-
Notifications
You must be signed in to change notification settings - Fork 0
/
Lecture04.tex
executable file
·171 lines (150 loc) · 7 KB
/
Lecture04.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
\stepcounter{lecture}
\setcounter{lecture}{4}
\sektion{Lecture 4}
Today's lecture was given by Christian.
\section*{Generating Functions}
So far we have $Sym(M_1,M_2)\hookrightarrow\lag(\bar M_1\times
M_2)$. The idea of generating functions is this. You start with
a function on $M_1\times M_2$, and by some differentiation
process, you get a lagrangian manifold, and then you check if it
corresponds to a symplectomorphism. Starting with a function and
getting a morphism is easy (because it is differentiation), and
the other way is hard.
Let's say that $M=T^*X$ and $\mu\in \W^1(X)$. Then we have seen
that $\im \mu\in \lag(M)$ if and only if $d\mu=0$. Thus, we can
just start with a zero form (a function), and differentiate it to
get a lagrangian manifold.
Now consider the case where $M_1=T^*X_1$ and $M_2=T^*X_2$. Then
we have $\overline{T^*X_1}\times T^*X_2\xleftarrow{\sigma}
T^*(X_1\times X_2)$ given by $\sigma((x_1,x_2),(\xi_1,\xi_2)) =
((x_1,-\xi_1),(x_2,\xi_2))$. This is called the Schwartz
transform. Now we can consider $f\in C^\infty(X_1\times X_2)$,
then $\im df \in \lag(\bar M_1\times M_2)$. How do we check if
this is the graph of some symplectomorphism, $\phi$? We must
have that $\phi(x_1,-\xi_1)=(x_2,\xi_2)$, which happens if and
only if $\xi_1=-d_1f, \xi_2=d_2f$\footnote{These are the natural
projections of $df$.} That is,
\begin{align*}
\xi_{1i} &= -\frac{\partial f}{\partial x_1^i}(x_1,x_2)\\
\xi_{2i} &= \frac{\partial f}{\partial x_2^i}(x_1,x_2)
\end{align*}
By the implicit function theorem, this is locally solvable for
$x_2$ when $\det \left| \frac{\partial^2 f}{\partial x_1^i \partial x_2^i}
\right|\not=0$. It is hard to tell when we can solve this
globally ... you have to check it separately.
\underline{Fibre-preserving diffeomorphisms}: A symplectomorphism
\[\xymatrix{
T^*X\ar[r]^\phi \ar[d]^\pi & T^*X\ar[d]^\pi\\
X\ar[r]^\psi & X
}\]
is fibre-preserving if this diagram commutes for some $\psi$.
Examples:
\begin{itemize}
\item[(1)] Take $\psi:X\to X$ diffeomorphism, and let
$\phi=T^*\psi$. How do we see this is a symplectomorphism? The
canonical 1-form does not depend on coordinates.
\item[(2)] \emph{Fibre translation.} Take $\psi=\id_X$, and let
$\phi:(x,\xi)\mapsto (x,\xi+\mu(x))$, where $\mu\in \W^1(X)$.
Then you can check that $\phi^*\alpha = \alpha + \pi^*\mu$ (you
can see this using the local description of $\alpha$):
$\xi_i(\alpha) = \frac{\partial}{\partial x} \lrcorner\alpha$.
\[
\w = -d\alpha \buildrel{!}\over{=} \phi^*\w = \phi^*(-d(\alpha
+ \pi^*\mu)) = \w+\pi^*d\mu \Leftrightarrow d\mu =0 \ (\mu=df)
\]
\item[(3)] Take $\phi:T^*X\to T^*X$ a symplectomorphism which is
fibre-preserving. Then this induces $\psi:X\to X$. We can
check that
\[
\phi = \underbrace{(\phi\circ T^*\psi)}_{\text{of type 1}}\circ
\underbrace{T^*\psi^{-1}}_{\text{of type 2}}
\]
\item[(4)] Consider $M_1 = (T^*X,\w=-d\alpha)$ and $M_2 =
(T^*X,\w_B = \w+\pi^* B$ where $B\in \W^2(X)$. Then $\w_B$ is
symplectic if and only if $B$ is closed. To check that this is
non-degenerate, check
\[
\w_{Bij} = {{\frac{\partial}{\partial x^i}\ \frac{\partial}{\partial
p_i}}\atop{\matrix{B_ij}{I}{-I}{0}}}
\]
\begin{itemize}
\item[(1)] $\w\mapsto \phi^* = \w$
\item[(2)] $\w \mapsto \phi^*\w = \w + \pi^*d\mu$ is a
symplectomorphism if and only if there is a $\mu\in \W^1(X)$
such that $B=d\mu$. This is not always possible when
$H^2(X,\RR)\not=0$. (This $\mu$ is often called $A$)
\end{itemize}
\item[(5)] If the manifolds are the same, then $\aut
(M,\w)\hookrightarrow \lag(\bar M\times M)$. The first is a Lie
group, so we can study the Lie algebra. The Lie algebra of a lie
group $G$ is $T_eG$ as a vector space. Take $t\mapsto \phi_t$ to be a
smooth curve in $\aut(M,\w)$ such that $\phi_0=\id_M$. By smooth
we mean that $\phi:M\times \RR\to M$ is smooth.
