Beruflich Dokumente
Kultur Dokumente
PUCP
Contents
2
Introduction
3
Chapter 1
| {z
alternating) multilinear mappings from V V} to R. Certainly, for
k times
k > n, we have
k V = 0 (1.1.1)
0 V = R, 1 V = V , dim n V = 1. (1.1.2)
4
A nonvanishing element of n V defines an orientation of V . By taking
antisymmetric multiplication, V = nk=0 k V becomes an algebra with its
Z- grading induced by its degree.
We say that a bilinear form : V V R is nondegenerate if , for v V ,
(v, ) = 0 implies that v = 0.
We say that a bilinear form g : V V R is a scalar product (or
Euclidean metric) on V if g is symmetric and positive, i.e., for any u, v V ,
1 = 2 . (1.1.3)
is an isomorphism.
5
Proof. Let h , i be a scalar product on V . Then there exists an antisym-
metric invertible endomorphism A End(V ) such that
( , ) = h , A i . (1.1.5)
Since
det A = det At = (1)n det A
(1.1.6)
LL ((l1 , l1 ) , (l2 , l2 )) = hl1 , l2 i hl2 , l1 i . (1.1.7)
n n
R2n , 0 := Rn Rn , R R
6
the standard symplectic space.
Furthermore, since R2n ' Cn , by replacing z = x + iy in (1.1.7) and (1.1.8),
we obtain
n
iX
st = dzk dz k (1.1.9)
2 k=1
as the standard symplectic form for Cn .
1. W is symplectic if W W = {0};
2. W is isotropic if W W ;
3. W is coisotropic if W W ;
4. W is Lagrangian if W = W .
(V, ) = (W, |W ) W , |W .
(1.1.12)
7
As A is invertible, by (1.1.13), we get the first equation of (1.1.11), in par-
ticular we have dim W = dim W . But by definition of W , we have
W W . This means the second equation of (1.1.11) holds.
If W is symplectic, then W W = W W = {0}, thus W is
symplectic. Now we get (1.1.12) by the first equation of (1.1.11)
(J, J) = (, ) (1.1.15)
(, ) = g (, J) ,
(J, J) = g (J, ) = g , J t .
(1.1.16)
g , J t = g (, J) = , J 2 = (, ) .
(1.1.17)
8
1.1.3 Symplectic vector bundles
ji := j 1 m m
i : Ui Uj K Ui Uj K , ji (x, w) = (x, ji (x, w)) ,
9
Definition 1.10. Let V be a real vector bundle on M , we say that (V, ) is
a symplectic vector bundle on M if C (M, 2 V ) and for any x M ,
(Vx , x ) is a symplectic vector space.
: End(V ) T11 (V ) = { : V V R}
B (B) : V V R
10
For a submanifold W of a symplectic manifold (M, ), we call W a
symplectic (resp. isotropic, coisotropic, Lagrangian) submanifold if for any
x M , Tx W is a symplectic (resp. isotropic, coisotropic, Lagrangian) sub-
space of (Tx M, x ).
A diffeomorphism : M N is called a symplectic diffeomorphism
or symplectomorphism) for two symplectic manifolds (M, ), (N, 1 ) if
1 = .
Let J C (M, End(T M )) be an almost complex structure on a symplec-
tic manifold (M, ), then we say J is a compatible almost complex structure
if (, J) defines a J-invariant Riemannian metric on T M .
Let (M, ) be a symplectic manifold. By Proposition 1.3, M has even
dimension. Let dim M = 2n. Then n 6= 0 2n (T M ) induces a canonical
orientation on M .
Set
L
T = d. (1.1.20)
L
Then T is a closed 2-form on T L. Let : U L V Rn ,
q (x1 = 1 (q), , xn = n (q)) be a local coordinate, then { x j } is a local
frame of T L, and {dxj } is a local frame of T L which gives the trivialization
of T L on U . Thus
!
X
T L V Rn , q, pi (dxi ) (x1 , xn , p1 , , pn ) (1.1.21)
i
11
is the induced local coordinate of T L|U , and { x j , p j } is a local frame of
T (T L).
Xi x i + Pi p i , we have
P
For X =
i
n
* n +
X X
h, Xi(x,p) = pi X i = pi dxi , X . (1.1.22)
i=1 i=1
Notice that the flow t corresponding to the Liouville vector field is such
that t = et , that is along the flow the symplectic form is rescaled expo-
nentially.
