to use the Brieskorn graph which corresponds to removing the tree edges
meeting at o
0
in the polytope. However, in the more general case, it is advantageous
to keep all edges and faces labelled by 1s. We shall refer to such polytopes G(a) as
labelled Brieskorn polytopes. We get the following generalization of Theorem 10.3.4:
Theorem 10.3.5: Let 1(a) = 1(o
0
. o
1
. o
2
. o
3
) be a BP link with labelled Brieskorn
polytope G(a). Then
H
2
(1(a). Z) = Z

(Z
m
0
)
2g
0
(Z
m
1
)
2g
1
(Z
m
2
)
2g
2
(Z
m
3
)
2g
3
.
where
(i) m = (:
0
. :
1
. :
2
. :
3
) are the reduced indices of the vertices a. i.e.,
:
i
is a part of o
i
which is not a part of any o
j
in the remaining 3 vertices.
(ii) g = (o
0
. o
1
. o
2
. o
3
). where the integer 2o
i
2 is computed from the face
opposite to the vertex o
i
and it is the product of the edge numbers
divided by the face number minus the sum of the edge numbers.
(iii) + 3 t is computed as the sum of the six edge numbers minus the
sum of four products of edge numbers divided by the face numbers
(one such term for each face).
The decoding of the labelled Brieskorn polytope is just calculating the triple
(m. g. ) which yields H
2
(1(a). Z). Hence, Theorem 10.3.5 splits the calculation
with Orliks algorithm into two stages: labelling and decoding.
a = G(a) = H
2
(1(a). Z) .
Note that :
i
s are always positive integers and, by construction, they must be
pairwise relatively prime. If some :
i
= 1 then Z
mi
is trivial. Likewise, 2o
i
is
a non-negative integer and if some o
i
= 0 then Z
0
mi
is trivial regardless of the
corresponding :
i
.
Exercise 10.5: Prove Theorem 10.3.5.
Remark 10.3.1: In the case of more general links one can organize the information
coming from Orliks algorithm the same way it is presented in Theorem 10.3.5.
The combinatorial information is slightly more complicated as it depends on the
(u. v)-data. It can, however, be organized in a labelled rational Brieskorn polytope
G(u; v) in such a way that H
2
(1
f
. Z) is computed in almost the same fashion, i.e.,
the second arrow in (u; v) = G(u. v) = H
2
(1
f
. Z) is essentially the same as in
the BP case.
Exercise 10.6: In analogy with the BP case dene G(u; v) and describe Orliks
Algorithm as labelling and decoding: (u; v) = G(u. v) = H
2
(1
f
. Z).
Example 10.3.6: We illustrate how to decode the labelled Brieskorn polytope by
considering a = (o
0
. o
1
. o
2
. o
3
) = (27. 9. 3. 3). Then the reduced indices are m =
(:
0
. :
1
. :
2
. :
3
) = (3. 1. 1. 1). There is one edge labelled by 9 and ve others
labelled by 3. All four faces are labelled by 3 and the interior has t = 81. Now, one
can easily see that g = (o
0
. o
1
. o
2
. o
3
) is given by 2g (2. 2. 2. 2) = (0. 0. 12. 12).
Hence, g = (1. 1. 7. 7). Finally, + 3 = (9 + 5 3) 2(27 + 9) + 81 = 30. Thus
H
2
(1. Z) = Z
30
Z
2
3
. Let us revisit one of the links of Example 9.1 but write a
using the prime number decomposition a = (2 3 13. 2 5 17. 3 11 19. 7 17 19).
Then the reduced indices m = (13. 5. 11. 7). There are four vertices labelled by
356 10. SASAKIAN GEOMETRY IN DIMENSIONS THREE AND FIVE
2. 3. 19. 17 and two labelled by 1. All four faces are labelled by 1 and t = 1938. Our
theorem easily gives g = (144. 18. 8. 1) while
+ 3 = (2 + 3 + 17 + 19 + 1 + 1) (2 3 + 17 19 + 3 19 + 2 17) + 1938 = 1561.
Generally speaking, the polytopes with more edges labelled by 1 lead to a
simpler recipe for the second homology. For example, when
/.-, ()*+
a
0
G(a)=
/.-, ()*+
a
1
/.-, ()*+
a2
/.-, ()*+
a3
then the formulas of Theorem 10.3.5 easily reduce to H
2
(1(a). Z) = Z

. where
= (gcd(o
1
. o
3
) 1)(gcd(o
0
. o
2
) 1). The special examples corresponding to this
type are the positive BP links 1(2. 2. j. c) of Table B.4.4.
A similar analysis can be done with any of the Types in Table B.5.1. Since a
complete picture is quite involved we content ourselves by considering only a few
more examples of Type II in Table B.5.1.
Example 10.3.7: For a 4-tuple a = (o
0
. o
1
. o
2
. o
3
) of positive integers consider the
weighted homogeneous polynomial
1(z) = .
a0
0
+.
a1
1
+.
a2
2
+.
0
.
a3
3
.
This is a simplest modication of a BP type which is listed in Table B.5.1 as Type
II. One can easily see that
w = (do
0
. do
1
. do
2
. (o
0
1)do
3
o
0
).
Hence, the link 1
f
is Fano if and only if
1
o
0
+
1
o
1
+
1
o
2
+
1
o
3

1
o
0
o
3
1 .
Just as in the BP case one can enumerate all the examples. However, many of
them are merely deformations of the BP links given in the tables in Appendix B.4
(the duplications we have already mentioned). Below are some interesting examples
which are not of BP type
` a w d
n`

(n + 1. n + 1. n + 1. 1) (1. 1. 1. n) n + 1
2`

#`
k
(3. 3. 3. /). (/. 2) = 1 (/. /. /. 2) 3/
`

#`
k
(3. 6. 2. /). (/. 2) = 1 (2/. /. 3/. 4) 6/
`
k
(6. 2. 3. /). (/. 5) = 1 (/. 3/. 2/. 5) 6/
`

#`
k
(4. 2. 4. /). (/. 3) = 1 (/. 2/. /. 3) 4/
2`
4
(10. 5. 2. 4) (4. 8. 20. 9) 40
2`
4
(5. 5. 4. 2) (4. 4. 8. 5) 20
`

#2`
4
(6. 6. 2. 4) (4. 4. 12. 5) 24
3`
3
(4. 4. 3. 2) (3. 3. 4. 3) 12
n`
2
(2n + 1. 2n + 1. 2. 2) (2. 2. (2n + 1). 2n) 2(2n + 1)
One could consider an arbitrary link 1
f
of some weighted homogenous polyno-
mial 1 of degree d and with weights w = (n
0
. n
1
. n
2
. n
3
) of Table B.5.1.
Open Problem 10.3.1: Enumerate all positive non-BP links of Table B.5.1, i.e.,
all links of Type II-XIX with [w[d 0, and compute H
2
(1
f
. Z) for each such link,
thus determining the corresponding Smale manifold. Note that we could exclude all
10.3. SASAKIAN LINKS IN DIMENSION FIVE 357
the well-formed links as these have no torsion in H
2
. (A classication of well-formed
links, under some additional assumptions, was carried out in [BGN03b]).
10.3.2. Null Sasakian Links. In the null case there appears to be only a -
nite number of compact simply connected manifolds admitting null Sasakian struc-
tures in each dimension. The key here is Proposition 4.7.11 of Kollar which states
that a Seifert bundle over a Calabi-Yau orbifold necessarily has no branch divisor.
Before discussing null Sasakian links outright, we give two general results concern-
ing null Sasakian structures in dimension ve. The rst of these results shows that
niteness indeed holds in dimension ve. This result is closely related to Theorem
5.4.5, but here we give a proof that is independent of that theorem. Both Theorem
10.3.8 and Corollary 10.3.11 were rst given in [BGM06].
Theorem 10.3.8: Let o be a null Sasakian structure on a compact simply con-
nected 5-manifold `. Then 2 /
2
(`) 21 and H
2
(`. Z) is torsion-free, and ` is
dieomorphic to /`

= /(o
2
o
3
). where / = /
2
(`). Furthermore, if /
2
(`) = 21
then o is regular and ` is dieomorphic to 21(o
2
o
3
). and A = `T

is a K3
surface.
Proof. A null Sasakian structure must be quasi-regular by (iii) of Theorem
8.1.14. So ` is a o
1
-orbibundle over a orbifold K3 surface. By Proposition 4.7.11
there are no branch divisors and 1
orb

1
X
is trivial. So the geometric genus
j
g
(A) and the transverse geometric genus j
g
(T

) both equal 1. This implies that

/
2
(A) is at least 3. so /
2
(`) is at least 2. This give the lower bound. The upper
bound comes from the fact that a smooth K3 is a minimal resolution of the singular
K3 surface A as shown in Corollary 10.4 of [Kol05]. This corollary also shows that
if /
2
(`) = 21. then A must be a smooth K3 surface, which in the absence of branch
divisors implies that the Sasakian structure is regular. That H
2
(`. Z) is torsion
free follows from Proposition 4.7.11 and Theorem 4.7.14. Since c
1
(T

) = 0. the
manifold ` is spin by Theorem 7.5.27, and so must be of the form given by the
Smale-Barden classication Theorem 10.2.3.
We mention that the lower bound in Theorem 10.3.8 holds with the weaker
condition H
1
(`. R) = 0. We now have
Corollary 10.3.9: o
5
and o
2
o
3
or any quotient by a nite group do not admit
null Sasakian structures.
Compact 5-manifolds ` with null Sasakian structures can be easily obtained
from the list of 95 orbifold K3 surfaces given in the Table B.1 in Appendix B.
Interestingly, we see from this table that there are gaps in the range of possible
null Sasakian 5-manifolds given by Theorem 10.3.8, namely /
2
(`) = 2. 17 do not
occur.
Example 10.3.10: As discussed in Section 5.4.2, in 1979 Reid [Rei80] produced
a list of 95 (possibly singular) K3 surfaces A given as well-formed hypersurfaces
in weighted projective space CP(n
0
. n
1
. n
2
. n
3
). This list is given in Table B.1
of Appendix B. With each of Reids examples one can consider the associated 5-
dimensional link 1
f
which by Theorem 10.3.8 are all dieomorphic to some /-fold
connected sum of o
2
o
3
. The integer / = /
2
(1
f
) can be computed in each case by
the Milnor-Orlik procedure described in Section 9.3.1, and Table B.1 includes this
information giving /
2
(A) which equals /+1. For instance 1(4. 4. 4. 4) 21(o
2
o
3
)
is regular and it is a circle bundle over the smooth 13-surface A
4
CP
3
. The only
358 10. SASAKIAN GEOMETRY IN DIMENSIONS THREE AND FIVE
other regular circle bundle over a non-singular 13 surface on Reids list is also a
BP link, namely 1(6. 6. 6. 2) 21(o
2
o
3
). which is a regular circle bundle over
A
6
CP(1. 1. 1. 3) given by the vanishing of 1(z) = .
6
0
+.
6
1
+.
6
2
+.
2
3
. However, A
4
and A
6
are not the same as complex algebraic varieties, and the induced Sasakian
structures on 1(4. 4. 4. 4) and 1(6. 6. 6. 2) belong to inequivalent deformation classes
(of Sasakian structures).
As one sees from the Table B.1 of Appendix B all of Reids examples have
3 /
2
(1
f
) 21. Hence, for instance, the link 1
f
with 1(z) = .
8
0
+.
6
1
+.
3
2
.
0
+.
3
3
must
be dieomorphic to 3(o
2
o
3
), whereas, the BP link 1(2. 3. 12. 12) is dieomorphic
to 20(o
2
o
3
). It is interesting that neither /
2
(1) = 2 nor 17 occur on Reids list, so
we would need to consider possible singular K3 surfaces that are not hypersurfaces
in CP(n
0
. n
1
. n
2
. n
3
) to ll these gaps. For instance, Fletcher gives a list of 84
examples of codimension 2 complete intersections [IF00], but here he assumes only
canonical singularities and that [w[ 100. so there could be even more examples.
Exercise 10.7: Show that all the null links of Table B.5.1 (i.e., the ones with
[w[ = d) are equivalent to Reids K3 surfaces of Table B.1.
Summarizing the results for the computation of the second Betti number of all
95 orbifold K3 surfaces given in Table B.1 of Appendix B we have
Corollary 10.3.11: /(o
2
o
3
) admit null Sasakian structures for 3 / 21 and
/ = 17. Furthermore, the count of the number of null Sasaki-Seifert structures up
to deformation given by Reids list can be read o from Table B.1.
The question which of the simply connected compact spin 5-manifolds on
Smales list admit null Sasakian structures is related to classifying all orbifold K3-
surfaces which is still open. The more modest interesting question is whether / = 2
or 17 can occur. This leads us to:
Open Problem 10.3.2: Find examples of null Sasakian structures on 2(o
2
o
3
)
and 17(o
2
o
3
). or prove that none can exist.
In higher dimensions there are many, though probably nite, null Sasakian
structures. Interesting examples include the over 6000 Calabi-Yau orbifolds in
complex dimension 3 [CLS90].
10.3.3. Negative Sasakian Links. We nish this section with a discussion
of negative Sasakian structure. These appear to be the least constrained in terms
of what kind of H
2
(`. Z) one should expect when ` is a Smale manifold. On the
other hand, other than examples we do not really have any classication result.
The Brieskorn-Pham links produce already very many examples. The simplest
ones are circle bundles over the Fermat hypersurfaces of degree d in P
3
which are
described in Example 5.4.1. For d 5 we get regular negative Sasakian structures
on /(o
2
o
3
) for / = (d 2)(d
2
2d + 2) + 1. These begin with / = 52 and grow
rapidly.
Example 10.3.12: Consider 1(/. /. / + 1. j). Such a link is negative as long as
/. j 3 or with / = 3 and j 12. Moreover, 1(/. /. / + 1. j) is a 5-sphere as long
as j is prime to both / and / + 1 which can easily be arranged.
Proposition 10.3.13: Every negative Sasakian structure on o
5
is non-regular and
there exists innitely many inequivalent negative Sasakian structures.
10.3. SASAKIAN LINKS IN DIMENSION FIVE 359
Proof. The existence of innitely many non-regular negative Sasakian struc-
tures on o
5
follows immediately from Example 10.3.12. To prove the rst statement
we assume that o
5
has a regular negative Sasakian structure. Then we have a circle
bundle o
5
A with c
1
(A) negative,
1
(A) = 0. and /
2
(A) = 1. This is impossible
by a Theorem of Yau, cf. Theorem V.1.1 of [BPVdV84].
More generally one should be able to examine the existence of negative Sasakian
structures on Smale-Barden manifolds, and in particular, on rational homology
spheres. Unlike the positive case where we have Kollars classication Theorems
10.2.19 and 10.2.25, not much is known in general. However, we have
Theorem 10.3.14: Let ` o
B
be a rational homology sphere. Then ` is either
positive and the torsion in H
2
(`. Z) is restricted by theorem 10.2.19 or ` is neg-
ative. There exists innitely many such manifolds which admit negative Sasakian
structures but no positive Sasakian structure. There exists innitely many positive
rational homology spheres which also admit negative Sasakian structures.
Proof. To show that there are negative rational homology 5-spheres which
cannot admit any positive Seifert bration one can consider the BP link 1(j. j. j. /)
with (j. /) = 1 and 13j + 1/ < 1. By Orliks formula this has H
2
(`. Z)
tors
=
(Z
k
)
d
, where d = j(j 3) + 2. When j 3 and / 1 this produces torsion
disallowed by Kollars Theorem 10.2.19.
To get examples of rational homology 5-spheres which admit both positive and
negative Sasaki-Seifert structures we consider the links of [BG02]. One can easily
see, for example, that `
k
admits both positive and negative Sasakian structures
for every prime /.
Open Problem 10.3.3: Determine which simply connected rational homology 5-
spheres admit negative Sasakian structures.
Open Problem 10.3.4: Classify the negative Sasaki-Seifert structures on simply
connected rational homology 5-spheres.
These problems are very dicult in general, and Problem 10.3.4 is clearly more
dicult than Problem 10.3.3. Nevertheless, it is rather easy to produce examples
of negative Sasakian structures on /(o
2
o
3
); however, we havent been able to
nd one nice sequence which does this, or even a sequence that works for odd /.
The examples below give the essential ideas.
Example 10.3.15: This is example 10.3.12 with j = / + 1. Now the BP link is
1(/. /. / + 1. / + 1) and it is easy to see that the induced Sasakian structures are
negative as long as / 4. Here the Milnor-Orlik method gives /
2
(1(/. /. / +1. / +
1)) = /(/ 1). This gives negative Sasakian -Einstein structures on /(o
2
o
3
)
for innitely many /. but with larger and larger gaps as / grows. For low values of
/. for example, / = 4 and 5. we get negative Sasakian structures on 12(o
2
o
3
).
and 20(o
2
o
3
). respectively.
Example 10.3.16: Consider the BP links 1(j. c. :. jc:) such that j. c. : Z
+
are
pairwise relatively prime. The transverse space is a surface A
pqr
P(j:. c:. jc. 1)
and it is automatically well-formed, i.e., it has only isolated orbifold singularities.
Hence, 1(j. c. :. jc:) is dieomorphic to /(o
2
o
3
). where
/ = /
2
(1(j. c. :. jc:)) = (jc: jc j: c: 1) +j +c +: .
360 10. SASAKIAN GEOMETRY IN DIMENSIONS THREE AND FIVE
Now, 1(j. c. :. jc:) is negative when the term in parenthesis is positive, i.e., when
jc + j: + c: + 1 < jc:. To simplify we can consider 1(2. 3. :. 6:) with : prime to
both 3 and 2. 1(2. 3. 7. 42) is one of the examples on Reids list. But from : 7 we
get innitely many examples with second Betti number /
2
= 2(: 1). For example
for : = 11. we see that 1(2. 3. 11. 66) is dieomorphic to 20(o
2
o
3
).
Notice that both Examples 10.3.15 and 10.3.16 produce negative Sasakian
structures on /(o
2
o
3
) for even / only. We havent yet found a nice series that
gives metrics for odd values of /. However, we can obtain negative Sasakian struc-
tures for odd / if we consider the more general weighted homogeneous polynomials.
Example 10.3.17: Here we give a few examples only to illustrate the method.
A much more extensive list can be generated with the aid of a computer a la
[JK01a, BGN03b]. We begin with two inequivalent twin negative Sasakian
structures on o
2
o
3
. Consider the weighted homogeneous polynomials
.
21
0
.
1
+.
5
1
.
2
+.
3
2
.
0
+.
2
3
. .
21
0
.
1
+.
5
1
.
0
+.
3
2
.
1
+.
2
3
.
Both have degree 316 and both give links dieomorphic to o
2
o
3
. The weight
vectors are w = (13. 43. 101. 158) and w = (11. 61. 85. 158). respectively. So [w[
d = 1 giving negative Sasakian structures on o
2
o
3
.
Similarly, the polynomials
.
20
0
+.
3
1
.
3
+.
3
2
.
1
+.
2
3
.
0
.
10
0
+.
4
1
.
2
+.
2
2
.
3
+.
2
3
.
0
have degree 40 and 159. respectively. They give negative Sasakian structures on
7(o
2
o
3
) and 2(o
2
o
3
).
We have
Corollary 10.3.18: 7(o
2
o
3
). 12(o
2
o
3
) and 20(o
2
o
3
) all admit positive,
null, and negative Sasakian structures.
Open Problem 10.3.5: Show that /(o
2
o
3
) admit negative Sasakian structures
for all /. or show that this is not true, and determine precisely for which / it is
true.
Open Problem 10.3.6: Determine which torsion groups correspond to Smale-
Barden manifolds admitting negative Sasakian structures.
10.4. Regular Sasakian Structures on 5-Manifolds
We begin with the homogenous case. Every compact Sasakian homogeneous
manifold is regular, in particular. Furthermore, the base of the circle bundle has
to be a compact homogenous Kahler surface. Using Theorem 8.3.6 and Example
7.6.10 we easily get
Theorem 10.4.1: Let ` o
B
be a homogenous Sasakian space. Then ` is one
of the following
(i) o
5
with the standard Sasakian structure bering over CP
2
.
(ii) one of the Wang-Ziller manifolds `
1,1
k,l
of Example 7.6.10 bering over
CP
1
CP
1
and dieomorphic to o
2
o
3
.
Since the classication of compact complex surfaces uses the concept of mini-
mal models we do the same for regular Sasakian structures on compact 5-manifolds.
Let o = (. . . o) be a regular Sasakian structure on a compact 5-manifold `
5
.
10.4. REGULAR SASAKIAN STRUCTURES ON 5-MANIFOLDS 361
Then by Theorem 7.5.1 `
5
is the total space of an o
1
-bundle over a smooth com-
pact polarized algebraic surface A
2
. The polarization is determined by a choice of
integral Kahler class. The regular Sasakian structures on compact 5-manifolds are
classied by picking out the algebraic surfaces in the Enriques-Kodaira classication
[BPVdV84, BHPVdV04] in complex surface theory. We have
Theorem 10.4.2: Let o = (. . . o) be a regular Sasakian structure on a com-
pact 5-manifold `
5
. Then `
5
is the total space of an o
1
-bundle over a projective
algebraic surface A
2
which is obtained by blowing-up / = 0. 1. . . .-times one of the
following minimal surfaces:
(i) a minimal rational surface: CP
2
or a Hirzebruch surface S
n
with n
0. n = 1.
(ii) a ruled surface of genus o 1.
(iii) an algebraic K3 surface.
(iv) an Enriques surface.
(v) an algebraic 2-torus (Abelian variety of dimension 2).
(vi) a hyperelliptic surface.
(vii) an algebraic minimal properly elliptic surface.
(viii) a minimal surface of general type.
Recall that a compact complex surface is minimal if it does not contain a (1)-
curve. Thus, we can say that a regular Sasakian structure o on `
5
is minimal
if A
2
is minimal. In order to describe regular Sasakian structures on simply con-
nected 5-manifolds, we rst discuss which algebraic complex surfaces are simply
connected. First we briey discuss elliptic brations following [GS99]. The com-
plex surface CP
2
#9CP
2
admits elliptic brations of the form : CP
2
#9CP
2
CP
1
with generic bres an elliptic curve, and as such it is denoted by 1(1). Given el-
liptic brations
i
: o
i
C
i
we can form the bre sum o
1
#
f
o
2
as follows: Let t
i
be points of C
i
with generic bres 1
i
=
1
i
(t
i
), and take regular neighborhoods

i
of these bres in o
i
of the form 1
2
T
2
. Glue the manifolds o
i

i
together
with an orientation reversing dieomorphism on the boundary 3-tori T
3
i
. This gives
an elliptic bration : o
1
#
f
o
2
C
1
#C
2
. If one of the original elliptic brations
contains a cusp bre, then the resulting 4-manifold is independent of the gluing dif-
feomorphism. Now it is known that 1(1) admits elliptic brations containing cusp
bres, so we obtain well dened 4-manifolds by dening 1(n) = 1(n 1)#
f
1(1)
inductively. Note that 1(2) = 13. A properly elliptic surface is an elliptic surface
with Kodaira dimension one.
We are interested in which Smale-Barden manifolds admit regular Sasakian
structures. The main point is a result of Geiges [Gei91] which says that any
simply connected 5-manifold that admits a free circle action has no torsion.
Theorem 10.4.3: Let ` B. Then
(i) ` oT
R
B
if and only if ` = o
5
. /(o
2
o
3
). or A

#(/ 1)(o
2
o
3
),
where / = /
2
(`) 1.
(ii) oT
R
B
= oT
R
B
/( = oT
R
B
( = o
R
.
(iii) (oT
R
B
)
0
= (o
R
)
0
.
Proof. If ` admits a smooth free circle action then it is the total space of a
principal o
1
-bundle over a smooth base 7. By the long exact homotopy sequence
7 is simply connected, and by Poincare duality H
3
(7. Z) = 0. Then the Gysin
sequence implies that H
3
(`. Z) is torsion-free, and this is the only possible torsion
362 10. SASAKIAN GEOMETRY IN DIMENSIONS THREE AND FIVE
in a simply connected 5-manifold. The converse will be proved below by proving
that the manifolds of Theorem 10.2.3 without torsion all admit regular Sasakian
structures. This follows immediately from Proposition 10.4.4 below.
We shall give some examples below of regular Sasakian structures. Recall from
Denition 7.5.24 that there are four types of Sasakian structures, positive, negative,
null, and indenite. We have
Proposition 10.4.4: Let o = (. . . o) be a regular Sasakian structure on a
Smale-Barden manifold `. Then
(i) if o is null Sasakian then ` = 21(o
2
o
3
).
(ii) if o is positive then ` = o
5
. /(o
2
o
3
). or ` = A

#(/ 1)(o
2
o
3
),
where 1 / 8.
(iii) if o is indenite then ` = /(o
2
o
3
). or ` = A

#(/ 1)(o
2
o
3
)
for / 1.
Conversely, regular null Sasakian structures exist on 21(o
2
o
3
); regular posi-
tive Sasakian structures exist on all the Smale-Barden manifolds o
5
. /(o
2
o
3
). and
A

#(/ 1)(o
2
o
3
), where 1 / 8; and regular indenite Sasakian structures
exist on /(o
2
o
3
). and A

#(/ 1)(o
2
o
3
) for all / 1.
Proof. Any regular Sasakian manifold must be one listed in Theorem 10.4.3.
Regular null Sasakian structures occur only on 21(o
2
o
3
) since the only smooth
algebraic surface with c
1
= 0 is a K3 surface. Regular positive Sasakian structures
are all circle bundles over smooth del Pezzo surfaces which are completely classied
(see Example 3.5.11). It is straightforward to see that positive Sasakian structures
exist on all the manifolds indicated in the proposition by choosing the o
1
-bundle
appropriately as in Example 10.4.5. As for regular indenite Sasakian structures it
is clear that they occur on o
1
-bundles over the equivariant blow-up of Hirzebruch
surfaces. It is also clear that the Picard group of the base must have rank at least
two, so o
5
is excluded. The existence of regular Sasakian structures of the type
indicated and on the indicated manifolds will then follow from the construction of
circle bundles over Hirzebruch surfaces and their blow-ups. This construction is
given in Examples 10.4.5 and Theorem 10.4.12 below.
Example 10.4.5: [Circle Bundles over Hirzebruch Surfaces] Our discussion
here follows that in [BGO07]. We construct innite families of deformation classes
of toric Sasakian structures on circle bundles over Hirzebruch surfaces. The total
space ` of such circle bundles have /
1
(`) = 0 and /
2
(`) = 1. It is well-known
[Bar65] that there are precisely two simply connected closed 5-manifolds with
/
2
= 1. They are o
2
o
3
and the non-trivial o
3
-bundle over o
2
. denoted A

by
Barden. They are distinguished by their second Stiefel-Whitney class n
2
. that is,
by whether they are spin or not, or equivalently by their Barden invariant i(`).
Recall the Hirzebruch surfaces S
n
[GH78b] are realized as the projectivizations
of the sum of two line bundles over CP
1
. namely
S
n
= P

O O(n)

.
They are dieomorphic to CP
1
CP
1
if n is even, and to the blow-up of CP
2
at
one point, which we denote as

CP
2
. if n is odd. For n = 0 and 1 we get CP
1
CP
1
and

CP
2
. respectively. Now Pic(S
n
) Z Z. and we can take the Poincare duals
of a section of O(n) and the homology class of the bre as its generators. The
10.4. REGULAR SASAKIAN STRUCTURES ON 5-MANIFOLDS 363
corresponding divisors can be represented by rational curves which we denote by
C and 1. respectively satisfying
C C = n. 1 1 = 0. C 1 = 1 .
Let
1
and
2
denote the Poincare duals of C and 1, respectively. The classes
1
and
2
can be represented by (1. 1) forms
1
and
2
. respectively, so that the (1. 1)
form
l
1
,l
2
= |
1

1
+|
2

2
determines a circle bundle over S
n
whose rst Chern class
is [
l
1
,l
2
]. Since we want the total space our bundle to be simply connected, we
need to choose gcd(|
1
. |
2
) = 1. We thus have circle bundles depending on a triple of
integers (|
1
. |
2
. n). with n non-negative,
o
1
`
l
1
,l
2
,n

S
n
.
Now in order that `
l
1
,l
2
,n
admit a Sasakian structure it is necessary that
l
1
,l
2
be
a positive (1. 1)-form, that is,
l
1
,l
2
must lie within the Kahler cone /(S
n
). The
conditions for positivity are by Nakais criterion Theorem 3.5.15,
(i)
2
l1,l2
0.
(ii)

D

l1,l2
0 for all holomorphic curves 1.
which in our case give |
1
. |
2
0.
Next we determine the dieomorphism type of `
l
1
,l
2
,n
. Since the Kahler class
[
l
1
,l
2
] transgresses to the derivative of the contact form, d
l
1
,l
2
. we see that

1
=
|
2
and

2
= |
1
. where is a generator of H
2
(`
l1,l2,n
. Z) Z. Now the rst
Chern class of o
n
is [GH78b]
(10.4.1) c
1
(o
n
) = 2
1
(n 2)
2
.
which pulls back to the basic rst Chern class c
1
(T

) on `
l
1
,l
2
,n
. So the rst Chern
class of the contact line bundle T is given by
(10.4.2) c
1
(T) = [2|
2
|
1
(2 n)] .
Since `
l
1
,l
2
,n
has a regular contact structure, there is no torsion in H
2
(`
l
1
,l
2
,n
. Z).
Thus, by Theorem 10.2.3 `
l
1
,l
2
,n
is either `

or A

. depending on whether
`
l
1
,l
2
,n
is spin or not. But we have n
2
(`
l
1
,l
2
,n
) n|
1
mod 2. so `
l
1
,l
2
,n
is
dieomorphic to `

= o
2
o
3
if n|
1
is even, and to A

if n|
1
is odd. Choosing
n 2 gives countably innite sequences of regular indenite Sasakian structures
on o
2
o
3
and A

. It is easy to see [Ful93, BGO07] that for every such triple

(|
1
. |
2
. n) of positive integers `
l
1
,l
2
,n
has a toric Sasakian structure, that is a toric
contact manifold with a compatible Sasakian structure; see Sections 8.4 and 8.5.
The case of regular negative Sasakian structures is notably missing from Propo-
sition 10.4.4. The classication problem for regular negative Sasakian structures
appears to be quite dicult owing to the fact that very little is known about the
existence of simply connected algebraic surfaces of general type with low Picard
number. So we have
Open Problem 10.4.1: Classify the Smale-Barden manifolds that admit regular
negative Sasakian structures.
Even though the classication problem seems to be intractable, it is quite easy
to generate examples.
Example 10.4.6: Regular Negative Sasakian Structures. These are all circle bun-
dles over surfaces of general type. As indicated in Section 5.4.1 there are many
of these, but it appears to be dicult to obtain regular sequences that give all.
364 10. SASAKIAN GEOMETRY IN DIMENSIONS THREE AND FIVE
By constructing the circle bundles over the complex surfaces exhibited in Ta-
bles 3 and 4 we obtain regular negative Sasakian structures on /(o
2
o
3
) with
/ = 18|(| 1) + 6. 4|(4| 3) + 3. and / =
1
2
|(27| 31) for all | 2. Many other
sequences can be obtained by considering BP links with relatively prime weights.
The simplest of these sequences gives the Fermat hypersurfaces of degree greater
than 4 discussed in Example 5.4.1, but the series begins with / = 52. It appears
quite dicult to nd examples for low values of /. The lowest / known to the au-
thors admitting a regular negative Sasakian structure is / = 8. This comes from
circle bundles over a Barlow surface [Bar85] which is a complex surface of general
type that is homeomorphic to CP
2
blown-up at 8 points.
Example 10.4.7: [Circle Bundle over Fake CP
2
s] A fake projective plane is a
smooth compact complex surface A
2
having the same Betti numbers as CP
2
, but
not CP
2
. It is known that such a surface is a projective algebraic surface of general
type and a quotient of the open unit ball in C
2
by a cocompact torsion-free discrete
subgroup of the projectivization 1l(2. 1) of ol(2. 1). In 1979 Mumford [Mum79]
(see also [BPVdV84], pg. 136) gave the rst example of such a surface using p-adic
analysis, and recently there has been a classication by Prasad and Yeung [PY05].
Now
1
(A
2
) is innite, and c
1
(A
2
) < 0. Choose the Kahler class of A
2
to be the
positive generator in H
2
(A
2
. Z) Z. and let : `
5
A
2
denote the circle bundle
whose rst Chern class is this class. By the long exact homotopy sequence,
1
(`
5
)
is also innite, and by the rational Gysin sequence `
5
is a rational homology
sphere. Since c
1
(A
2
) pulls back to c
1
(T

). `
5
has a negative canonical Sasakian
structure.
This example proves the existence of regular negative Sasakian structures on
rational homology 5-spheres `
5
with an innite fundamental group. Our next
result is another consequence Yaus celebrated theorem [Yau77] (see also Theorem
V.1.1 of [BPVdV84]). It implies that if such a structure exists on a rational
homology 5-sphere, it must have innite fundamental group.
Theorem 10.4.8: Let o be a regular Sasakian structure on a rational homology
5-sphere `
5
. Then either
(i) o is positive, and `
5
is o
5
or a lens space 1(j. 1. 1) = o
5
Z
p
with o in
the standard o-deformation class.
(ii) or o is negative in which case
1
(`
5
) is innite.
Proof. First by Proposition 7.5.29 the Sasakian structure is either positive or
negative. Since o is regular, `
5
is the total space of an o
1
-bundle over a compact
complex surface A
2
which by the rational Gysin sequence must have /
1
= 0 and
/
2
= 1. By Theorem V.1.1 of [BPVdV84] A
2
is either isomorphic to CP
2
or

1
(A
2
) is innite. If A
2
CP
2
then `
5
is the total space of the circle bundle
over CP
2
with Chern class equal jo. where j is a positive integer and o is the
positive generator of H
2
(CP
2
. Z). In this case `
5
is o
5
if j = 1 and the lens
space 1(j. 1. 1) for j = 2. 3. . . . . Since any two Kahler forms on CP
2
dene the
same cohomology class up to a constant, one obtains the standard o-deformation
class on `
5
. Furthermore, the Sasakian structure o is positive since c
1
(T

) is the
pullback of c
1
(CP
2
) which is positive.
If o is negative then c
1
(T

) < 0. and this is the pullback of c

1
(A
2
). Thus,
c
1
(A
2
) is negative so again by Theorem V.1.1 of [BPVdV84]
1
(A
2
) must be
innite. But then
1
(`
5
) is also innite by the long exact homotopy sequence.
10.4. REGULAR SASAKIAN STRUCTURES ON 5-MANIFOLDS 365
We have an immediate
Corollary 10.4.9: There is a unique regular Sasaki-Seifert structure on o
5
and
the lens spaces 1(j. 1. 1). and it contains the standard round metric.
10.4.1. Toric Sasakian Structures on 5-manifolds. Smale-Barden mani-
folds admitting an eective action of a three-dimensional torus T
3
have been studied
by McGavran [McG77], who incorrectly ruled out the non-spin case, and by Oh
[Oh83] who obtained the complete classication. This was based on the earlier clas-
sication of torus actions on 4-manifolds by Orlik and Raymond [OR70, OR74].
The main result in Oh [Oh83] is
Theorem 10.4.10: Let ` B have an eective T
3
-action. Then ` is dieomor-
phic to o
5
. /`

or A

#(/ 1)`

. where / = /
2
(`) 1. Conversely, all these
manifolds admit eective T
3
actions.
By using an o
1
-equivariant K-contact surgery, Yamazaki [Yam01] proved that
all Ohs toric 5-manifolds also admit compatible K-contact structures. Then by
invoking Corollary 8.5.7 we obtain
Theorem 10.4.11: Let ` be a closed simply connected 5-manifold with an eec-
tive T
3
-action. Then ` admits toric Sasakian structures and is dieomorphic to
o
5
. /`

or A

#(/ 1)`

. where / = /
2
(`) 1.
In [BGO07] we gave a construction of toric Sasakian structures on Smale-
Barden manifolds by constructing circle bundles over the T
2
-equivariant blow-ups
of Hirzebruch surfaces arriving at
Theorem 10.4.12: Let S
n,k
be the equivariant /-fold blow-up of the Hirzebruch
surface S
n
. Let
n,k
: `
n,k
S
n,k
be the circle bundle dened by integral Kahler
form
l2
. Then for each positive integer |
2
satisfying 2|
2
+n /. the manifold `
n,k
admits a toric regular Sasakian structure, `
n,k
is dieomorphic to /`

if n is
even, and if n is odd it is dieomorphic to A

#(/ 1)`

. Thus, every regular

contact 5-manifold admits toric regular indenite Sasakian structures.
Proof. It is well-known [Ful93] that the smooth toric complex surfaces are
all obtained by blowing-up Hirzebruch surfaces at the xed points of the T
2
action.
Begin with a Hirzebruch surface o
n
and blow-up o
n
at one of the 4 xed points
of the T
2
action. This gives a smooth toric surface o
n,1
which can be represented
by a Delzant polytope with 5 vertices. Repeat this procedure inductively to obtain
smooth algebraic toric surfaces o
n,k
whose Delzant polytope has / + 4 vertices.
Choose a Kahler class [] lying on the Neron-Severi lattice, and construct the circle
bundle
n,k
: `
n,k
o
n,k
whose Euler class is []. The Kahler form can be chosen
to be invariant under the T
2
action, and we can choose a T
2
invariant connection
in
n,k
: `
n,k
o
n,k
whose curvature form satises d =

n,k
. This gives a
regular Sasakian structure o = (. . . o) on `
n,k
. and by Theorem 10.4.3 `
n,k
is
dieomorphic to either (/ + 1)`

or A

#/`

. Since H
1
(o
n,k
. Z) = 0 the torus
action T
2
is Hamiltonian, and thus lifts to a T
2
action in the automorphism group
of the Sasakian structure o by Corollary 8.1.9. Together with the circle group
generated by the Reeb eld this makes o a regular toric Sasakian structure.
It remains to show that all of the manifolds in B with no torsion in H
2
occur.
For this we need to compute the second Stiefel-Whitney class n
2
(`
n,k
) which is
the mod 2 reduction of the rst Chern class of contact bundle T
n,k
of `
n,k
. First
366 10. SASAKIAN GEOMETRY IN DIMENSIONS THREE AND FIVE
we give a Kahler form of the complex manifolds o
n,k
constructed in the paragraph
above. Let
l
1
,l
2
.
1
.
2
denote the proper transform of
l
1
,l
2
.
1
.
2
respectively,
where
l1,l2
is the Kahler form of Example 10.4.5. Then the Kahler form on o
n,k
can be written as
(10.4.3)
l1,...,l
k+2
=

i
|
i

i
.
where
i+2
is the (1. 1)-form representing the Poincare duals
i+2
of the exceptional
divisors 1
i
. Then writing the Kahler class as

k
i=1
|
i

i
, the positivity condition
becomes
(10.4.4) 0 <
k

i=1
|
i

i

i=1
|
i

i
= |
1
(2|
2
+n|
1
)
k

i=1
|
2
i+1
.
So
l1
,...,l
k+2
denes a Kahler metric if this inequality is satised. Here we have used
the fact the exceptional divisor is in the kernel of the corresponding blow-up map.
Dene the integer valued / + 2-vector l = (|
1
. . . . . |
k+2
). It is convenient to choose
l = (1. |
2
. 1. . . . . 1) in which case the positivity condition becomes 2|
2
+ n /. For
simplicity we denote the corresponding Kahler form by
l
2
.
Next we need to compute the rst Chern class of T
n,k
. This is

n,k
c
1
(o
n,k
)
modulo the transgression of the Kahler class on o
n,k
. that is, modulo the relation

n,k

k+2
i=1
|
i

i
= 0. Let
1
. . . . .
k+1
be a basis for H
2
(`
n,k
. Z). and write

n,k

i
=

k+1
j=1
:
ij

j
. We need to choose the / + 2 by / + 1 matrix (:
ij
) such that
(10.4.5)
k+2

i=1
|
i
:
ij
= 0.
Now using equation (10.4.1) we have
(10.4.6) c
1
(o
n,k
) = 2
1
(n 2)
2

i=1

k+2
which gives
(10.4.7)

n,k
c
1
(o
n,k
) = 2

j
:
1j

j
(n 2)

j
:
2j

j

k+2

i=3

j
:
ij

j
.
We now make a judicious choice of the matrix (:
ij
).
(10.4.8) (:
ij
) =

|
2
2 2 2
1 0 0 0
0 2
.
.
.
.
.
.
.
.
.
.
.
. 0
.
.
.
0 0
.
.
.
.
.
. 2 0
0 0 0 2

The orthogonality condition (10.4.5) is satised and equation (10.4.6) becomes

n,k
c
1
(o
n,k
) = 2|
2

1
6(
2
+ +
k+1
) (n 2)
1
.
It follows that n
2
(`
n,k
) n mod 2.
10.4. REGULAR SASAKIAN STRUCTURES ON 5-MANIFOLDS 367
Such toric contact structures are very numerous. Here for purposes of illustra-
tion we consider only the case of circle bundles over Hirzebruch surfaces. A similar
construction should be possible for the general case, but it is much more tedious.
Let F
l1,l2,n
denote the deformation class of Sasakian structures dened by the above
bundle construction. Then from Example 10.4.5 it is easy to show [BGO07]:
Theorem 10.4.13: For each triple of positive integers (|
1
. |
2
. :) satisfying gcd(|
1
. |
2
) =
1. the manifold o
2
o
3
admits the deformation classes F
l
1
,l
2
,2m
and F
2l
1
,l
2
,2m+1
of
regular toric Sasakian structures. For each triple of positive integers (|
1
. |
2
. :) satis-
fying gcd(|
1
. |
2
) = 1. the manifold A

admits the deformation classes F

2l
1
1,l
2
,2m1
of regular toric Sasakian structures.
In fact each contact structure can have inequivalent toric contact structures cor-
responding to the number of non-conjugate maximal tori in the contactomorphism
group Con(o
2
o
3
. ) or Con(A

. ) [Kar03, Ler03b, BGO07]. For simplicity

we consider the case n = 2: so S
2m
is dieomorphic to o
2
o
2
. For i = 1. 2 we
let
i
denote the classes in H
2
(o
2
o
2
. Z) given by pulling back the volume form
on the i
th
factor. Writing the Kahler class [] = o
1

1
+ o
2

2
in terms of the this
basis, we see that
o
1
= |
1
:+|
2
. o
2
= |
1
.
and the positivity condition becomes o
1
:o
2
0. We denote the corresponding
deformation classes of toric Sasakian structures on o
2
o
3
by F(o
1
. o
2
. :). The
integers o
1
. o
2
are written as o. / in [Kar03] and [Ler03b]. In terms of the o
i
the
rst Chern class 10.4.2 simplies to c
1
(T) = 2(o
1
o
2
). Thus, F(o
1
. o
2
. :) and
F(o
t
1
. o
t
2
. :
t
) belong to non-isomorphic contact structures if o
t
1
o
t
2
= o
1
o
2
. The
following theorem is due to Lerman [Ler03b].
Theorem 10.4.14: For every pair of relatively prime integers (o
1
. o
2
) there are

a2
a
1
| inequivalent regular toric Sasakian structures on o
2
o
3
having the same
contact form
a
1
,a
2
. However, for each integer : = 0. . . . .
a
2
a
1
| the structures
F(o
1
. o
2
. :) are inequivalent as toric contact structures.
CHAPTER 11
Sasaki-Einstein Geometry
In some sense this chapter is the heart of our book, hence, its length. The
question of the existence of Einstein metrics on compact odd-dimensional mani-
folds has been and remains our main motivation in learning and further developing
Sasakian geometry. Since the time we began our study of Sasaki-Einstein manifolds
in the early nineties, there has been an enormous progress in proving the existence
of such geometries. In 1987 Besse asked
1
: Are manifolds admitting an Einstein
metric rather scarce or numerous ? The rst interesting odd (Sasakian) dimension
is dimension ve, since in dimension three an Einstein manifold must be of constant
curvature. At the time the question was posed the only Einstein metrics known to
exist on compact simply connected 5-manifolds were: the round sphere metric on
o
5
, the homogeneous Kobayashi-Tanno metric on o
2
o
3
, the product metric on
o
2
o
3
, and the symmetric metric on ol(3)oO(3). The subject of Einstein man-
ifolds has over the years been so dynamic that new important results appear almost
continuously. In order to cope with that phenomenon Besses book was appended
by a series of addenda which listed several new results, most of them not even
published at the time. One of them (see [Bes87], Addendum A) describes a result
of Wang and Ziller [WZ86] where it is shown, that using the Kobayashi bundle
construction [Kob63], one can obtain innitely many inequivalent Einstein metrics
exist on o
2
o
2m+1
for each :. as well as on the non-trivial o
2m+1
-bundle over o
2
.
So the oodgates were opened and it was soon realized that Einstein metrics tend
to exist in great profusion. About the same time the existence of Kahler-Einstein
metrics on certain smooth del Pezzo surfaces [TY87, Tia87, Siu88] was given
with the complete picture being understood somewhat later; all del Pezzo surfaces
save the one and two point blow-ups of CP
2
admit Kahler-Einstein metrics [Tia90].
Then using the Kobayashi bundle construction this result automatically gives rise
to new families of Sasaki-Einstein metrics on 5-manifolds [FK89]. They naturally
generalize the Kobayashi-Tanno metric on o
2
o
3
. Apart from the manifolds al-
ready mentioned, Tians theorem added six more members to the club of compact
5-manifolds admitting Einstein metric of positive scalar curvature. Throughout the
nineties this club remained very exclusive with just a total of merely nine members.
The current chapter of our book describes most advances that have been made in
the last few several years. Like Besse we are facing a dicult problem: new results
seem to appear with increasing frequency. Some of the material we have included
here has not yet been published. Today, 20 years after the publication [Bes87],
Einstein metrics are known to exist on all /-fold connected sums /`

= /(o
2
o
3
).
on innitely many rational homology 5-spheres `
p
, as well as on innitely many
connected sums of these. Most of these are actually Sasaki-Einstein. For many
1
See [Bes87], Chapter 0.I: The Main Problems Today.
369
370 11. SASAKI-EINSTEIN GEOMETRY
manifolds it appears that innite sequences of Einstein metrics occur and with
large moduli. After 20 years the answer to the Besses question appears to be: Ein-
stein metrics in each dimension are numerous. It is quite conceivable that every
compact simply connected 5-manifold admits an Einstein metric. We should also
mention here that at least part of the reason for this explosion of interest in Sasaki-
Einstein metrics comes from Physics. As we discuss in Chapter 14 Sasaki-Einstein
metrics admit Killing spinors, something of great interest to physicists working in
conformal quantum eld theory.
11.1. Foundations of Sasaki-Einstein Geometry
Although we are mainly interested in Sasaki-Einstein geometry, there is a nat-
ural generalization of the Einstein condition in the category of Sasakian geometry,
the so-called -Einstein metrics. It is convenient to present the foundations in this
more general setting. Sasaki--Einstein structures are treated in more detail in
Section 11.8.
Denition 11.1.1: A contact metric structure o = (. . . o) on ` is said to be
-Einstein if there are constants o. / such that Ric
g
= oo +/ .
Clearly, o is Einstein when / = 0. Little seems to be known concerning Einstein
metrics (or more generally -Einstein metrics) in the general contact metric case
(see, however, Proposition 11.1.6 below), and as we shall see by Theorem 11.1.7 be-
low that K-contact Einstein implies Sasakian, so we restrict ourselves momentarily
to the Sasakian case. Our rst result gives necessary conditions for the existence
of Sasaki--Einstein metrics on a Sasakian manifold `. There are restrictions on
both the type of contact structure T as well as the topology of the manifold ` in
order for it to admit a Sasaki--Einstein structure.
Theorem 11.1.2: Let o = (. . . o) be a Sasakian structure on `. Then a nec-
essary condition that there is a type II deformation of o to a Sasaki--Einstein
structure in F(T

) is that there exists a real number o such that c

1
(T

) = o[d]
B
.
The underlying contact bundle T must have vanishing real rst Chern class, i.e.,
c
1
(T) is a torsion class. In particular, a simply connected manifold admitting a
Sasaki--Einstein structure must be spin.
Proof. This follows from the discussion in Sections 7.5.1 and 7.5.2. The de-
tails are left as an exercise.
Exercise 11.1: Prove Theorem 11.1.2.
Remarks 11.1.1: The fact that a simply connected Sasaki-Einstein manifold is spin
was rst noticed by Moroianu [Mor97]. When F(T

) is positive or zero Theorem

11.1.2 holds under the weaker condition that o = (. . . o) is K-contact, but
Theorem 11.1.7 below shows that this is a mute point. The type II deformation
(7.5.9) always exists when F(T

) is negative or zero by the transverse Yau Theorem

7.5.19 of El Kacimi-Alaoui.
We shall make use of the following important theorem which for the regular case
is due to Kobayashi [Kob63]. The general quasi-regular case appeared in [BG00b].
Recall from Theorem 7.1.3 that if ` is a quasi-regular K-contact manifold with
compact leaves, the space of leaves `T

is an almost Kahler orbifold (Z. /).

Theorem 11.1.3: Let ` be a compact manifold of dimension 2n+1 with a quasi-
regular K-contact structure (. . . o). Then
11.1. FOUNDATIONS OF SASAKI-EINSTEIN GEOMETRY 371
(i) (. . . o) is -Einstein if and only if the almost Kahler orbifold (Z. /)
is Einstein.
(ii) o is Einstein if and only if / is Einstein with scalar curvature 4n(n+1).
In particular, o is Sasaki-Einstein if and only if / is Kahler-Einstein with
scalar curvature 4n(n + 1).
Proof. An immediate consequence of Theorem 7.3.12 is: a K-contact struc-
ture (. . . o) is -Einstein if and only if the transverse metric o
T
is Einstein,
and o is Einstein if and only if o
T
is Einstein with Einstein constant 2n + 2. From
Theorem 7.1.3 there is an o
1
orbibundle : `Z such that the almost Kahler
orbifold (Z. /) satises

/ = o
T
. Furthermore, the Ricci curvature Ric
h
of the
orbifold metric / pulls back to Ric
T
. so (i) follows. Moreover, o will be Einstein if
and only if o
T
is Einstein with Einstein constant 2n + 2. But this means that the
scalar curvature is 2n(2n + 2).
An immediate corollary of Proposition 7.3.9 or Theorem 7.1.3 and the well-
known Myers Theorem is
Corollary 11.1.4: If a K-contact structure (. . . o) on `
2n+1
is Einstein, then
the Einstein constant equals 2n. Hence, if `
2n+1
is complete, then it is compact
with diameter less than or equal to and with nite fundamental group.
Recall our discussion of both Riemannian and symplectic cones in Section 6.5.
In particular Proposition 6.5.5 gives a one to one correspondence between contact
metric structures on a manifold ` and almost Kahler structures on the cone C(`).
Furthermore, by Proposition 6.5.6 a contact metric structure (. . . o) is K-contact
if and only if the vector eld i pseudo-holomorphic with respect to the almost
complex structure on C(`). We are interested in the case when the cone metric
o = d:
2
+ :
2
o satises the Einstein condition Ric
g
= o. The following important
lemma seems to be well-known as folklore.
Lemma 11.1.5: Let (`. o) be a Riemannian manifold of dimension n. and con-
sider (C(`) = ` R
+
. o) the cone on ` with metric o = d:
2
+ :
2
o. Then if
o is Einstein, it is Ricci-at, and o is Ricci-at if and only if o is Einstein with
Einstein constant n 1.
Proof. Let
i

n
i=1
be a local orthonormal coframe for (`. o). then we obtain
a local orthonormal coframe

for the cone metric o on ` R

+
by setting

i
= :
i
.
0
= d: .
where 0 j n. We employ the convention that the Greek indices range from 0 to
n. whereas, the Latin indices range from 1 to n, and we identify ` with ` 1.
Consider the Cartan structure equations (1.3.2)
d

= 0 .
d

<

.
Here

denotes the connection 1-forms with respect to the Levi-Civita connection

on `R
+
which satises

= 0. and

1

denotes the Riemannian curvature.

The rst Cartan equations for the cone metric gives
i
0
=
i
. and that
i
j
are the
372 11. SASAKI-EINSTEIN GEOMETRY
connection 1-forms for the Levi-Civita connection with respect to o on `. The
second equations then imply

1
i
0
= 0

1
i
jkl
=
1
:
2
1
i
jkl

k[i

j]l
.
where
ik
denotes the Kronecker delta symbol, and the brackets mean antisym-
metrize over i and ,. Forming the Ricci tensor one now easily proves the result.
Notice that if ` has dimension 2n+1 and (C(`). o) is Einstein then (`. o) is
Einstein with Einstein constant 2n. Thus, if o belongs to a contact metric structure
o = (. . . o) and (C(`). o) is Einstein, then o = (. . . o) is actually K-contact
by Proposition 7.3.9. Summarizing,
Proposition 11.1.6: If the metric cone (C(`). o) on a contact metric manifold
is Einstein, then the contact metric structure o = (. . . o) is K-contact and
Einstein.
The question now arises concerning the consequences of adding to the hypothe-
sis of Theorem 11.1.3 the condition that o be Sasakian. Remarkably such an added
hypothesis is not needed when F(T

) is positive or zero. The following theorem

proved by the authors in [BG01a] can be thought of as an odd-dimensional ver-
sion of the well-known Goldberg Conjecture [Gol69]. It says that if a K-contact
structure is Einstein it is Sasakian, that is the underlying almost CR-structure is
necessarily integrable.
Theorem 11.1.7: Let (`. . . . o) be a compact K-contact manifold. Then if o
is Einstein (`. . . . o) is Sasakian.
Proof. We rst prove the theorem under the assumption that is quasi-
regular. By Theorem 7.1.3 ` is the total space of a principal o
1
orbibundle over
a compact almost Kahler orbifold Z. Furthermore, by Theorem 11.1.3 the induced
metric / on Z is almost Kahler-Einstein which has positive scalar curvature, since
o has positive scalar curvature. Now since Sekigawas [Sek1, Sek2] proof of the
Goldberg conjecture in the case of nonnegative scalar curvature only involves lo-
cal curvature computations together with a Bochner type argument using Stokes
Theorem, it carries over to the case of a compact orbifold. So the almost complex
structure on Z is integrable, and (Z. /) is Kahler-Einstein. It then follows from
Theorem 7.1.3 that (`. . . . o) is normal, hence, Sasaki-Einstein. This proves
the result under the assumption of quasi-regularity.
Now assume that (`. . . . o) is K-contact and Einstein, but not quasi-regular.
Then by Proposition 2.6.3 the Reeb vector eld lies in the commutative Lie subal-
gebra t(`. T

) a(`. o) which has dimension / 1. Thus, there exists a sequence

of quasi-regular contact forms
j
and Reeb vector elds
j
t(`. T

) that ap-
proximate (. ) in the compact-open C

topology. (In what follows we use this

topology on the space of smooth sections of all tensor bundles.) Explicitly, there is
a monotonically decreasing sequence c
j

1
with lim
j
c
j
= 0 such that
(11.1.1)
j
= 1(c
j
).
j
= +
j
. 1(c
j
) =
1
1 +(
j
)
.
where 1(c
j
) are positive functions in C

(`) that satisfy lim

j
1(c
j
) = 1. Clearly

j
t(`. T

) and lim
j

j
= 0. Moreover, ker
j
= ker = T. so we have the same
11.1. FOUNDATIONS OF SASAKI-EINSTEIN GEOMETRY 373
underlying contact structure. We also have the following easily veried relations
for the induced contact endomorphisms
j
:
(11.1.2)
j
=
1
1 +(
j
)

j
= 1(c
j
)
j
.
This implies that
j

j
= 0 and that the almost complex structure J on T remains
unchanged. However, the induced metrics become
o
j
= 1(c
j
)o
1
1(c
j
)
2
+d1(c
j
) 1(c
j
)d (
j
) 1l .
For c
j
small enough o
j
are well dened Riemannian metrics on ` which converge
to o. and can easily be seen to satisfy the compatibility conditions
o
j
(
j
A.
j
Y ) = o
j
(A. Y )
j
(A)
j
(Y ) .
Moreover, since
j
t a(`. ). it follows that the functions 1(c
j
) C

(`)
T
.
where C

(`)
T
denotes the subalgebra of C

(`) invariant under the action of

the torus T. Thus, from equation (11.1.2) we have
L
j

j
= 0 .
Hence, (`.
j
.
j
.
j
. o
j
) is a sequence of quasi-regular K-contact structures on
` whose limit with respect to the compact-open C

topology is the original K-

contact Einstein structure (`. . . . o). Now the metrics o
j
are not Einstein, but
their Ricci tensor can be seen to satisfy
Ric
g
j
=
j
o
j
+(c
j
.
j
. o) .
where (c
j
.
j
. o) is a traceless symmetric 2-tensor eld depending on c
j
.
j
. o that
tends to 0 with c
j
. and
j
C

(`) satisfy lim

j

j
= 2n.
Now there is a sequence of orbifold Riemannian submersions
j
: `Z
j
,
where (Z
j
. /
j
) are a sequence of compact almost Kahler orbifolds satisfying

j
/
j
=
1(c
j
)o
1
. Moreover, it follows from the above limits that the scalar curvatures of
the /
j
are all positive. Notice that in Sekigawas proof [Sek2] of the positive scalar
curvature Goldberg conjecture, the Einstein condition is not used until Section 4
of [Sek2]. Following [Sek2] and making the necessary adjustments to our situation,
we nd that there are nonnegative numbers
j
and nonnegative smooth functions
1
j
such that
(11.1.3)

1
j
+
:
j
n
[[

j

J
j
[[
2
j
+
1
2n
[[

j

J
j
[[
4
j

j

j
.
where

j
.

J
j
. :
j
.
j
and [[ [[
j
are the Levi-Civita connection, almost complex
structure, scalar curvature, volume element, and Riemannian norm, respectively on
(Z
j
. /
j
). Now since the metrics o. o
j
are bundle-like the leaves of the characteristic
foliation are geodesics and the ONeill tensors T and vanish [Ton]. Moreover, for
any K-contact manifold of dimension 2n + 1 the ONeill tensor satises [[[[
2
=
o(. ) = 2n. Thus, we have the relation between the functions
j
on ` and the
scalar curvatures :
j
on Z
j
:
:
j
= (2n + 1)
j
+ 2n.
So that lim
j
:
j
= 2n+(2n+1) lim
j

j
= 4n(n+1) . Furthermore, we have lim
j

j
= 0.
Thus, since 1
j
(see [Sek2]) and :
j
are nonnegative for each ,. the estimate 11.1.3
374 11. SASAKI-EINSTEIN GEOMETRY
implies the estimate
[[

j

J
j
[[

j

t
j
.
where
t
j
are nonnegative numbers satisfying lim
j

t
j
= 0. Now for each , the hori-
zontal lift of

J
j
is the horizontal projection (
j
J
j
)
h
= (
j

j
)
h
. where
j
. J
j
.
j
are the corresponding Levi-Civita connection and tensor elds with respect to the
metrics on `. But on ` J
j
= J for all , and we have
[[(J)
h
[[ = lim
j
[[(
j
J)
h
[[
j
lim
j

t
j
= 0 .
where [[ [[
j
is the Riemannian norm with respect to o
j
. So the almost CR structure
on T is integrable which by Proposition 6.5.14 implies that (`. . . . o) is Sasaki-
Einstein.
Exercise 11.2: Prove the following result of Okumura [Oku62]: If the Ricci tensor
of a Sasakian manifold (`. . . . o) is parallel, then o is Einstein.
An alternative proof of Theorem 11.1.7 has been given recently by Apostolov,
Draghici, and Moroianu [ADM06]. Instead of using the transverse geometry as
in [BG01a], they work on the metric cone C(`) by using a Weitzenbock-type
formula to show that if (C(`). d:
2
+ :
2
o) is almost Kahler and Einstein it must
be Kahler. They also notice that Theorem 11.1.7 cannot be improved in dimension
3 by assuming the weaker hypothesis that (`. . . . o) is only a contact metric
structure. A simple counterexample is given by the 3-torus Example 6.1.23 with the
at metric. However, this counterexample does not generalize to higher dimension,
for Blair [Bla76b] has shown that a contact metric structure in dimension 5 or
higher cannot be at. Thus, it is still an open question as to whether the hypothesis
of Theorem 11.1.7 can be weakened to contact metric structures in dimension 5 and
greater.
There is another important characterization of the Sasaki-Einstein condition
which can be formulated as a corollary of Lemma 11.1.5:
Corollary 11.1.8: Let (`. o) be a Riemannian manifold. Then the metric o
is Sasaki-Einstein if and only if the cone metric o is Calabi-Yau, i.e., (C(o). o)
is Kahler Ricci-at. In particular, it follows that the restricted holonomy group
Hol
0
( o) ol(n + 1).
Now Hol
0
( o) is the normal subgroup of the full holonomy group Hol( o) that is
the component connected to the identity. There is a canonical epimorphism

1
(`) =
1
(C(`)) Hol( o)Hol
0
( o) .
so if ` is simply connected its structure group reduces to 1 ol(n) and it will
admit a spin structure. So for Sasaki-Einstein manifolds we have a generalization
of the last statement of Theorem 11.1.2.
Corollary 11.1.9: Let ` be a Sasaki-Einstein manifold such that the full holo-
nomy group of the cone metric Hol( o) is contained in ol(n + 1). Then ` admits
a spin structure.
We give some examples that illustrate the complications in the presence of
fundamental group. The hypothesis of this corollary is not necessary as the second
example shows.
Example 11.1.10: The real projective space ` = RP
2n+1
with its canonical metric
is Sasaki-Einstein, and the cone C(`) = (C
n+1
`0)Z
2
with the usual antipodal
11.1. FOUNDATIONS OF SASAKI-EINSTEIN GEOMETRY 375
identication. We have Hol( o)
1
(`) Z
2
. When n is odd the antipodal
map is in ol(n + 1). so ` = RP
2n+1
admits a spin structure. But when n
is even the antipodal map does not lie in ol(n + 1). which obstructs a further
reduction of the structure group. In this case it is well-known that ` = RP
2n+1
does not admit a spin structure. In fact the generator of Hol( o) Z
2
is the
obstruction. There are many other similar examples. An example that shows that
the hypothesis in Corollary 11.1.9 is not necessary is the following: Consider the
lens space 1(j; . c
1
. . . . . c
n
) o
2n+1
Z
p
where the c
i
s are relatively prime to j.
The action on C
n+1
` 0 is generated by (.
0
. .
1
. . . . . .
n
) (.
0
.
q
1
.
1
. . . . .
q
n
.
n
).
where is a primitive j
th
root of unity. It is known [Fra87] that if j is odd,
1(j; c
1
. . . . . c
n
) admits a spin structure. However, if

i
c
i
+1 is not divisible by j.
the holonomy group Hol( o) Z
p
does not lie in ol(n + 1).
Now combining the results from Theorems 7.5.1 and 11.1.7 we have
Theorem 11.1.11: Let o = (. . . o) be a quasi-regular K-contact structure on
a compact manifold ` of dimension 2n+1. and let Z = (7. ) denote the orbifold
structure on the space of leaves 7 of the characteristic foliation. If o is Einstein,
then o is Sasakian and
(i) Z is an orbifold log Fano variety with an orbifold Kahler-Einstein metric
/ with scalar curvature 4n(n + 1) and the projection : `Z is a
Seifert o
1
-bundle (equivalently and o
1
-orbibundle) with geodesic bres.
(ii) If o is regular then : `7 is a principal circle bundle over the smooth
Fano manifold 7 with a Kahler-Einstein metric.
The Inversion Theorem 7.5.2 in the Sasaki-Einstein setting gives
Theorem 11.1.12: Let Z be a compact Fano orbifold with
orb
1
(Z) = 0. Let :
`Z be the o
1
-orbibundle whose rst Chern class is
c
orb
1
()
Ind()
. Then
orb
1
(`) = 0
and ` is smooth if the local uniformizing groups of Z inject into o
1
. Furthermore,
there is a metric o on the total space ` such that ` is a compact simply connected
Sasakian manifold with Ric
g
2. If Z has an orbifold Kahler-Einstein metric /
then there is a unique metric o on ` given by o =

/ + which is Sasaki-
Einstein.
11.1.1. Homogeneous Sasaki-Einstein Manifolds. There is one special
case where the hypothesis of Theorem 11.1.7 can be weakened from a K-contact
structure to an arbitrary metric contact structure, namely for compact homoge-
neous contact manifolds (`. ). This is a consequence of Theorem 8.3.6. From the
proof of that theorem we may assume that (`. ) is homogeneous under a compact
Lie group G. We then let G
c
denote its complexication. We have
Theorem 11.1.13: Let (`. . G) be a compact homogeneous contact manifold with
Reeb vector eld . Suppose also that (`. . G) admits a compatible homogeneous
Einstein metric o. Then (. . o. G) is a homogeneous Sasaki-Einstein structure and
is an o
1
-bundle over a generalized ag manifold G
c
1. Conversely, given any gen-
eralized ag manifold G
c
1 there is a circle bundle : `G
c
1 whose total
space ` is a simply connected homogeneous Sasaki-Einstein manifold, and all ho-
mogeneous Sasaki-Einstein manifolds are of the form `Z
k
for some integer / 1.
Proof. The rst statement is an immediate corollary of Theorem 8.3.6. To
prove the converse we notice that a theorem of Matsushima [Bes87] says every
376 11. SASAKI-EINSTEIN GEOMETRY
G
c
1 admits a G invariant positive Kahler-Einstein metric with Kahler form .
Moreover, there is a subgroup l G such that G
c
1 = Gl. One then chooses
any circle bundle given by an integral class lying on the positive ray in the Kahler
cone dened by . One then obtains a Sasaki-Einstein metric on the total space `
of this circle bundle by Theorem 11.1.3. By construction one easily sees that this
metric is homogeneous. For each G
c
1 and each / Z
+
we get a homogeneous
o
1
-bundle over G
c
1 with fundamental group
1
= Z
k
. It can be obtained as a
discrete Z
k
-quotient of the unique simply connected model.
Note that the Stiefel manifold \
2
(R
n+1
) of Example 9.3.20 is a homogeneous
Sasaki-Einstein manifold that can be realized as the link of a quadric.
A somewhat weaker version of Theorem 11.1.13 was rst given in [BG00b].
The following corollary lists all the possible G1 in the rst three dimensions
[BG99, BG00b]:
Corollary 11.1.14: Let (`. o) be a compact homogeneous Sasaki-Einstein mani-
fold of dimension 2n + 1. Then ` is a circle bundle over
(i) CP
1
when n = 1. implying ` = o
3
;
(ii) CP
2
or CP
1
CP
1
when n = 2. implying ` = o
5
or ` = o
2
o
3
;
(iii) CP
3
. CP
1
CP
2
. CP
1
CP
1
CP
1
. the complex ag 1
3,2,1
= ol(3)T
2
.
or the real Grassmannian G:
2
(R
5
) when n = 3, implying ` = o
7
. the
Wang-Ziller manifold `
1,2
2,3
. an o
2
o
3
bundle over o
2
. the 3-Sasakian
homogeneous manifold ol(3)o
1
. or the real Stiefel manifold \
2
(R
5
).
respectively.
Exercise 11.3: Using the classication of low-dimensional homogeneous Einstein
manifolds [BK06] give the list of all homogeneous Sasaki-Einstein manifolds in
dimensions 9 and 11.
To end this subsection we mention a Sasaki-Einstein version of a sphere theorem
Theorem 11.1.15: Let ` be a compact simply connected Sasaki-Einstein manifold
of positive sectional curvature. Then ` is isometric to the round unit sphere.
It appears that the theorem was established for Sasaki-Einstein 5-manifolds
with additional assumption on sectional curvature 13 and in its full generality
by Moskal [Mos66]. However, Moskals thesis has never been published. On the
other hand, the proof of this theorem can be found in the Sasaki lecture notes (cf.
[Sas68], Theorem 45.1), which are perhaps better known, but have never really
been published either.
11.1.2. Moduli of Sasaki-Einstein Structures. Recall that Proposition
8.2.6 arms that the innitesimal deformations of the transverse holomorphic struc-
ture of the characteristic foliation of a quasi-regular Sasakian structure can be de-
scribed in terms of two pieces, the deformations of the underlying complex structure
which are elements of H
1
(Z.

). and the deformation of the characteristic foliation

itself which can be characterized by the Sasaki cone t
+
k
(T. J) = cr
+
(T. J)CR(T. J)
as described by Corollary 8.2.14. We believe that a Sasaki-Einstein structure, if it
exists, will be unique within t
+
k
(T. J) with a xed J. This is formulated as Conjec-
ture 11.1.18 below. So the deformations that we consider here come from deforming
the complex structure of the base orbifold. These deformations were discussed in
Section 5.5. We shall make use of
11.1. FOUNDATIONS OF SASAKI-EINSTEIN GEOMETRY 377
Proposition 11.1.16: Let (Z
1
. /
1
) and (Z
2
. /
2
) be two Kahler-Einstein Fano orb-
ifolds and `
1
and `
2
the corresponding Sasaki-Einstein manifolds. Assume also
that `
1
is not a metric of constant sectional curvature one. Let : `
1
`
2
be
an isometry. Then there is an isometry

: Z
1
Z
2
which is either holomorphic
or anti-holomorphic, such that the following diagram commutes:
`
1

`
2

2
Z
1

Z
2
.
Moreover,

determines up to the o
1
-action induced by the Reeb vector eld.
Proof. Since is an isometry we can assume that `
2
= `
1
= ` and o
2
=
o
1
= o. Since o does not have constant curvature, Lemma 8.1.17 implies that either

o
2
= o
1
.

o
2
= o
c
1
the conjugate Sasakian structure, o
c
1
= (
1
.
1
.
1
. o
1
).
or

o
2
and o
1
form part of a 3-Sasakian structure dened by Denition 13.1.9
below. In the rst case Aut(o
1
). and in the second case it induces complex
conjugation on the base, that is Z
2
=

Z
1
with the same orbifold metric. Finally
when

o
2
and o
1
form part of a 3-Sasakian structure, Z
1
and Z
2
are isomorphic
as complex Kahler-Einstein contact orbifolds by Theorem 13.3.1 below.
Conversely, any such biholomorphism or anti-biholomorphism

lifts to an or-
bibundle map : `
1
`
2
uniquely up to the o
1
-action given by the Reeb vector
eld.
It is clear that our procedure produces inequivalent Sasaki-Einstein structures,
but Proposition 11.1.16 implies a much stronger result, namely
Theorem 11.1.17: Let ` be a smooth compact manifold and let o = (. .
t
. o
t
)
be a family of quasi-regular Sasaki-Einstein structures on ` induced by a continu-
ous family of inequivalent complex orbifolds Z
t
with Kahler-Einstein metrics. Then
the metrics o
t
are inequivalent as Einstein metrics.
Proof. The constant sectional curvature metric is rigid as an Einstein metric
[Bes87] and is covered by an o
1
-bundle over complex projective space. There are
no families of inequivalent complex structures on projective space, so none of the o
t
can have constant sectional curvature. So Proposition 11.1.16 applies to conclude
that the o
t
are all inequivalent as Einstein metrics.
Theorems 11.1.17 and 5.5.7 can be used to give a lower bound on the dimension
of the moduli space of Einstein metrics on many manifolds. We give many such
examples later in the chapter. Notice also that Theorem 11.1.17 only addresses
the moduli problem for quasi-regular structures. Nothing appears to be known
about the moduli problem for Sasaki-Einstein structures in the irregular case. As
we shall see irregular Sasaki-Einstein metrics do exist by Theorem 11.4.5 below. As
mentioned previously an interesting question is whether one can deform through
Sasaki-Einstein metrics by deforming the Reeb vector eld within the Sasaki cone
t
+
k
(T. J). We think not, but we really have no good reason other than the lack of
any example.
Conjecture 11.1.18: Sasaki-Einstein metrics occur for at most one t
+
k
(T. J).
378 11. SASAKI-EINSTEIN GEOMETRY
This is certainly true when / = 1 by Theorem 11.8.2 below.
11.2. Extremal Sasakian Metrics
In this section we dene extremal metrics in the context of Sasakian geometry.
These
2
were rst introduced in [BGS06] and they naturally generalized the notion
of both Sasaki-Einstein metrics and Sasaki--Einstein metrics. Just as in the Kahler
case the concept is very useful in studying the obstruction theory which will be
discussed in further sections. We begin by introducing a Riemannian functional
whose critical points will describe an important subset of Sasakian structures in
T(.

J). Our presentation follows [BGS06] closely.
We denote by M(.

J) the set of all compatible Riemannian Sasakian metrics
arising from structures in F(.

J), and dene the energy functional 1 : M(.

J)
R by
(11.2.1) 1(o) =

M
:
2
g
dvol
g
which is just the 1
2
-norm of the scalar curvature of o.
The variation of a metric in M(.

J) depends upon the two functions and
of Proposition 7.5.7. However, the transverse Kahler metric varies as a function of
only, and does so within a xed basic cohomology class. The critical point of
(11.2.1), should it exist, will allow us to x the parameter , which is determined
by the condition that d = d + i

be an extremal transverse Kahler metric

[Cal82]. The remaining gauge function parameter represents nothing more than
a change of coordinates in the representation of the form of the Sasaki structure
in question. Thus, nding a critical point of 1 produces an extremal representative
of F(.

J).
11.2.1. Variational Formulae and the Euler-Lagrange Equations. In
order to derive the Euler-Lagrange equation of (11.2.1), we describe the innitesimal
variations of the volume form, Ricci tensor, and scalar curvature, as we deform the
metric in M(.

J).
Proposition 11.2.1: Let (.
t
.
t
. o
t
) be a path in F(.

J) that starts at (. . . o)
when t = 0. and is such that d
t
= d +ti

for certain basic function . and for

t suciently small. Then we have the expansions
dvol
t
= (1
t
2

B
))dvol +O(t
2
) .

t
= ti

1
2

B
+

+O(t
2
) .
:
t
= :
T
2n t

1
2

2
B
+ 2(
T
. i

+O(t
2
) .
for the volume form, Ricci form, and scalar curvature of o
t
. respectively. Here, the
geometric terms without sub-index are those corresponding to the starting metric o.
Proof. The proof in [BGS06] is a straightforward adaption of [Sim04] to
the Sasaki case.
2
In [BGS06] such metrics were called canonical Sasakian metrics. But we believe that this
terminology can be misleading, suggesting that it refers to Sasakian structures that are canonical
or anticanonical as given by Denition 7.5.24. So we prefer to call such metrics extremal (cf.
Denition 11.2.4 below) which is also more in agreement with the standard usage.
11.2. EXTREMAL SASAKIAN METRICS 379
Associated to any Sasakian structure (. . . o) in T(.

J). we introduce a
basic dierential operator 1
B
g
of order 4 whose kernel consists of basic functions
with transverse holomorphic gradient. Given a basic function : ` C. we
consider the vector eld
#
dened by the identity
(11.2.2) o(
#
. ) =

.
Thus, we obtain the (1. 0)-component of the gradient of . a vector eld that,
generally speaking, is not transversally holomorphic. In order to ensure that, we
would need to impose the condition

#
= 0. that is equivalent to the fourth-order
equation
(11.2.3) (

#
)

#
= 0 .
because '. (

#
)

#
`
L
2 = |

#
|
2
L
2
.
We have that
(11.2.4) 1
B
g
= (

#
)

#
=
1
4
(
2
B
+ 4(
T
. i

) + 2(:
T
)
#
) .
The functions on ` that are transversally constant are always in the kernel of 1
B
g
.
The only functions of this type that are basic are the constants. Thus, the kernel
of 1
B
g
has dimension at least 1.
Proposition 11.2.2: The rst derivative of 1 at o M(.

J) in the direction of
the deformation dened by (. ) is given by
d
dt
1(o
t
) [
t=0
= 4

M
(:
T
2n)

#
)

dvol .
Proof. This result follows readily from the fact that : = :
T
2n, the varia-
tional formulae of Proposition 11.2.1, and identity (11.2.4).
Theorem 11.2.3: A Sasaki metric o M(.

J) is a critical point of the energy
functional 1 of (11.2.1) if and only if the basic vector eld
#
g
:
T
g
=
#
g
:
g
is transver-
sally holomorphic.
So as in the Kahler case the Euler-Lagrange equations boil down to holomor-
phicity of the gradient of the scalar curvature. We are thus led to our fundamental
denition:
Denition 11.2.4: We say that o = (. . . o) is an extremal Sasakian struc-
ture in T(.

J) or that o is an extremal Sasakian metric if o satises the
condition of Theorem 11.2.3, that is if and only if o is transversally extremal.
11.2.2. Sasaki-Futaki Invariants. In this section we adapt the construction
of the well-known Futaki invariant to the Sasakian category. Again our presentation
follows [BGS06] closely. Given any Sasakian structure (. . . o) F(.

J). the
metric o is an element of M(.

J) whose transverse Ricci form
T
is basic. We
dene the Ricci potential
g
as the function in the basic Hodge decomposition of

T
given by

T
=
T
h
+i

g
.
where
T
h
is the harmonic representative of the basic cohomology class in H
2
B
(T

)
represented by
T
. Notice that if G
T
g
is the Greens operator of the transverse
metric, we have that

g
= G
T
g
(:
T
g
) = G
T
g
(:
T
g
2n) = G
T
g
(:
g
) = G
g
(: :
g,0
) .
380 11. SASAKI-EINSTEIN GEOMETRY
where :
g
and G
g
are the scalar curvature and Greens operator of o, and :
g,0
is the
projection of :
g
onto the constants. The sequence of equalities above follows from
the fact that :
g
= :
T
g
2n is a basic function. Thus, the Ricci potential
g
is itself
a basic function.
On a Sasakian manifold we have a transverse holomorphic structure, and thus
from Denition 2.5.22 the Lie algebra h
T
(T

.

J) of transversally holomorphic vector
elds. So in analogy to the Kahler case equation (5.3.2) we have
Denition 11.2.5: With F(.

J) xed we dene a map SF
,

J
: h
T
(T

.

J)C by
SF(A)
,

J
=

A(
g
)dvol
g
.
We call SF = SF
,

J
the Sasaki-Futaki invariant.
Since
g
is basic, the integrand in this expression can be fully written in terms
of the transversally holomorphic realization
X
=
1
2
(A
1
iJ(A
1
)) of A. Our next
result shows that SF is well-dened as a Lie algebra character.
Proposition 11.2.6: The map SF
,

J
only depends on the basic cohomology class
[d]
B
(equivalently only on the space F(.

J)) and not on the particular Sasakian
structure in F(.

J).
Proof. We take a path o
t
in M(.

J) starting at o for which the transverse
Kahler form is of the form
d
t
= d +ti

.
with the ane parameter a basic function. From the identity
B

g
= :
T
g,0
:
T
.
we see that the variation

g
of
g
satises the relation
2(i

. i

g
)
g
+
B

g
=

:
T
=
1
2

2
B
+ 2(
T
. i

)
g
.
Hence,

g

1

g
dvol
g
=
1
2

B
+ 2G
T
g
(
T
h
. i

)
g
.
where is the volume of ` in the metric.
Since
T
h
is harmonic, the last summand in the right side can be written as
2G
T
g
(

(
T
h
))). For convenience, let us set =

(
T
h
). Hence,
d
dt

A(
t
)dvol
gt
=

1
2

B
2G
T
g
(

))
1
2

dvol
g
.
By the Ricci identity for the transverse metric, we have that
1
2
(
B
)

=

,
+
,
(

,
+ (:
T
h
)

) =

,
+
,

,
+

.
and so, after minor simplications, we conclude that
d
dt

A(
t
)dvol
gt
=

X
(
,

,
)

dvol
g
+

X
(

2(G
g

)
,
) dvol
g
.
where
X
is the (1. 0)-component of A
1
.
11.2. EXTREMAL SASAKIAN METRICS 381
The rst summand on the right above is zero because
X
is holomorphic. This
is just a consequence of Stokes theorem. The second summand is also zero since
we have

X
(

2(G
g

)
,
) dvol
g
=

(
B
G
T
g
.

X
)dvol
g
+

(2

G
T
g
.

X
)dvol
g
.
and
B
G
T
g
= 0 while

X
= 0. Here, of course,

X
is the (0. 1)-basic form
corresponding the (1. 0)-vector eld
X
.
Proposition 11.2.7: The following hold
(i) SF(A) = 0 for any section A of the trivial line bundle 1

.
(ii) SF([A. Y ]) = 0 for any pair of vector elds A. Y h
T
(T

.

J).
We leave the proof as
Exercise 11.4: Prove Proposition 11.2.7.
Our next proposition extends to extremal Sasakian metrics a now well-known
result in Kahler geometry originally due to Futaki [Fut83]. The Sasaki version
here is analogous to the expanded version of Futakis result presented by Calabi
[Cal85].
Theorem 11.2.8: Let o = (. . . o) be an extremal Sasakian structure in F(.

J).
Then, the metric o has constant scalar curvature if and only if SF( ) = 0.
Proof. In one direction the statement is obvious: a constant scalar curvature
Sasakian metric has trivial Ricci potential function, so the Sasaki-Futaki invariant
SF vanishes on every A h
T
(T

.

J).
In order to prove the converse, we rst observe that if A h
T
(T

.

J) is a
transversally holomorphic vector eld of the form A =
#
g
1 for some basic function
1. then
SF(A) =

#
g
1(G
g
:
g
)dvol
g
= 2

1.

G
T
g
:
g
)
g
dvol
g
= 2

1(

g

G
T
g
:
g
)dvol
g
.
because the scalar curvature :
g
is a basic function also. Since 2

=
B
. we
conclude that
SF(
#
g
1) =

1(:
g
:
g,0
)dvol
g
.
Now, if the Sasakian metric o is a critical point of the energy function 1 in (11.2.1),
then
#
g
: =
#
:
T
=
#
g
:
g
is a transversally holomorphic vector eld, and we
conclude that
SF(
#
g
:
T
) =

(:
g
:
g,0
)
2
dvol
g
.
Thus, if SF( ) = 0. then :
g
must be constant.
Corollary 11.2.9: Let (. . . o) be an extremal Sasakian structure in F(.

J).
Then the metric o M(.

J) is Sasaki-Einstein if and only if the basic rst Chern
class satises 2c
1
((T

)) = 2n[d]
B
. and the Sasaki-Futaki invariant SF
,

J
van-
ishes.
382 11. SASAKI-EINSTEIN GEOMETRY
Proof. If o is Sasaki-Einstein, then o must have positive Einstein constant 2n.
and all the implications follow readily. Conversely, if o is extremal in F(.

J) and
SF
,

J
vanishes, its scalar curvature is constant. It follows that the scalar curvature
of the transverse metric is constant also, and this implies that
g
+ 2d =
T
is
transversally harmonic. As this form represents 2c
1
(T

). which is also represented

by a constant multiple of d. the uniqueness of the harmonic representative of a
class implies that
T
= d for some 0. It then follows from this that the Ricci
tensor Ric
g
is a constant multiple of o. Thus, o is Einstein.
Example 11.2.10: The only example so far where one has an explicit computation
of the Sasaki-Futaki invariant is for the weighted Sasakian structures in the standard
CR structure on the sphere o
2n+1
. and even for this case the computations are
somewhat involved. We refer to Examples 7.1.12 and 8.4.39 for a discussion of
these weighted spheres and their toric geometry. Indeed, in [BGS06] the Sasaki-
Futaki invariant, SF

w
,

J
w
. is computed for an arbitrary weight vector w (R
+
)
n+1
.
and it is shown that for the standard CR structure on o
2n+1
the entire Sasaki
cone t
+
n+1
(T. J) of Corollary 8.2.14 is represented by extremal Sasakian metrics.
Moreover, an explicit expression for SF

w
,

J
w
is given which shows that it vanishes
if and only if w = n(1. . . . . 1) for some real number n. Moreover, the only Sasaki-
Einstein metric is when n = 1. proving the veracity of Conjecture 11.1.18 for
this case. Note that identifying the Lie algebra t
n+1
with its dual t

n+1
gives an
identication of the Sasaki cone t
+
n+1
with the moment cone minus the cone point,
C() 0.
11.3. Further Obstructions to Sasaki-Einstein Structures
We have already discussed several obstructions to the existence of a Sasaki-
Einstein structure (in fact, of a Sasaki--Einstein structure). First there is Theo-
rem 11.1.2 which says in order for a Sasakian manifold (`. o) to admit a Sasaki-
Einstein structure, the rst Chern class c
1
(T) of the contact bundle T of o must
be a torsion class in H
2
(`. Z). Moreover, the basic rst Chern class c
1
(T

) must
be positive. Both of these are conditions on the underlying contact CR structure.
But the rst actually gives a topological obstruction, namely if
1
(`) = id. then
n
2
(`) must vanish for ` to admit a Sasaki-Einstein structure. If these obstruc-
tions vanish then Corollary 11.2.9 above says that the non-vanishing of the Sasaki-
Futaki invariant is another obstruction to the existence of a Sasaki-Einstein metric
within the class F(.

J). We now give three more obstructions the rst of which is
a straightforward generalization of Theorem 5.3.1 as well as Calabis generalization
[Cal85]. The other two obstructions are quite new, arising from recent observations
of Gauntlett, Martelli, Sparks, and Yau [GMSY06].
11.3.1. Calabi-Matsushima-Lichnerowicz Obstruction. First notice that
the Lie algebra h
T
(.

J) is innite-dimensional, since the sections (1

) form a sub-
algebra. However, owing to (i) of Proposition 11.2.7 it is really only the quotient
algebra h
T
(.

J)(1

) that is of interest. We are now ready to extend to the

Sasakian context a result of Calabi [Cal85] on the structure of the algebra of holo-
morphic vector elds of a Kahler manifold with an extremal metric. This result is
an extension of work of Lichnerowicz [Lic57] on constant scalar curvature metrics,
itself an extension of a result of Matsushima [Mat57b] in the Kahler-Einstein case.
11.3. FURTHER OBSTRUCTIONS TO SASAKI-EINSTEIN STRUCTURES 383
We also refer the reader to the theorem for harmonic Kahler foliations in [NT88],
which is relevant in this context.
Consider a Sasaki structure o = (. . . o) in F(.

J). Let H
B
g
be the space of
basic functions in the kernel of the operator 1
B
g
in (11.2.4), and consider the map
(11.3.1)
#
g
: H
B
g
h
T
(.

J)(1

) .
where
#
g
is the operator dened in (11.2.2). We use the Sasaki metric o to identify
the quotient space in the right side above with the holomorphic sections of (T. J).
which we shall refer to from here on as h(T. J). We also dene the operator

1
B
g
on
H
B
g
by

1
B
g
= 1
B
g
. It follows that
(11.3.2) (

1
B
g
1
B
g
) =
#
g
:
g

#
g
:
g
.
where :
g
is the scalar curvature of o. a basic function.
The image h
0

= H
B
g
C of the map (11.3.1) is an ideal in h(T. J). and can be
identied with the space of holomorphic elds that have non-empty zero set. The
quotient algebra h(T. J)h
0
is Abelian.
Suppose now that (. . . o) is an extremal representative of F(.

J). so that o
is an extremal Sasakian metric. Let z
0
be the image under
#
g
of the set of purely
imaginary functions in H
B
g
. This is just the space of Killing elds for the transverse
metric o
T
that are of the form J
g
T . H
B
g
. Furthermore, by (11.3.2) we see
that the complexication z
0

Jz
0
coincides with the commutator of
#
g
:
g
:
z
0

Jz
0
= A h(T. J) : [A.
#
g
:
g
] = 0 .
Notice that we are implicitly using the fact here that
#
g
:
g
denes a section of
h(T. J). as :
g
is a basic function. We come to the main result of this section which
rst appeared in [BGS06].
Theorem 11.3.1: Suppose that there exists an extremal Sasakian structure (. . . o)
in F(.

J). Let H
B
g
be the space of basic functions in the kernel of the operator 1
B
g
in
(11.2.4), and let h
0
be the image of the map (11.3.1). Then we have the orthogonal
decomposition
h
T
(.

J)(1

)

= h(T. J) = a h
0
.
where a is the algebra of parallel vector elds relative to the transverse metric o
T
.
Furthermore,
h
0
= z
0

Jz
0
(
>0
h

) .
where z
0
is the image of the purely imaginary elements of H
B
g
under
#
g
, and h

=
A h
T
(.

J)(1

) : [A.
#
g
:
g
] = A. In particular, the Lie algebra aut(

J. o
T
)
of Killing vector elds for the transverse metric o
T
is equal to
aut(

J. o
T
) = a z
0

= a aut(. . . o) .
The presence of the algebra a above does not contradict the absence of non-
trivial parallel vector elds on a closed Sasakian manifold. A vector eld can be
parallel with respect to o
T
without being parallel with respect to o.
Proof. We sketch the main points, as the argument is an adaptation to our
situation of that in [Cal85, Sim04]. Given a section A in h(T. J). we look at the
Hodge decomposition of the (0. 1)-form that corresponds to it via the metric o
T
. It
is

-closed, and both, its harmonic and

components, are the dual of holomorphic
elds. The vector eld dual to the harmonic component is o
T
-parallel. Since o
T
is
384 11. SASAKI-EINSTEIN GEOMETRY
an extremal metric, the operators 1
B
g
and

1
B
g
commute. We then restrict

1
B
g
to H
B
g
.
and use the resulting eigenspace decomposition together with the identity (11.3.2) to
prove the rst part of the theorem. The identication of z
0
with aut(. . . o)
is essentially a consequence of the exact sequence 8.1.4.
Remark 11.3.1: This result obstructs the existence of special extremal Sasakian
structures in the same way it does in the Kahlerian case. For instance, let (. )
be the one-point or two-point blow-up of CP
2
. and consider the circle bundle
: ` with Euler class []. Now ` has a natural Sasakian structure
o = (. . . o) satisfying

= d, then F(.

J) contains no Sasakian structure
(. .

. o) with o a metric of constant scalar curvature. By Theorem 11.3.1 the
structure of h
T
(T

.

J)(1

) obstructs it.
11.3.2. Bishops Obstruction. In the following two subsections we will briey
discuss two obstructions recently discovered by Gauntlett, Martelli, Sparks, and
Yau [GMSY06]. It appears that the new obstructions are non-trivial only in the
case when the fundamental foliation is not regular. For this reason they have gone
unnoticed as obstructions to the existence of both positive Kahler-Einstein metrics
and Sasaki-Einstein metrics. The rst obstruction is quite interesting arising from
a classical inequality due to Bishop which rst appeared in a short note of Bishop
in the Notices of the American Mathematical Society in 1963, and then appeared in
the 1964 book of Bishop and Crittenden which has been recently reprinted by AMS
Chelsea [BC01]. An extended version due to Gromov [Gro81, Gro99] appears in
most recent texts in Riemannian geometry, cf. [Pet98, Ber03]. We only need the
original Bishops inequality.
Theorem 11.3.2: Let (`
n
. o) be a compact Einstein manifold of positive Einstein
constant normalized so that the metric cone (C(`) = R
+
`. o = d:
2
+ :
2
o) is
Ricci-at, so Ric(o) = (n 1)o by Lemma 11.1.5. Then
vol(`. o) vol(o
n
. o
can
) .
where o
can
is the unit round sphere metric on o
n
.
Now, let `
2n+1
be a compact manifold with Sasakian structure o = (. . . o).
Recall that by Proposition 7.5.10 the volume of any Sasakian metric in F() is the
same. On the other hand, if one includes transverse homotheties and consider
F(T

). one can scale the volume of a Sasakian metric to an arbitrary chosen value.
However, it follows from Theorem 11.8.2 below that if a Sasaki-Einstein metric in
F(T

) exists, it is unique within the transverse homothety class.

Denition 11.3.3: Let `
2n+1
be a compact manifold with a Sasakian structure
o = (. . . o). The ratio
() =
vol(`
2n+1
. o)
vol(o
2n+1
. o
can
)
is called the normalized volume of the Sasakian structure o.
Let us rephrase Theorem 11.3.2 for the special case of Sasakian structures.
Corollary 11.3.4: Let `
2n+1
be a compact manifold with Sasakian structure o =
(. . . o). If there is a Sasaki-Einstein metric in F(T

). for some
t
= c then
(
t
) 1.
11.3. FURTHER OBSTRUCTIONS TO SASAKI-EINSTEIN STRUCTURES 385
Of course, a priori it is not clear that the inequality of Corollary 11.3.4 should
lead to any obstruction. However, we know from Proposition 7.5.10 that the volume
of any Sasakian structure (. . . o) F() is the same. Suppose that one could
compute the volume of any, hence all structures in F(). If such a computation
were to yield a normalized volume greater than one, Corollary 11.3.4 would imply
that none of the Sasakian structures in F() could be Sasaki-Einstein. Indeed, this
was recently observed to be the case in [GMSY06]. We stress here that one needs
the techniques of [MSY06a, MSY06b] to actually compute the volumes in many
cases. In the theory surrounding Maldacenas AdS/CFT conjecture the normalized
volume of the Sasaki-Einstein horizon metric is related to the coupling constants
of the dual theory. This will be discussed in the last chapter. In [MSY06a]
Martelli, Sparks and Yau describe the methods of computing the volumes of Sasaki-
Einstein metrics in some special situations.
Example 11.3.5: Let 1

(w; d) be the link of an isolated hypersurface singularity

with weight vector w = (n
0
. . . . . n
n
) and degree d. Since the singularity is isolated
the surface A
d
CP
n
(w) is quasi-smooth. The physicists have computed the
normalized volume (w; d. n) of such a link and have shown that
(11.3.3) (w; d. n) =
d([w[ d)
n
n
n

n
i=0
n
i
.
This formula was rst derived by Bergman and Herzog [BH02] under the assump-
tion that the link is well-formed, but this assumption is unnecessary as was noticed
in [MSY06b]. Hence, we have [GMSY06]
Theorem 11.3.6: Let 1

(w; d) be a smooth link with the Sasakian structure de-

ned by diagram (9.2.3). If
d([w[ d)
n
n
n
n

i=0
n
i
then 1

(w; d) cannot admit any Sasaki-Einstein structure in F(T

(w; d)).
We shall see explicitly in one of the next sections that this obstruction rules
out Sasaki-Einstein metrics in case of many links. Here we give two examples.
Example 11.3.7: Consider the link 1(2. 2. . . . . 2. j) o
2m+1
. where n = : 2.
When : and j are both odd that this is Kervaire sphere. Assume j 2 is odd.
Then we have d = 2j and 1 = j: 2j + 2. Hence, the Bishop obstruction enters
when
(j:2j + 2)
m
:
m
j
m1
.
The same inequality is obtained for j even. For any xed : at j = 2 we always have
(2:2)
m
< :
m
2
m1
. However, at some j = j
0
(:) the left hand side dominates.
For instance, when : = 3. we have (j + 2)
3
27j
2
which happens when j 20.
Another way of saying this is that 1(2. 2. 2. j) has its normalized volume bigger
than 1 for j 20. Bishop obstruction works even better for larger :. When : = 4
we have (j + 1)
4
16j
3
which is satised from j 11. When : = 5 we get j 8
and so on.
Example 11.3.8: Another interesting application of the Theorem 11.3.6 is the case
of the weighted projective space CP(w). Take 1(z) = .
0
, for example. This is a
homogeneous polynomial of degree d = n
0
and its vanishing denes a hyperplane
in CP
n
(w) which is again a weighted projective space. In particular, 1

(w; n
0
)
386 11. SASAKI-EINSTEIN GEOMETRY
is simply the weighted Sasakian sphere with weights w = (n
1
. . . . . n
n
) bering
over the weighted projective space CP
n1
( w) of one lower dimension. The normal-
ized volume formula of a hypothetical Sasaki-Einstein metric gives the following
beautiful expression
(
w
) = vol(o
2n1
w
) =

1
n

n
i=1
n
i
n

n
i=1
n
i

n
.
As the arithmetic average is greater than or equal to the geometric average we get,
Proposition 11.3.9: Let w = (n
0
. . . . . n
n
) and o
w
be the weighted Sasakian struc-
ture on o
2n+1
. Then the n
th
root of the normalized volume (T

w
) of a hypothetical
Sasaki-Einstein metric equals the ratio of the arithmetic to the geometric averages
of all the weights. In particular, Bishops inequality obstructs the existence of a
Sasaki-Einstein metric in F(T

w
) unless all the weights are equal, in which case we
get the standard Sasakian structure on the unit sphere.
11.3.3. Lichnerowiczs Obstruction. One of the rst results concerning
eigenvalues of the Laplace operator on a compact Riemannian manifold was proved
by Lichnerowicz. It is buried on page 135 of his famous book [Lic57]. The second
part of the statement is due to Obata [Oba62]. For further discussion see [Ber03].
Theorem 11.3.10: Let (`
n
. o) be a compact Riemannian manifold with Ricci
curvature Ric(o) n 1. Then the rst non-zero eigenvalue
1
of the Laplacian
satises
1
n. Furthermore, the equality
1
= n happens only for manifolds
isometric to the sphere.
It easily follows from the above theorem that
Corollary 11.3.11: Let `
2n+1
be a compact manifold with a Sasaki-Einstein struc-
ture o = (. . . o). Then the rst non-zero eigenvalue
1
of the Laplace operator

g
is bounded

1
2n + 1
and
1
= 2n+1 if and only if o is the standard Sasaki-Einstein structure on o
2n+1
.
Following [GMSY06] we shall see that Lichnerowicz bound naturally leads to
an obstruction to the existence of Sasaki-Einstein metrics on certain links. Let `
be a Sasakian manifold of dimension 2n1 with Sasakian structure o = (. . . o)
and let Y = C(`) be the associated cone with the induced Kahler structure. Let
1 be a holomorphic function on Y with
(11.3.4)

1 = ci1 .
where c 0 is a real constant called charge of 1 with respect to . Since 1 is
holomorphic, this immediately implies that 1 = :
c
1. where

1 is homogeneous of
degree zero under the Euler eld = :
r
. In particular,

1 is the pull-back to Y of
a function on the base of the cone `. Moreover, since (Y. o) is Kahler,
Y
1 = 0.
where
Y
=
2
Y
is the Laplacian on (Y. o). For a metric cone, this is related to
the Laplacian on the base (`. o) at : = 1 by
(11.3.5)
Y
=
1
:
2

M

1
:
2n1

:
2n1

:

.
From this, one sees that
11.3. FURTHER OBSTRUCTIONS TO SASAKI-EINSTEIN STRUCTURES 387
(11.3.6)
M

1 =

1 .
where = c[c +(2n2)]. Thus any holomorphic function 1 of positive charge w.r.t
, or, equivalently, degree zero under . corresponds to an eigenfunction of the
Laplacian on the base `. The charge c is then related simply to the eigenvalue
by the above formula.
Let us now assume that, (Y. o) is Ricci-at Kahler, or equivalently that (`. o)
is Sasaki-Einstein. The rst non-zero eigenvalue
1
of
M
is bounded from below
as
1
2n1. This is of course Lichnerowiczs
1
-theorem. In terms of the charge
c the Lichnerowiczs bound becomes c 1. This leads to a simple new holomorphic
obstruction to the existence of Sasaki-Einstein metrics on many links as well as an
obstruction to the existence of Kahler Ricci-at Calabi-Yau cones.
Theorem 11.3.12: Let ` be a (2n 1)-dimensional manifold with a positive
Sasakian structure o = (. . . o). and let Y be the cone with base ` and its
induced Kahler structure. Suppose also that the complex structure induced by o on
the cone Y has c
1
(Y ) = 0. and there exists a holomorphic function 1 on Y of charge
0 < c 1 with respect to the Reeb vector eld . Then Y admits no Ricci-at cone
metric compatible with the Kahler structure and Reeb vector eld unless c = 1. in
which case the cone Y is smooth and biholomorphic to C
n
. In particular, the space
of Sasakian structures F(T

) on ` does not have any Sasaki-Einstein structure

unless c = 1 and (`. o) is the standard round sphere.
It was shown in [GMSY06] by using a well-known result of Kobayashi and
Ochiai [KO73] that this new obstruction is non-trivial only in the non-regular
case. Again, we consider some simple examples of links.
Example 11.3.13: Let 1

(w; d) be a link of isolated hypersurface singularities with

weight w = (n
0
. . . . . n
n
) and of degree d. Let us further assume that the link is a
smooth manifold, i.e., the surface A
d
CP
n
(w) is quasi-smooth. Further assume
that n
0
n
n
. It is easy to see that there exists a holomorphic function on
the cone with charge at least
(11.3.7) c =
nn
0
[w[ d
.
Hence, we get
Theorem 11.3.14: Let 1

(w; d) be a smooth link with the Sasakian structure de-

ned by diagram (9.2.3). If the index
1 = [w[ d nn
0
then 1

(w; d) cannot admit any Sasaki-Einstein structure in F(T

(w;d)
).
We shall see explicitly in one of the next sections that this obstruction rules
out Sasaki-Einstein metrics in the case of many links. Here we just show that some
classical Brieskorn and Kervaire spheres are obstructed.
Example 11.3.15: Consider the Brieskorn spheres of dimension 2: + 1 given by
BP links 1(2. . . . . 2. 3. 6/ + 6) o
2m+3
. where n = : + 1 and : is odd. Since
d = 6(6/+5). w = (6. 2(6/+5). 3(6/+5). . . . . 3(6/+5)) and 1 = (6/+5)(3:4)+6.
Hence, the Lichnerowicz obstruction excludes links with 1 6(:+ 1) which holds
388 11. SASAKI-EINSTEIN GEOMETRY
for any / 0 and any : 3. Note the when : is even the corresponding link is
dieomorphic to the standard sphere o
2m+1
and it is still obstructed.
The second is a generalization of the example considered in [GMSY06]. Con-
sider the link 1(2. 2. . . . . 2. j) o
2m+1
, where n = : 2. When : and j are both
odd that this is Kervaire sphere. Assume rst j 2 is odd (the inequality one gets
at the end does not depend on this assumption). Then d = 2j and w = (2. j. . . . . j)
and 1 = j(: 2) + 2. The Lichnerowicz obstruction excludes links with 1 2:.
This gives
2 < j < 2
:1
:2
which can possibly be Sasaki-Einstein. When : = 3 this leaves 1(2. 2. 2. 3) and
1(2. 2. 2. 4). As j = 4 realizes the equality it would have to be o
5
and it is not
so that it can also be excluded. This leaves 1(2. 2. 2. 3). When : = 4 the only
possibility is 1(2. 2. 2. 2. 3) in which case the equality holds so we can exclude this
link because it is a rational homology sphere but not o
7
. Hence, we get
Proposition 11.3.16: The positive BP links 1(2. . . . . 2. 3. 6/+5) cannot admit any
Sasaki-Einstein structure in F((w; d)). If j 2 the positive BP links 1(2. . . . . 2. j)
cannot admit any Sasaki-Einstein structure in F((w; d)) with the possible exception
of 1(2. 2. 2. 3) o
5
.
Remark 11.3.2: Note, that just as it is the case with the Bishop Obstruction the
weighted Sasakian structure o
w
cannot admit a Sasaki-Einstein metric in F(
w
)
by Lichnerowicz obstruction, unless all weights are equal. In particular, we get
three dierent obstructions to the existence of a Kahler-Einstein metric on weighted
projective spaces.
11.4. Sasaki-Einstein Metrics in Dimensions Five
The only compact Sasaki-Einstein manifolds in dimension 3 are spherical space
forms. Dimension 5 is thus the rst interesting case from the point of view of the
existence or non-existence of any Sasaki-Einstein metrics. Indeed, as we shall see,
one encounters a true cornucopia of interesting examples even on very simple man-
ifolds such as o
5
or o
2
o
3
which is `

in the Smale-Barden notation employed

in Chapter 10. At the same time, at least in some special cases, one would hope for
some partial classication results. The question of which Smale manifolds admit
a quasi-regular Sasaki-Einstein structure is a question about the nature of the in-
clusions (11.4.1). As discussed in chapter 10 the set of positive Sasakian manifolds
o
+
B
is already known to be a relatively small subset. A complete answer would
require a classication of cyclic log del Pezzo surfaces which at the orbibundle level
should describe o
+QR
B
B
0
. In addition one needs an orbifold version of the Tians
theorem which would answer the questions about the nature of the rst inclusion
oc
QR
o
+QR
B
. The two problems are quite dierent in nature: the rst has to do
with positivity in the presence of quotient singularities, the second with stability.
Both tasks appear to be quite hard. So a complete description of all Smale mani-
folds which admit some Sasaki-Einstein structure is probably out of reach at this
time. The problem of describing the moduli space of all Sasaki-Einstein structures
on a given Smale manifold ` appears even more challenging. In the notation of
Chapter 10, ultimately we would like to understand all the inclusions
(11.4.1) oc
R
oc
QR
(o
+
B
)
0
B
0
.
11.4. SASAKI-EINSTEIN METRICS IN DIMENSIONS FIVE 389
We devote this section to a brief discussion of all that is known about them. For
organizational purposes we divide our discussion according to whether a Smale
manifold ` has torsion in H
2
(`. Z) or not. Moreover, since Sasaki-Einstein struc-
tures are ubiquitous on /(o
2
o
3
) we consider separately the case of toric Sasaki-
Einstein structures from the non-toric case. We treat the latter rst.
11.4.1. Non-toric Sasaki-Einstein Structures on /(o
2
o
3
). We begin
with the case of regular Sasaki-Einstein manifolds, i.e., the inclusion oc
R
B
0
.
By (ii) of Theorem 10.2.26 any ` o
R
B
has no torsion in H
2
(`. Z). Furthermore,
the Einstein condition eliminates the appearance of A

. so we are left with o

5
and
/(o
2
o
3
). and these must bre over a smooth del Pezzo surface with a Kahler-
Einstein metric. But as discussed at the end of Section 5.2 [Tia90] determined
precisely which compact complex surfaces admit positive Kahler-Einstein metrics,
so we can easily obtain a complete classication of regular Sasaki-Einstein struc-
tures. Furthermore, we actually get a complete description of the moduli of regular
Sasaki-Einstein structures in each case due to Tians Theorem 5.2.18. More pre-
cisely [FK89, BFGK91, BG00b]
Theorem 11.4.1: Let ` be a simply connected ve manifold with a regular Sasaki-
Einstein structure. Then ` is dieomorphic to /`

= /(o
2
o
3
). / 0 (recall
that / = 0 means o
5
) with 0 / 8. Moreover,
(i) for each / = 0. 1. 3. 4. there is precisely one regular Sasaki-Einstein struc-
ture on /(o
2
o
3
) up to isometries,
(ii) for each 5 | 8 there is a 4(/ 4) parameter family of inequivalent
regular Sasaki-Einstein structures on /(o
2
o
3
).
The regular Sasaki-Einstein structures on o
5
and o
2
o
3
are homogeneous and
by Corollary 11.1.14 (ii) these are the only homogeneous simply connected Sasaki-
Einstein 5-manifolds. The Sasaki-Einstein metric on `

= T
1
(o
3
) was discovered
independently by Kobayashi [Kob63] and Tanno [Tan79] and we will call it the
Kobayashi-Tanno metric. It follows from the classication of toric smooth del Pezzo
surfaces that these two and 3(o
2
o
3
) are the only regular toric simply connected
Sasaki-Einstein manifolds in dimension 5.
Turning to the quasi-regular case, although we can nd huge numbers of families
of Sasaki-Einstein structures on /(o
2
o
3
). there is no complete classication since
there is no classication of log del Pezzo orbifold surfaces admitting Kahler-Einstein
orbifold metrics even in the case of no branch divisors. With no branch divisors
present it follows from Theorem 4.7.14 that the total space of an o
1
-orbibundle is
dieomorphic to some /(o
2
o
3
) for some / = 0. 1. . . . (again / = 0 means o
5
).
There are actually three distinct constructions of Sasaki-Einstein structures on
these manifolds. One method uses the examples of log del Pezzo surfaces obtained
as quasi-smooth well-formed hypersurfaces A
d
CP
3
(w) of degree d and index
1 = [w[ d in weighted projective spaces. The techniques developed in Chapter 5
then can be used to show that some of these admit orbifold Kahler-Einstein metrics.
By Corollary 4.7.15 the well-formedness condition implies that the total space `
of the orbibundle has no torsion in H
2
(`. Z). The rst examples of quasi-regular
but not regular metrics on o
2
o
3
and 2(o
2
o
3
) were obtained via this method in
[BG01b]. Later Johnson and Kollar classied all anti-canonically embedded (index
1 = 1) log del Pezzo hypersurfaces in CP
3
(w) which can possibly be shown to admit
an orbifold Kahler-Einstein metric using these methods [JK01b]. Their analysis
390 11. SASAKI-EINSTEIN GEOMETRY
produced an extensive list of available examples and an even longer list of possible
candidates. We further extended this classication to a higher index 1 < 10 in
[BGN02a, BGN03b]. The combined results of [Ara02, BG03] give two families
of quasi-regular non-regular Sasaki-Einstein structures on 8(o
2
o
3
). All the rele-
vant log del Pezzo surfaces are described in section 5.4 and in the tables of Appendix
B.2. Put together they give families of quasi-regular Sasaki-Einstein structures on
/(o
2
o
3
) for 0 < / < 9. The construction of [BGK05] with slightly improved es-
timates of [GK05] gives nite number of additional examples in the / = 0 (sphere)
case. Finally, in [Kol07] Kollar uses the more general approach of Seifert bundles
to prove that for every integer / 6 there are innitely many 2(/ 1)-dimensional
families of Einstein metrics on /(o
2
o
3
). Let us summarize all these results in
the following theorem. For each / we give the minimum number of known distinct
Sasaki-Einstein structures as well as the (real) dimension of the space of eective
parameters of the largest known family. This gives a lower bound on the dimension
of the premoduli space of Einstein metrics on these manifolds. Summarizing, we
have
Theorem 11.4.2: Consider the Smale manifolds /`

= /(o
2
o
3
). We have
/`

oc
QR
for all / 0. Moreover
(i) o
5
admits at least 80 distinct non-regular quasi-regular Sasaki-Einstein
Seifert structures with the largest known family of dimension 10.
(ii) o
2
o
3
admits at least 14 distinct non-regular quasi-regular Sasaki-
Einstein Seifert structures.
(iii) With the exception of / = 2. 8 the manifolds /`

= /(o
2
o
3
) admit
innitely many distinct families of inequivalent non-regular quasi-regular
Sasaki-Einstein Seifert structures. The largest family depends on 2(/1)
parameters.
(iv) 2(o
2
o
3
) admits at least two distinct 2 parameter families plus 21 in-
equivalent non-regular quasi-regular Sasaki-Einstein structures.
(v) 8(o
2
o
3
) admits innitely many inequivalent non-regular quasi-regular
Sasaki-Einstein structures. The largest family has 16 parameters.
Proof. The proof of this theorem combines Theorems 5.4.16 and 11.1.12 to-
gether with the analysis in Section 5.5 and subsection 11.1.2 as well as Kollars
results in [Kol07] for / 9. The details are left to the reader.
It is very likely that 2(o
2
o
3
) admits innitely many inequivalent Sasaki-
Einstein structures and, if so, could be combined with the statement for general
/. There are good candidates of three countable families of examples given by the
following sequences of log del Pezzo surfaces:
A
12k+11
CP
2
(8. 4/ +1. 4/ +3. 4/ +5) . A
12k+11
CP
2
(9. 3/ +2. 3/ +5. 6/ +1) .
A
36k+24
CP
2
(6. 6/ + 5. 12/ + 8. 18/ + 9) .
They can be found in the rst table of Appendix B.2. We were not able to prove
the existence of Kahler-Einstein metrics in either case but we believe that these
examples can be shown to admit such metrics via a more subtle analysis similar to
the one introduced in [BGN03b].
On the other hand 8(o
2
o
3
) appears to be somewhat special in that the
largest known family has dimension 16 and not 14 as predicted by (iii) of Theorem
11.4. SASAKI-EINSTEIN METRICS IN DIMENSIONS FIVE 391
11.4.2. Indeed, the examples of Kollar in [Kol07] give distinct innite families de-
pending on 14 parameters. But the two log del Pezzo surfaces A
10
CP
2
(1. 2. 3. 5)
and A
15
CP
2
(1. 3. 5. 7) of index 1 are actually known to be Kahler-Einstein and
they give two distinct 16-dimensional families of non-regular Sasaki-Einstein struc-
tures. Notice in addition that the regular Sasakian structure on 8(o
2
o
3
) also
has dimension 16.
11.4.2. Toric Sasaki-Einstein Geometry. It follows from Ohs Classica-
tion Theorem 10.4.10 and Theorem 11.1.2 that toric Sasaki-Einstein structures can
only occur on the Smale manifolds /`

= /(o
2
o
3
). In addition, there are inn-
itely many distinct Sasaki-Einstein structures on o
2
o
3
but unlike the examples
used to establish Theorem 11.4.2 they have to do with Sasakian toric geometry
and arise via a completely dierent construction. Considering just the well-formed
quasi-smooth hypersurfaces in CP
3
(w) produces only the 14 examples of Theorem
11.4.2 with 4 more likely candidates all listed in Table B.3.2. The rst toric exam-
ples were discovered by Gauntlett, Martelli, Sparks, and Waldram. It follows that
o
2
o
3
admits innitely many distinct quasi-regular toric Sasaki-Einstein struc-
tures [GMSW04b]. Toric geometry of these examples was further explored in
[MS05, MSY06a, MSY06b].
Consider the symplectic reduction of C
n
(or equivalently the Sasakian reduction
of o
2n1
) as discussed in Section 8.5 by a /-dimensional torus T
k
. Every complex
representation of a T
k
on C
n
can be described by an exact sequence
0T
k
f
T
n
T
nk
0 .
The monomorphism 1

can be represented by the diagonal matrix

1

(
1
. . . . .
k
) = diag

i=1

a
i
1
i
. . . . .
k

i=1

a
i
n
i

.
where (
1
. ...
k
) o
1
o
1
= T
k
are the complex coordinates on T
k
. and
o
i

Z are the coecients of a / n integral weight matrix

k,n
(Z). The
following proposition follows easily from Theorem 8.5.3 or 8.5.4.
Proposition 11.4.3: Let A() = (C
n
` 0)T
k
() denote the Kahler quotient of
the standard at Kahler structure on (C
n
`0) by the weighted Hamiltonian T
k
-action
with an integer weight matrix . Consider the Kahler moment map
(11.4.2) j
i

(z) =
n

=1
o
i

[.

