Let $A$ be an antisymmetric matrix.

Usually, one proves that for a Grassmann integral of the type, $$\int d\psi d\theta \exp( \psi^T A \theta) = \det(A)$$ where $\psi$ and $\theta$ are vectors of Grassmann variables.

I have a different problem which I feel should trivially reduce to the same, but it can't figure it out:

$$\int d\psi d\theta \exp( 1/2 \psi^T A \psi + 1/2 \theta^T A \theta)$$

I think it should be possible to show that my expression is equivalent to the upper one, but the necessary variable substitution escapes me.

  • $\begingroup$ The statement in the question(v1) would in general need the assumption that the matrix $A=-A^T$ is antisymmetric. $\endgroup$ – Qmechanic Jan 27 '12 at 22:03
  • $\begingroup$ Ah, okay. I know the version below doesn't require this assumption because the symmetric part of $A$ gives 0 contribution to the Grassmann integral anyway $\endgroup$ – Lagerbaer Jan 27 '12 at 22:05
  • $\begingroup$ In the first Grassmann integral of the question(v1) the symmetric part of $A$ contributes non-trivially. $\endgroup$ – Qmechanic Jan 27 '12 at 22:08
  • $\begingroup$ Okay. From this I take that I cannot transform the version 2 into version 1. But with the answer RE the Pfaffian that doesn't matter. $\endgroup$ – Lagerbaer Jan 27 '12 at 22:11

The second integral factors into the product of integrals over $\psi$ and $\theta$. In your notation, you want to show that the $\int d\psi \mbox{exp}(1/2 \psi^TA \psi) = \mbox{Pfaff}(A)$, the Pfaffian of $A$. Then use the fact that $\mbox{Pfaff}(A)^2 = \mbox{det}(A)$.

| cite | improve this answer | |
  • $\begingroup$ Ah, I think I already did that, I just had no clue what a Pfaffian was, and got stuck at proving Pfaff(A)^2 = det(A). $\endgroup$ – Lagerbaer Jan 27 '12 at 22:01
  • $\begingroup$ ...where the answer(v1) has implicitly used at one point that $A$ is antisymmetric, either in the definition of the Pfaffian, or in the very last equality... $\endgroup$ – Qmechanic Jan 28 '12 at 9:43

Define the Grassman variables $\sqrt{2}\eta=\psi+\phi$ and $\sqrt{2}\chi=\psi-\phi$.

Then $$ \eta^{T} A \chi = {1\over 2} (\psi+\phi)^{T} A (\psi - \phi) = {1\over 2}(\psi^T A\psi - \phi^T A \phi)$$

Since the middle terms cancel by antisymmetry of A and the Grassmann property. The result proves the equality of the two actions, so that the determinant of A is equal to the product of Pfaffians.

This is not quite what you wanted--- there is a minus sign in the above. But the minus sign only gives an overall minus to the whole expression if the dimension of A is odd, and it is necessary.

| cite | improve this answer | |
  • $\begingroup$ Thanks, that's a good idea. How about I define them as $\psi + i\phi$ and $\psi - i\phi$? Then, the middle terms should still cancel, but I'll get a $+$ sign. Also, for an antisymmetric matrix of odd dimension, the determinant is always 0, no? $\endgroup$ – Lagerbaer Jan 29 '12 at 18:06
  • $\begingroup$ @Lagerbaer: Yes, the "i" works in the action, but it changes the sign of the measure in the odd dimension case (because the $d\eta$ s get an "i" factor, so it is equivalent to what I did above. $\endgroup$ – Ron Maimon Jan 30 '12 at 2:21

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.