12
be the spaces of smooth (r, q)-forms on M .
On local holomorphic coordinates (z1 , , zn ) with zj = xj + iyj , we denote
1 1
= i , = +i , (1.2.3)
zj 2 xj yj z j 2 xj yj
dzj = dxj + idyj , dz j = dxj idyj .
which verify the Leibniz rule for and . Besides, we have the decomposition
2
d = + , 2 = = + = 0. (1.2.6)
(X, Y ) = g(JX, Y ).
is a closed 2-form.
13
J is integrable, that is J is a complex structure.
i : Ui Cn
z0 zi zi+1 zn
[z] ,..., , ,...
zi zi zi zi
defines an holomorphic local coordinate of CPn .
eF S,z a real 2-form in Cn+1 \ {0} defined by
Let
1
log kzk2 .
eF S,z =
2
Equivalently,
1
log |z0 |2 . . . + |zn |2
eF S,z =
2
1
= log (z0 z 0 + . . . + zn z n )
2
n
1 X
= dz j log (z0 z 0 + . . . + zn z n )
2 j=0
z j
1 h i
= dz 0 log (z0 z 0 + . . . + zn z n ) + . . . + dz n log (z0 z 0 + . . . + zn z n )
2 z 0 z n
1 h z0 zn i
= dz 0 + . . . + dz n
2 z0 z 0 + . . . + zn z n z0 z 0 + . . . + zn z n
n
1 X z0 zn
= dzk dz 0 + . . . + dz n
2 k=0 zk z0 z 0 + . . . + zn z n z0 z 0 + . . . + zn z n
Pn Pn Pn
z dz j=0 zj dz j
1 k=0 dz k dz k k=0 k k
= . (1.2.7)
2 kzk2 kzk4
14
Let U be an open set in CPn and : U Cn+1 \ {0} an holomorphic section,
that is, is an holomorphic map with = idU .
Claim 1.19.1.
eF S does not depend on the section .
Thus,
1
eF S,[z] =
eF S,1 ([z])
1
k1 ([z]) k2
= log
2
1
kf ([z]) ([z]) k2
= log
2
1 2 1
log k ([z]) k2
= log |f ([z]) | +
2 2
1
|f ([z]) |2 +
= log eF S,[z] . (1.2.8)
2
15
Besides,
2
log |f | = dz log f f
z
1
= dz ff
f f z
1 f f
= dz .f + f.
f f z z
1 f
= dz f.
f f z
1 f
= dz .
f z
1 f
= . dz dz
z f z
1 f 1 f
= . + . dz dz
z f z f z z
= 0, (1.2.9)
Therefore, by denoting F S :=
eF S , the previous claim implies that
F S independient of the election of the section , and since these sections
exist locally, F S is a global differential form in CPn .
0 : U0 Cn
[z] (w1 , . . . , wn ) := w
zi
with wi = z0
. Thus, for the section
: U0 Cn+1 \ {0}
[z] (1, w)
16
we obtain the following expression for F S respect to the coordinate 0 :
1
log 1 + kwk2 .
F S = (1.2.10)
2
and dF S = d (
eF S ) = de
F S = 0.
A = 0A . (1.2.12)
First of all, we observe that Un+1 acts transitively on CPn since CPn
=
S 2n+1 /S 1 and Un+1 acts transitively on S 2n+1 , which follows from the fact
that every unit vector can be extended to an orthonormal basis in R2n+2 '
Cn+1 and consequently, given two orthonormal bases in Cn+1 , the linear
transformation which carries one basis to another corresponds to a unitary
matrix.
Claim 1.19.2. A
eF S =
eF S
17
Proof.
A
eF S,z =
eF S,A (z)
1
log kA (z)k2
=
2
1
log kAzk2
=
2
1
log kzk2
=
2
=
eF S,z ,
where the fourth equality follows from the fact that A is a unitary matrix.
Claim 1.19.3. F S =
eF S
Proof. Since F S does not depend on the election of the section, we can use
the section 0 related to the chart 0 defined in the coordinate open set U0
as we did in (1.2.10) to obtain:
F S,z = F S,(z)
1 z1 zn
= log 1 + k ,..., k2
2 z0 z0
kzk2
1
= log
2 |z0 |2
1 2 1
= log kzk log |z0 |2
2 2
=
eF S,z ,
where the last equality follows from (1.2.9), since z0 is an holomorphic func-
tion defined in U0 and take values in C .