[
2
. i = 1. . . . . / .
If all minor / / determinants of are non-zero then A() = C(Y ()) is a
cone on a compact Sasakian orbifold Y () of dimension 2(n /) 1 which is the
Sasakian reduction of the standard Sasakian structure on o
2n1
. In addition, the
projectivization of A() dened by Z() = A()C

is a Kahler reduction of the

complex projective space CP
n1
by a Hamiltonian T
k
-action dened by and it is
the transverse space of the Sasakian structure on Y () induced by the quotient. If
(11.4.3)

o
i

= 0. i = 1. . . . . /
392 11. SASAKI-EINSTEIN GEOMETRY
then c
1
(A()) = c
1
(T) = 0. In particular, the orbibundle Y ()Z() is anti-
canonical. Moreover, the cone C(Y ()), its Sasakian base Y (), and the transverse
space Z() are all toric orbifolds.
Remark 11.4.1: The conditions on the matrix that assure that Y () is a smooth
manifold are straightforward to work out. They involve gcd conditions on certain
minor determinants of .
This proposition is nicely summarized by the reduction Diagram 8.5.3 which
in the current context becomes
(11.4.4)
CP
n1
o
2n1
C
n
` (0)

Z() Y () C(Y ()).
Both the toric geometry and the topology of Y () depend on . Furthermore,
Y () comes equipped with a family of Sasakian structures. When n / = 3.
assuming that Y () is simply connected (which is an additional condition on ),
we must have :(o
2
o
3
) for some : /. We will be mostly interested in the case
when : = /.
Example 11.4.4: This is a continuation of Example 8.5.5. Let us rephrase this
in terms of the notation of Proposition 11.4.3. So we consider = (/. /. |. |).
The quotient Y () is the Wang-Ziller manifold `
1,1
k,l
o
2
o
3
which bres over
Z() CP
1
CP
1
. From Example 10.4.5, the rst Chern class of the contact
bundle T is c
1
(T) = 2(/ |) where is a generator of H
2
(`
1,1
k,l
. Z). Thus, we see
that the only way to get an anticanonical Sasakian structure is to have / = |. This
gives anticanonical structures on o
2
o
3
and its Z
k
quotients (o
2
o
3
)Z
k
. These
are known to admit Sasaki-Einstein metrics. In fact, the Sasaki-Einstein metric on
o
2
o
3
is just the homogeneous Kobayashi-Tanno metric of Theorem 11.4.1.
Let us generalize this construction to an arbitrary circle reduction. We take
matrix of Proposition 11.4.3 to be = (p. q) = (j
1
. j
2
. c
1
. c
2
). where
j
1
. j
2
. c
1
. c
2
Z
+
. We consider complex coordinates (r
1
. r
2
. n
1
. n
2
) C
4
with the
corresponding circle action (r
1
. r
2
. n
1
. n
2
) (c
ip
1

r
1
. c
ip
2

r
2
. c
iq
1

n
1
. c
iq
2

n
2
).
The moment map in these coordinates corresponding to this circle action is
j = j
1
[r
1
[
2
+j
2
[r
2
[
2
c
1
[n
1
[
2
c
2
[n
2
[
2
.
So j
1
(0) is again dieomorphic to o
3
o
3
. If we denote the innitesimal generators
of the circle action in the coordinates r
i
by H
i
and those for n
i
by 1
i
. then the
innitesimal generator of this circle action is j
1
H
1
+ j
2
H
2
c
1
1
1
c
2
1
2
. and
this circle will act freely on o
3
o
3
as long as gcd(j
i
. c
j
) = 1 for all i. , = 1. 2.
The Reeb vector eld for the induced Sasakian structure can be taken as 1 =
j
1
H
1
+j
2
H
2
+c
1
1
1
+c
2
1
2
. So the induced Sasakian structure on `
p,q
o
2
o
3
is an orbibundle over CP(j
1
. j
2
) CP(c
1
. c
2
). If we let and denote generators
for H
2
orb
(CP(j
1
. j
2
). Z) and H
2
orb
(CP(c
1
. c
2
). Z). respectively, we see that the rst
Chern class of CP(j
1
. j
2
) CP(c
1
. c
2
) is c
1
= (j
1
+ j
2
) + (c
1
+ c
2
). So pulling
back to `
p,q
we nd c
1
(T) =

c
1
= (j
1
+j
2
c
1
c
2
). where is a generator for
H
2
(o
2
o
3
. Z) Z. So the Sasakian structure is anticanonical if and only if j
1
+j
2
=
c
1
+c
2
in agreement with Proposition 11.4.3. Recall from our previous discussion,
cf. Example 11.2.10, that weighted projective spaces with unequal weights do not
admit Sasaki-Einstein metrics. Thus, the Sasakian structure associated with this
orbibundle does not admit a compatible Sasaki-Einstein metric.
11.4. SASAKI-EINSTEIN METRICS IN DIMENSIONS FIVE 393
Nevertheless, Gauntlett, Martelli, Sparks, and Waldram [GMSW04b] gave
an explicit construction of a Sasaki-Einstein metric that is related to these toric
examples, but with a dierent Reeb vector eld. In fact, they gave the rst examples
of irregular Sasaki-Einstein metrics. They proved the following remarkable result
in the special case when = (p. q) = (j. j. j + c. j c). where j and c are
relatively prime nonnegative integers with j c. (The general case for / = 1 was
treated later in [CLPP05, MS05], see Remark 11.4.1 below). Let us simplify
notation a bit by writing A() = A
p,q
and Y () = Y
p,q
. Then we get:
Theorem 11.4.5: The 5-manifolds Y
p,q
are dieomorphic to o
2
o
3
. Furthermore,
Y
p,q
admits a Sasaki-Einstein structure o
p,q
= (
p,q
.
p,q
.
p,q
. o
p,q
) for any co-prime
j. c such that c < j. With the exception of (j. c) = (1. 0) all the metrics o
p,q
are
of cohomogeneity one. The isometry groups equal oO(3) l(1)
2
when both j. c
are odd, and l(2) l(1) when one of them is even, respectively. The Sasaki-
Einstein structure o
p,q
is quasi-regular if and only if
(11.4.5) 4j
2
3c
2
= n
2
which happens for innitely many j. c. Otherwise, the Sasaki-Einstein structure is
irregular of rank 2. The normalized volume of (Y
p,q
. o
p,q
) is given by the following
formula
(11.4.6) (Y
p,q
) =
c
2
[2j + (4j
2
3c
2
)
1/2
]
3j
2
[3c
2
2j
2
+j(4j
2
3c
2
)
1/2
]
and it is rational if and only if the Sasaki-Einstein structure is quasi-regular. Lastly,
o
1,0
is the homogeneous Kobayashi-Tanno metric on o
2
o
3
.
Proof. The original proof of this theorem can be found in [GMSW04b] and
the toric aspect of the metric was later described in [MS06]. The argument relies
on the explicit local expression for these metrics. Let us begin with the following
two-parameter family of local metrics
d:
2
=
1 cn
6
(d
2
+ sin
2
d
2
) +
1
n(n)c(n)
dn
2
+
c(n)
9
(d cos d)
2
+ n(n) [d +1(n)(d cos d)]
2
d:
2
B
+n(n)[d +]
2
. (11.4.7)
where
n(n) =
2(o n
2
)
1 cn
.
c(n) =
o 3n
2
+ 2cn
3
o n
2
.
1(n) =
oc 2n +n
2
c
6(o n
2
)
. (11.4.8)
and d:
2
B
is a transverse metric. We stress that d:
2
B
is not the transverse Kahler-
Einstein metric of d:
2
but rather some transverse metric. This will be discussed
later. When c = 0 one can show that this has the local form of the Kobayashi-Tanno
metric on o
2
o
3
. In such a case, it can also be seen that (j. c) = (1. 0) and the
metric is simply obtained via the Sasakian reduction of o
7
with zero momentum
level. When c = 0 one can scale the constant so that c = 1. which we shall assume
from now on. One can easily show that (11.4.7) is Einstein with Ric(o) = 4o for all
values of the constants o. c. To show that the metric is Sasaki-Einstein one needs
394 11. SASAKI-EINSTEIN GEOMETRY
to exhibit the Reeb vector eld which will be done later. At this stage we have
a one-parameter family of metrics. We need to examine for what values of the
constant o one can get a complete metric on a compact manifold.
The compactication argument is based on the properties of the solutions of
the cubic polynomial Q(n) = o 3n
2
+ 2n
3
. The analysis of [GMSW04b] rst
showed that the transverse space 1 can be made into a smooth complete compact
manifold with appropriate choices for the ranges of the coordinates. In particular,
for o (0. 1) one can take the ranges of the coordinates (. . n. ) to be 0 .
0 2, n
1
n n
2
. 0 2 so that the base space 1 is an axially
squashed o
2
-bundle over the round 2-sphere. The latter is parameterized by . .
with being an azimuthal coordinate on the axially squashed o
2
bre. This bundle
is topologically trivial, i.e., 1

= o
2
o
2
. The range of n is xed so that 1 n 0.
o n
2
0. n(n) 0. c(n) 0. Specically, n
i
are two zeroes of c(n). i.e., are two
roots of the cubic Q(n). We have
Proposition 11.4.6: The metric extends to a complete Sasaki-Einstein metric
on o
2
o
3
if and only if 0 < o < 1 and the polynomial Q(n) has three roots
n
1
< 0 < n
2
n
3
such that n
1
n
2
Q.
Proof. If 0 < o < 1 there are three real roots of Q(n). one negative n
1
and
two positive n
2
n
3
. The values n = n
1
. n
2
then correspond to the south and north
poles of the axially squashed o
2
bre. One can now choose the period of so to
describe a principle o
1
-bundle over 1. For this to happen the periods must be
commensurate
(11.4.9) 1
1
= /j. 1
2
= /c
with the periods 1
i
. i = 1. 2. given by
(11.4.10) 1
i
=
1
2

Ci
d.
where C
1
and C
2
give the standard basis for H
2
(1. Z) = Z Z. In this case, one
may take 0 2/ and the ve-dimensional space is then the total space of an
o
1
bration over 1, with Chern numbers j and c over the two 2-cycles. An explicit
calculation shows that
(11.4.11) n
2
n
1
=
3c
2j
and
(11.4.12) / =
c
3c
2
2j
2
+j(4j
2
3c
2
)
1/2
.
Moreover, as a function of the parameter (n
2
n
1
)(o) is a monotonically increasing,
taking the range (0. 32). Thus for any j. c N such that 0 < cj < 1 there
exists o (0. 1) for which (11.4.11) is satised. Furthermore, for any such j and
c which are relatively prime, the total space of the circle bundle over 1 is simply
connected. We denote this manifold by `
p,q
. A standard argument shows that
H
2
(`
p,q
. Z) = Z and, hence, `
p,q
must be dieomorphic to o
2
o
3
.
Clearly, the Lie algebra isom(`
p,q
. o
p,q
) = su(2) u(1) u(1) but the isometry
group depends on the parity of j and c. For any j. c there are three commuting
11.4. SASAKI-EINSTEIN METRICS IN DIMENSIONS FIVE 395
Killing vectors

. and

. where / so that the all three generators have

period 2. In particular, the Sasaki-Einstein manifolds `
p,q
are toric. One can
identify the toric structure of `
p,q
with that of the quotient Y
p,q
by seeing that
their Delzant polytopes are equivalent [MS06]. Of course the metric o
p,q
is not
the Sasakian quotient metric unless (j. c) = (1. 0). When (j. c) = (1. 0). one can
easily see that the isometry group is oO(3) l(1)
2
when both j. c are odd and
l(2) l(1) otherwise. In either case G = Isom(Y
p,q
. o
p,q
) makes Y
p,q
= o
2
o
3
into a cohomogeneity one manifold. In fact, one can see that there is a map 1 :
Y
p,q
[n
1
. n
2
] such that in the local chart (. . . ; n) of equation (11.4.7) 1 is
projection onto the n-coordinate. Now for n (n
1
. n
2
) we have 1
1
(n) is either
oO(3) l(1) or ol(2) l(1) depending on the pair (j. c). These are the principal
orbits. Whereas, n = n
1
and n = n
2
corresponds to the singular orbits.
Let us change the local coordinates so that
(11.4.13) (. ) = (6
t
6.
t
) .
In this new chart the metric (11.4.7) can be expressed as
(11.4.14)
d:
2
=
1 n
6
(d
2
+ sin
2
d
2
) +
dn
2
n(n)c(n)
+
1
36
n(n)c(n)(d + cos d)
2
+
1
9
[d
t
cos d +n(d + cos d)]
2
= d:
2
KE
+ (
1
3
d
t
+)
2
.
where now d:
2
KE
is the transverse Kahler-Einstein metric with Kahler form
(11.4.15) d =
1 n
3
sin d d +
1
3
dn (d + cos d) .
Hence, (
1
3
d
t
+)
2
= , where 3 = d
t
cos d+n(d +cos d) is the dual
1-form to the Reeb vector eld
(11.4.16) = 3

t
= 3

1
6

= 3

1
2/

.
In particular, =
p,q
has closed orbits if and only if / is rational which is equivalent
to the Diophantine equation (11.4.5). If / is irrational the generic orbits do not close
and their closure is the 2-torus generated by [. ]. Hence, rank of these
metrics is equal to 2. The normalized volume (Y
p,q
) was calculated in [MS06].
As mentioned earlier, Y
1,0
is just the homogeneous Kobayashi-Tanno metric on
o
2
o
3
which is both toric and regular. The next simplest example is Y
2,1
which,
as a toric contact (Sasakian) manifold, is a circle bundle over the blow up of CP
2
at one point 1
1
= CP
2
#CP
2
[MS06]. As 1
1
cannot admit any Kahler-Einstein
metric, Kobayashis bundle construction cannot give a compatible Sasaki-Einstein
structure. But there is a choice of a Reeb vector eld in the torus which makes
it possible to give Y
2,1
a Sasaki-Einstein metric. The Sasaki-Einstein structure on
Y
2,1
is not quasi-regular and this is the rst such example in the literature. Hence,
o
2
o
3
admits innitely many toric quasi-regular Sasaki-Einstein structures and
innitely many toric irregular Sasaki-Einstein structures of rank 2. We have the
following generalization of Theorem 11.4.5 due to [FOW06, CFO07]:
Theorem 11.4.7: Let Y () be as in Proposition 11.4.3. Then Y () admits a
toric Sasaki-Einstein structure which is unique up to a transverse biholomorphism.
396 11. SASAKI-EINSTEIN GEOMETRY
This existence of a Sasaki-Einstein metric is proved in [FOW06] although the
authors do not draw all the conclusions regarding possible toric Sasaki-Einstein man-
ifolds that can be obtained. They give one interesting example of an irregular
Sasaki-Einstein structure which generalizes the Y
2,1
example of [MS05] in the fol-
lowing sense: One considers a regular positive Sasakian structure on the anticanoni-
cal circle bundle over the del Pezzo surface CP
2
#2CP
2
which gives a toric Sasakian
structure on 2(o
2
o
3
). The regular Sasakian structure on 2(o
2
o
3
) cannot
have any Sasaki-Einstein metric as its existence is obstructed by both the Sasaki-
Futaki invariant, Corollary 11.2.9, and by the Sasakian version of the Lichnerowicz-
Matsushima Theorem 11.3.1. However, as it is with Y
2,1
Futaki, Ono and Wang
[FOW06] show that one can deform the regular structure to a unique irregu-
lar Sasaki-Einstein structure. A slightly dierent version of the Theorem 11.4.7 is
proved in [CFO07] where uniqueness is also established. Cho, Futaki and Ono work
with toric diagrams rather than with Kahler (Sasakian) quotients which amounts
to the same thing by Delzants construction. We should add that the results of
[CFO07] apply to the toric Sasaki-Einstein manifolds in general dimension and
not just in dimension 5.
Corollary 11.4.8: The manifolds /(o
2
o
3
) admit an innite family of toric
Sasaki-Einstein structures for each / 1.
As in the / = 1 case one would expect innitely many quasi-regular and in-
nitely many irregular such Sasaki-Einstein structures for each satisfying all the
condition.
Remarks 11.4.1: The general anticanonical circle reduction was considered inde-
pendently in two recent papers, [CLPP05, MS05]. There it was shown that for
= p = (j
1
. j
2
. c
1
. c
2
), with j
i
. c
i
Z
+
, j
1
+ j
2
= c
1
+ c
2
, and gcd(j
i
. c
j
) = 1
for all i. , = 1. 2. the 5-manifold Y () o
2
o
3
admits a Sasaki-Einstein structure
which coincides with that on Y
p,q
when j
1
= j
2
= j and c
1
= j c. c
2
= j + c.
In [CLPP05] this family is denoted by 1
5
(o. /. c). where p = (o. /. c. o / +c)
and they write the metric explicitly. However, in this case it appears to be harder
(though, in principle, possible) to write down the condition under which the Sasaki-
Einstein Reeb vector eld = (o. /. c) is quasi-regular. A priori, it is not even clear
whether the quasi-regularity condition has any additional solutions beyond those
obtained for the subfamily Y
p,q
. Moreover, it follows from [CFO07] that the met-
rics of [CLPP05, MS05] describe all possible toric Sasaki-Einstein structures on
o
2
o
3
.
There have been similar constructions of a two-parameter family A
p,q
of toric
Sasaki-Einstein metrics on 2(o
2
o
3
) [HKW05], and another two-parameter fam-
ily, called 7
p,q
. on 3(o
2
o
3
) [OY06c]. All these examples, and many more, can be
obtained as special cases of Theorem 11.4.7 as they are all Y () for some choice of .
The Y
p,q
. 1
5
(o. /. c). A
p,q
and 7
p,q
metrics have received a lot of attention because
of the role such Sasaki-Einstein manifolds play in the AdS/CFT Duality Conjecture.
They created an avalanche of papers studying the properties of these metrics from
the physics perspective [ABCC06, OY06c, OY06b, OY06a, KSY05, HEK05,
BZ05, BBC05, BFZ05, SZ05, HKW05, BLMPZ05, BHK05, BFH
+
05,
Pal05, HSY04]. The AdS/CFT duality will be briey discussed in the last chap-
ter of this book.
11.4. SASAKI-EINSTEIN METRICS IN DIMENSIONS FIVE 397
A recent construction of van Coevering shows that there many more examples
of quasi-regular toric Sasaki-Einstein structures on some Smale manifolds /(o
2
o
3
)
[vC06]. His construction is quite dierent from the one previously discussed. It
involves nding toric Sasakian submanifolds of the toric (in the quaternionic sense)
3-Sasakian 7-manifolds o() discussed in Section 13.7.1. Van Coevering proves
Theorem 11.4.9: /(o
2
o
3
) admits innite families of distinct toric quasi-regular
Sasaki-Einstein structures for each odd / 1.
Proof. We only sketch the construction referring the reader to [vC06]. Con-
sider any toric QK orbifold O (see Section 12.5 for the classication and denition)
with its twistor space Z. Now, Z is a complex 3-fold with a complex contact struc-
ture and Kahler-Einstein metric. Note that Z is not toric as a complex variety.
However, it can be shown that one can always choose a hypersurface A Z which
is a symmetric toric log del Pezzo surface. It follows from an orbifold generalization
of the result of Batyrev and Selivanova [BS99]that if Z is a symmetric toric log del
Pezzo surface, the Futaki invariant must vanish implying that Z admits an orbifold
Kahler-Einstein metric. The theorem then follows from the explicit construction
of all toric QK orbifolds o = o() as QK quotients. For any admissible matrix
(see Denition 13.7.3) we get the following diagram:
(11.4.17) `
5
()

o
7
()

A
2
()

Z
3
()

O
4
() .
where the vertical maps are orbifold Riemannian submersions discussed in Chap-
ters 7 and 13, while the horizontal maps are Sasakian and Kahlerian embeddings,
respectively. They depend on the weight matrix and are explicitly obtained by
exhibiting A() as a hypersurface in the twistor space Z(). However, the Kahler-
Einstein metric on the toric del Pezzo surface A() does not come from the explicit
Kahler-Einstein metrics on Z() and the Sasaki-Einstein metric on the orbibundle
`() does not come from the explicit 3-Sasakian metric on o().
Example 11.4.10: We describe two elementary examples of this construction. We
begin with the case when O = o
4
with the standard round sphere metric which
is self-dual and Einstein. Its twistor space is Z = CP
3
and we get the following
special case of the diagram 11.4.17
398 11. SASAKI-EINSTEIN GEOMETRY
(11.4.18)
` = o
2
o
3

o
7

A = CP
1
CP
1
CP
3

o
4
.
where the second horizontal map is the standard realization of del Pezzo surface
CP
1
CP
1
. Actually, the Sasakian embedding o
2
o
3
o
7
has been studied
in [OP99] and it produces a Sasakian -Einstein metric on o
2
o
3
discovered
by Tanno. This metric can then be deformed to the Kobayashi-Tanno metric via
a transverse homothety. Of course, this metric is Sasaki-Einstein as mentioned
previously.
The second example is more general and better illustrates the power of the
method. Consider the family of Galicki-Lawson orbifolds O(p) for a triple of posi-
tive, pairwise prime integers p = (j
1
. j
2
. j
3
). The diagram 11.4.17 becomes then
(11.4.19) `(p)

o(p)

A(p)

Z(p)

O(p).
where the second horizontal map is described explicitly in [vC06]. The case of
p = (1. 1. 1) is special as then O(p) = CP
2
with the Fubini-Study metric and all
the ve spaces involved are smooth and compact Einstein manifolds. We have
(11.4.20) o(3) = 3(o
2
o
3
)

ol(3)l(1)

A = CP
2
#3CP
2
1
2,1
= ol(3)T
2

CP
2
.
where the del Pezzo surface A = CP
2
#3CP
2
is realized as hypersurface in a ag
manifold 1
2,1
. Alternatively, combining this with the twistor quotient description
of the homogeneous ag, one can describe A = CP
2
#3CP
2
as the usual Kahler
quotient of CP
1
CP
1
CP
1
by a circle action (see, for instance, Example 7.3.11
in [Fut88]). As pointed out by Futaki the quotient metric is not Einstein. On the
other hand the existence of a unique Kahler-Einstein metric in this case was shown
by Tian and Yau [TY87] and by Siu [Siu88]. More generally, A(p) is a log del
Pezzo surface with /
2
= 4 which can be considered as weighted deformation of the
smooth surface A = CP
2
#3CP
2
just as the twistor space Z(j) is a weighted ag
11.4. SASAKI-EINSTEIN METRICS IN DIMENSIONS FIVE 399
manifold [BGM94a, BG97b]. The 5-manifolds `(p). by Smales classication,
must all be dieomorphic to 3(o
2
o
3
), but the toric structure and the Seifert
bration `(p)A(p) depend on p and we get an innite family of toric quasi-
regular Sasaki-Einstein structures on 3(o
2
o
3
). Taking an admissible /
k,k+2
gives innite families of such structures on (2/ + 1)(o
2
o
3
).
It should be noted that van Coevering method fails to produce any toric Sasaki-
Einstein metrics on 2/(o
2
o
3
) which are also toric as Sasakian and contact man-
ifolds and they admit toric Sasaki-Einstein metrics for each / 1.
Open Problem 11.4.1: The Sasaki-Einstein manifolds Y
p,q
are of cohomogeneity
one. A very natural problem is to classify all Sasaki-Einstein 5-manifolds of cohomo-
geneity one. Any such space, if simply connected, must be dieomorphic to either
o
2
o
3
or o
5
. It is likely that all cohomogeneity one Sasaki-Einstein structures
on o
2
o
3
are indeed give by the manifolds Y
p,q
. However, the question regarding
existence of cohomogeneity one Sasaki-Einstein metrics on o
5
is more subtle. Con-
sider 1(2. 2. 2. /). for example. For any / 1 odd this gives a cohomogeneity one
Sasakian structure on o
5
. However, for all but 1(2. 2. 2. 3). Sasaki-Einstein struc-
tures are obstructed by Proposition 11.3.16. The case of 1(2. 2. 2. 3) was studied in
[GMSY06] but the existence problem remains open.
Exercise 11.5: Compute the group diagram H 1

. 1
+
G for each of the
manifolds Y
p,q
.
Exercise 11.6: Let / be odd. The isometry group of the Sasakian structure on the
link 1(2. 2. 2. /) o
5
is oO(3) l(1). Give the cohomogeneity one group diagram
for this action on o
5
. The transverse space A
2k
CP(2. /. /. /) of this structure
is a cohomogeneity one orbifold Fano surface under the action of oO(3). Compute
the cohomogeneity one diagram of the oO(3) action on A
2k
and compare it with
the diagrams of the Exercise 1.6.34.
11.4.3. Sasaki-Einstein Structures on General Smale Manifolds. We
consider the problem of determining which Smale manifolds with non-vanishing
torsion in H
2
(`. Z) admit Sasaki-Einstein metrics. Clearly, a necessary condition
is that ` must admit an anticanonical Sasakian Seifert structure so that all the
restrictions of Theorems 10.2.19 and 10.2.25 apply. We begin by considering the
case of Sasaki-Einstein metrics on rational homology spheres. The rst examples of
Sasaki-Einstein metrics occurring on simply connected rational homology spheres
appeared in [BG06a] (the preprint version of this paper appeared in November,
2003), but the denitive work is that of Kollar [Kol05, Kol06b]. It is quite
remarkable that almost all rational homology 5-spheres with a positive Sasakian
structure are known to admit a Sasaki-Einstein structure.
Theorem 11.4.11: Let ` (o
+
B
)
0
. When H
2
(`. Z) = (Z
5
)
4
. (Z
4
)
2
. (Z
3
)
8
or
(Z
m
)
2
. : = 30/. : 12 the moduli space of quasi-regular Sasaki-Einstein struc-
tures on ` is non-empty and can be completely described in each case. In addition,
there are examples of Sasaki-Einstein structures on other positive rational homology
spheres of Theorem 10.2.19. More precisely, we have
(i) If ` = `
k
. / 12 for some / not divisible by 30 then ` admits between
0 and 4 positive Sasakian Seifert structures, depending on / mod 30
which are given by the following links
1(3. 3. 3. /). (/. 3. ) = 1.
400 11. SASAKI-EINSTEIN GEOMETRY
1(2. 3. 6. /). (/. 6) = 1.
1(2. 4. 4. /). (/. 2. ) = 1.
1

(/. 2/. 3/. 5; 6/). (/. 5) = 1.

Each such Seifert structure yields a unique 2-parameter family of Sasaki-
Einstein metrics naturally parameterized by the moduli space of genus 1
curves.
(ii) 2`
5
has a unique positive Sasakian Seifert structure given by the BP link
1(2. 5. 6. 6) which yields a 6-dimensional family of Sasaki-Einstein struc-
tures naturally parameterized by the moduli of genus 2 curves.
(iii) 4`
3
has a unique positive Sasakian Seifert structure given by the BP link
1(2. 3. 10. 10) which yields a 14-dimensional family of Sasaki-Einstein
structures naturally parameterized by the moduli of hyperelliptic genus 4
curves.
(iv) 2`
4
has exactly 2 distinct positive Sasakian Seifert structures given by
the links 1

(4. 8. 9. 20; 40) and 1

(4. 4. 5. 8; 20). In both cases the family

of Sasaki-Einstein metrics is naturally parameterized by the moduli space
of pairs (C. j). where : CCP
1
is a genus 2 curves and j C is not
a branch point of .
(v) The manifolds 2`
3
. 3`
3
. and `
k
. with 1 < / < 12 all admit Sasaki-
Einstein structures.
Proof. To prove (i) in the rst three cases it is enough to apply Theorem
5.4.11. This was already done in [BG06a, BG05]. In particular, also `
7
. `
8
. `
10
.
and `
11
admit Sasaki-Einstein structure which proves part of (v). In [Kol05]
Kollar improves it slightly to show that `
5
also admits a Sasaki-Einstein struc-
ture. In the case of 1

(/. 2/. 3/. 5; 6/). (/. 5) = 1. Kollar shows that it admits a

Sasaki-Einstein metric for / 3. That gives the fourth series in (i) and `
4
. `
6
. `
9
in (v).
Existence in part (ii) follows from the link representation and Theorem 5.4.11.
Kollar shows in [Kol05] the completeness of the construction.
Existence in part (iii) is proved by Kollar in [Kol05]. Let 1
n
CP
n+1
be the
cone over the degree n rational normal curve and C 1
n
a smooth intersection
of 1
n
with a quadric. Kollar shows that one gets a Kahler-Einstein metric in
the following cases: (Z. ) = (1
3
.
4
5
C). (1
3
.
3
4
C). (1
4
.
2
3
C) and (1
5
.
2
3
C). Now,
`(1
5
.
2
3
C) can be easily identied with 1(2. 3. 10. 10) and the uniqueness and
completeness of the construction is established in [Kol06b].
Existence in part (iv) is rst proved by Kollar in [Kol05] for 1

(4. 4. 5. 8; 20)
by realizing that this link is a Seifert bration over (Q.
3
4
C). where Q is the quadric
cone and a smooth degree 5 curve in Q. The existence of a Sasaki-Einstein metric
on 1

(4. 4. 5. 8; 20) as well as the uniqueness and completeness of the construction

is argued in [Kol06b].
The existence for 2`
3
was established in [Kol05] by the following example:
Let S
n
be the Hirzebruch surface with a negative section 1 S
n
with 1
2
= n
and ber 1. 2`
3
can be obtained as a Seifert bundle over (S
1
.
2
3
1 +
1
2
1). where
1 [21+41[. In the 3`
3
case, the existence was established in [BG06a] by taking
the BP link 1(3. 4. 4. 4) and using Theorem 5.4.11. Alternatively Kollar shows that
3`
3
is a Seifert bundle over (1
4
.
2
3
C).
11.4. SASAKI-EINSTEIN METRICS IN DIMENSIONS FIVE 401
To nish (v) for `
3
take the link 1

(3. 3. 4. 6; 12). The estimates in [BG06a]

show that this has Sasaki-Einstein structure. For `
2
see the remark in [Kol05],
Proof 9.6.

Hence, with the exception of n`

2
. n 2. each rational homology sphere in
(o
+
B
)
0
is known to admit a Sasaki-Einstein structure.
Open Problem 11.4.2: The methods used by Kollar in [Kol05] should make it
possible to produce Sasaki-Einstein metrics on n`
2
for small values of n 1. The
question whether Sasaki-Einstein metrics can be found for any n is probably much
harder.
In the case of Theorem 10.2.25 when both /
2
(`) 0 and the H
2
(`. Z)
tor
are
non-trivial we have the following existence result
Theorem 11.4.12: Let ` (o
+
B
)
0
and / = /
2
(`) 0.
(i) If `(Z. ) is one of the Seifert brations listed in the table of The-
orem 10.2.25 (i) then ` admits at least a two-dimensional family of
compatible Sasaki-Einstein structures.
(ii) `

#2`
4
has only one positive Seifert bration giving rise to a fam-
ily of Sasaki-Einstein structures. This structure is realized by the link
1

(4. 4. 5. 12; 24). The family is naturally parameterized by the moduli

space of pairs (C. j). where : CCP
1
is a genus 2 curve and j C is
not a branch point of .
(iii) `

#n`
2
admits a Sasaki-Einstein structure for all n 1.
Proof. For the BP links of the Theorem 10.2.25 (i) one can use simple esti-
mates of Theorem 5.4.11 and this partial result was obtained in [BG05]. This gives
the Sasaki-Einstein metrics only in 6 out of the 15 cases. In all the cases the exis-
tence is established in [Kol05]. Observe, however, that some BP links give Sasaki-
Einstein structures also for : < 12. For example, 1(3. 3. 3. 3:) and 1(2. 4. 4. 4:)
give Sasaki-Einstein structures on 6`

#`
m
and 7`

#`
m
for any : 2. The
link 1(2. 3. 6. 6:) gives a Sasaki-Einstein structure on 8`

#`
m
for any : 4,
while the link 1(2. 4. 4. 2:) gives a Sasaki-Einstein structure on 3`

#`
m
. for
any odd : 5 (see Example 74 of [BG05]).
For part (ii) one easily sees that the link is realized by the polynomial 1
6
(.
0
. .
1
)+
1
1
(.
0
. .
1
).
4
2
+ .
2
3
. where 1
i
is a polynomial of degree i and gives `

#2`
4
. The
remainder of the proof can be found in [Kol06b].
Part (iii) is proved in [Kol05], where Kollar gives the following example: Let
S
n
be the minimal ruled surface with a negative section 1 S
n
with 1
2
= n and
ber 1. Take a smooth curve C [21+(2n+3)1[ which is transverse to 1. Then
o(C) = n+2 and (S
n
. (1
1
2
)C +(1
1
m
)1) is a log Del Pezzo surface for : 2n.
These give examples with H
2
(`. Z) = Z (Z2)
2n
for any n 2.
Corollary 11.4.13: The table below summarizes existence of Sasaki-Einstein met-
rics on Smale manifolds. The rst column lists all cases when a Sasaki-Einstein met-
ric might exist. That is, any Smale manifold not listed there does not admit any
Sasaki-Einstein structure. The second column lists the cases when the existence is
known.
402 11. SASAKI-EINSTEIN GEOMETRY
` (B
+
)
0
S-E
/`

. / 0 any /
8`

#`
m
. : 2 : 4
7`

#`
m
. : 2 : 2
6`

#`
m
. : 2 : 2
5`

#`
m
. : 2 : 11
4`

#`
m
. : 2 : 4
3`

#`
m
. : 2 : = 7. 9 or : 10
2`

#`
m
. : 2 : 11
`

#`
m
. : 2 : 11
`
m
. : 2 : 2
2`
5
. 2`
4
. 4`
3
. `

#2`
4
yes
/`

#2`
3
/ = 0
/`

#3`
3
. / 0 / = 0
/`

#n`
2
. / 0. n 0 (/. n) = (0. 1) or (1. n). n 0
/`

#`
m
, / 8. 2 < : < 12
Question 11.4.1: The case of H
2
(`. Z)
tor
= (Z
5
)
4
. (Z
4
)
2
. (Z
3
)
8
and (Z
m
)
2
. :
12 is completely understood in terms of the existence and the moduli of quasi-
regular Sasaki-Einstein structures. Let us therefore summarize some outstand-
ing questions/conjectures which describe what we do not yet know about Sasaki-
Einstein geometry of Smale manifolds. Let ` B
0
.
(i) Suppose / = /
2
(`) 9. We conjecture that ` admits a Sasaki-Einstein
structure if and only if H
2
(`. Z)
tor
= 0, i.e., ` is dieomorphic to
/(o
2
o
3
) where existence is known.
(ii) Suppose ` admits a Sasaki-Einstein structure with dim(aut(. . )) 1.
We conjecture that in such a case H
2
(`. Z)
tor
= 0. Recall the Corollary
11.4.8 stating that /(o
2
o
3
) admits toric Sasaki-Einstein structures for
any / 0.
(iii) We conjecture that /`

#`
m
admit Sasaki-Einstein structures for all
0 / 8 and for all : 2 giving 31 missing cases of Corollary 11.4.13.
(iv) What can one say about Sasaki-Einstein structures on /`

#2`
3
and
/`

#3`
3
? Is / 0 possible and, if so, which are the possible values it
takes?
(v) What can one say about Sasaki-Einstein structures on /`

#:`
2
and
/`

#:`
2
? Is / = 1 possible and, if so, which are the possible values
it takes?
Remark 11.4.2: Some Smale-Barden manifolds admit other Einstein metrics. This
is known in four cases: A
1
= ol(3)oO(3). o
5
. A

and `

. A
1
is a symmetric
space and the metric is Einstein. But A
1
is not even almost contact. o
2
o
3
is
well-known to have innitely many inequivalent homogeneous Einstein metrics dis-
covered by Wang and Ziller [WZ86, WZ90]. o
5
and o
2
o
3
have innitely many
inequivalent Einstein metrics of cohomogeneity one discovered by Bohm [Boh98].
Finally, we mention that both `

= o
2
o
3
and as well as A

. the non-trivial
o
3
-bundle over o
2
. admit innitely many Einstein but non Sasaki-Einstein metrics
that were recently constructed by the physicists [GHY04, HSY05, LPP04].
11.5. SASAKI-EINSTEIN METRICS ON HOMOTOPY SPHERES 403
11.5. Sasaki-Einstein Metrics on Homotopy Spheres
Let us recall that until very recently Einstein metrics on spheres appeared
to be very rare. Any sphere o
n
admits a metric of constant sectional curvature,
known as the standard metric (actually a homothety class of metrics) on o
n
. These
metrics are homogeneous and Einstein, that is the Ricci curvature is a constant
multiple of the metric. The spheres o
4m+3
, : 1 are known to have another
oj(: + 1)-homogeneous Einstein metric discovered by Jensen [Jen73]. In addi-
tion, o
15
has a third Spin(9)-invariant homogeneous Einstein metric discovered by
Bourguignon and Karcher [BK78]. In 1982 Ziller proved that these are the only
homogeneous Einstein metrics on spheres [Zil82]. No other Einstein metrics on
spheres were known until 1998 when Bohm constructed innite sequences of non-
isometric Einstein metrics, of positive scalar curvature, on o
5
. o
6
. o
7
. o
8
. and o
9
[Boh98]. Bohms metrics are of cohomogeneity one and they are not only the rst
inhomogeneous Einstein metrics on spheres but also the rst non-standard Einstein
metrics on even-dimensional spheres. Then in September of 2003 the authors in
joint work with Janos Kollar (Evan Thomas helped us with the computer programs)
[BGK05, BGKT05] announced the existence of a plethora of Einstein metrics on
spheres in all odd dimensions greater than three, including exotic sphere which
bound parallelizable manifolds. A short review and history of our work appears in
[BG06d]. This section consists of a synopsis of our work. The main result is
Theorem 11.5.1: Let 1(a) = 1(o
0
. . . . . o
n
) be a Brieskorn-Pham link.
(i) 1(a) has a Sasaki-Einstein metric if
1 <
n

i=0
1
o
i
< 1 +
n
n 1
min
i,j

1
o
i
.
1
/
i
/
j

.
where the /
i
o
i
are dened before Theorem 5.4.11. 1(a) is homeo-
morphic to o
2n1
if and only if the conditions of Theorem (9.3.18) are
satised. The dieomorphism type can be determined as in Section 9.3.2.
(ii) If a = (o
0
. . . . . o
n
) are pairwise relatively prime. Then 1(a) is home-
omorphic to o
2n1
and it admits a Sasaki-Einstein metric if and only
if
1 <
n

i=0
1
o
i
< 1 +nmin
i

1
o
i

.
(iii) Given two sequences a and a
t
satisfying condition (i), the manifolds 1(a)
and 1(a
t
) are isometric if and only if a is a permutation of a
t
.
Proof. Parts (i) follows from Theorems 11.1.3 and 5.4.11. The existence in
part (ii) follows from the improved Ghigi-Kollar estimate of Theorem 5.4.12. The
fact that this estimate is sharp
3
follows from the Lichnerowicz obstruction of The-
orem 11.3.14 applied to this specic BP link. Part (iii) will follow by Proposition
11.1.16 and Theorem 13.3.1 below if one can show that the corresponding Kahler-
Einstein orbifolds A
d
of Theorem 5.4.11 admit no holomorphic contact structure.
This is guaranteed by the following lemma whose proof is essentially found in
[BGK05]
3
We thank Janos Kollar for pointing this out to us.
404 11. SASAKI-EINSTEIN GEOMETRY
Lemma 11.5.2: Let 1(a) be a BP link and suppose the following estimate holds:
n

i=0
1
o
i
< 1 +
n
2
min
i

1
o
i

.
Then the induced Sasakian Seifert structure on 1(a) is not part of a 3-Sasakian
structure.

Remark 11.5.1: It is interesting that the estimate of Lemma 11.5.2 lies between
the two estimates in Theorem 11.5.1. This allows for the intriguing possibility that
3-Sasakian structures could reside on some links of type (ii). We also remark that
for any link 1(a) to admit a 3-Sasakian structure n must be even. Thus, (iii) of
Theorem 11.5.1 holds for the sequences of case (ii) if n is odd, or if n is even and
a satises the estimate of Lemma 11.5.2.
Using the theory described in Section 9.3.2 one can use Theorem 11.5.1 to
enumerate all :-tuples a in every dimension which give homotopy spheres and
then, depending on : determine the dieomorphism type of the sphere 1(a). This
was done in [BGK05, BGKT05] for case (i) of the theorem. Case (ii) of the
theorem provides additional examples [GK05].
Corollary 11.5.3: On o
5
we obtain 80 families of Sasaki-Einstein metrics that
are inequivalent as Einstein structures. Some of these admit non-trivial continuous
Sasaki-Einstein deformations, the largest of which depends on 10 real parameters.
Proof. We get 68 examples of BP links which satisfy the rst inequality of the
Theorem 11.5.1. They are all positive BP links of Table B.4.3. Possible moduli can
be determined using the results of Sections 5.5 and 11.1.2. The largest family comes
from Example 5.5.8 which has (real) dimension 10. Out of the 63 sporadic spheres
in Table B.4.3 all but 1(2. 3. 7. 7) and 1(2. 3. 11. 11) can be shown to have a Sasaki-
Einstein structure using Theorem 11.5.1 (i). The remaining 7 examples are the
BP links 1(2. 3. 4. 5) and 1(2. 3. 5. /). / = 7. 8. 9. 11. 13. 14. The second inequality
gives an additional 12 examples. These are the BP links 1(2. 3. 5. /) which admit
Sasaki-Einstein structure for / = 17. 19. 23. 29. 31. 37. 41. 43. 47. 49. 53. 59. None of
these 12. however, admit moduli.
In the next odd dimension the situation becomes much more interesting. A
computer search [BGKT05] found 8. 610 distinct families of Sasaki-Einstein struc-
tures on the standard o
7
=
28
and exotic 7-spheres
i
which satisfy the inequality
(i) of Theorem 11.5.1. By the results of Kervaire and Milnor [KM63] as discussed
in Section 9.4, there are 28 oriented dieomorphism types of topological 7-spheres.
(15 types if we ignore orientation.) Then Theorem 9.4.6 together with Brieskorns
signature formula 9.4.2 allows one to decide which 1(a) corresponds to which ho-
motopy sphere by computing the Hirzebruch signature of the 8-manifold \
8
a
. This
computation was performed by a computer in [BGKT05] for the 8. 610 cases show-
ing that each oriented dieomorphism type contains several hundred distinct fami-
lies of Sasaki-Einstein metrics. There, the number of distinct families ranged from
230 on
25
to 452 on
22
. However, the results of [GK05] increase these 8. 610
cases by more than 1000. so these numbers increase somewhat. We have
Corollary 11.5.4: Each of the 28 oriented dieomorphism classes on o
7
admit
several hundred inequivalent families of Sasaki-Einstein structures, some of them
11.5. SASAKI-EINSTEIN METRICS ON HOMOTOPY SPHERES 405
(in each dieomorphism class) depending on moduli. The largest of these families
depends on 82 real parameters and occurs on the standard sphere.
Proof. One can shows that already a = (2. 3. 7. 41. j). 43 j 256 real-
izes all 28 dieomorphism types. This is an easy signature computation. In fact,
there are nitely many 5-tuples which are homology 7-spheres and satisfy either
of the inequalities of Theorem 11.5.1 giving hundreds of examples in each dieo-
morphism class. The example with the biggest moduli is given by the sequence
a = (2. 3. 7. 43. 43 31) which is the standard 7-sphere with a 2(43 2) = 82-
dimensional family of Sasaki-Einstein metrics. The Sasaki-Einstein metrics arising
from case (ii) of Theorem 11.5.1 do not contribute moduli.
Remark 11.5.2: Let us mention here that any orientation reversing dieomorphism
takes a Sasaki-Einstein metric into an Einstein metric, but not necessarily a Sasaki-
Einstein metric, since the Sasakian structure xes the orientation.
Since Milnors discovery of exotic spheres [Mil56b] the study of special Rie-
mannian metrics on them has always attracted a lot of attention. Perhaps the
most intriguing question is whether exotic spheres admit metrics of positive sec-
tional curvature. This problem remains open. In 1974 Gromoll and Meyer wrote
down a metric of non-negative sectional curvature on one of the Milnor spheres
[GM74]. More recently it has been observed by Grove and Ziller that all exotic
7-spheres which are o
3
bundles over o
4
admit cohomogeneity one metrics of non-
negative sectional curvature [GZ00]. But it is not known if any of these metrics
can be deformed to a metric of strictly positive curvature.
Lastly, we should add that although heretofore it was unknown whether Ein-
stein metrics existed on exotic spheres, Einstein metrics have been known to exist on
manifolds which are homeomorphic but not dieomorphic. In dimension 7 there are
even examples of homogeneous Einstein metrics with this property [KS88]. Kreck
and Stolz nd that there are 7-dimensional manifolds with the maximal number of
28 smooth structures, each of which admits an Einstein metric with positive scalar
curvature. Corollary 11.5.4 establishes the same result for 7-spheres. In dimension
4n + 1 we easily get
Theorem 11.5.5: For n 2 both the standard and Kervaire spheres of dimen-
sion (4n + 1) admit continuous families of inequivalent Sasaki-Einstein metrics.
The number of inequivalent families as well as the number of moduli grows doubly
exponentially with dimension.
Proof. Let c
i
be the Sylvester sequence of equation (5.4.3), consider se-
quences of the form
a = (o
0
= 2c
0
. . . . . o
m2
= 2c
m2
. o
m1
= 2. o
m
) .
where o
m
is relatively prime to all the other o
i
s. By an easy computation, condition
(i) of Theorem 11.5.1 is satised if 2c
m2
< o
m
< 2c
m1
2. The relatively prime
condition is harder to pin down, but it certainly holds if in addition o
m
is a prime
number. By the prime number theorem, the number of primes in the interval
[c
m1
. 2c
m1
] is about
c
m1
log c
m1

(1.264)
2
m2
2
m1
log 1.264
(1.264)
2
m1
4(m1)
.
406 11. SASAKI-EINSTEIN GEOMETRY
so it is still doubly exponential in :. By Theorem 9.4.3, for even :. 1(a) the
standard sphere if o
m
1 mod 8 and the Kervaire sphere if o
m
3 mod 8.
It is easy to check for all values of : that we get at least one solution of both
types.
A partial computer search yielded more than 3 10
6
cases for o
9
and more
than 10
9
cases for o
13
. including a 21300113901610-dimensional family, see Ex-
ample 11.5.10. The only Einstein metric on o
13
known before was the standard
one.
In the remaining case of n 3 mod 4 the situation is more complicated. Recall
that for these values of n the group /1
n+1
is quite large and we do not know how to
show that every member of it admits a Sasaki-Einstein structure, since our methods
do not apply to the examples given in [Bri66]. We believe, however, that this is
true:
Conjecture 11.5.6: All odd-dimensional homotopy spheres which bound paralleliz-
able manifolds admit Sasaki-Einstein metrics.
Theorem 11.5.7: The above conjecture is true in odd dimensions 1(4) and it is
true in dimension 7. 11. and 15.
Proof. In all dimensions 1(4) it follows from Proposition 11.5.5. In dimension
7 it is the Corollary 11.5.4. In dimensions 11 and 15 the proof follows a computer
search of links satisfying the estimates of Theorem 11.5.1(i). This was described in
[BGKT05].
Although we do not know the distribution of solutions between dierent dif-
feomorphism types, just as in the case of dimensions 1(4) in the remaining odd
dimensions Theorem 11.5.1(i) likewise yields doubly exponential growth.
Example 11.5.8: Consider sequences of 1(a) of the form
a = (o
1
= c
1
. . . . . o
m1
= c
m1
. o
m
) .
The troublesome part of the inequalities of Theorem 11.5.1(i) is the computation
of the /
i
. However /
i
o
i
thus it is sucient to satisfy the following stronger
restriction:
1 <
m

i=1
1
o
i
< 1 +
:1
:2
min
i,j

1
o
i
o
j

= 1 +
:1
:2

1
o
m1
o
m
.
By direct computation this is satised if c
m
c
m1
< o
m
< c
m
. At least a third of
these numbers are relatively prime to o
1
= 2 and to o
2
= 3. thus we conclude:
Proposition 11.5.9: The Theorem 11.5.1(i) yields at least
1
3
(c
m1
1)
1
3
(1.264)
2
m1
0.5
inequivalent families of Sasaki-Einstein metrics on (standard and exotic) (2:3)-
spheres.
If 2: 3 1 mod 4 then by Theorem 9.4.3, all these metrics are on the
standard sphere. If 2:3 3 mod 4 then all these metrics are on both standard
and exotic spheres but we cannot say anything in general about their distribution.
Finally, let us consider examples with the largest moduli.
11.6. THE SASAKI-EINSTEIN SEMI-GROUP 407
Example 11.5.10: We consider the sequences
a = (o
1
= c
1
. . . . . o
m1
= c
m1
. o
m
= (c
m1
2)c
m1
) .
Any two of them are relatively prime, except for gcd(o
m1
. o
m
) = c
m1
. and the
inequalities of Theorem 5.4.11 are satised. The Brieskorn-Pham polynomial has
weighted homogeneous perturbations of the form
r
a1
1
+ +r
am2
m2
+G(r
m1
. r
c
m1
2
m
) .
where G is any homogeneous polynomial of degree c
m1
. Up to coordinate changes,
these form a family of complex dimension c
m1
2. This example proves
Proposition 11.5.11: Our methods yield an at least 2(c
m1
2) 2((1.264)
2
m1

2.5)-dimensional (real) family of pairwise inequivalent Sasaki-Einstein metrics on

some (standard or exotic) (2:3)-sphere.
As before, if 2:3 1(4) then these metrics are on the standard sphere.
11.6. The Sasaki-Einstein semi-group
In this section we apply the join construction described in Section 7.6.2 to
the set of quasi-regular Sasaki-Einstein orbifolds. It is based on the toral bundle
construction of Wang and Ziller [WZ86, WZ90] for Einstein manifolds. Here we
see that the Einstein condition pins down the pair of integers (/
1
. /
2
) of equation
(7.6.2).
Denition 11.6.1: We denote by oc the subset of oO with
orb
1
= 0 and whose
Sasakian structures are also Einstein, by oc
s
the subset of oc that are smooth
manifolds, and by {c oc
s
the subset of regular Sasaki-Einstein manifolds. All
these subsets are given the subspace topology.
The condition
orb
1
= 1 is made to avoid complications. Now since every
element o oc is an o
1
-orbibundle over a compact Fano orbifold, Denition
4.4.24 of Fano index can be applied to oc. We have
Denition 11.6.2: The (Fano) index of an element o oc. denoted by Ind(o).
is dened to be the Fano index of the orbifold Z = oo
1
. Furthermore, given two
elements o
1
. o
2
oc. the relative indices |
i
of the pair (o
1
. o
2
) by
|
i
=
Ind(o
i
)
gcd(Ind(o
1
). Ind(o
2
))
for i = 1. 2.
The natural grading of oO given by equation (7.6.1) is clearly inherited by the
subset oc. that is,
oc =

n=1
oc
2n+1
.
and similarly for oc
s
and {c. For the case n = 1 it follows from a theorem of
Hamilton [Bes87] that oc
s
3
= {c
3
= jt. namely o
3
with its standard Sasaki-
Einstein structure.
We now want to restrict our join map
k
1
,k
2
of equation (7.6.2) to oc
2n
1
+1

oc
2n
2
+1
and ask when can we choose the pair (/
1
. /
2
) of relatively prime integers
such that the image of
k
1
,k
2
lies in oc
2(n
1
+n
2
)+1
? Given o
1
. o
2
oc of dimension
2n
1
+1 and 2n
2
+1 respectively, we know that their respective spaces of leaves Z
1
and Z
2
are Kahler-Einstein Fano orbifolds of complex dimension n
1
. n
2
. respectively.
408 11. SASAKI-EINSTEIN GEOMETRY
The index of Z
1
Z
2
is just gcd(Ind(Z
1
). Ind(Z
2
)). Thus, by Theorem 11.1.12 the
o
1
-orbibundle on Z
1
Z
2
whose rst Chern class is
(11.6.1)
c
1
(Z
1
Z
2
)
Ind(Z
1
Z
2
)
=
c
1
(Z
1
) +c
1
(Z
2
)
gcd(Ind(Z
1
). Ind(Z
2
))
is simply connected in the orbifold sense. So the appropriate o
1
-orbibundle over
Z
1
Z
2
is the one whose orbifold rst Chern class is given by equation (11.6.1).
Thus, if
i
denotes a Kahler form on Z
i
whose class [
i
] H
2
orb
(Z. Z) is indivisible,
we see that
c
1
(Z
1
Z
2
) = c
1
(Z
1
)+c
1
(Z
1
) = Ind(Z
1
)[
1
]+Ind(Z
2
)[
2
] = Ind(Z
1
Z
2
)[|
1

1
+|
2

2
] .
Hence, we choose the pair (/
1
. /
2
) to be the pair of relative indices (|
1
. |
2
). In terms
of the Ricci form this translates to

l
1

1
+l
2

2
= 2Ind(Z
1
Z
2
)(|
1

1
+|
2

2
) .
so |
1

1
+|
2

2
is Kahler-Einstein. We now only need to rescale the Kahler-Einstein
metric to obtain one whose scalar curvature is 4(n
1
+ n
2
)(n
1
+ n
2
+ 1). Then by
Theorem 11.1.12 we get a Sasaki-Einstein metric on the total space o
1

l
1
,l
2
o
2
of
the bundle whose orbifold rst Chern class is [|
1

1
+ |
2

2
]. We have constructed
a multiplication on oc. dened by =
l1,l2
. and refer to it as the join. It is
clear from the construction that if o
1
. o
2
oc then also o
1
o
2
oc with the
Sasaki-Einstein structure as dened above, and that this procedure can be iterated
in such a way that is associative. It is also commutative up to isomorphism.
Furthermore, is continuous in both factors. We have arrived at:
Theorem 11.6.3: The join operation =
l1,l2
gives oc the structure of a com-
mutative associative topological semi-group.
Actually we are interested in the subset oc
s
of smooth Sasaki-Einstein mani-
folds. However, as in Proposition 7.6.6 it is not true that oc
s
is a subsemi-group.
From the discussion above and Proposition 7.6.6 it is easy to see that
Proposition 11.6.4: Let o
i
oc
s
with orders
i
. and relative indices |
i
. respec-
tively. Then
(i) Ind(o
1
o
2
) = gcd(Ind(o
1
). Ind(o
2
)).
(ii) o
1
o
2
oc
s
if and only if gcd(
1
|
2
.
2
|
1
) = 1.
(iii) {c is a subsemi-group.
(iv) If o
1
is regular (i.e.,
1
= 1) and Ind(o
2
) divides Ind(o
1
). then o
1
o
2

oc
s
independently of
2
.
Before further discussion, we adapt Denition 7.6.11 to the current situation.
Denition 11.6.5: We say that o oc is oc-reducible if it can be written as
o = o
1
o
2
for some o
1
. o
2
oc. o is oc-irreducible if it is not oc-reducible.
Since a Sasaki-Einstein structure determines a unique Sasaki-Seifert structure
we have
Proposition 11.6.6: An element o oc is oc-reducible if and only if it is o-
reducible.
Using this it is easy to see that in terms of Sasaki-Einstein geometry Theorem
8.3.9 becomes
11.7. SASAKI-EINSTEIN METRICS IN DIMENSIONS SEVEN AND HIGHER 409
Theorem 11.6.7: The subset Hc of simply connected homogeneous Sasaki-Einstein
manifolds forms a subsemi-group of {c. Furthermore, a simply connected Sasaki-
Einstein homogeneous manifold corresponding to a semi-simple Lie group G is oc-
irreducible if and only if G is simple.
It is clear that up to dieomorphism there is precisely one regular oc-reducible
simply connected 5-manifold, namely the homogeneous Stiefel manifold \
2
(R
4
)
o
2
o
3
o
3
o
3
with the Kobayashi-Tanno metric. When ` is not simply
connected any reducible Sasaki-Einstein manifold would have to be a quotient of
o
2
o
3
with the Kobayashi-Tanno metric. Proposition 11.6.4 will allow us to
construct smooth examples of quasi-regular Sasaki-Einstein manifolds in all odd
dimensions greater than or equal to 7. The next section is devoted to describing
such examples.
11.7. Sasaki-Einstein Metrics in Dimensions Seven and Higher
There are several dierent methods for proving the existence of Sasaki-Einstein
metrics in dimension 7 and higher. First we mention that in the homogeneous case
Theorem 11.1.13 gives us a complete classication. Next if one knows the existence
of a Kahler-Einstein metric on Fano manifold, one can construct the appropriate
circle bundle. But this is more or less hit or miss in higher dimensions, and not
much is known generally about the existence of Kahler-Einstein metric on Fano
manifolds of high dimension. Nevertheless, the physicists [GMSW04a] have been
able to construct explicit examples in some cases. Another method involving so-
called 3-Sasakian manifolds will be discussed separately Chapter 13.
The two somewhat systematic methods which have enjoyed some success in
producing Sasaki-Einstein metrics in high dimension are: (1) the join method de-
scribed in the previous section, and (2) links of isolated hypersurface singularities.
As far as (2) is concerned we have already dedicated an entire section to proving
the existence of Sasaki-Einstein metrics on homotopy spheres in odd dimensions.
These two methods are quite complimentary in the sense that (2) only gives ex-
amples on highly connected ((n 1)-connected (2n + 1)-dimensional) manifolds;
whereas, method (1) always generates 2-dimensional cohomology classes.
Our join construction or more explicitly Theorem 11.6.3 allows us to construct
new Sasaki-Einstein structures from old ones. The price that we pay is that we
always increase dimension. For example if ` is an odd-dimensional manifold then
Theorem 11.6.3 allows to propagate any Sasaki-Einstein metric on ` to the next
odd dimension by taking a join of ` with o
3
with its standard round Sasakian
metric. Of course not every such join will be a smooth manifold but it will nev-
ertheless have an orbifold Sasaki-Einstein structure. In many cases, however, the
join produces smooth examples as indicated by Proposition 11.6.4 and as we shall
soon see. We begin the discussion with the regular case.
The classication of regular Sasaki-Einstein 5-manifolds in Section 11.4 imme-
diately gives a classication of the regular oc-reducible 7-manifolds.
Proposition 11.7.1: Any simply connected regular oc-reducible 7-manifold is one
of the following:
o
3
o
3
o
3
. o
3
o
5
. o
3
o
k
for 3 / 8. where o
k
is one of the Sasaki-Einstein circle bundles over the del
Pezzo surface discussed in Theorem 11.4.1.
410 11. SASAKI-EINSTEIN GEOMETRY
These examples have already been noted in [BFGK91]. Non simply connected
examples are obtained by dividing by a cyclic subgroup of the circle generated by
the characteristic vector eld. Notice also that for 5 / 8 the 7-manifolds o
3
o
k
have continuous families of Sasaki-Einstein structures on them. We shall discuss
these shortly, but rst we mention that in the list given in Proposition 11.7.1 only
the rst two are homogeneous as seen in (iii) of Corollary 11.1.14.
The above theorem brings into the picture 3 additional Sasaki-Einstein spaces
which are oc-irreducible. One more such example occurs as a toric Sasaki-Einstein
manifold [Mab87]
Proposition 11.7.2: The simply connected regular toric Sasaki-Einstein 7-manifolds
are precisely one of the following:
(i) the reducible Sasaki-Einstein manifolds o
3
o
3
o
3
. o
3
o
5
or o
3
o
3
,
(ii) the Kobayashi bundle over the Fano 3-fold P(O
CP
1
CP
1O
CP
1
CP
1(1. 1)).
Other regular oc-irreducible 7-manifolds can be constructed as links (Fermat
hypersurfaces, cf. Example 11.7.15 below) and we will consider them in the fol-
lowing subsection. So far a complete classication of regular Sasaki-Einstein 7-
manifolds is lacking. Although the classication of Fano 3-folds is complete, it is
still not known precisely which ones admit Kahler-Einstein metrics. Actually the
classication of Fano 3-folds has an interesting history. This classication was be-
gun by Fano, but he missed a degree 22 3-fold \
22
in CP
13
. It was then found by
Iskovskikh [Isk79] in his study of Fanos work, but a mistake was made and not all
were found. Mukai and Umemura [MU83] (See also [IP99]) produced a \
22
that is
an equivariant compactication of o1(2. C)I that was missed by Iskovskikh. Here I
is the icosahedral group. Later Prokhorov (see Proposition 4.3.11 of [IP99]) showed
that the Mukai-Umemura \
22
completes the Fano-Iskovskikh classication of Fano
3-folds. Interestingly Tian [Tia97, Tia00] showed that there are deformations of
the Mukai-Umemura \
22
which do not admit a Kahler-Einstein structure, giving
a counterexample to the folklore conjecture that every compact Kahler manifold
with no holomorphic vector elds admits a compatible Kahler-Einstein metric.
Thus, we have a blueprint for a complete classication of regular Sasaki-Einstein
7-manifolds. Determine which Fano 3-folds admit Kahler-Einstein metrics and con-
struct the corresponding anti-canonical circle bundle. Here we content ourselves by
quoting some general structure results which follow from the classication together
with results of Mori and Mukai [MM82].
Proposition 11.7.3: Let o be a compact regular Sasaki-Einstein 7-manifold. Then
(i) /
2
(o) 9.
(ii) If /
2
(o) 5 then o is reducible. Explicitly, o o
3
o
k
. where 4 / 8.
Notice that the upper bound on the second Betti number for regular Sasaki-
Einstein 7-manifolds is realized by o
3
o
8
. and there is a continuous 16-dimensional
family of Sasaki-Einstein structures on these manifolds. It is interesting to contem-
plate whether it is generally true that the bound on /
2
is realized by a oc-reducible
element in {c. If this were true then the regular Sasaki-Einstein 9-manifold with
the largest second Betti number would be o
8
o
8
with /
2
= 17. The answer proba-
bly lies in Mori theory. We end this section with a niteness theorem that follows
from Theorem 7.5.34.
Theorem 11.7.4: There are at most a nite number of deformation types of reg-
ular Sasaki-Einstein manifolds in any given dimension.
11.7. SASAKI-EINSTEIN METRICS IN DIMENSIONS SEVEN AND HIGHER 411
11.7.1. Reducible Sasaki-Einstein Manifolds. We now give a brief gen-
eral discussion of the join construction. We know that Theorem 11.6.3 constructs
Sasaki-Einstein metrics on a new orbifold, and one can impose the condition of
Proposition 11.6.4 to assure that we obtain a smooth manifold. Once this is ac-
complished the more dicult part is to determine the homeomorphism or even
dieomorphism type of the new manifold. Here we outline a procedure for deter-
mining the cohomology groups by studying the spectral sequences in the commu-
tative diagram 7.6.5 where the pair (/
1
. /
2
) is now the relative index pair (|
1
. |
2
).
so our diagram becomes
(11.7.1)
`
1
`
2
`
1
`
2
1o
1

`
1
`
2
1Z
1
1Z
2
1o
1
1o
1
.
We reiterate that the important point in the analysis of this diagram is that the
dierentials in the spectral sequence of the top bration are determined through
naturality by the dierentials in the sequence of the bottom bration. If we have
complete information about the brations in the lower row, we can get completely
cohomological information about the top row. Here we present one such result
using this technique and refer the reader to [BG00b] for more examples.
Theorem 11.7.5: Let o
k
denote the circle bundles over CP
2
#/CP
2
determined by
the positive generator on the ray containing rst Chern class c
1
(CP
2
#/CP
2
). Then
the integral cohomology ring of the Sasaki-Einstein 7-manifolds o
3
o
k
is given by
H
q
(o
3
o
k
. Z)

Z if c = 0. 7;
Z
k+1
if c = 2. 5;
Z
k
2
if c = 4;
0 if otherwise,
with the ring relations determined by
i

j
= 0. :
2
= 0. 2
i
: = 0. where
i
. :
are the / + 1 two classes with i = 1. . . . . /. Moreover, if 3 / 8 the manifolds
o
3
o
k
admit Sasaki-Einstein metrics.
Proof. We know that o
k
is dieomorphic to /(o
2
o
3
). By the Kobayashi-
Ochiai Theorem [KO73] we have Ind(o
k
) = 1. Thus, we have (|
1
. |
2
) = (2. 1). By
considering the 1
2
term of the spectral sequence of the bration in the lower row
of Diagram 11.7.1, we can compute the relevant dierentials by naturality. Letting
n denote orientation class of o
3
. and
i
the 3-classes in o
k
. we nd d
4
(n) = :
2
. and
d
2
(
i
) = 2:
i
. This gives rise to torsion classes 1
2,2
3
Z
k
2
which survive to
1

. The remainder now follows by naturality and Poincare duality. See [BG00b]
for more details.
It is interesting to replace the manifolds o
k
by one of their weighted counter-
parts in which case there are many more examples with Sasaki-Einstein metrics.
For example, we can consider the join o
3

2,1
/(o
2
o
3
)
w
or o
3

1,1
/(o
2
o
3
)
w
.
where we write o
k
(w) = /(o
2
o
3
)
w
for any of the Sasaki-Einstein manifolds
consider in [BGN03b, BGN02b, BG03, Kol07] with Fano index 1 = 1 in the
rst case, and 1 = 2 in the second. This gives Sasaki-Einstein metrics on simply
412 11. SASAKI-EINSTEIN GEOMETRY
connected 7-manifolds o
3
o
k
(w) whose rational cohomology can be determined
as in the proof of Theorem 11.7.5.
Corollary 11.7.6: There are innitely many deformation classes of Sasaki-Einstein
structures on simply connected 7-manifolds o
3
o
k
(w) with the rational cohomology
of
(/ + 1)(o
2
o
5
) .
This process can clearly be iterated as discussed in [BG00b]. For example,
by the iterated join of (n + 1) copies of o
3
one gets simply connected manifolds
of dimension 2n + 7 admitting Sasaki-Einstein structures and having the rational
cohomology of
n times
. .. .
o
2
o
2
(/ + 1)(o
2
o
5
) .
Together with our previous results in dimension ve this gives
Corollary 11.7.7: There exist Sasaki-Einstein structures on simply connected man-
ifolds having arbitrarily large second Betti number in all odd dimensions greater
three.
Let us now consider the case o
3
o
5
w
where o
5
w
is one of the 80 weighted 5-
spheres that admit Sasaki-Einstein metrics determined by Corollary 11.5.3. Rather
than using the method outlined above, we can apply Proposition 7.6.7 directly
giving sphere (or lens space) bundles over o
2
. This will allow us to determine the
homeomorphism type in many cases. Let 1
5
(|
2
) denote the lens space o
5
C
l
2
where
C
l
2
Z
l
2
is the cyclic subgroup of the circle group o
1
w
generated by the Reeb vector
eld of the corresponding Sasakian structure, and |
1
is the reduced Fano index of
o
5
w
with respect to o
3
. It is straightforward to compute |
2
= |
2
(w) as a function of
w. Then we have
Theorem 11.7.8: If gcd(|
2
.
2
) = 1 then o
3

l1,l2
o
5
w
is the total space of the bre
bundle over o
2
with bre the lens space 1
5
(|
2
). and it admits Sasaki-Einstein met-
rics. In particular, for 16 dierent weight vectors w, the manifold o
3

2,1
o
5
w
is
homeomorphic to o
2
o
5
and admits Sasaki-Einstein metrics including one 10-
dimensional family. Moreover, for three dierent weight vectors w, the manifold
o
3

1,1
o
5
w
is homeomorphic to o
2
o
5
and admits Sasaki-Einstein metrics.
Proof. As mentioned above this is constructed using Proposition 7.6.7 with
`
1
= o
3
with its standard round sphere Sasaki-Einstein structure, and `
2
one
of the Sasaki-Einstein structures on o
5
mentioned above. So the rst statement
follows. To prove the second statement we need to compute the relative indices.
Since 1(o
3
) = 2 for the standard Sasakian structure on o
3
, we need to consider
two cases, namely when 1(o
5
w
) = 1. or 2. In both cases the relative index |
2
= 1.
so we have o
3

l
1
,1
o
5
w
which is an o
5
bundle over o
2
. These are classied by their
second Stiefel-Whitney class n
2
. But by construction the orbifold rst Chern class
of CP
1
Z is proportional to the rst Chern class of the o
1
-orbibundle dening
o
3

l
1
,1
o
5
w
. This implies that n
2
(o
3

l
1
,1
o
5
w
) vanishes by Theorem 11.1.2. Now
all of the Sasakian structures on o
5
can be represented as links of Brieskorn-Pham
polynomials of the form 1 = .
a
0
0
+ .
a
1
1
+ .
a
2
2
+ .
a
3
3
or deformations thereof. So
we need to compute the Fano index 1 for the 80 cases in Corollary 11.5.3. As
mentioned in the proof of that corollary 68 were found in [BGK05], and 12 found
11.7. SASAKI-EINSTEIN METRICS IN DIMENSIONS SEVEN AND HIGHER 413
in [GK05]. For Brieskorn-Pham links the Fano index takes the form
1 = [w[ d = lcm(o
0
. o
1
. o
2
. o
3
)

i=0
1
o
i
1

.
It is easy to write a Maple program to determine the Fano index of the 80 cases.
There are 16 with 1 = 1 and 3 with 1 = 2 giving 19 in all. For example the 10
parameter family given by Example 5.5.8 has a = (2. 3. 7. 35). and one easily sees
that 1 = 1.
Remarks 11.7.1: We remark that only the o
5
w
found in [BGK05] give rise to
Sasaki-Einstein metrics on o
2
o
5
; the ones found in [GK05] all have Fano index
greater than 2 and give non-trivial lens spaces. In fact the largest |
2
obtained is
89. Generally, assuming gcd(|
2
.
2
) = 1, an easy spectral sequence argument shows
that the manifolds o
3

l1,l2
o
5
w
are simply connected with the rational homology
type of o
2
o
5
. but with H
4
(o
3

l
1
,l
2
o
5
w
. Z) Z
l
2
.
11.7.2. Irreducible Sasaki-Einstein Manifolds: Rational Homology
Spheres. Here we discuss the problem of existence of Sasaki-Einstein metrics that
are oc-irreducible, so we need to rely on methods other than the join construction.
The most common method is that of construction Sasakian structures on links of
hypersurface singularities of weighted homogeneous polynomials. This is similar to
the 5-dimensional links discussed in the previous section. We begin with the case
of rational homology spheres which appeared in [BG06a]. By Corollary 7.6.12
rational homology spheres are always oc-irreducible.
Let 1 = 1(.
1
. . . . . .
m
) be a quasi-smooth weighted homogeneous polynomial of
degree d
f
in : complex variables, and let 1
f
denote its link. Let w
f
= (n
1
. . . . . n
m
)
be the corresponding weight vector. We consider branched covers given by equa-
tion (9.3.20). Then 1
g
is a /-fold branched cover of o
2m+1
branched over the
link 1
f
. The degree of 1
g
is d
g
= lcm(/. d
f
). and the weight vector is w
g
=

d
f
gcd(k,d
f
)
.
k
gcd(k,d
f
)
w
f

. We shall always assume that / 2, since the linear case

/ = 1 is a hyperplane in a weighted projective space. Notice that if gcd(/. d) 1 and
gcd(/. n
i
) = 1 for each i = 1. . . . . :. then any common factor of / and d must divide
n
i
for each i. and so will be an overall common factor which gives an equivalent link.
So hereafter we shall assume that gcd(/. d) = 1. Theorem 9.3.17 says that the link
1
g
is a rational homology sphere such that the order [H
m1
(1
g
. Z)[ = /
bm2(L
f
)
.
where /
m2
(1
f
) is the (:2)
nd
Betti number of 1
f
. Let A = A
k,d,w
denote the
quotient orbifold of 1
g
by the natural o
1
action. It is easy to see that A
k,d,w
is
Fano if and only if
(11.7.2) /([w
f
[ d
f
) +d
f
0 .
It is easy to see that the klt condition (5.2.13) cannot be satised when [w
f
[d
f
0.
First we consider the case w
f
d = 0. Then the klt condition (5.2.13) becomes
(11.7.3) (:1)d < :/ min
i
n
i
.
This is clearly satised for / large enough, namely for /
m1
m
d max

1
w
i

. We
are particularly interested in the case of perturbations of Brieskorn-Pham (BP)
singularities where we can establish the existence results via Theorem 5.4.11. We
414 11. SASAKI-EINSTEIN GEOMETRY
will apply it to the weighted homogeneous polynomials of the form
1(.
0
. . . . . .
m
) =
m

i=0
.
a
i
i
+tj(.
0
. . . . . .
m
) .
where t C. and n(j) = lcm(o
0
. . . . . o
m
) which is the degree of the polynomial.
We impose a condition on the zero set Y (a. j) = 1
1
(0). namely the genericity
condition (GC) of Example 4.6.7.
Example 11.7.9: [Branched Covers of Calabi-Yau hypersurfaces] First we consider
/-fold branched covers of Fermat-Calabi-Yau hypersurfaces
(11.7.4) .
k
0
+.
m
1
+ .
m
m
= 0
with gcd(/. :) = 1. For : 3 the link `
2m1
k
is a simply connected nontrivial
rational homology sphere. The (:2)-nd Betti number of the Calabi-Yau link is
/
m2
= (1)
m

1 +
(1 :)
m
1
:

.
So by Theorem 9.3.17 [H
m1
(`
2m1
k
. Z)[ = /
b
m2
. For example, in dimension 7
(: = 4) we get [H
2
(`
7
k
. Z)[ = /
21
whereas, in dimension 9 we have [H
2
(`
9
k
. Z)[ =
/
204
.
It is easy to see that the klt condition in Theorem 5.4.11 is satised if /
:(:1) in which case the rational homology spheres `
2m1
k
admits a family of
Sasaki-Einstein metrics. From Theorem 5.5.7 one can easily determine the number
of eective real parameters j, namely
j = 2

2:1
:

2:
2
.
By Sterlings formula one sees that j grows exponentially with :. In dimension 7 we
get 38 real parameters, while in dimension 9 we have 202 eective real parameters.
In the orbifold category there are many Calabi-Yau hypersurfaces in weighted
projective spaces P(w). For example, one can consider w = (1. . . . . 1. : 1) of
degree d = 2(: 1). or w = (1. . . . . 1. : 2. : 2) of degree d = 3(: 2). etc.
For rational homology spheres of dimension 7. we obtain examples from branched
covers of Reids list (cf. [IF00]) of 95 log K3 surfaces, and in dimension 9 from
branched covers of the over 6000 Calabi-Yau orbifolds in complex dimension 3
[CLS90].
Example 11.7.10: [Branched Covers in the Canonical Case] We consider
links 1
F
of branched covers of a canonical Fermat hypersurface of the form
1 = .
k
0
+.
l
1
+ +.
l
m
= 0
with | :+ 1 and gcd(/. |) = 1. Here the quotient A will be Fano if and only if
/ <
|
| :
.
Also for any branched cover we are only interested in / 2. Combining this with
the Fano condition gives an upper bound on | for xed :. namely | < 2:. Thus,
we get the range for |
(11.7.5) :+ 1 | 2:1 .
11.7. SASAKI-EINSTEIN METRICS IN DIMENSIONS SEVEN AND HIGHER 415
The link 1
F
= `
2m1
k
is a rational homology sphere with [H
m1
(`
2m1
k
. Z)[ =
/
b
m2
. where
(11.7.6) /
m2
= (1)
m

1 +
(1 |)
m
1
|

.
We now look at the klt condition, that is, the right hand inequality of Theorem
5.4.11. We see that 1
F
will have Sasaki-Einstein metrics if we choose / to satisfy
(11.7.7)
(:1)|
2
(:1)|(| :) +:
< / <
|
| :
.
As above we can compute the number of eective parameters j to be
(11.7.8) j = 2

:+| 1
|

2:
2
.
As a special case we consider | = : + 1. In this case the only solution to the
inequality 11.7.7 is / = :. This gives Sasaki-Einstein metrics on rational homology
spheres depending on 2(

2m
m+1

:
2
) eective real parameters.
Other solutions can be worked out, for example, the singularity
.
k
0
+.
2m
1
+ +.
2m
m1
+.
2
m
= 0 .
This satises the inequalities of Theorem 5.4.11 if we choose / = 2:1.
Theorem 11.7.11: There exists continuous parameter families of Sasaki-Einstein
metrics on innitely many simply connected rational homology spheres in every
odd dimension greater than 3. For some of these families the number of eective
parameters grows exponentially with dimension.
Example 11.7.12: In [BGN02a] we studied sporadic rational homology 7-spheres
which admit Sasaki-Einstein metric that were extracted from a classication of well-
formed orbifold Fano hypersurfaces anti-canonically embedded in CP
4
(w) which
can be shown to admit Kahler-Einstein metrics using the estimate of Corollary
5.4.8. This classication was carried out by Johnson and Kollar in [JK01a] via
a computer search. We constructed a table of these sporadic rational homology
7-spheres which is part of an earlier preprint version of this article available as
math.DG/0108113.
Our table lists the weights w = (n
0
. n
1
. n
2
. n
3
. n
4
). degree d, Milnor number
j. and the order of H
3
(`
7
w,d
. Z) of the rational homology 7-sphere `
7
w,d
. It lists
`
7
w,d
s in increasing order of their weights beginning with n
0
. The rst entry has
weights (17. 34. 75. 125. 175) while the last has weights (357. 388. 2231. 2975. 5593).
In the rst entry the order of H
3
is huge, 17
12
a number over 500 trillion, while the
lowest order of H
3
is 13
2
= 169. In general the orders of H
3
tend to be quite large.
From a quick perusal of the table, it is easy to notice the existence of twins.
These are rational homology 7-spheres with the same degree d, Milnor number j
and order of H
3
. Twins often occur as adjacent listings with the same n
0
. but this
is not always the case as with twins d = [H
3
[ = 10881. j = 10880 and n
0
= 101 and
109. and the twins d = [H
3
[ = 7777 with n
0
= 141 and n
0
= 167. Twins may also be
members of a larger set, such as the septuplets with d = [H
3
[ = 5761 and j = 5760.
These have n
0
= 157. 157. 185. 205. 214. 253. 271. respectively. Since twins have
the same Milnor number, it is tempting to conjecture that twins correspond to
homeomorphic or even dieomorphic links, but we have no proof yet.
416 11. SASAKI-EINSTEIN GEOMETRY
11.7.3. Other Sasaki-Einstein Links. Our method of using links of hy-
persurface singularities to produce examples of Sasaki-Einstein metrics on homo-
topy spheres and rational homology spheres can also be used for general higher-
dimensional manifolds. The more dicult problem is determining the dieomor-
phism or even homeomorphism type of the manifold. We have written a computer
program that computes the relevant Betti number of the link as well as the torsion
using Orliks Conjecture 9.3.15. So it can be applied to Brieskorn-Pham links or
the other type of link where Orliks algorithm is known to hold, cf. Proposition
9.3.16. The list of 4442 well-formed quasismooth Q-Fano 3-folds of index one found
by Johnson and Kollar [JK01a] contains 1936 3-folds where the klt estimates of
Corollary 5.4.8 are satised, and so admit a Kahler-Einstein metric. Then The-
orem 11.1.12 implies that the corresponding link 1

(n
0
. n
1
. n
2
. n
3
. n
4
; d) admits
Sasaki-Einstein metrics. More generally, we have
Theorem 11.7.13: Let 1

(n
0
. . . . . n
n
; d) be the link of an isolated hypersurface
singularity of degree d, with weight vector w = (n
0
. . . . . n
n
). and Fano index 1.
Suppose also that the estimate
d1 <
n
(n 1)
min
i,j
n
i
n
j

holds. Then 1

(n
0
. . . . . n
n
; d) admits a Sasaki-Einstein metric.
As usual the above estimate is far from optimal. For example as we saw the
estimate in Theorem 5.4.11 works better in the case of BP polynomials. Given
this to be the case we conclude our discussion by giving some examples. For each
example a dierent klt estimate is used.
Example 11.7.14: Here is a pair of twins taken from the end of the Johnson-
Kollar list. The klt estimate used here is the one of Theorem 11.7.13. The two
links are 1

(355. 631. 5279. 12528; 37584) and 1

(407. 547. 5311. 12528; 37584). Our

computer program shows that they both have /
3
= 2. and Orliks algorithm suggests
that there is no torsion. However, we do not know whether Orliks Conjecture 9.3.15
holds in this case. If it does then these two links are homeomorphic to 2(o
3
o
4
)
by Theorem 9.4.10. It would be interesting to compute their dieomorphism type.
They would be dieomorphic to 2(o
3
o
4
)#
7
(w) where
7
(w) is a homotopy
7-sphere in /1
8
which in principal depends on the weight vector w.
Next we consider examples of regular Sasaki-Einstein manifolds. These are cir-
cle bundles over the well-known Fermat hypersurfaces described in Example 5.2.6.
Example 11.7.15: [Links of Fermat Hypersurfaces]. First there are the well-
known Stiefel manifolds \
2
(R
n+1
) of 2-frames in R
n+1
. These are circle bundles
over the complex quadrics 1
n,2
. and they admit Sasaki-Einstein metrics since their
base spaces 1
n,2
= G
2
(R
n+1
) admit Kahler-Einstein metrics. Furthermore, they
are all homogeneous Sasaki-Einstein, and G
2
(R
n+1
) is a generalized ag manifold
whose parabolic subgroup 1 is maximal. The manifolds \
2
(R
2m+1
) are rational
homology spheres with H
2m
(\
2
(R
2m+1
). Z) = Z
2
. whereas, the manifolds \
2
(R
2m
)
have H
2m2
(\
2
(R
2m
). Z) = Z.
It was seen in Example 5.2.6 the 1
n,p
are Fano when j < n + 1 and that
the Nadel-Sui-Tian klt conditions hold for all Fermat hypersurfaces satisfying
n
2

j n. The BP link o
p,n
= 1(j. . . . . j) then admits regular Sasaki-Einstein metrics
for each j in the range
n
2
j n. The topology can be determined from the
11.8. SASAKIAN -EINSTEIN METRICS 417
well-known equation for the n
th
Betti number
/
n
= /
n
(j) = (1)
n1

1 +
(1 j)
n+1
1
j

.
The torsion can be computed by Orliks algorithm by Proposition 9.3.16. This
suggests the general result
H
n
(o
p,n
. Z) =

Z
bn
Z
p
if n is even;
Z
bn
if n is odd.
For example in dimension 7 our computer program gives H
4
(o
4,4
. Z) = Z
60
Z
4
and H
4
(o
3,4
. Z) = Z
10
Z
3
. whereas, in dimension 9 we have H
5
(o
5,5
. Z) = Z
820
and H
5
(o
4,5
. Z) = Z
183
.
Our last example is a branched cover of a Fermat hypersurface.
Example 11.7.16: [Branched Covers of Fermat Hypersurfaces]. We consider the
link 1(:. n) = 1(:n. n. . . . . n) with BP polynomial 1 = .
mn
0
+.
n
1
+ .
n
n
of degree
:n. It is easy to see that the conditions of Theorem 5.4.11 are satised when : n.
The formula for the Betti number is straightforward, but somewhat messy and left
as an exercise. Furthermore, Orliks Conjecture holds in this case by Proposition
9.3.16. In dimension 7 our computer program suggests H
4
(1(:. 4). Z) = Z
60

Z
4m
20Z
m
. We have checked that this formula is correct for about 10 dierent
values of :. and leave its direct verication as an exercise. Similarly in dimension
9 our program suggests H
5
(1(:. 5). Z) = Z
820
204Z
m
.
Many other similar examples can be worked out.
Exercise 11.7: Discuss the Sasaki-Einstein moduli problem for the links of Ex-
ample 11.7.15.
Exercise 11.8: Discuss the Sasaki-Einstein moduli problem for the links of Ex-
ample 11.7.16.
11.8. Sasakian -Einstein Metrics
Here we discuss in some detail the less well-known case of -Einstein metrics
that were dened in Denition 11.1.1. These metrics were introduced and studied
by Okumura [Oku62] and then named by Sasaki [Sas65] in his lecture notes in
1965. A more recent study of such metrics can be found in [BGM06].
For -Einstein metrics we are mainly interested in K-contact and Sasakian
structures. More generally we could have dened an -Einstein structure as in Def-
inition 11.1.1, but where o and / are smooth functions on `. However, analogously
to the Einstein case, we have
Proposition 11.8.1: Let (. . . o) be a K-contact structure on `. and suppose
also that o satises
Ric
g
= oo +/
for smooth functions o. /. Then if n 1 the functions o and / are constants.
Proof. For an -Einstein metric we have : = (2n + 1)o + /; whereas, since o
is also K-contact we have o + / = 2n. Thus, : = 2n(o + 1) and / = 2n o. and
it suces to show that : is constant. Now the contracted Bianchi identities are
418 11. SASAKI-EINSTEIN GEOMETRY
d: = 2Ric
g
where : is the scalar curvature and is the codierential (cf.[Bes87]).
So
d: = 2Ric
g
= 2

oo + (2n o)

= 2d
B
o .
Here we have used the fact that on tensor elds : the divergence operator satises
:(A
1
. . . . . A
r
) =

i
(
Ei
:)(1
i
. A
1
. . . . . A
r
) .
where 1
i
is a local orthonormal frame. But since d: = 2ndo the result follows.
Exercise 11.9: Fill in the missing details in the proof of Proposition 11.8.1 by
showing that ((2n o) ) = (o) and (oo) = (o) +d
B
o.
Proposition 11.8.1 was rst proved by Okumura [Oku62] when o = (. . . o)
is Sasakian. Tanno [Tan79] also showed that for a Sasakian -Einstein metric o
with o 2. there is a transverse homothety transformation 7.3.10 such that the
resulting metric is Sasaki-Einstein. However, owing to Theorem 11.1.7 we can do
better [BG01a].
Theorem 11.8.2: Let (. . . o) be an -Einstein K-contact structure with o
2. Then (. . . o) is Sasakian, and if o 2 there exists a unique transverse
homothety whose resulting structure is Sasaki-Einstein.
Proof. Since o is -Einstein it follows from (ii) of Theorem 7.3.12 that the
transverse metric o
T
is Einstein with Einstein constant o +2. Since this is nonneg-
ative it follows from Theorem 11.1.7 that the almost CR structure is integrable
4
.
Thus, o is Sasakian. Now applying a transverse homothety 7.3.10 we get
(
t
.
t
.
t
. o
t
) = (
1
. . . o + (
2
) )
which is Sasakian by Proposition 7.3.20 is also Sasakian. Again using (ii) of Theo-
rem 7.3.12 we have for A. Y sections of T
Ric
g
(A. Y ) = Ric
g

T
(A. Y ) 2o
t
(A. Y ) = Ric
g
T
(A. Y ) 2o
T
(A. Y )
= Ric
g
(A. Y ) + 2o(A. Y ) 2o(A. Y ) = (o + 2 2)o
T
(A. Y )
=
o + 2 2

o
t
(A. Y ) .
where the second equality holds since the Ricci curvature of o
T
is invariant under a
homothety. Since o
t
is Sasakian Ric
g
(A.
t
) = 2no
t
(A.
t
) by Theorem 7.3.12. So
o
t
is Sasaki-Einstein if and only if o 2 and =
a+2
2n+2
.
Notice that in the 3-dimensional case (n = 1) Proposition 11.8.1 doesnt
hold. Moreover, any K-contact manifold is automatically Sasakian in dimension
3, since two-dimensional almost complex structures are automatically integrable.
A Sasakian 3-manifold ` is always -Einstein in the wider sense of the denition
where o. / are functions on `. Hence, in this book we dene a Sasakian 3-manifold
to be -Einstein as in (11.1.1). On the other hand, when ` is a metric contact
manifold one can study metrics for which Ric
g
= oo + / , with o. / C

(`)
(see [Bla02]). In dimension 3 a Sasakian -Einstein structure is quite restrictive.
4
In Theorem 11.1.7 this is proved for the case a > 2 only, but it follows easily that the
a = 2 case also holds, since Sekigawas proof holds in the Ricci at case as well (cf. [BG01a]
for details).
11.8. SASAKIAN -EINSTEIN METRICS 419
Proposition 11.8.3: A three-dimensional K-contact manifold is -Einstein if and
only if it is Sasakian with constant -sectional curvature.
Proof. The if part, which holds in all dimensions, follows easily from Tannos
classication theorem [Tan69b]. To prove the only if part we choose a local or-
thonormal bases A. A. for the Sasakian structure o = (. . . o). Now in three
dimensions the Ricci curvature and the sectional curvature are related by [Pet98]
Ric
g
(A. A) = 1(A. A) +1(A. ) = 1(A. A) + 1 .
Ric
g
(A. A) = 1(A. A) +1(A. ) = 1(A. A) + 1 .
Ric
g
(. ) = 1(A. ) +1(A. ) = 2.
So if o is -Einstein we have
Ric
g
= oo + (2 o) .
and this gives 1(A. A) = o 1.
Thus, up to transverse homothety the -Einstein metrics on a three-dimensional
manifold are the uniformizing metrics given in Theorem 10.1.3.
Now if o is a Sasakian--Einstein metric on `. then
T
= o
1
for some constant
o. so [
T
] = o[d]. Thus, as a real cohomology class c
1
(T) =

[
T
] = 0. and from
Corollary 7.5.26 we have
Theorem 11.8.4: Let (`. o. . ) be a compact Sasakian manifold with basic rst
Chern class c
1
(T

) proportional to [d]
B
. Suppose further that ` has a Sasakian-
-Einstein metric in the same deformation class F(T

) as c
1
(T

). Then, c
1
(T) is
a torsion class. In particular, if ` is simply connected then c
1
(T) = 0.
Hence, a non-torsion c
1
(T) is the obstruction to the existence of a Sasakian--
Einstein metric. In particular in the case when o 2 there is a canonical variation
to a Sasaki-Einstein metric[Tan, BG2], so c
1
(T) is an obstruction to the existence
of a Sasaki-Einstein metric on ` in the given deformation class.
Thus, Theorems 7.5.27 and 11.8.4 give
Corollary 11.8.5: Let (`. o. . ) be a compact Sasakian--Einstein manifold with
H
1
(`. Z)
tor
= 0. Then ` is spin. In particular, a compact simply connected
Sasakian--Einstein manifold is spin.
We are interested in the transverse Calabi problem and its solution, Theorem
7.5.19 due to El Kacimi-Alaoui. From this theorem it is straightforward to obtain
the transverse version of the well-known Aubin-Yau result Theorem 5.2.2, namely
Theorem 11.8.6: If the basic rst Chern class c
1
(T

) H
2
B
(T

) is zero or can
be represented by a negative denite (1. 1)-form, then there exists a Sasakian--
Einstein metric g on ` in the given -deformation class with o = 2 in the rst
case and o < 2 in the second. In both cases c
1
(T) is a torsion class.
This theorem says that all null Sasakian and negative Sasakian structures ad-
mit the corresponding Sasakian--Einstein metrics. This applies to Reids list of 95
K3 surfaces, as well as all the examples of both null and negative Sasakian struc-
tures discussed in Section 10.3. Indeed, it is easy to construct examples of negative
Sasakian -Einstein metrics on many manifolds in arbitrary odd dimensions, al-
though it appears much harder to obtain a systematic procedure as in the positive
case. Here we mention some results concerning homotopy spheres [BGM06].
420 11. SASAKI-EINSTEIN GEOMETRY
Example 11.8.7: Let the integers :
i
, i = 1. . . . . 2: be any pairwise prime positive
integers. Then the (4:+1)-dimensional link 1(2. 2:
1
. . . . . 2:
2m
. o) is dieomorphic
to the standard sphere if o 1(8) and to the Kervaire sphere if o 3(8).
The link will be negative when

i
1
ri
<
a2
a
which is easily satised for :
i
s large
enough. This gives
Theorem 11.8.8: In any dimension 4: + 1 both the standard and the Kervaire
spheres admit innitely many inequivalent negative Sasakian -Einstein structures.
In dimension 4:+3 one can construct examples of negative Sasakian structures
should exists on all homotopy spheres which bound parallelizable manifolds. That is
to say, all BP links that by Brieskorn Graph Theorem are homeomorphic to a sphere
of dimension 4: + 3 and are negative should easily contain all possible oriented
dieomorphism types. We have checked it in dimension 7 with the following
Theorem 11.8.9: All 28 oriented dieomorphism types of homotopy 7-spheres
admit negative Sasakian -Einstein structures. Hence, they all admit Lorentzian
Sasaki-Einstein structures.
Proof. Consider 1(/. /. /. /+1. j). where j is prime to both / and /+1. The
link 1 is a homotopy 7-sphere and it is negative for / and j large enough. Using
the computer codes of [BGKT05], it is easy to check that such links realize all 28
oriented dieomorphism types.
Similarly, one should be able to show that the homotopy spheres in dimensions
11 and 15 that bound parallelizable manifolds admit negative Sasakian structures.
11.8.1. Lorentzian-Sasakian Geometry. There is a canonical way to asso-
ciate a Lorentzian signature metric to any Sasakian manifold. Let o = (. . . o)
be a Sasakian structure on a manifold `. Then one easily sees that
(11.8.1) o
L
= o 2 = d ( 1l)
denes a pseudo-Riemannian metric with signature (+. . . . . +. ) on `. and it is
clearly canonically associated to the Sasakian structure. In fact every Sasakian
structure has both a Riemannian and a Lorentzian metric associated to it in a
natural way. Notice that as with Riemannian metrics the same Lorentzian metric
is associated to both o and its conjugate o
c
. Recall [Bau00, Boh03] the following
Denition 11.8.10: Let (`. o) be a Lorentzian manifold of dimension 2n+1 and
let be a time-like Killing vector eld such that o(. ) = 1. We say that ` is
Sasakian if (A) =
X
satises the condition (
X
)(Y ) = o(A. )Y +o(A. Y )
and Sasaki-Einstein if, in addition, o is Einstein.
Equivalently, we can require the metric cone ((`) = (R
+
`. dt
2
+:
2
o. d(:
2
))
to be pseudo-Kahler of signature (2. 2n). In the Sasaki-Einstein case the cone met-
ric is pseudo-Calabi-Yau and the Einstein constant must equal to 2n. This is in
complete analogy with the Riemannian case.
Recall from Example 1.4.11 that a manifold admits a Lorentzian metric if and
only if it has a nowhere vanishing vector eld.
CHAPTER 12
Quaternionic Kahler and Hyperkahler Manifolds
Quaternions were rst described by the Sir William Rowan Hamilton
1
in 1843.
Hamilton believed that his invention, like complex numbers, should play a funda-
mental role in mathematics as well as in physics. The jury is perhaps still out
on what, if any, importance should quaternions have in describing our physical
world. But there is little or no doubt that they have earned an important place in
Riemannian and algebraic geometry. Following Hitchin [Hit92] we would like to
argue that todays rich theory of quaternionic manifolds, in some sense, vindicates
Hamiltons conviction.
In this chapter we will recall some basic results concerning various quaternionic
geometries which were introduced briey from the point of view of G-structures in
Example 1.4.18. Our main focus will be on positive quaternionic Kahler (QK) and
hyeprkahler (HK) manifolds, as these two geometries are of special importance in
the description and understanding of 3-Sasakian structures, the main topic of our
next chapter. It would be impossible here, in a single chapter, to give a complete
account of what is currently known about QK and HK spaces. Each case would re-
quire a separate monograph. Our goal is to describe some of the properties of such
manifolds relevant to Sasakian Geometry. Quaternionic Kahler geometry is tradi-
tionally dened by the reduction of the holonomy group Hol(`. o) to a subgroup of
oj(n)oj(1) oO(4n. R). Observe that oj(1) oj(1) oO(4) so any oriented Rie-
mannian 4-manifold has this property. It is generally accepted and, as we shall see
later, quite natural, to extend this denition in dimension 4 via an additional cur-
vature condition: an oriented Riemannian manifold (`
4
. o) is said to be QK if the
metric o is self-dual or anti-self-dual and Einstein. Interest in QK manifolds and this
holonomy denition dates back to the celebrated Berger Theorem 1.4.8. The Lie
group oj(n)oj(1) appears on Bergers list of possible restricted holonomy groups of
an oriented Riemannian manifold (`. o) which is neither locally a product nor lo-
cally symmetric. In particular, the holonomy reduction implies that QK manifolds
are always Einstein [Ber66], though their geometric nature very much depends on
the sign of the scalar curvature. The model example of a QK manifold with positive
scalar curvature (positive QK manifold) is that of the quaternionic projective space
HP
n
. The model example of a QK manifold with negative scalar curvature (negative
QK manifold) is that of the quaternionic hyperbolic ball HH
n
. The rst attempts
to study QK manifolds span over a decade and date back to the works of Bonan
[Bon64], Kraines [Kra65, Kra66], Wolf [Wol65], Alekseevsky [Ale68, Ale75],
Gray [Gra69a], Ishihara and Konishi [IK72, Ish73, Ish74, Kon75]. At this
early stage the departing point was the holonomy reduction and eorts to under-
stand what kind of geometric structures on the manifold would naturally lead to
1
As a tribute, in this chapter, as in most literature on quaternionic structures, H stands for
Sir William Hamilton.
421
422 12. QUATERNIONIC K

AHLER AND HYPERK

AHLER MANIFOLDS
such a holonomy reduction. For example, it appears that, independently, Bonan
and Kraines were the rst to consider the fundamental 4-form of the quaternionic
structure and deduced some topological information. Wolf and Alekseevsky studied
and classied symmetric and some homogeneous examples, respectively. Ishihara
and Konishi explored some special geometric properties of such manifolds and their
relation to the 3-Sasakian spaces. The true revolution, however, came in the early
eighties. Salamon [Sal82] and, independently, Berard Bergery [BB82] realized
that QK manifolds can be studied in the language of algebraic and holomorphic
geometry. Their twistor correspondence was a generalization of the beautiful Pen-
rose twistor space construction in dimension 4 to the case of QK manifold of any
quaternionic dimension. The power of the twistor correspondence which allows for
applying algebraic geometry when dealing with problems involving positive QK
manifolds will be illustrated by many results described in this chapter.
When the scalar curvature vanishes a QK manifold is necessarily locally hy-
perkahler. In the language of holonomy the hyperkahler manifolds are character-
ized by the reduction of the holonomy group Hol(`. o) to a subgroup of oj(n)
oO(4n. R). In this sense HK geometry is a special case of QK geometry and, just
as in the previous case, we nd the Lie group oj(n) on Bergers list. The model
example of HK geometry is that of a quaternionic vector space H
n
with the at
metric. The rst study of hyperkahler manifolds appears to be that of Wakakuwa
[Wak58] who gave an example in local coordinates, but the name hyperkahler,
as well the name hypercomplex, is due to Calabi [Cal79] who constructed com-
plete hyperkahler metrics on the cotangent bundle T

CP
n
. Of course, Yaus famous
proof [Yau77] of the Calabi conjecture provides the 13 surface with a hyperkahler
structure. Hyperkahler manifolds are special cases of Calabi-Yau manifolds, and
so compact examples are important to mirror symmetry. There are several recent
books [VK99, GHJ03, NW04] treating hyperkahler manifolds. In this book we
are more interested in non-compact hyperkahler manifolds, especially hyperkahler
cones.
12.1. Quaternionic Geometry of H
n
and HP
n
The purpose of this section is to describe quaternionic geometries of some model
examples of quaternionic manifolds. We will do it in considerable detail using terms
which, in greater generality, will only be dened later. The quaternions H are the
associative, non-commutative real algebra
H = n [ n = n
0
+n
1
i
1
+n
2
i
2
+n
3
i
3
. n
a
R R
4
.
The imaginary units are often denoted by i
1
. i
2
. i
3
= i. ,. /. The imaginary
quaternions Im(H) = span(i
1
. i
2
. i
3
) R
3
and the multiplication rules are given by
the formula
(12.1.1) i
a
i
b
=
ab
+
3

c=1
c
abc
i
c
.
We dene the quaternionic conjugate c and the norm [n[ by
n = n
0

a=1
n
a
i
a
. and [n[
2
=
3

a=0
(n
a
)
2
.
12.1. QUATERNIONIC GEOMETRY OF H
n
AND HP
n
423
The non-zero quaternions H ` 0 = H

= G1(1. H) from a group isomorphic to

R
+
oj(1). where oj(1) is the subgroup of unit quaternions and the isomorphism
is given explicitly by the map n ([n[. n[n[). The group of unit quaternions
(12.1.2) oj(1) = H

G1(1. H) [ = 1 .
as a manifold, is just the unit 3-sphere in R
4
. Furthermore, we have the group
isomorphism 1 : oj(1) ol(2) explicitly given by
(12.1.3) 1() =

0
+
1
i
1

2
+
3
i
1

2
+
3
i
1

0

1
i
1

.
It is known that ojin(4) = oj(1) oj(1) and oO(4) oj(1)oj(1), where cus-
tomarily oj(1)oj(1) denotes the quotient of oj(1) oj(1) by the diagonal Z
2
.
This is yet another group isomorphism between classical groups which can be
explained using the quaternionic geometry of H R
4
. Consider the action of
G = oj(1)
+
oj(1)

on H given by
(12.1.4)
(,)
(n) = n

.
We assume the convention that the oj(1)
+
factor acts by the left quaternionic
multiplication while the oj(1)

factor acts from the right. Clearly, the two actions

commute and the Z
2
subgroup generated by (1. 1) acts trivially. The quotient
acts on R
4
preserving the Euclidean metric and orientation. This is the special
orthogonal group oO(4). It is worthwhile to write this action on R
4
. The oj(1)
+
part is given by the following group homomorphism
+
: oj(1) oO(4):
(12.1.5)
+
() =

=
0
1l
4
+
1
1
+
1
+
2
1
+
2
+
3
1
+
3
.
where the matrices 1
+
i
=
+
(c
i
)
(12.1.6)
1
+
1
=

0 1 0 0
1 0 0 0
0 0 0 1
0 0 1 0

. 1
+
2
=

0 0 1 0
0 0 0 1
1 0 0 0
0 1 0 0

. 1
+
3
=

0 0 0 1
0 0 1 0
0 1 0 0
1 0 0 0

.
give a globally dened hypercomplex structure I
+
= 1
+
1
. 1
+
2
. 1
+
3
on R
4
. For a
purely imaginary = in oj(1) one sets 1
+
() =
+
() and gets the whole
o
2
-family of complex structures. We obtain the left hyperkahler structure on H
by further setting o
0

+
= o
0
i
1

+
1
i
2

+
2
i
3

+
3
= dn d n. where the
multiplication in H is used to interpret the left hand side as an H-valued tensor.
This gives the standard Euclidean metric o
0
and the three symplectic forms
(12.1.7)
+
a
= dn
b
dn
c
+dn
0
dn
a
.
where (o. /. c) is any cyclic permutation of (1. 2. 3). We can also introduce an H-
valued dierential 2-form
(12.1.8)
+
= i
1

+
1
+i
2

+
2
+i
3

+
3
= Im(dn d n) =
1
2
dn d n.
424 12. QUATERNIONIC K

AHLER AND HYPERK

AHLER MANIFOLDS
The 2-from dn d n is purely imaginary as = (1)
pq
. where j. c are the
respective degrees. The oj(1)

part is given by

: oj(1) oO(4) with

(12.1.9)

() =

=
0
1l
4
+
1
1

1
+
2
1

2
+
3
1

3
.
The matrices 1

i
=

(c
i
)
(12.1.10)
1

1
=

0 1 0 0
1 0 0 0
0 0 0 1
0 0 1 0

. 1

2
=

0 0 1 0
0 0 0 1
1 0 0 0
0 1 0 0

. 1

3
=

0 0 0 1
0 0 1 0
0 1 0 0
1 0 0 0

.
give a globally dened hypercomplex structure I

= 1

1
. 1

2
. 1

3
. Furthermore,
with the Euclidean metric one gets the right hyperkahler structure on H by setting
o
0
+

= o
0
+i

1
+i
2

2
+i
3

3
= d n dn. This gives
(12.1.11)

= i
1

1
+i
2

2
+i
3

3
= Im(d n dn) =
1
2
d n dn.
where, as before, we get the three symplectic forms
(12.1.12)

a
= dn
b
dn
c
dn
0
dn
a
.
for each cyclic permutation of (1. 2. 3). These are clearly fundamental 2-forms
associated to the complex structures 1

a
. Note that by construction, for any (. )
one has [
+
().

