-
Notifications
You must be signed in to change notification settings - Fork 0
/
Lecture09.tex
executable file
·172 lines (144 loc) · 6.59 KB
/
Lecture09.tex
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
\stepcounter{lecture}
\setcounter{lecture}{9}
\sektion{Lecture 9 - Compatibility of Symplectic \\ and Almost Complex Structures}
Correction from last time: We said $\W^K = \bigoplus_{p+q=k}
\W^{p,q}$, then $d\W^{p,q} \subseteq \W^{p+1,q}+\W^{p,q+1}$.
This used the fact that $d\w = \partial \w + \bar\partial \w$.
It is only true when the complex structure is integrable. In
local coordinates $\w = \sum a_{IJ} \theta^I\wedge \theta^J$.
When we took $d$ of this thing, we forgot to take $d$ of the
$\theta^i$'s.
Let $\theta\in \W^{1,0}$, then to prove that $d\theta\in
\W^{2,0}+\W^{1,1}$, it suffices to show that $(d\theta)(x,y)=0$
if $x$ and $y$ have type $0,1$. We can calculate this using
\begin{align*}
d\theta(x,y) &= \underbrace{x\theta(y)}_0 - \underbrace{y\theta(x)}_0 - \theta([x,y])
\end{align*}
and the last term is zero if $T^{0,1}$ is involutive.
Once you have that, you can deduce that
$\partial^2=\bar\partial^2=0$.
There is a real analog of this. Suppose that $TM=E_1\oplus E_2$,
then we get that $T^*M\simeq E^*_1\oplus E^*_2$. Then we have
\[
\W^kM = \bigoplus_{p+q=k} \Gamma(\wedge^p
E^*_1)\otimes_{C^\infty M}
\Gamma(\wedge^q E^*_2)
\]
you get this structure on a ``bifoliated'' manifold (if both
$E_1$ and $E_2$ are involutive).
\vspace{5mm}
\[\xymatrix{
V \ar[r]^J \ar[rd]_{\tilde \w} & V \ar[d]^{\tilde g}\\
& V
}\]
One way to state compatibility is to say that
\begin{equation*}
\w(Jx,Jy)=\w(x,y) \tag{$\ast$}
\end{equation*}
We define $g(x,y)=\tilde g(x)(y) = \tilde \w(J^{-1}x,y) =
-\w(Jx,y)$. $(\ast)$ implies that $g$ is symmetric such that
$g(Jx,Jy)=g(x,y)$. Any two of these define the third. This
triple is called a \emph{pseudohermitian structure}. It is
called \emph{hermitian} if $g$ is positive definite.
Consider $\RR^2$, with coordinates $(p,q)$, and
$J(\pder{}{q})=\pder{}{p}$. Let $z=q+ip$, and $\w=dq\wedge dp$.
Then we have that
\[
g\left(\pder{}{q},\pder{}{q}\right) = -\w\left(J\pder{}{q},\pder{}{q}\right) = 1
\]
Good, so we have the correct sign.
If we are on a symplectic [almost complex] manifold, can we find
compatible almost complex [almost symplectic\footnote{Not
necessarily closed.}] structure. The answers are yes.
Consider the anulus in $\CC$, with the inner and outer circles
glued radially. This is a complex manifold, topologically $S^1\times
S^1$, which has a symplectic structure.
Figure 1
Now if we do the same thing in $\CC^n$, we get a $S^{2n-1}\times
S^1$, which cannot have a symplectic structure.
If $J$ is integrable, we get a K\"ahler structure.
Locally, every complex structure looks like $\CC^n$, and every
symplectic structure looks like $\RR^{2n}$. But when we put the
two together, the $g$ has weird twists and doesn't have to be
equivalent to the usual metric.
What can we say anything about the geodesic flows of such things.
I dunno. It can be tied to the behavior of the laplacian (on
functions), but I've never seen anything about this.
If we think of almost complex structure as
Figure 2
then for $\w \in \wedge^2 V^*$, we can complexify it to
$\w_\CC\in \wedge^2 V^*_\CC$ by requiring it to be complex
bilinear.