18
n
1 X
F S,z0 = dzk dz k (1.2.13)
2 k=1
Claim 1.19.4. 0
A F S = F S
0 0
A F S,[z] = A (0
eF S ),[z] =
eF S,(0 0 )([z])
A
=
eF S,(0 0 )(z) .
A
From (1.2.12),
0 eF S,(0 A )(z) = 0
A F S,[z] = eF S,(A )(z)
= F S,(A )(z)
= F S,A (z)
=
eF S,A (z)
= A
eF S,z .
0
A F S,[z] =
eF S,z
= F S,z
= F S,[z] .
19
Chapter 2
In [1], Lerman defines the notion of a symplectic cone and its relationship
with a given contact structure as a base space. It is stated as follows.
the action of the real line R expands the symplectic form exponentially.
That is, = e , where denotes the diffeomorphism defined by
R.
GM M M
20
(g, m) 7 (g m, m)
is proper.
2. the Lie derivative of the symplectic form with respect to the vector
field X is again : LX = ,
Remark 2.4. From Definition 1.17, we note that X is a Liouville vector field
for the symplectic cone (M, , X).
21
2.2 Contact manifolds and contact transfor-
mations
Definition 2.7. A 1-form on a manifold B is a contact form if the following
two conditions hold:
Proof. Note first that since f is positive then in particular nowhere zero,
kerf = ker. Thus to show that f is contact, it is enough to check that
d(f )|D is nondegenerate, where D = ker = kerf . Now d(f ) = df +f d
22
and |D = 0. Therefore d(f )|D = f d|D . But f is nowhere zero and d|D is
nondegenerate because is a contact form by assumption. Thus d(f )|D is
nondegenerate.
23
Under the same conditions of the previous theorem, there is a 1-form
0 = dz ni=1 y i dxi which is the standard contact form in R2n+1 where
P
Remark 2.18.
= U 0 (U d0 )n
= U 0 U (d0 )n
6= 0.
24
contact transformations forms a subpseudogroup denoted by sCon .
Therefore, we can expose a more general definition of a contact manifold in
terms of contact transformations. .
LX 0 = f 0
Definition 2.22. Let (B1 , D1 = ker1 ) and (B2 , D2 = ker2 ) be two co-
orientable contact manifolds. A diffeomorphism : B1 B2 is a con-
tactomorphism if the differential d maps D1 to D2 preserving the co-
orientations. That is, 2 = f 1 for some positive function f .
25
Remark 2.25.
G
Dp = D = ker
pM
di = d(fij j )
= dfij j + fij dj .
Therefore
= (fij )n (dj )n |D
26
Consequently, if n is odd , we obtain from (2.2.3) that the sign of the volume
form depends only on D but not on the choice of , so the contact structure
D induces a natural orientation on B.
And in case n is even, from (2.2.2) we will have that D is orientable, which
means that B is co-orientable. In this case, B has a strict contact structure
if and only if we can choose the fij all positive, that is, B is orientable.
1. () = 1
2. y d = 0.
X X y ( (d)n ).
b y (d)n = 0. (2.2.5)
n
()(d)
b = (d)n .
27
Thus ()
b = 1.
= (b y d) (d)n1 + d (b y (d)n1 )
= n(b y d) (d)n1
28
1-forms i defined in Ui by
X tj
i = dxi + dxj
j=06=j
t i
X tj
= dxj .
j
ti
In Ui Uj , we have that
ti
j = i ,
tj
and this defines the contact line bundle L which is non-trivial since it is
induced by the tautological line bundle on RPn . Hence, there is no globally
defined contact 1-form on M which defines the contact structure, that is, the
contact structure is not strict. We can obtain that
n+1
n ti
j (dj ) = i (di )n ,
tj
where dx
c means the dx is omitted.
Now the 1-form = ni=0 (xi dxi + yi dyi ) is normal to the sphere S 2n+1 and
P
29
we get that
n
X
(d)n = 2n n! (xi )2 + (yi )2 dx0 dy0 . . . dxn dyn .
i=0
Definition 2.31. Let B be a strict contact manifold, and let Con(B, D) de-
note the group of global contact transformations, that is, the subgroup of the
group Diff (B) of diffeomorphisms of B that leaves the contact distribution
D invariant. Alternatively fixing a contact form such that D = ker, then
Con(B, D) can be characterized as
With the 1-form fixed we are also interested in the subgroup Con(B, ) of
global strict contact transformations defined by the condition = .
