7
$\begingroup$

The Duistermaat Heckman localization formula states how integrals over symplectic spaces with Hamiltonian $U(1)$ group actions.

$$ \int_M \frac{\omega^n}{n!} e^{-\mu} = \sum_{x_i \text{ fixed}} \frac{e^{-\mu(x_i)}}{e(x_i)} $$

Here $M$ is a symplectic manifold and here are a few invariants:

  • $\omega$ is the symplectic form
  • $\mu$ is the moment map of the $U(1)$ rotation action
  • $x_i$ are the fixed points of the group action
  • $e(x_i)$ is the product of the weights of the rotation at fixed point $x_i$

In a few cases, we can produce formulas an engineer might recognize:

  • $M = \mathbb{R}^2 \simeq \mathbb{C}$ and $\omega = dx \wedge dy$
  • $U(1)$ the the rotation $z = x_1 + i x_2 \mapsto e^{i\theta}z$, the moment map is $\mu(z) = |z|^2 = x_1^2 + x_2^2$
  • Certainly $\mathbb{R}^{2n}$ is analogous.

The fixed point of the rotation is $x_i = 0$. This leads to the formula for the Gaussian integral:

$$ \int (dx \wedge dy) \,e^{-(x^2 + y^2)} = \frac{e^{-\mu(0)}}{e(0)} = \frac{1}{2\pi}$$

I'm not actually sure that $e(0) = 2\pi$ but I'm guessing. There's not too many cases were all the ingredients are known. It is possible to write a manifold (or orbifold or variety) with moment map $\mu$.

My understanding is that we can rarely write $\omega$ down explicitly. For the sphere $S^2 = \{ x^2 + y^2 + z^2 = 1\}$ there's a symplectic form $\omega = \frac{dx \wedge dy}{z}$ which is invariant under rotating aroun the $z$ axis.

Are there other cases where the DH formula specializes in a particularly nice way? I'm guessing that typically we bypass explicitly computing $\omega$.

$\endgroup$
  • $\begingroup$ Flipping through this article of Álvaro Pelayo, but still inconclusive. Also ICM 2006 article of Michéle Vergne. $\endgroup$ – john mangual Aug 7 '18 at 18:44
5
$\begingroup$

Nice examples are worked out in Audin (2004, §VI.3.d), Arvanitoyeorgos (1999)(pdf), McDuff-Salamon (1998, §5.6). It’s not true that $\omega$ is rarely explicit: e.g. on all coadjoint orbits (including $\smash{S^2}$) it is, and DH gives a formula of Harish-Chandra for their Fourier transforms; see Berline-Vergne (1983), Vergne (1983), or Guillemin-Sternberg (1984, ends of §§33 and 34).

$\endgroup$
  • $\begingroup$ i believe you. i have to read these and verify what you are saying $\endgroup$ – john mangual Aug 8 '18 at 20:57

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.