()] = 0 and the product

+
()

() oO(4). In particular,
the two hypercomplex structures I
+
and I

commute. The hyperkahler structure

(o
0
. 1

a
.

a
) is preserved by oj(1)
+
(hypekahler isometry) while oj(1)

acts by
rotating the complex structures on o
2
. The role of oj(1)
+
and oj(1)

reverses
for (o
0
. 1
+
a
.
+
a
). With only little extra eort one can compactify this example
to see that another Lie group l(2) is a compact manifold with two commuting
hypercomplex structures, though l(2) admits no hyperkahler metric.
Remark 12.1.1: Consider the group of integers Z acting on H by translations of
the real axis. The action preserves both hypercomplex structures and the metric,
hence, the quotient HZ o
1
R
3
is also a at hyperkahler manifold with innite
fundamental group
1
= Z. To indicate the dierence, we will write the at metric
in this case as o
0
= d
2
+dx dx replacing r
0
with the angle coordinate .
Example 12.1.1: [Quaternionic vector spaces]. Much of the above discussion
extends to H
n
= u = (n
1
. . . . . n
n
) [ n
j
= n
0
j
+n
1
j
i
1
+n
2
j
i
2
+n
3
j
i
3
H. , = 1. . . . . n.
Here and from now on we will choose to work with the left hyperkahler structure
on H
n
, i.e., with the symplectic 2-forms given by
(12.1.13) o
0
=
n

j=1
d n
j
dn
j
so that
(12.1.14) o
0
=
n

j=1
3

a=0
(dr
a
j
)
2
.
a
=
n

j=1

dr
b
j
dr
c
j
dr
0
j
dr
a
j

12.1. QUATERNIONIC GEOMETRY OF H

n
AND HP
n
425
for any cyclic permutation (o. /. c) of (1. 2. 3). The corresponding hypercomplex
structure is then given by left multiplication by

i
1
.

i
2
.

i
3
= i. ,. / with the
standard basis as in 12.1.10, where 0. 1 are now matrices of size n n.
We associate to o
0
a quaternionic Hermitian inner product
(12.1.15) 1(u. v) =
n

j=1
n
j

j
and dene
(12.1.16) oj(n) = G1(n. H) ['u. v` = 'n. ` .
Now, oj(n) oj(1) acts on H
n
by
(12.1.17)
(A,)
(u) = u

with oj(n)oj(1) acting eectively. Clearly, oj(n)oj(1) is now a subgroup of

oO(4n). The group oj(n) assumes the role of oj(1)

and it acts by hyperkahler

isometries, while oj(1) is the previous oj(1)

and rotates the complex structures.

We will also work with complex coordinates (z. w) on H
n
C
2n
writing
(12.1.18) u = z +, w = (x
0
+ix
1
) +,(x
2
ix
3
) .
With such conventions we obtain
(12.1.19) = i
1
+
+
, = i
1
+ (
2
+i
3
), =
1
2

j
d n
j
dn
j
.
where
(12.1.20)
1
=
i
2
n

j=1

d.
j
d .
j
+dn
j
d n
j

.
+
=
n

j=1
dn
j
d.
j
.
Comparing with Example 3.1.11 we recognize (o
0
.
1
) as the standard Hermitian
metric and Hermitian form on C
2n
. In addition, the (2. 0)-from
+
is a complex
symplectic form so that
1
n!

n
+
= dn
1
dn
n
d.
1
d.
n
is the standard holomorphic volume form on C
2n
.
Example 12.1.2: Quaternionic projective space. We now use the left H

action
on H
n
to introduce another model space of quaternionic geometry.
Denition 12.1.3: The quaternionic collectivization
HP
n
= P
H
(H
n+1
) = (H
n+1
` 0)H

dened with respect to the left action of H

on H
n+1
is called the quaternionic
projective n-space.
Let o
4n+3
= u H
n+1
[ 1(u. u) = 1 be the unit sphere in H
n+1
. The group
oj(n+1) acts on o
4n+3
transitively with the isotropy at every point oj(n). Hence,
(12.1.21) o
4n+3
=
oj(n + 1)
oj(n)
426 12. QUATERNIONIC K

AHLER AND HYPERK

AHLER MANIFOLDS
is a homogeneous space and the induced metric is of constant sectional curvature
1. Note, that the oj(1) subgroup of H

acts on the sphere and we get the natural

identication
(12.1.22) HP
n
= o
4n+3
oj(1)
oj(n + 1
oj(n) oj(1)
.
so we observe that HP
n
is actually a compact rank one symmetric spaces. If, in
addition, we make a choice 1 R

we can also dene three more

projective spaces associated to HP
n
.
Denition 12.1.4: Let H
n+1
be the quaternionic vector space and HP
n
the asso-
ciated quaternionic projective space. We dene
(i) Z = P
C
(H
n+1
) = (H
n+1
` 0)C

.
(ii) o = P
R
(H
n+1
) = (H
n+1
` 0)R

.
(iii) | = P
Z
2
(H
n+1
) = (H
n+1
` 0)Z
2
.
The spaces Z. o. | are called the twistor space, the Konishi bundle, and the
Swann bundle of HP
n
. respectively.
As homogeneous spaces we have
Z = CP
2n+1

oj(n + 1)
oj(n) l(1)
. o = RP
4n+3

oj(n + 1)
oj(n) Z
2
.
Proposition 12.1.5: Let HP
n
be the quaternionic projective space. We have the
following natural bre bundles dened by 1 R

(i) H

= o
2
Z HP
n
.
(ii) H

= oO(3) o HP
n
.
(iii) H

Z
2
| HP
n
.
(iv) C

= o
1
o Z .
(v) C

Z
2
| Z .
(vi) R

Z
2
= R
+
| o .
The six bundles of this proposition are the six arrows in the following diagram
(12.1.23) (H
n
` 0)Z
2

CP
2n1

RP
4n1
.

.
.
.
.
.
.
.
.
.
.
.
.
HP
n1
We shall see later in this chapter that all these bundles exist in a more general
setting. However, the following is a very special property of HP
n
and has to do
with vanishing of a certain secondary characteristic class, the Marchiafava-Romani
class c dened in Denition 12.2.1 below. This class clearly vanishes for HP
n
since
H
2
(HP
n
. Z) = 0.
Proposition 12.1.6: With the exception of the rst one, all the bundles of the
previous proposition admit global Z
2
-lifting.
The existence of the bundle oj(1) o
4n+3
HP
n
means that the structure
group of HP
n
is can be lifted from oj(n)oj(1) to oj(n) oj(1). We will now
12.1. QUATERNIONIC GEOMETRY OF H
n
AND HP
n
427
construct an atlas on HP
n
. Consider homogeneous coordinates [n
0
. . . . . n
n
] HP
n
.
These are dened in analogy with homogeneous charts on a complex projective
space by the equivalence of non-zero vectors in H
n+1
. with u u
t
meaning u = u
t
,
for some H

. Let
(12.1.24) l
j
= [n
0
. . . . . n
n
] HP
n
[ n
j
= 0
and consider the maps
j
: l
j
H
n
dened by

j
([n
0
. . n
n
]) = (n
0
n
1
j
. . . . . n
j1
n
1
j
. n
j+1
n
1
j
. . . . . n
n
n
1
j
) . , = 0. . . . . n.
Now / = l
j
.
j

j=0,...,n
is clearly an atlas on HP
n
giving it a structure of dier-
entiable manifold. Consider the inhomogeneous quaternionic coordinates
r
(j)
i
= n
i
n
1
j
. i = ,. , = 0. . . . . n.
on l
j
and H-valued 1-forms
(12.1.25) dr
(j)
i
= (dn
i
r
(j)
i
dn
j
)n
1
j
. i = ,. , = 0. . . . . n.
At each r HP
n
the forms dr
(j)
i
dene an isomorphism T
x
HP
n
H
n
of quater-
nionic vector spaces, and thus a local section
(j)
(l
j
. 1

(HP
n
)) of the principal
coframe bundle 1

(HP
n
) HP
n
. Let
(k)
(l
k
. 1

(HP
n
)) be another such local
section and consider l
k
l
j
. An easy computation shows that at any r l
k
l
j
(12.1.26) dr
(k)
i
= (dr
(j)
i
r
(k)
i
dr
(j)
k
)[r
(j)
k
]
1
. i = / .
Note that by convention r
(j)
j
= 1. dr
(j)
j
= 0. The equations 12.1.26 imply that
pointwise in l
k
l
j
(12.1.27)
(k)
=
(j)
c .
where c has its values in G1(1. H) and in G1(n. H) G1(4n. R). The group
G1(n. H) G1(1. H) does not act eectively, but there is an eective action of the
quotient group (G1(n. H) G1(1. H))R

= G1(n. H)oj(1). Thus the structure

group of HP
n
reduces to G1(n. H)oj(1). We are now ready to give HP
n
a Rie-
mannian metric which is induced by the at metric on H
n+1
` 0. We can write
the quaternionic Hermitian form in homogeneous coordinates as
(12.1.28) o
0

1
i
1

2
i
2

3
i
3
=
4
[u[
2

j
dn
j
dn
j

4
[u[
4

j,k
(n
j
dn
j
)(dn
k
n
k
) .
Note that the above equation denes the metric on HP
n
as well as the three local
2-forms
1
.
2
.
3
which are local sections of a 3-dimensional vector subbundle
O
2
T

HP
n
. Using the language of H-valued forms we can introduce
(12.1.29) =
4
[u[
2

dn

dn

4
[u[
4

(n

dn

) (dn

) .
with = . so that is purely imaginary. The constant c is equal to the so-
called quaternionic sectional curvature which generalizes the notion of holomorphic
sectional curvature in complex geometry. The quaternionic Kahler 4-form is then
given by
(12.1.30) = = .
It is real and closed. We have the following
428 12. QUATERNIONIC K

AHLER AND HYPERK

AHLER MANIFOLDS
Theorem 12.1.7: The 4-form is parallel. When n 1 the holonomy group
Hol(o
0
) oj(n)oj(1). When n = 1 HP
1
o
4
and the metric o
0
is simply the
metric of constant sectional curvature on o
4
which is self-dual and Einstein.
12.2. Quaternionic Kahler Metrics
Let ` be a smooth 4n-dimensional manifold (n 1). Recall from Example
1.4.18 that ` is almost quaternionic if there is a 3-dimensional subbundle O
End(T`) with the property that at each point r ` there is a basis of local
sections 1
1
. 1
2
. 1
3
of O satisfying the quaternion algebra, that is
(12.2.1) 1
i
1
j
=
ij
1l +c
ijk
1
k
.
This denition is equivalent to ` admitting a G-structure with G = G1(n. H)oj(1).
Note that any oriented 4-manifold admits such a structure, but in higher dimen-
sions there are obstructions to admitting an almost quaternionic structure as we
now describe.
Suppose now that ` carries a Riemannian metric o adapted to the quaternion
structure in the sense that each point on ` has a neighborhood such that any local
section 1 of O is a local isometry, i.e.,
(12.2.2) o(1A. 1Y ) = o(A. Y ) .
for any local vector elds A. Y. Adapted metrics always exists and the resulting
triple (`. O. o) is called an almost quaternionic Hermitian manifold, giving a fur-
ther reduction of the structure group to oj(n)oj(1). Given an adapted metric we
obtain a subbundle O

2
T

` which associates to each local section 1 of O the

local non-degenerate 2-form dened by
(12.2.3)
I
(\. \) = o(1A. Y ). A. Y T
x
` .
The oj(n)oj(1)-structure is a principal oj(n)oj(1)-bundle 1 over ` and as
such it can be regarded as an element of the cohomology group H
1
(`. oj(n)oj(1))
with coecients in the sheaf oj(n)oj(1) of smooth oj(n)oj(1)-valued functions.
The short exact sequence
(12.2.4) 0 Z
2
oj(n) oj(1) oj(n)oj(1) 0
gives rise to the homomorphism
(12.2.5) : H
1
(`. oj(n)oj(1)) H
2
(`. Z
2
) .
We have
Denition 12.2.1: Let c = (1) H
2
(`. Z
2
). Then c is called the Marchiafava-
Romani class of (`. O. o).
The Marchiafava-Romani class was introduced in [MR75] and it is the obstruc-
tion to lifting 1 to the oj(n) oj(1) bundle. When n = 1 the sequence (12.2.4)
becomes
(12.2.6) 0 Z
2
ojin(4) oO(4) 0
and it follows that c equals the second Stiefel-Whitney class n
2
(`). For n 1 we
can identify c = n
2
(O) with the second Stiefel-Whitney class of the vector bundle
O. Furthermore, we get [Sal82, MR75]
Proposition 12.2.2: Let (`
4n
. O. o) be an almost quaternionic Hermitian mani-
fold. Then n
2
(`) nc(2).
12.2. QUATERNIONIC K

AHLER METRICS 429

In particular, the Marchiafava-Romani class c is the second Stiefel-Whitney
class of ` if its dimension is 4(8). In complementary dimensions we get
Corollary 12.2.3: Any almost quaternionic manifold ` of dimension 0(8) is spin.
Example 12.2.4: Consider the complex projective spaces CP
4n
of real dimension
8n. The rst Chern class c
1
(CP
4n
) = (4n + 1) where is a positive generator of
H
2
(CP
4n
. Z). Since n
2
is the mod 2 reduction of c
1
, the manifold CP
4n
is not spin,
and so by Corollary 12.2.3 CP
4n
cannot admit an almost quaternionic structure.
The full obstruction theory for oj(n)oj(1)-structures (even in the 8-dimensional
case) is subtle and not completely understood. See, for example, the article by

Cadek and Vanzura [

CV98] where they prove

Theorem 12.2.5: Let ` be a compact oriented 8-manifold. If the conditions
n
2
(`) = n
6
(`) = 0. 4j
2
(`) j
2
1
(`) 8c(`) = 0. j
2
1
(`) + 4c(`) 0(16)
hold, then ` admits an almost quaternionic structure.
Not all these conditions are necessary, however. We know that the vanishing
of n
2
is necessary, and

Cadek and Vanzura remarked that the middle condition is
necessary. However, they also noticed that Borel and Hirzebruch [BH58] had com-
puted the mod 2 cohomology ring of the quaternionic Kahler manifold G
2
oO(4)
showing that n
6
= 0 which implies that the vanishing of n
6
is not necessary.
Suppose 1
1
. 1
2
. 1
3
are locally dened smooth sections of Owhich satisfy 12.2.1
at each point. Then these form a local orthonormal frame for O with respect to the
standard metric '. 1` =
1
2n
Tr(
t
1) on End(T`). Let
i

i=1,2,3
be the basis of
2-forms corresponding under 12.2.3. The associated exterior 4-form
(12.2.7) =
3

i=1

i

i
is invariant under a change of frame and thus globally dened on `. It is non-
degenerate in the sense that
n
is nowhere vanishing on `. The group oj(n)oj(1)
is precisely the stabilizer of the form in G1(4n. R). The form is called the
fundamental 4-form of the almost quaternion Hermitian structure (`. O. . o).
Denition 12.2.6: An almost quaternionic structure (`
4n
. O) with n 1 is 1-
integrable if ` admits a torsion-free connection
C
preserving the quaternionic
structure O. In such a case (`
4n
. O) is called a quaternionic structure on `
4n
,
and if it has an adapted Riemannian metric o. the triple (`
4n
. O. o) is called a
quaternionic Hermitian manifold.
The case of real dimension 4. that is n = 1. is given in Denition 12.2.12 below.
Recall that the connection
C
is not unique. The obstruction to 1-integrability
has been studied by Salamon in [Sal86]. In the 4-dimensional case, there is no
obstruction as G = G1(1. H)oj(1) = R
+
oO(4) so that G-structure is equiva-
lent to a choice of orientation and conformal class. In particular, the Levi-Civita
connection of any compatible metric preserves the G-structure and has no torsion.
But in higher dimensions, there are non-trivial obstructions.
Here we shall be interested in a very special class of quaternionic Hermitian
manifolds, namely the case where a torsion-free quaternionic connection
C
is also
a metric connection. In this case it must be the Levi-Civita connection.
430 12. QUATERNIONIC K

AHLER AND HYPERK

AHLER MANIFOLDS
Denition 12.2.7: An almost quaternionic Hermitian manifold (`
4n
. O. . o) of
quaternionic dimension n 1 is called quaternionic Kahler (QK) if
C
coin-
cides with the Levi-Civita connection, or alternatively, if the holonomy group Hol(o)
lies in oj(n)oj(1).
We also refer to such a manifold as one with a quaternionic Kahler structure.
We can easily see the holonomy denition to be equivalent to the following
Proposition 12.2.8: An almost quaternionic Hermitian manifold (`
4n
. O. . o).
n 1. is quaternionic Kahler if = 0. where denotes the Levi-Civita connec-
tion of o. In particular, an almost quaternionic Hermitian manifold (`
4n
. O. . o).
n 1. is quaternionic Kahler if it admits a parallel 4-form which is in the same
G1(4n. R)-orbit as at each point r `.
The hypothesis = 0 clearly implies that d = 0. Surprisingly, the following
theorem was proved by Swann [Swa89]:
Theorem 12.2.9: An almost quaternionic Hermitian (`
4n
. O. . o) of quater-
nionic dimension n 2 whose fundamental 4-form is closed is quaternionic
Kahler.
The geometry of almost quaternionic Hermitian 8-manifolds is somewhat richer
as there are examples of such spaces for which the fundamental 4-form is closed
but not parallel. Swann showed [Swa91] that
Theorem 12.2.10: An almost quaternionic Hermitian 8-manifold is quaternionic
Kahler if and only if the fundamental 4-form is closed and the algebraic ideal
generated by the subbundle O

2
T

` is a dierential ideal.
We now investigate some curvature properties of quaternionic Kahler manifolds.
Let
1
.
2
.
3
be a local orthonormal frame eld for O

2
T

`. If is parallel
we get
3

i=1
(
i
)
i
= 0
from which it follows that
(12.2.8)
i
=
3

j=1

ij

j
.
where the
ij
are 1-forms which satisfy
(12.2.9)
ij

ji
i. , = 1. 2. 3 .
This means in particular that the subspace (O

) (
2
T

`) is preserved by
the Levi-Civita connection. The equations (12.2.8) were considered by Ishihara
[Ish74]. The matrix
(12.2.10) =

0 o
12
o
13
o
12
0 o
23
o
13
o
23
0

is the connection 1-form with respect to the local frame eld

1
.
2
.
3
. The cur-
vature of this induced connection represents a component of the Riemann curvature
tensor 1 and is given by
1 = d .
12.2. QUATERNIONIC K

AHLER METRICS 431

Using the facts that d
i
=

3
j=1

ij

j
and d
2

i
= 0, one deduces that
(12.2.11) 1
ij
= d
ij

il

lj
=

k
c
ijk

k
for some constant . Since O

is an oriented 3-dimensional bundle, there is a canon-

ical identication SkewEnd(O

)

= O

via the cross-product. Using this identica-

tion we can consider 1 as a map 1 :
2
T` SkewEnd(O

)

= O

2
T`
and, as such, equation 12.2.11 simply states that 1 =
C
. where
C
denotes
pointwise orthogonal projection
C
:
2
T` O

. The full Riemann curvature

tensor 1 of QK manifold viewed as a symmetric endomorphism 1 :
2
T`

2
T` (curvature operator), has the property that
(12.2.12) 1 [
C
= Id
C
.
where is a positive multiple of the scalar curvature : on `.
We will now use Equation 12.2.12 to extend our denition of quaternionic
Kahler manifolds to 4-dimensional spaces. Recall that the problem in dimension
four is that the structure group oj(1)oj(1) is isomorphic to the orthogonal group
oO(4) which just describes generic four dimensional oriented Riemannian geometry.
So the problem is caused by a certain low dimensional isomorphism of Lie groups.
Remarkably this same isomorphism of Lie groups provides us with the solution as
well. Now the Lie algebra so(n) and
2
(R
n
) are isomorphic as oO(n) modules. So
in dimension four we have a splitting so(4) = su(2) su(2) giving rise to a splitting
(12.2.13)
2
T` =
2
+

.
where
2

are precisely the eigenspaces of the Hodge star operator . The bundles

2
+
and
2

are known as the bundles of self-dual and anti-self-dual 2-forms, respec-

tively. Reversing orientation interchanges the self-dual and anti-self-dual 2-forms.
(Note also that in dimension 4, the condition that = 0 is trivially satised since
is the volume form). Fixing an orientation and identifying O

with
2

we have
Denition 12.2.11: An oriented Riemannian 4-manifold (`. o) is called quater-
nionic Kahler if condition 12.2.12 holds.
With the reverse orientation Q

is identied with
2
+
. Relative to the decom-
position (12.2.13), the curvature operator 1 can be represented by the matrix
(12.2.14) 1 =

\
+
+
s
12
1l Ric
0
Ric
0
\

+
s
12
1l

.
where \

are the self-dual and anti-self-dual Weyl curvatures, Ric

0
is the trace-free
part of the Ricci curvature, and : is the scalar curvature. So Equation (12.2.12)
implies that Denition 12.2.11 is equivalent to
Denition 12.2.12: A 4-dimensional oriented Riemannian manifold (`. o) is
quaternionic Kahler if and only if it is self-dual (i.e., \

= 0) or anti-self-
dual (i.e., \
+
= 0) and Einstein (i.e., Ric
0
= 0). More generally, and oriented
4-manifold (`. o) is quaternionic if \

= 0 or \
+
= 0.
Remark 12.2.1: It follows from Denition 4.2.15 that Denitions 12.2.6, 12.2.7
and 12.2.12 work equally well in the case of orbifolds. Thus, it makes perfect sense
to talk about quaternionic or quaternionic Kahler orbifolds. These will play an
important role in Sections 12.4 and 12.5 as well as Chapter 13.
432 12. QUATERNIONIC K

AHLER AND HYPERK

AHLER MANIFOLDS
Remark 12.2.2: Recall that changing the orientation of a quaternionic 4-manifold
interchanges \
+
and \

. We shall stick with the more usual convention by saying

that a QK 4-manifold or orbifold with non-zero scalar curvature is self-dual and
Einstein, thus xing the orientation. This becomes particularly important when
one adds a complex structure, since a complex structure xes the orientation. The
complex manifold CP
2
is self-dual and Einstein, not anti-self-dual; whereas, a 13
surface is anti-self-dual and Einstein. Neither of these manifolds are complex if one
reverses the orientation. Indeed, there are very few compact complex surfaces that
are complex with respect to the reverse orientation [Kot97].
There are at least two more justications for adopting the Denition 12.2.12.
One is the theory of quaternionic submanifolds of QK manifolds. ` is called
a quaternionic submanifold if for each r . T
x
is an H

-submodule of T
x
`.
Marchiafava observed that a 4-dimensional submanifold of a QK manifold is neces-
sarily self-dual and Einstein. The other justication comes in the theory of quater-
nionic Kahler reduction which will be discussed in a later section.
We now return to the general case. Following Salamon [Sal82] we decompose
the Riemannian curvature on a QK manifold in terms of its irreducible pieces
under the group oj(n)oj(1). We can write the cotangent bundle as T

` = 1H
where 1 and H are locally dened vector bundles on ` that transform as the
standard representations of oj(n) and oj(1). respectively. Although 1 and H
are not globally dened bundles their symmetric products o
2
(1). o
2
(H) and anti-
symmetric products
2
1.
2
H are. Note that o
2
(1) and o
2
(H) transform as
the adjoint representations of oj(n) and oj(1). respectively, so the vector bundle
= o
2
(1) +o
2
(H) transforms as the adjoint representation of oj(n)oj(1). So the
(4. 0) Riemannian curvature tensor 1 can be thought of as a section of
o
2
() = o
2
(o
2
(1)) +o(o
2
(1) o
2
(H)) +o
2
(o
2
(H))
satisfying the rst Bianchi identities.
We already mentioned that the quaternionic projective space is quite special as
it is the only example of a compact quaternionic Kahler manifold which admits an
integrable oj(n)oj(1)-structure. The curvature tensor of the canonical symmetric
metric on HP
n
plays a key role in the more general setting. The following result
which is due to Alekseevsky [Ale68] is presented in the form of Salamon [Sal82]
to which we refer for a proof. This proof can also be found in [Bes87].
Theorem 12.2.13: Let (`
4n
. O. . o) be a quaternionic Kahler manifold. The
Riemann curvature tensor can be written as
1 = :1
1
+1
0
.
where : is the scalar curvature, 1
1
is the curvature tensor of quaternionic projective
space HP
n
. and 1
0
is a section of o
2
(o
2
(1)).
The section 1
0
behaves like the curvature tensor of a hyperkahler manifold. In
particular, 1
0
has zero Ricci curvature and 1
1
has a traceless Ricci curvature, so
we get the following result due to Berger [Ber66]
Corollary 12.2.14: Any QK manifold is Einstein. A QK manifold with vanish-
ing scalar curvature : is locally hyperkahler, i.e., the restricted holonomy group
Hol
0
(o) oj(n).
Theorem 12.2.13 and Corollary 12.2.14 are of fundamental importance to any
further study of quaternionic Kahler manifolds. They imply that QK geometry
12.2. QUATERNIONIC K

AHLER METRICS 433

splits into three cases, positive, negative, and null or the hyperkahler case. We will
discuss some basic properties of hyperkahler metrics in the last four sections of this
chapter, so until then we assume that scalar curvature is not zero. Furthermore, we
almost exclusively discuss the positive QK case, as it is this case that has a strong
connection with the main theme of this book. We end this section by dening three
important bundles that will play an important role in the next chapter.
Denition 12.2.15: Let (`. O) be an almost quaternionic manifold. Let o(`)
be the oO(3)-principal bundle associated to O. This principal bundle is called the
Konishi bundle of `. We dene the following associated bundles o
SO(3)
1
(i) |(`) = o(`)
SO(3)
1. where 1 = H

Z
2
.
(ii) Z(`) = o(`)
SO(3)
1. where 1 = o
2
is the unit sphere in O.
The bundles |(`). Z(`) are called the Swann bundle, and the twistor space
of `. respectively.
The bundle o(`) was rst described in Konishi [Kon75] for quaternionic
Kahler manifolds, and the bundle |(`) by Swann [Swa91], again for QK mani-
folds. The twistor space Z(`) takes its name from Penroses twistor theory, cf. the
two volume set [PR87, PR88] and references therein. The twistor space construc-
tion used here has its origins in Penroses non-linear graviton [Pen76]. It plays
an important role in understanding the geometry of both quaternionic and quater-
nionic Kahler manifolds. Here is why. Since the bundle Z is just the unit sphere in
O. each point . Z represents an almost complex structure 1() =
1
1
1
+
2
1
2
+
3
1
3
as in Section 12.1. Thus, a smooth section : of Z over an open set l ` is an
almost complex structure on l. So we can think of the twistor space as a bundle
of almost complex structures on `. Global sections do not exist generally, so ` is
not almost complex. However, if 1 denotes the vertical subbundle of TZ consisting
of tangent vectors to the bres 1 = o
2
. a choice of quaternionic connection
C
determines an equivariant splitting TZ = 1 H. So we obtain an almost com-
plex structure on Z by adding the standard complex structure 1
0
on o
2
to 1()
at each point . = ((.). ) making Z an almost complex manifold. Moreover, the
antipodal map on the bres induces an anti-holomorphic involution : ZZ.
Then the main result concerning twistor spaces is the following theorem due to
Atiyah, Hitchin, and Singer [AHS78] in quaternionic dimension one, and Salamon
[Sal84] for quaternionic dimension greater than one which encodes the quaternionic
geometry of ` in the complex geometry of Z.
Theorem 12.2.16: Let ` be a quaternionic manifold (orbifold). Then the twistor
space Z(`) is a complex manifold (orbifold). Moreover, the bres of : Z`
are rational curves whose normal bundle is 2nO(1) and Z has a free anti-holomorphic
involution that is the antipodal map on the bres.
For four dimensional manifolds (quaternionic dimension one) the converse is
true, that is, if the induced almost complex structure on Z is integrable, then
the conformal structure is self-dual (\

= 0). However, in higher dimension the

integrability of Z(`) only implies the vanishing of a piece of the torsion of
C
.
We end this section with a brief discussion of some quaternionic manifolds
in dimension 4. The manifolds o
4
= HP
1
. CP
2
. 13. T
4
are all well-known to be
quaternionic. In fact, o
4
is self-dual Einstein with one orientation and anti-self-dual
Einstein with the other, CP
2
is self-dual Einstein, and 13 is anti-self-dual Einstein
with the standard orientation induced by the complex structure. Of course T
4
.
434 12. QUATERNIONIC K

AHLER AND HYPERK

AHLER MANIFOLDS
being at, is self-dual, anti-self-dual and Einstein with either orientation. There
has been much work on self-dual and anti-self-dual structures on 4-manifolds over
the years, and it is not our purpose here to describe what is known. Indeed many
examples of such manifolds as well as orbifolds will make their appearance either
explicitly or implicitly in the present monograph. Usually they occur enjoying some
other property, such as being Einstein, or Kahler. We mention here only some
results of a more general nature. First, there is the existence of self-dual structures
on the connected sums /CP
2
for / 1 [Poo86, DF89, Flo91, LeB91b, PP95,
Joy95] as well as other simply connected 4-manifolds that are neither Einstein nor
complex. Second, Taubes [Tau92] has proven a type of stability theorem that says
that given any smooth oriented 4-manifold ` then `#/CP
2
admits a self-dual
conformal structure for / large enough.
12.3. Positive Quaternionic Kahler Manifolds and Symmetries
All known complete positive quaternionic Kahler manifolds are symmetric spaces
(see Conjecture 12.3.7 below), and Salamon [Sal82] showed that through dimension
16 any positive QK manifold must have a fairly large isometry group. Moreover,
a bit earlier Alekseevsky proved that all homogeneous positive QK manifolds must
be symmetric [Ale75]. These spaces had been classied by Wolf [Wol65] and they
are often called Wolf spaces. There is precisely one for each simple Lie algebra and
we have
Theorem 12.3.1: Let ` be a compact homogeneous positive QK manifold Then
` = GH is precisely one of the following:
oj(n + 1)
oj(n) oj(1)
.
ol(:)
o

l(:2)l(2)
.
oO(/)
oO(/ 4)oO(4)
.
G
2
oO(4)
.
1
4
oj(3)oj(1)
.
1
6
ol(6)oj(1)
.
1
7
Spin(12)oj(1)
.
1
8
1
7
oj(1)
.
Here n 0. oj(0) denotes the trivial group, : 3, and / 7. In particular, each
such ` is a symmetric space and, there is one-to-one correspondence between the
simple Lie algebras and positive homogeneous QK manifolds.
We remark that the integral cohomology group H
2
(`. Z) vanishes for ` =
HP
n
and it is Z for the complex Grassmannian ` = Gr
2
(C
n+2
). In all other cases
we have H
2
(`. Z) = Z
2
. The main results concerning positive QK manifolds are
due to LeBrun and Salamon. Before embarking into a description of their work we
give an innitesimal rigidity theorem due to LeBrun [LeB88].
Theorem 12.3.2: Let (`. O. . o) be a compact positive QK manifold. If o
t
is a
family of positive QK metrics of xed volume depending smoothly on R such that
o
0
= o. Then there exists a family of dieomorphisms 1
t
: ` ` depending
smoothly on t such that o
t
= 1

t
o.
This result is far from true for hyperkahler or negative QK manifolds. Mod-
uli in the hyperkahler case is well-known, and LeBrun has shown that the moduli
space of complete negative QK metrics on H
n
is innite-dimensional [LeB91c].
LeBrun and Salamon [LS94, LeB93] have strengthen LeBruns innitesimal rigid-
ity Theorem 12.3.2 to a strong rigidity result. They give two theorems, the rst is
a niteness theorem, and second severely restricts the topological type of positive
12.3. POSITIVE QUATERNIONIC K

AHLER MANIFOLDS AND SYMMETRIES 435

QK manifolds. We should also mention that the rigidity results given below break
down entirely in the case of compact positive QK orbifolds.
Theorem 12.3.3: For each positive integer n there are up to isometries and rescal-
ings only nitely many compact positive quaternionic Kahler manifolds of dimen-
sion 4n.
outline of Proof. The proof of this theorem relies heavily on Mori theory
applied to the twistor space Z(`). In the positive QK case Theorem 12.2.16 was
strengthened in [Sal82]. We state this as a lemma together with another result
from [LS94].
Lemma 12.3.4: Let `
4n+3
be a positive quaternionic Kahler manifold. Then its
twistor space Z(`) is a Fano manifold with a positive Kahler-Einstein metric and
a complex contact structure. Moreover, two positive quaternionic Kahler structures
are homothetic if and only if their twistor spaces are biholomorphic.
Since Z(`) has a complex contact structure c
1
(`) is divisible by n + 1. Then by
contracting extremal rays Wisniewski [Wis91] shows that /
2
(Z(`)) = 1 with the
exception of three cases. Only one of these cases admits a complex contact struc-
ture, namely the ag variety P(T

CP
n+1
) which is the twistor space of the complex
Grassmannian Gr
2
(C
n+2
). So we can conclude from Lemma 12.3.4 that /
2
(`) = 0
unless ` = Gr
2
(C
n+2
) with its symmetric space metric. In the remaining cases we
are dealing with Fano manifolds with Picard number one. We can make use of the
rational connectedness theorem which imply that there are only a nite number of
deformation types of smooth Fano varieties. This was proven for Picard number one
in [Cam91, Nad91], and more generally without the Picard number restriction
in [Cam92, KMM92]. Then in line with Theorem 12.3.2 LeBrun and Salamon
show that there is only a nite number of Fano contact manifolds up to biholo-
morphism. We refer to [LS94, LeB93] for details. Then the second statement of
Lemma 12.3.4 implies that there are only a nite number of QK manifolds up to
homotheties.
This proof says a lot about the topology of compact positive QK manifolds.
Here we collect the results of [LS94] together with what was known earlier about
the topology of positive QK manifolds.
Theorem 12.3.5: Let ` be a compact positive QK manifold. Then
(i)
1
(`) = 0;
(ii)
2
(`) =

0 if ` is isometric to HP
n
,
Z if ` is isometric to Gr
2
(C
n+2
),
nite containing Z
2
otherwise;
(iii) /
2k+1
(`) = 0 for all / 0;
(iv) /
4i
(`) 0 for 0 i n;
(v) /
2i
(`) /
2i4
(`) 0 for 2 i n;
(vi)

n1
r=0
[6:(n 1 :) (n 1)(n 3)]/
2r
(`) =
1
2
n(n 1)/
2n
(`).
Proof. To prove (i) we see that by Lemma 12.3.4 Z(`) is Fano so by Theorem
3.6.9 it is simply connected. Then (i) follows by the long exact homotopy sequence
applied to the bration o
2
Z(`)`. For (ii) we recall in the proof of Theo-
rem 12.3.3 /
2
(`) = 0 unless ` = Gr
2
(C
n+2
) in which case
2
(`) = Z. But this to-
gether with (i) and universal coecients imply that if /
2
(`) = 0 then H
2
(`. Z
2
) is
436 12. QUATERNIONIC K

AHLER AND HYPERK

AHLER MANIFOLDS
the 2-torsion of H
2
(`. Z). and this is non-vanishing when the Marchiafava-Romani
class c is non-vanishing. So the remainder of (ii) then follows from the following
result of Salamon [Sal82]:
Lemma 12.3.6: Let (`. O. o) be a compact positive QK manifold with vanishing
Marchiafava-Romani class c. Then (`. o) (HP
n
. o
0
) with its canonical symmetric
metric o
0
.
Part (iii) which is due to Salamon [Sal82] is a consequence of the fact that the
twistor space Z(`) of a QK manifold has only (j. j)-type cohomology which in
turn is a consequence of the Kodaira-Nakano Vanishing Theorem 3.5.8 and a gen-
eralization due to Akizuki and Nagano [AN54]. Part (iv) follows immediately from
the non-degeneracy of the closed 4-form of Equation (12.2.7). Part (v) follows
from the quaternionic version of the Lefschetz theory [Kra65, Bon82] as described
for example in Section 3.3. (vi) follows from index theory computations which we
refer to Section 5 of [LS94].
Remarks 12.3.1: Parts (iv) and (v) hold for any compact QK manifold, not only
positive ones. However, in the positive case we shall see in Proposition 13.5.5
below that the numbers
2i
= /
2i
/
2i4
are precisely the even Betti numbers of
the principal oO(3)-bundle associated to the quaternionic bundle O.
In the absence of any counterexamples, Theorems 12.3.3 and 12.3.5 strongly
points towards
Conjecture 12.3.7: All compact positive QK manifolds are symmetric.
The above conjecture was rst formulated in [LS94] and we will refer to it
as the LeBrun-Salamon Conjecture. Beyond Theorems 12.3.3 and 12.3.5 there are
several other results showing the conjecture to be true in some special cases. We
shall collect all these results in the following
Theorem 12.3.8: Let (`
4n
. O. . o) be a compact positive QK manifold. Then
` is a symmetric space if
(i) n 3.
(ii) n = 4 and /
4
= 1.
Proof. The statement in (i) dates back to the Hitchins proof that all compact
self-dual and Einstein manifolds of positive scalar curvature must be isometric to
either o
4
with the standard constant curvature metric or CP
2
with the Fubini-
Study metric [Hit81] (see also [FK82, Bes87]). For n = 2 the result was proved
by Poon and Salamon [PS91]. The proof was greatly simplied in [LS94] using
the rigidity results of Theorem 12.3.5. The n = 3 case is a recent result of Herrera
and Herrera [HH02a, HH02b]. Their proof uses an old result which estimates
the size of the isometry group of ` in lower dimension [Sal82]. The dimension
of the isometry group of a positive QK 12-manifold must be at least 6 and the
dimension of the isometry group of a positive QK 16-manifold must be at least 8.
In particular, when n = 3 the manifold ` admits an isometric circle action. Using
some deep results concerning

(`) genus of non-spin manifolds with nite
2
(`)
and smooth circle actions Herrera and Herrera prove that
Lemma 12.3.9: Let ` be a positive QK 12-manifold which is not Gr
2
(C
5
). Then

(`) = 0.
12.3. POSITIVE QUATERNIONIC K

AHLER MANIFOLDS AND SYMMETRIES 437

The result is then a consequence of the vanishing of

(`) and Theorem 12.3.5.
The result in (ii) follows from the estimate on the dimension of the isometry group
in this case and Betti number constraints of Theorem 12.3.5 [GS96].
Remark 12.3.1: The argument of [HH02a] does not work in 16-dimensional case
because all QK manifolds of quaternionic dimension 4 are automatically spin. Nev-
ertheless, the estimate on the size of the isometry group together with all the known
results can most likely be used to construct a proof of the LeBrun-Salamon Conjec-
ture in this case. However, as pointed out by Salamon in [Sal99], the biggest gap
in any potential geometric proof of this conjecture is the conundrum of whether a
QK manifold of quaternionic dimension n 4 has any non-trivial Killing vector
elds.
There is another approach to the LeBrun-Salamon Conjecture which proceeds
via the algebraic geometry of the twistor space Z(`) and uses Lemma 12.3.4.
The following, apparently stronger, conjecture was suggested by Beauville [Bea05,
Bea98]
Conjecture 12.3.10: Any compact Fano manifold with a complex contact struc-
ture is homogeneous.
This, of course implies the LeBrun-Salamon Conjecture. Several years ago there
were some attempts to use algebraic geometry to prove this result. Wisniewski
even briey claimed the proof of the conjecture but later Campana found a gap in
Wisniewskis argument. Campana briey claimed to have bridged that gap but later
also withdrew the claim. Hence, as of the time of writing this monograph, both the
Beauville Conjecture 12.3.10 and LeBrun-Salamon Conjecture 12.3.7 remain open.
Let rk(`) be the symmetry rank of ` dened as the rank of its isometry group
Isom(`. o), i.e., the dimension of the maximal Abelian subgroup in Isom(`. o).
Bielawski [Bie99] proved that a positive QK manifold of quaternionic dimension
n with rk(`) n + 1 is isometric to HP
n
or to the Grassmannian G:
2
(C
n+2
).
Recently Fang proved several rigidity theorems for positive quaternionic Kahler
manifolds in terms its symmetry rank [Fan04]. Fangs result slightly enhances
Bielawskis theorem.
Theorem 12.3.11: Let (`
4n
. O. . o) be a compact positive quaternionic Kahler
manifold. Then the isometry group Isom(`. o) has rank at most (n + 1), and `
is isometric to HP
n
or G:
2
(C
n+2
) if rk(`) n 2 and n 10.
This theorem is quite interesting and apparently rather deep. It follows from
several dierent results. First recall that a quaternionic submanifold is one that pre-
serves the quaternionic structure. It is a well-known result of Gray that [Gra69a]
Proposition 12.3.12: Any quaternionic submanifold in a quaternionic Kahler
manifold is totally geodesic and quaternionic Kahler.
In [Fan04] Fang proves the following rigidity results for positive QK manifold
Theorem 12.3.13: Let (`
4n
. O. . o) be a positive QK manifold. Assume 1 =
(1
1
. 1
2
) : ` `. where =
1

2
and 1
i
:
i
` are quaternionic
immersions of compact quaternionic Kahler manifolds of dimensions 4n
i
, i = 1. 2.
Let be the diagonal of ` ` and set : = n
1
+n
2
. Then
(i) If : n. then 1
1
() is non-empty.
(ii) If : n + 1. then 1
1
() is connected.
438 12. QUATERNIONIC K

AHLER AND HYPERK

AHLER MANIFOLDS
(iii) If 1 is an embedding, then for i :n there is a natural isomorphism,

i
(
1
.
1

2
)
i
(`.
2
) and a surjection for i = :n + 1.
Remark 12.3.2: The study of homogeneous negative QK manifold is more deli-
cate. There are of course the non-compact duals of the Wolf spaces. Alekseevsky
showed that there are also non-symmetric homogeneous examples. He obtained
a classication of such spaces under the assumption that the symmetry group is
completely solvable[Ale75]. We will not describe these spaces here referring the
interested reader to an extensive review on this subject by Cortes [Cor00]. All
these are typically called Alekseevskian spaces. Much later de Wit and Van Proyen
discovered a gap in Alekseevskys classication [dWVP92] while considering some
supersymmetric -model coupled to supergravity. They lled in the gap and also
claimed that there should be no other homogeneous examples. Inspired by this
work Cortes [Cor96] provided a Lie algebraic proof lling the gap in Alekseevskys
original paper. A proof that all negative QK manifolds are the known Alekseevskian
spaces is, however, still lacking.
12.4. Quaternionic Kahler Reduction
In this section we describe an of the symmetry reduction method described in
Section 8.4 for QK manifolds. Just as in the case of hyperkahler quotients which
are introduced later in Section 12.8, the ideas originated in physics of supersym-
metric eld theories. In 1983 Witten and Bagger observed that matter coupled
to 4-dimensional supergravity theory with = 2 supersymmetries requires the
scalar elds of the coupling to be local coordinates on a negative scalar curvature
QK manifold [BW83]. Later more general Lagrangians of such theories were con-
structed and studied. In particular, some elements of the symmetry reduction can
be found in [dWLP
+
84, dWLVP85]. However, mathematical formulation of the
theory of quaternionic Kahler quotients and its application to the case of positive
QK manifolds was developed later in [Gal87a, GL88].
To begin we consider the spaces
p
(O

) (
p
T

` O

) of smooth exterior
j-forms on ` with values in the bundle O

. The connection given on O

induces
a de Rham sequence
(12.4.1)
0
(O

)
d

=

1
(O

)
d

2
(O

)
d

such that
(12.4.2) d

(1) = 1(1)
for 1
0
(O

).
Consider now the Lie group
Aut(`. O. . o) o Isom(`. o) : o

=
and its Lie algebra
aut(`. O. . o) \ isom(`. o) :
V
= 0
which is a Lie subalgebra of the Lie algebra isom(`. o) of Killing vector elds on
`. We have the following immediate consequence
Proposition 12.4.1: Let (`. O. . o) be a QK manifold of non-zero scalar cur-
vature. Then aut(`
4n
. O. . o) = isom(`
4n
. o). It follows that any one parameter
subgroup H Isom(`
4n
. o) is also a subgroup of Aut(`
4n
. O. . o).
12.4. QUATERNIONIC K

AHLER REDUCTION 439

Proof. When ` is symmetric all statements follow by inspection. When `
is not locally symmetric the holonomy Lie algebra hol = sp(n) sp(1). Since ` is
irreducible, by a theorem of Kostant [Kos55] any Killing vector eld normalizes the
holonomy algebra and in particular the sp(1)-factor which denes the quaternionic
structure O. Hence isom(`. o) normalizes Q and therefore any Killing vector eld
\ preserves . The rest follows from the fact that both groups are compact Lie
groups.
The full isometry group may contain discrete isometries which do not lie on
any one-parameter subgroup and these may not preserve the quaternionic 4-from
. To each \ isom(`. o) we associate the O

-valued 1-form

V

1
(O

)
dened in terms of a local frame
1
.
2
.
3
by
(12.4.3)
V

i
(\
i
)
i
.
Clearly
V
remains invariant under local change of frame eld (i.e., under local
gauge transformations). We have [GL88]
Theorem 12.4.2: Assume that the scalar curvature of (`. O. . o) is not zero.
Then to each \ isom(`. o) there corresponds a unique section
0
(O

) such
that
(12.4.4) =
V
.
In fact, under the canonical bundle isometry : SkewEnd(O

) O

. is given
explicitly by the formula
(12.4.5) =
1

(
V

V
) .
where is the constant positive multiple of the scalar curvature dened by 12.2.12.
We observe now that by the uniqueness in Theorem 12.4.2, the map \
transforms naturally under the group of automorphisms, specically for o
Aut(`. O. . o) and \ isom(`. o) we have
(12.4.6)
g(V )
= o

() .
where o

(r) = o

(o
1
(r))

and where o denotes the map induced by o on the

bundle O

2
T`. Note also that o

\ = Ad
g
(\ ). Hence, 12.4.6 means that the
diagram
(12.4.7)
isom(`. o)
Ad
g
isom(`. o)

0
(O

)
g

0
(O

) .
commutes.
Suppose now that G Isom(`. o) is a compact connected Lie subgroup with
corresponding Lie algebra g.
440 12. QUATERNIONIC K

AHLER AND HYPERK

AHLER MANIFOLDS
Denition 12.4.3: The moment map associated to G is the section of the
bundle g

O

= Hom(g. O

) whose value at a point r is the homomorphism

\ (r).
From the equivariance above we see immediately that the moment map is G-
equivariant. Since the action of G in the bundle g

is linear on the bers, it

preserves the zero section. Consequently the set
(12.4.8) =
1
(0) = r ` : (r) = 0
is G-invariant. We have the following reduction theorem due to Galicki and Lawson
[GL88].
Theorem 12.4.4: Let (`. O. . o) be a quaternionic Kahler manifold with non-
zero scalar curvature. Let G Isom(`. o) be a compact connected subgroup with
moment map . Let
0
denote the G-invariant subset of = r ` : (r) = 0.
where j intersects the zero section transversally and where G acts locally freely.
Then

` =
0
G is a quaternionic Kahler orbifold.
The statement here is a slight generalization of the main theorem in [GL88].
The proof of the theorem proceeds along the lines discussed in [GL88] with the
exception that the locally free action allows us to use Molinos Theorem 2.5.11 to
conclude that the quotient has an orbifold structure. We rephrase an important
special case of Theorem 12.4.4 as:
Corollary 12.4.5: Let ` be as above and suppose G

= T
k
Isom(`. o) is a
/-torus subgroup generated by a vector elds \
k
isom(`. o). If \
1
. . . \
k
is
a /-plane eld at all points r . then G is a compact quaternionic Kahler
orbifold.
Example 12.4.6: Consider the o
1
-action dened on HP
n
in homogeneous coordi-
nates as follows

([n
0
. . . . . n
n
]) = [n
0
. . . . . n
n
] .
where is a complex unit. Recall that the projectivization is by right multiplication
u uc as described in Section 12.1. In a local trivialization we can identify the
local frame
1
.
2
.
3
with the imaginary quaternions i. ,. /, respectively, in which
case the moment map for the o
1
-action can be written as
(12.4.9) = uiu =
n

j=0
n
j
in
j
.
So the zero set
1
(0) is invariant under right multiplication by H

and so cuts out a

real codimension 3 subvariety of HP
n
. Moreover, one easily checks that it is a smooth
embedded submanifold that is invariant under the left action of l(n+1) by u u.
The isotropy subgroup of the point [u] = [1. ,. 0. . 0]
1
(0) is identied with
ol(2)l(n1). So
1
(0) is the homogeneous space l(n+1)

ol(2)l(n1)

.
which identies the quotient
1
(0)o
1
with the complex Grassmannian

` =
Gr
2
(C
n+1
).
This can easily be generalized to the case of a weighted circle action

([n
0
. . . . . n
n
]) = [
p
0
n
0
. . . . .
p
n
n
n
] .
12.4. QUATERNIONIC K

AHLER REDUCTION 441

in which case the moment map becomes
(12.4.10) =
n

j=0
j
j
n
j
in
j
.
In this case the quotients
1
(0)o
1
are quaternionic Kahler orbifolds. The orbifold
stratication is analyzed for the special case j
0
= c and j
1
= j
n
= j in [GL88].
Theorem 12.4.4 can be used to obtain many examples of compact QK orbifolds.
When the reduced space is 4-dimensional it is automatically self-dual and Einstein.
The only complete positive QK manifolds in dimension 4 are HP
1
o
4
or CP
2
with their standard symmetric space metrics. In this context the more interesting
quotients are those with orbifold singularities. They will be discussed in the next
section.
Just as in the symplectic case one can study more singular QK quotients with-
out assuming that the action of the quotient group G on =
1
(0) is locally
free. A detailed study of this more general situation was done by Dancer and Swann
[DS97a]. Let (`. O. . o) be a QK manifold and let G be a connected Lie group
acting smoothly and properly on ` preserving QK structure with moment map .
For any subgroup H G we denote by `
H
the set of points in ` xed by H and
`
H
the set of points whose isotropy subgroups are exactly H. Further we write
`
(H)
for the set of points whose isotropy subgroups are conjugate to H in G. Now,
if `
H
is not empty then H must be compact. It follows that both `
H
`
H
`
are smooth manifolds. If (H) is the normalizer of H in G then 1 = (H)H acts
freely and properly on `
H
with the quotient `
H
1 = `
(H)
G a smooth manifold.
Thus ` decomposes into the union of `
(H)
G. where (H) runs over all conjugacy
classes of stabilizers. We dene
(12.4.11)

`
H
=

1
(0) `
(H)
G
=

1
(0) `
H
1
.
One can rst show that the stratication of ` by orbit types induces the strati-
cation of

` into a union of smooth manifolds, i.e.,
Theorem 12.4.7: Let H G be a subgroup so that `
H
is not-empty. Then

`
H
is a smooth manifold.
However, not all the pieces

`
H
have QK structures. Their geometry depends
on the way H acts on O. To be more precise, let r `
H
and consider the dierential
action of H on T
x
`. Since H acts preserving the quaternionic structure we have the
representation Hoj(n)oj(1) which induces the representation : HoO(3).
The group H acts on O R
3
via the composition of with the standard three-
dimensional representation. If r. n `
H
are on the same path-component then
parallel transport along any path joining r to n denes an H-equivariant isomor-
phism T
x
` T
y
`. It follows that the representation : HoO(3) is equivalent
at all points on a path component of `
H
. Hence, the image (H) < oO(3). up to
isomorphism, is the same on 1-orbits. There are four possibilities [DS97a]:
(H) is trivial

`
H
is a QK manifold,
(H) = Z
k
. / 1 each path component of

`
H
is covered by a Kahler
manifold,
(H) is nite but not cyclic `
H
is totally real in ` (

`
H
is real),
(H) = oO(2) or oO(3)

`
H
is empty.
442 12. QUATERNIONIC K

AHLER AND HYPERK

AHLER MANIFOLDS
Even simple examples show that the stratication of the quotient by orbit type can
include all of the three possible pieces. However, there is a coarser stratication of

` in which all pieces are in fact QK. Let `

[H]
be the set of points in `. where the
identity component of the stabilizer equals H and `
([H])
the set of points whose
stabilizer has identity component conjugate to H in G.
Theorem 12.4.8: The union

` =

HG
(
1
(0) `
([H])
)
G
taken over all compact connected subgroups H G induces a canonical decomposi-
tion of

` into a union of QK orbifolds each of which is a nite quotient of a QK
manifold.
Proof. The key observation is that `
[H]
is an open submanifold in `
H
.
where `
H
is a smooth QK submanifold of `. Hence, itself `
[H]
is a smooth QK
manifold. The restriction of to `
[H]
is the moment map for the locally free
action of 1. hence (
1
(0) `
[H]
)1 = (
1
(0) `
([H])
)G is a QK orbifold by
Theorem 12.4.4.
We nish this chapter with a brief discussion of Morse theory on QK manifolds.
The idea to consider 1 = [[|[
2
as the Morse function is quite natural as suggested in
analogy with Kirwans work on symplectic quotients [Kir84, Kir98]. The function
1 was rst introduced by Battaglia in [Bat96b, Bat99] and more recently also in
[ACDVP03]. The motivation behind [ACDVP03] was the fact that the so-called
BPS sates in 5-dimensional supergravity theory correspond to gradient ows on a
product ` . where ` is a negative QK manifold and is a special Kahler
space. Such ows are generated by certain energy function 1 which is nothing
but the square of the moment map 1 = [[|[
2
. Battaglia was interested mostly in
the positive QK case and, it appears the authors of [ACDVP03] were not aware
of her work. We now describe some of the Battaglias results. Recall that a Morse
function 1 is called equivariantly perfect over Q if the equivariant Morse equalities
hold, that is if

1
t
(`) =

1
t
(
1
(0)) +

F
1
t
(1) .
where the sum ranges over the set of connected components of the xed point
set,
F
is the index of 1. and

1
t
is the equivariant Poincare polynomial for the
equivariant cohomology with coecients in Q. Battaglia proves
Theorem 12.4.9: Let (`
4n
. O. . o) be a positive QK manifold acted on isomet-
rically by o
1
. Then the non-degenerate Morse function 1 = ||
2
is equivariantly
perfect over Q. The critical set of 1 is the union of the zero set 1
1
(0) =
1
(0)
and the xed point set of the circle action.
Moreover, the zero set
1
(0) is connected, and a xed point component is
either contained in
1
(0) or does not intersect with
1
(0).
Proposition 12.4.10: Let `
4n
be a positive QK manifold acted on isometrically
by o
1
. Then every connected component of the xed point set, not contained in

1
(0). is a Kahler submanifold of ``
1
(0) of real dimension less than or equal
to 2n whose Morse index is at least 2n. with respect to the function 1.
12.5. COMPACT QUATERNIONIC K

AHLER ORBIFOLDS 443

In [Bat99] Battaglia uses Morse theory to improve the results obtained earlier
in [Bat96b]. She shows that the quotient in Example 12.4.6 is unique in the
following sense.
Theorem 12.4.11: Let (`
4n
. O. . o) be a positive QK manifold acted on isomet-
rically by o
1
. Suppose o
1
acts freely on =
1
(0). Then `
4n
is homotopic to
HP
n
with the quotient

` = Gr
2
(C
n+1
).
12.5. Compact Quaternionic Kahler Orbifolds
As already indicated the method of QK reduction enjoys much success if one
allows the quotient ` of be a QK orbifold. The price that is paid is the loss of the
rigidity described in Section 12.3. Perhaps a more interesting observation about
such a generalization is that when ` has orbifold singularities the total space of
the orbifold Konishi bundle o(`), which is a principal orbibundle with structure
group G = oO(3) or G = oj(1). may actually be smooth. That should not come
as a surprise to the reader familiar with earlier chapters of our book. In fact,
this happens exactly (just as it does in the case of orbifold circle \ -bundles) when
the orbifold uniformizing groups are subgroups of the structure group G. In this
section we introduce some examples of compact positive QK orbifolds and discuss
some obvious classication problems.
The rst examples of positive QK orbifolds were introduced in 1987 by Galicki
and Lawson [Gal87a, GL88]. We will briey describe the construction slightly
generalizing the original example. The key to the construction is Corollary 12.4.5
of the previous section. Consider (`. o) = (HP
n
. o
can
) and an arbitrary reduction
of ` be a /-dimensional Abelian subgroup of the isometry group. Such a reduction
is associated to a choice
(12.5.1) H = T
k
T
max
= T
n+1
l(n + 1) oj(n + 1) .
where oj(n + 1) = Isom(HP
n
. o
can
) and T
max
= T
n+1
is the maximal torus sub-
group. One can always choose T
n+1
to be the set of diagonal matrices in the
unitary group l(n + 1). Any rational subtorus H is then determined by a collec-
tion of a non-zero integer vectors
1
. . . . .
n+1
generating R
n+1k
. These can be
put together as a matrix
n+1k,n+1
(Z). Dually, we can consider a matrix

k,n+1
(Z) whose column vectors
1
. . . . .
n+1
generate R
k
. This gives the
exact sequence of Lie algebras
(12.5.2) 0 h

R
n+1

R
n+1k
0 .
and its dual
(12.5.3) 0 R
n+1k

R
n+1

0 .
where (e
i
) = e
i
=
i
R
n+1k
and

(c
i
) = e
i
=
i
R
k
with e
1
. . . . . e
n+1

being the standard basis in R

n+1
. There is corresponding exact sequence at the
group level 1HT
n+1
T
n+1k
1 and the subtorus H is identied with the
image of the homomorphism 1

: T
k
T
n+1
(12.5.4) 1

(
1
. . . . .
k
) = diag

j=1

a
j
1
j
. . . . .
k

j=1

a
j
n+1
j

.
444 12. QUATERNIONIC K

AHLER AND HYPERK

AHLER MANIFOLDS
where (o
j
i
) = . Now, with each we associate H acting on HP
n
. the QK moment
map j

, and the zero level set () = j

1

(0) HP
n
. It is elementary to check
when the condition of Corollary 12.4.5 is satised. We have arrived at
Theorem 12.5.1: Suppose all / / minor determinants of do not vanish, i.e.,
any collection of / column vectors of are linearly independent. Then the reduced
space O() =

`() is a compact positive QK orbifold. Furthermore, the Lie
algebra isom(O(). o()) contains (n + 1 /) commuting Killing vector elds. In
particular, when / = n1. O() is a compact self-dual Einstein orbifold of positive
scalar curvature and 2 commuting Killing vector elds.
The rational cohomology of these orbifolds were computed independently in
[BGMR98] and [Bie97]. In particular, we have
Corollary 12.5.2: There exist compact toric positive self-dual Einstein orbifolds
with arbitrary second Betti number.
Remark 12.5.1: The case originally considered in [Gal87a, GL88] corresponds
to / = 1 and = (c. j. . . . . j) with = (1. . . . . 1) being the canonical quotient of
Example 12.4.6. More general cases were analyzed only much later in [BGM94a]
( = (j
1
. . . . . j
n+1
)) and [BGMR98] (an arbitrary ), where it was realized that
the Konishi orbibundle of such orbifolds can often be a smooth manifold carrying
a natural Einstein metric. We shall return to a detailed analysis of these examples
in the next chapter after we dene 3-Sasakian manifolds.
We now specialize to the case of compact 4-dimensional QK orbifolds (`. o),
i.e., 4-orbifolds with self-dual conformal structure with an Einstein metric o of
positive scalar curvature. Recall that when ` is smooth it must be isomorphic to
o
4
or CP
2
. On the other hand, Theorem 12.5.1 alone provides plenty of examples
of such spaces [GL88]. As orbifolds, some of them are the familiar examples of
weighted projective spaces introduced in Chapter 4.
Proposition 12.5.3: Let O(p) be O() of Theorem 12.5.1 with = (j
1
. j
2
. j
3
) =
p. i.e., O(p) is a QK reduction of HP
2
by the isometric circle action with weights
p. In addition assume that all j
i
s are positive integers such that gcd(j
1
. j
2
. j
3
) = 1.
Then
(i) there is smooth orbifold equivalence
O(p)

CP
2
p
1
+p
2
2
,
p
2
+p
3
2
,
p
3
+p
1
2
. when j
i
is odd for all i.
CP
2
p1+p2,p2+p3,p3+p1
. otherwise;
(ii) the metrics o(p) dened by QK reduction are inhomogeneous unless p =
(1. 1. 1) in which case we get the Fubini-Study metric on CP
2
;
(iii) the QK metrics o(p) are Hermitian with respect to the standard complex
structure on the corresponding weighted projective space.
Proof. To proof (i) one needs to identify the level set of the moment map
with o
5
which is easily done. We refer to Section 13.7.4, where it is shown that
the level set of the 3-Sasakian moment map is dieomorphic to the Stiefel manifold
\
2
(C
3
) = l(3)l(1). Now, in terms of the QK quotient, the level set must be
l(3)l(2) o
5
. The result follows by observing that the circle action on v = zw
(which can be though as a coordinate on the 5-sphere) has weights (j
2
+ j
3
. j
3
+
j
1
. j
1
+j
2
). (ii) can be proved by using the relation between the local form of any
positive toric metric given in Theorem 12.5.5 and QK toric quotients, This relation
12.5. COMPACT QUATERNIONIC K

AHLER ORBIFOLDS 445

was established in [CP02]. In particular, it easily follows that the metric is of
cohomogeneity two for all distinct weights, and of cohomogeneity one if exactly two
weights are equal. The case of equal weight gives the Fubini-Study metric which
is symmetric. Finally (iii) follows from the result of Apostolov and Gauduchon
[AG02].

The above orbifolds are also quite interesting for another reason. In addition to
being self-dual and Einstein the metric o(p) is often of positive sectional curvature.
This curvature property of O(p) was discovered by Dearricott [Dea04, Dea05] and
by Blazic and Vukmirovic [BV04]. First we have the following result Dearricott
Theorem 12.5.4: Let O(p) be the Galicki-Lawson orbifold with j
1
j
2
j
3
.
Then the self-dual Einstein metric o(p) is of positive sectional curvature if and
only if
(12.5.5)
3
(j
1
+j
2
+j
3
. j
1
j
2
+j
3
. j
1
+j
2
j
3
. j
1
j
2
j
3
) 4j
3
3
.
where
3
denotes the third symmetric polynomial in 4 variables.
The paper of Blazic and Vukmirovic uses quite dierent methods. In fact
their main theorem is a generalization of the Galicki-Lawson examples to the case
of pseudo-Riemannian metrics of split signature (+. +. . ). where the quotient
construction involves paraquaternions. However, the curvature calculations apply
to the Riemannian case as well. For the l(2)-symmetric orbifolds O(p) = O(j. c. c)
case they calculate the pinching constants and get the following result.
Theorem 12.5.5: The self-dual Einstein metric on O(p) = O(j. c. c) has positive
sectional curvature if j
2
< c
2

2 < 2j
2

2 and then at every point r O(j. c. c)

the sectional curvature is /-pinched with 0 < / < 1 and
1
4

3
4
c
2
j
2
c
2
+j
2
/
1
4
+
3
4
c
4
j
4
c
4
+j
4
. j c.
1
4

3
4
j
4
c
4
j
4
+c
4
/
1
4

3
4
j
2
c
2
j
2
+c
2
. c j.
Clearly, /
1
4
when j c. Furthermore, / =
1
4
if and only if j = c = 1 in which
case O CP
2
is symmetric.
When
n1,n+1
(Z) the orbifold structure of O() is more involved. How-
ever, as each O() has two commuting Killing vectors, locally these metrics are
described by the following results of Calderbank and Pedersen [CP02]
Theorem 12.5.6: Let 1(. ) be a solution of the linear dierential equation
1

+1

=
31
4
2
on some open subset of the half-space 0. and consider the metric o(. . . )
given by
o =
1
2
4
2
(1
2

+1
2

)
41
2
d
2
+d
2

2
+

(1 21

) 21

2
+

21

+ (1 + 21

2
1
2

1
2
4
2
(1
2

+1
2

)
.
(12.5.6)
where =

d and = (d + d)

. Then
446 12. QUATERNIONIC K

AHLER AND HYPERK

AHLER MANIFOLDS
(i) On the open set where 1
2
4
2
(1
2

+ 1
2

), o is a self-dual Einstein
metric of positive scalar curvature, whereas on the open set where 0 <
1
2
< 4
2
(1
2

+ 1
2

). o is a self-dual Einstein metric of negative scalar

curvature.
(ii) Any self-dual Einstein metric of non-zero scalar curvature with two lin-
early independent commuting Killing elds arises locally in this way (i.e.,
in a neighborhood of any point, it is of the form (12.5.6) up to a constant
multiple).
This theorem, together with the explicit construction of Theorem 12.5.1 leads
quite naturally to the question: Are all compact positive QK orbifolds admitting
two commuting Killing vectors obtained via some QK reduction O() of HP
n
? A
partial answer to this question in the case when the Konishi bundle of O() is
smooth (which is an extra condition on ) was provided by Bielawski in [Bie99].
We shall discuss his result later in the context of smooth toric 3-Sasakian mani-
fold. More recently, via a more careful analysis of orbifold singularities Calderbank
and Singer proved the following [CS06a]
Theorem 12.5.7: Let (O. o) be a compact self-dual Einstein 4-orbifold of positive
scalar curvature whose isometry group contains a 2-torus. Then, up to orbifold
coverings, (O. o) is isometric to a quaternionic Kahler quotient of quaternionic
projective space HP
n
, for some n 1. by a (n1)-dimensional subtorus of oj(n+1).
There is yet another family of orbifold metrics due to Hitchin [Hit95a, Hit96,
Hit95b]. These metrics come from solutions of the Painleve VI equation and as
such were also introduced by Tod [Tod94]. We describe these metrics in some
details here and come back to them once more in the next chapter. Consider the
space \ dened by
(12.5.7) \ = B
3,3
(R) [ B
T
= B. Tr(B) = 0
of traceless symmetric 3 3 matrices with inner product 'B
1
. B
2
` = tr(B
1
B
2
).
Clearly, \ R
5
and oO(3) acts on \ by conjugation B o
1
Bo. o oO(3) and
the unit sphere ` = o
4
in \ can be described as matrices in \ whose eigenvalues

1
.
2
.
3
satisfy
(12.5.8)
3

i=1

i
= 0 = 1
3

i=1

2
i
.
This action has cohomogeneity one and is the Z
2
-quotient of the rst case described
in Example 1.6.34. The associated group diagram has the structure
(12.5.9) oO(3)
1

= O(2)
j

.
.
.
.
.
.
.
.
.
.
.
1
+
= O(2)
j
+

1 = Z
2
Z
2
h

h+

.
.
.
.
.
.
.
.
.
.
.
.
.
with the generic orbit oO(3)1. where 1 = Z
2
Z
2
oO(3) is the subgroup of
diagonal matrices which is the stabilizer of any generic point. The two degenerate
orbits 1

= oO(3)1

RP
2
are both Veronese surfaces in o
4
that correspond
12.5. COMPACT QUATERNIONIC K

AHLER ORBIFOLDS 447

to the subset of matrices with two equal eigenvalues. If two eigenvalues are equal,
then they must be equal to
1

6
. and the two signs correspond to the two orbits 1

and the subgroups 1

= O(2). The diagram 12.5.9 gives rise to the decomposition

(12.5.10) o
4
= RP
2

(0. +) (oO(3)1)

RP
2
.
Explicitly, we can parameterize the conic 12.5.8 by observing that (
1

2
)
2
+
3(
1
+
2
)
2
= 2 so that
(12.5.11)

1
(t) =

2
2
(t) +
1

6
(t) .
2
(t) =

2
2
(t) +
1

6
(t) .
3
(t) = 2(t) .
where
2
+
2
= 1. Choose the standard rational parameterization
(t) =
2t
t
2
+ 1
. (t) =
1 t
2
t
2
+ 1
.
Note that t = 0 gives
1
=
2
=
1

6
and t = + gives the other degenerate
orbit
2
=
2
=
1

6
so that we have an explicit dieomorphism o
4
` 1

. 1
+

(0. +) oO(3)1 given by (t. o) o
1
(t)o. where o oO(3) and (t) =
diag(
1
(t).
2
(t).
3
(t)).
Any oO(3)-invariant metric on o
4
` 1
+
. 1

denes an invariant metric on

each orbit oO(3)1. It follows that any such metric must be of the form
(12.5.12) o = 1(t)dt
2
+ [T
1
(t)]
2

2
1
+ [T
2
(t)]
2

2
2
+ [T
3
(t)]
2

2
3
.
where
1
.
2
.
3
is the basis of Maurer-Cartan invariant one-forms dual to the
standard basis of the Lie algebra so(3). The equations for the most general self-
dual Einstein metric with non-zero scalar curvature and in the diagonal form
(12.5.12) has been derived by Tod [Tod94]. It follows that
(12.5.13) 1o =
dr
2
r(r 1)
+

2
1
1
2
1
+
(1 r)
2
2
1
2
2
+
r
2
3
1
2
3
.
where 1
1
(r). 1
2
(r). 1
3
(r) satisfy the following rst order system of ODEs
(12.5.14)
d1
1
dr
=
1
2
1
3
r(1 r)
.
d1
2
dr
=
1
3
1
3
r
.
d1
3
dr
=
1
1
1
2
r(1 r)
and the conformal factor
(12.5.15)
41 =
8r1
2
1
1
2
2
1
2
3
+ 21
1
1
2
1
3
(r(1
2
1
+1
2
2
) (1 41
2
3
)(1
2
2
(1 r)1
2
1
))
(r1
1
1
2
+ 21
3
(1
2
2
(1 r)1
2
1
))
2
.
The expression for the conformal factor is algebraic in r. 1
1
. 1
2
. 1
3
so that the
problem reduces to solving the system (12.5.14). It turns out that this system can
be reduced to a single second order ODE: the Painleve VI equation
(12.5.16)
d
2
n
dr
2
=
1
2

1
n
+
1
n 1
+
1
n r

dn
dr

1
r
+
1
r 1
+
1
n r

dn
dr
+
+
n(n 1)(n r)
r
2
(r 1)
2

+
r
n
2
+
r 1
(n 1)
2
+
r(r 1)
(n r)
2

.
where (. . . ) = (18. 18. 18. 38). One can dene an auxiliary variable . by
(12.5.17)
dn
dr
=
n(n 1)(n r)
r(r 1)

2.
1
2n
+
1
2(n 1)
+
1
2(n r)

448 12. QUATERNIONIC K

AHLER AND HYPERK

AHLER MANIFOLDS
which then allows one to express the original functions 1
1
. 1
2
. 1
3
in terms of any
solution n = n(r) of the equation 12.5.16:
(12.5.18)
1
2
1
=
n(n 1)(n r)
2
r(1 r)

.
1
2(n 1)

.
1
2n

.
1
2
1
=
n
2
(n 1)(n r)
r

.
1
2(n r)

.
1
2(n 1)

.
1
2
1
=
n(n 1)
2
(n r)
1 r

.
1
2n

.
1
2(n r)

.
In a series of papers Hitchin analyzed the Painleve VI equation giving an alge-
braic geometry description of the solutions in terms of isomonodromic deforma-
tions [Hit95a, Hit96, Hit95b]. In particular, any such solution can be described
in terms of a meromorphic function on an elliptic curve

C with a zero of order /
at a chosen point 1 and a pole of order / at a point 1. Hitchins description
gives explicit formulas for the coecients of the metric 1
1
. 1
2
. 1
3
in terms of the
elliptic functions. In particular, Hitchin shows
Theorem 12.5.8: Choose an integer / 3 and consider the oO(3)-invariant met-
ric o
k
dened on (1. ) oO(3)1 by the formula (12.5.13) via the corresponding
solution of the Painleve VI equation with the metric coecients 1
i
= 1
i
(r. /),
i = 1. 2. 3.
(i) The metric o
k
is a positive denite self-dual Einstein of positive scalar
curvature for all 1 < r < .
(ii) The metric o
k
extends smoothly over at r = 1 over 1

= RP
2
and
as r o
k
acquires an orbifold singularity with angle
2
k2
around
1
+
= RP
2
.
Hence, for any integer / 3 the metric o
k
can be interpreted as an orbifold metric
on O
k
= 1

(1. )oO(3)1

1
+
o
4
. where (O
k
. o
k
) is a compact self-dual
Einstein orbifold of positive scalar curvature.
We shall return to these cohomogeneity one orbifold metrics in the next chapter
when we consider the Konishi bundle over O
k
and the twistor space Z(O
k
). Here
we discuss the metric for some lower values of /. To each / one associates a solution
of the Painleve VI equation n = n(:) and we write the metric o
k
as in 12.5.12
(12.5.19) o
k
= 1(:; /)d:
2
+ [T
1
(:; /)]
2

2
1
+ [T
2
(:; /)]
2

2
2
+ [T
2
(:; /)]
2

2
3
.
Explicit computation shows that for / = 3 the metric is given by the following
solution of the Painleve VI equation
n =
:
2
(2:
2
+ 5: + 2)
(2: + 1)(:
2
+: + 1)
. with r =
:
3
(: + 2)
2: + 1
.
The components of o
3
can easily be calculated with 1(:; 3) = 3(1 +: +:
2
)
2
and
T
2
1
(:; 3) =
3(1 + 2:)
2
(1 +: +:
2
)
2
. T
2
2
(:; 3) =
3(1 :
2
)
2
(1 +: +:
2
)
2
. T
2
3
(:; 3) =
3:
2
(2 +:)
2
(1 +: +:
2
)
2
.
In the arc length coordinates this metric can be easily transformed to
o
3
= dt
2
+ 4 sin
2
t
2
1
+ 4 sin
2
(23 t)
2
2
+ 4 sin
2
(t + 23)
2
3
.
which shows that o
3
is the standard metric on o
4
written in triaxial form. Hence,
the orbifold (O
3
. o
3
) is actually non-singular and the metric is the standard one
12.5. COMPACT QUATERNIONIC K

AHLER ORBIFOLDS 449

on o
4
. For / = 4 the metric comes from the following solution of the Painleve VI
equation
n = :. with r = :
2
.
so that
1(:; 4) =
1
4:(1 +:)
. T
2
1
(:; 4) =
1
1 +:
. T
2
2
(:; 4) =
(1 :)
2
(1 +:)
2
. T
2
3
(:; 4) =
:
1 +:
.
In an arc length parameterization o
4
becomes
(12.5.20) o
4
= dt
2
+ sin
2
t
2
1
+ cos
2
2t
2
2
+ cos
2
t
2
3
.
which is indeed locally the Fubini-Study metric on CP
2
. The orbifold (O
4
. o
k
) has

orb
1
= Z
2
and its universal cover is (CP
2
. o
FS
). This corresponds to the second case
of Example 1.6.34.
Just to illustrate how complicated the metric coecients get for larger values
of /, following Hitchin [Hit96] we also give explicit formulas for / = 6. 8. For / = 6
(the orbifold singularity at angle 2) one gets
n =
:(:
2
+: + 1)
(2: + 1)
. with r =
:
3
(2: + 2)
: + 1
.
This yields
1(:; 6) =
(1 +: +:
2
)
: (: + 2)
2
(2: + 1)
2
.
[T
1
(:; 6)]
2
=
(1 +: +:
2
)
(: + 2) (2: + 1)
2
.
[T
2
(:; 6)]
2
=
(:
2
1)
2
(1 +: +:
2
) (: + 2) (2: + 1)
.
[T
3
(:; 6)]
2
=
: (1 +: +:
2
)
(: + 2)
2
(2: + 1)
.
This gives the metric o
6
of Equation 12.5.19 for the range 1 < : < .
For / = 8 (the orbifold singularity at angle 3) we get
n =
4:(3:
2
2: + 1)
(: + 1)(1 :)
3
(:
2
+ 2: + 3)
. with r =

2:
1 :
2

4
.
This yields
1(:; 8) =
(1 +:)(:
2
+ 2: 1)(3 2: +:
2
)(1 2: + 3:
2
)(1 + 2: + 3:
2
)
(1 :) : (1 +:
2
)(:
2
2: 1) (3 + 2: +:
2
)
2
(:
4
6:
2
+ 1)
.
[T
1
(:; 8)]
2
=
(1 +:
2
)(3 2: +:
2
)(1 2: :
2
)
2
(1 + 2: + 3:
2
)
(1 + 2: :
2
)
2
(3 + 2: +:
2
)
2
(1 2: + 3:
2
)
.
[T
2
(:; 8)]
2
=
(1 :)(1 +:)
3
(3 2: +:
2
)(1 2: + 3:
2
)
(1 + 2: :
2
)(3 + 2: +:
2
)
2
(1 + 2: + 3:
2
)
.
[T
3
(:; 8)]
2
=
4:(: + 1)
2
(1 2: + 3:
2
)(1 + 2: + 3:
2
)
(1 + 2: :
2
)(3 2: +:
2
)(3 + 2: +:
2
)
2
.
This metric o
8
of Equation 12.5.19 becomes positive denite for the range

21 <
: < 1.
The Theorems of Hitchin [Hit95b], Calderbank and Pedersen [CP02], and
Calderbank and Singer [CS06a] are milestones in the broader problem of the clas-
sication of all compact self-dual Einstein 4-orbifolds. It seems plausible that these
450 12. QUATERNIONIC K

AHLER AND HYPERK

AHLER MANIFOLDS
are the only cohomogeneity one compact positive self-dual Einstein 4-orbifolds, but
a proof is lacking so far.
Open Problem 12.5.1: Classify all compact positive self-dual Einstein 4-orbifolds
with a cohomogeneity one action of a Lie group.
If one adds to this classifying the cohomogeneity two actions of ol(2) one
arrives at
Open Problem 12.5.2: Classify all compact positive self-dual Einstein 4-orbifolds
with at least a 2-dimensional isometry group.
The problem of nding examples of compact positive self-dual Einstein 4-
orbifolds without any assumption about symmetries seems quite intractable. No
compact orbifolds without any Killing vector elds are known at this time. How-
ever, there are two examples of positive QK 4-orbifolds with one Killing vector,
both obtained via symmetry reduction.
Example 12.5.9: [QK Extension of Kronheimer Quotients] The rst orbifold
examples of positive QK orbifold metrics which are not toric were constructed in
[GN92]. The construction is a quaternionic Kahler modication of the Kronheimer
construction of hyperkahler ALE spaces discussed later in Section 12.10. We use
the notation there to explain the result. Let oj(1) be a discrete subgroup.
Consider the quaternionic projective space P
H
(H
]]
H). The Kronheimer group
1() acts on H
]]
as in 12.10.3. Let us consider a homomorphism / : 1()oj(1).
Such a homomorphism extends the action of 1() to P
H
(H
]]
H) via
(12.5.21) o [u. n
0
] = [o u. /(o)n
0
]. u H
]]
. n
0
H. o 1() .
where u o u is the Kronheimers action discussed in 12.10.3. Consider the
map d/ : ksp(1) of Lie algebras. We can think of d/ k

:j(1) by setting
'd/. A` = d/(A). A k. Let us denote the new action by 1(; /) and the QK
reduction of P
H
(H
]]
H) by O(. /). We have the following [GN92]
Theorem 12.5.10: The QK reduced space O(. /) is a compact positive self-dual
and Einstein orbifold if d/ k

:j(1) is in the Kronheimers good set of Denition

12.10.4.
Note that in the case of = Z
n
the construction and Theorem 12.5.10 merely
give subfamilies of toric examples discussed earlier. However, in all cases when
is non-Abelian we get families of positive self-dual Einstein orbifold metrics which
are not toric.
Example 12.5.11: [Abelian Quotients of Real Grassmannians] Consider the
positive QK structure on the Grasmannian of oriented 4-planes in R
n
. The isometry
group oO(n) of the symmetric space Gr
+
4
(R
n
) contains a torus and one can examine
possible QK reductions of Gr
+
4
(R
n
). These are described in detail at the level of the
Konishi bundle in Section 13.9. In particular, there are only two possible Abelian
quotients of Gr
+
4
(R
n
) for which the reduced space is 4-dimensional: (i) T
3
-reduction
of Gr
+
4
(R
8
) and (ii) T
2
-reduction of Gr
+
4
(R
7
) (see Proposition 13.9.2). Both lead
to non-trivial examples of positive self-dual Einstein orbifold metrics which are not
toric and they were rst introduced in [BGP02] and later studied in [Bis07]. In
particular, with the notation of Denition 13.9.1 and Proposition 13.9.2 we have
the following
12.5. COMPACT QUATERNIONIC K

AHLER ORBIFOLDS 451

Theorem 12.5.12: Let the weight matrices
1
2,3

2,3
(Z) and
0
3,4

3,4
(Z)
describe the choices of T
2
T
3
oO(7) and T
3
T
4
oO(8). respectively.
Let O(
1
2,3
) and O(
0
3,4
) denote the corresponding QK reductions of Gr
+
4
(R
7
) and
Gr
+
4
(R
8
). We have
(i) If all three 2 2 minor determinants of
1
2,3
are non-zero then O(
1
2,3
)
is a compact 4-orbifold.
(ii) If all four 3 3 minor determinants of
0
3,4
are non-zero then O(
0
3,4
)
is a compact 4-orbifold.
In both case we get compact orbifold families of positive self-dual Einstein metrics
with a one-dimensional isometry group.
With all the available examples one can naturally begin asking questions about
geometric properties of such metrics. It turns out that a pivotal role in understand-
ing such metrics is played by the non-linear PDE
(12.5.22) n
xx
+n
yy
+ (c
u
)
zz
= 0 .
This equation was rst described in [BF82] as providing solutions to the self-dual
Einstein equations with zero scalar curvature and one Killing vector eld of rota-
tional type. In [BF82] it was shown that the zero scalar curvature (or vacuum)
self-dual Einstein equations admitting one Killing vector eld amounts to solving ei-
ther the well-known three dimensional Laplace equation or Equation 12.5.22. Those
Killing elds that led to the three dimensional Laplace equation were called trans-
lational Killing elds, whereas, those leading to Equation 12.5.22 were called rota-
tional. The translational Killing elds have self-dual covariant derivative and are
well understood [TW79]. For example, they give rise to the well-known Gibbons-
Hawking Ansatz [GH78a]. On the other hand Equation 12.5.22 has proven to be
very resistent in oering up explicit solutions [Fin01]. Nevertheless, it has appeared
in a variety of settings, for example LeBrun [LeB91b] used it in his construction
of self-dual metrics on the connected sums of CP
2
. (See also [PP98b, AG02]
for further development in terms of Hermitian-Einstein geometry). Moreover, it
can be viewed as an innite dimensional version of the better known Toda lat-
tice equation associated with the Lie algebra of type
n
. and so it has become
known as the ol()-Toda eld equation [War90, Sav89] or alternatively the
Boyer-Finley equation [FKS02]. Its importance for us at this stage lies in the re-
markable observation made by Tod [Tod97] that nding solutions to the self-dual
Einstein equations with non-zero scalar curvature can be reduced to solving Equa-
tion (12.5.22). This equation as with the full self-dual or anti-self-dual Einstein
equations is related to integrability questions, innite sequences of conservations
laws, and twistor theory, which we briey discuss in Section 12.7. Actually there
are several cases where there are known implicit solutions to Equation (4.6.2). It
would be interesting to see if one could turn implicit solutions of Theorem 12.5.12
and Theorem 12.5.12 into explicit solutions of Equation (4.6.2).
Most 4-dimensional Einstein metrics appear as Riemannian metrics adapted to
some other geometric structure, self-dual (or anti-self-dual) metrics, and/or Kahler
metrics are perhaps the best known examples. Here is an interesting Venn dia-
gram taken from Tod [Tod97], where the special intersecting regions deserve some
comment.
Let us now discuss the overlapping areas of Figure 1. All of the labelled areas
can be related to Sasakian geometry by taking an appropriate o
1
or oO(3) bundle or
452 12. QUATERNIONIC K

AHLER AND HYPERK

AHLER MANIFOLDS
W
+
= 0 (or W

= 0)
Einstein Kahler
C
A
B D
.................................................................................................................................................................................................................................................................................................................................................................................................................................................................. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..
................................................................................................................................................................................................................................................................................................................................................................................................................................................................. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ................................................................................................................................................................................................................................................................................................................................................................................................................................................................. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..
Figure 1. Tods Venn diagram of special metrics in dimension four.
orbibundle. In the area labelled C are the generic Kahler-Einstein 4-manifolds that
were treated in Chapter 5. In the regions labelled A and D the complex structure
chooses an orientation which breaks the equivalence between being self-dual or anti-
self-dual. Thus, in both region A and region D one must consider the two cases
separately. The two cases for region A consists of self-dual Kahler-Einstein, and
anti-self-dual Kahler Einstein. In the category of compact manifolds the former
consists of CP
2
with the Fubini-Study metric and compact quotients of Hermitian
hyperbolic space with the Bergman metric [ADM96, Boy88b, Kod87, KS93b],
whereas, the later consists only of at tori, 13 and Enriques surfaces, which are
all locally hyperkahler and have zero scalar curvature. Many more examples of
both self-dual Kahler-Einstein and anti-self-dual Kahler-Einstein appear when one
allows orbifold singularities [Bry01, ACG06], for example in the hyperkahler case
Reids list of 95 singular 13 given in Appendix B.1 and discussed in Section 5.4.2
occur. Likewise, complete metrics on non-compact 4-manifolds are plentiful in
all regions of the above diagram. We discuss hyperkahler geometry more fully
in Sections 12.7 through 12.10 with many 4-dimensional examples. In region D
many researchers [Bou81, Bry01, Che78, Der83, Ito84] have obtained results
concerning self-dual Kahler manifolds. Anti-self-dual Kahler metrics automatically
have zero scalar curvature [Der83]. Moreover, which possible compact complex
surfaces can admit such metrics have been delineated [Boy86]. There has been
much work [LeB91d, LS93, KP95, Tod95, Dan96, KLP97, RS05, DF06]
within the last fteen years or so in proving the existence of scalar at Kahler
metrics on 4-manifolds. Finally region B are the self-dual metrics which are not
Kahler. Here we are interested mainly in the case of positive scalar curvature, since
only these admit oO(3) or ol(2) orbibundles whose total space has a 3-Sasakian
structure. These have been discussed in detail above. The world of negative scalar
curvature self-dual Einstein metrics is fascinating with spectacular abundance of
complete metrics on non-compact manifolds [Ped86, Gal87b, Gal91, LeB91c,
Hit95b, Biq00, Biq02, CP02, CS04, BCGP05, Duc06]. The subject is worthy
of a separate book. We concentrate on the positive QK manifolds here as they are
bases of Konishi orbibundles of 3-Sasakian spaces considered in the next chapter.
However, negative QK manifolds are also related to 3-Sasakian (and not just semi-
Riemannian 3-Sasakian geometry) as rst observed by Biquard [Biq99].
12.6. HYPERCOMPLEX AND HYPERHERMITIAN STRUCTURES 453
12.6. Hypercomplex and Hyperhermitian Structures
Recall from Example 1.4.19 the following
Denition 12.6.1: A smooth manifold ` is said to be almost hypercomplex
if it admits a G1(n. H)-structure.
Alternatively, an almost hypercomplex structure is an almost quaternionic
structure such that the subbundle O End(T`) is trivial. Thus, O has a global
orthonormal frame 1
1
. 1
2
. 1
3
whose elements satisfy Equation (12.2.1). Such an
orthonormal frame can be viewed as a map 1 : R
3
O satisfying 1(e
a
) = 1
a
.
where e
a

3
a=1
is the standard basis for R
3
. So given any two points . o
2
we
can write Equation (12.2.1) in terms of arbitrary frames of O as
(12.6.1) 1() 1(
t
) = '.
t
`1l +1(
t
) .
where '.
t
` is the standard inner product in R
3
and
t
is the cross-product.
So hypercomplex structure provides ` with an o
2
s worth of complex structures.
We denote the family of complex structures satisfying Equation (12.6.1) by I and
refer to it as a hypercomplex structure.
Let o be a metric on ` such that
(12.6.2) o(1()A. 1()Y ) = o(A. Y ) .
for any o
2
nd A. Y (T`). Such a metric is said to be adapted to the
hypercomplex structure I and the pair (I. o) is called a hyperhermitian structure.
It is easy to see that such a metric always exists.
Denition 12.6.2: An almost hypercomplex manifold (`. I. o) with an adapted
metric o is called an almost hyperhermitian manifold.
As discussed in Chapter 1 an almost hyperhermitian structure is equivalent to
a reduction of the G1(n. H)-bundle to the subgroup oj(n). Obata showed that
every almost hypercomplex manifold `
4n
admits a canonical G1(n. H)-invariant
connection [Oba66], called the Obata connection.
Denition 12.6.3: An almost hypercomplex manifold (`. I) is called hypercom-
plex if all complex structures 1(). o
2
are integrable. A hypercomplex manifold
with an adapted metric is called hyperhermitian.
Actually, if any two orthogonal almost complex structures in the almost hyper-
complex structure are integrable then all the complex structures 1() are integrable
[Oba66, Sal89]. In the hypercomplex case integrability can be expressed in sev-
eral dierent ways. For example, the Obata connection in general has non-trivial
torsion. But on a hypercomplex manifold this unique connection is torsion-free.
So an alternative denition of a hypercomplex structure is that it is an almost hy-
percomplex structure such that the Obata connection is torsion-free. In the lowest
dimension compact hyperhermitian 4-manifolds were classied by Boyer [Boy88a]
who proved
Theorem 12.6.4: Let (`. I. o) be a compact hyperhermitian 4-manifold. Then
(`. I. o) is conformally equivalent to one of the following
(i) a 4-torus with its at metric,
(ii) a K3 surface with a Kahler Ricci-at metric,
(iii) a coordinate quaternionic Hopf surface with its standard locally confor-
mally at metric.
454 12. QUATERNIONIC K

AHLER AND HYPERK

AHLER MANIFOLDS
In higher dimensions there are many examples of hypercomplex structures, but
no classication results so far. For example, it is known which Lie groups admit such
structures [SSTVP88, Joy92, BDM96]. The simplest example here is G = l(2)
which as a compact complex surface is a Hopf surface, and it actually admits two
commuting hypercomplex structures. More generally [Joy92]
Theorem 12.6.5: Let G be a compact Lie group. Then there exist an integer
0 / max3. rk(G) such that l(1)
k
G has a homogeneous hypercomplex
structure.
There is a natural construction of hypercomplex structures on the total space of
circle bundles over any 3-Sasakian manifold ` [BGM98a] (See the next chapter for
a description of 3-Sasakian structures). For example, any trivial bundle o
1
` then
admits locally conformally hyperkahler structures that are automatically hypercom-
plex. However, non-trivial circle bundles give more interesting results. For example,
large families of hypercomplex structures were shown [BGM94b, BGM96a] to
exist on the complex Stiefel manifolds V
C
n,2
of 2-frames in C
n
. (See also [Bat96a]).
There is also a good deformation theory for hypercomplex structures [PP98a].
Further discussion of hypercomplex structures and their relation to quaternionic
geometry can be found in [Joy92, AM96a, AM96b, PPS98].
Here we recall the quotient construction of Joyce [Joy91] which we shall use
later. Let (`. I) be a hypercomplex manifold. We dene the automorphism group
Aut(`. I) of (`. I) by
(12.6.3) Aut(`. I) = Diff(`) [

1() = 1()

for all o
2
.
Since a hypercomplex structure is a G-structure of nite type, it follows that the
group Aut(`. I) is a Lie group. Let H be a Lie subgroup of H Aut(`. I).
Then H acts on ` as complex automorphisms with respect to any of the complex
structures in 1()
S
2.
Denition 12.6.6: Let (`. I) be a hypercomplex manifold. Given a compact Lie
subgroup H Aut(`. I) a hypercomplex moment map is any H-equivariant
map = i
1
j
1
+ i
2
j
2
+ i
3
j
3
: ` h

sp(1) satisfying both of the following

conditions
(i) 1
1
dj
1
= 1
2
dj
2
= 1
3
dj
3
. where 1
a
acts on sections (T

` h

).
(ii) For any non-zero element h and the induced vector eld A

(T`)
1
1
dj
1
(A

) = 0 on `.
Note that the condition (i) of 12.6.6 is equivalent to requiring that the complex
valued function j
a
+ ij
b
on a complex manifold (`. 1
c
) be holomorphic function
with respect to the complex structure 1
c
for any cyclic permutation (o. /. c) of
(1. 2. 3). Joyce proves the following [Joy91]
Theorem 12.6.7: Let (`. I) be a hypercomplex manifold H a any compact Lie
group in Aut(`. I). Choose any hypercomplex moment map and let =
1
i
1
+

2
i
2
+
3
i
3
h

sp(1). where all three

i
are in the center of h

. Suppose the
H-action on

=
1
() has only nite isotropy groups and

`() =

H is an
orbifold. Then

`() has a naturally induced hypercomplex structure.
The hypercomplex quotient construction can be used to build many examples
of hypercomplex manifolds as we shall see in Chapter 13. The main point here
is that unlike in the case of hyperkahler reduction which will be dened in the
12.7. HYPERK

AHLER MANIFOLDS 455

following sections the hypercomplex reduction is much more exible in the way one
chooses the associated moment map.
Since a hypercomplex manifold ` is quaternionic, it has a twistor space Z(`)
which satises all the properties of Theorem 12.2.16. But also in the hypercomplex
case the trivialization of Ogives a trivialization of the twistor space Z(`) = o
2
`
as smooth manifolds, but not as complex manifolds. Nevertheless, the projection j
onto the rst factor is holomorphic, and we have a double bration
(12.6.4)
Z(`)
p
, `
CP
1
`
which gives a correspondence: points o
2
CP
1
correspond to complex struc-
tures 1() on ` in the given hypercomplex structure I; points r ` correspond
to rational curves in Z(`) with normal bundle 2nO(1). called twistor lines.
12.7. Hyperkahler Manifolds
Given a an almost hyperhermitian manifold (`. I. o) we can use the metric to
dene the 2-forms
(12.7.1) ()(A. Y ) = o(1()A. Y ). A. Y (T`) .
In particular, given a hypercomplex structure I and choosing the basis 1
1
. 1
2
. 1
3

we get the three fundamental 2-forms

1
.
2
.
3
which trivializes the subbundle
O

. By analogy with the almost Kahler case consider

Denition 12.7.1: An almost hyperhermitian manifold (`. I. o) is called almost
hyperkahler if the associated fundamental 2-forms are closed and it is called hy-
perkahler (HK) if the associated 2-forms are parallel with respect to the Levi-Civita
connection of o.
Unlike the Kahler case an almost HK manifold must automatically be HK
[Hit87a]. In fact we have the following equivalent characterization of hyperkahler
manifolds
Theorem 12.7.2: Let (`
4n
. I. o) be an almost hyperhermitian manifold with the
fundamental 2-forms
a
(A. Y ) = o(1
a
A. Y ). o = 1. 2. 3. Then the following condi-
tions are equivalent
(i) (`
4n
. I. o) is hyperkahler,
(ii) (`
4n
. I. o) is almost hyperkahler,
(iii) (`
4n
. I. o) is 1-integrable.
(iv) 1
1
= 1
2
= 1
3
= 0.
(v) Hol(o) oj(n).
In particular, an HK manifold is Kahler with respect to any choice of complex
structure in 1(). and the holonomy reduction implies that HK manifolds must
be Ricci-at. Of course the 4-form =

a

a

a
is parallel so that any HK
manifold is also QK, only the quaternionic bundle O on ` is trivial and the scalar
curvature vanishes. The following diagrams describes how HK geometry relates to
other quaternionic geometries discussed in previous sections
In dimension 4 the situation is special. Since oj(1) ol(2) the HK condition
is equivalent to asking that `
4
be Kahler and Ricci-at. From the decomposition
456 12. QUATERNIONIC K

AHLER AND HYPERK

AHLER MANIFOLDS
Quaternionic Kahler
Sp(n)Sp(1)
= 0

Quaternionic
GL(n, H)H

Oproiu connection

n > 1

Hyperkahler
Sp(n)

1
=
2
=
3
=0

Hypercomplex
GL(n, H)
Obata connection
Figure 2. Quaternionic geometries in dimension 8.
of the Riemann curvature in (12.2.14) we see that the only non-zero component is
\

and such manifolds are also sometimes called half-at.

W
+
= 0 or W

= 0,
Einstein
Sp(1)Sp(1)

W
+
= 0 or W

= 0,
Conformal
R

SO(4)

n = 1

Hyperkahler
Sp(1)

Hypercomplex
H

Figure 3. Quaternionic 4-manifolds.