If we apply $\w_\CC$ to $V^{0,1}$, since $V^{0,1}$ is the graph
of $J$, we have
\begin{align*}
\w_\CC(x+iJx,v+iJv) &= \underbrace{\w(x,v)-\w(Jx,Jv)} +
i(\underbrace{\w(x,Jv)+\w(Jx,v)})\\
&=0 \quad \text{if{f} $J,\w$ compatible}
\end{align*}
So another way to say $J,\w$ compatible is to say that $V^{0,1}$
is lagrangian in $V_\CC$. So looking for compatible almost
complex structure on a symplectic manifold is the same as looking
for lagrangian somethings, and on a K\"ahler manifold, it is the
same as looking for lagrangian bi-foliations of the complexification.
What about the metric? We do a little computation. Remember
that under the projection from $V^{1,0}$ to $V$, $V$ gets the
structure of a complex vector space.
\begin{align*}
\w_\CC(\underbrace{v-iJv}_{1,0},\underbrace{{w+iJw}}_{0,1}) &=
\w(v,w)+\w(Jv,Jw) + i(\w(v,Jw) - \w(Jv,w))\\
&= 2\w(v,w) + 2i g(v,w)
\end{align*}
So by taking the complexified form and restricting it to
$V^{1,0}$, we get a hermitian form $\frac{1}{2i}\w_\CC = g -
i\w$. If we set $v=w$, we just get the $g$ part.
\vspace{2mm}
On $V\oplus V^*$, we have two natural bilinear forms:
$\w_{\pm}((x,\alpha),(y,\beta)) = \alpha(y)\pm \beta(x)$. If we
have a form $\w\in \wedge^2 V$, then the graph of $\tilde \w$ is
Dirac (is killed by $\w_+$). On a manifold, we have $TM\oplus T^*M$. Suppose we
are given $\w\in \W^2M$, then the graph of $\tilde \w$ is a
Dirac structure on the bundle $TM$ (i.e., is lagrangian
sub-bundle of $TM\oplus T^*M$).
T.~Courant: Bracket on section of $TM\oplus T^*M$ defined by
\[
[[(X,\alpha),(Y,\beta)]] = ([X,Y],\L_X\beta-\L_Y\alpha)
\]
If we stop here, this is a semi-direct product of lie algebras
(because $[\L_X,\L_Y] = \L_{[X,Y]}$). But it doesn't have the
property that something is closed under it if and only if it is
the graph of a closed form. So we change it to
\[
[[(X,\alpha),(Y,\beta)]] = ([X,Y],\L_X\beta-\L_Y\alpha+\frac{1}{2}d(i_Y\alpha-i_X\beta))
\]
This is no longer a lie algebra (doesn't satisfy Jacobi) and is not a semi-direct
product. It has the property that the graph of $\tilde \w$ is
closed under this bracket if and only if $d\w=0$. A lagrangian
sub-bundle of $(TM\oplus T^*M,\w_+)$ is called an \emph{almost
dirac} structure, and an almost dirac structure whose sections
are closed under bracket is called a \emph{dirac structure}.
\begin{theorem}
If $E$ is a Dirac structure, then the restriction to $E$ of the
bracket satisfies the Jacobi identity. That is, $E$ is a lie
algebra under $[[\ ,\, ]]$.
\end{theorem}
\[\xymatrix{
E \ar@{^(->}[r] \ar[dr] & TM\oplus T^*M \ar[d]^\rho\\ & TM
}\]
sends $[[\ ,\, ]]$ to $[\ ,\,]$.
Suppose you have $\tilde \pi:T^*M\to TM$, which corresponds to
$\pi\in \wedge^2 T^{**}M = \wedge^2 TM$. Given $\pi$, we get an
almost Dirac structure, which is the graph of $\tilde \pi$. What
is the condition on $\pi$ that is equivalent to this being a
Dirac structure? That is, when are things like $(\tilde \pi
\alpha,\alpha)$ closed under the bracket?
\begin{theorem}
Given $\pi\in \wedge^2 TM$, define $\{f,g\} = \pi(df,dg)$ on
$C^\infty M$. Then the graph of $\tilde \pi$ is a Dirac
structure if and only if $\{\ ,\, \}$ satisfies the Jacobi
identity. In this case, we call $\pi$ a \emph{Poisson
Structure}.
\end{theorem}
If you have a symplectic structure $\tilde\w:TM\to T^*M$, it is
invertible, so we get a $\tilde \pi$. Define $\{f,g\} =
\w(\tilde\w^{-1} df,\tilde\w^{-1}dg)$