Let $\Sigma$ be a closed, orientable surface.

Then the cotangent bundle $T^*\Sigma$ has a canonical symplectic form $\omega$, given as the derivative of the tautological Liouville one-form. We can modify it to a "magnetic" form by adding some two-form $\sigma$ on the base to the symplectic form.

The notation $T^*_\sigma \Sigma$ will denote the "magnetic cotangent bundle", i.e. the cotangent bundle equipped with a symplectic form $\omega + \sigma$.

Given this, my (rather broad) question is the following: in what cases (i.e. varying $\sigma$ or the genus of $\Sigma$) is it known that a small neighborhood of the zero section in $T^*_\sigma \Sigma$ symplectically embeds into a closed symplectic $4$-manifold?

For the purposes of this question, we will suppose that $\sigma \neq 0$. Otherwise, there are many examples with $\sigma = 0$, as by the Weinstein neighborhood theorem, we can just take a neighborhood of an embedded Lagrangian $\Sigma$ in a closed symplectic $4$-manifold.


This can always be done.

Let's first treat the case when $\Sigma$ is not a torus. Then take any symplectic $4$-manifold $(M,\omega)$ where $\Sigma$ can be embedded as a Lagrangian surface. Now, take a small neighbourhood $U$ of $\Sigma\subset M$ that is symplectomorphic to a neighbourhood the zero section in $T^*\Sigma$. Let $\pi: U\to \Sigma$ be the corresponding projection. Now take $\pi^* \sigma$ on $U$ and extend it to a closed two form $\sigma'$ on $M$. This is always possible, since $\Sigma^2=-\chi(\Sigma)\ne 0$, so we don't have cohomological obstructions. Finally for some large $t$ the form $t\omega+\sigma'$ will be symplectic and it will induce on $U$ the desired magnetic form.

In case $\Sigma=T^2$ one can do everything directly on $T^4=T^2\times T^2$. We take standard symplectic $T^4=\mathbb R^4/\mathbb Z^4$ (with the form $\omega=dx_1\wedge dy_1 +dx_2\wedge dy_2$) and choose Lagrangian $T^2$ that is given by the $(x_1,x_2)$-plane. Project $T^4$ to this $T^2$ and pullback $\mathbb \sigma$ to $T^4$ from this $ T^2$. Then $\omega+\pi^*\sigma$ does the job.

  • $\begingroup$ Why does $t \omega + \sigma'$ give you the magnetic form $\omega_{\textrm{can}} + \sigma$ on $U$ (seen as a subset of $T^* \Sigma$)? Don't you need $t = 1$ for this? $\endgroup$ – Tobias Diez Oct 16 '20 at 11:10
  • 1
    $\begingroup$ Tobias, you don't really need $t=1$ for this. The point is that if you consider a map from the cotangent bundle to itself that sends $\alpha\to t\alpha$, the canonical form $\omega$ pulls back to $t\omega$. So, any cotangent bundle with $\omega_{can}$ is simplectomorphic to one with $t\omega_{can}$. At the same time, this operation of scaling the fibers by $t$ doesn't affect the pullback of $\sigma$. Hopefully this answers your question. $\endgroup$ – Dmitri Panov Oct 16 '20 at 12:01
  • $\begingroup$ Thanks for the explanation. That construction works indeed, so you just silently suppressed the fiber resealing in your answer. $\endgroup$ – Tobias Diez Oct 17 '20 at 11:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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