Michel Goze, Yusupdjan Khakimdjanov - Nilpotent Lie Algebras PDF

Nilpotent Lie Algebras
Mathematics and Its Applications
Managing Editor:
M.HAZEWINKEL
Centre for Mathematics and Computer Science, Amsterdam, The Netherlands
Volume 361
Nilpotent Lie Algebras
by
Michel Ooze
Department ofMathematics,
Universite de Haute Alsace,
Mulhouse-Colmar, France
and
Yusupdjan Khakimdjanov
lrutitute of Mathematics,
Uwekistan Academy of Sciences,
Tachkent, Uzbekistan
SPRINGER-SCIENCE+BUSINESS MEDIA, B.V.

A C.I.P. Catalogue record for this book is available from the Library of Congress
ISBN 978-90-481-4671-0 ISBN 978-94-017-2432-6 (eBook)

DOI 10.1007/978-94-017-2432-6
Printed on acid-free paper
All Rights Reserved

© 1996 Springer Science+Business Media Dordrecht
Originally published by Kluwer Academic Publishers in 1996
No part of the material protected by this copyright notice may be reproduced or
utilized in any form or by any means, electronic or mechanical,
including photocopying, recording or by any information storage and
retrieval system, without written pemrission from the copyright owner.
TABLE OF CONTENTS
PREFACE xiii
Chapter 1 : Lie algebras. Generalities 1
SI Lie algebras. Generalities 1

1.1. Nodons of Ue algebras 1
1.2. Examples 2
1.3. Ue subalgebras 3
1.4. classical Ue algebras 3
IS. Ideals 4
I.6. Quodent Ue algebra 5
1.7. Homomorphisms 5
I.8. Direct sum of Ue algebras 7
§II Derivations of Lie algebras 7
ILl. Definldon and examples 7
11.2. Inner derlvadons 8
11.3. Characterlsdc ideals 9
11.4. Derlvadons of a cbrect sum of Ue algebras 9
11.5. Semidirect surD. of Ue algebras 10
SIll Nilpotent and solvable Lie algebras 11
111.1. Derlved sequences, centml sequences 11
111.2. DefiniUon of solvable Ue algebras 12
111.3. Definidon of nilpotent Ue algebras 13
111.4. Engel's Theorem 15
111.5. Ue's Theorem 17
111.6. Cartan Crlterlon for a solvable Ue algebra 20
SIV Semisimple Lie algebra 21
IV.l. Semisimphcity and radical 21
IV.2. KiUingfonn 22
IV.3. Complete reducibility of representadons 23
IVA. Reductive Ue algebras 24
IVS. Levi's theorem 25
§V On the classification of complex semisimple Lie algebras 28
V.l. Regular elemenls. Cartan subalgebras 28
V.2. Root systems 29
vi
V.3. An order relation on the set of the roots fl 30

VA. Weyl Bases 31
V.5. On the c1assificaUon of complex simple Ue algebras 32
§ VI The nilradical 34
VI. 1. Definition 34
V1.2. On the algebraic Ue algebras 35
§ VII The classical invariants of nilpotent Lie algebras 35
VII.1. The dimension of characteristic ideals and the nilindex 36
VII.2. The charac1erlsUc sequence 37
VII.3. The rank of a nilpotent Ue algebra 38
VIlA. Other invariants 38
Chapter 2 : Some classes of nilpotent Lie algebras 40
Filiform Lie algebras 40

1.1. Basic notions. Examples 40
1.2. Graded filifonn Ue algebras 42
§ II Two-step nilpotent Lie algebras 43
11.1. Definition and Examples 43
11.2. On the structure of the tw'o-step nilpotent Ue algebras 44
11.3 On the classification of the two-step nilpotent Ue
algebras 45
§ III Characteristically nilpotent Lie algebras 46
III.l. On Jacobson's theorem 46
I1Iol. Charac1erlzation of characteristically nilpotent Ue
algebras 47
I1I.3. Examples 49
IlIA. Direct sum of characteristically nilpotent Ue algebras 49
§ IV Standard Lie algebras 50
IV.l. Parabolic subalgebras of a semisimple algebra 50
IV.2. Standard subalgebra 51
IV.3. Nilpotent standard algebras 53
IVA. Structure of the nonnalizer of a nilpotent standard
subalgebra 53
IV.5. On the nilpotent algebras of maximal rank 56
IV.6. Complete standard Ue algebras 56
§V On the classification of nilpotent complex Lie algebras 57
V.I. The classificaUon in dimensions less than 6 57
V.2. The classificaUon in dimension 7 60
V.3. other classifications 76
vii
Chapter 3 : Cohomology of Lie algebras 77
§I Basic notions 77
1.1. 9-modules and representations 77
1.2. The space of cochains 78
1.3. The coboundary operator 79
104. The cohomology space 80
1.5. Exact sequence 81
§II Some interpretations of the space W(g, V) for i = 0,1,2,3 84
11.1. The spaces HOC9,V) 84
11.2. The spaces HIC9,V) 84
11.3. The spaces HZC9, V) 86
1104. The spaces H'C9,9) 89
§I11 Cohomology of filtered and graded Lie algebras 91
III. 1. Graded Ue algebras. Filtered Ue algebras 91
11.2. Graded and filtered 9-modules 92
111.3. Graduation and filtration of the cohomology spaces 93
§ IV Spectral sequences 95
IV.l. Definition 95
IV.2. Some properties 95
IV.3. Exact sequence associated to a filtered complex 97
§V The Hochschild-Se"e spectral sequence 99
§ VI. Cohomological calculus and computers 101
VI.1. The MATHEMATICA Program 102
V1.2. How to use this program 110
Chapter 4 : Cohomology of some nilpotent Lie algebras 111
§I Derivations of some filiform algebras 111

1.1. The algebra of derivations of Ln 112
1.2. The algebra of derivations of Wn 114
1.3. The algebra of derivations of Qn (n - 2k+ n 115
104. The algebra of derivations of Rn 117
1.5. Derivations of the standard nilpotent algebra in
9 - slCr+ 1,(C) 118
§ II Cohomology of filiform Lie algebras 121
§ III Cohomology of the nilradicals of parabolic algebras 123
111.1. A decomposition of the groups HiCn,n), 1- 1,2 124
111.2. The calculation of the spaces Htcn,9/n) 130
viii
111.3. PcuUcu1ar case : n is the nllradical of a Borel subalgebra 146

III.4. PcuUcular case : n is an Heisenberg algebra 147
IllS. Description of the space H2fondCn,n) 147
IIl.6. The space H2(n, n) : case of the nllradical of a Borel
subalgebra. 154
III.7. The calculus ofH2Cn,n) when n is the Heisenberg
algebra 157
§ IV Derivations of some infinite dimensional topologically
nilpotent Lie algebras 159
IV.1. InfinitlHiimensional standard algebras 159
IV.2. A topology in a standard ndalgebra 159
IV.3. The weighting derivations of L 161
IVA. Res1rlctlon to the finite-dimensional subalgebras 161
IV.5. The structure of the algebra DerOJ 164
.§V Cohomology of the infinite Lie algebra of vector fields
of the real straightline 167
§ VI On the cohomology of nilpotent Lie algebras in small
dimension 169
Chapter 5 : The algebraic variety of the laws of Lie algebras 170
§I The vector space of tensors 1"n2,1 170

1.1. Definition 170
1.2. Action of GLCn,C). Classificatlon of the tensors 171
1.3. Rigid tensors. Open orbits 172
§ II The variety Ln of the Lie algebraic laws 172
11.1. tie algebraic laws on Cn 172
11.2. The algebraiC variety Ln 173
11.3. The scheme Ln 174
1104. The ac1ion of GLCn,C). Fibratlon by orbits 177
11.5. Open orbits. Laws of rigid tie algebras 179
11.6. The dimension of orbits 180
11.7. The components of Ln 181
§ III Contractions. Deformations 182
III.l.Contractions in LD 182
IIl.2. The deformations 183
IIl3. Equivalent deformations 186
§ IV Perturbations. Notions of infinitesimal algebra 187
IV.l. The infinitesimals in cn 187
ix
IV.2. Theorem of decompositton of a limited. vector of CD 190

IV.3. About the compactness of the Grassmannian manifold 196
IVA. Some problems of perturbations 199
§V The tangent geometry to r,D 202
V.1. Cohomological spaces and tangent spaces 202
V.2. Resolutton of the equatton of deformattons 202
V.3. The equatton of perturbations 206
§ VI The components of LD 213
VI.l. The components of LD for n ~ 7 213
VI.2. Rigid Ue algebras 218
§ VII Construction of solvable rigid Lie algebras 224
VII.l. The decomposability of rigid algebras 225
VII.2. tinear systems of roots associated to a rigid Ue algebra 226
VII.3. Rigid Ue algebras whose nilradical is filiform 229
VIlA. Classification when the torus is of dimension 2 231
VII.5. Classification when the torus is l-dimensional 231
VII.6. On the classification of rigid Ue algebras 232
VII. 7. classification of solvable rigid laws in small dimension 240
§ VIII Study of the variety Ln in the neighborhood of the
nilradical of a parabolic subalgebra of a simple complex
Lie algebra 243
Chapter 6 : Variety of nilpotent Lie algebras 250
§I The tangent space of the variety of nilpotent Lie

algebras 250
§ II On the filiform components of the variety ND 252
§ III On the reducibility of the variety Nn, n ~ 12 256
§ IV Description of a irreducible component ofND+l
containing Rn 259
§V Description of an irreducible component ofND+l
containing W D 263
§ VI Study of the variety ND p in a neighborhood of
a nilradical of a parabolic subalgebra 269
VI.1. On the orbit of n 269
VI.2. On the Zariski tangent space at the point n to the
variety ND 271
x
VI.3. On the irreducible components ofNOp containing 11. 272

VIA. Particular case : 11. is the nilradical of a Borel subalgebra 274
§ VII. On the components of the variety N° 275

VII.1. A nonfilifono component 275
VII.2. An estimation of the number of components 277
Chapter 7 : Characteristically nilpotent Lie algebras 281
§I Characteristically nilpotent filiform Lie algebras 281

§ II Characteristically nilpotent Lie algebras in the variety N° 286
§ III On the nonfiliform characteristically nilpotent Lie
algebras 289
III.1. Construction of some families of nonfilifono
characteristically nilpotent tie algebras 289
III.2. The study of special cases 292
111.3. Characteristically nilpotent tie algebras in the variety
N~ ~
Chapter 8 : Applications to differential geometry.

The nilmanifolds 295
§I A short introduction to the theory of Lie groups 296

1.1. Ue groups 296
1.2. Correspondence between Ue groups and Ue algebras 296
1.3. The left invariant geometry. Interpretation of I:t 298
§ II The nilmanifolds 300
11.1. Definition of nilmanifolds 300
11.2 Vnifono subgroups 300
11.3. Existence of cUscrete unifono subgroups 301
11.4. Rational nilpotent Ue algebras 301
§ III Contact and symplectic geometry on nilpotent Lie
algebras 302
III.1. The Cartan class of an element of 9· 302
III.2. Contact Ue algebras. A characterization of the
Heisenberg algebra 304
111.3. Ue algebras With a symplectic structure 306
§ IV Left invariant metrics on nilpotent Lie groups 313
IV.1. Left inVariant metlics 313
xi
IV.2. Classification of left Invariant me1l1cs on the

3-dimenslonai Heisenberg group 314
IV.3. Some formulas of Riemannian geometry 317
IVA. Some formulas for left Invariant metries 319
IV.5. Left invariant metries on nilpotent Ue groups 320
§V Classification of left invariant metrics on the Heisenberg
group 323
BIBLIOGRAPHY 329
INDEX 335
PREFACE
Nilpotent Ue algebras have played an Important role over the last ye!US : either
In the domain at Algebra when one considers Its role In the classlftcation problems of
Ue algebras, or In the domain of geometry since one knows the place of nilmanlfolds
In the Illustration, the description and representation of specific situations.
The first fondamental results In the study of nilpotent Ue algebras are obvlsouly,
due to Umlauf. In his thesis (leipZig, 1991), he presented the first non trlvlal
classifications. The systematic study of real and complex nilpotent Ue algebras was
Independently begun by D1xmler and Morozov. Complete classifications In dimension
less than or equal to six were given and the problems regarding superior dimensions
brought to light, such as problems related to the existence from seven up, of an
infinity of non Isomorphic complex nilpotent Ue algebras. One can also find these
losts (for complex and real algebras) In the books about differential geometry by
Vranceanu. A more formal approach within the frame of algebraiC geometry was
developed by Michele Vergne. The variety of Ue algebraiC laws Is an affine algebraic
variety and the nilpotent laws constitute a Zarlski's closed subset In this view the role
of Irreduclbl~ components appears naturally as well the determination or estimate of
their numbers.
Theoritical physiCiSts, Interested In the links between diverse mechanics have
developed the Idea of contractions of Ue algebras (Segal, Inonu, Wlgner). That Idea was
In fact very convenient In the determination of components. Others tried to combine
it with the formal notion about deformations which was Introduced by Gerstenhaber.
xiv
Truly the link between these two notions was only described a few years ago. One can
read it in this book. The linear approach to defonnations is situated in the frame of
spaces of cohomology: a deformation is a fonnal series which fonnally saUsfies
jacobi's identities. The linear part of it is thus acocycle for wlues in the adjoint module
(see Chevalley cohomology). Hence the problem of the existence in a given cocycle,
of a defonnation having that cocycle as its linear part. In this frame rigid algebras
appear also naturally. They are the ones which are invariant by defonnations. Nljenhuis
and Richardson have shown that the laws which have a trivial second group of
Chevalley's cohomology are rigid. However they are not all of the same kind :
NiJenhuis and Richardson give an example of a rigid law having a non-trivial
cohomology (in fact, as we wt1l see it here, the class of rigid laws with a non trivial
cohomology Is far from being small).
It Is Important to know about rigid laws : the orbit (t.e the set of all Isomorphic
laws) of a rigid is an open subset in the variety of Ue algebraiC laws and its closure is an
irreducible component of Zarlski. An estimate of the number of isomorphic classes
of rigid laws gives thus an estimate of the number of irreducible components. This
approach has been taken by Michele Vergne in the frame of nilpotent laws. She has
developed a cohomology adapted to nilpotent defonnations, described components
for large dimensions and introduced inCidentally the notion of flUfonn algebras.
Unfortunately we do not know yet of rigid nilpotent laws, and the problem of the
existence of rigid nilpotent Ue algebras In the subvariety of nilpotent laws is also being
considered.
The aim of this book is to put together all the results which were obtained
either In the cohomological study of nilpotent algebras or In the problems of
classification and the problems of deformations. The notion of derivations of
nilpotent Ue algebras is also eludied It elips naturally into the cohomological domain
(derivations are linted to the first group of cohomology). Their interest Is not only
algebraic: Riemanian spaces with a left invariant homogeneous metric often have an
isometric group the algebra of which can be identified with one of orthogonal
derivations. The study of infinitesimal frame : it has the advantage of simplyfylng the
approach and thus eliminate its formal aspect. It also allows to solve the geometrical
problems : what are the conditions for a cocycle (or for afonnal tangent vector) to be
Integrable, mean the first term of a deformation (or so that this formal vector be a
vector of the cone of tangents). One can In the infinitesimal scape describe the duality
contraction - deformation perfecdy. We focus also on two classes of nilpotent Ue
xv
algebras, which are more or less models of nilpotent Ue algebras: those are filiform
algebras (the least commutative ones) and the characteristically nilpotent Ue algebras
(all the derivations are nilpotent).
This book was made possible thanks for the collaboration of Professor Yu.
Khakimdjanov, who worked at the UHA as a visiting professor during the yeatS 1991-
94. We hope that this fruitful exchange will continue ....
We would like to thank the Mathematischer Forschungsinstitut of oberwolfach

(Germany) in allowing us for granting us accomodation and access to their library on
different occasions.
We are also thankful to Professor Otto Kegel for the many suggestions and.
fruitful discussions and to Professor Michael Haz.ewinkel for its interest to our work.
Finally, we would like to thank Mrs Fabienne Garuz for the patience with which
she typed our manuscript
Colmar - Mulhouse
23 october 1994
CHAPTER 1
LIE ALGEBRAS. GENERALITIES
The aim of this chapter is to recall some fundamental notions concerning finite-
dimensional Lie algebras. We also present the principal classes of these algebras and
mainly the more interesting classes of nilpotent Lie algebras. First we study Lie
algebras on an arbitrary field, although the larger part of the book is devoted to
complex nilpotent Lie algebras.
I. LIE ALGEBRAS. GENERALITIES
1.1. Notions of Lie algebras
Definition 1. A Lie algebra. q over a field K is a vector space on K with a bilinear

mapping qxq ~9 denoted (x,y) ~ [x,y] and called the bra.cketof q andsatisfying:
en [x,x] - 0 , V XE q ,
(2) [[x,y],Z] + [[y,Z],x] + [[z,x],y] - 0 , V x, y, Z E 9 .
Remarks
1. The identity (2) is called the Jacobi identity.
2 Chapter 1
2. The relation CI) implies the anticommutativtty of the multiplication of 9 :

[X, y] - - [y, x] ';f x, YE 9.
In fact, we have
0- fx+y ,x+yJ - [x,y] + [y,x].
Conversely, if the characteristic of the field K is different from 2, anticommutativtty
of the bracket implies [x,x] - 0 .
1.2. Examples
1; Every vector space a with the bracket [x,y] - 0, for all x and y in a, is a Ue algebra.,
called an Abe11an Ue algebra
2. Let V be the real vector space Ii'. The bracket defined by the cross-product of the
vectors of Ji3 defines on Ji3 a non-Abelian Ue algebra.
3. Let noCK) be the space of n-square matrices on K. The multiplication
[A,B] - AB - BA
satisfies conditions 1 and 2. Then, nnCK) with this bracket is a Ue algebra, and.
denoted as glCn,K).
4. Let V be a vector space on K. The vector space End V of the endomorphisms of

V with the bracket [f,g] - fog - go f is a Ue algebra, denoted as gICV,K) Cor glCv) if
the field is clearly specified). This Ue algebra is called the hint of general linear
algebra.
5. Let A be an associative algebra on K whose multiplication is 0 • We can provtde A

for the structure of a Ue algebra by putting
[x,y] - x 0 y - yo x , ';f x, yEA.
Note that examples 3 and 4 are of this type.
6. Let V be a C2k+D-dimensional vector space and Cel '''., e2k+I) a basis ofv.
The brackets defined by
Generalities 3
[e21 , e21+11 - e 1 for i - 1,..., k.

Cother brackets except those obtained by anticommutativity are 0) endows V with the
structure of a lie algebra.
This lie algebra plays an important role in the theory of nilpotent lie algebras.
It is called the Heisenberg algebra and noted Hk .
1.3. Lie sub algebras
Definition 2. Let c:I be a Ue algebra. on K. A subspace c:ll of c:I is called a Ue

subalgebra. jf [x, yJ E c:ll whenever x, y E c:ll' We note that aUe subalgebra. of a Ue
algebra. js also a Ue algebra. for the induced multiplication.
Examples
1. Every subspace of an Abelian lie algebra is an Abelian subalgebra.

2. The subspace of glCn,K) constituted of all upper triangular matrices ~r with a,.j -0
if i < J is a subalgebra of glCn,K).
3. The subspace of glCn,K) constituted of all diagonal matrices is also a subalgebra.
1.4. Classical Lie algebras
1. The special linear Ue algebra. slCn,K) , K - ~ or tC .

The set of matrices A = Ca lr of g1Cn,K) with trace zero, that is
n
trA- L~-O'
1-1
is a lie subalgebra of glCn,K).
2. The special orthogonal Ue algebra. so(n,K), K - mor «: .

Let soCn,K) be the space
soCn,K) = {A - (3.jJ) E gl(n,K) : tA - - A}
where tA is the transposed matrix of A.

4 Chapter 1
3 The symplectic Ue algebra sp(2n,K), K - mor C .

Let sp(2n,K) a {A - (3.ij) E gl(2n,K) : tAJ - - JA}
where J- ( 0 Id n ).
-Id n 0
Every matrix A of sp(2n,K) can be decomposed as
A-(~~) ,
where N and P are n-square symmetric matrices and the matrices M and Q satisfy
tM - - Q. Then, sp(2n, K) is a Ue subalgebra of glCn,K) called a symplectic algebra..
Remark. It is easy to enlarge these examples by considerin'g, not only the set of
matrices, but endomorphisms of agiven finite-dimensional vector space. Therefore,
we can define the following Ue algebras
(1) sl(Y) - {f E End V : tref) - O}, where trCf) is the trace of f, that is the sum of the
diagonal entries of a matrix of f (this is independent of the choice of basis of D.
(2)SO(Y)-{fE EndV/<f(v),w>--<v,f(w», VV,WE Vanddetf-l} where<,>

is a nondegenerate symetric bilinear form and det f is the determinant of f.
(3) sp(V) - {f E End V / <Q(f(v),w) - - Q(v/(w)) >, V v, WE V}, where n is a

nondegenerate bilinear antisymmetric form on V with dim V - 2n.
1.5. Ideals
Definition 3. A Ue subalgebra 1- of q is called an ideal of q if [1-,q] c 1- that is
[x,y] E 1- for all x E 1- and yE q.
of course, there are some Ue subalgebras which are not ideals.
Examples
1. Let q be a Ue algebra. Then {OJ is an ideal.
Generalities 5
2. Let 9 be a Ue algebra. The center

Z(9) - {x E 9 : [z, yJ - 0 'V YE .p
is an ideal of 9.
3. Let fa. be a subalgebra of a Ue algebra 9. The normalizer of fa. in 9

NIJCh) - {x E 9 : [x, h] c h}
is a subalgebra of 9 containing h. Then h is an ideal of Nlh).
4. Let 9 - glen, CC) and h - slen, CC). Then fa. is an ideal of 9. In fact, every matrix A can
be written as A - al + A l' where A1 is in slen, CC) and I is the identity matrix. Let B
be in slen, CC). We have
[A, BJ - [aI + AI' B] - [AI' BJ E s1Cn, tC:) •
Remark. If 'L and 1 are ideals of aUe algebra 9, then 'L + 1 - {x + y, X E 'L and y E 1} ,
['L, 1J- {l: [xl> Y2J; xl E 'L, Y2 E 1} and 'L (] 1 are ideals of 9·
1.6. Quotient Lie algebra
Let 9 be a Ue algebra and 'L an ideal of 9. The quotient vector space 9/'L is a Ue
algebra for the multiplication defined by [x + 'L , Y + 'L] - [x, yJ + 'L . This is
unambiguous since if x + 'L = x' + 'L, Y+ 'L = y' + 'L, then [x' + 'L, y' + 'LJ- [x, yJ + 'L.
1.7. Homomorphisms
Let 9 and 91 be two Ue algebras on K. A linear mapping

91 <p: 9 ~
is called a homomorphism if <p[x,yJ- [q(x), cpCy)] for all x and yin 9. We note that the
first bracket is in 9 and the second is in 91 .
As usual, <p is an isomorphism if it is a bijectif homomorphism and <p is an auto-

6 Chapter 1
morphism if cp is an isomorphism and 9 - 91 .
Examples
1. Let V be a n-dimensional K vector space. Then if we choose a basis of V, the

mapping
cp : gleV) ~ glCn, K)
f ~ <pCf),
where <pCf) is the matrix of f corresponding to the given basis, Is an Isomorphism
between these two lie algebras.
For this, we find an isomorphism between sl(v) and slen, K), between so(V) and
so(n, K), and also between sp(V) and sp(2p, K) (n - 2p).
2. Let 1.. be an ideal of the lie algebra 9. The canonical map

1t:9~9!L
x~x+1..
Is a homomorphism of I} onto 9/1... Its Kernel

Ker 1t - {x E 9: 1t(x) - o}
Is equal to 1...
Proposition 1. Let cp : 9 ~ 91 be a homomorphism of Ue algebras. Then Ker cp is an

ideal of 9 and for every ideal 1.. of 9, 1.. c Ker cp, there exists an unique homomorphJsm
'V : 1}11.. ~ 91 such that the following diagram
cp
9 --------------7 91
1t
commutes, where 1t is the canonical homomorphism.
We note that, if 1.. - Ker cp, then 9/Ker cp is isomorphic to 1m cp, the Image of cpo The
Generalities 7
proof is standard and very easy.
1.8. Direct sum of Lie algebras
Let 91 and 92 be two tie algebras over K. We can build a new tie algebra by
9 - 91 + 92 - {(xl' X2): Xl E 91 ,X2 E 92}
with the multiplication
[(xl' X2) , (Yl> Y2)) - ([Xl' X2) ,[Y1, Y2 D .
We can remark that 91 and I:h are ideals of 9. We shall say that 1:1 is the direct sum of
91 and 92 and we shall denote 9 - 91 E9 92 .
It is easy to generalize this construction in order to define the direct sum

9 - 91 E9 92 E9 ... E9 9 p
of the tie algebras 91 , ... , 9 p over K.
Now suppose that 1. and J are ideals of a given tie algebra 9, such that 9 -'L+ J (direct
sum of vector subspaces). Then, 9 Is isomorphic to the direct sum 1. E9 J .
II. DERIVATIONS OF LIE ALGEBRAS
11.1. Definition and examples
Definition 4. Let 9 be a Ue algebra over K. A linear endomorphism f of 9 is ca1Jed

a derivation of 9 if it satisfies
[((x), yl+ [X, fey)] - fix, y)- 0, 'if x, Y E 9.
It is easy to see that the set Der 9 of all derivations of 9 Is a vector subspace of End 9.
Proposition 2. Der 9 is a Ue algebra over K for the bracket [f, g) - fog - g 0 f .

8 Chapter 1
Proof.
(f 0 g - g 0 f) [x, Y1 - fog [x, Y1- g 0 f [x,y1
- f[[gCx),yl + [x,g(y)]J - g[[fCx),yl + [x,fCy)]]
- [f ogCx),y] + fgCx),fCy)] + [fCx),gCy)]+ [x,f ogCy)]
- [go fCx),y] - [fcx),fCy)] - [gCx),fCy)]- [x,g 0 fCy)]
- [Cfog-gof)(x),y]+ [x,Cfog-gof) (y)]
Then, Der q is a subalgebra of gl(q).
Examples
1. If q is an Abelian lie algebra, then every endomorphism of q is a derivation.
2. Let q be a lie algebraandf an endomorphism of q satisfying ([x, yl- 0, 'V x, Y E q,

and f(q) c ZCq). Then f is a derivation of q.
11.2. Inner derivations
Let q be a lie algebra over K and x E q. The endomorphism ofq defined by y ~ [x, yl
is denoted ad x. The Jacobi identity implies that ad x is a derivation of q for all x In q,
called ithe nner derivation. All others derivations of q are called outer.
Proposition 3. The linear map x ~ ad x is a homomorphism of q into the lie

algebra Der q.
In fact ad [x, yl Cy) - [[x,yl, z] - [X, [y,z]] - [y, [x,zll

- Cad x 0 ad y - ad y 0 ad x)(y)
- [ad x, ad y](z)
Proposition 4. The set of inner derivations of q is an Jdeal of Der q.
Proof. Let f be in Der q and x in q. We have, for all yin q,

Generalities 9
[f, ad x] (y) - f [x, yl- ad x (f(y))

- [f(x), y] + [X, fey)] - [X, fey)]
- [f(x), y]
- adCf(x)) (y)
Then [f, ad xl - ad (f(x)) for all x in 9 .
11.3. Characteristic ideals
An ideal" of a lie algebra 9 is called characteristic if fCL) is included in " for all
derivations f of 9. We note that every ideal is invariant for any inner derivation of 9.
But, generally, it is not invariant for any outer derivation. We can see easily that every
characteristic ideal J of a characteristic ideal" of 9 is also a characteristic ideal of 9.
likewise, if "1 and"2 are characteristic ideals of 9, ["1' 'L:z] is also a characteristic ideal
of9·
11.4. Derivations of a direct sum of Lie algebras
We wantto compute the derivations of a finite direct sum 9 - EEl 9t of lie algebras 9i
over K. We can consider every derivation fl of 91 as a derivation of 9 by :
flx) - flCxl' ... , xJ - fl:x~ for all x = CXI' ... , xJ in q.
Lemma. For every i, j such that 1 ~ i , j ~ p, any endomorphism f of q satisfying

f(9k) - 0 if k;t:i, f(q0 c Z(9j) and f[(ql), (91)] = 0 lsaderJvation ofq.
The proof is obvious.

The set of all derivations of q described in this lemma is denoted DCq!, 9j).
Theorem 1. If 9 is a direct sum q = EE>~-1 ql of Ue algebras over K, then

10 Chapter 1
Proof. From the lemma, we have
Der 9 => Ee~-1 Der 91 E9 (Eel.. P(91 ,-9j»)
Now we prove the converse inclusion:
Let fbe in DeqJ. Then f - ~f-l fl + ~f"l fij where fl EEnd 91 ,flJ EHom (91 , 9V are de-
fined by fl - 1t1 0 f 0 1t1 ' flj - 1tj 0 f 0 1t1 where 1t1 is the canonical projection of 9 on
91' These endomorphisms fb flJ are derivations of 9. We shall prove that flj E DC91,9)).
Let x - [y, zl in [91' 911 . Then
*
flj(x) - flj [y, z] - [flj (y), z] + [y, flj (z)] - 0 if i j and flj [91 ,91] - 0 .
Suppose now that flj(91) is not contained in 2(9j) . There is x '# 0 in 91 such that y-f lj (x)
yeo 2(9j) .WechooseZin9j Sadsfying[z,yJ '#O.Then
[flj' adz] (x) - [z, y] '# 0
and the nontrivial derivation [flj,ad z] is in D(91 , 9j). However, this derivation is inner
(Proposition 4). This Is Impossible If I '# j.
11.5. Semi direct sum of Lie algebras
Let 91 and 92 be two Ue algebras over K. A IJe algebra 9 over K Is called a semidirect
sum of 91 by 92 If 9 - 91 + 92 (direct sum of vector spaces) where 91 is an ideal
and 92 a subalgebra of 9·
Notation. We shall also denote the semldirect sum by

9 - 91 E9 92
and In this decomposition first the Ideal Is always layed.
So, we have
The direct sum corresponds to the particular case where 92 Is also an Ideal. We use
the same notation which we will summarize If it is necessary.
Generalities 11
Then the restriction of ad x to 91 Is a derivation of 91' This defines a homomorphism

9: 92~ Der91
x ~adx.
Conversely, the datum of the tie algebras 91 ,92 and of a homomorphism

9 :92 ~ Der91
of tie algebras permits the defining of a tie algebra 9 - 91 + 92 (direct sum of vector
space) with the bracket:
[(Xl. X2) , (Ylo Y2)] - {[Xl> Y1] + 9(X2) (Y1) - 9(Y2) (xd , [X2, Y2]) .
In this manner, 9 appears as the semi direct sum 9 - 91 ffi 92' The trivial case 9 - 0
corresponds to the direct sum of 91 and 92 .
III. NILPOTENT AND SOLVABLE LIE ALGEBRAS
111.1. Derived sequences, central sequences
Let 9 be a tie algebra over K (K - m or tC). We put D09 - 9; D ~ - [9,9] and,

more generally, Dk+~ - [D"9, D~], for every k ~ O. All these subspaces are ideals
of 9 and we have the following decreasing sequence, called the derived sequence
of 9 :
9 - D09 ::> D~ ::> ••• ::> D~ ::> •••
The descending central sequence of 9 :
9 - C09 ::> C~::> ... ::> ~ ::> •••
is defined with the follOwing ideals:
V'k~o
C09 - 9 ,
{
Ck+19 - [Ck9 ,9] . }
Note that this sequence decreases more slowly that the derived sequence and the tie
algebra ct+19/ct+~ Is an ideal of ct9/C i+~ contained in its center.
The ascending central sequence is given by

12 Chapter 1
Co(9) - 0,
Ck(9) - {X E 9 : [x, 9) c C k-l (9)} V k ~ 1.
We have
{OJ - Co(9) c C1(9) c ... c Ck (9) c ...
We note that C{9) is the center of 9 and Ck+1(9) is the reciprocal image for the
canonical embedding of 9 on 9ICk(9) of the center of 9ICk(~ .
111.2. Definition of solvable Lie algebras
Definition 5. A Liealgebra 9 isca11edsolvableifthere is an integer less k such that

n~ - {O}.
Examples.
1. Every Abelian Ue algebra is solvable.
2. The Ue algebra of the affine group of a straight line is solvable because it is

generated by two independent vectors x and y satisfying [x, y ] - x.
3. The subalgebra of gICn,tC) composed of triangular matrices is solvable.
4. Let V be a vector space of dimension n and D - (0 - V 0 c V 1 C ... C V n - V)

a flag of V that is a sequence of fitted-together vector subspaces. Let
beD) - {f E End V : f(v i) C Vi Vi}. Then b (D)is a solvable Ue algebra for the
bracket
[f,g)-fog-gof.
The matrices of the elements of beD) relative to an adapted basis of the flag are
triangular. This algebra is similar to that described in example 3.
Proposition 5. The Lie algebra 9 is solvable if and only if there is a descending

sequence ofideals of 9
Generalities 13
such that [Ij ,IU c 11+1 for O:S: J ~ k-l .
Proof. For the necessary condition, we take Ij - nJc:J.

Sufficient condition: we have 1) IIJ C 11 and, more generally, 1) jlJ c Ij .
As Ik - {O}, then nke;, is equal to O.
Proposition 6. Let IJ be a Ue algebra.

(a) If IJ is solvable, every subalgebra and every quotient algebra are solvable.
(b) If 1 is a solvable ideal of 9 such that 1Jft. is solvable, then IJ is also solvable.
(c) If 1 and J are solvable ideals of IJ, then 1 + J is also solvable.
Proof.
(a) If L is a sub algebra of IJ, then nil is contained in n'IJ.
(b) Let 1t: Ij ~ Ij / 1 be the canonical homomorphism. As it exists, an integer k such
as1)K(IJ/t) -0, then 1tlDKIJ)-0. Thus DKlJct andas t issolvable, DKIj isalso
solvable and therefore IJ is solvable.
(c) We can deduce this result from the property (b) usingthe fact that the ideal
1+J / J is isomorphic to 1 / 1nJ .
Property (c) implies that there exists a unique maximal solvable ideal of IJ. This ideal is
called the radical of IJ.
111.3. Definition of nilpotent Lie algebras
Definition 6. A Ue algebra 9 is called nilpotent if there is an integer k such that

cklJ - {O}. The smallestlnteger k such that eke;, - {O} is called the nillndex(or the
nilpotency class) of IJ.
Examples.
1. Every Abelian algebra is nilpotent With a nilindex equal to 1.
14 Chapter 1
2. The Heisenberg algebra Hk defined in the basis (el' e 2 , ••• , e2k+ 1) by

[e 21- 1 ' e21 ] - e2k+1 i - 1 ,..., k
(the undefined brackets being nul) has a nilindex equal to 2.
3. The n-dimensional algebra defined in a basis (e 1 , •••, en) by the brackets
[e 1 ' ei J - e'+1 2 $; i $; n-1
is nilpotent with a nilindex equal to n - 1.
4. The subalgebra of gICn,K) made up by the upper triangular nilpotent matrices (that
is to say, the matrices A - (~~ such that ~J - 0 for i $; j) is nilpotent.
5. Let V be a vectorial space and D - tvo c V1 c ... C Vn - V) is a flag of length n.
The Ue algebra
nCD)- {fE End(V) : f(Vi)CVi_l}
is nilpotent. We note that the endomorphisms belonging to neD) are written in an

adapted basis to the flag D in a triangular form.
Proposition 7. Let 9 be a Ue algebra.

(a) If 9 is nilpotent, then every subalgebra, every quotient algebra is nilpotent.
(b) Let Z(9) be the center of 9. Then if 9!Z(~ is nilpotent, 9 is also nilpotent.
(c) If 9 is nilpotent, its center Z(9) is notnr1vial.
(d) If 1. and J are nilpotent ideals of 9, then 1. n J and 1. + J also are nilpotent.
Proof.
(a) This is the same as the proof in the previous section.
(b) Suppose that 9/2(9) is nilpotent. There is an integer k such that 0'<:9!Z(9)) - 0 so
cI<cJ is contained in Z(9) and dt+ 19 - {OJ.
(c) If g is nilpotent with a nilindex equal to k, then ck+19 '# {OJ and Z(9) => dt-~.
(d) Suppose that the nilindex of I is k and this one of J is 1. From the Jacobi identities,
CPCI+J) is included in C£P121+ 1Cj) where [p/2J denotes the integer part of p/2. As k-l is
less than dim 9 for p - 2dim 9, then CP+l(l+J) - {OJ.
Proposition 8. Every nilpotent Ue algebra is solvable.

Generalities 15
I I
Infact, C 9 cD 9 .
Proposition 9. A lie algebra 9 is nilpotent if and only if there is a descending

sequence ofideals
9 - 10 ::) 11 ::) ... ::) Ik - {OJ
suchas [9, Ii] c li+l 0 ~j ~k-l .
The proof Is similar to the one shown in the solvable case.
Proposition 10. A lie algebra 9 Is nilpotent with a nillndex k if and only if for every
"0, Xl' ... , xk E 9, we have
[XO,[Xl,[X2, ... [Xk-l,Xk]]] ...... ]-0.
This follows directly from the definition.
111.4. Engel's Theorem
Let 9 be a Lie algebra and let X E 9. Recall that ad X is the endomorphism of 9

defined by ad x (y) ~ [x , y ].
Theorem 2. A lie algebra 9 Is nilpotent if and only if ad x is nilpotent for every x

in 9.
The proof is based on the follOwing lemma (Engel's lemma) :
Lemma 1. Let V be a nontrivial vector space on K and 9 a subalgebra of glCY).

Suppose that every x in 9 is a nilpotent endomorphism of V. Then there is
v"# 0, V E V, such that xCv) - 0, V X E 9.
Proof of the lemma. We use an induction on the dimension n of 9. The lemma is

true for n - O. Suppose that it is true for a dimension less than n. We begin by
16 Chapter 1
showing that 9 contains a codimension 1 ideal. Let fI. be a subalgebra of dimension

m of 9. Let x E fl.; it is a nilpotent endomorphism of V. This implies that the
endomorphism ad x of fI. is also nilpotent Indeed, the terms of the decomposition
of (ad x)P(y) have the form x'yxk with i+k - p. As xP - 0 for some power p, ad x
is also nilpotent. Now let oex) : 0/fI. ~ 91fI. the endomorphism defined from ad x
by projection on the quotient. As ad x is nilpotent, o(x) is also nilpotent. Then,
by hypothesis, there is ~ YE 9 / fI. such that a(x)(Y) = 0 for all x in fl.. Let Yin 9 be a
representation of Y. Then [x, yJ is in fI. for all x E fI. andfl. is an ideal of a m+l-
dimensional subalgebra of 9 . If we iterate this process, we find a codimension 1 ideal
of 9. Let fI. be this ideal. We choose a E 9 - fl.. The set U - {v E V : xCv) - 0 'it x
E fl.} is not nul. As xCC(v)) - [x, a]Cv) - 0 , 'it v E U. The space U is a-invariant By
the induction hypothesis, there Is u E U , U'l; 0 such as c(u) - o. As xCu) - 0 'it x E U
and 9 - fI. EB (C a, the lemma Is proved.
Proof of the theorem. The necessary condition follows from Pproposition 6. For
proving the sufficient condition, we use an induction. Suppose that the theorem is
true for the tie algebras of dimensions less than n. Let 9 be an n-dimensional tie
algebra such that ad x is nilpotent for all x. From Engel's lemma, it exists that y E 9
, y*"O such that [x,y] - 0 'itx E 9. This prove that the center Z(9) of 9 is nontrivial.
Let 91 the quotient algebra 9 1 Z(9) , by hypothesis 91 Is nilpotent and from
Proposition 7, 9 Is nilpotent.
Corollary 1. Let 9 be aUe algebra, and 1. an idea/of 9. If the quotientUe algebra

9 / 1. is nilpotent and if f or all x in 9, the restriction of ad x to 1. is nilpotent,
then the Ue algebra 9 is also nilpotent
The proof Is contained in the proof of Engel's theorem.
Corollary 2. Let 9 be a Ue subalgebra of giCn,K) whose elements are nilpotent. Then

9 is nilpotent.
Generalities 17
111.5. Lie's theorem
The field K is algebraic closed and of characteristic O.
Let I} be a Ue algebra. on K and V a vector space on K. A linear representation of

I} on V is a homomorphism of the algebras
4>:I}~glCV),
i.e. is a linear map satisfying

4l(x , y]9 - [ctl(x) , ctl(y)]91M =4>(x) 0 (y) - (y) 0 (x)
with x and y in I} ([ , ]9 is the bracket in I}). Then the vector space V is a

I}-module by putting
x.v- ctl(xXv) x E I} and VE V .
Naturally, we have an equivalence between the notions of the linear representations of

I} and the I}-modules.
Theorem 3 ( Lie's theorem). Let I} be a solvable lie algebra on K and a linear
representation of I} In a vector space V. Then there exists a flag D - (V - V0::> VI
::> ... ::> V n - {OJ ) of V such that b(D) ::><l>Cg) where beD) is the lie algebra of
endomorphlsms of V related with the flag D.
This algebra. &CD) is described in example 4 (1.2). We can also enunciate the theorem
as :
Theorem 4. With the hypothesis of lie's theorem, if V =I- {O}, there is a vector v =I- 0
in V such that
«XXv) - 'i.I.x)v , 'v'XEI}.
where '). is a linear form on I}.
The eqUivalence between these theorems can be proved using an induction on the
dimension of V.
18 Chapter 1
Proof of Theorem 3.
First, we prove the following lemma:
Lemma. Let 9 be a Ue algebra on K and t an Ideal of 9. Let V be a 9-

module and v E V, V¢ 0 such that, for every a E t, one has a. v - 'J...W. v where A.
Is a given linear fonn on '1. Then 1..( [x~]) - 0 "t x E 9 and a E 'L.
Proof of lemma.One chooses a vector x E 9 . Let VI be the subspace of V

generated by the vector x.v, x.(x.v) - x2·v, ... , x(x ...(x.v)) - xi-l·v. These subspaces
define the ascending sequence
{O} - Vo C VI C ... C Vi-l C Vi.
Let P be the smallest integer satisfying Vp - Vp+l . The sequence (VI) Is stationary
from Vp' One has dim Vp - p .We shall show that axl.v == A.(a)xl (mod VI) for any I ~ 0
and a In '1. (Recall that xl.v - x.x.. .x. v and ,f!>. v - v.) This relation Is true for I - O.
Suppose that the relation Is true for every 10 ,to ~ I - 1 ~ O. Then
axl.v - a.x.x1-1.v - x.a.xI-1.v - [x~lxl-l.v - x.A.(a)xI-l.v + x.vl_1 - A.[X~lxl-l.v + w l_1
with vl_1 and w l_1 In VI_I'
As X.VI_l C VI and t,VI C VI, we have X.VI_I E VI and A{x~] XI-I.V E VI.
Then
a Xl.v == x.A.(a) XI-I.V (mod VI)
== A.(a) XI.V (mod VI) .
ThIs proves that axl.v == A.(a)x1.V (mod VI)' So, the matrix of «Il(a) related with a basis of
V adapted to the flag of V, where «Il(a)(v) - a.v, Is triangular. The elements of the
diagonal are equal to ACa). Then Tr (<<Il(a)) - nA.(a), and as [a,x] E I, "t x E 9, we have
Tr [a,x) - nA.[a,x). But Tr [a,n) - Tr(<<Il(a)o(x) - «Il(x) 0 (a») - 0 , and the linear
form A. satisfies A.[a,x) - o. This proves the lemma.
Returning to the proof of Theorem 3, one uses an induction on the dimension of 9.

If dim 9 - 0, the theorem is true. Suppose dim 9 ¢ O. As 9 Is solvable, we have
D9- DI9-[9,9]C9 and D9 ¢9.

Let '1 be a codimenslon 1 subspace of 9 and t containing D9. It Is an Ideal of 9
Generalities 19
because 1. => n9 => [t , 9] . From the hypothesis, there is a nontrivial vector v in V and
a linear fonn A. on 1. such that
fJ)(a)(v) - A.(a)v , V' ae 1..
We have W - { we V such that fl»(a) (v) - A.(a) (av) V' a e 1. }. It is a nontrivial
subspace of V. As before in the previous lemma, we have
A{x,a] - 0 V'xe 9 and ae 1. .
But
fJ)(a) 0 fJ)(x) (w) - tll(a,x] (w) + q,(x) 0 <I(a) (w)
- A{a,x] w + q,(x) (A(a)w)
- A.(a) fJ)(x) (w) ,
then fl»(xXw) e W and W is invarlantin respect to the action of 9. Let x e 9, x Ii!: 1.. As
fJ)(x)(W) is contained in W, there is an eigenvector vo ;/: 0, Vo Ii!: W. This vector is also
an eigenvector for every linear operator fl»(y), y e Kx + I - 9. This proves the lemma.
Consequences
1. We can deduce from Ue's theorem the existence of a basis such that the matrices of
the endomorphisms with respect to this basis are triangular.
2. We consider the adjoint representation of 9. Then as before in the previous case,
there is a basis of 9 such that every adjoint operator ad x admits a matricial
triangular representation with respect to this basis.
Corollary 3. The Ue algebra 9 is solvable if and only if the derived algebra n(9) is
nilpotent
Proof. We suppose that 9 is solvable. From Ue's theorem applied to the adjoint
representation, we deduce that there exists a sequence
9=>1.1=>~=>"'=>'Ln-(0} .
where 1., are ideals of 9. If x e n~, then ad xC1.,) is contained in 1.'+1' The endo-
morphism ad x is nilpotent. Its restriction to n 19 is also nilpotent. From Engel's
theorem, :1)19 is nilpotent
20 Chapter 1
111.6. Carlan Criterion for a solvable Lie algebra
Theorem 5. Let 9 be a subaJgebra of 9ICv). where V is a flnite-d.imenslonal space

over K. Then 9 is solvable if and only if
TrCxy) - 0 'v' x E 9 and 'v' y E D 9 .
Proof. Suppose that TrCxy) - 0 for all y In 9 and x In D9. For proving the solvability of
9. It is sufficient to prove that D9 is nilpotent. From Engel's lemma. we can prove that
x Is nilpotent In D9. This Is a consequence of the following lemma:
Lemma. Let 9 be a subaJgebra of glM. and M - {x E glM: [x. 9] c [9. 9]} . If x E M

satisfies TrCxy) - 0 for all yin M. then x is nilpotent
Proof of the lemma. As x Is an endormorphism of V. It can be written as x - Xs + len.

where Xs Is the semisimple part and len is the nilpotent part with Des. len] - o. Suppose
An be the corresponding
that (VI' ...• v.) is a basis of eigenvectors of Xs and let AI' ...•
eigenvalues. Let E be the Q-vector space spanned by AI' ... , An. We have to prove that
any linear function f: E ~ Q Is zero. Let f be In E-. We consldery E glCv) such that
y Cv~ - f CA~ VI and we note by (<pql the basis of glCv) defined from 'Pt(..v~- vJ •
'Pt(..vk) - 0, k :1= j . Then ad Xs Cy) - (<Jlt - cpj) CPtj and ad y CCJliY - (f(W - fCcpy) CPtj • From
the Lagrange interpolation. we can find a polynomial POO in K[X1 without a constant
term such that PCCl>t - ~ - fCW - fCcpy for all i, j. and ad y - P(ad Xs). As ad Xs is the
semisimple part of ad x. from the Jordan decomposition of endomorphisms. we can
write ad Xs as a polynomial In ad x without a constant term. Therefore. ad y is also a
polynomial In ad x without a constantterm. From the hypothesis, ad xC~ Is in
[9. 91 and this implies that ad Y(9) Is in [9, 91. thus Y is In M and TrCx.y) - 0 . So we get
1: 'A.tfC~ - 0 and f (1: A.f.C~) - 1: fC~ - 0 . The number fC~ is rational. this gives
fC'A.t)-Oandf-O.
We can return to the proof of the Cartan criterion.

As 9 eM. the lemma implies that every x In [9,9] with TrCx.y ) - 0 "if Y E M Is nllpo-
Generalities 21
tent. But, Tr([u, v] z) - Tr(u[v, z]) - Tr([v, z]u) for every u, v, z in gl(V) . In particular,
if z eM, u, v e q, we have [v, z] e [q,q] and Tr([v, z] u) - o. This gives Tr([u, v] z)
- 0 and Tr(x y) - 0 'r:j y eM. From the lemma, x is nilpotent this ends the proof
after we have seen that the converse is a direct consequence of Ue's theorem.
Corollary. Let q be a Lie algebra satisfying TrCad x ad y) - 0 for aD x e [q, qJ and

ye q. Then q is solvable.
IV. SEMISIMPLE LIE ALGEBRAS
The principal topics of this book concern the study of nilpotent tie algebras. So we
shall give only a short survey of the theory of semisimple tie algebras Cthe reader
interested in this subject can read the classical books, for example [SelJ). The aim of
this survey is to present the necessary tools in order to study some important classes
of nilpotent subalgebras of semisimple complex Ue algebras.
Throughout this section and in those following (except where otherwise

specified), the field K is the complex field (C.
IV.l. Semisimplicity and radical
Definition 7. A non-Abelian Lie algebra q is called simple if its only ideals are q
and {a}.
Examples
1. ste2, tC) is simple.
2. We will see later that every classical tie algebraslCn, tC), soCn, «:;), spC2n, «:;) Csee 104)
are simple.
22 Chapter 1
Recall that the radical of a Ue algebra9 is the unique maximal solvable ideal of 9. It Is
denoted by rad 9.
Definition 8. The Ue algebra 9 Is called semIsJmple If rad 9 - (OJ.
Proposition 11. Let 9 be a Ue algebra. Then .yrad 9 Is semIslmple.
Corollary. If 9 Is a semisimple Ue algebra, then It doesn't have nonzero AbelIan

Ideals.
IV.2. Killing form
Let 9 be a Ue algebra. A bilinear form B : 9 x 9 -+ IC is said 10 be invariant if we have :

B[x,y],z)+B{x,[y,Z])-O forall X,y,ZE 9.
For example, the bilinear form BCx,y) - TrCad x 0 ad y) is invariant and symmetric.
This form is called the KIlllng form.
Proposition 12. Let 'L be an Ideal of the Ue algebra 9. Then the orthogonal space 'L.L
of 'L with respect to he KIllIng form B Is an Ideal of 9.
This results direcdy from the invariance of B.
Theorem 6 (Cartan Killing Criterion). A Ue algebra 9 Is semisimple If and onlyIfIts

KJlllng form B Is nondegenerate.
Proof. Let 'L be the Kernel of B. It is an ideal of 9. Suppose that 9 is semisimple. If

x E 'L, then BCx, y) - TrCad x 0 ad y) - 0 for all y in 9, hence, in particular, if y is In
D 9 - [9, 91. By Cartan's criterion, the subalgebra ad 'L of gl(9) is solvable. Since
ad'L - 'L/Z(9), then 'L is also solvable. As 9 is semisimple, 'L - {OJ.
Conversely. Suppose B is nondegenerate; we take an Abelian ideal Q. of 9. We
Generalities 23
consider f - ad x 0 ad y for x E a. and y E 9. Then a.::::> f(9) and f(a.) - {O}, this gives
f2 - 0 and TrCD - o. This proves that a. c 'L = Ker B = {O}.
Corollary 1. Let 9 be a semisimple Ue algebra and 'L an ideal of 9. Then 9 - 'L $ 'Ll.
(direct sum of Ue algebras) where V is the orthogonal of 'L with respect to the
Killing form of 9.
Corollary 2. A semisimple Ue algebra is isomorphic to a direct sum of simple Ue

algebras.
Corollary 3. If9 is semisimple, then 9 - D9 - [9,91.
Proposition 13. Every derivation of a semisimple lie algebra is inner.
Proof. Let B be the Killing fonn of a semisimple tie algebra 9. It is nondegenerated.

Let f be a derivation of 9 and q, the linear fonn on 9 defined by <Il(x) - trCad leo D.
There is a vector u in 9 such that Beu, x) - <Il(x) for all x in 9. We put g - f - ad u. It is a
derivation of g and we have
trCad x 0 g) - trCad x 0 D - trCad x 0 ad u)

- <Il(x) - BCx, u) - 0 for all x in 9
Then BCgCx), y) - trCad f!f.x) 0 ad y)
- trC[g, ad xl 0 ad y)
- trCgoadxoady-adxogoady)
- trCgo ad xo ad y- go ad yo ad x)
- trCg 0 [ad x, ad y])
- trCg 0 ad[x, y]) - 0,
Hence, g(x) - 0 for all x in 9 and f - ad u. This proves the proposition.
IV.3. Complete reducibility of representations
First, we recall the notions of representations of tie algebra 9 and the corresponding
24 Chapter 1
notion of 9 -modules.
Let 9 be a Ue algebra. A representation of 9 In a complex vector space V Is a
homomorphism of Ue algebras cI> : 9 ~ gICV). A vector space V endowed with an
operation 9 x V ~ V (denoted x . v) is called a 9-module if we have
Q) (ax + by)v - a(xv)+ b(yv) , a, be <C, x, y E 9, v E V ,
Cit) x(av + bw) - a(xv) + bCxw) , a, b E <C, x, Y E 9, v E V ,
aID [x, yJv - xyv - yxv, x, Y E 9, v E V .
The notions of the representations of 9 in V andg-modules are eqUivalent: if cI> is a

representation of 9 in V, then V is a 9-module for the operator x.v - cl>(x) (v) ;
conversely, If V is a 9-module, then the map cI>: 9 ~ glM given by cI>Cx) (v) - xv is a
representation of 9 in V.
Definition 9. A 9-module V is called simple if V ":t: {OJ and V has no submodules

other than {OJ and V which is called semisimple if it is the direct sum of simple
submodules.
In the. language of representations, these notions correspond to irreducible and

completely reducible representations.
Remark. 9 may be considered as a9-module for the operation xy - [x, yl. Note that 9
may be asemlslmple 9-module without being a semisimple Ue algebra; for example
in the case 9 - <C.
Theorem 7 (H. Weyl). Let9 be a semisimple Uea1gebra, and V afinlte-c:Umensional

9-module. Then V is semisimple.
Proof. See, for example, [Se 11 or [Hu].
IV.4. Reductive Lie algebras
We can define the nllradicalof the Ue algebra 9 as the largest nilpotent Ideal of 9. We
Generalities 25
can also define the notion of the nilpotent radical of 9 as the intersection of the
kernels of the finite dimension irreducible representations of 9. This ideal is
contained in the nilradical and we can prove lhat it coincides with:09 n rad 9.
Definition 10. A Ue algebra is called reductive if its adjoint representation is

completely reducible.
Proposition 14. Let9 be a Ue algebra. The follOwing conditions are equivalent:

(a) 9 is reductive,
(b) :09 is semJslmple,
(c) 9 - :09 E9 2(9),
(d) rad 9 - 2(9),
(e) the nilpotent radical of 9 is (OJ.
IV.S Levi 's Theorem
Theorem 8 (Levi). Every Ue algebra 9 is the semi-direct sum ofits radical rad 9 and a
semisimple subalgebra s :
9 - rad 9 E9 s.
The proof of this theorem follows Bourbaki [Bou 1].
Proof.
1st case. : [9, rad 9]- {OJ • Then rad 9 is the center of 9 and the adjoint representation
of 9 is semlsimple (we can Identify this representation with the representation of
91Z(9) = 9/rad 9 which Is semisimple). Then:09 is also semisimple and 9 - rad 9 E9
:09·
2fJd case: [9, rad 9] '* (OJ and the only ideals of 9 contained In rad 9 are {OJ and rad 9.
Then [9, rad 9] - rad 9 • [rad 9. rad 9] - (OJ and the center of 9 is (OJ.
'*
Let M be defined by M = (f E End 9: f(9) c rad 9; f(x) - A.(f)x A.(f) 0 V X E rad 9}
26 Chapter 1
and N - (f E End 9: f(9) c rad 9 andf(rad9) - o} .

Therefore, N is of codimension lin M. Let 0 be the representation of 9 in End 9 given
by
o(x) f - [ad x, f] 'V x E 9, 'V f E End 9 .
We have o(x) M c N , for all x in 9 . Moreover, if x E rad 9, YE 9 and f E M then

(o(x) f) (y) - - A(f) [x, y] (because rad 9is Abelian)
- - adCA,Cf) x) Cy) .
As the center of 9 is (OJ, the mapping x ~ ad x defines a bijection <p of rad 9 onto a
subspace P of End 9. stable under aCq). As rad 9 is Abelian, Pc N and oCx)Mc P for
all x in rad 9. The representation of 9 on V - M/p derived from 0 is zero on rad 9. It
defines a representation 0 1 of the semisimple algebra 9/rad 9 on V. For all
y E 9/rad 9 cj (y) (V) c N /P. As dim V - dim N /p - 1 , there exists fo E M with
A(fo) - -1 and o(x) fo E P for all x in 9. The mapping x ~ "(Pl(cr(X) fo) is linear
o(x) fo E P for all x in 9. The mapping x ~ "(Pl(O(X) fo) is a linear mapping of 9 into
mapping of 9 into rad 9 . Then its restriction on rad 9 is the identity mapping.
Hence the Kernel 5 of this mapping satisfies 9 - rad 9 + 5 Cvectorial direct sum). As 5
- {x E 9: oCx) fo - O}, 5 is a suba1gebra of 9.
General case. We process by induction on the dimension n of rad 9.

If n - 0, the theorem is clear. Suppose the theorem is true for all Ue algebras whose
radicals have a dimension less than dim rad 9. Followingthe first case, It suffices to
consider the case where [9, rad 9] '# (OJ. But [9, rad 9] corresponds to the nilradJca10f
9, it is nilpotent, andits center c is nontrivial. We choose a minimal nonzero ideal
m of 9 contained in c. The case m - rad 9 corresponds to the 2m case. We suppose
m '# rad 9 and consider 7t : 9 ~ 9/m - 9 the canonical mapping. Then
rad 91 - rad 9/m . By the induction hypothesis, 91 - rad 91 E9 51 where 51 is
semisimple.
Let Fa. be Fa. - i- l (51) . It is a subalgebra of 9 containing m, such that him- 51 is semi-
simple. Its radical Is m and by the induction hypothesis, Fa. - m E9 5, where 5 is
semlslmple. This implies 9 - rad 9 E9 Fa. - rad 9 + m + 5 - rad 9 E9 5. This gives the
Generalities 27
Levi decomposiUon.
We call the semisimple Ue subalgebragiven in the Levi's theorem a Levi suba1gebra.

of 9.
Remarks. 1. If 51 and 52 are two semisimple subalgebras of 9 such that
then there exists an automorphism (J of 9 such that 0(5 1) - 52' Moreover, we can
choose (J of the form (J - expCad x), where x in the nilradical of 9. This result is due to
Malcev.
2. From Levi's theorem, the general description of Ue algebras is deduced. from the
description of semisimple Ue algebras, solvable Ue algebras, and the representation
of semisimple Ue algebras as derivations of the radical. The class of semisimple Ue
algebras is well known (see later). But the class of solvable and nilpotent Ue algebras is
very difficult to describe. Moreover, solvable Ue algebras exist that are never radical
of an unsolvable Ue algebra wuh a nontrivial representation of the Levi subalgebra.
Example. Let 9 be the 7-d.imension nilpotent Ue algebra defined by
[el' el] - el+1 , i - 2 ,...,6,

[e2 ,e,] - e7 ,
[e, ,e4] - - e7 ,
[e2 , e4] - e6 ,
[e2 ,e,] - e6 + e7 .
We shall see that every derivation of 9 is nilpotent Then Der 9 does not contain a
semisimple subalgebra.
28 Chapter 1
V. ON THE CLASSIFICATION OF COMPLEX SEMISIMPLE LIE

ALGEBRAS
By classification, in this book we mean the classification up to an isomorphism of Ue

algebras. In this section we resume the classical work of Killing and carran about the
classification of complex semisimple Ue algebras. Recall that the ground field is IC and
that the Ue algebras considered are finite-dimensional.
V.l. Regular elements. Cartan sub algebras
Let 9 be asemisimple Ue algebra and x E 9, x;t o. We let 9\ denote the nilspace of

ad x - A.Id j that is
9\ - {y E 9 j Cad x - A. Id)p Cy) - 0 for some p} ,
where A. E IC.
Definition 11. The element x E 9 Is ca1led regular If the dimension of the space 9 0 x
18mlnlmal:
dim 90x - min (dim 90y).
ye ':J
Proposition 15. If x Is regular, then 9° x Is Abelian and Is equal to Its nonnallzer In 9.
Recall that the normalizer of 9° x in 9 is the set of the elements y of 9 such that
[y , cfx] c cfx .
Definition 12. A Ue subalgebra h. of 9 is called a CaFtan subaJgebra of 9 if there 18

a regular element xln 9 such that h.- 9 0 x.
We can note that there always exists a nontrivial Carlan subalgebras for every
semisimple Ue algebra. Moreover, all the Carlan subalgebras of a semisimple Ue
Generalities 29
algebra 9 are isomorphic. They are also conjugated in the following sense: let hI and
h2 be two Cartan subalgebras ; there is an automorphism cr of 9 of the form exp(ad y)
with Y E 9 such that ~ - o(h l )·
Consequences
1. All the Cartan subalgebras of 9 have the same dimension r. This dimension is calied
thenmkof 9.
2. A Cartan subalgebra h of 9 is a maximal Abelian subalgebra of 9. Every element of

h is semisimple, i.e. the endomorphisms ad x for all x in h are semisimple. As these
endomorphisms commute, it can be diagonalize together.
3. A maximal Abelian subalgebraof 9 is not usually a Cartan subalgebraof 9. However,

a maximal Abelian subalgebra of 9 whose elements are semisimple, is a Cartan
subalgebra.
V.2. Root systems
Throughout this section,9 denotes a complex semisimple Ue algebra and h a fixed

Cartan subalgebra of 9. A root a of 9 (relative to h) is a nonzero element a E h· (the
dual of h) such that the space rf'" defined by
9!X - {x E 9 : [X, hJ = aCb) x V h E h}
is not (0).
If a is a root, we call 9 !X a elgensubspace of 9, and an element of 9!X is said to have a

weight a.
We have:
1. 90 - h .
2. [9(1,9 13] c 9(1+13 if a + ~ is a root,
[9(1,9 13] - {O} if not

3. Let A be the set of the roots of 9. Then
I} - fI. E9 ~ I}IX CVectorial direct sum).

lXed
4. Dim I}IX - 1 forall a In A.
5. If B Is the Killing form of I}, then B(I}IX, 1}1i) - 0 when a + ~ *0 .