When `
4
is a compact HK manifold then, up to cover, it must be either a K3
surface or a at torus. On the other hand, if we do not insist on compactness the
question of the classication of such metrics remains wide open. Only partial clas-
sication results are known. All these metrics are important in General Relativity
Theory as they are vacuum solutions (Ricci-at) of the Euclidean Einstein equa-
tions. Such solutions are called gravitational instantons. They will all be described
as certain quotients in the next section.
Proposition 12.7.3: Let (`
4n
. I. o) be an HK manifold and consider the Kahler
structure (1
1
. o.
1
). The complex 2-form
+
=
2
+ i
3
is of type (2. 0) and holo-
morphic, i.e., it is a holomorphic symplectic form on `.
Proof. Let (l; .
1
. . . . . .
2n
) be a holomorphic local chart on ` with respect to
the complex structure 1
1
. Consider the 2-form
+
(A. 7) = o(1
2
A. Y ) +io(1
3
A. Y )
and extend it by linearity to the complexied tangent bundle T` C. Setting
A =

zj
we compute for any vector eld Y

+
(
z
j
. Y ) = o

1
2

.
j
. Y

+io

1
3

.
j
. Y

= io

1
2
1
1

.
j
. Y

+io

1
3

.
j
. Y

= 0 .
12.7. HYPERK

AHLER MANIFOLDS 457

since 1
2
1
1
= 1
3
which implies that
+
is a (2. 0)-form. It is holomorphic since it
is closed.
Note that the (0. 2)-form conjugate under 1
1
to
+
is

=
2
i
3
. Proposi-
tion 12.7.3 can easily be generalized to an arbitrary oriented orthonormal 3-frame

1
.
2
.
3
. namely
Proposition 12.7.4: Let (`
4n
. I. o) be an HK manifold and consider the Kahler
structure (1(
1
). o. (
1
)). The complex 2-form
+
(
1
) = (
2
) +i(
3
) is a holo-
morphic symplectic form on ` with respect to the complex structure 1(
1
).
We now consider briey the twistor space Z(`) of a hyperkahler manifold
`. For some references here see [HKLR87, Joy00, Sal86]. Since a hyperkahler
structure is hypercomplex, the twistor space Z(`) of a hyperkahler manifold sat-
ises all the properties of Theorem 12.2.16 as well as the correspondence given by
Diagram 12.6.4. But when ` is hyperkahler the complex manifold (`. 1(
1
)).
which is the bre over
1
o
2
CP
1
of the holomorphic bration j. can be viewed
as a divisor in Z(`) with a holomorphic symplectic form
+
(
1
). This gives a
twisted holomorphic 2-form c on Z(`) that is a section of j

O(2)

O

. where

is the bundle O

on ` pulled back to Z(`) under the right hand projection in

diagram 12.6.4. To express this holomorphic data explicitly we choose the standard
basis e
1
. e
2
. e
3
of R
3
. and write a point o
2
as =

a
e
a
. These standard
coordinates are related to the complex ane coordinate t C CP
1
o
2
we have
[HKLR87]
(
1
.
2
.
3
) =

1 [t[
2
1 +[t[
2
.
t +

t
1 +[t[
2
. i
t

t
1 +[t[
2

.
Then the complex structure 1

on Z(`) as an endomorphism of the tangent space

T
(x,t)
Z(`) = T
x
` T
t
o
2
becomes
(12.7.2) 1

= 1() +1
0
=
1 [t[
2
1 +[t[
2
1
1
+
t +

t
1 +[t[
2
1
2
+i
t

t
1 +[t[
2
1
3
+1
0
.
where 1
0
denotes the standard complex structure on the tangent space T
t
o
2
given
by multiplication by i. Of course, there is a similar expression for the other chart
with ane coordinate : =
1
t
centered about (1. 0. 0). Now the twisted holomorphic
2-form c on Z(`) is written as
(12.7.3) c =
+
+ 2t
1
t
2

.
There is a converse to this twistor space construction [HKLR87], but rst some
notation. Given a complex manifold Z with a holomorphic bration j : Z CP
1
.
we let T
F
denote the vertical subbundle of TZ, that is, the kernel of the dierential
j

: TZTCP
1
.
Theorem 12.7.5: Let (Z. J) be a complex manifold of complex dimension (2n+1)
equipped with the following data
(i) a holomorphic projection j : Z CP
1
.
(ii) a holomorphic section c of j

O(2)
2
T

F
which restricts to a holomor-
phic symplectic form on the bres of j.
(iii) a free antiholomorphic involution : Z Z that satises

(c) = c.
and j =o j. where o is the antipodal map on o
2
.
458 12. QUATERNIONIC K

AHLER AND HYPERK

AHLER MANIFOLDS
Let ` be the set of all rational curves C in Z with normal bundle 2nO(1) and
(C) = C. Then ` is a hypercomplex manifold with a natural pseudo-HK metric
o. If o is positive denite then ` is HK.
12.8. Hyperkahler Quotients
In this section we review the generalization of the Marsden-Weinstein construc-
tion described in Section 8.4 to HK manifolds with hyperholomorphic isometries.
Such reductions were rst considered by Lindstrom and Rocek as early as in 1983
[LR83], while investigating the so-called 4-dimensional = 2 globally supersym-
metric -model theories. It is known that target manifolds of such -models are HK.
Lindstrom and Rocek observed that one can gauge away hyperholomorphic sym-
metries. In the process one introduces auxiliary gauge elds without kinetic terms
in the Lagrangian, i.e., hyperholomorphic Killing vectors. The Euler-Lagrange
equation for such elds are algebraic (moment map equations) and eliminating
these elds leads to a new = 2 supersymmetric model theory, hence, a new
HK metric. A few years later Hitchin gave the rigorous mathematical descrip-
tion of what is now known as hyperkahler reduction [HKLR87]. We will describe
this construction and some of the basic examples as it provides a blueprint for
much of the material of the next chapter. Let (`. I. o) be an HK manifold and
G Aut(`. I. o) Isom(`. o) be a Lie group acting smoothly and properly on `
by preserving the hyperkahler structure. Then G acts by symplectomorphism pre-
serving symplectic forms
a
. o = 1. 2. 3. Suppose the G-action is Hamiltonian with
respect to each symplectic form
a
. We will call such an action hyperhamiltonian.
Denition 12.8.1: A hyperkahler manifold (`. I. o) together with an eective
hyperhamiltonian G-action is called a hyperhamiltonian G-manifold.
As discussed in section 8.4.1 such an action gives rise to three G-equivariant
symplectic moment maps j
a
: ` g

. We can assemble these maps together to

get
Denition 12.8.2: Let (`. I. o) be a hyperkahler G-manifold. The map =
(j
1
. j
2
. j
3
) = i
1
j
1
+i
2
j
2
+i
3
j
3
(12.8.1) = ` g

sp(1)
is called the hyperkahler moment map for the action of G.
We have the following natural generalization of the Marsden-Weinstein sym-
plectic reduction theorem [HKLR87]
Theorem 12.8.3: Let (`. I. o) be HK and G be a hyperhamiltonian action on `
with the HK moment map : ` g

sp(1). Let = (
1
.
2

3
) g

sp(1) be
any element xed by the co-adjoint action of G on its Lie co-algebra g

. Suppose
is a regular value of so that =
1
() ` is a manifold. Suppose further
that the orbit space

`() =
1
()G is a manifold (orbifold). Then

`() is an
HK manifold (orbifold) with the HK structure induced from ` via inclusion and
projection maps.
Proof. We only sketch proof here. The manifold (`. 1
1
.
1
. o) is a Kahler
manifold. The HK reduction can be seen as a two step process: First, we consider
the function
(12.8.2) j
+
= j
2
+i
1
j
3
.
12.8. HYPERK

AHLER QUOTIENTS 459

which is easily seen to be holomorphic on (`. 1
1
. o). Thus the set
+
= j
1
+
(0) is a
complex subspace of (`. 1
1
. o). in particular it must be Kahler. Note that
+
need
not be a smooth manifold, it is sucient that it be smooth in some H-invariant
open neighborhood
t
+
such that
t
+

+
. The action of G restricts to

+
with the Kahler moment map j
1
:
+
g

. Hence, the reduced space

`
is nothing but a Kahler reduction of
+
by the action of G. In particular,

` is
Kahler with the complex structure

1
1
induced from ` by the quotient construction.
Now, the result follows by observing that the same argument applies to 1
2
and 1
3
.
and

` is therefore Kahler with respect to all three complex structures

1
1
.

1
2
.

1
3
.
One can easily check that the induced complex structures satisfy the quaternionic
relations.
We remark that, just as in the symplectic case, one can consider more general
singular quotients. This was done by Dancer and Swann [DS97a, Swa97] who
showed
Theorem 12.8.4: Let (`. I(). o) be a hyperhamiltonian G-manifold with the
moment map : `g

sp(1). Furthermore, suppose G acts smoothly and

properly on `. Let `
(H)
denote the stratum consisting of orbits of type H < G.
Then
H
=
1
(0) `
(H)
is a manifold and the orbit space

`
H
=

1
(0) `
(H)
G
has a natural HK structure. Consequently, the reduced space

` =
1
(0)G is a
disjoint union of HK manifolds

` =

H<G

`
H
.
The proof is a corollary of the Sjamaar-Lerman Theorem 8.4.3. In particular,
each stratum is a smooth HK manifold but unlike the symplectic case it is not clear
whether the quotient is decomposable in the sense of Goretsky-MacPherson.
The rest of this section is devoted to introducing some basic examples of HK
reduction of the at model H
n
. We will often use the pair = (j
1
. j
+
) to describe
the moment map in complex coordinates on (`. 1
1
.
1
. o) relative to 1
1
. Since we
single out 1
1
we will also use i. ,. / for the quaternions i
1
. i
2
. i
3
.
Example 12.8.5: [Calabi metrics on T

CP
n
] Let u = z +w, H
n
C
n
C
n
and consider the diagonal action of G = o
1
oj(n) given by left multiplication
with o(t) = c
it
. The moment map for this action is
(12.8.3) (u) = ij
1
+j
+
, =
n

k=1
n
k
in
k
.
In complex charts we get
(12.8.4) j
1
(w. z) = i
n

j=1
([n
j
[
2
[.
j
[
2
) . j
+
(w. z) = 2i

j
n
j
.
j
.
and the circle action reads (w. z) (c
it
w. c
it
z). One could consider an arbitrary
level set of the moment map. However, as oj(1)
+
is a symmetry of the at HK
metric, one can use it to choose the value of to be a constant multiple of i.
460 12. QUATERNIONIC K

AHLER AND HYPERK

AHLER MANIFOLDS
Further scaling the metric shows that it is sucient to consider =
1
(i) and

+
= j
1
+
(i) which are described as
(12.8.5) =

(w. z) C
n
C
n
[
n

j=1
([n
j
[
2
[.
j
[
2
) = 1.

j
n
j
.
j
= 0

.
(12.8.6)
+
=

(w. z) C
n
C
n
[

j
n
j
.
j
= 0

.
Let

` =
1
(i)o
1
. We rst want to identify

` with the Kahler reduction of
+
(or a G-invariant open set
t
+

+
). For an appropriate choice of
t
+
its Kahler
reduction
+
will be an algebraic quotient of
t
+

+
by the complexication C

of G = o
1
. Note, however, that
t
+
is not compact so we cannot rely on Kirwans
theorems in this setting[Kir84]
2
. Nevertheless, we get the following identication
(12.8.7)

` =
1
(i)o
1

t
+
C

.
where
t
+
= (w. z) C
n
C
n
[

j
n
j
.
j
= 0. w = 0. and C

acts by
(w. z) (w.

z). Thus,

` is the holomorphic cotangent bundle T

CP
n1
. It
turns out that the HK metric obtained on T

CP
n1
via this reduction is isometric
to the Calabi metric [Cal79], the rst non-at example of a complete HK manifold.
An = 2 supersymmetric -model description of the metric is due to Lindstrom
and Rocek [LR83] and, in the above language it appears in [HKLR87]. An explicit
expression for this metric in dimension 4 was discovered by Eguchi and Hanson
[EH79] and it is called the Eguchi-Hanson gravitational instanton.
Example 12.8.6: This example involves a non-compact hyperhamiltonian group
action of G = R on HH
n
H
n+1
dened for any p = (j
1
. . . . . j
n
) R
n
by
(12.8.8)
p
(n
0
. n
1
. . . . . n
n
) = (n
0
t. c
p
1
it
n
1
. . . . . c
p
n
it
n
n
) .
with the moment map
(12.8.9)
p
(u) = ij
1
+j
+
, = 2Im(n
0
) +
n

k=1
ij
k
n
k
in
k
.
We can always shift to the zero-level set and then the moment map equations can
be solved by writing
(12.8.10) Im(n
0
) =
n

k=1
j
k
n
k
in
k
.
This action is free and proper on H H
n
. In particular it is free and proper on

1
p
(0) =
p
. Denote the quotient manifold by `(p) =
1
p
(0)R. First, note
that `(p) is dieomorphic to H
n
. This follows from the observation that the set
o = u
1
p
(0) [ Re(n
0
) = 0 is a global slice for this action. The induced
HK metric o(p) can easily be calculated and o(0) = o
0
is the at metric. In
dimension 4. this metric o() depends on one parameter and when = 0 we get
`(0) isomorphic to H C
2
with the standard at metric. Hence, (`(). o())
is a smooth 1-parameter family of HK deformations of the Euclidean metric. The
2
This is not a special feature of this example. On the contrary, this is what typically happens
with HK reductions of H
n
by compact hyperhamiltonian G-actions.
12.9. TORIC HYPERK

AHLER METRICS 461

metric o() is called the Taub-NUT
3
gravitational instanton and it has interesting
history. Just as the famous Schwarzschild metric, or Kerr solution, the metric
appears rst in the Lorenzian signature. One can always perform the so-called
Wick rotation changing t it which locally gives a Riemannian metric with
similar properties. However, there is no reason for the Riemannian metric to extend
globally to a complete metric on some manifold. This is fairly rare and happens
only in special situations. It was Hawking who observed that this indeed is the
case for the Lorenzian Taub-NUT solution, giving rise to a complete Ricci at
metric on R
4
[Haw77]. For some time thereafter the metric was not really fully
understood, there being claims in the literature is that this metric was not Kahler.
We should point out that this is the only known complete Ricci-at Kahler metric
on C
2
apart from the standard one [LeB91a]. If one imposes a Euclidean volume
growth condition the only known example of such complete Ricci-at Kahler (or
just Ricci-at) metric on C
2
is the Euclidean metric.
12.9. Toric Hyperkahler Metrics
It is easy to see that previous two examples fall into a special category of com-
plete hyperkahler manifolds: they admit n commuting hyperholomorphic Killing
vector elds, where n is the quaternionic dimension. Following Bielawski and
Dancer we consider [BD00, Bie99]
Denition 12.9.1: An HK manifold (`
4n
. I. o) is locally toric if it admits
n commuting hyperholomorphic Killing vector elds, linearly independent at each
point r `
4n
. i.e., locally ` admits a free action of R
n
by hyperholomorphic
isometries. Furthermore, (`
4n
. I. o) is said to be a toric HK manifold if it is a
hyperhamiltonian T
n
-space.
We emphasize that toric is understood here to be in the quaternionic sense.
A local description of such metrics in dimension 4 is due to Gibbons and Hawking
[GH78a] and in arbitrary dimension 4n it has been generalized by Lindstrom and
Rocek [LR83]. The so-called Legendre transform method developed by Lindstrom
and Rocek associates a 4n-dimensional hyperkahler metric with n commuting hy-
perholomorphic Killing vectors to every real-valued function 1 on an open subset
| R
3
R
n
which is harmonic on any ane 3-dimensional subspace 1 of the
form R
3
Rv, v R
n
(such functions are sometimes called polyharmonic). The
construction proceeds as follows: Let us identify R
3
R
n
with R
n
C
n
and let
(x. z) R
n
C
n
be coordinates on |. Given any polyharmonic function 1(z. z) on
| we consider a real-valued function
(12.9.1) 1(u. u. z. z) = 1(x. z. z) 2
n

i=1
(n
i
+ n
i
)r
i
.
where the r
i
are determined by
(12.9.2)
1
r
i
= 2(n
i
+ n
i
) .
It is an elementary exercise to check that polyharmonicity of 1 turns 1 into a
Kahler potential of an HK metric. Furthermore, if we set y = i( u u) then
3
The acronym NUT stands for Newmann-Unti-Tamburino and has become standard termi-
nology for describing a certain type coordinate singularity of the Einstein equations.
462 12. QUATERNIONIC K

AHLER AND HYPERK

AHLER MANIFOLDS
A
i
= n
i
. i = 1. . . . . n yield n commuting hyperholomorphic Killing vector
elds with the corresponding hyperkahler moment maps
(12.9.3)
i
= (j
i
1
. j
i
+
) = (r
i
. .
i
) . i = 1. . . . . n.
One can show that, relative to local coordinates (y. x. z). the HK metric takes
the form [PP88]
(12.9.4) o =

i,j

ij
(dr
i
dr
j
+d.
i
d .
j
) + (
1
)
ij
(dn
i
+
i
)(dn
j
+
j
)

.
where
ij
=
1
4
1
xixj
and
j
=

1
2

l
(1
xj z
l
d .
l
1
xjz
l
d.
l
). The functions
ij
are also polyharmonic. The n n matrix [
ij
] locally determines the hyperkahler
and hyperhamiltonian structure. When n = 1 this is well-known Gibbons-Hawking
Ansatz. Renaming (n
1
. r
1
. .
1
) = (t. r
1
. r
2
+ir
3
) = (t. x).
1
= dx.
11
= \ we
can write the metric in a more familiar form
(12.9.5) o = \ (dx dx) +\
1
(dt + dx)
2
.
where gradV = curl. In particular, \ is a solution of the Laplace equation so that
we can write
(12.9.6) \ (r
1
. r
2
. r
3
) = +
k

i=1
:
i
[x a
i
[
.
When = 0 these metrics are called /-center gravitational multi-instantons. The
rst two values / = 1. 2 correspond to the Euclidean and Eguchi-Hanson metrics,
respectively. For larger values of / it is not easy to determine when the metric
is actually complete and even harder to see what the manifold `
k
on which it is
dened is. When = 1 we get the so-called /-center Taub-NUT gravitational multi-
instantons with / = 1 corresponding to the Euclidean Taub-NUT metric discussed
in Example 12.8.6.
The two basic examples of this construction are at o
1
-invariant metrics on
o
1
R
3
and on H. In the rst case we have
(12.9.7) 1(r. .. .) = 2r
2
. .
and, consequently, 1. while in the second case
(12.9.8) 1(r. .. .) = rln(r +:) : .
where :
2
= r
2
+ . . with = 14:. More general forms are given in [BD00]. In
the latter, the functions 1 and the metrics for hyperkahler quotients of at vector
spaces are computed. They are essentially obtained by taking linear combina-
tions and compositions with linear maps of the solution (12.9.8). Bielawski shows
[Bie99] that, in the case of a complete metric, the only other possibility is adding
a linear combination of (12.9.7), which corresponds to a Taub-NUT deformation of
Denition 12.9.4.
For an HK metric of the form (12.5.6) taking hyperkahler quotients by subtori is
simple. The moment map equations are now linear (in r
i
. .
i
), and the hyperkahler
quotient corresponds to restricting the function 1 to an appropriate ane subspace
of R
3
R
n
. In fact, the requirement that 1 be polyharmonic is a consequence of
the fact that we must be able to take hyperkahler quotients by any subtorus.
An explanation of this construction in terms of twistors was given by Hitchin,
Karlhede, Lindstrom and Rocek [HKLR87]. In particular, they have shown that
12.9. TORIC HYPERK

AHLER METRICS 463

any hyperkahler 4n-manifold with a free hyperhamiltonian R
n
-action which extends
to a C
n
-action with respect to each complex structure and such that the moment
map is surjective is given by the Legendre transform. In fact, one can show [Bie99]
that the Legendre transform provides a complete local description of such metrics,
i.e.,
Proposition 12.9.2: Let (`
4n
. o) be a locally toric HK manifold. Then o is locally
given by Equation 12.9.4.
The key to the understanding of the global properties of such metrics is the
HK quotient construction. The relevant spaces were rst introduced in [HKLR87],
but the global properties of such metrics were studied only much later by Bielawski
and Dancer [BD00] culminating in a complete classication result by Bielawski
[Bie99].
We will discuss this classication here as it will be important in the next chap-
ter. First, we would like to identify hyperhamiltonian G-manifolds which are the
same in the sense of the following denition.
Denition 12.9.3: Let `. `
t
be two hyperhamiltonian G-manifolds and let ,

be the chosen moment maps. We say that ` and `

t
are isomorphic as hy-
perkahler G-manifolds, if there is a hyperholomorphic G-equivariant isometry 1 :
` `
t
such that =
t
1.
Secondly, there is a natural relation between the at metric and the Taub-NUT
metric on H. This is, however, only an example of a more general correspondence.
The construction of the Example 12.8.6 suggest the following denition.
Denition 12.9.4: Let `
4n
be a connected complete HK manifold of nite topo-
logical type with an eective hyperhamiltonian action of G = R
p
T
np
. A Taub-
NUT deformation (of order :) of ` is the hyperkahler quotient of ` H
m
by
R
m
. where R
m
acts on ` via an injective linear map : R
m
Lie (T
n
) = R
n
.
Note that such a deformation `
t
is canonically T
n
-equivariantly dieomorphic
to ` by a dieomorphism 1 which respects the hyperkahler moment maps j. j
t
.
i.e., j = j
t
1. Bielawski shows that up to G-equivariant isometry and Taub-NUT
deformations one can restrict attention to the following HK quotients studied in
detail in [BD00]. We consider a T
k
T
m+1
oj(:+1) action on H
m+1
dened
via exact sequence 0 h

R
m+1

R
m+1k
0. and its dual, as in
(12.5.2, 12.5.3), and with :+1 / = n. This is a hyperhamiltonian T
k
-action for
any choice of the weight matrix
k,m+1
(Z) or, alternatively,
n,m+1
(Z).
We can consider the HK moment map for this action j

: H
m+1
R
3
R
k
(12.9.9) j

(u) =
m+1

=1
( n

in

+c .
where
(12.9.10) c =
m+1

=1

and

=
1

i +
2

, +
3

/ are purely imaginary quaternions (or vectors in R

3
) for
= 1. . . . . : + 1. In particular, c R
3
R
k
is simply an arbitrary choice of the
constant in the denition of the HK moment map. Let us denote by the moment
464 12. QUATERNIONIC K

AHLER AND HYPERK

AHLER MANIFOLDS
level set data, i.e.,

1
i +
2
, +
3
/ = (
1
. . . . .
m+1
) R
3
R
m+1
.
Consider now the HK quotient space, i.e.,
(12.9.11) `(. c) `(. ) = j
1
(0)T
k
.
Clearly, the HK quotient `(. ) locally inherits a hyperhamiltonian T
n
-
action. However, it can be singular and the stratication depends on the choice of
the quotient data. Note that either choosing (. ) or (. c) completely determines
both the T
k
action and the associated HK quotient. In fact, these two descriptions
are dual to one another and they are both useful. Consider the following codimen-
sion three ane subspaces in R
3
R
n
:
(12.9.12) H

= (r
1
. r
2
. r
3
) R
3
R
n
; 'r
i
.

` =
i

. i = 1. 2. 3 .
= 1. . . . . :+ 1 and

R
n
is the
th
column of .
Bielawski and Dancer prove the following theorem [BD00]
Theorem 12.9.5: Suppose the column vectors

of
n,m+1
(Z) are primitive
integer vectors spanning R
n
. Suppose R
3
R
m+1
is such that H

are all distinct.

Then the HK quotient `(. ) is smooth if and only if
(i) every collection of n + 1 of the H

have empty intersection, and

(ii) whenever some n of the H
k
1
. . . . . H
k
n
have non-empty intersection, then
the set
k
1
. . . . .
k
n
is a Z-basis for Z
n
.
The rst condition is sucient for `(. ) to be an HK orbifold with at worst
Abelian quotient singularities.
If condition (i) of Theorem 12.9.5 holds the orbifolds `(. ) and `(.

) are
homeomorphic, and if

= (
1
. 0. 0). they are dieomorphic. In particular, one can
always set
2
=
3
= 0 while investigating the topology of such quotients. Bielawski
and Dancer have given a formula for the Betti numbers of `(. ) in terms of
arrangements of certain hyperplanes. Consider the collection of hyperplanes
H
1

= y R
n
; 'r.

` =
1

. i = 1. 2. 3 .
i.e., H
1

in R
n
dened by restricting H

to the rst coordinate in R

3
factor. These
hyperplanes divide R
n
into a nite family of closed convex polyhedra. Let / be
the polyhedral complex consisting of all faces of all dimensions of these polyhedra
and let ( be the polyhedral complex consisting of all bounded polyhedra in /. We
have the following
Theorem 12.9.6: Let ` = `(. ) be a toric HK orbifold of dimension 4n.
(i) Then ` is simply connected and H
j
(`. Q) = 0 for , odd;
(ii) /
2p
(`) =

n
i=p
(1)
ip

i
p

d
i
. where the integer d
i
denotes the number of
i-dimensional elements of the complex (;
(iii) if ` = `(. ) is a smooth manifold, there is a ring isomorphism
H

(`. Z) Z[n
1
. . . . . n
N
]. where n
i
are the rst Chern classes of certain
complex line bundles on `.
Parts (i) and (ii) of this theorem are due to Bielawski and Dancer [BD00]
while (iii) is due to Konno [Kon00]. In [Bie99] it is shown that the quotients
` = `(. ) are essential part of the classication of locally toric HK manifolds.
On can show that
12.10. ALE SPACES AND OTHER HYPERK

AHLER QUOTIENTS 465

Theorem 12.9.7: Let `
4n
be a connected complete hyperkahler manifold of nite
topological type with an eective hyperhamiltonian action of G = R
p
T
np
. Then
(i) If ` is simply connected and j = 0. then ` is isomorphic, as a hyper-
hamiltonian hyperkahler T
n
-manifold, to a hyperkahler quotient of some
at H
d
H
m
. : n. by T
dn
R
m
.
(ii) If ` is simply connected and j 0. then ` is isomorphic, as a hyper-
hamiltonian hyperkahler G-manifold, to the product of a at H
p
and a
4(n j)-dimensional manifold described in part (i).
(iii) If ` is not simply connected, then ` is the product of a at (o
1
R
3

l
.
1 | n. and a 4(n |)-dimensional manifold described in part (ii).
In the case of dimension four we can be more specic.
Corollary 12.9.8: Let ` be a simply connected 4-dimensional complete hyperkahler
manifold with a non-trivial hyperhamiltonian vector eld. If /
2
(`) = / 0, then
` is isometric either to an ALE-space of type
k
(i.e., a multi-Eguchi-Hanson
space) or to its Taub-NUT-like deformation (i.e., to the hyperkahler quotient by R
of the product of such a space with H). If /
2
(`) = 0. then ` is either the at H
or it is the Taub-NUT metric on R
4
.
In particular, Bielawski concludes the following HK analogue of the Delzants
theorem for complex toric manifolds:
Theorem 12.9.9: Complete connected HK hyperhamiltonian T
n
-manifolds of -
nite topological type and dimension 4n are classied, up to Taub-NUT deforma-
tions, by arrangements of codimension 3 ane subspaces H

in R
3
R
n
dened as
in (12.9.12) by
H

= (r
1
. r
2
. r
3
) R
3
R
n
; 'r
i
.

` =
i

. i = 1. 2. 3
for some nite collection of vectors

in R
n
and scalars
i

, i = 1. 2. 3. such that,
for any j R
3
R
n
. the set

; j H

is part of a Z-basis of Z
n
.
We end this section by briey mentioning some fascinating recent work by
Hausel, Nakajima, Strurmfels and others that relates hyperkahler geometry, and
toric hyperkahler structures in particular to combinatorics and representation the-
ory [HS02, Nak98, Nak99] as well as to the famous ADHM construction of in-
stanton moduli spaces [BM93b, Nak99, Hau06] and to number theory [Hau05,
Hau06].
12.10. ALE Spaces and Other Hyperkahler Quotients
In this last section we shall describe some other examples of hyperkahler met-
rics with particular focus on examples relevant to 3-Sasakian geometry. Our rst
goal is to describe quotient construction of ALE spaces discovered by Kronheimer
[Kro89a, Kro89b].
12.10.1. Classical McKay Correspondence. We begin by recalling some
elementary facts about discrete subgroups of ol(2), their representation, and the
classical McKay correspondence. These were already discussed in the proof of
Theorem 10.1.5, where they were denoted by: Z
n
. D

n
. T

. O

. I

. Let H = C
2
be
the standard complex 2-dimensional representation of ol(2). In particular, H gives
a representation of each ol(2). Let
0
.
1
. . . . .
r
be the set of irreducible
466 12. QUATERNIONIC K

AHLER AND HYPERK

AHLER MANIFOLDS
representation of with
0
the trivial representation. Then
(12.10.1)
i
H =
r

i=0
o
ij

j
.
McKay observed that there is a relation between the matrix = (o
ij
), whose
entries are all either 0 or 1, is the adjacency matrix of the extended and Dynkin
diagram of type A-D-E. If

C is the Cartan matrix of the extended Dynkin diagram
then +

C = 21l
r+1
. Let
1
. . . . .
r
be the simple roots of the root system of the
associated Lie algebra. Let
0
be the negative of the highest root. McKay further
noticed that

0
=
r

1
n
i

i
.
where n
i
is the dimension of the representation
i
[McK80, McK81]. The regular
representation of decomposes as
=

i
C
n
i

i
.
Separately for each . all of this information can now be encoded in a labelled
extended simply-laced Dynkin graph.
Z
n+1

A
n
. n 0 :
'&%$ !"#
1
@A BC
'&%$ !"#
1
'&%$ !"#
1
'&%$ !"#
1
D

n

D
n
. n 4 :
'&%$ !"#
1

'&%$ !"#
1
'&%$ !"#
2
'&%$ !"#
2
'&%$ !"#
2
'&%$ !"#
2
.
.
.

'&%$ !"#
1
.
.
.
'&%$ !"#
1
T

E
6
:
'&%$ !"#
1
'&%$ !"#
2
'&%$ !"#
1
'&%$ !"#
2
'&%$ !"#
3
'&%$ !"#
2
'&%$ !"#
1
O

E
7
:
'&%$ !"#
2
'&%$ !"#
1
'&%$ !"#
2
'&%$ !"#
3
'&%$ !"#
4
'&%$ !"#
3
'&%$ !"#
2
'&%$ !"#
1
I

E
8
:
'&%$ !"#
3
'&%$ !"#
2
'&%$ !"#
4
'&%$ !"#
6
'&%$ !"#
5
'&%$ !"#
4
'&%$ !"#
3
'&%$ !"#
2
'&%$ !"#
1
The vertices of the diagram correspond to the irreducible representations
i
with the
numbers in each vertex giving the dimension of that representation n
i
. The usual
Dynkin diagram is obtained from the extended one by removing one vertex which
corresponds to the trivial representation
0
. In particular, McKays observations
show that
r

i=0
n
2
i
= [[.

ij
2o
ij
n
i
n
j
= 4[[ .
12.10. ALE SPACES AND OTHER HYPERK

AHLER QUOTIENTS 467

Example 12.10.1: Consider the example of = Z
n+1
. Let
n+1
= 1. Any irre-
ducible representation of Z
n+1
is one-dimensional and
i
(r) =
i
r. r C. Now, the
two-dimensional representation H gives H()(r. n) = (r. n). Clearly, H =
1

n
so that
i
H =
i+1

i1
. The vertex
i
of the McKay graph is joined by an
edge to the vertices
i+1
and
i1
. This denes the extended Dynkin diagram

A
n
.
Exercise 12.1: Consider the example of = D

n
. As a subgroup of ol(2) the
binary dihedral group
4
is generated by two matrices
(12.10.2)

0 1
1 0

0
0

.
2(n2)
= 1 .
Note that for n = 3 we simply get the cyclic group Z
4
. This gives the represen-
tation H. In particular, D

4
is the group of quaternions Q = 1. i. ,. /.
D

4
has three non-trivial one-dimensional irreducible representation
1
.
2
.
3
and
one two-dimensional representation
4
. Derive the McKays graph for D

4
as in
Example 12.10.1. Repeat this for any binary dihedral group D

n
.
12.10.2. Geometric McKay Correspondence and Kleinian Singulari-
ties. McKays observation is closely related to the algebraic geometry of Kleinian
singularities which relation we shall explain next. This is often referred to in the
literature as the geometric McKay correspondence.
Denition 12.10.2: For oj(1) a nite subgroup, the quotient variety A =
C
2
= SpecC[r. n]

is called a Kleinian singularity (also known as a simple

surface singularity, or a rational double point, or A-D-E type singularity).
The quotient can be embedded as a hypersurface A C
3
with an isolated
singularity at the origin with the dening equation 1(.
0
. .
1
. .
2
) = 0 determined by
the conjugacy class of . These polynomials have already appeared in the table of
Remark 10.1.1. Suppose now : `A = C
2
is a crepant resolution. Then the
divisor =
1
(0) is the dual of the associated Dynkin diagram in the following
sense: the vertices of the Dynkin diagram correspond to rational curves 1
i
with
self-intersection 2. Two curves intersect transversally at one point if and only if
the corresponding vertices are joined by an edge in the Dynkin diagram. Otherwise
they do not intersect. The collection of these curves 1
1
. . . . . 1
r
forms a basis for
H
2
(`. Z). The intersection form with respect to this basis is the negative of the
Cartan matrix C.
12.10.3. Kronheimer-McKay Correspondence and Hyperkahler ALE
Spaces. Hitchin observed that in the case of = Z
n
the crepant resolution
: `C
2
Z
n
admits a family of complete hyperkahler metrics [Hit79]. In
fact, locally these metrics are produced via the Gibbons-Hawking Ansatz. Using
twistor methods Hitchin showed that Gibbons-Hawking gravitational instantons,
as hyperkahler ALE spaces, are the minimal resolution of singularity C
2
Z
n
. In
particular, the minimal resolution of the singularity C
2
Z
2
is the cotangent bun-
dle T

CP
1
and the hyperkahler metric is the Eguchi-Hansom metric. Hitchin then
conjectured that such metrics should exist for all other spaces : `C
2
. It
was only after the discovery of the hyperkahler reduction and description of its
mathematical foundations in [HKLR87] that the conjecture was nally proved by
Kronheimer [Kro89a, Kro89b]. Kronheimer generalizes the quotient construction
4
Recall that our notation is not completely standard. Our binary dihedral group D

n
=
Z
2(n2)
Z
2
has order 4(n 2) and not 4n.
468 12. QUATERNIONIC K

AHLER AND HYPERK

AHLER MANIFOLDS
described in [HKLR87] for = Z
n
to the arbitrary group oj(1). Again, the
absolute key is the McKay correspondence. Not surprisingly, the quotient can be
completely described by the extended Dynkin diagram associated to . In a way,
Kronheimers quotient construction is the third McKays correspondence. To
make our statement more precise late us begin with a precise denition of an ALE
space.
Denition 12.10.3: Let oj(1) and let (H. I
+
. o
0
) be the standard at (left)
hyperkahler structure on H dened in Section 12.1. Let : : HR
0
be the
radius function on H. We say that an HK manifold (`. I. o) is Asymptotically
Locally Euclidean (ALE), and asymptotic to H, if there exists a compact
subset A ` and a map : ` ` AH that is a dieomorphism between
` ` A and r H[:(r) 1 for some 1 0 such that

k
(

(o) o
0
) = O(:
4k
) .
k
(

(1()) 1

0
()) = O(:
4k
)
as : and / 0. where is the Levi-Civita connection of the at metric o
0
.
Consider any extended Dynkin diagram

(). With each vertex of

() we
associate the unitary group l(n
i
) and with each edge the vector space H
ninj
=
Hom(C
ni
. C
nj
) Hom(C
nj
. C
ni
). One can think of each edge as the vector space

ni,nj
(H) of quaternionic matrices. For each , we dene
(12.10.3) H
]]
=

i,j
o
ij
Hom(C
n
i
. C
n
j
) . G() = l(n
0
) l(n
r
) .
The group G() acts naturally on H
]]
and the action is hyperhamiltonian
with appropriately dened at hyperkahler structure on each edge. However, the
action is not eective, hence, we take the quotient 1() = G()T. where T is
the central l(1) G(). The action of 1() is then eective and it denes the
hyperkahler moment map
(12.10.4) j

: H
]]
k

sp(1) .
Denition 12.10.4: We say that 1()-invariant element k

sp(1) is in a
good set if the 1()-action on the -level set j
1

() of the moment map is free.

The notion of the good set is generic. The set 7 of 1()-invariant elements in
k

can be identied with the dual of the center. Kronheimer shows that is not
in the good set if 1

sp(1) 7 sp(1). where 1

are the walls of the Weyl

chamber. The reduction gives the following theorem [Kro89a, Kro89b]
Theorem 12.10.5: Let k

sp(1) be a G()-invariant element. Let `(. )

be the hyperkahler reduction of H
]]
by the action of 1() and with the momentum
level set . Then `(. 0) C
2
and `(. ) is a hyperkahler orbifold for any .
In addition, when is in the good set j
1

() the smooth manifold `(. ) gives a

family of complete ALE hyperkahler metrics on the crepant resolution of singularity
C
2
.
When is not in the good set, the HK orbifold `(. )C
2
is a partial
resolution of the quotient singularity. In [Kro89b] Kronheimer shows that his
construction is also complete. That is any hyperkahler ALE space of Denition
12.10.3 is obtained as such quotient. The quotient metrics are known in local charts
for = Z
n
(the Gibbons-Hawking Ansatz, see (12.9.5)) as they always have non-
trivial isometries. For the non-Abelian the ALE metrics have no Killing vectors.
12.10. ALE SPACES AND OTHER HYPERK

AHLER QUOTIENTS 469

Recently, Cherkis and Hitchin gave explicit formulas for the ALE gravitational
instantons in the binary dihedral case [CH05].
Remark 12.10.1: The Kleinian singularities and discrete groups of ol(2) also give
one-to-one correspondence with all compact 3-Sasakian manifolds in dimension 3.
We shall discuss this in the next chapter.
12.10.4. Other Hyperkahler Metrics. In the last 25 years hyperkahler ge-
ometry has become an important eld of Riemannian geometry. Already over a
decade ago, in a Seminaire Bourbaki review article Hitchin points out that the
richness of the theory of hyperkahler manifolds, in some sense, vindicates Hamil-
tons conviction that quaternions should play a fundamental role in mathematics
and physics [Hit92]. As it happens, many new ideas in this eld have came from
mathematical physics. In this chapter we have only covered a small number of
selected topics most relevant to the material of Chapter 13. We would like to end
it with a brief discussion of several subjects we were unable to introduce. We re-
fer the interested reader to several books and review articles about the subject
[Hit87b, AH88, Hit95a, Dan99, VK99].
[Instantons, Monopoles, and Stable Pairs] Many hyperkahler metrics emerge
in the description of the geometry of various moduli spaces. In 1983 Atiyah and
Bott made a fundamental observation that the moduli space of self-dual Yang-Mills
connections over a Riemann surface can be described as an innite dimensional
Kahler quotient [AB83]. In such picture one rst equips the space of all connec-
tions with a structure of an innite dimensional Kahler manifold. On it acts an
innite dimensional group of gauge transformations and the moment map for this
action is precisely the self-duality equation. Hence, the Kahler quotient, which
in turns out to be nite-dimensional, is the space of self-dual Yang-Mils connec-
tions modulo the gauge equivalence. When a Riemann surface is replaced by a
four-dimensional manifold (such as R
4
or o
4
. for example) the space of connections
can be given a structure of an innite-dimensional hyperkahler manifold. The HK
moment map for the gauge group action produces the self-duality equations and
the HK quotient can be naturally identied with the moduli space of instantons.
This picture explains why the /-instanton moduli spaces over o
4
carry a natu-
ral HK structure [AHDM78]. Yang-Mills connections on other 4-manifolds were
studied since. For instance, Kronheimer and Nakajima considered instantons over
the ALE spaces discussed earlier [KN90, Nak90]. The innite dimensional HK
quotient picture is inherently present in a variety of dierent moduli problems. One
important case is the moduli space of solutions of the Bogomolny equations on a
3-manifold. These are known as monopoles. They can be viewed as Yang-Mills con-
nections which a translational symmetry. Many complete hyperkahler metrics have
been constructed as solutions of the Bogomolny equations or the related Nahms
equations [Nah82, AH88, Dan93, Dan94]. For instance, in the ol(2) case,
the universal cover

`
0
k
of the moduli space of charge / monopoles with a xed
center is a complete hyperkahler manifold of dimension 4(/1) [AH88]. The four-
dimensional case of

`
0
2
gives the famous oO(3)-invariant Atiyah-Hitchin monopole
metric [AH85]. Like in the Taub-NUT case the oO(3)-action acts by rotations on
the 2-sphere of complex structures I but the metric is quite dierent as it is not
toric: it has no hyperhamiltonian Killing vectors. Finally, we mention the mod-
uli space of R
2
-invariant Yang-Mills equation. Such reduction naturally leads to
470 12. QUATERNIONIC K

AHLER AND HYPERK

AHLER MANIFOLDS
the moduli space of the so-called stable pairs or Yang-Mills-Higgs elds over an
arbitrary Riemann surface and was considered by Hitchin in [Hit87b].
[Hyperkahler Manifolds of Type

and 1

] The rst example of a com-

plete hyperkahler manifold of innite topological type was obtained by Anderson,
Kronheimer and LeBrun [AKL89]. They showed that one can take / limit
in the Gibbons-Hawking Ansatz (12.9.5) under an appropriate assumption about
the distribution of the mass centers. Later these metrics were considered from an
algebraic viewpoint by Goto who also showed that a similar limit can be taken in
the D

n
ALE case [Got98].
[Hyperkahler Deformations of ALE Spaces] Taub-NUT deformations of the
ALE gravitational instantons corresponding to = Z
n
were discussed in Section
12.9. These metrics are no longer ALE but the are asymptotically locally at.
Other ALE spaces do not admit any Taub-NUT deformations in the sense of Def-
inition 12.9.4 as they have no hyperholomorphic isometries. But the ALE spaces
of the binary dihedral group admit deformations similar to Taub-NUT deforma-
tions. Recall, that we can view the Taub-NUT deformation of the Z
n
ALE space as
follows: consider the space H

and replace one H-factor with the at hyperkahler

manifold o
1
R
3
modifying the action of 1() to be the translation on o
1
. Con-
sider a non-Abelian and suppose we could replace the Euclidean metric on some
edge H
n
i
n
j
with a non-Euclidean hyperkahler metric on H
n
i
n
j
which, however, ad-
mits hyperhamiltonian action of l(n
i
) l(n
j
). Dancer observed that there is one
such case: a complete l(2) l(2)-invariant hyperkahler metric on H
4
considered
as T

G(2. C) [Dan93]. The metric is obtained as a monopole moduli space by

solving Nahms equations. Dancer shows that replacing the at H
4
with one copy
of ` = T

G(2. C) together with its monopole metric and then performing the HK
quotient of Kronheimer gives non-trivial deformations of the ALE metrics for each
D

n
. n 4. It is not clear if similar deformations exist for any of the ALE spaces of
the E-series.
[Hyperkahler Metric on Coadjoint Orbits] Kronheimer showed that there
is a natural hyperkahler metric on a regular semisimple coadjoint orbits of a com-
plex group Lie group G
C
[Kro90a]. Kronheimer also proved that nilpotent orbits
admit hyperkahler structure [Kro90b]. Later Biquard [Biq96] and independently
Kovalev [Kov96] showed that there is a hyperkahler structure on any coadjoint
orbit of G
C
.
[Hyperkahler Metrics on Cotangent Bundles] The rst explicit non-trivial
example of a hyperkahler metric is the Calabi metric on the cotangent bundle
T

CP
n
. Already in 1983 Lindstrom and Rocek constructed a hyperkahler met-
ric on the cotangent bundle T

Gr
n
(C
n+m
) which is realized as a reduction of the
at space H
m(n+m)
by a hyperhamiltonian action of l(:) [LR83]. In fact, Lind-
strom and Rocek derived an explicit formula for the Kahler potential of this metric
generalizing the formula given by Calabi [Cal79]. It is not surprising that Calabi-
Lindstrom-Rocek metrics are only a special case. The hyperkahler metrics on coad-
joint orbits are complete if and only if the orbit is semisimple. These orbits are
then dieomorphic to the cotangent bundle of ag manifolds for G. In such cases
one can write down the metric and Kahler potentials explicitly [BG97a, DS97b].
[Compact Hyperkahler Manifolds] Much work has been done on the geometry
of compact hyperkahler manifolds with many new examples in higher dimensions.
12.10. ALE SPACES AND OTHER HYPERK

AHLER QUOTIENTS 471

We refer the reader to several extensive reviews on the subject and references therein
[GHJ03, VK99, NW04].
[Quaternionic Geometries with Torsion] These structures arose from the
attempts of physicists to incorporate the so-called bosonic Wess-Zumino-Witten
term [WZ71, Wit83] in -models with (extended) supersymmetry. The idea of
considering torsion connections in such -models dates back to the early 80ties
(see [HS84] (Kahler with torsion) and [GHR84] (hyperkahler with torsion)). The
manifolds involved are not Kahler or hyperkahler, but Hermitian and hyperher-
mitian. In 1996 Howe and Papadopoulos introduced a formal denition of the
so-called HKT geometry (hyperkahler with torsion) and studied twistor spaces of
such manifolds [HP96]. This sparked a considerable interest in such models also
among mathematicians (cf. [GP00, GGP03]). Today physicists and mathemati-
cians alike continue studying KT (Kahler with torsion), CYT (Calabi-Yau with
torsion), HKT, and even QKT (quaternionic Kahler with torsion) geometries. This
subject, although very interesting, goes far beyond the scope and the main focus
of our book. We refer the reader to a couple of extensive review articles on the
mathematical foundations, the history, and the bibliography [Gra04, Agr06].
CHAPTER 13
3-Sasakian Manifolds
In this chapter we consider a special kind of Sasaki-Einstein manifolds which
we shall call manifolds (several dierent names have been introduced and used in
the earlier literature on this subject). In 1970 Kuo [Kuo70] rened the notion
of a Sasakian structure and introduced manifolds with Sasakian 3-structures (see
also [KT70], [TY70]). Independently, the same concept was invented by Udriste
[Udr69]. Between 1970 and 1975 this new kind of geometry was investigated al-
most exclusively by a group of Japanese geometers, including Ishihara, Kashiwada,
Konishi, Kuo, Tachibana, Tanno, and Yu. Already in [Kuo70] we learn that the
3-Sasakian geometry has some interesting topological implications. Using earlier re-
sults of Tachibana about the harmonic forms on compact Sasakian spaces [Tac65],
Kuo showed that odd Betti numbers up to the middle dimension must be divisible
by 4. In 1971 Kashiwada observed that every 3-Sasakian manifold is Einstein with
a positive Einstein constant [Kas71]. In the same year Tanno proved an interest-
ing theorem about the structure of the isometry group of every 3-Sasakian space
[Tan70]. In a related paper he studied a natural 3-dimensional foliation on such
spaces showing that, if the foliation is regular, then the space of leaves is an Ein-
stein manifold of positive scalar curvature [Tan71]. Tanno clearly points to the
importance of the analogy with the quaternionic Hopf bration o
3
o
7
o
4
.
but does not go any further. In fact, Kashiwadas paper mentions a conjecture
speculating that every 3-Sasakian manifold is of constant curvature [Kas71]. She
attributed this conjecture to Tanno and, at the time, these were the only known
examples.
Very soon after, however, it became clear that such a conjecture could not
possibly be true. This is due to a couple of papers by Ishihara and Konishi [IK72],
[Ish73]. They made a fundamental observation that the space of leaves of the
natural 3-dimensional foliations mentioned above has a quaternionic structure,
part of which is the Einstein metric discovered by Tanno. This led Ishihara to
an independent study of this sister geometry: quaternionic Kahler manifolds
[Ish74]. His paper is very well-known and is almost always cited as the source of
the explicit coordinate description of quaternionic Kahler geometry. Among other
results Ishihara showed that his denition implies that the holonomy group of the
metric is a subgroup of oj(n)oj(1). thus providing an important connection with
the earlier studies of such manifolds by Alekseevsky [Ale68], Bonan [Bon67], Gray
[Gra69a], Kraines [Kra65], and Wolf [Wol65]. In 1975 Konishi [Kon75] proved
the existence of a Sasakian 3-structure on a natural principal oO(3)-bundle over any
quaternionic Kahler manifold of positive scalar curvature. This, with the symmetric
examples of Wolf, gives precisely all of the homogeneous 3-Sasakian spaces. Yet,
at the time they did not appear explicitly and escaped any systematic study until
much later.
473
474 13. 3-SASAKIAN MANIFOLDS
In fact, 1975 seems to be the year when 3-Sasakian manifolds are relegated to an
almost complete obscurity which lasted for about 15 years. From that point on the
two sisters fair very dierently. The extent of this can be best illustrated by the
famous book on Einstein manifolds by Besse [Bes87]. The book appeared in 1987
and provided the reader with an excellent, up-to-date, and very complete account
of what was known about Einstein manifolds 10 years ago. But one is left in the
dark when trying to nd references to any of the papers on 3-Sasakian manifolds we
have cited; 3-Sasakian manifolds are never mentioned in Besse. The other sister,
on the contrary, received a lot of space in a separate chapter. Actually Einstein
metrics on Konishis bundle do appear in Besse (see [Bes87] 14.85, 14.86) precisely
in the context of the oO(3)-bundles over positive quaternionic Kahler manifolds
as a consequence of a theorem of Berard-Bergery ([Bes87], 9.73). Obviously, the
absence of 3-Sasakian spaces in Besses book was the result rather than the cause
of this obscurity. One could even say it was justied by the lack of any interest-
ing examples. The authors have puzzled over this phenomenon without any sound
explanation. One can only speculate that it is the holonomy reduction that made
quaternionic Kahler manifolds so much more attractive an object. Signicantly, the
holonomy group of a 3-Sasakian manifold never reduces to a proper subgroup of
the special orthogonal group. And when in 1981 Salamon [Sal82, Sal86], indepen-
dently with Berard-Bergery [Bes87], generalized Penroses twistor construction for
self-dual 4-manifolds introducing the twistor space over an arbitrary quaternionic
Kahler manifold, the research on quaternionic Kahler geometry ourished, fuelled
by powerful tools from complex algebraic geometry.
13.1. Almost Hypercontact Manifolds and 3-Sasakian Structures
We rst describe some special kinds of almost contact and contact structures
which are closely related to the quaternionic geometries discussed in the previous
chapter.
Denition 13.1.1: Let ` be a manifold with a family of almost contact struc-
tures (). (). () parameterized by points = (
1
.
2
.
3
) o
2
R
3
on
a unit sphere. We say that (). (). ()
S
2 is an almost hypercontact
structure on ` if
() (
t
) () (
t
) = (
t
) (
t
)1l .
()(
t
) = (
t
). () (
t
) = (
t
)
for any two .
t
o
2
. Furthermore, a Riemannian metric o on ` is said to be
compatible with (or associated to) the almost hypercontact structure if
o(()A. ()Y ) = o(A. Y ) ()(A)()(Y )
for all o
2
. In such a case (). (). (). o
S
2 is called an almost hy-
percontact metric structure on `.
The above denition is a natural extension of the standard notion of the almost
contact metric structure given in Denition 6.3.1. Just as before one can easily show
that compatible metrics always exist. We can recover the standard denition of the
almost contact (metric) 3-structure due to [Kuo70] via a choice of an arbitrary
orthonormal basis c
1
. c
2
. c
3
in R
3
and by setting (c
a
) =
a
. (c
a
) =
a
. and
(c
a
) =
a
. Hence,
13.1. ALMOST HYPERCONTACT MANIFOLDS AND 3-SASAKIAN STRUCTURES 475
Denition 13.1.2: An almost contact (metric) 3-structure
a
.
a
.
a
. (o)
3
a=1
on ` is an almost hypercontact (metric) structure together with a choice of an or-
thonormal frame on R
3
.
We should note here that our denition allows for two dierent sign convention
which depend on the orientation of the chosen frame c
1
. c
2
. c
3
. Both conventions
can be found in the literature. The original denition in [Kuo70] corresponds to
choosing left-handed frame. In this chapter we will always choose the opposite: the
standard right-handed basis.
One can easily extend the results of Section 6.5 by considering the cone C(`) =
R
+
` together with
(13.1.1) 1()Y = ()Y +()(Y ). 1() = () . o
2
.
This clearly denes an almost hypercomplex structure I on C(`) which is invariant
under and such that the vector eld 1() is tangent to ` for each o
2
.
In particular, a manifold with an almost hypercontact structure must be of
dimension 4n + 3. Conversely, if C(`) is a cone with an almost hypercomplex
structure which is -invariant and such that 1() is tangent to ` for all o
2
we can dene an almost hypercontact structure on ` by taking () = 1(). For
each we then let 1
()
denote the trivial line bundle generated by (). we have
an exact sequence
01
()
T`Q0 .
and there is a one-to-one correspondence between the splittings of this exact se-
quence and 1-forms () on ` that satisfy () () = 1. This correspondence
is given by () ker (). Then the tensors () are given by
(13.1.2) () =

1(). on ker () .
0 . on 1
()
.
One arrives at the following straightforward generalization of Proposition 6.5.2
Proposition 13.1.3: There is a one-to-one correspondence between almost hyper-
contact structures (). (). (). o
S
2 on ` and -invariant almost hyper-
complex structures I on C(`) for which the vector eld 1() is tangent to `
and Equation 13.1.2 holds for each o
2
. Furthermore, if o is a Riemannian
metric on ` that is compatible with the almost hypercontact structure, then the
cone metric / = d:
2
+ :
2
o is hyperhermitian with respect to the almost hypercom-
plex structure I. and the one-to-one correspondence above induces a one-to-one
correspondence between almost hypercontact metric structures on ` and almost
hyperhermitian cone structures on C(`).
The following proposition is in complete analogy with 6.3.2 and it was already
observed by Kuo.
Proposition 13.1.4: There is a one-to-one correspondence between almost hy-
percontact metric structures (). (). (). o
S
2 on ` and reductions of the
frame bundle to the group oj(n) 1l
3
.
In particular, a 3-manifold admits such a structure if and only if it is paralleliz-
able. It is known that every compact orientable 3-manifold is parallelizable and,
hence,
476 13. 3-SASAKIAN MANIFOLDS
Corollary 13.1.5: Every compact orientable 3-manifold admits an almost contact
3-structure.
In higher dimensions there are other obstructions. For example, almost hyper-
complex manifolds are strict almost contact manifolds, so Theorem 6.2.7 implies
that the odd Stiefel-Whitney classes an all integral Stiefel-Whitney classes must
vanish, but the reduction to oj(n) implies more.
Proposition 13.1.6: Let `
4n+3
be a compact manifold with an almost contact 3-
structure. Then the only possible non-vanishing Stiefel-Whitney classes are n
4i
(`).
In particular, ` is a spin manifold. Furthermore, all the integral Stiefel-Whitney
classes \
i
(`) must vanish.
Proof. Because oj(n) ol(2n) l(2n) Theorem 6.2.7 implies every-
thing but the vanishing of n
4i+2
. For this we notice that T` splits as T
a
1l.
where T
a
= ker
a
for any o = 1. 2. 3. The group of the vector bundle T
a
is
oj(n) 1l
2
. So its characteristic classes are pulled back from the universal classes
in :
i
H
4i
(1oj(n). Z). See Appendix A. By the Universal Coecients Theo-
rem H

(1oj(n). Z
2
) is also generated by classes in dimension 4i. In particular the
classes n
4i+2
must vanish.
The reduction of the structure group to oj(n)1l
3
is equivalent to the existence
of a section of the associated bundle to 1(`) with ber 1
4n+3
= oO(4n+3)oj(n).
In particular, one needs to compute certain cohomology groups H
i
(`.
i1
(1
4n+3
)).
This can be done, at least in principle. In dimension 7 the situation is somewhat
special due to the following theorem [FKMS97]
Theorem 13.1.7: Let ` be a compact smooth 7-manifold and let ol(2) ol(3)
G
2
Spin(7). where the three subgroups of Spin(7) are the stability subgroups xing
one, two, or three spinors, respectively. Then the following conditions are equiva-
lent:
(i) ` admits reduction of the structure group to ol(2).
(ii) ` admits reduction of the structure group to ol(3).
(iii) ` admits reduction of the structure group to G
2
.
(iv) n
1
(`) = n
2
(`) = 0. i.e., ` admits a spin structure.
In particular, every compact oriented spin 7-manifold admits an almost hypercontact
structure.
The above theorem follows from the fact that any compact orientable 7-manifold
admits two independent vector elds [Tho69] which, on a spin manifold, can be
used to dene three linearly independent spinors at every point of : `
7
. Hence,
`
7
must admit three orthogonal spinor elds which can be used to dene ol(2)-
structure [FKMS97]. It is perhaps worth remarking that we get the following
interesting corollary.
Corollary 13.1.8: Let ` be a compact oriented 7-manifold. If ` is spin then the
integral third Stiefel-Whitney class \
3
(`) must vanish.
The above, by inspection, is also a property of compact smooth orientable 5-
manifolds which follows from the Barden-Smale Classication Theorem 10.2.3. But
the result here is much stronger as we do not need to assume simple connectivity.
From the point of view of Proposition 13.1.3 there several special cases one can
consider. We have
13.1. ALMOST HYPERCONTACT MANIFOLDS AND 3-SASAKIAN STRUCTURES 477
Denition 13.1.9: Let (). (). (). o
S
2 be an almost hypercontact metric
structure on ` with the corresponding -invariant almost hyperhermitian structure
I on C(`). We say that S = (). (). (). o
S
2 is a
(i) metric contact 3-structure if (C(`). I. o) is almost hyperkahler,
(ii) 3-Sasakian if (C(`). I. o) is hyperkahler.
We denote by S a 3-Sasakian structure on `. Alternatively, relative to a
choice of a right-handed orthonormal basis c
1
. c
2
. c
3
in R
3
we set o
a
= o(c
a
)
and S = o
1
. o
2
. o
3
and also refer to this triple of Sasakian structures as the
3-Sasakian structure.
Remark 13.1.1: Any almost HK structure is automatically HK (see Theorem
12.7.2) so it is elementary that any metric contact 3-structure is automatically
3-Sasakian. This was rst observed by Kashiwada [Kas01] and it eliminates the
need for the term hypercontact (as in almost contact vs. contact) altogether.
In particular, also the notion of the so-called 3-K-contact structure exists only as
part of a 3-Sasakian structure which was proved by Tanno [Tan96] in dimension
7. and by Jelonek [Jel01] in dimensions 15.
Remarks 13.1.1: We must warn the reader that there several other inequivalent
denitions of hypercontact, or quaternionic contact structures. For example, Geiges
and Thomas [GT95] dene a hypercontact structure as a triple of contact forms

1
.
2
.
3
and a Riemannian metric o compatible with an almost contact 3-
structure so that the 2-forms
a
(A. Y ) = o(A.
a
Y ) = d
a
(A. Y ) for all o = 1. 2. 3
and all A. Y (T`). This denition coincides with ours under the additional
requirement that the contact forms
a
=
a
. But without that assumption the
notion is much weaker. In particular, the Geiges-Thomas hypercontact property
is preserved by taking connected sums. In contrast, connected sums of 3-Sasakian
manifolds are typically not 3-Sasakian even for crude topological reasons (Betti
number constraints). The Geiges-Thomas hypercontact structures were also inves-
tigated by Banyaga [Ban96, Ban97].
Another generalization which has led to important discoveries that are beyond
the scope of this book is due to Biquard [Biq99, Biq00] who observed that the
conformal innity of a complete QK metric on a 4n-dimensional manifold with
boundary has a codimension 3 distribution on the boundary. Biquard used this to
dene what he called a quaternionic contact structure showing that in dimensions
greater than 7 his quaternionic contact structure is always the conformal innity of
some QK metric. In dimension 7, however, this is no longer automatic. Duchemin
solved the problem by proving a criterion for quaternionic contact structures to
be the conformal innity of an 8-dimensional QK metric[Duc06]. The 3-Sasakian
structures described in the present text appear as special cases of Biquards more
general theory. More recently similar structures were also studied by Alekseevsky
and Kamishima in [AK04], where they are called (para-)quaternionic CR struc-
tures.
Remark 13.1.2: Even though a hypercontact manifold is automatically 3-Sasakian
there are several meaningful weakenings of the 3-Sasakian condition. Starting with
an almost hypercontact metric structure one can consider the following conditions:
(i) The integrability conditions for the associated almost hypercomplex struc-
ture on the cone C(`) were studied by Kehowski [Keh04] and in such
a case an almost hypercontact metric structure is called hypernormal.
478 13. 3-SASAKIAN MANIFOLDS
(ii) When d
a
= d
a
= 0. such a structure on ` is said to be almost
cosymplectic 3-structure, and if the almost hypercontact structure is hy-
pernormal then ` is said to be cosymplectic 3-structure. It is easy to
see that an almost cosymplectic 3-structure is necessarily cosymplectic
[FIP04].
(iii) When d
a
= 0 and the almost hypercontact structure is hypernormal
then ` is said to be quasi-3-Sasakian.
(iv) Another possible generalization is by localizing [MO00]. That is, one
considers each of the tensors elds in the 3-structure
a
.
a
.
a
to be
sections of a 3-dimensional vector bundle.
Remark 13.1.3: As a 3-Sasakian manifold has not just 3 but an o
2
s worth of
Sasakian structures, in analogy with the term hyperkahler, the name hypersasakian
would be more consistent. However, most of the existing literature (including all
papers the two authors have written on the subject) uses the name Sasakian 3-
structure or 3-Sasakian structure. We feel it is too late at this stage to start using
anything else in its place.
13.2. Basic Properties
Let ` be a 3-Sasakian manifold so that C(`) is hyperkahler with a hypercom-
plex structure 1(). As in the Sasakian case we dene () = 1()(:
r
) for each
o
2
and show that these can be used to reconstruct the 3-structure S on `.
One important consequence of the quaternionic structure on C(`) is the relations
it implies for the 2-sphere of the characteristic vector elds () :
Proposition 13.2.1: If S = (). (). (). o
S
2 is a 3-Sasakian structure
on ` then o((). (
t
)) =
t
. [(). ()] = 2(
t
) and () = ().
Conversely, if ` has three Sasakian structures o
1
. o
2
. o
3
with characteristic vec-
tor elds
1
.
2
.
3
such that o(
a
.
b
) =
ab
and [
a
.
b
] = 2c
abc

c
. then S =
o
1
. o
2
. o
3
is 3-Sasakian.
Exercise 13.1: Prove Proposition 13.2.1.
In particular, Proposition 13.2.1 implies that the characteristic vector elds

1
.
2
.
3
generate a local isometric oj(1)-action on `. Also combining Proposi-
tion 13.2.1 together with Lemma 8.1.17 implies
Corollary 13.2.2: Suppose (`. o) admits two Sasakian structures o
1
. o
2
with
characteristic vector elds
1
.
2
. respectively. Suppose o(
1
.
2
) is a constant dif-
ferent from 1. Then ` admits a 3-Sasakian structure.
Since a hyperkahler manifold is Ricci-at, Proposition 11.1.5 implies
Corollary 13.2.3: Every 3-Sasakian manifold (`. S) of dimension 4n+3 is Ein-
stein with Einstein constant = 2(2n+1). Moreover, if ` is complete it is compact
with nite fundamental group.
The important result that every 3-Sasakian manifold is Einstein was rst ob-
tained by Kashiwada [Kas71] using tensorial methods.
If (`. S) is compact the characteristic vector elds
1
.
2
.
3
are complete and
dene a 3-dimensional foliation T
Q
on `. The leaves of this foliation are necessarily
compact as
1
.
2
.
3
denes a locally free oj(1) action on `. Hence, the foliation
T
Q
is automatically quasi-regular and the space of leaves is a compact orbifold. We
shall denote it by O. In addition, for any o
2
we can consider the characteristic
13.2. BASIC PROPERTIES 479
vector eld () associated with the direction . Since () generates a 1-parameter
subgroup of o
1

oj(1). it denes a 1-dimensional foliation T

T
Q
` with
compact leaves. Hence, () is quasi-regular and denes a locally free circle action
on `.
Denition 13.2.4: Let (`. S) be a compact 3-Sasakian manifold with 3-Sasakian
structure S. We say that S is regular if the Sasakian structures o() are regular
for all o
2
.
Remark 13.2.1: When dim(`) 3 the above denition is equivalent to regularity
of the foliation T
Q
. When dim(`) = 3 the leaf space of the foliation T
Q
is a single
point so that T
Q
is always regular. However, o is not always regular.
We have the following observation of Tanno [Tan71] concerning regularity
properties of the foliations T

T
Q
which is essentially a corollary of Theorem
13.3.1 below.
Proposition 13.2.5: Let (`. S) be a compact 3-Sasakian manifold. Every Sasakian
structure o() S is quasi-regular. Moreover, if o() is regular for some =
0

o
2
then it is regular for all . Furthermore, if S is regular then either all the leaves
are dieomorphic to oO(3) or all the leaves are dieomorphic to o
3
.
Actually in the regular case it follows from a deeper result of Salamon [Sal82]
that all leaves are dieomorphic to o
3
in precisely one case, namely when ` =
o
4n+3
.
Remark 13.2.2: Note that every Sasaki-Einstein 3-manifold must also have a 3-
Sasakian structure. This is because in dimension four Ricci-at and Kahler is
equivalent to hyperkahler. Every compact 3-Sasakian 3-manifold must be a space
of constant curvature 1. Hence, ` is covered by a unit round 3-sphere and, in
fact, it is always the homogeneous spherical space form o
3
. where is a discrete
subgroup of oj(1) [Sas72]. The homogeneous spherical space forms in dimension 3
are well-known. They are oj(1). where is one of the nite subgroups of oj(1)
discussed in detail in Sections 10.1.1 and 12.10.1. The only regular 3-Sasakian
manifolds in dimension 3 are o
3
and oO(3). Compare this with the classication
by Belgun and Geiges of all Sasakian 3-manifolds given in Section 10.1.
A Sasaki-Einstein structure on a 3-Sasakian manifold does not have to be a
part of the 3-Sasakian structure. Several hundred Sasaki-Einstein metrics exist on
o
7
by Corollary 11.5.4 none of which are part of a 3-Sasakian structure which can
be proved using Exercise 13.2 below. It is worth observing that the only examples
that we are aware so far involve spheres and their quotients. It is possible that some
of the 3-Sasakian rational homology 7-spheres of Theorem 13.8.3 are dieomorphic
to some of the Sasaki-Einstein rational homology 7-spheres discussed in Example
11.7.12. For example if we take | in Theorem 13.8.3 to be one of the primes listed in
the table referred to in Example 11.7.12, then the two rational homology 7-spheres
will have the same cohomology ring, but we have no way of knowing at this time
whether they are even homeomorphic. However, Corollary 13.3.2 below implies
that the Sasaki-Einstein metrics of Example 11.7.12 cannot be part of a 3-Sasakian
structure since they all have orbifold Fano index one. Of course, the previously
mentioned examples refer to distinct Riemannian metrics. Here is an example of
distinct Sasakian structures sharing the same Riemannian metric.
480 13. 3-SASAKIAN MANIFOLDS
Example 13.2.6: Recall that Lemma 8.1.17 says that the only Riemannian met-
ric that can share two distinct Sasakian structures that are neither conjugates nor
part of the same 3-Sasakian structure are metrics of constant sectional curvature
one. In fact it is easy to see that the constant sectional curvature one metric o
0
on o
4n+3
admits two distinct 3-Sasakian structures dened by whether the oj(1)-
action as unit quaternions is taken from the right or left. Now consider the lens
space Z
k
`o
4n+3
obtained by quotienting by the cyclic subgroup Z
k
of the diag-
onal (oj(1) oj(n). This can represented by multiplication from the left by
a diagonal matrix with diagonal entries o
1
oj(1) such that
k
= 1. This
Z
k
commutes with the right oj(1)-action, so the right 3-Sasakian structure S
on o
4n+3
descends to a 3-Sasakian structure on the quotient Z
k
`o
4n+3
. However,
the left 3-Sasakian structure on o
4n+3
does not descend to Z
k
`o
4n+3
since the
centralizer of Z
k
in (oj(1)) is a diagonal o
1
Aut(o
l
). Nevertheless, the left
Sasakian structure determined by this o
1
does descend to a left Sasakian struc-
ture o
l
on Z
k
`o
4n+3
. This left Sasakian structure is actually regular while none
of the Sasakian structures of the right 3-Sasakian structure can be regular unless
/ = 1. 2.
13.3. The Fundamental Foliations T

and T
Q
In this section we discuss the two fundamental foliations associated to 3-Sasakian
manifolds and describe their consequences. We begin with
13.3.1. The 1-dimensional Foliation T

. Fixing a Sasakian structure, say

o
1
= (
1
.
1
.
1
) in the 3-Sasakian structure S, the space of leaves of the corre-
sponding characteristic foliation T
1
is a compact complex orbifold (Z
1
. 1
1
) which
by Corollary 13.2.3 and Theorem 11.1.3 admits a compatible Kahler-Einstein met-
ric. The subbundle T
1
= ker
1
of T` together with
1
[
11
dene the CR structure
on `, and it is related to the complex structure 1
1
on Z
1
by 1
1
A = ([
11
)

A. where

A denotes the horizontal lift of the vector eld A on Z

1
. Actually a 3-Sasakian
structure gives a special kind of CR structure, namely, a CR structure with a
compatible holomorphic contact structure. Notice that the complex valued one
form on ` dened by
+
=
2
+ i
3
has type (1. 0) on `. Moreover, one checks
that
+
is holomorphic with respect to the CR structure. Although the 1-form

+
is not invariant under the circle action generated by
1
. the trivial complex
line bundle 1
+
generated by
+
is invariant. Thus, the complex line bundle 1
+
pushes down to a nontrivial complex orbibundle L on Z
1
. Let \ denote the one
dimensional complex vector space generated by 1
+
. Writing the circle action as
exp (i
1
) shows that \ is the representation with character c
2i
. and since `
is a principal o
1
-orbibundle over Z
1
. the twisted product L `
S
1 \ is a holo-
morphic line orbibundle on the complex orbifold (Z
1
. 1
1
). Now we can dene a map
of orbibundles : T
(1,0)
Z
1
L by (A) =
+
(

A). Notice that (A) is not a
function on Z
1
but a section of L. Now a straightforward computation shows that

+
(d
+
)
n
is a nowhere vanishing section of
(2n+1,0)
T
1
on `. and thus (d)
n
is a nowhere vanishing section of /
orb

1
L
n+1
. implying the relation L
n+1
/
orb

1
= 1l
in Pic
orb
(Z). This gives Z
1
a complex contact structure by Proposition 6.1.14. It
is straightforward to perform this construction for an arbitrary choice of oriented
orthonormal 3-frame (
1
.
2
.
3
). Choosing the complex structure 1(
1
) we dene

+
(
1
) = (
2
) + i(
3
). Thus, we obtain a complex contact structure
1
on
13.3. THE FUNDAMENTAL FOLIATIONS J

AND J
Q
481
the complex orbifold (Z

1
. 1(
1
)) with a compatible orbifold Kahler-Einstein met-
ric /

1
. It is easy to see that there is a oj(1) such that o

1
=

o
1
. and
that induces an isomorphism of the corresponding complex orbifolds with their
induced structures, viz. (Z
1
. 1(
1
).
1
. /
1
) (Z
1
. 1
1
. . /). We have arrived at
[BGM93, BG97b]:
Theorem 13.3.1: Let (`. S) be a complete 3-Sasakian manifold, choose a direc-
tion o
2
. and let Z

denote the space of leaves of the corresponding foliation T

.
Then Z

is a compact locally cyclic Fano contact orbifold with a Kahler-Einstein

metric /

of scalar curvature 8(2n + 1)(n + 1) such that the natural projection

: ` Z

is an orbifold Riemannian submersion with respect to the Riemann-

ian metrics o on ` and /

on Z

. Furthermore, dierent .

o
2
give an
isomorphism (Z

. 1().

. /

) (Z

. 1(
t
).

. /

).
The space Z

turns out to be the twistor space dened in Denition 12.2.15,

and in [BG97b] we named it, perhaps erroneously, the twistor space of the 3-
Sasakian manifold (`. S). Actually, there is another object that could merit the
name the twistor space of `. namely the trivial 2-sphere bundle o
2
` with the
structure induced from the twistor space o
2
C(`) of the hyperkahler cone.
There is a nice corollary that follows from the proof of Theorem 13.3.1.
Corollary 13.3.2: Let Z be the twistor space of a complete 3-Sasakian manifold
`
4n+3
. Then its orbifold Fano index is divisible by n + 1.
Exercise 13.2: Let A
f
be a weighted homogeneous hypersurface in CP(w) of
complex dimension 2n +1. Show that if 1
X
f
= [w[ d < (n +1)min
i
n
i
then A
f
can not admit a complex contact structure.
Example 13.3.3: As observed in Example 13.2.6 the only manifold that has two
distinct 3-Sasakian structures sharing the same Riemannian metric is the sphere
o
4n+3
with its constant sectional curvature one metric. A related result was proven
by LeBrun in [LeB95] where he shows that the only positive Kahler-Einstein man-
ifold that admits two distinct complex contact structures is CP
2n+1
. Recall that
the manifold Z
k
`o
4n+3
of Example 13.2.6 has a non-regular 3-Sasakian structure.
Since the right and left circle actions commute we see that its twistor space is the
orbifold Z
k
`CP
2n+1
with
orb
1
= Z
k
. Note also that the quotient of Z
k
`o
4n+3
by
the left circle action is just CP
2n+1
.
Recall from Theorem 12.2.16 that the twistor space of a quaternionic manifold
is ruled by rational curves. The same is true in our case as long as one allows for
singularities. We have
Proposition 13.3.4: The twistor space Z of a 3-Sasakian manifold is ruled by a
real family of rational curves C with possible singularities on the singular locus of
Z. All the curves C are simply-connected, but
orb
1
(C) can be a non-trivial cyclic
group.
For any line orbibundle L we let

L denote L minus its zero section. Recall
Kawasakis notion [Kaw78] of proper from Remark 4.2.2.
Proposition 13.3.5: Let Z be the twistor space of a 3-Sasakian manifold ` of
dimension 4n + 3. and assume that
orb
1
(Z) = 0. If the contact line orbibundle L
(or equivalently its dual L
1
) has a root in Pic
orb
(Z). then it must be a square
root, namely L
1
2
. Moreover, in this case if both

L and

L
1
2
are proper in the sense
482 13. 3-SASAKIAN MANIFOLDS
of Kawasaki, then we must have Z = P
2n+1
. In particular, this holds if the total
space of

L is smooth.
Proof. Since Pic
orb
(Z) is torsion free. By Proposition 13.3.4 Z is ruled by
rational curves C which on the singular locus take the form `P
1
. Now the restric-
tion L
1
[C is O(2) which is an orbibundle if C is singular. In either case it has
only a square root namely the tautological orbibundle O(1). Since these curves
C cover Z this proves the rst statement. The second statement follows from a
modication of an argument due to Kobayashi and Ochiai [KO73] and used by
Salamon [Sal82]. The main point is that since

L.

L
1
2
are proper and it follows
that we can apply Kawasakis Riemann-Roch Theorem [Kaw79] together with the
Kodaira-Baily Vanishing Theorem 4.4.28 to arbitrary powers of the line orbibundle
L
1
2
to give (n + 1)(2n + 3) innitesimal automorphisms of the complex contact
structure on Z. Since
orb
1
(Z) = 0. these integrate to global automorphisms on Z
and the result follows. See [BG97b] for details.
Remark 13.3.1: There is an error in the statement of Proposition 4.3 of [BG97b].
The error is in leaving out the assumptions that
orb
1
(Z) is trivial and that the
contact line bundle is proper. Example 13.3.6 below shows that the conclusion in
Proposition 13.3.5 does not necessarily hold if the hypothesis
orb
1
(Z) = 0 is omit-
ted. Likewise, Example 13.3.7 below gives a counterexample when the condition
that L be proper is omitted.
Example 13.3.6: Consider the 3-Sasakian lens space 1(j; c) = Z
p
`o
7
constructed
as follows: o
7
is the unit sphere in the quaternionic vector space H
2
with quater-
nionic coordinates n
1
. n
2
. The action of Z
p
is the left action dened by (n
1
. n
2
)
(n
1
.
q
n
2
). where
p
= 1 and j and c are relatively prime positive integers. If
j = 2: for some integer : then id is an element of Z
2m
. so the 3-Sasakian mani-
folds 1(2:; c) and 1(:; c) both have the same twistor space, namely Z = Z
m
`P
3
.
and
orb
1
(Z) Z
m
. There are clearly many similar examples in all dimensions equal
to 3 mod 4.
Example 13.3.7: Consider the 3-Sasakian 7-manifolds o(j
1
. j
2
. j
3
) described in
Section 13.7.4 below, where the j
i
s are pairwise relatively prime, and precisely one
of the j
i
s is even, say j
1
. o(j
1
. j
2
. j
3
) is simply-connected and its twistor space
Z(j
1
. j
2
. j
3
) has
orb
1
(Z(j
1
. j
2
. j
3
)) = 0. Now there is a Z
2
acting on o(j
1
. j
2
. j
3
).
but not freely, which acts as the identity on Z(j
1
. j
2
. j
3
). Thus, Z
2
`o(j
1
. j
2
. j
3
)
has Z(j
1
. j
2
. j
3
) as its twistor space, and as an orbibundle Z
2
`o(j
1
. j
2
. j
3
)
Z(j
1
. j
2
. j
3
) is not proper in the sense of Kawasaki. Thus, the orbibundle

L is not
proper, and Kawasakis Riemann-Roch theorem [Kaw79] cannot be applied.
We now wish to formulate a converse to Theorem 13.3.1.
Denition 13.3.8: A compact Fano contact orbifold Z is said to be good if the to-
tal space of the principal circle bundle ` associated with the contact line orbibundle
L is a smooth compact manifold.
The following theorem given in [BG97b] can be thought of as an orbifold
version of a theorem of LeBrun [LeB95]:
Theorem 13.3.9: A good compact Fano contact orbifold Z is the twistor space
associated to a compact 3-Sasakian manifold if and only if it admits a compatible
Kahler-Einstein metric /.
13.3. THE FUNDAMENTAL FOLIATIONS J

AND J
Q
483
Proof. The only if part is Theorem 13.3.1, so we assume that Z is a good
Fano contact orbifold with a compatible Kahler-Einstein orbifold metric /. Choose
the scale of / so that the scalar curvature is 8(2n + 1)(n + 1). Let : ` Z
denote the principal orbifold circle bundle associated to L. It is a smooth compact
submanifold embedded in the dual of the contact line orbibundle L
1
. The Kahler-
Einstein metric / has Ricci form = 4(n + 1). where is the Kahler form on
Z. and represents the orbifold rst Chern class of /
1

. Let
1
be the connection
in : ` Z with curvature form 2

. Then the Riemannian metric o

M
on `
can be dened by o
M
=

/ +
2
1
. It follows from the Inversion Theorem 11.1.12
that o
M
is Sasaki-Einstein. As in Proposition 2.2.4 of [Swa91] the orbibundle
L
(1,0)
Z has a section such that the Kahler-Einstein metric / decomposes as
/ = [[
2
+/
D
. where /
D
is a metric in the codimension one orbibundle 1 = ker .
Let us write

=
+
. Since ` is a circle bundle in L
1
. the contact bundle L
trivializes when pulled back to `. This together with the condition that (d)
n
is nowhere vanishing on Z implies that
+
is a nowhere vanishing complex valued
1-form on `. So the metric o
M
on ` can be written as
o
M
=
2
1
+[
+
[
2
+

/
D
.
We claim that this metric is 3-Sasakian. To see this consider the total space of the
dual of the contact line orbibundle minus its 0 section which is ` R
+
= C(`).
Put the cone metric d:
2
+ :
2
o on C(`). The natural C

action on C(`) induces

homotheties of this metric. Now using a standard Weitzenbock argument, LeBrun
[LeB95] shows that C(`) has a parallel holomorphic symplectic structure and his
argument works just as well in our case. Let denote the pullback of the contact
form to C(`) which is a holomorphic 1-form on ` that is homogeneous of degree
1 with respect to the C

action. Thus = d is a holomorphic symplectic form

on C(`) which is parallel with respect to the Levi-Civita connection of the cone
metric. Hence, (C(`). d:
2
+ :
2
o) is hyperkahler. Furthermore, if 1
a

3
a=1
denote
hyperkahler endomorphisms on C(`).
2
.
3
are the real and imaginary parts of
. and
1
is the pullback of
1
to C(`). then LeBrun shows that

1
1
1
=
2
1
2
=
3
1
3
.
It then follows from our previous work [BGM93] that o is 3-Sasakian. But by
construction Z is the space of leaves of the foliation generated by
1
. so Z must be
the twistor space of the compact 3-Sasakian manifold `.
13.3.2. The 3-dimensional Foliation T
Q
. We now discuss the geometric
properties of the 3-dimensional foliation T
Q
generated by the characteristic vector
elds
1
.
2
.
3
. Basic references are [BGM94a, BG99]. First we make note of
some of the properties of the ONeill tensor elds and T discussed in Section
2.5.1. Since T
a
T
Q
for each o many of the properties satised by the Sasakian
characteristic foliation T
a
hold for the 3-Sasakian foliation T
Q
. In particular, the
ONeill tensor T
Q
vanishes for T
Q
. The only potentially non-vanishing components
satisfy T
Q
a

b
= (
a

b
)
h
= (
b

a
)
h
= 0. and T
Q
a
A = (
a
A)
v
= 0. For the
tensor of ONeill we notice that Equations (2.5.15)-(2.5.18) imply a direct analogy
with the T
a
foliation. Using the convention that A. Y are sections of H. we have
the 3-Sasakian version of Lemma 7.3.5:
Lemma 13.3.10: For a 3-Sasakian manifold the following identities hold:
(i) T
Q
= 0 .
484 13. 3-SASAKIAN MANIFOLDS
(ii)
Q
X
Y =

3
a=1
d
a
(A. Y )
a
.
(iii)
Q
X

a
=
a
A .
(iv) [
Q
[
2
= 12n.
Exercise 13.3: Prove Lemma 13.3.10 and give the relation between ONeills ten-
sor
a
for the foliation T
a
. and ONeill tensor
Q
for T
Q
.
We summarize some of the basic results in
Proposition 13.3.11: Let (`. S) be a 3-Sasakian manifold such that the char-
acteristic vector elds
1
.
2
.
3
are complete. Let T
Q
denote the canonical three
dimensional foliation on `. Then
(i) The foliation T
Q
is Riemannian, or equivalently the metric o is bundle-
like.
(ii) The leaves of T
Q
are totally geodesic spherical space forms `o
3
of con-
stant curvature one, where oj(1) = ol(2) is a nite subgroup.
(iii) The 3-Sasakian structure on ` restricts to a 3-Sasakian structure on
each leaf.
(iv) The generic leaves are either ol(2) or oO(3).
Proof. The proof of (i), (ii), and (iii) follow from the basic relations for 3-
Sasakian manifolds and is left to the reader. To prove (iv) we notice that the
foliation T
Q
is regular restricted to the generic stratum `
0
. By (ii) and regularity
there is a nite subgroup ol(2) such that the leaves of this restricted foliation
are all dieomorphic to `o
3
. which is 3-Sasakian by (iii). Now the regularity of
T
Q
on `
0
implies that its leaves must all be regular with respect to the foliation
generated by
1
. But a result of Tanno [Tan70] says that the only regular 3-
Sasakian 3-manifolds have = id or Z
2
. in which case (iv) follows.
Example 13.3.12: Consider the 3-Sasakian lens space 1(j; c) = Z
p
`o
7
of Example
13.3.6. If j is odd then id is not an element of Z
p
so the generic leaf of the
foliation T
Q
is o
3
. The singular stratum consists of two leaves both of the form
Z
p
`o
3
with leaf holonomy group Z
p
. These two leaves are described by n
2
= 0 and
n
1
= 0. respectively. If j is even then id is an element of Z
p
. so the generic leaf
is ol(2)Z
2
= oO(3). and the leaf holonomy of the two singular leaves is Z
p
2
.
The next theorem was rst proved by Ishihara [Ish74] in the regular case. Its
general version described here was proved in [BGM94a].
Theorem 13.3.13: Let (`. S) be a 3-Sasakian manifold of dimension 4n+3 such
that the characteristic vector elds
1
.
2
.
3
are complete. Then the space of leaves
`T
Q
has the structure of a quaternionic Kahler orbifold (O. o
c
) of dimension
4n such that the natural projection : ` O is a principal orbibundle with
group ol(2) or oO(3) and a Riemannian orbifold submersion such that the scalar
curvature of o
c
is 16n(n +2). Furthermore, = (
1
.
2
.
3
) is a connection in this
principal orbibundle.
Proof. We can split T` = 1
3
H. where 1
3
is the subbundle spanned by
the Lie algebra su(2) of characteristic vector elds
1
.
2
.
3
and the horizontal
bundle is the orthogonal complement H = 1

3
. Let /
a
=
a
[
1
be the restriction
of characteristic endomorphisms. One can easily see that
/
a
/
b
=
ab
1l +

c
c
abc
/
c
.
13.3. THE FUNDAMENTAL FOLIATIONS J

AND J
Q
485
It follows that H is pointwise a quaternionic vector space and O is a compact
quaternionic orbifold. We must show that the metric o
c
obtained from o by the
orbifold Riemannian submersion : ` O has its holonomy group reduced to a
subgroup of oj(n) oj(1). This can be done by constructing a parallel 4-form on
O. Consider
a
= d
a
and

a
=
a
+

b,c
c
abc

b

c
.
It is easy to see that the 4-form =

a
is horizontal and oj(1)-invariant.
It follows that there is a unique 4-form

on the orbifold O invariant under the
action of the local uniformizing groups such that

= . One can show that

is parallel on O using standard tensor computation with ONeill formulas (see

[BGM93, GL88] for details). In the case n = 1 the parallelism of the 4-form
does not further restrict Riemannian geometry of O. However, one can show that
(O. o
c
) is a compact self-dual Einstein orbifold. Self-duality follows easily from
the fact that O is quaternionic. The fact that the metric is Einstein is a simple
computation and in the regular case can be found in [Tan71].
There is an important inversion theorem of Theorem 13.3.13 originally in the
regular case due to Konishi [Kon75]. By now there are several proofs of this,
all of them related. Given a quaternionic Kahler orbifold O one can construct
the Salamon twistor space Z and then get ` from the Inversion Theorem 13.3.9
[BG97b]. Another approach would be to construct the orbifold version of Swanns
bundle [Swa91] on O and then use the results of [BGM93] to obtain `. Here our
proof is essentially that of Konishis, only slightly modied to handle the orbifold
situation.
Theorem 13.3.14: Let (O. o
c
) be a quaternionic Kahler orbifold of dimension 4n
with positive scalar curvature 16n(n+2). Then there is a principal oO(3) orbibundle
over O whose total space ` admits a 3-Sasakian structure with scalar curvature
2(2n + 1)(4n + 3).
Proof. Let O denote the orbisubbundle of End TO describing the quater-
nionic structure. Let

l
i
be local uniformizing neighborhoods that cover O and
1
(i)
a
a local framing of O on

l
i
that satises
1
(i)
a
1
(i)
b
=
ab
1l +c
abc
1
(i)
c
.
Since O is quaternionic Kahler there are 1-forms
(i)
a
on each

l
i
such that
1
(i)
a
= c
abc

(i)
b
1
(i)
c
.
Now the structure group of the orbibundle End TO is oj(n)oj(1). and that of the
orbisubbundle O is oO(3). Let : ` O denote the principal oO(3) orbibundle
associated to O. The local 1-forms
(i)
a
are the components of an so(3) connection

(i)
=

3
a=1

(i)
a
c
a
. where c
a
denotes the standard basis of so(3) which satises
the Lie bracket relations [c
a
. c
b
] = 2c
abc
c
c
. The local connection forms satisfy the
well-known relations

(i)
= ad
g
ij

(j)
+o
1
ij
do
ij
486 13. 3-SASAKIAN MANIFOLDS
in

l
i

l
j
for some smooth map o
ij
:

l
i

l
j
oO(3). Furthermore, from [Ish74]
one checks that the curvature forms
(13.3.1)
(i)
a
= d
(i)
a
+c
abc

(i)
b

(i)
c
satisfy the relation 2o
c
(1
(i)
a
A. Y ) =
(i)
a
(A. Y ). Now on each

l
i
there exists a
smooth local section
i
:

l
i
` and on ` there is a global 1-form
a
such that

(i)
a
=

a
. On ` we dene a Riemannian metric by
o =

o
c
+
3

a=1

a
.
By construction the vector elds
a
generating the oO(3) action on ` are dual to
the forms
a
with respect to this metric, viz. o
M
(A.
a
) =
a
(A) for any vector eld
A on `. Also by construction the vector elds
a
are Killing elds with respect to
the metric o. Now dene the (1. 1) -tensor eld
a
=
a
. Since the
a
are mutually
orthogonal vector elds of unit length on `. one easily checks that
a

b
= c
abc

c
.
Thus
a
splits as

a
= /
a
+c
abc

c
.
One then checks using 13.3.1 that on each open set
1
(l
i
). /
a
equals the hori-
zontal component of (
i
)

1
(i)
a
. From this one then shows that

a

b

b
= c
abc

ab
1l .
If the homomorphisms
i
:
i
oO(3) are injective the total space ` will be
a smooth 3-Sasakian manifold. We summarize our discussion of the fundamental
foliations associated to a 3-Sasakian structure with the following
Denition 13.3.15: With every 3-Sasakian manifold (orbifold) (`. S) we asso-
ciated the diamond diagram (`. S) of ` and three more Einstein orbifold
spaces canonically associated to `:
(13.3.2) C(`)
.

1Q

O
of orbifold brations, where
(i) ` C(`) is the inclusion of ` into the hyperk ahler cone C(`) .
(ii) C(`)O. C(`)Z . C(`)` are the Swann orbibundles,
(iii) `O is the Konishi orbibundle,
(iv) `Z is the twistor bration.
Konishis construction gives an oO(3) orbibundle over O. In the case that O is a
smooth manifold there is a well-known obstruction to lifting this bundle to an oj(1)
bundle, the Marchiafava-Romani class c of Denition 12.2.1. This obstruction also
occurs when O is an orbifold as long as one uses Haeigers orbifold cohomology.
The class c is the image of the connecting homomorphism
(13.3.3) : H
1
orb
(O. () H
2
orb
(O. Z
2
).
13.3. THE FUNDAMENTAL FOLIATIONS J

AND J
Q
487
where ( is the sheaf of germs of smooth orbifold maps from open sets of O to the
group oj(n)oj(1). Since every 3-Sasakian structure S has canonically associated to
it a unique quaternionic Kahler orbifold, we can consider the Marchiafava-Romani
class c to be an invariant of the 3-Sasakian structure S in which case we shall
sometimes write c(S).
If following Salamon [Sal82] we write TO C E H. then c is the second
Stiefel-Whitney class n
2
of the orbibundle o
2
(H) over O. We have
Proposition 13.3.16: The principal oO(3)-orbibundle of Theorem 13.3.14 lifts to
a principal oj(1) orbibundle if and only if c H
2
orb
(O. Z
2
) vanishes. Moreover,
when c = 0 the 3-Sasakian structure on the total space ` of the oO(3) orbibundle
lifts to the total space `
t
of the oj(1) orbibundle.
Thus, in the case that c = 0 there are precisely two 3-Sasakian orbifolds `. `
t
corresponding to the quaternionic Kahler orbifold O. Let Z denote the twistor
space of the orbifold O. Then likewise, since o
2
oO(3)o
1
oj(1)o
1
the two
3-Sasakian orbifolds ` and `
t
have the same twistor space Z. When O is a smooth
manifold a result of Salamon [Sal82] says that c = 0 if and only if the quaternionic
Kahler manifold O is quaternionic projective space. If we impose the condition
that the orbifolds ` and `
t
are smooth manifolds, there is a similar result.
Theorem 13.3.17: If two 3-Sasakian manifolds (`. S) and (`
t
. S

) are associ-
ated to the same quaternionic Kahler orbifold O or, equivalently, the same twistor
space Z. then both ` and `
t
have the same universal covering space

`

`
t
which is the standard 3-Sasakian sphere.
Proof. We work with the twistor space Z. Now ` and `
t
are unit circle
bundles in the line orbibundles L
1
and (L
t
)
1
respectively. Moreover, since `
t
is a double cover of `. it follows that L = (L
t
)
2
. Consider the universal orbifold
cover

Z of Z with
orb
1
(

Z) = 0. Pull back the orbibundles L and L
t
to orbibundles

L and

L
t
on

Z respectively. These bundles satisfy

L
t
=

L
1
2
. By construction and
naturality of the covering maps

L is the contact line bundle on

Z. Moreover,

`
and

`
t
which are the total spaces of the pullbacks of ` and `
t
to

Z are both
smooth manifolds since they cover smooth manifolds. Thus, the orbibundles

L
t
and

L are proper, so we can apply Proposition 13.3.5 giving

Z P
2n+1
. It follows that

`
t
o
4n+3
.
Remark 13.3.2: Konishi also considers the case when the quaternionic Kahler
manifold has negative scalar curvature. This gives a Sasakian 3-structure on `
with indenite signature (3. 4n).
13.3.3. The Second Einstein Metric. Of course, by an Einstein metric we
actually mean a homothety class of Einstein metrics. In this section we shall show by
using a theorem of Berard-Bergery [Bes87] that every 3-Sasakian manifold, as well
as its twistor space, has at least two distinct homothety classes of Einstein metrics.
The method involves the canonical variation [Bes87] associated with Riemannian
submersions. Due to the local nature of the calculations involved this construction
holds equally well for orbifold Riemannian submersions. The canonical variation is
constructed as follows [Bes87]:
Let : ` 1 be an orbifold Riemannian submersion with o the Riemannian
metric on `. Let 1
3
and H denote the vertical and horizontal subbundles of the
488 13. 3-SASAKIAN MANIFOLDS
tangent bundle T`. For each real number t 0 we construct a one parameter
family o
t
of Riemannian metrics on ` by dening
(13.3.4) o
t
[
\
3
= to[
\
3
. o
t
[
1
= o[
1
. o
t
(1
3
. H) = 0 .
So for each t 0 we have an orbifold Riemannian submersion with the same base
space. Furthermore, if the bers of o are totally geodesic, so are the bers of o
t
. We
apply the canonical variation to the orbifold Riemannian submersion : ` O.
The metric as well as other objects on O will be denoted with a check such as o.
Theorem 13.3.18: Every 3-Sasakian manifold ` admits a second Einstein metric
of positive scalar curvature. Furthermore, the twistor space Z also admits a second
orbifold Einstein metric which is Hermitian-Einstein, but not Kahler-Einstein.
Proof. We apply the canonical variation to the orbifold Riemannian submer-
sion : ` O. According to the Berard-Bergery Theorem [Bes87], 9.73 there
are several conditions to check. First, the connection H must be a Yang-Mills
connection. By (i) of Lemma 13.3.10, the ONeill tensor T
Q
vanishes. So the
Yang-Mills condition is [Bes87]
(13.3.5)

i
o((
E
i

Q
)
E
i
A.
a
) = 0
for each o = 1. 2. 3 and where 1
i
is a local orthonormal frame of H. A is any
horizontal vector eld, and
Q
is ONeills tensor for T
Q
. Actually in our case
of a Sasakian metric we can prove the stronger condition o((
E
i

a
)
E
i
A.
a
) = 0
for all i = 1. . . . . n and all o = 1. 2. 3 by using (ii) of Lemma 7.3.10 and Theorem
7.3.16. Then using the relation between
Q
and
a
Equation 13.3.5 follows. Second
[
Q
[
2
must be constant which is (iv) of Lemma 13.3.10. The nal condition to be
satised is 9.73e of [Bes87]. Using the easily veried relations o(
X
i
.
X
j
) = 3
ij
and o(
a
.
b
) = 4n
ab
this condition becomes (

)
2

(12n + 18) 0. where

and

are the Einstein constants for O and the bers, respectively. Since in our
case

= 4(n + 2) and

= 2. we see that the inequality is satised.
The last statement, due also to Berard-Bergery, follows also by applying the
canonical variation to the twistor space orbifold bration
t
: ZO. See 14.84
of [Bes87].
The scalar curvature of any metric o
t
in the canonical variation of the 3-
Sasakian metric o is given by the formula :
t
= 16n(n + 2) + 6t 12nt [Bes87].
Moreover, the value of t that gives the second Einstein metric is t
0
=

=
1
2n+3
.
A similar computation can be done for the second Einstein metric on Z. We shall
return to these second Einstein metrics in Chapter 14.
13.3.4. 3-Sasakian Metrics and Positive Sectional Curvature. The only
3-Sasakian metrics of positive sectional curvature are the standard metrics on
spheres. On the other hand, Dearricott has shown that there is a relation between
certain 3-Sasakian metrics and metrics of positive sectional curvature [Dea04,
Dea05]. More precisely he has proven
Theorem 13.3.19: Let (`. S) be a 3-Sasakian manifold and let (O. o
c
) be the
associated quaternionic Kahler orbifold. If o
c
is of positive sectional curvature
then the canonical variation o
t
of the 3-Sasakian metric o also has positive sectional
curvature for suciently small t.
13.3. THE FUNDAMENTAL FOLIATIONS J

AND J
Q
489
The proof of this theorem can be found in [Dea04] and it easily follows from
the relation between sectional curvatures of the original metric and its canonical
variation. The condition on o
c
being itself of positive sectional curvature is very
strong and allows few examples. All of the known examples are the weighted pro-
jective spaces of Theorem 12.5.4 in dimension 4. These do lead to a new geometric
construction of positive curvature metrics on some Eschenburg spaces which will
be discussed later. Unfortunately, in higher dimensions, the only known example
of quaternionic Kahler spaces of positive sectional curvature are HP
n
.
13.3.5. Invariants and the Classication of 3-Sasakian Structures. We
consider the question of equivalence of 3-Sasakian manifolds. Note that a 3-Sasakian
structure S =
a
.
a
.
a

3
a=1
on a manifold ` is determined completely by the
metric o and the characteristic vector elds
1
.
2
.
3
.
Denition 13.3.20: Two 3-Sasakian manifolds (`. S) and (`
t
. S

) are said to
be isomorphic if there exist a dieomorphism 1 : ` `
t
and an oj(1)
such that 1

o
t
= o and
t
a
= (Ad

a
. where Ad denotes the adjoint action of
oj(1) on its Lie algebra sp(1).
In practice we shall always choose a basis
t
a
of the 3-Sasakian structure on
`
t
so that 1

a
=
t
a
. Now given such a dieomorphism 1 : ` `
t
it is
clear that the corresponding foliations are 1-related, that is that 1

T
1
= T
t
1
and
1

T
Q
= T
t
Q
. This induces a commutative diagram of orbifold dieomorphisms
(13.3.6)
`
F
`
t

Z
F1
Z
t

O
F
Q
O
t
.
This implies that if 1
x
is the leaf of T
Q
at r O. then 1(1
x
) is the leaf at
1
Q
(r) O
t
. that is, 1
t
F
Q
(x)
= 1(1
x
). Let G(1) denote the leaf holonomy group of
the leaf 1. Then we have G(1
t
F
Q
(x)
) G(1
x
). More generally let ((`) denote the
holonomy groupoid of the foliation T
Q
. By Theorem 2.5.10 ((`) is a Lie groupoid
and a smooth Hausdor manifold of dimension 4n + 6. We have
Proposition 13.3.21: Let 1 : ` `
t
be an isomorphism of 3-Sasakian man-
ifolds. Then 1 induces an isomorphism 1

: ((`) ((`
t
) of Lie groupoids.
We are interested in the unordered list (
1
.
2
. . . .) of holonomy groups in ((`)
up to abstract isomorphism. This list is nite if ` is complete and it provides
important invariants of a 3-Sasakian manifold. Since the leaves of the foliation T
Q
are all spherical space forms, the groups
i
are all either subgroups of oj(1) or all
subgroups of oO(3). depending on whether the Marchiafava-Romani class c of the
quaternionic Kahler orbifold O is 0 or 1. respectively. Notice that it follows from
its denition and 13.3.6 above that the class c is an invariant of the 3-Sasakian
structure on the manifold `. Indeed, c can be identied with a certain secondary
characteristic class of the foliation T
Q
. Thus, the Marchiafava-Romani class splits
490 13. 3-SASAKIAN MANIFOLDS
the isomorphism classes S of 3-Sasakian manifolds into the disjoint union S
0
+ S
1
depending on whether c is 0 of 1. A further rough classication scheme is given by
Denition 13.3.22: ` is said to be:
(i) regular if all the
i
are the identity.
(ii) of cyclic type if all the
i
are cyclic.
(iii) of dihedral type if all the
i
are either cyclic or dihedral or binary
dihedral with at least one
i
non-Abelian.
(iv) of polyhedral type if at least one of the
i
is one of the polyhedral
groups, tetrahedral, octahedral, or icosahedral (or the corresponding bi-
nary double covers) groups.
At this writing the only known examples of 3-Sasakian manifolds of polyhedral
or dihedral type are the spherical space forms `o
4n+3
and `RP
4n+3
. where is a
binary polyhedral or a binary dihedral group in the rst case and a polyhedral or a
dihedral group in the second. The action is that induced by the diagonal action of
on the quaternionic vector space H
n+1
from the left. All the known non-regular
3-Sasakian manifolds are of cyclic type and are discussed in detail in Section 13.7.
Open Problem 13.3.1: Prove the existence of 3-Sasakian structures of polyhedral
or dihedral type that are not space forms.
Theorem 13.3.23: Let ` be a complete regular 3-Sasakian manifold with c = 0.
Then ` o
4n+3
or RP
4n+3
.
For more results about regular 3-Sasakian manifolds see Section 13.4. Next we
consider an important innitesimal rigidity result. In the regular case this rigidity
is a simple consequence of the results of LeBrun [LeB88] and Nagatomo [Nag92]
(see [GS96]). In the general case it was proved by Pedersen and Poon [PP99].
Theorem 13.3.24: Complete 3-Sasakian manifolds are innitesimally rigid.
Proof. The deformation theory of 3-Sasakian manifolds is tied to the defor-
mation theory of hypercomplex manifolds studied previously in [PP98a]. Let `
be a complete 3-Sasakian manifold. Then the compact manifold o
1
` has a
natural hypercomplex structure [BGM94a]. Thus, its twistor space \ is compact
and bers holomorphically over CP
1
. Moreover, there is a holomorphic foliation on
\ whose leaves are elliptic Hopf surfaces, and whose space of leaves is the twistor
space Z associated to `. The geometry of the corresponding deformation theory
is as follows. Deformations (`
t
. o
t
) of the 3-Sasakian structure (`
0
. o
0
) on `
correspond to deformations of the hypercomplex structure on o
1
` of the form
o
1
`
t
. In turn these deformations correspond to deformations of the holomorphic
bration j : \ CP
1
. Thus, there are natural projections:
(13.3.7) \
p

CP
1
Z .
where each ber of j is a divisor in \ dieomorphic to o
1
` and is an
orbifold submersion whose leaves are elliptic Hopf surfaces. Now the product map
j : \ CP
1
Z is an orbifold submersion whose leaves are elliptic curves.
13.4. HOMOGENEOUS 3-SASAKIAN MANIFOLDS 491
The dierential of j induces the exact sequence of sheaves
0 O
W

W

CP
1 0 .
where O
W
denotes the structure sheaf of \ and denotes the holomorphic tangent
sheaf. Then using standard techniques together with the Kodaira-Baily vanishing
theorem and the orbifold version of the Akizuki-Nagano vanishing theorem, Peder-
sen and Poon show that the virtual parameter space for 3-Sasakian deformations
lies in
(13.3.8) H
0
(Z. O

) H
1
(1. O
F
) H
0
(Z.

) H
1
(1. O
F
)
H
1
(Z.