30
Proof. We have that
0 = 0 + ([X, ]) . (2.2.7)
Then
0 = L[X,] + 0. (2.2.8)
0 = 0 + [X, ] y d. (2.2.9)
Proposition 2.33. Let (B, D = ker) be a contact manifold. The linear map
from contact vector fields to smooth functions given by X f X := (X) is
one-to-one and onto.
() = 1.
31
For every section v of D B and every vector field X in B,
0 = LX 0 = LX ((v))
Now, let X be a contact vector field, then ([X, v]) = 0 which, according to
the previous equation, it leads to
(LX ) (v) = 0
(X y d + d (X y )) (v) = 0
Let us define the linear map from contact vector fields to smooth functions
by
X f X := (X).
Thus, for every section v of D and for every contact vector field X in B,
= df X (v). (2.2.11)
Moreover,
X X fX y d = f X y d
= 0, (2.2.12)
= df X |D , (2.2.13)
32
for every contact vector field X in B. Consequently, if we assume that
f X = f Y for every contact vector fields X, Y in B, it follows from what we
observed in (2.2.10) and (2.2.13) that X = Y in D, and, as T B = D R
where X = X (X) + (X) with (X) R, we obtain that the linear
map is 1-1.
We are going to see that the linear map is onto. Indeed, for every f C (B)
and from (2.2.13), there exists a unique section Xf0 of D, such that:
Xf := Xf0 + f ;
we observe that
= (Xf0 ) + f = f. (2.2.15)
From (2.2.14), (2.2.15) and the definition of the Reeb vector field :
LXf (w) = d(f )(w1 ) + d(f )(w2 ) d(f )(w1 ) + (Xf y d)(w2 )
33
Now, as w2 R, we can write w2 = t for some t R, then:
LXf (w) = LXf (w2 ) = LXf ((w2 )) ([Xf , w2 ])
= ([Xf , w2 ]) . (2.2.17)
([Xf ,w2 ])
where g = t
C (B) and (w) = t, which implies that Xf is a
contact vector field in B.
34
Then on an intersection U U
! !
X X
s0 s0 = s (s s ) s (s s )
6= 6=
!
X
= s s (s s ) + (s s ) (s s )
6=,
X
= s s ( )(s s )
= 0.
Proposition 2.35. Let (M, , X) be a symplectic cone, let B be its base and
let : M B denote the projection. Pick a trivialization : B R M .
Then = d(et ) where t is a coordinate on R and is a contact form
on B. Conversely, if is a contact form on B then (B R, d(et ), t ) is a
symplectic cone.
:BRM
35
As d = 0 and LX = , then
= LX = d(X y ). (2.3.2)
LX = d(X y ) + X y d = X y d
= . (2.3.3)
As (X) = 0 and X =
t
, we obtain that = t (1-forms in B R) does
not depend on dt. So we can set
Y = (a11,2,...,2n+2 ) 2n+1
+ ... + (a1,2,...,2n+2 ) 2n+2
+ (a1,2,...,2n+2 ) , (2.3.5)
x1 x2n+1 t
36
Consequently, by taking the exterior derivative in (2.3.7) and in view of the
identification made above of in B:
for every p B.
1
Let us prove that is a contact form in B. By setting n = 2
dim M 1,
we know that n+1 6= 0 in M , and since (2.3.8) and the fact that is a
trivialization, we obtain :
n+1
d(et ) 6= 0. (2.3.9)
LX = d( y d(et )) = d( y (et dt + et d)) = d(et + 0) = ,
t t
37
Let (B, ) be a strict manifold of dimension 2n + 1 with Reeb vector field
, and M its symplectic cone. On M we define S(M, ) as the group of
symplectomorphisms of (M, ), and S0 (M, ) the subgroup of S(M, ) that
commutes with homotheties, which are the ones that satisfy = e with
Diff(M ) and R.
where is the Liouville vector field which generates the flow of the homoth-
eties.
According to the definitions given in Definition 2.31,
where
C (B) = {f C (B)|t f = f where t is the flow generated by the Reeb vector field }.