6. The KIlling form restricted to fI. Is nondegenerated. This last property Implies
that for every a In A, there Is a vector Hex In fI. such that BCH, H~ - a(H) for aU
H In fl..
7. If ae A,then -ae A and we have

[I}IX, I} -ex] - CHcx ' aCHcx) *0 .
V.3. An order relation on the set of roots d
Let r be the rank of the semlslmple Ue algebra I}. It Is possible to find a basis
(ai' ... , a) of fl.- such that
1. <Xt e A , I - 1 ,..., r ,
2. Every root a can be decomposed as a sum
a - kl <Xt + k2 <Xz + ... + ~ afl
and the coefficients kt are all positive Integers or all negative Integers.
A root system S - (ai' ... , a) satisfying the previous conditions Is called a system of
simple roots.
Now, we fix a system of simple roots. The elements of this system are called simple
roots. A root a Is called positive Crespo negative) If the coefficients of a In
decomposition In the simple roots are positive Crespo negative).
Generalities 31
We denote A+ the set of positive roots, and A_ the set of negative roots. Then:
A - A+ u A_, A_ - - A+ .
From a fixed simple roots system S - Cex l , ... , ex), we can provide fl.- with a partial
relation by stating that y > v <=> y - v has a decomposition as a sum of simple
roots exl with positive coefficients. of course, A Inherits this partial order relation.
V.4. Weyl Bases
We keep the notation of the preceding section. We construct a basis of 9 adapted to

the decomposition In elgensubspaces
9 - fI. EEl ( EEl + 9 a ) EEl ( EEl _ 9 a )

aeA aeA
and to the chosen system of simple roots S - Cexl' ... , ex;J.
For each I, we put HI - H IX! and we choose elements XI E 91X!, YI E 9- IX! such thaI
[xl' YI] - HI . Finally, we put nCi, j) - CXtCH~ . The matrix formed by the numbers nCi,p Is
called the Canan matrix of the given system s. We note that nCi,p Is an integer ~ 0 If
*
I j and n(l, I) - 2.
Theorem 9. The vectors CHI' ~, YI) I - 1 ,... , r satisfy the fol1owing conditions
1. n+ - 1: 9 a is generated by the vectors XI ,

aeA+
n- - 1: 9a is generated by the vectors Y I ,

aeA-
9 is generated by (X h Y h HI) ,
2. [Hh HU - 0 ; [Hh Xu - n(I,j) Xj; [Hh YU- - n(i,J) YI '

*
[X I, Y I] - HI; [X h YU - 0 If I j,
3. Cad x l}nCl,j)+l ('9 = 0, 1* j,
Cad y l)-n(l,j)+l (Y~ - 0, 1* j .
The Weyt basis is a basis of 9 defined by the vectors (XI' YI' H~ and the above
32 Chapter 1
relations. Its vectors are in h. or in the eigensubspace 9 a . If Xa is the corresponding

vector in the Weyl basis, we have
[xa' xpl - 0 ifa + ~ eo d,
[xu' xpl - Nu,pX a+p if a + ~ Ed,
with Na,p - - N_a,_p and CNa,~)2:::; qC1-p)aCH a )/2 where p and q are defined as this :
~ + na is a root for each n, p S n S q and ~ + ka is not in d if k eo [p, ql.
V.S. On the classification of complex simple Lie algebras.
Based on particular properties of the roots, the following properties are

concentrated in the Dynkin diagram.
Let S - Cal' ... , a) be a simple roots system. These roots will be the vertices of a
graph. Two vertices will be formed by nCi, j) . n(j, 1) edges. These integer numbers are
equal to 0, 1, 2 or 3. We put the inequality sign> on the multiple edge to indicate
which of the two adjacent roots is the longer; the absence of an inequality sign means
that two adjacent roots have the same length. The length of a root a l is given by
< ai' a l > where < , > is the restricted Killingform to fl.., Thus, the graph labelled is
called the Dynkin diagram.
Theorem 10. All the Dynkin diagrams of the simple complex LIe algebras are the
foHowing:
Type of algebra Dynkio diagram
0<1 0<2 0,.-1 Or

Ar(r2:1) 0--0- - - --0--0
0<1 0<2 0<3 0',.-1 Q r

Bn (n 2: 3) ~---~
0<1 0<2 O'r-l Q r

en (n 2: 2) 0--0- - - - ---o::::¢:O
Generalities 33
al a2 a3 ~ ar_1
Dn (n 2: 4) 0----0--0- - --
a r
al a3 a4 as 0<6
~
£6
0<1 0<3 0<4 o<s 0<6 0<7

£7
~
al 0<3 0<4 as 016 0<7 0<8
£8
I
~
0<2
0<1 0<2 013 a4

F4 ~
a1 a2
G2 <::l=$O
Remark. Indeed the Dynkin diagrams An "'" Ea correspond to roots systems and
therefore to simple complex tie algebras, We note that two isomorphic simple tie
algebras admit the same diagram and, conversely, two simple tie algebras having the
same Dynkin diagram are isomorphic, Hence, the classification of complex simple tie
algebras is deduced from theorem 9, It is sufficient to give the corresponding tie
algebra for each diagram :
- type Ar : slCr+ 1,«:) ,

- type Br : so(2r+ 1,«:) ,
- type Cr : spe2r) ,
- type Dr : so(2r,(C:) ,
34 Chapter 1
These algebras are described in Section 104. The diagrams E6, E7, E8, F4 and G2
correspond to the exceptional simple Lie algebras. These algebra admit also matrlcial
representations. For example, the simple tie algebra G2 may be presented as a
derivation algebra of a Cayley algebra.
VI. THE NILRADICAL
Vl.l. Definition
Theorem 11. Let q be a Lie algebra and r - rad q its radical. The following four sets
are identical :
(a) the largest nilpotent ideal ofq,
(b) the largest nilpotent ideal of rad q,
(c) the set of x E rad q such that ad qX is nilpotent,
(d) the set of x E rad q such that <dr x is nilpotent
The ideal defined in this theorem is called the nilradlcal of q and denoted nil q.
Proposition 16. Let q be a Lie algebra and f a derivation of q. Then:

f(q) n rad q c nil q .
This proposition is based on the follOWing relation :
ad (f(x» - [f, ad x] E f (Der q) n rad (Der q) for all x in q.
Corollary
1. The nilpotent radical Dq n rad q of q coincides with [rad q, qJ, hence it is
included in nil q .
2. For every f of Der q, f(rad g) c nil q .
Generalities 35
VI.2. On the algebraic Lie algebras
Algebraic Lie algebras are the Lie algebras of algebraic Lie groups (see [Bo]). They
constitute an interesting class of Lie algebras containing, in particular, semlsimple Lie
algebras and nilpotent Lie algebras. The algebra of all derivations of a Lie algebra Is also
algebraiC. For these Lie algebras, we have a similar decomposition to levi's
decomposition.
Theorem 12. Let 9 be an algebraiC Ue algebra.. Then there is a reductive subalgebra. c;a.
such that 9 - ni 19 EB c;a. (semidirect sum).
VB. THE CLASSICAL INVARIANTS OF NILPOTENT LIE

ALGEBRAS
The classification up to isomorphism of the Lie algebras is fundamental and a very

difficult problem. It is probably one of the first problems that we encounte when
understanding the structure of a set of Lie algebras. In the follOwing sections, we shall
give an algebraic structure to this set in order to present a general survey on the
behaviour of Lie algebras. From the geometrical point of view, the classification of Lie
algebras corresponds to a fibration of this set, the fiber being the isomorphic classes.
We have seen that the classification of semisimple complex Lie algebras has
been known ever since the works of Cartan and Killing. Modulo the study of
derivations of solvable Lie algebras, the classification of the Lie algebras may be
reduced by the Levi theorem to the classification of solvable Lie algebras and, in
particular, of nilpotent Lie algebras. This shows that the study of nilpotent and
solvable Lie algebras is very important. But it is very difficultto distinguish two non
isomorphic solvable or nilpotent Lie algebras. Then it is interesting to know some
invariants of these Lie algebras which are easy to present and to compute.
36 Chapter 1
VII.t. The dimension of characteristic ideals and the nilindex
Let 9 be a complex finite-dimensional nilpotent lie algebra. of course, dim 9 is the

simplest invariant. Then it is natural to classify the lie algebras of a given dimension.
Currently, we know the classifications of complex nilpotent lie algebras of
dimensions less than 8. These classifications are given in the next section. It is clear
that the elaboration of these classifications Is made using a finer invariant than the
dimension of 9.
co The nilindex of 9
By definition, the nilindex of 9 is the smallest positif integer s such that

CS 9 - {O}.
in particular, if s is the nilindex of 9, we have Cad X)S - 0 for all X in 9.

We can also consider the minimum of the integer k such that Cad x)k - 0 for all X in 9.
Examples
1. If s - 1, then 9 is Abelian. It is unique for a given dimension.
2. If s - 2, then we say that 9 Is two-step nilpotent. The Heisenberg algebra is of this

type. We note that the classification of two-step nilpotent Ue algebras Is unknown for
dimensions greater than 9 (see Chapter 2).
3. If s - dim 9 - 1, we say that 9 is filiform. The study of these algebras can also be
found In the Chapter 2.
(10 The dimension of the characteristic ideals

Consider the derived sequence, the descending central sequence, and the ascending
central sequence :
Generalities 37
o 1 s
9 - C 9::::> C 9::::> ... ::::> C 9 - {O},
{O} - C~ c C19 c ... c Cs9 - 9 .
We note cit - dim DlcJ, c1- dim ClcJ andc,- dim C8 .

Then the sequence of the dimensions Cn, d l , ... , ~-l' cl , ... , cs - l , c l , ... , Cs-l) Is an
invariant of 9. The classification of complex nilpotent Ue algebras of dimension 6 (and
less than 6) can be obtained using only this invariant
VII.2. The characteristic sequence
Let X be in 9. We denote c(X) the ordered sequence of a similitude invariant of the

nilpotent operator ad X, which is the ordered sequence of the dimension of the
Jordan blocks of this operator. In the set of these sequences, we can use the
lexicographical order:
(Cl' C;2, ••. , c s ) ~ (db ... , dp) <=> 3 i such that cJ - ~ for J d; .
Let c(9) be defined by c(9) - Max (c(X)).

XEIj-DIj
This sequence is an invariant of 9 called the characteristic sequence.
Examples
1. c(9) = (n-l, 1)
There is X in 9 such that (ad X)n-2 "* 0, and the Jordan form of the matrix of ad X is
o 1 0 ... 0
o 0 1 ... 0
o 0 0 ... 1
o 0 0 ... 0
38 Chapter 1
The nilindex of 9 is equal to n-l and 9 is filifonn.
2. c(9) - 0, 1, ... , 1)
For every X in 9 - 1)9, the Jordan matrix of ad X is zero. Then 9 is Abelian.
3. c(9) - (2, 1, ... , 1)

This case corresponds to the Heisenberg algebra.
The classification of 7-dimensional complex nilpotent Ue algebras can be made using

only this invariant (see the next Chapter).
VII.3. The rank of a nilpotent Lie algebra
In the Ue algebra Der 9 of the derivations of the nilpotent Ue algebra 9, we consider

the maximal Abelian subalgebras constituted from semisimple elements. From
Mostow's result [Mo], these algebras are conjugate under the action of the group of the
inner automorphism of 9. These Abelian algebras are sometimes called the maximal
torus of derivations. So, the dimension of the maximal torus of Der 9 is an invariant of
9. It is called the rank of 9.
We note that if the rank of 9 is maximal, it is equal the dimension of 9 - 1)9.
VII.4. Other invariants
Here are some of the invariants used in this book:
CD the dimensions of cohomologtcal spaces (for the trivial values in a cohomology, the
cohomology with coefficients 9-module, in the adjoint module, etc.),
Generalities 39
(10 the characteristic of the tie algebra of derivations,
cuo the dimension of the center,
(Iv) Dixmier's invariant. It is the dimension of a maximal orbit for the coadjoint
representation (see Chapter 7).
CHAPTER 2
SOME CLASSES OF NILPOTENT

LIE ALGEBRAS
This chapter is devoted to some important classes of nilpotent Lie algebras. We

also consider the problem of classification for the small dimensions.
I. FILIFORM LIE ALGEBRAS
1.1. Basic Notions. Examples
Let 9 be a nilpotent Lie algebra of d!menslon n. Let C~ be the central descendJog

sequence of 9 .
Definition 1. The Ue algebra 9 is called filifonn if

k
d!mC9-n-k-l ror k~l.
We can give another characterization of filiform Lie algebras.
Proposition 1. The Ue algebra 9 is filJfonn if and only if there is a nonzero vector X

in 9 -:09 such that the characteristic sequence cOO - Cn-l, 1).
Nilpotent Lie Algebras 41
Proof. Recall that cOO is the ordered sequence of the dimension of Jordan blocks of
ad X. If c(X) - (n - 1 , D, there is a Jordan basis (X - Xl' ... , Xn) of ad X satisfying
[Xl , ~] - ~-l for i - 3, ... , n
and C 19 is generated by {X 2 , X3 , .•• , Xn_l} for 2$; i. Then the tie algebra 9 is filifonn.
Conversely, we choose a basis {X o' Xl' ... , ~) corresponding to the filtration
9 - CO(9) :::> Cr9) :::> Cl:9) :::> ••• :::> Cn -\9) .
We can choose this basis so as to satisfy c(Xo) - (n - 1 , D.
Remark. The filifonn tie algebra have a maximal nilindex. These algebras are the "less"
nilpotent.
Examples
1. Let (,n be the (n + D-dimensional tie algebra defined by [XO' XI] - XI+I for i - 1,
... , n - 1 where (X o' ... , Xn) is a basis of (,n' the undefined brackets being zero. This
algebra is filifonn. This is, in a certain manner, the simplest filifonn Ue algebra.
2. Let Qn be the (n + D-dimensional nilpotent Ue algebra defined in the basis (X o' ... ,
Xn) by
[Xo ,XI] = XI+I , i - 1, ... , n - 1 ; [XI, Xn-t] = (- 1)IXn , i = 1, ..., n - 1 .
This algebra is also filifonn.
3. Let :Rnbe defined in the basis (X o, ... , Xn) by

[Xo , XI] - XI+I , i = 1, ..., n - 1 [X I, XI] - XI+2 ,i - 2, ... , n - 1 .
It is evidently filifonn.
4. Let Wn be the tie algebra whose brackets, in the basis (X o, ... , Xn), are:
°,
[X XI] - XI+l , i - 1, ..., n - 1 ,
6(i-l)!(j-l) <.. <
[XI,XJ]= (" ')' XI+j+I, 1_I<J_n-1 i+j+1$;n.
1+J •
This algebra is filifonn. It is isomorphic to the tie algebra defined by :

[YI,YU=(j-i)YI+J i+j$;n+1, l$;i<j$; (n-l)
which is a finite quotient of the nilpotent part of the Witt algebra.
42 Chapter 2
We shall see later that there are only two, up to isomorphism, (n+ I)-dimensional Ue
algebras, n ~ 11, whose brackets satisfy the relations:
[X 0 , XI] - XI+ 1, i-I, ... , n - 1
[XI,X~-aljXI+j+1 with l~i<j~n-l and i+j+l~n.
These Ue algebras are Ron and Wn.
1.2. Graded filiform Lie algebras
Let 9 be a filiform Ue algebra. It is naturally filtered by the ideals CI-~ of the

descending sequence (one put CIc:J - 9 for i ~ 0). Then we can associate to a filiform Ue
algebra 9 a graded Ue algebra, noted gr 9 ,which is also filiform. This algebra Is
defined by
We denote CI- 19 / C~ by 91. Then, we have gr 9 - 91 + 92 + ... + 9n with dim Ih - 2,

dim 9,-1 for 2 ~ i ~ n and [9" 9 jl c 91+J for i+j ~ n.
Lemma. There is a homogeneous basis (X o ' Xl' ..., ~) of gr 9 such that

X 0, XI E 91 , XI E 91 i - 2, ... , n ,
[XO,X I]-XI+I, l~i~n-l ; [Xo,Xn]-O,

[X 1 ,X 2]-0 ; [XI,Xj]-O if l~i<J and i+j*n,
[XI, Xn-I] - (- 1) I <xX n with <X - 0 if n is even.
Proof. Let Y I be a nonzero vector in gr 9 for eachi, 1 ~ i ~ o. There Is Y E 91' such

that
[V, YI] - fl (Y) YI+l ,
the linear map fl : 9t -+ ~ being not identically null. There is Xo E 91 such that
fl(X o) * 0, 'if i, 1 ~ i ~ n-l . One can choose the vectors XI E 91 ' with XI - AI YI'
AI * 0 ,so that they satisfy the conditions
[XO , XI] - XI+l, 1 S; is; n-l, [Xo, Xn] - O.

As we have dim 91 - 2, then [Xo, X2 ] - X3 and [Xl' X2 ] - ~. We can affinn that there
exists a unique Xl E 91 satisfying rx 1 ' X2] - 0 . The other relations are detennined by
induction on the dimension of 9. We consider the basis CX o' ... , Xn) of gr 9 defined
above. Then CX o ' X l' ... , Xn_1) is a basis of the filiform graded algebra gCl~/(CXn
satisfying the relations of the lemma. Suppose that n is odd_ We have [XI' XJ]- 0, if
i+j < n-I . One put rxl , Xn-t] - a l Xn . jacobi's identities imply a l - - a l+1 and then
a l - C- 1)1a. This gives the sought-after basis_ Suppose n even. Than one has
[XI, Xn-I-1] - (- 1) I aX n-1 [XI, Xn-t] - al Xn
and the Jacobi Identities imply
(-1) la - al+1 + al
and, thus, al - (- 1)2+1 (a1 + (i-I) a) . As we have anl2 - 0, we deduce
a1 - (I-I) a and al- (_1)1+1 (i-I) a .
The Jacobi identities for the vectors (X 1 , X (nl2) _ 1 , X n-2) implies a - 0 and al - 0 .
Corollary. If n is odd, then there are only, up to isomorphism, two (n+ 1)-
dimensional graded filiform Lie algebras. These algebras are Ln and Qn _
II. TWO-STEP NILPOTENT LIE ALGEBRAS
11.1. Definition and Examples
Definition 2. A nilpotent Lie algebra 9 is called two-step nilpotent if it satisfies

C29 = {o}.
Examples
1. The Abelian tie algebra is two-step nilpotent.

2. The Heisenberg algebra Hp of dimension 2p+ 1 is defined by the brackets :
44 Chapter 2
[Xl, X2] = [X3 , ~] = ••. - [X2p-l, X2p] - X2p+ 1
This Ue algebra is two-step nilpotent. It is, in a way, a model within these two-step
nilpotent Ue algebras. It is characterized by the following property :
Proposition 2. Every Ue algebra satisfying Z(I:O - C1(9) and dim Z(~ - 1 is

isomorphic to the Heisenberg algebra.
11.2. On the structure of the two-step nilpotent Lie algebras
Let V be a vector space complementary of CI(9) in 9 : 9 - Cl(9) ffi V . Note that

s - dim V - dim 9/C19 is the minimum number of generators of 9. As 9 is two-step
nilpotent, the derived algebra CI(9) is contained in the center Z(9) .
Let U be the subspace U - ZC~ n V . It is an Abelian ideal of 9. Consider a vector
space complementary of U in V . We have
9 - CI(9) ffi U ffi W, Z(9) = CI(9) ffi U.
We can deduce that h. - C1(9) ffi W is a two-step nilpotent subalgebra of 9 satisfying

ZCh.) - CIC~ and CICh.) - C1(9) .
Then every two-step nilpotent Ue algebra is a trivial extension of a two-step nilpotent
algebra whose center is the derived algebra extended by an Abelian ideal. So one can
suppose that the algebra 9 satisfies the condition: C I(~ - Z(9) .
Consider eX I' ... , ~, YI' ... , Yn-p) as a basis of 9 adapted to the decomposition:
Obviously, we have [xl' XJ] - [XI' Yk] - 0 . Then the Ue algebra 9 is entirely defined
by the only brackets : p
[Yi> YJJ = L a~J Xk 1 ~i<j ~n-p .
k-I
jacobi's identities are all satisfied and the constant aklJ are free. Intrinsically this means
that the structure of 9 is defined by a surjective linear map
cp: NV ~ Z(9) .
The dual point of view permits a concrete analysiS. The dual map
cp. : Z(9)· ~ A2CU)·
Is Injective. The data of this mapping Cand then the data of the bracket of 9) Is the same
as the data of a subspace of dimension p (p - dim ZCrp) in A2y . If Cal' ... , ap, ~1'
... , ~n-p) is the dual basis of a given basis (Xl' ... , Xp' Y 1, •.• , Yn-p), then the structural
equations of 9 Cthe dual Jacobi conditions) are :
dUk - L
J.g<j~-p
~j 13t,,~ .
The bilinear forms C9 1 - da 1 , ••• , 9k - dak) are two-forms on Y . They generate the
subspace cp"C ZCrp") . It appears clearly, in this writing, that the data of the two-forms
91 is eqUivalent to the data of the constants Cakij).
11.3. On the classification of the two-step nilpotent Lie algebras
At first Sight, the problem of classifying two-step nilpotent tie algebras appears to be
easy to solve. Recall that this problem was first approached by Umlauf In 1891 In his
thesis. Today, it is always an open problem. At the end of this chapter, we will give the
rare results concerning this problem. If we introduce a new and strong hypothesis as
the two-step nilpotency, we can hope for a simplification of this problem. Indeed,
this is not the case. From the previous results, we can affirm :
Proposition 3. The classification up to isomorphism of two-step nilpotent Lie algebras

is equivalent to the classification of the orbits corresponding to the actien of the linear
group G LCp,!C:) in the Grassmannian of the 2-spaces in the vector space A2C!C:P).
Now we consider the characteristic sequence of a two-step nilpotent tie algebra 9 . As

the nilindex of 9 is 2 Cif 9 is not Abelian), then the characteristic sequence of 9 has
the follOwing form :
s(9) - C2 , 2 , 2 ,... , 2 , 1 , 1 ,..., 1) .
Consider the class of 2-step nilpotent whose characteristic sequence is C2, 2 , ... , 2 , 1).
This class determines an open set in the algebraiC set of two-step nilpotent tie
algebras. Suppose that dim 9 - 2p+ 1. In this case, dim Z(9) - P and the brackets
of 9 are
46 Chapter 2
[VI' Yj ] -l:k-1,p a.l\rk' 1 SI.<jSp+l

where (X 11 ..., Xp, Y11 ..., Yp+ 1) is a basis of 9 and (X 11 •••, Xp) a basis of Z(9) •
We can verify that the parameters at) j are free (all the Jacobi conditions are satisfied).
Then this class of algebra is parametrized by the a\ j and is isomorphic to the vector
space Alt{Cp) of the alternating bilinear forms on CP with values on CP. SO the
classification of these two-step nilpotent tie algebras is eqUivalent to the classification
of the bilinear forms on CP with values on CP and contains the classification of all the
tie algebras of dimension p. We claim that this classification is impossible to establish.
III. CHARACTERISTICALLY NILPOTENT LIE ALGEBRAS
111.1. On Jacobson's theorem
Theorem 1. Every Lie algebra on a fIeld of characterlstlc zero havlng a

nondegenerated clerJvatlon Is nIlpotent
Comments. A derivation f of a tie algebra 9 is defined by

df(X, Y) - [ fOO, Y ] + [ X, feY) ] - f [ x, Y 1 - 0
(the operator d is related with the Chevalley cohomology). This derivation is
nondegenerated if the linear map f is nondegenerated (the determinant is not 0). For
example, if 9 is semisimple, every derivation is inner, i.e. is it can be written as f - ad
X for some X in 9. Such a derivation admits a nontrivial kernel, then every derivation
in a semisimple tie algebra is degenerated.
In his work, Jacobson presents the foUowingproblem : Is every nilpotent tie algebra
admitting a nondegenerated derivation 7
The answer to this problem is negative.
The first example of a nilpotent algebra whose derivations are degenerated was given
by Dixmier and tister in 1957 [D.L.]. It is the follOwing 8-dimensional tie algebra whose
brackets are
[Xl> X2l - Xs ' [~, X4l-~,
[Xl' X3l - X() , [X2, X()l - -X7 ,
[Xl' X4l - X7 ' [~, X4l - -XS '
[Xl' Xsl - -Xs , [X3, XSl - -X7 ,
[~, X 3l- X 8 , [X4,~l--X8'
This example was the starting point of a study of a new class of tie algebras called
characteristically nilpotent tie algebras.
Definition 3. Let 9 be a nilpotent Lie algebra and Der 9 its derivations algebra. The
Lie algebra 9 is called characteristically nilpotent if every derivation f in Der 9 is a
nilpotent endomorphism.
111.2. Characterization of characteristically nilpotent Lie algebras
Let 9[1) = Der if..9) - { Y E 9 such that Y - f(X), f E Der 9 ,X E 9 } and, more gener-
ally,
tit] = Der 9 (tit-I]), i > 1 .
The Ue algebra 9 is characteristically nilpotent If and only if there is k E JRII such that
9 1kl - {OJ.
This sequence 9[1) generalizes the central descending sequence; here we use the set of
all derivations instead of the set of inner derivations.
Theorem 2 [LTl. Let 9 be a nilpotent Lie algebra of a dimension more than 2. Then 9
is characteristically nilpotent if and only if the Lie algebra Der 9 is nilpotent
Note that the nilpotence of Der 9 doesn't directly imply the nilpotence of its
elements, but only the nilpotence of the adjoint operators. The proof of the theorem
is based on the follOwing lemma.
Lemma. Let L be a nilpotent Lie algebra such that L - 'Ll €a 'Lz where 'Ll and 'L2
48 Chapter 2
are ideals, and'Ll is a centra/ideal. Then Der L is nonniJpotent.

Proof. As 'Ll Is centtal, one has ['Ll' L1 - {a}. Let x '# 0 be In'Ll, and U such that
'Ll - U E9 rc:{x}. As 'L:z Is an ideal of L, it is nilpotent Then Its center is nontrivial. Let
be y '# 0 in ZC'Lz). Let fl andf2 be the derivations of L defined by
flC'Lz) - 0 , flCx) - Y , flCU) - 0,
f2C'Lz) - 0 , f2(x) - x , f2CU) - 0 .
Then we have [£1' f21- fl and Der 9 Is not nilpotent.
Proof of the theorem. It is obvious that if 9 is characteristically nilpotent, then Der 9

is nilpotent. Now we prove the converse. Suppose Der 9 nilpotent and dim 9 > 1.
Let <X be in the dual space of Der 9, and let Va be the subspace
Va - ( X E 9 such that 3 mEN with Cf-<xCf) Id)mCX) - 0 , 'V f E Der 9 }.
If Va:j!: {a} , we call <X a weightof Der 9. Note by n the set of weights of Der 9.
One has
9 - E9 Va
aeD
and [Va,V~]cVa+~ If <X+~E n or [Va,V~]-{O} Ifnot.

As every ad X is a derivation of 9, we deduce that each space Va is.an ideal of 9.
Onededuces:
[Va,V~] eVa n V~ n Va+~'
Then, [Va,V~]-{O} when <x'#o or ~'#O.
We consider the spaces
and 12 - Vo .
Then 9 is a direct vectorial sum of'Ll and'L2, which are ideals of 9, 'Ll being centtal.
The previous lemma implies 'L t - {OJ or'L2 - {OJ.
if'Ll - {OJ, then 9 - 'L2 - Vo and every derivation of 9 is nilpotent.
if'L2 - (OJ, then 9 - 'L l . This implies that 9 Is Abelian and Der 9 - glCn,rc:).
As Der 9 Is nilpotent, we deduce that n - 1, and we find a contradiction.
111.3. Examples
co The smallest dimension where we meet an example of complex characteristically

nilpotent Ue algebras is 7. This example has been given by Favre [Pa]:
[X 1, X2] - X3, [Xto~] - X 7,
[X to X3] - Xi, [X 2 ,X3 ] - ~,

[Xl'~] - X5, [X2'~] - [X2,X5] - - [X 3,Xi] - X 7·
[X 1,X5] - ~,
Note that in dimension 7 there is one parameter family of nonisomorphic

characteristically nilpotent Ue algebras. Contrary to Favre's example, these algebras are
not filiform. This family is denoted by n.,2,U in the list which will be presented in
Section V.2 of this chapter.
(li) In [Bou 11 we can read the follOwing Ue algebra :

[X to X2] - X 3, [X 2,X3] = X5, [X3'~] - - X 7 + Xs,
[X lo X3] - Xi, [X2'~] = ~, [X 3,X5 ] = - Xs·

[Xto~] - X5, [X 2,X5] - X 7,
[X to X5] - ~, [X2'~] - 2 Xs ,
[Xl'~] = Xs,
[X 1,X7] - Xs ,
This algebra is characteristically nilpotent and not filiform.
111.4. Direct sum of characteristically nilpotent Lie algebras
Theorem 3. Let 9 - E9Pi~l 9i be a Lie algebra where the 9i are ideals of 9· Then 9
is characteristically nilpotent if and only if each ideal 91 is characteristically nilpotent.
50 Chapter 2
Proof. We have seen, In a previous section, that
Der (~ 9i) = .~ (Der 9i) E9 ($i"j :I)(9i.9J») •

1=1 1=1
where
:1)<9" 9j) - { f E End 9 , f(9iJ - 0 if k "* I, f(91) c Z(9t) and fi91' 9 j1 - 0 }.

We suppose that each 91 is characteristically nilpotent. In this case, we have
Z(91) c [91' 9j1. Indeed, if It Is not the case, there Is a nonzero vector X E Z(9) with
X ~ [c;J" 9 j1. If U is a complementary space of (CX In 91 ' we define the derivation f by
f(X) - X and feu) - o. This derivation Is nonnilpotent. So we have Z( 91) c [91' 911.
One deduces :
mIf fl E Der (91) and fj E Der C9j) with I "* j, then fl 0 fj - fj 0 fl

(lI) flo f2 - 0 iff 1 E :1)(91' 9j) and f2E :l)CI.Jtq 91)'
As we have also fCZC~) c Z(9) for all f E Der 9, we deduce that every derivation is
nilpotent.
p
Conversely, we suppose that 9 - $ 9i is characteristically nilpotent.
i=1
Each derivation of 91 can be extended naturally to a derivation of 9. Then this
derivation Is nilpotent.
N. STANDARD LIE ALGEBRAS
N.l. Parabolic subalgebras of a semisimple algebra
Let 9 be a semislmple complex Ue algebra.

Definition 4. A subalgebra 90 of 9 Is called parabolic If It contains a Borel subalgebra
of 9.
Recall what is a Borel subalgebra. First we fix the notation : fI. is a Cartan subalgebra of
9, A the set of roots corresponding to fl., S a basis of A Cthe simple roots), A+
Crespo A_) the set of positive Crespo negative) roots Crecall that A - A+ u A_ ) and
9 a - {X E 9 / [X, Hl - cxCH)X, 'v' HE fl.} .
The space 9 a has a dimension one. We choose a nonnull vector Xa in 9 a . A Borel
subalgebra & of 9 is a solvable maximal subalgebra of 9. It is conjugate, up to an
inner automorphism, with a subalgebra of the follOwing type :
&' - fI. E9 L
9a·
aeL\..
The parabolic subalgebraare determined, up to an inner automorphism, by a subset

Sl of S. Let Al be the set of roots whose decomposition on S contains only
elements of S - Sl . One denotes A2 - A - AI' A2 + - A2 n A+ . Then the Ue algebra
p - fI. E9 L 9a
aetJ.2+utJ.,
is paraboliC and every parabolic subalgebra is conjugated with an algebra p.

We note that the nilradical of p is
n= L 9a
aetJ.2+
and its reductive part is r - fI. E9 L 9a

aetJ.,
N.2. Standard sub algebra
Definition 5. A subalgebra of a semisimple Ue algebra is called standard if its normal-

izer is a parabolic subalgebra.
These algebras were first studied by G.B. Gurevich [GUll but in the case where the
simple algebra 9 is of type Ar • The motivation of this study is founded on the theory
of complex homogeneous spaces: let M be a compact homogeneous space M - G/H
G being a complex Ue group and M a closed subgroup. If 9 and fI. are the lie
algebras corresponding to G and H, the normalizer of fI. in 9 is parabolic. This
result has been established by Tits. It permits us to translate the study of
52 Chapter 2
homogeneous complex spaces in the study of standard subalgebras. In order to

simplify the writing, we shall call every standard subalgebra of a semisimple tie algebra
standatd aJsebra.
We note that, for every standard algebra t (not necessarly nilpotent) whose nonnalizer
has the follOwing fonn
(which is the canonical fonn of the parabolic algebras), one has :
t - 11.1 e 1 9« where I:i' c I:i and 11.1 c 11. •

aeD.·
In the preceding chapter, we introduced that the partial order relation on the dual 11. *
of the Cartan subalgebra 11. : COl ~ IDz ¢::> 0)1 - IDz is a linear combination of simple roots
with nonnegative coefficients.
Proposition 4. Let t be a standard aJsebra such that Its nonnalizer can be expressed
as 11. e 19a. Suppose that a and ~ are posItive roots with a S ~ .
aeD.lu~+
d the subspace 9a is Included in t, then 91i is also In t.
Proof. We consider in each subspace 9 a a nonzero vector Xa. As 'Y ~ a , one can write
'Y - a + 'Yl + ... + "I'k where the 'Yi are positive roots. One can order again the terms of
this sum 'Y - 'YiJ + ... + 'Ylo + a + 'Yio+l + ... + r.. '
such that every partially connected sum
beginning with 'Yt, are positive roolS. Each vector Xli is in the nonnaliz.er of t. Then the
image of the vector X a by ad X n. 0 ad X"A._1 0 ••• 0 ad X t".1 0 ad X "fIl + ... + 'YI,. is in the
space 9'Y" This proves the proposition.
A description of standard algebras is presented in the general case in urn 1]. Here we
study only the nilpotent standard algebras.
Nilpotent Ue Algebras 53
N.3. Nilpotent standard algebras
Let R be a subset of 8+ that consists of pairwise incomparable roots (for the partial
order in h.' given in the preceding section). One puts :
Rl- {ae 8+/a;;::~ for ~e R}.
The subspace n - L 9a is a nilpotent subalgebra of 9.
aER,
The normalizer of n contains a Borel subalgebra. Then n is a nilpotent standard

subalgebraof 9. We will say that n is the nilpotent standardsubalgebraassociated to
R. This process permits us to construct more easily such subalgebras. The follOwing
theorem shows that every standard algebra is of this type. So one obtains a complete
description of these algebras.
Theorem 4. Let n be a nilpotent standard subalgebra whose nonnalizer has the

fonn 11.1 ffi L 9a· Then there is a subset R of 8+ whose elements are two
aE.11u.12
by two incomparable and n is the stand.ard. nilpotent subalgebra associated to R.
Proof. As n is an ideal of its normalizer, it is contained in the radical of this

normalizer. But n can be written as n - L
9a for some 8' C 8. One deduce
aEtJ.'
8' C 8 2 +C 8+. Let R be the subset of 8' constituted of all minimal elements, always for
the partial order law. These elements are pairwise incomparable. Then R defines n.
Corollary. Every standard nilpotent algebra is conjugated to a nilpotent stand.ard.

algebra associated to a set R of roots pairwise incomparable.
N.4. Structure of the normalizer of a nilpotent standard subalgebra
Definition 6. A basic root a E S is calJed extremal for ~ e 8+ ifit satisfies a ~ ~ or

~-ae 8.
If ~ e ~ , then we note sP the set of extremal roots for ~.

54 Chapter 2
Lemma 1. Suppose that ai' ~, a" a 1 + ~, a 1 + ~ +a, are roots such that a 1 + a,
and ~+ a, are nonzero. Then one of these comblnal1ons a 1 +a, or ~ + a, is a
root.
The proof Is a direct consequence of the JacoblldenUty concerning (Xal,Xa2,Xa,).
Lemma 2. Let a and ~ be two posIl1ve roots such that a + ~ E A. If 1 E S Is not

extremal for a, and for ~, then 1 Is not extremal for a + ~.
Proof. We suppose 1 to be extremal for a + ~. As 1 E S, then 1 ~ a + ~ and a + ~ -

1 E A. From lemma 1, a - 1 or ~ - 1 Is a root and 1 Is extremal for a or for ~. This
Is Impossible.
Lemma 3. Let a and ~ be two roots such that the decomposIl1on of a as the sum of
the basIc roots contains all the roots of sli Cthe set of extremal roots of ~) with a
mull1plicity greater than that of ~. Then a ~ ~ .
Proof. We put a - ~ + 11 + ... + 'Yk- ~c ...- ~ with 'YI and ~ In S and 'Y1 ~ ~ for
every I and j. By hypothesis, the roots ~1' ... , ~ are not In sli. Suppose now that
a is not greater Cor equal to) than ~ Ci.e. m ~ 1). We are going to show that v - ~ + 'Y1
+ ... + 'Yk Is a root of A and that ~1' .•. , l;.n are not in Sv. For k - O. the result is
obvious. We suppose k ~ 1 and we make an inducUon on k. One has :
k m
C-) (~.~) = (~,a) + ~ (~.-'Yi) + ~ (~.~J > 0,
1-1 j-1
where C ,) is the usual product on the set of the roots associated with the Killing-
Cartan form.
m
As (p, (X) ~ 0 and ~ (~.~j) ~ 0 Cif not ~ - (X and ~ -1: ~j are roots, this Is opposite to
j-1
m ~ 1 or to ~ 1 ••••• ~ m ~ S~. from C-) there Is an integer r. 1 ~ r ~ k such that one has
(~ , - 'Yr) > 0 . So ~ + 'Yr E A. From Lemma 2. ~ 1 ,...• ~m ~ sP + Yr. We apply the
induction hypothesIs for proving the previous assertion.
m
So. one has (V,v) =(v,a) + ~ (V'~i) > 0 and ~lo ••• ,~m ~ sv. One deduce (v,a) > 0
1-1
and ~ - V- a - ~ 1 + ... + ~m is In the set /l. We can suppose that each paxUal sum
~1 + ... + ~ of~ is In A. Let" - ~1 + ... + ~-l" Then a - -" -~ + v. From Lemma 2,
one has V - " - V - ~1 - '" - ~_I Is aroot. By a simple induction. we canafflnn that
V - ~I E A. As ~ It!: sv, this Is Impossible. We have proved the lemma.
Theorem S. Let n be a standard nilpotent subalgebra associated to a set R of

pairwise Incomparable roots. Then the normalIZer N(n) of n (It Is parabolic because
n Is standard) is defined by the subset
SI- U SPeS.
PeR
Proof. Let r be a parabolic subalgebra associated 10 the system SI and let n be
n - ~ 9a CR 1 is defined above). From Lemma 3, we can write r e N(n) .

aeR,
Suppose that r ~ N(n). As A - Al U A2 • there Is a E A2+ such that X-a E N(n).
This shows that the decomposition of a in simple roots contains one element V of
SI (by definition. v Is an extremal element for ~ E R). But N(n) Is a parabolic
subalgebra. Then X-v E N(n). More so we have ~ - v E A or ~= v. If ~ - V E A,
then
[Xj30 X-v] - N 'X~-v E n With N ~ o. If ~ =v, then [Xj30 X-v] - [Xv. X-v] E n, but this
vector is also in the Cartan subalgebra. These two cases lead to a contradiction.
Corollary. Every system ReA+ of palnvlse Incomparable roots deflnes a nilpotent

stanciaJd subalgebra
n- L 9a
whose norma.lJzB Is the parabolic subalgebra associated 10 the subsystem
SI - U Sp.
PeR
Conversely, every nilpotent standard subalgebra Is conjugated to a subalgebra n
associated to a subsystem ReA+ of pairwise Incomparable roots.

56 Chapter 2
&5. On the nilpotent algebras of maximal rank

Let 9 be a semisimple complex lle algebra and 9+ - L 9a its nilpotent part. If 11. Is
aet..+
a standard nilpotent lle algebra, then the quotient 9+/11. Is also nilpotent
This process permits us to construct a class of nilpotent Ue algebras. Every Ue algebra
of this class, and every standard nilpotent tie algebra Is given by a subset R c 11+ that
consists of pairwise incomparable roots. If we consider the subset R c 11+ whose
elements are pairwise incomparable and containing an the decomposition In simple
roots) at least three distinct simple roots, then we obtain a subclass of nilpotent lle
algebras. These algebras are nothing but the nilpotent lle algebras with maximal rank
studied by G. Favre and L Santharoubane [FS.].
&6. Complete Standard Lie algebras
Definition 7. A standard nilpotent algebra 11. Is called complete If It Is the nlbadlcaJ of

Its normalIzer.
It is easy to see that such an algebra is conjugated to a subalgebra assodated to a subset

of simple roots.
Theorem 6. Let 11. be a standard nilpotent Ue algebra assocIated to a system R c 11+

of pairwise Incomparable roots. The follOwing propositions are equivalent
CO 11. is an intersection of complete standard nilpotent subalgebrasassoclated to
some subsystems of S
CU) every element ~ E R can be expressed as cx l + ... + cxm with cx l ' ... , cxm E S
and cxl'¢: cxj for i ¢ j.
Proof: CO => CD).

We put
R- {~lo""~k} and CPR = {~ {cxd I

1=1
CXj E s(3. ,k} .
i - l, ...
Every element S· of «I>R Is a subset of S and defines a standard nilpotent complete

subalgebrans' We easdyshow that n ens' for all S· E cIIR Then n 1- nns, for an
0 0
s· E ~ saUsfiesn c n 1• Suppose that a E .1+, Xu E n 1 andXu E n. The

decomposition of a in simple roots does not con13Jn roots of S~ when ~ E R (if
not, we shall have Xu En, from Lemma 3).
k
Consider S· -U i=l{ai} of cIIR. where a:l: ai for i - 1,...,k (recall that al E sl'), then
X a Ens' ; this Is contrary to the choice of Xa.
(ii) ~ (i).
Let be n - ()-l,nnl where I1t is a complete nilpotent stanadard subalgebra associated.
to the system SI of simple roots. Consider a root ~E R. Suppose that Its
decomposition ~ - klal + ... + kmam contains acoefficient kl strlcdy greater than 1.
As the vector X~ is in n - ()-l,nnl , there is a root 11 in SI such that ~ ~ 11 for alii, 1 ~ i
~ n. Then for the root ~ - al + ... + <Xu! E R+, we shall also have ~. ~ 110 Thus, the
vector X~ Is In each nl for alii and, 1hus,1s In n. This Is Impossible because po < ~o
Note. If the semislmple algebra 9 Is of type An' condition en) is necessarily satisfied
The decomposition of n as an Intersection of the complete subalgebra Is always true.
This result was proved by Gurevich [GU].
V. ON THE CLASSIFICATION OF NILPOTENT COMPLEX

UEALGEBRAS
V.I. The classification in dimensions less than 6
The classification given In this section (and in the follOwing sections) concerns only
complex lJe algebras. By convention, the undefined brackets are null, except those
which emanate from the anticommutativlty.
58 Chapter 2
Dimension 1
fl, 11 - A.t - the Abeban algebra of dimension 1.
Dimension 2
fl,1 2 - A.t 2 - the Abelian algebra of dJmension 2.
Dimension 3
fl,l' - A.t' - the Abelian algebra of dimension 3,

fl,2' - HI - the Heisenberg algebra defined by [XI,X2] - X,.
From dJmension 4, we don't write the decomposable algebras, i.e. the tie algebras
which are dJrect sums of proper ideals.
Dimension 4
Dimension 5
fl,1 5 : [XhX5] - ~; [Xl'~] - X,; [XhX,] - ~,

fl,2 5 : [XI'X5 ] - ~ ; [XI'~] - X,; [XI'X,] - X2 ;
[X 4 ,Xs] - ~.
These two laws are the 5 dimensional ftlifonn laws.
fI,,5 : [X1'XS] - ~ ; [Xl,X,] - X2 ; [X2,X,] - ~ ,

fl,4 5 : [XI'XS] - ~ ; [X2'X,] - ~ ,
fl, S5 : [XI'XS] - ~ ; [XI'X,] - X2 ; [X"XS] - ~ ,
fl,6 5 : [XI'Xs] - ~; [XI,X,] - X2 ,
Dimension 6
nl 6 : [X I ,X61 - Xs; [X1,X51 - ~; [XI,~1 - X, ; [XI,X,1 - X2 ,

ni : [XI,X61 - X5 ; [XIoX51 - ~; [XI,~1 - X, ; [XI,X,1 - ~ ;
[X5,x.s1 - X2 ,
n,6 : [XI,x.s1 - Xs; [X1,Xs1 - ~; [XI,~1 - X, ; [XI,X,1 - X2 ;
[X"x.s1 - X2 ; [X4,X51 - - X2 ,
ni : [XI'x.s] - X5 ; [X1'X5] - ~; [XI'~] - X, ; [XI,X,1 - ~;
[X5,x.s1 - X, ; [X4'x.s] - X2 ,
n5 6 : [XI'x.s] - Xs; [XI,Xs1 - ~; [XI'~] - X, ; [XI'X,] - X2 ;
[X"x.s1 - X2 ; [X4,X5] - - ~ .
These laws are the Alifonn laws.
n6 6 : [XI,x.s1 - Xs; [X1,X5] - ~; [XI,~1 - X, ; [X2'x.s] - ~ ; [X2'X5] - - X,

n6 7 : [XI,x.s1 - Xs; [XIoXs1 - ~; [XI,~1 - X, ; [X2,x.s1 - X, ,
n6 8 : [XI'x.s] - Xs ; [X1'X5] - ~; [XI'~] - X, ; [X2,x.s1 - X2 ;
[X5,x.s1 - X, '
n6 9 : [XI,x.s1 - X5 ; [X1,X5] - ~; [XI,X,1 - ~ ; [X5,x.s1 - X, ;
[X4'x.s] - ~ ,
n6 10 : [XI'x.s] - X5 ; [X I,Xs1 - X,; [X"x.s1 - X2 ; [X4, X51 - ~ ,
n6 11 : [XI'x.s] - X5 ; [X1,Xs] - ~; [XI,X,1 - X2 ; [X2'X, - ~ ,
n6 12 : [XI,x.s] - X5 ; [X 1,Xs1 - ~; [X2X,1 - ~ ,
n6 1' : [X I ,x.s1 - X5 ; [X1,Xs1 - ~; [XI,~1 - X, ; [X5,x.s1 - ~ ,
n6 14 : [Xl,x.s] - X5 ; [X 5,x.s1 - X,; [X 4,x.s] - X2 ,
n6 15 : [XI,x.s] - X5 ; [XI,~1 - X,; [X5,x.s1 - X, ; [X 4,x.s1 - X2 ,
n6 16 : [XI,x.s1 - X5 ; [XI'~] - X,; [X5,x.s1 - ~ ,
n6 17 : [XI,x.s1 - X5 ; [XI.Xs1 - ~; [X5,x.s1 - ~ ; [X"x.s1 - ~ ,
n6 18 : [XI,x.s1 - X5 ; [XI,X51 - ~; [X"x.s1 - X2 ; [X5,x.s1 - X2 ,
n6 19 : [X 1 ,x.s1 - X5 ; [Xl'~] - X,; [X2,x.s1 - X, '
n620 : [XI,x.s1 - X5 ; [XI'~] - X,; [X4,x.s1 - X2 .
60 Chapter 2
V.2. The classification in dimension 7
The previous tables show that, for dimensions less than 6, there is a finite number of
Isomorphism classes of complex nilpotent tie algebras. But, for dimension 7 and
more, we are going to find families with one or more parameters of nonisomorphic
tie algebras. This Is the same as saying that the variety of Isomorphism classes has
nontrivial dimension. Finding one-parameter families of Isomorphic classes of 7
dimensional nilpotent algebra is not the only difficulty. There Is, in a variety of
isomorphic classes, some isolated points. So every table of dimension 7 is very
difficult to establsh. Now, we have 3 or 4 tables given by various authors.
Unfortunately, It is not easy to compare these lists because of the techniques used
and the Invariants are very different Ccharacteristic sequence, rank, extension). The
links between the invariants used are not evident. We begin by giving all the one-
parameter families.
Theorem 7. There are six families with one parameter nilpotent Ue algebras of
dJmensJon 7, palrwlse not isomorphic:
"'7 1,a : [X1,Xt] - Xt-l i - 3, ... ,7

[X4,X7] - aX 2 [XS,X7] - Cl+a)X,
[X SX6] - ~ [X6,X7] - Cl+a)~
"'72,a : [Xl'Xt] - Xt-l 1-4,...,7

[X2'~] - X3 [X 2,X7] - X3 + ~
[XSX 7] - aX3 [X6,X7] - aX 4 + X2
"'73,a : [X1,Xt] - Xt-l 1-3,4,5,7

[X 4,XS] - ~ [X 4,X7] - aX 2
[X SX6] - X2 [Xs,X7] - Cl+a)~
"'74,a : [X1,Xt] - Xt-l 1-3,4,5,7

[X 4,X7] - aX 2 [XS'~] - X2
[XsX 7] - Cl+a)X3
n 7s ,a : [Xl'~] - ~-l 1-3,4,5,7

[Xs'~] - [X6'X7 ] = X 2 [X 4 ,Xs] - cxX 2
[XSX 7] - X,
1-4,6,7
[X 4 ,X7 ] - X2 [X2'~] - cxXs
[X 2X 7] - [X,X 4 ]:;:: Xs
Remark: Each Ue algebra of the family ~l,a Is filiform, and the Ue algebras of the
family n 72 ,a are characterlsdca1ly nilpotent
The classification of 7-d.imenslonal nilpotent Ue algebras presented here is established

usingthe characterlsdc sequence as the prlncipal Invariant We don't give anything
concerning the proof, the nonisomorphism between two algebras can be seen by
using isomorphisms preservirag the Jordan form of the characterlsdc vector, and
isomorphisms transforming the characterlsdc vectors into characterlsdc vectors.
Lie algebras whose characteristic sequence is c(9) - (6,1). Filiform Lie algebras.
n 7 1,a : [Xl,Xt] - xt-l ' 3 SIS 7 ,

[X4,Xs] - ~ ; [X S,X7] - C1+a)X, a¢O;
[XSX6] - X2 ; [X6,X7] - C1+a)~ ,
n 72 .• [Xl,Xt] - Xt-l , 3SiS7,

[X 4,X7] - X2 ;
[X6,X7] - ~,
n'
7 .. [Xl,Xt] - Xt-l , 3SiS7,
[X 4,X7] - X2 ;
[X6,X7] - X2 + ~ ,
62 Chapter 2
fI, 7" . [X1,Xt] - xt-l ' 3SIS7,

[X",X 7] - X2 ; [Xs'~] - - X2 ;
[X SX 7] - X2 ; [X 6,X,] - X3 ,
fI,,5 : [X1,Xt] - xt-l ' 3SIS7,

[XS,X,] - X2 ; [X6,X,] - X2 + X3 ;
,
fI, 6 .. [X1,Xt] - Xt-l ' 3SiS7,
[Xs,X,] - X2 , [X6'X,] - - X3 ,
fI, , .
7 • [X1,Xt] - xt-l ' 3SiS7,
[X6'X,] - X2 ,
fI, ,8 .. [Xl,Xt] - Xt-l , 3:S1:S7.
Lie algebra with characteristic sequence C(Ij) - (5,1,1)
,
fI, 9 .• [X1,Xt] - Xt-l ' 4 SIS 7 ,
[X2,X6] - t X3 [X2,X,]- X3 - t X"
[X 5,X6] - tX3 [X5,X,]- X3 - t X4

[X6,X,] - t t
X3 + X3
fl,7 10 : [Xl'~]- ~-1 ' o4SIS7,

[X 4 ,X,] - ~ ; [Xs'~] - -~ ;
[X 5X,] - x,; [X 6,X,] - X" '
fI"l1 : [Xl'~] - ~-1 ' 4 SIS 7 ,

[X",X7] - ~ ;
[X 6,X,] - X, ,
"7}2 : [X},Xt] - xt-l ' 4 S t S 7 ,

[X",X7] - X2 ;
"7 13•a : [X},Xt] - xt-l ' 4 SiS 7 ,

[X2'~] - X3 ; [X 2'X7] - X3 + ~;
[X 5.X 7] - ax3 ; [X 6'X7] - ~ + X2 .
"7 14 : [X},Xt] - xt-l ' 4StS 7 ,

[X2'~] - X3 ; [X2'X7] - X" ;
[X 5.X 7] - X3 ; [X 6'X7] - X2 + X4 •
"7 15 : [X},Xt] - xt-l 4SiS7,

[X2'~] - X3 [X 2 'X7 ] - ~
[X6,X7] - X3
,,/6: [Xl'~]- ~-l' 4SiS7,

[X2,X6] - t X3 [X 2,X7] - -~ X3 + t X4
[X5,X7] - ~ X3 -t X" [X6 X7] - -.1X2 _.1 X3 +.1 X" _.1 X5

, 2 2 4 2
[X5,X6] - -t X3
"7 17 : [X},Xt] - xt-l ' 4 SiS 7 ,
[X2,X6] - t X3
[X5,X6] - - t X3
[X5,X7] - ~ X3+ t ~
"7}8 : [X},'G] - xt-l ' 4 SIS 7 ,

[X 2'X7] - X3 ; [X5'~] - - X3 ;
[X 5X 7] - X3 - X" ; [X 6'X7] - X2 - 2X3 + ~ - X5 •
64 Chapter 2
"'7 19 : [X1'~] - x;'-l ' 4 SiS; 7 ,

[X 2,X,] - X3 ; [X5'~] - X3 ;
[X 5X,] - ~ ; [X 6,X,] - X2 + X5 •
"'7 20 : [X1'~] - x;'-l ' 4 SiS 7 ,

[X 2,X,] - X3 ;
"',z1 : [X1'~] - x;'-l ' 4 S; i S 7 ,

[X2,X7 ] - X3 ; [X5'~] - X3 ;
[X5X 7] - X3 + Xi ; [X 6, X,] - X2 + ~ + X5 •
"',22 : [X1'~] - x;'-l ' 4 SiS; 7 ,

[X5'~] - X3 ;
[X6X,] - X2 + ~ + X5 •
"',23 : [X1'~] - Xt-l ' 4 SiS 7 ,

[X2'~] - X3 ; [X 2,X,] - X4 ;
[X5'~] - X3 ; [X6,X7 ] - X2 + X5 ;
[X 5X,] - X4 ,
"',z4 : [X1'~] - Xt-l ' 4 SiS 7 ,

[X5'~] - X3 ;
[X6,X,] - X2 + X5 '
"',z5: [X1'~]- Xt-l ' 4SiS7

[X 5X,] - X3 ; [X6,X,] - X2 + ~.
"',26 : [X1'~] - x;'-l ' 4 SiS 7 ,

[X6,X7 ] - X2 ,
"'7 2 ' : [X1'~]- Xt-l ' 4SiS7,

[X2'~] - X3 ; [X 2,X,] - X4 ;
[X5'~] - X3 ; [X6,X,] - X4 + X5 ;
n7 28 : [X 1 ,Xt1 - xt-l ' 4:S; i:S; 7 ,

[X2,~1 - X3 ; [X 2,X 71 - X3 + X 4 ;
[Xs,~1 - X3 ; [X 6,X71 - Xs;
[XsX71 - X 4 ,
nl9 : [X 1,Xt1 - xt-l ' 4:S; i:S; 7 ,

[X 2,X 71 - X3 ; [Xs,~1 - X3 ;
[XsX71 - X 4 ; [X 6, X 71 - X s '
n7 30 : [X 1,Xt1 - xt-l ' 4:S; i:S; 7 ,

[X2,~1 - X3 ;
[X s ,X 71 = X4 ; [X 6,X71 - Xs;
[X s ,X 71 - X3 ,
n7 31 : [X 1 ,Xt1- xt-l ' 4StS7,

[X 2,X 71 - X3 ; [X5,~1 - X 3 ;
n7 32 : [X 1 ,Xt1 - xt-l ' 4 SiS 7 ,

[X 2,X7] - X3 ; [Xs,~] - X3 ;
[X s , X 71 - ~; [X6,X7 1 -XS'
n7 33 : [X 1 ,Xt1 - xt-l ' 4 SiS 7 ,

[X 2,XS] = X3 ; [X2'~] - X 4 ;
[X 2,X 71 - X5 ; [X 6,X7 ] -X3'
n7 34 : [X 1 ,Xt1 - xt-l ' 4:S; i :s; 7 ,

[X 2,XS] - X3 ;
[X 2, X 71 - X3+XS '
66 Chapter 2
n7'S : [Xl,Xt] - Xt-l ' 4 SiS 7 ,

[X 2 ,Xs] - X, ;
[X2'X7] - Xs '
n 7,6: [X1,Xt] - Xt-l ' 4 SiS 7 ,

[X2'Xti] - X, ; [X 2 'X7 ] - Xt;
[XS,X7] - X, ; [X6,X7]- Xt.
n 7'7 : [Xl'Xt] - Xt-l , o4SiS7,

[X 2,Xti] - X, ; [X 2'X7]- X,+Xt.
,
n ,8·. [X1,Xt] - Xt-l , 4SiS7,
[X 2,Xti]- X, ; [X 2'X7]- X,+Xt •
n 7'9·. [X1'Xt] - Xt-l , 4!S1!S7,

[X2'X7]- X, ; [X S,X7] - X, ;
[X6,X7]- Xt .
n 740·. [X 1,Xt] - Xt-l , 4SIS7,

[X 2,X,]- X, ; [X6,X7] - X, .
n7 41 : [X 1,Xt] - Xt-l o4SIS7,

[X 2,X,] - X, ,
n42 .,.44
7 .. - .,.'6 e a 1
7 .. - n16 ea 1 n 743 .. - .,.26 ea1
, .
.,.4S. - .,.46 ea1 , .
.,.46 • - ",S6 ea1
Lie algebras whose characteristic sequence is c(t;:J) - (4,1,1,1)

n7 47 : [Xl'~] - xt-l 5SIS7,

[X 2 ,X3 ] - ~ ; [X2'~] - ~ ;
[X 2,x7] - X5 ; [X 6'X7] - ~.
n7 48 : [Xl>~] - xt-l 5SIS7,

[X 2 'X3] - ~ ;
[X 2,x7] - X5 '
n7 49 : [Xl'~] - xt-l 5SIS7,

[X2'~] - ~
[X 3,x7] - ~.
n 750 : [Xl'~] - xt-l 5SiS7,

[X 2 'X7] - ~ ;
n7 51 : [Xl'~] - xt-l 5SiS7,

[X 2 'X3 ] - ~ ;
n7 52 : [Xl'~] - xt-l 5SiS7,

[X 2 'X3] - X4 ,
n 753 •• 6
-n6 ffia l
n 756 .• -nl~ffial
n 759 •• - n2S ffi a2
Lie algebras whose characteristic sequence is c(9) - (4,2,1)
n 760 : [Xl'~] - xt-l i - 3,4,5,7 ,

[X 3'X5] - ~ ;
[X 5,x7] - ~,
68 Chapter 2
117 61 : [Xl'~] - Xt-l , i - 3,4,5,7 ,

[X 3,XS] - ~ , [X 4,xs] - X7 ,
ni2 : [Xl'~] - Xt-l , i - 3,4,5,7 ,

[X 4,XS] - ~; [X 4.X 7] - aX2;
[Xs'~] = X2 ; [X S.X 7] - C1+a)~ a¢O,
n763 : [Xl'~] - Xt-l , i - 3,4,5,7 ,

[X 4,XS] - ~ , [X 4,x7] - X2 ;
[X S.X7] - X3 '
n 764 : [Xl'~] - Xt-l , i - 3,4,5,7 ,

[X 4,XS] - ~ , [X S,x7] - Xti .
n 76S : [Xl'~] - Xt-l , i - 3,4,5,7 ,

[X 4,XS] - ~ , [X S.X 7] - X2 •
n 766·. [X 1,Xt] - Xt-l , i - 3,4,5,7 ,

[X 4,XS] - ~ , [X S.X 7] - X2 + Xti .
n7 67 : [Xl'~] = Xt-l , i - 3,4,5,7 ,

[X 4,XS] - ~ ,
n7 68 : [Xl'~] - Xt-l , i - 3,4,5,7 ,

[X 4,XS] = X2 ; [X S·X 7] - Xti ;
[X6,X 7] - X2 .
n 769·. [Xl'~] - Xt-l , i - 3,4,5,7 ,

[X S,X7] - ~ , [X6,X 7] - x 2 .
n 770 .a : [Xl'~] - Xt-l , i - 3,4,5,7 ,

[X 4,XS] - aX2 . [XS.X6] - X2 ,
n 7 71·• [Xl'~] - xt-l ' 1- 3,4,5,7 ,

[X 4,XS] - X2 ,
n 7 72·• [Xl'~] - xt-l ' 1- 3,4,5,7 ,

[X6,X 7 ] - x2 ;
n 7 73 : [Xl'~] = xt-l' 1 = 3,4,5,7 ,

[X 4,X7] - X2 , [X S,x6] - X2 '
[X 5,x7] - 2X3 + ~;
n 7 74·• [Xl'~] = XI-l , i = 3,4,5,7 ,

[X 4 'X7 ] - X2 , [X S,x7] - X3 + ~ ;
n 77S·• [Xl'~] = xt-l , i - 3,4,5,7 ,

[Xs,~] - X2 , [X S,x7] - X3 + ~;
n 7 76.• [Xl!~] - xt-l , i - 3,4,5,7 ,

[X 4 'X5] - X2 , [Xs,X 7] = ~;
n 7 77·• [Xl!~] - xt-l , i - 3,4,5,7 ,

[X S,X7] - ~ ;
n/ 8 ,a: [Xl'~] - xt-l , 1 = 3,4,5,7 ,

[X 4 'X7 ] = aX2 ' [X S,x6] - X2 ,
[X 5'X7] - C1+a)X3 a;>!: °;
n 7 79·• [Xl'~] = xt-l , 1 - 3,4,5,7 ,
[X 4 ,XS] = X2 , [X 4 ,x7] - X2 ,
[Xs'~] - X2 , [X S'X 7] - 2X3 ;
70 Chapter 2
n. 780·. [X1'~] - xt-l , 1- 3,4,5,7 ,

[X 4,X7] - X2 , [X 5,X 7] - X3 ;
n. 781.. [X1'~] - xt-l , 1- 3,4,5,7 ,

[X 4,X7] - X2 , [X5'X 7] - X2 ;
n. 782·. [X1'~] - xt-l , 1 - 3,4,5,7 ,

[X 4,X5] a X2 ,
n. 783·. [X1'~] - xt-l , i - 3,4,5,7 ,

[X5,X7] - X2 ;
n. 784.. [X1,~1- xt-l , 1- 3,4,5,7 ;
Lie algebras whose characteristic is C(9) - (3,3,1)
n. 785·. [X1'~] - xt-l , i - 3,4,6,7 ,

[X 4,X7] - X2 ;
n. 786.. [X 1,'G] - xt-l , i - 3,4,6,7 ,

[X 4,X7] - X2 , [X 3,x4] - X5 ;
n. 787·. [X1,~1- "t-l , 1- 3,4,6,7 ;
n. 788·. [Xl'~] - "t-l , 1- 3,4,6,7 ,

[X 3'X4] - X5 ;
n. 789·. [Xl'~] - "t-l 1- 3,4,6,7 ,

[X 3'X4] - X5 ' [X 6,x7] - ~;
n. 790.. [Xl'~] - "t-l 1- 3,4,6,7 ,

[X 3,X7] - X5 , [X",x7] - ~;
n 791.. [Xl'~] - Xt-l 1- 3,4,6,7 ,

[X3'~] - X2 + X5 , [X 3,x7] - X5 '
[X",X7] - ~, [X 6,x7] - X5 ;
n 792.. [Xl'~] - Xt-l 1- 3,4,6,7 ,

[X3'~] - X2 + X5 ' [X 3,X 7] - X5 '
[X",X7] - ~;
n 793·. [Xl'~] - Xt-l 1 - 3,4,6,7 ,

[X3'~] - X2 , [X6,x7] - X5 ;
n 79"·. [Xl'~] - Xt-l i - 3,4,6,7 ,

[X3'~] - X2 , [X6,X 7] - X2 + X5 ;
n 795·. [Xl'~] - Xj-l 1- 3,4,6,7 ,

[X",~] - X2 , [X",x7] - X3 + X5 '
[~,X7] - X5;
n 796.. [Xl'~] - Xt-l 1 = 3,4,6,7 ,

[X3'~] - X2 , [X6,x7] - X2 ;
n 797·. [Xl'~] - Xt-l 1- 3,4,6,7 ,

[X3'~] - X2 , [X 3'X 7] - X5 '
[~,X6] - 'S;
n 798.. [Xl'~] - Xj-l 1- 3,4,6,7 ,

[X3'~] - X5 ; [X 3'X 7] - X2 '
[~,X6] - X2 ;
72 Chapter 2
n 799·• [Xl'Xt] - xt-1 1- 3.4.6.7 •

[X3'Xt] - Xs • [X4,x7] - ~.
[~,x7] - XS;
n 7100 : [Xl'Xt] - xt-1 1- 3.4.6.7 •

[X3, X7] - X2 • [Xt,x6] - X2 ;
n7 101 : [X1.Xt] - xt-1 1- 3.4,6.7 •

[X 3.Xt] - X2 ;
Lie algebras whose characteristic sequence is c(9) - (3,2,1,1)
n 7102 : [X1.Xt] - xt-1 1- 4,6.7 •

[X 4.X7] - X2 ;
n 7103 : [X1.Xt] - xt-1 1- 4,6,7 ,

[X 4.X7] - Xs •
n7 104 : [X1.Xt] - xt-1 i - 4.6,7 ,

[X6,X7] - X2 ;
n 710S : [X1.Xt] - xt-1 1- 4.6,7 •

[X 4.X7] - X2 ,
[X3,X 4] - Xs ;
n 7106 : [Xl'Xt] - xt-1 i - 4,6,7 ,
[X3'Xt] - Xs '
n 7107 : [Xl'Xt] - xt-1 1- 4.6,7 •

[X 2'X7] - X3 ,
n 7108·. [Xl'~]- xt-l i - 4,6,7 ,

[X2'~]- X3 , [X 2,x7]- X5 ;
n7 l09 : [Xl'~]- xt-l i - 4,6,7 ,

[X 2'X7]- X3 [X 2,x4] - X5 ;
n7 110 : [Xl'~]- xt-l i - 4,6,7 ,

[X2'~]- X5 ;
n7 111 : [XlI~]- xt-l i - 4,6,7 ,

[X 2'X7] = X5 ;
n7 112 : [Xl'~] = xt-l i = 4,6,7 ,

[X2'~]- X5 , [X6,X 7] = X5 ;
n 7 113 : [Xl'~] = xt-l i = 4,6,7 ,

[X2'X7] =X5 ' [X 6'X 7] = X5 ;
n7 114 : [Xl'~]- xt-l i = 4,6,7 ,

[X 2'X7] =X5 [X3'X 4]- X5 ;
n7 115 : [Xl'~]- xt-l i = 4,6,7 ,

[X2'~] =~ , [X 2,x3] - X5 ;
n7 116 : [Xl'~] = Xi-l i - 4,6,7 ,

[X2'X7] - X3 ' [X6'X 7]- X3 ;
n 7 117 : [Xl'~] = xt-l i = 4,6,7 ,

[X2'~] =X3 ;
n7 118 : [Xl'~] = xt-l i - 4,6,7 ,

[X2,~]-X3 ' [X 3,x4] = X5 ;
74 Chapter 2
t1.7 11' : [X1,~J - xt-l 1- 4,6,7 ,

[X2,X-.J - X3 ' [X6,x7J - X, ;
t1.7 12O : [X1,~J - xt-l 1- 4,6,7 ,

[X 2,X3J -Xs ;
t1. 7121 : [X1,~J - xt-l 1- 4,6,7 ,

[X 2,X-.J - X3 ; [X4,X 7J - Xs ;
t1. 7122 : [X l' x.J - X.-l 1- 4,6,7 ,
[X2,X-.J - Xs ' [X3,x7J - Xs '

[X4,x7J - Xs ;
t1.7123 : [X1,~J - xt-l 1- 4,6,7 ,

[X 2,X7J -X3 ' [X 4,x7J - Xs ;
t1.7124 : [X1,~J - xt-l 1- ':,6,7 ,

[X3,X-.J -Xs ' [X 4,x7J - X2 ;
t1.712S : [Xl,~J - xt-l 1 - 4,6,7 ,

[X 2,X,J - Xs , [X2,X 4J - X, + ~ ,
[X4,x7J - Xs;
t1. 7126 : [X1,~J - xt-l i - 4,6,7 ,

[X2,X3J - Xs ' [X 2,x4J - X3 + ~;
t1. 7127 : [X1,~J - xt-l 1- 4,6,7 ,

[X2,X-.J - axs , [X2,x7J - Xs '
[X 3,X-.J - Xs , [X4,x7J - X2 a~O;
t1. 7128 : [XlI~J - xt-l 1- 4,6,7 ,

[X2,X-.J - Xs ' [X3,x4J - Xs '
"'7129 : - ",Ii eal _130 .