) H
0
(1. O
F
) H
1
(\. j

CP
1) .
where 1 is the generic elliptic Hopf surface o
1
oj(1). One then analyzes each
summand of 13.3.8 to show that there are no 3-Sasakian deformations. For exam-
ple, possible deformations lying in the last summand vanish by results of Horikawa,
while 3-Sasakian deformations lying in the second and third summands must pre-
serve the complex contact structure on Z. There are no such deformations in the
third summand by the Kodaira-Baily vanishing theorem. Elements in the second
summand correspond to complex contact transformations that are invariant under
the l(1) l(1) action coming from a discrete quotient of the C

principal action
on L. and there are no such elements. Finally, elements of the rst summand cor-
respond to scale changes in the o
1
factor of o
1
oj(1) and these hypercomplex
deformations do not come from 3-Sasakian ones.
While this theorem says that there is no innitesimal moduli, there can be
discrete moduli of 3-Sasakian structures. The two known examples of 3-Sasakian
manifolds, where the moduli consists at least of two structures is dues to [CEZ07]
and will be discussed later.
13.4. Homogeneous 3-Sasakian Manifolds
Since a 3-Sasakian structure consists of a 2-spheres worth of Sasakian struc-
tures or alternatively 3 orthogonal Sasakian structures, and each Sasakian structure
has an isometry group of at least one dimension, it follows that every 3-Sasakian
manifold (`. S) has a nontrivial isometry group Isom(`. o) of dimension at least
three. We wish to study the automorphism group of a 3-Sasakian structure and its
relation to Isom(`. o). The most natural denition is:
Denition 13.4.1: Let (`. S) be a 3-Sasakian manifold. We dene the auto-
morphism group Aut(`. S) of the 3-Sasakian structure by
Aut(`. S) =

S
2
Aut(o()) .
(`. S) is said to be a 3-Sasakian homogeneous space if Aut(`. S) acts tran-
sitively on `.
We now wish to give the 3-Sasakian analogue of Theorem 8.1.18. We begin
analyzing the structure of Isom(`. o) of a 3-Sasakian manifold with two lemmas.
Lemma 13.4.2: Let (`. S) be a 3-Sasakian manifold. Then
(i) there is a chain Aut(`. S) Aut(o()) Isom(`. o) of closed Lie
subgroups for any o
2
. and
(ii) Aut(`. S) = Aut(o(
1
)) Aut(o(
2
)) if
1
=
2
.
492 13. 3-SASAKIAN MANIFOLDS
(iii) the subgroup
G
3
=

S
2
G() Isom(`. o) .
where G() is the one parameter subgroup generated by (), is iso-
morphic to either oj(1) or oO(3), and commutes with the subgroup
Aut(`. S).
Proof. The subgroup property in (i) is clear, and Proposition 8.1.1 says that
Aut(o()) is a closed Lie subgroup of Isom(`. o) and the remainder follows from
(ii). By Corollary 13.2.2 it is enough to consider any two Sasakian structures
o(
1
). o(
2
) with
1
=
2
implying (ii). It follows from Proposition 13.2.1 that
the Lie algebra of G() is sp(1); hence, G() is isomorphic to either oj(1) or oO(3).
Since the one parameter groups G() lie in the center of Aut(o()) for each , the
groups G
3
and Aut(`. S) commute.
The second lemma gives a characterization of Aut(`. S).
Lemma 13.4.3: Let (`. S) be a 3-Sasakian manifold, and let Isom(`. o).
Then the following are equivalent
(i)

a
=
a
for all o = 1. 2. 3.
(ii)

a
=
a
for all o = 1. 2. 3.
(iii)

a
=
a

for all o = 1. 2. 3.
(iv) Aut(`. S).
Proof. That (ii) is equivalent to (iv) follows from Proposition 8.1.1 and (ii)
of Lemma 13.4.2. (i) and (ii) are equivalent by an easy duality argument. But (i)
and (iii) are equivalent since
a
=
a
for all o = 1. 2. 3 and preserves .
Exercise 13.4: Prove directly the innitesimal version of Lemma 13.4.3, namely.
that for any Killing vector eld A the following conditions are equivalent:
(i)
X

a
= 0. for o = 1. 2. 3.
(ii)
X

a
= 0. for o = 1. 2. 3.
(iii)
X

a
= 0. for o = 1. 2. 3.
(iv) A aut(`. S).
We are now ready for our main theorem concerning the structure of Isom(`. o).
Theorem 13.4.4: Let (`. S) be a complete 3-Sasakian manifold. If (`. S) is not
of constant sectional curvature, then
(i) isom(`. o) = aut(`. S) sp(1).
(ii) either Isom(`. o) = Aut(`. S) oj(1) or Isom(`. o) = Aut(`. S)
oO(3);
whereas if (`. o) does have constant sectional curvature then either Aut(`. S)
oj(1) or Aut(`. S) oO(3) is a proper subgroup of Isom(`. o).
Proof. Since (`. o) is complete ` is compact, so the Lie algebra isom(`. o)
is reductive. By Lemmas 8.1.17 and 13.4.2, we know that dim Isom(`. o) =
dim Aut(`. S) + 3. But then using the commutativity discussed above together
with the fact that isom(`. o) is reductive we obtain (i). But then this implies that
Isom(`. o) =

Aut(`. S) oj(1)

.
13.4. HOMOGENEOUS 3-SASAKIAN MANIFOLDS 493
where is a nite normal subgroup. By Lemma 13.4.2 Aut(`. S) oj(1) = id.
and Aut(`. S) is a subgroup of Isom(`. o). so must be a normal subgroup of
oj(1). But the only normal subgroups of oj(1) are the central Z
2
and the identity
which implies (ii).
Proposition 13.4.5: Let (`. S) be a 3-Sasakian homogeneous space. Then all
leaves are dieomorphic and O = `T
Q
is a quaternionic Kahler manifold, where
the natural projection : `O is a locally trivial Riemannian bration. Further-
more, Aut(`. S) passes to the quotient and acts transitively on the space of leaves
O.
Proof. Let : Aut(`. S) `` denote the action map so that, for each
o Aut(`. S).
a
= (o. ) is a dieomorphism of ` to itself. Theorem 13.4.4
implies that the isometry group Isom(`. o) contains Aut(`. S) G
3
. where G
3
is
either oj(1) or oO(3). Since the Killing vector elds
a
for o = 1. 2. 3 are both the
innitesimal generators of the group G
3
and a basis for the vertical distribution 1,
it follows that G
3
acts transitively on each leaf with isotropy subgroup of a point
some nite subgroup G
3
. Now let j
1
and j
2
be any two points of ` and let L
1
and L
2
denote the corresponding leaves (not necessarily distinct) through j
1
and j
2
,
respectively. Since Aut(`. S) acts transitively on `, there exists an o Aut(`. S)
such that
a
(j
1
) = j
2
. Now
a
restricted to L
1
maps L
1
dieomorphically onto
its image, and, since the oj(1) factor acts transitively on each leaf and commutes
with Aut(`. S), the image of
a
lies in L
2
. But the same holds for the inverse map

a
1 with L
1
and L
2
interchanged, so the leaves must be dieomorphic. Thus, the
leaf holonomy is trivial and : ``T
Q
= O is a locally trivial Riemannian
bration. The fact that the space of leaves O is a quaternionic Kahler manifold
now follows from Ishiharas Theorem 13.3.13. Since Aut(`. S) commutes with G
3
,
it follows that Aut(`. S) passes to the quotient and acts transitively on O.
The following classication theorem now follows from Proposition 13.4.5, the
result of Alekseevsky which states that all homogeneous quaternionic Kahler man-
ifolds of positive scalar curvature are symmetric [Ale75], and from Wolfs classi-
cation of homogeneous QK manifolds [Wol65] or from Boothbys classication of
complex homogeneous contact manifolds [Boo61]
Theorem 13.4.6: Let (`. S) be a 3-Sasakian homogeneous space. Then ` =
GH is precisely one of the following:
oj(n + 1)
oj(n)
.
oj(n + 1)
oj(n)Z
2
.
ol(:)
o

l(:2)l(1)
.
oO(/)
oO(/ 4)oj(1)
.
G
2
oj(1)
.
1
4
oj(3)
.
1
6
ol(6)
.
1
7
Spin(12)
.
1
8
1
7
.
Here n 0, oj(0) denotes the trivial group, : 3, and / 7. Hence, there is
one-to-one correspondence between the simple Lie algebras and the simply-connected
3-Sasakian homogeneous manifolds.
The dimension of the 3-Sasakian homogeneous spaces are as follows: the rst
two have dimension 4n + 3, while the second has dimension 4: 5 with : 3.
and the third 4/ 13 with / 7; the dimension of the 3-Sasakian homogeneous
spaces corresponding to the exceptional Lie groups are in the order presented
11. 31. 43. 67. 115. respectively.
494 13. 3-SASAKIAN MANIFOLDS
Below we give the fundamental diagram (GH) (G) for each 3-Sasakian
homogeneous space of Theorem 13.4.6
(13.4.1)
R
+
GH
`
G(Hl(1))

GH.
`
GHoj(1)
where G(Hl(1)) are precisely the complex homogeneous contact manifolds clas-
sied by Boothby [Boo61] and G(Hoj(1)) are the Wolf spaces [Wol65].
Remark 13.4.1: Note that a homogeneous 3-Sasakian manifold is necessarily simply-
connected with the exception of the real projective space. This is in sharp con-
trast with the Sasaki-Einstein case. Also notice that a 3-Sasakian manifold can
be Riemannian homogeneous (i.e., the full isometry group acts transitively) but
not 3-Sasakian homogeneous. This is true for the lens spaces `o
3
. with [[ 2.
For example, Z
k
`o
3
. / 2. is a homogeneous Sasaki-Einstein manifold but not
3-Sasakian homogeneous.
Theorem 13.4.6 does not specify what the 3-Sasakian metric on the coset GH
is. In Section 13.6.2 we will describe a quotient construction of the 3-Sasakian
homogeneous spaces with G = ol(n + 1) and G = oO(n + 1). Here we quote a
theorem of Bielawski [Bie96], which gives an explicit description of these metrics
in all cases.
Theorem 13.4.7: Let ` = GH be one of the spaces in Theorem 13.4.6. and let
g = h m be the corresponding decomposition of the Lie algebras. Then there is a
natural decomposition m = sp(1) m
t
and the metric o on ` is given in terms of
the scalar product on m
[[:[[
2
= '. `
1
2
':
t
. :
t
` .
where sp(1). :
t
m
t
. and '. ` is the Killing form on g. In particular, the
metric o is not naturally reductive with respect to the homogeneous structure on `.
Remark 13.4.2: In the case when ` is of constant curvature the canonical metric
on o
4n+3
(or RP
4n+3
) is not the standard homogeneous metric on the homogeneous
space oj(n + 1)oj(n) (or oj(n + 1)oj(n) Z
2
) with respect to the reductive
decomposition sp(n + 1) sp(n) + m. It is, of course, the standard homogeneous
metric with respect to the naturally reductive decomposition o(4n + 4) o(4n +
3) + m. This is quite special to the sphere and orthogonal group. In general the
3-Sasakian homogeneous metrics are not naturally reductive with respect to any
reductive decomposition.
13.5. 3-Sasakian Cohomology
In this section we describe some cohomological properties of 3-Sasakian man-
ifolds `. We discuss a vanishing theorem and then derive a relation between the
Betti numbers of ` and the Betti numbers of the QK orbifolds O. We conclude
with various implications of these relations in the regular case.
13.5. 3-SASAKIAN COHOMOLOGY 495
13.5.1. A Vanishing Theorem. Let (`. S) be a compact 3-Sasakian man-
ifold of dimension 4n + 3 and 3-Sasakian structure S. Throughout this section we
shall suppose that j 2n+1. Referring to the splitting of the tangent bundle of `
into T` = 1
3
H. we shall say that a j-form n
p
(`) has bidegree (i. ji) if it
is a section of the subbundle of

p
T

` isomorphic to the dual of

i
1
3

pi
H.
In particular, n is called 3-horizontal if it has bidegree (0. j). or equivalently if

a
|n = 0 for o = 1. 2. 3. An element
p
(`) is called invariant if /

= for
all / oj(1). In the regular case, there is a principal oj(1)-bundle : ` O.
and is both 3-horizontal and invariant if and only if it is the pullback

of a
form on the QK base O. Now the curvature forms

a
dened in the proof of
Theorem 13.3.13 are horizontal with respect to the foliation T
Q
. The Killing elds

a
transform according to the adjoint representation of G
3
, and the same is true of
the associated triples
a
, d
a
. and
a
. If / G
3
. we may write
/

a
=

b
/
ab

b
. o = 1. 2. 3.
where /
ab
are components of the image of / in G
3
. The 3-forms
(13.5.1) =
1

2

3
. =

a
=

a
d
a
+ 6
have respective bidegrees (3. 0). (1. 2). and are clearly invariant. Their exterior
derivatives are
(13.5.2) d =
1

2

3
+
2

3

1
+
3

1

2
. d = + 2d.
where the 4-form is dened in 12.2.7. In fact, is the canonical 4-form determined
by the quaternionic structure of Theorem 13.3.13 of the subbundle H. and is the
pullback of the fundamental 4-form

on the quaternionic Kahler orbifold O.
Proposition 7.4.13 implies that any harmonic j-form with j 2n + 1 on the
compact 3-Sasakian manifold ` is 3-horizontal. We can apply the construction in
Lemma 7.4.12 so as to obtain
a
: H
p
(`) H
p
(`). o = 1. 2. 3. j 2n + 1.
and
(1
a
n)(A
1
. A
2
. . . . . A
p
) = n(
a
A
1
.
a
A
2
. . . . .
a
A
p
) .
As we showed in the proof of Theorem 7.4.11 each 1
a
denes a complex structure on
the space of odd degree harmonic j-forms. Now, using the basic identities satised
by S we can prove the following quaternionic extension of this property to get
the following result originally due to Kuo [Kuo70]:
Proposition 13.5.1: Let 1
a
: H
p
(`) H
p
(`). o = 1. 2. 3. and j 2n+1. Then
1
b
1
a
= (
ab
)
p
I +

c
(c
abc
)
p
1
c
.
In particular, when j is odd, 1
1
. 1
2
. 1
3
denes an almost quaternionic structure
on the vector space H
p
(`).
We are now ready to prove the main theorem of this section.
Theorem 13.5.2: Let (`. S) be a compact 3-Sasakian manifold of dimension 4n+
3, and let n H
p
(`) with j 2n + 1.
(i) If j is odd then n = 0. that is H
p
(`) = 0.
(ii) If j is even then 1
a
n = n for o = 1. 2. 3.
496 13. 3-SASAKIAN MANIFOLDS
Proof. Let n H
p
(`). We shall in fact show that 1
1
n = 1
2
n irrespective of
whether j is even or odd; the result then follows from the identities in Proposition
13.5.1 and symmetry between the indices 1. 2. 3. We may choose an isometry /
oj(1) so that /

1
=
2
. Both n and 1
1
n are harmonic, so /

n = n and
(1
1
n)(A
1
. . . . . A
p
) = (/

(1
1
n))(A
1
. . . . . A
p
) = n((/

1
)(A
1
). . . . . (/

1
)(A
p
)) =
= n(
2
A
1
. . . . .
2
A
p
) = (1
2
n)(A
1
. . . . . A
p
) .
Corollary 13.5.3: Let (`. S) be a compact 3-Sasakian manifold of dimension
4n + 3. Then the odd Betti numbers /
2k+1
of ` are all zero for 0 / n.
Obviously Corollary 13.5.3 does not apply to compact Sasakian or even Sasaki-
Einstein manifolds as we saw in Chapters 10 and 11. In contrast to the Sasaki-
Einstein case, for example, when ` is highly connected we get
Corollary 13.5.4: Let (`. S) be a 3-Sasakian manifold that is highly connected.
Then ` is a rational homology sphere. In particular, a 3-Sasakian manifold that
is a the link of a complete intersection must be a rational homology sphere.
The only known example to date of a 3-Sasakian manifold that is the link of an
isolated hypersurface singularity or a complete intersection are certain space forms
in dimension 4n + 3.
13.5.2. 3-Sasakian Cohomology as Primitive Cohomology. We are go-
ing to consider the connection between the cohomology of ` and that of O by
applying the Vanishing Theorem 13.5.2 and a rational Gysin sequence argument to
the diagram of orbifold bundles of (`):
(13.5.3)
` Z.

O
The horizontal arrow is equivalent to the study of the transverse Kahler geom-
etry treated in Chapter 7. In particular, Theorem 7.4.14 gives the relation between
the Betti numbers /
p
(`) of ` and the basic Betti numbers /
B
p
(T

). In the quasi-
regular case the basic Betti numbers are just the Betti numbers of Z by Proposition
7.2.2. So we have /
p
(`) = /
p
(Z) /
p2
(Z) for j 2n + 1. The analogue of this
for the vertical arrow is
Proposition 13.5.5: Let (`. S) be a compact 3-Sasakian manifold of dimension
4n + 3 and let O = `T
3
. Then /
2p
(`) = /
2p
(O) /
2p4
(O) for j 2n + 1.
Proof. The result follows by applying the rational Gysin sequence to the
orbifold bration L ` O using Corollary 4.3.8. We have
H
i
(`. Q) H
i3
(O. Q)

H
i+1
(O. Q) H
i+1
(`. Q) H
i2
(`. Q)
and the statement of the proposition follows easily from the vanishing of the odd
Betti numbers of `.
13.5. 3-SASAKIAN COHOMOLOGY 497
Recall the vector space of basic primitive harmonic j-forms {
p
(T

) dened in
Section 7.2.2. In the quasi-regular case {
p
(T

) is isomorphic to vector space of

primitive harmonic j-forms H
p
0
(Z. Q) of the orbifold Z = `T

which in turn is
isomorphic to the cokernel of the injective mapping 1

: H
p2
(Z) H
p
(Z). j
2n dened by wedging with the Kahler 2-form. We dene the primitive Betti
numbers /
0
p
(Z) of Z as the dimension of H
p
0
(Z). Then as usual the primitive Betti
numbers of Z are just the Betti numbers of ` and it follows from the fact that
for, 0 : 2n + 1, an :-form on ` is harmonic if and only if it is the lift of
a primitive harmonic form on Z [BG97b]. Similarly, for j 2n + 2 the vector
space of primitive harmonic j-forms H
p
0
(O. Q) of the orbifold O is isomorphic to
the cokernel of the injective mapping 1
c
: H
p4
(O) H
p
(O) dened by wedging
with the quaternionic Kahler 4-form . The injectivity of this mapping is well-
known in the smooth case [Bon67, Fuj87, Kra65] and it extends to the orbifold
case. We dene the primitive Betti numbers /
0
p
(O) of O as the dimension of H
p
0
(O).
Proposition 13.5.5 says that the primitive Betti numbers of O are the usual Betti
numbers of `. Again, Proposition 13.5.5 is a consequence of the fact that an :-form
on ` is harmonic if and only if it is the lift of a primitive harmonic form on O.
0 : 2n + 1.
13.5.3. Regular 3-Sasakian Cohomology, Finiteness, and Rigidity. In
this Section we shall assume that ` is regular and, hence, both Z and O are smooth.
In this case, using the results of the previous section, one can easily translate all
the results about strong rigidity of positive quaternionic Kahler manifolds given
in Section 12.3 to compact regular 3-Sasakian manifolds. In particular, Theorem
12.3.3 and the rst part of Theorem 12.3.5 translate to
Proposition 13.5.6: Let (`. S) be a compact regular 3-Sasakian manifold of di-
mension 4n + 3. Then
1
(`) = 0 unless ` = RP
4n+3
and

2
(`) =

Z i ` = ol(n + 2)o(l(n) l(1)),

1initc otherwise.
Furthermore, up to isometries, for each n 1 there are only nitely many regular
3-Sasakian manifolds `.
As with quaternionic Kahler manifolds, there are very interesting relations
among the Betti numbers of regular 3-Sasakian manifolds.
Proposition 13.5.7: The Betti numbers of a regular compact 3-Sasakian manifold
(`. S) of dimension 4n + 3 satisfy
(i) /
2
(`) 1. with equality if and only if ` = ol(| + 2)o(l(|) l(1)),
(ii)
n

k=1
/(n + 1 /)(n + 1 2/)/
2k
(`) = 0.
Proof. (i) follows from Proposition 13.5.6 and (ii) for Salamons relation on
Betti numbers of O in Theorem 12.3.5 (iv) using Proposition 13.5.5.
We give below a table for the coecients in item (ii) of Proposition 13.5.7.
Notice that the coecient in (ii) changes sign under the map / n / + 1.
So there is an interesting symmetry among the coecients. Table 13.1 gives the
relations for 2 n 9 as well as the 3 remaining dimensions of the 3-Sasakian
homogeneous manifolds corresponding to the exceptional simple Lie algebras. In
this case to save space we just write the coecients of the left hand side of the
498 13. 3-SASAKIAN MANIFOLDS
equality as the relation can be determined from those. We remark that these Betti
number relations do not hold in general for non-regular 3-Sasakian manifolds as
seen in Proposition 13.7.34 below.
n relation on Betti numbers or coecients thereof
2 /
2
= /
4
3 /
2
= /
6
4 2/
2
+/
4
= /
6
+ 2/
8
5 5/
2
+ 4/
4
= 4/
8
+ 5/
10
6 5/
2
+ 5/
4
+ 2/
6
= 2/
8
+ 5/
10
+ 5/
12
7 7/
2
+ 8/
4
+ 5/
6
= 5/
10
+ 8/
12
+ 7/
14
8 28/
2
+ 35/
4
+ 27/
6
+ 10/
8
= 10/
10
+ 27/
12
+ 35/
14
+ 28/
16
9 12/
2
+ 16/
4
+ 14/
6
+ 8/
8
= 8/
12
+ 14/
14
+ 16/
16
+ 12/
18
10 15. 21. 20. 14. 5
16 40. 65. 77. 78. 70. 55. 35. 12
28 126. 225. 299. 350. 380. 391. 385. 364. 330. 285. 231. 170. 104. 35
Table 13.1
Notice that one can satisfy (ii) of Proposition 13.5.7 by the duality condition
/
2k
(`) = /
2n2k+2
(`). Such 3-Sasakian manifolds were called balanced in [GS96].
Not all the 3-Sasakian homogeneous manifolds are balanced as can be seen by
looking at the Poincare polynomials given in Table 13.2 below.
Proposition 13.5.8: The Poincare polynomials of the homogeneous 3-Sasakian
manifolds are as given in Table 13.2.
G H 1(GH. t)
ol(n + 2) ol(n)
Z
n
T
1

n
i=0
(t
2i
+t
4n+32i
)
oO(2/ + 3) oO(2/ 1) ol(2)

k1
i=0
(t
4i
+t
8k14i
)
oj(n + 1) oj(n) 1 +t
4n+3
oO(2| + 4) oO(2|) ol(2) t
2l
+t
6l+3
+

l
i=0
(t
4i
+t
8l+34i
)
1
6
ol(6) 1 +t
6
+t
8
+t
12
+t
14
+t
20
+
1
7
Spin(12) 1 +t
8
+t
12
+t
16
+t
20
+t
24
+t
32
+
1
8
1
7
1 +t
12
+t
20
+t
24
+t
32
+t
36
+t
44
+
1
4
oj(3) 1 +t
8
+t
23
+t
31
G
2
ol(2) 1 +t
11
Table 13.2
Recall the LeBrun-Salamon Conjecture 12.3.7 which says that any positive
quaternionic Kahler manifold must be symmetric. This translates to
Conjecture 13.5.9: Every regular 3-Sasakian manifold is 3-Sasakian homoge-
neous.
The two well-known classication results for positive quaternionic Kahler man-
ifolds (see Theorem 12.3.8) easily translate into:
Theorem 13.5.10: Let (`. S) be a compact regular 3-Sasakian manifold of di-
mension 4n + 3. If n < 4 then ` = GH is homogeneous, and hence one of the
spaces listed in Theorem 13.4.6.
13.6. SYMMETRY REDUCTION 499
The n = 0 case is trivial and it was an observation made by Tanno [Tan71].
The n = 1 case is based on [Hit81, FK82] and it was rst observed in [FK90,
BFGK91]. The n = 2 case is based on [PS91] and was stated in [BGM93]. Fi-
nally, the n = 3 follows from more recent classication of compact positive quater-
nion Kahler manifolds [HH02a, HH02b]. One can compute the Poincare poly-
nomials of all the 3-Sasakian homogeneous manifolds of Theorem 13.4.6. This was
done in [GS96] to which we refer for the proof.
We end this section by giving the 3-Sasakian version of a theorem of Salamon
[GS96] the proof of which involves index theory. A condensed proof for the n = 3
case can be obtained using the result of [HH02a, HH02b] together with Table
13.2, but the proof of the Herrera-Herrera part of Theorem 12.3.8 itself uses index
theory.
Proposition 13.5.11: Let (`. S) be a regular compact 3-Sasakian manifold of
dimension 4n+3. If n = 3. 4 and /
4
= 0. then ` is either a sphere o
4n+3
or a real
projective space RP
4n+3
.
13.6. Symmetry Reduction
In this section we give a general 3-Sasakian reduction procedure which con-
structs new 3-Sasakian manifolds from a given 3-Sasakian manifold (`. S) with a
non-trivial 3-Sasakian isometry group Aut(`. S) [BGM94a]. Actually, this is a
reduction that is associated with a quadruple of spaces of the fundamental diagram
(`. S). At the level of the hyperkahler cone C(`) the reduction was discov-
ered by Lindstrom and Rocek [LR83] in the context of supersymmetric -model
and later rigorously described by Hitchin et al. in [HKLR87]. In the case of the
quaternionic Kahler base O the reduction was discovered by the second author and
Lawson [Gal87a, GL88]. In this section we restrict ourselves to describing the
general procedure of reduction together with the homogeneous case arising from
reduction by a circle group, as well as a brief description of the singular case. The
large class of 3-Sasakian toric manifolds obtained by reduction is relegated to a
separate section, namely Section 13.7. It should also be understood that every
3-Sasakian reduction gives as well a reduction procedure for each of the spaces of
the fundamental diagram (`. S).
13.6.1. The 3-Sasakian Moment Map. Let (`. S) be a 3-Sasakian mani-
fold with a nontrivial group Aut(`. S) of 3-Sasakian isometries. By the Denition
13.1.9, C(`) = ` R
+
is a hyperkahler manifold with respect to the cone metric
o. The isometry group Aut(`. S) extends to a group Aut(C(`). I. o)

= Aut(`. S)
of isometries on C(`) by dening each element to act trivially on R
+
. Further-
more, it follows easily from the denition of the complex structures 1
a
that these
isometries Aut(C(`). I. o) are hyperkahler; that is, they preserve the hyperkahler
structure on C(`). Any subgroup G Aut(C(`). I. o) gives rise to a hyperkahler
moment map j : `g

R
3
. where g denotes the Lie algebra of G and g

is its
dual (see the discussion of Section 12.8). Thus, we can dene a 3-Sasakian moment
map
(13.6.1) j
S
: ` g

R
3
by restriction j
S
= j [
S
. We denote the components of j
S
with respect to the
standard basis of R
3
. which we have identied with the imaginary quaternions,
by j
a
S
. Recall that ordinarily moment maps determined by Abelian group actions
500 13. 3-SASAKIAN MANIFOLDS
(in particular, those associated to 1-parameter groups) are only specied up to an
arbitrary constant. This is not the case for 3-Sasakian moment maps since we
require that the group oj(1) generated by the Sasakian vector elds
a
acts on the
level sets of j
S
. However, we shall see that 3-Sasakian moment maps are given by
a particularly simple expression.
Proposition 13.6.1: Let (`. S) be a 3-Sasakian manifold with a connected com-
pact Lie group G acting on ` by 3-Sasakian isometries. Let be an element of
the Lie algebra g of G and let A

denote the corresponding innitesimal isometry.

Then there is a unique 3-Sasakian moment map j
S
such that the zero set j
1
S
(0)
is invariant under the group oj(1) generated by the vector elds
a
. This moment
map is given by
'j
a
S
. ` =
1
2

a
(A

) .
Furthermore, the zero set j
1
S
(0) is G invariant.
Proof. Using the Denition 13.1.9 we can dene the 2-forms
a
S
on ` as
the restriction of the hyperkahler 2-forms
a
. Then any 3-Sasakian moment map
j
a
S
() determined by g satises 2dj
a
S
() = 2A

a
S
= A

d
a
. As A

is a 3-Sasakian innitesimal isometry, Exercise 13.4 implies that 2'j

a
S
. ` diers
from
a
(A

) by a constant depending on o and . One then uses the invariance of

the zero set j
1
S
(0) to show that these constants must vanish. See [BGM94a] for
details.
Henceforth by the 3-Sasakian moment map, we shall mean the moment map
j
S
determined in Proposition 13.6.1. Hence, the Denition 13.1.9 and Proposition
13.6.1 imply
Theorem 13.6.2: Let (`. S) be a 3-Sasakian manifold with a connected com-
pact Lie group G Aut(`. S) acting on ` smoothly and properly. Let j
S
be
the corresponding 3-Sasakian moment map and assume both that 0 is a regular
value of j
S
and that G acts freely on the submanifold j
1
S
(0). Furthermore, let
: j
1
S
(0) ` and : j
1
S
(0)j
1
S
(0)G denote the corresponding embedding
and submersion. Then (`G = j
1
S
(0)G. o) is a smooth 3-Sasakian manifold
of dimension 4(n dim g) + 3 with metric o and characteristic vector elds

a
(hence, the reduced 3-Sasakian structure

S) determined uniquely by the two condi-
tions

o =

o and

(
a
[ j
1
S
(0)) =

a
.
We conclude this part with the following fact concerning 3-Sasakian isometries
whose proof can be found in [BGM94a].
Proposition 13.6.3: Assume that the hypothesis of Theorem 13.6.2 holds. In ad-
dition assume that (`. S) is complete and hence compact. Let C(G) Aut(`. S)
denote the centralizer of G in Aut(`. S) and let C
0
(G) denote the subgroup of
C(G) given by the connected component of the identity. Then C
0
(G) acts on the
submanifold j
1
S
(0) as isometries with respect to the restricted metric

o and the
3-Sasakian isometry group Aut(`G.

S) of the quotient (`G.

S) determined in
Theorem 13.6.2 contains an isomorphic copy of C
0
(G). Furthermore, if C
0
(G) acts
transitively on `G. then `G is a 3-Sasakian homogeneous space.
It should be mentioned that it is not required that the 3-Sasakian isometry
group Aut(`G.

S) of the reduced space acts eectively.
13.6. SYMMETRY REDUCTION 501
13.6.2. Regular Quotients And Classical Homogeneous Metrics. We
now apply the reduction procedure given in Theorem 13.6.2 to the round unit sphere
o
4n+3
to explicitly construct the Riemannian metrics for the 3-Sasakian homoge-
neous manifolds arising from the simple classical Lie algebras. These metrics are
precisely the ones associated to the three innite families appearing in Theorem
13.4.6. The quotient construction applied to (o
4n+3
) explicitly describes all met-
rics in the fundamental diagrams (GH). where G is either the special unitary
ol(n + 1) or the orthogonal group oO(n + 1). To carry out this reduction we
must set some conventions. We describe the unit sphere o
4n+3
by its embedding
in at space and we represent an element u = (n
1
. . . . . n
n+1
) H
n+1
as a column
vector. The quaternionic components of this vector are denoted by u
0
for the real
component and by u
a
for the three imaginary components so that we can write
u = u
0
+ iu
1
+ ,u
2
+ /u
3
using the quaternionic units i. ,. /. We also dene
quaternionic conjugate u = u
0
iu
1
,u
2
/u
3
.
Now, the innitesimal generators of the subgroup oj(1) H

acting by the
right multiplication on u are the dening vector elds
a
for the Sasakian 3-
structure. These vector elds are given by

r
a
= u
0

u
a
u
a

u
0
c
abc
u
b

u
c
.
where the dot indicates sum over the vector components n
i
and the subscript :
means that these vector elds are the generators of the right action.
We will rst consider G = l(1) acting on the sphere o
4n+3
as follows

t
(u) = u. = c
2it
. u o
4n+3
.
Note that this action on the cone C(o
4n+3
) = H
n+1
` 0 has been already con-
sidered in the Example 12.8.5, where the corresponding hyperkahler moment map
was computed in (12.8.3). The 3-Sasakian moment map j
S
: o
4n+3
Rsp(1)
is the restriction
j
S
(u) =
n+1

=1
n

in

.
Let be the zero locus of j
S
. Using the complex vectors (x. y) C
n+1
C
n+1
as in
(12.8.5) one can then easily identify the zero-level set of the moment map with the
Stiefel manifold of complex 2-frames in C
n+1
, i.e., \
C
n+1,2
= l(n+1)l(n1).
The following proposition is an immediate consequence of the reduction theorem
Proposition 13.6.4: Let = j
1
S
(0) and : o
4n+3
be the inclusion. Then
is an embedding and (.

o
can
) is the complex homogeneous Stiefel manifold
\
C
2,n+1
= ol(n+1)ol(n1) of 2-frames in C
n+1
. Hence, the 3-Sasakian quotient
o
4n+3
l(1) = \
C
2,n+1
l(1) = ol(n +1)o(l(n 1) l(1)) with the 3-Sasakian
metric o given by inclusion and submersion : l(1), i.e.,

o
can
=

o.
Example 13.6.5: A similar construction can be carried out for the oj(1)-action
on o
4n+3
dened by the left multiplication of u by a unit quaternion , i.e.,

(u) = u. = 1. u o
4n+3
.
This action is free on o
4n+3
and the zero-level set of the corresponding moment
map can be identied with the real Stiefel manifold \
R
4,n+1
oO(n+1)oO(n3)
of oriented 4-frames in R
n+1
with n 4. Hence, the reduced space o
4n+3
oj(1) =
502 13. 3-SASAKIAN MANIFOLDS
SO(n+1)
SO(n3)Sp(1)
. For the more detailed and uniform description of the geometry of
these two quotients see [BGM94a].
13.6.3. The Structure of Singular Quotients. In this section we will de-
scribe a more general situation, when the zero-level set of the 3-Sasakian moment
map (13.6.1) is not necessarily smooth and the group action on the level set is not
necessarily locally free.
Let G be a Lie group acting smoothly and properly on a manifold o and let
H G be a subgroup. Using standard notation we will denote by o
H
o the set
of points in o. where the stability group is exactly equal to H and by o
(H)
o the
set of points with stabilizer conjugate to H in G. It follows than that the normalizer
(H) of H in G acts freely on o
(H)
. Then we have the following theorem due to
Dancer and Swann [DS97a]:
Theorem 13.6.6: Let (o. o) be a 3-Sasakian manifold with a connected compact
Lie group G acting on o smoothly and properly by 3-Sasakian isometries. Let j
S
be the corresponding 3-Sasakian moment map. Then the quotient j
1
S
(0)G is a
union of the smooth, 3-Sasakian manifolds (o
(H)
j
1
S
(0))G. where (H) runs over
the conjugacy classes of stabilizers of points in o.
Quite often o
(H)
does not meet the zero locus of the moment map. Then the
stratum (o
(H)
j
1
S
(0))G is empty.
Example 13.6.7: We start with the 3-Sasakian sphere o
4n+3
in the notation of the
previous section. But now we consider a dierent l(1)-action

p,q;m
t
(u) = (
p
n
1
. . . . .
p
n
m
.
q
n
m+1
. . . . .
q
n
n+1
). = c
2it
. u o
4n+3
.
where 0 : n + 1 and j. c Z. Let o(j. c; :) = o
4n+3
l(1) be the quotient.
(i) First, let j. c be relatively prime positive integers bigger than 1 and 2 :
n 1. Then, the stratied manifold o(j. c; :) consists of 3 strata. The stratum of
the highest dimension corresponding to H = id is an open incomplete 3-Sasakian
manifold. The two strata of lower dimension are easily seen to be the homogeneous
spaces: one with H = Z
q
is the homogeneous 3-Sasakian space of ol(n + 1 :)
and the one with H = Z
p
is the homogeneous 3-Sasakian space of ol(:). In this
case, o(j. c; :) is actually a compact 3-Sasakian orbifold and the stratication of
Theorem 7.3.1 coincides with the orbifold stratication.
(ii) Consider o(0. j; :). where j 1 and 2 : n 1. There are two strata
now: the stratum of the highest dimension corresponds to H = Z
p
and the second
stratum is just the sphere o
4m1
with H = l(1). The space o(0. j; :) is not an
orbifold but, as pointed out in [DS97a], it does have a length space structure.
(iii) Consider o(0. 1; n). Here H is either l(1) or trivial but the set o
4n+3
id
does
not meet the zero locus of the moment map. Hence, there is only one stratum and
o(0. 1; n) = o
4n1
.
(iv) Finally, consider o(0. 1; n 1). The stability group H is either l(1) or trivial.
The stratum corresponding to H = l(1) is the sphere o
4n5
. We leave it as an
exercise to the reader to show that o(0. 1; n 1) is an orbifold and that it can be
identied with o
4n1
Z
2
. where (n
1
. . . . . n
n
) o
4n1
. where Z
2
acts on the last
quaternionic coordinate by multiplication by 1.
Example 13.6.8: [Kobak-Swann Quotient] This is a less trivial example of a
singular quotient which we examine in considerable detail. Recall that G
2
oO(4)
is one of the three possible models of positive quaternionic Kahler manifolds in
13.6. SYMMETRY REDUCTION 503
dimension 8. It turns out that this space, as well as the associated homogeneous
3-Sasakian manifold can be obtained via symmetry reduction. Before we describe
the geometry of this quotient construction let us briey discuss some of the facts
about the two homogeneous Sasaki-Einstein geometries that are naturally associ-
ated with the exceptional Lie group G
2
. They both come from the classical Lie
group isomorphism between oO(4) G
2
and oj(1)

oj(1)
+
G
2
. The two
oj(1)

subgroups are very dierent. One of them has index 1 in G

2
and the
other one has index 3. Consequently, the quotients are not of the same homo-
topy type as can be seen from the exact sequence in homotopy for the bration
oj(1)

G
2
G
2
oj(1)

. In particular, the two spaces can be distin-

guished by their third homotopy groups being trivial in one case and Z
3
in the
other. One of these quotients, which we shall denote by G
2
oj(1)

is dieomor-
phic to the real Stiefel manifold \
7,2
(R) = oO(7)oO(5) of 2-frames in R
7
[HL82].
As \
7,2
(R) is 4-connected
3
(G
2
oj(1)

) = 0. The other quotient denoted here

by G
2
oj(1)
+
is one of the 11-dimensional 3-Sasakian homogeneous spaces and

3
(G
2
oj(1)
+
) = Z
3
. G
2
oj(1)
+
bers as a circle bundle over a generalized ag
Z = G
2
l(2)
+
. which in turn is well-known to be the twistor space of the ex-
ceptional 8-dimensional Wolf space G
2
oO(4). The second homogeneous Sasaki-
Einstein manifold is a circle bundle over the complex ag G
2
l(2)

which can be
identied with the complex quadric in the 6-dimensional complex projective space
CP
6
or, equivalently, the real Grassmannian Gr
2
(R
7
) = oO(7)oO(2) oO(5) of
oriented 2-planes in R
7
. We have the following diagram of Riemannian submersions:
(13.6.2)
G
2
`
G2
Sp(1)
+
G2
Sp(1)

\
7,2
(R)

Z =
G
2
U(2)+
G
2
U(2)
Gr
2
(R
7
)
`
G
2
SO(4)
Kobak and Swann [KS93a] proved the following:
Theorem 13.6.9: The quaternionic Kahler manifold Gr
4
(R
7
) admits a l(1)-action
such that the quaternionic Kahler quotient is a compact quaternionic Kahler orb-
ifold O = O
r
CP(2) = G
2
(oO(4) Z
3
).
One can easily reinterpret the Kobak-Swann construction in the language of
3-Sasakian geometry of the Konishi oO(3)-orbibundle over O (or, equivalently, the
hyperkahler geometry of the regular nilpotent orbit of sl(3. C)). In particular, we
have the following:
Theorem 13.6.10: The 3-Sasakian homogeneous manifold oO(7)oO(3) oj(1)
admits a l(1)-action such that the 3-Sasakian quotient is a compact 3-Sasakian
orbifold =
r
ol(3)l(1) = Z
3
`G
2
oj(1)
+
.
Theorem 13.6.10 is a straightforward translation of Theorem 13.6.9 into the
language of 3-Sasakian geometry and we could leave it at that. However, we will
504 13. 3-SASAKIAN MANIFOLDS
outline a constructive proof of this result as we will need it to describe an interesting
modication of this particular quotient [BGP02].
Using the Example 13.6.5 one can think of the homogeneous 3-Sasakian man-
ifold oO(7)oO(3) oj(1) as the 3-Sasakian reduction o
4n1
oj(1) : Let u =
(n
1
. . . . . n
7
) o
27
. Consider the oj(1) action given by multiplication by unit
quaternion oj(1) on the left that is
(13.6.3)

(u) = u.
In the i. ,. / basis the 3-Sasakian moment maps for this action read:
(13.6.4) j
i
(u) =
7

=1
n

in

. j
j
(u) =
7

=1
n

,n

. j
k
(u) =
7

=1
n

/n

.
Then, the common zero-locus of the moment maps
(13.6.5) = u o
4n1
[ j
i
(u) = j
j
(u) = j
k
(u) = 0
is the Stiefel manifold oO(7)oO(3) = \
4
(R
7
) of the orthonormal 4-frames
in R
7
and the corresponding 3-Sasakian quotient o = oj(1) is Konishis RP
3
-
bundle over the real Grassmannian of oriented 4-planes in R
7
. We can combine
Theorem 13.6.10 with this description to get
Corollary 13.6.11: The 3-Sasakian sphere o
27
admits an action of l(1) oj(1)
such that the 3-Sasakian quotient is a compact 3-Sasakian orbifold =
r

ol(3)l(1) = Z
3
`G
2
oj(1)
+
.
We now turn to the explicit description of the l(1) quotient. Consider the
following subgroups of the group of 3-Sasakian isometries of the 27-sphere:
(13.6.6) oj(7) oO(7) 1 oO(6) 1 l(3).
where l(1) l(3) is the central subgroup. Explicitly, 1 : [0. 2) oO(7)
1(t) =

1 0 0 0
0 (t) 0 0
0 0 (t) 0
0 0 0 (t)

. (t) =

cos t sin t
sin t cos t

oO(2).
The homomorphism 1(t) yields a l(1)-action on o
27
and, after performing oj(1)
reduction rst, on the homogeneous 3-Sasakian manifold oO(7)oO(3) oj(1) via
left multiplication 1(t)u and the associated 3-Sasakian moment map can be written
as
(13.6.7) (u) =

=1,2,3
(n
2
n
2+1
n
2+1
n
2
) .
Note here that (u) does not depend on the n
1
quaternionic coordinate.
Denition 13.6.12: Let us dene the zero level set of this new moment map in-
tersected with , that is

1
(0).
First we observe, following Kobak and Swann [KS93a] that
Lemma 13.6.13: The manifold

can be identied with l(1)G

2
= (o
1
G
2
)Z
3
.
where l(1) G
2
= Z
3
.
Proof. The argument is similar here to the one used by Kobak and Swann in
[KS93a] and it is based on the Proposition 1.10 of [HL82]. First, using the basis
13.6. SYMMETRY REDUCTION 505
i. ,. / of unit imaginary quaternions, we write n

= n
0

+ in
1

+ ,n
2

+ /n
3

and
introduce the 4 7 real matrix
(13.6.8) A =

n
0
1
n
0
2
n
0
3
n
0
4
n
0
5
n
0
6
n
0
7
n
1
1
n
1
2
n
1
3
n
1
4
n
1
5
n
1
6
n
1
7
n
2
1
n
2
2
n
2
3
n
2
4
n
2
5
n
2
6
n
2
7
n
3
1
n
3
2
n
3
3
n
3
4
n
3
5
n
3
6
n
3
7

1
0
1
1
1
2
1
3

.
where to make the connection with the notation in [KS93a] we also think of the
rows of A as purely imaginary octonions Im(O). In the standard basis of Im(O)
we write 1
a
= n
a
1
i + n
a
2
, + n
a
3
/ + n
a
4
c + n
a
5
ic + n
a
6
,c + n
a
7
/c. Let (o. /. c) = 'o/. c`
denote the 3-form dening the associative calibration [HL82] on Im(O). where
'. ` denotes the standard Euclidean inner product. Then writing =
1
i +
2
, +

3
/ a straightforward computation shows that for o = 1. 2. 3 and c
abc
the totally
antisymmetric tensor satisfying c
123
= 1
(13.6.9)
a
= 2'1
0
1
a
+c
abc
1
b
1
c
. i` .
where the summation convention on repeated indices is used. Now, u if and
only if the rows 1
0
. 1
1
. 1
2
. 1
3
of A form an orthonormal frame in R
7
Im(O). and
one can identify G
2
with a special kind of oriented orthonormal 4-frame, namely
those which are co-associative. This means that the 3-plane that is orthogonal to the
4-plane dened by the frame 1
0
. 1
1
. 1
2
. 1
3
is spanned by an associative subalgebra
of Im(O). Then one shows that these special 4-frames satisfy the l(1)-moment map
equation (u) = 0 and, hence, l(1) G
2

1
(0). As, l(1) G
2
= Z
3
it
is enough to show that by acting with l(1) one gets the whole
1
(0). The
argument is similar to the one presented in [KS93a]. (See [KS93a] Lemma 5.1
and the discussion that follows.)
Now, Theorems 13.6.9 and 13.6.10 and Corollary 13.6.11 all follow from the
above lemma as we get the quotient
(13.6.10) =

l(1) oj(1)

l(1) G
2
l(1) oj(1)
Z
3
`G
2
oj(1).
We will examine an interesting generalization of this quotient construction later in
Section 13.9.1.
Remark 13.6.1: The l(1) oj(1) action on the level set

is not locally free. If

we divide by oj(1) rst and consider the l(1)-action on the orbit space

oj(1)
this circle action is quasi-free. This means that there are only two kinds orbits:
regular orbits with the trivial isotropy group and singular orbits (points) where
the isotropy group is the whole l(1). In such cases the quotient space is often an
orbifold (or even a smooth manifold). The stratication of the Theorem 13.6.10 is
precisely with respect to the orbit types as will be seen in the next section.
Question 13.6.1: It has been pointed out by Kobak and Swann in [KS98] that
up to quotients by a nite group the 3-Sasakian manifold 1
4
oj(3) (or, equiva-
lently, the Wolf space 1
4
oj(3) oj(1)) can be also obtained via symmetry reduc-
tion, but the question remains open for the three homogeneous spaces of 1-series
1
6
ol(6). 1
7
ojin(12). and 1
7
1
8
. Can these be obtained via symmetry reduc-
tion of a sphere (up to a quotient by a nite group)?
506 13. 3-SASAKIAN MANIFOLDS
13.7. Toric 3-Sasakian Manifolds
In this section we shall describe the quotient construction of large families of
3-Sasakian manifolds (and orbifolds) o() with a 3-Sasakian structure S

which
are Konishi orbibundles over the toric orbifolds O() of Theorem 12.5.1. These
are characterized by the property that Aut(o(). S

) T
m
. where dim(o()) =
4:1. In particular, the hyperkahler cone C(o()) is toric in the sense of Denition
12.9.1 so that, following the ideas of [BD00], we shall call such 3-Sasakian manifolds
toric. Let us point out, however, that a toric 3-Sasakian manifold is not a toric
contact (Sasakian) manifold as dened in 8.4.32, but it is toric in the quaternionic
sense. Much of the treatment in this section is taken from [BGMR98, BGM98b]
and [BGM94a].
13.7.1. Toral Reductions of Spheres. Consider the torus action dened
at the beginning of Section 12.5, only lifted to H
n+1
and then restricted to the
embedded unit (4n + 3)-dimensional sphere o
4n+3
embedded in the quaternionic
vector space H
n+1
. Every quaternionic representation of a /-torus T
k
acting on
H
n+1
as a subgroup of oj(n + 1) can be described by an exact sequence (12.5.2)
or its dual sequence (12.5.3). The monomorphism 1

: T
k
oj(n + 1) can be
represented by the diagonal matrix as in Equation (12.5.4) with the weight matrix
= (o
l

)
l=1,...,k
=1,...,n+1

k,n+1
(Z).
Let c
l

k
l=1
denote the standard basis for t

k
R
k
. Then the 3-Sasakian moment
map j

: o
4n+3
t

k
R
3
of the /-torus action dened by 1

is given by j

l
j
l

c
l
with
j
l

(u) =

io
l

.
Let us further denote the triple (T
k
. 1

.
(1,...,
k
)
) by T
k
().
Denition 13.7.1: () = j
1

(0) and o() = o

4n+3
T
k
() = ()T
k
().
Let o
4n+3
H
denote all the points on the sphere, where the stability group H T
k
is exactly H. Because T
k
is Abelian o
4n+3
H
= o
4n+3
(H)
. Furthermore, let 1
H
= T
k
H
and denote by o(; H) = o
4n+3
H
()1
H
. Following Theorem 13.6.6 we have
Proposition 13.7.2: The quotient o() =

H
o(; H) is a disjoint union of 3-
Sasakian manifolds, where each stratum o(; H) is smooth.
We will be interested in the case when o() is a compact orbifold (all stability
groups H for which o
4n+3
H
() are non-empty are discrete) or a compact smooth
manifold (there is only one stratum). Necessary and sucient conditions for this to
happen can be expressed in terms of properties of the matrix . First observe that,
without loss of generality, we can assume that the rank of equals /. Otherwise,
one simply has an action of a torus of lower dimension and the whole problem
reduces to considering another weight matrix with fewer rows.
We introduce the following terminology: consider the

n
k

minor determinants
(13.7.1)
1...
k
= det

o
1
1
. . . o
1

k
.
.
.
.
.
.
o
k
1
. . . o
k

obtained by deleting n + 1 / columns of .

Denition 13.7.3: Let
k,n+1
(Z) be the weight matrix.
13.7. TORIC 3-SASAKIAN MANIFOLDS 507
(i) If

k
= 0. 1
1
< <
k
n + 1, then we say that is
non-degenerate.
(ii) Suppose is non-degenerate and let o be the /
th
determinantal divisor,
i.e., the ocd of all the / by / minor determinants
1
k
. Then is
said to be admissible if in addition we have
gcd(
2
k+1
. ....
1 s
k+1
. ....
1
k
) = o
for all sequences of length (k+1) such that 1
1
< <
s
< <

k+1
n + 1.
13.7.2. Equivalence Problem and Admissibility. Before we show how
these properties of the matrix impact on the geometry of the quotient o()
we need to discuss the notion of the equivalence of T
k
-actions on o
4n+3
and ob-
tain a normal form for admissible weight matrices. We are free to change bases
of the Lie algebra t
k
. This can be done by the group of unimodular matrices
G1(/. Z). Moreover, if we x a maximal torus T
n+1
of oj(n + 1). its normal-
izer, the Weyl group J(oj(n + 1))
n+1
> (Z
2
)
n+1
. preserves the 3-Sasakian
structure on o
4n+3
and intertwines the T
k
actions. Thus, there is an induced action
of G1(/. Z) J(oj(n + 1)) on the set of weight matrices
k,n+1
(Z). The group
G1(/. Z) acts on
k,n+1
(Z) by matrix multiplication from the left, and the Weyl
group J(oj(n + 1)) acts by permutation and overall sign changes of the columns.
Actually we want a slightly stronger notion of equivalence than that described
above. If the i
th
row of has a gcd d
i
greater than one, then by reparameterizing
the one-parameter subgroup
t
i
=
di
i
we obtain
a
j

i
= (
t
i
)
b
i

. where gcd/
i

= 1.
So the action obtained by using the matrix whose i
th
row is divided by its gcd d
i
is the same as the original action. The integers d
i
all divide the /th determinantal
divisor o. We say that a non-degenerate matrix is in reduced form (or simply re-
duced) if o = 1. The following easy lemma says that among non-degenerate matrices
it is sucient to consider matrices in a reduced form.
Lemma 13.7.4: Every non-degenerate weight matrix is equivalent to a matrix
in a reduced form.
Henceforth, we shall only consider matrices in a reduced form.
Denition 13.7.5: Let /
k,n+1
(Z)
k,n+1
(Z) denote the subset of reduced ad-
missible matrices. This subset is invariant under the G1(/. Z) J(oj(n + 1))-
action so the set /
k,n+1
(Z)G1(/. Z) J(oj(n +1)) of equivalence classes [] is
well-dened. We let c
k,n+1
(Z) /
k,n+1
(Z) denote a fundamental domain for the
action.
Our interest in /
k,n+1
(Z) is the following:
Theorem 13.7.6: Let o() be the quotient space of denition 13.7.1. Then
(i) if is non-degenerate, o() is an orbifold.
(ii) If is degenerate, then either o() is a singular stratied space which
is not an orbifold or it is an orbifold obtained by reduction of a lower
dimensional sphere o
4n4r1
by a torus T
kr
(
t
) or a nite quotient
of such, where 1 : / and
t
is non-degenerate. (When : = / the
quotient is the sphere o
4n4k1
).
(iii) Assuming that is non-degenerate o() is a smooth manifold if and only
if is admissible.
508 13. 3-SASAKIAN MANIFOLDS
One can easily see that the non-degeneracy of is not necessary for the quotient
space o() to be smooth or a compact orbifold (see Example 13.6.7(iv)). However,
Theorem 13.7.6(ii) shows that then we can reformulate the whole problem in terms
of another quotient and a new non-degenerate weight matrix
t
and can be found in
[BGM98b]. Theorem 13.7.6(iii) shows then the importance of admissible matrices
in the construction and it easily follows from the fact that non-degeneracy implies
that at most n/ quaternionic coordinates n
j
can simultaneously vanish on ()
[BGMR98].
Remark 13.7.1: Our discussion shows clearly that, if .
t
/
k,n+1
(Z) such that
[] = [
t
] then the quotients o() o(
t
) are equivalent as 3-Sasakian manifolds.
We believe that the converse of this is also true, though we will establish it later
only in certain cases.
13.7.3. Combinatorics and Admissibility. In general Theorem 13.7.6(iii)
is not yet an existence theorem, since /
k,n+1
(Z) could be empty. Indeed, for many
pairs (/. n) this is the case and we shall demonstrate this next.
Let /
k,n+1
(Z). Since is reduced there is a / by / minor determinant
that is odd. By permuting columns if necessary this minor can be taken to be the
rst / columns. Now consider the mod 2 reduction
k,n+1
(Z)
k,n+1
(Z
2
). We
have the following commutative diagram
(13.7.2)
G1(/. Z)
k,n+1
(Z)
k,n+1
(Z)

G1(/. Z
2
)
k,n+1
(Z
2
)
k,n+1
(Z
2
) .
Let

/
k,n+1
(Z
2
) denote the mod 2 reduction of /
k,n+1
(Z). Since the rst
/ by / minor determinant of is odd, the mod 2 reduction of this minor in

is
invertible. Thus, we can use the G1(/. Z
2
) action to put

in the form
(13.7.3)

=

1 0 . . . 0 o
1
k+1
. . . o
1
n+1
0 1 . . . 0 o
2
k+1
. . . o
2
n+1
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
0 0 1 o
k
k+1
. . . o
k
n+1

with o
i
j
Z
2
.
Lemma 13.7.7: The set /
k,n+1
(Z) is empty for n / + 1 and / 4.
Proof. The second admissibility condition is equivalent to the condition that
every / by / + 1 submatrix of

has rank /. By considering / 1 of the rst /
columns and 2 of last n+1 / columns, this condition implies (o
j
l
. o
j
m
) = (0. 0) for
all , = 1. . . . . /. and / + 1 | < : n + 1. Similarly, by considering / 2 of the
rst / columns and 3 of last n + 1 / columns, (13.7.3) implies
(13.7.4)

o
i
l
o
i
m
o
i
r
o
j
l
o
j
m
o
j
r

1 1 1
1 1 1

o
i
l
o
i
m
o
i
r
o
j
l
o
j
m
o
j
r

0 1 1
0 1 1

.
where the last inequality is understood to be up to column permutation. Hence, it
follows that, up to column and row permutations, that any four triples of the last
13.7. TORIC 3-SASAKIAN MANIFOLDS 509
n / columns of an admissible

must have the form
(13.7.5)

1 1 1
0 1 1
1 0 1
1 1 0

.
So we see that we cannot add another row without violating the above condi-
tions. It follows that / 4.
Similar analysis shows that
Lemma 13.7.8: The set /
k,n+1
(Z) is empty if / 1 and n / 4.
Remark 13.7.2: In view of the above lemmas and the fact that in the remainder
of this section we will be interested only in the smooth and compact quotients we
are left with the following possibilities:
(i) Trivial case of n = /. Then there are many admissible matrices but
dim(o()) = 3 and it follows that o() = o
3
Z
p
. where j = j()
depends on . This case is of little interest.
(ii) Bi-quotient geometry with / = 1 and n 1 arbitrary. Here is just
a row vector p. The admissibility condition means that the entries are
non-zero and pairwise relatively prime. The quotient o(p) turns out to
be a bi-quotient of the unitary group l(n + 1) and we shall discuss its
geometry and topology in the next subsection.
(iii) The most interesting, 7-dimensional case of / = n1. Here one easily sees
that there are many admissible matrices and we analyze the geometry
and topology of the quotients in a separate subsection. In particular,
when / = 1 we will completely determine the dieomorphism type of the
reduced space o(j
1
. j
2
. j
3
) showing that there are examples of 3-Sasakian
7-manifolds with more than one 3-Sasakian structure S.
(iv) Special quotients with: (/. n) = (2. 4). (2. 5). (3. 5). (3. 6). (4. 6). (4. 7).
These quotients are 11- or 15-dimensional and we give examples of ad-
missible weight matrices in each case. We shall show also that they
provide counterexamples to certain Betti number relations of Theorem
13.5.7 which are always satised in the regular case.
13.7.4. 3-Sasakian Structures as Bi-Quotients. When / = 1 we have
= p = (j
1
. . . . . j
n+1
) and we shall write o() = o(p). () = (p). 1

= 1
p
.
j

= j
p
. and
1
= . The quotients o(p) are generalizations of the homogeneous
examples discussed in Section 13.6.2. We get
/
1,n+1
(Z) = p (Z)
n+1
[ j
i
= 0 i = 1. . . . . n+1 and gcd(j
i
. j
j
) = 1 i = ,.
c
1,n+1
(Z) = p Z
n+1
[ 0 < j
1
j
n+1
and gcd(j
i
. j
j
) = 1 i = , .
Note that c
1,n+1
(Z) can be identied with a certain integral lattice in the positive
Weyl chamber in t

n+1
.
First, by studying the geometry of the foliations in the diagram (o(p))
[BGM96a] one can solve the equivalence problem in this case. We get [BGM96b]:
Proposition 13.7.9: Let n 2 and p. q /
1,n+1
(Z) so the quotients o(p) and
o(q) are smooth manifolds. Then o(p) o(q) are 3-Sasakian equivalent if and
only if [p] = [q].
510 13. 3-SASAKIAN MANIFOLDS
It is easy to see that for p /
1,n+1
(Z) the zero locus of the moment map (p)
is always dieomorphic to the Stiefel manifold \
C
2,n+1
of complex 2-frames in C
n+1
.
More generally,
Proposition 13.7.10: For each p (Z

)
n
the zero locus (p)is dieomorphic to
the complex Stiefel manifold \
C
n,2
.
Hence, we have the quotient o(p) = \
C
2,n+1
o
1
. We rst observe that one
can identify \
C
2,n+1
with the homogeneous space l(n + 1)l(n 1). Using this
identication we have
Proposition 13.7.11: For each p c
1,n+1
(Z). there is an equivalence o(p)
l(1)
p
`l(n +1)l(n 1) as smooth l(1)
p
l(n 1)-spaces, where the action of
l(1)
p
l(n 1) l(n + 1)
L
l(n + 1)
R
is given by the formula

p
,B
(W) = 1
p
()W

1l
2
O
O B

.
Here W l(n + 1) and (. B) o
1
l(n 1).
Note that the identication o(p) l(1)
p
`l(n+1)l(n1) is only true after
assuming that all the weights are positive, as the right-hand side is not invariant
under such sign changes. Proposition 13.7.11 shows that, in a way, the quotients
o(p) can be though of as a discrete bi-quotient deformation of the homogeneous
model o(1). Now let
p
: (p) o
4n+3
be the inclusion and
p
: (p)
o(p) be the Riemannian submersion of the moment map. Then the metric o(p) is
the unique metric on o(p) that satises

p
o
can
=

p
o(p). Similarly we dene the
other structure tensors of the 3-Sasakian structure S
p
. Using the geometry of the
inclusion
p
one can show the following [BGM96b, BGM96a]
Theorem 13.7.12: Let Aut(o(p). S
p
) be the group of 3-Sasakian isometries of
the manifold (orbifold)

o(p). S
p
) and let / be the number of 1s in p. Then the
connected component of Aut(o(p). S
p
) is o(l(/) l(1)
n+1k
). where we dene
l(0) = id. Thus, the connected component of the isometry group is the product
o(l(/) l(1)
n+1k
) oO(3) if the sums j
i
+j
j
are even for all 1 i. , n +1.
and o(l(/) l(1)
n+1k
) oj(1) otherwise.
In the case that p has no repeated 1s, the cohomogeneity can easily be deter-
mined, viz. [BGM96b]
Corollary 13.7.13: If the number of 1s in p is 0 or 1 then the dimension of the
principal orbit in o(p) equals n+3 and the cohomogeneity of o(p) is 3n4. In par-
ticular, the 7-dimensional family (o(p). S
p
) contains metrics o(j) of cohomogeneity
0. 1. and 2.
Combining Proposition 13.7.11 with techniques developed by Eschenburg [Esc82,
Esc92] in the study of certain 7-dimensional bi-quotients of ol(3) one can compute
the integral cohomology ring of o(p) [BGM94a]:
Theorem 13.7.14: Let n 2 p c
1,n+1
(Z). Then, as rings,
H

o(p). Z

Z[/
2
]
[/
n+1
2
= 0]
1[1
2n+1
]

{(p) .
Here the subscripts on /
2
and 1
2n+1
denote the cohomological dimension of each
generator. Furthermore, the relations {(p) are generated by
n
(p)/
n
2
= 0 and
13.7. TORIC 3-SASAKIAN MANIFOLDS 511
1
2n+1
/
n
2
= 0. where
n
(p) =

n+1
j=1
j
1
j
j
j
n+1
is the n
th
elementary symmet-
ric polynomial in the entries of p.
Notice that Theorem 13.7.14 shows that H
2n
(o(p); Z) = Z

n
(p)
and hence
has the following corollary.
Corollary 13.7.15: The quotients (o(p). o(p)) give innitely many homotopy in-
equivalent simply-connected compact inhomogeneous 3-Sasakian manifolds in di-
mension 4n 1 for every n 2. In fact, there are innite families that are not
homotopy equivalent to any homogeneous space.
Remark 13.7.3: Corollary 13.7.15 shows that the niteness results for regular 3-
Sasakian manifolds discussed in Section 13.5.3 fail for non-regular 3-Sasakian man-
ifolds. Moreover, combining our results with a well-known niteness theorem of
Anderson [And90] we have
Corollary 13.7.16: For each n 2 there are innitely many 3-Sasakian (4n1)-
manifolds with arbitrarily small injectivity radii.
13.7.5. S(p
1
, p
2
, p
3
), Eschenburg Spaces, and Dieomorphism Type.
We will now examine the n = 2 case in some detail. In particular we will show
how to determine the smooth structure. It can be seen that the spaces o(p) =
o(j
1
. j
2
. j
3
) are a special subfamily of more general bi-quotients of ol(3). We
have
Denition 13.7.17: Let k = (/
1
. /
2
. /
3
) Z
3
. l = (|
1
. |
2
. |
3
) Z
3
be such that
/
1
+/
2
+/
3
= |
1
+|
2
+|
3
. Consider the circle action o
1
k,l
dened on ol(3) by

k,l

(A) = diag(
k
1
.
k
2
.
k
3
) A diag(
k
1
.
k
2
.
k
3
) .
where A ol(3) and [[
2
= 1. Suppose (k. l) are chosen so that the associated
circle action o
1
k,l
is free. Then the quotient
`
k,l
= ol(3)o
1
k,l
.
is a smooth compact manifold and it is called Eschenburg bi-quotient.
The action is free if and only if diag(
k1
.
k2
.
k3
) is not conjugate in ol(3) to
diag(
l1
.
l2
.
l3
). This translates into the following conditions, which must all be
satised:
(13.7.6)
gcd(/
1
|
1
. /
2
|
2
) = 1. gcd(/
1
|
2
. /
2
|
1
) = 1.
gcd(/
1
|
1
. /
2
|
3
) = 1. gcd(/
1
|
2
. /
2
|
3
) = 1.
gcd(/
1
|
3
. /
2
|
1
) = 1. gcd(/
1
|
3
. /
2
|
2
) = 1.
It is easy to see compute the cohomology ring of `
k,l
. In particular, the non-
trivial cohomology groups are H
i
(`
k,l
. Z) = Z for i = 0. 2. 5. 7 and H
4
(`
k,l
. Z) =
Z
]r]
. where : = :(k. l) =
2
(k)
2
(l). These spaces were introduced by Eschen-
burg in 1982 showing that some of them admit positive sectional curvature metrics
[Esc82]. At that time these were the rst examples of inhomogeneous Riemannian
metrics with this property. More precisely, later in [Esc84], Eschenburg proved
that
Proposition 13.7.18: A bi-quotient metric on `
k,l
has positive sectional curva-
ture if and only if either |
i
[min/
1
. /
2
. /
3
. max/
1
. /
2
. /
3
] for each i = 1. 2. 3
or /
i
[min|
1
. |
2
. |
3
. max|
1
. |
2
. |
3
] for each i = 1. 2. 3.
512 13. 3-SASAKIAN MANIFOLDS
Let us note that a given Eschenburg space can be represented as a bi-quotient
in many dierent ways. For example, we can change the order of the /
i
s in k or
the |
t
i
: in l or switch all k with l. We can also simultaneously shift both k, l by
a = (o. o. o) or replace (k. l) by (k. l). All such operations do not change the bi-
quotient manifold. However, there is always a canonical choice. More importantly
after removing this redundance there are only nitely many Eschenburg bi-quotients
for each given : = :(k. l). More precisely
Proposition 13.7.19: Any positively curved Eschenburg space `
k,l
has a unique
representative of the following form (k[l) = (o. /. c + d o /[0. c. d). where o
/ c d 0. Furthermore, for each : = :(k. l) =
2
(k)
2
(l) there are nitely
many such representatives.
This large collection of spaces contains several interesting subfamilies. One of
them is the homogeneous family (k[l) = (:. n. n:[0. 0. 0) of the so-called Alo-
Wallach spaces `
7
mn
[AW75] which are of much interest since they admit Einstein
metrics of positive sectional curvature [Wan82]. The importance of these spaces
to 3-Sasakian geometry follows from the following simple observation [BGM94a]
Proposition 13.7.20: o(p) = `
k,l
with (k[l) = (j
1
. j
2
. j
3
[0. 0. j
1
+ j
2
+ j
3
). In
particular, all 7-manifolds o(p) are Eschenburg bi-quotients and they have metrics
of positive sectional curvature.
Exercise 13.5: Prove Proposition 13.7.20 by writing an explicit identication of
o(p) and the bi-quotient `(j
1
. j
2
. j
3
[0. 0. j
1
+j
2
+j
3
).
Note, however, that the positive sectional curvature bi-quotient metrics of Es-
chenburg has essentially nothing to do with the 3-Sasakian metric on o(p) which
necessarily always has negatively curved sections.
Because of their curvature properties the Eschenburg spaces `(k[l) have been
studied in great detail. Homotopy invariants for Eschenburg space have been
worked out independently by Kruggel [Kru98, Kru97] and Milgram [Mil00]. The
homeomorphism and dieomorphism classication was rst done for a certain sub-
class of Eschenburg spaces which include some of the o(p) spaces by Astey, Micha,
and Pastor [AMP97]. Later Kruggel [Kru05] obtained the dieomorphism and
homeomorphism classication of a more general class of Eschenburgs bi-quotients
by computing the Kreck-Stolz invariants [KS88]. In his construction one needs a
condition on (k. l). namely that the matrix o
ij
= /
i
|
j
contains at least one row
or column with pairwise relatively prime entries. Kruggel calls it condition (C) and
admits that he does not know if there are any Eschenburg spaces which satisfy it.
More recently Chinburg, Escher and Ziller found that there are many Eschenburg
spaces which satisfy this condition [CEZ07]. Note, however, that this condition is
automatically satised for the 3-Sasakian family o(p). Let us describe this classi-
cation and some of its important consequences. We closely follow the notation of
[CEZ07].
First, besides :(k. l) we can compute the rst Portraying class and the linking
number of `
k,l
. For the linking form we get
(13.7.7) Lk(k. l) = :: QZ.
13.7. TORIC 3-SASAKIAN MANIFOLDS 513
where : = :(k. l) =
3
(k)
3
(l). The Portraying class j
1
(`
k,l
) = j
1
(k. l) mod :
in Z
]r]
is
(13.7.8) j
1
(k. l) =
3

i=1
(/
i
|
1
)
2
+
3

i=1
(/
i
|
2
)
2
(|
1
|
2
)
2
.
Next one needs to compute two more QZ-invariants introduced by Kreck and Stolz
[KS88]. Let us assume that condition (C) holds for the ,
th
column, there are at
most three exceptional orbits for the o
1
k,l
action on the cobordism of ol(3) with
isotropy groups Z
]k1lj]
. Z
]k2lj]
and Z
]k3lj]
. After removing small equivariant
neighborhoods of these orbits the action becomes free and the quotient is a smooth
eight dimensional manifold \
k,l
with boundary (\
k,l
) = `
k,l
1
1
1
2
1
3
.
where the 1
i
are the following seven dimensional lens spaces.
(13.7.9)
1
1
= 1(/
1
|
j
; /
2
|
j
. /
3
|
j
. /
2
|
[j+1]
2
. /
3
|
[j+1]
2
) .
1
2
= 1(/
2
|
j
; /
1
|
j
. /
3
|
j
. /
1
|
[j+1]2
. /
3
|
[j+1]2
) .
1
3
= 1(/
3
|
j
; /
1
|
j
. /
2
|
j
. /
1
|
[j+1]2
. /
2
|
[j+1]2
) .
where we used the notation [n]
p
= : if n = j + : for : = 1. . . . . j. for the
residue class [n] modulo j.
Since the invariants are additive with respect to unions, we obtain :
i
(\
k,l
) =
:
i
(`
k,l
) + :
i
(1
1
) + :
i
(1
2
) + :
i
(1
3
) QZ. i = 1. 2 . Calculating :
i
(\
k,l
). yields
the following expressions for the Kreck-Stolz invariants, which hold in the case
condition (C) is satised for the ,
th
column.
(13.7.10)
:
1
(`
k,l
) =
4 [ :(k. l) (/
1
|
j
) (/
2
|
j
) (/
3
|
j
) [ c(k. l)
2
2
7
7 :(k. l) (/
1
|
j
) (/
2
|
j
) (/
3
|
j
)

i=1
:
1
(/
[i]
3
|
j
; /
[i+1]
3
|
j
. /
[i+2]
3
|
j
. /
[i+1]
3
|
[j+1]
2
. /
[i+2]
3
|
[j+1]
2
)
:
2
(1
k,l
) =
c(k. l) 2
2
4
3 :(k. l) (/
1
|
j
) (/
2
|
j
) (/
3
|
j
)

i=1
:
2
(/
[i]3
|
j
; /
[i+1]3
|
j
. /
[i+2]3
|
j
. /
[i+1]3
|
[j+1]2
. /
[i+2]3
|
[j+1]2
) .
where
c(k. l) = (/
1
|
j
)
2
+ (/
2
|
j
)
2
+ (/
3
|
j
)
2
+
+ (/
1
|
[j+1]
2
)
2
+ (/
2
|
[j+1]
2
)
2
+ (/
3
|
[j+1]
2
)
2
(|
j
|
[j+1]
2
)
2
and :
i
(j; j
1
. j
2
. j
3
. j
4
) = :
i
(1
p
(j
1
. j
2
. j
3
. j
4
)) QZ. i = 1. 2 are the Kreck-Stolz
invariants of the lens space 1
p
(j
1
. j
2
. j
3
. j
4
) = o
7
Z
p
. The freeness condition of
the action (13.7.6) implies that /
i
|
j
= 0 for i. , = 1. 2. 3. and, hence, the above
expressions for :
1
and :
2
are well-dened.
By the Atiyah-Patodi-Singer index theorem the Kreck-Stolz invariants can
also be expressed as linear combinations of eta-invariants. Calculating these eta-
invariants for the lens spaces, one obtains:
514 13. 3-SASAKIAN MANIFOLDS
:
1
(d; d
1
. d
2
. d
3
. d
4
) =
1
2
5
7 d
]d]1

k=1
4

j=1
cot

/d
j
d

+
1
2
4
d
]d]1

k=1
4

j=1
csc

/d
j
d

;
(13.7.11)
:
2
(d; d
1
. d
2
. d
3
. d
4
) =
1
2
4
d
]d]1

k=1
(c
2k
|d|
1)
4

j=1
csc

/d
j
d

.
For d = 1 these expressions are interpreted to be 0. These formulas only hold
in the case that condition (C) is satised for the ,
th
column. Similar formulas can be
derived in the case where the ,
th
row consists of relatively prime entries [CEZ07].
Using these invariants we now state the classication theorems by Kruggel for the
Eschenburg spaces [Kru05, CEZ07].
Theorem 13.7.21: Assume the Eschenburg spaces `
k,l
and `
k

,l
both satisfy
condition (C). Then
1. `
k,l
and `
k

,l
are (orientation preserving) homeomorphic if and only if
(o) [:(k. l)[ = [:(k
t
. l
t
)[ Z.
(/) Lk(/. |) Lk(k
t
. l
t
) QZ.
(c) j
1
(k. l) j
1
(k
t
. l
t
) Z
]r(k,l)]
.
(d) :
2
(`
k,l
) :
2
(`
k

,l
) QZ.
2. `
k,l
and `
k

,l
are (orientation preserving) dieomorphic if and only if
in addition
:
1
(`
k,l
) :
1
(`
k

,l
) QZ.
3. `
k,l
and `
k

,l
are (orientation preserving) homotopy equivalent if and
only if
(o) [:(k. l)[ = [:(k
t
. l
t
)[ Z.
(/) Lk(k. l) Lk(k
t
. l) QZ.
(c) :
22
(`
bfk,l
) :
22
(`
k

,l
) QZ.
where :
22
(`
k,l
) = 2 [:(k. l)[ :
2
(`
k,l
).
For the corresponding theorem in the orientation reversing case the linking
number and the Kreck-Stolz invariants change signs. Recall that in this theorem,
:(k. l) =
2
(k)
2
(l), j
1
(`
k,l
) = [2
1
(k)
2
6
2
(k)] n
2
Z
]r]
, and the equality
of the linking forms can replaced by the equality of the numbers :(k. l) =
3
(k)

3
(l) Z
]r(k,l)]
.
Remark 13.7.4: In [KS88] Kreck-Stolz used another invariant :
3
in the homeomor-
phism classication, and showed that :(k. l). :
2
and :
3
determine the homeomor-
phism type. Following [Kru05], the formula for the invariant :
3
for the Eschenburg
spaces is easily seen to be:
13.7. TORIC 3-SASAKIAN MANIFOLDS 515
:
3
(1
k,l
) =
c(k. l) 8
2
2
3 :(k. l) (/
1
|
j
) (/
2
|
j
) (/
3
|
j
)

i=1
:
3
(/
[i]
3
|
j
; /
[i+1]
3
|
j
. /
[i+2]
3
|
j
. /
[i+1]
3
|
[j+1]
2
. /
[i+2]
3
|
[j+1]
2
).
where
:
3
(j; j
1
. j
2
. j
3
. j
4
) =
1
2
4
j
]p]1

k=1
(c
4k
|p|
1)
4

j=1
csc

/j
j
j

.
Remark 13.7.5: In the case of 3-Sasakian manifold o(p) all relevant invariants
other than Kreck-Stolz invariants can be written in terms of the elementary sym-
metric functions
i
(p). We have : = :(p) =
2
(p). : = :(p) = 2
2
1
(p) 4
2
(p).
Theorem 13.7.21 can be used to obtain two important results concerning the
dierential topology of 3-Sasakian manifolds. Like in the Sasaki-Einstein case there
exists 3-Sasakian manifolds which are homeomorphic but not dieomorphic. Fur-
thermore, in spite of the local rigidity, there are non-isometric 3-Sasakian structures
on the same smooth manifold. Both results are expected, however, the examples
are relatively hard to nd, and, somewhat curiously the known ones occur in pairs.
Theorem 13.7.22: For : =
2
(p) 10
7
, there are exactly 138 pairs of 3-Sasakian
spaces o(p). o(p
t
) which are homeomorphic to each other, but not dieomorphic.
The rst few pairs, together with their relevant invariants, are listed in the
table below.
p :(p) j
1
(p) :(p) :
2
(p) :
1
(p)
(171. 164. 1) 28379 27139 335
2393
56758

82869
3178448
(223. 60. 53) 28379 27139 335
2393
56758

1104513
3178448
(362. 291. 37) 129503 12564 45679
80901
259006
69409
14504336
(423. 169. 98) 129503 12564 45679
80901
259006
5767541
14504336
(717. 362. 13) 273581 91230 196280
370663
1094324

393315
1094324
(761. 241. 90) 273581 91230 196280
370663
1094324
310179
1094324
(891. 368. 43) 382025 35741 334208
294993
1528100

74669
436600
(928. 191. 183) 382025 35741 334208
294993
1528100
1442017
3056200
(1265. 347. 2) 442179 6448 346023
115166
1326537

173889
611408
(1274. 311. 29) 442179 6448 346023
115166
1326537

21037
611408
The search in [CEZ07] was performed with help of a computer code which
originally had some errors. L. Florit further ran corrected C codes of [CEZ07]
to nd more examples with : =
2
(p) 1.5 10
7
. He found an additional 83
homeomorphic pairs and one new pair that is dieomorphic. In particular, we have
516 13. 3-SASAKIAN MANIFOLDS
Theorem 13.7.23: For : =
2
(p) 10
7
, there are two pairs of 3-Sasakian spaces
o(p). o(p
t
) which are dieomorphic to each other, but not isometric. They are
given by
(i) p = (2279. 1603. 384) and p
t
= (2528. 939. 799) with : = 5143925.
(ii) p = (4219. 2657. 217) and p
t
= (4637. 1669. 787) with : = 12701975.
13.7.6. 7-dimensional Toric 3-Sasakian Manifolds. First, let us observe
that as a consequence of Calderbank-Singer classication of all compact toric pos-
itive self-dual Einstein orbifolds in [CS06a] we have the following 3-Sasakian ana-
logue of Theorem 12.5.7.
Theorem 13.7.24: Any toric 7-dimensional compact 3-Sasakian orbifold is iso-
morphic to one of the quotients o(). where
k,k+2
(Z) is some non-singular
weight matrix.
In this case we can easily see that there are innitely many examples of admis-
sible weight matrices for which o() is smooth. The simplest family of examples
is given by matrices of the form
(13.7.12) =

1 0 . . . 0 o
1
/
1
0 1 . . . 0 o
2
/
2
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
0 0 1 o
k
/
k

.
for which we have
Proposition 13.7.25: Let / be a positive integer, and let
k,k+2
(Z) be as
in 13.7.12. Then /
k,k+2
(Z) if and only if (a. b) (Z

)
k
(Z

)
k
and the
components (o
i
. /
i
) are pairs of relatively prime integers for i = 1. . / such that
if for some pair i. , o
i
= o
j
or /
i
= /
j
then we must have /
i
= /
j
or o
i
= o
j
.
respectively.
Proposition 13.7.25 shows that /
k,k+2
(Z) is never empty and we have many
examples of compact smooth 7-dimensional quotients o() for arbitrary / 1.
Some of these examples were rst mentioned in [BGM93] and the idea of the
quotient is based on the result of [GN92]. As we shall not present here the complete
solution to the equivalence problem, we shall further assume that /
k,k+2
(Z) is
arbitrary and shall determine some important topological properties of the quotients
o(). More explicitly [BGMR98, BGM98a],
Theorem 13.7.26: Let /
k,k+2
(Z). Then
1
(o()) = 0 and
2
(o()) = Z
k
.
Proof. First note that the groups T
k+2
oj(1) and T
2
oj(1) act as isometry
groups on () and o(). respectively. Let us dene the following quotient spaces:
Q() = ()T
k+2
oj(1). 1() = ()oj(1) .
We have the following commutative diagram
(13.7.13)
() 1()

o() Q() .
13.7. TORIC 3-SASAKIAN MANIFOLDS 517
The top horizontal arrow and the left vertical arrow are principal bundles with
bers oj(1) and T
k
. respectively. The remaining arrows are not brations. The
right vertical arrow has generic bers T
k+2
. while the lower horizontal arrow has
generic bers T
2
oj(1) homeomorphic either to T
2
RP
3
or T
2
o
3
depending on
. The dimension of the orbit space Q() is 2.
The diculty is in proving that both () and 1() are 2-connected. Once
this is accomplished the result follows by applying the long exact homotopy se-
quence to the left vertical arrow in diagram 13.7.13.
Lemma 13.7.27: Both () and 1() are 2-connected.
To prove this lemma we construct a stratication giving a Leray spectral se-
quence whose dierentials can be analyzed. Let us dene the following subsets of
() : (Recall that, in this case, at most one quaternionic coordinate can vanish.)

0
() = u ()[ n

= 0 for some = 1. . . . . / + 2.

1
() = u ()[ for all = 1. . . . . / + 2. n

= 0 and there is a pair (n

. n

)
that lies on the same complex line in H.

2
() = u ()[ for all = 1. . . . . / + 2. n

= 0 and no pair (n

. n

)
lies on the same complex line in H .
Clearly, () =
0
().
1
().
2
() and
2
() is a dense open submanifold
of (). This stratication is compatible with the diagram 13.7.13 and induces
corresponding stratications
(13.7.14) 1() = 1
0
() . 1
1
() . 1
2
(). Q() = Q
0
() . Q
1
() . Q
2
().
The 1
i
() ber over the Q
i
() whose bers are tori T
k+i
. The strata are labelled
by the dimension of the cells in the resulting CW decomposition of Q(). Using
known results about cohomogeneity 2 actions [Bre93] one can easily prove:
Lemma 13.7.28: The following are true
(i) The orbit space Q() is homeomorphic to the closed disc

1
2
. and the
subset of singular orbits Q
1
().Q
0
() is homeomorphic to the boundary

1
2
o
1
.
(ii) Q
2
() is homeomorphic to the open disc 1
2
.
(iii) Q
1
() is homeomorphic to the disjoint union of / +2 copies of the open
unit interval.
(iv) Q
0
() is a set of / + 2 points.
After making certain choices Lemma 13.7.28 can be used to represent Q()
topologically as a polygon as illustrated in Figure 1.
Next one can easily show that
1
(1()) is Abelian; hence,
1
(1()) = H
1
(1()).
Now we claim that H
1
(1()) = H
2
(1()) = 0. and since the bers of the top hor-
izontal arrow of 13.7.13 are o
3
s this together with Hurewicz will prove Lemma
13.7.27. To prove this claim we dene Y
0
= Q
0
(). Y
1
= Q
0
() Q
1
(). and
Y
2
= Q(). Then, we lter 1() by A
i
=
1
(Y
i
) to obtain the increasing ltra-
tion
A
0
= 1
0
(). A
1
= 1
0
() 1
1
(). and A
2
= 1() .
The Leray spectral sequence associated to this ltration has 1
1
term given by
1
1
s,t

= H
s+t
(A
t
. A
t1
; Z)
518 13. 3-SASAKIAN MANIFOLDS
........................................................................................................................................................................................................................................................................................ ..................................................................................................................................................................................... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. ......................................................................

v
k
v
1
. .
v
k+1
v
2
v
k+2
.
.
.
e
k+1

2
e
k+2
e
1
e
k
Figure 1. The quotient space Q().
with dierential d
1
: H
s+t
(A
t
. A
t1
; Z) H
s+t1
(A
t1
. A
t2
; Z). where we use
the convention that A
1
= .
To compute these 1
1
terms notice that all the pairs (A
t
. A
t1
) are relative
manifolds so that one can apply the Alexander-Poincare duality theorem. Hence,
by 13.7.14
H
s
(A
0
; Z)

= H
s
(.
k+2
T
k
; Z) . (13.7.15)
H
s
(A
1
. A
0
; Z)

= H
k+2s
(.
k+2
T
k+1
; Z) . (13.7.16)
H
s
(A
2
. A
1
; Z)

= H
k+4s
(T
k+2
; Z) . (13.7.17)
where .
j
T
l
means the disjoint union of , copies of T
l
. The 1
1
s,t
term of the spec-
tral sequence is described by the diagram of Figure 2. The computation of the
dierentials is fairly tedious and we refer the reader to [BGM98a] for details.
............................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................

Z
k+2
Z
k+2
Z
k+2
Z
(k+1)(k+2)
Z
Z
k(k+2)
Z
(
k+1
2
)(k+2)
Z
(
k+2
2
)
Z
(
k
2
)(k+2)
t
s
Figure 2. The diagram for the 1
1
s,t
term.

13.7. TORIC 3-SASAKIAN MANIFOLDS 519

Theorem 13.7.26 together with the well-known Hurewicz Theorem determines
the integral cohomology groups of o() except for the torsion in H
4
(o(). Z).
Corollary 13.7.29: /
k,k+2
(Z). Then H
i
(o(). Z) are given as follows
H
0
H
1
H
2
H
3
H
4
H
5
H
6
H
7
Z 0 Z
k
0 G() Z
k
0 Z
where G() is a nite torsion group.
Recently Hepworth [Hep05a, Hep05b] determined the group G(). Rather
than using the Leray spectral sequence as in [BGMR98], he considered the Serre
spectral sequence of the associated Borel construction of the toral action. Let us
set some notation. Changing slightly the notation of Equation 13.7.1 we let
/
k,k+2
(Z). i. , 1. . . . . / +2 and
ij
be the determinant of the / / minor ob-
tained from by deleting i
th
and ,
th
columns. Set c
i
j
= (1)
i+j
sign(i,)sign(
ij
).
By construction ()o() is a T
k
-bundle, so we let r
p
H
2
(o(). Z) Z
k
be the rst Chern class associated to the i
th
factor of the T
k
-action. Then the r
p
freely generate H
2
(o(). Z). By abuse of notation we also let r
p
denote the gener-
ators of the cohomology ring of the classifying space 1T
k
so that H

(1T
k
. Z)
Z[r
1
. . . . . r
k
]. We also put
i
=

p
r
p

pi
. Then Hepworth proves
Theorem 13.7.30: G() is generated by the degree four symmetric products r
p
r
q

subject only to the relations

c
i
j

j
+c
j
k

k
+c
k
i

i
= 0
for i. ,. / distinct. Furthermore, G() is a nite group of order

[
i
1
j
1
[ [
i
k+1
j
k+1
[ .
where the summand indexed by i
1
. ,
1
. . . . . i
k+1
. ,
k+1
is included in the sum if and
only if the graph on the vertices 1. . . . . / + 2 with edges i
r
. ,
r
is a tree.
Since for an admissible the determinants [
ij
[ are nonzero, the sum above is
at least as large as the number of trees on the vertex set 1. . . . . / +2. A theorem
of Cayley states that the number of trees on a vertex set is precisely [[
]A]2
.
We therefore have:
Corollary 13.7.31: [H
4
(o(). Z)[ (/
2
(o()) + 2)
b
2
(S())
.
Let /
k,k+2
(Z) be as in (13.7.12). Theorem 13.7.30 implies that H
4
(o(). Z)
will have size at least as large as each of the [o
i
[, [/
j
[. Therefore for each /, one
can construct an innite sequence of for which the groups H
4
(o(). Z) are of
unbounded size. Hence, just as in the / = 1 case we get
Corollary 13.7.32: There are innitely many homotopy types of o() for each
xed second Betti number.
In addition, Hepworth computes the stable tangent bundle and, hence, the
Portraying classes of the o().
Theorem 13.7.33:
To() c
4k+1
R

= 2
k+2

p=1
|
p
.
520 13. 3-SASAKIAN MANIFOLDS
where, for 1 j (/ +2), |
p
is the complex line bundle over o() with rst Chern
class
p
. Consequently
j
1
(o()) = 2
k+2

p=1

2
p
.
In addition, in his thesis [Hep05a] Hepworth shows that the manifolds o() are
cobordant to much simpler spaces whose second Betti number equals 1. This suggest
that, just as in the case of o(p). one could obtain a dieomorphism classication
of o() by computing their generalized Kreck-Stolz invariants.
Proposition 13.7.34: Let /
k,k+2
(Z) so that o() is a smooth manifold. Let
Z() and O() be the associated twistor space and quaternionic Kahler orbifold,
respectively. Then we have /
2
(o()) = /
2
(O()) = /
2
(Z()) 1 = /.
This shows that inequality /
2
1 in Proposition 13.5.7 does not hold for non-
regular 3-Sasakian manifolds. Finally, we give several interesting corollaries.
Corollary 13.7.35: There exists a simply connected 3-Sasakian 7-manifold for
every rational homology type allowed by Corollary 13.5.3.
Our next corollary follows from the results of this section and a remarkable
theorem of Gromov [Gro81]:
Corollary 13.7.36: For any non-positive real number there are innitely many
3-Sasakian 7-manifolds which do not admit metrics whose sectional curvatures are
all greater than or equal to .
Corollary 13.7.37: There exist Q-factorial contact Fano 3-folds A with /
2
(A) = |
for any positive integer |.
The last corollary should be contrasted to the smooth case, where Mori and
Mukai have proven that /
2
10 [MM82]:
Corollary 13.7.38: If the second Betti number /
2
(o()) = / 3. the 3-Sasakian
manifolds o() are not homotopy equivalent to any homogeneous space.
13.7.7. Higher Dimensional Toric 3-Sasakian Manifolds. As in dimen-
sion 7, we have the following classication theorem of Bielawski [Bie99]:
Theorem 13.7.39: Let (`. S) be a toric 3-Sasakian manifold of dimension 4n1.
Then (`. S) is isomorphic as a 3-Sasakian T
n
-manifold to a 3-Sasakian quotient
of a sphere by a torus, that is to a o() for some .
This result is a weaker than the similar result of Theorem 13.7.24 as following
Bielawski we assume that (`. S) is a smooth manifold. On the other hand it is
likely that this assumption can be removed. Note that this theorem includes the
degenerate case when the quotient is a sphere or a discrete quotient of such. The
Betti numbers of a 3-Sasakian orbifold obtained by a toral quotient of a sphere were
computed by Bielawski [Bie97] using dierent techniques than the ones employed
in the previous section. The following result follows easily from Theorem 12.9.6:
Theorem 13.7.40: Let
k,n+1
(Z) be non-degenerate so that o() is a com-
pact 3-Sasakian orbifold of dimension 4(n /) + 3. Then we have
(13.7.18) /
2i
(o()) =

/ +i 1
/

. i < n + 1 /.
13.7. TORIC 3-SASAKIAN MANIFOLDS 521
Furthermore, the Betti number constraints of Proposition 13.5.7(ii) can hold
for o() if and only if / = 1.
Combining Theorems 13.7.39 and 13.7.40 with Lemmas 13.7.7 and 13.7.8 which
give obstructions to smoothness gives the somewhat surprising result [BGM98b],
Theorem 13.7.41: Let (`. S) be a toric 3-Sasakian manifold.
(i) If the dimension of ` is 19 or greater, then /
2
(`) 1.
(ii) If the dimension of ` is 11 or 15, then /
2
(`) 4.
(iii) If /
2
(`) 4. then the dimension of ` is 7.
A corollary due to Bielawski [Bie99] is:
Corollary 13.7.42: Let (`. S) be a regular toric 3-Sasakian manifold. Then `
is one of the 3-Sasakian homogeneous spaces o
4n1
. RP
4n1
or
SU(n)
S

U(n2)U(1)

.
Next we give an explicit construction of toric 3-Sasakian manifolds not elimi-
nated by Theorem 13.7.41. It is enough to show that /
4,8
and /
4,7
are not empty
as the rest follow by deletion of rows of the corresponding /
4,
. We shall
present two three parameter families of solutions, namely

1
=

1 1 1
2 1 1 + 2
l
1 16 1 + 2
m
1 3 2c

.
2
=

1 1 1 2
2 1 1 + 2
l

1
1 16 1 + 2
2n
3
1 3 2c
t
1

.
where |. |
t
. :. n Z
+
. and c. c
t
Z.
With the aid of MAPLE symbolic manipulation program [BGM98b], we nd
Lemma 13.7.43: Let = 2(31c + 6 + 19 2
l1
7 2
m1
).
(i)
1
=

1l
4

1

is admissible if and only if c = 0 and is not divisible by 3.

and = 0 and is not divisible by 7. 19 nor 31.
(ii)
2
=

1l
4

2

is admissible if and only if c and all minor determi-

nants of
2
are non-vanishing, and c
t
0(3). |
t
0(4). c
t
5(7).
and 11. 19. 37. 71 do not divide det = 19 2
2n
63 148c
t
11 2
l

. and
the following conditions hold:
(13.7.19)
gcd(3. 4c
t
+ 2
l

+ 1. 2c
t
2
l

1) = 1
gcd(7. 2
2n+1
2
l

+ 1. 3 2
l

+ 2
2n
+ 4) = 1
gcd(19. 2
2n
2
l

+4
15. 3 2
l

+ 2
2n
+ 4) = 1
gcd(25. 32c
t
3 2
2n
3. 6c
t
+ 2
2n
+ 1) = 1 .
The conditions in this proposition guarantee that the quotient spaces denoted
by o(c. |. :) and o(c
t
. |
t
. n) are smooth manifolds of dimension 11 and 15. respec-
tively.
It is routine to verify that the three parameter innite family given by
c 14(21). | 1(5). : (c) mod(18) .
where 2
(c)
22(31c + 6)(18) satises the conditions in (i) of Lemma 13.7.43.
This gives examples in dimension 11. (Notice that as 2 is a primitive root of 19
the equation dening (c) has a unique solution (mod 18) for each value of c.)
522 13. 3-SASAKIAN MANIFOLDS
Similarly, it is straightforward to verify that the innite family given by
c
t
= 2. |
t
= 1. n 21(90) .
satises the conditions (ii) of Lemma 13.7.43. We have arrived at [BGM98b]:
Theorem 13.7.44: There exist toric 3-Sasakian manifolds o of dimensions 11
and 15 with /
2
(o) = 2. 3. 4. Consequently, the Betti number relations of (ii) of
Proposition 13.5.7 do not hold generally. More explicitly there are compact 11-
dimensional 3-Sasakian manifolds for which /
2
= /
4
, and compact 15-dimensional
3-Sasakian manifolds for which /
2
= /
6
.
13.8. Cohomogeneity One 3-Sasakian 7-Manifolds
In this section we discuss all known complete cohomogeneity one 3-Sasakian
metrics in dimension 7. Locally, the classication of such metrics follows from
Hitchins classication of self-dual Einstein metrics with an isometric action of
ol(2) [Hit96, Hit95a, Hit95b]. In particular, as we shall see, there are at least
two families of such manifolds: the l(2)-invariant family o(c. 1. 1) of the toric 3-
Sasakian spaces and another non-toric oO(3)-invariant family of metrics obtained
from the Hitchins metrics on the orbifolds O
k
described in Section 12.5. Globally,
however, the question is more subtle. Nevertheless, it is quite reasonable to
Conjecture 13.8.1: Let (`. S) be a compact simply connected 3-Sasakian orbifold
of dimension 7 and of cohomogeneity one under the action of Aut(`. S). Then
(`. S) is one of the two innite families of spaces
(i) the toric 3-Sasakian manifold o(c. 1. 1) as the Konishi orbibundle over
the Galicki-Lawson orbifold O(c. 1. 1) of Proposition 12.5.3
(ii) the Konishi bundle o
k
over non-toric Hitchin orbifolds O
k
of Theorem
12.5.8.
The toric 3-Sasakian 7-manifolds o(c. 1. 1) were already discussed in the pre-
vious section. The Hitchin metrics likewise have smooth cohomogeneity one 3-
Sasakian manifolds associated to them. They were recently considered in [GWZ05]
as possible candidates of 7-manifolds admitting cohomogeneity one metrics of pos-
itive sectional curvature. We describe the geometry and topology of these new
examples. We closely follow [GWZ05] in our exposition. Let us recall that Equa-
tion 1.6.9 and its associated Group Diagram 1.6.10 determine the manifold `.
We rst consider several innite families of 7-manifolds with cohomogeneity one
actions. Let G = oj(1) oj(1). where we think of oj(1) as the group of unit
quaternions. Consider the following circle subgroups of G:
(13.8.1) C
i
1
(p,q)
= (c
pi
1

. c
qi
1

) [ R. C
i
2
(p,q)
= (c
pi
2

. c
qi
2

) [ R.
where (j. c) are integers, called slopes. Let Q denotes the quaternion group gener-
ated by i
1
. i
2
. i
3
and H = Z
2
Z
4
= (1. 1). (i
1
. i
1
) G.
Denition 13.8.2: Let ` = `
7
(p

,q

),(p
+
,q
+
)
be dened by the group diagram
Q 1

. 1
+
G. where 1

. 1
+
= C
i1
(p,q)
Q. C
i2
(p+,q+)
Q, where
(j

. c

) as well as (j
+
. c
+
) are relatively prime odd integers. Furthermore, let
1
l
= `
(1,1),(2l1,2l+1)
.
Here 1

1
0

= Z
2
. where the second component is generated by (i
2
. i
2
) on the
left and (i
1
. i
1
) on the right, up to signs (of both coordinates). The embedding of
13.8. COHOMOGENEITY ONE 3-SASAKIAN 7-MANIFOLDS 523
Q G is determined by the slopes and is always diagonal Q. up to sign changes
in both coordinates.
The action of oj(1) oj(1) is only almost eective, i.e., G and Q have a nite
normal, hence, central subgroup in common. For example, the manifolds 1
l
have
the eective group diagram
Z
2
Z
2
O(2). O(2) oO(4).
It is easy to see that 1
1
= o
7
with the well-known cohomogeneity one action of
oO(4). Grove, Wilking, and Ziller [GWZ05] prove
Theorem 13.8.3: The manifolds ` = `
(p,q),(p+,q+)
are 2-connected. If
p
q
=

p
+
q+
their cohomology ring is determined by
3
(`) = Z
k
with / = (j
2

c
2
+
j
2
+
c
2

)8.
Otherwise H
3
(`. Z) = H
4
(`. Z) = Z. In particular, the manifolds 1
l
are all
rational homology spheres with
1
(1
l
) =
2
(1
l
) = 0 and
3
(1
l
) = Z
l
.
In order to describe all Konishi bundles over Hitchin orbifolds we need yet
another innite family of cohomogeneity one 7-manifolds.
Denition 13.8.4: Let =
6
(p

,q

),(p
+
,q
+
)
be the cohomogeneity one 7-manifold
dened by the group diagram H 1

. 1
+
= C
i
1
(p

,q

)
H. C
j
(p
+
,q
+
)
H G
with (j

. c

) as well as (j
+
. c
+
) relatively prime, j
+
even and j

. c

. c
+
odd.
Furthermore, let Q
l
= `
(1,1),(l,l+1)
.
Here the component groups 1

1
0

are determined by the fact that (i

1
. i
1
)
1
0

and (1. 1) 1
+
0
. As in the previous example the oj(1) oj(1)-action is only
almost eective. For example, the manifolds Q
k
have the eective group diagram
Z
2
oO(2). O(2) oO(3) oO(3) .
where the groups 1

are embedded in dierent blocks in each component of oO(3)

oO(3). It is easy to see that 1
1
= o(1. 1. 1) with the well-known cohomogeneity
one action of oO(3) oO(3). The integral cohomology groups of these manifolds
was given in [GWZ05].
Theorem 13.8.5: The manifolds =
(p

,q

),(p
+
,q
+
)
are simply connected with
H
2
(. Z) = Z. H
3
(. Z) = 0 and H
4
(. Z) = Z
k
with / = j
2

c
2
+
j
2
+
c
2

.
In particular, for Q
l
we have H
2
(Q
l
. Z) = Z and H
4
(Q
l
. Z) = Z
2l+1
. We will
be only interested in the manifolds 1
l
and Q
l
. Note that 1
1
and Q
1
are precisely
the two simply connected homogeneous (hence, regular) compact 3-Sasakian 7-
manifolds. They both admit cohomogeneity one subactions and, from this point
of view, naturally appear as rst in two series of cohomogeneity one spaces, under
actions of oO(4) and oO(3) oO(3), respectively. It turns out that although these
two series are very dierent, they have quotients which can be described in a uniform
fashion. These quotients are precisely the Konishi bundles over Hitchin orbifolds
O
k
described in Theorem 12.5.8. We can already see the pattern by considering the
rst two examples. 1
1
o
7
is an oj(1) bundle over o
4
. The Konishi bundle, on
the other hand, is the oO(3)-bundle over o
4
which is RP
7
1
1
Z
2
. The manifold
Q
1
is and oO(3)-bundles over CP
2
and it is the Konishi bundle. But Q
1
admits a
smooth 7
2
quotient which is a Konishi orbibundle over CP
2
Z
2
= O
4
. Recall from
section 12.5 that indeed, the Hitchin orbifold O
4
is not simply connected and its
universal cover is CP
2
with the standard Fubini-Study metric. This picture repeats
itself in higher dimensions: the Hitchin orbifolds O
2l+1
are simply connected and
524 13. 3-SASAKIAN MANIFOLDS
the Konishi bundle o
2l+1
O
2l+1
admits a lift to an oj(1)-bundle which turns out
to be 1
l
. The Hitchin orbifolds O
2l+2
are not simply-connected, and neither are
the Konishi bundles o
2l+2
O
2l+2
. The universal cover of o
2l+2
is again an oO(3)
orbibundle over the orbifold universal cover of O
2l+2
.
Before we consider the Konishi orbibundles let us dene the cohomogeneity one
7-manifolds H
k
dened by the group diagram
(13.8.2) Z
2
Z
2
1
0

H. 1
0
+
H oO(3) oO(3) .
where the identity components 1
0

oO(2) depend on integers (j. c) which de-

scribe the slope of their embedding into a maximal torus of oO(3) oO(3). They
are (1. 1) for 1
0

embedded into the lower 22 block and (/. /+2) for 1

0
+
embedded
into the upper 2 2 block. That also determines the embedding of H = Z
2
Z
2
.
In principle, it is clear that the Hitchin metrics give rise to 3-Sasakian orbifold
metrics on a seven dimensional orbifold o
k
which is the total space of the oO(3)-
orbibundle o
k
O
k
. The cohomogeneity one action by oO(3) on the base admits
a lift to the total space o
k
which commutes with the almost free principal orbifold
oO(3) action. The joint action by oO(3)oO(3) on o
k
is, hence, an isometric co-
homogeneity one action. In general, one would expect the metric on o
k
to have
orbifold singularities since the base does. However, a simple argument shows that
this is not the case [GWZ05]
Theorem 13.8.6: For each /, the total space o
k
of the Konishi bundle correspond-
ing to the self-dual Hitchin orbifold O
k
is a smooth 3-Sasakian manifold.
Proof. Note that the singular orbit 1
+
in O
k
. / 4 must be totally geodesic.
Indeed, being an orbifold singularity, one can locally lift the metric on a normal
slice 1
2
to RP
2
to its (/ 2)-fold branched cover

1 1 with an isometric action
by Z
k2
such that

1Z
k2
= 1. Hence the singular orbit is a xed point set of a
locally dened group action and thus totally geodesic. The oO(3) principle bundle
o
k
is smooth over all smooth orbits in o
k
. If it had orbifold singularities, they would
consist of an oO(3)oO(3) orbit which projects to 1
+
. and be again totally geodesic
by the same argument. This ve dimensional orbit would now be 3-Sasakian with
respect to the natural oO(3) action on o
k
, since it is totally geodesic and contains
all oO(3) orbits. But the quotient is 2-dimensional which contradicts the fact that
the base of such a manifold has dimension divisible by 4.
The following theorem was proven by Grove, Wilking and Ziller and it estab-
lishes the relation between the Konishi orbibundles o
k
over the Hitchin orbifolds
O
k
and the cohomogeneity one manifolds 1
l
and Q
l
. We just quote the result and
refer the reader to the proof in [GWZ05].
Theorem 13.8.7: The manifolds 1
l
and Q
l
are equivariantly dieomorphic to the
universal covers of the 3-Sasakian manifolds o
2l+1
and o
2l+2
. respectively.
In Theorem 11.7.11 we saw that there are innitely many rational homology
spheres in every odd dimension greater than three which admit Sasaki-Einstein struc-
tures. As an immediate consequence of the Theorem 13.8.7 we get
Corollary 13.8.8: There are innitely many rational homology 7-spheres which
admit a 3-Sasakian structure.
Open Problem 13.8.1: Note that H
2
(o(j. 1. 1). Z) = H
2
(Q
p
. Z) = Z. and the
torsion H
4
(o(j. 1. 1). Z) = H
4
(Q
p
. Z) = Z
2p+1
. Metrically the two manifolds are
13.9. NON-TORIC 3-SASAKIAN MANIFOLDS IN DIMENSION 11 AND 15 525
dierent, one being toric and the other not. It would be interesting to calculate the
Kreck-Stolz invariants for the Q
p
family to see if these spaces can be dieomorphic
to some of the o(j. 1. 1). However, this appears to be a rather dicult problem.
Question 13.8.1: The 7-manifolds 1
k
. Q
k
with their induced 3-Sasakian structure
and the Hitchin orbifolds O
k
appear to be the only known inhomogeneous examples
which are not obtained via the symmetry reduction method. Is it indeed the case
that these metrics cannot be obtained as quotients?
13.9. Non-Toric 3-Sasakian Manifolds in Dimension 11 and 15
We now consider some examples of non-Abelian reductions. Example 13.6.8
suggests that one should consider reductions by the group T
k
oj(1) which is
equivalent to Abelian reductions of the diamond diagram associated to the classical
group oO(n). At the level of the quaternionic Kahler geometry such quotients are
just Abelian reductions of the real Grassmannian Gr
+
4
(R
n
) of oriented 4-planes
in R
n+1
. We should point out that Abelian reductions of (ol(n)) can produce
only subfamilies of the toric examples discussed earlier. This is because (ol(n))
is itself an o
1
reduction of (oj(n)). Hence, other than the exceptional cases,
(oO(n)) is clearly of special interest from this point of view. There are two cases
to consider: oO(2:). : 3 and oO(2: + 1). : 2. In both cases we need to
choose a proper /-dimensional subtorus T
k
of the maximal torus T
m
. We can treat
both cases simultaneously by introducing a parameter = 0. 1. and choosing
T
k
T
m
oO(2:+).
Denition 13.9.1: Let

k,m

k,m
(Z) 0. 1 be a weight matrix with / =
1. . . . . : 1 and : 3 . We denote by o(

k,m
) the 3-Sasakian reduction
of oO(2: + )oO(2: + 3) oj(1) by an action of T
k
oO(2: + ) de-
termined by

k,m
. Equivalently, o(

k,m
) is the 3-Sasakian reduction of the unit
sphere o
4(2m+)1
by the action of T
k
oj(1). Furthermore, we say that
(i)

k,m
is non-singular if the corresponding T
k
-action on the level set of
the 3-Sasakian moment map is locally free,
(ii)

k,m
is admissible if, in addition to

k,m
being non-singular, the cor-
responding T
k
-action on the level set of the 3-Sasakian moment map is
free.
Generically, the quotient o(

k,m
) will be of dimension 4(2:/ 4 + ) + 3.
In the case when the weight matrix is non-singular the reduced space will automat-
ically be a compact 3-Sasakian orbifold, while admissible matrices yield smooth
3-Sasakian manifolds. It is interesting to observe, that unlike the case of toric
quotients, in each xed dimension we only have nitely many choices for the
triple (; /. :). In that triple the pair (; :) determines the diamond diagram
(oO(2:+)) for the reduction while / is the dimension of the torus that one is
quotienting by.
Proposition 13.9.2: Let dim(o(

k,m
)) = n = 4j + 3 with / = 1. . . . . : 1 and
: 3 . Then for each n = 4j + 3 7 the number of possible triples (; /. :)
is j + 2. whereas, there are two for n = 7. We list the triples for the rst three
dimensions.
(i) if n = 7 then (; /. :) = (0; 3. 4). (1; 2. 3),
(ii) if n = 11 then (; /. :) = (0; 2. 4). (0; 4. 5). (1; 1. 3). (1; 3. 4),
526 13. 3-SASAKIAN MANIFOLDS
(iii) if n = 15 then (; /. :) = (0; 1. 4). (0; 3. 5). (0; 5. 6). (1; 2. 4). (1; 4. 5).
Proof. Fixed n means xed j. and the dimension formula for the quotient
gives j = 2: / 4 + . Writing | = : / we have : + | = j + 4 . The
number of possible values of | equals the number of possible values of /. Assuming
that j 1 we see that there are [
p+3
2
] possible triples if = 0, and [
p+2
2
] if = 1.
This gives j +2 total. When j = 1 the constraint : 3 eliminates (0. 1. 3). so
there are only two. It is easy to delineate the rst several cases.
In principle, it is not clear that there should exist any admissible matrices,
though it is rather obvious that there should be many non-singular ones. The
following result was established in [BGP02, Bis07].
Proposition 13.9.3: There are no admissible matrices among
0
3,4
and
1
2,3
.
On the other hand, as we saw in Theorem 12.5.12, there are non-singular ma-
trices
0
3,4
and
1
2,3
which yield two families of examples of positive self-dual and
Einstein orbifold metrics with a one-dimensional symmetry group. The combina-
torics of admissibility in these cases appears to be even more restrictive than in the
toric case. We do not present the solution to this problem here, but only show that
there are some cases when

k,m
is admissible. In the remainder of this section we
shall consider the two special cases of Proposition 13.9.2 with / = 1:
1
1,3
and
0
1,4
that were treated in [BGP02].
13.9.1.
1
1,3
: 11-Dimensional Examples. We rst consider a generaliza-
tion of the Kobak-Swann Example 13.6.8. In particular, we consider
1
1,3
=
(j
1
. j
2
. j
3
) p Z
3
. We determine the conditions under which
1
1,3
is admis-
sible. Dene the homomorphism 1
p
: T
1
oO(7)
1
p
(t) =

1 0 0 0
0 (j
1
t) 0 0
0 0 (j
2
t) 0
0 0 0 (j
3
t)

. (j
i
t) =

cos(j
i
t) sin(j
i
t)
sin(j
i
t) cos(j
i
t)

.
Note that for j
1
= j
2
= j
3
= 1 we recover the Example 13.6.8 so this case is a
very natural generalization. Furthermore, we use the same notation as in Example
13.6.8 regarding the oj(1) reduction of o
27
. The homomorphism 1
p
(t) yields a
circle action on the homogeneous 3-Sasakian manifold oO(7)oO(3) oj(1) via
left multiplication 1
p
(t)u with the 3-Sasakian moment map
(13.9.1)
p
(n
1
. . . . . n
7
) =

=1,2,3
j

(n
2
n
2+1
n
2+1
n
2
) .
Observe that without loss of generality we can assume all weights to be non-negative
as j
i
can be changed to j
i
by renaming the quaternions in the associated pair
(n
2i
. n
2i+1
). We begin the analysis of this quotient by considering the zero level set
of the moment map

(p) \
7,4
(R)
1
p
(0) o
27
.
We want to consider a stratication of the level set

(p) that will allow us to

analyze the quotient space
(13.9.2) o(
1
1,3
)
11
(p) =

(p) \
7,4
(R)
1
p
(0)
oj(1) l(1)
p
.
13.9. NON-TORIC 3-SASAKIAN MANIFOLDS IN DIMENSION 11 AND 15 527
Since

(p) is a submanifold of the Stiefel manifold \

4
(R
7
) at most 3 quaternionic
coordinates can vanish on

(p). So setting various quaternionic coordinates equal

to zero determines a stratication of

(p) in which the strata of minimal dimension

play an important role. We call these strata vertices although, as we shall see, they
each have two connected components. We now give a series of lemmas to determine
conditions on the integers (j
1
. j
2
. j
3
) that guarantee that
1
1,3
be admissible. The
proofs are straightforward and can be found in [BGP02].
Lemma 13.9.4: Let 0 < j
1
< j
2
< j
3
. At a vertex neither n
1
nor any of the three
pairs of quaternions (n
2i
. n
2i+1
), i = 1. 2. 3 can vanish. Thus, there are precisely
eight vertices and they are all dieomorphic to O(4).
Our analysis suggests the importance of the following strata:
o
0
= u

(p) [ n
1
= 0 .
o
1
= u

(p) [ n
1
= 0 .
o
2
= u

(p) [ some quaternionic pair (n

2i
. n
2i+1
) vanishes .
o
3
= u

(p) [ no quaternionic pair (n

2i
. n
2i+1
) vanishes .
Then Lemma 13.9.4 easily implies that
Corollary 13.9.5: Let 0 < j
1
< j
2
< j
3
. Then
(i) o
0
o
1
= o
2
o
3
=

(p).
(ii) o
0
o
1
= o
2
o
3
= .
(iii) o
0
o
2
= .
(iv) o
2
o
1
.
(v) o
0
o
3
.
Notice that (iii) fails if j
i
= j
j
for some i = ,. In particular it fails for the
level set

of Denition 13.6.12 in the previous section, and this is the reason that
the quotient
11
of 13.6.10 is not smooth. We now are ready to give necessary
conditions to guarantee a smooth quotient.
Lemma 13.9.6: Let p = (j
1
. j
2
. j
3
) (Z
+
)
3
. Then
(i) if the (j
1
. j
2
. j
3
) are pairwise relatively prime, then the isotropy group of
the oj(1) l(1)
p
action at every point of o
1
is the identity;
(ii) if gcd(j
1
j
2
. j
1
j
3
) = 1. then the isotropy group of the oj(1) l(1)
p
action at every point of o
0
is the identity.
It now follows immediately from Corollary 13.9.5, Lemma 13.9.6, and Denition
13.9.1 of admissibility that
Theorem 13.9.7: The weight matrix
1
1,3
p Z
3
+
is admissible if and only if
0 < j
1
< j
2
< j
3
, gcd(j
i
. j
j
) = 1 for all i < ,, and gcd(j
1
j
2
. j
1
j
3
) = 1.
Note that there are innitely many admissible weight matrices
1
1,3
p Z
3
+
.
For example we can take p = (2/ 1. 2/. 2/ + 1). where / Z
+
. Thus, there are
innite families of smooth quotients o(
1
1,3
)
11
(p) and innite families of the
associated triples
11
(p) Z(p) O(p) with their (orbifold) Einstein metrics.
The geometry of the smooth families
11
(p) is rather interesting. First we
observe that these spaces cannot be toric. This can be seen in several dierent
ways, for example by careful analysis of the associated foliations. One can also
generalize the analysis of [BGM96a] to show that the only isometries of the level
528 13. 3-SASAKIAN MANIFOLDS
set of the moment map

(p) o
27
R
28
can come from the restriction of the
isometries of the Euclidean space R
28
. From this we conclude
Theorem 13.9.8: Let
1
1,3
p be admissible so that
11
(p) is a smooth compact
3-Sasakian 11-manifold. Then the Lie algebra Aut(
11
(p). S
p
) of the group of 3-
Sasakian isometries of (p) is isomorphic to R
2
sp(1). In particular, all such
quotients are non-toric.
Up to this point the topology of the manifolds
11
(p) remains largely unde-
termined.
Open Problem 13.9.1: Determine the cohomology ring of the 3-Sasakian mani-
folds
11
(p).
13.9.2.
0
1,4
: 15-Dimensional Examples. Consider now the reduction de-
ned by
0
1,4
p = (j
1
. j
2
. j
3
. j
4
) and dene the following homomorphism
1
p
(t) =

(j
1
t) 0 0 0
0 (j
2
t) 0 0
0 0 (j
3
t) 0
0 0 0 (j
4
t)

. (j
i
t) =

cos(j
i
t) sin(j
i
t)
sin(j
i
t) cos(j
i
t)

.
As before we can choose all weights to be non-negative. Further note that at most
one of the weights can vanish. Using techniques similar to the ones described in
the previous section one can show that
Theorem 13.9.9: Let
0
1,4
(j
1
. j
2
. j
3
. j
4
) Z
4
. Then
0
1,4
is admissible if and
only if 0 j
1
< j
2
< j
3
< j
4
, gcd(j
i
. j
j
. j
k
) = 1 and gcd(j
i
j
j
. j
i
j
k
) = 1 for
any triple in p.
Proof. First, let us assume that all of the weights are non-negative. Consider
a triple, say (j
1
. j
2
. j
3
). Set n
7
= n
8
= 0. Then the analysis of the previous section
shows that the three must be distinct and that we must have gcd(j
1
j
2
. j
1
j
2
) =
1. However, we no longer need the three weights to be pairwise relatively prime
as one cannot set two of the quaternionic pairs (n
2i1
. n
2i
). i = 1. 2. 3. 4 equal to
(0. 0) at the same time. One such quaternionic pair can vanish; hence, we need
gcd(j
1
. j
2
. j
3
) = 1 to get a free action. The analysis in the case when j
1
= 0 is
similar. Then one sees that the triple (j
2
. j
3
. j
4
) has to be admissible in the sense
of the Theorem 13.9.7.
As before we have we denote the quotient manifold corresponding to an admis-
sible
0
1,4
= (j
1
. j
2
. j
3
. j
4
) by
15
(p).
Open Problem 13.9.2: Determine the cohomology ring of the 3-Sasakian mani-
folds
15
(p).
Question 13.9.1: Can

k,m
be admissible for other values of (; /. :) with :
3 ?
CHAPTER 14
Sasakian Structures, Killing Spinors, and
Supersymmetry
The main purpose of this chapter is to provide a connection (non-rigorous) of
Sasaki-Einstein and related geometries to the theoretical physics of supersymmetric
quantum eld theory, especially to string theory and M-theory. There have been
many recent attempts to describe these important notions of theoretical physics in
precise mathematical terms. This enormous task is far beyond the scope of this
book, so we refer the reader to recent monographs and references therein [DEF
+
99,
Var04, Jos01, AJPS97]. Here we content ourselves with providing the main
theorems and results (often without proofs) concerning Killing spinors, and then
give a brief heuristic discussion in Section 14.5 describing the relation with physics.
In particular, we discuss some additional properties of Sasaki-Einstein manifolds
that relate Killing spinors to supersymmetry.
14.1. The Dirac Operator and Killing Spinors
We begin with a denition of spinor bundles and the bundle of Cliord algebras
of a vector bundle. Here we only give a quick denition and refer to [LM89, Fri00]
for details. Recall that the Cliord algebra C|(R
n
) over R
n
can be dened as the
quotient algebra of the tensor algebra T (R
n
) by the two-sided ideal 1 generated by
elements of the form +c() where c is a quadratic form on R
n
.
Denition 14.1.1: Let 1 be a vector bundle with inner product '. ` on a smooth
manifold `, and let T (`) denote the tensor bundle over 1. The Cliord bundle
of 1 is the quotient bundle C|(1) = T (`)1(1) where 1 is the bundle of ideals
(two-sided) generated pointwise by elements of the form +'. ` with 1
x
.
A real spinor bundle o(1) of 1 is a bundle of modules over the Cliord bundle
C|(1). Similarly, a complex spinor bundle is a bundle of complex modules over
the complexication C|(1) C.
As vector bundles C|(1) is isomorphic to the exterior bundle (1). but their
algebraic structures are dierent. The importance of C|(1) is that it contains the
spin group ojin(n), the universal (double) covering group of the orthogonal group
oO(n). so one obtains all the representations of ojin(n) by studying representations
of C|(1). We assume that the vector bundle 1 admits a spin structure as discussed
in Chapter 1, so n
2
(1) = 0. We are interested mainly in the case when (`. o) is a
Riemannian spin manifold and 1 = T` in which case we write o(`) instead of
o(T`). The Levi-Civita connection on T` induces a connection, also denoted
. on any of the spinor bundles o(`). or more appropriately on the sections
(o(`)).
529
530 14. SASAKIAN STRUCTURES, KILLING SPINORS, AND SUPERSYMMETRY
Denition 14.1.2: Let (`
n
. o) be a Riemannian spin manifold and let o(`) be
any spinor bundle. The Dirac operator is the rst order dierential operator
1 : (o(`))(o(`)) dened by
1 =
n

j=1
1
j

E
j
.
where 1
j
is a local orthonormal frame and denotes Cliord multiplication.
The Dirac operator, of course originating with the famous Dirac equation de-
scribing fermions in theoretical physics, was brought into mathematics by Atiyah
and Singer in [AS63]. Then Lichnerowicz [Lic63a] proved his famous result that
a Riemannian spin manifold with positive scalar curvature must have vanishing

-
genus. An interesting question on any spin manifold is: what are the eigenvectors
of the Dirac operator. In this regard the main objects of interest consists of special
sections of certain spinor bundles called Killing spinor elds or just Killing spinors
for short. Specically, (cf. [BFGK91, Fri00])
Denition 14.1.3: Let (`. o) be a complete n-dimensional Riemannian spin man-
ifold, and let o(`) be a spin bundle (real or complex) on ` and a smooth section
of o(`). We say that is a Killing spinor if for every vector eld A there is
C. called Killing number, such that

X
= A .
Here A denotes the Cliord product of A and . We say that is imaginary
when Im(C

). is parallel if = 0 and is real

1
if Re(C

).
We shall see shortly that the three possibilities for the Killing number : real,
imaginary, or 0. are the only possibilities. The name Killing spinor derives from
the fact that if is a non-trivial Killing spinor and is real, the vector eld
(14.1.1) A

=
n

j=1
o(. 1
j
)1
j
is a Killing vector eld for the metric o (which, of course, can be zero). If is a
Killing spinor on an n-dimensional spin manifold, then
(14.1.2) 1 =
n

j=1
1
j

Ej
=
n

j=1
1
j
1
j
= n .
So Killing spinors are eigenvectors of the Dirac operator with eigenvalue n. In
1980 Friedrich [Fri80] proved the following remarkable theorem:
Theorem 14.1.4: Let (`
n
. o) be a Riemannian spin manifold which admits a
non-trivial Killing spinor . Then (`
n
. o) is Einstein with scalar curvature : =
4n(n 1)
2
.
A proof of this is a straightforward curvature computation which can be found
in either of the books [BFGK91, Fri00]. It also uses the fact that a non-trivial
Killing spinor vanishes nowhere. It follows immediately from Theorem 14.1.4 that
must be one of the three types mentioned in Denition 14.1.3. So if the Killing
number is real then (`. o) must be a positive Einstein manifold. In particular, if
1
Here the standard terminology real and imaginary Killing spinors can be somewhat mis-
leading. The Killing spinor is usually a section of a complex spinor bundle. So a real Killing
spinor just means that is real.
14.1. THE DIRAC OPERATOR AND KILLING SPINORS 531
` is complete, then it is compact. On the other hand if the Killing number is pure
imaginary, Friedrich shows that ` must be non-compact.
The existence of Killing spinors not only puts restrictions on the Ricci curva-
ture, but also on both the Riemannian and the Weyl curvature operators [BFGK91].
Proposition 14.1.5: Let (`
n
. o) be a Riemannian spin manifold. Let be a
Killing spinor on ` with Killing number and let {. J :
2
`
2
` be the
Riemann and Weyl curvature operators, respectively. Then for any vector eld A
and any 2-from we have
J() = 0 ; (14.1.3)
(
X
J)() = 2

A J()

; (14.1.4)
({() + 4
2
) = 0 ; (14.1.5)
(
X
{)() = 2

A {() + 4
2
(A)

. (14.1.6)
These curvature equations can be used to prove (see [BFGK91] or [Fri00])
Theorem 14.1.6: Let (`
n
. o) be a connected Riemannian spin manifold admit-
ting a non-trivial Killing spinor with = 0. Then (`. o) is locally irreducible.
Furthermore, if ` is locally symmetric, or n 4. then ` is a space of constant
sectional curvature equal 4
2
.
Friedrichs main objective in [Fri80] was an improvement of Lichnerowiczs
estimate in [Lic63a] for the eigenvalues of the Dirac operator. Indeed, Friedrich
proves that the eigenvalues of the Dirac operator on any compact manifold satisfy
the estimate
(14.1.7)
2

1
4
n:
0
n 1
.
where :
0
is the minimum of the scalar curvature on `. Thus, Killing spinors
are eigenvectors that realize equality in equation (14.1.7). Friedrich also proves the
converse that any eigenvector of 1 realizing the equality must be a Killing spinor
with
(14.1.8) =
1
2

:
0
n(n 1)
.
Example 14.1.7: [Spheres] In the case of the round sphere (o
n
. o
0
) equality in
equation (14.1.7) is always attained. So normalizing such that :
0
= n(n 1). and
using the Bars Correspondence Theorem 14.2.1 below the number of corresponding
real Killing spinors equals the number of constant spinors on R
n+1
with the at
metric. The latter is well known (see the appendix of [PR88]) to be 2
|n/2|
for each
of the values =
1
2
. where n2| is the largest integer less than or equal to n2.
Remark 14.1.1: Actually (without making the connection to Sasakian geometry)
already in [Fri80] Friedrich gives a non-spherical example of a compact 5-manifold
with a real Killing spinor: ` = oO(4)oO(2) with its homogeneous Kobayashi-
Tanno Sasaki-Einstein structure.
We now wish to relate Killing spinors to the main theme of this book, Sasakian
geometry. First notice that if a Sasakian manifold `
2n+1
admits a Killing spinor,
Theorem 14.1.4 says it must be Sasaki-Einstein, so the scalar curvature :
0
=
2n(2n + 1), and equation (14.1.8) implies that =
1
2
. We have the following
result of Friedrich and Kath [FK90]
532 14. SASAKIAN STRUCTURES, KILLING SPINORS, AND SUPERSYMMETRY
Theorem 14.1.8: Every simply connected Sasaki-Einstein manifold admits non-
trivial real Killing spinors. Furthermore,
(i) if ` has dimension 4:+1 then (`. o) admits exactly one Killing spinor
for each of the values =
1
2
.
(ii) if ` has dimension 4:+3 then (`. o) admits at least two Killing spinors
for one of the values =
1
2
.
Outline of Proof. (Details can be found in [FK90] or the book [BFGK91].)
Every simply connected Sasaki-Einstein manifold is spin by Theorem 11.1.2, so `
has a spin bundle o(`). Given a xed Sasakian structure o = (. . . o) we con-
sider two subbundles c

(o) of o(`) dened by

(14.1.9) c

(o) = o(`) [ (2A +

A) = 0. A (T`) .
Set

X
=
X

1
2
A . A straightforward computation shows that

preserves the
subbundles c

and denes a connection there. Moreover, by standard curvature

computations it can be shown that the connection

is at in c

(o). So it has
covariantly constant sections which are precisely the Killing spinors. One then uses
some representation theory of Spin(2n + 1) to compute the dimensions of c
+
(o)
and c

(o) proving the result.