Proof. con(B, ) ' C (B) : In fact, we observe that we can use the
same linear map X (X), in this case, for every X con(B, ), cf.
38
Proposition 2.33 . Consequently, we have that the linear map is 1-1.
From Proposition 2.33, we obtain that con(B, D) ' C (B).
Then, there exists X con(B, D), that is, LX = h for some h
C (B), such that f = (X). Let us prove that our map is onto.
Let f C (B) , that is,
L ((X)) = 0
0 + ([, X]) = 0
(LX ) () = (h) ()
LX (()) ([X, ]) = h
0 = h,
s0 (M, ) con(B, )
X XB
39
where the V i s and W j s are the coefficients for X and , respectively.
By calculating we obtain that:
!
2n+2
a11,...,2n+2
2
a1,...,2n+2 a1,...,2n+2
1
=0
t x t x2 t t
2n+2
a11,...,2n+2 a1,...,2n+2
= 0 = = , (2.3.13)
t t
which means that our vector field X has no any coefficient in the t
coordinate. So we can set XB := X.
It only remains to prove that our map is well defined. In fact, let t be
the flow generated by X, the trivialization taken in Proposition 2.35
and t the flow generated by the Liouville vector field .
Since t = ,
(t ) =
(t ) = d(e )
t ( ) = d(e )
t d(e ) = d(e )
d t e = d(e ).
(2.3.14)
d t e e = 0,
(2.3.15)
t e e = df (2.3.16)
40
is in the center of con(B, ): This is exactly what we proved in Lemma
2.32.
41
Chapter 3
The notion of contact reduction arises from the natural interplay between the
symplectic cones and contact manifolds. As expected, the notion of symplec-
tic reduction plays a key role in understanding the concept of reduction at
the level of contact structures. We begin this chapter explaining symplectic
reduction performed on the complex projective space CPn . This example will
be very useful, at the level of contact reduction, especially for contact toric
manifolds. First some preliminaries.
42
1. For every X g, let
X : M R
X (p) := h(p), Xi
Then
dX = X # y .
Adg = g .
Proof. Let us recall that , due to the nondegeneracy of , for every 1-form
, there is a unique vector field such that
y = .
As H is invariant,
H exp(tX) (p) = H(p)
43
for every X g. Besides, as is a moment map for , it follows that
dX = X # , so by differentiating over t = 0 we have
0 = dH(p) X # (p)
= LX # H
= {H, X }
= (dH , X # )
= (X # , dH )
= dX (p) dH (p)
where XH = dH and the last equality follows from the fact that is a
moment map of . This implies that (Ft (p)) = (p) for every p M and
Ft flow of dH .
h(p), i = # y (p),
0 = # y d + d( # y )
= # y + d( # y ).
44
That is,
d( # y ) = # y
(Adg1 )# = g ( # )
d
(Adg )#
p = exp tAdg (p)|t=0
dt
d
= g(exp t)g1 (p)|t=0
dt
d
(g exp t ) g 1 p
=
dt
(g )g1 p # g1 p
=
g1 # p .
=
45
Consequently, we obtain that
(Adg1 )# y (p) = g ( # ) y (p)
g ( # ) y g (p)
=
g # y (p)
=
# y (g (p))
=
: S1 C C
eit , z 7 eit z.
46
Thus, from (3.1.1) and (3.1.2):
X# = .
If we want X # to be expressed in z and z coordinates,
z z
X# = . + .
z z
= (iz). + (iz)
z z
=i z z . (3.1.3)
z z
Additionally, from (1.2.3), (3.1.3) becomes, in real coordinates
# 1 1
X = i (x + iy) i (x iy) +i
2 x y 2 x y
= i i x +i y
y x
=x y . (3.1.4)
y x
Since (1.1.9), = 2i dz dz in C, and from (3.1.3):
#
i
X y (.) = i z z y dz dz
z z 2 (.)
1
= z z y (dz dz)
2 z z (.)
1
= dz(z z ) .dz (.) dz(.) .dz (z z )
2 z z z z
1
= zdz (.) + zdz(.)
2
1
= (zdz + zdz)
2 (.)
1
= d (z.z) . (3.1.5)
2 (.)
47
Next, we state one important result that determines how to produce from
coisotropic submanifolds of a symplectic manifold (the zero set of a moment
map coming from a Hamiltonian action) symplectic quotients.