$\frac{d}{dt}\phi_t(x)|_{t=0} = \sqrt{x}$, then
\[
0 = \frac{d}{dt}(\phi^*_t\w - w)|_0 = \L_v \w = 0.
\]
A vector field with $\L_v\w=0$ is called \emph{symplectic}. The
collection of symplectic vector fields is the lie algebra of
$\aut(\bar M\times M)$.
Is $\L_{[v,w]}\w=0$? $\L_v = d\circ i_v + i_v\circ d$ and
$i_{[v,w]} = \L_v i_w - i_w\L_v$. Thus we compute
\begin{align*}
\L_{[v,w]}\w &= (d\circ i_{[v,w]} + i_{[v,w]}\circ d)\w \\
&= d(\L_v i_w - i_w\L_v\lrcorner)\w\\
&= d(\L_vi_w \w) \\
&= \L_v (d i_w \w)& (\L_v d = d\L_v) \\
&= \L_v(\L_w - i_v\circ d)\w = 0
\end{align*}
so this really is a lie algebra. Note that $\L_v\w = 0$ if and
only if $d(i_v\w) = 0$ (e.g., for $i_v\w = df$ for a function
$f$). Recall that we have $\tilde \w:TX\to T^*X$ ... note that
$\tilde \w(v) = i_v\w$. Thus, we can solve $\tilde \w(v) =df$,
so $X_{f} = v = \tilde \w^{-1}(df)$. This is called \emph{the
hamiltonian vector field generated by $f$}.
Locally, you can always find $f\in C^\infty(U_X)$, so symplectic
vector fields are often called \emph{locally hamiltonian}.
So we have the following picture:
\[\xymatrix{
f\ar[r]^{\text{``differ''}} & X_f\ar[r]^{\text{\tiny integrate}} &
flow\, \phi_t \ar[r] & graph\, \phi_t\in \lag(\bar M\times M)
}\]
\[
\begin{pspicture}(-.3,0)(3.5,3.1)
\psline(0,0)(0,3) \rput(-.3,2.8){$\bar M$}
\psline(0,0)(3,0) \rput(3.3,0){$M$}
\psline(0,0)(3,3) \rput(2.1,2.8){$\id_M$}
\pscurve(-.1,.1)(.5,1)(2,1.5)(2.7,2.6)(3,2.9) \rput(3.1,2.6){$\phi_t$}
\psline[linecolor=darkgray]{->}(.75,.75)(.5,1)
\psline[linecolor=darkgray]{->}(.45,.45)(.21,.69)
\psline[linecolor=darkgray]{->}(1.9,1.9)(2.15,1.65)
\psline[linecolor=darkgray]{->}(1.6,1.6)(1.8,1.4)
\psline[linecolor=darkgray]{->}(2.2,2.2)(2.37,2.03)
\pscurve{->}(1.8,.8)(1,1)(.7,.9) \rput(2.1,.7){$X_f$}
\end{pspicture}
\]
What are the generators of these lagrangian submanifolds?
\end{itemize}
\underline{A glimpse of Hamilton-Jacobi theory}: From mechanics:
Call the base manifold $Q$, with coordiantes $q^i$, called the
configuration space. Then $T^*Q$ is phase space, with
coordinates $q^i, p_i(\alpha) = \frac{\partial}{\partial
q^i}\lrcorner \alpha$ (the second is \emph{momentum}). Then we
have $H\in C^\infty(T^*Q)$, which we usually think of as energy.
Then the flow, $\phi_t$, generated by $X_H$ is given by
$\phi_t:(q^i,p_i)\mapsto (\bar q^i,\bar p_i)$. Then we can think
about the generating functions of the lagrangian submanifolds
given by the $\phi_t$s. Call them $S:Q\times Q\times \RR\to
\RR$, $S=S(q,\bar q,t) = \int_q^{\bar q} \L\, dt$ (this is
integrating the Lagrangian) ... more or less $H=qp-\L$. Then
conservation of energy says what? $X_H H=0$
(follows from $\w$ being skew-symmetric). Thus, $E=H(q,p) = H(\bar q^i,\bar
p_i) = [[H(d_2S)=E]]$ since
\begin{align*}
p_i &= -\frac{\partial S}{\partial q^i}\\
\bar p^i &= \frac{\partial S}{\partial \bar q^i}
\end{align*}
The boxed equation is what you have to solve to get at $S$. If
you are lucky, then it will be some partial differential
equation. If you're not lucky, then you'll have some arbitrary
function of derivatives.