We have the following:
Corollary 14.1.9: Let (`. o) be a compact Sasaki-Einstein manifold of dimen-
sion 2: + 1. Then (`. o) is locally symmetric if and only if (`. o) is of constant
curvature. Moreover, Hol(o) = oO(2: + 1) and (`. o) is locally irreducible as a
Riemannian manifold.
Proof. If necessary, go to the universal cover

`. This is a compact simply
connected Sasaki-Einstein manifold; hence, it admits a non-trivial Killing spinor by
Theorem 14.1.8. The rst statement then follows from the Theorem 14.1.6. The
second statement follows from the Berger Theorem 1.4.8. Since ` has dimension
2: + 1 the only possibilities for Hol(o) are oO(2: + 1) and G
2
. But the latter is
Ricci at, so it cannot be Sasaki-Einstein.
Friedrich and Kath began their investigation in dimension 5 [FK89] where
they showed that a simply-connected compact 5-manifold which admits a Killing
spinor must be Sasaki-Einstein. In dimension 7 they showed that there are exactly
three possibilities: weak G
2
-manifolds, Sasaki-Einstein manifolds which are not 3-
Sasakian, and 3-Sasakian manifolds [FK90]. Later Grunewald gave a description of
6-manifolds admitting Killing spinors [Gru90]. We should add an earlier result of
Hijazi who showed that the only 8-dimensional manifold with Killing spinors must
be the round sphere [Hij86]. By 1990 a decade of research by many people slowly
identied all the ingredients of a classication of such manifolds in terms of their un-
derlying geometric structures. The pieces of the puzzle consisting of round spheres
in any dimension, Sasaki-Einstein manifolds in odd dimensions, nearly Kahler man-
ifolds in dimension 6, and weak G
2
-holonomy manifolds in dimension 7 were all in
place with plenty of interesting examples to go around [BFGK91]. What remained
at that stage was to show that in even dimensions greater than 8 there is noth-
ing else but the round spheres, while in odd dimensions greater than 7 the only
such examples must be Sasaki-Einstein. The missing piece of the puzzle was -
nally uncovered by Bar: real Killing spinors on ` correspond to parallel spinors on
14.2. REAL KILLING SPINORS, HOLONOMY AND B

ARS CORRESPONDENCE 533

the cone C(`) [Bar93]. A bit earlier Wang [Wan89] had shown that on a sim-
ply connected complete Riemannian spin manifold the existence of parallel spinors
correspond to reduced holonomy. This led Bar to an elegant description of the
geometry of manifolds admitting real Killing spinors (in any dimension) in terms of
special holonomies of the associated cones. We refer to the correspondence between
real Killing spinors on ` and parallel spinors on the cone C(`) (equivalently re-
duced holonomy) as Bars correspondence. In particular, this correspondence not
only answered the last remaining open questions, but also allowed for simple unied
proofs of most of the theorems obtained earlier.
14.2. Real Killing Spinors, Holonomy and Bars Correspondence
As mentioned the Bar correspondence relates real Killing spinors on a com-
pact Riemannian spin manifold (`. o) to parallel spinors on the Riemannian cone
(C(`). o). We now make this statement precise.
Theorem 14.2.1: Let (`
n
. o) be a compact Riemannian spin manifold and let
(C(`
n
). o) be its Riemannian cone. Then there is a one to one correspondence
between real Killing spinors on (`
n
. o) with =
1
2
and parallel spinors on
(C(`
n
). o).
Proof. The existence of a parallel spinor on (C(`
n
). o) implies that o is Ricci
at by Theorem 14.1.4. Then by Lemma 11.1.5 (`
n
. o) is Einstein with scalar
curvature : = n(n 1). So any Killing spinors must have =
1
2
by equation
(14.1.8). As in the proof of Theorem 14.1.8,

X
=
X

1
2
A denes a connection
in the spin bundle o(`). The connection 1-forms

of

are related to the

connection 1-form of the Levi-Civita connection by

=
1
2
. where is a 1-
form called the soldering form. This can be interpreted as a connection with values
in the Lie algebra of spn(n +1) = spn(n) R
n
. and pulls back to the Levi-Civita
connection in the spin bundles on the cone (C(`
n
). o). So parallel spinors on the
cone correspond to parallel spinors on (`. o) with respect to the connection

which correspond precisely to real Killing spinors with respect to the Levi-Civita
connection.
Now we have the following denition:
Denition 14.2.2: We say that a Riemannian spin manifold (`. o) is of type
(j. c) if it carries exactly j linearly independent real Killing spinors with 0 and
exactly c linearly independent real Killing spinors with < 0.
The following theorem has an interesting history. As mentioned above it
was Bar [Bar93] who recognized the correspondence between real Killing spinors
on (`. o) and parallel spinors on the Riemannian cone (C(`). o). The relation
between parallel spinors and reduced holonomy was anticipated in the work of
Hitchin [Hit74] and Bonan [Bon66], but was formalized in the 1989 paper of
Wang [Wan89]. It has also been generalized to the non-simply connected case in
[Wan95, MS00].
Theorem 14.2.3: Let (`
n
. o) be a complete simply connected Riemannian spin
manifold, and let Hol( o) be the holonomy group of the Riemannian cone (C(`). o).
Then (`
n
. o) admits a non-trivial real Killing spinor with (`
n
. o) of type (j. c) if
and only if

dim`. Hol( o). (j. c)

is one of the 6 possible triples listed in the table

below:
534 14. SASAKIAN STRUCTURES, KILLING SPINORS, AND SUPERSYMMETRY
dim(`) Hol( o) type (j. c)
n id (2
|n/2|
. 2
|n/2|
)
4:+ 1 ol(2:+ 1) (1. 1)
4:+ 3 ol(2:+ 2) (2. 0)
4:+ 3 oj(:+ 1) (:+ 2. 0)
7 ojin(7) (1. 0)
6 G
2
(1. 1)
Here : 1. and n 1.
Outline of Proof. Since (`. o) is complete and has a non-trivial real Killing
spinor, it is compact by Theorem 14.1.4. It then follows from a theorem of Gallot
[Gal79] that if the Riemannian cone (C(`). o) has reducible holonomy it must be
at. So we can apply Bergers Theorem 1.4.8. Now Wang [Wan89] used the spinor
representations of the possible irreducible holonomy groups on Bergers list to give
the correspondence between these holonomy groups and the existence of parallel
spinors. First he showed that the groups listed in Table 1.4.7 that are not on the
above table do not admit parallel spinors. Then upon decomposing the spin rep-
resentation of the group in question into irreducible pieces, the number of parallel
spinors corresponds to the multiplicity of the trivial representation. Wang computes
this in all but the rst line of the table when (C(`). o) is at. In this case (`. o) is
a round sphere as discussed in Example 14.1.7, so the number of linearly indepen-
dent constant spinors is (2
|n/2|
. 2
|n/2|
). By Bars Correspondence Theorem 14.2.1
real Killing spinors on (`. o) correspond precisely to parallel spinors on (C(`). o).
Note that the hypothesis of completeness in Wangs theorem [Wan89] is not neces-
sary, so that the correspondence between the holonomy groups and parallel spinors
holds equally well on Riemannian cones. However, the completeness assumption on
(`. o) guarantees the irreducibility of the cone (C(`). o) as mentioned above.
Let us briey discuss the types of geometry involved in each case of this the-
orem. As mentioned in the above proof the rst line of the table corresponds to
the round spheres. The next three lines correspond to Sasaki-Einstein geometry,
so Theorem 14.2.3 generalizes the Friedrich-Kath Theorem 14.1.8 in this case. The
last of these three lines corresponds precisely to 3-Sasakian geometry by Denition
13.1.9. Finally the two cases whose cones have exceptional holonomy will be dis-
cussed in more detail in Section 14.3.1 below. Suce it here to mention that it
was observed by Bryant and Salamon [BS89] that a cone on a weak holonomy G
2
manifold has its own holonomy in ojin(7). It is interesting to note that Theorem
14.2.3 generalizes the result of Hijazi in dimension eight mentioned earlier as well
as part of the last statement in Theorem 14.1.6, namely
Corollary 14.2.4: Let (`
2n
. o) be a complete simply connected Riemannian spin
manifold of dimension 2n with n = 3 admitting a non-trivial real Killing spinor.
Then ` is isometric to the round sphere.
14.3. GEOMETRIES ASSOCIATED WITH 3-SASAKIAN 7-MANIFOLDS 535
We end this section with a brief discussion of the non-simply connected case.
Here we consider two additional cases for Hol( o), namely ol(2: + 2) Z
2
and
oj(2) Z
d
. See [Wan95, MS00] for the list of possibilities.
Example 14.2.5: Hol( o) = ol(2:) Z
2
. Consider the (4: 1)-dimensional
Stiefel manifold \
2
(R
2m+1
) with its homogeneous Sasaki-Einstein metric. As de-
scribed in Example 9.3.23 the quotient manifold `
4m1

of \
2
(R
2m+1
) by the
free involution induced from complex conjugation has an Einstein metric which
is locally Sasakian. The cone C(`
4m1

) is not Kahler and its holonomy is

Hol( o) = ol(2:+2) Z
2
. According to Wang [Wan95] C(`
4m1

) admits a spin
structure with precisely one parallel spinor if and only if : is even, and according
to Moroianu and Semmelmann [MS00] C(`
4m1

) admits exactly two spin struc-

tures each with precisely one parallel spinor if : is even. Thus, by Theorem 14.2.1
`
4m1

admits exactly two spin structures each with exactly one Killing spinor if
and only if : is even.
Example 14.2.6: Consider a 3-Sasakian manifold (`
4n1
. S) and choose a Reeb
vector eld (). Let C
m
be the cyclic subgroup of order : 2 of the circle group
generated by (). Assume that : is relatively prime to the order (S) of S and
that the generic bre of the fundamental 3-dimensional foliation T
Q
is oO(3). so
that C
m
acts freely on `
4n1
. This last condition on the generic bre is easy to
satisfy; for example, it holds for any of the 3-Sasakian homogeneous spaces other
than the standard round sphere, as well as the biquotients described in Section
13.7.4 when all the j
j
s are odd. (To handle the case when the generic bre is oj(1)
we simply need to divide : by two when it is even). Since C
m
is not in the center
of oO(3). the quotient `
4n1
C
m
is not 3-Sasakian. However, C
m
does preserve
the Sasakian structure determined by (). so `
4n1
C
m
is Sasaki-Einstein. The
cone C(`
4n1
C
m
) has holonomy oj(n) Z
m
, and admits precisely
n+1
m
parallel
spinors if and only if : divides n + 1 [Wan95, MS00]. Thus, by Theorem 14.2.1
`
4n1
C
m
admits precisely
n+1
m
Killing spinors when : divides n + 1.
14.3. Geometries Associated with 3-Sasakian 7-manifolds
It is most remarkable that to each 4n-dimensional positive QK metric (O. o
c
)
(even just locally) one can associate nine other Einstein metrics in dimensions 4n+/,
/ = 1. 2. 3. 4. Alternatively, one could say that each 3-Sasakian metric (`. o) canon-
ically denes an additional nine Einstein metrics in various dimensions. We have
already encountered all of these metrics. First there are the four geometries of the
diamond diagram (`. S) of Denition 13.3.15. Then ` and Z admit additional
squashed Einstein metrics discussed in Section 13.3.3. Thus we get ve Einstein
metrics with positive Einstein constants: (O. o
c
). (`. o). (`
t
. o
t
). (Z. /). (Z
t
. /
t
).
Of course ` `
t
and Z Z
t
as smooth manifolds (orbifolds) but they are dier-
ent as Riemannian manifolds (orbifolds), hence, the notation. Let us scale all these
metrics so that the Einstein constant equals the dimension of the total space minus
1. Note that any 3-Sasakian metric already has this property. In the other four
cases this is a choice of scale which is quite natural due to Lemma 11.1.5. However,
note that this is not the scale one gets for (Z. /). and (O. o
c
) via the Riemannian
submersion from (`. o). Now, in each case one can consider its Riemannian cone
which will be Ricci-at by Lemma 11.1.5. We thus obtain ve Ricci-at metrics
on the corresponding Riemannian cones. In addition, one can also take (iterated)
sine-cone metrics (see Equation (14.4.1) below) on the same ve base spaces. These
536 14. SASAKIAN STRUCTURES, KILLING SPINORS, AND SUPERSYMMETRY
metrics are all Einstein of positive scalar curvature (cf. Theorem 14.4.2). Let us
summarize all this with the following extension of (`. S):
(14.3.1) C(Z
t
)
Z
t ?
_

`
t
.

C(`
t
)
O



C(O)
C(Z) Z
?
_

C(`)
There would perhaps be nothing special about all these 10 (and many more
by iterating the sine-cone construction) geometries beyond what has already been
discussed in Chapter 13. This is indeed true when dim(`) 7. However, when
dim(`) = 7, or, alternatively, when O is a positive self-dual Einstein orbifold metric
(more generally, just a local metric of this type) some of the metrics occurring in
diagram (14.3.1) have additional properties. We shall list all of them rst. For the
moment, let us assume that (`. o) is a compact 3-Sasakian 7-manifold, then the
following hold:
(i) (O. o
c
) is a positive self-dual Einstein manifold (orbifold). We will think
of it as the source of all the other geometries.
(ii) (C(O). dt
2
+t
2
o
c
) is a 5-dimensional Ricci-at cone with base O.
(iii) (Z. /) is the orbifold twistor space of O.
(iv) (Z
t
. /
t
) is a nearly-Kahler manifold (orbifold).
(v) (`. o) is the 3-Sasakian manifold.
(vi) (`
t
. o
t
) is a 7-manifold with weak G
2
structure.
(vii) (C(Z). dt
2
+t
2
/) is a 7-dimensional Ricci-at cone with base Z.
(viii) (C
s
(Z
t
). dt
2
+ sin
2
t/
t
) is a 7-manifold with weak G
2
structure.
(ix) (C(Z
t
). dt
2
+t
2
/
t
) has holonomy contained in G
2
.
(x) (C(`). dt
2
+t
2
o) is hyperkahler with holonomy contained in oj(2).
(xi) (C(`
t
). dt
2
+t
2
o
t
) has holonomy contained in ojin(7).
The cases (ii) and (vii) do not appear to have any special properties other than
Ricci-atness. The cases (i), (iii), (v), and (x) are the four geometries of (`. S).
The ve remaining cases are all very interesting from the point of view of the clas-
sication of Theorem 14.2.3. Indeed Z
t
and C(Z
t
) are examples of the structures
listed in the last row of the table while `
t
and C(`
t
) (as well as C
s
(Z
t
) and
C(C
s
(Z
t
))) give examples of the structures listed in the fth row. In particular,
our diagram (14.3.1) provides for a cornucopia of the orbifold examples in the rst
case and smooth manifolds in the latter.
14.3.1. Weak G
2
-Structures and ojin(7) Holonomy Cones. Recall, that
geometrically G
2
is dened to be the Lie group acting on the imaginary octonions
R
7
and preserving the 3-form
=
1

2

3
+
1
(
4

5

6

7
)
+
2
(
4

6

7

5
) +
3
(
4

7

5

6
).
(14.3.2)
where
i

7
i=1
is a xed orthonormal basis of the dual of R
7
. A G
2
structure
on a 7-manifold ` is, by denition, a reduction of the structure group of the
tangent bundle to G
2
. This is equivalent to the existence of a global 3-form
14.3. GEOMETRIES ASSOCIATED WITH 3-SASAKIAN 7-MANIFOLDS 537

3
(`) which may be written locally as 14.3.2. Such a 3-form denes an associated
Riemannian metric, an orientation class, and a spinor eld of constant length. The
following terminology is due to Gray [Gra71]:
Denition 14.3.1: Let (`. o) be a complete 7-dimensional Riemannian manifold.
We say that (`. o) has weak holonomy G
2
if there exist a global 3-form

3
(`) which locally can be written in terms of a local orthonormal basis as in
14.3.2, and d = c , where is the Hodge star operator associated to o and c is
a constant whose sign is xed by an orientation convention.
The equation d = c implies that is nearly parallel in the sense that
only a 1-dimensional component of is dierent from zero [FG82]. Thus, a weak
holonomy G
2
structure is sometimes called a nearly parallel G
2
structure. The case
of c = 0 is somewhat special. In particular, it is known [Sal89] that the condition
d = 0 = d is equivalent to the condition that be parallel, i.e., = 0 which
is equivalent to the condition that the metric o has holonomy group contained in
G
2
. The following theorem provides the connection with the previous discussion on
Killing spinors [Bar93]
Theorem 14.3.2: Let (`. o) be a complete 7-dimensional Riemannian manifold
with weak holonomy G
2
. Then the holonomy Hol( o) of the metric cone (C(`). o)
is contained in ojin(7). In particular, C(`) is Ricci-at and ` is Einstein with
positive Einstein constant = 6.
Remark 14.3.1: The sphere o
7
with its constant curvature metric is isometric to
the isotropy irreducible space ojin(7)G
2
. The fact that G
2
leaves invariant (up
to constants) a unique 3-form and a unique 4-form on R
7
implies immediately that
this space has weak holonomy G
2
.
Denition 14.3.3: Let (`. o) be a complete 7-dimensional Riemannian manifold.
We say that o is a proper G
2
-metric if Hol( o) = Spin(7).
We emphasize here that G
2
is the structure group of `. not the Riemannian
holonomy group. Specializing Theorem 14.2.3 to dimension 7 gives the following
theorem due to Friedrich and Kath [FK90].
Theorem 14.3.4: Let (`
7
. o) be a complete simply-connected Riemannian spin
manifold of dimension 7 admitting a non-trivial real Killing spinor with 0 or
< 0. Then there are four possibilities:
(i) (`
7
. o) is of type (1. 0) and it is a proper G
2
-manifold,
(ii) (`
7
. o) is of type (2. 0) and it is a Sasaki-Einstein manifold, but (`
7
. o)
is not 3-Sasakian,
(iii) (`
7
. o) is of type (3. 0) and it is 3-Sasakian,
(iv) (`
7
. o) = (o
7
. o
can
) and is of type (8. 8).
Conversely, if (`
7
. o) is a compact simply-connected proper G
2
-manifold then it
carries precisely one Killing spinor with 0. If (`
7
. o) is a compact simply-
connected Sasaki-Einstein 7-manifold which is not 3-Sasakian then ` carries pre-
cisely 2 linearly independent Killing spinors with 0. Finally, if (`
7
. o) is a
3-Sasakian 7-manifold, which is not of constant curvature, then ` carries precisely
3 linearly independent Killing spinors with 0.
Remark 14.3.2: The four possibilities of the Theorem 14.3.4 correspond to the
sequence of inclusions
Spin(7) ol(4) oj(2) id .
538 14. SASAKIAN STRUCTURES, KILLING SPINORS, AND SUPERSYMMETRY
All of the corresponding cases are examples of weak holonomy G
2
metrics. If
we exclude the trivial case when the associated cone is at, we have three types of
weak holonomy G
2
geometries. Following [FKMS97] we use the number of linearly
independent Killing spinors to classify the types of weak holonomy G
2
geometries.
We call these type I, II, and III corresponding to cases (i), (ii), and (iii) of Theorem
14.3.4, respectively.
We are now ready to describe the weak G
2
geometry of the `
t
C(`
t
) part
of Diagram 14.3.1 [GS96, FKMS97]:
Theorem 14.3.5: Let (`. S) be a 7-dimensional 3-Sasakian manifold. Then the
3-Sasakian metric o has weak holonomy G
2
. Moreover, the second Einstein metric
o
t
given by Theorem 13.3.18 and scaled so that the Einstein constant = 6 has
weak holonomy G
2
. In fact o
t
is a proper G
2
metric.
Proof. For the second Einstein metric o
t
we have three mutually orthonormal
1-forms
1
=

t
1
.
2
=

t
2
.
3
=

t
3
. where t is the parameter of the
canonical variation. Let
4
.
5
.
6
.
7
be local 1-forms spanning the annihilator
of the vertical subbundle 1
3
in T

o such that

1
= 2(
4

5

6

7
) .

2
= 2(
4

6

7

5
) .

3
= 2(
4

7

5

6
) .
Then the set
1
. . . . .
7
forms a local orthonormal coframe for the metric o
t
. In
terms of the 3-forms and of 13.5.1 we have =
1
2

t +

t
3
. One easily
sees that this is of the type of equation (14.3.2) and, therefore, denes a compatible
G
2
-structure. Moreover, a straightforward computation gives
d =
1
2

t +

t(t + 1)d. =
1
2
td
1
24
.
Thus, d = c is solved with

t = 1

5, and c = 12

5. So o
t
has weak
holonomy G
2
. That o
t
is a proper G
2
metric is due to [FKMS97]. The idea is
to use Theorem 14.3.4. Looking at the four possibilities given in that theorem,
we see that it suces to show that o
t
is not Sasaki-Einstein. The details are in
[FKMS97].
Example 14.3.6: 3-Sasakian 7-manifolds are plentiful as we have seen in Chapter
13. All of them give, by Theorem 14.3.5, examples of type I and type III geome-
tries. Examples of simply connected type I geometries that do not arise via Theorem
14.3.5 are the homogeneous Alo-Wallach spaces `
7
m,n
. (:. n) = (1. 1). discussed
in Section 13.7.4 as special case of Eschenburg biquotients [CMS96, BFGK91]
together with an isotropy irreducible homogeneous space dened as follows: Con-
sider the space H
2
of homogeneous polynomials of degree 2 in three real variables
(r
1
. r
2
. r
3
). As dim(H
2
) = 5 it gives rise to the embedding oO(3) oO(5). We take
` = oO(5)oO(3). This example was used by Bryant to get the rst 8-dimensional
metric with holonomy ojin(7) [Bry87]. Examples of type II geometries (Sasaki-
Einstein) are equally rich as seen in Chapter 11, Sections 11.5 and 11.7. In partic-
ular, there are hundreds of examples of type II weak holonomy G
2
metrics on each
of the 28 homotopy spheres in dimension 7.
14.3. GEOMETRIES ASSOCIATED WITH 3-SASAKIAN 7-MANIFOLDS 539
Remark 14.3.3: According to [CMS96] the Alo-Wallach manifold `
7
1,1
has three
Einstein metrics. One is the homogeneous 3-Sasakian metric. The second is the
proper G
2
metric of Theorem 14.3.5, while the third Einstein metric also has weak
holonomy G
2
most likely of type I but we could not positively exclude type II as a
possibility.
Open Problem 14.3.1: Classify all compact 7-manifolds with weak G
2
holonomy
of type I, II, or III, respectively.
Type III classication is the classication of all compact 3-Sasakian 7-manifolds,
or more generally all compact positive self-dual Einstein orbifolds. This is proba-
bly very hard. The case of 3-Sasakian 7-manifolds with vanishing aut(`. S) ap-
pears quite dicult. Type II classication (7-dimensional Sasaki-Einstein manifolds
which are not 3-Sasakian) is clearly completely out of reach at the moment. A clas-
sication of proper weak G
2
structures on a compact manifold that do not arise via
Theorem 14.3.5 would be very interesting and it is not clear how hard this problem
really is.
Remark 14.3.4: The holonomy ojin(7) cone metrics are plentiful but never com-
plete. However, some of these metrics can be deformed to complete holonomy
ojin(7) ones on non compact manifolds. The rst example was obtained in Bryant
and Salamon who observed that the spin bundle over o
4
with its canonical metric
carries a complete metric with holonomy ojin(7) [BS89]. Locally the metric was
later considered also in [GPP90]. More generally, spin orbibundles over positive
QK orbifolds also carry such complete orbifold metrics as observed by Bryant and
Salamon in [BS89]. Other complete examples were constructed later by physicists
[CGLP02, CGLP04, KY02a, KY02b]. Finally, the rst compact examples
were obtained in 1996 by Joyce [Joy96a, Joy99]. See Joyces book [Joy00] for an
excellent detailed exposition of the methods and the discussion of examples.
Open Problem 14.3.2: Let (`
7
. S) be any 3-Sasakian 7-manifold introduced in
Chapter 13 and let (`
7
. o
t
) be the associated proper weak holonomy G
2
squashed
metric. Consider the two Riemannian cones (C(`). dt
2
+t
2
o
t
). (C(`). dt
2
+t
2
o).
(i) When does the metric cone (C(`). dt
2
+t
2
o
t
) admit a complete holonomy
ojin(7) deformation?
(ii) When does the metric cone (C(`). dt
2
+t
2
o) admit a complete holonomy
oj(2) (hyperkahler) deformation?
The metric on the spin bundle o(o
4
) by Bryant and Salamon is a deformation
on the ojin(7) holonomy metric on the cone over the squashed metric on o
7
[CGLP02, CGLP04], so there are examples of such deformations regarding ques-
tion (i). Regarding (ii): Recall that every compact 3-Sasakian 3-manifold is isomet-
ric to o
3
and the metric cone is the at cone C
2
. Hence, one could think of (ii)
as a 7-dimensional analogue of the similar problem whose complete solution was
described in Section 12.10. There are non-trivial examples also in the higher dimen-
sional cases. The metric cone on the homogeneous 3-Sasakian manifold o(1. 1. 1)
admits complete HK deformations, namely the Calabi metric on T

CP
2
of Example
12.8.5. We do not know of any other examples at the moment.
14.3.2. Nearly Kahler 6-Manifolds and G
2
Holonomy Cones. In this
section we explain the geometry of the Z
t
C(Z
t
) part of the diagram 14.3.1.
Before we specialize to dimension 6 we begin with a more general introduction.
540 14. SASAKIAN STRUCTURES, KILLING SPINORS, AND SUPERSYMMETRY
Nearly Kahler manifolds were rst studied by Tachibana in [Tac59] and they appear
under the name of almost Tachibana spaces in Chapter VIII of the book [Yan65].
They were then rediscovered by Gray [Gra70] and given the name nearly Kahler
manifolds which by now is the accepted name.
Denition 14.3.7: A nearly Kahler manifold is an almost Hermitian manifold
(`. o. J. ) such that (
X
J)A = 0 for all tangent vectors A, where is the Levi-
Civita connection and J is the almost complex structure. One says that a nearly
Kahler manifold is strict if it is not Kahler.
This denition is equivalent to the condition
(14.3.3) (
X
J)Y + (
Y
J)A = 0
for all vector elds A. Y. which is to say that J is a Killing tensor eld. An
alternative characterization of nearly Kahler manifolds is given by
Proposition 14.3.8: An almost Hermitian manifold (`. o. J. ) is nearly Kahler
if and only if
=
1
3
d .
In particular, a strict nearly Kahler structure is never integrable.
Any nearly Kahler manifold can be locally decomposed as the product of a
Kahler manifold and a strict nearly Kahler manifold. Such a decomposition is
global in the simply connected case [Nag02b]. Hence, the study of nearly Kahler
manifolds reduces to the case of strict ones. In addition every nearly Kahler mani-
fold in dimension 4 must be Kahler so that the rst interesting dimension is six.
The following theorem establishes relationship between the twistor space ZO
of a quaternionic Kahler manifold (orbifold) and nearly Kahler geometry.
Theorem 14.3.9: Let : (Z. /)(O. o
c
) be the twistor space of a positive QK
manifold with its Kahler structure (J. /.
h
). Then Z admits a strict nearly Kahler
structure (J
1
. /
1
.
h
1
). If T` = 1 H is the natural splitting induced by then
(14.3.4) /[
\
= 2/
1
[
\
. /[
1
= /
1
[
1
=

(o
c
) .
(14.3.5) J[
\
= J
1
[
\
. J[
1
= J
1
[
1
.
Theorem 14.3.9 is due to Eels and Salamon when O is 4-dimensional. The
higher dimensional analogue was established in [AGI98] (see also [Nag02b]).
Remark 14.3.5: Observe that the metric of the nearly Kahler structure of Theorem
14.3.9, in general, is not Einstein. In particular, /
1
is not the squashed metric /
t
introduced in the diagram 14.3.1, unless dim(Z) = 6. In six dimensions, we can
scale /
1
so that it has scalar curvature : = 30 and then indeed /
1
= /
t
as one can
easily check.
Denition 14.3.10: Let ` = GH be a homogeneous space. We say that ` is
3-symmetric if G has an automorphism of order 3 such that G

0
H G

,
where G

is the xed point set of and G

0
is the identity component in G

0
.
We have the following two theorems concerning nearly Kahler homogeneous
Riemannian manifolds. The rst is due to Wolf and Gray in all dimensions but
six [WG68a, WG68b]. They also conjectured that the result is true for strict
nearly Kahler 6-manifolds. The Wolf-Gray conjecture was proved quite recently by
Butruille [But05, But06] which is the second theorem below.
14.3. GEOMETRIES ASSOCIATED WITH 3-SASAKIAN 7-MANIFOLDS 541
Theorem 14.3.11: Every compact homogeneous strict nearly Kahler manifold `
of dimension dierent than 6 is 3-symmetric.
Theorem 14.3.12: Let (`. o) be a strict nearly Kahler 6-dimensional Riemannian
homogeneous manifold. Then ` is isomorphic as a homogeneous space to a nite
quotient of GH. where G and H are one of the following:
(i) G = ol(2) ol(2) and H = id;
(ii) G = G
2
and H = ol(3). where metrically GH = o
6
the round sphere;
(iii) G = oj(2) and H = ol(2)l(1). where GH = CP
3
with its nearly
Kahler metric determined by Theorem 14.3.9;
(iv) G = ol(3) and H = T
2
, where GH is the ag manifold with its nearly
Kahler metric determined by Theorem 14.3.9.
Each of these manifolds carries a unique invariant nearly Kahler structure, up to
homothety.
In every dimension, the only known compact examples of nearly Kahler mani-
folds are 3-symmetric. On the other hand, Theorem 14.3.9 can be easily generalized
to the case of orbifolds so that there are plenty examples of compact inhomogeneous
strict nearly Kahler orbifolds in every dimension.
Theorem 14.3.13: Let ` be a compact simply-connected strict nearly Kahler
manifold. Then, in all dimensions, as a Riemannian manifold ` decomposes as a
product of
(i) 3-symmetric spaces,
(ii) twistor spaces of positive QK manifolds O such that O is not symmetric,
(iii) 6-dimensional strict nearly Kahler manifold other than the ones listed in
Theorem 14.3.12.
This theorem is due to Nagy [Nag02a], but our formulation uses the result of
Butruille together with the fact that the twistor spaces of all symmetric positive QK
manifolds are 3-symmetric. The LeBrun-Salamon conjecture can now be phrased
as follows
Conjecture 14.3.14: Any compact simply connected strict irreducible nearly Kahler
manifold (`. o) of dimension greater than 6 must be a 3-symmetric space.
In particular, the Conjecture 14.3.14 is automatically true in dimensions 4n
because of Nagys classication theorem and also true in dimensions 10 and 14
because all positive QK manifolds in dimension 8 and 12 are known. The third
case leads to an important
Open Problem 14.3.3: Classify all compact strict nearly Kahler manifolds in
dimension 6.
Dimension six is special not just because of the role they play in Theorem
14.3.13. They have several remarkable properties which we summarize in the fol-
lowing theorem.
Theorem 14.3.15: Let (`. J. o.
g
) be a compact strict nearly Kahler 6-manifold.
Then
(i) The metric o is Einstein of positive scalar curvature.
(ii) c
1
(`) = 0 and n
2
(`) = 0.
542 14. SASAKIAN STRUCTURES, KILLING SPINORS, AND SUPERSYMMETRY
(iii) If o is scaled so it has Einstein constant = 5 then the metric cone
(C(`). dt
2
t + t
2
o) has holonomy contained in G
2
. In particular, C(`)
is Ricci-at.
The rst property is due to Matsumoto [Mat72] while the second is due to
Gray [Gra76]. The last part is due to Bar [Bar93]. In fact, nearly Kahler 6-
manifolds is the geometry of the last row of the table of Theorem 14.2.3. More
precisely we have the following theorem proved by Grunewald [Gru90]:
Theorem 14.3.16: Let (`
6
. o) be a complete simply connected Riemannian spin
manifold of dimension 6 admitting a non-trivial Killing spinor with 0 or < 0.
Then there are two possibilities:
(i) (`. o) is of type (1. 1) and it is a strict nearly Kahler manifold,
(ii) (`. o) = (o
6
. o
can
) and is of type (8. 8).
Conversely, if (`. o) is a compact simply-connected strict nearly Kahler 6-manifold
of non-constant curvature then ` is of type (1. 1).
Compact strict nearly Kahler manifolds with isometries were investigated in
[MNS05] where it was shown that
Theorem 14.3.17: Let (`. J. o.
g
) be a compact strict nearly Kahler 6-manifold.
If ` admits a unit Killing vector eld, then up to nite cover ` is isometric to
o
3
o
3
with its standard nearly Kahler structure.
Remark 14.3.6: The rst example of a non-trivial G
2
holonomy metric was found
by Bryant [Bry87], who observed that a cone on the complex ag manifold U(3)T
3
carries an incomplete metric with G
2
-holonomy. The ag U(3)T
3
is the twistor
space of the complex projective plane CP
2
and as such it also has a strict nearly
Kahler structure. As explained in this section, this therefore is just one possible
example. One gets such non-trivial metrics also for the cones with bases CP
3
and
o
3
o
3
with their homogeneous strict nearly Kahler structures. Interestingly,
in some cases there exist complete metrics with G
2
holonomy which are smooth
deformations of the asymptotically conical ones. This fact was noticed by Bryant
and Salamon [BS89] who constructed complete examples of G
2
holonomy metrics
on bundles of self-dual 2-forms over CP
2
and o
4
. Replacing the base with any
positive QK orbifold O of Section 12.5 gives complete (in the orbifold sense) metrics
on orbibundles of self-dual 2-forms over O. Locally some of these metrics were
considered in [San03]. More complete examples of explicit G
2
holonomy metrics on
non-compact manifolds were obtained by Salamon [Sal04]. G
2
holonomy manifolds
with isometric circle actions were investigated by Apostolov and Salamon [AS04].
The rst compact examples are due to the ground breaking work of Joyce [Joy96b].
14.4. Geometries Associated with Sasaki-Einstein 5-manifolds
Like 3-Sasakian manifolds Sasaki-Einstein 5-manifolds are naturally associated
to other geometries introduced in the previous section. Of course, each such space
(`
5
. o) comes with its Calabi-Yau cone (C(`). o) and, if the Sasaki-Einstein struc-
ture o is quasi-regular, with its quotient log del Pezzo surface (Z. /). But as it
turns out, there are two more Einstein metrics associated to o. The examples of
this section also illustrate how the Theorem 14.2.3 and the Bars correspondence
break down when (`. o) is a manifold with Killing spinors which is, however, not
complete.
14.4. GEOMETRIES ASSOCIATED WITH SASAKI-EINSTEIN 5-MANIFOLDS 543
We begin by describing a relation between 5-dimensional Sasaki-Einstein struc-
tures and six dimensional nearly Kahler structures which was uncovered recently
in [FIMU06]. This relation involves the so-called sine cone. This is the warped
product metric on C
s
() = (0. ) of Denition 6.5.1 with (t) = sin
2
t. so the
Riemannian metric is
(14.4.1) o
s
= dt
2
+ sin
2
t o
N
.
We use the notation o
s
to distinguish this metric from the usual Riemannian cone
metric o. Of course this metric is not complete, but one can compactify ` obtaining
a very tractable stratied space

` = [0. ] with conical singularities at t = 0
and t = . Observe the following simple fact which shows that the Riemannian
cone on a sine cone is always a Riemannian product.
Lemma 14.4.1: Let (`. o) be a Riemannian manifold. Then the product metric
d:
2
= dr
2
+dn
2
+n
2
o on RC(`) can be identied with the iterated cone metric
on C(C
s
(`)).
Proof. Consider the map R
+
(0. )R R
+
given by polar coordinate
change (:. t) (r. n) = (: cos t. : sin t). where : 0 and t (0. ). We get
d:
2
= dr
2
+dn
2
+n
2
o = d:
2
+:
2
dt
2
+:
2
sin
2
to = d:
2
+:
2
(dt
2
+ sin
2
to) .
So the iterated Riemannian cone (C(C
s
(`)). d:
2
) has reducible holonomy
1 Hol(C(`)). The importance of sine-cones can be seen via the following easy
consequence of Lemmas 14.4.1 and 11.1.5:
Theorem 14.4.2: Let (
m
. o) be a an Einstein manifold with Einstein constant
:1. Then the sine-cone C
s
() with the metric o
s
given by Equation (14.4.1) is
Einstein with Einstein constant :.
One can iterate this construction. Upon each iteration we get an Einstein
metric in one dimension higher with an appropriately scaled Einstein constant.
However, each iteration also introduces more singularities. One exception is the
sphere metric (o
m
. o
can
) scaled in each iteration so that Einstein constant equals
: 1. The sine-cone metric gives the constant curvature metric with Einstein
constant :. In each iteration one gets only a part of the sphere, but it can be
completed to the usual round sphere metric on o
m+1
. The case of special interest
to us is when the Einstein manifold
m
of Theorem 14.4.2 is Sasaki-Einstein. Of
course, the Einstein constant is such that the sine-cone is Einstein, but now it
inherits more structure. For example, when : = 5 the metric on C
s
(
5
) will be
nearly Kahler as we discuss next. First compare Lemma 14.4.1 with the following
result in [Joy00], Propositions 11.1.1-2:
Proposition 14.4.3: Let (`
4
. o
4
) and (`
6
. o
6
) be Calabi-Yau manifolds. Let
(R
3
. d:
2
= dr
2
+dn
2
+d.
2
) and (R. d:
2
= dr
2
) be the Euclidean spaces. Then
(i) (R
3
`
4
. o = d:
2
+o
4
) has a natural G
2
structure and o has holonomy
Hol(o) 1l
3
ol(2) G
2
.
(ii) (R `
6
. o = d:
2
+ o
6
) has a natural G
2
structure and o has holonomy
Hol(o) 1 ol(3) G
2
.
As long as (`
4
. o
4
) and (`
6
. o
6
) are simply connected then the products
R
3
`
4
and R`
6
are simply connected G
2
-holonomy manifolds with reducible
holonomy groups and parallel Killing spinors. Note that this does not violate The-
orem 14.2.3 as these spaces are not Riemannian cones over complete Riemannian
544 14. SASAKIAN STRUCTURES, KILLING SPINORS, AND SUPERSYMMETRY
manifolds. Using (ii) of Proposition 14.4.3 we obtain the following corollary of
Theorem 14.4.2 rst obtained in [FIMU06]
Corollary 14.4.4: Let (
5
. o) be a Sasaki-Einstein manifold. Then the sine cone
C
s
(
5
) =
5
(0. ) with metric o
s
is nearly Kahler of Einstein constant = 5.
Furthermore o
s
approximates pure ol(3) holonomy metric near the cone points.
Using Corollary 14.4.4 we obtain a host of examples of nearly Kahler 6-manifolds
with conical singularities by choosing
5
to be any of the Sasaki-Einstein manifolds
discussed in detail in Section 11.4. For example, in this way we obtain nearly-Kahler
metrics on (0. ) where is any Smale manifold with a Sasaki-Einstein metric
such as o
5
or /(o
2
o
3
). etc. Note that every simply connected strict nearly Kahler
manifold has exactly two real Killing spinors. So, as long as
5
is simply connected
C
s
(
5
) will have two real Killing spinors. Using Theorem 14.4.2 the examples in
Chapter 11 in other dimensions also give new Einstein metrics on C
s
(
2n+1
). For
example, one obtains many positive Einstein metrics on
2n+1
(0. ) where
2n+1
is any homotopy sphere in /1
2n+2
. Of course, there are no Killing spinors unless
n = 2. Returning to the case of dimension 6, a somewhat more general converse
has been obtained in [FIMU06], namely
Theorem 14.4.5: Any totally geodesic hypersurface
5
of a nearly Kahler 6-
manifold `
6
admits a Sasaki-Einstein structure.
The method in [FIMU06] uses the recently developed notion of hypo ol(2)
structure due to Conti and Salamon [CS06b]. The study of sine cones appears to
have originated in the physics literature [BM03, ADHL03], but in one dimension
higher. Now recall the following result of Joyce (cf. [Joy00], Propositions 13.1.2-3)
Proposition 14.4.6: Let (`
6
. o
6
) and (`
7
. o
7
) be Calabi-Yau and G
2
-holonomy
manifolds, respectively. Let (R
2
. d:
2
= dr
2
+dn
2
) and (R. d:
2
= dr
2
) be Euclidean
spaces. Then
(i) (R
2
`
6
. o = d:
2
+o
6
) has a natural ojin(7) structure and o has holo-
nomy Hol(o) 1l
2
ol(3) ojin(7).
(ii) (R`
7
. o = d:
2
+o
7
) has a natural ojin(7) structure and o has holonomy
Hol(o) 1 G
2
ojin(7).
Again, if (`
6
. o
6
) and (`
7
. o
7
) are simply connected so are the ojin(7)-
manifolds R
2
`
6
and R`
7
so that they have parallel spinors. Not surprisingly,
in view of Lemma 14.4.1 and Proposition 14.4.6, the sine cone construction now re-
lates strict nearly Kahler geometry in dimension 6 to weak holonomy G
2
geometry
in dimension 7. More precisely [BM03]
Theorem 14.4.7: Let (
6
. o) be a strict nearly Kahler 6-manifold such that o
has Einstein constant
6
= 5. Then the manifold C
s
() =
6
(0. ) with its
sine cone metric o
s
has weak holonomy G
2
with Einstein constant
7
= 6 and it
approximates pure G
2
holonomy metric near the cone points.
Proof. Just as before, starting with (
6
. o
6
) we consider its metric cone
C(
6
) with the metric o = dn
2
+ n
2
o
6
and the product metric o
8
on R C(
6
).
With the above choice of the Einstein constant we see that o
8
= dr
2
+ dn
2
+ n
2
o
6
must have holonomy Hol(o
8
) 1 G
2
ojin(7). By Lemma 14.4.1 o
8
is a met-
ric cone on the metric o
7
= dt
2
+ sin
2
to
6
. which must, therefore, have weak G
2
holonomy and the Einstein constant
7
= 6.
14.5. GEOMETRIC STRUCTURES ON MANIFOLDS AND SUPERSYMMETRY 545
Again, any simply connected weak G
2
-manifold has at least one Killing spinor.
That real Killing spinor on C
s
(
6
) will lift to a parallel spinor on C(C
s
(
6
)) =
R C(
6
) which is a non-complete ojin(7)-manifold of holonomy inside 1 G
2
.
One can iterate the two cases by starting with a compact Sasaki-Einstein 5-manifold

5
and construct either the cone on the sine cone of
5
or the sine cone on the
sine cone of
5
to obtain a weak holonomy G
2
manifold. We list the Riemannian
manifolds coming from this construction that are irreducible.
Proposition 14.4.8: Let (
5
. o
5
) be a compact Sasaki-Einstein manifold which is
not of constant curvature. Then the following have irreducible holonomy groups:
(i) the manifold C(
5
) with the metric o
6
= dt
2
+t
2
o
5
has holonomy ol(3);
(ii) the manifold C
s
(
5
) =
5
(0. ) with metric o
6
= dt
2
+ sin
2
t o
5
is
strict nearly Kahler;
(iii) the manifold C
s
(C
s
(
5
)) =
5
(0. ) (0. ) with the metric o
7
=
d
2
+ sin
2
(dt
2
+ sin
2
t o
5
) has weak holonomy G
2
.
In addition we have the reducible cone metrics: C(C
s
(
5
)) = R C(
5
) has
holonomy in 1ol(3) G
2
and C(C
s
(C
s
(
5
))) = RC(C
s
(
5
)) = RRC(
5
)
has holonomy 1l
2
ol(3) 1G
2
ojin(7). If
5
is simply connected then G
5
. o
6
and o
7
admit two Killing spinors. For a generalization involving conformal factors
see [MO07].
Remark 14.4.1: Recall Remark 14.3.2. Note that when a weak holonomy G
2
metric is not complete then the type I-III classication is no longer valid. The
group ojin(7) has other subgroups than the ones listed there and we can consider
the following inclusions of (reducible) holonomies
Spin(7) G
2
1 ol(3) 1l
2
ol(2) 1l
3
1l
8
.
According to the Friedrich-Kath Theorem 14.3.4 the middle three cannot occur as
holonomies of Riemannian cones of complete 7-manifolds with Killing spinors. But
as the discussion of this section shows, they most certainly can occur as holonomy
groups of Riemannian cones of incomplete weak holonomy G
2
metrics. These met-
rics can be still separated into three types depending on the holonomy reduction:
say the ones that come from strict nearly Kahler manifolds are generically of type
1
s
while the ones that come from Sasaki-Einstein 5-manifolds via the iterated sine
cone construction are of type 11
s
and of type 111
s
when H ol(3) is some proper
non-trivial subgroup. On the other hand, it is not clear what is the relation between
the holonomy reduction and the actual number of the Killing spinors one gets in
each case.
14.5. Geometric Structures on Manifolds and Supersymmetry
Supersymmetry has emerged in physics as an attempt to unify the way physical
theories deal with bosonic and fermionic particles. Since its birth around the early
70ties it has come to dominate theoretical high energy physics (for a historical
perspective see [KS00] with the introduction by Kane and Shifman, and for a
mathematical treatment see [Var04]). This dominance is still ongoing in spite of the
fact that after almost 40 years there is no single experimental evidence that would
directly and convincingly prove or discover the existence of supersymmetry in
nature. On the other hand, especially in the last 20 years, supersymmetry has given
546 14. SASAKIAN STRUCTURES, KILLING SPINORS, AND SUPERSYMMETRY
birth to many beautiful mathematical theories. Gromov-Witten Theory, Seiberg-
Witten Theory, Rozansky-Witten Theory as well as the Mirror Duality Conjecture
are just a few of the more famous examples of important and deep mathematics
having its origins in the physics of various supersymmetric theories.
The intricate relationship between supersymmetry and geometric structures
on manifolds was recognized along the way the physics of supersymmetry slowly
evolved from its origins: rst globally supersymmetric eld theories (70ties) arose,
later came supergravity theory (80ties), which evolved into superstring theory and
conformal eld theory (late 80ties and 90ties), and nally into M-Theory and the
supersymmetric branes of today. At every step the rst theory would quickly
lead to various generalizations creating many dierent new ones: so it is as if after
discovering plain vanilla ice cream one would quickly nd oneself in an Italian ice
cream parlor confused and unable to decide which avor was the right choice for the
hot afternoon. This is a confusion that is possibly good for ones sense of taste, but
many physicists believe that there should be just one theory, the Grand Unied
Theory which describes our world at any level.
2
An interesting way out of this
conundrum is to suggest that even if two theories appear to be completely dierent,
if both are consistent and admissible, they actually do describe the same physical
world and, therefore, they should be dual to one another in a certain sense. This
gave rise to various duality conjectures such as the Mirror Symmetry Conjecture
or the AdS/CFT Duality Conjecture.
The rst observation of how supersymmetry can restrict the underlying geom-
etry was due to Zumino [Zum79] who discovered that globally = 1 supersym-
metric -models in d = 4 dimensions require that the bosonic elds (particles) of
the theory are local coordinates on a Kahler manifold. Later Alvarez-Gaume and
Friedman observed that = 2 supersymmetry requires that the -model manifold
be not just Kahler but hyperkahler [AGF81]. This relation between globally su-
persymmetric -models and complex manifolds was used by Lindstrom and Rocek
to discover the hyperkahler quotient construction in [LR83, HKLR87].
The late seventies witnessed a series of attempts to incorporate gravity into the
picture which quickly led to the discovery of various supergravity theories. Again
the = 1 supergravity-matter couplings in d = 4 dimensions require bosonic
matter elds to be coordinates on a Kahler manifold with some special properties
[WB82] while = 2 supergravity demands that the -model manifold be quater-
nionic Kahler [BW83]. The quaternionic underpinnings of the matter couplings
in supergravity theories lead to the discovery of quaternionic Kahler reduction in
[Gal87a, GL88].
At the same time manifolds with Killing spinors emerged as important play-
ers in the physics of the supergravity theory which in 1 = 11 dimensions was
rst predicted by Nahm [Nah78] and later constructed by Cremmer, Julia and
Scherk [CJS79]. The well-know Kaluza-Klein trick applied to a 1 = 11 supergrav-
ity model is a way of constructing various limiting compactications which would
2
Actually, string theory of today appears to oer a rather vast range of vacua (or possible
universes). Such possible predictions have been nicknamed the string landscape [Sus03]. This
fact has been seen as a drawback by some, but not all, physicists (see more recent discussion on
landscape and swampland in [Vaf05, OV06a]). The insistence that the Universe we experience,
and this on such a limited scale at best, is the only Universe, is largely a matter of philosophical
attitude towards science. See the recent book of Leonard Susskind on string theory and cosmic
landscape [Sus05].
14.5. GEOMETRIC STRUCTURES ON MANIFOLDS AND SUPERSYMMETRY 547
better describe the apparently four-dimensional physical world we observe. The
geometry of such a compactication is simply a Carteasian product R
3,1
`
7
,
where R
3,1
is the Minkowski space-time (or some other Lorenzian 4-manifold) and
`
7
is a compact manifold with so small a radius that its presence can be only
felt and observed at the quantum level. Many various models for `
7
were studied
in the late seventies which by the eighties had already accrued into a vast physics
literature (cf. the extensive three-volume monograph by Castellani, DAuria and
Fre [CDF91]). Most of the models assumed a homogeneous space structure on
`
7
= GH (see Chapter V.6 in [CDF91], for examples). Two things were of key
importance in terms of the required physical properties of the compactied theory.
First, the compact space `
7
, as a Riemannian manifold, had to be Einstein of pos-
itive scalar curvature. Second, although one could consider any compact Einstein
space for the compactication, the new theory would no longer be supersymmetric
unless (`
7
. o) admitted Killing spinor elds, and the number of them would be
exactly the number of residual supersymmetries of the compactied theory. For
that reason compactication models involving (o
7
. o
0
) were quite special as they
gave the maximally supersymmetric model. However, early on it was realized that
there are other, even homogeneous, 7-manifolds of interest. The oj(2)-invariant
Jensen metric on o
7
, or as physicists correctly nicknamed it, the squashed 7-sphere
is one of the examples. Indeed, Jensens metric admits exactly one Killing spinor
eld since it has weak holonomy G
2
. Of course, any of the Einstein geometries in
the table of Theorem 14.2.3 can be used to obtain such supersymmetric models.
The 1 = 11 supergravity theory only briey looked liked it was the Grand
Theory of Einsteins dream. It was soon realized that there are diculties with
getting from 1 = 11 supergravity to the standard model. The theory which was
to solve these and other problems was Superstring Theory and later M-Theory
(which is yet to be constructed). With the arrival of superstring theory and M-
theory, supersymmetry continues its truly remarkable inuence on many dierent
areas of mathematics and physics: from geometry to analysis and number theory.
For instance, once again ve, six, and seven-dimensional manifolds admitting real
Killing spinors have become of interest because of the so called AdS/CFT Dual-
ity. Such manifolds have emerged naturally in the context of j-brane solutions
in superstring theory. These so-called j-branes, near the horizon are modelled
by the pseudo-Riemannian geometry of the product adS
p+2
`, where adS
p+2
is the (j + 2)-dimensional anti-de-Sitter space (a Lorentzian version of a space of
constant sectional curvature) and (`. o) is a Riemannian manifold of dimension
d = 1 j 2. Here 1 is the dimension of the original supersymmetric theory. In
the most interesting cases of M2-branes, M5-branes, and D3-branes 1 equals either
11 (Mj-branes of M-theory) or 10 (Dj-branes in type IIA or type IIB string theory).
String theorists are particularly interested in those vacua of the form adS
p+2
`
that preserve some residual supersymmetry. It turns out that this requirement im-
poses constraints on the geometry of the Einstein manifold ` which is forced to
admit real Killing spinors. Depending on the dimension d, the possible geometries
of ` are as follows:
where the notation (j. j). which is common in the physics literature, represents the
ratio of the number of real Killing spinors of type (j. c) to the maximal number
of real Killing spinors that can occur in the given dimension. This maximum is,
of course, realized by the round sphere of that dimension. So this table is just a
548 14. SASAKIAN STRUCTURES, KILLING SPINORS, AND SUPERSYMMETRY
d Geometry of M (j. j)
any round sphere (1. 1)
7 weak holonomy G
2
(
1
8
. 0)
SasakiEinstein (
1
4
. 0)
3-Sasakian (
3
8
. 0)
6 nearly Kahler (
1
8
.
1
8
)
5 SasakiEinstein (
1
4
.
1
4
)
translation of the table of Theorem 14.2.3 for the special dimensions that occur in
the models used by the physicists.
Furthermore, given a j-brane solution of the above type, the interpolation be-
tween adS
p+2
` and R
p,1
((`) leads to a conjectured duality between the
supersymmetric background of the form adS
p+2
` and a (j + 1)-dimensional
superconformal eld theory of n coincident j-branes located at the conical singu-
larity of the R
p,1
((`) vacuum. This is a generalized version of the Maldacena
or AdS/CFT Conjecture [Mal99]. In the case of D3-branes of string theory the
relevant near horizon geometry is that of adS
5
`, where ` is a Sasaki-Einstein 5-
manifold. The D3-brane solution interpolates between adS
5
` and R
3,1
((`),
where the cone ((`) is a Calabi-Yau threefold. In its original version the Malda-
cena conjecture (also known as AdS/CFT duality) states that the t Hooft large n
limit of = 4 supersymmetric Yang-Mills theory with gauge group ol(n) is dual
to type IIB superstring theory on adS
5
o
5
[Mal99]. This conjecture was further
examined by Klebanov and Witten [KW99] for the type IIB theory on adS
5
T
1,1
,
where T
1,1
is the other homogeneous Sasaki-Einstein 5-manifold T
1,1
= o
2
o
3
and the Calabi-Yau 3-fold ((T
1,1
) is simply the quadric cone in C
4
. Using the well-
known fact that ((T
1,1
) is a Kahler quotient of C
4
(or, equivalently, that o
2
o
3
is a Sasaki-Einstein quotient of o
7
), a dual super Yang-Mills theory was proposed,
representing D3-branes at the conical singularities. In the framework of D3-branes
and the AdS/CFT duality the question of what are all the possible near horizon
geometries ` and ((`) might be of importance. Much of the interest in Sasaki-
Einstein manifolds is precisely due to the fact that each such explicit metric, among
other things, provides a useful model to test the AdS/CFT duality. That is why the
Y
p,q
manifolds of Section 11.4.2 attracted such attention (cf. the papers mentioned
in Remarks 11.4.1).
Remark 14.5.1: [G
2
holonomy manifolds unication scale and proton de-
cay] Until quite recently the interest in 7-manifolds with G
2
holonomy as a source
of possible physical models was tempered by the fact the Kaluza-Klein compact-
ications on smooth and complete manifolds of this type lead to models with no
charged particles. All this has dramatically changed in the last few years largely be-
cause of some new developments in M-theory. Perhaps the most compelling reasons
for reconsidering such 7-manifolds was oered by Atiyah and Witten who consid-
ered the dynamics on manifolds with G
2
holonomy which are asymptotically conical
[AW02]. The three models of cones on the homogeneous nearly Kahler manifolds
mentioned earlier are of particular interest, but Atiyah and Witten consider other
cases which include orbifold (quotient) singularities. Among other things they point
14.5. GEOMETRIC STRUCTURES ON MANIFOLDS AND SUPERSYMMETRY 549
to a very interesting connection between Kronheimers quotient construction of the
ALE metrics and asymptotically conical manifolds with G
2
-holonomy. To explain
the connection, consider Kronheimers construction for = Z
n+1
described in Sec-
tion 12.10. Suppose one chooses a circle o
1
k,l
U(1) 1(Z
n+1
) = U(1)
n
and then
one considers a 7-manifold obtained by performing Kronheimers HK quotient con-
struction with zero momentum level ( = 0) while forgetting the three moment
map equations corresponding to this particular circle. An equivalent way of look-
ing at this situation is to take the Kronheimer quotient with nonzero momentum
= o sp(1) but only for the moment map of the chosen circle o
1
k,l
(such is never
in the good set) and then consider the bration of singular Kronheimer quotients
over a 3-dimensional base parameter space. Algebraically this corresponds to a
partial resolution of the quotient singularity and this resolution depends on the
choice of o
1
k,l
, hence . This example was rst introduced in [AW02]. It can be
shown that the 7-manifold is actually a cone on the complex weighted projective
3-space with weights (/. /. |. |). where /+| = n+1. It then follows from the physical
model considered that such a cone should admit a metric with G
2
holonomy. How-
ever, unlike the homogeneous cones over the four homogeneous strict nearly Kahler
manifolds of Theorem 14.3.12, the metric in this case is not known explicitly. This
construction appears to dier from all previous geometric constructions of metrics
with G
2
holonomy. One can consider similar constructions for other choices of
o
1
1() [BB02].
In [FW03] using a specic models of M-theory compactications on mani-
folds with G
2
holonomy, Friedman and Witten address the fundamental questions
concerning the unication scale (i.e., the scale at which the Standard Model of
ol3) ol(2) l(1) unies in a single gauge group) and proton decay. The au-
thors point out that the results obtained are model dependent, but some of the
calculations and conclusions apply to a variety of dierent models.
Appendix A
A.1. Preliminaries on Groupoids
We briey review some properties of groupoids [CdSW99, Mac87, Moe02,
MM03, MS88, ALR06]. Here we follow [Moe02] closely.
Denition A.1.1: A groupoid ( is a small category in which every morphism
has an inverse.
So a groupoid ( consists of a set G
0
of objects and a set G
1
of arrows or mor-
phisms together with certain structure maps. For any groupoid there are canonical
maps : : G
1
G
0
and t : G
1
G
0
. called the source map and target map (some-
times the latter is called the range map), respectively, and dened by sending each
arrow o : rn to its source r and target n, respectively, i.e., :(o) = r. t(o) = n.
Next there is an associative multiplication, called multiplication or composition,
dened for any two arrows o. / G
1
such that t(o) = :(/) with :(/ o) = :(o) and
t(/ o) = t(/). Alternatively, if o : rn and / : n. are arrows from r to n,
and from n to .. respectively, then / o : r. is an arrow from r to .. Since,
composition is dened only when the target of the rst arrow equals the source of
the second, we consider the bered product G
1

G0
G
1
dened by
(A.1.1) G
1

G0
G
1
= (/. o) G
1
G
1
[ :(/) = t(o) .
Then composition denes a map : : G
1

G
0
G
1
G
1
by :(/. o) = / o. For each
object r G
0
there is an unit or identity arrow 1
x
G
1
dened by sending r to
itself, i.e., 1
x
r = r. The identity is a two-sided unit for composition, i.e., o 1
x
= o
and 1
x
/ = / for any o. / G
1
with :(o) = r and t(/) = r. The unit arrows dene
a map n : G
0
G
1
by n(r) = 1
x
.
Finally for each arrow o : rn in G
1
there is an inverse arrow o
1
: nr
which is a two-sided inverse for composition, i.e., o o
1
= 1
y
and o
1
o = 1
x
.
The inverse denes an involution : G
1
G
1
by (o) = o
1
.
Thus, a groupoid ( = (G
0
. G
1
. :. t. :. n. ) consists of two sets G
0
and G
1
together with ve structure maps :. t. :. n. . It is often convenient to think of G
0
as a base space and refer to ( as a groupoid over G
0
, written as G
1
= G
0
. where
the double arrow indicates the two maps, the source and target. The composition
o / in ( is often written as juxtaposition o/.
For any groupoid ( we have the following subsets of interest:
(
x
= o G
1
[ :(o) = r. (
y
= o G
1
[ t(o) = n .
(A.1.2)
(
y
x
= o G
1
[ :(o) = r. :(o) = n. (
W
Z
= o G
1
[ :(o) 7. t(o) \ .
551
552 APPENDIX A
Here 7 and \ are subsets of G
0
. Notice from this point of view that the composition
map : becomes a map on the subsets (
y
x
(
z
y
(
z
x
and that the subset (
x
x
G
1
is a group which acts on (
y
x
from the left. Morever, the group (
y
y
acts on (
y
x
from
the right. The group (
x
x
of arrows from r to itself is called the isotropy group at r
and often denoted by G
x
.
Denition A.1.2: A subgroupoid H of ( is dened by a pair of subsets H
0

G
0
and H
1
G
1
such that :(H
1
) H
0
. t(H
1
) H
0
. and H
1
is closed under
multiplication when it is dened and under inversion. If H
0
= G
0
then H is called
a wide subgroupoid. If for all r. n H
0
. H
y
x
= (
y
x
then H is said to be a full
subgroupoid.
Notice that the fact that H is closed under inversion implies that n(r) = 1
x
is
in H
1
for all r H
0
. i.e., the restriction of n to H
0
denes a map from H
0
to H
1
.
Next we give some simple examples of groupoids.
Example A.1.3: Some simple examples of groupoids are:
(i) The identity or trivial or unit groupoid: All arrows are units, so G
1
= G
0
as well. Any subset of this groupoid is a subgroupoid.
(ii) A group G is a groupoid with G
0
= pt. The subgroupoids of G are just
the subgroups of G and the empty set.
(iii) The pair groupoid { : G
0
= A is any set, and G
1
= A A with exactly
one arrow between any two points r. n G
0
. Then a subgroupoid of { is
a relation on A which is symmetric and transitive. A wide subgroupoid
is an equivalence relation on A.
The groupoids that we are interested in have a richer structure. In fact they
are smooth manifolds. This leads to:
Denition A.1.4: A topological groupoid is a groupoid ( in which both G
1
and
G
0
are topological spaces, and the structure maps are continuous. A Lie groupoid
is a topological groupoid for which G
1
is a smooth, but not necessarily Hausdor,
manifold, G
0
is a smooth Hausdor manifold, the above structure maps are dier-
entiable, and both the source and target maps are submersions.
For a complete treatment of Lie groupoids, see [Mac87]. In almost all cases
of interest to us, the groupoid ( will be a Lie groupoid, and G
1
will be Hausdor.
By abuse of notation we sometimes refer to the manifold G
1
as the groupoid (
and by the dimension of ( we mean the dimension of G
1
. For any Lie groupoid (
the subsets dened by A.1.2 above are smooth manifolds, and G
x
= (
x
x
is a Lie
group. Lie groupoids exist in abundance. The simple examples in Example A.1.3
have their Lie groupoid analogues.
(i) G
0
= ` is any manifold and all arrows are units, so G
1
= ` as well. (
is called the unit Lie groupoid.
(ii) A Lie group is a Lie groupoid with G
0
= pt.
(iii) G
0
= ` is any manifold, and G
1
= ` ` with exactly one arrow
between any two points r. n G
0
. This is called the pair Lie groupoid
and sometimes it is denoted by ` `.
There are two other Lie groupoids that are of much interest to us. The rst
is the holonomy groupoid ((`. T) of a foliation T that is discussed in detail in
Chapter 1. The second is the so-called action or translation groupoid.
A.1. PRELIMINARIES ON GROUPOIDS 553
Denition A.1.5: Let 1 be a Lie group acting smoothly on the manifold ` from
the right
3
. We dene a Lie groupoid ` 1 called the action or translation
groupoid as follows: the space of objects (` 1)
0
is ` and the space of arrows
(` 1)
1
is ` 1. An arrow (r. /) (` 1)
1
= ` 1 from r ` to
n ` is given by the right action n = r/. The source map is : : ` 1` is
projection onto the rst factor, while the target map t : ` 1` is dened by
t(r. /) = r/. Composition is dened by (r. /) (n. /
t
) = (r. //
t
) when n = r/. The
unit is n(r) = (r. c). where c is the identity in 1.
Next we give several denitions that are pertinent to understanding orbifolds
from the groupoid point of view [ALR06, Moe02].
Denition A.1.6: Let ( be a Lie (topological) groupoid.
(i) ( is called an etale groupoid if the source map : is a local dieomor-
phism (homeomorphism), and it is said to be eective if the isotropy
groups (
x
x
act eectively on (
0
;
(ii) ( is said to be proper if G
1
is Hausdor and the achor map (:. t) :
G
1
G
0
G
0
is proper;
(iii) ( is called a foliation groupoid if every isotropy group (
x
x
is discrete;
(iv) an orbifold groupoid is a proper etale Lie groupoid.
Note that the fact that the source map : is a local dieomorphism (home-
omorphism) implies that the other structure maps are dieomorphisms (homeo-
morphisms). These denitions are enough to capture the essence of an orbifold as
seen in Section 4.3. (For our orbifolds we should add that the isotropy groups act
eectively).
Following [Moe02] we now describe the important notion of equivalence of
groupoids. Let ( and H be groupoids. A homomorphism : (H of groupoids is
a pair
0
: G
0
H
0
and
1
: G
1
H
1
of maps that commute with all the structure
maps, e.g.,
0
: = :
1
. ctc. If ( and H are Lie (topological) groupoids, we demand
that the maps are smooth (continuous). In particular, is an isomorphism of Lie
(topological) groupoids if the maps
0
.
1
are dieomorphisms (homeomorphisms).
We call such groupoids strongly equivalent. For groupoids it is important to have a
weaker notion of equivalence. First we need
Denition A.1.7: Let : H( and : /( be homomorphisms of Lie
(topological) groupoids. The bered product H
(
/ is the Lie (topological)
groupoid whose objects are triples (n. o. .). where n H
0
. . 1
0
. and o : (n)(.)
is in G
1
. The morphisms (n. o. .)(n
t
. o
t
. .
t
) of H
(
/ are pairs (/. /) of mor-
phisms / : n n
t
in H
1
and / : . .
t
in 1
1
such that o
t
(/) = (/)o. This can
be represented by the following diagram:
n (n)
g
(.) .

(h)

(k)

k
n
t
(n
t
)
g

(.
t
) .
t
Composition in H
(
/ is dened by extending he diagram in the obvious way.
3
It is more common to use the left action, but we choose the action to be from the right in
order to follow the standard convention for principal bundles.
554 APPENDIX A
Denition A.1.8: A homomorphism of groupoids : (H is called an essen-
tial or weak equivalence if:
(i) the map t
1
: H
1

H0
G
0
H
0
is a surjective submersion.
(ii) the square
G
1

H
1

(s,t)

(s,t)
G
0
G
0

H
0
H
0
is a bered product.
Two such groupoids are said to be Morita equivalent if there exists a third
groupoid / such that ( / H are weak equivalences.
Condition (i) means that every object n H
0
is connected by a morphism
(/ : (r)n) H
1
to an object in the image of . Condition (ii) implies that
induces a dieomorphism G
1
(r. n)H
1
((r). (n)) from the subspace of all
morphisms from r to n in G
1
to the subspace of all morphisms (r) to (n) in H
1
.
One can check that Morita equivalence denes an equivalence relation. Although
Morita equivalent Lie groupoids can be very dierent, for example they can have
dierent dimensions and dierent connectivity properties, Morita equivalence does
preserve some essential geometric features as we shall see. For us the relationship
between etale Lie groupoids and orbifolds is crucial.
Next we describe the concept of an action of a topological groupoid on an
appropriate space [Mac87, CdSW99].
Denition A.1.9: Let ( be a topological groupoid over A G
0
. and let j : Y A
be a continuous map. Dene the space
G
1

X
Y = (o. n) G
1
Y [ :(o) = j(n) .
Then the action of ( on Y is dened by a map / : G
1

X
Y Y. satisfying
(i) j(/(o. n)) = t(o).
(ii) /(o /. n) = /(o. /(/. n)).
(iii) /(n(j(n)). n) = n.
The orbit of ( through n Y is the subset ((n) = /(o. n) [ o (
(y)
Y.
The map j is often called the moment map in analogy with the moment map of
symplectic geometry. The action map /(o. n). thought of as a left action, is often
written as o n. There are two natural actions of a groupoid on itself. The rst
is an action of ( on its set of arrows, G
1
. so we take Y = G
1
. j = t. and dene
the action by / : G
1

X
G
1
G
1
as /(o. /) = o / which is well dened since
:(o) = t(/). The second is an action on the set of objects, so we take Y = G
0
and
j = id. The action map is now dened by /(o. r) = t(o) which makes sense since
the source of o is r. Notice that the rst action is always free, while the second
is generally not free. A right action is dened in a completely analogous way. An
important example in which the convention is to use the right action is [Hae84]:
Denition A.1.10: A principal (-bundle over Y is a topological space 1 to-
gether with a continuous surjection : 1Y , and a continuous action / of (
on 1 with moment map j : 1A such that (/(o. c)) = (c) for c 1. The
A.2. THE CLASSIFYING SPACE OF A TOPOLOGICAL GROUPOID 555
action / is transitive on the bres of . Furthermore, each point n Y has an
open neighborhood l and a continuous section : l1 of such that the lo-
cal triviality condition holds, that is the map / : G
1

X
l
1
(l) dened by
/(o. n) = /(o. (n)) is a homeomorphism.
A.2. The Classifying Space of a Topological Groupoid
Classifying spaces are important in mathematics as spaces which capture all
objects with a particular behavior. In particular, we are interested in constructing
universal spaces for topological groupoids. Classifying spaces were rst constructed
for topological groups. Good references here are the book of Husemoller [Hus66],
and the papers of Milnor [Mil56a] and Dold [Dol63]. The classifying space theory
for groupoids was developed by Haeiger [Hae62, Hae71] and independently by
Buet and Lor [BL70]. In [Hae62] Haeiger approaches the subject through
E. Browns theory of representable homotopy functors, whereas, in [Hae71] and
[BL70] the authors generalize Milnors beautiful join construction of classifying
space for topological groups [Mil56a] to groupoids. It is this latter approach that
we present here. Another suitable approach to classifying spaces for groupoids, or
more generally to categories is Segals construction [Seg68, Moe02, ALR06] of
the nerve of a category. Segal shows this to be equivalent to the Milnor construction.
For any topological groupoid ( we consider the superdiagonal
(A.2.1) o1
n+1
(() = (r
0
. . . . . r
n
) G
n+1
1
[ t(r
i
) = t(r
j
) for all i. , = 0. . . . n
together with the standard n-simplex
n
with barycentric coordinates (t
0
. . . . . t
n
).
i.e., t
i
[0. 1] and

n
i=0
t
i
= 1. Write a point of o1
n+1
(()
n
as a 2n +2-tuple
(t
0
r
0
. t
1
r
1
. . . . . t
n
r
n
). We denote the n + 1-fold join by
(A.2.2) 1((n) = (
n+1 times
( .
Then as a set 1((n) is the identication space o1
n+1
(()
n
. where
(t
0
r
0
. t
1
r
1
. . . . . t
n
r
n
) (t
t
0
r
t
0
. t
t
1
r
t
1
. . . . . t
t
n
r
t
n
)
if t
i
= t
t
i
for all i and if t
i
0 then r
i
= r
t
i
. A point of 1((n) is denoted by
't. x` = 't
0
r
0
. . . . . t
n
r
n
`. On 1((n) there is a natural map which by abuse of
notation is also denoted by t
i
: 1((n)[0. 1]. It is dened to be the projection
map
(A.2.3) 't
0
r
0
. . . . . t
n
r
n
` t
i
.
On each inverse image t
1
i
((0. 1]) there are also natural projection maps r
i
dened
by
(A.2.4) 't
0
r
0
. . . . . t
n
r
n
` r
i
.
Notice that if any of the coordinates t
i
vanish then r
i
is not dened as a map. Now
the topology on 1((n) is dened to be the smallest topology such that the maps
t
i
and r
i
are continous. There are also inclusion maps : 1((n) 1((n + 1)
dened by
(A.2.5) ('t
0
r
0
. . . . . t
n
r
n
`) = 't
0
r
0
. . . . . t
n
r
n
. 0r
n+1
` .
Notice that this is independent of the choice of r
n+1
. The inclusions give rise to a
directed system and the direct limit
(A.2.6) 1( = lim

1((n)
556 APPENDIX A
with the direct limit topology. Notice also that we have not used all of the properties
of a topological groupoid. Indeed, we have only used the fact that ( consists of
topological spaces G
1
and G
0
together with a continuous map t : G
1
G
0
.
Now we dene an action of the topological groupoid ( on 1(. The moment
map j : 1((n)G
0
is dened by j('t. x`) = t(r
i
) which is a well dened point
of G
0
. since 1((n) is formed from the superdiagonal o1
n+1
((). The action / on
1((n) is that induced by the left action of ( on G
1
. namely,
/(o. 't. x`) = 't. ox` .
One easily checks that this satises the conditions of Denition A.1.9. The inclusion
maps are equivariant with respect to this action giving an action on 1(. It is also
clear since not all t
i
can vanish that this action is free on both 1((n) and 1(.
The quotient spaces are denoted by 1((n) and 1(. respectively, and the natural
projections
n
: 1((n)1((n) and : 1(1( are principal (-bundles. In
analogy with the case of principal bundles, the space 1( is called the classifying
space of principal (-structures which we now describe.
Denition A.2.1: Let 1 be a topological space and l
i

i1
an open cover of 1.
Then the cover l
i
is said to be numerable if there exists a locally nite partition
of unity n
i

i1
such that the closure of n
1
i
((0. 1]) is contained in l
i
for each i 1.
A principal (-bundle : 11 is said to be numerable if 1 has a numerable
cover l
i

i1
such that
1
(l
i
) is locally trivial for each i 1. A numerable
principal (-bundle : 11 is n-universal if for each C\-complex A with
dim n the pullback operation 1 1

1 induces a one-to-one correspondence

between homotopy classes of maps 1 : A1 and isomorphism classes of principal
(-bundles over A. In particular, 1 is called universal if it is n-universal for all n.
It is a standard result that a Hausdor space A is paracompact if and only if
every open cover is numerable. Moreover, a result of Miyazaki says that any C\-
complex is paracompact. So for any C\-complex every open cover is numerable.
It is easy to see that under any continuous map numerable bundles pullback to
numerable bundles. Moreover, pulling back numerable bundles by homotopic maps
gives isomorphic bundles. We now have the generalization to groupoids of Milnors
well-known result. It is due to Buet-Lor [BL70] and Haeiger [Hae71].
Theorem A.2.2: The bundle : 1(1( is a numerable universal principal
(-bundle.
So 1( classies principal (-bundles as well as Haeiger cocycles, but it is only
the Morita equivalence class of ( that is relevant.
Theorem A.2.3: Let ( and H be Morita equivalent topological groupoids, then
there classifying spaces 1( and 1H are weakly homotopy equivalent.
When ( is a connected topological group G, we recover the classical universal
principal G-bundle of Milnor [Mil56a] 1G1G. See also [Ste51] and Appen-
dix B of [LM89]. In particular, if G is a Lie group it suces to consider compact
Lie groups since any Lie groups is homotopy equivalent to its maximal compact
Lie group by the Iwasawa decomposition. Then isomorphism classes of principal
G bundles on a smooth manifold ` are in one-to-one correspondence with ho-
motopy classes [`. 1G] of maps c : `1G. Elements of the cohomology ring
H

(1G. ) for a ring are called universal characteristic classes. One obtains
A.2. THE CLASSIFYING SPACE OF A TOPOLOGICAL GROUPOID 557
characteristic classes on ` by pulling back the universal characteristic classes by
a classifying map c. We have the following important examples:
(i) The (universal) Stiefel-Whitney classes n
i
are the generators of the coho-
mology ring H

(1O(n). Z
2
), i.e., H

(1O(n). Z
2
) = Z
2
[n
1
. n
2
. . . . . n
n
].
(ii) H

(1oO(n). Z
2
) = Z
2
[n
2
. . . . . n
n
].
(iii) H

(1l(n). Z) = Z[c
1
. . . . . c
n
]. where c
i
H
2i
(1l(n). Z) is the i
th
Chern
class.
(iv) H

(1ol(n). Z) = Z[c
2
. . . . . c
n
].
(v) H

(1O(n). ) = [j
1
. . . . . j
|
n
2
|
]. where is an integral domain contain-
ing
1
2
such as Z[
1
2
] or Q. When = Q, j
i
H
4i
(1O(n). ) is called the
i
th
rational Pontrjagin class.
(vi) H

(1O(n). Z) = Z[j
1
. . . . . j
|
n
2
|
] Image(). where is the Bockstein
homomorphism : H
i
(1O(n). Z
2
)H
i+1
(1O(n). Z) induced by the
exact sequence 0Z
2
ZZ
2
0; here j
i
H
4i
(1O(n). Z) is
the i
th
integral Pontrjagin class.
(vii) H

(1oO(2n + 1). ) = [j
1
. . . . . j
n
]. where is an integral domain
containing
1
2
.
(viii) H

(1oO(2n). ) = [j
1
. . . . . j
n
. c](c
2
j
n
). where is an integral
domain containing
1
2
; here c H
2n
(1oO(2n). ) is called the Euler
class.
(ix) H

(1oj(n). Z) = Z[:
1
. . . . . :
n
]. where :
i
H
4i
(1oj(n). Z) is the image
of c
2i
under the natural map 1oj(n)1l(2n).
The corresponding characteristic classes of a vector bundle 1 on a manifold ` are
obtained by pulling back these universal classes by the classifying map c. When
1 = T`. the tangent bundle of `, the characteristic classes are invariants of the
manifold with certain additional structure and we write n
i
(`). c
i
(`), etc. For
a real vector bundle 1 it is common to dene the integral Pontrjagin classes by
j
i
(1) = (1)
i
c
2i
(1 C). Also of some interest are the integral Stiefel Whitney
classes \
i
dened to be the image of the ordinary Stiefel Whitney classes under
the Bockstein homomorphism, that is \
i
= (n
i1
). Since the kernel of is the
set of classes in H

(1O(n). Z
2
) that are the mod 2 reductions of integral classes,
\
i
vanishes if and only if n
i1
is the mod 2 reduction of an integral class.
Appendix B
B.1. Reids List of K3 Surfaces as hypersurfaces in CP
4
(w)
b
2
(X) X
22 X
4
P(1, 1, 1, 1), X
6
P(1, 1, 1, 3)
21 X
5
P(1, 1, 1, 2), X
12
P(1, 1, 4, 6)
20 X
8
P(1, 1, 2, 4), X
10
P(1, 1, 3, 5)
19 X
6
P(1, 1, 2, 2), X
7
P(1, 1, 2, 3), X
9
P(1, 1, 3, 4)
17 X
10
P(1, 2, 2, 5), X
18
P(1, 2, 6, 9)
16 X
8
P(1, 2, 2, 3), X
12
P(1, 2, 3, 6), X
14
P(1, 2, 4, 7),
X
16
P(1, 2, 5, 8)
15 X
9
P(1, 2, 3, 3), X
10
P(1, 2, 3, 4), X
11
P(1, 2, 3, 5),
X
12
P(1, 2, 4, 5), X
15
P(1, 2, 5, 7), X
24
P(1, 3, 8, 12)
14 X
16
P(1, 3, 4, 8), X
18
P(1, 3, 5, 9), X
22
P(1, 3, 7, 11)
X
30
P(1, 4, 10, 15)
X
12
P(1, 3, 4, 4), X
13
P(1, 3, 4, 5), X
14
P(2, 2, 3, 7)
13 X
15
P(1, 3, 4, 7), X
15
P(1, 3, 5, 6), X
20
P(1, 4, 5, 10)
X
21
P(1, 3, 7, 10), X
22
P(1, 4, 6, 11), X
28
P(1, 4, 9, 14)
X
36
P(1, 5, 12, 18), X
42
P(1, 6, 14, 21)
12 X
12
P(2, 2, 3, 5), X
16
P(1, 4, 5, 6), X
18
P(1, 4, 6, 7)
X
26
P(1, 5, 7, 13), X
30
P(1, 6, 8, 15),
11 X
12
P(2, 3, 3, 4), X
18
P(2, 3, 4, 9), X
21
P(1, 5, 7, 8),
X
24
P(1, 6, 8, 9), X
30
P(2, 3, 10, 15),
10 X
14
P(2, 3, 4, 5), X
16
P(2, 3, 4, 7), X
20
P(2, 3, 5, 10)
X
22
P(2, 4, 5, 11), X
24
P(2, 3, 7, 12), X
26
P(2, 3, 8, 13),
X
15
P(2, 3, 5, 5), X
15
P(3, 3, 4, 5), X
17
P(2, 3, 5, 7),
9 X
18
P(2, 3, 5, 8), X
20
P(2, 4, 5, 9), X
20
P(2, 4, 5, 9)
X
21
P(2, 3, 7, 9), X
24
P(2, 3, 8, 11), X
26
P(2, 5, 6, 13)
8 X
18
P(3, 4, 5, 6), X
20
P(2, 5, 6, 7), X
24
P(3, 4, 5, 12)
X
32
P(2, 5, 9, 16), X
42
P(3, 4, 14, 21),
X
19
P(3, 4, 5, 7), X
20
P(3, 4, 5, 8), X
21
P(3, 5, 6, 7)
7 X
27
P(2, 5, 9, 11), X
28
P(3, 4, 7, 14), X
30
P(4, 5, 6, 15)
X
34
P(3, 4, 10, 17), X
36
P(3, 4, 11, 18), X
48
P(3, 5, 16, 24)
X
24
P(3, 4, 7, 10), X
24
P(4, 5, 6, 9), X
30
P(3, 4, 10, 13)
6 X
32
P(4, 5, 7, 16), X
34
P(4, 6, 7, 15, 17), X
38
P(3, 5, 11, 19)
X
54
P(4, 5, 18, 27)
X
25
P(4, 5, 7, 9), X
27
P(5, 6, 7, 9), X
28
P(4, 6, 7, 11)
5 X
33
P(3, 5, 11, 14), X
38
P(5, 6, 8, 19), X
40
P(5, 7, 8, 20)
X
42
P(2, 5, 14, 21), X
44
P(4, 5, 13, 22), X
66
P(5, 6, 22, 33)
4 X
24
P(3, 6, 7, 8), X
30
P(5, 6, 8, 11), X
36
P(7, 8, 9, 12)
X
50
P(7, 8, 10, 25),
559
560 APPENDIX B
B.2. Dierential topology of 2/(o
3
o
4
) and 2/(o
5
o
6
)
Table B.2.1: 2/#(o
3
o
4
)
/
k
1
2
(/)
D
2
(k)
]bP8]
1 1 28 1
2 3 28 1
3 6 14
1
2
4 10 14
1
2
5 15 28 1
6 21 4
1
7
7 28 1
1
28
8 36 7
1
4
9 45 28 1
10 55 28 1
20 210 2
1
14
48 1176 1
1
28
50 1275 28 1
100 5050 14
1
2
496 123256 1
1
28
500 125250 14
1
2
Table B.2.2: 2/#(o
5
o
6
)
/
k
1
3
(/)
D
3
(k)
]bP
12
]
1 1 992 1
2 3 992 1
3 6 496
1
2
4 10 496
1
2
5 15 992 1
6 21 992 1
7 28 248
1
4
8 36 248
1
4
9 45 992 1
10 55 992 1
31 496 2
1
496
48 1176 124
1
8
50 1275 992 1
62 1953 32
1
31
124 7750 16
1
62
248 30876 8
1
124
496 123256 4
1
248
500 125250 496
1
2
992 492528 2
1
496
B.3. TABLES OF K

AHLER-EINSTEIN METRICS ON HYPERSURFACES 561

B.3. Tables of Kahler-Einstein metrics on hypersurfaces CP(w)
Table B.3.1 lists the innite series solutions appearing in Theorem 5.4.16, while
Table B.3.2 lists the sporadic ones. The last column of the tables indicate whether
the klt condition (which implies that Z
w
admits a Kahler-Einstein metric) holds
(Y) or is unknown (?). The computer search gives the following complete list:
Table B.3.1: Innite Series Examples of Z
w
of Index 1 1 10
1 w d /
2
K-E
1 (2. 2/ + 1. 2/ + 1. 4/ + 1) 8/ + 4 8 Y
2 (4. 2/ + 1. 2/ + 1. 4/) 8/ + 4 7 Y
2 (3. 3/ + 1. 6/ + 1. 9/ + 3) 18/ + 6 6 Y
2 (3. 3/ + 1. 6/ + 1. 9/) 18/ + 3 5 ?
2 (3. 3/. 3/ + 1. 3/ + 1) 9/ + 3 7 Y
2 (3. 3/ + 1. 3/ + 2. 3/ + 2) 9/ + 6 5 Y
2 (4. 2/ + 1. 4/ + 2. 6/ + 1) 12/ + 6 6 ?
4 (6. 6/ + 3. 6/ + 5. 6/ + 5) 18/ + 15 5 Y
4 (6. 6/ + 5. 12/ + 8. 18/ + 9) 36/ + 24 3 ?
4 (6. 6/ + 5. 12/ + 8. 18/ + 15) 36/ + 30 4 Y
6 (8. 4/ + 1. 4/ + 3. 4/ + 5) 12/ + 11 3 ?
6 (9. 3/ + 2. 3/ + 5. 6/ + 1) 12/ + 11 3 ?
562 APPENDIX B
Table B.3.2: Sporadic Examples of Z
w
of Index 1 1 10
I w Monomials of 1
w
d /
2
K-E
1 (1,2,3,5) .
10
0
. .
5
1
. .
3
2
.
1
. .
2
3
. . . . (17)

10 9 Y
1 (1,3,5,7) .
15
0
. .
5
1
. .
3
2
. .
2
3
.
0
. . . . (19) 15 9 Y
1 (1,3,5,8) .
16
0
. .
5
1
.
0
. .
3
2
.
0
. .
2
3
. . . . (20) 16 10 Y
1 (2,3,5,9) .
9
0
. .
6
1
. .
3
2
.
1
. .
2
3
. . . . (13) 18 7 Y
1 (3,3,5,5) o
5
(.
0
. .
1
). 1
3
(.
2
. .
3
) 15 5 Y
1 (3,5,7,11) .
6
0
.
2
. .
5
1
. .
2
2
.
3
. .
2
3
.
0
. . . . (8) 25 5 Y
1 (3,5,7,14) .
7
0
.
2
. .
5
1
.
0
. o
2
(.
2
2
. .
3
). . . . (9) 28 6 Y
1 (3,5,11,18) o
2
(.
6
0
. .
3
). .
5
1
.
2
. .
3
2
.
0
. . . . (10) 36 6 Y
1 (5,14,17,21) .
7
0
.
3
. .
4
1
. .
3
2
.
0
. .
2
3
.
1
. .
5
0
.
1
.
2
56 4 Y
1 (5,19,27,31) .
10
0
.
3
. .
4
1
.
0
. .
3
2
. .
2
3
.
1
. .
7
0
.
1
.
2
81 3 Y
1 (5,19,27,50) o
2
(.
10
0
. .
3
). .
5
1
.
0
. .
3
2
.
1
. .
7
0
.
2
1
.
3
100 4 Y
1 (7,11,27,37) .
10
0
.
1
. .
4
1
.
3
. .
3
2
. .
3
3
.
0
81 3 Y
1 (7,11,27,44) o
2
(.
4
1
. .
3
). .
11
0
.
1
. .
3
2
.
0
. .
4
0
.
3
1
.
2
88 4 Y
1 (9,15,17,20) .
5
0
.
1
. .
4
1
. .
3
2
.
0
. .
3
3
60 3 Y
1 (9,15,23,23) .
6
0
.
1
. .
4
1
.
0
. .
3
2
. .
2
2
.
3
. .
2
.
2
3
. .
3
3
69 5 Y
1 (11,29,39,49) .
8
0
.
2
. .
4
1
.
0
. .
3
2
. .
2
3
.
1
127 3 Y
1 (11. 49. 69. 128) .
17
0
.
2
. .
5
1
.
0
. .
4
2
. .
2
2
.
3
. .
2
3
256 2 Y
1 (13,23,35,57) .
8
0
.
1
. .
4
1
.
2
. .
2
2
.
3
. .
2
3
.
0
127 3 Y
1 (13. 35. 81. 128) .
17
0
.
1
. .
5
1
.
2
. .
3
2
.
0
. .
2
3
256 2 Y
2 (2,3,4,5) .
6
0
. .
4
1
. .
3
2
. .
2
3
.
0
. . . . (10) 12 5 ?
2 (2,3,4,7) .
7
0
. .
4
1
.
0
. .
3
2
.
0
. .
2
3
. . . . (11) 14 6 ?
2 (3,4,5,10) .
5
0
.
2
. .
5
1
. .
4
2
. .
2
3
. . . . (9) 20 5 Y
2 (3,4,6,7) o
3
(.
2
0
. .
2
). .
3
1
.
2
. .
2
3
.
1
. . . . (8) 18 6 ?
2 (3,4,10,15) .
10
0
. .
5
1
.
3
. .
3
2
. .
2
3
. . . . (10) 30 7 Y
2 (3,7,8,13) .
7
0
.
2
. .
3
1
.
2
. .
2
2
.
3
. .
2
3
.
0
. . . . (7) 29 5 ?
2 (3,10,11,19) .
10
0
.
3
. .
3
1
.
2
. .
2
2
.
3
. .
2
3
.
0
. . . . (7) 41 5 ?
2 (5,13,19,22) .
7
0
.
3
. .
4
1
.
0
. .
3
2
. .
2
3
.
1
. .
5
0
.
1
.
2
57 3 Y
2 (5,13,19,35) o
2
(.
7
0
. .
3
). .
5
1
.
0
. .
3
2
.
1
. .
5
0
.
2
1
.
2
70 3 Y
2 (6,9,10,13) .
6
0
. .
4
1
. .
3
2
.
0
. .
2
3
.
2
. .
3
0
.
2
1
36 4 Y
2 (7,8,19,25) .
7
0
.
1
. .
4
1
.
3
. .
3
2
. .
2
3
.
0
. .
2
0
.
3
1
.
2
57 3 Y
2 (7,8,19,32) .
8
0
.
1
. o
2
(.
4
1
. .
3
). .
3
2
.
0
. . . . (7) 64 4 Y
2 (9,12,13,16) .
4
0
.
1
. .
4
1
. .
3
2
.
0
. .
3
3
48 3 Y
2 (9,12,19,19) .
5
0
.
1
. .
4
1
.
0
. .
3
2
. .
2
2
.
3
. .
2
.
3
3
. .
3
3
57 5 Y
2 (9,19,24,31) .
9
0
. .
3
1
.
2
. .
3
2
.
0
. .
2
3
.
1
81 3 Y
2 (10,19,35,43) .
7
0
.
2
. .
5
1
.
0
. .
3
2
. .
2
3
.
1
105 3 Y
2 (11,21,28,47) .
7
0
.
2
. .
5
1
. .
3
2
.
1
. .
2
3
.
0
105 3 Y
2 (11,25,32,41) .
6
0
.
3
. .
3
1
.
2
. .
3
2
.
0
. .
2
3
.
1
107 3 Y
2 (11,25,34,43) .
10
0
. .
4
1
.
0
. .
2
2
.
3
. .
2
3
.
1
111 3 Y
2 (11. 43. 61. 113) .
15
0
.
2
. .
5
1
.
0
. .
3
2
.
1
. .
2
3
226 2 Y
2 (13,18,45,61) .
9
0
.
1
. .
5
1
.
2
. .
3
2
. .
2
3
.
0
135 3 Y
B.3. TABLES OF K