'''7 . - '" 146 ""al
"" "'l~1 : - ",1~ eal
"'l32: - ",2~ eal _ 133 .
'''7 . - ",1~ eal ",l34 : - ",1~ eal
",l3S : - ",1 " e ",13 _136.
'''7 . -",li ea l
Lie algebras whose characteristic sequence is c(lj) - (3,1,1,1,1)
"'7 137: [X 1,'G1 - xt-l , 1- 6,7 ,

[X2,X,1 - Xs ' [X4,x71 - Xs ;
",138
7 .. - ",3 S ea2 ",139
7
.•
- "'f e a
",140 .
7 • - ",t ea2
",141 .
7 . - "'$ e a2
Lie algebras whose characteristic sequence is c(lj) - (2,2,2,1)
"'7 142 : [Xl,~1 - X.-l , 1-3,5,7,

[X3, Xs 1 - X4 ' [Xs ,x71 - ~;
"'7 143 : [Xl'~1 - X.-l , 1- 3, 5, 7 ,

[X 3,Xs1 - ~ , [X s ,x71 - ~;
"'7 144 : [Xl,~1 - X.-l , 1- 3,5,7 ,

[XS,x71 - ~;
"'7 14S : [Xl'~1 - X.-l , 1- 3,5,7 ;
Lie algebras whose characteristic sequence is c(lj) - (2,2,1,1,1)
"'7 146 : [Xl'~1 - X.-l' 1- 5, 7,

[X2,~1 - X4 '
76 Chapter 2
n7 147 : [X1,Xt] - Xt-1 ' i - 5, 7 ,

[X 2 ,X3] - X4 ;
ni48 - nl 1 $ a n7150 - n24

6 ""
<J7
a1
nf51 - n? $a2
Lie algebras whose characteristic sequence is c(9) - (2,1,1,1,1,1)
n/52 : [Xl'X 7] ~ X6 '

[X4'~] - X6
n153 n154
7 7
Lie algebras whose characteristic sequence is c(9) - (1,1,1,1,1,1,1)
v'3. Other classifications
In the next chapter, we shall present other classifications. In particular we shall de-
scribe all the filiform Lie algebras of rank nonzero, i.e. the uncharacteristically
nilpotent tie algebras (see Chapter 7).
CHAPTER 3
COHOMOLOGY OF LIE ALGEBRAS
I. BASIC NOTIONS
1.1. Ij-modules and representations
In the first chapter, we glanced at the notions of representations of Lie algebras and
the corresponding notions of 9-modules. Here we will study all those notions which
can be utilized for the theories of cohomology. First, we recall all the definitions.
Let 9 be a Lie algebra on a field K. An n-dimensional vector space V over the same
field K is called a 9-module if a bilinear map <p : 9 x V ~ V satisfying the condition
q(lx, yl, v) - q(x,q(y, v)) - q(y, q(x,v)), 'V x, y E 9 and v E V,
is given. We write gv instead of q(g,v), if no disarray is possible. In other words, a
9-module V corresponds to a representation of 9 on V, i.e. a homomorphism of Lie
algebras ell : 9 ~ gl (V) defined by
cIICg) (v) - q(g, v) - gv .
For any IJ-modules V and W, the vector spaces V $ W, V ® W and Hom(V, W) can
be considered as 9-modules by putting:
g(v+ w) - gv+ gw,
78 ChapterS
g Cv ® w) - gv ® w + v ® gw ,
Cgf) Cv) - gCf(v)) - fCgv),
where ve V,weW,ge I}andfe HomCV, W).
The structure of al}-module on a tensor product Induces one on the exterior products
AJlV (the quotient of the tensor product ®PV - V ® ... ® V by the subspace generated
by all the v ® v). Note that AV - I A'V is a fini~mensional (because V has a finite
dimension) graded algebra over K with AOV - K, AIV - V. It is clear that Anv m
one-dimensional(n - dim V) andAkV - {OJ for k > n. If W is a subspace of V, then
APW can be identified canonically with a subspace of APV.
Examples
(t) Let V be a vector space over K. It can be considered as a I}-module for all Ue
algebras 9 by putting gv - 0 for all g In I} and v In V. This 9-module Is called trivial.
(ll) Let V be a IJ-module. An element v In V Is called I}-inw.dantlf gv - 0 for all gin I}.
Let be V9 the set of all9-lnvariant elements. Then V9 is al}-submodule of V, which m
trivial. This submodule Is maximal within the trivial 9-submodules of V.
OIl) Every Ue algebra I} can be considered as a I}-module by putting gl82 - [g1' 82] for
all gl and 82 In I}. The corresponding representation of I} on I} Is nothing but the
adjoint representation 4l(g1) (g2) - ad gl (g2). This example plays an Important role.
(Iv) We consider I} - glCn, K). The structure of a one-dimensional module on a vector

space EA (A. e K) Is given by fv - - A. TrC£) v , v e EA, where TrC£) denotes the trace of
the endomorphism f.
1.2. The space of cochains
Let 9 be a Ue algebra and V a I}-module. If p ~ 1, a p-dimensJonaJ cochain of I} with

values In V, Is a p-Ilnear alternating mapping of 9 P Into V. If P - 0, one defines a
zero-dlmenslonalcochaln of I} as a constant function from I} to V. We denote CP(I}, V)
Cohomology of Lie Algebras 79
as the space of the p-cochain of I}. Then, CPCI}, V) - Hom CAPIJ, V) for p ~ 1,
COCI}, V) - V, and forp < 0, we putCPCI}, V) - {OJ.
Let C·CI}, V) be the direct sum of CP(I}, V). The elements of C"CI}, V) are called cochains
of I} with values in V ; those of CPCI}, V) are said to be of degree p.
We provide the space CPCI}, V) of a I}-module structure :
(x4») (Xl, ... , Xp) - x4» (Xl, ... , Xp) - 1: 4»(Xl ,... , XI-l, [X, XI] , ... , Xp)
lSiSp
for aU X, Xl ,... , Xp in I} and 4» in CPCI}, V). This structure of al}-module can be expanded
to the space C"CI}, V). The corresponding representation wiD be denoted by S.
For all y in ':J, we denote iCy) as the endomorphism of C·CI}, V) which maps each
subspace CPCI}, V) into cP-l(I}, V) and given by
(iCy) 4») (Xl'''.' Xp-l) - «y ,Xl'···' Xp-l) .
Lemma 1. The representation e of 9 on C"C9, V) verifies :

Sex) 0 iCy) - iCy) 0 SCx) - i[x , yJ for all X and Y In I}.
The proof is a direct consequence of the definition of S.
1.3. The coboundary operator
Let 9 be a Ue algebra and V a 9-module. We define the endomorphism

d : C"C9, V) ~ C"CI}, V)
by putting
d«x) - x4» for 4» E COCI}, V), X E I}
d4»(xl ,..., Xp+l) - 1: (-If+ l (xs4>>)(Xl ,..., Xs ,..., xp+s) +

19Sp+l
for 4» in CP(I}, V), P ~ 1, and Xl'.'" "P+l in I}, where the symbol A over a letter means
that it is omitted. This endomorphism d maps CPCI}, V) into CP+l(I}, V). Its restriction
to CP(I}, V) wiD be denoted d p.
80 Chapter 3
Theorem 1. The endomorphism d verlfles d 2 - o.
First we prove some identities.
Lemma 2. 9Cx) - ICx) 0 d + do I(x) ,

do9Cx)- 9(x)od forallxin9.
Proof. The first Identity is nothing but the deflniUons of the endomorphisms d and i.
We prove the last by an induction on p. Let «I> be a O-cochain. We have
(9Cx) d«l» (y) - x(d«l» (y)- (d«l»([x,y» =x(y«l» -[x, y] «I>=y(x«l» - (d9(x) «1» (y)
We suppose now that the relation is true until the Indice p-l, i.e. dC9CX)«I» - 9Cx) d«l>
for all «I> In CP-1(9, V). One considers an element «I> of CP(9, V). We have
l(y)(d9Cx) - 9CX)d)«I>
- (9Cy) - diCy) )9(x).«I> + (i{[X,y] - 9Cx) I(y))}d«l>
- 9Cy)9(x)«I> + d(l([x,yJ) - 9Cx)lCy)) + (9([x,y]) - di([x,y])«I> - 9(x)(9Cy) - diCy)))«I>

- (9Cx)d - d9Cx)) (Ky)«I»
-0 forall yin 9.
So (d9Cx) - 9Cx)d)«I> - 0 and the ldenUty Is verified.
Proof of the Theorem. From lemmas 1 and 2, we have:

I(x)ddcll- (9Cx) - diCx) )d«l> = d9(x)«I> -d(e(x) -di(x))«I> = ddCl(x)«I».
o
One deduces: dd«l> - 0 'if «I> e C (9,V) and, by inducUon,
dd«l>- 0 'if«l>e CP(9,V).
So dd-o.
1.4. The cohomology space
The restriction ~ of the coboundary operator d to the space of the p-cochaine

CP(9, V) of the Ue algebra9 with values In V, has a kernel denoted ZP(9, V) whose
elements are called cocycJes of degree p (or p-cocycJes) with values in V.
The elements of the image BP(9, V) of ~-1 are called coboundarles of degree p (or p-
coboundarles) with values In V. From Theorem 1, we have BP(9,V) c ZP(9, V). The
quotient space
HP(9, V) - ZP(9, V) / BP(9, V)
Is called the cohomology space of 9 of degree p with values In V. If I - 0, we agree to

take BO(9, V) - 0, since there are no (-l)-cochalns. In this case, we have HO(9, V) -
ZO(9, V)' This space can be Identified with the subspace V9 of Invariant elements of V.
The direct sum of the HP(9, V) for p In it Is denoted by H"C9, V).
1.5. Exact sequence
Let a be a homomorphism of the 9-module V Into the 9-module W. Using this, we

can construct the homomorphism
<Xp: CP(9, V) ~ CP(9, W)
deflnedby
<Xp( «1» - a 0 «I>
for each p. Then a can be extended to a K-Ilnear mapping
a·: C·(9, V) ~C·(9, W).
It is easy to see that :
and this implies that

a·od-doa·.
This last relation pennits us to define a homomorphism of cohomology spaces
a- H·(9, V) ~ W(9, W) .
It is called a homomorphism associated to a .
Consider an exact sequence of 9-modules :
By exactness, we means that 1m a - Ker~, Ker a - ° and 1m ~- V3 •

82 Chapter 3
From this sequence, we can define another exact sequence of the cochain spaces :
a' 13'
o ~ C"(IJ, V1) ~ C"(IJ, V2) ~ C"(IJ, V3) ~ 0 .
In fact, let cI> be such that u"C cI» - 0, i.e. U 0 cI> - 0 . Suppose that cI> Is a p-cochain, then
vJ -
U(cI>(V1' ... , v~) - 0 for all v1 ,... , vp. As U is injective, «v1 ,... , 0 for all v 1•... , vp
and cI> - 0 . The exactness of the terms of the sequence could be verified In this way.
Passing to the cohomology spaces, we find the long sequence of homomorphisms
~ ~ So
o ~ HO (IJ,V1) ~Ho (IJ,V 2) ~Ho (IJ,V3) ~H1 (IJ,V1) ~
-131 51
~ H1 (IJ,V 2) ~ H1 (IJ,V3) ~H2 (IJ,V1) ~ ..... .
JVe begin by explaining these notations. The homomorphism al is obtained from

a: H"(IJ,V 1) ~ H"(IJ,V 2) by the restriction of HI(IJ,V 1). It is analogous for ~I. Now let q;
be In HP(g,V 3) and <pin ZP(g,V 3) be a representative of <p. As ~ is a surjective homomor-
phism, we can choose an element 'I' in CP(IJ,V 2) such that ~p('I') - <p. But d'l' is in
Zp+1(IJ,V2) . As d o~" - ~"o d, then d (~p'I') - ~p+1(d'l') - d<p - 0 and d'l' is In
Ker ~P+1 - 1m Up+1 . So there is a cochain pin Cp+1CIJ, V1) such that <X.p+1Cp) - d'l'. But
dCup+1Cr)) - up+ldp) - 0 then dp - o. Hence, p is in Zp+1CIJ,V 1) and the mapping Op+1 is
lefined by op(q;) - p, where p Is the class of p in HP+1(IJ,V 1). This Is well defined be
cause the class p of p doesn It depend of the choice of the representative <p In the class
q;. In fact, we suppose that <p + d v Is another representative of q;. The cochain '1'1 In
In CP(IJ,V2), whose Image by ~P Is <p + dv. can be written as '1'1 - 'I' + ell; . Then
d'l'l - d'l'and this defines the same cochain p in Cp+1CIJ,V1).
Theorem 2. Let
be an exact sequence of IJ-modules. Then the long sequence of homomorphisms

IXo
-IIa
o ~ HO(9,V 1) ~ HO(9,v2) ~HO(9,V,) ~
110 at -Ih lit

~Hl(9,Vl) ~ H 1(9,V2) ~Hl(9,V,) ~H2(9,Vl) ~ ...
Is exact
Proof. It Is sufficient to prove that ao Is Injecdve, 1m Sp - Ker ap+l and 1m /3 p - Ker lip .
By definldon,
HOC9,V1) - ZOC9,V 1) - {VE VI: 9.Vl - 0 , 'V 9 E 9} .
By hypothesis, (X Is Injecdve. Then ao Is also Injecdve.
Let q; be In HP(9,V,) . By the construcdon of lip, we have seq;) - pwith pin Hp+l(9,V 1) ,
and (Xp+l(P) - dw with ~(w) - <p, Win CP(9,V2). The class of cohomology of ap+l(p)
is nonzero and 1m Sp is contained In Ker ap+l .
Now we consider 'if in Ker ap+l and 'II as a representadve of \if. We have (Xp+l(W) - c41
where Ills an element of CP(9, V2). Furthennore, one has :
d(J3pCJ,D) - ~p+ICdJ,D - J3p+1C«P+1CW)) - o.
Then ~p(l1) is In ZP(9,V,). We denote by vthe class of /3p(l1) In HP(9,V,). By the deftni·
don of 8p, we have 8p(V) - 'if then 1m 8p - Ker ~ 1 .
We can prove the second reladon 1m /3 p - Ker lip by using the same arguments, and the
theorem is proved.
Corollary. Let W be a sub-'iJ-module oEv. The canonical homomozphlsms

i 11
W ~ V ~ V/W
deflne the exact long sequence
a;, if. 110 at

O~HO(9,W) ~HO(9,V) ~~HO(9,V/W) ~ Hl(9,W) ~Hl(9,V)
itt lit
~ HI (9,VlW) ~ H2 (9,W) ~ ... ~ .
84 Chapter 3
n. SOME INTERPRETATIONS OF THE SPACE Hi(Ij,V) FOR

i = 0, 1, 2, 3
11.1. The spaces HO(Ij,V)
We have seen that

HO(I},V)-Kerdo-{veV/g.v-o ,'v'gel}}.
This is the space of invariant vectors for I}.
Particular case. Suppose that V - I} is considered as I}-module adjoint. Then, the

space HOCI},I}) is the center of I} :
n.2. The spaces Hl(Ij,V)
11.2.1. V= K
We consider K as a trivial I}-module. The spaces HICI},K) are nothing but the
cohomological spaces for the De Rham cohomology of the left invariant forms on the
connected Ue group G associated to I}.
H~I},K)- Kerd l - {ae I}·:a{x,y] -0 'v'x,ye I}} .

In other words, HlCI},K) - elCI}/ [1},1}1, K).
11.2.2. V = I} considered as the adjoint I}-module
First interpretation. The space Z IeI},I}) Is the space of the derivations of I} and the
space Bl(9,9) is the space of the inner derivations of 9. Then, the space Hl(9,9) can
be interpreted as the space of the "outer" derivations of the tie algebra 9. The
particular case H 1(9,9) - {O} corresponds to the tie algebra 9 whose derivations are
inner. This is the case for semisimple tie algebras and for parabolic subalgebras of
semisimple tie algebras.
Second interpretation. The space H 1(9,9) can be interpreted as the set of the one-
dimensional extensions classes of the Ue algebra 9.
A one-dimensional extension of 9 is an exact sequence
of the homomorphism of tie algebras, where K is viewed as an Abelian tie algebra.

Two such sequences,
and
are eqUivalent if there is a homomorphism <p: ij ~ 91 of tie algebras such that the
diagram
Is commutative.
Let ex be a cocycle in 2 1(9,9). To this cocycle corresponds the one-dirnenslonal
extension of 'J :
112
o ~ 9 ~ ij ~ 9 E9 K ~ K ~ 0,
where i : 9 ~ q = 9 E9 K Is the natural Injection I(x) - (x, 0) and 1t2 Is the projection on K
86 ChapterS
'1C2u!', k) - k. The tie algebra law on 9 E9 K is given by:

[(x,a), (y,b)] - [x,y] '+ ba{x) -aa(y), 0).
The Jacobi identity for this bracket Is equivalent to a being a cocycle.

It is easy to see that with cohomologeous cocycles correspond equivalent one-
dimensional extensions. Indeed, let a '+ de be a cohomologeous cocycle to a, with e in
CO(9,9) -9. This cocycle defines a one-dlmensional extension
o ~9~~-9E9K ~K
and the bracket of 91 Is given by the bracket

[(x,a), (y,b)] - [x,y] '+ b(a,+de) (x)-a{a,+de)(y), 0).
The mapping cp: ij ~ iji defined by cp(x,a) - (x+a9,a) is a homomorphism of Ue algebras

which gives the equivalence between these two extensions.
11.3. The spaces H2(91 V)
11.3.1. The spaces H2(9,K)
We can look upon this space as a set of classes of the cohomology of the 2-left
invariant forms on the connected tie group G associated with 9. Recall that if G is
compact, then we obtain the De Rham cohomology of G. We can also interpret
H2(9,K) as the set of classes of one-dlmensional central extensions of the tie algebra 9.
A one-dlmenslonal central extension of 9 is an exact sequence
0~K~9~9~O
of Ue algebras and their homomorphisms, in which the image of the homomorphism

of K into Ij is contained in the center of 9.
Let a be in Z2(9, K). To a we can associate the extension
I 1t2
o ~K-?9-KE99 -?9~0
Cohomology of Us Algebras 87
where the bracket of ils given by

[(a, x). (b. y)] - (a [x. y] • [x. y])
We easily verify that the Jacobl's ldenddes for this bracket are equivalent to a be a
cocycle. and two cohomologeous cocycles correspond to equivalent extensions. The
trivial extension. i.e. the sequence which splits. corresponds to the zero of the space
H2(9.K)·
Remark. Let CCOI'.... COn) be a basis of 9- and ~ - 1: CIJt CIlj 'Ilk the structural
1\
equations of 9 (the definldon of a Ue algebra. by Its constants of structure or by Its

structural equations wiD be developed In Chapter 5). If 9 is In Z2(9. 10. then 9 Is a 2-
exterior form on 9 sadsfytng d9 - oed is the coboundary operator for the
cohomology H-C9.K); it corresponds to the differendal operator of differendal Cleft or
not) forms on the Ue group G associated to 9).
We put 9 - dmn+l. We define aUealgebra.91 of dimension dim 9 + 1. whose structural
'Ilk .1-1•...,0. and dCOn+ 1 - o.
equations are dCl\ - 1: CIJt CIlj 1\
This extension corresponds to the preceding extension.
III.3.2. The spaces H2(9,1j)
The space Z2(9.9) can be Interpreted as the space of the infinitesimal deformadons of
the Ue algebra 9. and two cohomologeous cocycles define their equivalenttnfiniteslmal
deformadons. This permits us to Interpret the space H2(9.9) as the set of Infinitesimal
deformadon classes of 9. We will devote one chapter for the general study of
deformadons. However. In order to understand better this interpretadon of H2(9.9).
we give here some ways of approaching the topic.
Let 9 be a Ue algebra whose bracket Is denoted by J1. A (fonnal) defonnation of 9 Is a

one-parameter family of bilinear alternated mappings on the vector space underlying
9 of the form :
J1t (x.y) - J1(x.y) + t<Pl (x.y) + t2<P2 (x.y) + ...
such that C0 <Ilt Is a bilinear altemated mapping on 9.
ClO the Jacobi condidons for J1t are fulfllied (formally).
88 ChapterS
This last condition means that all the coefficients of the fonnal serie
J.lt (J.lt (x,y),z) + J.ll (J.ll (y,z),x) + J.lt (J.lt (z.x),y)
are zero.
If we develop this fonnal identity, then the first relations are :
Cal) J.l(J.l(x,y),Z) + J.l(J.l(y,z),x) + J.l(J.l(z.x),y) - 0
It is the jacobi's identity for J.l.
Ca2) J.l(<P1 (x,y),z) + J.l(<Pl (y,z),x) + J.l(<P1 (z.x),y) + <PI (J.l(x,y),z) +
+ <P1 (J.l(y,z),x) + <PI (J.l(z,x),y) =d<Pl - o.
This relation is nothing but d<P1 - 0 and <PI is in 22(9,9).
So the first tenn <PI of the defonnation J.lt of J.l , called the infinitesimal defonnation
associated to J.lt Is an element of 22(9,9),
Two defonnations J.lt and Yt of J.l are called equivalenuf there is a family (ft) of
nondegenerate endomorphisms of the vector space 9 of the fonn
f t (x) - X + L
tl9 (x),
1>1
where 91 is an endomorphism of the vector space 9, such that
Yt (x,y) - f t- 1 (J.lt (f t (x),f. (y»), for all x and y in 9 .
If we consider two equivalentdefonnations J.lt and 1t of J.l, then their infinitesimal

defonnations <PI and '1'1 are cohomologeous cocycles, i.e. verify <PI - '1'1 E B2(9,9)·
Another interpretation. The aim of the defonnation theory is to give an interesting

tool for a local and topolOgical study of the variety of Ue algebras laws (see chapter 5).
But one big disadvantage a consequence of this approach, lives in the nature of the
defonnation : a defonnation is not an element of this variety by a parametrized curve
in this space throught out by the given point J.l corresponding to a given Ue algebra 9.
We can process this directly by considering only "infinitesimal" defonnations. For
don't become dumbfounded by the infinitesimal defonnation associated with a fonnal
defonnation, we call these little defonnations perturbations. It is more gratifying to
define this notion within the framework of Nonstandard Analysis. So we have the
notion of a law infinitely close to a given law J.l. Later, we will see that every
perturbation yof J.l can be written as
y(x,y) = Jl(x,y) + EICPI (X,y) + EIE2CP2(X,y) + ... + EIE2 ... EkCPk(X,y),

where k is less than (03-0 2)/2 (n - dim I}) and the bilinear alternated mappings <Pi are
linearly independent.
As y verifies the Jacobi identity Cit is a tie algebra law), this implies that CPI is in Z2cJl,J.t).
We again find the interpretation of Z2(Jl,J.t) : it is the space of the first terms called the
differential of perturbation y.
11.4. The spaces H3(9,9)
This space coincides with the space of the obstruction of the deformations of I}. We
consider an element cP of the space H 2(1},1}). Are the infinitesimal deformations of this
class derived with respect to some deformations of I}? It is not the general case and
we shall indicate some necessary conditions related to H3(1} ,I}) for this to take place.
Let Ilt be a deformation of the tie algebra law Jl of I} and CPl its infinitesimal
deformation.
By writing Ilt - Jl + t CPl + t2 'P2 + ... , and using jacobi's conditions for Ilt, we obtain
0- Jl 0 Jl + tJl 0 CPI + t2 (} CPI 0 CPI + Jl 0 CP2) + ... + tn (L

I+J-n
CPi 0 CPj + Jl 0 CPn) ,
where a 0 ~ is the 3-linear alternated mapping given by

(ao~) (x,y,z) - a(~(x,y), z) + a(~(y,z), x) + a(~(z,x), y) +
+ ~(a(x,y), z) + ~(a(y,z), x) + ~(a(z,x), y)
In particular, we have Jl 0 cP - dcp. The conditions related with CPl are

dcpl = 0
k+I=I+2i
CPlOCP2i+l+ L CPk 0 CPI + 1. CPI 0 CPI = - dcp21+2
k+I=I+2i 2
k"l
90 Chapter 3
The first relation shows that <PI is in Z2(9.cp. The second relation concerns the
3-cochains <PI 0 <PI . We can see that if <PI is in Z2(9.9). then <PI 0 <PI is In Z3(9.9). and the
cohomologlcal class of <PI 0 <Pl corresponds to the classical product on the c1asses of
cohomology. We denote this by [<PI ]2. So this second relation couldbe interpreted E
a necessary condition for <Pl E Z2(9.9) to be a infinitesimal deformation of a
deformation J.lt of !.liS that [<PI]2 - 0 .
A:J we can find examples showtng that this condition is not sufficient, we can read the
thUd relation as
<Pl 0 cpz - -d <Il3 •
This shows that the 3-cochaiil <Pl cpz is a 3-coboundary (it is a 3-cocycle because cpz
0
verifies d cpz - <PI 0 <PI ). Thus. the cohomologlcal class of <PI 0 cpz vanishes in H3(9.9).
From the construction of cpz. this class doesn't depend of the choice of cpz. It is the
MMassey cube" of the class of <PI . We can denote It by [<Pl ]3 • Continuing the argument,
we see that for the existence of the deformation it is necessary that all the Massey
products of <Pl vanish; all of these powers are in the space H3(9.9). (Note that the
ith..power product is defined when the kth..power products are zero for k < I).
Conclusions. The conditions related to <Pl show that all Massey powers are zero. It is
clear that all these conditions are satisfied when H3(9.9) - 0 . We can conclude that if
H3(9.9) - O. every cocyc1e <Pl in H2(9.9) is the infinitesimal deformation of a
deformation (formal) of the law !.l of 9.
An element <P in <pl is called integrable if there is a deformation J.lt of !.l such that
J.lt-!.l+t<pl+t2 <p2+ ...
Remarks.
1. In fact, It is sufficient that a subspace of H3(9.9) vanishes for the integrability of any
cocycle <Pl In Z2(9.9). This subspace H3~9.1}) has been studied by Rauch [Raul.
2. We can also write the necessary and sufficient conditions for the integrability of an
Cohomology of Ue Algebras 91
element CP1 of Z2(9,9) In perturbation theory. Note that, In this case, we can entirely
solve the conditions system on CPl. As we can prove that the conditions of integrability
for perturbation theory and deformation theory are the same, we find another
approach of the space H3~9,9) (see Chapter 5).
III. COHOMOLOGY OF FILTERED AND GRADED LIE ALGEBRAS
111.1. Graded Lie algebras. Filtered Lie algebras
DeftnlUon 1. A Ue algebra 9 Is called Z-graded If It admIts a decomposltlon

9 - ES1ez91
where the subspace 91 satlshes [9109,] c91+j foraJ1 i and J In Z.
Note that 90 is a subalgebra. of 9.
DeftnlUon 2. A Ue algebra 9 Is called filrered If It satlshes :
m 9 -U1ez Sh where Sl/sasubspaceof 9 foraJ11
(U) [SI,sj] cSt+j , foraJ1 i and JIn Z

(HI) SI cSj , when I> J (In this case ,the fllll'atlon Is called ascenclJng).
A filtration 9 -UleZ SI of 9 is called finlte if there are n 1 and n2 In Z such that
Sl - 9 if I s: n1 ,
Sl - {oJ if I ~ n2 .
Every Z-gradua1ion on 9 defines a filtration by putting

92 ChapterS
This filtration is associated to the graduation 9 = EEl 91 . Conversely, if 9 is filtered, i.e.

9 - UleZiSh we construct a graduated lie algebra, denoted as gr 9, whose underlying
subspace is isomorphic to 9, by putting
gr 9 - E9 Si with Si- ~ .
leZ Sl+l
The structure of a Ue algebra on gr 9 is given by

[x, Y]- [x, y] with xe Si and Ve Sj ,
x and y being representative of x and yin SI and SJ .
Remark. The filtration associated to the graduation of gr 9 is given by the sequence
{ Fk(gr 9) - $ Si} .
I~
It is isomorphic to the filtration of the initial lie algebra 9 given by the sequence Sk .
111.2. Graded and filtered 9-modules
Let 9 = $leZi 9 J be a graduated lie algebra and V a 9-module.
We say that V is a graded trmodule if it is a direct sum of subspaces VI : V - E9 le ll V I

suchthat 91.VJcVl+j forall I,j in i£ .
Now we consider a filtered lie algebra 9 - UJd SI , with the filtration descending.
A 9-module V is called a filtered 9-module if

CO V Is provided with a filtration
... Vj => Vj+l => ••• j e i£ with V - U VI

lell
(the filtration is descending).
This filtration is called finite if there are m 1 and m2 in Z:: such that
VI - V if i:S: m 1 ,
VI - {O} if i ~ m2 •
The link between filtration and graduation is established in the same way as for Ue
algebras :
(1) Let V - EBlell V I a graded IJ-module. We put Fk(V) - EBI~ VI .
Then V - Ukel£ Fk(V) is a filtered IJ-module associated to the graded IJ-module V,

V - EBld Vb IJ being considered as the filtered lie algebra associated to the given graded
lie algebra.
(2) Let V - Utel£ VI be a filtered IJ-module. We put
gr V - EB VJ with VJ - VJ/ VJ+l .

Jell
We consider the graded Ue algebra gr 9 - Ei1 le l! Si. We provide gr V with a structure
of a graded gr IJ-module :
Let ij E S; and vE VJ ; we put ij.v = IJv (this is correcdy defined).
In fact, the filtration associated to this graded gr IJ-module is isomorphic to the

filtration of the IJ-module V.
111.3. Graduation and filtration of the cohomology spaces
Let IJ - EBleiB IJI be a graded Ue algebra and V = <:9 1e11: VI a graded IJ-module.
We can graduate the space of the j-cochains by putting CJ (IJ,v) - <:9 Cl(IJ,V), where
kell
c~ (IJ,V) - {c E CJ(IJ,V) : C(3.tl, ... ,a~) E Vh+ ...+lj+k, al E IJI} .
The coboundary operator d is in respect to the graduation, i.e. :
d(cl(IJ,v») c cr 1 (IJ,V) .
This permits us to graduate the cocyde and cobord. spaces :
94 Chapter 3
ZJ(9,V) - E9 Zl(9,V),
keZ
sJ(9,V) = E9 Bl(9,V),
keZ
and then the cohomology spaces :

HJ(9,V) = E9 Hl(9,V), where Hl(9,V) - Zl(9,V) / Bl(9,V) .
keZ
Remuk. The equalities are understood in an enlarged sense for avoid the use of
isomorphisms.
Now we consider the filtrations.

Let 9 - UleZ SI be a filtered. Ue algebra and V - UleZ VI a filtered 9-module.
We filter the space of cochains C~9,V) by putting:
d(9,V) - U Fk d(9,V) ,
keZ
where
As d(Fk c'(9,V» c Fk CJ+l(9, V), we can provide the spaces of the cocycles of the coboun-
daries and of cohomology by the filtration:
ZJ(9,V) - U Fk zJ(9,V) ,
keZ
BJ(9,V) - U Fk sJ{Ij,V) ,
keZ
H~Ij,V) = E9 Fk H1(Ij,V) wi1h Fk H~Ij,V) - FkzJ / FksJ .

keZ
Note: The correspondence between graduation and filtration that we have considered
for the 9-modules and Ue algebras, passes on the spaces of cocycles, coboundaries,
and cohomology.
N. SPECTRAL SEQUENCES
N.I. Definition
In this section we take a quick glance at the theory of spectral sequences.
Definition 3. A spectral sequence is a sequence (Er,dr)rdf, where Er is a module (on a

given ring) and <1r an endomorphism of Er,such that
(1) Er - $ Ef,q,
p,qeZ
dy 0 <1r - 0,
dr(Ef,q) c ErrA-r+l,
PA Ker dP,q
(4) E - r with dfA is the restriction of d r to Ef,q .
r+l 1m d p-q,q+r-l
r
This ends the theory.
Remark. We often add the condition EtA - (OJ if one of the two indices Cor the two) is
negative. Such a spectral sequence is denoted as being of the first quadrant
N.2. Some properties
Let m be an integer. We consider a spectral sequence (Er,dy)reN such that El,q - {O},
if we have p < m or q < m (the case m - 0 corresponds to the spectral sequence of the
first quadrant).
The condition (4) of the definition of aspectral means that Er+l is the cohomology of
the complex (Er,dr). Then Er+ IP,q is a subquotient of EtA. We show that EtA is also a
subquotient of EoP,q. For this we construct the sequence :
o c Bci,q c sf,q c ... c S~A c Z:,q c ... c zf,q c zci,q c Eci,q
such that:
96 Chapter 3
E p,q =ZrP,q =zf,q

r+1 -sp,q --Sp,q
r r
We put Zt,q - Ker tY',q,

st,q - 1m d,.p-q,q+r-l,
rn p,q
Tr : Z rp,q -+ S
Z f,q
p,q =Er+p,q1 the canonical prol--tion
J"'.... ,
r
zf,q - (cpC,q)-1 (z f,q),
Bf,q - (CP6,qr 1 (sr,q) .

Hence, we already have the following sequence :
To continue the construction, one puts :

-p,q
li\P,q:
T1 zP,q 2 =~
1 -+EP,q -p,q the canonical prolAction
J""
S1
after having identified the two last spaces
zf,q - ("<Pf,q)-l (zf,q),
Bf,q - ("<pf,q)-l (sf,q),

and iterating this process. To finish, we denote :
Z_- nZr,q,
r
As Et,q - 0, when p < m or q < m, we have:

-SpM
p,q - -Sp-M+1·"
p,q - - -S _,
p,q
-p,q - -p,q - - -p,q

Zq-M+1 Zq-M+2 ... Z_,
then
E p,q = Zr,q = E p,q
- -Bf,q - r+1
when r ~ max (q+ I-M,p-M) .
N.3. Exact sequence associated to a filtered complex
We have previously defined the notion of a filtered 9-module. Here we use the more
general notion of a filtered module when the action of 9 doesn't occur.
DeftnlUon 4. A filtered module is a module V provided by a family of submodules

(FkVlkeZ such that FsV::) FkV if k> s Cdescending filtration).
In this work, we consider only the filtrations satisfying:
FrV - V for r < m , mE Zi: .
DeftnlUon 5. A complex of modules

dO d1 d2
A:O~Ao~Al~A2~ ...
is called filtered if all the modules AP are filtered and if the coboundary operators
respect these filtrations, that is :
{ forall p,
}
We can extract from such complex of modules the follOwing complexes:
dO d1 d2
Aq:O~AOq~Alq~A2q~ ... ,
and Aq+ 1 is a subcomplex of Aq and Am - A.
We note H"CA) as the cohomology of the complex A, and H"CAq) as the cohomology
of the complex Aq :
HPCA) = ZP(A)
BPCA)
98 Chapter 3
To unburden the notadons, we write :

ZP(resp. BP) the space ZPCA) (resp. BPCA)),
and
Z8(resp. B8) the space ZPCAp> (resp. BPCA~)
We put
This Is the canonical Image of HP(A~ In HP(A). We deduce a filtradon of HPCA).
Now we can construct a spectral sequence associated to this filtradon.

Let r be In 2. For each p and q In 2 satisfying p ~ m and p + q ~ 0, we put:
Zl',q - Ar n (dP+<lyl (AJ.;<t+"I),
P n dp+q-l (A q-r·
Brp,q - AP+q P+q-l)
We note that Zrp,q contains BP,q

r-l and Zr-l
p+l,q-l.
Let be Ef,q - Zf,q / B p,q + Z p+l,q-l .

r-l r-l
As d,p+q maps Zp,qln Z p+r,q-r+l and BP,q + Z p+r,q-l ln B P+r,q-r+l + Zp+r+l,q-r then we
r r r-l r-l r-l r-l'
reclaim, by passing through a quodent, a morphism
df,q : Ef,q ~ ErP+r,q-r+ 1 •
In pardcular :
Z p+l,q-l - A p+q
-1 p+l '
BP,q - A p+q
-1 p+l '
Z p,q - A p+l
o P'
then E6,q - App+q / ~~
The sequence is weD constructed.

V. THE HOCHSCHILD-SERRE SPECTRAL SEQUENCE
Let I} be a tie algebra, h. a subalgebra of g, and V a IJ-module. We put

FPCJ>+<I(I},V) - {c E CP+<l(I},V) / c(aL... ,ap+q) - 0 for aL· .. ,aq+l E hI
We have cr(l},v) - F<tr(I},V):J ... :J FrCr(I},V) :J Fr+lcr(I},V) - 0 . Then
dFPCJ>+<I(I},V) c FPCp+q+~I},V) .
So we obtain a filtration of the complex C"(I},y). We denote by (E ~,q,dr,q) the spectral

sequence corresponding to this filtration Cthe study of these spectral sequences has
been developed in the previous section).
Then
iss isomorphic to
and we have
E p,O - HP(I} k'Y)

2 '"
In this notation, H·CI},h;V) denotes the cohomology of I} modulo h. with values in V (or
relative cohomology). The corresponding cochains verify:
c(al ,... , aq) - 0 if al E h. and dc(al ,... , aq+l) - 0 if al E h. .
It is the same as saying
cq(I},h.;V) - Hom k(A\~)' Y) .
The spectral sequence constructed in this way is called the Hochschild-Serre sequence
[H-S).
100 Chapter 3
Applications
1. The Hochschild-Serre exact sequence
Let 'L be an ideal of I} and v a I}-module. We suppose that

Hn('L,V) - (O) for all n satisfying 0 < n < m and m ~ 1 .
We consider the set HmC'L,V)1J of I}-invarlant elements of Hmc'L,V). Each element of
HmC'L,V)1J has a representative which is the restriction to 'L of a cochatn cr in CmCI},V)
such that dcr belongs to Hm+l(~ , v'L ) (where v'L - {v E V: a.v - 0 Va E 'L}) .
This element doesn't depend upon the choice of HmCI},V)IJ. We denote tm+l as the
homomorphism
tm+l: Hn\'L,V)9 --+ Hm+i~' V'L)
that we have constructed. Therefore, we have the exact sequence :
o --+ Hm (.'l,
'L
v'l)
--+ H m (I},V) --+ Hm ('L,V)9 --+ Hm+l (.'l,V'L) --+ H m+1 (I},V)
1m rm tm+l 'L 1m+l
In this sequence, the homomorphism 1m is defined by looking upon the cochains of
2.
'L
in v'l as cochains of I} in V.
As for the homomorphism rm ' this is a homomorphism of restriction.
We suppose now that Hn(l,v) - 0 for 2 ~n ~m and m> 1. In this case, we have
the following exact sequence
Here the homomorphism r'm is the restriction to the first argument of the chosen
cocycle representing the given cohomological class for the ideal I. The
homomorp hi sm d 12 corresponds to d 2: E2m-l,m --+ E2m+lO
' .
In particular, if
p - dim(f) and q - dimCO,
then
2. Factorisation theorem
Suppose now that V is a finite-dimensionallJ-module such that IJ/'L is semisimple.

Then
Example. Let IJ - tC IS E9 s1(2,tC) be the semidirect sum corresponding to the irreduc-
ible representation of slC2,tC) In tC 15. We put V - IJ and 'L - tC 15 (considered as an

Abelian Ue algebrn.). Then, we have H2(IJ,IJ) ~ o.
This computation has been done by Richardson to describe a rigid tie algebrn. (see
Chapter 5) whose second cohomology space H2(IJ,IJ) is not trivial.
VI. COHOMOLOGICAL CALCULUS AND COMPUTERS
To end this chapter we shall discuss some loglcials of formal calculus which have been
developed. For example, Mapple and Mathematlca, etc. These IOglcials permit us to use
personal computers for the study of the cohomology of tie algebras. of course, the
calculus related with the cohomology is linear, but the dimension of the underlying
spaces and the rank of used matrices are very big, and this implies a quick limitation.
With these IOglcials we can determine without toil, not only the dimensions of the
cohomologlcal spaces, but also the basis of these spaces and also for dimensions of
large enough tie algebras (this depends only of the possibility of the computer). Here
we present a progrn.m using Mathematica which permits :
C1) the verification of the Jacobi Identities;
(2) the calculus of the dimensions of B2(IJ,IJ) and z2(lJ, IJ} ;
(3) the determination of a basis of H2(IJ, IJ) ;
C4) the verification of the nilpotency.
102 Chapter 3
VI.l The MA THEMATlCA Program

BeginPackage [" ALII' ,,]
mu: :usage =
"muL._] is the ley of Lie algebra"
Adj : : usage =
"Adj [V] compute the adj oint of the vector y,
Adj [k] is the adjoint of the vector x[k],"
Jacobi: :usage "Jacobi[ ] gives the Jacobi conditions,"
XilPot: :usage "XilPot[ ] gives the nilpotence conditions,"
DilllB2: :usage "DimB2[ ] computes the dimension of B2(mu.mu),"
DimZ2: :usage "DimZ2[ ] computes the dimension of Z2(mu.mu) ,"
DimH2: :usage "DimB2[] cOllputes the dimension of B2(mu.mu),"
BaseDer: :usage "BaseDer[] determines a base of Der(mu),"
BaseB2: : usage "BaseB2[ ] determines a base of B2(mu.mu),"
BaseZ2: :usage "BaseZ2[ ] determines a base of Z2(mu.mu),"
BaseB2: :usage "BaseB2[ ] determines a base of B2(mu.mu),"
·{'x. 'n};
Begin[" 'Private''']
mu[O. x_] := 0;
mu[x_. 0] := 0;
mu[x_. x_] := 0;
mu[x_. y_] := -Jau[y. x] I; OrderedQ[{y.x}];
mu[x_+y_. z_] := mu[x. z] + mu[y. z];
mu[z_. x_+y_] := muEz. x] + muEz. y];
mu[a_ x_, y_] := a mu[x,yl.
mu[x_. &_ y_] := a au[x,Y]j
(* ADJOIITS *)
Coord [i_Integer. v_] := v I, Table[x[j]->If[j==i. 1. 0]. {j.1.n}]

Coord [v_] := Table[Coord[i. v]. {i. 1. n}]
Adj[i_Integer] := Adj[i] = Adj[x[i]]

Adj[v_] :=
Module [{adjv. fl.
adjv = Array[O. {n.n}];
For[f = 1. f <= n. f++. adjv[[f]] Coord [mu [v. x[f]]J];
Transpose [adjv]
(* JACOBI COIOITIOIS *)
Jacobi[] :=
Module [{i, j, k, LRel} ,
LRel = {};
For[i = 1, i <= n-2, i++,
For[j = i+1, j <= n-1, j++,
For[k = j+1, k <= n, k++,
LRel = Union[LRel, Jacobi[i, j, k]]
];
] ;
] ;
LRel
]
Jacobi [i_, j_, k_] :=

Module [{LRel} ,
LRel = Coord [mu[mu [x [i] , x[j]], x[k]] +
mu[mu[x[j], x[k]], x[i]] +
mu[mu[x[k], x[i]], x[j]]
] ;
Select [LRel, Function[!lumberQ[#]]]
]
(* IILPOTEICE COIDITIOIS *)
lilPot[] :=
Module[{i, LRel, cadj} ,
LRel = {};
For[i = 1, i <= n, i++, LRel = Union[LRel, lilPot[i]]];
LRel
lilPot[i_Integer] :=
Module [{pol , lrel, laux, lambda},
pol = Det [lambda IdentityMatrix[n] - Adj [i]];
Select [Expand[CoefficientList [pol, lambda]],
Function [!lumberQ [#]]]
104 ChapterS
(* RESOLUTIOR OF LlREAR SYSTEMS .... *)
Desp[ee_]:=
Module [{eex, lv, v, rs},
eex = Expand[ee];
Iv = Reverse[Flattenrvariables[eex]]];
If[Length[lv] > 0,
v = Iv[[l]];
rs = Flatten[Solve[eex == 0, v]] [[1]];
Set [Evaluate[rs[[l]]] , rs[[2]]]
]
(* DIMERSIOR OF SECORD COHOMOLOGICAL GROUP *)
DimB2[ ]:=
Module[{A, a, i, j, f, Df, Base},
Base=Array[x, {n}];
A=Array[a, {n,n}];
f[v_]:=Coord[v] . A . Base;
Df[i_, j_] := mu[f[x[i]], x[j]] +
mu[x[i], f[x[j]]] -
f[mu[x[i], x[j]]];
For [i=l, i<=n, i++,

For[j=i+1, j<=n, j++,
Map [Desp, Coord[Of[i,j]]]
]
] ;
n·2-Length[Flatten[Variables[A]]]
Cohomology of Ue Algebras 105
DimZ2[ ]:=
Hodule[{A, TA, a, i, j, k, f, Df, Base},
Base = Array[x, n];
A = Array [a, {n,n,n}];
TA = Transpose[A, {1,3,2}];
f[u_, v_] := Coord[u] . TA . Coord[v] . Base;
For[i = 1, i <= n - 1, i++,

For[j = i + 1, <= n, j++,
For[k = 1, k <= n, k++,
a[j ,i,k] = -a[i,j,k]
]
];
For[i = 1, i <= n, i++,

For[k = 1, k <= n, k++,
a[i,i,k] = 0
] ;
Df[i_Integer, j_Integer, k_Integer] :=

mu[f[x[i], x[j]], x[k]] +
mu[f[x[j], x[k]], xCi]] +
mu[f[x[k], x[i]], x[j]] +
f[mu[x[i], x[j]], x[k]] +
f[mu[x[j], x[k]], xCi]] +
f[mu[x[k], x[i]], x[j]];
For[i = 1, i <= n - 2, i++,

For[j = i + 1, j <= n - 1, j++,
For[k = j + 1, k <= n, k++,
Hap [Desp , Coord[Df[i,j,k]]]
]
] ;
Length [Flatt en [Variables [A]]]
DimH2I] := DimZ2[ ] - DilIlB2[ ]

106 Chapt6r3
(* BASE OF H2 *)
LDif[genV_List. genL_List] :=
Module [{dim. v. p. 1n1. i. j. base. genv. sol}.
d = Dimensions [First [genV]] ;
base = RowReduce[Map[Flatten. genL]];
genv = Map[Flatten. genV];
dim = Length[genv[[l]]];
p = U;
For[i = 1. i <= Length[base]. i++.
j 1;
While[j <= dim AA base[[i. j]] == O. j++];
If[j > dim. base = Drop[base. {ill. AppendTo[p. j]];
];
ini = Length[base] + 1;
For[i = 1. i <= Length[genV]. i++.

v = genv[[i]] - Sum[genv[[i. p[[j]]]] * base[[j]].
{j. 1. Length[base]}
] ;
j = 1;
While[j <= dim AA v[[j]] O. j++];
If[j <= dim.
v = l/v[[j]] * v;
AppendTo[p. j];
AppendTo[base. v];
];
] ;
sol = base[[Range[ini. Length[base]]]];

For[i = Length[d]. i> 1. i--.
sol = Map[Function[Partition[#. d[[i]]]]. sol]
] ;
sol
BaseDer [] : =
Module [{A. a. i. j. f. Df. base}.
base =
Array[x. in}];
A = Array [a. {n.n}];
f[v_] := Coord[v] . A . base;
Df[L. j_] := mu[f[x[i]], x[j]] +
mu[x [i] • f[x[j]]] -
f[mu[x[i]. x[j]]];
For[i = 1. i <= n. i++.

For[j = i+1. j <= n. j++.
Map [Desp. Coord[Df[i.j]]]
] ;
ParToBase [A]
ParToBase[lExPar_List] :=
Module [{lPar. dim. i. j}.
lPar = Flatten[Variables[lExPar]];
dim = Length[lPar];
VectorBase[i_] := lExPar I.
Table[lPar[[j]] -> If[j == i. 1.0].
{j. 1. dim}
] ;
Array [VectorBase. dimJ
BaseB2[] :=
Module[{BBaseEnd}.
BaseEnd = Map[Function[Partition[#. n]].
IdentityMatrix[n-2]
] ;
Map [Delta. LDif[BaseEnd. BaseDer[ ]]]
108 Chapter 3
Delta[A_List] :=
Module[{f, Df, base, Dfij},
base = Array[x, {n}];
f[v_] := Coord[v] . A . base;
Df[i_, j_] : = mu[f [x [i)], x [j]] +
mu[x[i] , f[x[j]]] -
f [mu [x [i) , x[j]]] ;
Table[If[k == 1, Dfij Df[i,j]];

Coord [k , Dfij],
{i,l,n}, {j,l,n}, {k,l,n}
]
BaseZ2[ ] :=
Module [{A, TA, a, i, j, k, f, Df, Base},
Base = Array[x, n];
A = Array[a, {n,n,n}];
TA = Transpose[A, {1,3,2}];
f[u_, v_] := Coord[u] . TA . Coord[v] . Base;
For[i = 1, i <= n - 1, i++,

For [j = i + 1, j <= n, j ++ ,
For [k=l , k<=n, k++,
a[j ,i,k]=-a[i,j ,k]
]
];
For[i = 1, i <= n, i++,

For[k = 1, k <= n, k++,
a[i,i,k] = 0
J;
Df[i_Integer, j_Integer, k_Integer] :=

mu[f[x[i], x[j]], x[k]] +
mu[f[x[j], x[k]] , xCi]] +
mu[f[x [k], x [i]] , x [j]] +
f[mu[x[i] , x[j]], x[k]] +
f[mu[x[j], x[k]] , xCi]] +
f [mu[x [k] , x[i]], x[j]];
For[i = 1, i <= n - 2, i++,

For [j = i + 1, j <= n - 1, j++,
For[k = j + 1, k <= n, k++,
Kap[Desp, Coord[D:t[i,j ,It]]]
]
] ;
ParToBase [A]
]
BaseH2[] := LDif[BaseZ2[], BaseB2[ ]]
End[ ]
EndPackage [ ]
110 ChapterS
VI.2. How to use this program
First, one Introduces the constants of the structure by giving the dimensions of the Ue
algebra and writing the nontrivial brackets :
mu[x!l],x[jll - C1jkx[k] .
For example, for the flve-dimenslonal Heisenberg algebra, we write:
n - 5,
mu[X[2],x[3]] - x[l],
mu[x[4],x[5]] - x[l].
Now, the procedure Jacobi[ I returns all the Jacobi equations which are not satisfied. In
this example, Jacobi[ ] gives { l.
If we want compute the dimension of B2 or Z2 or H2, It Is suffldent to write dimB2[ ]
or dImZ2[ ] . or dImH2[ ]
To expllddy obtain a basis for these spaces, write baseH2I ], or baseZ2l lor baseB2[ ].
CHAPTER 4
COHOMOLOGY OF SOME
NILPOTENT LIE ALGEBRAS
Today a few concrete results concerning cohomological calculaUons on nilpotent lie

algebras are revealed. In this chapter we present these calculations tor some particular
but nevertheless Important classes of nilpotent lie algebras: I.e. the filiform algebras
and the nibadtca1s of parabolic algebras. We restrict ourselves here to the cohomology
whose coefficients are In the adjoint module; some results concerning the
cohomology with values In the field are given In the last chapter. As the lie algebra of
derivaUons related with the first cohomology grouP. we endeavour to describe this
algebra.
I. DERIVATIONS OF SOME FILIFORM ALGEBRAS
Recall (see Chapter 2) that a (n+ 1)-dimenslon filiform algebra has a maximal Index of
nilpotency equal to n. Among all these algebras. only four play an Important role. They
are described In the basis (eo.el' ...•en ) by :
Ln : [eo.el] - el+l , 1- 1....,0-1 ,
Qn : [eo.el] - el+l • I - 1•...,0-1 and n-2k+ 1 ,

[el.en-l] - (-1) len , 1- 1,...,0-1,
112 Chapter 4
Rn: [eo,e,] - e'+l, 1- 1,00.,n-l,

[e l,e,] - e'+2 , I - 2,00.,n-2 ,
W n: [eo,e,] - e2+1, i - 1,00.,n-l,

6(1-1)!(j-1)!(j-1) with { 1 ~ IJ ~ n-2
[e"e~ - (I+j)! e'+f+l
I+J+l~n.
1.1. The algebra of derivations of Ln
Proposition 1. The linear-mappings ad eo, ad el ,00', ad en_l ,h2' h3' ... ,bn-l, tl' t2 and t3
of the space Ln, with
hk(eJ-e'+k forany l~f~n-k,
tl(e,) - e, forany 1 ~i ~n , tl (eo) - 0 ,

t2 (eo) - eo , tz (eJ - (i -1)e, forany 2 ~f ~n , t3 (eo) - el ,
fonn a basis of Der Ln.
Proof. It Is easy to verify that these mappings are the derivations of In and that they are
hneadyindependent
Let d e Der Ln . The Ue algebra Ln admits a natural graduation Ln - E9ie z(Ln>, , where
(In), is the space generated by the vectors (e,) for 2 ~ i ~ n, while (In)l is generated by
eo and e l . The other spaces are reduced to (0). This graduation is associated to the
descending central sequence (see Chapter 2). The derivation d can be decomposed
with respect to the graduation:
d- do + d l + 00. + ~ with ~ E DerIn .
Consider do e Oer In. It is obvious that

Soeo + Sloe! if k - 0
ddek) - ( SleO + SileO if k - 1
6tA if 2~k~n
) ,
CohorrrJIogyof some ~ Uealgebras 113
for some 90. 9'0, 9t. 9' 11 92 ,•••, On e K . It follows from the equality
do [a, b] -[do (a), b] -[a, do (b)] - 0,
for a - e 1 and b - ek for 2 S k < n, that S' 1 - o. This same equatton, with a - eo and
b - ek for 1 S k < n, yields 9k+1- 90 + 9k . Considering the values CO, 1, 0), 0, 0, 0), and
(0,0, 1) for the trlpleCS o• 9 1, 9'0)' we arrive at t l , ~,and '3, respectively. Thus, we have
do - 91tl + 92t2 + 9'0 t::s .
Now consider d k e Der Ln with k ~ 1 . It is obvious that
dk(e!l _ { 'tOek+l for i - 0 } ,

'tlek+1 for 1 Si Sn-k
for some 'to, 'tl ,..., 'to.k' Put
d'k - dk+ 'toadek-'tl hk .
Then, d'kCeO) - d'k(el) - O. We will show that d'k - O. Indeed, suppose d'k -:I- 0 .
Then there exists an integer m such that d'kCem) -:I- 0 and d'k(e~ - 0 for I < m
(clearly m ~ 2) . This Implies that
d'k!:eo, em.l] - [d'k(eO) ,em-I] -[eo, d'k(em-l)] - d'k(e m) -:I- 0 ,
which is impossible, proving the proposition.
Corollary 1. dim Der L - 2n + 1 .
Corollary 2. dim WCI., L) - n + 1 .
In fact
In fact B2 (Ln,Ln) - {df for f e End(Ln)} ,
where dfCx,y)--tIx,y]+[fCx),y]+[x,fCy)] x,ye Ln.

114 Chapter 4
1.2. The algebra of derivations of W n
Proposition 2. Let h, t1, ~,t, be the endomorphism of Wn defined by

h{et}-{i+l}et O~I ~n
tl{et} - en tde.) - 0 1*1

t2 (e.) - e n·3+1 1- 1,2,3 t2 (e.) - 0 I ~4
t3 (el) - e n-2+1 1- 1,2 t3 (e.) - 0 I ~3 .

Then, the endomorphism ad eo, ad el' ... , ad en-I' h, tl' t2 and t, form a basIs of
DerW n ·
Proof. These endomorphisms are Independents and belong to DerW n. Let d be a

derivation of W n. We can grade Wn by putting
We put W n - EDleZ {W n>1 ,where {W n)1 Is the space generated by el-l for I ~ 1 and
{W n )j- {OJ for J ~ n+2 or J ~O. This graduation Is not associated to the filtraUon given
by the filtration descending central sequence. The derivation d can be decomposed
as d - d_ 1 + do + ... + do with
diE Der{W n>1 - 1 :s I :s n ,
dl{W n>j C (W n)l+j .
Note that the Ideals CICWn) are conserved by the derivations; these Ideals are
characteristic. In the fonnal decomposition of d, d - Ea1ezd l , we have dk - 0 for k ~ -2 .
By the same reasoning, we have d.ICel) - 0 for 1 :s 1:S n. But as d. 1 Is a derivation, we

don't have d_ICel) - aeo except the case a - o. This shows that d. 1 - o. So d - do +
d l + ... + do. By adding, If necessary, a linear combination of the derivations ad eo, ... ,
ad en-I' hand tl ,we can suppose that
d- do + d l + ... + do-2
and dCe~ - 0 .
We put dk(ej) - ajkek+j for o:s k:s n-2 and 1~ j ~ n-k .
The Identity
dk(e\oej] - [~et~ej] + [el.dk(eJ)]
gives
Co/xxnoIogy of some nilpotent Lie a/gebIas 115
at = ai- ... - akn_k for I - 0
r
and
a = 0 with 0 ~ k ~ n-4 and k;t 1 for I - 1 and j- 2 .
This proves the proposition.
Corollary 1. Dim DerCWn) - n + 4 .
Corollary 3. Dim BleWn' Wn) - n 2 + n - 3 .
1.3. The algebra of derivations of Q n (n =21<+1)
Proposition 3. Let tl' tz, ~, h5, ..., hzk-l be the elements of EndCQn) defined by
tlCeO) - - e1
tlCe~ - el if 1 SiS n-l
tlCe n) - 2en
tleo) - eo + e 1
t2Ce~ - (i-1)el if 1 SIS n-l
tlerJ - Cn-2)en
hsCe~ - el+s if 1 SiS n-s

hsCe~ - 0 if not.
Then the endomorphisms ad e l Ci - O, ... ,n-1), tl' t2 and h 3 , h 5 , ... , h 2k-1 form a basis
of DerCQn)'
Proof. The independence and the existence of these endomorphisms in DerCQn) can
be easily verified.
One grades Qn by the filtration associated to the descending central sequence Cas for
Ln}. If d is in DerCQrJ, it can be decomposed as
d - do + ... + ~-l
with d l E DerCQ~ and dICQn)J c CQn)I+J for 0 SiS n-l and is j S n .
The derivation do verifies
116 Chapter 4
doCe o) - no eo + ~o e l
doCe l ) - al eo + ~l e l
doCe~ - Clj el if
As dd:eLel] - [dd.elbel] + [eLdd.el)] , one obtains d l - 0 .
This Identity, calculated for the pair Ceo,e~, gives a l+ l - ~l + lao (for i - 1,... ,n-2) and
Un - no + a n-l - ~o Cfor I - n-1). For the pair Cel,en_l)' one obtains an - ~l + an-I.
Then, all the coefficients can be ex primed with only a o and ~l. The derivations
corresponding to Cno - 1, ~l - 0) and Cno - 0, ~l - 1) are nothing but t2 and t l .
Now, consider the derivations ds with n-l ~ s ~ l.
Onehas:
dsCeo) - no e s + I
if l~i~n-k
ifnot .
Even If we add to ds a multiple of ad e s ' we can suppose <Xo - o. The Identity of the
derivation applies to the pairs Ceo, e~, and with 1 ~ i < n-s-l gives:
a l - <Xz - ••• - an-s .
The same identity applies to the pairs Cel,e~ with 1~, j ~ n-s, i+j - n-s, but for
s < n-l gives :
(-I)lal - (-I)lal where I+j - n-s.
If s Is even, we have necessary al - 0 and ds - o. If s Is odd., the Identity Is obviously
satisfied because I and j have the same parity (recall that n Is odd).
Then ds - 0 If s is even,
- a l es+1
dsCe~ lfsisodd.
We can conclude that
dzs'+l - h2S '+1 with 1 ~ s' ~ k-l and recall that d l - al ad eo.
The proposition Is proved.
Corollary 1. Dim DerCQ2k+l) - 3k+3.

Cchom:JIogyof some ni/potent Lie aJgebIas 117
Corollary 3. Dim B2(Q2k+l. Q2k+l) - 4k2 + 5k + 1.
1.4. The algebra of derivations of Rn
Proposition 4. Let t, h2' h, •...• ho-l be the endomorphisms of Rn defJned by

t(e.) - (1+ 1~ • O:S:ISn,
hk(eo - el+k , 1 SIS n-k, hk(eO) - 0 •

hen the endomorphisms ad e t CI - O. 1, ...• n-1), t. ~ •...• h n_l constitute a basis of
Dez(R~.
Proof. Note that these endomorphlsms are the derivations of ~ and form a free
linear system. Let d be a derivation of Rn. We consider the graduation of Rn given
by Rn - E9 te z(Rn>t , where (Rn>t is 1he subspace of Rn generated by the vector el-l • 1-1..n+]
and (Rn)j - (OJ for J S 0 and J ~ n+2. Again. we note that this graduation Is not
associated to the filtration defined by the central descending sequence. Each Ideal
Cf(R.~Is stable for the derivations andls generated by the vectors e l+l •... , en assoon
as I ~ 1. Then. the derivation d can be written:
d - dol + do + ... + ~ •
with d t e DerCRn) , dt:Rn:>j c CRn)t+j
Consider do e Dez(R~. We have

doCe~ - (Xt et , I - O. 1, ... , n .
The Identity
do[eo,et] - [do(eo),et] + [eo,d(eO].
where 1:S: IS n-1 gives

(Xt+l - (Xo + (Xi • 1 SI Sn-1.
The same Identity, applied to the pair Cel,e~ with 2:S I :s n-2. gives
(Xt+2 - (Xl + (Xi , 2 SIS n-2 .
Then «t - 0+ 1)«0 and do - «ot. Now, we consider ~ e Dez(R~. where 2:S k :s

n-1. By addJng, If it Is necessary, alinear combination of the derivations ad e l • ad e 2•...•
ad en-2' we can suppose that ~(eo) - o.
As.wehave:
118 Chapter 4
dk(el) - a~I+1 ' 1 ~j ~n-k ,

the identity of the derivation applied to the pair (eo,ey, gives
a l - a z - ... - an-k ,
then ~ - <It hk .
Note also that ~ - a ad e n-l for some a in te. Now, it Is sufficient to consider only
the derivation d l . By addition, ifit is necessary, of a derivation of the type a ad eo, we
can suppose that dl(e l ) - o. This shows that d l is defined as follow:
dl(el) - alel+l , 0 ~ i ~ n-1 , where al - 0 .
The identity of the derivation applied to the pair (eo,e~, where i - 1, 2, ..., n-2, gives:
aZ=al=O and al-Cj-2)ao if 3~j~n-l.
The same identity, applied to the pair (el,e Z)' gives CXz - a 4 • Then, we deduce a o - 0
and d l - 0 . Q.E.D.
1.5. Derivations of the standard nilpotent algebra in fj = sl(r+ 1, C)
Let n be a standard subalgebra of 9 - glCr+ 1,te) defined by asubset R of d+ composed

of roots pairwise incomparable. We then have:
n - L 9a, where d' - {a E d + I 3 'Y E R with a ~ 'Y} .
aeA'
Let Fa. be the Cartan subalgebra of 9 fonned of all the diagonal matrices of 9. The
algebra of derivation Der n of the algebra n is naturally provided with a structure of
h.-module (see Chapter 3) :
(h.dXx) - [h,d(x)] - d([h,x]) .
Then Der n is the vectorial direct sum of the weighting subspaces VA' where the
elements of VA are the vectors having A. as their weight. If d Is anontrlvlal element of
VA' then d(Xa ) - c .Xp, where A. -
- a. This pennlts us to consider only the weighting
~
derivations for the description of Der n, I.e. the derivations belonging to VA for a
weight A. of the type A- ~ - a, where a, ~ are In d.
119
Let NlJn the nonnalizer of n In 9. We put

Dl- {de Endn/d-adx where xeNlJn}.
D2- Id e Endn/ d 1[n,nJ - 0 • den) CZ(9)} .
It Is clear that D 1 and D2 are subsets of Der n .
Let S - (a 1.CXz • .... «rl be the ordered subset of the simple roots Csee Chapter 1). One
designates by ~ the root <Xf + ... + ~1 If I < J. or - «j - «j+l - ... - «t-l if I > J.
Lemma 1. Letn be astandaltlsubalgebra of I) deflnedbyR.lfR doesn'tcontaln the

roots of the fonn al,k or «S,r+l' then
Der n - Dl + D2 .
Proof. If n is Abelian. the theorem Is obvious. Suppose that n is not Abelian. Let d be
aderlvation with weight A. - P- a, where a and p are In .:1'. We disUnguishthree cases
CO 1.-p-a-O;
(II) 1.- p-ae.:1;
ClIO 1.-p-a~.:1.
Case (i). For all Indices i such that Xat....1 is In n. we choose an element h In Fa. such that
ad h (X at....1) - f (X at....1). Then the derivation d ' - f - ad h is rendered as zero for
every vector XCXi,rtl of n. We can suppose, by adding, of necessity, a derivation of D2,
that d' is zero on aU the vectors Xa of (n-{n-nD n ZCn). By using the fact that d' is a
derivation, we can easdy prove that d'CXa ) - 0 for every Xa of n.
Case (ii). Let A. - P- a - ak,s' By adding, if it is necessary, a linear combination of

derivations of D2 and ad ej. we can suppose that dCX~ - 0 if «S,m where m is
't -
the smallest Indice such that Xa..m is In n at is easy to see that such a root exists). Then
we verify that dCX) - 0 for every Xv In n.
Case (iii). Let d be a derivation of n such that dCXa) - ad X~, p - a e .:1, Xa ' X~ e n.
Besides, suppose that d is not In D 2 • Then
A. - (1.cp+acp+l +...+ a't - a p - ... - au) with <p S 't S P S u
or
120 Chapter 4
A. - (alP+alP+I +...+ a't - a p - ..• - au) with <P < 't < p-l .
In each case, we can write A. as a difference of two roots A. - ~ - a, and in a more

distinguished manner:
~ - (alP+alP+I +...+ a't) and a - (al + ... + an)
or
~ - (alP +...+ ai-I) and a - (al+l +... + an)
in the first case, and
~ - ± (alP +... + ai-I) and a - ±(at+l + ...+ al.I)

in the second case.
Except in the case where R admits a root of the form al,k - a l +...+ ak-l (or of the
form alP,f+l - alP +...+ a r ) andif a - al +..+ at (respectively, a - at +...+ ~), in other
cases, we can always find a root vector x.., in n such that
This proves the lemma.
Let n be a standard nilpotent subalgebra of 9 defined by R, where R contains an

element of the form ~,k. We put
R - {a lP1,kl,alpz,k2 ' ... , acp...,km} ,
where 1 - <Pl < CP.z < ... < 'Pm S r , kl < k2 < ... < ksn S r+ 1 , 'Pi S kl .
If t, s, p are integers satisfying
kl < t ~ S k2 , P > <Pm , Xs,p En,
then the endomorphism of n defined by
~X (Xl,.) - Xs,p ,d(X (Xis) - Xl,:
is in Der n. We designate by D, the set of all these derivations.
Now suppose that R contains an element of the fonn acp,r+l" We put
R - {a lPl,kl,alpz,k2 ' ... , alPm.r+l }
with <Pl < CP.z < ... < <Pm S r , kl < k2 < ... < ksn - r+ 1 , 'Pi S kl .
If t, s, p are integers satisfying
Is kl , 'Pm-l < t S p S'Pm , x"t En,
CohomoIogyd some nIpoIent I..ie algebras 121
then the endomorphism of n defined by

dex.,r+1) - X~ • dC~,r+1) - X),t
is a derivation of n. We note by 04 the set composed of all these derivations.
Theorem 1. Let n be a standard nilpotent subalgebra 0/9 defined by a system R of

A+ composed of roots pairwise incomparable. Then
Der n - 01 + 02 + 0, + 04 .
This theorem can be obtained direcdy by using the lemma and the definitions of 0,
and 04.
II. COHOMOLOGY OF FILIFORM LIE ALGEBRAS
In this secdon. we shall describe the cohomology spaces with values In the adjoint
module of the "model" filiform tie algebra which is denoted ~ and which Is deflned,
in the basis eeO.e 1•••• ,ert. by the brackets [eo.ell - e l+ 1 1- 1•...•n. If F;:2e~.~) are the
elements of the filtration obtained by the central descending and ascending sequences:
then the space Fo2 2el.n.Ln) Is Identified as the "tangent spaces" at the point 1.0 to the set
of nilpotent tie algebras. This geometrical description will be developed In a later
chapter.
We define 2-cochalns <l>tJ of a tie algebra 1.0 In the adjoint representations by

setting <l>tJ eeo.e~ - ej for I. j - 1 ,.... n . It Is obvious that these cochalns are
cocycles. Therefore. for a description of the cocycle space 2 2eLn.Ln) it suffices to
describe only the cocycles cp given by <pCeo.e~ - 0 for all 1 :s: I :s: n . Since
Ln- E9 (L.JI Is a graded algebra. we can restrtct ourselves to cocycles cp with ql{Ln)1 (Ln~ c(L.Jk
We remark that a 2-cochaln cp satisfying q(eo • eJ - 0 is a cocycle if and only if
122 Chapter 4
for all 1 ~ i , j ~ n Ofr > n, then we set e r - 0).
Proposition 5. Let (k,z) be a pair of Integers such that 1 S k S n-l,2k S s S n ; then

there Is one and only one element 'I' E Z2(Ln,Ln) satisfying
'I'(el,el+l)- {
el if i-k,
0 if not , }
\jJCeo,e~ - 0 ,IS i S n .
Proof. Let 'I' be a cocycle satisfying the hypothesis of the proposition, and we have :
'I'(ehel) - adeo('I'(ehel-l» -'I'(el+hel-l) .
This relation shows that one of the following conditions is satisfied
'll(ehel) - 0 if
(1) k<l<j;
(2) I<j~k;
(3) k-I > j-k-l .