(F S , CPn )
F S = i . (3.2.2)
48
The diagram above represents a symplectic reduction of Cn+1 , where S 1 acts
on CPn with the following moment map:
: Cn+1 R
kzk2 1
z 7 + .
2 2
In fact, let
n+1
1 X
= dzi dz i
2 i=1
n+1
X
= dxi dyi
i=1
n+1
X
= ri dri di
i=1
where
j : S 1 Tn+1 = S 1 . . . S 1
49
which is equivalent to say that
X# = + + ... +
1 2 n+1
n+1
X
=i zk zk . (3.2.3)
k=1
zk z k
Now, let us consider
: Cn+1 R
kzk2
z 7 + ct,
2
since
n+1
1 X 2
d = d( r )
2 i=1 i
we have that
n+1
!
X
X # y (v) = (X # , v)
ri dri di
i=1
n+1
!
X
= ri dri (v)
i=1
1X 2
= d(ri ) (v) = d(v).
2
Then, the action is Hamiltonian with moment map .
1
Besides, by taking as the constant,
2
kzk2 1
1 n+1
(0) = z C / + =0
2 2
= S 2n+1 ,
where
: Cn+1 R
kzk2 1
z 7 + .
2 2
Consequently, 1 (0)/S 1 = S 2n+1 /S 1 = CPn , that is, CPn is the symplectic
reduction of Cn+1 .
50
3.3 Contact moment maps
An alternative way to define a cooriented contact structure for a manifold is
stated in terms of the annihilator of a certain distribution of T B which is
going to be useful in understanding the way G acts in T B.
Then
D = {0} {f | f > 0} {f | f < 0}.
for all x B and all vectors X in the Lie algebra g of G, where, as above,
d
X denotes the vector field induced by X : X x = |
dt t=0
(exp tX) x
51
Intuitively, let f C (B). If we set = ef 0 where 0 D, then
d = d(ef ) 0 + ef d 0
= ef (df 0 + d 0 )
= ef d 0
h (x), Xi = (X x )
h (x), Xi = (X x ) .
If we assume that = d, in D:
d = d(ef ) = d(ef ) + ef d
= ef d
ef = ef . (3.3.1)
Remark 3.10. From this last equation it is clear that the moment map de-
pends upon the 1-form (to be more precise, it depends on the conformal
52
class of the contact form) and not on the contact structure. In [11], Lerman
proposes the definition of a universal moment map which depends on the
contact structure and not only on the contact form. This generalisation of
the contact map will be explained in the next subsection. Nevertheless, the
restricted notion of a contact moment map given in Definition 3.9 will suffice
to exhibit examples of contact reduction in Chapter 3.
g : T B T B
Tp B Tg (p) B
g (p), g .
(p, )
Thus
g ()(v) = (g v)
= g v H + v V
= g (v H ) + g (v V )
= v H + g (v V )
= g (v V )
53
Therefore, the action preserves D . On the other hand, for every D+
we
have that g ()(X) = ((g ) (X)) > 0 because the G action preserves the
cooriented structure and (X v ) > 0 where X v is the vertical vector of X.
The restriction = |D+ of the moment map for the action of G on T B
to D+ depends only on the action of the group and on the contact structure.
Moreover, since : T B g is given by the formula
h(q, p), Xi = p, X q
54
if and only if X is nowhere zero on the level set 1
(0). Hence, in this
Proof. We compute
d = d((X)) = LX X y d = X y d. (3.4.1)
along the 0level of , the fact that (d)n 6= 0 and (3.4.1) gives us that
0 is a regular value of if and only if X is nowhere zero in the level set
1
(0).
lies in this kernel, and if kerdp had dimension more than 1, it would imply
that (d)n will be zero in some point in 1
(0).
= 0
for every Yp , Zp in Tp (1
(0)) kerp with d(X p , Zp ) and d(X p , Yp ) both
by restricting to hyperplanes in Tp (1
(0)) complementary to X p . Sim-
55
such hyperplanes. It follows, as claimed, that induces a contact form on
1 1
(0)/S .
