3
$\begingroup$

Let $\phi$ be a homogeneous symplectomorphism of tangent bundle $\dot{T}^*M=T^*M-0_M$ and let $\alpha_M$ be the canonical Liouville 1-form of $\dot{T}^*M$. Then is it true that $\phi^*\alpha_M=\alpha_M$?

$\endgroup$
5
  • 3
    $\begingroup$ If you mean that $\alpha_M$ is equivariant with respect to the $\mathbb R^\times$-action of rescaling the fibers, this follows from $\alpha_M = \iota_{E}\omega$, where $E$ is the generator of this action and therefore preserved by any equivariant map. $\endgroup$ Sep 30, 2019 at 9:00
  • $\begingroup$ @Bertram Arnold: You mean that $\phi$ is equivariant with..., not $\alpha_M$ (otherwise you are right of course). $\endgroup$
    – abx
    Sep 30, 2019 at 19:00
  • $\begingroup$ @Bertram Arnold: I mean that $\phi$ is $\mathbb{R}^{\times}$-equivariant with respect to the action and I don't know whether $\alpha_M$ is $\mathbb{R}^{\times}$-equivariant. $\endgroup$
    – SoYu
    Oct 1, 2019 at 1:46
  • $\begingroup$ you mean 'cotangent' instead of 'tangent' $\endgroup$
    – sanette
    Oct 1, 2019 at 20:15
  • $\begingroup$ @sanette: Oh, thanks. I corrected. $\endgroup$
    – SoYu
    Oct 4, 2019 at 1:06

1 Answer 1

4
$\begingroup$

Yes this is true. You can prove it in Darboux coordinates $(x,\xi)$. Let $\phi=(\phi_1,\phi_2)$ be the symplectomorphism. Since it is homogeneous, $\phi(x,t\xi) = (\phi_1(x,\xi), t\phi_2(x,\xi))$ and differentiating wrt $t$ you get

$$ \partial_\xi\phi_1 \cdot \xi = 0$$ and $$ \partial_\xi\phi_2 \cdot \xi = \phi_2$$ (these are Euler identities).

On the other hand, $\phi$ is a symplectomorphism, so $d\phi_2 \wedge d\phi_1 = d\xi \wedge d x$. Applying these 2-forms to a pair of vectors $(U,V)$ of the form $$U=(0,\xi) \text{ and } V=(v,0)$$ you get $$d\xi \wedge d x (U,V) = \xi\cdot v $$ while $$d\phi_2 \wedge d\phi_1 (U,V) = (\partial_\xi \phi_2 \cdot \xi)(\partial_x \phi_1 \cdot v) - (\partial_\xi \phi_1 \cdot \xi)(\partial_x\phi_2\cdot v).$$ The equality between these 2-forms, combined with the Euler identities, gives $$\phi_2 \partial_x\phi_1\cdot v = \xi\cdot v,\quad \forall v,$$ which is the same as $$\phi^* (\xi d x) = \xi d x.$$ So the Liouville 1-form $\xi d x$ is preserved.

Alternative, coordinate free proof: you want to use Bertram Arnold's claim, which can be proved as follows: (since you were asking whether $\alpha$ was homogeneous).

  1. We denote the dilation action on the cotangent fibers by $h_t(\beta) =e^t \beta$, where $\beta$ is a 1-form on $M$. Differentiating with respect to $t$, you see that $h_t$ is the flow of the Euler vector field, which is tangent to the vertical fibers: $E(\beta) = (0,\beta)$. Hence $\iota_E \alpha = 0$.

  2. Now there is a nice characterization of the Liouville 1-form by the tautological formula $\beta^* \alpha = \beta$, for any 1-form $\beta$ on $M$ viewed as a map $M\to T^*M$. Hence $\beta^*(h_t^*\alpha) = (h_t\circ\beta)^*\alpha = h_t\circ\beta = e^t\beta$, which implies that $h_t^*\alpha = e^t \alpha$.

  3. You differentiate this last identity at $t=0$ to obtain $$\mathcal{L}_E\alpha = \alpha.$$ The Cartan formula together with (1.) yields $\iota_E d\alpha = \alpha$.

  4. Let $\phi$ be a homogeneous symplectomorphism: $\phi\circ h_t = h_t\circ \phi$, and $\phi^*(d\alpha) = d\alpha$. From the first equality you obtain that the operators $\iota_E$ and $\phi^*$ commute. Hence

$$ \phi^*\alpha = \phi^*(\iota_E d\alpha) = \iota_E \phi^* (d\alpha) = \iota_E d\alpha = \alpha. $$

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.