Let 1 ~ I ~ k < j ~ n. Then an induction with respect to (j-k-D - (k-O - i+j - 2k-l
shows that
'I'(el,el) - (_I)k-l C k-~_l (adeo~+1-2k-l el.
This Is deduced from the above relation and from the classical rule
C k-lJ-k-2 + C k-I-l
J-k-2 -
C k-I
J-k-l·
This shows the existence and the unicity of the cocycle '1'.
We now construct the cocycles 'l'k,s for aliI ~ k ~ n-l and2k ~ s~ n (we need2k ~ s in
order to ensure that q{el' e n +l) - 0 for aliI ~ i ~ n) by putting
es If 1- k }
'l'k,s(ehel+ 1) - {
o if I *k
Proposition 6. The cocycles <j>IJ and 'l'k,s with 1 S i < j ~ n and 2k+ 1 ~ 1 ~ n form a
basis of FOZ2(~,~).
Proof. Let ~ E FOZ2(L ,L ) be a cocycle compatible with the grading that satisfies
CohomoIogyof some ~ Liea/geblas 123
Ii CCIn)f,CIn)f) c CIn)f+J . SubtracUng, If necessary, the cocycles <Pt", we may assume that
IiCeo,e.> - 0 for all 1 SIS n. If IiCef,el +1) - 0 for all 1 SIS n-l ,then we have Ii - o.
Let k be the minimal positive Integer such that IiCek,ek+1) - A.es with A.;e 0, A. E C and
s ~ 2k+ 1. Then the cocycle Ii' - Ii - A.'I'k,s sallsftes the condition ~. Cek,ek+1) - O. If
~. ;e 0, then we repeat the argument for Ii' . Continuing in this way, we eventually find
that ~ is a linear combination of cocycles of the form 'IIk,s. The linear independence of
the cocycles 'PtJ and 'l'k,s Is obvious, and the proof Is complete.
Corollary 1. The dimension of the space FoZ2CL,IJ is equal to C3n2-4n+ 1)/4 If n is odd,
and to (30 2-40)/4 If n If even.
Corollary 2. The dimension of the space FoH2CL,IJ is equal to (o2-2n-3)/4 If n is odd,

and to (o2-2n-4)/4 if n is even. A basis of FoH2CL,IJ is formed by the cohomology
classes of the cocycJes 'IIk,s where 1 ~ k ~ n, 4 ~ s ~ n and s ~ 2k + 1.
The proof of Corollary 2 can be deduced from the fact that the cocycle <Pt", 1 ~ I < j ~
n, and the cocycle '1'1" are elements of the space FoB2CLo,Ln). In fact, fPtJ - df and
'1'1" - dg, where the endomorphisms f and g are defined by
fCe m) - e)-l If m - I ,
fCe m) - 0 if not,
~em) -eo If m - I ,
~em) - 0 ifnot .
The cohomology classes of the other elements of FoZ2CLo,Ln) are nonzero and form a
free system.
ITI. COHOMOLOGY OF THE NILRADICALS OF PARABOUC

ALGEBRAS
This study gives rise to the works of Kostant which gave a description of the
cohomology space H"Cn,V), where fl. is the nihadical of a parabolic subalgebra p of a
complex semlslmple Ue algebra 9, and V is an Irreducible 9-module restricted to fl.. If
124 Chapter 4
9 is simple, then for V, one can consider the Ue algebra9 with respect to the adjoint
action, and we obtain a description of H-Cn,9). However, for many applications Cthe
study of the automorphisms of the Ue algebra n, the problem of classifying Ue
algebras with a given radical, and algebraiC Ue algebras with a given nIlra.dlcal, the study
of the deformations of the Ue algebra n, and the study of a variety of Ue structures of a
vector space) it would be useful to have a description of the cohomology space
H-Cn,n). As Kostant observed by modifying his arguments somewhat, one could
obtain a description of the space H1Cm,m), where m is the nllradicalof a Borel
subalgebra of a Ue algebra 9. This would give a description of the derivation algebra
Derm.
In the follOwing two sections, we describe H 1Cn,n) and H2Cn,n) for the nUradical n of
an arbitrary parabolic subalgebra p of a complex simple Ue algebra 9. The prototype
Is the Heisenberg algebra. To save on language, we will speak of the nllradical of
parabolic algebras instead of the nllradical of parabolic subalgebras of simple
complex algebras.
111.1. A decomposition of the groups Hi(n,n) i = 1,2
ill.l.l. The root filtration
Let us recall some classical notions. Let 9 be acomplex semisimple Ue algebra of finite
dimension, fI. a Cartan subalgebra of it, /1 a root system with respect to fl., and.
S a basis of /1 Ca system of simple roots). We have Introduced a partial ordering in fl.-
relative to which the positive elements are linear combinations of roots In S with
positive coefficients. This ordering divldes /1 Into aset of positive roots /1+ and aset of
negative roots /1-, so that /1 - /1+ u /1- .
We denote by 9 a the root space corresponding to the root <X.. We shall assume that for
each <X. E /1, some nonzero element e a E 9 a Is fixed.
Let Sl be a subsystem of s, /10 the set of roots that can be expressed In terms of S1'
Cohomology of some nilpotent Lie algebras 125
and fl.' = (fl. - fl.o) n fl.+. Then the subalgebra p - h. G) L I}a is parabolic, and
ae~u/1'
Is a reductive Levi subalgebra of It, and Its nilradical has the fonn fI. - L I}a
<xe/1'
Recall that any parabolic subalgebra Is conjugate to the subalgebra p for some
subsystem SI of S.
Let S - {aI' ... , an} and let ~ E fl.. We consider a decomposition kl a l + ... + k n an of
the root ~ Into simple roots. We call the Integer lC~) - kl + ... + k n the length of the
root ~. We consider the subset S -SI - {<XI,. ... ,<XI.} . We call the integer In(~) - ~, + ... + kl,
the fl.-length (or the S-Sllength) of ~. In the case S - S-Sl ' the concepts of length and
the fl.-length of roots obviously coincide. Since an arbitrary root decomposes into
simple roots with either positive or negative coefficients, we have
for any ~1' ~2 , ~1 + ~2 E fl. .

We introduce the follOwing notation:
fl.1=I~E fl./ln(~)=i} ; I}i - L I}a, i~O
The additivity of In shows that I} - El1ieZl}i is a lZ-graded lie algebra, where
Recall that the length of a negative root is negative, then the I}I for I < 0 are not
necessary zero. In fact, if I}i ~ {OJ for i > 0, then I}-t ~ {OJ .
We call the resulting grading the natural grading of the lie algebra I} with respect to fl..
The subalgebra fI. = G) fl.i Is also ll-graded if one sets
IE!
Setting
126 Chapter 4
Fk(9) - $ 9i , Fk(n) - $ ni ,
i~k i.,k
we obtain the filtrations FCip and Fen) of the tie algebras 9 and n. We note that Fen)
coincides with the natural filtration of the nilpotent tie algebra n, i.e.,
Fin) - n , Fk(n) - [n,Fk_tn)] .
Remark. All subspaces 91' I E 3::, are s-modules with respect to the adjoint action Cthe
subalgebra s coincides with 90)' Since s is reductive, these s-modules are semisimple.
In addition, the n-modules n, 9, and 9/n are semisimple s-modules. In what follows,
we identify the s-module 9J'n with the submodule $ iSO 9i in 9 complementary to n.
Now, the gradings of the tie algebras n and 9 define 3::-gradings in the spaces of
cocoons, cocycles, and coboundarles :
Cl (n,t) - $ C/ (n,t) ;
iEZ
zl (n,t) - $ Zi (n,t) ; B(n,t) - $ B! (n,t) ;
iEZ iEZ
where t - n, 9, or 91'n. This grading is compatible with the coboundary operator, and
we also obtain a 3::-grading of the cohomology space:
HI (n,t) - $ Hli (n,t) .
iEZ
Remark. In the space of cochains Cl(n,t,) where t - n, 9 or 9/n ap-module structure

can be introduced in the usual way :
(ac) (Xl, ... 'Xl) - ac(Xh ... ,Xj) - 2. C(XI'''' ,(a,xs], ... ,Xl) ,
ISsSj
where a E p and c E Cl(n,t). The spaces of cocycles ZJ(n,t) and coboundarles BJ(n,t)
are p-submodules of the p-module Cl(n,t), and the space Hl(n,t) also acquires a p-
module structure. All these p-modules can also be regarded as s-modules with respect
to the same action, which are semisimple.
Cohomology of some nilpotent Lie aJgebras 127
III.1.2. The decomposition theorem
Consider a short exact sequence

<P IJI
O~n~1} ~.yn~O,
where <p Is an Imbedding and 'II Is the natural homomorphism. This sequence generates
a long exact sequence
.
Ii 1-1
.
<PJ
.
IJIJ
... ~ HJ-l (n,.yn) ~ HJ (n,n) ~ HJ (n,l}) ~ HJ (n, .yn) ~ ....
We will note the space 1m <P*J c HJ(n,l}) by HJfund(n, n)_ It is called the fundamental
cohomology space. We can consider the space HJfunin, n) as a sub-s-module of
HJ(n,n) by considering a supplementary of 1m d·l-l ( HJ(n,l}/n)) in HJ(n,n) .
Theorem 2_ Let n be the nilradical of a parabolic suba1gebra p of a complex simple

Ue algebra I}, defined by means of a subsystem SI c S and different from those
listed in Table 1. Then
(a) H1(n,n) == s E9 Hl(n,l}) , and

(b) H2(n,n) = HJ (n,l}.n) E9 H2fund (n,n)
Proof. To prove (b), itis enough to show that

0·1: HI (n, .yn) ~ H2 (n,n)
is injective, and to prove (a) it is enough to verify that
H~n, I}) - Hlund(n,n)
and that 0.0 Is injective.
(a) According to a result of Kostant(Ko), as representatives of the primitive elements

of the simple components of the semisimple s-module Hl(n,~, one can take the
follOwing system of cocycles (which we shall henceforth identify with the
corresponding elements in Hl(n,~ : {fa I ex E S-SI}' where
128 Chapter 4
TABLE!
IJ 5 51
Al (Xl
An ,n ~2 (Xl
An,n ~2 (Xn
Cn,n ~2 (Xl
TABLE 2
IJ 5 51
An ,n ~2 (Xl
C n ,n ~2 (Xl, (X2
CohorooIogyd some ~ Liea/geblas 129
f.(e,)- {
eSg(II) , if 'Y - a
o, if 'Y¢ a }.
Sa Is a reflection of the space s" that carries A into Itself and such that sa(a) - - a .
Definition 1. A root a of S 1s caJJed sJngular if O-a 1s a root where 8 1s maximal.
If a Is not a singular root, then a Is not connected with a minimal root (- 8) in an

extended Oynkin diagram; in this case, siID - 8, and we have f E HlfundCn,n). If a Is a
singular root, then saCID - 8 - 2a for a simple Ue algebra '} of type Cn , and
sa(8) - 8 - a for all remaining types of simple Ue algebras '} . In all cases, except those
listed in Table l,saCID E !J.' ,I.e., e su (lI) E nandsof E Hlfund(n,n).
The spaces of n.-Invariants elements in n and in'} with respect to the adjoint action
coincide (up to imbedding In '}) and are equal to the center zen). This means that
HO(n,tp - HOCn,n) and the mapping 8"0 in the sequence Is Injective.
(b) We consider the given long exact sequence. We know that
1m CP"l - Hlfund(n,n) ,
and, from the proof of (a),
H1 fund(n,n) - H1Cn,,}).
This means that the mapping cpo 1 is surjective and 1m'll" 1 - O. Since the long sequence
is exact, we have Ker 8"1 - 1m'll" 1 - 0, i.e., the mapping 8" 1 is injective. This proves
the theorem.
Remark. It is well known that Z1(n,n) coincides with the derivation algebra Der n.
Theorem 1, in essence, gives a description of Der n as an s-module.
Moreover, Theorem 1 reduces the descriptions of H2(n,n) to the description of

W(n, ,}/n) and H2 fund(n,n) (except for the cases listed In Table n. It turns out that for
all n, except for those listed in Tables 1 and 2,
130 Chapter 4
In addition, it turns out that H2 fund(n,n) contains all the cocycles in H2(n,n) that
preserve the filtration, I.e.,
Now we examine the cases listed In Table 1 and describe the space H2(n,n). We
denote simple roots by a1 , ..., Un and number them according to Chapter 1.
(a) 9 - A1 and S Sl - (a 1) •
The Ue algebra n is one-dimenstonal, and so H2(n,n) - 0 .

(b) 9 - An , en ~ 2) and S Sl - {a1} or {Un}.
The Ue algebra n is Abelian andH2(n,n) consists of all skew-symmetric bilinear maps
f:nAn-+n.
(c) 9 - Cn , en ~ 2) and S Sl - {a 1} •
The Ue algebra n is isomorphic to the Heisenberg algebra Hn-1 andone can Imbed it
Into An as the nilradical of the parabolic subalgebra defined by the subset S,
Sl - {a1,Un} and use Theorem 1.
At the end of this chapter, we will carefully detail his example.
111.2. The calculation of the spaces Hl(n,q!n)
As shown in the previous section, the description of the space H2(n,n) reduces to the
calculal10n of H1(n,CJ/n). In this section, we shall prove a number of lemmas that are
necessary for the description of H1(n,CJ/n) . In order to understand the ideaof the
calculations, we shall briefly describe their order.
(i) Some properties of cocyc1es

We saw that the spaces
ckfn,t) , Zkfn,t) , Bkfn,t) , Hkfn,t)
are semlslmple s-modules with respect to the usual action. As the n-module t, we can
take n, 9 or 91n. These spaces can also be regarded as fl.-modules with respect to the
CohomoIogyof some nIpotent Lie algebras 131
same action which are semisimple. We call elements of the components of the
decomposition of these h.-modules into simple submodules weIght elements
Ccochains, cocycles, coboundarles, cohomologies). Let
be an element of Ck/n,rj). For any h e h., we have
(h.g) (ea1, ... ,eak) - [h,ep] - ±

~1
al (h)ep - (p(h) - ±
~I
al (h»)ep .
n
This means that g is a weight cochain with P- L (XI • Since any cochatn (resp. cocycle,
1
coboundary, cohomology) in our case is a linear combination of cochains of the form
C·), to describe the space of cochatns Ccocycles, coboundaries, cohomologies) it m
enough to find all the weight cochatns that correspond to a weight A. of the form
L
~- ~ ai, where aland ~are roots,e lXl e n,andell e t.
Lemma 1. Let f be an element of C l(n, 9) such that fen) does not lie ~ FI 9 for
some i and let df(n) c FI 9 (d Is the coboundaryoperator). Then there exists
an element ae n l such that fCa) ¢o.
Proof. Assume the contrary, let fen l ) - 0 and let r be the smallest integer such that
fen~ ¢ O. We choose an element x e nr for which fex) ¢ O. Since x can be expressed ll'!iI
a linear combination of root vectors and since f is linear, we can assume that x - e p'
where ~ is a root with n-Iength r. Since r> 1, we can decompose ~ into the sum
~I + ~2 ,where ~I e Ak, ~2 e Am' 1 S k, m < r , and k + m - r. We have
By the hypothesis of the lemma, df (elh,elJJ E F1 9. Since ~epl'e/32] - NploP2 ' f (e p) is anoo-
zero element that does not lie in Ff9, this is possible only if f (e 13,) ¢ 0 or f (e~) ¢ O.
But this contradicts the fact that r is minimal. This proves the lemma.
Corollary 1. Let f be a nonzero cocycle In 2 1(n,91n). Then there exists an element

ae n l , such that rca) ¢ 0, where a can be chosen asarootelement
132 Chapter 4
Lemma 2. Let f be a weight element in C1-tCn,9), where I ~ 2, and letdf be a nonzero

element 012 2Cn,9) that takes Its vaJuesin n. Then there exists e a E n 1 such that either
fCe a) - ae_a or fCe a) - ae_~, where e~ E n, a + ~ E ~ and a E C.
Proof. By Lemma 1, there exists an e a E n1 such that fCea) "# o. Since f is a weight map
In C 1-1Cn,9), we have fCe a) - ae_~ for some ~ E nl_1 and for some nonzero a E C . By
the hypothesis of the lemma, dfCea,e~ E n. On the other hand, df E a2_ICn,9) and
dfCe a . e~ E no = s. which is not contained in n. Therefore, dfCe a . e~ - 0, and so
f[ea,e~]-[f(ea),e~]-[ea,f(e~)]-O .
If a + ~ ~ L1, then - a [e_~,e~] - [ea,f(e~)] - 0 • Since the first tenn on the left hand
side of this equality Is not equal to zero and lies in fI., the second tenn must also lie in fl..
Consequently, f(e~ - be_a for some nonzero b E C. We have ahp - bha , where hy
denotes the element leY' e_yl in fl., 'Y E L1. But this equality Is possible only If a - ~ and
a-b. This proves the lemma.
Corollary 2. Let f be a weight cocyc1e in 21 -I(n,9/n) where I ~ 2. Then there exists

an e a E n 1 such that either f(e a) - ae_ a , or f(e a) - ae_~ where e p E n, ex + ~ E L1
and aE C
Lemma 3. Let f be a map in C 1.<n,9), where I ~ - 1. Then df Is a cocycJe In 2 2(n,9)

that takes Its values in n.
Proof. Itis c1earthatf (F2n) eF1n - nandf (n1) eno - s. Therefore, df(e~l,e~2) En
for any e ~1 ' e ~ En. This proves the lemma.
This shows that the space 21fn,9/n) Is perfectly described.
Now we construct some cocycles in the space 21_ 2(n,o/n) which, as will be proved in
the follOwing subsections, generate this space.
Lemma 4. Let e a E nl be such that ,for any root y E do, we have a + y ~ d, and let
fEe lCn,9/n) be such that fCe a) - e_ a and fCe~) - 0 for any~, ~ =F- a. Then
f E ZlCn,o/n).
Proof. It is enough to verify the containment dfCev,e,;> E n, in the case when one of the
roots is equal to a Cif v = 't = a, then obviously df(e a , ea) = 0). We have
dfCea,e,;> - [e_a,etl .
The element [e_a,etl can fail to lie in n only when e t E n l and't - a E d. Let't - a - y.
Since e t , e a E n l , we have y E do and a + y - 't E d, which contradicts the hypothesis
of the lemma. This proves the lemma.
Remark 6. For the hypothesis of the lemma to hold, it is necessary and sufficient that
a E S- 51 and that all the simple roots connected with a in the Dynkin diagram also
lie in S-SI' In the case when n is the nilradical of a Borel subaigebraCi.e., in the case
when 5 - SI - S), the hypothesis of the lemma holds for all elements e a E n l .
Lemma 5. Let e a E n 1 be such that the set Ra of roots y E do with a + y E dis non-
empty, where for any yE Ra we have 2a+ yE d, and let f E c l (n,9/n) be such
that
f (e a) - e_a, f (e2a+y) - ~. [e_a,e a+y]
a,a+y
with f (e ~) - 0 for ~ =F- a ,2a + y. If 2a + y = a + (a + y) is the unique decomposition i
into the sum of two roots in dl forany y E R a , then f E ZI(n, 9/n) .
Proof. It is enough to show that dfCev,e,;> E n in the case when one of the roots v or 't
Csay v) is equal to a. We have
If e t E nl, where i ~ 2, then obviously df(ea,e t ) En. But if 't E dl and't =F- a + y,
where y E do, then df(ea,e t } = 0 E n. Let't - a + 1- Then
df(ea,et) - Na,a+yf(ea+t)-[e_a,et]
by the hypothesis of the lemma. This proves the lemma.
134 Chapter 4
Remark. In the case when fI. is the nilradical of a Borel subalgebra, there do not exist
elements e a E fl.I that satisfy the hypothesis of Lemma 5. Such elements also do not
exist in the case 9 - slCfI.,C).
Lemma 6. Let e a ' e~ E fl. I, a+ ~ E A, be such that ~ + "( eo A for any "( E Ao, andlet
f E CI(fI.,o/fI.) be such that
f(ea) - e_~ , f(ea+~) - t-. [e_~,e~] ,
a,~
with f(e~ - 0 or a ¢ a, a +~. Then f E ZI(fI.,o/fI.).
Proof. A direct check shows that df(ea,e~) - 0 E fI. . If neither of the roots 't or v is
equal 10 a or ~, then obviously dfCe't,e) - 0 E fl.. It remains to examine the case
case when t or v (say t) is equal 10 a (or ~) . We have
dfCea,e) - - [e_~,e).
This expression can fail to lie in fI. only in the case when e y E n I and v - ~ - "( E Ao .
But this case is impossible Since, by the hypothesis of the lemma, ~ + "( eo A for any ~
where "( E ~ . This proves the lemma.
Lemma 7. Let e ex ' e~ E fl.1' a + ~ E A, be such that the set Rp ofroots "( E Ao with
~ + "( E A is nonempty, where for any "( E RW we have a + ~ + "( E A (except,
perhaps, for the case a - ~ + "(J, and let
f(ea) - e_~ , f(£ex+~y) - ~. [e_~,e~+y]
ex,ex+y
with f(e o) - 0 for a¢ a, a+ ~ + "( where "(E R~ U {O}. If a+ ~ + "(- a+ (~+"() is the
unique decomposition into the sum of two roots in Al for any "( E R~, then we have
fEZ I(fI.,o/fI.).
Proof. It is clear that df(ey,e~ - 0 for all v and 't, neither of which is equal 10 a or
a + ~ + "( , where "( E ~. Let v - a and 't ¢ a + ~ + ,,(, where "( E R~. We have
If e't E fl.I ' where i 2: 2 , then obviously dCea,e) E fl.. But if e't E fl.I and't ¢ ~ + "( , where
"( E ~, then f(ea,e) - o. Lett - ~ + ,,(for some "(E Ao. In this case also
CohomoIogyof some nlpotent Lie algebras 135
d(ea,e~y) - Na,~yf (ea+~y) -[e_~,e~y] - 0

by the hypothesis of the lemma. In the remaining case, v - a + ~ + y , 't * a, we have
d(ev,e·tl- c [e."e't] e 11.
for some constant c. This proves the lemma.
Remark. In the case when 11. is the nUradical of a Borel subalgebra., Lemma 6 is satisfied
by any simple roots a , ~ e S, connected by an edge in the Dynkin diagram, and there
do not exist roots a and ~ that satisfy Lemma 7.
Lemma 8. Let be e a e 11.1 be such that the set Raofroots ye .10, where a + yea,
Is nonempty but there exists y e Ra such that 2a + y ~ .1 and let f be a weighted
element In c1(n.,91n.) such that fCe a ) =ae a , a* O. Then f ~ Z1(n.,9/n.) .
Proof. We choose a root y e .10 so that a + 'Yea but 2a + y ~ l!.. We have

df(ea,ea+y} - a[e_a,ea+Y] - [e a, f(ea+y}] .
If fCe a+.( * 0, then, bearing in mind that f has weight - 2a, we have - a + yea and
fCe a +.( - be_a+y for some b * O. But this contradicts Lemma 2, since Ca + y) + Ca - y) -
2a is not in .1. Therefore, fCea+-? - 0 and
df(ea,ea+y} - -aN_a,a+yey* a ,
where e y ~ 11.. This proves the lemma.
Lemma 9. Let e a e 11.1 and e pen., a + ~ e a, be such that the set R p of all the roots
ye U.11 with ~ + ye a' Is nonempty, but there exlstsye R~ such that a+{3 +r ~ .1,
ISO
a * ~ + Y andlet f be a welghtelement In Z1(n., 9/11.) such that fCea) - 3e-IJ' Then
f e B1(n., 9/11.).
Proof. We consider an element g - b ad e-aL-~ in Bl(n,lJIn.) and we choose b so that

g(ea}-f(ea}.Fortheelement fl-f-gandforaroot ye R~ with l3+ye .1andalso
~+ a + y ~ .1, a * ~ + y, we have
136 Chapter 4
elf l(e a,ej3+y) - - cN a,-a+y e y e: n,

whence f e: Z I Cn, 9/n ) , which contradicts the hypothesis. This proves the lemma
(elf - elf1) in the case under consideration.
Let f le13+1 - O. Since a + C~ + y) - Y= a + ~ E .1. and a + (~ + y) e: 8., we have a - y E .1.

and
df'(e a_y.ej3+y) - Na-y,l3+y f'(ea+~) -[f'(ea_y),e~+y] ,
df'(e a,ej3) - Na,j3 f'(e a +j3) -[ea, f'(e~)] .
If at least one of these expressions is nonzero, then f' e: Zl(n,lJIn) ,which is impos-
sible. Therefore, both expressions are zero. If f' (ea+~)"* 0, then
f1(e a_y) a be_l3+y"* 0, f1(e~) - ce_a "* 0,
and we have
[e+y, el3+y] - N[ea, e_a]
for some nonzero number N, i.e., hj3+y - - N ha where a"* ~+ y, which is impossible.
Therefore f'Ce a +j3) - 0, whence f' Ce~), f'Cej3+1 - O.
Let us assume that f'Ce e) -le_v for some ee E n 1, where 1"* O. Since f'Ce a) - f'Ce~) - 0,
we have 9 "* a, ~. Since f is a weight element, we have 9 + v = a + ~ and a + ~ - 9 = V
E A By jacobi's identity, either a - 9 E .1. or ~ - 9 E d.

In the first case, 9 - a + 't and v = ~ - 't for some 't E do, and we have
df'(ee,e v) - -1 [e't_j3 , e~_'t] - [ea+'t, f'(ej3_'t)] .
This expression cannot be zero, since hl3--'t + cha+'t "* 0 in view of the fact that we have
~ - 't "* a + 't Cotherwise a + ~ - 2Ca + 't) E .1., which is impossible). This means that
e1f'Cee, e) e: nand f e: ZlCn, 9/n). We have a similar situation in the second case. Thus,
if for some ee E n 1 we have f'Ce e) "* 0 , we obtain a contradiction. Let f'Ce e) - 0 for all ee
E n 1. According to the corollary of Lemma 1, f' - o. This is eqUivalent to the fact that
f - g E B1Cn, 9/n). This proves the lemma.
Lemma 10. Let e a E n 1 and e~ En, a + ~ E .1., and suppose that there exists
yE (~dl) U{OJ such that~ + y, a+ ~+ y E .1. and a+ ~+ y- 't+ v forsome't and v
in A' different from a and ~ + y. Then anycocyclef in Zl (n,9/n) such that f (ea) - e_j3
is cohomologous to zero.
CohorooIogy of some nilpotent Lie aIgebtas 137
Proof. We first consider the case when a. + ~ - 't + V , where 't, v E d' ,'t *- a. , ~, and
we consider an element g - a ad. e_a_~ in B1(n, 9/n) ; we choose the number a so that
f(e~ - gCe~ (it is clear that 't == a + ~ - v). For the cocycle f' - f - 8, we have
f'(e t, e v) - Nt,v f'(ea+~) -[e t , f'(e v)] .
Since f' is a weight cocycle With weight -a. -~, we have

f'(e) - be--'t - h
f'(ea+~)
for some b E CC and h h. Thus, the element f'(et,e), which lies in n, lies in h as well
E
and is therefore equal to zero. This is possible in only two cases:

(a) f'(ea+~) - f'(e) - °;
(b)
f'(ea+~) - Nht ' N E CC •
In the case (a), for any 9 E d 1 we have f'(e a) - 0, and by Lemma 1 we have f' - 0, i.e.,
f- g E B 1(n, o/n). For if we assume that f'(e~ *- 0, then J.l == a + ~ - 9 E d' and we have
f'(ea,e l1) - NIhil + N:ila E 11- n h- °
for nonzero N1 and N2, which is impossible, since J.l *- 9 .
In case (b), we have
f'(ea,e~) - Noh t - N'h~ - N'ha
for some constarlts No, N' and N". This expression lies in n n h and is therefore equal
to zero, i.e.
(1)
for some nonzero numbers N 1 and N2 (if, for example, N1 = 0, then h t - N2ha' where
t *- a., which is impossible).
Similar arguments for v instead. of 't lead. to the fact that either
f E B1(n,9/n), or
hv - N3h~ + N4 ha (2)
for some nonzero numbers N3 and N4.

We can assume that the root vectors in the Lie algebra are chosen so that
n
ha - l: kl(a)ha.
1-1
,
(3)
if
138 Chapter 4
n
L kt(alat ,
a - 1-1 (4)
is a decomposition of the positive root ainto simple roots (the kt(a) are nonnegaUve
integers). It follows from (1)....(4) that
t - NIP + N2 a, v - N,P + N4 a ,
where, since t + V - a + P and t c# a ,p, we have N, - 1 - N1 and N4 - 1 - N2 .
In addition, the numbers NI , N2 , N" andN4 are different from zero and one, and we
can assume they are rational.
Let us show that all the simple roots in the decomposition of a + P into simple roots
occur with mUltiplicity ~ 2. To do this, we consider several possible cases :
(1) O<N I , N2 < 1 . In this case, 0<N"N 4 < I, andtherootstandv,each

separately, con1ain all the simple roots in the decomposition of a + P . This means that
their sum t + V - a + P con1alns all these simple roots with multiplicity ~ 2.
(2) 0 < NI < 1 and N2 > 1 All simple roots having mUltiplicity 1 in the
decomposition of P are con1ained in a as well (otherwise these simple roots would
occur in the decomposition of t with nonintegra! coefficients, which is impossible). In
addition, since 0 < N, < 1 andN4 < 0, from the second equality in (.5), all the simple
roots in the decomposition of a are con1ainedin P (since v is a positive root, all the
simple roots occur in the decomposition with mUltiplicity ~ 0). The simple roots that
occur in the decomposition of P but not in the decomposition of a, must have
mUltiplicity ~ 2 in P (otherwise, since 0 < N I < I, these roots would occur in the
decomposition of t with nonintegra! coefficients). Thus, a + P con1ains all simple roots
in its decomposition with multiplicity ~ 2.
(3) N1 < 0 and 0 < N2 < 1. In this case N, - 1 - N I > 1 and 0 < N" < 1, and, by
analogywith case (2), all the simple roots in the decomposition of a + P occur with
multiplicity ~ 2.
(4) N1 < 0 and N2 > 1. In this case from the first equality in (.5), all roots in the
decomposition of b occur in the decomposition of a. Since N, > 0 and N" < 0, by the
CohoImIogyof sane nIpotent Lie aJgebtas 139
second equality In (13), all roots In the decomposition of a occur In the decomposition
of 13. Thus, a and 13 have the same simple roots in their decompositions Into simple
roots. Consequendy, they all occur with mUltiplicity ~ 21n a + 13.
The cases N l' N2 > I and N l' N2 < 0 are impossible, since 't, v < a + 13 (with respect to
the partial order introduced in h.") and since 't, v E A+.
We note that there existthree cases In all when a root from A+ in the decomposition
Into simple roots contains all slnlple roots with multiplicity ~ 2 :
m 9 - E8 anda+ 13= 8is a maximal root;
CiO 9 - FIj and a + p = 8is a maximal root;
CliO 9 - G 2 and a + P= 8 is a maximal root
In all cases a slngularroot (see, for example, [KH2]), i.e., a simple root connected with a
minimal root in an extended Dynkin diagram), occurs in the decomposition of 8 with
multiplicity 2, but all other positive roots occur with multiplicity ~ 1. This means that
in the decomposition of the roots, a, 13, 't, and v, a singular root occurs with
mUltiplicity 1. This is possible only In the case when we have N 1 + N2 - 1 in (5).
Since N1 ¢ N2 (otherwise a + 13 - 2't E A, which is impossible), It follows that
Nl N2 I-N1-N2¢O
II-N1 I-N2
and we have
13 - N't 't + N' 2 V a - C1 - N'l) 't + C1 - N' 2) v , (6)
where N' 1- I-Nl N'2 - N1 N'l + N'2 - I

N2 -N1 , N 1-N2 ' .
Let us consider a simple root in the decomposition of 8 that occurs with mUltiplicity 3
(such roots exist in all three cases). This simple root ao can occur in the
decomposition of a and 13 with multiplicities 0 and 3 or 1 and 2 (up to rearrangement).
In the first case, 't contains (in the decomposition into simple roots) <Xo with
multiplicity lor 2 (multiplicities 0 and3 are impossible, since N2 ¢ 0, 1). Thus, up to
replacing the equalities (5) by (6), we can assume that a and 13 contain ao with
multipliCity I and 2, respectively. Then the number N 1 + 2N2 , which gives the
140 Chapter 4
multiplicity of the root 0.0 in the decomposition of't, can be equal to 0, 1, 2, or 3. Since
N1 + N2 - 1, the only possible case is N2 - 2, N1 - - 1 (or N2 - -1, N1 - 2). Assuming
'3 "# G 2, we consider a simple root ~o in the decomposition of () occurring with
multiplicity 4. Denoting the multiplicities of the occurrence of ~o in the
decomposition of ~ and <X. by x and y, respectively, we have
-x+2y-ao, 2x-y-bo ,
where the pair (~, b o) can assume the values (4, 0), (3, 1), (1,3), and (0,4).
For all these values, system (7) does not admit integral solutions. We have concluded,
therefore, that case (b) is impossible if '3 "# G 2.
In the case '3 - G 2, the only situation in which (5) and (6) hold is
<X. - 0. 1 + 0.2 ' ~ - 2 0. 1 + ~ , 't - 0.2 ' V - 3 0. 1 + 0.2
(we use Bourbaki's notation [BOU31 for simple roots).
Butin this case, for 'Y - - 0. 1 we have ~ + 'Y E ~ and <X. + ~ + 'Y ~ ~,andby Lemma 9,
f' is a nonzero element in H1(n, .yn).
Thus, we have proved the lemma in the case when <X. + ~ = 't + v, where 't, v E ~. and
't "# 0., ~.
Now let <X. + ~ + 'Y - 't + v, where 'Y E U1g) ~I and 't"# <X. , ~ + 'Y but the root <X. + ~
cannot be represented as a different sum of this type. We consider a cocycle g ,
g - a ad e-a_p in Z len, '3/n) cohomologous to zero and choose a such that
f(ep+y) - g(e a +y). For the difference f' - f - g we have
f'(e,(,e v ) - N 1f'(e a+p+y) - [f'(e,(), e v] - [e,(, f'(e v )] - 0
If fl (e'() "# 0 (or fl (e v )"# 0), then - <X. - ~ +t - - A. E ~ and f'(e'() - ce_a_p+,( for
some nonzero c E (C. This means that <X. + ~ = 't + A. is a different representation of
<X. + ~ as a sum of two roots in ~, which is impossible by hypothesis. Therefore,
and we have
CohorrcIogyof some nHpotent Lie a/gebIas 141
which is impossible only if f'(e a) - O.

For any 9 in .1.1 different from a, we also have f'(es) - 0 (otherwise A - a + ~ - 9 E Il'
and a + ~ = 9 + A, which is impossible by hypothesis). According to Lemma 1, the

cocyc1e f' is equal to zero ; i.e., f - f' + g lies in B1(n, 'lIn). This proves the lemma.
Corollary. H1~n, o/n) - 0 if i < - 2.
Proof. Let f be a nonzero (weight) cocycle in Z\(n, 'lIn) where i < - 2. According
to Lemma 1, there exist a E III and ~ E UI~IlI, a + ~ E Il, such that f(e a ) - ae_~,
a;t: o. We consider the decomposition ~ - ~1 + ~2' where ~1' ~2 E Il'. If one of the ~1
or ~2 (say ~1) is equal to a, then ~ - P2 - a E Il' although a + ~ - ~2 - 2a eo Il.

According to Lemma 9, f is cohomologous to zero, and in this case, the coroI1ary is
proved. Butif neither ~1 nor ~2 is equal to a, then either a + ~1 E Il or a + ~2 E
Il (since one has a + ~ - a + P1 + P2 Ell). Setting 1: - a + P1 and v - ~2 in the first case

and 1: - a + ~2 and v - ~1 in the second case, we can apply Lemma 10. This proves the
coroI1ary.
The proof of the foI1owing lemma is similar to the proof of Lemma 10, and we omit it
Lemma 11. Let e a E n1 be such that the set R a of roots y E .1.0 with a + y EllIs non-
empty, where for any y ERa' we have 2a + YEA, and we suppose that there exists Y
E Ra u {OJ such that2a + Y- 1: + v for some 1: and v in III different from a and a + y. If
f E Zl(n, 9/n) is a weightcocyde such thatf(e a ) - e_ a , then f E B1(n,9/n).
III.2.2. Description of the space Hl(n, o/'n)
For any root vEA l ' we set

Ry-fyE Ilo/v+YE ~
and we introduce the follOwing notation:
no=((a,hs)/aE III , hs-[es,e_s] , 9E S},
n1={(a,-~)laE III , ~E Ilo},
142 Chapter 4
n 2 ={(a,-a)/aE..1\, Ra- 0 },
(a,-a) I a E..1\ , Rg ¢ 0 , 2a+1 Ed , 2a+1 ¢ 't+v }

{
n3 = where 't, v E ..11 and 't ¢ ex, a+y for all 1 E Ra ,
n4={(a,-~)/a,~E..1\ , a+~E..1 , RI3=0},

(a,-~)/a,~E..1\ , a+~E..1 , RI3¢0, a+~+1E..1 )
n5 - (
- a+~+y¢ HV where 't,v E ..1\ and 't ¢ a, ~+1 for all 1 E RI3 .
Remarks. 1. In the case S - Sl - S (I.e., in the case when n is the nUradJcal of a Borel
subalgebra), we have n 1 - n, - Os - 0. In addition, the pair (a,-a) lies in ~ if and only
if a E S, and the pair Ca, -~) lies in !l.t if and only if a and ~ lie in S and are connected
by an edge in the Dynkin diagram.
2. For classical Ue algebras I} the conditions 2a + 1 ¢ 't + v and a + ~ + 1 ¢ 't+V in the

definitions of ~ and Os are superfluous. They hold If the other conditions hold.
For each element 01 E ~, where o:s; i :s; 5, we consider the follOwing cochains
f OJ, in c 1 (n,91n) defined as follows :
f ,",Je~)
...,~ . _ { h a, if 't - a } , where 0)0 - (a,ho) E no ;
0, if't;t: a
e-j3, if 't - a
If 't¢a
CohorooIogy of some nilpotent Lie aIgebtas 143
e~ If't-a
0, If 't * a, 2a+y
e--J3' If't-a
0, If 't * a, a+~
if't-a
,where 0)5 - (a, -~) E Os;
0, if 't * a, a+~y
Let
5
where 1 ~ i ~ 6, and let E - UI-O EI . It follows from Lemmas 3 - 7 that each element of E is
contained in Z l(n, IJ/n) and Is not cohomologous to zero. These elements define
nonzero elements In Hl(n, IJ/n). In what follows, for simplidty, we shall assume that
E c H~n., rI n} and identify cocycles that differ by an element in B~n., rI n} .
Theorem 3. Let n be the nllradical of a paraboliC subalgebra p of a complex

simple Ue algebra IJ definedbymeansofa subsystem Sl c S. Then Hl(n., rln) - <E>.
144 Chapter 4
Proof. Let 9 be a nonzero weight element of weight A. in HI(n, 9/n). According to

Lemmas 9 and 10 (see also the corollary to the latter),9 lies in either HI_I(n, 9/n) or
HI_ 2(n,o/n) . In the first case, obviously, 9 E <Eo, El> and the theorem is true.
Let 9 E HI_ 2(n, cy'n) . We consider a root vector e a "f;. 0 (the existence of such an
element follows from Lemma 1). Then 9(e a) - ae_~ for some nonzero a E c: and for a
root ~ E d l with - a - ~ = A.. By Lemma 2, two cases are possible:
(a) ~ = a;
(b) a + ~ E d.
Ca) According to Lemmas 8 and 11, this case can also be divided into two subcases :
(al) Ra - 0;
Ca2) Ra "# 0, but 2a + 'Y E d and 2a + 'Y "# 't + v, where 't, v E d l and 't, v "f;. a ,
a + 'Y for all 'Y ERa·
In subcase (al), we choose an element f in <E2> such that f(e a ) - ae_a> and we
consider the difference g' - g - f .
In subcase~) we choose an analogous element f in < E3> and consider the difference
g-g- f . In both subcases, g is an element of Hl(n, 9/n) that vanishes at e a and
continues to vanish at those elements e t E nl at which g vanishes.
(b) According to Lemmas 9 and 10, this case also splits into two subcases, and in each
of these subcases one can choose an element f in <E4> or < E5> such that the
difference g' - g - f vanishes at e a and continues to vanish at all elements e t E n l for
which gCeJ - o.
Thus, in all cases, the number of root vectors in n l at which g' vanishes is one less
than the same number for g. If g'(eJ "f;. 0 for some e t E n l , then, repeating the
argument, we can find an f 1 in E2 u E3 u E4 U ES for which g" - g - f 1 vanishes at e"C.
Continuing in this way, we find an element ik) - ik-n - fk_l in HI(n, 9/n) that
vanishes on all of n 1.
According to Lemma 1, this is possible only if ik) - o. This means that g can be
expressed as a linear combination of elements in E2, E3• E4• and E5. This proves the
theorem.
CohonrJIogyd some niIpoIerI Lie algebras 145
111.2.3. A Basis of the space H1(n, c.:J/n)
The elements of E, as shown in Theorem 2, are linear generators of the space

H1(n, .yn). In general, they can be linearly dependent in H1(n, c.:J/n), although as
elements of Zl(n, c.:J/n) they are linearly independent (as is easy to see). Below, we
shall prove several assertions that make it possible to find a basis of H l(n, c.:J/n) and
sharpen Theorem 3.
We consider a partition E - u EAo of E into nonempty subsets EAo consisting of
mappings with the same weight A..
Proposition 7. (a) If A. Is not a root, then EAo Is linearly Independent In H l(n, c.:J/n)
(b) If A.Isarootln A_1' then the system EAoIsJinearlydependentln H1(n, c.:J/n) .
(c) If A. is a root In ..:1_2' then EAols linearly lndependentln H1(n, c.:J/n) If and only If
MAo - a" u Ll:s where MAo - {(a. - 13) / a.,f3 E ..:11 ,A. - -a. - f3}.
Proof. (a) The weight elements of B1(n, c.:J/n) obviously have the form g - a ad e a '
where a. E u,so ..:1,. Therefore, any nonzero linear combination of elements in EAo
cannot lie in B1(n, c.:J/n) ,since A. G!: ..:1.
(b) Let g - adeAo E B1(n, c.:J/n), where A. E ..:1_ 1. For any root a. in ..:11 for which
a. + A. - - 13 E ..:1, the pair (a.,-f3) lies in 0 1. Therefore, one can find a linear combination
f of elements in Eo UE1 such that the values of the maps f and g are equal at all root
vectors e a E n 1 . We consider their difference f' - f -g, whichliesinZ 1(n, c.:J/n). The
value of f' at any element in n 1, by the choice of f', is equal to zero. According to
Lemma 1, this is possible only if f'(n)cn, i.e. if f' - o. Thus a nonzero linear
combination of elements in EA.lies in B1(n, c.:J/n) , i.e. is equal to zero in H1(n, c.:J/n).
(c) The proof of the sufficiency is completely analogous to the proof of (b). Let us
prove the neceSSity. If MAo ~ Ut u Os, then there exists a pair (a.,-f3) in MAo' where
A. - -a. - f3 , that does not lie in Ut u Os. For this pair, we have fi,.e a ) ~ N e~ ~ 0 , and
any nonzero linear combination f of elements in E1 is equal to zero at ea. Since any
weight element of Weight A. in B1(n, c.:J/n) is a multiple of g,f cannot lie in B1(n, c.:J/n);
that is, EAo is linearly independent This proves the lemma.
146 Chapter 4
Corollary 1 (from the proof). If A is a root and EA is linearly dependent (J.e., either
AE d..1' or A E A-2 and MA - 0.. u as), then :
(a) If A E A-2, any collection E' A that l1es properly in EA is linearly Independent
(b) If A E A_I any collection E' A'::) Eo n EA that lies properly in EA is linearly
Independent.
Corollary 2. If n is the nilradical of a Borel subalgebra. in a complex simple Ue

algebra. 9, then 0 1 =a, =as =0, and so Et - E, - Es - O. In addition, any weight
Aof acocycle inHI_ 2(n, o/n) which isaroot is equalto the sum of two slmple roots.
Consequendy, MA - 0...
111.3. particular case : n is the nilradical of a Borel subalgebra
where f Is an arbitrary chosen element of EA'

The next theorem follows from Proposldon 7 and Its corollary.
Theorem 4. Let n be the nllradicaJ of a paraboliC subaJgebra. of a complex simple

Ue algebra. 9 defined by a subsystem SI c s, and let A be the set of weights (with
respect to the adjoint action of h) of the cocycles in E. Then EI - UAeA E I A Is a basis
of the space HI(n, o/n).
Corollary 1. If 9 is a simple Ue algebra. of rank n and n is the nllradicaJ of a Borel

subalgebra., then dim Hlen, 9/n) - n 2 + n - 1 .
Indeed, In this case A - (-S) U A_2 U (-2S) and

147
n, If A. E - S, n-l, If A.E -S,

IE~ - { 2, if A. E a_2,
1, If A. E - 2S,
}, IE I ~ - { 1, if A. E a_2,
1, If A. E - 2S,
}
and we have dim H1{n,o/l1) - n(n -1) + (n -1) + n - n 2 + n -1.
111.4. Particular case: 11. is an Heisenberg algebra
Let Hk be an Heisenberg algebra with k 2: 2. It is isomorphic to the nilradJcal n of a

parabolic subalgebra of a simple algebra 9 of type Ar (r - k+ 1),· defined by
Sl - (a.1'a.~·
It is easy to verify that, in this case, we have
~=n~=~=n~=0.
Then the space H1(n,IJ/n) is generated by the vectors of Eo u E1. As

cardinal Cal) - 2Cr-l),
cardinal Cao) - r2 - 3r + 2,
cardinal (S) - r,
we have
cardinal (E) - 2.-3 - 6r2 + Sr - 4 .
We deduce
cardinal (U).eA E I~.) - (2r3 - 6r2 + Sr - 4) - 2{r-l} - 2r3 - 6r2 + 6r - 2 ,

and
dim H1(n,IJ/n) - 2r3 - 6r2 + 6r - 2 .
111.5. Description of the space H2fond (n,n)
This description is based on the study of the s-module Hl(n,V), where V is a simple
9-module whose structure has been given in Kostant's works (see [KOD.
148 Chapter 4
m.S.l. Description of the s-module Hi (n,V)
Let n be the nilradical of a parabolic subalgebra of the simple complex algebra fJ,
defied by SI c S. Consider a simple fJ-module V of dominating weight g (see Chapter
1) and the corresponding Weyl group of fJ noted W.
We have seen that the spaces cl(n,V), Zl(n,V), Blen,V) and Hl(n,V) are provided to
structures of s-modules, where s is the LevI subalgebra of fJ given by
We put
U a - a(A-) n A+, where a E W ,
W1- { a E W I U a c A' - ~ AI} ,

w J l - wJnWl,
P-~L'Y·
yeA
We have
HJ(n,V) - L Wa,
aeW.
where Wa Is the simple s-module with domlnatlng weight ;a = a(p + A.) - P (A. is
the weight of V).
Let vo(J..) be a vector of V which Is nonzero and which has a weight equal to 00..) and let
(al ,... , ay be the ordered elements of U(J"
Then the cohomology classes of the cocycIe fa defined by
f a{X al ,..., XCIj) - va(A.)
is a primitive vector (I.e. anonzero vector corresponding to the dominating weight) of
the s-module W(J"
CohorTKJ/ogyof SO/7I9 nipotent Lie algebras 149
Remarks. 1. For the description of these structures, Kostant shows that the
dominating weight ~ - O<p+/J - p of the s-modules Wa are all different from each
other and distinct to the dominating weights of the simple components of the s-
module &n,V).
2. As ZJ(n, V) is a semislmple s-module, there is a supplementary s-module to the

s-module &n.,V). This can be identified by HJ(n.,V).
m.S.2. The space H2fond{n.,n.)
We recall that aroot a is singular if 3 - a E S, where 31s a maximal root. The simple tie
algebras of type Ar (with r > 2) owns two singular roots denoted a l and a r in
Bourbaki's terminology. The other types have only one singular root. But in these
cases, the mUltiplicity in the decomposition of 3 of the singular root is equal to two.
Now If ~ is a positive root, then 3 - Pis a root if and only if the singular root (or one
singular root in the Ar case) belongs to the decomposition of p, and Its multipliCity is
equal to 1 otherwise ~ - 3.
Following is the table of maximal roots, of singular roots and the expanded Dynkin
diagram for each type of simple complex tie algebras. Also we precise a number y(9)
which we wID use later.
150 Chapter 4
TABLE 1
Type of Singular
Dyukin extended diagram Maximal root y(g)
algebra roots
An
(n ~ 2) ~
al a2 On-Ian
al +a2+···+a" Ql,On 5
,
al a2 a3 0,,_1 an
Bn
(n~ 3)
L---~ al + 2a2 + ... + 2a" a2 3
-6
-6 al a2 Qn-lOn
en o::::;x:o---o- -- ~ ~ 2al + ···+2a,,_1 +a" al 4
(n ~ 2)
(n~4)
Dn
T---<::::-'-6
al + 2a 2 + ... + 2a n_2+
+an_1 +an
a2 3
al a3 a4 aG as
Es al + 2a2 + 2a3 + 3a4+
~
a2 2
+2a5+ as
-6 al a3 a4 aG a6 a7
E7 2al + 2a2 + 3aa + 4a4+
"
~
al 2
+3aG+ 2a6 + a7
la2
al a3 a4 0'5 a6 a7 as
Es ,.. ,.. ,... ,... 2al + 3a2 + 4a3 + 60'4+
I
"
~ "
~ ~ ~
f'o
as 2
+5a5 + 4as + 3a7 + 2as
la2 -6
-6 al a2 a3 a4
2al + 3a2 + 4a3 + 2a4
F4 ~ al 2
al a2 -6
G2 (;E3I$O--O 3al + 2a2 aa
CchomoIogyof some nlpotent Lie aJgebtas 151
We need some notions which were Introduced before.

W2 - raE W / card. Ua - 2}.
ThIs set Is composed of the elements a of the Weyl group which are written as
CJ - SP1,s1iz where Ih, ~2 e S, ~1 *" ~2
<here sp designates the associated reflexlon to the root ~).

As for the elements of W21 - W2 n WI, they are of the fonn a - s PI 0 s liz' where the
the roots ~1 and ~2 satisfy one of the conditions
ea:> 13 1 + 132 eo & , 13 1, 132 e S - SI'

Cb) ~1 + ~2 e & , ~1 e S - SI .
In the first case, U a - {~1' ~2} andln the second case U a - {~1' ~2 + kI3 1}, where k Is
the largest Integer such that ~2 + k13 1 e A.
Proposition 8. The cardJnaJ of W21 ls equal to
r(r-l)
- - + 2r - tl - t2 + t,
2
where tl ls the number or roots of S - SI whlch are on the boundazy of the Dynkin
dlagzam, ~ the number of pairs of roots of S - SI whlch are fastenedln the Dynkin
diagram and t, - 0 for the Ue aJsebras IJ of type A r, Br' Cr' 02' F4 and t, - 1 for the
other types.
The proof Is obvious after we have noted that If ~1 and ~2 satisfy the conditions of case
Ca), the reflexlons s PI and s liz commute.
Let W2 x be the subset of W21 defined by

W~ - (a e WZtl a(a) e &" •
In particular, If a e W2nO the vector Xa(8) e n, where aIs a maximal root

Lemma 1. If IJ ls of type Cr or Ar (r it 2) andJf S - SI contBlns only one sJngular root
orJf IJ - ~ and S-SI - {a.l'~} then, W2 n - 0.
152 Chapter 4
Proof. We begin with the case 9 - A2.

Then W21 contain only two elements sal 0 SIJ;l and s IJ;l 0 Sal which, when applied to
the maximal root a - a1 + ~, give a negaUve root So w2 n - 0.

In the other cases, W 2 1 - { S ~l 0 s \J2} , where ~ 1 E S 1 is a singular root and ~2 is fastened
to ~1 in the Dynkin diagram. This set is reduced to only one element (see Table D. We
have
a- ~ 1 if 9 - Ar r > 2
a-2~1 If 9-Cr r2:2
}.
As the roots a - ~1 (In the cases 9 - ~) and a- 2~1 (In the other cases) don't contain In
their decomposition elements of S - St. then SPl oS\J2(a) f S. Thus, W2n. - 0.
Lemma 2. If 9 - Ar(r 2: 3) andlf Sl - {a 1,<Xrl then

W1t. - {sal 0 Sa:z , SUr 0 SUr_l} .
The proof Is established by considering the tables of roots before the description of
roots tied to the elements of W2r We use same considerations for the follOwing
lemmas :
Lemma 4. If 9 - Cr Cr 2: 3) and If S - Sl - {al'~}' then

W~ - {sal 0 slJ;l , Sa:z 0 sa,} .
Lemma 5. In all the other cases, W2 n. - W2 1.
We can deduce the follOwing two theorems.
Theorem 4. Let n. be the nilradica1 of a paraboliC subalgebra of a simple complex Ue

algebra which don't belong to Table 2. Then
CohorooIogy of some nilpotent Lie algebras 153
TABLE 2
9 S SI wt ~l j}z Y
A2 al,a2 0
al 0
An On
0
An aI, On IPal . IPa2 al al + IX2 0- al

~ IPa n • IPan-1 On On-I + On
0- On
C2 al,a2 IPal . IPa2 al 2al + IX2 IX2
C3 al,a2 IPal . IPa2 al al+ a2 21X2+Oa

IPa2· IPa3 a2 2al+ 21X2+a3
21X2 + Oa
Cn al.a2 IPal . IPa2 al al + IX2 21X2 + Oa

n~ IPa2· IPa3 IX2 IX2 + a3
0
Cn al 0
~
154 Chapter 4
Theorem 6. If 9 is in the list of Table 2, then the s-modu/e H2 fondCn,n) is the direct
sum of s-submodules generated by the cocycles VI of the following form :
(e~l' ep.z) ~ eJ .
Now, one fixes for each complex simple tie algebra the Integer Y(9) whose value Is
given In the Table I.
Theorem 7. Suppose that cardCS 1) > Y(9). Then

H2 o(n,n) - 0 , F/i2(n,n) == H2 fond(n,n) for 1- O.
Proof. Let 0" - S~l 0 Sp.z In w'2n , ~1 and ~21n S.

We have
Va - {~1' ~2 + k~I}'
O(~) - ~ - ~ ~2 - kl ~1' where kl and k2 are the greatest Integers such that
~2 + k~I' ~ - k2 ~2 ' ~ - k2 ~1 - kl ~1 are roots. By using the roots tables, we can show
that, if card(SI) > Y(9), then
ISl(O"(~» =ISl(~) -lsl(k2~2) -ISl(kl~ 1)
and
ISl(O"(~» >ISl(~l) +ISl(~2 + k~I)'
where lSI «X) appoints the S - S 1 length of (X • This means that
ISl(o(~)) > IsJ~tl + lSI

where ~1 and ~2 are the roots such that Va - {~1' ~2 + k~l} .
We deduces that
W~ - (0) and W~- U W~.

120
Now this theorem becomes a consequence of the two previous theorems.
111.6. The space H2(n,n): case of the nilradical of a Borel subalgebra
We suppose that n is the nilradJcal of a Borel subalgebra CS 1 - 121). In this case, the
previous theorem can be formulated more precisely.
CohomoIogyd some n4:lotetJt Lie a/gebIas 155
Theorem 8. Ca)If 9 is a simple Ue algebra of rank r whose type is different to AI for

I - 1,2, 3, 4,5, B" ell D" and G2 then H2 oCn,n) - 0, FoH2Cn,n) - F1H2Cn,n) -
H2 fondCn,n) and dUn H2 fondCn,n) - Cr2 + r -2)/2
(b) If 9 is not of type poinred In Ca), then the cocyc1es defJned by cp( epl,ep) - ey ,
whose llst is presenred in Table 3. determine the basis of H2ICn,n) and H2fondCn,n).
TABLE 3
IJ VI,II W" 1\ Pl y HtJ, n.n)
VI "'I-~ a, ex< 1X2I-CX3

v. "','liz. a, ar+a. a.-+aa+a. W!o~.nl
Y] ~'IIz, CJA (X)foCJA CZI-HlI2+CX3

A~
1IJ "',CIIl, CZI CX3 a.-HI3+ex<
1IJ ..... 1I\x4 a. a. al+U2I-U3

H~dn.nl
IIj 'Paz-lAx, a. U21-U3 Ii
III ",,.CIIl. U3 U21-U3 Ii
VI ",.. CIIlI a. atl-a. CX3+a<+US

H~.nl
v. ~CIIl, a. CX4I-a. a,+U2I-U3
1IJ "'1-CIIl. a, ar+a. a:IX3+1l11+115

1IJ lParlAx, a, a. a.-+aa+a. H~dn.nl
Us 'Pa,.1I\x4 a. a.+-a. a!CJo+CIII+a.
lit "',-'IIz, a, CX3 a:IX3+IlII+US
"' "'I-~ a, a. ~+ao+a.
A, '- lPa.-CIIl, a. Cl2tCX3 Ii

117 'Pa.. IAx, a. a. a!CJo+CIII+a.
H~.nl
1IJ ..... 1iIx2 U3 U21-U3 II
1.11 U3 as+a. II
"'''~
UID lPu,-lAx, U3 a. a!CJo+CIII+a.
UII 'Pa.-.a, a. ar+a. Ii
Uu lPa..1I\x4 a. a. Ii H~.nl
VI "',CRX. al ar+a. 2a.+CIII H~~.nl
c, 1IJ lPa,-1Ax1 U3 at 2a.+CIII

H~dn.nl
1IJ ",.. 'liz, a. 0I+2a. Ii
IIj lPu,-1iIx2 U3 CIII+U3 Ii H~.nl

156 Chapter 4
TABLE 3 (follow up)
9 VI, lit WZ ~I ~ Y H1I( n,n)
VI 'P()tZ·'P()t1 (Xz (Xl+(XZ 2(X3+<X4 H~o:(n,n)
UI 'P()tI·'P()tZ (Xl (Xl+(X2 2(X2+2aa+1X4 H~{n,n)
112 'P<X4·'P()t3 (Xl (X3 2(X2+2aa+1X4
II3 'P()tI·'P<X4 (Xl IX4 2<X2+21Xl+1X4 H~{n,n)

III 'P()t3·'P<X4 (X3 1X4+2(X3 a
C4
Il5 'P()tZ·'P()t3 (X2 (X2+(X3 a
Us 'P()t3·'P()t2 (X3 (X2+(X3 a H14f:n,n)
117 'P<X4·'P()t3 <X4 (X3+<X4 a
Us 'P()t2·'P<X4 (Xz IX4 a H1s{n,n)
VI 'P()tI·'PCX2 (Xl (Xl+(XZ (X2+2a3 H~o{n,n)
UI 'P()tZ·'PCXI (Xz (Xl+(XZ (Xl+(X2+2a3

H11{n,n)
112 'P()t2·'PCX3 (Xz (X2+(X3 (Xl+(X2+2a3
II3 'P0t1·'P()t3 (Xl (X3 a H~:(n,n)
VI 'P()tI"!P()tZ (Xl (Xl+(XZ (X2+2a3
Vz 'P()t3·!PCX2 (X3 (X2+(X3 (Xl+(X2+<X4 H1o:(n,n)
83 v3 'P<X4·!p()tz !l4 (X2+!l4 (Xl+(X2+(X3
Ul 'P()t2·!P()t1 (Xz (Xl+(X2 (X1+a2+IXl+1X4
112 'P()t2·!P()t3 (Xz (X2+(X3 (XJ+(X2+IXl+1X4 H11{n,n)
II3 'P()t2·!PCX4 (Xz (X2+!l4 (XJ+(X2+1Xl+1X4
III 'P()t1·!P()t3 (Xl (X3 a

115 'P()tI·!P<X4 (Xl !l4 a H13{n,n)
Us 'P()t3·!P<X4 (X3 !l4 a
Gz UI 'P()tI"!P()tz (Xl (Xl+(XZ (Xl+3a2 H1.:(fi,n)

Cohomology of some nilpotent Lie a/gebIas 157
Proof. Point (a) emanates direcdy from Theorem 3 (III.5.2) and from Proposition
(III.5.2).
Point (b) emanates from Theorems 1 and 2 CIII.5.2).
Conclusion. We have decomposed the space H2(n,n) as :
for every nilradical of a parabolic subalgebra (with some exceptions). But each spaces
H2 fond(n,n) and Hl(n,9/n ) is very precisely described in the previous sections. If we
concentrate this study on the case when n is the nilradical of a Borel subalgebra, and if
9 is not of the types indicated in the previous theorem at the point (a), we reclaim the
theorem of Leger-Luks.
Theorem 9.
dim H2(n,n) - } (r2 + r - 2) + r2 + r -1 - } (r2 + r) - 2
111.7. The calculus of H 2(n,n) when n is the Heisenberg algebra
We continue the description of the cohomology of the Heisenberg algebra worked

out in Section IlIA. Let Hk be the (2k + n-dimensional Heisenberg algebra. It can be
described as the nilradical of a parabolic subalgebra of the simple algebra Ak+l when
we choose SI - (Ul,Uk+l). We have seen that dim Hl(n,9/n! - 2r3 - 6r2 + 6r - 2.
This permits us to compose the dimension of the space H2 fond(n,n) from the
previous description,
H2fond(n,n) - W 01 EB W 02' where 01 =Sill 0 sll2 , 02 =SIXk+1 0 SIXk
as soon as k ~ 2 (see Lemma 2, III.52 and Theorem 1, 111.5.2).
Note that dim W01 - dim W02 •

158 Chapter 4
The dimension of the simple s-module W 0"1 whose dominating Weight Ais equal to
A - Co - a l ) - al - Ca l + ~) - 0 - 3a l - a2 can be comptuted from the Weyl fonnula
dim Val - IT CAY))

(1+ - '- -
kCk2_l)
')'Ed~ (p,y) 3
Then, we have
Theorem 10.
Remark. By a direct calculus, we can give the dimension of the space H2CH 1,H l ) Cwe
can also find the dimension of these spaces by considering H 1 as the nilradical of a
Borel subalgebra of A2. As for the algebra H2 , the calculus is also made using the Weyl
fonnula.
Theorem 11.
Proof. In effect we have dim Z2CH 1,H l ) - 0 , dim B2CH 1,H l ) - 3 and dim Von - 2 if
k- 2.
Cohomology of some nilpotent Lie aIgebtas 159
IV. DERIVATIONS OF SOME INFINITE-DIMENSIONAL

TOPOGICALLY NILPOTENT LIE ALGEBRAS
IV.1. Infinite-dimensional standard algebras
We can describe the algebra of derivations of asubalgebra of the Infinlte-dimenslonal

Ue algebra M composed of Infinite complex matrices A - Caq:> for I,j O!: 1 that contain
only a finite number of nonzero elements. First, we expand the notion of a Borel
subalgebra of cartan subalgebras for M.
Let B be the subalgebra of M composed of upper triangular matrices, I.e. the matrices
A- Caq) such that <1jj - 0 for I > j. This subalgebra Is called a Borel subalgebra of M.
Definition 2. A Ue subalgebra L of M is called standard if its nonnallzer

NorCL) - {X EM: [X,L] C L}
contains the Borel subalgebra B. The subalgebra. L is called a standard nilsubalgebra
If each matrix of L is nilpotent.
Note that a standard nilsubalgebra L of M Is not generally nilpotent In the classical

sense, I.e. the descending sequence Is stationary. Butevery element of L is a nilpotent
matrix. With an abuse of language, we shall speak of L as a standard nilalgebra.
N.2. A topology in a standard nil algebra
Let H be the subalgebraof M composed of diagonal matrices. It plays the role of a

Cartan subalgebra.
Definition 3. A root a of M with respect to H is a linear fonn of H satisfying

aCA)-all-ajj Ci:;t:j).
We denote by a the set of these roots.