Lemma 3.12. The moment map is equivariant with respect to the given
G-action on B and the coadjoint action of G on g , that is,
Proof.
d
X gm = (exp (tX) g m) |t=0
dt
d
gg 1 exp (tX) g m |t=0
=
dt
d 1
= gm g exp (tX) g m|t=0
dt
1
= gm ad g (X) ,
m
(g m)(X) = gm (X gm )
1
= gm gm ad g (X)
m
= (g )m ad g 1 (X)
m
= m ad g 1 (X)
m
= (m) ad g 1 (X)
= g ( (m)) (X).
56
(b) The flow of the Reeb vector field preserves the level sets of .
(c) If (p) = 0, then Tp (G p), the tangent space to the orbit through p,
is an isotropic subspace of the symplectic vector space (kerp , dp ).
Let v Tp B, thus
d Xp (v) + d(X p , v) = 0 (3.4.3)
Thus, dp () = 0 which means that the flow of the Reeb vector field
preserves the level sets of .
57
(c) The tangent space Tp (G p) is spanned by vectors of the form X p with
X lying in g by the isomorphism between g and Te G . In particular, it is
a subspace of kerp , since (p) = 0 and then p (X p ) = h (p), Xi = 0.
If we take v = Y p for some Y g we obtain from (a) that
dp (Y p , X p ) =p (Y p ), X
d
= p (exp(tY ) p) |t=0 , X
dt
d
= ( (exp(tY ) p)) |t=0 , X
dt
d
= (exp (tY ) ( (p))) , X
dt
= 0,
where the third equality is a consequence of the chain rule and the
fourth is obtained by applying the Lemma 3.12.
to Tp 1
(0) , as is not in kerp and by (b) the flow of preserves
Tp (G p) and kerp Tp 1
(0)
58
are of complementary dimension in kerp .
From (a), for every v kerp Tp 1 1
(0) , p (0) and X p
= 0.
Thus,
(Tp (G p)) kerp Tp 1
(0) .
Proof. First of all, we are going to show that G acts locally freely on the
level set 1
(0). Indeed, let us call Gp = {g G | g.p = p} the isotropy
Im dp = g .
for every p 1
(0). On the other hand, Im dp Ann gp since for every
59
hT, Xi = hdp (v), Xi
= d(v, X p )
= 0,
1
(0) ,moreover those isotropy groups are finite since we are assuming that
G is compact.
Afterwards, let us show that induces a contact action on the quotient if
the action is free. In fact, if we assume that the action of G on B is free and
since G is compact, 1
(0)/G is a manifold. Since we have that g = for
In fact, the examples that will be presented in the next section, are contact
toric manifolds, these are manifolds with a large group of automorphisms
which allows the manifold to admit very symmetric groups acting on them
60
in an appropriate fashion. We have the following definition.
Remark 3.16. Lemma 3.13 reveals an important difference between the con-
tact and the symplectic case: in the proof of d) on the lemma, one notices
that other possible regular value besides zero can not ensure that Tp (G p) is
a subspace of kerp . So contact reduction, stated as theorem 3.14 only works
for zero as regular value. There is a variation of this notion, given by Willet
in [20] where it is possible to contactify quotients for certain non-zero regular
values.
Let S 1 act on S 7 by
: S1 S7 S7
61
The associated fundamental vector field of this action is (in real coordinates),
X 0 = x0 y0 x1 y1 + x2 y2 + x3 y3 .
y0 x0 y1 x1 y2 x2 y3 x3
We can proceed by the same way as we did in Example 3.4 and we will obtain
that the moment map : S 7 R is then stated (up to a factor 12 ) as
and
Thus,
1 1 3
3 1
(0) = S S . (3.5.2)
2 2
Clearly, 0 is a regular value for . Thus, the reduced space can be iden-
tified with (S 3 S 3 ) /S 1 which, by the contact reduction theorem 3.14, is a
contact manifold.
Let us identify more explicitly the manifold (S 3 S 3 ) /S 1 . In order to
do this, let us consider the following diffeomorphism
F : S3 S3 S3 S3
(z0 , z1 , z2 , z3 ) 7 (z0 z3 + z1 z2 , z0 z2 z1 z3 , z2 , z3 )
62
and the following S 1 action
: S1 S7 S7
= z0 z3 + z1 z2 , z0 z2 z1 z3 , eit z2 , eit z3
(3.5.3)
Besides,
F (z0 , z1 , z2 , z3 ) = (z0 z3 + z1 z2 , z0 z2 z1 z3 , z2 , z3 )
= z0 z3 + z1 z2 , z0 z2 z1 z3 , eit z2 , eit z3 ,
(3.5.4)
S 3 S 3 /S 1
= S 3 S 3 /S 1 = S 3 S 2.
