A symplectomorphism on the cotangent bundle of the circle which is not hamiltonian

201 Views Asked by At

An example our professor gave in class for us to work out was that on $M = T^*S^1$, endowed with the canonical symplectic form on the cotangent bundle, the map $\varphi: M \to M$ given by $\varphi(q, p) = (q, p+1)$ is a symplectomorphism, even homotopic to the identity, but not hamiltonian.

By our definition, a symplectomorphism $\varphi$ is hamiltonian if it is of the form $\phi_1^H$, where $\phi_t^H$ is the flow of the hamiltonian vector field corresponding to some (possibly time-dependent) $H \in C^\infty([0,1] \times M)$. The sign convention we use in the definition of hamiltonian vector fields $X_t^H$ is that we require $$ \omega(X_t^H, \cdot) = -dH_t. $$

Idea: I know that an isotopy $\phi_t: M \to M$ is symplectic iff $\imath_{X_t}\omega$ is closed, where $X_t$ is the vector field it defines by $\dot{\phi_t} = X_t(\phi_t)$, and hamiltonian iff it is exact. If I set $\phi_t(q, p) = (q, p + t)$ for $t \in [0,1]$, I get such an isotopy from the identity to $\varphi$, so I have to compute $X_t$ and show $\imath_{X_t}\omega$ defines a nontrivial class in $H^1(M; \mathbb{R})$. My problem lies mainly with the description of the symplectic form on $M$.

Problems:

  • Description of $T(T^*S^1)$: We could realize $S^1 \subset \mathbb{C}$ as the unit circle. Then for $q \in S^1$, its tangent space is $T_qS^1 = (\mathbb{R}q)^\perp = \mathbb{R}iq$. The cotangent space thus consists of all linear functionals of the form $p = \langle \beta i q, \cdot \rangle$ for $\beta \in \mathbb{R}$. Then a general path $\gamma$ in $T^*S^1$ should look like $(q(t), p(t))$, where $q$ is a path in $S^1$, so $\langle q(t), q(t) \rangle = 1$, and $p(t)$ is the linear functional $\langle \beta(t)iq(t), \cdot\rangle$ for some real function $\beta$.

Taking the derivative of the constraints on $\gamma$, we get $0 = 2\langle \dot{q}(0), q\rangle$ and $\dot{p}(0) = \langle \dot{\beta}(0)iq + \beta i \dot{q}(0), \cdot \rangle$, where $(q(0), p(0)) = (q, p)$. Hence we should get $$ T_{(q, p)}(T^*S^1) = \{ (v, \langle w i q + \beta i v, \cdot \rangle) \mid v \in T_qS^1 , w \in \mathbb{R}, p = \langle \beta i q, \cdot \rangle \}. $$
Identifying $w$ with the functional it defines above, let us just write $((q,p),(v,w)) \in TM$. Then we have $\dot{\phi_t}(q, p) = ((q, p),(0,1))$ for all $t$, so $\phi_t$ is the flow of the vector field $X(q, p) = (0,1)$. Does this make sense to work with/ Is my description correct so far?

  • Description of $\omega$: The tautological 1-form is $\lambda_{(q,p)}(v) = p(D\pi(q, p)[v])$ in general, for $\pi: T^*Q \to Q$ the projection. So here, we should have $\lambda_{(q, p)}((v,w)) = p(v) = \langle \beta i q, v \rangle$. To compute the exterior differential, I would have used two paths $\gamma_1, \gamma_2: (-\varepsilon, \varepsilon) \to T^*S^1$ as above, and then used the coordinate free expression of $d$ $$ d\lambda(\dot{\gamma_1}, \dot{\gamma_2}) = \dot{\gamma_1}(\lambda(\dot{\gamma_2})) - \dot{\gamma_2}(\lambda(\dot{\gamma_1})) - \lambda([\dot{\gamma_1}, \dot{\gamma_2}]). $$ If $\gamma_i = (q_i, p_i)$ as above, then $\lambda(\dot{\gamma_i})$ would be the function (in $t$) $p_i(\dot{q_i})$. To compute the expression for $d\lambda$, I need to know how $\dot{\gamma_i}$ acts on this function; I realized that this probably does not make sense as $p_i(\dot{q_i})$ really is a function from $\mathbb{R} \to \mathbb{R}$ and not $M \to \mathbb{R}$. Hence, how do I go about computing this differential? And incidentally, is there a coordinate free version of the exterior differential that lets me do computations with tangent vectors as derivatives of paths?

This is the strategy I came up with so far. Any help would be appreciated, be it with regards to my thoughts above or another approach! I have seen this question, but the explanations are too brief for me to understand.