160 Chapter 4
Definition4. A root (X E A is simple if (X(A) - ~I - ~+l.l+l.We denote this simple

Tootby (XI'
The set of simple roots is naturally denoted by S. So S - {(Xl. ~ '(X3 , ••• , <Xr ,•••). Each
root (X - (XIj' where (XIEA) - a. -an, can be decomposed into the sum of simple roots :
<Xjj - (XI + (XI+I + ... + (Xj_l if i < j and

If i > j .
The unit matrix (root vector) associated to a root (X is denoted Xa and the
unidimensional space generated by Xa is denoted Ma' We define a partial order on the
set H' of the linear form on H by putting A. > 0 if and only If A. can be expressed as a
linear combination of simple roots with positive coefficients. It Is clear that each root
Is equal to a linear combination of simple roots either with the coefficient 1 or with the
coefficient -I, so that we obtain a patltlon of A Into positive (A+) and negative (A-)
roots: A - A+ U A-.
Let R be the subset of A+ that consists of the roots which are pairwise Incomparable
roots for the previous partial order relation. We put L - Ll3eb' M 13 where A' is the set
A' - {~E A+ 13 YERas ~ ~ Y} .
It Is clear that L Is a standard nllsubalgebra. Note that every standard nilsubalgebra of M
can be obtained as this, from some set R of roots pairwise Incomparable.
We consider in L, the natural filtration corresponding to the descending central
sequence
CO (L) ~ L::J C l (L) ::J C2 (L) ::J ...... ,
where C'CL) - [L,CI-I(L)] , i> 0 .
We provide L with a topology, considering the C'CL) as a basis of the neighborhood of

O. If R is a finite set, the tie algebra L is nilpotent. Otherwise, L Is not nilpotent but for
every neighborhood U of 0, there is an integer n such that CDL c U. In this case, we
will say that L is topolOgically nilpotent.
N.3. The weighting derivations of L
Let L be the standard nilsubalgebra detennined by a subset R c ~+ of pairwise

incomparable roots.
Definition 5. An operator f: L ~ L Is called a weighting operator with weight A E W

If [h,f] - ACh)f with h E H and where [, ] designates the bracket in the Ue algebra of
endomorph/sms of the vector space L.
For example, if f is a weighting operator satisfying fCX a) - axf3' where a:¢: 0, then its
weight A is equal to ~ - (X.. Every operator of L can be fonnally decomposed Into a
sum Ceventually infinite) of weighting operators.
Lemma 1. Let d be a derivation of the algebra L. If d - :E <it is a decomposition of d

in the sum of the weighting operators, then each <it is a derivation of L.
Proof. For all Xa and Xf3 of L, we have

d(X a'X[3]- [d(X a),Xf3] + [Xa,d(X[3)].
The vectors d(Xa,X[3] and [d{X a ),X f3] + [Xa,d{X(3)] can be decomposed in a finite number
of root vectors. This means that only a finite number of weighting components of d
are not equal to zero. But, two weighting operators, having distinct weights, transfonn
a given root vector into two distinct root vectors Clf the images are not zero). This
proves the lemma.
This lemma permits us to restrict the study of derivations of L to the study of

weighting derivations.
NA. Restriction to the finite-dimensional subalgebras
Let MCn) be the subalgebra of M composed of matrices A - Caty satisfying aq - 0 for

i > n and j > n. This subalgebra can be identified as the Lie algebra glCn,«:) of matrices
of order n. Put LCn) - L r. MCn). It is a standard subalgebra of gICn,«:) Cbefore we have
162 Chapter 4
identified M(n) to fl,(n,C)) and it is defined by the system R(n) - {<XtJ E R/ iJ:S; n} (see
Chapter 2). It is clear that the derived subalgebra of LCn) is nothing but D(L) (') LCn),
and we have
Lemma 1. If d is a derivation of L such that d(L(n» c L(n) then the restriction of d to

Ln denoted by dlLCN) is a derivation of LCn).
Lemma 2. Let d be a nonzero weighting derivation of L .Then there is an integer N

such that dlLCn) is a derivation of LCn) for all n, n ~ N.
Proof. From Lemma 1, It Is sufficient to find an Integer n such that d(L(n» c L(n).
Let Xa be a root vector such that d(Xa) - ax~ with a :#: o. We consider the follOwing
three possible cases for weight 'Y - ~ - a of the derivation d:
First case. 'Y - 0

We choose n so that Xa E LCn).
Second case. 'Y :#: 0 and 'Y E A.

CO a - a lJ and ~ - ak,j' where I < j, k < j and I:#: k.
We have 'Y - 13 - a - ak,!.
Let be n - maxCi,k). For every root a', Wsuch that ~' - a' - y, we have Xa' E LCn) and
X~' E Un) and dI LCn) E DerCLCn)).
(Ii) a- <XjJ ' ~ - 131,k, where i < j , I < k and j:#: k.

Here we have y - ~ - a - aj,k • For n, we choose the maximum between j and k.
Third case. 'Y II: .::\ u {OJ.

The weight ycan be decomposed under the following forms:
± (as + «'+1 + ... + «t - a)- ... - a.;; with s ~ t < I ~ r ,

or ± (as + a s+1 + ... + «t +a) + ... + a.;; with s ~ t < 1-1.
In each case, 'Y can be represented as a difference y - ~ - a of two roots, and this at
most as two distinct manners :
CohomoIogyof some ntJotent Lie aIgebtas 163
p - ± (a. + a.+1 + ... +~) and a - ± (al + ... + ~

or p - ± (a. + ... + al-1) and a - ± (~+1 + ... + ai>
in the first described cases and
p - ± (a. + ... + ~ and a - ± (~+1 + ... + al-1)
in the other case.

We choose N so as the root vectors Xa and Xli are in LCN).
Lemma 3. Letd be a weJghtingdedvation of L which is not of the form ad x for any

x E M. Then there is N such that d/L(n) *" ad. Y, V YE MCn) for aJ1 n, n > N.
Proof. Let 'Y be the weight of d. If 'Y is neither zero nor a root, it can be represented as a
difference of two roots In no more than two different manners (see the proof of the
preceding lemma). We choose N in such a way that the root vectors correspond to the
roots defined by 'Y lying in LCN). We deduce the lemma from this.
Suppose that 'Y is a root and suppose that the lemma is not verified. Then for all N,
there is n> N such that d/LCn) - ad Y , YE M(n). As d is weighting, with Weight 'Y, then
d/LCn) - c.ad. X'Y' c E te. By hypothesis, d is not equal to c.ad X.., Then there is y E L
such that dey) :f:. c.ad. X.f.y). But these two matrices lie in a space LCn o) for sufficiendy
large number n. Even if we choose N - no, a con1radlction Is unveiled
The case where 'Y - 0 can be proved analogously. The lemma is proved.
Lemma 4. Let dbe a weightingdedvatlon of L such that d/[L,LJ:f:. o. Then there exists
a poSitive integer N such that d/LCn) is a dedvatlon of LCn) such that d/[LCn),LCn)J :f:. 0
foraJ1 n, n > N.
Proof. Let d[x,yJ *" 0, x, Y E L. We choose N as so to the matrices x and y lie in L(N).
From Lemma 2, d/ LCN) is a derivation of L(N).
Lemma 5. Let d be a weJghtingc1erJvation of L such that deL) cL Z(L), where Z(L) ~
the center of L. Then there exists an integer N such that d(LCn)) cL Z(LCn))
164 Chapter 4
for all n, n ~ N.
Proof. let x E L such that dCx) - y ~ Z(I.). let z be in L saUsfying [y,zl :¢: O. We can find
an Integer N such that x, y and z lie In LCN). By virtue of lemma 4, we can regard d/LCN)
as a derivation. Moreover d/LCNXx) - y ~ Z(LCn)). The lemma is proved.
N.S. The structure of the algebra Der(L)
let M be the IJe algebra of matrices A - (aq), IJ ~ 1, such that each line and each column
contains only a finite number of nonzero elements. Then L Is a subalgebra of M. We
note N ~L) as the nonnallzer of L In M. If x E N ~I.), the operator ad x Is an endomor-
phism of L which belongs to Der CU. Then the set D 1 - {ad x / x E N M<L)} Is a subset of
Der (I.). We put D2 - {d E EndCU / dCL) c Z(U and d/[L,L] - O} .

Each element of D2 is a derivation of L. Note that, If R is infinite, L is not nilpotent,
Z(I.) - 0 and D2 - (OJ.
Theorem 12. Let L be a standard nilsubalgebra of M defJned by a subsystem RcA+

ofpairwise incomparable roots and let R not contain any roots of the fonn al,k' Then
Der L - Dl + D2.
Proof. let d be a nontrivial derivation of L. Suppose that d doesn't belong to D1 + D2.

From Lemmas 2 and 3, there is an integer N such that d/LCn) is a nontrivial derivation
of LCn), for all n, n > N. The standard nilsubalgebra LCn) of glCn,CC) Is defined by a finite
system R(n) of roots which doesn't contain any roots of the type al,k.It follows from
Lemmas 3, 4, and 5, that the derivation d/LCn) doesn't belong to
D1nDer(LCn)) + D2nDer(LCn))
for sufficiendy large n. However, in the finite dimension case, there are no such exist
derivations (see Section 1.5). This gives the theorem.
Suppose now that R Is a finite set and contains roots of the form a 1,k' We order the
elements of R as a 1J1 , a 12J2 ,••• ,aIoJo, where 1 < ~ < ... < ~. For every pair of roots
a- <X'J and ~ - ~k,m' where it S j S k S jr and ~ < m, such that Xa and Xp lie In L, we
Coht:JnrJIogy of some nilpotent Lie aJgebtas 165
define the operator of L, denoted da,p' by

dap(Xa) - Xp , dap(Xa\,J - Xaj,m and dap(Xy) - 0 where Y"" a and Y"" al,k,
Direct verification shows that d ap E Der L. We denote D3 as the linear subspace

spanned by these derivations.
Theorem 13. Let L be a standard nilsubalgebra. ofM defined by a finite system R C ~+
of pairwise incomparable roots, Suppose that R contains a root al,k' Then

Der L - Dl + D2 + D3•
Proof. Suppose that there is d "" 0, d E Der L such that d ~ Dl + D2 + D3 . We can

assume that d is a weighting derivation. From Lemma 2, there is an integer N such that
dlLCn) E Der{LCn)) for a1ln, n :<!: N. Then the standard nilsubalgebra LCn) of gICn,C) is
determined by the system RCn) and dlLCn) is a derivation of Der LCn) which doesn't
belong to CD l + D2 + D3) n Der LCn). From Section 1.5, this brings a contraction.
QED.
Finally, suppose that R is infinite. This implies the nonnilpotence of L, L is still

topologically nilpotent, and its center lCI..) is reduced to {O}.
Lemma 1. Letd be a weightingderiva.tion such that dCXa) ~ axb with /3 "" a, /3 - a ~ ~.

Then d O. m
Proof. We put a - a lJ and /3 - ak,m' By hypothesis, j "" m and i "" k. We consider the
root Y- am,s and we choose s to be large enough so that Xy ELand s > j. This choice is
possible because R contains an infinity of elements. As the set of inner derivations is
an ideal of DerCI..), then [d,adX~ - d'is an inner derivation. We have
d'CXa) - Cd 0 adXy - adXy 0 d)(Xa ) - ax't
where 'C - Y+ /3 ~ ak s' The difference 'C - a is not a root because k "" i and j "" s. If a"" 0,
we obtain a contradiction, d' being an inner derivation.
Lemma 2. Let d be a weighting derivation such that dCXa) - axp where y - /3 - a E ~
and Xy E L. Then d - c.ad Xy-

166 Chapter 4
Proof. Let a - <X.tJ' and ~ - ak,m' By assumption of the lemma, we have m - j or k-

i. We can suppose, without restriction, that a - <ltJ and ~ - akJ' So 'Y - ak,i' Let X't a root
vector of L such that 'Y + 't E d but 't *- a. We only need to show that dCX~ -
c.ad.X.y 00. The root 't can be either of the fonn al,s' or al,k' We will be satisfied to
examine only the first case.We have
dCX~ - b XH'Y' c.adX.fX~ - c X't+y'
We consider the derivation d 1 - adXo' where 9 - <Xj,s Cif s < j , then 9 - CI.s~ and
d' - [d,dll. Then
d'CXa) - dodICXa) - d 1cdCXa ) - dCX~ - d 1CXp) - b Xm - C X't+v - Cb - c)XHV .
The weight of the derivation d' is equal to 't + 'Y - a - ak,s - alJ' where k *- i and s *- j
Ca *- 't and ~ *- a) andis not a root. The weighting derivation d' has a weight which Is
neither a root nor a zero. From Lemma 1, d' - 0 and c - o. This gives the lemma.
Theorem 14. Let L be a standardsubalgebraof M defined by an infinite system R of

pairwise incomparable roots. Then Oer L - 0 1,
Proof. If R doesn't contain a root of the fonn a 1,k, then Theorem 13 follows from
Theorem 11 and the fact that ZCL) - 0 .
Let al,jl E R andlet d be a nonzero weighting derivation that doesn't lie in 0 1, From
Lemma 3 (IVA), there is an Integer N such that el/LCn) E OerCLCn)) andel/LCn) *- adx for
all x E LCn) and for all n, n ~ N. From the studies of derivations of fInlte-dlmensional
standard algebras C§ 1.5), the follOwing two cases are possible:
CO d/[LCn),LCn)] - 0 and d(1.(n) c Z(1.(n) ,
CiO cVLCn)(Xa) - aXp ; cVLCn)(Xa l) - a.Xp-
cVLCn)(XY) - 0 where a - alJ' ~ - ak,m ,
a' = al,k , W= aj,m , 'Y *- a and 'Y *- a' ,
h ~ j ~ k < 12 and is < m .
In this case, LCn) is supposed as being defined by the system
RCn) - {al,jl ,a i2j2' ... , a i,j,}' with 1 < i2 < ... < is·
Coho!ooIogy of some n8potent Lie algebras 167
We choose an integer n sufflciendy large for be sure that [L(n),L(n)] n Z(L(n)) - {OJ and
(X ~ ~ (it is possible that R is infinite). Then the weight of d is not equal to zero. From
Lemma 1, if ~ - (X ~ .6, then d - 0, which is inconsistent
If ~ - (X E .6, from Lemma 2, then d - c ad Xy and this Is Impossible because we have

supposedd e 01.
V. COHOMOLOGY OF THE INFINITE LIE ALGEBRA OF VECTOR

FIELDS OF THE REAL STRAIGHTLINE
Consider the vector space V of Infinite dimension provided with a countable basis
{eo,e 1, .•. ,ek, ...J. It is possible to make a structure of a tie algebra, noted Lo , from the
bracket
We define the subalgebra Lk of Lo as the subalgebra generated by the vectors

(ek,ek+l, ... ,en ,•• .). These algebra are provided with a natural graduation: the weight of
vector ej being the integer j. From this, we deduce a graduation of the cohomologlcal
spaces.
Remarks. (1) The algebras Lt are topolOgically nilpotent for the topology associated to
the filtration generated by this graduation.
(2) A representation of these algebras is given by putting el - Xl+l ~ .

ax
For this, we consider these algebras as subalgebras of the algebra or vector fields on m.
Proposition 9. The spaces Hl(Ll'L 1) and HI-l(L:z,CC) are isomorphed asgraded. vectors
spaces.
168 Chapter 4
We accept this proposition without proof; for an example, one can read [FU].
As Goncharova has composed all the cohomological spaces of ~ with values in (C (see
also [FUD, one deduces the follOwing proposition.
Proposition 10.
In fact this proposition has been proved direcdy by A. Fialovski [FI]. More precisely,
he gave a precise description of the cocycles of a basis of H2(L v L1).
Let C2, C3 and C4 be cocycles of weights -2, -3 and -4 defined by
i - 2, 3 and J~ 4 }
otherwise
L 3 (el,ej)-0 'r;f J
L3 (e2,e3) - 8e2
L3 (e2,e4) - 4e3
L3 (e3,e4) - - 10e4
L3 (et,ej)- {
(_1)1 ct2 (J - i + 3) ej+I-3, as i - 2,3,4 and
0, otherwise
f-4(el,ej)-O 'd J
f-4 (e2,e3) - - 14el f-4 (e2,e4) - 0 ; f-4 (e2,e5) - 8e3
f -4 (e3,e4) - - 24e3 f-4 (e3,e5) - - 16e4
f-4 (e4,e5) - 18e5
CohorooIogyof some nIpotent Lie aJgebtas 169
Then 1he classes of these cocyc1es fonn a basis of H2CL1.L1).
VI. ON THE COHOMOLOGY OF NILPOTENT LIE ALGEBRAS IN

SMALL DIMENSION
By using the MATHEMATICA program given In Chapter 3. we can easely compute. for
each nilpotent Ue algebras n of dimension less than 7. the dimensions of the spaces
Hl(n,n) for 1- 0, 1 and 2 and also the dimension of the spacesHl(n,CHor
dim n. Also this programm pennlts to obtain more Infonnations as the basis of these
spaces.
CHAPTER 5
THE ALGEBRAIC VARIETY OF THE LAWS

OF LIE ALGEBRAS
I. THE VECTOR SPACE OF TENSORS TIl 2,1
1.1. Definition
A tensor of type (2,1) on (CD Is a bilinear mapping
with values In (CD.

We denote by TD2 ,1 the set of alternated tensors. It Is provided for a structure of a
complex vector space of dintenslon Cn3 - n2)/2
We fix a basis {ell ... , eJ of (CD. The tensor T Is defined by Its constarlts of the structure
Tkljglvenby
°
T(ebeJ) - L T~J ek .
k-l
The antlsymmetry condition can be expressed by

The variety of lie Algebras 171
Identification of T with its structure constants determines, once the base {eJ is fixed,
n'-n2
an isomorphism between fi2•1 and (C 2 •
1.2. Action of GL(n,C). Classification of the tensors
Definition 1. Two tensors T and T' of Tn2 •1 are called isomorphic if it exists
f E GLCn,(C) such that
TI (x,y) - f- 1 (T(fCx),fCy») , "i/ x, Y E (Cn .
To simplify, we will note that T' - f-l 0 T(f, f). This formula corresponds to that for a
change of basis. So the two tensors T and l' are isomorphic when the two tensors
have the same constants of structure in comparison with the two bases of (Cn.
We will denote by OCT) (or G(T)lf it is necessary to define the group G) the orbit ofT
under the action of GLCn,(C), i.e. is the set of isomorphic tensors of T.
To classify up an isomorphism, the tensors of (Cn consists in determininge the whole

orbits. currently, only a few results are known about the classification. This can be
easily explained. If we note T1, ••• , Tn as the components of the tensor T, every TI is a
bilinear alternated form. The description of the tensors of P2.1 consists of
simultaneously classifying n bilinear alternated forms. Except for the case where TI
commutes, we know almost nothing.about this classification.
Proposition 1. If T - CTk~ isa tensor of P 2 •1 andif f - Ulq) is an isomorphism of(Cn,

then the constants of structure of T' - f- 1 0 TCf, f) verify
where Cb~ is the matrix of f-l.

172 Chapter 5
1.3. Rigid tensors. Open orbits
We provide the vector space Tn2,l with a euclidean topology.
Definition 2. A tensor T E P2,1 is said to be rigid if its orbitOC!) is open in P 2,1'
Example. Let To be the tensor of T2 2,1 given by TOCe 1,e2) - el' We showthat To is
rigid. Every tensor of T of T2 2,l is expressed by
If f is the isomorphism with matrix Cat? ' then the tensor T' - f-1 0 TCf, f) is written as
If Ca,b) *" CO,O), we can always find an isomorphism uy so that we have T'Ce 1,e2) -
TOCel,e2) - e1' Particularly, if T is in a neighborhood of To, i.e. is if Ca,b) is neaI(1,O), then
Ca,b) *" CO,O) and T is isomorphic to To, which shows the rigidity of this tensor.
Consequence. Every tensor T of T2 2,l is either isomorphic to the nul tensor, or
isomorphic to the tensor To'
II. THEVARIETY L n OF THE LIE ALGEBRAIC LAWS
11.1. Lie algebraic laws on C n
Definition 3. A lie algebraic law Jl on (Cn is a tensor of Tn2 ,1 verifying the Jacobi
identity.
~~x,y),z) + ~~y,z),x) + ~~z,x),y) - 0
for every x, y and z in (Cn.
We fix, once for ever, a basis Cel) of (Cn. We will note Ckij as the structure constants of
The Variety of Lie Algebras 173
J1 relative to this basis. Jacobi's Identity can be reduced to the polynomial e<pidons
system:
n
(S) 1:
1-1
cIIJCslk+c1jkcslI+c1kiCsIJ-O, lSsSn, lSI<j<kSn.
So we can consider a tie algebraic law J1 as a tensor J1 - (Ck~, the CklJ of which verifies
(S).
Remark.Until now, we have only spoken of tie algebras and a tie algebra has been
defined like a space vector 9 , provlded with a square brackets [ , 1verifying the Jacobi
Identity. If one identifies the vectorial 9 to ten and the brackets to a tie algebraiC Jaw,
the tie algebra Is nothing but a pair 9 - CJ1,te n), where J1 is a tie algebraic law. It Is clear
that most of the time, we wIllldentlfy 9 to its law.
11.2. The algebraic variety r,n
One ldentlftes the law J1 with Its structure constants. The set lJl of the tie algebraic laws
~
Is then the subset of ~ 2 - T'2,1 defined by the system of polynomial equations (S).
n n"-n2
So L Is provlded with a structure of the algebraic variety Imbedded In tC2 .
The topology of Ln.

";-cr n
As a subset of ~ ,r; can be provided with the classical topology. It Is almost natural
to provlde Ln with the Zarlskl topology which Is less fine than the preceding topology.
We recall that a Zarlskl closed set Is defined by a finite number of polynomial
equations. In this chapter, somewhat contrary to the uses, we will consider Ln with a
metric topology. When It will be necessary to restrict oneself to the Zarlskl topology,
we will predse It
174 ChapterS
Examples.
1. The variety (,2
Since every tensorT - CT 112, T2 l2) verifies (S), we Immediately deduce that 0 - 1"22,1'
Here the algebraiC variety L2 is nothing but a complex plane of dimension 2.
2. The variety L3
A tensor ~ - (c~J 1 sIs J s 3 and 1 S k s 3 veri8es the Jacobi conditions, If and only If we
have
C l 12 C223 + C l 13 C'2' -C I 23 (C'13 + C2l2 ) - 0,

(Sy c2a (C'23 - C l 12 ) + Cia C212 - C223 C'23 - 0 ,
C'12 (C 223 + ell') - C l 12 C'l3 - C 212 C'23 - 0 .
The variety {,3 Is an algebraiC variety of «:9 defined by the system s,. We will see later
that this variety Is a reunion of two irreducible algebraiC components. The tangent
geometry of {,3 is more difficult to describe, because of the existence of singular
points. For example, the point ~ defined by C3 12 - I, c klj - 0 for the other Indices I, J,
k (~ corresponds to the Heisenberg law H3) Is a singular point
To Interpret better some geometric or algebraiC proprieties of the variety Ln, It Is

convenient to consider (}1 as an affine scheme. In the next paragraph, we will present
briefly these notions of algebraic geometry.
11.3. The scheme r,n
1. Spectrum and scheme
Let be R a commutative ring. Its spectrum Is the set

SpeC(R) - {I c R where I is a prime ideal I of R different to R} .
(A prime ideal is an ideal which verifies pq E I ~ pel or q E I.) One provides Spec(R)
with a topology whose basis of open sets is given by
SpeC(R), - {I E SpeC(R) / f fi!I: I}.

On this topological space one constructs a sheaf of functions. Let I be a prime ideal of
Rand RI the localized of R in I :
RI - { ~/ a E R, s E R - I} modulo the equivalence relal10n :
il_ a' <=> 3 s· E R - I such that Sl(as' - a's) - 0 .

s s'
Let U be an open set of Spec(R). We define
reu) - {s : U ~ II RI such that S(0 E RI} .

leU
One suppose, moreover, that s is locally a quotient of elements of R :

'tI lEU , 3 VCO c U neighborllood in I and a and b in R such that
'tIJEVCO, bfi!l:J so s(J)-~inR.J.
Approximately, it remains to define the regular functions on Spec(R). The fibre reo of
the corresponding sheaf above the point I corresponds to RI.
The affine scheme is the datum of Spec(R) and its sheaf.
2. The affine scheme Ln
We consider the ring R - C[X1, ... , XN ] of complex polynomials with N - (o3_n2:>12. Let
J be the ideal of R generated by the Jacobi polynomials of the system (S) and let
A - R/1. The scheme Ln is an affine scheme constructed from Spec A. We will not
distinguish, at least in the notal1ons, the scheme Ln with the variety lJl.
176 Chapter 5
It Is interesting to consider Ln as a scheme which allows us to guess some of the

properties of the set of tie algebra laws which are expressed very clearly in the
language of the schemes.
We recall the notion of reduced scheme.
Definition 4. A scheme CSpecCR),n 1s reduced 1f the local dng RCI) has no n11potent
1deal for all I in SpecCR).
This means that doesn't exist non trivial polynomial f such that Cf)p is in the Ideal of
theJacobi.
3. The tangent space to the scheme Ln
Let J.lo - (aktf be a point of the space Ln. The tangent space in J.lo to the scheme LD Cwe
can identify J.lo with a point of the scheme) is defined by the linear equations which
correspond to the homogeneous part with degree 1 of the Jacobi equations centred
on the point J.lo.
Examples.
(1) If J.lo - CO) is the Abelian law, so the tangent space Is eN, where N-Cn'-n 2)/2
because there is no component of degree lin the Jacobi equations.
(2) J.lo - (a'12 - 1, aklj - 0 otherwise) e V. The equations translated on this point are
written
e l12 e22, + e l13 e'2' - e 12, (e'l' + e2l2 ) - 0,

e2l, (e'23 - e l12) + ell' e2l2 - e223 e'13 - 0 ,
- (e 323 + ell') + e212 (e 223 + e113) + (e l12 - e'23) e212 - 0,
and the equations of the plan tangent to the scheme L' on J.lo are
The Variety of Ue AJgebtaS 177
It is an 8-plane.
Theorem 1. The tangent plane at the poInt J.Io to the scheme LO coIncIdes with the
space Z2(J.Io,J,Io) of the 2-cocycles for the cohomology of J.Io with values In the adjoint
module.
In the notation Z2(J.Io,J.Io) we identify, to agree with the precedJng chapters, the law J.Io
with 1he Ue algebra 90 - (J.Io,ICO).
Proof. Let <p E Z2(J.Io,J.Io). Then

&p(e"eJ,ek) - 0, 'If i, J,k .
We put ",e"ej) - 1: a~J ek . The identity hereunder is equivalent to the system
which is nothing else than the linearization of (S) translated. on 1he point J.Io - (C~.
11.4. The action of GL(n,C). Fibration by orbits
The action of GL(n,IC) on 'Jb2,1 restricted to lJ1 induces an action on the variety of the
IJe algebra laws : two laws Jll and Jl2 are isomorphic, if it exists, f E G L(n,IC), such that
Jl2 - (--10 J1Cf,O.
We denote by O(ID the orbit of Jl correspondJng to this action.
Lemma 1. For every Jl E I,n Cor in 1'02,1) the orbIt OCID Is provided with the
structure of a dJfferentiable mando/d.
Effectively, OCID is the image through the action of the Ue group GL(n,lC) of the point
178 ChapterS
~ (considered here as a point of the vector space Tn2,1). The Isotropy subgroup of ~,
I.e. the subgroup of GLCn,C) defined by
Iso{~) - {f E GL(n,C) / ~ - (-I 0 ~(f,f)} ,
colnddes with the group of automorphlsms of~. Then It Is a closed subgroup and the
orbit Is Isomorphic with the homogeneous space GLCn,C) / IsoCJD. This proves the
lemma.
Theorem 2. Let ~ be in Ln. Then the tangent space at the point ~ on the orbit O(JD
colnddes with the space of coboundaries B2(~,JD.
Reca1I that the elements of B2(~,JD are coboundaries of the form !p - 'Ov,.f, where f
Is in g1Cn,tC) with
'Ol1f(X , V,) - ~f(X) , Y) + ~X , f(Y» - f(~X , V»~ .
Proof. Let ~l be a point close to ~ in O(JD (for the topology of the manifold O(JD which
coincides with the topology inducedby TD2,1) In OCJD. Then ~I - (-I 0 ~f,f) with f
near to the Identity In g1Cn,C). We put f - Id + g, where g Is a linear mapping given by a
matrix which belongs to the neighborhood of 0 in gICn,C). Knowing that, one can WIire
f- I - Id + h with h E yeo) c gICn,C),

we have
~t(x,y) - ~x,y) + ~g(x1y) + ~x,g(y» + ~g(x),g(y» + h~x,y) +
+ h~g(x1y) + h(x,g(y» + h~g(x1Y) .
As Cld + h) oCld + g) - Id+ h + g+ hog -Id ,one deduces h - -g-h og, so
~I(X,y) - ~(x,y) - 'O!18(x,y) + e,
where e Is an uinflnireslmal" of order 2.

This shows that the tangent vectors in ~ to O(JD have the form 'Ol1g. Hence this proves
the theorem.
The Variety of Us Algebras 179
11.5. Open Orbits. Laws of rigid Lie algebras
The algebraic variety Ln can be provided with two structures of topological spaces by
considering either the metric topology or the Zariski topology which, recall, is less
fine than the first. We can also consider the orbits, either like differentiable manifolds
provided with the metric topology or like topological spaces provided with the
Zariski topology.
Proposition 2. The two topologic spaces OCJ,D (metric topology) and OCJ,D (ZariskJ
topology) are homeomorphic.
This results from the fact that O(J,D is differentiable linear (see, for example, Mumford
"the Red Book").
Definition 5. A Ue algebra Jaw /1 of Ln is said to be rigid if its orbit is a Zariski open

set of Ln.
According to the preceding proposition, it means that 0(/1) is open for the metric
topology.
Proposition 3. Only a flnited number of open orbits (or of isomorphic classes of rigid
laws) exist in Ln.
Effectively, all algebraiC varieties are decomposed to a finite number of irreducible

algebraic components. Now, if 0(/1) is a Zariski open set, its closure, for the Zariski
topology, is an algebraic component of Ln. From which we obtain the proposition.
Examples.
1. Only one open orbit exists in L2, which is defined in /1 by
/1(el,e2) - 0
(Recall the implicit convention: the undefined brackets are nul).
In this case, L2 is hreducible and coincides with 0(/1) .
180 ChapterS
2. We consider in 0 the law defined by

IlCe 1,e2) - e3 '
IlCe3,e 1) - e2 '
IlCe2,e3) - e 1 •
It is the simple Ue algebra law slC2,C). We will see later that all simple laws are rigid.
So 0(1l) is a component of L3. The following law
Ill(e 1,e2) - e2 ,
Ill(e 1,e3) - e3 '
is notin 0(1l). Hence, L3 is not irreducible (see §. V).
11.6. The dimension of orbits
Let JIo be in LD. The orbit OCJlo) identifies itself with G LCn,C)/AutCJIo), where AUtCJ.Io) is
the group of isomorphisms of JIo. We have
dim O(llo) - n 2 - dim Aut(llo) .
We note that the Ue algebra of the Ue group AUtCJIo) is nothing more than the algebra
of derivations DerCJIo).
Proposition 4. For all every JIo E Ln, we have
dim O(llo) - n 2 - dim Der(llo) .
Examples.
Cl) If Ilo is Abelian, DerCllo) - glCn,C). Then
The Variety of Lie AJgebtas 181
dim 0(110) - 0 .
(2) If Ilo Is the law of the Heisenberg algebra HI of dimension 3, then
dim Der(J1o) - 6 ,
dim 0(110) - 3 .
(3) If Ilo is the law of simple algebra sl(2,C), then dim Der(J1o) - 3 and dim 0(1lo) - 6.
11.7. The components of (,n
We can see very easdy that, after dimension 3, the variety LD is not yet irreducible. The
problem of determining the components Is also naturally presented.
Unfortunately, this problem Is a little difficult to tackle. We wdl see at the end of this
chapter that this determination Is only explicit for dimensions less than 7. The
problem is dlsplacedby an estimation of this number of components. But here also,
we need refined techniques, and the approaches are difficult. Nevertheless, the next
proposition allows us to envisage an estimation of the number of components.
Proposition 5. Let J.1o be a rigid law in {JI. Then the closure in the Zariski sense
O(llo)of the open orbit O(llo) Is an l1reduclble component of LD.
But there exist component given by nonrigid laws. We can consider some "rigid"
families parametrized by one, two or several parameters. In theses caces, the orbits of
theses families give components. The foHowing proposition precise such orbits.
Proposition 6. Let A be an irreduclble algebraiC set In Ck and we consider a subset

{llcJ aeA of LD formed by the laws of two pairwise nonlsomorphlcs. If this family is
rigid (J.e. the union r of the orbits of the Jaws Ila is an open set), then the closure 01
rls an l1reduclble component of Ln.
182 Chapter 5
m. CONTRACTIONS. DEFORMATIONS
111.1. Contractions in Ln.
Let ~ be a point of LnandCf~peli a sequence of isomorphisms ofCn(wtth, to simplify

the notations, fo - Id). For all pEN, the law ~p - f-lp 0 ~ Cfp x 9 is in
the orbit of ~. But the limit (!~ ~p) perhaps does not exist in 'j'i2,1 • If this limit does
exist, itis inside Ln and is called a conlraclion of ~. We will note it, for example as ~oo.
So
Proposition 7. Let ~oo be a conlraclion of ~, then ~oo Is In the closure of the orbit
of~.
This proceeds direcdy from the definition of ~oo.
Corollary. Let C be an lrredudble component conra/nlng ~. If ~oo Is a contTaclion

of~, then ~oo E C.
In fact, if ~o E C, O(~o) C C and O(~o) E C.
Consequence. If ~oo is a conlJaction of ~, then
Remark. The notion of contraction (sometimes called degeneration) has been

introduced by theorician physicians, more precisely by Segal, for the purpose of
understanding the relationship between relativistic and classical mechanics. Effectively,
if we consider the IJe algebra of a Lorentz group as an algebra parametrlezed by the
speed of light, this contracts itself by stretching the speed to infinity, to the Galileo'
The Variety of Lie AIgebtas 183
algebra. This noUon of contracUon was developed someUmes later by Inonu and
Wigner with the parUculariZaUon : In their works, they suppose that one subalgebra.
remains fixed throughout the contracdon. As this noUon Is alitde bit restricUve, It has
been generalized by Saletan, Levi-Nahas, and some others. If we carry to extremes the
generalizations of Inonu and Wigner, we will probably discover the original noUon of
contra.cUons!
111.2. The deformations
The nodon of deformadon Is supposed to be the "dual" of the nodon of contra.cdon. It

was also Inlroduced to dispose of a topological tool. Unfortunately, the definldon of a
deformaUon, as was InlUally Introduced by Gerstenhaber and then developed by
Nljenhuls and Rlchardson,lays In a formal concept. However, as Its use has shown,
tteself to be most effecUve, we will conserve It and try, in the next chapters, to study
some propriedes in the case of nUpotent laws. But the problem of the contra.cUon
duality-d.eformadon stays enUre. We will tackle It and perhaps produce a convincing
saludon about It, by Inlroduclng a nodon which Is closer to the reality than that of
deformadon. This we will call perturbations. But returning to the defonnaUons, we
have the follOwing deflnldon.
Definition 6. A deformation ofalaw ~ of (,n is a formal sequence wilba parametert

Ilt - Ilo + ~ t l q>1 , where the q>1 are In T'2,1 and verifying the formal identity ofJacobi.
19S"
Ilt 0 J.1t - 1: t1+j [( q>to<PJ + q>f'<!ll) + q>~ q>1+j] + 1: t2I q>101 - 0 .

19+j~.. 19~..
We define the notations which intervene In this deflnldon.

If q> and 'I' are in 'Jll2 ,1' we note q> 0'1' the trilinear applicadon altemated with wlues In
Cn defined by
q> 0 'I'(X,Y,Z) - <A'\jI(X,Y~Z) + <A'\jI(Y,Z~X) + <A'\jI(Z,x~ Y) +

+ \jI(<A'X,Y~Z)+ \jI(<A'Y,Z~X)+ \jI(<A'Z,X~Y) •
184 ChapterS
In particular, we have
~cP - J.Io 0 cP - cP 0 J.Io as soon as J.Io E (,n and cP E "2,1 and Il 0 Il- 0
Is nothing more than 1he Jacobi condition.
Example. A linear deformation Is given by
(I.e. CPt - 0 If I ~ 2).

In this case, 1he formal Jacobi condition Is written as
which implies
Proposition 8. A 2-cocycJe CP1 for a given Jaw J.Io dellnes a Ilnear defonnation if and
onJylfcp1 lsalawof (,n.
Proposition 9. Let f.1t - J.1o + 1: t l CPt be a defonnation of J.Io. Then the Ilrst tenn CP1ls
a 2-cocycJe for J.Io, i.e. 8J.1O CPl - 0
Effectively, 811o'P1 corresponds to the coefficient of the linear part of the formal Jacobi
Identity.
Remark. We can always consider a deformation f.1t of J.Io as a tie algebra on the field
C((O) of the formal sedes In t. This,moreover, permits us to generalize this notion In
consldedng not C((t)) but the dng of formal sedes with some variables as the space of
1he scalars. So one Is led to the very general notion of deformations parametrized by
C-complete local algebras A, so AlmAn Is of finite dimenSion, where mA Is the
maximal ideal of A.
This generalization allows us to foresee a tie with the notion of contIaction. But we
prefer to present here amore direct aspect and most topologic than the perturbation
for tackling this problem.
Proposition 10. So /J.t Is a deformation of J.Io. Then there exists a neighborhood V

of 0 In IC such that 'I u E V, the series
J.lu-J.Io+l:d<llt
conve.r.ges in 'JD2,1 •
This results from the Artin theorem relative to the polynomial developments with
coeffidents In Integer series.
This proposition justifies, at least to a certain extent, our desire to conserve as a

definition of defonnation, the original notion. We have seen that the coefficient of the
linear part of a deformation of J.Io belongs to the space Z2(J.Io,J.Io). Now this space
coinddes with the tangent plan in J.Io to the scheme LD. The fonnal Jacobi conditions
relative to /J.t are eqUivalent to the infinite system :
(N.R)
The system (N.R) can be Interpreted as the necessary and suffidentconditions for a
vector <P1 of the tangent plan of Zariski in J.Io to the scheme Ln being the first term of a
deformation.
If this system admits a solution, we wiD say that the linear development J.Io + t <P1 is
Integrable. We also will say that <P1 is integrable.
186 Chapter 5
We suppose that there exists a neighborhood U of ~ In Ln (for the metric topology

for example) such that every point ~ of U Is the sum of a convergent deformation. In
this case, the system (N .R) describes the sufficient and necessary conditions for a
vector tangent to the scheme In ~ to be a tangent vector In ~ to the variety (}l (this
means a vector of the cone of the tangents).
Theorem 3. Let ~ E (}l such that there exists an open U containing ~ whose
elements are sums of convergentdeformatlons of ~. Then a vector tangent<Pl to the
scheme (}lin ~ is tangent to the variety Ln if and only If there is a sequence of
tensors (<1>2, ... , CJln, ...) of Tn2,1 as CN.R) Is satIsfled.
Remark. We wilisee, by studingthe perturbations of~, that the hypothesis relative

to the existence of this is always satisfied and that the system eN .R) is equivalent to a
system which contains a finite number of equations.
Consequence. Let T2 be the smallest vector subspace containing the cone of the
tangents to the variety Ln in ~. If T2 Is strlcdy contalnedln Z2(~,~) , then the
scheme Ln Is not reduced In ~.
IIIol. Equivalent deformations
Let fCt) be a C[[t]] automorphism of a set of formal series (which Is a C[[tll-moduie).
This automorphism is written
where every fils an endomorphism of Cn.

The deformation ~. t defined by
where ft-1 Is the formal Inverse of ft' Is called a deformation equivalent to the
deformation J.lt of 1he law JJo.
A deformation equlvalentto the null deformation J.lt - JJo ofJJo Is usuaIIycalledatrlvlal

deformation.
We PERTURBATIONS. NOTIONS OF INFINITESIMAL ALGEBRA
To give a precise sense to the object which represents a law close to a given law, and to
tum the formal obstacle of the deformations, we will Introduce the notion of
perturbation, which is nothing more than a law infinitely close to a given IawCthen it Is
a concept marvellously adapted. to the metric topology of LR). This notion Is based on
the notion of inftniteslmals. We will briefly recall the few basic notions of the theory of
the complex infinitesimal.
Wel. The infinitesimals in en
1) A short survey of the Nelson axioms
The infinlteslmals are Introduced In an axiomatic way into the classical axiom of the set
theory. First, we will describe this by summing up the consequences of this theory on
complex space vectors.
In the fteldof complex numbers C, there exist both standard and nonslandard
elements. Every sca1arconstructed. by formulas ofclassical mathematics is standard.;
for example I, 10, 10- 10, i are slandards Cusuallyevery set classically defined Is standard;
and also 1he fteklC Is Slandard.).
Axiom I : OdeaU?allon). There exIsts an element of C which Is nonstandard.
If a nonstandard element eXiSts, there Is an Infinite number of them, otherwise these

188 Chapter 5
nonstandard elements would be standard and Axiom I would be at fault On the other
hand, there does not exist any subset of C constltutedby all the nonstandard elements
of C , because these elements are not, by nature, constructed by the formulas of the
classical set theory.
Definition 7. An element £ of C Is said to be Inflnlteslmallf It verlfles lEi S Izl for all

standard ZE C. We will Wl1te £; o.
An elementoolnC Is infinitely large If 00- 1 ;0.
An elementa Is limited If It Is not Infinitely large, although If It Is not lnflnltesimal, one
will say that It Is appreclable.
Axiom S (standardization). Every limited element a of C Is Infinitely close to a

unique standard element noted by Oa, and called the shadow of a.
Saying that the two elements a and b of C are infinitely close, means that their
dJfference is infinitesimal ; we will write a ; b.
Axiom T (transference). About every theorem of classical mathematlc, one can

deduce a nonstandard theorem supposing that every variable Is standard and vice
vema.
Example. Let (n be the theorem: "every complex polynomial admlls a root".
of course, this affirmation keeps a theorem with in the nonstandard frame, because
this extension is conservative: every theorem keeps a theorem in the extension and
every theorem of the extension expressible in the language of the initial theory, is a
theorem of this. If we suppose every variable to be standaJd, we will write : "every
complex standard polynomial admits a standard root".
The axiom of transfer permits us to affirm that this sentence is also a theorem
(non expressible within the classical theory of the sets) ; without this axiom, we only
would write "every complex standaJd polynomial admits a root", without we can
The variety of lie Algebras 189
pretend that the root Is standard.
We will currently use the transfer In "the opposite sense". For example, we suppose
that we want to prove C'J) i we can prove within the (nonstandard) frame, that every
standard polynomial admits a standard root. With the transfer principle, we deduce
that (1) Is true. The rewards of this approach reflect perfectly the qulntescence of the
nonstandard method: to demonstrate the transference of (1), we can use all the tools
that are produced by nonstandard analysis, which are the classical tools (within the
nonstandard framework we can dlrecdy demonstration en), to which are added the
specific Nonstandard tools. As we are richer the proof of en transfered can only be
more simple.
2) The algebra of the infinitesimals
We easily prove the natural rules of calculus concerning the Infinitesimal and infinitely
large elements :
CI»+CI»-CI»,
Cl»xCl»-CI» ,
LxCl»-CI»,
L+L-L,
lJ+CI»-lJ,
lJxlJ-lJ ,
where CI» designates every Infinitesimal, L designates every limited and lJ an Infinitely
large. On the other hand, one can not pretend about the result of the operations of the
type:
CI»/CI»,
lJ+lJ,
Cl»xlJ,
which cOJTeSpond to the famous undetermined foons of elementary analysiS.

190 ChapterS
3) Infinitesimal vectorial algebra
We consider a standard integer n. Then the vector space Cn is standard. AccordJng to

the transfer axiom. there exists a standard basis ; all the vectorial calculus in Cn will be
established with respect to this standard basis. So a standard vector will have standard
components with respect to this basis.
Definition 8. A vector Y of Cn is calledlnfiniteslmalif its components (v1.v2 ..... v.)

are lnflniteslmal.lt is said to be limited If the components are limited. and lnflnitely
large If one of the components is lnflnltely large.
We consider a standard norm In Cn (for example. that associated to Ihe usual Hermitian
product). If Y is Infinitesimal (resp. limited, resp. infinitely large). its norm is infinitesi-
mal Cresp. limited. resp. infinitely large).
Shadow of a limited vector.
We suppose Y is limited. From the axiom S. each of its components VI admits a

shadow 0v1 and the standard vector (0 v l' •..• 0 v.) is called the shadow of Y ; it is noted
ov.
Usually. the vectors Y and oy are linearly Independent. The aim of the next section is
to determine the relations between these two vectors.
lY.2. Theorem of decomposition of a limited vector of Cn
This theorem is the basis of all calculus relative to the perturbations of a point in a
vectorial plane.
Theorem 4. Let Y be a limited vector of cn and OV its shadow. We suppose these

two vecton to be lndependent. We have the follOwing decomposition:
with Eft JnfJnJtesJmaJ In C,

(0 £1' ~, ••• ,
(lDV1, V2, •••, VkstandardandJinearlyJndependentJn CD.
7bJs decomposltion Is unique up to an equivalence :

If
V - oV + £lV 1 + £1£2V 2 + ... + £1£2 ..• £kVk-
- oV+ (llW1 + (l1(l2W2 + ... + (l1(l2 ••• «sWs
are two decompositions oiV, then we have :
(Ok-s,
I
(H) WI -1:
)-1
;JIVj with all standard and all *0,
Comments
1) The integer k determined by a decomposition of V and which does not depend on

the choice of the decomposition, Is called the length of V. If V Is standam, It has as Its
length o. Further, the length Is always one standam smaller or ~ to n, which Is the
dimension of the vectorial frame.
2) The vector V defines In a single way the flag (V l'CV l' V2)' ... , CV l' V 2' ..• , V k)). This
proceeds direcdy from relations Cu). The most esthetic illustration of this geometry
linked to alimlted vector Is based on the Euclidean plan. If we take a limited point M
situated on a standam smooth curve, the flag linked to this point is nothing other than
the flag given by the tangent vector, the osculator plane, the second osculator plane,
etc.
3) Geometric Interpretation of the decomposition
The flag (V 1,eV1,V2), ..•, (V1, V2 , ••. , ViJ) which is canonically linked to the vector V, Is
defined as follows: we consider a Cnonstandam vectorial) Straight line with a director
192 Chapter 5
vector V-OV. It Is Infinitely close to a unique standard straight line whose director
vector is V 1. We consider now a vector plane containing these straight lines. Let us
note that If these two straight lines are the same, the length of the decomposition is 1.
Otherwise, the plan is well determined. It is standard or not. If it is (standard), the
length of the decomposition is 2 and (Vi' V2) is a standard basis of this plane. If it Is
nonstandald, It Is infinitely close to a unI<Jle standard. plane Cits shadow) and (V l' V2) Is
a basis of this standard. plane, etc.
This permits us to interpret geometrically the integer k linked to V.
Proposition ll. The length of V colndcies with the dJmensIon of the smallest standard
vectodal subspace containing the InfInItesImal vector V-OV .
Proof of the theorem.
1) Existence of the decomposition
We can suppose V to be infinitesimal, otherwise we replace it by V-OV. Its

components (Vi' v2 ' ••• , vn) are allinflniteslmal. Let £1 be the largest module of the
components Vi. The quantities v l£l are all limited and we can wIlte :
with ~ standard (eventually null) and Wi == O. Let us note V 1 as the standard vector
(ai' a.:z, ... , ~ and Withe infinitesimal vector (wi' w2' ••• , wri. By construction, the
vector V 1 is not zero, the component corresponding to £1 is equal to 1. If Wi Is zero,
decomposition Is achieved, it has a length equal to 1. Otherwise we repeat the above
mentioned procedure to the vector Wi whose component corresponding to £1 Is
zero, that reduces by 1 the dimension of the space In which we operate. By

construction, the vectors V 1 and Wi are lineary independents. As the vectorial frame
is of finite dimenSion, the proceedings stop before n steps (at least).
The Variety of Ue Algebras 193
2) Equivalence of two decompositions
Let us consider two decompositions of V :
and let us suppose k 2: 1. The vector VI is nonzero.
Lemma. The inflnitesimals el and (Xl are equivalent, which means (Xl leI is
appreciable (this implies that (Xl is nonnuID.
Proof of the lemma. If (Xl leI is not appreciable, it is infinitely large or infinitesimal.
In the first case, we divide the identity en by (Xl and then take the shadow of the two
members. We get WI - 0, which is the contrary to the hypothesis made on the vectors
of the decomposition. In the second case, it is the vector V I which is zero. So the
lemma is proved.
As el and (Xl are equivalent, we can write
with al standard nonzero and a.l == 0 .

We get
and, after division by el'
So, the two vectors are equal and limited and their shadows are equal, this gives:
V I - a I + W I and a 1nonnuU standard.

194 ChapterS
Relation (2) Is reduced to
If ~ - O. we recover a trivial linear relation of the Independant vectors W 1 • W2 •...• Wn'

This gives all - 0 and «z - O. In this case. the equivalence Is proved.
Let us suppose ~ "#. O. We pose a 12 - (all + (11)«Z' Thi infinitesimal equivalent to «z.
The identity is written :
We are led to comparing the Infinitesimais Ez. all and a 21. The different situations are
condensed In the next scheme:
af£2=0 and a}I£2=0 => V2-0.

af£2 appreciable and a}I£2 = 0 => (W 1oV2) dependent.
af£2 infinllely large and ail£2 = 0 => W1 - 0 •
af£2 infinilelylarge and ail£2appreciable => W1-0.
af£2 = 0 and a}I£2 Infinllely large => W2 - 0 •
af£2 appreciable and a}I£2infinitelylarge => W2-0.
af£2 infinitely large and a}I£2 Infinitely large => (W 10 V2) dependent.
All these cases disprove the hypothesis. The only possibilities are
a1ie2limlted and ail£2 appreciable.
We can put
a}-£2(a}+al) with a~=o and alstandard nonnul.

af- £2(a/ + al> with ai =0 and aJ standard.
The variety of Us Algebras 195
This gives the relat10ns :
V2 ai W 1 + ai W 2 ,
-
a1 - £la l + £l£~i + £1£2a [.

£,V, + ... + £, ... £kVk - aiW 1+ atw2+ a.fW,+ afa..W .. + ... + afa.. ·.. a.Ws (5)
We are now led to comparing the Inflnlte5lmals .:" a 21, a 22 and a 2,. The hypothesis
relative to the vecto!"S V 1 and W 1 Implies that the quantities a2/e, are not Infinitely
large. So .:, Is equivalent to one of the a2 1 , this gives the linear relations between V,
and W 1 for I S 3. This process decreases the number of terms In the left part of the
Identity (5). Before r steps (r S k) , we get :
k-1
Vr-1- ~ a1r-1W1 wl1h a!~l;tOands1andard.
1-1
In particular, If r - It, then we have
£lV k - ar-tw 1 + ai-tw2 + ... + ak~itwk-1 + <Xk(ali+ ak~i)Wk + ... +

+ a~ali + «k~i) ak+ 1 ... asW s (we suppose that k s s)
We put
The above mentioned identity is written
with
196 ChapterS
After division by ~, we find
But the vectors W l' ... , Ws are standard and lineary independent. As Vk is also
standard, its decomposition in the frame WI' ... , Ws has standard coefficients. Every
scalar A.. , i - 1, ... , k and A.kak+1' <Xs is standard. As the a1 are inflnitesimals,
•••, A.kak+1 ...
we must have ak+1 - 0 , this gives the equivalence between the two decompositions.
1Y..3. About the compacmess of the Grassmannian manifold
1) A little bit more of Non-Standard: the I.S.T. Theory
In the first section, we exposed just enough for an infinitesimal study of the
perturbation problems of Ue algebra laws or, more generally, of the perturbation
problems in linear algebra and in algebraic geometry. of course, nonstandard analysts
does not confine itself only to these simple studies. The axiomatic presentation,
established by E. Nelson and called I.S.T. (for Internal Set Theory), is a universal
presentation of nonstandard analysis. But it can be good in precise application of
A.N .s., like here in the study of the perturbations of algebraiC laws, to group and to
understand what, in the Nelson axiomatic, playa fundamental role.
Let us take, for example, the properties of elementary classical analysis; the Nelson
axiomatic of course permits a wonderful use of inflnitesimals ; but somebody wtJl
reproach us for using too rich an axiom for the use we want, hence the temptation to
find a minimal model (but with risks limited to its own preoccupations). The
infinitesimals in the algebraiC problems which interest us, in fact intervene only in
situations of comparison, and the essentialness of the Nelson axiomatic consists in the
existence of inflnitesimals and into taking the shadowoflimited elements. Perhaps the
most characteristic example of this situation concerns the interpretations of the
famous theorem of Hironaka : that we can summarize as follows: let us suppose that
we give a standard number p of infinitesimal complex which are solutions of a

standard number q of polynomial standard relations; then all of these infinitesimals
can be compared to an infinitesimal Eo This means that they are WIitten like a standard
sum of the powers of Eo This example stresses the role which the infinitesimals take In
algebra: they have only a minor interest in the computational field, but they are very
important in the comparative approximations: the theorem of decomposition Is
made in this perspective.
We will go on to briefly fonnulate Nelson's axiom to gratify the lower of the

quanttficators.
AXIOM I (Idealization)
So R. is an internal (or classical) relation:
~F, 3x - XF [R.(x,y) 'rJ y e F] <=> 3 x{R.(x,y), 'ff y]
vhere 'ff resumes •for all standard •and ~. for all finite standards I
This is translated by : let R. be a classical formula; for finding an element x satisfying
R.Cx,y) for all standarr:l y it is necessary and suffldent that for all finite and standarr:l
parts F wecanflnd x - XF such thatR.Cx,F).
As applications, we deduce from (I) the existence of the infinitesimals in C (that was
announced as an axiom In the first part).
Consequence. There is a finite set containing all the standatds.
This most surprising consequence pennits us to show that all infinite standard sets
contain at least a standard element or that in a standard finite set, all elements are
standard. This will be the key for the study of rigid algebras by infinitesimal methods.
Now let us give the reason for this : every algebraiC variety admits a finite number of
irreducible components; if this variety is standard, then the components are standard.
198 Chapter 5
AXIOM S (Standardization)
Let E be a standard set and P a standard or nonstandard property; then
35 A , 'if' x (x e A ¢:> x e E and PCx)
This axiom permits us to construct standard elements from an internal propriety (i.e.
is expressible in a classical language, which does not use either the standard predicate
or its derivative) or external propriety (which is not internal). From this axiom, we
deduce the existence of the shadow of a limited complex element
AXIOM T (Transference)
This is nothing but the axiom previously enunciated.
2) Shadow of a vector subspace
Theorem 5. Let V be a vector subspace of (Cn (n standard). We suppose that V Is of

dimenslon p. Then the shadowof V, whlchls the only standard set whose standard
elements are shadows of the limited elements of V, is also a vector subspace of (CD
with the same dimenslons as V.
Proof. As the sum and the external product are continuous operations, we can write
for limited vectors of V : "(x+y) - Ox + oy andif ex. is alimited scalar, "(ax) - °ex.Ox. This
permits us to provide OY with the structure of a vector space. To look for its
dimension, one considers a standard scalar product in (CD and an orthonormal basis of
V. Every vector of this basis is limited, admits a shadow, and its orthogonality implies
that these shadows are independent and form a basis of OY .
Remark. This theorem does not translate anything other than the compactness of the
Grassmannian manifold, although here no topology of this variety has been defined.
The Variety of Lie AIgebtas 199
N.4. Some problems of perturbations
1) Perturbations of polynomials
We consider the standard. vector space CD[Xl of complex polynomials of standard

degreen.
Definition 9. Let POO be a standard polynomial ofCD[Xl. We call pertwbatlon of P(X)

for every polynomial ofCD[Xlinfinitely close to P(x), i.e. its coefficlents are lnflnitely
close of those to P(X).
We choose a standard. basis of CD[Xl. If Q(X) is a perturbation of P(x), it admits a

decomposition. As an application, we immediady deduce the theorem of the
continuity of the roots with respect to the coefficients, which is enunciated.here lB
follows:
Theorem 6. The roots of Q(X) are infinitely close to the roots of P(X) and the
multiplicity of the (standard) root of P(X) is the sum of the multiplicIties of the roots
of QeX) which are inflnltely close to the given root of peX).
This theorem is the basis of complex algebraic geometry, which one easily understand
by its easy formulation and by the evident proof in the nonstandard. frame that the
approach of some problems of algebraic geometry, via the N.S.A., can be easily made
(see, for example, [G04D.
2) Perturbations of linear operators
Let E be a complex (or real) standard. vector space with a finite and standard
dimension. The space EndCE) of the endomorphisms of E is also standard and we can
define, as previously, a natural notion of perturbation in this space. Then we can
observe the role of the decomposition for the limited vectors of EndCE).
200 Chapter 5
Definition 10. Let f be a standanl endomorphism of E A perturbation of f is an

endomorphisme gin EndCE) verifying gOO;; fOO for evetyX standazd In E.
Some proprieties of perturbations
( 0 If g Is a perturbation off, It verifies gOO ;; f(X) for all X limited In E. In fact, a linear
mapping being conUnuous, we have for all limited vectors X : fCOX) ;; f(X). From this
wededuce:
gOO ;; f/..0X) ;; f(OX) ;; f(X) .
00 We consider as1andard basis (e~ of E. Then, If gls a perturbation off, the matrix of
g relative to (e~ Is infinitely close to the s1andard matrix of f. In parUcular, the
characteristic polynomials of g and f are infinitely close. From the preceding remarks,
we can affirm
Proposition 12. The eigenvalues of g are Infinitelydose to the eigenvalues of f. If v ~

a limited eigenvector of g assoclated to the eigenvalue A., then ov Is an eigenvector of
f associated to the eigenvalue 0A..
Within the framework of the nons1andard analysis, we are led to call s1andard a Ue
algebra whose cons1ants of structure relative to a S1andard basis of (Cn (n S1andard) are
s1andard..
This can Induce some confusion with the notion of s1andard nilpotent Ue algebra
Introduced in Chapter 2. It is difficult to modify the terminology of one or other of
these notions without loosing the relation with the eXisting literature. As the risk of
confusion between s1andard algebras (about the viewpoint of N.S.A.) and standan:l
algebras (like the nilpotent algebras of a parabolic algebra), absolutely excluded in this
treatise (we will neither talk about standard algebra, nor of s1andard nonstandan:l
algebra !!D, we will conserve for the two notions the word s1andard when necessary,
we explain what it means.
The Variety of lie Algebras 201
Definition U. Let n be standard. A Ue algebra. Jaw ~ on Cn Is standard, If Its constants

of structure relative to a standard basIs ofCn are standard. A perturbation J.1 of J.Io Is a
Ue algebra. on Cn verifying:
J.1(X,Y) == J.1o(X.Y) 'V X, Y standard In Cn •
2) Decomposition of a perturbation
Let J.1 be a perturbation of the standanllaw ~ on cn. As these laws are elements of the
standanl vector space Tn2 ,1 of tensors (2.1) on cn, we can apply the theorem of the
decomposition of limited vectors.
Theorem 7. Let J.1 be a perturbation of the standard Ue algebra. J.1o on cn. Then there
are Independent standard mappings 'Pl ••••• <ilk in 'Jh2 ,1 and there are Infiniteslmals
£1' ••• , Et such that
J.1 - J.1o + £1'P1 + £}£2'P2 + ... + £1 ... EkCJlk •
We will see In the next section the relations which must verify the tensors CJlt , so the
Jacobi conditions are satisfied for J.1. We note now the difference between the
deformations and the perturbations. not from the conceptual viewpoint (these two
objects must represent a close law), but from the viewpOint of the presentation: a
perturbation is a law of Ln and is written as a finite sum of Independent tensors and a
deformation Is an Infinite formal series.
3) Duality between deformation and contraction
Let f be an Isomorphism (not standanl of cn) and ~ a standanllaw of Ln. The law
J.1- f- l 0 ~(f.f) Is certainly nonstandanlandls Isomorphic to ~. If this law is limited
(I.e. admits constants of structure limited with respect to astandanlbasls of CD), then
It admits a (standani) shadow 1hatwe will denote J.1oo j It Is a law of Ln.
202 ChapterS
Proposition 13. The law Iloo is a contraction of ~.
This results from the definition of the contraction. So, conserving the above-
mentioned notations, we note that If Iloo (which Is standard) Is a contraction of the
slandaJd law ~, then there Is a perturbation Il of Iloo Isomorphic to the standard law ~
Theorem 8. Let Il be a nonstandard limited law of 1Jl. Then the shadow of Il is a

contraction of a law ~ if and only if Ilis isomorphic to a standard law.
V, THE TANGENT GEOMETR YTO Ln
v'l. Cohomological spaces and tangent spaces
Let Il be a law of (,11 andOCw Its orbit. We have seen that the tangent spaceTIlOCw to
the variety O(W at the point Il, Is Identified with the space B2(/l,W.
The space of cocycles Z2(Il,W Is Identified with the tangent space at the point /l to the
scheme (,11, otherwise Z2(Il,W Is the fonnal tangent space at /l to 1Jl.
We must now detennlne the equations of the tangent plane to the variety Ln at a
regular point and the equations of the tangent cone In a singular point
v'2. Resolution of the equation of deformations
Let J.It - /l + tCPl + t 2CJl2 + ... + tllc!>n + ... be adefonnation of the law /l of Ln. We write that
The Valiety of Lie Algebras 203
J1t formally verifies the Jacobi identity. First, we recall the notations in1roduced
previously. If Cj) and 'I' are in Tn2,l' \Ve note by Cj) ° 'l'the trllineu mapping vitb. values in
Cn defined by
cpo'!'(X, Y,Z) - q('!'(X, Y1Z) + q('!'(Y,Z1X) + q('!'(Z ,X1Y) +

+ ,!,(q(X,Y1Z) + ,!,(q(Y,Z1X) + '!'(q(Z,X1 Y).
In particuw, Cj) E (,nifandonlyif Cj)oCj) - O. More, if ~ E (,Randif Cj) E TR2,I' then
~o Cj) - allCj).
Now the condition J.lt ° J1t - 0 is equivalent to
J.IP~ + t~Cj)1 + tz(.l2 Cj)]'oCj)1 + ~CJl2) +...+ tzp( l+j-2p

1: <!lI0'PJ + .l
2
pOlPp) +
l<j
+ tzP+l( 1:
l+j-2p+l
Cj)10Cj)j ) + ... - 0,
l<j
which leads to the next system :
~Cj)I- 0,
t Cj)]'oCj)1 + ~CJl2 ,
CE.D.)
1:
l+j-2p
1 Cj)pOCj)p + a llCj)2p - 0 ,
Cj)MlJ + -2
kp
1:
l+j-2p+l
Cj)t«Pj + a llCj)2p+l - 0 .
The system CE.D.) , which has an infinity of equations, can be interpreted as the
necessuyandsufficiantcondltionsthat an element Cj)1 E Z2C~,~ is the first term of a
one parameter famdy, i.e. a defonnatlon of the law ~.
204 Chapter 5
For example, the first obstruction is
i.e. [CPl ° CPl~ - 0, where [ 13 designates the cohomology class in the space H 'C"qD.
We suppose the first obstruction to be cleared up. Then the second obstruction is
written
This is interpreted in the followtngmanner : we give an element CPl E 'Jn2 ,l such that
511CPl - o. The first obstruction is :
Let us choose <Jl2 such that CPPCPl + ~<Jl2 - 0 .
In order that CPl is the first term of a deformation, It is necessary that [CPl ° <Jl.2~ - 0 for
every representative <Jl.2 of the coboundary CPl 0 CPl. It is clear that this does not depend
on the choice of <Jl.2.
Applications.
1. CPl defines an infinitesimal deformation of Jl if and only if
511CPl - 0 ,
CPPCPl - o.
2. CPl is the first term of a deformation of degree 2 Ci.e. all the coefficients of tl are null
for i ~ 3) if and only if
511CPl - 0 ,
[CPPCPl] - 0 ,
[CPlOCP2] - 0 V CP2 ,such that CPlOCPl - 5<p2 •

CP;!l <Jl2 - 0 •
Theorem 9. Let CPl be in Z2CJ.l.,J,D. Then the obstructions for CPl' being the first term
of a defonnation, are In the space H3CJ.I.,~.
Proof. Let J.I.t - J.I. + tCPl + ... + tPcPp + .... Let us suppose that all obstructions up to order
k are cleared up, that is
J.l.toJ.l.t - tk+ l'llk+ 1 + tk+2'11k+2 + ... , .
where
'lip - L
I+j-p
CPloCPj + o)1CPP (+ -21 CPpOCPp if p is even)
Ij + ... + 'Ps0C/>s+l
with i + J - k + 1 and k - 2s or to
0- O)1CPk+l + CPl 0 CPk + ... + CPs-l ° CPs+l + ~ CPs 0 CPs
ifk - 2s-1.
In order that these equations be satisfied, it is necessary that
I+j~+l
y. CJ\0'PI (or 1: qlj0'PI + -2
1 'PsoCJls)
I<j
is a coboundary B3CJ.I.,J,D.
We can show direcdy that, if the obstructions are satisfied up to the order k, then
1: qljo{j>j (or 1: qlj0'PI + ~ CJlsoC/ls) is a cocycle of Z~J.I.,J.I.).
This proves the theorem.