AHLER-EINSTEIN METRICS ON HYPERSURFACES 563

Table B.3.2: (cont.) Sporadic Examples of Z
w
of Index 1 1 10
Index w Monomials of 1
w
d /
2
K-E
2 (13,20,29,47) .
6
0
.
3
. .
3
1
.
3
. .
3
2
.
1
. .
2
3
.
0
107 3 Y
2 (13,20,31,49) .
7
0
.
1
. .
4
1
.
2
. .
3
2
.
3
. .
2
3
.
0
111 3 Y
2 (13,31,71,113) .
15
0
.
1
. .
5
1
.
2
. .
3
2
.
0
. .
2
3
226 2 Y
2 (14,17,29,41) .
5
0
.
3
. .
5
1
.
0
. .
2
2
.
3
. .
2
3
.
1
99 3 Y
3 (5,7,11,13) .
4
0
.
3
. .
4
1
.
0
. .
3
2
. .
2
3
.
1
. .
3
0
.
1
.
2
33 3 ?
3 (5,7,11,20) o
2
(.
4
0
. .
3
). .
5
1
.
0
. .
3
2
.
1
. .
3
0
.
2
1
40 4 Y
3 (11,21,29,37) .
6
0
.
2
. .
4
1
.
0
. .
2
2
.
3
. .
2
3
.
1
95 3 Y
3 (11,37,53,98) .
13
0
.
2
. .
5
1
.
0
. .
3
2
.
1
. .
2
3
196 2 Y
3 (13,17,27,41) .
6
0
.
1
. .
4
1
.
2
. .
2
2
.
3
. .
2
3
.
0
95 3 Y
3 (13,27,61,98) .
13
0
.
1
. .
5
1
.
2
. .
3
2
.
0
. .
2
3
196 2 Y
3 (15,19,43,74) .
7
0
.
2
. .
7
1
.
0
. .
3
2
.
1
. .
2
3
148 2 Y
4 (5,6,8,9) .
3
0
.
3
. .
4
1
. .
3
2
. .
2
3
.
1
. .
2
0
.
1
.
2
24 3 ?
4 (5,6,8,15) o
2
(.
3
0
. .
3
). .
5
1
. .
3
2
.
1
. .
2
0
.
2
1
.
2
30 4 ?
4 (9,11,12,17) .
5
0
. .
3
1
.
2
. .
3
2
.
0
. .
2
3
.
1
45 3 ?
4 (10,13,25,31) .
5
0
.
2
. .
5
1
.
0
. .
3
2
. .
2
3
.
1
75 3 Y
4 (11,17,20,27) .
4
0
.
3
. .
3
1
.
2
. .
3
2
.
0
. .
2
3
.
1
71 3 ?
4 (11,17,24,31) .
5
0
.
2
. .
4
1
.
0
. .
2
2
.
3
. .
2
3
.
1
79 3 Y
4 (11,31,45,83) .
11
0
.
2
. .
5
1
.
0
. .
3
2
.
1
. .
2
3
166 2 Y
4 (13,14,19,29) .
4
0
.
2
. .
3
1
.
3
. .
3
2
.
1
. .
2
3
.
0
71 2 ?
4 (13,14,23,33) .
5
0
.
1
. .
4
1
.
2
. .
2
2
.
3
. .
2
3
.
0
79 3 Y
4 (13,23,51,83) .
11
0
.
1
. .
5
1
.
2
. .
3
2
.
0
. .
2
3
166 2 Y
5 (11,13,19,25) .
4
0
.
2
. .
4
1
.
0
. .
2
2
.
3
. .
2
3
.
1
63 3 ?
5 (11,25,37,68) .
9
0
.
2
. .
5
1
.
0
. .
3
2
.
1
. .
2
3
136 2 Y
5 (13,19,41,68) .
9
0
.
1
. .
5
1
.
2
. .
3
2
.
0
. .
2
3
136 2 Y
6 (7,10,15,19) .
5
0
.
1
. .
3
1
.
2
. .
3
2
. .
2
3
.
0
45 3 ?
6 (11,19,29,53) .
7
0
.
2
. .
5
1
.
0
. .
4
2
. .
2
3
106 2 Y
6 (13,15,31,53) .
7
0
.
1
. .
5
1
.
2
. .
3
2
.
0
. .
2
3
106 2 Y
7 (11,13,21,38) .
5
0
.
2
. .
5
1
.
0
. .
3
2
.
1
. .
2
3
76 2 Y
8 (7,11,13,23) .
5
0
.
1
. .
3
1
.
2
. .
3
2
.
0
. .
2
3
46 2 ?
8 (7,18,27,37) .
9
0
.
1
. .
3
1
.
2
. .
3
2
. .
2
3
.
0
81 3 ?
9 (7,15,19,32) .
7
0
.
1
. .
3
1
.
2
. .
3
2
.
0
. .
2
3
64 2 ?
10 (7,19,25,41) .
9
0
.
1
. .
3
1
.
2
. .
3
2
.
0
. .
2
3
82 2 ?
10 (7,26,39,55) .
13
0
.
1
. .
3
1
.
2
. .
3
2
. .
2
3
.
0
117 3 ?
* (for lack of space only the total number of monomial terms in 1
w
is indicated)
The computer program indicates that there are neither series solutions nor
sporadic solutions satisfying the hypothesis of Theorem 5.4.16 for 1 10. In fact,
an easy argument shows that there are no such solutions for suciently large 1.
4
4
The code for the C program used to generate the tables of the Theorem 5.4.16 are available
at: http://www.math.unm.edu/galicki/papers/publications.html.
564 APPENDIX B
B.4. Positive Breiskorn-Pham Links in Dimension 5
Table B.4.1: Positive BP Links with /
2
(`) = 0, H
2
(1(a). Z) = 0
` 1(a)
SJ
`
2
1(2. 3. 9. 15), 1(2. 3. 3. /), / = 3(6)
`
3
1(2. 3. 8. 20), 1(2. 3. 4. /), / = 4. 8(12)
`
4
1(3. 3. 3. 4), 1(3. 3. 4. 9) 2
`
5
1(3. 3. 5. 6), 1(2. 3. 5. /), / = 6. 12. 18. 24(30),
1(2. 4. 5. 8), 1(2. 4. 5. 12), 1(2. 4. 5. 16)
1(2. 4. 7. 8), 1(2. 4. 6. 7), 1(2. 3. 7. 36),
`
7
1(2. 3. 7. 30), 1(2. 3. 7. 24), 1(2. 3. 7. 18), 10
1(2. 3. 7. 12), 1(3. 3. 3. 7), 1(2. 4. 4. 7),
1(2. 3. 6. 7)
`
8
1(3. 3. 3. 8) 1
`
9
1(2. 4. 4. 9) 1
`
10
1(3. 3. 3. 10) 1
`
11
1(2. 3. 11. 12), 1(3. 3. 3. 11), 1(2. 4. 4. 11), 5
1(2. 3. 6. 11), 1(2. 4. 6. 11)
`
k
, / 12, / = 1. 5(6) 1(2. 3. 6. /), 1(3. 3. 3. /), 1(2. 4. 4. /) 3
`
k
, / 12, / = 2. 4(6) 1(3. 3. 3. /) 1
`
k
, / 12, / = 3(6) 1(2. 4. 4. /) 1
2`
3
1(2. 3. 5. /), / = 10. 20(30)
3`
3
1(2. 3. 8. 8). 1(2. 3. 8. 16), 1(3. 4. 4. 4). 5
1(2. 3. 7. 14), 1(2. 3. 7. 28)
4`
3
1(2. 3. 10. 10) 1
2`
5
1(2. 5. 6. 6) 1
4`
2
1(2. 3. 5. /), / = 15(30)
6`
2
1(2. 5. 5. 5). 1(2. 3. 7. 21) 2
7`
2
1(2. 3. 9. 9) 1
B.4. POSITIVE BP LINKS 565
Table B.4.2: Positive BP Links with /
2
(`) 0, H
2
(1(a). Z)
tor
= 0
` 1(a)
SJ
2`

#`
2
1(2. 3. 4. /), / = 6(12)
1(2. 4. 6. 9), 1(2. 3. 6. 8)
2`

#`
4
1(2. 3. 8. 18), 1(2. 3. 6. 8) 2
2`

#`
5
1(2. 3. 10. 12), 1(2. 3. 6. 10) 2
2`

#`
k
, (3. /) = 1, / 6 1(2. 3. 6. 2/) 1
3`

#`
3
1(2. 4. 6. 8), 1(2. 4. 4. 6) 2
3`

#`
k
, (2. /) = 1, / 4 1(2. 4. 4. 2/) 1
4`

#`
2
1(3. 3. 4. 6) 1
4`

#`
3
1(2. 3. 9. 12), 1(2. 3. 6. 9) 2
4`

#`
k
, (2. /) = 1, / 4 1(2. 3. 6. 3/) 1
4`

#2`
2
1(2. 4. 5. 10) 1
5`

#2`
2
1(2. 4. 6. 6) 1
6`

#`
2
1(3. 3. 3. 6), 1(2. 3. 8. 12) 2
6`

#`
k
, / 2 1(3. 3. 3. 3/) 1
7`

#`
k
, / 1 1(2. 4. 4. 4/) 1
8`

#`
k
, / 1 1(2. 3. 6. 6/) 1
566 APPENDIX B
Table B.4.3: Positive BP Links dieomorphic to o
5
1(2. 2. j. c) (j. c) = 1
1(2. 3. 3. /) / = 1. 5(6)
1(2. 3. 4. /) / = 1. 2. 5. 7. 10. 11(12)
1(2. 3. 5. /) / = 0. 6. 10. 12. 15. 18. 20. 24(30)
1(2. 3. 7. /) 6 < / < 42, / = 12. 14. 18. 21. 24. 28. 30. 36
1(2. 3. 8. /) / = 10. 11. 13. 14. 17. 19. 22. 23
1(2. 3. 9. /) / = 11. 13. 17
1(2. 3. 10. /) / = 11. 13. 14
1(2. 3. 11. /) / = 11. 13
1(2. 4. 5. /) / = 6. 7. 9. 11. 13. 14. 17. 18. 19
1(2. 4. 6. /) / = 7. 11
1(2. 4. 7. /) / = 9
1(2. 5. 5. /) / = 7. 9
1(2. 5. 6. /) / = 7
1(3. 3. 4. /) / = 5. 7. 11
1(3. 3. 5. /) / = 7
1(3. 4. 4. /) / = 5
Table B.4.4: Positive BP Links with /
2
(`) 0, H
2
(1(a). Z)
tor
= 0
` 1(a)
(n 1)`

1(2. 2. j. c), (j. c) = n 1

2`

1(2. 3. 3. /), / = 2. 4(6), 1(2. 3. 4. /), / = 3. 9(12)

4`

1(2. 3. 3. /), / = 0(6)

6`

1(2. 3. 4. /), / = 0(12)

8`

1(2. 3. 5. /), / = 0(30)

`

1(2. 4. 6. 10)
2`

1(2. 3. 8. /), / = 9. 15. 21, 1(2. 3. 9. /), / = 10. 14. 16

1(3. 3. 4. 10)
4`

1(2. 4. 5. /), / = 5. 15, 1(2. 5. 5. /), / = 6. 8

6`

1(2. 4. 7. 7), 1(3. 3. 4. /), / = 4. 8

8`

1(3. 3. 5. 5)
B.5. YAU-YU LINKS 567
B.5. The Yau-Yu Links in Dimensions 5
Table B.5.1: The Yau-Yu Links in Dimensions 5
Type 1(.
0
. .
1
. .
2
. .
3
) [w[degree(1)
I (BP) .
a
0
+.
b
1
+.
c
2
+.
d
3
1
a
+
1
b
+
1
c
+
1
d
II .
a
0
+.
b
1
+.
c
2
+.
2
.
d
3
1
a
+
1
b
+
1
c
+
c1
cd
III .
a
0
+.
b
1
+.
c
2
.
3
+.
2
.
d
3
1
a
+
1
b
+
d1
cd1
+
c1
cd1
IV .
a
0
+.
0
.
b
1
+.
c
2
+.
2
.
d
3
1
a
+
a1
ab
+
1
c
+
c1
cd
V .
a
0
.
1
+.
0
.
b
1
+.
c
2
+.
2
.
d
3
b1
ab1
+
a1
ab1
+
1
c
+
c1
cd
VI .
a
0
.
1
+.
0
.
b
1
+.
c
2
.
3
+.
2
.
d
3
b1
ab1
+
a1
ab1
+
d1
cd1
+
c1
dc
VII .
a
0
+.
b
1
+.
1
.
c
2
+.
2
.
d
3
1
a
+
1
b
+
b1
bc
+
[b(c1)+1]
bcd
VIII .
a
0
+.
b
1
+.
1
.
c
2
+.
1
.
d
3
+.
p
2
.
q
3
1
a
+
1
b
+
b1
bc
+
b1
bd
p(b1)
bc
+
q(b1)
bc
= 1
IX .
a
0
+.
b
1
.
3
+.
c
2
.
3
+.
1
.
d
3
+.
p
1
.
q
2
1
a
+
(d1)
bd1
+
b(d1)
c(bd1)
+
b1
bd1
p(d1)
bd1
+
qb(d1)
c(bd1)
= 1
X .
a
0
+.
b
1
.
2
+.
c
2
.
3
+.
1
.
d
3
1
a
+
[d(c1)+1]
bcd+1
+
[b(d1)+1]
bcd+1
+
[c(b1)+1]
bcd+1
XI .
a
0
+.
0
.
b
1
+.
1
.
c
2
+.
2
.
d
3
1
a
+
a1
ab
+
[a(b1)+1]
abc
+
[ab(c1)+(a1)]
abcd
XII .
a
0
+.
0
.
b
1
+.
0
.
c
2
+.
1
.
d
3
+.
p
1
.
q
2
1
a
+
a1
ab
+
a1
ac
+
[a(b1)+1]
abd
p(a1)
ab
+
q(a1)
ac
= 1
XIII .
a
0
+.
0
.
b
1
+.
1
.
c
2
+.
1
.
d
3
+.
p
2
.
q
3
1
a
+
a1
ab
+
a1
ac
+
[a(b1)+1]
abd
p[a(b1)+1]
abc
+
q[a(b1)+1]
abd
= 1
XIV .
a
0
+.
0
.
b
1
+.
0
.
c
2
+.
0
.
d
3
+.
p
1
.
q
2
+.
r
2
.
s
3
1
a
+
a1
ab
+
a1
ac
+
a1
ad
p(a1)
ab
+
q(a1)
ac
= 1 =
r(a1)
ac
+
s(a1)
ad
XV .
a
0
.
1
+.
0
.
b
1
+.
0
.
c
2
+.
2
.
d
3
+.
p
1
.
q
2
b1
ab1
+
a1
ab1
+
b(a1)
c(ab1)
p(a1)
ab1
+
qb(a1)
c(ab1)
= 1 +
[c(ab1)b(a1)]
cd(ab1)
XVI .
a
0
.
1
+.
0
.
b
1
+.
0
.
c
2
+.
0
.
d
3
+.
p
1
.
q
2
+.
r
2
.
s
3
(b1)
ab1
+
(a1)
ab1
+
b(a1)
c(ab1)
+
b(a1)
d(ab1)
p(a1)
ab1
+
qb(a1)
c(ab1)
= 1 =
r(a1)
ac
+
s(a1)
ad
XVII .
a
0
.
1
+.
0
.
b
1
+.
1
.
c
2
+.
0
.
d
3
+.
p
1
.
q
3
+.
r
0
.
s
2
b1
ab1
+
a1
ab1
+
a(b1)
c(ab1)
+
b(a1)
d(ab1)
p(a1)
ab1
+
qb(a1)
d(ab1)
= 1 =
r(b1)
ab1
+
sa(b1)
c(ab1)
XVIII .
a
0
.
2
+.
0
.
b
1
+.
1
.
c
2
+.
1
.
d
3
+.
p
2
.
q
3
[b(c1)+1]
abc+1
+
[c(a1)+1]
abc+1
p[a(b1)+1]
abc+1
+
qc[a(b1)+1]
d(abc+1)
= 1 +
[a(b1)+1]
c(abc+1)
+
c[a(b1)+1]
d(abc+1)
XIX .
a
0
.
3
+.
0
.
b
1
+.
c
2
.
3
+.
2
.
d
3
[b(d(c1)+1)1]
abcd1
+
[d(c(a1)+1)1]
abcd1
+
[a(b(d1)+1)1]
abcd1
+
[c(a(b1)+1)1]
abcd1
568 APPENDIX B
Bibliography
[AB68] M. F. Atiyah and R. Bott, A Lefschetz xed point formula for elliptic complexes.
II. Applications, Ann. of Math. (2) 88 (1968), 451491. MR 0232406 (38 #731)
[AB83] , The Yang-Mills equations over Riemann surfaces, Philos. Trans. Roy. Soc.
London Ser. A 308 (1983), no. 1505, 523615. MR 702806 (85k:14006)
[Abb84] E. Abbena, An example of an almost Kahler manifold which is not Kahlerian, Boll.
Un. Mat. Ital. A (6) 3 (1984), no. 3, 383392. MR 86a:53036
[ABC
+
96] J. Amoros, M. Burger, K. Corlette, D. Kotschick, and D. Toledo, Fundamen-
tal groups of compact Kahler manifolds, Mathematical Surveys and Monographs,
vol. 44, American Mathematical Society, Providence, RI, 1996. MR 97d:32037
[ABCC06] R. Argurio, M. Bertolini, C. Closset, and S Cremonesi, On stable non-
supersymmetric vacua at the bottom of cascading theories, J. High Energy Phys.
(2006), no. 9, 030, 21 pp. (electronic). MR 2257499
[Abe76] K. Abe, Some examples of non-regular almost contact structures on exotic spheres,
Tohoku Math. J. (2) 28 (1976), no. 3, 429435. MR 0423247 (54 #11227)
[Abe77] , On a generalization of the Hopf bration. I. Contact structures on the
generalized Brieskorn manifolds, Tohoku Math. J. (2) 29 (1977), no. 3, 335374.
MR 0464253 (57 #4187)
[Abr01] M. Abreu, Kahler metrics on toric orbifolds, J. Dierential Geom. 58 (2001), no. 1,
151187. MR 1895351 (2003b:53046)
[ACDVP03] D. V. Alekseevsky, V. Cortes, C. Devchand, and A. Van Proeyen, Flows on
quaternionic-K ahler and very special real manifolds, Comm. Math. Phys. 238
(2003), no. 3, 525543. MR 1993384 (2004m:53080)
[ACG06] V. Apostolov, D. M. J. Calderbank, and P. Gauduchon, Hamiltonian 2-forms in
K ahler geometry. I. General theory, J. Dierential Geom. 73 (2006), no. 3, 359412.
MR 2228318
[ADHL03] B. Acharya, F. Denef, C. Hofman, and N. Lambert, Freund-Rubin revisited,
preprint; arXiv:hep-th/0308046 (2003).
[ADM96] V. Apostolov, J. Davidov, and O. Muskarov, Compact self-dual Hermitian surfaces,
Trans. Amer. Math. Soc. 348 (1996), no. 8, 30513063. MR 1348147 (96j:53040)
[ADM06] V. Apostolov, T. Draghici, and A. Moroianu, The odd-dimensional Goldberg Con-
jecture, Math. Nachr. 279 (2006), no. 9-10, 948952. MR MR2242959
[AE75] K. Abe and J. Erbacher, Nonregular contact structures on Brieskorn manifolds,
Bull. Amer. Math. Soc. 81 (1975), 407409. MR 0417974 (54 #6019)
[AG86] V. A. Alexiev and G. T. Ganchev, On the classication of the almost contact metric
manifolds, Mathematics and mathematical education (Bulgarian) (Sunny Beach
(Sl

nchev Bryag), 1986), Publ. House Bulgar. Acad. Sci., Soa, 1986, pp. 155161.
MR 872914 (88c:53035a)
[AG02] V. Apostolov and P. Gauduchon, Selfdual Einstein Hermitian four-manifolds,
Ann. Sc. Norm. Super. Pisa Cl. Sci. (5) 1 (2002), no. 1, 203243. MR 1994808
(2004i:53048)
[AGF81] L. Alvarez-Gaume and D. Z. Freedman, Geometrical structure and ultraviolet nite-
ness in the supersymmetric -model, Comm. Math. Phys. 80 (1981), no. 3, 443451.
MR 626710 (82i:81143)
[AGI98] B. Alexandrov, G. Grantcharov, and S. Ivanov, Curvature properties of twistor
spaces of quaternionic Kahler manifolds, J. Geom. 62 (1998), no. 1-2, 112.
MR 1631453 (2000b:53066)
569
570 BIBLIOGRAPHY
[AGP06] C. Arezzo, A. Ghigi, and G. P. Pirola, Symmetries, quotients and Kahler-Einstein
metrics, J. Reine Angew. Math. 591 (2006), 177200. MR 2212883
[Agr06] I. Agricola, The SRNI Lectures on non-integrable geometries with torsion, preprint;
arXiv:math.DG/0606705 (2006).
[AGV72] M. Artin, A. Grothendieck, and J. L. Verdier, Theorie des topos et cohomologie etale
des schemas. Tome 1: Theorie des topos, Springer-Verlag, Berlin, 1972, Seminaire
de Geometrie Algebrique du Bois-Marie 19631964 (SGA 4), Dirige par M. Artin,
A. Grothendieck, et J. L. Verdier. Avec la collaboration de N. Bourbaki, P. Deligne
et B. Saint-Donat, Lecture Notes in Mathematics, Vol. 269. MR 50 #7130
[AH85] M. F. Atiyah and N. J. Hitchin, Low energy scattering of nonabelian monopoles,
Phys. Lett. A 107 (1985), no. 1, 2125. MR 778313 (86e:53053)
[AH88] M. Atiyah and N. Hitchin, The geometry and dynamics of magnetic monopoles, M.
B. Porter Lectures, Princeton University Press, Princeton, NJ, 1988. MR 934202
(89k:53067)
[AHDM78] M. F. Atiyah, N. J. Hitchin, V. G. Drinfel

d, and Yu. I. Manin, Construction of

instantons, Phys. Lett. A 65 (1978), no. 3, 185187. MR 598562 (82g:81049)
[AHS78] M. F. Atiyah, N. J. Hitchin, and I. M. Singer, Self-duality in four-dimensional
Riemannian geometry, Proc. Roy. Soc. London Ser. A 362 (1978), no. 1711, 425
461. MR 506229 (80d:53023)
[AJPS97] S. Albeverio, J. Jost, S. Paycha, and S. Scarlatti, A mathematical introduction to
string theory, London Mathematical Society Lecture Note Series, vol. 225, Cam-
bridge University Press, Cambridge, 1997, Variational problems, geometric and
probabilistic methods. MR 1480168 (99d:58178)
[AK94] N. ACampo and D. Kotschick, Contact structures, foliations, and the fundamental
group, Bull. London Math. Soc. 26 (1994), no. 1, 102106. MR 94j:57027
[AK04] D. Alekseevsky and Y. Kamishima, Quaternionic and para-quaternionic CR struc-
ture on (4n+3)-dimensional manifolds, Cent. Eur. J. Math. 2 (2004), no. 5, 732753
(electronic). MR 2172049 (2006i:53069)
[Akh90] D. N. Akhiezer, Homogeneous complex manifolds, Several Complex Variables IV
(G.M. Khenkin, ed.), Encyclopedia of mathematical Sciences, Vol. 10, Springer-
Verlag, Berlin, 1990, Algebraic aspects of complex analysis, A translation of Sovre-
mennye problemy matematiki. Fundamentalnye napravleniya, Tom 10, Akad. Nauk
SSSR, Vsesoyuz. Inst. Nauchn. i Tekhn. Inform., Moscow, 1986 [MR 87m:32003],
Translation by J. Leiterer and J. Nunemacher, Translation edited by S. G. Gindikin
and G. M. Khenkin, pp. 223275, 283. MR 88i:32047
[Akh95] , Lie group actions in complex analysis, Aspects of Mathematics, E27,
Friedr. Vieweg & Sohn, Braunschweig, 1995. MR 1334091 (96g:32051)
[AKL89] M. T. Anderson, P. B. Kronheimer, and C. LeBrun, Complete Ricci-at Kahler
manifolds of innite topological type, Comm. Math. Phys. 125 (1989), no. 4, 637
642. MR 1024931 (91a:53073)
[Alb89] C. Albert, Le theor`eme de reduction de Marsden-Weinstein en geometrie cosym-
plectique et de contact, J. Geom. Phys. 6 (1989), no. 4, 627649. MR 91k:58033
[Ale68] D. V. Alekseevski, Riemannian spaces with unusual holonomy groups, Funkcional.
Anal. i Prilozen 2 (1968), no. 2, 110. MR 37 #6868
[Ale75] , Classication of quaternionic spaces with transitive solvable group of mo-
tions, Izv. Akad. Nauk SSSR Ser. Mat. 39 (1975), no. 2, 315362, 472. MR 53
#6465
[Ale90] D. V. Alekseevski, Contact homogeneous spaces, Funktsional. Anal. i Prilozhen.
24 (1990), no. 4, 7475. MR 1092805 (91j:53027)
[ALR06] A. Adem, J. Leida, and Y. Ruan, Orbifolds and stringy topology, Cambridge Uni-
versity Press (to appear), 2006.
[AM78] R. Abraham and J. E. Marsden, Foundations of mechanics, Benjamin/Cummings
Publishing Co. Inc. Advanced Book Program, Reading, Mass., 1978, Second edition,
revised and enlarged, With the assistance of Tudor Rat iu and Richard Cushman.
MR 81e:58025
[AM96a] D. V. Alekseevsky and S. Marchiafava, Hypercomplex structures on quaternionic
manifolds, New developments in dierential geometry (Debrecen, 1994), Math.
BIBLIOGRAPHY 571
Appl., vol. 350, Kluwer Acad. Publ., Dordrecht, 1996, pp. 119. MR 1377270
(96m:53052)
[AM96b] , Quaternionic structures on a manifold and subordinated structures, Ann.
Mat. Pura Appl. (4) 171 (1996), 205273. MR 1441871 (98j:53033)
[AMP97] L. Astey, E. Micha, and G. Pastor, Homeomorphism and dieomorphism types of
Eschenburg spaces, Dierential Geom. Appl. 7 (1997), no. 1, 4150. MR 98h:53072
[AN54] Y. Akizuki and S. Nakano, Note on Kodaira-Spencers proof of Lefschetz theorems,
Proc. Japan Acad. 30 (1954), 266272. MR 0066694 (16,619a)
[And90] M. T. Anderson, Convergence and rigidity of manifolds under Ricci curvature
bounds, Invent. Math. 102 (1990), no. 2, 429445. MR 92c:53024
[Ano69] D. V. Anosov, Geodesic ows on closed Riemann manifolds with negative curva-
ture., Proceedings of the Steklov Institute of Mathematics, No. 90 (1967). Translated
from the Russian by S. Feder, American Mathematical Society, Providence, R.I.,
1969. MR 39 #3527
[Ara02] C. Araujo, Kahler-Einstein metrics for some quasi-smooth log del Pezzo sur-
faces, Trans. Amer. Math. Soc. 354 (2002), no. 11, 43034312 (electronic).
MR 2003g:32039
[Arn78] V. I. Arnold, Mathematical methods of classical mechanics, Springer-Verlag, New
York, 1978, Translated from the Russian by K. Vogtmann and A. Weinstein, Grad-
uate Texts in Mathematics, 60. MR 57 #14033b
[AS53] W. Ambrose and I. M. Singer, A theorem on holonomy, Trans. Amer. Math. Soc.
75 (1953), 428443. MR 0063739 (16,172b)
[AS63] M. F. Atiyah and I. M. Singer, The index of elliptic operators on compact manifolds,
Bull. Amer. Math. Soc. 69 (1963), 422433. MR 0157392 (28 #626)
[AS04] V. Apostolov and S. Salamon, K ahler reduction of metrics with holonomy G
2
,
Comm. Math. Phys. 246 (2004), no. 1, 4361. MR 2044890 (2005e:53070)
[Ati82] M. F. Atiyah, Convexity and commuting Hamiltonians, Bull. London Math. Soc.
14 (1982), no. 1, 115. MR 642416 (83e:53037)
[Aub76] T. Aubin,

Equations du type Monge-Amp`ere sur les varietes kahleriennes com-
pactes, C. R. Acad. Sci. Paris Ser. A-B 283 (1976), no. 3, Aiii, A119A121. MR 55
#6496
[Aub78] ,

Equations du type Monge-Amp`ere sur les varietes kahleriennes compactes,
Bull. Sci. Math. (2) 102 (1978), no. 1, 6395. MR 494932 (81d:53047)
[Aub82] , Nonlinear analysis on manifolds. Monge-Amp`ere equations, Grundlehren
der Mathematischen Wissenschaften [Fundamental Principles of Mathematical Sci-
ences], vol. 252, Springer-Verlag, New York, 1982. MR 85j:58002
[Aub84] , Reduction du cas positif de lequation de Monge-Amp`ere sur les varietes
k ahleriennes compactes `a la demonstration dune inegalite, J. Funct. Anal. 57
(1984), no. 2, 143153. MR 749521 (85k:58084)
[Aub98] , Some nonlinear problems in Riemannian geometry, Springer Monographs
in Mathematics, Springer-Verlag, Berlin, 1998. MR 1636569 (99i:58001)
[AW75] S. Alo and N. R. Wallach, An innite family of distinct 7-manifolds admitting
positively curved Riemannian structures, Bull. Amer. Math. Soc. 81 (1975), 9397.
MR 51 #6851
[AW02] M. Atiyah and E. Witten, M-theory dynamics on a manifold of G
2
holonomy, Adv.
Theor. Math. Phys. 6 (2002), no. 1, 1106. MR 1992874 (2004f:53046)
[Bai56] W. L. Baily, The decomposition theorem for V -manifolds, Amer. J. Math. 78 (1956),
862888. MR 20 #6537
[Bai57] , On the imbedding of V -manifolds in projective space, Amer. J. Math. 79
(1957), 403430. MR 20 #6538
[Bal06] W. Ballmann, Lectures on Kahler manifolds, ESI Lectures in Mathematics and
Physics, European Mathematical Society (EMS), Z urich, 2006. MR 2243012
[Ban90] A. Banyaga, A note on Weinsteins conjecture, Proc. Amer. Math. Soc. 109 (1990),
no. 3, 855858. MR 90m:58170
[Ban96] , Instantons and hypercontact structures, J. Geom. Phys. 19 (1996), no. 3,
267276. MR 1397411 (97g:53025)
572 BIBLIOGRAPHY
[Ban97] , The structure of classical dieomorphism groups, Mathematics and its
Applications, vol. 400, Kluwer Academic Publishers Group, Dordrecht, 1997.
MR 1445290 (98h:22024)
[Bar65] D. Barden, Simply connected ve-manifolds, Ann. of Math. (2) 82 (1965), 365385.
MR 32 #1714
[Bar85] R. Barlow, A simply connected surface of general type with pg = 0, Invent. Math.
79 (1985), no. 2, 293301. MR 778128 (87a:14033)
[Bar93] Ch. Bar, Real Killing spinors and holonomy, Comm. Math. Phys. 154 (1993), no. 3,
509521. MR 94i:53042
[Bat96a] F. Battaglia, A hypercomplex Stiefel manifold, Dierential Geom. Appl. 6 (1996),
no. 2, 121128. MR 1395025 (97g:53084)
[Bat96b] , S
1
-quotients of quaternion-Kahler manifolds, Proc. Amer. Math. Soc. 124
(1996), no. 7, 21852192. MR 1307492 (96i:53045)
[Bat99] , Circle actions and Morse theory on quaternion-Kahler manifolds, J. Lon-
don Math. Soc. (2) 59 (1999), no. 1, 345358. MR 1688506 (2000h:53060)
[Bau00] H. Baum, Twistor and Killing spinors in Lorentzian geometry, Global analysis
and harmonic analysis (Marseille-Luminy, 1999), Semin. Congr., vol. 4, Soc. Math.
France, Paris, 2000, pp. 3552. MR 2002d:53060
[BB73] A. Bialynicki-Birula, Some theorems on actions of algebraic groups, Ann. of Math.
(2) 98 (1973), 480497. MR 0366940 (51 #3186)
[BB82] L. Berard-Bergery, Sur de nouvelles varietes riemanniennes dEinstein, Institut

Elie Cartan, 6, Inst.

Elie Cartan, vol. 6, Univ. Nancy, Nancy, 1982, pp. 160.
MR 727843 (85b:53048)
[BB02] P. Berglund and A. Brandhuber, Matter from G
2
manifolds, Nuclear Phys. B 641
(2002), no. 1-2, 351375. MR 1928190 (2004f:53055)
[BBC05] M. Bertolini, F. Bigazzi, and A. L. Cotrone, Supersymmetry breaking at the end
of a cascade of Seiberg dualities, Phys. Rev. D (3) 72 (2005), no. 6, 061902, 5.
MR 2182338 (2006e:81194)
[BC01] R. L. Bishop and R. J. Crittenden, Geometry of manifolds, AMS Chelsea Publish-
ing, Providence, RI, 2001, Reprint of the 1964 original. MR 1852066 (2002d:53001)
[BCGP05] C. P. Boyer, D. M. J. Calderbank, K. Galicki, and P. Piccinni, Toric self-dual
Einstein metrics as quotients, Comm. Math. Phys. 253 (2005), no. 2, 337370.
MR 2140252
[BD00] R. Bielawski and A. S. Dancer, The geometry and topology of toric hyperkahler
manifolds, Comm. Anal. Geom. 8 (2000), no. 4, 727760. MR 2002c:53078
[BDM96] M. L. Barberis and I. Dotti Miatello, Hypercomplex structures on a class of solv-
able Lie groups, Quart. J. Math. Oxford Ser. (2) 47 (1996), no. 188, 389404.
MR 1460231 (98k:53055)
[BE69] M. Berger and D. Ebin, Some decompositions of the space of symmetric tensors on
a Riemannian manifold, J. Dierential Geometry 3 (1969), 379392. MR 0266084
(42 #993)
[Bea98] A. Beauville, Fano contact manifolds and nilpotent orbits, Comment. Math. Helv.
73 (1998), no. 4, 566583. MR 1639888 (99e:14046)
[Bea05] , Riemannian holonomy and algebraic geometry, preprint,
math.AG/9902110 (2005).
[Bel00] F. A. Belgun, On the metric structure of non-Kahler complex surfaces, Math. Ann.
317 (2000), no. 1, 140. MR 2002c:32027
[Bel01] , Normal CR structures on compact 3-manifolds, Math. Z. 238 (2001), no. 3,
441460. MR 2002k:32065
[Ber55] M. Berger, Sur les groupes dholonomie homog`ene des varietes `a connexion ane
et des varietes riemanniennes, Bull. Soc. Math. France 83 (1955), 279330.
MR 18,149a
[Ber66] , Remarques sur les groupes dholonomie des varietes riemanniennes, C. R.
Acad. Sci. Paris Ser. A-B 262 (1966), A1316A1318. MR 0200860 (34 #746)
[Ber03] , A panoramic view of Riemannian geometry, Springer-Verlag, Berlin, 2003.
MR 2002701 (2004h:53001)
BIBLIOGRAPHY 573
[Bes87] A. L. Besse, Einstein manifolds, Ergebnisse der Mathematik und ihrer Grenzgebiete
(3) [Results in Mathematics and Related Areas (3)], vol. 10, Springer-Verlag, Berlin,
1987. MR 88f:53087
[BF82] C. P. Boyer and J. D. Finley, III, Killing vectors in self-dual, Euclidean Einstein
spaces, J. Math. Phys. 23 (1982), no. 6, 11261130. MR 660020 (84f:53064)
[BFGK91] H. Baum, T. Friedrich, R. Grunewald, and I. Kath, Twistors and Killing spinors on
Riemannian manifolds, Teubner-Texte zur Mathematik [Teubner Texts in Math-
ematics], vol. 124, B. G. Teubner Verlagsgesellschaft mbH, Stuttgart, 1991, With
German, French and Russian summaries. MR 94a:53077
[BFH
+
05] S. Benvenuti, S. Franco, A. Hanany, D. Martelli, and J. Sparks, An innite family
of superconformal quiver gauge theories with Sasaki-Einstein duals, J. High Energy
Phys. (2005), no. 6, 064, 26 pp. (electronic). MR 2158305 (2006h:81276)
[BFZ05] A. Butti, D. Forcella, and A. Zaaroni, The dual superconformal theory for L
p,q,r
manifolds, J. High Energy Phys. (2005), no. 9, 018, 39 pp. (electronic). MR 2171349
(2006h:81204)
[BG67] D. E. Blair and S. I. Goldberg, Topology of almost contact manifolds, J. Dierential
Geometry 1 (1967), 347354. MR 37 #2128
[BG97a] O. Biquard and P. Gauduchon, Hyper-K ahler metrics on cotangent bundles of Her-
mitian symmetric spaces, Geometry and physics (Aarhus, 1995), Lecture Notes in
Pure and Appl. Math., vol. 184, Dekker, New York, 1997, pp. 287298. MR 1423175
(97k:53041)
[BG97b] C. P. Boyer and K. Galicki, The twistor space of a 3-Sasakian manifold, Internat.
J. Math. 8 (1997), no. 1, 3160. MR 98e:53072
[BG99] , 3-Sasakian manifolds, Surveys in dierential geometry: essays on Einstein
manifolds, Surv. Dier. Geom., VI, Int. Press, Boston, MA, 1999, pp. 123184.
MR 2001m:53076
[BG00a] , A note on toric contact geometry, J. Geom. Phys. 35 (2000), no. 4, 288
298. MR 2001h:53124
[BG00b] , On Sasakian-Einstein geometry, Internat. J. Math. 11 (2000), no. 7, 873
909. MR 2001k:53081
[BG01a] , Einstein manifolds and contact geometry, Proc. Amer. Math. Soc. 129
(2001), no. 8, 24192430 (electronic). MR 2001m:53071
[BG01b] , New Einstein metrics in dimension ve, J. Dierential Geom. 57 (2001),
no. 3, 443463. MR 2003b:53047
[BG02] , Rational homology 5-spheres with positive Ricci curvature, Math. Res.
Lett. 9 (2002), no. 4, 521528. MR 2003g:53044
[BG03] , New Einstein metrics on 8#(S
2
S
3
), Dierential Geom. Appl. 19 (2003),
no. 2, 245251. MR 2 002 662
[BG04] D. Burns and V. Guillemin, Potential functions and actions of tori on Kahler man-
ifolds, Comm. Anal. Geom. 12 (2004), no. 1-2, 281303. MR 2074879 (2005e:53133)
[BG05] C. P. Boyer and K. Galicki, Sasakian geometry, hypersurface singularities, and
Einstein metrics, Rend. Circ. Mat. Palermo (2) Suppl. (2005), no. 75, suppl., 57
87. MR 2152356
[BG06a] , Einstein Metrics on Rational Homology Spheres, J. Dierential Geom. 74
(2006), no. 3, 353362. MR 2269781
[BG06b] , Erratum and addendum for rational homology 5-spheres with positive
Ricci curvature, Math. Res. Lett. 13 (2006), no. 3, 463465.
[BG06c] , Highly connected manifolds with positive Ricci curvature, Geom. Topol. 10
(2006), 22192235(electronic).
[BG06d] , Sasakian geometry and Einstein metrics on spheres, Perspectives in Rie-
mannian Geometry, CRM Proceedings and Lecture Notes, vol. 40, Amer. Math.
Soc., Providence, RI, 2006, pp. 4761. MR 2237105
[BGK05] C. P. Boyer, K. Galicki, and J. Kollar, Einstein metrics on spheres, Ann. of Math.
(2) 162 (2005), no. 1, 557580. MR MR2178969 (2006j:53058)
[BGKT05] C. P. Boyer, K. Galicki, J. Kollar, and E. Thomas, Einstein metrics on exotic
spheres in dimensions 7, 11, and 15, Experiment. Math. 14 (2005), no. 1, 5964.
MR MR2146519 (2006a:53042)
574 BIBLIOGRAPHY
[BGM93] C. P. Boyer, K. Galicki, and B. M. Mann, Quaternionic reduction and Einstein
manifolds, Comm. Anal. Geom. 1 (1993), no. 2, 229279. MR 95c:53056
[BGM94a] , The geometry and topology of 3-Sasakian manifolds, J. Reine Angew.
Math. 455 (1994), 183220. MR 96e:53057
[BGM94b] , Some new examples of compact inhomogeneous hypercomplex manifolds,
Math. Res. Lett. 1 (1994), no. 5, 531538. MR 95f:53083
[BGM96a] , Hypercomplex structures on Stiefel manifolds, Ann. Global Anal. Geom.
14 (1996), no. 1, 81105. MR 97e:53084
[BGM96b] , On strongly inhomogeneous Einstein manifolds, Bull. London Math. Soc.
28 (1996), no. 4, 401408. MR 97a:53069
[BGM98a] , Hypercomplex structures from 3-Sasakian structures, J. Reine Angew.
Math. 501 (1998), 115141. MR 99g:53049
[BGM98b] , A note on smooth toral reductions of spheres, Manuscripta Math. 95
(1998), no. 2, 149158. MR 99c:53042
[BGM06] C. P. Boyer, K. Galicki, and P. Matzeu, On eta-Einstein Sasakian geometry, Comm.
Math. Phys. 262 (2006), no. 1, 177208. MR MR2200887
[BGMR98] C. P. Boyer, K. Galicki, B. M. Mann, and E. G. Rees, Compact 3-Sasakian 7-
manifolds with arbitrary second Betti number, Invent. Math. 131 (1998), no. 2,
321344. MR 99b:53066
[BGN02a] C. P. Boyer, K. Galicki, and M. Nakamaye, Einstein metrics on rational homology 7-
spheres, Ann. Inst. Fourier (Grenoble) 52 (2002), no. 5, 15691584. MR 2003j:53056
[BGN02b] , Sasakian-Einstein structures on 9#(S
2
S
3
), Trans. Amer. Math. Soc.
354 (2002), no. 8, 29832996 (electronic). MR 2003g:53061
[BGN03a] , On positive Sasakian geometry, Geom. Dedicata 101 (2003), 93102.
MR MR2017897 (2005a:53072)
[BGN03b] , On the geometry of Sasakian-Einstein 5-manifolds, Math. Ann. 325
(2003), no. 3, 485524. MR 2004b:53061
[BGN03c] , Sasakian geometry, homotopy spheres and positive Ricci curvature, Topol-
ogy 42 (2003), no. 5, 9811002. MR 1 978 045
[BGO07] C. P. Boyer, K. Galicki, and L. Ornea, Constructions in Sasakian Geometry,
preprint, arXiv:math.DG/0602233, to appear in Mathematische Zeitscrift (2007).
[BGP02] C. P. Boyer, K. Galicki, and P. Piccinni, 3-Sasakian geometry, nilpotent orbits,
and exceptional quotients, Ann. Global Anal. Geom. 21 (2002), no. 1, 85110.
MR 2003d:53076
[BGS06] C. P. Boyer, K. Galicki, and S. Simanca, Canonical Sasakian metrics,
math.DG/0604325 (2006).
[BH58] A. Borel and F. Hirzebruch, Characteristic classes and homogeneous spaces. I,
Amer. J. Math. 80 (1958), 458538. MR 0102800 (21 #1586)
[BH99] M. R. Bridson and A. Haeiger, Metric spaces of non-positive curvature,
Grundlehren der Mathematischen Wissenschaften [Fundamental Principles of
Mathematical Sciences], vol. 319, Springer-Verlag, Berlin, 1999. MR 1744486
(2000k:53038)
[BH02] A. Bergman and C. P. Herzog, The volume of some non-spherical horizons and
the AdS/CFT correspondence, J. High Energy Phys. (2002), no. 1, No. 30, 24.
MR 1900111 (2003e:83065)
[BHK05] S. Benvenuti, A. Hanany, and P. Kazakopoulos, The toric phases of the Y
p,q
quiv-
ers, J. High Energy Phys. (2005), no. 7, 021, 13 pp. (electronic). MR 2163034
(2006h:81277)
[BHPVdV04] W. P. Barth, K. Hulek, C. A. M. Peters, and A. Van de Ven, Compact complex
surfaces, second ed., Ergebnisse der Mathematik und ihrer Grenzgebiete. 3. Folge.
A Series of Modern Surveys in Mathematics [Results in Mathematics and Related
Areas. 3rd Series. A Series of Modern Surveys in Mathematics], vol. 4, Springer-
Verlag, Berlin, 2004. MR 2030225 (2004m:14070)
[Bie96] R. Bielawski, On the hyper-K ahler metrics associated to singularities of nilpotent
varieties, Ann. Global Anal. Geom. 14 (1996), no. 2, 177191. MR 97e:53083
[Bie97] , Betti numbers of 3-Sasakian quotients of spheres by tori, Bull. London
Math. Soc. 29 (1997), no. 6, 731736. MR 98j:53051
BIBLIOGRAPHY 575
[Bie99] , Complete hyper-K ahler 4n-manifolds with a local tri-Hamiltonian R
n
-
action, Math. Ann. 314 (1999), no. 3, 505528. MR 2001c:53058
[Biq96] O. Biquard, Sur les equations de Nahm et la structure de Poisson des alg`ebres de
Lie semi-simples complexes, Math. Ann. 304 (1996), no. 2, 253276. MR 1371766
(97c:53066)
[Biq99] , Quaternionic contact structures, Quaternionic structures in mathematics
and physics (Rome, 1999), Univ. Studi Roma La Sapienza, Rome, 1999, pp. 2330
(electronic). MR 1848655 (2002h:53079)
[Biq00] , Metriques dEinstein asymptotiquement symetriques, Asterisque (2000),
no. 265, vi+109. MR 1760319 (2001k:53079)
[Biq02] , Metriques autoduales sur la boule, Invent. Math. 148 (2002), no. 3, 545
607. MR 1908060 (2003i:53058)
[Bis07] L. Bisconti, Positive self-dual Einstein orbifolds with one-dimensional isometry
group.
[BK78] J.-P. Bourguignon and H. Karcher, Curvature operators: pinching estimates and
geometric examples, Ann. Sci.

Ecole Norm. Sup. (4) 11 (1978), no. 1, 7192. MR 58
#12829
[BK06] C. Bohm and M. M. Kerr, Low-dimensional homogeneous Einstein manifolds,
Trans. Amer. Math. Soc. 358 (2006), no. 4, 14551468 (electronic). MR 2186982
(2006g:53056)
[BL70] J.-P. Buet and J.-C. Lor, Une construction dun universal pour une classe assez
large de -structures, C. R. Acad. Sci. Paris Ser. A-B 270 (1970), A640A642.
MR 0271942 (42 #6823)
[Bla70] D. E. Blair, Geometry of manifolds with structural group U(n)O(s), J. Dierential
Geometry 4 (1970), 155167. MR 0267501 (42 #2403)
[Bla76a] , Contact manifolds in Riemannian geometry, Springer-Verlag, Berlin, 1976,
Lecture Notes in Mathematics, Vol. 509. MR 57 #7444
[Bla76b] , On the non-existence of at contact metric structures, Tohoku Math. J.
(2) 28 (1976), no. 3, 373379. MR 55 #4029
[Bla77] , Two remarks on contact metric structures, Tohoku Math. J. (2) 29 (1977),
no. 3, 319324. MR 57 #4043
[Bla02] , Riemannian geometry of contact and symplectic manifolds, Progress
in Mathematics, vol. 203, Birkhauser Boston Inc., Boston, MA, 2002.
MR 2002m:53120
[BLMPZ05] B. A. Burrington, J. T. Liu, M. Mahato, and L. A. Pando Zayas, Towards su-
pergravity duals of chiral symmetry breaking in Sasaki-Einstein cascading quiver
theories, J. High Energy Phys. (2005), no. 7, 019, 24 pp. (electronic). MR 2163035
[BM87] S. Bando and T. Mabuchi, Uniqueness of Einstein Kahler metrics modulo connected
group actions, Algebraic geometry, Sendai, 1985, Adv. Stud. Pure Math., vol. 10,
North-Holland, Amsterdam, 1987, pp. 1140. MR 89c:53029
[BM93a] A. Banyaga and P. Molino, Geometrie des formes de contact compl`etement
integrables de type toriques, Seminaire Gaston Darboux de Geometrie et Topolo-
gie Dierentielle, 19911992 (Montpellier), Univ. Montpellier II, Montpellier, 1993,
pp. 125. MR 94e:53029
[BM93b] C. P. Boyer and B. M. Mann, The hyper-K ahler geometry of the ADHM construc-
tion and quaternionic geometric invariant theory, Dierential geometry: geome-
try in mathematical physics and related topics (Los Angeles, CA, 1990), Proc.
Sympos. Pure Math., vol. 54, Amer. Math. Soc., Providence, RI, 1993, pp. 4583.
MR 1216529 (94d:58026)
[BM96] A. Banyaga and P. Molino, Complete integrability in contact geometry, preprint
(1996), 125.
[BM03] A. Bilal and S. Metzger, Compact weak G
2
-manifolds with conical singularities,
Nuclear Phys. B 663 (2003), no. 1-2, 343364. MR 1986853 (2004d:53055)
[Boh98] C. Bohm, Inhomogeneous Einstein metrics on low-dimensional spheres and other
low-dimensional spaces, Invent. Math. 134 (1998), no. 1, 145176. MR 99i:53046
[Boh03] C. Bohle, Killing spinors on Lorentzian manifolds, J. Geom. Phys. 45 (2003), no. 3-
4, 285308. MR 1 952 661
576 BIBLIOGRAPHY
[Bon64] E. Bonan, Tenseur de structure dune variete presque quaternionienne, C. R. Acad.
Sci. Paris 259 (1964), 4548. MR 0166741 (29 #4014)
[Bon66] , Sur des varietes riemanniennes `a groupe dholonomie G
2
ou spin (7), C.
R. Acad. Sci. Paris Ser. A-B 262 (1966), A127A129. MR 0196668 (33 #4855)
[Bon67] , Sur les G-structures de type quaternionien, Cahiers Topologie Geom.
Dierentielle 9 (1967), 389461. MR 38 #1624
[Bon82] , Sur lalg`ebre exterieure dune variete presque hermitienne quaternionique,
C. R. Acad. Sci. Paris Ser. I Math. 295 (1982), no. 2, 115118. MR 676377
(84e:53077)
[Boo61] W. M. Boothby, Homogeneous complex contact manifolds, Proc. Sympos. Pure
Math., Vol. III, American Mathematical Society, Providence, R. I., 1961, pp. 144
154. MR 0124863 (23 #A2173)
[Boo80] , On compact homogeneous contact manifolds, Bull. Inst. Math. Acad. Sinica
8 (1980), no. 2-3, part 2, 341351. MR 595542 (82a:53052)
[Bor54] A. Borel, Kahlerian coset spaces of semisimple Lie groups, Proc. Nat. Acad. Sci.
U. S. A. 40 (1954), 11471151. MR 17,1108e
[Bor60] , Density properties for certain subgroups of semi-simple groups without
compact components, Ann. of Math. (2) 72 (1960), 179188. MR 0123639 (23
#A964)
[Bot54] R. Bott, Nondegenerate critical manifolds, Ann. of Math. (2) 60 (1954), 248261.
MR 0064399 (16,276f)
[Bou81] J.-P. Bourguignon, Les varietes de dimension 4 `a signature non nulle dont la cour-
bure est harmonique sont dEinstein, Invent. Math. 63 (1981), no. 2, 263286.
MR 610539 (82g:53051)
[Bou89] N. Bourbaki, Commutative algebra. Chapters 17, Elements of Mathematics
(Berlin), Springer-Verlag, Berlin, 1989, Translated from the French, Reprint of the
1972 edition. MR 979760 (90a:13001)
[Bou02] F. Bourgeois, Odd dimensional tori are contact manifolds, Int. Math. Res. Not.
(2002), no. 30, 15711574. MR 2003f:53157
[Boy86] C. P. Boyer, Conformal duality and compact complex surfaces, Math. Ann. 274
(1986), no. 3, 517526. MR 87i:53068
[Boy88a] , A note on hyper-Hermitian four-manifolds, Proc. Amer. Math. Soc. 102
(1988), no. 1, 157164. MR 915736 (89c:53049)
[Boy88b] , Self-dual and anti-self-dual Hermitian metrics on compact complex sur-
faces, Mathematics and general relativity (Santa Cruz, CA, 1986), Contemp.
Math., vol. 71, Amer. Math. Soc., Providence, RI, 1988, pp. 105114. MR 954411
(89h:53127)
[BPVdV84] W. Barth, C. Peters, and A. Van de Ven, Compact complex surfaces, Ergebnisse
der Mathematik und ihrer Grenzgebiete (3) [Results in Mathematics and Related
Areas (3)], vol. 4, Springer-Verlag, Berlin, 1984. MR 86c:32026
[BR62] A. Borel and R. Remmert,

Uber kompakte homogene Kahlersche Mannigfaltigkeiten,
Math. Ann. 145 (1961/1962), 429439. MR 0145557 (26 #3088)
[BR86] M. Beltrametti and L. Robbiano, Introduction to the theory of weighted projective
spaces, Exposition. Math. 4 (1986), no. 2, 111162. MR 88h:14052
[Bre72] G. E. Bredon, Introduction to compact transformation groups, Academic Press,
New York, 1972, Pure and Applied Mathematics, Vol. 46. MR 54 #1265
[Bre93] , Topology and geometry, Graduate Texts in Mathematics, vol. 139, Springer-
Verlag, New York, 1993. MR 94d:55001
[Bre97] , Sheaf theory, second ed., Graduate Texts in Mathematics, vol. 170,
Springer-Verlag, New York, 1997. MR MR1481706 (98g:55005)
[Bri66] E. Brieskorn, Beispiele zur Dierentialtopologie von Singularitaten, Invent. Math.
2 (1966), 114. MR 34 #6788
[Bri68] , Rationale Singularitaten komplexer Flachen, Invent. Math. 4 (1967/1968),
336358. MR 0222084 (36 #5136)
[Bri00] , Singularities in the work of Friedrich Hirzebruch, Surveys in dierential
geometry, Surv. Dier. Geom., VII, Int. Press, Somerville, MA, 2000, pp. 1760.
MR 1919421 (2003g:32002)
BIBLIOGRAPHY 577
[Bro73] W. Browder, Cobordism invariants, the Kervaire invariant and xed point free
involutions, Trans. Amer. Math. Soc. 178 (1973), 193225. MR 0324717 (48 #3067)
[Bro82] K. S. Brown, Cohomology of groups, Graduate Texts in Mathematics, vol. 87,
Springer-Verlag, New York, 1982. MR 672956 (83k:20002)
[Bru70] G. Brumel, The homotopy groups of BPL and PL/O. III, Michigan Math. J. 17
(1970), 217224. MR 0271938 (42 #6819)
[Bry87] R. L. Bryant, Metrics with exceptional holonomy, Ann. of Math. (2) 126 (1987),
no. 3, 525576. MR 89b:53084
[Bry91] , Two exotic holonomies in dimension four, path geometries, and twistor
theory, Complex geometry and Lie theory (Sundance, UT, 1989), Proc. Sym-
pos. Pure Math., vol. 53, Amer. Math. Soc., Providence, RI, 1991, pp. 3388.
MR 1141197 (93e:53030)
[Bry96] , Classical, exceptional, and exotic holonomies: a status report, Actes de
la Table Ronde de Geometrie Dierentielle (Luminy, 1992), Semin. Congr., vol. 1,
Soc. Math. France, Paris, 1996, pp. 93165. MR 1427757 (98c:53037)
[Bry01] , Bochner-K ahler metrics, J. Amer. Math. Soc. 14 (2001), no. 3, 623715
(electronic). MR 1824987 (2002i:53096)
[BS89] R. L. Bryant and S. M. Salamon, On the construction of some complete metrics
with exceptional holonomy, Duke Math. J. 58 (1989), no. 3, 829850. MR 90i:53055
[BS99] V. V. Batyrev and E. N. Selivanova, Einstein-Kahler metrics on symmetric toric
Fano manifolds, J. Reine Angew. Math. 512 (1999), 225236. MR 2000j:32038
[BT82] R. Bott and L. W. Tu, Dierential forms in algebraic topology, Graduate Texts in
Mathematics, vol. 82, Springer-Verlag, New York, 1982. MR 658304 (83i:57016)
[But05] J.-B. Butruille, Classication des varietes approximativement kahleriennes ho-
mog`enes, Ann. Global Anal. Geom. 27 (2005), no. 3, 201225. MR 2158165
(2006f:53060)
[But06] , Homogeneous nearly Kahler manifolds, preprint; arXiv:math.DG/0512655
(2006).
[BV04] N. Blazic and S. Vukmirovic, Examples of self-dual, Einstein metrics of (2, 2)-
signature, Math. Scand. 94 (2004), no. 1, 6374. MR 2032336 (2004j:53056)
[BW58] W. M. Boothby and H. C. Wang, On contact manifolds, Ann. of Math. (2) 68
(1958), 721734. MR 22 #3015
[BW83] J. Bagger and E. Witten, Matter couplings in N = 2 supergravity, Nuclear Phys. B
222 (1983), no. 1, 110. MR 708440 (84k:83050)
[BZ05] A. Butti and A. Zaaroni, R-charges from toric diagrams and the equivalence of
a-maximization and Z-minimization, J. High Energy Phys. (2005), no. 11, 019, 42
pp. (electronic). MR 2187558
[Cal56] E. Calabi, The space of kahler metrics, Proceedings of the International Congress
of Mathematicians, Vol. 1, 2 (Amsterdam 2, 1954) (Amsterdam), North-Holland,
1956, pp. 206207.
[Cal57] , On Kahler manifolds with vanishing canonical class, Algebraic geometry
and topology. A symposium in honor of S. Lefschetz, Princeton University Press,
Princeton, N. J., 1957, pp. 7889. MR 19,62b
[Cal79] , Metriques kahleriennes et bres holomorphes, Ann. Sci.

Ecole Norm. Sup.
(4) 12 (1979), no. 2, 269294. MR 83m:32033
[Cal82] , Extremal Kahler metrics, Seminar on Dierential Geometry, Ann. of
Math. Stud., vol. 102, Princeton Univ. Press, Princeton, N.J., 1982, pp. 259290.
MR 83i:53088
[Cal85] , Extremal Kahler metrics. II, Dierential geometry and complex analysis,
Springer, Berlin, 1985, pp. 95114. MR 780039 (86h:53067)
[Cam91] F. Campana, Une version geometrique generalisee du theor`eme du produit de
Nadel, Bull. Soc. Math. France 119 (1991), no. 4, 479493. MR 1136848 (93h:14029)
[Cam92] , Connexite rationnelle des varietes de Fano, Ann. Sci.

Ecole Norm. Sup.
(4) 25 (1992), no. 5, 539545. MR 1191735 (93k:14050)
[Cam04] , Orbifolds, special varieties and classication theory, Ann. Inst. Fourier
(Grenoble) 54 (2004), no. 3, 499630. MR 2097416
[Car84a] Y. Carri`ere, Flots riemanniens, Asterisque (1984), no. 116, 3152, Transversal
structure of foliations (Toulouse, 1982). MR 86m:58125a
578 BIBLIOGRAPHY
[Car84b] , Les proprietes topologiques des ots riemanniens retrouvees `a laide
du theor`eme des varietes presque plates, Math. Z. 186 (1984), no. 3, 393400.
MR 86f:53048
[CC00] A. Candel and L. Conlon, Foliations. I, Graduate Studies in Mathematics, vol. 23,
American Mathematical Society, Providence, RI, 2000. MR 2002f:57058
[CC03] , Foliations. II, Graduate Studies in Mathematics, vol. 60, American Math-
ematical Society, Providence, RI, 2003. MR 1 994 394
[CDF91] L. Castellani, R. DAuria, and P. Fre, Supergravity and superstrings. A geomet-
ric perspective. Vol. 1-3, World Scientic Publishing Co. Inc., Teaneck, NJ, 1991,
Mathematical foundations. MR 1120022 (92k:83127a)
[CDG02] D. M. J. Calderbank, L. David, and P. Gauduchon, The Guillemin formula and
K ahler metrics on toric symplectic manifolds, J. Symplectic Geom. 1 (2002), no. 4,
767784. MR 2039163 (2005a:53145)
[CdS01] A. Cannas da Silva, Lectures on symplectic geometry, Lecture Notes in Mathemat-
ics, vol. 1764, Springer-Verlag, Berlin, 2001. MR 1853077 (2002i:53105)
[CdS03] , Symplectic toric manifolds, Symplectic geometry of integrable Hamiltonian
systems (Barcelona, 2001), Adv. Courses Math. CRM Barcelona, Birkhauser, Basel,
2003, pp. 85173. MR 2000746
[CdSW99] A. Cannas da Silva and A. Weinstein, Geometric models for noncommutative alge-
bras, Berkeley Mathematics Lecture Notes, vol. 10, American Mathematical Society,
Providence, RI, 1999. MR 2001m:58013
[CEZ07] T. Chinburg, Ch. Escher, and W. Ziller, Topological properties of Eschenburg spaces
and 3-Sasakian manifolds, Mathematische Annalen (2007), to appear.
[CFdL85] L. A. Cordero, M. Fernandez, and M. de Leon, Examples of compact non-Kahler
almost Kahler manifolds, Proc. Amer. Math. Soc. 95 (1985), no. 2, 280286.
MR 86m:53049
[CFG86] L. A. Cordero, M. Fernandez, and A. Gray, Symplectic manifolds with no Kahler
structure, Topology 25 (1986), no. 3, 375380. MR 87j:53051
[CFG93] , The failure of complex and symplectic manifolds to be Kahlerian, Dieren-
tial geometry: geometry in mathematical physics and related topics (Los Angeles,
CA, 1990), Proc. Sympos. Pure Math., vol. 54, Amer. Math. Soc., Providence, RI,
1993, pp. 107123. MR 94e:53071
[CFO07] K. Cho, A. Futaki, and H. Ono, Uniqueness and examples of compact toric Sasaki-
Einstein metrics, preprint; arXiv:math.DG/0701122 (2007).
[CG90] D. Chinea and C. Gonzalez, A classication of almost contact metric manifolds,
Ann. Mat. Pura Appl. (4) 156 (1990), 1536. MR 1080209 (91i:53047)
[CGLP02] M. Cvetic, G. W. Gibbons, H. Lu, and C. N. Pope, Cohomogeneity one manifolds of
Spin(7) and G
2
holonomy, Ann. Physics 300 (2002), no. 2, 139184. MR 1921784
(2003m:53074)
[CGLP04] M. Cvetic, G. W. Gibbons, H. L u, and C. N. Pope, New cohomogeneity one metrics
with Spin(7) holonomy, J. Geom. Phys. 49 (2004), no. 3-4, 350365. MR 2076750
(2005f:53074)
[CH05] S. A. Cherkis and N. J. Hitchin, Gravitational instantons of type D
k
, Comm. Math.
Phys. 260 (2005), no. 2, 299317. MR 2177322 (2007a:53095)
[Che46] S. S. Chern, Characteristic classes of Hermitian manifolds, Ann. of Math. (2) 47
(1946), 85121. MR 0015793 (7,470b)
[Che78] B.-Y. Chen, Some topological obstructions to Bochner-Kaehler metrics and their
applications, J. Dierential Geom. 13 (1978), no. 4, 547558 (1979). MR 570217
(81f:32037)
[Che03] J. Cheeger, Degeneration of Einstein metrics and metrics with special holonomy,
Surveys in dierential geometry, Vol. VIII (Boston, MA, 2002), Surv. Dier. Geom.,
VIII, Int. Press, Somerville, MA, 2003, pp. 2973. MR 2039985 (2005e:53060)
[Che07] I. Cheltsov, Fano varieties with many selfmaps, preprint; arXiv:math.AG/0611881
(2007).
[CJS79] E. Cremmer, B. Julia, and J. Scherk, Supergravity in eleven dimensions, Phys. Lett.
B 76 (1979), 409412.
BIBLIOGRAPHY 579
[CK99] D. A. Cox and S. Katz, Mirror symmetry and algebraic geometry, Mathematical
Surveys and Monographs, vol. 68, American Mathematical Society, Providence, RI,
1999. MR 1677117 (2000d:14048)
[CL85] C. Camacho and N. A. Lins, Geometric theory of foliations, Birkhauser Boston
Inc., Boston, MA, 1985, Translated from the Portuguese by Sue E. Goodman.
MR 87a:57029
[CL97] F. Catanese and C. LeBrun, On the scalar curvature of Einstein manifolds, Math.
Res. Lett. 4 (1997), no. 6, 843854. MR 1492124 (98k:53057)
[CLOT03] O. Cornea, G. Lupton, J. Oprea, and D. Tanre, Lusternik-Schnirelmann category,
Mathematical Surveys and Monographs, vol. 103, American Mathematical Society,
Providence, RI, 2003. MR 1 990 857
[CLPP05] M. Cvetic, H. L u, D. N. Page, and C. N. Pope, New Einstein-Sasaki spaces in ve
and higher dimensions, Phys. Rev. Lett. 95 (2005), no. 7, 071101, 4. MR 2167018
[CLS90] P. Candelas, M. Lynker, and R. Schimmrigk, Calabi-Yau manifolds in weighted P
4
,
Nuclear Phys. B 341 (1990), no. 2, 383402. MR 91m:14062
[CM74] S. S. Chern and J. K. Moser, Real hypersurfaces in complex manifolds, Acta Math.
133 (1974), 219271. MR 0425155 (54 #13112)
[CM92] D. Chinea and J. C. Marrero, Classication of almost contact metric structures,
Rev. Roumaine Math. Pures Appl. 37 (1992), no. 3, 199211. MR 1172273
(93g:53044)
[CM00] M. Crainic and I. Moerdijk, A homology theory for etale groupoids, J. Reine Angew.
Math. 521 (2000), 2546. MR 1752294 (2001f:58039)
[CMS96] F. M. Cabrera, M. D. Monar, and A. F. Swann, Classication of G
2
-structures, J.
London Math. Soc. (2) 53 (1996), no. 2, 407416. MR 97e:53047
[Con74] L. Conlon, Transversally parallelizable foliations of codimension two, Trans. Amer.
Math. Soc. 194 (1974), 79102, erratum, ibid. 207 (1975), 406. MR 51 #6844
[Cor96] V. Cortes, Alekseevskian spaces, Dierential Geom. Appl. 6 (1996), no. 2, 129168.
MR 1395026 (97m:53079)
[Cor00] , A new construction of homogeneous quaternionic manifolds and related
geometric structures, Mem. Amer. Math. Soc. 147 (2000), no. 700, viii+63.
MR 1708628 (2001e:53047)
[CP02] D. M. J. Calderbank and H. Pedersen, Selfdual Einstein metrics with torus symme-
try, J. Dierential Geom. 60 (2002), no. 3, 485521. MR 1950174 (2003m:53065)
[CR04] W. Chen and Y. Ruan, A new cohomology theory of orbifold, Comm. Math. Phys.
248 (2004), no. 1, 131. MR MR2104605 (2005j:57036)
[Cro01] D. Crowley, The classication of highly connected manifolds in dimension 7 and
15, Indiana University Thesis (2001).
[CS04] D. M. J. Calderbank and M. A. Singer, Einstein metrics and complex singularities,
Invent. Math. 156 (2004), no. 2, 405443. MR 2052611 (2005h:53064)
[CS06a] , Toric self-dual Einstein metrics on compact orbifolds, Duke Math. J. 133
(2006), no. 2, 237258. MR 2225692
[CS06b] D. Conti and S. Salamon, Generalized Killing spinors in dimension 5, preprint;
arXiv:math.DG/0508375, to appear in Trans. Amer. Math. Soc. (2006).
[CT06] J. Cheeger and G. Tian, Curvature and injectivity radius estimates for Einstein 4-
manifolds, J. Amer. Math. Soc. 19 (2006), no. 2, 487525 (electronic). MR 2188134
(2006i:53042)
[

CV98] M.

Cadek and J. Vanzura, Almost quaternionic structures on eight-manifolds, Os-
aka J. Math. 35 (1998), no. 1, 165190. MR 1621248 (99f:57031)
[CZ06] H.-D. Cao and X.-P. Zhou, The complete proof of the Poincare and geometrization
conjectures application of the Hamilton-Perelman theory of the Ricci ow, Asian
J. Math. 10 (2006), no. 2, 165492.
[Dan93] A. S. Dancer, Dihedral singularities and gravitational instantons, J. Geom. Phys.
12 (1993), no. 2, 7791. MR 1231230 (94g:53038)
[Dan94] , A family of hyper-K ahler manifolds, Quart. J. Math. Oxford Ser. (2) 45
(1994), no. 180, 463478. MR 1315458 (96a:53059)
[Dan96] , Scalar-at Kahler metrics with SU(2) symmetry, J. Reine Angew. Math.
479 (1996), 99120. MR 1414390 (97m:53080)
580 BIBLIOGRAPHY
[Dan99] , Hyper-Kahler manifolds, Surveys in dierential geometry: essays on Ein-
stein manifolds, Surv. Dier. Geom., VI, Int. Press, Boston, MA, 1999, pp. 1538.
MR 1798605 (2001k:53087)
[Dav05] L. David, The Bochner-at cone of a CR manifold, preprint; math.DG/0512604
(2005).
[Dea04] O. Dearricott, Positive sectional curvature on 3-Sasakian manifolds, Ann. Global
Anal. Geom. 25 (2004), no. 1, 5972. MR 2033140 (2005f:53060)
[Dea05] , Positively curved self-dual Einstein metrics on weighted projective planes,
Ann. Global Anal. Geom. 27 (2005), no. 1, 7986. MR 2130534
[DEF
+
99] P. Deligne, P. Etingof, D. S. Freed, L. C. Jerey, D. Kazhdan, J. W. Morgan,
D. R. Morrison, and E. Witten (eds.), Quantum elds and strings: a course for
mathematicians. Vol. 1, 2, American Mathematical Society, Providence, RI, 1999,
Material from the Special Year on Quantum Field Theory held at the Institute for
Advanced Study, Princeton, NJ, 19961997. MR 1701618 (2000e:81010)
[Del75] C. Delorme, Espaces projectifs anisotropes, Bull. Soc. Math. France 103 (1975),
no. 2, 203223. MR 0404277 (53 #8080a)
[Del88] T. Delzant, Hamiltoniens periodiques et images convexes de lapplication moment,
Bull. Soc. Math. France 116 (1988), no. 3, 315339. MR 984900 (90b:58069)
[Dem96] J.-P. Demailly, L
2
vanishing theorems for positive line bundles and adjunction the-
ory, Transcendental methods in algebraic geometry (Cetraro, 1994), Lecture Notes
in Math., vol. 1646, Springer, Berlin, 1996, pp. 197. MR 1603616 (99k:32051)
[Dem01] , Multiplier ideal sheaves and analytic methods in algebraic geometry, School
on Vanishing Theorems and Eective Results in Algebraic Geometry (Trieste, 2000),
ICTP Lect. Notes, vol. 6, Abdus Salam Int. Cent. Theoret. Phys., Trieste, 2001,
pp. 1148. MR 1919457 (2003f:32020)
[Der83] A. Derdzi nski, Self-dual Kahler manifolds and Einstein manifolds of dimension
four, Compositio Math. 49 (1983), no. 3, 405433. MR 707181 (84h:53060)
[DF89] S. Donaldson and R. Friedman, Connected sums of self-dual manifolds and de-
formations of singular spaces, Nonlinearity 2 (1989), no. 2, 197239. MR 994091
(90e:32027)
[DF06] S. Donaldson and J. Fine, Toric anti-self-dual 4-manifolds via complex geometry,
Math. Ann. 336 (2006), no. 2, 281309. MR 2244374
[DGMS75] P. Deligne, P. Griths, J. Morgan, and D. Sullivan, Real homotopy theory of Kahler
manifolds, Invent. Math. 29 (1975), no. 3, 245274. MR 52 #3584
[Dim92] A. Dimca, Singularities and topology of hypersurfaces, Universitext, Springer-
Verlag, New York, 1992. MR 94b:32058
[DK75] A. Durfee and L. Kauman, Periodicity of branched cyclic covers, Math. Ann. 218
(1975), no. 2, 157174. MR 52 #6731
[DK79] T. Duchamp and M. Kalka, Deformation theory for holomorphic foliations, J. Dif-
ferential Geom. 14 (1979), no. 3, 317337 (1980). MR 82b:57019
[DK80] , Stability theorems for holomorphic foliations, Trans. Amer. Math. Soc. 260
(1980), no. 1, 255266. MR 81f:57022
[DK84] , Holomorphic foliations and deformations of the Hopf foliation, Pacic J.
Math. 112 (1984), no. 1, 6981. MR 85m:57019
[DK90] S. K. Donaldson and P. B. Kronheimer, The geometry of four-manifolds, Oxford
Mathematical Monographs, The Clarendon Press Oxford University Press, New
York, 1990, Oxford Science Publications. MR 1079726 (92a:57036)
[DK00] J. J. Duistermaat and J. A. C. Kolk, Lie groups, Universitext, Springer-Verlag,
Berlin, 2000. MR 2001j:22008
[DK01] J-P. Demailly and J. Kollar, Semi-continuity of complex singularity exponents and
K ahler-Einstein metrics on Fano orbifolds, Ann. Sci.

Ecole Norm. Sup. (4) 34
(2001), no. 4, 525556. MR 2002e:32032
[DMR82] A. Daz Miranda and A. Reventos, Homogeneous contact compact manifolds and
homogeneous symplectic manifolds, Bull. Sci. Math. (2) 106 (1982), no. 4, 337350.
MR 688194 (84d:53034)
[dMT97] S. F. B. de Moraes and C. Tomei, Moment maps on symplectic cones, Pacic J.
Math. 181 (1997), no. 2, 357375. MR 98m:58038
BIBLIOGRAPHY 581
[DO98] S. Dragomir and L. Ornea, Locally conformal Kahler geometry, Progress in Math-
ematics, vol. 155, Birkhauser Boston Inc., Boston, MA, 1998. MR 1481969
(99a:53081)
[DO06] O. Dragulete and L. Ornea, Non-zero contact and Sasakian reduction, Dierential
Geom. Appl. 24 (2006), no. 3, 260270. MR MR2216940
[Dol53] P. Dolbeault, Sur la cohomologie des varietes analytiques complexes, C. R. Acad.
Sci. Paris 236 (1953), 175177. MR 0052771 (14,673a)
[Dol63] A. Dold, Partitions of unity in the theory of brations, Ann. of Math. (2) 78 (1963),
223255. MR 0155330 (27 #5264)
[Dol82] I. Dolgachev, Weighted projective varieties, Group actions and vector elds (Van-
couver, B.C., 1981), Lecture Notes in Math., vol. 956, Springer, Berlin, 1982, pp. 34
71. MR 85g:14060
[DOR02] O. Dragulete, L. Ornea, and T. S. Ratiu, Cosphere bundle reduction in contact ge-
ometry, J. Symplectic Geom. 1 (2002), no. 4, 695714. MR 2039161 (2004m:53141)
[DP04] J.-P. Demailly and M. Paun, Numerical characterization of the Kahler cone of
a compact Kahler manifold, Ann. of Math. (2) 159 (2004), no. 3, 12471274.
MR 2113021
[DS97a] A. Dancer and A. Swann, The geometry of singular quaternionic Kahler quotients,
Internat. J. Math. 8 (1997), no. 5, 595610. MR 1468352 (98m:53062)
[DS97b] A. S. Dancer and R. Szoke, Symmetric spaces, adapted complex structures and
hyper-Kahler structures, Quart. J. Math. Oxford Ser. (2) 48 (1997), no. 189, 2738.
MR 1439696 (98e:53083)
[DT92] W. Y. Ding and G. Tian, Kahler-Einstein metrics and the generalized Futaki in-
variant, Invent. Math. 110 (1992), no. 2, 315335. MR 93m:53039
[DT06] S. Dragomir and G. Tomassini, Dierential geometry and analysis on CR manifolds,
Progress in Mathematics, vol. 246, Birkhauser Boston Inc., Boston, MA, 2006.
MR 2214654
[Duc06] D. Duchemin, Quaternionic contsct structures in dimension 7, Ann. Inst. Fourier
(Grenoble) 56 (2006), no. 4, 851885.
[Dur71] A. H. Durfee, Dieomorphism Classication of Isolated Hypersurfaces Singularities,
Cornell University Thesis (1971).
[Dur77] , Bilinear and quadratic forms on torsion modules, Advances in Math. 25
(1977), no. 2, 133164. MR 0480333 (58 #506)
[dWLP
+
84] B. de Wit, P. G. Lauwers, R. Philippe, S. Q. Su, and A. Van Proeyen, Gauge and
matter elds coupled to N = 2 supergravity, Phys. Lett. B 134 (1984), no. 1-2,
3743. MR 729799 (84m:83067)
[dWLVP85] B. de Wit, Lauwers, and A. Van Proeyen, Lagrangians in N = 2 supergravity-matter
systems, Nucl. Phys. B 255 (1985), 569601.
[dWVP92] B. de Wit and A. Van Proeyen, Special geometry, cubic polynomials and homo-
geneous quaternionic spaces, Comm. Math. Phys. 149 (1992), no. 2, 307333.
MR 1186031 (94a:53079)
[EH79] T. Eguchi and A. J. Hanson, Self-dual solutions to Euclidean gravity, Ann. Physics
120 (1979), no. 1, 82106. MR 540896 (80h:83011)
[Ehr53] C. Ehresmann, Introduction `a la theorie des structures innitesimales et des pseu-
dogroupes de Lie, Colloque de topologie et geometrie dierentielle, Strasbourg,
1952, no. 11, La Biblioth`eque Nationale et Universitaire de Strasbourg, 1953, p. 16.
MR 15,828e
[EKA90] A. El Kacimi-Alaoui, Operateurs transversalement elliptiques sur un feuilletage rie-
mannien et applications, Compositio Math. 73 (1990), no. 1, 57106. MR 91f:58089
[EKAH86] A. El Kacimi-Alaoui and G. Hector, Decomposition de Hodge basique pour un
feuilletage riemannien, Ann. Inst. Fourier (Grenoble) 36 (1986), no. 3, 207227.
MR 87m:57029
[EKAN93] A. El Kacimi-Alaoui and M. Nicolau, On the topological invariance of the basic
cohomology, Math. Ann. 295 (1993), no. 4, 627634. MR 1214951 (94c:57046)
[Eli89] Y. Eliashberg, Classication of overtwisted contact structures on 3-manifolds, In-
vent. Math. 98 (1989), no. 3, 623637. MR 1022310 (90k:53064)
[Eli92] , Contact 3-manifolds twenty years since J. Martinets work, Ann. Inst.
Fourier (Grenoble) 42 (1992), no. 1-2, 165192. MR 1162559 (93k:57029)
582 BIBLIOGRAPHY
[Eli98] , Invariants in contact topology, Proceedings of the International Congress
of Mathematicians, Vol. II (Berlin, 1998), no. Extra Vol. II, 1998, pp. 327338
(electronic). MR 2000a:57068
[Eps72] D. B. A. Epstein, Periodic ows on three-manifolds, Ann. of Math. (2) 95 (1972),
6682. MR 0288785 (44 #5981)
[ER44] C. Ehresmann and G. Reeb, Sur les champs delements de contact de dimension
p compl`etement integrables dans une variete continuement dierentiable Vn, C. R.
Acad. Sci. Paris 218 (1944), 955957. MR 0014757 (7,327g)
[Esc82] J.-H. Eschenburg, New examples of manifolds with strictly positive curvature, In-
vent. Math. 66 (1982), no. 3, 469480. MR 83i:53061
[Esc84] , Freie isometrische Aktionen auf kompakten Lie-Gruppen mit positiv
gekr ummten Orbitraumen, Schriftenreihe des Mathematischen Instituts der Uni-
versit at M unster, 2. Serie [Series of the Mathematical Institute of the University of
M unster, Series 2], vol. 32, Universit at M unster Mathematisches Institut, M unster,
1984. MR 86a:53045
[Esc92] , Cohomology of biquotients, Manuscripta Math. 75 (1992), no. 2, 151166.
MR 93e:57070
[Fan04] F. Fang, Positive quaternionic Kahler manifolds and symmetry rank, J. Reine
Angew. Math. 576 (2004), 149165. MR 2099202 (2005i:53060)
[FG82] M. Fernandez and A. Gray, Riemannian manifolds with structure group G
2
, Ann.
Mat. Pura Appl. (4) 132 (1982), 1945 (1983). MR 84e:53056
[FGG88] M. Fernandez, M. J. Gotay, and A. Gray, Compact parallelizable four-dimensional
symplectic and complex manifolds, Proc. Amer. Math. Soc. 103 (1988), no. 4, 1209
1212. MR 90a:53039
[FIMU06] M. Fernandez, S. Ivanov, V Mu noz, and L. Ugarte, Nearly hypo structures
and compact nearly Kahler 6-manifolds with conical singularities, preprint;
arXiv:math.DG/0602160 (2006).
[Fin77] R. Fintushel, Circle actions on simply connected 4-manifolds, Trans. Amer. Math.
Soc. 230 (1977), 147171. MR 0458456 (56 #16659)
[Fin78] , Classication of circle actions on 4-manifolds, Trans. Amer. Math. Soc.
242 (1978), 377390. MR 496815 (81e:57036)
[Fin01] J. D. Finley, III, Diculties with the SDi(2) Toda equation, Backlund and Dar-
boux transformations. The geometry of solitons (Halifax, NS, 1999), CRM Proc.
Lecture Notes, vol. 29, Amer. Math. Soc., Providence, RI, 2001, pp. 217224.
MR 1870919 (2002k:37148)
[FIP04] M. Falcitelli, S. Ianus, and A. M. Pastore, Riemannian submersions and related
topics, World Scientic Publishing Co. Inc., River Edge, NJ, 2004. MR 2110043
[FK82] T. Friedrich and H. Kurke, Compact four-dimensional self-dual Einstein manifolds
with positive scalar curvature, Math. Nachr. 106 (1982), 271299. MR 84b:53043
[FK89] T. Friedrich and I. Kath, Einstein manifolds of dimension ve with small rst
eigenvalue of the Dirac operator, J. Dierential Geom. 29 (1989), no. 2, 263279.
MR 982174 (90e:58158)
[FK90] , 7-dimensional compact Riemannian manifolds with Killing spinors,
Comm. Math. Phys. 133 (1990), no. 3, 543561. MR 91j:58170
[FKMS97] T. Friedrich, I. Kath, A. Moroianu, and U. Semmelmann, On nearly parallel G
2
-
structures, J. Geom. Phys. 23 (1997), no. 3-4, 259286. MR 98j:53053
[FKS02] E. V. Ferapontov, D. A. Korotkin, and V. A. Shramchenko, Boyer-Finley equation
and systems of hydrodynamic type, Classical Quantum Gravity 19 (2002), no. 24,
L205L210. MR 1956302 (2003m:83024)
[Flo91] A. Floer, Self-dual conformal structures on lCP
2
, J. Dierential Geom. 33 (1991),
no. 2, 551573. MR 1094469 (92e:53049)
[FM94] R. Friedman and J. W. Morgan, Smooth four-manifolds and complex surfaces,
Ergebnisse der Mathematik und ihrer Grenzgebiete (3) [Results in Mathematics and
Related Areas (3)], vol. 27, Springer-Verlag, Berlin, 1994. MR 1288304 (95m:57046)
[FOW06] A. Futaki, H. Ono, and G. Wang, Transverse Kahler geometry of Sasaki manifolds
and toric Sasaki-Einstein manifolds, preprint: arXiv:math.DG/0607586 (2006).
[Fra87] A. Franc, Spin structures and Killing spinors on lens spaces, J. Geom. Phys. 4
(1987), no. 3, 277287. MR 957015 (90e:57047)
BIBLIOGRAPHY 583
[Fri80] T. Friedrich, Der erste Eigenwert des Dirac-Operators einer kompakten, Rie-
mannschen Mannigfaltigkeit nichtnegativer Skalarkr ummung, Math. Nachr. 97
(1980), 117146. MR 82g:58088
[Fri00] , Dirac operators in Riemannian geometry, Graduate Studies in Mathemat-
ics, vol. 25, American Mathematical Society, Providence, RI, 2000, Translated from
the 1997 German original by Andreas Nestke. MR 1777332 (2001c:58017)
[Fuj66] T. Fujitani, Complex-valued dierential forms on normal contact Riemannian man-
ifolds, Tohoku Math. J. (2) 18 (1966), 349361. MR 35 #3596
[Fuj72] A. Fujimoto, Theory of G-structures, Study Group of Geometry, Department of
Applied Mathematics, College of Liberal Arts and Science, Okayama University,
Okayama, 1972, English edition, translated from the original Japanese, Publications
of the Study Group of Geometry, Vol. 1. MR 50 #1161
[Fuj87] A. Fujiki, On the de Rham cohomology group of a compact Kahler symplectic man-
ifold, Algebraic geometry, Sendai, 1985, Adv. Stud. Pure Math., vol. 10, North-
Holland, Amsterdam, 1987, pp. 105165. MR 90d:53083
[Ful84] W. Fulton, Intersection theory, Ergebnisse der Mathematik und ihrer Grenzgebiete
(3) [Results in Mathematics and Related Areas (3)], vol. 2, Springer-Verlag, Berlin,
1984. MR 732620 (85k:14004)
[Ful93] , Introduction to toric varieties, Annals of Mathematics Studies, vol. 131,
Princeton University Press, Princeton, NJ, 1993, The William H. Roever Lectures
in Geometry. MR 94g:14028
[Fur86] M. Furushima, Singular del Pezzo surfaces and analytic compactications of
3-dimensional complex ane space C
3
, Nagoya Math. J. 104 (1986), 128.
MR 868434 (88m:32051)
[Fut83] A. Futaki, An obstruction to the existence of Einstein Kahler metrics, Invent. Math.
73 (1983), no. 3, 437443. MR 84j:53072
[Fut88] , Kahler-Einstein metrics and integral invariants, Lecture Notes in Mathe-
matics, vol. 1314, Springer-Verlag, Berlin, 1988. MR 90a:53053
[FW03] T. Friedmann and E. Witten, Unication scale, proton decay, and manifolds of
G
2
holonomy, Adv. Theor. Math. Phys. 7 (2003), no. 4, 577617. MR 2039032
(2005h:81279)
[Gal79] S. Gallot,

Equations dierentielles caracteristiques de la sph`ere, Ann. Sci.

Ecole
Norm. Sup. (4) 12 (1979), no. 2, 235267. MR 543217 (80h:58051)
[Gal87a] K. Galicki, A generalization of the momentum mapping construction for quater-
nionic Kahler manifolds, Comm. Math. Phys. 108 (1987), no. 1, 117138.
MR 88f:53088
[Gal87b] , New matter couplings in N = 2 supergravity, Nuclear Phys. B 289 (1987),
no. 2, 573588. MR 88j:53076
[Gal91] , Multi-centre metrics with negative cosmological constant, Classical Quan-
tum Gravity 8 (1991), no. 8, 15291543. MR 92i:53040
[Gei91] H. Geiges, Contact structures on 1-connected 5-manifolds, Mathematika 38 (1991),
no. 2, 303311 (1992). MR 93e:57042
[Gei97a] , Constructions of contact manifolds, Math. Proc. Cambridge Philos. Soc.
121 (1997), no. 3, 455464. MR 1434654 (98f:53027)
[Gei97b] , Normal contact structures on 3-manifolds, Tohoku Math. J. (2) 49 (1997),
no. 3, 415422. MR 98h:53046
[Gei06] , Contact geometry, Handbook of dierential geometry. Vol. II,
Elsevier/North-Holland, Amsterdam, 2006, pp. 315382. MR 2194671
[GGP03] D. Grantcharov, G. Grantcharov, and Y.S. Poon, Calabi-Yau connections with tor-
sion on toric bundles, preprint; arXiv:math.DG/0306207; to appear in J. Dieren-
tial Geometry (2003).
[GH78a] G. W. Gibbons and S. W. Hawking, Gravitational multi-instantons, Phys. Lett. B
78 (1978), 430432. MR 0465052 (57 #4965)
[GH78b] P. Griths and J. Harris, Principles of algebraic geometry, Wiley-Interscience [John
Wiley & Sons], New York, 1978, Pure and Applied Mathematics. MR 80b:14001
[GH06] K. Guruprasad and A. Haeiger, Closed geodesics on orbifolds, Topology 45 (2006),
no. 3, 611641. MR 2218759
584 BIBLIOGRAPHY
[GHJ03] M. Gross, D. Huybrechts, and D. Joyce, Calabi-Yau manifolds and related geome-
tries, Universitext, Springer-Verlag, Berlin, 2003, Lectures from the Summer School
held in Nordfjordeid, June 2001. MR 1963559 (2004c:14075)
[GHR84] S. J. Gates, Jr., C. M. Hull, and M. Rocek, Twisted multiplets and new supersym-
metric nonlinear -models, Nuclear Phys. B 248 (1984), no. 1, 157186. MR 776369
(87b:81108)
[GHS83] J. Girbau, A. Haeiger, and D. Sundararaman, On deformations of transversely
holomorphic foliations, J. Reine Angew. Math. 345 (1983), 122147. MR 84j:32026
[GHV73] W. Greub, S. Halperin, and R. Vanstone, Connections, curvature, and cohomology.
Vol. II: Lie groups, principal bundles, and characteristic classes, Academic Press
[A subsidiary of Harcourt Brace Jovanovich, Publishers], New York-London, 1973,
Pure and Applied Mathematics, Vol. 47-II. MR 49 #1424
[GHY04] G. W. Gibbons, S. A. Hartnoll, and Y. Yasui, Properties of some ve-dimensional
Einstein metrics, Classical Quantum Gravity 21 (2004), no. 19, 46974730.
MR 2094200 (2006b:53058)
[Gif69a] C. H. Gien, Smooth homotopy projective spaces, Bull. Amer. Math. Soc. 75 (1969),
509513. MR 0239607 (39 #964)
[Gif69b] , Weakly complex involutions and cobordism of projective spaces, Ann. of
Math. (2) 90 (1969), 418432. MR 0253346 (40 #6561)
[GK05] A Ghigi and J. Kollar, Kahler-Einstein metrics on orbifolds and Einstein metrics
on Spheres, preprint; arXiv:math.DG/0507289 (2005).
[GKN] M. Godli nski, W. Kopczy nski, and P. Nurowski.
[GKP89] R. L. Graham, D. E. Knuth, and O. Patashnik, Concrete mathematics, Addison-
Wesley Publishing Company Advanced Book Program, Reading, MA, 1989, A foun-
dation for computer science. MR 91f:00001
[GL80] M. Gromov and H. B. Lawson, Jr., The classication of simply connected mani-
folds of positive scalar curvature, Ann. of Math. (2) 111 (1980), no. 3, 423434.
MR 577131 (81h:53036)
[GL88] K. Galicki and H. B. Lawson, Jr., Quaternionic reduction and quaternionic orb-
ifolds, Math. Ann. 282 (1988), no. 1, 121. MR 89m:53075
[GL02] V. Guillemin and E. Lerman, Melrose-Uhlmann projectors, the metaplectic rep-
resentation and symplectic cuts, J. Dierential Geom. 61 (2002), no. 3, 365396.
MR 2004e:53126
[GLS96] V. Guillemin, E. Lerman, and S. Sternberg, Symplectic brations and multiplicity
diagrams, Cambridge University Press, Cambridge, 1996. MR 1414677 (98d:58074)
[GM74] D. Gromoll and W. Meyer, An exotic sphere with nonnegative sectional curvature,
Ann. of Math. (2) 100 (1974), 401406. MR 51 #11347
[GM80] X. Gomez-Mont, Transversal holomorphic structures, J. Dierential Geom. 15
(1980), no. 2, 161185 (1981). MR 82j:53065
[GM88] M. Goresky and R. MacPherson, Stratied Morse theory, Ergebnisse der Mathe-
matik und ihrer Grenzgebiete (3) [Results in Mathematics and Related Areas (3)],
vol. 14, Springer-Verlag, Berlin, 1988. MR 932724 (90d:57039)
[GMSW04a] J. P. Gauntlett, D. Martelli, J. Sparks, and D. Waldram, A new innite class of
Sasaki-Einstein manifolds, Adv. Theor. Math. Phys. 8 (2004), no. 6, 9871000.
MR 2194373
[GMSW04b] , Sasaki-Einstein metrics on S
2
S
3
, Adv. Theor. Math. Phys. 8 (2004),
no. 4, 711734. MR 2141499
[GMSY06] J. P. Gauntlett, D. Martelli, J. Sparks, and S.-T. Yau, Obstructions to the existence
of Sasaki-Einstein metrics, preprint; arXiv:hep-th/0607080 (2006).
[GN92] K. Galicki and T. Nitta, Nonzero scalar curvature generalizations of the ALE hyper-
K ahler metrics, J. Math. Phys. 33 (1992), no. 5, 17651771. MR 93d:53056
[GO98] P. Gauduchon and L. Ornea, Locally conformally Kahler metrics on Hopf surfaces,
Ann. Inst. Fourier (Grenoble) 48 (1998), no. 4, 11071127. MR 2000g:53088
[GO01] G. Grantcharov and L. Ornea, Reduction of Sasakian manifolds, J. Math. Phys. 42
(2001), no. 8, 38093816. MR 1845220 (2002e:53060)
[God58] R. Godement, Topologie algebrique et theorie des faisceaux, Actualites Sci. Ind. No.
1252. Publ. Math. Univ. Strasbourg. No. 13, Hermann, Paris, 1958. MR 0102797
(21 #1583)
BIBLIOGRAPHY 585
[God91] C. Godbillon, Feuilletages, Progress in Mathematics, vol. 98, Birkhauser Verlag,
Basel, 1991,

Etudes geometriques. [Geometric studies], With a preface by G. Reeb.
MR 93i:57038
[Gol65] S. I. Goldberg, Rigidite de varietes de contact `a courbure positive, C. R. Acad. Sci.
Paris 261 (1965), 19361939. MR 0188950 (32 #6377)
[Gol67] , Rigidity of positively curved contact manifolds, J. London Math. Soc. 42
(1967), 257263. MR 0232325 (38 #650)
[Gol69] , Integrability of almost Kaehler manifolds, Proc. Amer. Math. Soc. 21
(1969), 96100. MR 38 #6514
[Gol02] R. F. Goldin, An eective algorithm for the cohomology ring of symplectic reduc-
tions, Geom. Funct. Anal. 12 (2002), no. 3, 567583. MR 1924372 (2003m:53148)
[Gom95] R. E. Gompf, A new construction of symplectic manifolds, Ann. of Math. (2) 142
(1995), no. 3, 527595. MR 96j:57025
[Gor80] V. V. Gorbacevic, Compact homogeneous manifolds of small dimension, Geometric
methods in problems of algebra and analysis, No. 2 (Russian), Yaroslav. Gos. Univ.,
Yaroslavl

, 1980, pp. 3760, 161. MR 617386 (82j:53081)

[Got98] R. Goto, On hyper-K ahler manifolds of type A

and D

, Comm. Math. Phys.

198 (1998), no. 2, 469491. MR 1672449 (2000f:53062)
[GP00] G. Grantcharov and Y-S. Poon, Geometry of hyper-K ahler connections with torsion,
Comm. Math. Phys. 213 (2000), no. 1, 1937. MR 1782143 (2002a:53059)
[GPP90] G. W. Gibbons, D. N. Page, and C. N. Pope, Einstein metrics on S
3
, R
3
, and R
4
bundles, Comm. Math. Phys. 127 (1990), no. 3, 529553. MR 1040893 (91f:53039)
[GR65] R. C. Gunning and H. Rossi, Analytic functions of several complex variables,
Prentice-Hall Inc., Englewood Clis, N.J., 1965. MR 0180696 (31 #4927)
[GR84] H. Grauert and R. Remmert, Coherent analytic sheaves, Grundlehren der Mathe-
matischen Wissenschaften [Fundamental Principles of Mathematical Sciences], vol.
265, Springer-Verlag, Berlin, 1984. MR 755331 (86a:32001)
[Gra59] J. W. Gray, Some global properties of contact structures, Ann. of Math. (2) 69
(1959), 421450. MR 22 #3016
[Gra67] A. Gray, Pseudo-Riemannian almost product manifolds and submersions, J. Math.
Mech. 16 (1967), 715737. MR 34 #5018
[Gra69a] , A note on manifolds whose holonomy group is a subgroup of Sp(n) Sp(1),
Michigan Math. J. 16 (1969), 125128. MR 39 #6226
[Gra69b] , Six dimensional almost complex manifolds dened by means of three-fold
vector cross products, Tohoku Math. J. (2) 21 (1969), 614620. MR 0261515 (41
#6128)
[Gra70] , Nearly Kahler manifolds, J. Dierential Geometry 4 (1970), 283309.
MR 0267502 (42 #2404)
[Gra71] , Weak holonomy groups, Math. Z. 123 (1971), 290300. MR 45 #2614
[Gra76] , The structure of nearly Kahler manifolds, Math. Ann. 223 (1976), no. 3,
233248. MR 0417965 (54 #6010)
[Gra04] G. Grantcharov, Hyper-Hermitian manifolds and connections with skew-symmetric
torsion, Cliord algebras (Cookeville, TN, 2002), Prog. Math. Phys., vol. 34,
Birkhauser Boston, Boston, MA, 2004, pp. 167183. MR 2025979 (2004m:53082)
[Gro61] A. Grothendieck,

Elements de geometrie algebrique. II.