63
which has zero as a regular value for any (0 , . . . , n1 ) such that 0 n1 6=
0, gcd (0 , . . . , n1 ) = 1 and at least two s have different signs.
Now, let us take 0 = = k = a and k+1 = = n1 = b, a, b Z+
relative prime. Then , by the same procedure made in the previous example,
p p
1 (0)
= S 2k+1 a/a + b S 2(nk)1 b/a + b ,
64
p,q
Example 3.19. The Wang-Ziller manifold Mk,l given in [19] is defined as
the total space of the S 1 -bundle over CPp CPq whose Euler class is k1 +l2
where 1 and 2 are the positive generators of H 2 (CPp ) and H 2 (CPq ), re-
spectively and k and l are integers.
In the 5-dimensional case, Wang and Ziller obtained that, for p = q = 1, the
1,1
manifolds Mk,l are diffeomorphic to S 3 S 2 . To achive this, they used ar-
guments involving calculations of espectral sequences and a famous theorem
of Smale on the clasification of 5-dimensional manifolds in [16] . They show
that these manifolds are spin and simply connected and
1,1
H 2 (Mk,l , Z) = Z, hence applying Smale theorem, they concluded that all
these manifolds are diffeomorphic to S 3 S 2 . For a detailed argument cf.
[19] or Appendix in [7]. We note that the previous example we studied is a
Wang-Ziller manifold with weights k = 1 and l = 1.
p : S1 S5 S1 S5
(z0 , z1 , z2 , z3 ) 7 (z0 )k , z1 , z2 , z3
65
p
S1 S5 S1 S5
1 2
p
(S 1 S 5 )/S 1 1 (0)/S 1
G : S1 S5 S1 S5
(z0 , z1 , z2 , z3 ) 7 (z0 , z0 z1 , z0 z2 , z0 z3 )
2 : (z0 , z1 , z2 , z3 ) 7 (eit z0 , z1 , z2 , z3 )
satisfy that:
= (eit z0 , z0 z1 , z0 z2 , z0 z3 )
2 G(z0 , z1 , z2 , z3 ) = 2 (z0 , z0 z1 , z0 z2 , z0 z3 )
= (eit z0 , z0 z1 , z0 z2 , z0 z3 ).
1 (0)/S 1
= (S 1 S 5 )/S 1 /Zk
= S 5 /Zk .
(3.5.10)
66
References
[1] Audin, M., Cannas da Silva, A., Lerman, E. (2003), Symplectic geometry
of integrable Hamiltonian systems, Birkhauser Verlag, Basel ; Boston.
[4] Boyer, Ch., Galicki, K. (2000), A note on toric contact geometry , Jour-
nal of Geometry and Physics, 35, no 4, 288-298, doi: 10.1016/S0393-
0440(99)00078-9.
[6] Futaki, A., Ono, H., Wang, G. (2009), Transverse Kahler geometry of
Sasaki manifolds and toric Sasaki-Einstein manifolds , J. Differential
Geom., 83, no 3, 585-636, doi: 10.4310/jdg/1264601036.
[7] Gauntlett, J., Martelli, D., Sparks, J., Waldram, D. (2004), Sasaki-
Einstein Metrics on S 2 S 3 , Adv. Theor. Math. Phys., no 8, 711-734
doi: 10.4310/ATMP.2004.
67
[8] Geiges, H. (2008), An Introduction to Contact Topology, Cambridge Uni-
versity Press; USA.
[12] Libermann, P., Marle C.M. (1987), Symplectic Geometry and Analytical
Mechanics , D. Reidel Publishing Co.
[13] Martelli , D., Sparks ,J. (2005), Toric Geometry, SasakiEinstein Man-
ifolds and a New Infinite Class of AdS/CFT Duals, Communications in
Mathematical Physics , 262, no 1, 51-89, doi: 10.1007/s00220-005-1425-
3.
68
[18] Voisin, C. (2002), Hodge Theory and Complex Algebraic Geometry I ,
Cambridge University Press; USA.
[20] Willet, C. (2002), Contact Reduction, Trans. Amer. Math. Soc. 354 ,
4245-4260, doi: 10.1090/S0002-9947-02-03045-3.
69