206 Chapter 5
Corollary. If H3(J.I.,~ - O. every element 'PI In ZZ(J.I..~ 1s the f1rst term of a

defomatlon of J.I..
V.3. The equations of perturbations
Let J.I.' be a perturbation of a Ue standaJd algebra J10 which Is decomposed. In
where the CI\ are standard and linearly independent In ')hz, I" Let us write that J.I.' verifies
the Jacobi Identities
J.l.1 - J.l.1 ° J.l.1 - J.l.0 J.I. + £ IS IL'P 1 + ••• + £1 ..• £kSIL'Pk + ~ell'Pl ° 'PI + •.. +
+ ell·" £k'Pl ° 'Pk + ~ell elz 'Pz ° 'Pz + ... + ell elz ..• £k 'P2 ° 'Pk + ... +
+ ... + } ell"· elk 'Pk ° 'Pk .
As every mapping Is standaJd. this equation represents a lineary combination with

nonstandaJd coefficients of standaJd vectors In the space of the trilinear mapping (or
of the 3 cochains). This equation Is reduced to J.I. ° J.I. - 0 and
~'PI- o.
(E.P.) £z SIL'PZ + ·.. +£Z·.. £k SlL'Pk+2"£I'PI°'PI+·"+
1
2
z z
1. £I£Z···£k'Pk°'Pk -0 •
So the first term 'PI of a perturbation Is an element of Z2(J.I.,~. The system (E.P.) Is
Interpreted like this.
Theorem 10. An element 'PI of ZZ(J.I.,~ 1s the f1rst term of a perturbation J.I.' of J.I. if
and only1f there are b1llnearstandard 1ndependent forms cpz ••••• 'PIt andlnfin1teslmals
£1 •••.• ex such that (E.P.) 1ssatisfled.
The variety of Lie algebras 207
Remark. One will note that the resolution of the obstructions to the prolongation of a
cocycle Into a perturbation, makes to occur only a finite number of mapping, but the
prolongaUon of this cocycle in a defonnation is traduced by an infinity of conditions.
Theorem 11. Nontandard linear equation (E.P.) equIvalent to a standard finIte linear
system.
Proof. We will solve the equation (E.P.) by playing on the index k of the
decomposition of J.1' • This integer is called the length of J.1'.
1st case. The length of J.1' Is I,
In this case, we have
(E.P.)
So an element CPl of Z2(J.1,J.1} Is the fIrst term of a perturbation with length Ilf and
only If CPl E Ln.
(We will compare this result with that of the infinitesimal defonnation.)
2nd case. J.1' has a length 2:
We have
J.1' - J.1 + ElCPl + EIE2'P2 and
aJ.1cP1 - 0
(E.P.)
208 Chapter 5
Let us suppose £2 ~ o. Then a lL'P2 - 0 implies 'P1 0 'P1 - 'P1 0 'P2 - 'P2 0 'P2 - 0 and
J.1 + £1'P1 is a perturbation of J.L
Suppose a l1'P2 ~ 0 . Then the vectors ('P1 0'P1 , 'P1 0<1>2 , 'P2 0'P2 , a l1<1>2) fonn a system of
rank 1. Effectively, if the rank of these vectors is equal to 3, there is a linear relation
with standard coefficients between these vectors :
a1 'P1 0 'P1 + a2 'P1 0 'P2 + a, 'P2 0 'P2 + a ... aIL 'P2 - 0 , al standard.
This gives a1 - ~ a1E1 , a2 - a1E1E2 , a, - ~ alE lEI , a... - a1E2 .
As q == 0 and a, Is standard, this Is impossible.

So the rank is at most 2. By the same arguments, one shows that the rank cannot be
equal to 2. Then the rank is 1 and we have :
and E2 + a£1 + bE1£2 + c£~l- 0 .

In this case, (E.P) is eqUivalent to
'P1 0 'P1 - aalL'P2 ,
'P 1 0 'P2 - ba l1'P2 •
'P2 0 'P2 - Ca l1'P2 •
E1 (a+ bE2+ CEl) - E2·
Before we interpret this system, let us introduce some notations relative to the
products of cohomology classes.
Lemma 1. Let q> E Z2 (JI.,JI.). Then q> 0 q> E Z' (JI.,JI.) .
Let us recall that a 3-00chain ,is closed if
o - a"cll(X },X2.X,.X4) -
=JI.(X 1.cll(X 2. X,.X4» - JI.(X 2.cll(X 1.X,.X4» +
+ JI.(X ,.cll(X ltX2.X4» - JI.(X4.cll(X ltX2.X,» - cll(JI.(X 1.X2).X,.X4) +
+ cll(JI.(X },X,).X2.X4) - cll(JI.(X ltX4).X2.X4) - cll(JI.(X2.X,).XltX4) +
+ cll(JI.(X2.X4).X1.X,) - cll(JI.(X,.X4).XltX2) •
for all vectors Xl' Xz. x,. X4·

The lemma is proved by replacing, with q> 0 q> and by using the hypothesis a"q> - O.
We will note [q>21, as the class of cohomologis of q> 0 q>. This class exists only if [q>lz (the
class of q> in H2(JI.,JL) exists.
Lemma 2. Let [q>21, - 0 and let 'I' be a representative of q> 0 q> in 8'(JI.,JL). i.e.
a"'I' = q> 0 q>. Then q> 0 'I' E Z'(JI.,JL).
The proof is long and technical. It is not reproduced here.
As the class of q> 0 'I' only depends on q> • we will denote it [<P'1,. This class exists only
iHq>21, -o. Let us suppose that [<P'1, - o. Let p be suchthat q>o q> = ap. Then we have
a(q>o p.+ q>o '1'0 V) - o. The class of cohomology does not depend on the choice of p
(nor of V). One will denote it [qtl1,. The system (E.P) for the perturbations of length 2
is also interpreted as :
Proposition 14. Let q>1 E Z2(JI., JI.). Then q>1 is the first tetm of a perturbation of
length 2 If and only If
(1) [qrJ - [cp'J - [q>4J - 0,
(2) the tank of the representatives of these classes in 8'(JI.,JL) is equal to 1.
210 Chapter 5
Remarks
1. The first terms of the perturbations of length 1 are directing vectors of the
generative lines to (J1 in ~ One can so write again the results relative to the
perturbations of length 1 :
Let be 'P1 E Z2(Ji.JD. Then 'P1 Is a gene13live lJne In Ji to (,D If and only If 'P1 0 'P1 - o.
2. In the case of the perturbations of length 2. the term 'P1 represents the tangent
vector to the curve described by the laws Ji- Jio + E1 'P1 + E1 E2 'P2. where 'P1 and 'P2
are fixed and c 1 - E2 / (a+bE2+cEl) Cfrom the equation E.P J. Such a curve Is a rational
CUIve at is cubic). the tangent vector of which In Jils given by 'Pl.
2. case. Geneta1 case

Let us consider the equation of perturbations
~
E2 ull'P2 ~
+ ... + E2 ... Ekull'Pk 1
+ 2"E1'P1 122
0 'P1 + ... + 2"E1E2 ... Ek'Pk o 'Pk - o.
This equation represents a linear combination of the standard vectors 'PI 0 'Pj wuh
Infinitesimal coefficients.
Definition 12. The rank of Jils the rank of vectors {8J.l~ •...•8J.l'Pit • 'PI 0 'P1 .'P1 0 ~ •••••
'Pit-1 0 'Pit. 'Pit 0 'Pit}.
Theorem 12. Let Ji be a perturbation of length k. Then the rank of Jils equal to the
rank of the vectors (<Pt 0 ~ 1 SIS j S k-l.
Proof. Let co be a linear form whose kernel contains the vectors (<1\ 0 ~ ,
1 SIS j S k-l. Let us apply co to the linear combination defined by (E.D.) :

0- E1 ... Ekco('Pl ° 'Pk) + E1EiE3 ... Ekco('P2 ° 'Pk) +...+ E1Ei ... E~co('Pk ° 'Pk) +
+ 2E~~CI>2) + 2E2,E~~'P3) +...+ 2E2 ... Eko:(~CJlk).
After division by the Infinitesimal having the smallest module and by taking the
shadow, one obtains
Then the vectors 511OCP2, ... , CPk 0 k are In the kernel of 0) for every fonn 0) whose kernel
contains the vectors (CPI 0 CPJ) lSI S j S k-l. Then the vectors (511Ot) I - 2 •...,k and (CPI 0 CPk)
1- l •...,k. are in the space generated by (CPlo CPJ) l:S i :S j S k-l. This proves the theorem.
The relations of dependence which are consequences of this theorem are :
CPl 0CPl - a1511CP2 •

a12CP 1 0 CP2 + a22CP2 0 CP2 + a32CP 1 0 cP 1 - a2511CP3 •
a1+k-lCPI 0 CPk + ... + a2+k-lCPk-l 0 CPk-l + l:

I<k-l
blJCPI 0 cpJ - ak-15 I1CPk-l .
J<k-l
Consequence. Let CPl be a vector of the tangentfonnal plane In ~ to Ln. Then If CPl Is a
tangent vector to a curve In Ln passing through ~ (or CPl Is a vector of the tangent
line), we have:
Theorem 13. Let CPl E Z2(~.~) and Jet CPl be a tangent vector at ~ to Ln. Then we
have:
Example. Perturbations and contractions
Let ~o be astandardlawandf an isomorphism of GLCn). The law~ - f-l 0 ~ 0 fxf

defined by ~CX,y) - f-l(~CfCX), fCy)), Is isomorphic to the standard law ~o. If this Jaw
has a shadow ~oo then. by definition, ~oois a contraction of ~ and~ is a perturbation of
~oo Isomorphic to the standard law ~o. Thus,

212 Chapter 5
A perturbalion ~ of a standatdJaw ~.. is related with a contraction If and onlylf ~ Is

isomorphic to a standazd law J1o. In this case, ~.. Is a contraction of J1o.
Example: The Heisenberg case

Let us consider the Heisenberg algebra of dimension 2p+ 1. denoted by Hp and
deOnedby
A Ue algebra fJ of dimension 2p+ 1 is provided with acontact form if there is a basiS

exl ••••• ~P+l) such that
~eX2' ~ - ... - ~X2p. X2p+l) - Xl'
1: C~rk
~(X"XI) - k>1 •
Let us consider the isomorphism f of (C2p+l defined by
f (Xl) - E2X l •
f(Xv- EXI with E::O. i-2....,2p+l .
The Jaw ~' defined by ~' - f-l 0 ~ 0 f x f verifies
~' (X2. X,) - ... - ~' (X2p,X2p+l) - Xl.
~' (X.,XJ) -1: EkC~rk. where Ek is a positive power of E.
The shadow of~' is the HeiSenberg algebra. We w1I1 deduce
Theorem 14. Every Ue algebra of dJmenslon 2p+ 1 provided with a contact fonn can
be contracted on the HeIsenberg algebra Hp'
The perturbation ~' of the Jaw of Heisenberg algebra has for its first term the cocycle
CPl defined by
CPl (X2,X,) - ... - cP (X2p,X2p+ 1) - 0 ,

CPl (X 1,X,) - 0 ,
So we have a description of vectors tangent to the corresponding point to the

Heisenberg algebra.
VL THE COMPONENTS OF (,n
VI.l. The components of rP for n ~ 7
As 1..0 is an algebraic variety, it is the union of a finite number of algebraic irreducible

components. One proposes to determine these components for the dimension n
less than 7. The expressed results essentially lay in the next proposition:
Proposition 15. Let F be an irreducJble closed set of 1..0 and ~ a point of F such
that dim F - dim Z2(~,~. Then
(1) ~ is a simple point of the scheme 1..0
(2) F is the only irreducible component containing ~.
Proof. Let ~ E F and let C be an irreducible component containingF. As Z2(~,~ is
identified to the Zariski tangent space of 1..0 (that is, the tangent space in ~ to the
scheme 1..°), the Zariski tangent space in C at the point ~ is contained in Z2(~,~. Then
dim C :s dim Z2(~,~. But
dim F - dim Z2(~,~ :s dim C :s dim Z2(~,~
which gives dim F - dim C and F - C . QED.
214 Chapter 5
In the following classification we have need of some notation:

Notation. sl(2,IC) E9 0 1 designates the irreducible representation of slC2,IC) on the
module 0 1•
Theorem 15. list of the components of Ln
2 2 --
(1) n-2 L is irreductible : L - O(a2) ,
where ~ is the solvable algebra defined by [Xl'~] - Xl" The dimension of the
component is 2.
(2) n - 3 0 is the union of 2 components.

CD O(s1(2,1C» (the Zarlski closure of the orbit of sIC2,IC) (dim 6),
(ll) the set R3 of solvable algebras (dim 6).
n-4 L4 is the union of 4 irreducible components

CD O(s1(2,1C) E9 IC) (dim 12),
CiO O(r2 E9 r2) (dim 12),

CliO F 4, where F4 is the set of tie algebras containing one Abelian ideal of
dimension 3 (dim 12),
Civ) G <I where G <I is the set of tie algebras containing one noncommutative
nilpotent ideal of dimension 3 (dim 12).
n - 5 L5 allows 7 irreducible components
CD O(s1(2,1C) E9 r2) (direct sum) (dim 20),
CiO O(s1(2,1C) E9 02)' where 02 is the Irreducible slC2,C:>-module of

CliO O(J.l.5) ,where J.l.5 is defined by
J.l.5 (e2 ,e3) - el ; J.l.5 (el ,e4) - el J.l.5 (e2 ,e5) - e2 ,
Il.s (e, ,eof) - e, ; Ji.2 (e, ,e.s) - - e.s (dim 20).
Ov) Ks, where Ks Is the set of solvable Ue algebras whose nilrad!calls

commutative and Is of dimension 3 (dim 20,
(v) F.s, where F S Is the set of Ue algebras con1a!nlng one Abelian Ideal
(vi) G.s, where G.s Is the set of Ue algebras whose radical Is fIlIfonn and of
(vii) M.s, where M.s Is the set of Ue algebras whose nilrad!calls of dimension
4 and s Isomorphic to the direct sum HI ED C, where Hils the
3-dimenslonal Heisenberg algebras.
n - 6 (,6 admits 17 irreducible components
(I) O(sl(2,C) ED sl(2,C» (dim 30),
Of) O(sl(2,C) ED HI) (semldirect sum obtained from the action of sl(2,C)
on HI) (dim 30),
Ov) LO (sl(2,C) ED D2) (dim 30) (D2 is the iIreducible sl(2,C) module of
dimension 2),
(vI) RO (Cof) (dim 32),
(vii) RO (H 1 ED C) (dim 30,
(viii) RO (n.of). where n.of Is the 4-dimenslonal fIllfonn algebra (dim 30),
Ox) RO (C.s) (dim 30),
216 Chapter 5
(xi) RO (",4 E9 C) (direct sum) (dim 30),
(XU) RO (n.5) for every 5-d1mensional ndpotent algebra "" (k - 1,.••,6)

(dim 30),
where RO(a.) (resp. Lo(a.) ) Is the set of solvable Ue algebras Cresp. of Ue algebras )
I} - "' E9 t, where", Is the nilradlcal, t Is Abelian and reductive on 9 Cresp.
I} - "' E9 t E9 s, where s is a Levi subalgebraJ whose maximal Ideal with a nilpotent
tadicalls Isomorphic to a..
n - 7 L7 admits 491rredutlble components
co O(sl(2,C) E9 G2 E9 G2), where G2 is the solvable law of dim 2 (dim 42),
CID O(sl(2,C) E9 sl(2,C) E9 C) (dim 42),
CliO O(sl(2,C) E9 04), where 04 is the irreducible sl(2,C)-module of
Gv) O(sl(2,C) E9 02 E9 02) (dim 37) (02 is the irreducible s1C2,C)-module),
(v) LO (sl(2,C) E9 C') (dim 42),
(vI) LO (sl(2,C) E9 H 1) (direct sum) (dim 42),
(vii) LO (sl(2,C) E9 02 E9 C) (dim 42),
(viiI) LO (sl(2,C) E9 0,) (dim 42),
(ix) LO (sl(2,C) E9 H 1) (semidirect sum) (dim 42),

(xV RO (H 1 E9 «:;) (direct sum) (dim 42),
(dim 45),
(xiii) RO (H 1 E9 «:;2) (direct sum) (dim 44),
(xiv) RO (fI,4 E9 «:; ) (dim 43),
(xv) RO (WI) i-I, ... ,5 (see the classification of 5-dimensional nilpotent

algebras 5) (dim 42),
(xvi) RO (fI,61) for every 6-dimensional nilpotent Ue algebra (dim 42).
Sketch of the proof. One begins to determine the components of the variety RD of
solvable Ue algebras. Let us agaJn take the notations introduced in this theorem. Let I}
be aUe algebra and a - D19 E9 fI" where fl,is the nilradical of 9. It is clear that a is the
greatest ideal of 9 whose radical is nilpotent. We denote by L"(a) Crespo R"(a)) the set
of the laws of 1,0 (resp. R D) which have a as the greatest ideal with a nilpotent radical.
Definition 13. A lie algebra I} is decomposable if it can be written I} - 90 E9 n E9 s

where s is a semisimpJe subalgebra, 90 E9 n the radical, n the nJlradical, and 90
Abelian, reductive in 9 and verifying [90 ,5] - o.
In the next section, we will see the part which these algebras play in the
detennination of rigid Ue algebras.
Now, we note that if n :5; 7, then allUe algebras 9 E L"(a) are close (in the Zariski
sense) to a decomposable algebra of L"(a). This is verified by using the classification
of nilpotent Ue algebras of dimensions less than 6. We are led to introduce the subset
LO D(a) (resp. ~ D(a)) of L"(a) (resp. ~(a)) formed by decomposable algebras. The
previous remark mentions the density of LO(a) in L"(a).
218 Chapter 5
Lemma 1. The components of Rn(a) for n S 7 are the closed subset ROCa) with
dim a< n.
Proof. One directly verifies that there does not exist, for n S 7, any component which
contains only nilpotent laws.
The fact that the Rg(a) are component results from proposition CVI.l).
As for the determination of components of Ln, this is made by studying the spaces
LO nCa) and the determination of the orbits of the algebra 9 containing sIC2,IC) as Ue
subalgebras, and constructed from the irreductible representations of sIC2,IC).
VI.2. Rigid Lie algebras
VI.2.1. Definition 14. Let be ~ E Ln. One says that ~ is rigid if its orbit O(~) is a
Zariski open set
Equivalent definition. Let ~ be a standard law of Ln. Then ~ is rigid if every

perturbation is isomorphic to it a.e. O(~) is open for the metric topology).
We have previously shown the equivalence of these definitions. The interest of rigid
Ue algebras in the determination of the components Is revealed in the next
proposition.
Proposition 16. If ~ is rigid, then the Zartski closure of its orbit is an irreductible
component of the variety Ln.
Corollary
1. There does exist only a finite number (for a fixed dimension) of isomorphic dasses
of rigid laws.
2. Every rigid law is isomorphic to a standard law.
We now propose to determine the rigid laws. As there Is only a finite number of
Isomorphic classes, we have here a good family to classify.
VI.2.2. The Theorem of NljenhuIs and RIchaIdaon.
Theorem 16. Let J.l E Ln such that H2(J.l,J.L} - O. Then J.lis tlgld.
Proof. We have seen that the formal tangent space In J.l to Ln Is Z2(J.l,J.L} and the
tangent space In J.l to O(J.l) Is B2(J.l,J.L}. If H2(J.l,J.L} - 0, then Z2(J.l,J.L} Is the tangent space
to Ln. As It Is equal to B2(J.l,J.L}, so the theorem of the Inverse functions shows that 0 (J.l)
Is open In Ln. From where we have proved the theorem.
Remark. The converse is false.

Indeed, consider the law Jlo on L11 defined by
J.lo (X,X1) - lXI, 1- 0, ...,9,

J.lo (XO,XI) - XI , 1-4,5, ...,9,
J.lo (X 11 XI) - XI+1 , I - 2,4,5, ..., 8 (J.lo (X lIX,) - 0) ,
J.lo (X2,XI) - XI+2 , 1- 4,5, ..., 7 ,
the other brackets are null (here (X,XO,X1, ..., Xli) Is a basis of ell) .This law Is dgld.
Effectively, let us consider a perturbation J.l of Jlo. The linear o~rator adllX Is Infinite-
ly close to the operator ad iJC. As this last one Is diagonalizable and has for eigenvalues
CO, 0, 1, ... , 9), then mllx admits for eigenvalues (0, A", A. 1' ... , ~), where A, Is infinitely
close to l. But every nontrivial standa!d vector Y - aXo + bx verifies ~YCx.) - c,X,
, I - 1,... ,9 and C,:F. o. The Jacobi conditions Imply that A" - 0 and mllx Is
diagonaliZable, the eigenvalues being (0,0, A.1' ""~) With A, == I. One chooses abasls
(X, Uo, Ul""'U~ of eigenvectors of m~X verifying U, == X, ,I - 0, ... ,9. Jacobi'S
r
Identities show that the vector J.LCu1, U Isanelgenvectorfor the eigenvalues A, +\ If
this vector Is nonnull, then A, + A., Is an eigenvalue Infinitely close to (I + p. One
deduces the next system:
220 Chapter 5
Ao-O,
t..l + t..2 - t..3, t..2 + t..4 - t..6,
t..l + t..4 - t..5, t..2 + t..5 - t..7,
Note that IlCUI,Uy"# 0 as soon as lloC~,xy"# o. The resolution of this linear system is
written
t..2 - 2t..1'
t..'\ - 3t..1'
t..1 - t..4 + (i-l)t..l' i - 5, ...,9.
One can chooseanewbaseCY, Yo, U1 , ... , U9) so Y== X, Yo == Uo and

Il(Y, UI) - iUl , i - 0, 1, ...,9,
Il(Y o, UI) - 0, i-l,2,3,

Il(Y 0, U I) - UI , i - 4, 5, ... , 9 ,
Il(UI. UI) - all UI+j, with all == 1 , i - 2, 4, 5, ... , 8 ,
a21 == 1 , i - 4, 5, 6, 7 .
One considere a new change of basis
The constants of the structure of Il verify ~4 - ~I - 1 for i - 2, 4, 5, 6, 7, 8. The Jacobi

conditions imply ~5 - ~6 - ~7 - 1 and an - ~3 - aq - 0 for 3 :s; is; j. So Il is
isomorphic to 110 and the last one is rigid.
Now we can calculate the dimension of H2(f.1o,f.1o) by using either a program of formal
calculus,or a reduction via the Hochschild-Serre series, to the calculus of H2(n,n),
where n is the nilra.cUcal. One show
dim H2(~o'f.1o) - 1 .
Conclusion. There are rigid Ue algebras whose second group of cohomology Is not
null.
Let us note that the above-mentioned example is the most simple example known of
rigid algebra with nonnull cohomology. We could show directly that all rigid Ue
algebra of a dimension less than 8 have a trivial group of cohomologies.
VI.2.3. The rigidity of a semisimple complex Lie algebra. Rigidity of Borel

subalgebras
The rigidity of a semisimple complex Ue algebra is a consequence of the theorem of

Nijenhuis and Richardson. Indeed, one has the next result :
Proposition 17. If 9 Is a semlslmple complex Ue algebra, then dim H2(9,f;J) - o.
We don't expose the classical proof of this theorem. We prefer to prove directly the
rigidity of 9 to understand better the phenomen on which Implies the rigidity of
simple and semisimple Ue algebras.
Theorem 17. Every semlslmple complex Ue algebra Is rigid
Proof. Let f.1o be a law of a semisimple complex tie algebra. Then every perturbation
~ of f.1o is also semisimple. Effectively, If Ko (resp. K) designates a KillingCartan form
of f.1o (resp. K), the form K verifies K == Ko. Then it is nondegenerated and verifies
OK - Ko. As ~is semisimple, ltadmits a Weyl base (U a , Va' Za)aeD verifying:
222 Chapter 5
J.I.(U a,Ya) - a{H a)Z a ,

J.I.(Ua,Za) - - a{Ha)Ya ,
J.I.(Ya,Za) - 2 Ua ,
and the scalars a(H~, and other constants of structure only depend on the weights of
the roots. One knows that a(Ha) ¢ 0 and that
K(Ya,Zp) - 0, K(Ya,Yp) - 0,
K(Ua,Zp) - 0, K(Ya,Y a ) - -2,
K(Ua,Yp)-O, K(Za,Zp)-O,
K(Ua,U a ) - - a(HaL K(Za,Za) - -2.
Let (au ..., ar) be a basis of roots. Then K(U(Jj,UCXj) - 0 as soon as i ¢ j.
The vectors UIXI (i - 1, ...,r) , Ya (a E A+) and Za (a E A+) generate the Ue algebra J.L
This basis is orthogonal for K. It also generates the real compact form J.l.1I of J.I. and the
Killing form KII of J.l.1I correspond to the restriction of J.I. to J.l.II, the real law
corresponding to J.I.. Remember that KII is a scalar product
Lemma. The constants of structure of a simple Ue algebra are standard.
For example we can refer to the preceding chapters. The constants of structure of
simple algebras only depend on the roots, and on the lengths of these roots.
Lemma. The shadow of the Weyl basis of J.l.ls a Weyl basis of Ilo.
We show, firstly, that °CJ.l.II) is the real compact form of Ilo. Effectively, if 911 Crespo
90 II) designates the real compact subalgebra of the algebra 9 Crespo 90), we have :
and °CKII) is nondegenerated on 09 11• Let us denote this form by Ko ll . As

Ko IICx,x) :; KCx,x) for all standard vector, we deduce that Ko II is defined and negative
on 0 (911). Then 0 (911) is the real compact form of 90.
Let us show now that the vectors (U a,Y a,Za) are limited and noninfinitesimals.
We consider a standard norm on c n. We put
a -IUaI , b -IVai, c -lZai .
The vectors
U' a - U a yl a _ Ya Z' a _ Z a
a ' b ' c
have a nonnuU shadow in the algebra. 0 (9 11). But
Il~OUla,oY'a):; ~U'a,Y'a) - ah a(Ha) Z'a .
As aCHa) is standard and strictly positive, c/ab is limited. We show also that b/ac and
c/ab are limited. Let us suppose that c/ab is infinitesimal. Then
But
KJi(llo (OUla,OY' a ), OZ'a) - KolI(OUla,llo(OYla,OZ'a» - 20 (b~) KolI(OUla,OUla) .
As KolI(OUla,OUla) < 0, then 0 (b~) - o.

As 0 - KolI(oY'a,llo (OUla,OZ'a» - 0 (2:) KolI(oYla,OYla)a(Ha)
then o(~)-o.
In fact, we can show that every scalar b~' ~, ah is infinitesimal as soon as one of
them is.
We deduce, by doing the products two by two, that the scalars i, t, tare infinitesi-
mall. But K(U' a,U'a) - l K(U a,U a ) - -.1.. a(H cJ:; O. This implies that Kom(U 'a,U' a)-O
a2 il
and U' a - 0 ; that is impossible. So the numbers ah
,~, h~ are limited and not infi-
nitesimal. It is the same for lIa, lib, and lIc . Then the vectors U a' Yao Za have
224 Chapter 5
nonnulshadows in 130m - 09 11. As the shadows of the vectors of the Weyl basis Ua,
Ya , Za ' a > 0 are orthogonal for the scalar product (- Ko m), they are also
independent As dim 90 II - dim 9 11 , then we have a basis of 90 m and a complex basis
for 90. So the constants of structure of 9 with respect to the Weyl basis are standard.
They stay by placing their shadows. We recover, therefore, a Weyl basis of 90. This
proves the lemma.
The theorem is deduced direcdy from this proof. The constants of structure of 9 with
respect to the Weyl basis are equal to the constants of structure of 90 with respect to
the Weyl basiS, given by the shadows of the Weyl basis of 9. Therefore, these two
algebras are isomorphic.
Corollary. Let 90 be a semisimple complex Ue algebra. Every Borel subalgebra b of

90 is rigid.
In fact, b is defined from the roots system. As this is rigid (from the precedent
lemma), the rigidity of b will be deduced from this.
Remark. The rigidity of the Borel subalgebras is shown classically by the Nijenhuis-
Richardson Theorem. But the nullity of H2 is, to our mind, a consequence of a
relatively long and han:! calculus.
vn. CONSTRUCTION OF SOLVABLE RIGID LIE ALGEBRAS
In the preceding section, we have shown the importance of the rigid algebras in the
study of the components of a variety of Ue algebra laws. We propose to approach
here the classification of solvable rigid Ue algebras and to develop a technique for the
construction of these algebras.
VII.t. The decomposability of rigid' algebras
Let 9 be a solvable nonnilpotent tie algebra.
Definition 15. The Ue algebra 9 is called decomposable if there is a decomposition

9 - t E9 n, where n is the nilradical of9 and t an Abelian subalgebra whose elements
are ad 9-semisimple.
Examples.
1. Every algebraiC tie algebra is decomposable.
2. Let 9 be the tie algebra defined by the next law:
Jl(X1, X2) - Jl(X3 ,X4) - ... - Jl(X2p-1 ,A2p) - Xl

Jl(X2 , X3) - - ~ X3
Jl(X2 , X4) - - ~ X4 + X3
Jl(X2 , X2p-1) - - ~ X2p-1
Jl(X 2 , X2p) = - ~ X2p + X2p-1
The nilradical of 9 is generated by (Xl' X3, ... , X2p). It is isomorphic to the abelian
algebra (C2p-l. Every supplementary subalgebraof n is one dimensional. It is abelian
and possesses a basis fonned by a vector of the fonn X2 + U with U En. As adOS +
U) is never diagonalizable, this algebra is not decomposable.
Let us remark that this algebra is not rigid. In fact, consider the perturbation Jl' of Jl
defined by
Jl' (X 1 , X2) - Jl' (X3 , X4) - ...- Jl' (X2p-l , X2p) - Xl

Jl'(X2 ,X3) - ( -~+ £1) X3 ,
Jl' (X2 , X4) - ( -~ - £1) X4 + X3 ,

J.1'(X2' X2p-1) - (-~+ £p) X2p-1,
J.1'(X2 ,X2p) - (-~- £p ) X2p + X2p-1.
226 Chapter 5
This algebra.~· Is decomposable (ad.~X2 is dJagonalizable) and is nonlsomorphic to ~.
Theorem 18. If 9 is a solvable rigid Ue algebra, it is decomposable.
We wtlladmltthis structure theorem, the proof given by carles essentiallyisteclmlcal

and unfortunately does not permit us to understand part of the decomposability in
the rigidity propriety.
We will use this result for the construction of rigid solvable Cnonnllpotent) lle
algebras.
VII.2. Linear systems of roots associated to a rigid Lie algebra
Let ~ be a law of Ln whose associated lle algebra. 9 Is solvable, non nilpotent and
decomposable.
Definition 16. A vector X of t is said to be regular, if the dimension of the linear

space V000 - Ker (ad. X) is minimal :
dim Vooo - min{dim Ker(ad. Y), Yet}.
As ad. X Is diagonalizable and n Is ad. X-invariant, one can find a basis (Xl' X2 , ••• ,
Xp+q ,..., Xn - X) of eigenvectors of ad. X such that Cx 1 ' X2 ,... , Xp+ct Is a basis of n,
(X p+q +1 ,••• , Xn) a baSis of t and Cx p+1 ,••., Xn) a basis of VoCX).
Definition 17. Let X be a regular vector of the decomposable Ue algebra. 9. The

linear system of roots associated to Cx 1 ,... , Xn) is the linearsystem to n-l variables
"t whoseequatlonsare "t + "J - xk if the component of the vector ~x., Xl) on Xkls
nonnull.
We will note this system by SOO or, more simply, S, when the regular vector X Is
fixed.
Theorem 19 (ofthe rank), If the Ue algebra IJ Is rigid, then for allreguJar vector X,
one has rankCSCX») - dim n - 1 ,
Proof, In fact, the eigenvalues A.I of ad X are solutions of SCX); this results in the Jacobi
equations:
~CX, ~cx., ~») - CA.l +~) ~CX.,~) ,
As the Ue algebra IJ is decomposable, none of the operators ad Y for YEt is trivial,

As these operators are simultaneously diagonalizahle, their eigenvalues determine
solutions of the system SCX), It results from this, that the rank of SCX) is less than
dim n-1.
Now let us suppose that we have rgCSOO) < dim n - 1 - p+q-1.
1st case , SCX) admits a solution Cal' <X.z "'" a p + l "'" an-l) such that the vector Cap + l
"'" an-I) is nonnuU, Then there is anonisomorphic perturbation of ~ In fact, let <p be a
bilinear alternated mapping given by cpCX , XI) - <ltXI i-I"", n-1; the perturbation
~ + e<p where e is an infinitesimal complex, is not isomorphic to ~, So IJ is not rigid,
2- case , The solutions of SCX) verifies a l - 0 for i - p+1 "'" n-1. The constants of
structure of ~ can be written like this :
~(X I ,Xj) - al X for i - p+q+ 1 "'" n and j - 1 "", P with aJ '" 0 when i S j S P ,
~(XI,Xj) - L
lSkSp
~jXk for i-p+l"",p+q and j-1"",p,
~(X I ,Xj) - 0 for p+q+ 1 SiS n and p+ 1 S j S n ,
~XI , Xj) - L
lSkSp+q
a1\j Xk for 1 S i < j S P with a~j - 0 if aln + aln ",a~ ,
228 Chapter 5
As rgCS) < dim ........1, there exits a solution of S in the fonn of
v - (~1 ,..., (3p ,0 ,...,0) which is independent of the vectors :
Vp+q+l - (a1p+q+1 , ••• , aPp+q+1' 0,0,...,0) ,..., Vn - (a~ ,..., aPn ,0,...,0) .
Let cp be the bilinear alternative mapping defined by q(X, Xl) - ~IXI i-I,... , P , the
other products being null. We will show that the perturbation J1 + ecp is not
isomorphic to J1 as soon as e is a not zero infinitesimal. Suppose that 9 is rigid; the law
J1 + Ecp is in the orbit of J1 and cp represents a tangent vector to this orbit at the point JL
As the tangent space to the orbit coincides with the space B2(J1,J!} of2-coboundartes
for the Chevalley cohomology of 9, we must have cp - alLh where aIL is the
coboundary operator of this cohomology and h an endomorphism of c n . We put
h(X) - l: hi Xt (recall that X - Xn)

As aj.lh(X , Xt) - 131 Xt ,IS i S P and ~h(X, Xt) - 0 , p+l ~ i ~ n , we obtain
p+q n
~I - l: blal, + l: blalJ ISISp,

p+1 p+q+1
p+q
o-l:bkalkj ISi,jSp,
p+l
p+q
o - l: hk alkj p+ 1 S i ,j S p+q
p+1
This shows that the vector Y -1:;"1 hi XI verifies J.1(Y, XI) - ~I XI for i-I, ..., P
and J.1(y, XI) for I ~ p+ 1 . As for the vector Z - Y - ~+q+1 hi X" It is In the nilradica1
and verifies
This is impossible and the vectors V, Vp+q+ 1 ,... , Vn are not linearly independent
Corollary 1. II 9 is rigid, the rank of the linear system 01 roots does not depend of
the choice of the vectors (Xl ,... , ~) .
Corollary 2. If 9 is rigid, there is a basis (~+q+l ,... , Xn )of t such that the
eigenvalues 01 the operators ad. X. are integers.
In fact, if we give a basis of t, the eigenvalues of the operators ad. X, for every X in this
basis, are solutions of the system seX). As this has integer coefficients, we deduce the
corollary about this.
Remark. The above-mentioned corollary does not imply the rationality of the
solvable rigid lie algebras. In the next section, we will give an example of nonrational
rigid algebra.
Corollary 3. If IJ - n E9 t is rigid, then t is the maximal torus 01 derivations of n.
VIT.3 Rigid Lie algebras whose nilradical is filiform
In this section, we will consider the consequences of the previous theorem on the
classification of solvable rigid lie algebras whose nilradical is filiform. The choice of
this propriety concerning the nilradical is not haphazard: a good approach of the
classification of the nilpotent algebras is studying of the characteristic sequence which
is an invariant related to the problem of the perturbations. Then the characteristic
sequence is classified by a natural order relation and the filiform nilpotent lie algebras
correspond to the nilpotent algebras whose characteristic sequence is maximal. So
the classification of the rigids begins naturally by considering by filiform nilradical.
230 Chapter 5
Definition 18. The nllrac:l1caJ n of is fIlifonn, if there is a vector U in the cone

I}
n - [n, n) such that the inwu1ants of similitude of the nilpotent operator mnU are
equa]to (dim n-l, 1).
In this whole secdon, we wtll suppose that I} Is solvable. not nilpotent, decomposable
and the nllradical n Is filiform. Let us suppose that there Is a basis (Xl' .... Xn) of I}
such that
(a) X - Xn is a regular vector of I}.
(b) (X 1 ..... Xk) is a basis of nand (Xk+l ..... Xn) - X is a basis of t

(c) J1(X 1 •••. , XI) - XI+l for 2 SIS k-l .
Such a basis always exists. If the nilradical n Is a defonoadon of the fillfono graded
algebra.1n defined by a cocycle whose minimal homogeneity degree Is at least ecpd
to 1 en is not a defonoadon of 1he flilfono algebra. Qn)'
Note that. recendy • we have examlnated all rigid Ue algebras whose nllradical is
fllifono and eliminated this previous hypothesis concerning the basis ([G.A)). But the
method Is the same. This general case is based on the classlftcadon of flUfono Ue
algebras having a nonnull rank. We prefer to present here the parUcular case
correponding to the existence of an operator ad X which Is diagonalible on t and
jordanizable on n.
Proposition 18. Let I} be a solvable rigid. Ue algebra whose nilradlcalis fIlifonn and
which is provided with an adapted basis. Then the external torus is of dimension 1
or 2.
Proof. Consider the linear system of roots S corresponding to the basis of the
previous lemma. Its rank is greater or equal to n-'. Indeed, from the definidon of this
system, the equaUons Xl + "I - "1+1 • 2 SiS k-l • fono a subsystem of S. More.
S contains the equaUons Xk+J - O. 1 S JS n-k-l (see the proof of the theorem of
rank). SO the rank is greater than by n-'.
vn.4 Classification when the torus is of dimension 2
Theorem 20. Every solvable rigid lie algebra of dimenslon n, n 2: 4, whose nJbadJcal Is
flMorm and of codJmenslon 2, Is IsomorphJc to the lie algebra defIned by
~(Xn , XI) - IXI 1 SIS n-2 ; ~(Xn.l' XI) - XI 2 SIS n-2 ;

~(X 1 , XI) - XI+l 2 SIS n-3
the undefIned brackets are null.
As the rank of every system of roots Is equal to n-3, the nontrivial brackets are the
square brackets resulUng from the ftlifonnlty of the nilradica1.
The proof of the rigidity of the above-menUoned law,just as the other laws which
appear In this work, Is direct : one sbJdies a perturbaUon, we compute the rank of the
assoda1e<ilinear system and one concludes that the perturbaUon Is Isomorphic to the
given standard law.
vns Classification when the torus is I-dimensional
Theorem 21. Let 1.:1 be a rigId lie algebra whose nilradicalis fl1Jform and of
codimenslon 1. There Is a regular vector X E t such that the eIgenvalues of the
operator ad X are (1, k, k+ 1 ,..., n+k-3, 0) where k Is a posltlve Integer.
Proof. Effecuvely, let X be a regular vector and (XI' ...' Xn_l ) abasisofelgenvectors
of ad X given by Lemma 1. Let (AI ,... , ~_I ,0) be the corresponding eigenvalues. If
Al - 0, the Jacobi ldenUUes Imply At - A, for 2 ~ I, j ~ 0-1. As 1.:1 Is nonnilpotent,
one of the eigenvalues Is nonnull. This shows that ~ ¢ 0 and we have ~(Xl' XJ) - 0
2 ~ I < j . The rank of the associated linear system Is equal to n-2 and the Ue
algebra 1.:1 cannot be rigid. So Al ¢ 0 and the eigenvalues At are, two by two disUnct
232 Chapter 5
From the theorem of the rank, we deduce the existence of two integers i and j such
that !lCX1 , Xj) * 0 . As the product of two eigenvectors is also an eigenvector, there is
an index to such that !lCX 2 , XjO) is a nonnull eigenvector of ad X. The index h of the
corresponding eigenvalue ~ verifies h ~ jo.
Corollary. If 9 Is rigid, then k ~ n-4 en - dim 9).
VII.6. On the classification of rigid Lie algebras
VII.6.1. A parametrization of these algebras
Let us remember the hypothesis (H) : 9 is rIIJId solvable and its nilradical n is of
codimension 1 (other cases have been studied). We note X as a regular vector and
suppose that its eigenvectors (1, k , k+ 1 ,... , n+k-3 ,0) with k E if" . We suppose
thatk ~n-4.
Theorem 22.
1) If n - k+4 and n O!: 6 then 9 is Isomorphic to the Ue algebra defined by :
[Xn' XI] - X I ; [Xn, XI] - (k+i-2)XI , 2 S; i S; n-1 ; [X I , XI] - X I+I , 2 S; i S; n-2
[X2 , X,] - Xn-I .
2) If n - k+5 and n O!: 7 then 9 is isomorphic to the next Ue algebra :

[Xn, XI] - X I ; [Xn, XI] - (k+i-2)XI , 2 S; i S; n-1 ; [X I ,XI] - XI+I . 2 S; i S; n-2
[X2' X3] - Xn-2 ; [X2' x..] - Xn-I .
Remarks
CO The proof of this theorem only goes through the resolution of the equations
applied to the A; deduced from the Jacobi equations.
cli) This theorem gives a start to the classification of rigid Ue algebras whose nilradica1
is filiform and of codimension 1. The case of a codimension greater or equal to 2 has
been deduced. In the previous section. The case of the codimenslon 0 (of rigid filiform
algebras) has been studied by Carles. In an unpublished work, Carles demonstrated
that such algebras do not exist (see also [GAD.
(iiI) As for the general classification of rigid Ue algebras whose nilradicals are of
codimenslon 1, it Is necessary to examine only the rools system (1 • k • k+ 1 •...• n+k-2
• 0). As k Is an Integer. the most natural process Is to vary this parameter k In the finite
Interval determlnated by the above-mentioned theorem.
VTI.6.2. Case k = 1
A direct calculaUon permits us to demonstrate that rigid Ue algebras associated to the

system O. 1. 2 •..., n-2) do not exist
VTI.6.3. Case k = 2
These algebras have been partially studied by Bratzlavsky and Carles. In [BR21.
Bratzlavsky considers the nilpotent Ue algebras n defined by [~. XJl - ~J XIJ with
I+J < n-l - dim n and supposes the follOwing conditions: ~J ¢ 0 for 2 S J S n-'
and ~ ¢ o. Then he construcls the solvable algebras of dimension n whose
nilradical Is n by putting [X. Xl] - I~ for I - 1 •...• n-l . These algebras are
decomposable by construction (9 - n E9 tCX) and are complete : HO(9 • 9) -
H 1(9 .9) - (O}. carles shows that. for n ~ 12, rigids exisls with H2(9. 9) ¢ (O} • These
algebras. In the terminology of the previous section. correspond to the system 0 • 2 •
2 •...• n-l .0). We want to generalize this work.
Let 9 - n E9 t be asolvable decomposable Ue algebra of dimension n such that lis

nilradicalls of I-d.imenslonal, and admitting a vector X whose eigenvalues of adx are
C1 • 2 " •.••• n-l .0).
234 Chapter 5
Theorem 23.
(j) If n S 5, 9 is not rigid.
(Ii) If n - 6 and if 9 is rigid, so it is isomorphic to the next lie algebra
[X , XI] - IXI I - 1 ,..., 5 ,

[X 1 , XI] - XI+l i - 2,3,4
[X2 , X3] - X5 .
(Iii) If n - 7 and if 9 is rigid, it is isomorphic to the next algebra
[X1,XI]-X I+1 , 2SiS5,

[X2 ,XI] - XI+2, i - 3 ,4.
Iv) If 8 S n S 10, the algebra 9 is not rigid.
Proof. We know that there is a basis (Xl' ... ' Xn_l ,Xl such that [X, XI) - i XI for
1 SiS n-l and [Xl' XI) - Xl+l for 2 SiS n-2. This theorem is a consequence ofthe
theorem of the rank.
Now let us suppose n ~ 11 .
We put A.I - al+1I+2 where alJ is defined by [X I , XJJ - a1rl+j. As we have

CltJ - ~J+l +~+lJ '
all the constants of structure are defined from A.t, for i - I , ... , p-l, as soon as
n - 2p+ 2 or n - 2p+ 1 about its parity. In particular, all the Jacobi identities are only
expressed with A.t • And these identities l:: [XI , [~ , Xk J1 - 0 are trivial as soon as
i+j+k> dim n. In the next section, we will resolve the system of polynomial Jacobi
identities by considering the integer i+j+k as a parameter. From this, we will say that
the integer p - i+j+k is the weight of the identity l:: [xl , [~ , Xk)] - O.
Lemma. The JacobI Identities of weight p are the same for all the Ue algebras
verlfylng the given hypothesis as soon as dim n ~ p.
Although Us demonstradon Is trivial, this lemma Is very important. It \VIII permit us to

classify the rigid laws by extension. Resoludon of the system of the Jacobi equadons in
dimension n \VIII depend on the resoludon in a dimension inferior to n.
Jacobi equations with weight less than 12
pS8 no reIa1ions
p-9 - 2AI A,3 + 3A,i - A,2 A,3 - 0

p-l0 mo more reladons
P - 11
P - 12
Resolutions. There is a trivial soludon corresponding to A, - O. From the rank

theorem, the corresponding Ue algebras are not rigid. So we suppose that one of the
constants A, is nonnull and by a change of basts, this constant can be taken as being
equal to 1. If n - 11 or 12, the Jacobi system is only reduced to equa1ions with weight
less or equal to 11. It is an algebraiC system with 3 variables (one of the A, Is equal to
1) and 2 equadons. There is an undetermined parameter A,. As these A, are
parameters of perturbadon, the corresponding laws are not rigid.
The resoludon of the system n - 13 (p - 12) gives the next result:
Theorem 24. The only rlgJd Ue algebras of dJmenslon 13 verlfylng the hypothesIs (H)
are lsomorphJc to one of the foDowing algebras
o ria: [X, Xi] - oc. 1 SiS 12 ; [Xl, Xt] - Xt+l 2 SIS 11

[X2' Xt] - Xt+2 3:Si S 10
236 Chapter 5
It) ri13: [X, Xt] -txt 1 SIS 12 i [Xl, Xt] - Xt+1 2 SIS 11 ;
[X2 , X3] - X5 ; [X2'~] - X6 ; [X2' X5] - 9/10X7 ; [X2' X6] - 4!5Xs ;
[X2 , X7] - 517'X, ; [X2' Xs] - 9/14XlO ; [X2'~] - 7/12Xl l ;
[X2' XlO] - 8/15X12 i [X3'~] - VIOX7 ; [X3' XS] - VI0Xs ;

[X3 , X6] - 3/3~ ; [X3' X7] - V l4X lO ; [X3' Xs] - 5/84Xu ;
[X3 , ~] - V20X12 ; [X4' X5] - V7O~ ; [X4' X6] - V70XlO ;
[X4' X7] - V 84X u ; [X4' Xs] - VI05X12 ; [X5' X6] - V420Xll ;
[X5 , X7] - V420X12 .
These three Ue algebras are two by two nonlsomoIphic and verify :
Proof. If Al '# ° then we can 1ake Al - 1 and the solutions of the system are :
CAl' ~. ~. A4) - (1,0,0,0) which corresponds to the algebra 9 113
CAl'~'~' A4) - (1,VI0,V70,V420) which corresponds to the algebra 9213
If Al - 0, then ~ - ~ - °and the solutions of the system are reduced to :

(Alo A2, A3, A4) - (0,0,0,0) and (Alo A2, A3, A4) - (0,0,0,1) . The first system of
solution does not correspond to anyone rigid algebra, the second one to the algebra
9313.
The second cohomology group is detennlned from the equations In At. The tangent
plane on a given point of the variety Imbedded In C4, parametrized by At and whose
equations are given by the Jacobi polynomials, Is defined by the linear part of these
equations after a lranslation of the origin to given point. For example, the tangent
plane to the pOintCl,O, 0,0) is determined by putting al - Al - I, <Xi - At •1- 2,3,4.
The equations of Jacobi are written :
- 2~ + (2° degree) - °,
2a4 + (2° degree) - °,
- 2a4 + (2° degree) - °.

The Zarlski tangent plane is of codimension 2, and the orbit of the law associaled to
(1,0,0,0) is of dimension 1. This corresponds with the straight line (a, 0, 0, 0). We
deduce dimCH2(lil,' 9113)) - 1.
Jacobi equations with weight less than 14. Preparation of the induction
Jacobi equations with weight p S 14 are parametrized by (AI' ... ' A5), ; also these
parameters are solutions of the equations of weight p S 12. These last ones being
solved, we will write only the simplified equations of weight 13 and 14 by considering
the solutions found
If (Alo A2, A3, A4) - (1,0,0,0), so A5 - °.

If (Al' A2, A3, A4) - (1, VI0, V90, V420), then A5 - V42 .
If (Ab A2J A" AJ - (0, 0, 0, AJ, the equations of weight p - 13 are reduced to :
- A4(- 10A4 + A5) - 0 .
Then the solutions are (Al' A2. A3. A4, A5) - (0,0,0,0,0) which corresponds with
the nonrlgld algebra
(Al' A2, A3, A4, A5) - (0,0,0,0,1) which corresponds to another
rlgid nonlsomorphic algebra which Is noled 9'14 '
Remark. The hypotheses made by carles and 8ratdavsky correspond in the above-
mentioned study to the case Al - 1 .
238 Chapter 5
The study of Jacobi equaUons of weight 14, by the previous remark, is reduced. to the
study of the case A,1 - 0 ; this implies ~ - A., - 0 . These equaUons are wlitten :
A,4 (70A,4 - 19A,S) - 0 ,
A,4 (- 20A,4 + 5A,s) -0,

A,4 (t0).,4 - 4).,5) - 0 .
Then A,4 - 0 and the only soluUon (A,I' A,2, A,3, A,4, A,s) - (0,0,0,0, 1) deflnes the
rlgid tie algebra 9 3 15 •
Conclusion. The 14-dimenslonal rigid lie algebras are isomorphic to one of the four
fol1owing algebras :
9 114 - tED f13 ' 9 214 - tED w13 ' 9'14 and 94 15 .
The 15-dimenslonal rigid lie algebras are isomorphic to one of the three foJJowing
algebras:
9 115 - tED f14 , 9 215 - tED w14 and 9'15
All these algebras are two by two nonisomorphic.
The theorem of classification
The previous process shows how to pass from classiflcadon In dimension n to

classlflcadon In dimension n+ 1 by extension of the Jacobi system. The initialization of
this inducUon Is made in dimension 13 and 14.
Theorem 25. Every solvable rigid lie algebra 9 - n ED t such that the nJlradica1 n Is
fllifonn and ofcodJmension 1 and admitting a regular vector X whose roots are equal
to Cl, 2, 3, ..., n, 0), n ~ 12, is isomorphic to one of the fol1owing lie algebras :
ali) case n - 2p + 1
rf2P+2 associated to (AI, ... , Ap-I) - (0, ... ,0, 1) ,
l:(i2P+2 associated to (Al> ..., Ap-I) - (0, ... ,0, 1, (p-l) (p-2) /2) ;
ali) casen-2p+2,
rf2P+2 associated to (Al> ..., Ap-I) - (0,0, ... , 0, 1) .
All of these algebras are two by two nonisomorphic.
Proof. We suppose that Al - ° and make an induction on n - dim n .
1st case: n - 2p . We suppose that the Jacobi system only admits as asolution CAl' ... '
Ap-~ - CO, ... , 0, I) or (AI' ... ' Ap-2) - (0, ... , 0, 0) . This is the only rigid Ue algebra
corresponding to (Ap ... , Ap-2) - (0, ..., 0, I). Then, if n-2p+ 1, the solutions (AI, ... ,Ap-
2,Ap-I) of the Jacobi identities must verify Al - ... - AP-3 - 0. The equations of weight p
- 2p+l are reduced to AP-2 [Cp-l)Cp-2) AP-2 - 2 Ap-ll- °and this implies CAl'···' Ap-2'
Ap-l) - (0, ... , 0, 0) or (AI' ... ' Ap-2' Ap-l) - (0, ... , 0, 1, (p-2)(p-l)12). The algebras
corresponding to the two last solutions are rigid.
2° case: n - 2p + 1
The hypothesis of induction is : the only solutions of the Jacobi equations are CAl' ... '
Ap-2' '1>-1) - (0, ... , 0, 0) or (0, 0, ... , 1) or CAl' ... ' Ap-2' ~-l) - CO, ... , 0, 1, (p-2)(p-I)/2).
Only the algebras corresponding to the solutions (AI' ... ' Ap-2' Ap _l) - (0, 0, ... , I) and
(Ap ... , '1>-2' Ap-l) - (0, ... , 0, 1, (p-2)(p-l)l2) are rigid. Let us take n - 2p+2. The
equaUons of weight p - 2p+ 1 are reduced to :
240 Chapter 5
2P -I )
(
AP-2 I.
I-I
(p-i)(p-i-l) AP-2 - 2(p-l)Ap-1 -0,
which gives '1r2 - 0 .
VB.7 Oassification of solvable rigid laws in small dimension
We can direcdy verify that there Is no nilpotent rigid Ue algebra of dimension less or
equal to 8. The approach developed in the previous section permits us to construct an
of the solvable rigid Ue algebras, as soon as the dimension is no greater than 8. In fact,
we can determine the other rigid Ue algebras; the only obstacle we can meet comes
from the large number of nonlsomorphic rigid algebras which appear.
Notation: In the next list, the regular vector is noted X and the eigenvectors
associated.1D the eigenvalue ~ - i are noted YI ' Y'. , ....
Dimension 2
Dimension 4
Dimension 5
J1l(X,Y o)-O,
J1l (X , YI) - YI, J1l (Yo, yll) - yll , J1s1 (Y I, yll) - y 2 ,
J1l (X , yll) - yll , J1sl (y0 , Y2) - Y2 , J1sl (X, Y2) - 2Y2 ·
The variety of Ue algebras 241
Dimension 6
III (X , Yo) - 0, III (Yo, y 1) - 0, III (y 1, yll) - y 2 ,

III (X , y 1) - y 1 , III (Yo, yll) - yll , III (Y 1 , y 2) - y, ,
III (X , yll) - yll , III (Yo, y 2) - y 2 ,
III (X , y 2) - 2Y 2 , III (Yo, y,) - y, ,
III (X , y,) - 3Y, ,
116 (X , Y 1) - Y2 , 116 (Y 1 , Y2) - Y, ,

116 (X , Y 2) - 2Y 2 , 116 (Y 1 , Y,) - Y4 ,
116 (X , Y,) - 3Y, , 116 (Y 1 , Y4) - Y5 ,
116 (X , Y 4) - 4Y 4 ,
116 (X, Y5) - 5Y 5,
Dimension 7
1 1 X 1
117 - 115 1l2'
Ill(X, Yo) - 0, IlHy 0 , yll) - yll , III (Y 1, yll) - yI2 ,

III (X , Y 1) - Y1 , III (Yo, y12) - y12' III (Y 1 , y12) - Y, ,
IlHx , yII) - yl! , Ill(Yo, V,) - Y" III (yl 1 , y 2) - Y, ,
III (X , Y2) - 2Y 2 ,
III (X, y12) - 2y12,
Ill(X, V,) - 3Y"
242 Chapter 5
~f (X, Yo) - 0, ~f (Yo, ylo) - ylo, ~f(YI' ylo) - ylI'
~f (X , y 1) - y1 , ~f (Yo, ylI) - yll , ~f(y 1 , ylI) - y 2 ,
~f (X , ylI) - yll , ~'(Y

7 0 ,yll)
1 - 2Y" 1 , ~f (ylo, ylI) - ylI'
~f (X , y. 1) - yll 1 , ~f (Yo, y 2) - y 2,
~f (X , y 2) - 2Y 2 ,
~f (X, YO) - Yo, ~f (Yo, ylI) - yll , ~i (y 1 , ylI) - y 2,
~f (X , y 1) - y1, ~74 (Yo, y 2) - y 2 • ~f (Y 1 • y 2) - y, •

~f (X , ylI) - yll , ~f (Yo, V,) - y" ~f(YI,Y')-Y4'
~f (X, Y2) - 2Y2, ~i (Yo, Y4) - y 4 ,
~f (X , y,) - 3Y, ,
~71x , Y4) - 4Y4
~.f (X , y 1) - y1• ~.f (Y 1. y 2) - y, • ~.f(Y2'Y')-Y7'

~.f (X • y 1) - y1, ~.f (Y 1 , y,) - y, , ~.f (Y2, yl,) - Y7,
~.f(x.y,)-3Y" ~.f (Y 1 , y 4) - y 5 •
~j (X, yl,) - 3yl"
~.f (X • y 4) - 4Y 4 ,
~.f (X , y 5) - 5Y 5 •
~t (X , y 1) - y1, ~t(YI,Y')-Y4' ~t(Y"Y4)-Y7'

~t (X , y,) - 3Y, , ~t (Y 1 , y 4) - y 5 ,
~t (X , y 4) - 4Y 4 , ~t(YI.Y5)-Y6,
~t (X, y 5) - 5Y 5, ~t(YI,Y6)-Y7'
~t (X, Y6) - 6Y6,
~t (X , y 7) - 7Y 7 .
vm. STUDY OF THE VARIETY Ln IN THE NEIGHBORHOOD

OF THE NILRADICAL OF A PARABOLIC SUBALGEBRA
OF A SAMPLE COMPLEX LIE ALGEBRA
We use here the notations of chapter 4, where we have given a description of the
space H2(n,n) in decomposing it in to the sum of the two subspaces H1(n,9/n) and
H2 fund(n,n).
Let us first study the integrability of certain infinitesimal defonnations which belong
to Hl(n,91n).
Let GrC9,n) be the Grasmannian of the subspaces of 9 of dimension n, where

n - dim n. The subset of GrC9,n) fonned by the IJe subalgebras of 9 Is an algebraic
variety. Let us denote it by M. The tangent space to the variety M at the point n is
identified by a subspace of the space Z l(n,9I'n). This gives us an idea for finding the
prolongation of an infinitesimal defonnation <p E ZI(n,9/n) as a family of IJe
subalgebras of 9.
1. Let P be the ntlindex of nand n - dim n.
Lemma 1. Let a E a 1, such that ~ - { y E .:10 I a + yea} - 0 andlet m- 1: CC Xtp

for all roots of a' - {a} at is dear that rt:J.ls an ideal ofcodimension 1 of n). Then the
root vector X-a nonnallzes m, that is [x..a , m 1c m.
Proof. Let Xy be a root vector of m. Let us consider the difference y-a. As ~ - 0

and y - a E ~ , two cases are possible:
(a) y-aEa,
(b) y-aE~ where i~1.
in the case (a), we have [X-a,Xy 1- 0 E m.

In the case (b), we have [x..a,Xy 1 - C_a,y X_a+y E m because y-a :j:. a. This gives the
following lemma.
244 Chapter5
Lemma 2. Every lnflnlteslm.al deformation cp of the form dfml with CI\ En., i- 0, 1
(see chapter 4), is linearly Integrable (It means ~ + 'I' verifIes the Jacobi conditions,
where ~ Is the law of the Ue algebra n).
Proof. Let cp - df mo ,where COo - (a,he) E 0 0 , and we consider the ideal m of

codimension 1 of the Ue algebra n (see lemma I). We put
Yt - Xa - t.[X p,X_p] ,
where t E C. It is clear that Yt , normalizes m. Then the endomorphism ad Ytlm Is a

derlvation of m. This derlvatlon defines an extension n t of the Uealgebra m , which
is the semldirect product C.Yt e m with
[Y t ,x] - ad YtCx) .
ConsIder a law ~ of Ue algebra on the space n defined by
We denote nIt as the obtained algebra.

The mapping, which associates to (b.fPt,X) the vector b.Xa + X, where b E C, is an
Isomorphism of n t on nIt and J11 t verlfies the Jacobi Identities. The family of Ue
algebra nIt constructed from this Is a linear deformation whose differential is
dfmo .
of the same manner, we show that all of the infinitesimal deformation df cot ,where
WI E 01, is also a linear deformation. This proves the lemma.
Lemma 3. Every JniJnlteslmal deformation cp - df O>z' where CI>z E O 2 verifIes cp 0 cp - o.

The proof of this lemma Is based on Lemma 1 and is analogous to the proof of
Lemma 2.
The variety of Ue algebras 245
Lemma 4. Let a, peS, with a + peA, Rp - 0 and let A(a,p) be a subset of A'
formed by the roots, which can be expressed by a and p. Then the subspace
m- L I}a
aeA'-A(a,~)
of the Ue algebra n is an ideal of n. Further, the vector X_~ of the Ue algebra I}

normalizes m.
Proof. It is clear that the subset A' - A(a,p) of A' is closed with respect to the
addition of the roots. It is also clear that 't+ v e A' - A(a,p), If't e A', v e A' - A(a,p).
Then m is an ideal of n. Suppose that x.~ does not normalize the ideal m. Then there
is a root vector Xv so v - p e Ao u A(a,p). We remark that the case v - p e Ao
Is excluded, because R~ - 0. The case v - p e A(a,p) is also excluded,since
ve A'- A(a,p).
Lemma 5. Every infinitesimal deformation <p - df (J)j with a, peS 1 ,a + 2P II! A i s

linearly Integrable, i.e. <p 0 <p - O.
Proof. Let <p-df{J)j, where co4-(a.,-p),and a.PeSl, a+pe A,a.+2PE A ,and

let A(a, P) be the set of positive roots which can be expressed using only
a and p. For tee fixed, let us consider the subspace Q.t C I} formed by elements
of the form
where aye C for 'YeA. We can easily show, that Q.t is a subalgebra of I}. Let
m- L I}a.
aeAI-A(a,~)
246 Chapter5
Then m is an ideal of n and from Lemma 4, an arbitrary element X E Q., normalizes

the ideal m . So adX1m is a derivaUon of m. Denote n, the semi-direct sum m E9G,
de8nedby [X,Y] - adX(Y), where X E Q." Y E m. Consider an algebraic law on the
space n deftned by :
We designate the obtained algebra by ~ ,.