Etude globale elementaire
de quelques classes de morphismes, Inst. Hautes

Etudes Sci. Publ. Math. (1961),
no. 8, 222. MR 36 #177b
[Gro68] , Cohomologie locale des faisceaux coherents et theor`emes de Lefschetz lo-
caux et globaux (SGA 2), North-Holland Publishing Co., Amsterdam, 1968, Aug-
mente dun expose par Mich`ele Raynaud, Seminaire de Geometrie Algebrique du
Bois-Marie, 1962, Advanced Studies in Pure Mathematics, Vol. 2. MR 57 #16294
[Gro81] M. Gromov, Curvature, diameter and Betti numbers, Comment. Math. Helv. 56
(1981), no. 2, 179195. MR 82k:53062
[Gro99] , Metric structures for Riemannian and non-Riemannian spaces, Progress
in Mathematics, vol. 152, Birkhauser Boston Inc., Boston, MA, 1999, Based on
the 1981 French original [ MR0682063 (85e:53051)], With appendices by M. Katz,
P. Pansu and S. Semmes, Translated from the French by Sean Michael Bates.
MR 1699320 (2000d:53065)
586 BIBLIOGRAPHY
[Gru90] R. Grunewald, Six-dimensional Riemannian manifolds with a real Killing spinor,
Ann. Global Anal. Geom. 8 (1990), no. 1, 4359. MR 1075238 (92a:58146)
[GS66] V. Guillemin and S. Sternberg, Deformation theory of pseudogroup structures,
Mem. Amer. Math. Soc. No. 64 (1966), 80. MR 35 #2302
[GS82a] , Convexity properties of the moment mapping, Invent. Math. 67 (1982),
no. 3, 491513. MR 664117 (83m:58037)
[GS82b] , Homogeneous quantization and multiplicities of group representations, J.
Funct. Anal. 47 (1982), no. 3, 344380. MR 84d:58034
[GS84] , Symplectic techniques in physics, Cambridge University Press, Cambridge,
1984. MR 770935 (86f:58054)
[GS96] K. Galicki and S. Salamon, Betti numbers of 3-Sasakian manifolds, Geom. Dedicata
63 (1996), no. 1, 4568. MR 98e:53075
[GS99] R. E. Gompf and A. I. Stipsicz, 4-manifolds and Kirby calculus, Graduate Studies
in Mathematics, vol. 20, American Mathematical Society, Providence, RI, 1999.
MR 1707327 (2000h:57038)
[GT95] H. Geiges and C. B. Thomas, Hypercontact manifolds, J. London Math. Soc. (2)
51 (1995), no. 2, 342352. MR 96d:53048
[Gui94a] V. Guillemin, Kaehler structures on toric varieties, J. Dierential Geom. 40 (1994),
no. 2, 285309. MR 1293656 (95h:32029)
[Gui94b] , Moment maps and combinatorial invariants of Hamiltonian T
n
-spaces,
Progress in Mathematics, vol. 122, Birkhauser Boston Inc., Boston, MA, 1994.
MR 1301331 (96e:58064)
[GWZ05] K. Grove, B. Wilking, and W. Ziller, Positively curved manifolds cohomogeneity
one manifolds and 3-Sasakian geometry, math.DG/0511464 (2005).
[GY70] S. I. Goldberg and K. Yano, On normal globally framed f-manifolds, Tohoku Math.
J. (2) 22 (1970), 362370. MR 0305295 (46 #4425)
[GY71] , Globally framed f-manifolds, Illinois J. Math. 15 (1971), 456474.
MR 0278247 (43 #3978)
[GZ00] K. Grove and W. Ziller, Curvature and symmetry of Milnor spheres, Ann. of Math.
(2) 152 (2000), no. 1, 331367. MR 2001i:53047
[GZ02] , Cohomogeneity one manifolds with positive Ricci curvature, Invent. Math.
149 (2002), no. 3, 619646. MR 1923478 (2004b:53052)
[Hae58] A. Haeiger, Structures feuilletees et cohomologie `a valeur dans un faisceau de
groupodes, Comment. Math. Helv. 32 (1958), 248329. MR 20 #6702
[Hae62] , Varietes feuilletees, Ann. Scuola Norm. Sup. Pisa (3) 16 (1962), 367397.
MR 32 #6487
[Hae70] , Feuilletages sur les varietes ouvertes, Topology 9 (1970), 183194. MR 41
#7709
[Hae71] , Homotopy and integrability, ManifoldsAmsterdam 1970 (Proc. Nuc
Summer School), Lecture Notes in Mathematics, Vol. 197, Springer, Berlin, 1971,
pp. 133163. MR 44 #2251
[Hae84] , Groupodes dholonomie et classiants, Asterisque (1984), no. 116, 7097,
Transversal structure of foliations (Toulouse, 1982). MR 86c:57026a
[Ham82] R. S. Hamilton, Three-manifolds with positive Ricci curvature, J. Dierential Geom.
17 (1982), no. 2, 255306. MR 664497 (84a:53050)
[Har77] R. Hartshorne, Algebraic geometry, Springer-Verlag, New York, 1977, Graduate
Texts in Mathematics, No. 52. MR 57 #3116
[Har92] J. Harris, Algebraic geometry, Graduate Texts in Mathematics, vol. 133, Springer-
Verlag, New York, 1992, A rst course. MR 1182558 (93j:14001)
[Hat63] Y. Hatakeyama, Some notes on dierentiable manifolds with almost contact struc-
tures, Osaka Math. J. (2) 15 (1963), 176181. MR 27 #705
[Hau05] T. Hausel, Mirror symmetry and Langlands duality in the non-abelian Hodge the-
ory of a curve, Geometric methods in algebra and number theory, Progr. Math.,
vol. 235, Birkhauser Boston, Boston, MA, 2005, pp. 193217. MR 2166085
(2006h:14055)
[Hau06] , Betti numbers of holomorphic symplectic quotients via arithmetic Fourier
transform, Proc. Natl. Acad. Sci. USA 103 (2006), no. 16, 61206124 (electronic).
MR 2221039
BIBLIOGRAPHY 587
[Haw77] S. W. Hawking, Gravitational instantons, Phys. Lett. A 60 (1977), no. 2, 8183.
MR 0465052 (57 #4965)
[HD84] A. Haeiger and Quach Ngoc Du, Appendice: une presentation du groupe fonda-
mental dune orbifold, Asterisque (1984), no. 116, 98107, Transversal structure of
foliations (Toulouse, 1982). MR 755164 (86c:57026b)
[HE73] S. W. Hawking and G. F. R. Ellis, The large scale structure of space-time, Cam-
bridge University Press, London, 1973, Cambridge Monographs on Mathematical
Physics, No. 1. MR 54 #12154
[HEK05] C. P. Herzog, Q. J. Ejaz, and I. R. Klebanov, Cascading RG ows from new Sasaki-
Einstein manifolds, J. High Energy Phys. (2005), no. 2, 009, 22 pp. (electronic).
MR 2140502
[Hep05a] R. Hepworth, Generalized Kreck-Stolz invariants and the topology of certain 3-
Sasakian 7-manifolds, Edinburgh University Thesis (2005).
[Hep05b] , The topology of certain 3-Sasakian 7-manifolds, math.AT/0511735 (2005).
[Her96] G. Hernandez, On hyper f-structures, Math. Ann. 306 (1996), no. 2, 205230.
MR 1411345 (97m:53057)
[HH86] G. Hector and U. Hirsch, Introduction to the geometry of foliations. Part A, sec-
ond ed., Aspects of Mathematics, vol. 1, Friedr. Vieweg & Sohn, Braunschweig,
1986, Foliations on compact surfaces, fundamentals for arbitrary codimension, and
holonomy. MR 88a:57048
[HH02a] H. Herrera and R. Herrera, Classication of positive quaternion-Kahler 12-
manifolds, C. R. Math. Acad. Sci. Paris 334 (2002), no. 1, 4346. MR 2003e:53056
[HH02b] ,

A-genus on non-spin manifolds with S
1
actions and the classication of
positive quaternion-Kahler 12-manifolds, J. Dierential Geom. 61 (2002), no. 3,
341364. MR 1 979 364
[Hij86] O. Hijazi, Caracterisation de la sph`ere par les premi`eres valeurs propres de
loperateur de Dirac en dimensions 3, 4, 7 et 8, C. R. Acad. Sci. Paris Ser. I
Math. 303 (1986), no. 9, 417419. MR 862207 (88b:58142)
[Hir66] F. Hirzebruch, Topological methods in algebraic geometry, Third enlarged edi-
tion. New appendix and translation from the second German edition by R. L.
E. Schwarzenberger, with an additional section by A. Borel. Die Grundlehren der
Mathematischen Wissenschaften, Band 131, Springer-Verlag New York, Inc., New
York, 1966. MR 0202713 (34 #2573)
[Hir67] , Singularities and exotic spheres, Seminaire Bourbaki, Vol. 10, Soc. Math.
France, Paris, 1967, pp. Exp. No. 314, 1332. MR 1 610 436
[Hir71] , Pontrjagin classes of rational homology manifolds and the signature of
some ane hypersurfaces, Proceedings of Liverpool Singularities Symposium, II
(1969/1970) (Berlin), Springer, 1971, pp. 207212. Lecture Notes in Math., Vol.
209. MR 49 #1525
[Hit74] N. J. Hitchin, Harmonic spinors, Advances in Math. 14 (1974), 155. MR 50
#11332
[Hit79] , Polygons and gravitons, Math. Proc. Cambridge Philos. Soc. 85 (1979),
no. 3, 465476. MR 520463 (80h:53047)
[Hit81] , Kahlerian twistor spaces, Proc. London Math. Soc. (3) 43 (1981), no. 1,
133150. MR 84b:32014
[Hit87a] , Monopoles, minimal surfaces and algebraic curves, Seminaire de
Mathematiques Superieures [Seminar on Higher Mathematics], vol. 105, Presses
de lUniversite de Montreal, Montreal, QC, 1987. MR 935967 (89h:58043)
[Hit87b] , The self-duality equations on a Riemann surface, Proc. London Math. Soc.
(3) 55 (1987), no. 1, 59126. MR 887284 (89a:32021)
[Hit92] , Hyper-K ahler manifolds, Asterisque (1992), no. 206, Exp. No. 748, 3, 137
166, Seminaire Bourbaki, Vol. 1991/92. MR 1206066 (94f:53087)
[Hit95a] , Poncelet polygons and the Painleve equations, Geometry and analysis
(Bombay, 1992), Tata Inst. Fund. Res., Bombay, 1995, pp. 151185. MR 97d:32042
[Hit95b] , Twistor spaces, Einstein metrics and isomonodromic deformations, J. Dif-
ferential Geom. 42 (1995), no. 1, 30112. MR 96g:53057
588 BIBLIOGRAPHY
[Hit96] , A new family of Einstein metrics, Manifolds and geometry (Pisa, 1993),
Sympos. Math., XXXVI, Cambridge Univ. Press, Cambridge, 1996, pp. 190222.
MR 97d:53050
[HKK
+
03] K. Hori, S. Katz, A. Klemm, R. Pandharipande, R. Thomas, C. Vafa, R. Vakil,
and E. Zaslow, Mirror symmetry, Clay Mathematics Monographs, vol. 1, American
Mathematical Society, Providence, RI, 2003, With a preface by Vafa. MR 2003030
(2004g:14042)
[HKLR87] N. J. Hitchin, A. Karlhede, U. Lindstrom, and M. Rocek, Hyper-Kahler metrics and
supersymmetry, Comm. Math. Phys. 108 (1987), no. 4, 535589. MR 88g:53048
[HKW05] A. Hanany, P. Kazakopoulos, and B. Wecht, A new innite class of quiver gauge
theories, J. High Energy Phys. (2005), no. 8, 054, 30 pp. (electronic). MR 2166231
(2006h:81286)
[HL82] R. Harvey and H. B. Lawson, Jr., Calibrated geometries, Acta Math. 148 (1982),
47157. MR 666108 (85i:53058)
[HM04] Andre Henriques and David S. Metzler, Presentations of noneective orbifolds,
Trans. Amer. Math. Soc. 356 (2004), no. 6, 24812499 (electronic). MR MR2048526
[Hoe07] C. A. Hoelscher, Classication of cohomogeneity one manifolds in low dimensions,
University of Pennsylvania Ph.D. Thesis.
[Hon06] K. Honda, The topology and geometry of contact structures in dimension three,
math.GT/0601144 (2006).
[HOT63] Y. Hatakeyama, Y. Ogawa, and S. Tanno, Some properties of manifolds with contact
metric structure, Tohoku Math. J. (2) 15 (1963), 4248. MR 26 #4295
[HP96] P. S. Howe and G. Papadopoulos, Twistor spaces for hyper-Kahler manifolds with
torsion, Phys. Lett. B 379 (1996), no. 1-4, 8086. MR 1396267 (97h:53073)
[HS84] P. S. Howe and G. Sierra, Two-dimensional supersymmetric nonlinear -models
with torsion, Phys. Lett. B 148 (1984), no. 6, 451455. MR 769268 (86g:81121)
[HS91] A. Haeiger and

E. Salem, Actions of tori on orbifolds, Ann. Global Anal. Geom.
9 (1991), no. 1, 3759. MR 92f:57047
[HS02] T. Hausel and B. Sturmfels, Toric hyperK ahler varieties, Doc. Math. 7 (2002),
495534 (electronic). MR 2015052 (2004i:53054)
[HS82] I. Hasegawa and M. Seino, Some remarks on Sasakian geometryapplications of
Myers theorem and the canonical ane connection, J. Hokkaido Univ. Ed. Sect.
II A 32 (1981/82), no. 1, 17. MR 84j:53055
[HSY04] Y. Hashimoto, M. Sakaguchi, and Y. Yasui, Sasaki-Einstein twist of Kerr-AdS black
holes, Phys. Lett. B 600 (2004), no. 3-4, 270274. MR 2093038 (2005h:83235)
[HSY05] , New innite series of Einstein metrics on sphere bundles from AdS black
holes, Comm. Math. Phys. 257 (2005), no. 2, 273285. MR 2164598
[Hus66] D. Husemoller, Fibre bundles, McGraw-Hill Book Co., New York, 1966. MR 37
#4821
[HZ74] F. Hirzebruch and D. Zagier, The Atiyah-Singer theorem and elementary number
theory, Publish or Perish Inc., Boston, Mass., 1974, Mathematics Lecture Series,
No. 3. MR 58 #31291
[IF00] A. R. Iano-Fletcher, Working with weighted complete intersections, Explicit bira-
tional geometry of 3-folds, London Math. Soc. Lecture Note Ser., vol. 281, Cam-
bridge Univ. Press, Cambridge, 2000, pp. 101173. MR 2001k:14089
[IK72] S. Ishihara and M. Konishi, Fibred Riemannian spaces with Sasakian 3-structure,
Dierential geometry (in honor of Kentaro Yano), Kinokuniya, Tokyo, 1972,
pp. 179194. MR 48 #4964
[IP99] V. A. Iskovskikh and Yu. G. Prokhorov, Fano varieties, Algebraic geometry, V,
Encyclopaedia Math. Sci., vol. 47, Springer, Berlin, 1999, pp. 1247. MR 1668579
(2000b:14051b)
[Ish73] S. Ishihara, Quaternion Kahlerian manifolds and bred Riemannian spaces with
Sasakian 3-structure, Kodai Math. Sem. Rep. 25 (1973), 321329. MR 48 #2943
[Ish74] , Quaternion Kahlerian manifolds, J. Dierential Geometry 9 (1974), 483
500. MR 50 #1184
[Isk77] V. A. Iskovskih, Fano threefolds. I, Izv. Akad. Nauk SSSR Ser. Mat. 41 (1977),
no. 3, 516562, 717. MR 80c:14023a
BIBLIOGRAPHY 589
[Isk78] , Fano threefolds. II, Izv. Akad. Nauk SSSR Ser. Mat. 42 (1978), no. 3,
506549. MR 80c:14023b
[Isk79] , Anticanonical models of three-dimensional algebraic varieties, Current
problems in mathematics, Vol. 12 (Russian); (English translation: J. Soviet Math.
13 (1980), 745-814), VINITI, Moscow, 1979, pp. 59157, 239 (loose errata).
MR 537685 (81i:14026b)
[Ito84] M. Itoh, Self-duality of Kahler surfaces, Compositio Math. 51 (1984), no. 2, 265
273. MR 739738 (85m:53079)
[Ito97] , Odd-dimensional tori and contact structure, Proc. Japan Acad. Ser. A
Math. Sci. 73 (1997), no. 4, 5859. MR 98h:53048
[IY82] H. Inoue and K. Yano, The Gromov invariant of negatively curved manifolds, Topol-
ogy 21 (1982), no. 1, 8389. MR 82k:53091
[Jac90] H. Jacobowitz, An introduction to CR structures, Mathematical Surveys and
Monographs, vol. 32, American Mathematical Society, Providence, RI, 1990.
MR 93h:32023
[Jel01] W. Jelonek, Positive and negative 3-K-contact structures, Proc. Amer. Math. Soc.
129 (2001), no. 1, 247256. MR 1694866 (2001c:53061)
[Jen73] G. R. Jensen, Einstein metrics on principal bre bundles, J. Dierential Geometry
8 (1973), 599614. MR 50 #5694
[JK01a] J. M. Johnson and J. Kollar, Fano hypersurfaces in weighted projective 4-spaces,
Experiment. Math. 10 (2001), no. 1, 151158. MR 2002a:14048
[JK01b] , K ahler-Einstein metrics on log del Pezzo surfaces in weighted projective
3-spaces, Ann. Inst. Fourier (Grenoble) 51 (2001), no. 1, 6979. MR 2002b:32041
[Jos01] J. Jost, Bosonic strings: a mathematical treatment, AMS/IP Studies in Ad-
vanced Mathematics, vol. 21, American Mathematical Society, Providence, RI, 2001.
MR 1999211 (2004h:81002)
[Joy91] D. D. Joyce, The hypercomplex quotient and the quaternionic quotient, Math. Ann.
290 (1991), no. 2, 323340. MR 92f:53052
[Joy92] , Compact hypercomplex and quaternionic manifolds, J. Dierential Geom.
35 (1992), no. 3, 743761. MR 1163458 (93g:53064)
[Joy95] , Explicit construction of self-dual 4-manifolds, Duke Math. J. 77 (1995),
no. 3, 519552. MR 1324633 (96d:53049)
[Joy96a] , Compact 8-manifolds with holonomy Spin(7), Invent. Math. 123 (1996),
no. 3, 507552. MR 1383960 (97d:53052)
[Joy96b] , Compact Riemannian 7-manifolds with holonomy G
2
. I, II, J. Dierential
Geom. 43 (1996), no. 2, 291328, 329375. MR 1424428 (97m:53084)
[Joy99] , A new construction of compact 8-manifolds with holonomy Spin(7), J.
Dierential Geom. 53 (1999), no. 1, 89130. MR 1776092 (2002a:53063)
[Joy00] , Compact manifolds with special holonomy, Oxford Mathematical Mono-
graphs, Oxford University Press, Oxford, 2000. MR 2001k:53093
[Kah33] E. Kahler,

Uber eine bemerkenswerte Hermitesche Metrik, Abh. Math. Sem. Ham-
burg Univ. 9 (1933), 173186.
[Kah03] , Mathematische Werke/Mathematical works, Walter de Gruyter & Co.,
Berlin, 2003, Edited by Rolf Berndt and Oswald Riemenschneider. MR 2229234
[Kar03] Y. Karshon, Maximal tori in the symplectomorphism groups of Hirzebruch surfaces,
Math. Res. Lett. 10 (2003), no. 1, 125132. MR 1960129 (2004f:53101)
[Kas71] T. Kashiwada, A note on a Riemannian space with Sasakian 3-structure, Natur.
Sci. Rep. Ochanomizu Univ. 22 (1971), 12. MR 46 #2586
[Kas01] , On a contact 3-structure, Math. Z. 238 (2001), no. 4, 829832.
MR 2002h:53136
[Kaw78] T. Kawasaki, The signature theorem for V -manifolds, Topology 17 (1978), no. 1,
7583. MR 57 #14072
[Kaw79] , The Riemann-Roch theorem for complex V -manifolds, Osaka J. Math. 16
(1979), no. 1, 151159. MR 80f:58042
[Kaw81] , The index of elliptic operators over V -manifolds, Nagoya Math. J. 84
(1981), 135157. MR 83i:58095
[Keh04] W. Kehowski, Hypernormal manifolds, University of New Mexico Thesis (2004).
590 BIBLIOGRAPHY
[Kir84] F. Kirwan, Cohomology of quotients in symplectic and algebraic geometry, Mathe-
matical Notes, vol. 31, Princeton University Press, Princeton, NJ, 1984. MR 766741
(86i:58050)
[Kir98] , Momentum maps and reduction in algebraic geometry, Dierential Geom.
Appl. 9 (1998), no. 1-2, 135171, Symplectic geometry. MR 1636303 (99e:58072)
[KKS78] D. Kazhdan, B. Kostant, and S. Sternberg, Hamiltonian group actions and dynam-
ical systems of Calogero type, Comm. Pure Appl. Math. 31 (1978), no. 4, 481507.
MR 0478225 (57 #17711)
[KL06] B. Kleiner and J. Lott, Notes on Perelmans papers, preprint;
arXiv:math.DG/0605667 (2006).
[KLP97] J. Kim, C. LeBrun, and M. Pontecorvo, Scalar-at Kahler surfaces of all genera,
J. Reine Angew. Math. 486 (1997), 6995. MR 1450751 (98m:32045)
[KM63] M. A. Kervaire and J. W. Milnor, Groups of homotopy spheres. I, Ann. of Math.
(2) 77 (1963), 504537. MR 26 #5584
[KM98] J. Kollar and S. Mori, Birational geometry of algebraic varieties, Cambridge Tracts
in Mathematics, vol. 134, Cambridge University Press, Cambridge, 1998, With the
collaboration of C. H. Clemens and A. Corti, Translated from the 1998 Japanese
original. MR 2000b:14018
[KM99] S. Keel and J. McKernan, Rational curves on quasi-projective surfaces, Mem. Amer.
Math. Soc. 140 (1999), no. 669, viii+153. MR 1610249 (99m:14068)
[KMM92] J. Kollar, Y. Miyaoka, and S. Mori, Rational connectedness and boundedness of
Fano manifolds, J. Dierential Geom. 36 (1992), no. 3, 765779. MR 94g:14021
[KN63] S. Kobayashi and K. Nomizu, Foundations of dierential geometry. Vol I, Inter-
science Publishers, a division of John Wiley & Sons, New York-Lond on, 1963.
MR 27 #2945
[KN69] , Foundations of dierential geometry. Vol. II, Interscience Tracts in Pure
and Applied Mathematics, No. 15 Vol. II, Interscience Publishers John Wiley &
Sons, Inc., New York-London-Sydney, 1969. MR MR0238225 (38 #6501)
[KN77] L. H. Kauman and W. D. Neumann, Products of knots, branched brations and
sums of singularities, Topology 16 (1977), no. 4, 369393. MR 0488073 (58 #7644)
[KN90] P. B. Kronheimer and H. Nakajima, Yang-Mills instantons on ALE gravitational
instantons, Math. Ann. 288 (1990), no. 2, 263307. MR 1075769 (92e:58038)
[KO73] S. Kobayashi and T. Ochiai, Characterizations of complex projective spaces and
hyperquadrics, J. Math. Kyoto Univ. 13 (1973), 3147. MR 47 #5293
[Kob59a] S. Kobayashi, On the automorphism group of a certain class of algebraic manifolds.,
Tohoku Math. J. (2) 11 (1959), 184190. MR 0112159 (22 #3014)
[Kob59b] , Remarks on complex contact manifolds, Proc. Amer. Math. Soc. 10 (1959),
164167. MR 0111061 (22 #1925)
[Kob61] , On compact Kahler manifolds with positive denite Ricci tensor, Ann. of
Math. (2) 74 (1961), 570574. MR 24 #A2922
[Kob63] , Topology of positively pinched Kaehler manifolds, Tohoku Math. J. (2) 15
(1963), 121139. MR 27 #4185
[Kob72] , Transformation groups in dierential geometry, Springer-Verlag, New
York, 1972, Ergebnisse der Mathematik und ihrer Grenzgebiete, Band 70. MR 50
#8360
[Kob87] , Dierential geometry of complex vector bundles, Publications of the Math-
ematical Society of Japan, vol. 15, Princeton University Press, Princeton, NJ, 1987,
Kano Memorial Lectures, 5. MR 89e:53100
[Kod60a] K. Kodaira, On compact complex analytic surfaces. I, Ann. of Math. (2) 71 (1960),
111152. MR 24 #A2396
[Kod60b] , On deformations of some complex psuedo-group structures, Ann. of Math.
(2) 71 (1960), 224302. MR 22 #5992
[Kod86] , Complex manifolds and deformation of complex structures, Grundlehren
der Mathematischen Wissenschaften [Fundamental Principles of Mathematical Sci-
ences], vol. 283, Springer-Verlag, New York, 1986, Translated from the Japanese by
Kazuo Akao, With an appendix by Daisuke Fujiwara. MR 87d:32040
[Kod87] T. Koda, Self-dual and anti-self-dual Hermitian surfaces, Kodai Math. J. 10 (1987),
no. 3, 335342. MR 929993 (89a:53053)
BIBLIOGRAPHY 591
[Koi83] N. Koiso, Einstein metrics and complex structures, Invent. Math. 73 (1983), no. 1,
71106. MR 707349 (85d:58018)
[Kol92] J. Kollar, Flips and abundance for algebraic threefolds, Societe Mathematique de
France, Paris, 1992, Papers from the Second Summer Seminar on Algebraic Geom-
etry held at the University of Utah, Salt Lake City, Utah, August 1991, Asterisque
No. 211 (1992). MR 94f:14013
[Kol96] , Rational curves on algebraic varieties, Ergebnisse der Mathematik und
ihrer Grenzgebiete. 3. Folge. A Series of Modern Surveys in Mathematics [Results
in Mathematics and Related Areas. 3rd Series. A Series of Modern Surveys in Math-
ematics], vol. 32, Springer-Verlag, Berlin, 1996. MR 98c:14001
[Kol97] , Singularities of pairs, Algebraic geometrySanta Cruz 1995, Proc. Sym-
pos. Pure Math., vol. 62, Amer. Math. Soc., Providence, RI, 1997, pp. 221287.
MR 99m:14033
[Kol04] , Seifert Gm-bundles, preprint; arXiv:math.AG/0404386 (2004).
[Kol05] J. Kollar, Einstein metrics on ve-dimensional Seifert bundles, J. Geom. Anal. 15
(2005), no. 3, 445476. MR MR2190241
[Kol06a] , Circle actions on simply connected 5-manifolds, Topology 45 (2006), no. 3,
643671. MR 2218760
[Kol06b] J. Kollar, Positive Sasakian structures on 5-manifolds, preprint;
arXiv:math.DG/0612524 (2006).
[Kol07] , Einstein metrics on connected sums of S
2
S
3
, J. Dierential Geom. 77
(2007), no. 2, 259272. MR 2269781
[Kon00] H. Konno, Cohomology rings of toric hyperk ahler manifolds, Internat. J. Math. 11
(2000), no. 8, 10011026. MR 1797675 (2001k:53089)
[Kon75] M. Konishi, On manifolds with Sasakian 3-structure over quaternion Kaehler man-
ifolds, Kodai Math. Sem. Rep. 26 (1974/75), 194200. MR 51 #13951
[Kos55] B. Kostant, Holonomy and the Lie algebra of innitesimal motions of a Riemannian
manifold, Trans. Amer. Math. Soc. 80 (1955), 528542. MR 0084825 (18,930a)
[Kot97] D. Kotschick, Orientations and geometrisations of compact complex surfaces, Bull.
London Math. Soc. 29 (1997), no. 2, 145149. MR 1425990 (97k:32047)
[Kot98a] , Einstein metrics and smooth structures, Geom. Topol. 2 (1998), 110 (elec-
tronic). MR 1489351 (99h:57060)
[Kot98b] , On the Gromov-Hitchin-Thorpe inequality, C. R. Acad. Sci. Paris Ser. I
Math. 326 (1998), no. 6, 727731. MR 1641774 (99j:53068)
[Kou76] A. G. Kouchnirenko, Poly`edres de Newton et nombres de Milnor, Invent. Math. 32
(1976), no. 1, 131. MR 0419433 (54 #7454)
[Kov96] A. G. Kovalev, Nahms equations and complex adjoint orbits, Quart. J. Math.
Oxford Ser. (2) 47 (1996), no. 185, 4158. MR MR1380949 (97c:53070)
[KP95] J. Kim and M. Pontecorvo, A new method of constructing scalar-at Kahler sur-
faces, J. Dierential Geom. 41 (1995), no. 2, 449477. MR 1331974 (96c:32032)
[Kra65] V. Y. Kraines, Topology of quaternionic manifolds, Bull. Amer. Math. Soc. 71
(1965), 526527. MR 33 #6653
[Kra66] , Topology of quaternionic manifolds, Trans. Amer. Math. Soc. 122 (1966),
357367. MR 0192513 (33 #738)
[Kro89a] P. B. Kronheimer, The construction of ALE spaces as hyper-Kahler quotients, J.
Dierential Geom. 29 (1989), no. 3, 665683. MR 992334 (90d:53055)
[Kro89b] , A Torelli-type theorem for gravitational instantons, J. Dierential Geom.
29 (1989), no. 3, 685697. MR 992335 (90d:53056)
[Kro90a] , A hyper-K ahlerian structure on coadjoint orbits of a semisimple complex
group, J. London Math. Soc. (2) 42 (1990), no. 2, 193208. MR 1083440 (92b:53031)
[Kro90b] , Instantons and the geometry of the nilpotent variety, J. Dierential Geom.
32 (1990), no. 2, 473490. MR 1072915 (91m:58021)
[Kru97] B. Kruggel, A homotopy classication of certain 7-manifolds, Trans. Amer. Math.
Soc. 349 (1997), no. 7, 28272843. MR 97m:55012
[Kru98] , Kreck-Stolz invariants, normal invariants and the homotopy classication
of generalised Wallach spaces, Quart. J. Math. Oxford Ser. (2) 49 (1998), no. 196,
469485. MR 2000a:55018
592 BIBLIOGRAPHY
[Kru05] , Homeomorphism and dieomorphism classication of Eschenburg spaces,
Q. J. Math. 56 (2005), no. 4, 553577. MR MR2182466 (2006h:53045)
[KS58] K. Kodaira and D. C. Spencer, On deformations of complex analytic structures. I,
II, Ann. of Math. (2) 67 (1958), 328466. MR 22 #3009
[KS60] , On deformations of complex analytic structures. III. Stability theorems for
complex structures, Ann. of Math. (2) 71 (1960), 4376. MR 22 #5991
[KS72] A. Kumpera and D. Spencer, Lie equations. Vol. I: General theory, Princeton Uni-
versity Press, Princeton, N.J., 1972, Annals of Mathematics Studies, No. 73. MR 52
#1805
[KS88] M. Kreck and S. Stolz, A dieomorphism classication of 7-dimensional homoge-
neous Einstein manifolds with SU(3) SU(2) U(1)-symmetry, Ann. of Math. (2)
127 (1988), no. 2, 373388. MR 89c:57042
[KS93a] P. Z. Kobak and A. Swann, Quaternionic geometry of a nilpotent variety, Math.
Ann. 297 (1993), no. 4, 747764. MR 1245417 (94j:53057)
[KS93b] T. Koda and K. Sekigawa, Self-dual Einstein Hermitian surfaces, Progress in dif-
ferential geometry, Adv. Stud. Pure Math., vol. 22, Math. Soc. Japan, Tokyo, 1993,
pp. 123131. MR 1274944 (95b:53056)
[KS98] P. Z. Kobak and A. Swann, Exceptional hyperk ahler reductions, Twistor Newsletter
44 (1998), 2326.
[KS00] G. Kane and M. Shifman (eds.), The supersymmetric world, World Scientic Pub-
lishing Co. Inc., River Edge, NJ, 2000, The beginnings of the theory. MR 1887931
(2003g:81253)
[KSY05] H. Kihara, M. Sakaguchi, and Y. Yasui, Scalar Laplacian on Sasaki-Einstein mani-
folds Y
p,q
, Phys. Lett. B 621 (2005), no. 3-4, 288294. MR 2151750 (2006h:58038)
[KT70] Y. Y. Kuo and S. Tachibana, On the distribution appeared in contact 3-structure,
Taita J. Math. 2 (1970), 1724. MR 46 #8115
[KT87] Franz W. Kamber and Philippe Tondeur, de Rham-Hodge theory for Riemannian
foliations, Math. Ann. 277 (1987), no. 3, 415431. MR 89d:53070
[Kuo70] Y. Kuo, On almost contact 3-structure, Tohoku Math. J. (2) 22 (1970), 325332.
MR 43 #3956
[Kur71] M. Kuranishi, Deformations of compact complex manifolds, Les Presses de
lUniversite de Montreal, Montreal, Que., 1971, Seminaire de Mathematiques
Superieures, No. 39 (

Ete 1969). MR 50 #7588

[KW99] I. R. Klebanov and E. Witten, Superconformal eld theory on threebranes at a
Calabi-Yau singularity, Nuclear Phys. B 536 (1999), no. 1-2, 199218. MR 1666725
(99k:81253)
[KY02a] H. Kanno and Y. Yasui, On Spin(7) holonomy metric based on SU(3)/U(1). I, J.
Geom. Phys. 43 (2002), no. 4, 293309. MR 1929908 (2003m:53076)
[KY02b] , On Spin(7) holonomy metric based on SU(3)/U(1). II, J. Geom. Phys. 43
(2002), no. 4, 310326. MR 1929909 (2003m:53077)
[Lan00] T. Lance, Dierentiable structures on manifolds, Surveys on surgery theory, Vol. 1,
Ann. of Math. Stud., vol. 145, Princeton Univ. Press, Princeton, NJ, 2000, pp. 73
104. MR 2001d:57035
[Laz04a] R. Lazarsfeld, Positivity in algebraic geometry. I, Ergebnisse der Mathematik und
ihrer Grenzgebiete. 3. Folge. A Series of Modern Surveys in Mathematics [Results in
Mathematics and Related Areas. 3rd Series. A Series of Modern Surveys in Math-
ematics], vol. 48, Springer-Verlag, Berlin, 2004, Classical setting: line bundles and
linear series. MR 2095471 (2005k:14001a)
[Laz04b] , Positivity in algebraic geometry. II, Ergebnisse der Mathematik und ihrer
Grenzgebiete. 3. Folge. A Series of Modern Surveys in Mathematics [Results in
Mathematics and Related Areas. 3rd Series. A Series of Modern Surveys in Math-
ematics], vol. 49, Springer-Verlag, Berlin, 2004, Positivity for vector bundles, and
multiplier ideals. MR 2095472 (2005k:14001b)
[LB70] A. Lascoux and M. Berger, Varietes Kahleriennes compactes, Lecture Notes in
Mathematics, Vol. 154, Springer-Verlag, Berlin, 1970. MR 0278248 (43 #3979)
[LB92] H. Lange and Ch. Birkenhake, Complex abelian varieties, Grundlehren der Mathe-
matischen Wissenschaften [Fundamental Principles of Mathematical Sciences], vol.
302, Springer-Verlag, Berlin, 1992. MR 94j:14001
BIBLIOGRAPHY 593
[LdM71] S. Lopez de Medrano, Involutions on manifolds, Springer-Verlag, New York, 1971,
Ergebnisse der Mathematik und ihrer Grenzgebiete, Band 59. MR 0298698 (45
#7747)
[LeB88] C. LeBrun, A rigidity theorem for quaternionic-Kahler manifolds, Proc. Amer.
Math. Soc. 103 (1988), no. 4, 12051208. MR 89h:53105
[LeB91a] , Complete Ricci-at Kahler metrics on C
n
need not be at, Several com-
plex variables and complex geometry, Part 2 (Santa Cruz, CA, 1989), Proc. Sym-
pos. Pure Math., vol. 52, Amer. Math. Soc., Providence, RI, 1991, pp. 297304.
MR 1128554 (93a:53038)
[LeB91b] , Explicit self-dual metrics on CP
2
# #CP
2
, J. Dierential Geom. 34
(1991), no. 1, 223253. MR 1114461 (92g:53040)
[LeB91c] , On complete quaternionic-Kahler manifolds, Duke Math. J. 63 (1991),
no. 3, 723743. MR 92i:53042
[LeB91d] , Scalar-at Kahler metrics on blown-up ruled surfaces, J. Reine Angew.
Math. 420 (1991), 161177. MR 1124569 (92i:53066)
[LeB93] , A niteness theorem for quaternionic-Kahler manifolds with positive scalar
curvature, The Penrose transform and analytic cohomology in representation the-
ory (South Hadley, MA, 1992), Contemp. Math., vol. 154, Amer. Math. Soc., Prov-
idence, RI, 1993, pp. 89101. MR 1 246 379
[LeB95] , Fano manifolds, contact structures, and quaternionic geometry, Internat.
J. Math. 6 (1995), no. 3, 419437. MR 96c:53108
[LeB96] , Four-manifolds without Einstein metrics, Math. Res. Lett. 3 (1996), no. 2,
133147. MR 1386835 (97a:53072)
[LeB99] , Four-dimensional Einstein manifolds, and beyond, Surveys in dierential
geometry: essays on Einstein manifolds, Surv. Dier. Geom., VI, Int. Press, Boston,
MA, 1999, pp. 247285. MR 1798613 (2001m:53072)
[LeB01] , Ricci curvature, minimal volumes, and Seiberg-Witten theory, Invent.
Math. 145 (2001), no. 2, 279316. MR 1872548 (2002h:53061)
[LeB03] , Einstein metrics, four-manifolds, and dierential topology, Surveys in dif-
ferential geometry, Vol. VIII (Boston, MA, 2002), Surv. Dier. Geom., VIII, Int.
Press, Somerville, MA, 2003, pp. 235255. MR 2039991 (2005g:53078)
[Lee96] J. M. Lee, CR manifolds with noncompact connected automorphism groups, J.
Geom. Anal. 6 (1996), no. 1, 7990. MR 1402387 (97h:32013)
[Ler02a] E. Lerman, Contact toric manifolds, J. Symplectic Geom. 1 (2002), no. 4, 785828.
MR 2 039 164
[Ler02b] , A convexity theorem for torus actions on contact manifolds, Illinois J.
Math. 46 (2002), no. 1, 171184. MR 2003g:53146
[Ler03a] , Geodesic ows and contact toric manifolds, Symplectic geometry of
integrable Hamiltonian systems (Barcelona, 2001), Adv. Courses Math. CRM
Barcelona, Birkhauser, Basel, 2003, pp. 175225. MR 2 000 747
[Ler03b] , Maximal tori in the contactomorphism groups of circle bundles over Hirze-
bruch surfaces, Math. Res. Lett. 10 (2003), no. 1, 133144. MR 2004g:53097
[Ler04a] , Contact ber bundles, J. Geom. Phys. 49 (2004), no. 1, 5266. MR 2077244
[Ler04b] , Homotopy groups of K-contact toric manifolds, Trans. Amer. Math. Soc.
356 (2004), no. 10, 40754083 (electronic). MR 2 058 839
[Lib54] P. Libermann, Sur le probl`eme dequivalence de certaines structures innitesimales,
Ann. Mat. Pura Appl. (4) 36 (1954), 27120. MR 0066020 (16,520c)
[Lic53] A. Lichnerowicz, Sur les espaces homog`enes kahleriens, C. R. Acad. Sci. Paris 237
(1953), 695697. MR 0066016 (16,519b)
[Lic57] , Sur les transformations analytiques des varietes kahleriennes compactes,
C. R. Acad. Sci. Paris 244 (1957), 30113013. MR 0094479 (20 #996)
[Lic63a] , Spineurs harmoniques, C. R. Acad. Sci. Paris 257 (1963), 79. MR 0156292
(27 #6218)
[Lic63b] , Theor`emes de reductivite sur des alg`ebres dautomorphismes, Rend. Mat.
e Appl. (5) 22 (1963), 197244. MR 0157329 (28 #564)
[Lie77] S. Lie, Geometrie der Ber uhrungs transformationen, corrected ed., Chelsea Publish-
ing Co., New York, 1977, With editorial assistance by Georg Scheers. MR 0460049
(57 #45)
594 BIBLIOGRAPHY
[LM76] R. Lutz and C. Meckert, Structures de contact sur certaines sph`eres exotiques, C.
R. Acad. Sci. Paris Ser. A-B 282 (1976), no. 11, Aii, A591A593. MR 0397612 (53
#1471)
[LM87] P. Libermann and Ch.-M. Marle, Symplectic geometry and analytical mechanics,
Mathematics and its Applications, vol. 35, D. Reidel Publishing Co., Dordrecht,
1987, Translated from the French by Bertram Eugene Schwarzbach. MR 88c:58016
[LM89] H. B. Lawson, Jr. and M.-L. Michelsohn, Spin geometry, Princeton Mathematical
Series, vol. 38, Princeton University Press, Princeton, NJ, 1989. MR 91g:53001
[Loo84] E. J. N. Looijenga, Isolated singular points on complete intersections, London Math-
ematical Society Lecture Note Series, vol. 77, Cambridge University Press, Cam-
bridge, 1984. MR 747303 (86a:32021)
[Loo01] F. Loose, Reduction in contact geometry, J. Lie Theory 11 (2001), no. 1, 922.
MR 1828281 (2002g:53147)
[LPP04] H. L u, D. N. Page, and C. N. Pope, New inhomogeneous Einstein metrics on sphere
bundles over Einstein-Kahler manifolds, Phys. Lett. B 593 (2004), no. 1-4, 218226.
MR 2076718 (2005f:53063)
[LR83] U. Lindstrom and M. Rocek, Scalar tensor duality and N = 1, 2 nonlinear -
models, Nuclear Phys. B 222 (1983), no. 2, 285308. MR 710273 (85d:81113)
[LS93] C. LeBrun and M. Singer, Existence and deformation theory for scalar-at Kahler
metrics on compact complex surfaces, Invent. Math. 112 (1993), no. 2, 273313.
MR 1213104 (94e:53070)
[LS94] C. LeBrun and S. Salamon, Strong rigidity of positive quaternion-Kahler manifolds,
Invent. Math. 118 (1994), no. 1, 109132. MR 95k:53059
[LT97] E. Lerman and S. Tolman, Hamiltonian torus actions on symplectic orbifolds
and toric varieties, Trans. Amer. Math. Soc. 349 (1997), no. 10, 42014230.
MR 98a:57043
[Lut79] R. Lutz, Sur la geometrie des structures de contact invariantes, Ann. Inst. Fourier
(Grenoble) 29 (1979), no. 1, xvii, 283306. MR 82j:53067
[LW99] C. LeBrun and M. Wang (eds.), Surveys in dierential geometry: essays on Einstein
manifolds, Surveys in Dierential Geometry, VI, International Press, Boston, MA,
1999, Lectures on geometry and topology, sponsored by Lehigh Universitys Journal
of Dierential Geometry. MR 1798603 (2001f:53003)
[LW01] E. Lerman and C. Willett, The topological structure of contact and symplectic
quotients, Internat. Math. Res. Notices (2001), no. 1, 3352. MR 2001j:53112
[Mab87] T. Mabuchi, Einstein-Kahler forms, Futaki invariants and convex geometry on
toric Fano varieties, Osaka J. Math. 24 (1987), no. 4, 705737. MR 89e:53074
[Mac87] K. Mackenzie, Lie groupoids and Lie algebroids in dierential geometry, London
Mathematical Society Lecture Note Series, vol. 124, Cambridge University Press,
Cambridge, 1987. MR 89g:58225
[Mal99] J. Maldacena, The large-N limit of superconformal eld theories and supergravity,
Internat. J. Theoret. Phys. 38 (1999), no. 4, 11131133, Quantum gravity in the
southern cone (Bariloche, 1998). MR 2001h:81246
[Man93] L. Manivel, Un theor`eme de prolongement L
2
de sections holomorphes dun bre
hermitien, Math. Z. 212 (1993), no. 1, 107122. MR 94e:32050
[Mar58] A. Markov, Unsolvability of certain problems in topology, Dokl. Akad. Nauk SSSR
123 (1958), 978980. MR 0103455 (21 #2224)
[Mar59] E. Martinelli, Variet`a a struttura quaternionale generalizzata, Atti Accad. Naz.
Lincei. Rend. Cl. Sci. Fis. Mat. Nat. (8) 26 (1959), 353362. MR 0144288 (26
#1835)
[Mar60] A. A. Markov, Insolubility of the problem of homeomorphy, Proc. Internat. Congress
Math. 1958, Cambridge Univ. Press, New York, 1960, pp. 300306. MR 0115160
(22 #5962)
[Mar70] S. Marchiafava, Sulle variet`a a struttura quaternionale generalizzata, Rend. Mat.
(6) 3 (1970), 529545. MR 0276892 (43 #2632)
[Mar71] J. Martinet, Formes de contact sur les varietes de dimension 3, Proceedings of Liv-
erpool Singularities Symposium, II (1969/1970) (Berlin), Springer, 1971, pp. 142
163. Lecture Notes in Math., Vol. 209. MR 0350771 (50 #3263)
BIBLIOGRAPHY 595
[Mat57a] Y. Matsushima, Sur la structure du groupe dhomeomorphismes analytiques dune
certaine variete kahlerienne, Nagoya Math. J. 11 (1957), 145150. MR 0094478
(20 #995)
[Mat57b] , Sur les espaces homog`enes kahleriens dun groupe de Lie reductif, Nagoya
Math. J. 11 (1957), 5360. MR 19,315c
[Mat72] M. Matsumoto, On 6-dimensional almost Tachibana spaces, Tensor (N.S.) 23
(1972), 250252. MR 0312431 (47 #989)
[Mat02] K. Matsuki, Introduction to the Mori program, Universitext, Springer-Verlag, New
York, 2002. MR MR1875410 (2002m:14011)
[McD84] D. McDu, Examples of simply-connected symplectic non-Kahlerian manifolds, J.
Dierential Geom. 20 (1984), no. 1, 267277. MR 772133 (86c:57036)
[McG77] D. McGavran, T
n
-actions on simply connected (n + 2)-manifolds, Pacic J. Math.
71 (1977), no. 2, 487497. MR 0461542 (57 #1527)
[McK80] J. McKay, Graphs, singularities, and nite groups, The Santa Cruz Conference on
Finite Groups (Univ. California, Santa Cruz, Calif., 1979), Proc. Sympos. Pure
Math., vol. 37, Amer. Math. Soc., Providence, R.I., 1980, pp. 183186. MR 604577
(82e:20014)
[McK81] , Cartan matrices, nite groups of quaternions, and Kleinian singularities,
Proc. Amer. Math. Soc. 81 (1981), no. 1, 153154. MR 589160 (81j:20018)
[Mey73] K. R. Meyer, Symmetries and integrals in mechanics, Dynamical systems (Proc.
Sympos., Univ. Bahia, Salvador, 1971), Academic Press, New York, 1973, pp. 259
272. MR 0331427 (48 #9760)
[MFK94] D. Mumford, J. Fogarty, and F. Kirwan, Geometric invariant theory, third ed.,
Ergebnisse der Mathematik und ihrer Grenzgebiete (2) [Results in Mathematics and
Related Areas (2)], vol. 34, Springer-Verlag, Berlin, 1994. MR 1304906 (95m:14012)
[Mil56a] J. Milnor, Construction of universal bundles. II, Ann. of Math. (2) 63 (1956),
430436. MR 17,1120a
[Mil56b] , On manifolds homeomorphic to the 7-sphere, Ann. of Math. (2) 64 (1956),
399405. MR 18,498d
[Mil63] , Morse theory, Based on lecture notes by M. Spivak and R. Wells. Annals
of Mathematics Studies, No. 51, Princeton University Press, Princeton, N.J., 1963.
MR 29 #634
[Mil65] , Lectures on the h-cobordism theorem, Notes by L. Siebenmann and J.
Sondow, Princeton University Press, Princeton, N.J., 1965. MR 0190942 (32 #8352)
[Mil68] , Singular points of complex hypersurfaces, Annals of Mathematics Studies,
No. 61, Princeton University Press, Princeton, N.J., 1968. MR 39 #969
[Mil75] , On the 3-dimensional Brieskorn manifolds M(p, q, r), Knots, groups, and
3-manifolds (Papers dedicated to the memory of R. H. Fox), Princeton Univ. Press,
Princeton, N. J., 1975, pp. 175225. Ann. of Math. Studies, No. 84. MR 54 #6169
[Mil00] R. J. Milgram, The classication of Alo-Wallach manifolds and their generaliza-
tions, Surveys on surgery theory, Vol. 1, Ann. of Math. Stud., vol. 145, Princeton
Univ. Press, Princeton, NJ, 2000, pp. 379407. MR 2000m:57056
[MM64] H. Matsumura and P. Monsky, On the automorphisms of hypersurfaces, J. Math.
Kyoto Univ. 3 (1963/1964), 347361. MR 0168559 (29 #5819)
[MM03] I. Moerdijk and J. Mrcun, Introduction to foliations and Lie groupoids, Cambridge
Studies in Advanced Mathematics, vol. 91, Cambridge University Press, Cambridge,
2003. MR 2012261
[MM82] S. Mori and S. Mukai, Classication of Fano 3-folds with B
2
2, Manuscripta
Math. 36 (1981/82), no. 2, 147162. MR 83f:14032
[MNS05] A. Moroianu, P.-A. Nagy, and U. Semmelmann, Unit Killing vector elds on nearly
K ahler manifolds, Internat. J. Math. 16 (2005), no. 3, 281301. MR 2130627
(2006e:53056)
[MO70] J. Milnor and P. Orlik, Isolated singularities dened by weighted homogeneous poly-
nomials, Topology 9 (1970), 385393. MR 45 #2757
[MO00] P. Matzeu and L. Ornea, Local almost contact metric 3-structures, Publ. Math.
Debrecen 57 (2000), no. 3-4, 499508. MR 1798730 (2002a:53057)
[MO07] A. Moroianu and L. Ornea, Conformally Einstein products and nearly Kahler man-
ifolds, preprint; arXiv:math.DG/0610599 (2007).
596 BIBLIOGRAPHY
[Moe02] I. Moerdijk, Orbifolds as groupoids: an introduction, Orbifolds in mathematics
and physics (Madison, WI, 2001), Contemp. Math., vol. 310, Amer. Math. Soc.,
Providence, RI, 2002, pp. 205222. MR 2004c:22003
[Mol75] P. Molino, Sur la geometrie transverse des feuilletages, Ann. Inst. Fourier (Greno-
ble) 25 (1975), no. 2, xiv, 279284. MR 52 #11945
[Mol77] , Theorie des G-structures: le probl`eme dequivalence, Springer-Verlag,
Berlin, 1977, Notes redigees avec la collaboration de F. Toupine, Lecture Notes
in Mathematics, Vol. 588. MR 58 #24419
[Mol79] , Feuilletages riemanniens sur les varietes compactes; champs de Killing
transverses, C. R. Acad. Sci. Paris Ser. A-B 289 (1979), no. 7, A421A423.
MR 80j:57026
[Mol88] , Riemannian foliations, Progress in Mathematics, vol. 73, Birkhauser
Boston Inc., Boston, MA, 1988, Translated from the French by Grant Cairns,
With appendices by Cairns, Y. Carri`ere,

E. Ghys, E. Salem and V. Sergiescu.
MR 89b:53054
[Mor75] S. Morita, A topological classication of complex structures on S
1
S
2n1
, Topol-
ogy 14 (1975), 1322. MR 0405444 (53 #9237)
[Mor97] A. Moroianu, Parallel and Killing spinors on Spin
c
manifolds, Comm. Math. Phys.
187 (1997), no. 2, 417427. MR 98i:58245
[Mos57] P. S. Mostert, On a compact Lie group acting on a manifold, Ann. of Math. (2) 65
(1957), 447455. MR 0085460 (19,44b)
[Mos65] J. Moser, On the volume elements on a manifold, Trans. Amer. Math. Soc. 120
(1965), 286294. MR 32 #409
[Mos66] E. Moskal, Contact manifolds of positive curvature, University of Illinois Thesis
(1966).
[MP97] I. Moerdijk and D. A. Pronk, Orbifolds, sheaves and groupoids, K-Theory 12
(1997), no. 1, 321. MR 98i:22004
[MR75] S. Marchiafava and G. Romani, Sui brati con struttura quaternionale generalizzata,
Ann. Mat. Pura Appl. (4) 107 (1975), 131157 (1976). MR 53 #6558
[MS39] S. B. Myers and N. E. Steenrod, The group of isometries of a Riemannian manifold,
Ann. of Math. (2) 40 (1939), no. 2, 400416. MR 1503467
[MS74] J. W. Milnor and J. D. Stashe, Characteristic classes, Princeton University Press,
Princeton, N. J., 1974, Annals of Mathematics Studies, No. 76. MR 0440554 (55
#13428)
[MS85] P. Molino and V. Sergiescu, Deux remarques sur les ots riemanniens, Manuscripta
Math. 51 (1985), no. 1-3, 145161. MR 86h:53035
[MS88] C. C. Moore and C. Schochet, Global analysis on foliated spaces, Mathemati-
cal Sciences Research Institute Publications, vol. 9, Springer-Verlag, New York,
1988, With appendices by S. Hurder, Moore, Schochet and Robert J. Zimmer.
MR 89h:58184
[MS98] D. McDu and D. Salamon, Introduction to symplectic topology, second ed., Oxford
Mathematical Monographs, The Clarendon Press Oxford University Press, New
York, 1998. MR 2000g:53098
[MS99] S. Merkulov and L. Schwachhofer, Classication of irreducible holonomies of
torsion-free ane connections, Ann. of Math. (2) 150 (1999), no. 1, 77149.
MR 1715321 (2000m:53065a)
[MS00] A. Moroianu and U. Semmelmann, Parallel spinors and holonomy groups, J. Math.
Phys. 41 (2000), no. 4, 23952402. MR 1751897 (2001f:53099)
[MS05] D. Martelli and J. Sparks, Toric Sasaki-Einstein metrics on S
2
S
3
, Phys. Lett.
B 621 (2005), no. 1-2, 208212. MR 2152673
[MS06] , Toric geometry, Sasaki-Einstein manifolds and a new innite class of
AdS/CFT duals, Comm. Math. Phys. 262 (2006), no. 1, 5189. MR 2200882
[MSY56] D. Montgomery, H. Samelson, and C. T. Yang, Groups on E
n
with (n 2)-
dimensional orbits, Proc. Amer. Math. Soc. 7 (1956), 719728. MR 0078643
(17,1224a)
[MSY06a] D. Martelli, J. Sparks, and S.-T. Yau, The geometric dual of a-maximisation for
toric Sasaki-Einstein manifolds, Comm. Math. Phys. 268 (2006), no. 1, 3965.
MR 2249795
BIBLIOGRAPHY 597
[MSY06b] , Sasaki-Einstein manifolds and volume minimisation, preprint; arXiv:hep-
th/0603021 (2006).
[MT06] J. W. Morgan and G. Tian, Ricci ow and the Poincare Conjecture, preprint;
arXiv:math.DG/0607607 (2006).
[MU83] S. Mukai and H. Umemura, Minimal rational threefolds, Algebraic geometry
(Tokyo/Kyoto, 1982), Lecture Notes in Math., vol. 1016, Springer, Berlin, 1983,
pp. 490518. MR 726439 (85c:14027)
[Mum61] D. Mumford, The topology of normal singularities of an algebraic surface and a
criterion for simplicity, Inst. Hautes

Etudes Sci. Publ. Math. (1961), no. 9, 522.
MR 0153682 (27 #3643)
[Mum79] , An algebraic surface with K ample, (K
2
) = 9, p
g
= q = 0, Amer. J. Math.
101 (1979), no. 1, 233244. MR 527834 (80j:14032)
[MW74] J. Marsden and A. Weinstein, Reduction of symplectic manifolds with symmetry,
Rep. Mathematical Phys. 5 (1974), no. 1, 121130. MR 0402819 (53 #6633)
[MW88] K. Mikami and A. Weinstein, Moments and reduction for symplectic groupoids,
Publ. Res. Inst. Math. Sci. 24 (1988), no. 1, 121140. MR 944869 (90c:58060)
[MY06] G. Marinescu and N. Yeganefar, Embeddability of some strongly pseudoconvex CR
manifolds, math.CV/0403044 (2006).
[MZ55] D. Montgomery and L. Zippin, Topological transformation groups, Interscience Pub-
lishers, New York-London, 1955. MR 0073104 (17,383b)
[MZ88] M. Miyanishi and D.-Q. Zhang, Gorenstein log del Pezzo surfaces of rank one, J.
Algebra 118 (1988), no. 1, 6384. MR 961326 (89i:14033)
[Nad89] A. M. Nadel, Multiplier ideal sheaves and existence of Kahler-Einstein metrics of
positive scalar curvature, Proc. Nat. Acad. Sci. U.S.A. 86 (1989), no. 19, 72997300.
MR 1015491 (90k:32061)
[Nad90] , Multiplier ideal sheaves and Kahler-Einstein metrics of positive scalar
curvature, Ann. of Math. (2) 132 (1990), no. 3, 549596. MR 92d:32038
[Nad91] , The boundedness of degree of Fano varieties with Picard number one, J.
Amer. Math. Soc. 4 (1991), no. 4, 681692. MR 1115788 (93g:14048)
[Nag92] Y. Nagatomo, Rigidity of c
1
-self-dual connections on quaternionic Kahler mani-
folds, J. Math. Phys. 33 (1992), no. 12, 40204025. MR 93m:53028
[Nag02a] P.-A. Nagy, Nearly Kahler geometry and Riemannian foliations, Asian J. Math. 6
(2002), no. 3, 481504. MR 1946344 (2003m:53043)
[Nag02b] , On nearly-Kahler geometry, Ann. Global Anal. Geom. 22 (2002), no. 2,
167178. MR 1923275 (2003g:53073)
[Nah78] W. Nahm, Supersymmetries and their representations, Nuclear Phys. B 135 (1978),
149166.
[Nah82] , The construction of all self-dual multimonopoles by the ADHM method,
Monopoles in quantum eld theory (Trieste, 1981), World Sci. Publishing, Singa-
pore, 1982, pp. 8794. MR 766754 (86e:53058)
[Nak90] H. Nakajima, Moduli spaces of anti-self-dual connections on ALE gravitational
instantons, Invent. Math. 102 (1990), no. 2, 267303. MR 1074476 (92a:58024)
[Nak98] , Quiver varieties and Kac-Moody algebras, Duke Math. J. 91 (1998), no. 3,
515560. MR MR1604167 (99b:17033)
[Nak99] , Lectures on Hilbert schemes of points on surfaces, University Lecture Se-
ries, vol. 18, American Mathematical Society, Providence, RI, 1999. MR 1711344
(2001b:14007)
[NN57] A. Newlander and L. Nirenberg, Complex analytic coordinates in almost complex
manifolds, Ann. of Math. (2) 65 (1957), 391404. MR 19,577a
[NT88] S. Nishikawa and P. Tondeur, Transversal innitesimal automorphisms for har-
monic Kahler foliations, Tohoku Math. J. (2) 40 (1988), no. 4, 599611. MR 972248
(89j:53031)
[NW04] M. Nieper-Wikirchen, Chern numbers and Rozansky-Witten invariants of compact
hyper-Kahler manifolds, World Scientic Publishing Co. Inc., River Edge, NJ, 2004.
MR 2110899 (2005h:53070)
[Oba56] M. Obata, Ane connections on manifolds with almost complex, quaternion or
Hermitian structure, Jap. J. Math. 26 (1956), 4377. MR 0095290 (20 #1796a)
598 BIBLIOGRAPHY
[Oba62] , Certain conditions for a Riemannian manifold to be iosometric with a
sphere, J. Math. Soc. Japan 14 (1962), 333340. MR 0142086 (25 #5479)
[Oba66] , Riemannian manifolds admitting a solution of a certain system of dier-
ential equations, Proc. U.S.-Japan Seminar in Dierential Geometry (Kyoto, 1965),
Nippon Hyoronsha, Tokyo, 1966, pp. 101114. MR 35 #7263
[Oda88] T. Oda, Convex bodies and algebraic geometry, Ergebnisse der Mathematik und
ihrer Grenzgebiete (3) [Results in Mathematics and Related Areas (3)], vol. 15,
Springer-Verlag, Berlin, 1988, An introduction to the theory of toric varieties, Trans-
lated from the Japanese. MR 922894 (88m:14038)
[Oh83] H. S. Oh, Toral actions on 5-manifolds, Trans. Amer. Math. Soc. 278 (1983), no. 1,
233252. MR 697072 (85b:57043)
[Oku62] M. Okumura, Some remarks on space with a certain contact structure, Tohoku
Math. J. (2) 14 (1962), 135145. MR 26 #708
[Oku68] , On contact metric immersion, Kodai Math. Sem. Rep. 20 (1968), 389409.
MR 0264564 (41 #9156)
[Olm05] C. Olmos, A geometric proof of the Berger holonomy theorem, Ann. of Math. (2)
161 (2005), no. 1, 579588. MR 2150392 (2006c:53045)
[ON66] B. ONeill, The fundamental equations of a submersion, Michigan Math. J. 13
(1966), 459469. MR 0200865 (34 #751)
[ON83] , Semi-Riemannian geometry, Pure and Applied Mathematics, vol. 103, Aca-
demic Press Inc. [Harcourt Brace Jovanovich Publishers], New York, 1983, With
applications to relativity. MR 85f:53002
[OP99] L. Ornea and P. Piccinni, Induced Hopf bundles and Einstein metrics, New devel-
opments in dierential geometry, Budapest 1996, Kluwer Acad. Publ., Dordrecht,
1999, pp. 295305. MR 99k:53099
[Opr77] V. Oproiu, Almost quaternal structures, An. Stiint . Univ. Al. I. Cuza Iasi Sect . I
a Mat. (N.S.) 23 (1977), no. 2, 287298. MR 0493860 (58 #12822)
[Opr84] , Integrability of almost quaternal structures, An. Stiint . Univ. Al. I. Cuza
Iasi Sect . I a Mat. 30 (1984), no. 5, 7584. MR 800155 (86k:53055)
[OR70] P. Orlik and F. Raymond, Actions of the torus on 4-manifolds. I, Trans. Amer.
Math. Soc. 152 (1970), 531559. MR 0268911 (42 #3808)
[OR74] , Actions of the torus on 4-manifolds. II, Topology 13 (1974), 89112.
MR 0348779 (50 #1274)
[OR77] P. Orlik and R. C. Randell, The monodromy of weighted homogeneous singularities,
Invent. Math. 39 (1977), no. 3, 199211. MR 57 #314
[OR04] J.-P. Ortega and T. S. Ratiu, Momentum maps and Hamiltonian reduction,
Progress in Mathematics, vol. 222, Birkhauser Boston Inc., Boston, MA, 2004.
MR 2021152 (2005a:53144)
[Orl70] P. Orlik, Weighted homogeneous polynomials and fundamental groups, Topology 9
(1970), 267273. MR 41 #6251
[Orl72a] , On the homology of weighted homogeneous manifolds, Proceedings of
the Second Conference on Compact Transformation Groups (Univ. Massachusetts,
Amherst, Mass., 1971), Part I (Berlin), Springer, 1972, pp. 260269. Lecture Notes
in Math., Vol. 298. MR 55 #3312
[Orl72b] , Seifert manifolds, Springer-Verlag, Berlin, 1972, Lecture Notes in Mathe-
matics, Vol. 291. MR 54 #13950
[OT87] T. Ohsawa and K. Takegoshi, On the extension of L
2
holomorphic functions, Math.
Z. 195 (1987), no. 2, 197204. MR 88g:32029
[OV05] L. Ornea and M. Verbitsky, An immersion theorem for Vaisman manifolds, Math.
Ann. 332 (2005), no. 1, 121143. MR 2139254
[OV06a] H. Oguri and C. Vafa, On the geometry of the string landscape and the swampland,
preprint; arXiv:hep-th/0605264 (2006).
[OV06b] L. Ornea and M. Verbitsky, Embeddings of compact Sasakian manifolds,
math.DG/0609617 (2006).
[OW71a] P. Orlik and P. Wagreich, Isolated singularities of algebraic surfaces with C

action,
Ann. of Math. (2) 93 (1971), 205228. MR 44 #1662
[OW71b] , Singularities of algebraic surfaces with C

action, Math. Ann. 193 (1971),

121135. MR 45 #1914
BIBLIOGRAPHY 599
[OW75] , Seifert n-manifolds, Invent. Math. 28 (1975), 137159. MR 50 #13596
[OW77] , Algebraic surfaces with k

-action, Acta Math. 138 (1977), no. 1-2, 4381.

MR 57 #336
[OY06a] T. Oota and Y. Yasui, Explicit toric metric on resolved Calabi-Yau cone, Phys.
Lett. B 639 (2006), no. 1, 5456. MR 2244850
[OY06b] , New example of innite family of quiver gauge theories, preprint;
arXiv:hep-th/0610092 (2006).
[OY06c] , Toric Sasaki-Einstein manifolds and Heun equations, Nuclear Phys. B 742
(2006), no. 1-3, 275294. MR 2219526
[Pal57] R. S. Palais, A global formulation of the Lie theory of transformation groups, Mem.
Amer. Math. Soc. No. 22 (1957), iii+123. MR 0121424 (22 #12162)
[Pal61] , On the existence of slices for actions of non-compact Lie groups, Ann. of
Math. (2) 73 (1961), 295323. MR 0126506 (23 #A3802)
[Pal05] S. S. Pal, A new Ricci-at geometry, Phys. Lett. B 614 (2005), no. 3-4, 201206.
MR 2136797 (2006g:53060)
[Par86] J. Parker, 4-dimensional G-manifolds with 3-dimensional orbits, Pacic J. Math.
125 (1986), no. 1, 187204. MR 860758 (88e:57033)
[Ped86] H. Pedersen, Einstein metrics, spinning top motions and monopoles, Math. Ann.
274 (1986), no. 1, 3559. MR 834105 (87i:53070)
[Pen76] R. Penrose, Nonlinear gravitons and curved twistor theory, General Relativity and
Gravitation 7 (1976), no. 1, 3152, The riddle of gravitationon the occasion of the
60th birthday of Peter G. Bergmann (Proc. Conf., Syracuse Univ., Syracuse, N. Y.,
1975). MR 0439004 (55 #11905)
[Per02] G. Pereleman, The entropy formula for the Ricci ow and its geometric applica-
tions, preprint; arXiv:math.DG/0211159 (2002).
[Per03a] , Finite extinction time for the solutions to the Ricci ow on certain three-
manifolds, preprint; arXiv:math.DG/0307245 (2003).
[Per03b] , Ricci ow with surgery on three-manifolds, preprint;
arXiv:math.DG/0303109 (2003).
[Pet98] P. Petersen, Riemannian geometry, Graduate Texts in Mathematics, vol. 171,
Springer-Verlag, New York, 1998. MR 98m:53001
[Pha65] F. Pham, Formules de Picard-Lefschetz generalisees et ramication des integrales,
Bull. Soc. Math. France 93 (1965), 333367. MR 33 #4064
[Poo86] Y. S. Poon, Compact self-dual manifolds with positive scalar curvature, J. Dier-
ential Geom. 24 (1986), no. 1, 97132. MR 857378 (88b:32022)
[PP88] H. Pedersen and Y. S. Poon, Hyper-Kahler metrics and a generalization of the
Bogomolny equations, Comm. Math. Phys. 117 (1988), no. 4, 569580. MR 953820
(89f:53069)
[PP95] , Equivariant connected sums of compact self-dual manifolds, Math. Ann.
301 (1995), no. 4, 717749. MR 1326765 (95m:53069)
[PP98a] H. Pedersen and Y.-S. Poon, Deformations of hypercomplex structures, J. Reine
Angew. Math. 499 (1998), 8199. MR 99h:32009
[PP98b] J. F. Pleba nski and M. Przanowski, Hermite-Einstein four-dimensional mani-
folds with symmetry, Classical Quantum Gravity 15 (1998), no. 6, 17211735.
MR 1628004 (99h:53063)
[PP99] H. Pedersen and Y.-S. Poon, A note on rigidity of 3-Sasakian manifolds, Proc.
Amer. Math. Soc. 127 (1999), no. 10, 30273034. MR 2000a:53075
[PPS98] H. Pedersen, Y. S. Poon, and A. F. Swann, Hypercomplex structures associated
to quaternionic manifolds, Dierential Geom. Appl. 9 (1998), no. 3, 273292.
MR 1661193 (99k:53062)
[PR87] R. Penrose and W. Rindler, Spinors and space-time. Vol. 1, Cambridge Monographs
on Mathematical Physics, Cambridge University Press, Cambridge, 1987, Two-
spinor calculus and relativistic elds. MR 917488 (88h:83009)
[PR88] , Spinors and space-time. Vol. 2, second ed., Cambridge Monographs on
Mathematical Physics, Cambridge University Press, Cambridge, 1988, Spinor and
twistor methods in space-time geometry. MR 944085 (89d:83010)
[Pri67] D. Prill, Local classication of quotients of complex manifolds by discontinuous
groups, Duke Math. J. 34 (1967), 375386. MR 0210944 (35 #1829)
600 BIBLIOGRAPHY
[Pro90] Yu. G. Prokhorov, Automorphism groups of Fano 3-folds, Uspekhi Mat. Nauk 45
(1990), no. 3(273), 195196. MR 1071944 (91j:14034)
[PS91] Y. S. Poon and S. M. Salamon, Quaternionic Kahler 8-manifolds with positive
scalar curvature, J. Dierential Geom. 33 (1991), no. 2, 363378. MR 92b:53071
[PY05] G. Prasad and S-K. Yeung, Fake projective planes, preprint;
arXiv:math.AG/0512115 (2005).
[Ran75] R. C. Randell, The homology of generalized Brieskorn manifolds, Topology 14
(1975), no. 4, 347355. MR 54 #1270
[Rav86] D. C. Ravenel, Complex cobordism and stable homotopy groups of spheres, Pure
and Applied Mathematics, vol. 121, Academic Press Inc., Orlando, FL, 1986.
MR 860042 (87j:55003)
[Ree52] G. Reeb, Sur certaines proprietes topologiques des trajectoires des syst`emes dy-
namiques, Acad. Roy. Belgique. Cl. Sci. Mem. Coll. in 8

27 (1952), no. 9, 64.

MR 0058202 (15,336b)
[Rei59a] B. L. Reinhart, Foliated manifolds with bundle-like metrics, Ann. of Math. (2) 69
(1959), 119132. MR 21 #6004
[Rei59b] , Harmonic integrals on foliated manifolds, Amer. J. Math. 81 (1959), 529
536. MR 0107280 (21 #6005)
[Rei80] M. Reid, Canonical 3-folds, Journees de Geometrie Algebrique dAngers, Juillet
1979/Algebraic Geometry, Angers, 1979, Sijtho & Noordho, Alphen aan den
Rijn, 1980, pp. 273310. MR 82i:14025
[Rei83] B. L. Reinhart, Dierential geometry of foliations, Ergebnisse der Mathematik und
ihrer Grenzgebiete [Results in Mathematics and Related Areas], vol. 99, Springer-
Verlag, Berlin, 1983, The fundamental integrability problem. MR 85i:53038
[Rol76] D. Rolfsen, Knots and links, Publish or Perish Inc., Berkeley, Calif., 1976, Mathe-
matics Lecture Series, No. 7. MR 0515288 (58 #24236)
[RS01] J. Rosenberg and S. Stolz, Metrics of positive scalar curvature and connections with
surgery, Surveys on surgery theory, Vol. 2, Ann. of Math. Stud., vol. 149, Princeton
Univ. Press, Princeton, NJ, 2001, pp. 353386. MR 1818778 (2002f:53054)
[RS05] Y. Rollin and M. Singer, Non-minimal scalar-at Kahler surfaces and parabolic
stability, Invent. Math. 162 (2005), no. 2, 235270. MR 2199006
[Ruk93] P. Rukimbira, Some remarks on R-contact ows, Ann. Global Anal. Geom. 11
(1993), no. 2, 165171. MR 94h:53043
[Ruk94] , The dimension of leaf closures of K-contact ows, Ann. Global Anal.
Geom. 12 (1994), no. 2, 103108. MR 95f:53072
[Ruk95a] , Chern-Hamiltons conjecture and K-contactness, Houston J. Math. 21
(1995), no. 4, 709718. MR 96m:53032
[Ruk95b] , Topology and closed characteristics of K-contact manifolds, Bull. Belg.
Math. Soc. Simon Stevin 2 (1995), no. 3, 349356. MR 96g:53044
[Ruk99] , On K-contact manifolds with minimal number of closed characteristics,
Proc. Amer. Math. Soc. 127 (1999), no. 11, 33453351. MR 2000d:53049
[Sae87] O. Saeki, Knotted homology spheres dened by weighted homogeneous polynomials,
J. Fac. Sci. Univ. Tokyo Sect. IA Math. 34 (1987), no. 1, 4350. MR 88c:57021
[Sal82] S. M. Salamon, Quaternionic Kahler manifolds, Invent. Math. 67 (1982), no. 1,
143171. MR 83k:53054
[Sal84] , Quaternionic structures and twistor spaces, Global Riemannian geome-
try (Durham, 1983), Ellis Horwood Ser. Math. Appl., Horwood, Chichester, 1984,
pp. 6574. MR 757207
[Sal86] , Dierential geometry of quaternionic manifolds, Ann. Sci.

Ecole Norm.
Sup. (4) 19 (1986), no. 1, 3155. MR 87m:53079
[Sal89] , Riemannian geometry and holonomy groups, Pitman Research Notes in
Mathematics Series, vol. 201, Longman Scientic & Technical, Harlow, 1989.
MR 90g:53058
[Sal99] , Quaternion-Kahler geometry, Surveys in dierential geometry: essays on
Einstein manifolds, Surv. Dier. Geom., VI, Int. Press, Boston, MA, 1999, pp. 83
121. MR 1798608 (2001m:53078)
BIBLIOGRAPHY 601
[Sal04] , Explicit metrics with holonomy G
2
, UK-Japan Winter School 2004
Geometry and Analysis Towards Quantum Theory, Sem. Math. Sci., vol. 30, Keio
Univ., Yokohama, 2004, pp. 5058. MR 2131362 (2005k:53070)
[San03] O. P. Santillan, A construction of G
2
holonomy spaces with torus symmetry, Nu-
clear Phys. B 660 (2003), no. 1-2, 169193. MR 1983318 (2004d:53057)
[Sas60] S. Sasaki, On dierentiable manifolds with certain structures which are closely
related to almost contact structure. I, Tohoku Math. J. (2) 12 (1960), 459476.
MR 23 #A591
[Sas65] , Almost contact manifolds, Part 1, Lecture Notes, Mathematical Institute,
Tohoku University (1965).
[Sas67] , Almost contact manifolds, Part 2, Lecture Notes, Mathematical Institute,
Tohoku University (1967).
[Sas68] , Almost contact manifolds, Part 3, Lecture Notes, Mathematical Institute,
Tohoku University (1968).
[Sas85] , Selected papers, Kinokuniya Company Ltd., Tokyo, 1985, With a foreword
by Shiing Shen Chern, Edited by Shun-ichi Tachibana. MR 834384 (87g:01064)
[Sas72] , Spherical space forms with normal contact metric 3-structure, J. Dieren-
tial Geometry 6 (1971/72), 307315. MR 46 #831
[Sat56] I. Satake, On a generalization of the notion of manifold, Proc. Nat. Acad. Sci.
U.S.A. 42 (1956), 359363. MR 18,144a
[Sat57] , The Gauss-Bonnet theorem for V -manifolds, J. Math. Soc. Japan 9 (1957),
464492. MR 20 #2022
[Sat77] H. Sato, Remarks concerning contact manifolds, Tohoku Math. J. 29 (1977), no. 4,
577584. MR 0458334 (56 #16537)
[Sav79] I. V. Savel

ev, Structure of the singularities of a class of complex hypersurfaces,

Mat. Zametki 25 (1979), no. 4, 497503, 635. MR 534292 (80g:32018)
[Sav89] M. V. Saveliev, Integro-dierential nonlinear equations and continual Lie algebras,
Comm. Math. Phys. 121 (1989), no. 2, 283290. MR 985400 (90d:58061)
[Sav02] N. Saveliev, Invariants for homology 3-spheres, Encyclopaedia of Mathematical
Sciences, vol. 140, Springer-Verlag, Berlin, 2002, Low-Dimensional Topology, 1.
MR 2004c:57026
[Sch29] J. A. Schouten,

Uber unitare Geometrie, Nederl. Acad. Wetensch. Indagationes
Math. 32 (1929), 457465.
[Sch95] R. Schoen, On the conformal and CR automorphism groups, Geom. Funct. Anal.
5 (1995), no. 2, 464481. MR 1334876 (96h:53047)
[Sco83] P. Scott, The geometries of 3-manifolds, Bull. London Math. Soc. 15 (1983), no. 5,
401487. MR 84m:57009
[Sco05] A. Scorpan, The wild world of 4-manifolds, American Mathematical Society, Prov-
idence, RI, 2005. MR 2136212
[Seg68] G. Segal, Classifying spaces and spectral sequences, Inst. Hautes

Etudes Sci. Publ.
Math. (1968), no. 34, 105112. MR 0232393 (38 #718)
[Sei32] H. Seifert, Topologie dreidimensionaler gefaserter Raume, Acta Math. 60 (1932),
148238.
[Ser55] J.-P. Serre, Un theor`eme de dualite, Comment. Math. Helv. 29 (1955), 926.
MR 0067489 (16,736d)
[SH62] S. Sasaki and Y. Hatakeyama, On dierentiable manifolds with contact metric
structures, J. Math. Soc. Japan 14 (1962), 249271. MR 25 #4458
[SH76] S. Sasaki and C. J. Hsu, On a property of Brieskorn manifolds, Tohoku Math. J.
(2) 28 (1976), no. 1, 6778. MR 0397614 (53 #1473)
[Sim91] S. R. Simanca, Kahler metrics of constant scalar curvature on bundles over CP
n1
,
Math. Ann. 291 (1991), no. 2, 239246. MR 1129363 (92m:53115)
[Sim92] , A note on extremal metrics of nonconstant scalar curvature, Israel J. Math.
78 (1992), no. 1, 8593. MR 1194961 (93k:58054)
[Sim04] , Canonical metrics on compact almost complex manifolds, Pub-
licacoes Matematicas do IMPA. [IMPA Mathematical Publications], Instituto de
Matematica Pura e Aplicada (IMPA), Rio de Janeiro, 2004. MR 2113907
602 BIBLIOGRAPHY
[Siu87] Y. T. Siu, Lectures on Hermitian-Einstein metrics for stable bundles and
K ahler-Einstein metrics, DMV Seminar, vol. 8, Birkhauser Verlag, Basel, 1987.
MR 89d:32020
[Siu88] , The existence of Kahler-Einstein metrics on manifolds with positive anti-
canonical line bundle and a suitable nite symmetry group, Ann. of Math. (2) 127
(1988), no. 3, 585627. MR 89e:58032
[Sja95] R. Sjamaar, Holomorphic slices, symplectic reduction and multiplicities of repre-
sentations, Ann. of Math. (2) 141 (1995), no. 1, 87129. MR 1314032 (96a:58098)
[SL91] R. Sjamaar and E. Lerman, Stratied symplectic spaces and reduction, Ann. of
Math. (2) 134 (1991), no. 2, 375422. MR 1127479 (92g:58036)
[Sma61] S. Smale, Generalized Poincares conjecture in dimensions greater than four, Ann.
of Math. (2) 74 (1961), 391406. MR 25 #580
[Sma62a] , On the structure of 5-manifolds, Ann. of Math. (2) 75 (1962), 3846.
MR 25 #4544
[Sma62b] , On the structure of manifolds, Amer. J. Math. 84 (1962), 387399.
MR 0153022 (27 #2991)
[Som75] A. J. Sommese, Quaternionic manifolds, Math. Ann. 212 (1974/75), 191214.
MR 0425827 (54 #13778)
[Sou70] J.-M. Souriau, Structure des syst`emes dynamiques, Maitrises de mathematiques,
Dunod, Paris, 1970. MR 0260238 (41 #4866)
[Sou97] , Structure of dynamical systems, Progress in Mathematics, vol. 149,
Birkhauser Boston Inc., Boston, MA, 1997, A symplectic view of physics, Trans-
lated from the French by C. H. Cushman-de Vries, Translation edited and with a
preface by R. H. Cushman and G. M. Tuynman. MR 1461545 (98i:58103)
[Spe62] D. C. Spencer, Deformation of structures on manifolds dened by transitive, con-
tinuous pseudogroups. I. Innitesimal deformations of structure, Ann. of Math. (2)
76 (1962), 306398. MR 27 #6287a
[SS65] I. M. Singer and S. Sternberg, The innite groups of Lie and Cartan. I. The tran-
sitive groups, J. Analyse Math. 15 (1965), 1114. MR 36 #911
[SS85] B. Shiman and A. J. Sommese, Vanishing theorems on complex manifolds,
Progress in Mathematics, vol. 56, Birkhauser Boston Inc., Boston, MA, 1985.
MR 782484 (86h:32048)
[SSTVP88] P. Spindel, A. Sevrin, W. Troost, and A. Van Proeyen, Extended supersymmetric -
models on group manifolds. I. The complex structures, Nuclear Phys. B 308 (1988),
no. 2-3, 662698. MR 967938 (89m:53138)
[ST71] M. Sebastiani and R. Thom, Un resultat sur la monodromie, Invent. Math. 13
(1971), 9096. MR 0293122 (45 #2201)
[ST76] S. Sasaki and T. Takahashi, Almost contact structures on Brieskorn manifolds,
Tohoku Math. J. (2) 28 (1976), no. 4, 619624. MR 0428340 (55 #1365)
[ST80] H. Seifert and W. Threlfall, Seifert and Threlfall: a textbook of topology, Pure and
Applied Mathematics, vol. 89, Academic Press Inc. [Harcourt Brace Jovanovich
Publishers], New York, 1980, Translated from the German edition of 1934 by
Michael A. Goldman, With a preface by Joan S. Birman, With Topology of 3-
dimensional bered spaces by Seifert, Translated from the German by Wolfgang
Heil. MR 575168 (82b:55001)
[Ste51] N. Steenrod, The Topology of Fibre Bundles, Princeton Mathematical Series, vol.
14, Princeton University Press, Princeton, N. J., 1951. MR 12,522b
[Ste83] S. Sternberg, Lectures on dierential geometry, second ed., Chelsea Publishing
Co., New York, 1983, With an appendix by Sternberg and Victor W. Guillemin.
MR 88f:58001
[Sto74] R. E. Stong, Contact manifolds, J. Dierential Geometry 9 (1974), 219238. MR 49
#6258
[Sto92] S. Stolz, Simply connected manifolds of positive scalar curvature, Ann. of Math.
(2) 136 (1992), no. 3, 511540. MR 1189863 (93i:57033)
[SU02] Y. Shimizu and K. Ueno, Advances in moduli theory, Translations of Mathemati-
cal Monographs, vol. 206, American Mathematical Society, Providence, RI, 2002,
Translated from the 1999 Japanese original, Iwanami Series in Modern Mathemat-
ics. MR 1865412 (2002j:14017)
BIBLIOGRAPHY 603
[Sul77] D. Sullivan, Innitesimal computations in topology, Inst. Hautes

Etudes Sci. Publ.
Math. (1977), no. 47, 269331 (1978). MR 0646078 (58 #31119)
[Sus03] L. Susskind, The anthropic landscape of string theory, preprint; hep-th/0302219
(2003).
[Sus05] , The cosmic landscape: String theory and the illusion of intelligent design,
Little, Brown and Company, 2005.
[SvD30] J. A. Schouten and D. van Dantzig,

Uber unitare Geometrie, Math. Ann. 103
(1930), no. 1, 319346. MR 1512625
[SW99] B. Steer and A. Wren, Grothendieck topology and the Picard group of a complex
orbifold, Homotopy invariant algebraic structures (Baltimore, MD, 1998), Contemp.
Math., vol. 239, Amer. Math. Soc., Providence, RI, 1999, pp. 251262. MR 1718086
(2000j:14029)
[Swa89] A. F. Swann, Quaternionic Kahler geometry and the fundamental 4-form, Pro-
ceedings of the Workshop on Curvature Geometry (Lancaster, 1989) (Lancaster),
ULDM Publ., 1989, pp. 165173. MR 1089891 (91m:53053)
[Swa91] A. Swann, Hyper-K ahler and quaternionic Kahler geometry, Math. Ann. 289
(1991), no. 3, 421450. MR 92c:53030
[Swa97] , Singular moment maps and quaternionic geometry, Symplectic singular-
ities and geometry of gauge elds (Warsaw, 1995), Banach Center Publ., vol. 39,
Polish Acad. Sci., Warsaw, 1997, pp. 143153. MR 1458656 (98m:53066)
[SY91] J.-P. Sha and D. Yang, Positive Ricci curvature on the connected sums of S
n
S
m
,
J. Dierential Geom. 33 (1991), no. 1, 127137. MR 92f:53048
[SZ05] K. Sfetsos and D. Zoakos, Supersymmetric solutions based on Y
p,q
and L
p,q,r
,
Phys. Lett. B 625 (2005), no. 1-2, 135144. MR 2168911 (2006g:53062)
[Tac59] S. Tachibana, On almost-analytic vectors in certain almost-Hermitian manifolds,
Tohoku Math. J. (2) 11 (1959), 351363. MR 0117753 (22 #8527)
[Tac65] , On harmonic tensors in compact Sasakian spaces, Tohoku Math. J. (2) 17
(1965), 271284. MR 32 #8288
[Tak78] T. Takahashi, Deformations of Sasakian structures and its application to the
Brieskorn manifolds, Tohoku Math. J. (2) 30 (1978), no. 1, 3743. MR 81e:53024
[Tak00] S. Takayama, Simple connectedness of weak Fano varieties, J. Algebraic Geom. 9
(2000), no. 2, 403407. MR 1735807 (2001d:14042)
[Tam92] I. Tamura, Topology of foliations: an introduction, Translations of Mathematical
Monographs, vol. 97, American Mathematical Society, Providence, RI, 1992, Trans-
lated from the 1976 Japanese edition and with an afterword by Kiki Hudson, With
a foreword by Takashi Tsuboi. MR 93c:57021
[Tan62] S. Tanno, Note on innitesimal transformations over contact manifolds, Tohoku
Math. J. (2) 14 (1962), 416430. MR 0145441 (26 #2972)
[Tan63a] , Some transformations on manifolds with almost contact and contact metric
structures, Tohoku Math. J. (2) 15 (1963), 140147. MR 0150716 (27 #703)
[Tan63b] , Some transformations on manifolds with almost contact and contact metric
structures. II, Tohoku Math. J. (2) 15 (1963), 322331. MR 0161277 (28 #4485)
[Tan64] , A remark on transformations of a K-contact manifold, Tohoku Math. J.
(2) 16 (1964), 173175. MR 0172210 (30 #2430)
[Tan67] , Harmonic forms and Betti numbers of certain contact Riemannian mani-
folds, J. Math. Soc. Japan 19 (1967), 308316. MR 35 #3604
[Tan68] , The topology of contact Riemannian manifolds, Illinois J. Math. 12 (1968),
700717. MR 38 #2803
[Tan69a] , The automorphism groups of almost contact Riemannian manifolds,
Tohoku Math. J. (2) 21 (1969), 2138. MR 39 #3428
[Tan69b] , Sasakian manifolds with constant -holomorphic sectional curvature,
Tohoku Math. J. (2) 21 (1969), 501507. MR 40 #4894
[Tan70] , On the isometry groups of Sasakian manifolds, J. Math. Soc. Japan 22
(1970), 579590. MR 42 #6755
[Tan71] , Killing vectors on contact Riemannian manifolds and berings related to
the Hopf brations, Tohoku Math. J. (2) 23 (1971), 313333. MR 44 #4681
604 BIBLIOGRAPHY
[Tan79] , Geodesic ows on C
L
-manifolds and Einstein metrics on S
3
S
2
, Minimal
submanifolds and geodesics (Proc. Japan-United States Sem., Tokyo, 1977), North-
Holland, Amsterdam, 1979, pp. 283292. MR 81g:58027
[Tan96] , Remarks on a triple of K-contact structures, Tohoku Math. J. (2) 48
(1996), no. 4, 519531. MR MR1419082 (97i:53055)
[Tau92] C. H. Taubes, The existence of anti-self-dual conformal structures, J. Dierential
Geom. 36 (1992), no. 1, 163253. MR 1168984 (93j:53063)
[Tho69] E. Thomas, Vector elds on manifolds, Bull. Amer. Math. Soc. 75 (1969), 643683.
MR 0242189 (39 #3522)
[Tho76] C. B. Thomas, Almost regular contact manifolds, J. Dierential Geometry 11
(1976), no. 4, 521533. MR 57 #4195
[Thu76] W. P. Thurston, Some simple examples of symplectic manifolds, Proc. Amer. Math.
Soc. 55 (1976), no. 2, 467468. MR 53 #6578
[Thu79] W. Thurston, The geometry and topology of 3-manifolds, Mimeographed Notes,
Princeton University, 1979.
[Thu97] W. P. Thurston, Three-dimensional geometry and topology. Vol. 1, Princeton Math-
ematical Series, vol. 35, Princeton University Press, Princeton, NJ, 1997, Edited by
S. Levy. MR 97m:57016
[Tia87] G. Tian, On Kahler-Einstein metrics on certain Kahler manifolds with C
1
(M) > 0,
Invent. Math. 89 (1987), no. 2, 225246. MR 88e:53069
[Tia90] , On Calabis conjecture for complex surfaces with positive rst Chern class,
Invent. Math. 101 (1990), no. 1, 101172. MR 91d:32042
[Tia97] , Kahler-Einstein metrics with positive scalar curvature, Invent. Math. 130
(1997), no. 1, 137. MR 99e:53065
[Tia99] , Kahler-Einstein manifolds of positive scalar curvature, Surveys in dier-
ential geometry: essays on Einstein manifolds, Surv. Dier. Geom., VI, Int. Press,
Boston, MA, 1999, pp. 6782. MR 2001m:32052
[Tia00] , Canonical metrics in Kahler geometry, Lectures in Mathematics
ETH Z urich, Birkhauser Verlag, Basel, 2000, Notes taken by Meike Akveld.
MR 2001j:32024
[TO66] S. Tachibana and Y. Ogawa, On the second Betti number of a compact Sasakian
space, Natur. Sci. Rep. Ochanomizu Univ. 17 (1966), 2732. MR 0214096 (35
#4947)
[Tod94] K. P. Tod, Self-dual Einstein metrics from the Painleve VI equation, Phys. Lett.
A 190 (1994), no. 3-4, 221224. MR 1285788 (95c:83020)
[Tod95] , Scalar-at Kahler and hyper-Kahler metrics from Painleve-III, Classical
Quantum Gravity 12 (1995), no. 6, 15351547. MR 1344288 (96f:53068)
[Tod97] , The SU()-Toda eld equation and special four-dimensional metrics, Ge-
ometry and physics (Aarhus, 1995), Lecture Notes in Pure and Appl. Math., vol.
184, Dekker, New York, 1997, pp. 307312. MR 1423177 (98a:53068)
[Ton97] P. Tondeur, Geometry of foliations, Monographs in Mathematics, vol. 90,
Birkhauser Verlag, Basel, 1997. MR 98d:53037
[Tsu88] H. Tsuji, Logarithmic Fano manifolds are simply connected, Tokyo J. Math. 11
(1988), no. 2, 359362. MR 976571 (90b:14050)
[TW79] K. P. Tod and R. S. Ward, Self-dual metrics with self-dual Killing vectors, Proc.
Roy. Soc. London Ser. A 368 (1979), no. 1734, 411427. MR 551471 (81c:83011)
[TW03] S. Tolman and J. Weitsman, The cohomology rings of symplectic quotients, Comm.
Anal. Geom. 11 (2003), no. 4, 751773. MR 2015175 (2004k:53140)
[TY70] S. Tachibana and W. N. Yu, On a Riemannian space admitting more than one
Sasakian structures, Tohoku Math. J. (2) 22 (1970), 536540. MR 43 #1086
[TY87] G. Tian and S.-T. Yau, Kahler-Einstein metrics on complex surfaces with C
1
> 0,
Comm. Math. Phys. 112 (1987), no. 1, 175203. MR 88k:32070
[Udr69] C. Udriste, Structures presque coquaternioniennes, Bull. Math. Soc. Sci. Math. R.
S. Roumanie (N.S.) 13(61) (1969), no. 4, 487507 (1970). MR 45 #5908
[Ust99] I. Ustilovsky, Innitely many contact structures on S
4m+1
, Internat. Math. Res.
Notices (1999), no. 14, 781791. MR 1704176 (2000f:57028)
[Vaf05] C. Vafa, The string landscape and the swampland, preprint; arXiv:hep-th/0509212
(2005).
BIBLIOGRAPHY 605
[Vai82] I. Vaisman, Generalized Hopf manifolds, Geom. Dedicata 13 (1982), no. 3, 231255.
MR 690671 (84g:53096)
[Var04] V. S. Varadarajan, Supersymmetry for mathematicians: an introduction, Courant
Lecture Notes in Mathematics, vol. 11, New York University Courant Institute of
Mathematical Sciences, New York, 2004. MR 2069561 (2005g:58011)
[vC06] C. van Coevering, Toric surfaces and Sasaki-Einstein 5-manifolds, SUNY at Stony
Brook Ph.D. Thesis; see also preprint, arXiv:math.DG/0607721 (2006).
[VK99] M. Verbitsky and D. Kaledin, Hyperk ahler manifolds, Mathematical Physics
(Somerville), vol. 12, International Press, Somerville, MA, 1999. MR 1815021
(2002f:53079)
[vK04] O. van Koert, Contact homology of Brieskorn manifolds, preprint;
arXiv:math.AG/0410208 (2004).
[Voi02] C. Voisin, Hodge theory and complex algebraic geometry. I, Cambridge Studies in
Advanced Mathematics, vol. 76, Cambridge University Press, Cambridge, 2002,
Translated from the French original by Leila Schneps. MR 1967689 (2004d:32020)
[Voi04] , On the homotopy types of compact Kahler and complex projective mani-
folds, Invent. Math. (2004), no. 157, 329343.
[Wad75] A. W. Wadsley, Geodesic foliations by circles, J. Dierential Geometry 10 (1975),
no. 4, 541549. MR 53 #4092
[Wak58] H. Wakakuwa, On Riemannian manifolds with homogeneous holonomy group
Sp(n), Tohoku Math. J. (2) 10 (1958), 274303. MR 0106488 (21 #5220)
[Wal62] C. T. C. Wall, Classication of (n 1)-connected 2n-manifolds, Ann. of Math. (2)
75 (1962), 163189. MR 0145540 (26 #3071)
[Wal67a] F. Waldhausen, Eine Klasse von 3-dimensionalen Mannigfaltigkeiten. I, II, Invent.
Math. 3 (1967), 308333; ibid. 4 (1967), 87117. MR 0235576 (38 #3880)
[Wal67b] C. T. C. Wall, Classication problems in dierential topology. VI. Classication of
(s 1)-connected (2s + 1)-manifolds, Topology 6 (1967), 273296. MR 35 #7343
[Wan54] H.-C. Wang, Closed manifolds with homogeneous complex structure, Amer. J. Math.
76 (1954), 132. MR 0066011 (16,518a)
[Wan82] M. Y. Wang, Some examples of homogeneous Einstein manifolds in dimension
seven, Duke Math. J. 49 (1982), no. 1, 2328. MR 83k:53069
[Wan89] , Parallel spinors and parallel forms, Ann. Global Anal. Geom. 7 (1989),
no. 1, 5968. MR 91g:53053
[Wan95] , On non-simply connected manifolds with non-trivial parallel spinors, Ann.
Global Anal. Geom. 13 (1995), no. 1, 3142. MR 96a:53066
[War90] R. S. Ward, Einstein-Weyl spaces and SU() Toda elds, Classical Quantum Grav-
ity 7 (1990), no. 4, L95L98. MR 1045295 (91g:83019)
[WB82] E. Witten and J. Bagger, Quantization of Newtons constant in certain supergravity
theories, Phys. Lett. B 115 (1982), no. 3, 202206. MR 669354 (83m:81107)
[Web77] S. M. Webster, On the transformation group of a real hypersurface, Trans. Amer.
Math. Soc. 231 (1977), no. 1, 179190. MR 0481085 (58 #1231)
[Web78] , Pseudo-Hermitian structures on a real hypersurface, J. Dierential Geom.
13 (1978), no. 1, 2541. MR 80e:32015
[Wei58] A. Weil, Introduction `a letude des varietes kahleriennes, Publications de lInstitut
de Mathematique de lUniversite de Nancago, VI. Actualites Sci. Ind. no. 1267,
Hermann, Paris, 1958. MR 0111056 (22 #1921)
[Wei77a] A. Weinstein, Lectures on symplectic manifolds, American Mathematical Society,
Providence, R.I., 1977, Expository lectures from the CBMS Regional Conference
held at the University of North Carolina, March 812, 1976, Regional Conference
Series in Mathematics, No. 29. MR 57 #4244
[Wei77b] , Symplectic V -manifolds, periodic orbits of Hamiltonian systems, and the
volume of certain Riemannian manifolds, Comm. Pure Appl. Math. 30 (1977),
no. 2, 265271. MR 56 #13260
[Wei79] , On the hypotheses of Rabinowitz periodic orbit theorems, J. Dierential
Equations 33 (1979), no. 3, 353358. MR 81a:58030b
[Wei80] , Fat bundles and symplectic manifolds, Adv. in Math. 37 (1980), no. 3,
239250. MR 82a:53038
606 BIBLIOGRAPHY
[Wel80] R. O. Wells, Jr., Dierential analysis on complex manifolds, second ed., Gradu-
ate Texts in Mathematics, vol. 65, Springer-Verlag, New York, 1980. MR 608414
(83f:58001)
[WG68a] J. A. Wolf and A. Gray, Homogeneous spaces dened by Lie group automorphisms.
I, J. Dierential Geometry 2 (1968), 77114. MR 0236328 (38 #4625a)
[WG68b] , Homogeneous spaces dened by Lie group automorphisms. II, J. Dierential
Geometry 2 (1968), 115159. MR 0236329 (38 #4625b)
[Wil72] D. L. Wilkens, Closed (s 1)-connected (2s + 1)-manifolds, s = 3, 7, Bull. London
Math. Soc. 4 (1972), 2731. MR 46 #6378
[Wil02] C. Willett, Contact reduction, Trans. Amer. Math. Soc. 354 (2002), no. 10, 4245
4260 (electronic). MR 1926873 (2003m:53152)
[Win83] H. E. Winkelnkemper, The graph of a foliation, Ann. Global Anal. Geom. 1 (1983),
no. 3, 5175. MR 85j:57043
[Wis91] J. A. Wisniewski, On Fano manifolds of large index, Manuscripta Math. 70 (1991),
no. 2, 145152. MR 1085628 (92d:14033)
[Wit83] E. Witten, Global aspects of current algebra, Nuclear Phys. B 223 (1983), no. 2,
422432. MR 717915 (86i:81103)
[Wol65] J. A. Wolf, Complex homogeneous contact manifolds and quaternionic symmetric
spaces, J. Math. Mech. 14 (1965), 10331047. MR 32 #3020
[Wra97] D. Wraith, Exotic spheres with positive Ricci curvature, J. Dierential Geom. 45
(1997), no. 3, 638649. MR 98i:53058
[WZ71] J. Wess and B. Zumino, Consequences of anomalous Ward identities, Phys. Lett.
37B (1971), 9597. MR 0342064 (49 #6810)
[WZ86] M. Y. Wang and W. Ziller, Einstein metrics with positive scalar curvature, Curva-
ture and topology of Riemannian manifolds (Katata, 1985), Lecture Notes in Math.,
vol. 1201, Springer, Berlin, 1986, pp. 319336. MR 859594 (87k:53114)
[WZ90] , Einstein metrics on principal torus bundles, J. Dierential Geom. 31
(1990), no. 1, 215248. MR 91f:53041
[Yam99] T. Yamazaki, A construction of K-contact manifolds by a ber join, Tohoku Math.
J. (2) 51 (1999), no. 4, 433446. MR 2001e:53094
[Yam01] , On a surgery of K-contact manifolds, Kodai Math. J. 24 (2001), no. 2,
214225. MR 2002c:57048
[Yan63] K. Yano, On a structure dened by a tensor eld f of type (1, 1) satisfying f
3
+f =
0, Tensor (N.S.) 14 (1963), 99109. MR 0159296 (28 #2513)
[Yan65] , Dierential geometry on complex and almost complex spaces, International
Series of Monographs in Pure and Applied Mathematics, Vol. 49, A Pergamon Press
Book. The Macmillan Co., New York, 1965. MR 0187181 (32 #4635)
[Yau77] S.-T. Yau, Calabis conjecture and some new results in algebraic geometry, Proc.
Nat. Acad. Sci. U.S.A. 74 (1977), no. 5, 17981799. MR 0451180 (56 #9467)
[Yau78] , On the Ricci curvature of a compact Kahler manifold and the complex
Monge-Amp`ere equation. I, Comm. Pure Appl. Math. 31 (1978), no. 3, 339411.
MR 81d:53045
[YK84] K. Yano and M. Kon, Structures on manifolds, Series in Pure Mathematics, vol. 3,
World Scientic Publishing Co., Singapore, 1984. MR 86g:53001
[YY02] S. S.-T. Yau and Y. Yu, Algebraic classication of rational CR structures on
topological 5-sphere with transversal holomorphic S
1
-action in C
4
, Math. Nachr.
246/247 (2002), 207233. MR 1944558 (2003j:32042)
[YY05] , Classication of 3-dimensional isolated rational hypersurface singular-
ities with C

-action, Rocky Mountain J. Math. 35 (2005), no. 5, 17951809.

MR 2206037 (2006j:32034)
[Zhe00] F. Zheng, Complex dierential geometry, AMS/IP Studies in Advanced Mathemat-
ics, vol. 18, American Mathematical Society, Providence, RI, 2000. MR 1777835
(2001i:32035)
[Zil82] W. Ziller, Homogeneous Einstein metrics on spheres and projective spaces, Math.
Ann. 259 (1982), no. 3, 351358. MR 84h:53062
[Zum79] B. Zumino, Supersymmetry and Kahler manifolds, Phys. Lett. B 87 (1979), 203
207.
Index
3-Sasakian diamond, 486
3-Sasakian moment map, 499
3-symmetric spaces, 540
D = 11 supergravity, 546
G-manifold, 38
G-structure, 23
Q-Cartier, 125
Q-factorial, 125, 157
-atlas, 37
i

-lemmas, 81
transverse, 239
etale space, 10
3-Sasakian, 473
action, 39
free, 39
local type, 44
locally free, 39
locally proper, 42
proper, 39
transitive, 42
adjunction, 142
AdS/CFT Duality Conjecture, 547
Alekseevskian spaces, 438
ample, 99
automorphism
G structure, 40
Barden invariant, 336
Barden theorem, 336
basic p-from, 59
basic cohomology, 59
Beauville Conjecture, 437
Berger theorem, 26
Betti numbers
basic, 216
Bishop theorem, 384
blowing-up, 97
Boothby-Wang Theorem, 209
bundle
associated, 17
frame, 15
holomorphic line, 95
holomorphic vector, 11
line, 17, 95
principal, 14
spin, 18
vector, 17
Calabi conjecture, 102, 154
Calabi metrics on T

CP
n
, 459
Calabi-Yau manifold, 104
Calabi-Yau orbifold, 129, 165
Calabi-Yau Theorem, 103, 133
characteristic
Euler Poincare, 93
orbifold Euler, 121
characteristic classes, 18
characteristic foliation
irregular, 188
non-regular, 188
quasi-regular, 188
regular, 188
characteristic hyperplane, 285
Chern classes, 243, 557
Chern classes, basic, 243
Chern form, 92
class
characteristic, 556
Chern, 92, 557
Euler, 557
integral Pontrjagin, 557
integral Stiefel Whitney, 557
orbifold canonical, 127
orbifold characteristic, 120
orbifold Chern class, 121
orbifold Euler, 121
orbifold rst Chern class, 128
orbifold Pontrjagin, 121
rational Pontrjagin, 557
real characteristic, 121
Stiefel Whitney, 557
Todd, 92
classifying space, 117, 555
cocycle, 15
cocyle
Haeiger, 53
cohomology
607
608 INDEX
group, 118
commuting sheaf, 212
complex structure
transverse, 68
cone, 201
metric, 201
sine, 543
symplectic, 202, 203
connection, 19
Koszul, 21
Levi Civita, 26
linear, 24
Obata, 35
Oproiu, 35
torsionfree, 24
contact
almost, 190192, 195, 202, 474, 475
complex, 184
distribution, 182
form, 180, 182
Hamiltonian, 190
line bundle, 182
manifold, 181
co-oriented, 183
metric manifold, 199
metric structure, 199
pseudogroup, 180
strict transformation group, 189
structure, 182
subbundle, 182
transformation, 180
innitesimal, 181
strict, 180, 181
transformation group, 189
contact type, 186
convex polytope, 279
covariant derivative, 21
exterior, 20
covariantly constant, 21
cuplength, 229
curvature, 221
2-form, 20
-sectional, 225
Ricci, 223
Riemannian, 26, 221
scalar, 372
sectional, 221, 225, 233
deformation
characteristic foliation, 269
transverse holomorphic structure, 267
type I, 269
type II, 240
dimension
algebraic, 96
Iataka, 96
Kodaira, 96
divisor, 100, 123
Q, 125
absolute, 126
Baily, 126
big, 102
branch, 125
Cartier, 100, 124
class group, 124
eective, 100
exceptional, 97
nef, 102
orbifold canonical, 127
ramication, 127
Weil, 100
Eguchi-Hanson metric, 460
Eilenberg-MacLane space, 117, 118
elliptic
Lie algebra, 41
elliptic bration, 361
equivalence
linear, 101
numerical, 101
Euler characteristic
basic, 216
basic holomorphic, 218
holomorphic, 92, 96, 104
Euler class, 557
exotic
contact, 324
fan, 282
bration
Boothby-Wang, 188
ow
Anosov, 70
Riemannian, 69
foliated atlas, 31
foliated coordinate chart, 31
foliation, 31
characteristic, 187, 196, 198200
Riemannian, 62
simple, 51, 55
singular Riemannian, 70
form
Kahler, 78
function
transition, 16
fundamental basic class, 216
genus, 93
basic arithmetic, 218
basic geometric, 218
geometric, 96
Todd, 92
gravitational instantons, 456
gravitational multi-instantons, 462
group
Neron-Severi, 95
Picard, 95
INDEX 609
groupoid, 38, 53
etale, 38, 115, 553
action, 552
holonomy, 54
Lie, 552
translation, 553
Hard Lefschetz Theorem
transverse, 218
Heisenberg group, 181, 229, 261, 275
Hirzebruch signature
basic, 218
Hodge numbers, 8991
basic, 218
Hodge Theorem, 89
holonomy
covering, 56
group, 22
leaf, 54
pseudogroup, 55
holonomy group (of a leaf), 55
holonomy groupoid, 54
homologous
a, 242, 248
hypercomplex, 35, 453
hyperhamiltonian T
n
-space, 461
hyperhamiltonian action, 458
hyperkahler, 36
hyperkahler manifold
Asymptotically Locally Euclidean, 468
locally toric, 461
hypersurface
weighted, 139
index
divisibility, 131
Fano, 131
ramication, 125
injections, 106, 109
integrable, 23
irregular, 188
isotropy subgroup, 106
K-contact, 200, 203, 206, 207, 211, 221
structure theorem, 212
K-contact manifold, 200
Kahler
almost, 203
Kahler cone, 102
Kahler form, 29
Kahler potential, 81
Killing spinor, 530
Kirwan map, 278
Kleinian singularities, 467
leaf
closure, 211
generic, 55
leaf holonomy, 107
leaves, 32
space of, 32
LeBrun-Salamon Conjecture, 436
Legendre transform, 461
lens space, 264
line bundle
canonical, 96
link, 299
local uniformizing groups, 106
local uniformizing system, 105
log del Pezzo surfaces, 128
log resolution, 157
Lusternik-Schnirelman theory, 232
Maldacena Conjecture, 548
manifold
3-Sasakian, 477
homogeneous, 42
almost Hermitian, 78
almost hyperkahler, 455
complex, 76
complex ag, 81
Fano, 98
Hermitian, 78
Hodge, 99
hyperkahler, 455
Kahler, 79
nearly Kahler, 539
polarized Kahler, 96
quaternionic Kahler, 429
Vaisman, 250
Wang-Ziller, 254, 255, 275, 294, 392
Marchiafava-Romani class, 428
McKay correspondence, 465
meromorphic function, 100
metric
-Einstein, 370, 417
canonical Kahler, 152
Einstein, 370
Hermitian, 78, 93
Kahler, 79
Lorentzian, 29, 420
warped product, 201
Milnor number, 300
minimal model program, 341
mirror symmetry, 166
moment map, 277
contact, 284
hypercomplex, 454
hyperkahler, 458
quaternionic Kahler, 439
symplectic cone, 283
Monge-Amp`ere equation, 104, 153
Nijenhuis tensor, 28, 76, 203205
normal, 203
number
Bernoulli, 313
Chern, 93
610 INDEX
orbifold Picard, 131
Picard, 95
ONeill tensor, 65, 220, 483
obstruction
Bishop, 384
Lichnerowicz, 386
orbibundle, 110
anticanionical, 127
canonical, 127
orbifold, 105
classifying space, 117
developable, 107
Fano, 128
polarized Kahler, 132
quaternionic, 431
orbifold charts, 105
orbifold G-structure, 114
orbisheaf, 108
multiplier ideal, 157, 158
structure, 109
orbit, 39
exceptional, 46
principal, 46
regular, 46
singular, 46
type, 43
order
K-contact, 207
of an orbifold, 107
p-brane, 547
Painleve VI equation, 446, 447
parallel translation, 22
perturbation
weighted homogeneous, 141
Picard group, 125
of an orbifold, 130
plurigenera, 96
Poincare homology sphere, 333
Poisson bracket, 277
polyhedral cone, 283
polynomial
Brieskorn-Pham, 130, 141, 167
weighted homogeneous, 139, 300
polytope, 279, 281
Delzant, 280
lt, 280
presheaf, 9
primitive cohomology, 91
prolongation, 41
pseudo-Riemannian metric, 29
pseudogroup, 36
transitive, 37
quasi-regular, 188
quasi-smoothness, 140
quaternionic Kahler, 35
quotient
hypercomplex, 454
hyperkahler, 458
rational map, 97
reduction
contact, 290
hypercomplex, 454
hyperkahler, 458
quaternionic Kahler, 440
singular symplectic, 277, 459
symplectic, 276, 277
Reeb ow, 200
Reeb type, 285
Reeb vector eld, 187, 188, 199, 208
irregular, 188
non-regular, 188
quasi-regular, 188
regular, 188
reection, 123
reection group, 123
regular, 188
Ricci form, 86, 243
Ricci form, transverse, 243
Ricci tensor, 85, 223, 243, 371
transverse, 243
Ricci tensor, transverse, 224
Riemann curvature tensor, 84
Riemannian
curvature, 26
Riemannian metric, 25
ringed space, 11, 123
Sasaki group, 236
Sasaki metrics
Einstein, 370
extremal, 378
Sasaki-Einstein manifolds, 369
Sasaki-Seifert structure
positive, 245
Sasakian, 206
Sasakian manifold, 206
Sasakian structure, 206
anticanonical, 245
canonical, 245
indenite, 245
negative, 245
null, 245
positive, 245, 248, 319
Seifert bundles, 144
Seifert bred 3-manifold, 144
sheaf, 9, 108
derived functor, 118
locally free, 11
multiplier ideal, 156
structure, 10
sheaf of groupoids, 38
simple normal crossings, 157
singular locus
orbifold, 106, 124, 125
INDEX 611
singularity
Kawamata log-terminal (klt), 158
log-canonical, 158
slice, 42
Smale theorem, 336
Smale-Barden classication, 335
spin manifold, 18
spinor, 18
stabilizer, 39
startication
smooth, 43
Stiefel Whitney classes, 557
Stiefel-Whitney class, 18
stratication
Whitney, 43
strictly pseudoconvex, 199
structure
3-Sasakian, 477
f, 34
almost contact 3-structure, 475
almost contact metric, 195
almost hypercontact, 474
almomst hypercontact, 36
almost complex, 27
almost contact, 33, 202
normal, 203
almost CR, 33, 199
almost Hermitian, 29
almost hypercomplex, 35
almost hyperhermitian, 36
almost hyperkahler, 455
almost product, 32
almost product Riemannian metric, 32
almost quaternionic, 34, 428
almost quaternionic Hermitian, 35, 428
almost Tachibana, 540
complex, 17, 76
complex contact, 184
conformal, 30
conformal symplectic, 31
conjugate Sasakian, 241
contact, 180
contact metric, 195
CR, 33, 199, 206
Haeiger, 32, 52
Hermitian, 29
homogeneous contact, 272
homogeneous K-contact, 273
homogeneous Sasakian, 273
hyper f-, 36
hypercomplex, 35
hyperhermitian, 36
hyperkahler, 36, 455
K-contact, 221
Kahler, 29
locally Sasakian, 265, 312
Lorentzian, 29
nearly Kahler, 539
pseudo-Riemannian, 29
pseudogroup, 36
quaternionic, 429
quaternionic Hermitian, 35, 429
quaternionic Kahler, 35, 429, 431
Riemannian, 17
Sasaki-Seifert, 242
Sasakian, 206, 225, 236
Seifert bred, 47
spin, 18
symplectic, 17, 30
weak G
2
, 536
proper, 537
subgroup
isotropy, 39
submersion, 32, 51
supersymmetry, 545
surface
complex, 93
del Pezzo, 98
Riemann, 93
Taub-NUT metrics, 460
taut, 196
tensor
Killing, 540
Tians -invariant, 155
toral rank, 71, 72, 212
transversal, 54
transverse geometry, 32
transverse holonomy groupoid, 58
transverse homothety, 228
type
(p, q) dierential form, 77
of a Sasakian structure, 245
of Killing spinors, 533
of weak holonomy G
2
, 538
UFD, 123
V-bundle, 110
V-manifold, 105
variety
algebraic, 82
analytic, 82
irreducible, 82
weighted, 139
vector
exponent, 141
vector bundle
compelx, 92
vector eld
foliate, 60
stransverse holomorphic, 69
weighted Sasakian sphere, 213
weighted sphere, 237
Weinstein conjecture, 231
well-formed, 136, 141
612 INDEX
Wolf spaces, 434
Wolf-Gray conjecture, 540