Let X be an element of Q., and let Y E m. Consider the mapping f of n, on n\,
deftnedby:
f(x,y) - ~ a.y. e y + y.
yeA(a,p)
It is an isomorphism and the law ~<lt verifies the Jacobi idenUUes ; so n<lt is a bnear
deformaUon of n.. This proves the lemma.
2. Let W2 be the subset of the Weyl group W defined in Chapter 4. An arbitrary

element w of W2 is writen w - Sa.sp • where a, ~ E S, sa and sp are the reflexlons
(symmetries) with respect to the elements a and ~.
Let be W(A-) ("\ A+ - {'Y1o'Y2} . Let us denote f., the cocycle deflnited by:
fw{XYI'X')'Z) - Xw(1I) ,
where ais the maximal root
Lemma 6. Let Sl - S (n Is the nllradlcal of a Borel subalgebra of cpo Then, every

cocycJe (Infinitesimal defonnation) f., E 2 2(n,n). where w E W 2 defines a linear
defonnation of n (that Is, It verJIJes the condition f.,o f., - 0).
Proof. Let w - sa.sp. where a. ~ E S. Let us consider the ideal

It Is of codimension 1 ideal of the Ue algebra "'. Two cases are possible:

(a) a+~E d,
(b) a + ~ t! d.
Suppose that a + ~ E d. It is easy to see that the endomorphism h of m defined by
is a derivation of the lie algebra m (m is the nilradica1 of the parabolic subalgebra

defined by Sl - S \ {ex} and we can use the description of Der m s1UclJed in Chapter 4).
Lett be a fixed number of c. We put
D - th+ adX a
We have D E Der m. The semicIJrect product C. D ED m, defined by
[D,X] - D(X) - th(X) + ad Xa(X) ,
where X E mdefines a lie algebra defloted "'t. Consider an algebra structure, noted
",mt ,on the linear space underlying the '" associated to the product
[X,Yh - [X,Y]n + tim (X,Y).

The mapping which associates to (b.D,X) the vector b.Xa + X, is an isomorphism of
fit on "''t. Then the Iaw[X,Y~ verifies the Jacobi Identities. The family of lie algebras
"''t obtaines as dlis gives the sought-after linear deformation.
Now suppose that : a + ~ t! d. It Is clear that at least one of the simple roots a, ~,for
example a, is not singular. We consider the mapping h, defined by h(X!y - Xm{&)" We

have h E Der m (see the description of the algebra of derivation of a nilradica1 of a
parabolic subalgebra given in Chapter 4). In the same manner as the previous case, we
can show that the deformation Is linear.
This gives the lemma.

248 Chapter5
Proposition 19. Let 11. be the nilradical of a Borel subalgebm of a simple Lie algebm 9
of lank> 1 (it means S1 - s, card(S) > 1). Then, there is a system of cocycles
(infinitesimal deformations)
{ <PI I i-I, ... , dim H2(11.,11.) }
with <PI 0 <PI - 0 such that the system formed by the cohomology classes of these
cocycles forms a basis of the space H2(11.,11.).
Proof. We choose a basis of the space Hl(11.,9/11.) formed by the classes of

cohomology of the next cocycles (this is possible from the description of the space
H 1(11.,9/11.) given in Chapter 4) :
df~ , where 0)0 - (a,he) , a, e e S,

dfro.z, where 0)2-(a,-a) , aeS,
df 014 , where 0)4 - (a,-~) , a, ~ e S , a + ~ e!!.. , a + 2~ ~ !!.. .
We can suppose that the condition a + 2~ ~ !!.. is fulfilled, because the element f 004 of
H 1(n,IJI11.) is cohomologous to f 00'4' where 0)'4 - (~, -a). Lemmas 2, 3, and 5 show
that all of these cocycles (infinitesimal deformations) are linearly integrable. From the
description of the space H2 fund(11.,11.) (see Chapter 4), every element z of this space is
expressed by
where 3w e «:. From Lemma 6, every cocycle fw verifies fwJw - o. This means that it is
linearly integrable. Then we can choose a basis of the space H2 fund(n.,11.) formed by
the linearly integrable cocycles. This gives the proposition.
Theorem 26. Let 11. the nilradical of a Borel subalgebm of 9. Then the tangent space
to the scheme Ln coincides to the tangent space to the corresponding reduced
scheme at the point 11..
The theorem proceeds from Proposition 19.

3. Here we consider (nonlinearly) a defonnatlon of the nilradical n of a Borel

subalgebra of 9 - slCr+l,e) on a simple tie subalgebraof 9. Let 9 - slCr+ l,e) and let
S - {a l ,... , a r }, where r > 1, the corresponding system of the simple roots. We put
z - d (f (IXI,-IXI) + f (1l2,-IX2) + ... + f (Ilr,-Ilr» ,
where (al,-al) E il2 (see chapter 4) .

Let tEe. We consider the subspacent of the sPace9, fonned by all the matrix (a~
with a, - - for 1 :s; I < j :s; r+ 1. Then n t Is a tie subalgebra of 9

tJ-1 • ~
isomorphic to the simple algebra so(r+ l,e) c 9, fonned by the antlsymetrlc matrix.
Then we obtain afamilyn t of the subalgebras of 9 verifying no - n, n t == soCr+ l,e) for
t #- o. This family gives us a defonnation of n such that its linear part Is z.
CHAPTER 6
VARIETY OF NILPOTENT LIE ALGEBRAS
I. THE TANGENT SPACE OF THE VARIETY OF NILPOTENT

LIE ALGEBRAS
Let Ln be the variety of n-dimensional complex Ue algebras (see Chapter 5). We

denote by Nnp the subset of Ln constituted from the nilpotent Ue algebras whose
nilmdex is less than p. This subset is defined by
It Is an algebraic subset of Ln. If we fix a basis of (Cn the laws of Ue algebras are
identified with their constants of structure which verify some polynomial relations
(see Chapter 5). The nilpotent laws whose nilindex is less or equal to P are also given
by polynomial relations. Then Nnpis aZariski closed subset of Ln. We alsodesignate
the corresponding affine schema by Nnp.
The nilindex of a n-dimensional nilpotent Ue algebra does not exceed n-l. Then
Nnn. l contains all the n-dimensional complex nilpotent Ue algebra laws. For
Simplicity, we note Nnn_l by Nn.
Variety of Nilpotent Lie Algebras 251
Let I} - (CD,/lO) be anllpotent tie algebra. We wantto detennlne the tangent space to
the point /lo END. Suppose that the nillndex of I} Is equal to p. We consider the
flltradon of I} given by the descending central sequence: FIJ - CI-ltj and the ftllradon
of I}, viewed as an adjolntl}-module, given by TIJ - Cp-I+11}. These fUtradons Induce
fUtradons on the spaces of cochains, cocycles, coboundarles and cohomology spaces
(we have studied all these spaces In Chapter 3). We follow the notadons of this
chapter. Let
J.LCt) - /lo + t<lll + t2~ + ...
be adefonnadon of/lo In the varletyNDq wlthq~p. We know that <Ill E Z2(/lo,/lJ. As
J.LCt) Is a law with a nillndex equal 10 q, we have
<Ill (X1,/lO(X2,/lO( .../lO(Xct b xq) ...) + /lo (x t.<IlI (X2,/lO(X3' /lo( ...Xq) ...»). .. )
+ ... + /lO(X1,/lO(X2,.../lO(Xq-2,<IlI(Xct 1, xq) ...») - O.
From this equality, we deduce the existence of a coboundary df E B2(J.1,~ such that
the cocycle '1'1 - <Ill - df verifies 'l'1(F11},FJIJ) c Fi+j+p-qIJ ,that Is <Ill E Fp-q Z2(/lo,/lO)
an this notadon I} Is the tie algebra associated to J.1o). Then, we have the following
proposidon
Proposition 1. Let I} - (CD,~ be a nilpotent Ue algebra of nJlJndex p In the variety

NDq and let <Ill be In Z2(/lo,~ . Then there Is a coboundaty df E B2C/lo,~ such that
<Ill - df - '1'1' where '1'1 Is a cocyc1e satisfying
Corollary. Letl} - Ce D ,J.1o) be a nilpotent Ue algebra with anJ1lndexequa1to p. Then

the Zariskl tangent space In I} to NDq Is the subspace W whose elements are the
cocyc1es <Il EZ2C/lo,J.1o) such that their cohomology classes are In Fp-qH2(/lo'~.
Remark. The space W Is the tangent space In J.1o to the schema NDq. If this schema Is
252 Chapter 6
reduced at the point ~ , then W coincides with the tangent space in J.In to the variety
Nnq .
II. ON THE FILIFORM COMPONENTS OF THE VARIETY Nn
In this section, we study some components of the varietyNn. Let j='n be the subset of
Nn of filiform Ue algebras. As f"n - Nn - Nnn-2' this subset is a Zariski open subset of
Nn. Each component of f"n determines a component of Nn.
Let L - (Cn,J1o) be a n-dimensional Ue algebra. Recall that the elements of Z2(~,~)

can be interpreted as infinitesimal deformations of J1o. The simplest case corresponds
to the case where the deformation is linear, i.e. the deformation has the form
J.l - ~ + V with V E Z2(J1o,J1o). As J.l verifies the Jacobi conditions, the cocycle VaSl
verifies these conditions, and V E {,II. We will denote by ~ the Ue algebra
corresponding to Il- J10 + vCthat is ~ - (Cn. ~ + V)).
Now consider the filiform Ue algebra Ln - (Cn+1, ~) definedin the basis (eO,e 1,... ,e.)
by
J.1o(eo, ei> - el+1' 1 s: is: n-1 .
This Ue algebra has been studied more precisely in Chapter 2.

Let T be the tangentspace at the point Ln to the schema Nn+1. From Proposition 1
and Corollary 2 (Chapter 4, § 11), we can affirm that this space is generated by the
cocyc1es 'IIk.3 + q> with 4 s: s s: k, 2k+ 1 s: s, q> E 8 2(Ln,Ln). The study of the variety Nn+ 1
in a neighborhood of the point Ln is traduced in terms of integrability of the elements
of T ; we can discover if each infinitesimal deformation of Ln can be prolonged in a
deformation.
Lemma 1. Let 'I' - I ~,s'l'k,s be In FoHZC~,V with ~,s"# 0 for a palr Ck, s) saJislyJng
s - 2k+ 1 < n . Then the InfinItesimal de/onnatlon 'I' Is not Integrable.
Proof. A cocycle 'l'is integrable if and only if the 3-cocycle 'I' 0 'I' belongs to B3C~,V.
Let ko be the smallest integer satisfying the hypothesis of the lemma. The description
of the space FoHzCLn,Ln) gives us ko ;j!: 2. Let us determine the coefficient of
ez~zin the expression ('1'.'1') (el.elQ,.eko+ tl- From the choice of ko, this coefficient Is equal
to (_1)~1 k oa1co.3 and is different to O. As the coefficient of eZko+zln 'Po 'P(eloekepeko+l)

is equal to 0 for all 'P E B3CLn,Ln), we have 'I' ° 'I' e: B3CLn,Ln). Then the Infinitesimal
deformation 'l'ls not Integrable.
Corollary. The schema Nn+l Is not reduced at the point Ln.
From Lemma 1, the study of Nn+l around the point Ln Is brought back to problems
of the integrability of Infinitesimal deformation 'P . These cocycles are combinations
of the cocycles 'l'k,s with 2k+ 1 < s < n and 'l'r,n with n - 2r+ 1 Cif n Is odd). We denote
by X the linear space generated by these cocycles. As these cocycles are linearly
Independents, every 'I' E X Is written
'I' - I ~-s' 'I'k-s .
Let M be the affine algebraiC variety In X defined by the relation 'I' 0 'I' - O. The
decomposition
FoHZ (Ln,Ln) - E9 HZ. (Ln,Ln)

20
of the space FoHZCLn,Ln) considered In Chapter 3, gives the decomposition of X :
with
254 Chapter 6
x, - H2,(Ln.Ln} If I ~ 1.
Xo - 0 It n Is even.
Xo - C('I1r,n) If n - 21'+-1 Is odd.
Let J10 be the law of Ln. A filiform law Il E :f'Il+l always admits an adap1ed basis (see
Chap1er 2) and then can be described as being of the form Il - J10 + ~ with ~ E
FOC2(Ln.Lo). ~(eo.e~ - o. 11le equadon (J1o+~)oCJ.1o+~) - 0 gives :
and C~ 0 J10 + J.1o 0 ~ + ~ o~) (eo.~. eV - C~ 0 J.1o + J10 o~) Ceo.e,.e~ - o.

Then ~ 0 J10 + J10 0 ~ - 0 and ~ 0 ~ - O. In consequence. we have ~ E FOZ2(Ln.Lo) and.
(Cn.JD - (Lo)ji . As dim B2(Ln.Lo) - n 2 • a basis of this space Is given by the cocycles q;
defined by f4'tf.eo.e~ - ej . Then ~ E M.
Lemma 2. Every fillform Ue algebra law Il E 1"n+1 Is Isomorphic to a law J10 + ~
obtalned by a Ilnear deformation of J10 associated to an element ~ E M.
The follOwing proposition reduces the study of filiform components of Nn to the

study of the components of M.
Proposition 2. Let C be an Irreducclble component of M. Then GLCC n+lXJ.1o+C) Is

an Irreducible component of Nn+l. Its dimension Is equal to n 2 + dim C. and the
mapping which assoclates C to Its Image In Nn+l Is bijective In the set of Irreducible
components of Nn+l meeting the open set 1"n+1.
Proof. Let C be an Irreducible component of M and Clan Irreducible component of

Nn+l containing the Irreducible set GLCcn+lXJ.1o+C). Then J10 E Cl. Consider the open
set U of Nn+l containing J10 and the laws Il such that the vectors
{eo. el. x'+l(JD - Cadlleo)' (el). 1:sa l:sa n-l} are linearly independent Then. we have
J.1(el.X2(J.1» - L (X1(J.1)XI(J.1) •
w
where the functions x~J.L,) are rational and defined on u. We put
el(J.L,) - (X3(J.L,)eo - el •
el+ 1(J.L,) - (adJ.1(eo))i e 1(J.L,) .
Then the elements eo. e 1(J.L,). e2(J.L,) ..... en(J.L,) determine an adapted basis of the algebra.
J.1 and the mapping J.1 -+ (eO.e1(J.L,), ....en(J.L,)) is a rational mapping. Then, we have
J.1 0 J.1- q(J.L,) 0 (J.1o+~(J.L,)) where q(J.L,) e GLCCn+l) and ~(J.L,) eM. The mapping
J.1 -+ ~J.L,) is rational. Thus. ~U n C1) is an bred.uclble set conlalning C. This bnpUes
then
C 1 - GLCCn+1) 0 (J.1o + C)
Conversely.ifc1lsanirreducible componentofNn+l cutdngthe open setoffiliform

laws. then C1 contains J.1o. The same arguments show that If C is an irreducible
componentconlalning ~Cl n u). then
Finally. if C is an in'educible component of M. we have
dim(GI.(C n+1) 0 (J.1o + C)) - dim(GI.(C n+1) 0 J.1o)+ dim C-

- (n+ 1)2 - dim Der(L n) + dim C - n 2 + dim C .
This gives the proposition.

256 Chapter 6
m. ON THE REDUCIBILITY OF THE VARIETY Nn I n ~ 12
In this section. we consider only the case n ~ 12. The study of the variety Nn for
n s: 11 will be presented. in section VII.
Let n ~ 12 and consider the following closed subselS of M :

(1) Ml Is the closed subset of M defined by the relations
aZ,6 - a3,8 - ... - an-z ,n - 0

Z
If n Is even. and by the relations
az,6 - a,,s - ... - an-3 ,n-l - an-l ,n - 0

Z Z
Ifn Is odd.
(2) Mz Is the closed subset of M defined by the relations
~.2k+Z - (4 + f)a k+l .2k+4 • k - 1.2..... [n~] .

If n Is odd, we add the relations concerning an-l
-,n
.
Z
(3) M3 Is the closed subset of M defined by
al,4 - az,6 - ... - an-4 n-Z - 0 •

Z '
If n Is even. and by
al,4 - az,6 - ... - an-3 ,n-l - 0 •

Z
Ifn Is odd.
(4) M4 ' which is defined only if n is odd, is the closed set of M given by
al,4 - a2,6 - ... - an-5 n-2 - 0 ,an-l - 0

2 ' --2--,n
Lemma 1. The closed subsets Ml , M2, M3 (and M 4,lf n 1s odd) are nontrivial subsets
of M, strictly contalned 1n M.
Proof. We put
'PI = 'If1,4 ,
(n-2)/2
q>z =L ak,2k+2 'lfk,2k+2 with al,4 - 1 ,
k-l
'P3 ='l'n-l n '

2 '
'P4 ='If!!:l. n-l (if n is odd) .

2 '
The existence of the filifonn algebras Rn and W n (Chapter 2) shows that the cocycles
'PI and q>z are linearly Integrable, i.e. 'PI 0 'PI - 0 and q>z 0 q>z - O. It is obvious that also
we have 'P3 0 'P3 - 0 and 'P4 0 'P4 - O. Then, 'Pi E MI for j - 1, 2, 3, 4. Moreover, 'Pi eo
Mj if 1"* J. This proves the lemma.
Lemma 2. We have M - Ml uM2 uM3 1fn1seven,and M - Ml uM 2 uM3 uM 4

if n1sodd.
Proof. Let 'If E M and consider the equalities
By writing that the coefficients of es, e lO ande u are zero, we obtain the follOwing
relations:
258 Chapter 6
- 3 a 22,6 + a,,8 a2,6 + 2 a 1,4 a',8 - 0 ,

6 a\8 - 4 a2,6 a,,8 - a',8 a4,lO + 2 a1,4 a4,lO - a2,6 a4,lO - 0 ,
- 4 a\8 + 3 a',8 a4,lO + 3 a2,6 a4,lO - 0 .
The solution of this system is the union of the three straight-lines in the 4-dimensional
spaces parametrized by (a1,4' az,6' a,,8' a4,l~ whose equations are (t,O,O,o), (0,0,0,t) and
(t, 1~ '7~ '4~O) .We note that the cocycle '11- 1: ak,s 'IIk,s E FoH2(Ln,Ln) with a1,4 -
-az.6 - ~,8 - 0 and ~.10 "# 0 is not integrable if n > 11. For verifying this, it is sufficient
to consider the coefficient of e l2 in ('110'11) (e1,e4,eS)' This coefficient is equal to C3.j,lO)2
and is not equal to zero. But an element of B'(In,In) doesn't possess this property.
By successively using the following relations
('II.'II)(ebe40eS) - 0 ,
('II.'II)(ebe.50e6) - 0 ,
where m - n;4 and using an induction, we find the following assertions:
(a) if n is even, the vector (a1,4, a2,6, ... , an~2 , n) is equal to one of the following vectors :
a1,+(I, 0, ... ,0) ,

311-2 n (0, 0, ... , 1) ,
2 •
(b) if n is odd, the vector (al,4, a2,6, ... ,ant, n-1 ,ant. n) is equal to one of the following:
a1,+(I, 0, ... ,0) ,

an:! n (0, 0, ... , 0, 1) ,

2 •
3n:l. n-I (0, 0, ..., 0, 1,0) ,

2 •
al.+(alo a2 ..... an:;- • 0) ,

where al- 1 , <Xk+I=hak , 1 ~k~nr- .
The lemma is proved..
Theorem 1. The variety :fm (and then also the variety Nm), with m ~ 12, contains at
least three Irreducible components If m Is odd, and at least four irreducible
components If m Is even.
This theorem is a consequence of Lemmas 1 and 2.
IV. DESCRIPTION OF AN IRREDUCIBLE COMPONENT OF Nn+l

CONTAINING ~
Suppose n ~ 11. We denote UI as the ZariskJ open subset of Ml' defined by the
inequall.on a l ... '¢ 0 (we use the previous notadons). It Is clear that Rn e UI'
Definition 1. A cocyc1e 'II - 1: ~ 'IIk,s belonging to FoH2(~,~ Is ca1led

nondegenerated In the layer ko If a.w .. '¢ 0 for some Integer s.
Lemma. Let 'II - 1: au, 'Ilk.. E UI be a nondegenerated cocyc1e In layer 2 andlet So

be the minimal value of the Indexs such thata2,so ¢ O. Then
(a) a3J - 0 , If 1< 2 So - '"
b) (cpo-3)a22,so
a3.2so-" - 2
a I."
260 Chapter 6
(c) ak,2k+l+r - 0 , If k> 3 ,IS r S 280 - 11 .
Proof. First, we show the following assertion:

" - 0 If k ~ 3 and if the degree of homogeneousness of the cocycIe "'k,s is less
than 280-11 (that is If s -2k -1 < 2so-11).
We suppose the converse and let", be a cocycle such that ~,s ~ 0, where
r - s - 2k - 1 < 280 - 11. We can suppose that r is the minimal value satlsfytngtbis
hypothesis and that k is the maximal value such that s - 2k - 1 - r. By writUng that the
coefficient of e 2k+l+r In ",o'!'Ce 1,ek_1,eIJ Is equal to 0, we obtain ~,4 ~,2k+l+r - O. As
~,4 ~ 0, then ~,2k+l+r - O. This is impossible. We deduce the case (a). To verlfy point
(b), It is sufficient to consider the relation ('1'0"') Ce 1,e2,e,) - 0 and to write that
the coefficient of e2s0-4 Is zero. To end the proof, we can show that ak,2k+1+r - 0 if
If k > 3 and r - 2 So - 11. Let k > 3 be and r - 2 So - 11. We consider the relation
<", 0"') (el'ek_1,eIJ - 0 • The previous assertion permits us to write ~,2k+l+r - O.
Lemma 2. Let 'II - 1: ~ 'IIk,s E U1 be a nondegenerated cocycle in layer 2 and

layer 3 and let so, to be the minimal values of the index sandt with a2,~ ~ 0, a."to ~ 0
(from Lemma 1 , we have to - 2 So - 4). Then
(a) a4) - 0 , JI I < 3 So - 8 ;

(280- 7) (sO-3)a'2,so
(b) , JI 3 So - 8 S n .
The proof of this lemma Is analogous to the previous lemma.
Lemma 3. let", - 1: at,s 'I'k,s be a cocycle of U1. Then a2,s - 0, If s S n; 7 .
Proof. If n ~ 14, the lemma [s obvious. We suppose now that the Index s satisfies
s S n; 7 wi1h a2,s ~ 0 and denote by So the minimal value of such Index. The relation
('1/ 0 '1/) (e he"eS) - 0 gives, by writing that the coefficient of e3S0-7 is zero,
and
(- to+3) a2,so a"to + 2a1..~a4,3s0-8 = 0 if s - 7 .
From Lemmas 1 and 2, we have a2,so - O.
Corollary. If n == 1 mod 3, then a..) - 0 .
This depends of the inequalities So > n~ 7 and 3so - 7 < n which can be verified simul-
taneously if So = n~8 .
Lemma 4. We consider the arbitrary scalars 0.1,4. ai,S • .... a 1,n .0.2,50 ..... a2,n satisfyins
So - [n;lO] and 0.1,4 ¢ 0 . Then there is only one cocycJe 'I' - 1: ak,s'l'k,s of U 1 such that
a1,! - 0.1,1 anda2J - a2J·
Proof. We put
n n n
'1/ - L 0.105 'l/1,s + L
s-4 S-So
0.205 '1/205 + L
8-210-4
X,,s 'I/,,s +X4,n 'l/4,n
We choose the values x3,s in order that the condition ('1/0'1/) (el'e 2,e,) - 0 is verified By
writting that the coefficients of vectors e28o-4, e2so-3' __ ., en are zero. we obtain the va-
lues of the variables X"S, where s - 2s 0-4, ... , n. They have the desired form. In
particular, we have:
If n == 1 mod 3, then So - n+8 and we put

~
262 Chapter 6
In other cases, we put X4,n - O. Lemmas 1, 2 and 3 show that the equalities
also are verified. As for as equalities
ey depend on the previous relations and of ('" 0 ",) (e 1,e2,e,) - O. The others relations
(",0 ",) (el,ej,e r ) can be verlfted directly. This gives the lemma.
Corollary 1. The variety U lis lneducible and Its dJmenslon Is equal to 2n - [n16] .
Corollary 2. The dimenslon of the tangent space at the point 0 to the variety U liS
equa/b3 (n - [Df]) lin EO or2 mod 3 and Is equal b3 (n- [n;7]) + 1 lin E 1 mod 3.
Corollary 3. The dimension of the tangent space at an arbitrary point of U 1

to the variety U lis equal to 2n _ [ n;16] .
Finally we have the following theorem :
Theorem 2. Let n ~ 11. The point Rn of the variety Nn+l Is a simple pOint. The
dlmenslon of the JrreducJble component e of N
n+l
passlngthrough Rnlsequa/b
n 2 + 2n - n;16. Moreover, aJJ the points (Inlt, ole with '" E Ulareslmple points.
v. DESCRIPTION OF AN IRREDUCIBLE COMPONENT OF

Nn+l CONTAINING Wn
We denote U2 as the Zariski open set of M2 defined by a l ,4 * o.

Lemma 1. Let be 2 S r S n-3. Then there Is a cocycJe '" - "'1 + "'r + "'r+1 +...+ "'n-3m U 2
such that "'r *0 and "'I E H~L.t.Ln).
Proof. We put
"'1 = L lSisk
~ "'1,21+2.
where
k- 0-:/ . al - 1 • aj - (4 + f)aj+l • J - l .....k-l .
k - n;2 • al - 1 • aj - (4 + f)aj+l • J - l •...,k-l .
Then "'1 "'1 -0

0 and (Ln)ljI\ - W n .
Let r > 2 and consider g E G - GLC(Cn+l) defined by
f!.eo) - eo •
f!.e~ - el + el+r ' i-I •...• n-r ;
f!.e~ - ej • j - n-r+ 1 •...• n .
We have g((L.t~\) - (L.t~ • where '" is a cocycle verifying the hypothesis of the previous
lemma. Now suppose that r - 2. In this case. it is sufficient to consider the linear
transformation h E G defined by the following relations
hCeo) - eo + el • hCel) - el •
hCe~ - [eo,hC~+~] + "'lCel.hCei-l))
to deduce the lemma.

264 Chapter 6
Lemma 2. Let 2 S r S n-l0 and n 2: 12 and consider the following cocycle given by
<PI =~ al,21+2 '1'1,21+2 with al,4 - 1, aj,2j+2 - 2

(4 + ~)aj+l,2j+4 , j - 1,..., n 4 . Then there
is a unique nonnull solution (up a constant factor) of the equation <PI ° X + X ° <PI - 0
belonging to H2P-n'L.t).
Proof. Let x be a cocycle
x = L XI 'l'1,21+l+r E H2r (Ln,Ln)

IS1Sm
with m - .1 (n-r-l) and verifying CJlIOX + XOCJlI - O.

2
By verifying this equation for the vectors Cel,e2,e3)' CeI,e3,e4)' Ce2,e3,e4)' we obtain the
following homogeneous system:
(XI,ir)Xl + (XI,ir)x2 + (XIJr)x3 + (XI,~r)x4 - 0,

(X2,ir)xl + (X2,ir)x2 + (X2,':r)x3 + (X2,~r)x4 - 0 ,
(X3,ir)xl + (X3,ir)x2 + (X3,':r)x3 + (X3,.{.r)x4 - 0 .
The numbers (XI/r) are defined by the formulas :
(X ir) - ..1..._ r+2 + 2r

I, 60 (r+4)(r+5)(r+6) (r+3)(r+4)(r+5)(r+6)'
(XI,ir) - - ~+ (r+;~:+6) ,
(XI,':r) - ~ ,
(XI,ir) - 0 ,
) 1 2 (r+2) + 6 (r-l)
(X2,tCr - 420 - (r+ 5)(r+6)(r+ 7)(r+8) , (r+3)(r+4)(r+5)(r+6)(r+7)(r+8) ,
(X2 ir) - i + ;--=-;-_2-;l-(r;-;+_2-,;:):-;-;--::-;-

, 420 (r+ 5)(r+6)(r+ 7)(r+8) ,
(X ':r) _ 4 + r+6
2, - 35 (r+ 7) (r+8) ,
a2.{.r) .11
, 35
a3,l.r) • 0 ,
i ) 1 2 (r+3) + 6 (r+1)
a3, r • 420 - (r+6)(r+7)(r+8)(r+9) (r+5)(r+6)(r+7)(r+8)(r+9)'
a ,.r). _.l.L + r+5

3, 420 (r+7)(r+8)(r+9)'
a2J,,,,"fr) • 2....
35
We have
al,i.r) al,,.r) al,.{.r)

a2,i.r) a2,,.r) a2,.{.r) *" 0
a3,ir) a3,,.r) a3,.{.r)
for all values r ~ 2. The verification of this assertion is reduced to the verification of the
nonexistence of integer roots greater than 2 of the foUowing polynomial:
fer) • 1392 (r+5)(r+6)(r+ 7)(r+8)(r+9) - 131040 (r+5)(r+6)2 (r+9) +

+ 720 720 (r+ 5)2 (r+6) + 18 000(r+4)(r+ 7)(r+8)(r+9) + 604 800 (r+4)(r+6)(r+9)-
- 423 360 (r+2)(r+9) - 2328 48O(r+3)(r+5) + 15301440 (r+6) .
The supremum of the positive roots of this polynomial is equal to 50. An easy
calculation shows that fer) *" 0 for all intergers r, 2 S r S 50. Then the solutions of this
system generate a l-dimensional vector space in the 4-dimension vector space. The
relaUons
permits us to compute "s, "6, ..., "m from the parameters xl' ":2, x" x4 • This means
that the space of the solutions of the homogeneous system
266 Chapter 6
is, at most, of dimension 1. The previous lemmas have shown that this dimension Is
exacdyone.
Let be 2 S r S n-10. We designate by CPr the cocycle of H2r(~'V, satisfying the

equation : <PI 0 x+ x 0 <PI - 0 and such that <pr(e l ,e2) - e 3+ r Lemma 2 assures the
existence of such a cocycle.
Lemma 3. Letn-9 S r S n-7, n ~ 11 and <PI the cocyc1edeflnedin Lemma 2. Then the
dimension of the space of solutions of the equaJIon <PI 0 X +X0 <PI - ObeJonging to
H2 r(~'~) is equal 10 2.
Proof. Let
x= L XI 'l'1,21+l+r E H2.{Ln,Ln)
lSISm
be a cocyc1e saUsfylng the equation <PI 0 x+ x0 <PI - 0 ( m - n-;-l) . In the case r - n-9,
this equation gives a system of 2 linear homogeneous equaUons at 4 indetennlnates
(see the proof of Lemma 2). In the cases r - n-8 and r - n-7, we obtain an
homogeneous linear equation at , indeterminates. In all cases, the dimension of the
space of the soluuons Is equal to 2. Moreover, we can take XI and ~ as free
parameters. This gives the lemma.
Let n-9 S; r S; n-7 and n ~ 11. We denote by CPr and <p'r the cocycles of H2 r(~'~)
satisfying the equation <PI 0 x + x 0 <PI - 0 and such that
'Pr(et.e2) - er+3 , cMe2le3) - 0 ,

<p',(et.e2) - 0 , <p',(e2le3) - er+S .
The existence of such cocycles Is deduced from Lemma ,. We put

%-6 ='I'1,n-' , CPn-S ='I'1,n-2 , %-4 ='I'1,n-1 ,

CPn-' ='I'1,n , cp'n-6 ='I'2,n-1 , cp'n-S ='I'2,n .
Lemma 4. We suppose that the hypothesis of Lemma 2 is satisfied. Let 1r+1 be a

cocycJe of Z'r+1CLn,Lr). Then there is a unique solution of the equation
CP1 0 X + X 0 CP1 - 1r+1 belonging to C2.-CLn,Lr) and such that xCe1,e2) - o.
The proof of Lemma 4 is analogous to that of Lemma 3.
Lemma S. Suppose that the hypothesis of Lemma 3 is satisfied and let 1r+1 be a
cocycJe of Z'r+1CLn,Ln). Then there is a unique solution of the equation
CP1 0 X + X 0 CP1 - 1r+1 belonging to C2.-CLn,Lr) and such that xCe1,e2) - xCe2,e3) - o.
See Lemma 3.
Let 'I' - '1'1 + '1'2 + ... + 'I' n-3 be a cocycle belonging to u 2 with '1'1 E H21 (Ln,Ln) .
Lemmas 2, 3,4, and5 show that the cocycle 'I' can be uniquely written 'I' = '1'0 + '1'-,
"here
'1'0 - ('I'o)t + ... + ('I'ok-3 ,

('I'o)r - ArCPr , 1:S r:S n-3 ,
'1" =('I'·h + ... + ('II·k-7 ,
and C'II-)r is the unique solution satisfying Lemma 4 or 5 if n-9 S r S n-7 with
1r+1 = L
I+j- r+1
('11I.'I'j+'IIj.'I'I).
The cocycle '1'0 is called the homogeneous part of the cocycle '1'. It is clear that every
cocycle is determined by Its homogeneous part, i.e., if we have a cocycle '1'0 which is a
linear combination of the cocycles
268 Chapter 6
<PI , 'P2 , ..., 'Pn-3 ,

<p'n-5 , <p'n-6 , <p'n-? , <p'n-S ,<p'n-9 ,
and if the coefficient of <PI is different of zero, then there is one and only one element
'II of U2 whose homogeneous part is '110.
This reasoning shows that the variety U2 can be defined by the follOwing relauons :
n-2
3.t,2I+r - (X1,l' al,3+r + 'Y1,l' , i - 2, ... , 2 ,IS r S n-l0 ;
Here the elements 'YI~ , 'Y3,r , 'Y4~ are polynomials whose variables are
al,3+r' (2 S r' S r-l) ,

a2J+r' (n-9 S r' S n-7 , r' S r)
These polynomials have a degree greater than 2. The linear combination of cocydes
<PI ,'P2 , ... , 'Pn-3 • !p'n-5 ' .u, !p'n-9 generates the tangent space at the point Ln to the variety
U2 • As La E G{AJ for every filifonn algebra A E ~+l • we have the follOwing theorem:
Theorem 3. Let n;:o; 11. There exists a unique irreducible component of the variety
Nn+l containing W n. This component is smooth (each point is a simple point) and
its dimension is equal to n 2 + n + 2.
Corollary. Let n;:o; 11 and C be an irreducible component of the variety Nn+l

containing W n. Then dim C - dim G(Wn) - 5.
Variety of Nilpotent LIe Algebras 269
VI. STUDY OF THE VARIETYN~ IN A NEIGHBORHOOD

OF A NILRADICAL OF A PARABOLIC SUBALGEBRA
Let I} be a complex simple Ue algebra,n the niIradJca1 of a parabolic subalgebra p of

I} defined by a subset S1 of the set of simple roots S (see Chapters 2 and 4). We
denote p as 1he nilindex of 11. and n its dimension.
VI.l. On the orbit of n
Theorem 4. Let A+0 - Ao n A+ and let p be the halfsum ofpositive roots lying in A+0
Then
(a) If 9 - Aror Cr andlfS 1 - {ru, where aisaslngularroot, we have
dim G(n) - r2 + n + dim Zen) - dim I} •
(b) In other cases,
dim G(n) - n 2 +n +dim Zen) - dim 9 - L fl+ (1+ sa(~~~;a;y) ,

aeSl ')'E.1o
where ais the maximal root and sa the reflexion (symmetry) with respect to a.
Proof. We define a homomorphism
f:p ~Dern
by putting fCx) - ad x In. This mapping is surjective in case (a) from the description
of Der 11. (see Chapter n. Now assertion (a) results in the equality
dim p - dim I} - dim 11. and Ker f - Zen). For case (b), we put
Ip -{adx/n , x E p}.
For a E S1' we denote d a as the endomorphism of 11. defined by

270 Chapter 6
We know that d a Is a primitive element of the dominating weight sa(lJ) - a of the s-

module Der n (see Chapter 4). We note Wa as the simple s-module generated by this
element From the description of Der n, we have
Dern-~e(e Wa).
aes,.
The centralizer of the Ue algebra n In p coincides with the center Zen). Then we have
dim Ip - dim p - dim zen) - dim I} - dim n - dim zen).
The dimension of the module Wa can be calculated. using the Weyl formula [SE1] :
dim Wa - fl (1 + (Sa(a)-a;y») .
+ (p,y)
yeAo
As dim G(n) - n 2 - dim Der n, we have the required. formula.
Corollary 1. If n Is the nOradlcaJ of a Borel suba1gebra, then

dim G(n) - n 2 - n - 2r + 1,
where r Is the rank of I}.
Corollary 2. Let Hk be the Heisenberg algebra of dimension 2k+ 1. Then

dim G(Hk) - 2k2 + k.
The corollary can be proved as this: consider the Inclusion of Hk In the a1gebral} -
Cr> where r - k + 1 (n - Hk if S1 - (al, where a is a singular root). From the theorem
under consideration, we have :
dim GCHk) - C2k+l)2 + C2k+O + 1- C2r2+r) - 2k2 + k .
VI.2. On the Zariski tangent space at the point n to the variety Nn
Consider the filtration of H2Cn,n) defined by the filtrations of the nilpotent Ue

algebra n and the adjoint module studied In Chapters 3 and 4. From the description of
H2Cn,n) given In Chapter 4, we have :
From Proposition I, we have the follOwing theorem.
Theorem 5. Let n be the nilradicaJ ofa parabollc subaJgebra ofa complex simple Ue
algebra, p be the nilindex of n, and n - dim n, w.lth p S n-3. Then the Zariskl
tangent space at the point n to the schema LD coincides w.lth the Zariskl tangent
space at the same point n to the subschema N Dp+2.
Remarks.
1. For some Ue algebras n, we can accurately state this theorem. For example, If
n - Hk Is the Heisenberg algebra with k > I, then
and the tangent space LD colnddes with the tangent space to NDP+l at the pOint n.
2. In the general case where n Is an arbitrary nilpotent Ue algebra with a nilindex p,

we can establish a weaker result :
If2p + 1 S n, then the tangentspace to (,n coincides w.lth the tangentspace to ND2p at
thepoint n.
272 Chapter 6
This result is deduced dJrectly from the relation F-pH2(n,n) - H2(n,n) (see Chapter 3).
VI.3. On the irreducible components of Nnp containing n
Theorem 6. Let n the nilradlca1 of a parabolic subalgebra of a simple Ue algebra 9,

deflned by a subset Sl of the set S of simple roots; let p be the nillndex of n, and.
n - dim n, card(SI) ~ 5. Then the schema Nnp is smooth at the point n. The
dimension of the vallery Nnp at the pointn is equal tD
where
W 2 being the subset constituted of the elements CO of the Weyl group W such that
coer) (').1+ cA' and card(CO(A-) (') .1+) =2, P being the half sum of positive roots,
PI the halfsum of the roots of .1+0 , and Qpls the last Idea/of the descendingcenllal
sequence of n (Qp - Cp-1n).
Proof. Let 'I' E FoH2(n,n) be an infinitesimal deformation. It is integrable because

cardCS 1) ~ 5 and 'I' 0 'I' - o. Then the billnearmapping J.lo + 'I' defines a point of Nnp'
where J.lois the law of the Ue algebra n. The mapplng
deflned by Kg,'I') - g(J.lo+'I') is a morphism of affine algebraiC varieties. A classical

calculation shows that for the differential of this morphism at the point Mo - CId,O),
the following fonnula Is satisfied :
(df)M1 (X,V) (x,y) - V(x,y) + X(J.lo(x,y» - J.lo(X(x),y) - J.lo(x,X(x» ,
where x, yen., X E End 11.. This impbes that the image of (df)Mo coincides with the
space constituted from the cocycles <p E 2 2(11.,11.) whose cohomologic class Is in
FoH2(11.,11.). This space Is the tangent space to the schema Nnp at the point 11.. As
the Image of (df)M1 belongs to the tangent space T(NRp) of the variety N np at the
point 11., we have
where (Nnp)red is the reduced schema. Moreover, the mapping (df)Mo Is surjective.
Let M - (g1' 'II) be a pOint of the variety G x FoH2(11.,11.) sufficiendynear to the point
Od,O) for the metric topology. It Is clear that
dim (Im(df~) ~ dim(Im(~) .
Moreover, we have
Im(df~ cToo (N~) .
If we consider another point sufficiendy close to Od,O), the dimensions of the

corresponding tangent spaces cannot increase. Then
dim (Im(~)S dim TOO (N~)S dim T{(Mol (N~) - dim (Im(dfMo) .
Then we have
dim (Im(df~) - dim (Im(dfMo) .

The dimension of Im(df) Is stable In a neighborhood of the point Mo. The rank of the
smooth mapping
274 Chapter 6
then is stahle in a neighborhood of the point Mo. In this neighborhood, the image of
the mapping f is a smooth subvariety whose dimension is equal to the dimension of
the tangent space of Nnp at the point f(M o). As the dimension of the variety Nn p at a
point M is less than or equal to the dimension of the tangent space at the same point,
and as this dimension is stahle in a neighborhood of Mo , we have
) - dim Tn (Nn)
dim (Nn pn p.
This means that the point n. of the variety Nnp is simple. Then, as the scheme Nnpis
reduced, it is smooth at the point n.. To determinate the dimension of the variety Nnp
, we note that
:Urn T(N~) - dim G(n.) + dim FoH2(n..n.) .
The condition Card(Sl) ~ 5 implies the follOwing isomorphisms:
(see Chapter 4). The description of the space H2(n..9) and the formula for the
dimension of G(n.) imply the required result.
VI.4. Particular case: n is the nilradical of a Borel sub algebra
In this particular case, Theorem 6 can be exacdy stated as :

Theorem 7. Let n be the nilradical of a Borel subalgebra of a complex simple Ue

algebra q of rank r > 1 and let p be the nilindex of n, n - dim n. Then N°p is a
smooth scheme at the pointn. If the type of q is A 2 , A3 or B2 , then the orbit G(n)
is an open set on the varietyN°p' The dimension of this orbit is equal ,respectively, to
one of the numbers 3,25,9. If the type of q is A4 , C3 , B3 , D4 or G2 , then the
dimension of the variety N°p at the point n is equal to m + n 2 - n - 2r + 1, where
the number m is one of the numbers 7,4,4,6, 1 respectively. In the other cases,
the dimension of the variety N°p at the point n is equal to:
n 2 -n-2r+ 1+~(r2+r-2).
VB. ON THE COMPONENTS OF THE VARIETY Nn
The number of irreducible algebraic components of the variety fJl of n-dimensional

complex Ue algebras Is greater than e nl4 . This boundary relies on the study of rigid
Ue algebras: the Zariski closure of the orbit of a rigid Ue algebra is an Irreducible
component of LO and two non Isomorphic rigid laws determine two distinct
components. The same problem, i.e. the determination of the number of
components also appears for N° because this variety is reducible when n ~7. But, In
this case, the approach in terms of rigid laws cannot be applied: currendy, we know
of no nilpotent Ue algebras In Lo (and in N° ). Then we can find another way to study
this problem; we propose constructing some nilpotent laws which separate the
components and to reveal some components associated to a given characteristic
sequence.
VII.l A nonfiliform component
The subset of filiform laws is an open set (irredUcible or not) in N°. From the
previous paragraph, we have:
If n S 6, the variety N° Is irreducible.

276 ChapterS
If n - 7, the variety N 7 is union component such that one of them contains the open
set of fillfonn laws. In particular, this shows lhat this open set is Irreducible.
If n ~ 7, the variety N° is reducible.

This last property has been proved by revealing at least two components In the open
set of fllifonn laws for n ~ 11 , n - 8, and n - 10. For the 9-dimensional case, this open
set is reducible. This implies the research of another component which does not cut
the set of fllifonn laws so that n is greater than 8.
Theorem 8. If n ~ 8, then there is 10 N° an irreducible component which does not

cut the open set of filiform laws.
The proof is supported by the results which follows. Let (Xl' X2, ... , X~ be a basis of
!CO and J.1 the n-dimensionaille algebra law defined by :
J.1(X loX,) - X'-l , ,,:s; I:S; n and 8:S; n ,

J.1(X2,X O) - X, ,
J.1(X n-3,Xn-l) - X, ,
J.1(X n-3,Xn) - X", ,
J.1(X n-2,Xn-l) - X", ,
J.1(X n-2,Xn) - 2Xs,
J.1(X n-I,X o) - 2X6 + X2 ,
the nondeflned brackets being null.
Proposition 3. The irreducible algebraiC component of Nn passing through this aw

J.1 does not cut the open set of filiform laws.
Proof. Let us consider a nilpotent perturbation J.1' of J.1 (see Chapter 5). As the
characteristic sequence of J.1is (n-2, 1, n, Its J.1' Is equal to (0-2, 1, 1) or (0-1, 1). In this
last case, J.1' Is filifonn. We want to prove that J.1' is flUfonn. It is sufflclentto prove
that If Its center Is one-dimenslonal. then J1' Is not fllifonn. Let us suppose that the
center of J1' Is one-d1rnenslonal. Consider a basis of this center given by a vector Y,
with Y, =X, . The derived subalgebra of J1 Is generated by (~-1'X,. X2). As this
...•
subalgebra Is also of dimension n-2, we can find a basis CY0-1' ... , Y" Y2) for this
satisfying YI = XI. As the vector X1 Is a characteristic vector for J1, there exists a
=
characteristic vector Y1 for J1' such that Y1 Xl and J1'CYl'Y~ -YI-4 for I ~4. This gives
J1' (Y 1, Y,) - ! elYI andJl.' (Y 1. Y2) - ! ~YI
I.? -
with al =0 and el =0. The vector
space generated by the vector Y, Is the center of J1. This gives £Ie, =0 for 1*"3 and
J1' (Y2.Y,) is Infinitely large. Then J1 is not filiform and the component does not cut
the open set of these filiform laws.
VII.2. An estimation of the number of components
Theorem 9. For n sufficlendy large, the number of components of Nn is at least of

ordern.
Let 9 0 be the n-dimensional complex Ue algebra whose law J1 is defined in the basis
(eo' ...• em_I' fl' ...• fk' g1' ...• gk' h1' ...• hkl. where n - 3k + m by :
This Ue algebra is adirect sum of the model filiform Ue algebra Ln and of k patterns
of the 3-dimenslonal Heisenberg algebra.
Lemma 1. dim Der (9~ - 2 k2 + 8 k + 2 m - 1 .
Corollary 1. The dimension of the orbit of 9n is equal to :

7 k 2 + 6 km + m2 - 8 k - 2 m + 1 .
Consider the follOwing 2-cocycles in Z2(90. 9~ given by :

278 Chapter 6
'Pr,s (else.) - (- 1)r-I Cktk_l el+j+s-2r-l

for 1 S r S m - 2 , s ~" and 2r + 1 Cm-6)/2 S sSm - 1
'IIr,s,t (f~s) - h t
for lSr,s,tSk and r¢s¢t
llr,s,t (fr,fs) - h t
for 1 S r, s, t S k and r ¢ s ¢ t
~r,s,t (g~s) - h t
for 1 S r, s, t S k and r ¢ s ¢ t
pt,u(eo.ftl - hu , p't,u(eo,gtl - hu
for 1St, uSn and t¢u
for 1St, uSn and t¢u ;
vt,u(f\l8u) - em-l , v't,u(fJu) - em-l , v"t,u(g..,g.) - em-l
for 1St, uSn and t¢u ;
the unspecified products being nuD.
Lemma 2. Let n be the vector subspace of Z2(9n,9n) generated by the previous

cocyc1es. Then for all 'II En,'ll ¢ 0, the law J.L + 'II of Nn Is not lsomoIphic to J.L.
Indeed, if J.L + 'IIis in the orbit of J.L, then'll is acoboundary, I.e. 'II =lif with f E gl(n).
As, no cocycle of n is a coboundary, the laws J.L + 'II are not isomorphic to J.L.
Now, let Co be an algebraic component passing through 9 n and containing the family
J.L + 'II, 'It 'II E n.

Variety of Nilpotent LIe Algebras 279
Proposition 4. We have :
dim Co ~ M(k ,m) - 2 k' + 7 k 2 + lZ.

~
m2 + 6 km - 10 k _12 m + 11.
8~·
Indeed, dim n - MCk, m). From the previous lemma, this number Is a minimum of
the number of nonorbltal parameters of Co.
Lemma 3. Let 9 be a Ue algebra. which is a direct sum of3 ideals : 9 - 91 E9 92 E9 9,.

Let XI -:j:. 0 E 91 and suppose that X, e: 2C9,), where 2C9~ is the center of 91. Then
billnearaltematedmapplng <p with vaJueson Cn satisfying q(X1,X2) - X, doesnot
belong to the space 2 2C9 .9).
In effect, consider Y E 9, such that ~X"Y) -:j:. 0, where JL Is the law of 9. Then
&P(X 1,X2 ,Y) -:j:. o.
We know that If Im Is the model filifonn algebra, we have dim 22CIm,Im) :s; 2m2 -
2m. By looking at the structure of 9n' we see that exists an Ideal 1m Isomorphic to Im.
On the other hand, the dimension of the components passing through 9n Is smaller
than dim 22C9n,9~. Now Lemma 3 Implies:
Proposition 5. Let C be an irreducible component of Nn conlalnlng 9n. Then :
Proof of the theorem 9. From Propositions 4 and 5, we can give the boundaries of
the dimensions of the component Co. Suppose now that k ~ m. Then, for n
suffidendy large, we have :
280 Chapter 6
N Ck-7, m+2]) < M Ck,n).
If k varies between n/3 and n/4, we reveal a minoration of the number of (not
filiform) components of Nn. This boundary is of the order n/74. This gives the
theorem.
CHAPTER 7
CHARACTERISTICALLY NILPOTENT
LIE ALGEBRAS
The definitions, the first properUes, andsome examples of characteristically nilpotent

Ue algebras have been given In Chapter 2 (see Section III In this chapter). Here we
study these algebras from a geometrical point of view, by considering them as points
of the variety Nn. This study leads us to a stupendous result: almost all nilpotent Ue
algebras are characteristically nllpotents. Then the determination of the
noncharacterlstlcally nllpotents seems natural. We conclude this chapter by giving and
describing this family of noncharacterlstlcally nilpotent filiform Ue algebras.
I. CHARACTERISTICALLY NILPOTENT FILIFORM LIE ALGEBRAS
In this section, we describe a necessary and sufficient condition for a filiform Ue

algebra A - (In)1jI with 'I' E FIH2(~, ~), n ~ 7, to be characteristically nilpotent (recall
that the notations are those of Chapter 4). We have seen that every filiform Ue algebra
Is isomorphic to a Ue algebra (In)1jI for a cocycle 'I' E F1H2CIn,In) if n is even, and
'I' E «: 'l'm,n + FtH2(Ln, Ln) , m - ¥ If n Is odd.
Let 'I' be a nonnuU cocycle belonging to
282 Chapter 7
F1H2 (Ln,Ln) - Ea HI2 (Ln,Ln) .

1:2:1
Then 'I' - 'l'r + 'l'rH + ... + '1'1 with '1'1 E HI~In,In) , 'l'r ~ 0 and r ~ 1. We have called. the
component 'l'r the sill coqcIe of '1'. It verifies 'l'r. 'l'r - 0 Then
Lemma 1. Let'l' be anonnuHcocycle,llneadylntegrable andbeJongingto F1H2CLn.,Ln.).

Then the sJ1J cocycle 'l'r of 'I' also is lineady Integrable.
Let A - (Ln.)", be the filiform Ue algebradeftnedfrom 'l'E F1H2CLn.,Ln.) andlet 'l'r

be the sill cocycle of '1'. The Ue algebra (Inl.r is called. the sill algebra of A - (In).,.
The main result of this secUon is the following theorem :
Theorem 1. Let 'I' be a nonnull cocyc1e linearly integrable and belonging to

F1H2cI'IlOLn). The LIe algebra A - (1.0)", is characteristlcally nilpotent if and only if A
is not Isomorphic to ils sJ1J algebra (Inlvr .
A large part of the proof of this theorem is based on the follOwing lemmas :
First, we prove some of the lemmas Intervening In the proof of this theorem.
Lemma 2. Let 'I' be a nonnull11neady Integrable cocycle belonging to F1H2CLn.,Ln.), and

'l'r ils sm cocycle. Consider a derivatlon d of the sm algebra and do lis
homogeneous component of degree 0, d being considered as an endomorphism of
the graded space
V- Ea VI , VI - (C eo+ (C elo VI - C ~,I- 2 ,..., n.
lSISn
l
91,el -
) { el, If 1 - 0 ,
o , If lSISn •
} 92 (el ) - { eo • If 1- 0 •
Cl+r)el • if lSiSn • }
Proof. Leta E Vk, bE Vm • Onehas
283
l~l ~l
[a.
do([a,b]) -[do(a},b] -[a,do(b}] - [1: d1(a},b] + 1: d 1(b}]-1: d1([a,b]) +
. l~l
+ 'IIr (d(a},b) + 'IIr(a,d(b}} -d('IIr(a,b» .
The left part of this equality belongs to the space Vk+rn as soon as the right part
belongs to the space ED V t . This means that do E Der Ln . The descrtpUon of the Ue
t>k+m
algebra Der In (see Chapter 4) permits us to write
where the endomorphisms t 1 , f:2 and ~ are defined by
tl(e.)-et, ISiSn, tdeo}-O,

t2(eO} - eo, t2(e.) - (i-l)et , 2 SiS n ,
t,(eO}-el, t,(e.)-O, i~1 ,
where (et> is the basis cOlTespOnding to the graduation of V.

Let 'IIr(ek' ek+1) - a~2k+l+r wi1h ak ~ o. As do E Der Ln, we have
d o('IIr(ek, ek+l» - 'IIr(dO(ek}, ek+l) - 'IIr(e k, d O(ek+1}) -
- 'IIr (1: dt(et} , ek+1) + 'IIr(ek' 1: dt(ek+1» -1: d t('IIr(ek, ek+1»

t~l ~l ~l
By comparing the coefficients to e2k+l+r in 1he two parts of this idendty, we see 1hat
A.l - (r+ I) ~ and therefore do - 'AtiJ2 + A-,6 1 • Q.E.D.
Let A - (In),!, be a filiform Ue algebra, with V ~ 0 and 'liE F1H2cIn,In). Consider a

dertvation d E Der A. By a similar reasoning as 1he one of Lemma 2, we can affirm that
do E Der In (also here we consider d as an endomorphism of the graded space
V - EB V t>. Then we have
284 Chapter 7
Lemma 3. Let A' - (LrJ'Vr be the sill algebra of the Ue algebra A - (Ln)'V . Then we
have d 'OE Der A'
Proof. We have
d'O ([a,b] + vr(a,b»- [dlO (a),b] - Vr(d I O(a),b) -[a,dlo (b)] - Vr (a,dlo (b» -
- (-d+d'O) ([a,b] + v(a,b» + [(d-dI O) (a),b] + V«d-d I O) (a),b) + [a,(d-dlo) (b)] +
+ V(a,(d-d lo) (b» + 1: (d lo (-VI (a,b» + VI (d i O(a),b) + VI (a,dlo (b))) .

I>r
Consider this Identity for a - es ' s ~ 1 and b - em , m ~ 1. Then the left hand Is In
V s+m+n as soon as the light hand belongs to E9 V s+m+r+ 1 . This Implies that the left part
1~1
Is equal to zero. Now consider this Identity for a - eo and b - em . Then we have
vrCa,b) - 0 . As d' 0 E Der In, also In this case, the left part Is equal to zero. This proves
that dlo E Der AI •
Lemma 4. Let A - <Ln)", be a f IJlform Ue algebra with 'I' E F1H2(Ln,!.n). Consider a

denvation de Der A. Then
d - Adl o + d 1 + ... + d n , with

dlo(eo) - eo , dlo(e.) - (l+r)el, 1- 1, ..., n ,
This lemma is a direct consequence of Lemmas 2 and 3.
Lemma 5. Let A - CIn)", be a filiform Ue algebra with 'I' E F1H2(In,In) and Jet
d - Ad' 0 + d 1 + ... + dy. be a derlvation of A with A ~ 0 ( we use the notations 01
Lemma 4). Then, there Is a basis fo' f l' ... , fn of the space V such that
[fo, f l] - f l+1 , 1 ~I ~n-l ;
d(fo) - 'Mo ; d(fJ - A{i+r)fl + A1fl+l + ... + An.Jn , 1 ~ I ~ 0-1 .

285
Proof. We have
d(eo) - A.eo + A1.~1 + ... + An,oen
We put fo - e o+ aIel + ... +~en and we choose the scalars a l •..., ~ for having
d(f~ - Afo . This Is possible because the numbers r and A are different to O. By
putting
we find 1he required basis.
Lemma 6. Suppose that the hypotheses of Lemma 5 are satisfied. Then there Is a basis
Zo, Zl' ..., Zn of the space V such that
[Zo ,ZI] - ZI+1 , 1- 1,..., n-l ;
d(Zo) - AZo, d(z0 - A(i+r)zl , I - 1 ,..., n-l .
Proof. We choose a basis fo, f 1, .••, fn given by Lemma 5 and we put
Zoo - fo'
Zl - f1 + a1f2 + ... + a n-1f n ,
Z2 - f2 + a1f, + ... + a n-2f n ,
-----------,
It Is easy to choose aI' ..., ao-l as such, as the basis Cz:t) satisfies the required
conditions.
Now we can retum to the proof of the theorem.
Proof of the theorem. Let A - (10).'1' be a filiform tie algebra with 'I' E F1H2c1o,1o)
and consider A' - (Ln>'I'r Its sill algebra. Suppose that A' Is Isomorphic to A There is
286 Chapter 7
a derivation d of A' defined by
d(eo} - eo, d(et} - (I+r)e. ,IS; I S; n .
This derivation Is no nilpotent. Then A' and thus A, Is not characteristically nilpotent.
As A - (10)'11 Is not characteristically nilpotent, there Is a derivation d of A which is
written d - A.d.' 0 + d 1 + ... + du (notations of Lemma 4). We choose a basis Zo, 2:1, ... , Zn
satisfying the conclusion of Lemma 6. In this basis, the Ue algebra A Is written ~
(10)'11' I.e. gCA) - (1o)'V' ' where g is In G UV) and Is defined by zt - gCel ) , I - 0, 1, ..., n).
The cocycle'll' also belongs to F1H2C1o,1o) and Its sill cocycle is the same as the sill
cocyde 'IIr of'll. Show that 'II' - 'IIr. If It Is not the case, we have
'If - 'IIr + 'lis + ... + 'lit , with 'lis ¢ 0 and s > r .

Then
d('IIs(a,b» -'IIs(d(a) ,b) -'IIs(a,d(b» - A.(s-r)'IIs(a,b) - 0
for aU a, b E (1o)'V' This Is Impossible and the theorem Is proved.
II. CHARAcrERISTICALLY NILPOTENT LIE ALGEBRAS IN

THE VARIETY Nn
Let A - (10)'11 be a ffilform Ue algebra, where 'II is anonnull cocycle of F1HZC1o,1o)

and let 'IIr be the sill cocycle of'll.
Lemma 1. We consJder s > r with s ":I: 2r. If there Is a nonnull cocycJe of
Hs2(Ln,Ln) Ii S2«L n)'IIr ,(LJ'IIr) ,then this cocyc1e Is unique up a nontrivial fa.ctDr.
Proof. Let
CharacterisIica Njpotsnt Lie Algebras 287
There is f E End{I.t~r such that

\(Is(a,b) - f(\(Ir(a,b» -\(Ir(f(a),b) -\(Ir(a,f(b» + f([a,b]) -[f(a),b]-[a,f(b)]
We can suppose that
where (ev is the given basis of V (Section D, because this basis satisfies
[ek,es] - 0 for 1 ~k,s ~n .
More, by adding, if necessary, the derivation Ak ad ek of the Ue algebra (Ln)'I'r to the

derivation f, we can suppose that this last satisfies f (eo) - <X el for some <X E c. We have
As q>"# 2r, we have f(ell - Ael+s+r ,IS is n-S+r and ",does not depend of the index i.
The lemma is proved.
Lemma 2. Let n ;;:: 7 andJet U a open serin the variety Nn with A E U Then there
always there exists a characteristically nilpotent Ue algebra belonging to U.
Proof. If A is characteristically nilpotent, the lemma is proved. Suppose that A is not.

Then from Theorem 1 (Section I), we can suppose that \(1- \(Ir with \(Ir E H2P"Il' In) .
We examine each of the follOwing cases :
(1) r < n-6 , n-6"# 2r ;

(2) r < n-6 , n-5"# 2r ;
(3) r - n-6 ;
(4) r> n-6 .
(1).The dimension of the space H2n~In,In) is equal to 2 ; moreover, every cocyde

of this space also is a cocyde of the Ue algebra (Ln)'I'r. From Lemma 1, there is a co-
cycle \(In-6 E H2n-6 (Ln,Ln) which does not belong to B2«Ln)'Vr' (Ln)'I'r) . Every linear
288 Chapter 7
combination of cocycles 't and 'l'n-6 where 't Is a cocycle linearly Integrable belonging
to F IH2(Ln, Ln), Is also linearly integrable. Then every tie algebra of the fonn (L n )1I/(t)
with'i'Ct) - 'l'r + t'l'n-6 , t :# 0 , Is not isomorphic to the sill algebra (Ln)Vr .
From Theorem 1, the tie algebra (Ln¥l) Is characteristically nilpotent. By considering the
nonnull values of t, when t tends to 0, t :# 0, we obtain characteristically nilpotent tie
algebras belonging to a full neighborhood of A.
(2) The proof Is similar to case (1).
(3)We consider a cocycle q> E H2n-5(Ln,Ln) and not cohomologous to zero, q> being
considered as a cocycle of the tie algebra (Ln)Vr (this Is possible from Lemma D.
Then the tie algebra (~lvF'P Is characteristically nilpotent for all values of t, t :# o.
(4) We can always flnd a cocycIe <pof H 2n_2(L n,Ln) which does not belong 10 S2«Ln)", ,(Ln)",).
For example, If r > n-5 and 'l'r - a'l'I,o-2 + ~'I'2,o, a :# 0, then we take cp - 'l'2,n-1. If r - n-5
with a - 0 , then we take cp - '1'1,0-3. The tie algebra (Ln)VFtCP Is characteristically nil-
potent for all values of t different to O. Q.E.D.
Theorem 2. Let n ~ 7 and C be an Irreducible component of the variety f'0+1 (lfn is

oddmoreover we suppose thatQn e: C). Then there Is anonempty Zariski open set of
C whose elements are characteristically nilpotent Lie algebras.
Remark. Finallye, by considering the classification of noncharacteristically nilpotent

tie algebras, we have could remove the hypothesis Qn e: C and prove the theorem for
any component [GH2]. We have carried this result to the end of this chapter.
Proof of Theorem 2. Recall that an external torus of the derivation of a tie algebra L is
an Abelian subalgebra of Der L whose elements are semlsimple. We know that there is
only a finite number of conjugation classes of maximal torus for a nilpotent tie algebra
L (see Chapter O. Let Tj , 0 $; I $; k, be representatives of these conjugation classes,
. the nu11 torus. Conslde r the closed subset NTIn+1 of N
To bemg n+1 whose elements are
n+1
he T I-invariant laws ; the union of the orbits G Ln+ 1(<<:) x NT I for 1 $; i $; k Is a construe·
tIble subset, i.e. a finite union of locally closed subsets. It Is complementary to a
subsetof Nn+ 1 fonned by characteristically nilpotent laws. Then this last subset is also
289
a constructible subset of ND+l. Now if the theorem is not true, there is a nonempty
open set of C formed by non characteristically nilpotent tie algebras. From the lemma
2, this is impossible. Q.E.D.
Corollary. There always exists a nonempty Zariski open set of Nm , m ~ 7. whose

every element is characteristically nilpotent.
This corollary is a direct consequence of Theorem 2 when m ~ 8. For m - 7, we can

construct directly a nonempty Zariski open set whose elements are characteristically
nilpotent For example, the orbit of the familly n7 I3a (see chapter 2).
III. ON THE NON FILIFORM CHARACTERISTICALLY

NILPOTENT LIE ALGEBRAS
Construction of some families of nonfiliform characteristically

III.l.
nilpotent Lie algebras
Let 9 be a simple Ue algebra such that its type is different to A I' A 2, A 3, A 4, A 5, B2, B3,
C3, C4, D4, and G 2 . Let n be the radical of a Borel subalgebra of 9 (we use here the
notations of Chapters 4 and 6). Recall that, in this case, the spaces FoH2(n,n) and
FIH~n, n) coincide. The follOwing cocycles fa ,[3 , (ex,~) E E ,defined by
eSas~li) , if (y,v) - (ex, Sa(~)) ,

o , i f (y,v);t:(ex,Sa(~)) , )
where E is the set formed of pairs (ex, ~) ,ex, ~ E S, give a basis of F IH2(n , n). In the
definition of E, we suppose that the pair (ex,~) and (~,ex) are identified when ex and ~ are
are not tied by a row in the Dynkio diagram and in this case f a, [3 - f [3,a. Moreover, we
290 Chapter 7
have
In the next Chapter (as In the previous chapter), we wiU Identify the cocycles with
their cohomology classes.
Let <p E FoH 2 (n.,n) . It Is easy to verify that <p 0 <p - 0, I.e. the cocycle <p Is linearly
Integrable. The IJe algebra. defined In the underlying space to n by the bracket
[a,b]cp - [a,bJn + <p(a,b)
Is denoted by nIP'
Theorem 3. Let n be the nOtadlcal of a Borel suba1gebra of a simple algebra 9 whose

type is different to Ar Cr - 1,..., 5), 8 2 , 8 3, C3, C4 and D4
Let <p- 1: 'A.rJro bean element of FoH2(n,n) such that 'A.ro:;l!:O forall OlE E. Then
IDEE
nIP Is a characteristically nilpotent Ue algebra.
Let d E Der nIP and d - do + d 1 +...+ '\>-1 be the representation of d as a sum of

homogeneous endomorphisms of the graded space C 1(n,n). Here, we consider the
graduation of n associated to the filtration of n given by the descending central
sequence (see Section III of Chapter 4). It is clear that do E Der n. From the
description of Der n (see Chapter 4), we have do - ad h , with h E fa.. Let
p-l
g-d-do -1:
1-1
dl
Then
g([a,b]) - [g(a),b] - [a,g(b)] + g[<p(a,b)] - cr{g(a),b] - <p(a,g(b»

+ do(<P(a,b» - <p(do(a),b) - <p(a,do(b» - 0 .
First, consider the case '} of type Ar ( r > 5). For the pairs
(a,b) - (eOlt , ellk+OIt+l) and (a,b) - (eOlt+l , eOlt+OIt+l)
1 < k < n-l, we have :
291
Ci) The projections of the elements [a,gCb)] and [gCa), bl on 91i' where a is a maximal
root, are null because a - <lk. and a - ak+l are notin A.
Cit) From the description of FoH2(n,n), we have q(a, b) - !..eli for some A E ce. Thus
gCep Ca),b)) - 0, because g - d - do - d 1+ ... + dp-l.
afO ep(g(a),b) - 0, because g(a) and be [n,n] .
av) ep(a,g(b» - 0, because g(b) E [nIn,n] .
By writing that the coefficient of eli is in the left part of the above identity, we obtain
AIXk.I1I<+1 (a(h) - 2ak(h) - ak+l (h») - 0 ,

AI1I<+I.11I< (a(h) - ak(h) - 2ak+l (h») - 0 .
Then ak(h) - ak+l (h) for all 1 < k < 1-1 , because Am * o. Moreover, we have
a l(h) + ar(h) - (r-5)ak(h) for all k, 1 < k < r .
Now consider the pairs (a,b) - (e lll ,ea,) and (a,b) - (ea,.2 ,ear) .
The same reasoning gives:
(a - al) (h) - al(h) - a3(h) - 0
(a - a r) (h) - a r-2 (h) - a r (h) - 0 .
This implies
a r (h) - al (h) - (r-3)a2 (h) - 0 ,

a I (h) - a r (h) + (r-3)a2 (h) - 0 ,
and
aah) - 0 , aiN - 0 . Then aih) - aah) - ... - a,(h) - 0
and, therefore, h - o. As we have supposed that do - ad h, do - 0 and any derivation

292 Chapter 7
d of the Ue algebra. nql Is nilpotent
For other Ue algebras 9 of a type different to Art we can show, uSlngthe same
arguments, that nip Is chara.cterisdcally nilpotent We will restrict ourselves to the case
9 - E6·
We consider always the above ldendty but for the pairs
(eal ,eal+~)' (e~, eal+~)' (e~, e~+llf)'

(elXf , elXf+a,) , (ea, , ea,+(6) , (eal' eaz)
This gives
S(h) - 2al (h) - a, (h) - S(h) - al (h) - 2a, (h) - S(h) - 2a, (h) - a4 (h) -
- S(h) - 2a4 (h) - as (h) - S(h) - 2as (h) - a6 (h) - S(h) - al (h) - 2a2 (h) - 0 ,
11rus
S(h) - ath) + 2aih) + 2a~) + 3<X4:h) + 2ajh) + a4h) .
Then al(h) - 0, 1 SIS 6 , this Implies that h - 0 and do - 0 .

Q.E.D.
III.2. The study of special cases
If n Is the nilradical of a Borel subalgebra. of simple algebras of type AI' I - 1, 2, 3, 4, B2 ,

B" C"C4 or D4 , the proof of Theorem 3 Is not correct In fact, In these cases, we have
FcfI~n,n) ::J:. Ftf~n,n) .
Moreover, the descripdon of the cocycles of the space FoH2(n,n) are, In these cases,
different to the descripdons of the other cases (see Chapter 1). However, we can
extend the results of Theorem 3 by studying each one of these pardcul3.r cases.
Characteristica Njpotent Lie AIgebtas 293
Theorem 4. Let 9 be a simple Ue algebra of the following types: A4, A 5, 8 3, C 3, C4,

D4, and let n be the nilradica1 ofa Borel subalgebra of9. Let {fro' ro E E} be a basis of
FoH2(n., n) described in Chapter 4 (see Table 3) . Consider
CPl - l: Aro fro with fro :;t 0 V ro E E ,

roeE
and <Jl:z given in Table 4, and characterized by the fact that J.I. + t CPl does not satisfy
the Jacobi conditions (J.I. is the law of n). Then the Ue algebra m whose BW 1: is the
defonnation 1: - J.I. + tCPl +t2CP2 with t:;t 0, is characteristically nilpotent.
Remark. If 9 is a tie algebra of type AI' ~, ~, 8 2, G2, then we have FoH 2(n,n) - O.
Tab1e4
CJlI
9 CPl ~
i~
e 1X4 /\ e ll2 ~ b~l

. ell-Ill
alvl + 3zV2 + 3:)V3 + N Ill' &-CXl
0
A4 + blul + b 2Uz + b3~ + b 4U4
e 1X4 /\ e ll3 ~ b~l
. ell-Ill
aI, b 3 :;t 0 v alb 4 :;t 0 N Ill' &-CXl
blUl + b:P2 + b3~

b lb 3
C3 e ll2 /\ e ll3 ~ • ell-Ill 0
b l ,b 3 :;t0 N Ill' &-CXl
alvl + blUl + b 2Uz + b3~ a lb 3

e ll\+1l2 /\ e ll2 ~ . ell-Ill
N IlI , &-CXl 0
C3
b lb 3
all b 3 :;t0 e ll2 /\ e ll3 ~ • ell-Ill
N Ill' &-CXl
alvl + l: blllj
1-1
ea, /\ e ll2 ~ alb 7
C4 . e l>-Il1-1l2 0
N 1l1+1l2 , &-CX r CX2
al,b 7 :;t0
294 Chapter 7
II1.3. Characteristically nilpotent Lie algebras in the varietyNnP
The following theorem is a direct consequence of Theorems 3 and 4 of this chapter

and 3 and 4 of Section VI, in Chapter 4.
Theorem 5. Let n be the nilradica1 of a Borel subalgebra of a simple algebralJ ofrank

r, ofa type different to A 1, A 2 , A 3 , B2 and G2 ; let p be the nilindex of n, n - dim n
and we consider F as the subset of NnP constituted from the characteristically
nilpotent Ue algebras described in Theorems 3 and 4. Then G(F) is a component
of the variety NnP and the subset of characteristically nilpotent Ue algebras
of G(F) contains a nonempty Zarlskl open. If IJ is of type A4, B 3, C 3 or 04 , then the
dimension ofG(F) is, respectively ,equal to 90,71,71, 131. In the other cases, we have
-- r2+r-2
dim G(F) - n 2 - n - 2r + 1 +--::--
2
(Recall that G(F) is the orbit of F with respect to the action of GL(n)).
CHAPTER 8
APPLICATIONS TO DIFFERENTIAL
GEOMETRY:
THE NILMANIFOLDS
Introduction
of course, Ue algebras arise in a natural way in the study of transformation groups in

differential geometry. But, nilpotent Ue algebras play an essential role In this field:
they permit the construction of many examples of concrete differential manifolds and,
more precisely, compact differential manifolds. These manifolds are called
nilmanlfolds and their differential calculus is, usually, nothing more than a linear
calculus on the corresponding nilpotent Ue algebra. For example, the classical
algebraic and topolOgical Invariants of a nilmanlfold, such as the De Rham
cohomologlcal classes, are entirely defined In the Ue algebra. This explains why the
exceptional examples of affine manifolds, of symplectic compact non-kahlerlan
manifolds, of naturally reductive Rlemanlann manifolds, and the counter examples of
Uclmerowicz's conjecture on the harmonic spaces are described as nilmanlfolds.
296 Chapter 8
I. A SHORT INTRODUCTION TO THE THEORY OF LIE

GROUPS
1.1. Lie groups
A tie group is, roughly speaking, an analytic manifold with a group structure such that
the group operations
(x,y) ~ Xy-l
are analytic.
If G is a tie group, then the connected component passing through the identity of G is
a connected tie subgroup of G.
We note by La: x ~ ax the left translation by a in G and Ra the right translation bya.
A vector field X on G is called left invariant if it satisfies
(LSXx - Xax ,
where La- is the Jacobian (or the tangent mapping) of f. As
Li; [X,y] - [La"x , La·Y] ,

the set of left invariant vector fields of G is a tie algebra, denoted 9 , called the tie
algebra of G. As a left invariant vector field X on G is known as soon as its value Xe in
the identity e of G of X is given, we can identify 9 to the vector tangent space TeG of
G in e, the bracket on TeG is deduced to the bracket of 9.
X,YE 9.
1.2. Correspondence between Lie groups and Lie algebras
The exponential mapping
exp:9 ~G
297
of the Ue algebra 9 of G In the Ue group G sends straight hnes throught the origin In 9
Into one-parameter subgroups of G. This mapping Is defined as this: let X E 9 .
There exists a unique analytic homomorphism 9 of IIlnlo G such that 9(0) - X. We put
exp X - 9(1). So we have exp t X - 9Ct) andlt Is a one-parameter subgroup of G. We
have the formula
expCt + s) X - exp t X exp s X
for all s. tell and all X In 9.

Let us lllustrate this theorem for the matrix group G. It Is well know that a matrix A
sufficiently close to the matrix identity I. Is expressible as a matrix exponential
functions
A- expX- I+X+X2+ ... +xn + ...

2! n!
We have
exp t X - I + t X + ~? X2 + ... + t'l -xn +... and exp IX - A • exp OX - I .

l- -
2! n!
This explains the choice of the word "exponential" for the mapping between 9 and Its
UegroupG.
Now let x - exp X and y - exp Y be two arbitrary elements of the Ue group G.
where X and Y are In the Ue algebra 9. The campbell-Hausdorff series:
(_I)m-l [xP1YQl xPmyq..]

xy - exp Xexp Y- X+Y+[X.Y]+ L m' ... •
PI.ct 1: (p I + ql) 1t~!
where the summation extends over all nonnegative Integers PI • q • i-I ..... m • m - 1.
2.... and PI + q':t: 0 (exceptm - 0 and 1 which have been written explicitly)deflnes the
multlphcation In a neighborhood of the groug G. Consequently. an arbitrary Ue algebra.
9 (in finite dimension) determines a unique Ue grouP. up to local isomorphism. such
that its Ue algebra is 9.
298 Chapter 8
In particular there is a one-to-one correspondence between a finite-dimensional tie

algebra and a connected, simply connected tie group.
An interesting problem consists in detennining the class 1"X9) of the tie groups having
9 as the Ue algebra. The class 1"X9) contains the unique simply connected Ue group
which appears as the universal covering group of 1"X9).
Example. Let 9 be the one-dimensional (Abelian) Ue algebra. Then 1"X9) contains the
additive group of reals (it is the universal covering) and the multiplicative group of
complex numbers (group of rotations of a circles).
Remark. Consider a Ue algebra 9 and G the corresponding simply connected tie

group. Then every Ue group H whose Ue algebra is 9, can be writen as H - G / r,
where r is a discrete subgroup of the center of G.
1.3. The left invariant geometry. Interpretation of It
The Ue algebra 9 of a tie group G has been defined as the vector space of left invariant
vector fields on G. The dual space rf can be identified as the left invariant pfaffian
fonns on G. Let ol be a differential fonn on G. It is left invariant if
where -f is the transposed mapping of f-.
This pennits us to define an exterior differential d on the vector space 9- by putting
dOl (X,Y) - - OlIX,Yl
where X, Y are in 9 and ol e 9 . We can also extend this differentiation for every p-Ieft
invariant fonn on G (i.e. for the elements of AI1J-).
The NIman1ok:Js 299
So we define a complex of cochains andcohomological spaces, denoted Hl(9) (see

Chapter 3). In particular, we have
H~9) - 9 '
H 1(9) _ 2 1(9) ,
B1(9)
B~9) - (df:finvarianton9)- (O) .
This gives H~9) - 2~9) .
Often this cohomology is called the cohomology of left invariant forms on the Ile
group G. In general, this cohomology doesn't coincides with the De Rham
cohomology of G. But, if G is a compact Ile group, then these cohomologies
coincide. The nilpotent case is examined in the next section.
Remark. The Maurer Cartan equations. We can read the constants of the structure of
a Ue algebra 9 by writing the Maurer Cartan equations on 9-.
Let (X 1 ,... , ~) be a basis of 9 such that
Consider (001 , ••• , COr) as the dual basis of 9-. Then, we have
dcok--1:C~jCOIACOj, k-l, ...,n,
and these equations again givethe constants of the structure of 9. Note that jacobi's
identities correspond to d(d~) - 0 k - 1 ,..., n .
300 Chapter 8
D. THE NILMANIFOLDS
Let G be a nilpotent Ue group; then its Ue algebra 9 is nilpotent. In this section, we

suppose that G is connected and simply connected.
11.1. Definition of nilmanifolds
Definition 1. Let rbe a dlscrete closedsubgroupofG. Then the quolientspace G/r

Is provided with the structure of a dlfferenlial manifold (of dlmenslon equal to the
dimension OJ. It Is called a nllmanlfold.
An interesting problem consists In detennlnlng the discrete closed subgroups r of G

such that the nilmanlfolds G/r are compact. A problem In the classification of these
manifolds naturally occurs. The answer Is well known in a few cases only (Euclidean
case, Heisenberg case).
D.2. Uniform subgroups
Definition 2. A subgroup r of G Is called a unJfOIIlJ subgroup If G/H Is compact.
Usually, the research of unlfonn subgroups passes through the research of lattice, i.e
discrete, subgroups such that G/r carries a finite Invariant measure. If every discrete
unlfonn subgroup is a lattice, the converse is not usually true (see, for example,
"Discrete subgroups of Ue groups" by Raghunatan [RAGD.
The situation is simpler In the nilpotent case.
Theorem 1. Let G be a simply connected nilpotent Ue groups and r a closed

subgroup of G . Then G/r Is compact If and only If r Is a lattice.
We can also characterize the unifonnity of a subgroup in tenns of representation. A

representation p : G -+ GUn,1I\) is nilpotent if p(x) is unipotent for an x E G. As Gis
301
connected and simply connected, such a representation always exists. It can be also
supposed free. If risaclosed subgroup of G, then G/r is compact if and onlyiffor
any representation p : G ~ GLCn, lID, pCG) andpCD have the same closure for the
Zarlski topology, in the algebraic group GLCn , lID.
So the existence of uniform groups of G is read through the matricial representation of

the nilpotent group G.
11.3. Existence of discrete uniform subgroups
Theorem 2. Let G a simply connected nilpotent Ue group and letIJ be its Ue algebra.
Then G admits a discrete unifonn subgroup if and only if IJ is a rational Ue algebra, i.e.
9 admits a basis with respect to which the constants of structure are rational.
The proof of this theorem can be read in [RAG J.
This theorem is very useful for constructingnilmanifolds. The rationality of anilpotent

Ue algebra is not very difficult to see. Nevertheless, the determination of the discrete
uniform subgroup of a rational nilpotent Ue group is a hard problem, just as the
problem of the classification of nilmanifolds.
11.4. Rational nilpotent Lie algebras
From the classifications of nilpotent Ue algebras of dimension less than seven, we can
deduce the following result
Proposition 1. Every complex (and reaD nilpotent Ue algebra of dimension less than 6
is rational. Every complex nilpotent nonrational Ue algebra of dimension 7 belongs to
a parametrized family of nonisomorphic Ue algebras.
302 Chapter 8
III. CONTACT AND SYMPLECTIC GEOMETRY ON NILPOTENT

LIE ALGEBRAS
III.t. The Cartan class of an element of9-
We have identified the elements of 9- with the left invariant forms on a Ue group G
whose Ue algebra is 9. Then we can see how, within this framework, the classical
differential invariants be have themselves. The most important of them is probably
the Cartan class of a pfaffian form.
Definition 3. Let ro be in 9°. The characteristic subalgebra of ro is the subalgebra of9

defined by
h.ro - {X e 9 , i(X) do> - o} ,
where i(X) dro is the element of 9° given by
i(X) dro (Y) - dro (X Y) - - ro [X, Y].

I
The class (or CartaIl class) of ro is the codimension in 9 of the vector space
Ker ronh.ro
Proposition 2. If 9 is a nilpotent Ue algebra, then every fonn ro, ro e 9° , ro ::/:. 0 has

an odd class.
Proof. Consider the 2-exterior form dro on 9. It is a bilinear alternated form on 9

defined by
dro eX I y) - - ro rx I yl.
303
Lemma. If the class of c.o Is odd, there Is a basis (c.o - c.o1, ID.2, ••• , COn) of '.t such that
dc.o - ~ 1\ co, + ... + ~ 1\ ~p+1'
If the class ofc.o Is even and equal to 2p, there Is a basis Cc.o- c.ol' ..., COn) of 9- such that
This lemma is a direct consequence of the classification of the alternated bilinear form
on a real or complex vector space. From this lemma, we can easilysee that the dual
space (Ker c.o f"I h oo )- is generated by (ID.2P+2' ..., COn) if the class is 2p+ 1, or by
(ID.2P+1' ..., COn) if the class is 2p.
Proof of the proposition. Let (c.o - c.ol , ... , COn) be a basis of 9 reducing to the canonical
form dc.o(Lemma 3). Suppose, first, that c.o E Z(9)-. If(X 1, X2 ' .••, Xn) is the basis of9
such that (c.o1, ... , COn) is the dual basis, then Xl E Z(9) and [X, Yl - 0, V Y . Thus,
c1Co)) - 2p implies dc.a::x, X2) - 1, i.e. c.o [X , X21-:t: 0 . This is impossible andclCc.o) =
2p+ 1. Suppose now that c.a::X) - 0 , V X E Z(I}), the matrix of the operator ad X I is
nilpotent and nontrivial (we suppose c.o -:t: 0). Jordan's reduction of this operator shows
that the equation [Xl' Yl - Xl + U ,where U is independent of Xl' has no solution.
Then dc.a::XI ' Y) - 0 for all Yand c.o has an odd class.
Remark. Suppose that 9 is rational. If G is the simply connected nilpotent Ue group

associated. to 9, then there is a closed discrete uniform subgroup r such that G/r is a
compact manifold. As c.o corresponds to a left invariant form on G, it passes through
the quotient G/r and defines a pfaffian form c.o' on G/r. It is clear that c.o' has an even
class. Let Hco be the Ue group corresponding to the Ue algebra h oo associated. to c.o. It
is closed in G and x(Hoo) f"I G/r is a compact manifold whose dimension is equal to
the class of c.o as soon as c.o is supposed to have an even class equal to 2p (x is the
projection on G/D. The Stokes theorem implies
304 ChapterS
J.n(H (0) 1"'1 G/r

dro- r
Ja (n(H (0) 1"'1 G/r)
00-0.
This denies the fact that (dro)P Is a volume fOnD on 1t(H00) n G/r but explains the
previous proposition.
111.2. Contact Lie algebras. A characterization of the Heisenberg algebra
Definition 4. A (2p+ I)-dimensional Ue algebra 9 Is called a contact Ue aJgebralf there

Is a contact fonn 00 on 9. I.e. Is a fonn 00 E 9· having a class equal to 2p+ 1.
Proposition 3. If 9 Is a contact lie algebra, then dim Z(9) S; 1.
Proof. Suppose that dim Z(9) ~ 2. Then there are two independent vectors Xl andX 2
such that [~Yl- o. ';f Y E 9 • j - 1 and2 . If 00 E 9·. then I(xpdoo - 0 for j - 1 and2
andd(ro) S; 2p.
Consequence. If 9 Is a nllpotent contact Ue algebra, then dim Z(9) - 1.
Example. The Heisenberg algebra lip Is a contact Ue algebra. Indeed. contacts of the
structure of lip are given by
1- 1..... P.
and if (001 ..... OO2p+l) is the dual basis. we have
Theorem 3. In the variety of Ue algebras r,2p+I. there exIsts a nelghborhood of the

HeIsenberg algebra lip meeting all the orbIts of contact Ue algebras of r,2p+1 and
meeting only these orbIts. Thls property charactertzes the HeIsenberg algebra.
305
Proof. Let 9 be a contact tie algebra. Let (Xl' ... ,X 2p+ l) be a basts of 9 such that the
dualbasls (001' ••• , CO:2p+l) satisfies
dro l - - 002 A 003 + ... + 002p A ID2p+1
The constants of structure corresponding are given by
[X2 , X3] - Xl + L C123~

I~
Consider the contraction given by the following sequence of Isomorphisms
The bracket of the colresponding tie algebra. 90 defined. by fo satisfies:
The limit-point 9 00 ' which is a contraction of 9 , is the Heisenberg algebra.. Then every
contact tie algebra. can be contracted on the Heisenberg algebra. This proves the first
part of the theorem. As the existence of a contact form is an "open" property, every
perturbation of the Heisenberg algebra. also is a contact tie algebra.. Then we can
306 ChapterS
characterize a neighborhood of lip In L2p+l. For proving that this property

characterizes the Heisenberg algebra, it Is sufficient to see that another "model" can be
contracted on lip and which is also a contraction of tip. If this model is a contraction
of lip , it is nilpotent and its characteristic sequence is less than the characteristic
sequence of lip and then It is equal to (2, 1, ... , 1) or (1, 1, ... , 1). But every nilpotent Ue
algebra having (2, 1, ... , 1) as a characteristic sequence is isomorphic to lip or fit E9
~p-O where "'j is the ~mensional Abelian algebra. As this model is a contact algebra,
we have I - P and we aga.in find the Heisenberg algebra. The case (1, 1, ... , 1) is
impossible because it corresponds to the Abelian algebra.
mol. Lie algebras with a symplectic structure
Frst, it is Important not to blend a Ue algebra with a symplectic structure and the
symplectic algebra spCn).
m.3.1 Definitiui'i
Definition 5 . An alternated bJIJnear form 9 on 9 Js called symplectic Jf Jt satisfies
(1) d9 - 0,
(2) 9P ¢O, where dim 9 - 2p .
Here d9 corresponds to the differential of the left Invariant 2-fonn 9. It Is given by
d9(X , Y , Z) - 9[X , Y] , Z) + 9[Y , Z] , X) + 9[Z , X] , Y) .
Example. If there is alinearfonn 0>0n9 of maximal class equal to 2p, then 9 - drolsa
symplectic fonn.
A Ue algebra provided with such a symplectic fonn Is called Frobeniusian.
Proposition 4. There Is no FrobenJusJan nJlpotent Ue algebra.

The NiImanifokJs 307
Indeed, the class of a linear fonn on a nilpotent tie algebra is always odd.
m.3.2. Classification of complex nilpotent Lie algebras with a symplectic structure
- dimension 2 The algebra is Abelian
e - COlA CO2 is symplectic
- dJmension 4 (we describe the constants of structure from the Maurer Cartan
equations)
dOlJ. - 0 i - 2, 3, 4
dco, - co 1 A CO2
dC04 - CO 1 A C03
(the trivial equations are not written).
- dJmension 6
dro, - COlA CO2
dC04 - CO 1 A CO,
dC05 - COl A C04 + CO2 A CO,

dC05 - CO 1 A C04 + CO2 1\ CO,
dco, - COl A CO2
dC04 - CO 1 A CO,
dro 5 - COl A C04
dro6 - COlA C05 + CO2 A CO,

308 ChapterS
"'6,5
dro, - COl A CO2
dro4 - COl A CO,
dros - CO 1 A C04
dro6 - COl A COs
"'6,6
d~ - C01ACO:z
dOlt - C01A~
dCOs- C02A~
dCll6 - C01AOlt + C02ACOs
"'6,7
dco, - CO 1 A CO2
dro4 - COl A CO,
dros - CO2 A CO,
~ - COl A C04 - CO2 A COs
"'6,8
dro, - COl A CO2
dro4 - CO 1 A CO,
dros - COl A C04
dro6 - CO2 A CO,
"'6,9
dro4 - CO 1 A CO2
droS - COl A CO, + CO2 A C04
dC06 - CO2 A COs + CO, A C04

309
d<.O", - <.0 1 " <.02
d<.05 - <.0 1 " <.0",
d<.06 - <.01" <.05 + <.02" <.03 + <.02" <.0",
d<.O", - <.0 1 " <.02
d<.05 - <.0 1 " <.0",
d<.06 - <.01" <.05 + <.02" <.03
d<.O", - <.0 1 " <.02
dc.o5 - <.0 1 " <.03 + <.0 1 " <.0",
d<.06 - <.02 " <.0",
d<.O", - <.0 1 " <.02
d<.05 - <.0 1 " <.03 + <.02 " <.0",
d<.06 - <.0 1 " <.0",
d<.O", - <.0 1 " <.02
d<.05 - <.0 1 " <.03 - <.02 " <.0",
d<.06 - <.01"<.04 + <.Oz"~
dc.o", - <.0 1 " <.02
d<.05 - <.0 1 " <.03
d<.06 - <.02 " <.0",

310 ChapterS
dol" - COl A CO2
dol s- COl A CO,
dol6 - COl A CO" + CO2 A CO,
dol" - COl A CO2
dcoS - COl A CO,
dol6 - COl A CO"
dol" - COl A CO2
dol s- COl A CO,
dol6 - CO2 A CO,
ns.l ell
dro, - co I A 0)2
dol" - 0) I A 0),
dol s- 0) I A 0)" + 0)2 A 0),

The NImanIok:Is 311
"5,2 ffi m
d0>3 - 0>1/\ 0>2
dro4 - 0>1/\ 0>3
dro5 - 0>1 /\ 0>4
III.3.3. Construction of a Lie algebra with a symplectic structure
The nol1on of double extension has been defined by Medina and Revoy [M-Rl for the
312 Chapter 8
construction of tie groups provided with invariant metrlcs and also with a symplectic
structure.
Let (91 , 0)1) be a Ue algebra with a symplectic structure. Consider f 1 as a derivation of

91. It induces a bilinear mapping «1>1 given by
X, Ye 91 .
This mapping «1>1 corresponds to a 2-cocycle for the scalar cohomology of the left
symmetric algebra 9 51' whose product satisfies
(x . y) . Z - x.(y . Z) - (y . x) . Z - Y.(X . Z)
and the Ue algebra structure 91 is related with the algebra structure on 9 51 by
[x , y] - X . Y - Y . X
(91 and915 have the same underlying vector space). Consider the adjoint operator f"l
Offl with respect to 0)1 :
We can find a 2-cocycle «I> of the tie algebra 91 for the cohomology W(91' 91) by
putting
X, Ye 91.
Let be 1.1 - !lW E9 91 the central extension of 91 by !lW (~ is the scalar field of 91)
defined by «I> and let 9 be the semidirect product
9- ~V €a 1.1
given by the bracket:

The NiImanifokJs 313
[V, ul- 0
[V , xl- - oo'(W , X)u - f 1(X) , X E 91'
where W is defined in the left symmetric algebra 91 s by
W-V2.
Such a tie algebra 9 is called a double extension of 91 following (W , f 1).
Theorem 4. Let (9 , (0) be a Lie algebra with a symplectic structure of dimension 2n

such that the center Z(9) Is not trivial. Then 9 Is a double extension of a Lie algebra 91
with a symplectic structure ofdimension 2n-2 follOwing a pair (00 , f 1) where 00 E 91
and f1 E Der 91"
Every nilpotent symplectic lie algebra can be obtained this way from the trivial tie
algebra of dimension O.
N. LEFT INVARIANT METRICS ON NILPOTENT LIE GROUPS
N.l. Left invariant metrics
Let G be a tie group and 9 its lie algebra.
Definition 6. A left invatiant Riemannian metric g on G is a Riemannian metric on G

such that the mapping
is an isometry.
Then the metric g verifies
The metric g defines a scalar product, also noted g, on the lie algebra 9. We can find a
314 ChapterS
basJs (001 ,... , 0\.) of It such that :
Infinitesimal version
A vector field X Is a Killing field of g if
(Lxg)(Y ,Z) - X(g(Y ,Z))- g([X , Y] ,Z)- g(Y ,[X, Z]) - 0 .
This can be wrltten as :
(Lxg)(Y ,Z) - g(Vx Y ,Z)+ g(y , Vz X) - 0,
where V is the covariant derivative associated. to the I.evt-CIvt1a connection.

The set of Killing vector fields Is provided with a Ue algebra structure, called the
algebra of Inflnlteslmallsometries. It Is associated to the Ue group ISO(G, g) of
lsometries of g which has a finite dimension.
Iv.2. Classification of left invariant Metrics on a 3-dimensional

Heisenberg group
Let ttl be a 3-d.imenslonal Heisenberg group. It Is the group of nilpotent matrices
( oIx1 yz 1
001
Proposition s. Every left lnvanant metric on ttl can be wntten :

315
For the proof, see the foUowing section.
Now, we study the IJe algebra of Killing veclor fields for the metric gA:
Let
be a Killing field for the metric gAo

The components ~ satisfy
a~l _0,
ax
~
A;2X."l + ( 1 + A;2X2 )~2
- - Ax -~3 - 0 ,
ay ay
This system Implies

316 ChapterS
Then, by integrating:
~ (
." - - x + 3
'l2~2)~1
az - 1\."2
"I
ay + 9(y ,z)
X2 ~1
.
Wededuce:
As Ibis polynomial Is identically nul, we have :
2
d ~1 _ 0 and dIv _ d9 _ 0 .
ayaz az az
Ukewlse, the substitution of the expressions of ~ and ~3 In the system give
dIv
- - 0; --~1.
de
ay ay
Then
~1- ay+b,
'If- d ,
The NImanifokJs 317
~2--ax+d
~3 - - M. X2 + Aa y2 + Aby + c
2 2
Conclusion. Every Killing vector field has the following fonn
The tie algebra of infinitesimal isometries is generated by the following 4 vector fields :
a
X- -+ Ay-
a Y--
a Z--,
a
ax az dy az
Note that the fields ex, Y ,Z) form basis of IJ - Hl and correspond to the left
translations.
N.3. Some formulas of Riemannian geometry
Let X and Y two vector fields on G. We define V x Y as the unique vector field on G
satisfying
2g(Vx Y , Z) - Xg(Y , Z) + Yg(X , Z) - 2g(X , Y) + g([X , Y] , Z) +

+ g([Z , X] , Y) + g([Z , Y] , X) , 'if Z E xeG) .
If we suppose g to be left invariant, this fonnula can be reduced to
2g(vx Y , Z) - g([X , Y] , Z) + g([Z , X] , Y) + g([Z , Y] ,X) .

318 ChapterS
These fonn of the Levi Civita connecUon V are given by
where(el ,... , en) is an orthononnal frame of G for the metric g. They verify:
If (ro1 ,.... ron) is the dual basis. we define the symbol :rJ kI by: •
rol- 1:.
k
:rJ kI cJt •
and the Cadan equations are :
The Riemann curvature tensor of the connection V associates to each pair of vector
fields X and Y the linear transfonnation
R(X. Y). (Z) - VxVyZ - VyVxZ - V[x. y)Z
- QV x • Vy] - V[x. y) (Z)

from vector fields to vector fields (of course. here all vector fields are supposed
smooth).
The transfonnation RCX • y) is skew-adjoint It defines an (O,4)..tensor given by
Rxvzv - g(R(X • Y) V ,Z) .
This tensor verifies
Rxvzv - - Rvxzv - Rzvxy
and Bianchi's Identities

319
Rxyzv + Ryzxv + RZXYV - 0 (lstidentity),
V xR YZVU + V yRZXVU + V zRxyvu - 0 (2nd identity) .
Remark. If g is a left invariant metric on G, then the first Bianchi identity corresponds
to the Jacobi identities on the tie algebra CJ. Indeed, these relations, written on a
othonormal left invariant basis (~, are relations concerning the constants of the
structure of g related to ~.
Finally, recall the expressions of the curvature forms and of the Ricci tensor. The
curvature forms of G are the 2-forms
They satisfy
The Ricci curvature is the trace of the linear transformation
Z -+ R(X , Y) Z .
The Ricci tensor is given by the components :
Plj- p(el' ej) - L

k-l
R(ek, el, ek, ej) ,
where (el ) is an orthonormal basis.
N.4. Some formulas for left invariant metries
Let g be a left invariant metric on G, CJ the tie algebra of G, {el} an orthonormal basis
of CJ, and Cl~ the corresponding constants of structure. Then
320 ChapterS
Rljpq - r pq r' js - r jq r' ps - r jp r'sp ,

RIJd - rid r' js -rjl r'ks - rjk r's1 ,
where TrC designates the trace of the operator C and where Cl is the operator whose
matrtx in the basis {ell is (C~yk,j.
lV.S. Left invariant metries on nilpotent Lie groups
First, we will look at the behavior of the Ricci curvature.
Recall that a nilpotent Ue algebra Is a unimodular Ue algebra that sa1lsfle Tr(adX) - 0

'V X E 9 . Then, the Ricci tensor is reduced to :
n
pg - _lTrCI CJ -ITr ICI CJ+ 1
2 2
1: (clkr d kr )
4 k,r-l
n
and the scalar curvature 't - 1: PH
1-1
is given by
n n
't- _11: T r Cp C p _1 1: Tr ICpC p .
2 p-l 4 p-l
Now we note by L- the adjoint (with respect to g) of a linear mapping L on 9. Of

course, we have iden1lfled g with. the associated scalar product on 9. The operator L is
antisymmetrtc if L· - - L In this case, we have gCi.(X) , y) - - g{X , L(Y)).
The Ninani/oIds 321
Lemma. If the operator ad. X Is antisymmetric, then

K(X,Y)~0,'VYE9·
and K(X, Y) - 0 JfandonlyJf gCX, [X, 9D - o.

The proof is obvious. We take X - el' Y - e2 where {el ,...e n} is an orthononnal basis.
If ad. e I is antisymmetric, then
Corollary. If X E Z(9) , then K(X, Y) ~ 0 'V Y E 9 .
Lemma. If ad. X Is antisymmetric, then p(X, Y) ~ 0 ,and p(X, Y) - 0 Is equivalent to

g(x ,D(9)) - 0 .
Lemma. If p{X, D(9)) - 0, then p(X, X) ~o . Theequalityissa/isfiedJfandonlyil

ad. X is antisymmetric.
From these corollaries, we deduce the follOWing theorem:
Theorem 5. Suppose that the Ue algebra 9 is nilpotent and not Abelian. For every left
invadant metric, there exist a direction X such that p(X, X) > 0 and a direction Y
such that p(Y, Y) < 0 .
Indeed, let X oF- 0 be in Z(9) (') [9,9] . Then p(X) > O. For finding a vector Y giving a
negative curvature, we can take a vector orthogonal to [9 ,9] but not belonging to Z(9).
Such a vector exists because 9 is nilpotent
The group of isometries
The determination of the Ue group of isometries for a left invariant metric is easy
enough. The computations result in a theorem of Wolf generalized by Wilson [WI]. Let
{ell be an orthonormal basis of 9. For every derivation f of 9, we consider the matrix
322 ChapterS
MCt) with respect to the basis {el}' Then f corresponds to an infinitesimal isometry if
MCf) is an orthogonal matrix.
Wilson's theorem affirms that every isometry for a left invariant metric g on the
nilpotent group G is in AutC9) the group of automorphisms of 9. This implies that the
infinitesimal isometries are derivations f of 9 such that a matrix of f related to an
orthonormal basis is orthogonal.
In the following section, we compute this group for the left invariantmetrics on the
Heisenberg group.
Another interesting example consists in considering a nilpotent tie group such that
every derivation is orthogonal with respect to a given basis of 9. A partial answeris
given by considering the quatemlonian Heisenberg algebra. This Ue algebra is given by
dOll - (Xl A (X2 + ~ A (X4 ,

d~ - (Xl A (X3 - ~ A (X4 ,
d~ - (Xl A (X4 + ~ A (X3 '
where (001, ~, ~ , (Xl ' ~ , ~ , (X4) Is the dual basis of .~t. Consider the corresponding
basis (X 1 , X2 , X3 ' Y 1 , Y 2 , Y3) In 9 and the metric
2 2 __ 2 2 2 2 2
g- 00 1 + 00 2 , 013 + (X I + (X 2 + (X 3 + (X 4 .
Let f be a derivation of 9. The matrix of f related to the basis (X 1 ' X2 ' X3 ' Y 1 ' Y2 ' Y3'
Y4) is
(X -a -b
a 2(X -c
b -c 2(X
0 0 0 (X -d -e -f
0 0 0 d (X -h -i
0 0 0 e h (X -j
0 0 0 f j (X
The NimanifoIds 323
and the corresponding local isometry is
o -a -b
a 0 -c
b -c 0
0 0 0 0 ~ -e -
0 0 0 d 0 -h -i
0 0 0 e h 0 -j
0 0 0 f 0
Then the isometry group is isomorphic to SO(3) x SO(4).
V. CLASSIFICATION OF LEFT INVARIANT METRICS ON THE

HEISENBERG GROUP
The Heisenberg group Kp is the matrlcial Ue group
1 Xl Xp Z
o o o YI
o 0 0 1 YP
o 0 0 o 1
There is a basis (Xl' ... , X p ' Yl' ... , Yp ' Z) of the Heisenberg algebra ~ whose
corresponding left invariant vector fields are
324 ChapterS
d
Z--,
dz
where (xt, YI' z) are the global coordinates of Kp defined above. These fields define the
structure of the tie algebra lfp because they verify
[XI' VI]- -Z , 1- 1, ...,p.
The left invariantforrn (a.1' ... , «po ~1' ..., ~P' m) which define the dual basis of ~ are
writen
Proposition 6. The Ue algebra DerCHp> of derivations of lfp Is Isomorphic to

sp(p, lID E9 m
2 p+1.
We have seen this In the previous chapter but from an other point of view. We again
give the proof using here the matricial representaUon~
Let f be a derivaUon of~. A matrix of f with respect to the basis (XI' VI' Z) has the
following fonn
M-( H~ )
where A, B. C, D are square matrices of the order p, B and C being symmetric, and D
saUsfying the relaUon D - - tA + a.I.
The decomposlUon
The /IIImanIokJs 325
M-(
A
B -tA
C 0
0 oo aI
0 0
0
J
000 p J.1 a
shows that Der(lfp) - sP(P.1i) e 112p+l.
This proves. in particular. that the group of automorphisms of ~ is isomorphic to

spCp. II) X A2p+l where A2p+l is Abelian. Now we can be interested to the classiftcaUon
of the left invariantmet1ics on Kp up to an automorphism. FollowingWilson. this
corresponds to the classification up an to isometry.
Theorem 6: Every left invariant invariant metric on Kp is equivalent up to an

automoIphism to one of the metdcs
with A.l ..... Ap E Ii - (O) . The famnygA" ...,Ap is irreducible.
Let g be a left invariant met1ic on Kpo It is written
where 9 is a linear combination of the <Xt. /3. and ro.

The action of the subgroup of AutCHJ corresponding to the mat1ices of the fonn
reduces the definite positive quadJatic fonn g to

326 ChapterS
and the quadratic form gl is positive in restriction to acomplementary of the center.

As DenHJ is isomorphic to sp(p, lID EB ]i2p+l , the problem Is reduced to the
reduction of the quadratic form gl modulo the symplectic group. Let A be the matrix
associated to gl with respect to the basis Co:l , ~f' Its eigenvalues are strictly positive. So
If E - ( ~ -01) , the eigenvalues of the matrix AE are complex without a real part.
From the classification of Williamson (see, for example, the book of Arnold on the
classical mechanic), we can reduce g to the form
This gives the theorem.
Application: A Riemannian characterization of the Heisenberg group
We begin by recalling the notion of a homogeneous naturally reductive space.

Consider a Riemannian space (M, g) and Ig its tie algebra of infinitesimallsometries.
Then M is a homogeneous naturally reductive space, M - G/K where G is a group of
Isometrles and K Is the isotropy subgroup at a point p of M if the tie algebra 9 of G
(which is a subalgebra of I~ satisfies
< [X , Y]m , Z > + < [X , Z]m , Y> - 0, '<:f X, Y, Z Em,
(here < ,> denotes the product induced on m from 9 by Identification of m with T.,M
and ~ the tie algebra of K).
Suppose now that M - Kp and g - gA.l....,Ap .
ProposiUon 7.For every left invariant metric gA.l,''''Ap , the Riemannian space
(Up, gA.lo....Ap) is homogeneous naturally reductive.
327
We consider the fields
;1- XI + YIZ , TIl - YI-XIZ , Z , i-I,..., p,

x2+y2
R - XIYI- YIYI- I 2 I Z , i-I,..., P .
If 1. is the suba1gebra of Der(IV generated by these vectors, the bracket of 1. verifies
other products are zero. Then 1. - ~ E9 m, where ~ Is the Abelian algebra generated
by the RI gives the naturally reductive decomposition.
Theorem 7. Let G be a nilpotent Ue group such that for every left lnvarlantmetrlc g,
the Riemannian space(G, g) is ahomogeneous naturally reductive space. Then the Ue
algebra I} ofG is a dJrect sum of Heisenberg and Abellan algebras.
The proof is based on the notion of Riemannian contractions. We study some models
of nilpotent Riemannian Ue groups and, by deformation, we can look all the other
nilpotent groups (see [GP2].
This theorem shows the importance of the riemannian Heisenberg group, because the
behavior of the isometry group Is analogous to the group of Isometries of the
Riemannian Abelian group which corresponds to the Euclidean geometry.
BIBLIOGRAPHY
fA 1] Ancochea-Bermudez).M. sabre la rlgldez de Algebras de Ue.

These, Madrid, 1984.
fAG 11 Ancochea-BermudezJ.M., Gaze M. Algebres de Ue rlgldes.

Proc. Konlnklljke Nederl. Acado van Wetenscha.ppen, 5&. A (1985), v. 88, N4, p. 397-415.
fAG 2] Ancochea-BermudezJ.M., Gaze M. dasslflcatlon des aJgebres de Ue flllformes de

dimension 8. '
Arch. Math., 1988, v. SO, 511-525.
fAG 3] Ancochea-Bermudez).M., Gaze M. classification des aJgebres de Ue nllpotentes
complexes de dimension 7.
Arch. Math., 1989, v. 52, n" 2, 157-185.
fAG 4] Ancochea-BermudeZ).M., Goze M. AJgebres de Ue rlgldes dont Ie nllradlcal est
fIllforme.
C.RASc. Paris, t 312, 5&. 1,21-24, 1991.
[AGH 11 Hakimjanov Yu, Ancochea-BermudeZ).M., Gaze M. Sur la reductivile de la variele

des aJgebres de Ue nllpotentes.
C.R.A.Sc. Paris, t 313 CD, 1991,59-62.
fAGH 2] Ancochea-Bermudez).M., Gaze M, Hakimjanov Yu. Sur les composantes de]a varlele
des aJgebres de Ue nllpotentes.
C.R.A.Sc. Paris, t 314 CD, 1992,657-660.
[BA] Bahturin Yu. Lectures on ue algebras.

Academic-Verlag, 1978.
[BO] Borel A. Unear algebraic Groups.

Benjamin, 1969.
[BOU 11 Bourbaki N. Groupes et aJgebres de Ue.

Ch. I, Hermann, 1960.
[BOU 2] Bourbaki N. Groupes et aJgebres de Ue.

Ch. II - III, Hermann, 1972.
[BOU3] Bourbaki N. Groupes et algebres de Ue.

Ch.IV-VI, Hermann, 1968.
[BOU4] Bourbaki N. Groupes et aJgebres de Ue.

ch. VII - VIII, Hermann, 1975.
[BR 1] BraI2lavsky F. Une algebre de Ue caraclerlstiquement nllpotente de dimension 6.

C.RASc. Paris, 5er. A, 1973, t 276,1035-1037.
[BR2] BraI2lavsky F. Sur les aJgebres admetlant un lOre d'aulOmorphlsme donne.

). of Algebra, 1974, v. 30, 305-316.
330 Bibliography
rCA 11 Carles R. Vadi!tI!s des aJgebres de Ue de dimension tnfeneure ou i!gaIe a 7.

C.R.ASc. PaIts, ser. A, 1979, t 289, 263-266.
rCA 2) carles R. Sur la structure des aJgebres de Ue rlgtdes.
Ann. lost Fourier, Grenoble, 1984, v. 33, p. 65-82.
[CA3) Carles R. Sur certa1nes classes d'aIgebres de Ue rlgtdes.

Math. Ann., 1985, v. 2n, 477-488.
rCA 4) Carles R. Sur les aJgebres de Ue canu:terlstlquement nllpoll:nll:s.

Preprint N5. Unlv. de PolUers, 1984.
[CO) Carles R., D1akire Y. Sur les wrli!tI!s d'aIgebres de Ue de dimension S 7.

J. of Algebra, 1984, v. 91, 53-63.
ICE) Cartan H., Ellenberg S. Homological algebla. 1956.
[CH 11 Chevalley C. TMorle des groupes de Ue.

Princeton, tome I, 1946.
(CH 2) Chevalley C. TMorle des groupes de Ue.

Tome II, Hermann, 1955.
(CH 3) Chevalley C. TMorle des groupes de Ue.

Tome III, Hermann, 1955.
[0111 Olxmler J. Sur les representallons unltalres des groupes de Ue nlJpoll:nll:s Ill.
Canad.J. Math., 1958, v. 10,321-348.
[012) Olxmler J. CohomolosJe des aIgebres de Ue nlJpoll:ntes.
Acta ScI. Mat szeged, 1955, v. 16, 246-250.
[OL) Olxmler J., Uster W.G. Derlvallons of nlJpoll:nt Ue algebras.

Proc. AM.s., 1957, v. 8,155-158.
roy) Dyer J.L. A nlJpoll:nt Ue algebra with nllpoll:nt automolphlsm group.

BuD. A.M.s., 1970, v. 76, Nl, 52-56.
[DO) Do2'Jas J. Sur les di!rlvallons des aJgebres de Ue.

C.R.ASc., PaIts, 8&. A, 1964, t 259, 2748-2750.
[fA 1) favre G. Une aJgebre de Ue canu:terlsllquement nlJpoll:nll: de dimension 1-
C.R.ASc., PaIts, ser. A, 19n, t 274,1338-1339.
[fA 2) favre G. Systeme de polds sur une aIgebre de Ue nllpoll:nll:.
Manuscript math. 1973, V. 9, 53-90.
[fS) Favre G., Santharoubane LJ. NlJpoll:nt Ue algebras of maximal rank and of classical
simple type.
Preprlnt.
Iff) Feigln B.L, FIa10wski A. CohomolosJe of the nllpoll:nt subaIgebras of current Ue

algebras.
sf. ScI. Math., Hungarlca, 1989; t 24, Nl, 1-9.
[fJ) FtaIowski A. Deformallons of a Ue algebra of ~tor fields on the lJne.
SOviet Math. Survey, 1983, v. 38, n- 1,201-202.
Bibliography 331
[FU] Fuks D. Cohomologte of infinite dimentional Ue algebras.

Nauka, 1984 Crusstan). English transl. Consultants bureau, 1986.
[GE] Gerstenhaher M. On the ddormatlons of rings and algebras.

Ann. of Math, 1964, v. 79, 59-103.
[GO] Goto M., Crosshans F.D. Semlslmple Ue algebras.
Marcel Dekker, 1978.
[GA] Gaze M. Ancochea BermudezJ,M. Classification of dgld Ue algebras whose nllradIcaJ Is

flllform.
Preprlnt Unlv Mulhouse 1995
[GO 11 Goze M. Bouyakoub K. Sur les aJgebres de Ue munies d'une forme symplectlque.
Rend. Unlv. Cagliart, V. LVII (1), 1987.
[GO 2] Goze M. Perturbatlons of Ue algebras In ddormatlon theory.

ASI NATO, Serle, C 247, 265-355.
[GO 3] Goze M. trude locale de ]a vartere des lois d'aJgebre de Ue.

These, 1982, Mulhouse
[GO 4] Goze M. Cdteres cohomologtques pour ]a dgldire de lois aJgebdques.
Bull. Soc. Math de Belgique, t XUII, 1991,33-42.
[GP] Plu P~ Goze M. Classlflcatlon des metrlques Invartantes agauche sur Ie [pOupe de
HeIsenbers·
Rend. Clrcolo Math. di Palermo, sec. II, 1990.
[GH 11 Go2E M., Hakimjanov Yu. Solvable and nilpotent Ue algebras.

Prepublication IRMA, 1994.
[GH 2] G02E M., Hakimjanov Yu. Sur les aJgebres de Ue nIlpotentes admettant un tore de
dedvatlons.
Manuscripta math. 1994, V. 84,115-124.
[GUI] Gulcharde A. Cohomologle des [pOupes topologtques et des alsebres de Ue.
CEDIC/FERNAND NATHAN, 1980.
[GUR 11 Gurevich G.B. The derivation algebra of an arbillaly standard nil-algebra.

Izv. wzov, matem., 1957, n° I, 103-120 (russian).
[GUR 2) Gurevich G.B. Slandard Ue algebra.

Matem. sb., 1954, v. 35 (m, 437-460.
[HA] f!.anlson D. Commutative algebras and cohomology.

Trans. A,M.s., 1962, v. 104, 191-204.
[HE) Helgason S. Differential geometry, Ue [pOups, and symmetric spaces.
Academic Press, 1978.
[HS) Hochschlld G., Sene J.P. Cohomology of Ue algebras.

Ann. of Math, 1953, v. 57, 591-603.
[HU) Humphreys J.E. introduction to Ue algebras and representation theory.
Sprlnger, 19n.
ij 11 Jacobson N. A note autommptJsms and derivations of Ue algebras.

Proc. A.M.s~ 1955, v. 6, 281-283.
332 Bibliography
U21 Jacobson N. Ue algebras.

Jnterscience Publ., 1962.
lKAI Kaplansky. Ue algebras and loca11y compact 8fOUPS.

U. Chicago Press, 1971.
IKOI Kostant B. Ue algebra cohomology and the generalJ2Z!d Borel-Wen theorem.

Ann. of math., 1961, v. 74, n° 2, 329-387.
IKH 1.J Khakimdjanov Yu. Standard subaJgebras of reductlv Ue algebras.

Moscov Unlv. Math. Bull., 1976, v. 29, N° 1-2.
IKH 2J Khakimdjanov Yu. Derivations of some nnpotent Ue algebras.

Isv. Vysch. Uchebn. Zaved., Math., 1976, N° 1, 100-110 (russian). EnglIsh transl.ln
Soviet Math. (20) 1976.
IKH 31 HaklmJanOV You (Khakimdjanov Yu). Cohomologle et deformations de ceztalnes

a/gebres de Ue nnpotentes.
Commun. Algebra, 1989, v. 16, N° 10, 2149-2192.
IKH 41 Khakimdjanov Yu. On the cohomology of some nIlpotent Ue algebras.

Mat Sbomlk, 1989, v. 180, N° 5, 658-675 (russian). English transl. In Math. of the
USSR Sbomlk 67 (1990).
IKH 51 Khakimdjanov Yu. On the characteristically nnpotent Ue algebras.

Algebra I10gika, 1989, v. 28, N° 6,722-736 (russian). EnglIsh transl.ln Algebra
and IDglc, 1989, v. 28, 475-485.
IKH 61 HaklmjanOV You (Khaklmdjanov Yu). Sur les varieles d'a/gebre de Ue.
Commun. Algebra, 1990, v. 18, N° 4, 1147-1187.
IKH 71 Khakimdjanov Yu. Characteristically nIlpotent Ue algebras.

Math. Sbomlk, 1990, v. 181, N° 5, 642-655 (russian). English transl. In Math.
USSR Sbomlk, 1991, v. 70, N° 1.
[KH 81 Khakimdjanov Yu. Derivations of some lnflnlte dimensional topologically nnpollmt

Ue algebras.
Math. Zametki, 1990, v. 48, N° 1, 128-133 (russian). EnglIsh transl.ln Math. Notes
at the Acad. of Sc. of USSR, 1991.
[KH 91 Haklmjanov You (Khaklmdjanov Yu). Varletes des lois d'algehres de Ue nilpotentes.
Geometriae dedicata, 1991, v. 40, N° 3, 269-295.
[KHOI OnIshchick A.L, Khakimdjanov Yu. Sem/dlrect sums of Ue algebras.

Mat. Zametki, 1975, v. 18, N° 1,31-40 (russian). English transl.ln Math. Notes at
Academy of Sc. of the USSR, 1975, V. 18.
[LNI Levy-Nahas M. Deformations and contractions of a Ue algebra.

J. Math. Phys., 1967, V. 8, 1211-1222.
[LEI Leger G. A note on the derivations of Ue algebra.

Proc. A.M.5., 1953, v. 4, 511-514.
[LL 11 LegerG~ Luks E. Cohomology theorems for Borel-Uke solvable Ue algebras In

arbItrary characteristic.
Canad.J. Math., 19n, v. 24, N° 6,1019-1026.
ILL 21 Leger G., Luks E. Cohomology and weight systems for nHpotent Ue algebras.
Bull. A.M.5., 1974, v. 50, N° 1, 77-80.
Bibliography 333
[LL 31 Leger G., Luks E. Cohomology of nilradicals of Borel subalgebras.

Trans. AM.s., 1974, v. 195,305-316.
[LT1 Leger G., T680 s. CharacteI1stlcaJJy nIlpotent lie algebras.

Duke Math. J., 1959, v. 26, N° 4, 623-628.
[LU 11 Luks E. lie algebra with only Inner deI1vations need be complete.
J. of Algebra, 1970, v. 15, N" 2, 280-282.
[LU 21 Luks E. What Is the typIcal nilpotent lie algebra ?
In ·Computers In nonassoclattve rings and algebras", Academic Press, 1977.
[Y 11 Yamaguchi S. DeI1vations and affine structures at some nilpotent lie algebras.

Mem. Fac. Sc. KyushuUnlv., 1980, v. 34, 151-170.
[Y 2] Yamaguchi S. On some classes of nIlpotent lie algebras and their automotphIsm

groups.
Mem. Fac. Sc. Kyushy Unlv., 1981, v. 35, 341-351.
[MA] Malcev AI. Solvable lie algebras.

Isv. Acad. Naak SSSR. Math. 1945, v. 9, 329-352 (russian). English trans!. In Transl.
of AMS N 27, 1950.
[MOR] Morosov v. ClassIfIcation of nIlpotent lie algebras of sIxth order.

Isv. Vysoh U. Zaved. Mat. N 4 (5),1958,161-171.
[MOS] Mostov G.D. Fully reductlble subgroups of algebraiC groups.

Amer. J. Math. 78, 1956.
IM-R] Medina A- Revoy P.Algebres de lie et prodult scalaire Invadant

Ann. ScI. Ecole Normale Sup. (4), 18, 1985,553-561.
[NR 1] Nljenhuls A., Richardson R.W. Deformations of lie algebra structures.

J. Math. Mech., 1967, V.
17,89-105.
[NR 21 Nljenhuls A., Richardson R.W. Cohomology and deformations of algebraic structures.
Bull. AM.S., 1964, V. 70,406-411.
[NR 31 Nljenhuls A., Richardson R.W. Cohomology and deformations In graded lie algebras.
Bull. AM.s., 1966, V. 72, Nl, 1-29.
[NR 41 Nljenhuls A, Richardson R.W. Commutative algebra cohomology and deformations

of lie and assocIative algebras.
J. of Algebra, 1968, V. 9, WI, 42-53.
[RAG] Raghunathan M.s. DIscrete Subgroups of lie groups.
Ergeb. der Math. n° 68, Springer Verlag.
[RAU] Rauch G. Vadatlons d'algebres de lie.

These doct Sc. Math., Parts, Unlv. Parts-Sud, 1973.
[RI 1] Richardson R.W. A I1gIdIty theorem for subalgebras of lie and assocIative algebra.
Illinois J. Math., 1967, V. 11, 92-110.
[R121 Richardson R.W. On the rigidity of semi-dIrect products of lie algebras.

pacific J. Math., 1967, V. 22, N° 2, 339-344.
334 Bibliography
[SAl Sato T. The derlvallons of the Ue aJsebra.

Tohokn Math.J., 1971, v. 23, 21-36.
[SALI Sale1an E. Contracllon of Ue groups.

J. Math. Phys. 1966, V. 2, 1-22.
[SAM I Samelson H. Nores on Ue Algebras.

Van Nosuand Reinhold Co.,New Yolk 1969.
[SAN 1 santharoubane LJ .. InfInlre famllles of nOpotent Ue algebras

J. Math. Soc. Japan ,35,3, 1983,515-519.
[SEMI 5emlna\re ·Sophus Ue". Theone des ~res de Ue. Topologle des groupes de Ue.
Parts : ENS. 1954-1955.
[SCI Schenkman E. On the denvallons algebra and the holomOlph of a nnpotent Ue

algebras.
Mem. A.MS., 1955, N° 14, 15-22.
[SEEI Seeley C. Seven-dlmenslonal nilpotent Ue aJsebras over the complex numbers.

thesis, Chicago, 1988.
[SE 11 Sene J.P. Ue algebras and Ue groups.
Benjamin, New York, 1965.
IT 11 T6g6 S. Dlmensons of the derlvallons algebras of Ue algebras

J. Sc. Hiroshima Unlv.• ser. A-I. 1967. V. 31. 17-23.
IT 21 T6g6 S. A theorem on characte1isllcally nnpotent Ue algebra.

J. Sc. Hiroshima Unlv., Ser. A. 1969. V. 33. 209-212.
IT 31 T6g6 S. On the derlvatlon algebras of Ue algebras.

Canad.J. Math., 1961, V. 13.201-216.
IT 41 T6g6 S. Deltvallons of Ue algebras.

J. Sc. Hiroshima Unlv.• ser. A-I, 1964. V. 28,133-158.
[V 11 Vergne M. Varlete des ~res de Ue nllporentes.
These de 3eme cycle, Parts, 1966.
[V 21 Vergne M~ Cohomologle des aJsebres de Ue nnpotentes. Appllcallon a I'etude de

la valtete des aJsebres de Ue nllporenres.
Bull. Soc. Math. france, 1970, V. 98, 81-116.
[WI Williams f. Laplace operators and the h-module structure of certain cohomology
sroups.
Trans. AMS., 1975, V. 197, N° I, 1-57.
[WI) Wilson E. Isometry groups of homogeneous nilmanlfolds.

Geom. dedicata., 1982, V. 12.
[WOl Wolf J.A On locally symetrJc spaces of non negatlve crnvature.

Comm. Math Helv, 37, 1963, 265-295.
INDEX
Abelian Lie algebras 2 filiform component 252

adjoint representation 78 filiform Lie algebra 36,40
algebraic Lie algebra 35 filtered Lie algebra 91
algebraic variety 170 filtered module 92
ascending central sequence 11 Frobeniusian Lie algebra 306
fundamental cohomological space 127
Borel subalgebra 51
bracket 1 general linear algebra 2
graded filiform Lie algebras 42
Cartan subalgebra 28 graded Lie algebras 91
Cartan Class 302 graded module 92
Cartan Criterion 20
Cartan Killing Criterion 22 Heisenberg algebra 3,14,47
Cartan matrix 31 Hochschild-Serre spectral 99
center (of Lie algebra) 5 sequence
characteristically nilpotent 46 homomorphism of Lie 5
Lie algebra algebras
characteristic ideals 9,36
characteristic sequence 37 ideal 4
classical Lie algebras 3 idealisation 187,197
coboundaries 81 inner derivation 8
coboundary operator 79 Internal Set Theory 196
cochain 78 invariants (of nilpotent Lie algebras) 35
cocycle 80 irreducible component 181
cohomology (of Lie algebras) 77 irreducible representation 24
component (of variety) 181 isomorphism (of Lie algebras) 5
contact geometry 302
contact Lie algebras 302 Jacobi identity 1
contraction 182
Killing form 22
decomposition (of perturbation) 87, 183
deformation 87,183 left invariant metric 313
derivation 7,111 Levi subalgebra 27
derived sequence 11 Levi's Theorem 25
descending central sequence 11 Lie algebra 1
direct sum of Lie algebras 7 Lie group 296
Dynkin diagram 32 linear deformation 184
Lie's Theorem 17
eigensubspace 29
Engel's Theorem 15 maximal tores 38
equivalent deformations 186 module (for Lie algebra) 24,77
exact sequence 81
exceptional Lie algebras 34
336
negative root 31 tangent space 202

Nelson axioms 187 tensor 170
nilindex 36 trace 4
nilpotent Lie algebras 13 transference 188
nilpotent radical 25 two-step nilpotent Lie algebra 43
nilradical 24,34
nilmanifold 300 uniform subgroup 300
Non standard Analysis 196
variety 170
obstruction 204
Open orbits 179 weight 29
orbits 177 Weyl basis 31
Weyl's formula 158
parabolic subalgebra 50 Weyl's Theorem 24
perturbation 88,187
positive roots 31
reductive Lie algebra 24

regular element 28
representation 23,77
rigid lie algebra 179
rigid tensor 172
root 29
root system 29
scheme 174
semidirect sum 10
semisimple Lie algebra 21,28
sill algebra 282
sill cocyc1e 282
simple lie algebra 21,32
simple module 24
simple root 30
singular root 129
solvable lie algebras 12
special linear algebra 3
special orthogonal Lie algebra 3
spectral sequences 95
standardisation 188
standard Lie algebra 50,52
subalgebra (of lie algebra) 3
symplectic geometry 306
symplectic Lie algebra 4
symplectic structure 306

Michel Goze, Yusupdjan Khakimdjanov - Nilpotent Lie Algebras PDF

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Michel Goze, Yusupdjan Khakimdjanov - Nilpotent Lie Algebras PDF

Hochgeladen von

Copyright:

Verfügbare Formate

Nilpotent Lie Algebras

Mathematics and Its Applications

SPRINGER-SCIENCE+BUSINESS MEDIA, B.V.

ISBN 978-90-481-4671-0 ISBN 978-94-017-2432-6 (eBook)

Printed on acid-free paper

All Rights Reserved

Chapter 1 : Lie algebras. Generalities 1

SI Lie algebras. Generalities 1

V.3. An order relation on the set of the roots fl 30

Chapter 2 : Some classes of nilpotent Lie algebras 40

Filiform Lie algebras 40

Chapter 3 : Cohomology of Lie algebras 77

Chapter 4 : Cohomology of some nilpotent Lie algebras 111

§I Derivations of some filiform algebras 111

111.3. PcuUcu1ar case : n is the nllradical of a Borel subalgebra 146

Chapter 5 : The algebraic variety of the laws of Lie algebras 170

§I The vector space of tensors 1"n2,1 170

IV.2. Theorem of decompositton of a limited. vector of CD 190

Chapter 6 : Variety of nilpotent Lie algebras 250

§I The tangent space of the variety of nilpotent Lie

VI.3. On the irreducible components ofNOp containing 11. 272

§ VII. On the components of the variety N° 275

Chapter 7 : Characteristically nilpotent Lie algebras 281

§I Characteristically nilpotent filiform Lie algebras 281

Chapter 8 : Applications to differential geometry.

§I A short introduction to the theory of Lie groups 296

IV.2. Classification of left Invariant me1l1cs on the

We would like to thank the Mathematischer Forschungsinstitut of oberwolfach

LIE ALGEBRAS. GENERALITIES

I. LIE ALGEBRAS. GENERALITIES

1.1. Notions of Lie algebras

Definition 1. A Lie algebra. q over a field K is a vector space on K with a bilinear

2. The relation CI) implies the anticommutativtty of the multiplication of 9 :

3. Let noCK) be the space of n-square matrices on K. The multiplication

4. Let V be a vector space on K. The vector space End V of the endomorphisms of

5. Let A be an associative algebra on K whose multiplication is 0 • We can provtde A

[e21 , e21+11 - e 1 for i - 1,..., k.

1.3. Lie sub algebras

Definition 2. Let c:I be a Ue algebra. on K. A subspace c:ll of c:I is called a Ue

1. Every subspace of an Abelian lie algebra is an Abelian subalgebra.

1.4. Classical Lie algebras

1. The special linear Ue algebra. slCn,K) , K - ~ or tC .

is a lie subalgebra of glCn,K).

2. The special orthogonal Ue algebra. so(n,K), K - mor «: .

where tA is the transposed matrix of A.

3 The symplectic Ue algebra sp(2n,K), K - mor C .

Every matrix A of sp(2n,K) can be decomposed as

(2)SO(Y)-{fE EndV/<f(v),w>--<v,f(w», VV,WE Vanddetf-l} where<,>

(3) sp(V) - {f E End V / <Q(f(v),w) - - Q(v/(w)) >, V v, WE V}, where n is a

Definition 3. A Ue subalgebra 1- of q is called an ideal of q if [1-,q] c 1- that is

[x,y] E 1- for all x E 1- and yE q.

of course, there are some Ue subalgebras which are not ideals.

2. Let 9 be a Ue algebra. The center

3. Let fa. be a subalgebra of a Ue algebra 9. The normalizer of fa. in 9

is a subalgebra of 9 containing h. Then h is an ideal of Nlh).

1.6. Quotient Lie algebra

Let 9 and 91 be two Ue algebras on K. A linear mapping

As usual, <p is an isomorphism if it is a bijectif homomorphism and <p is an auto-

morphism if cp is an isomorphism and 9 - 91 .

1. Let V be a n-dimensional K vector space. Then if we choose a basis of V, the

2. Let 1.. be an ideal of the lie algebra 9. The canonical map

Is a homomorphism of I} onto 9/1... Its Kernel

Proposition 1. Let cp : 9 ~ 91 be a homomorphism of Ue algebras. Then Ker cp is an