I am trying to derive a Ward-type identity between amplitudes involving $\bar\psi \sigma_{\mu\nu}\gamma_5\psi$, $\bar \psi \gamma_\mu \gamma_5 \psi$, and $\bar \psi \gamma_5 \psi$ in QCD (diagonal quark current-mass matrix). It should be the following:

$$\partial^\nu \langle \bar \psi (0) \sigma_{\mu\nu} \gamma_5 \psi(x)\rangle=-\partial_\mu \langle \bar \psi (0) i\gamma_5 \psi(x)\rangle+m\langle\bar\psi(0)\gamma_\mu\gamma_5 \psi(x)\rangle \tag{1}$$

How do I derive this? This is different than the normal chiral Ward identity, which is:

$$\partial^\nu\langle\bar\psi(x)\gamma_\nu\gamma_5\psi(x)\rangle=2m\langle\bar\psi(x)\gamma_5\psi(x)\rangle-\underset{\textrm{anomaly}}{\underbrace{\frac{N_f}{8\pi^2}\langle F(x) \tilde F (x)\rangle}}\tag{2}$$

I understand how to derive equation (2) - simply apply a (diagonal) chiral flavor transformation to the generating functional of QCD, then expand to first order (and take into account the anomaly, for example using the Fujikawa method). However I don't see how to do that for equation (1). The LHS of (2) is $\partial^\mu J_\mu^5(x)$ where $J_\mu^5(x)$ is the Noether current associated with the diagonal chiral flavor transformation. However on the LHS of (1), the divergence involves $\sigma_{\mu\nu}$ which is not the Noether current of any global symmetry.

Also, equation (2) involves field operators evaluated at a single point $x$, whereas equation (1) is point-split.

  • $\begingroup$ Interesting, I did not know about the first identity. Are you aware of any specific use it has? $\endgroup$
    –  Mio
    Jun 14, 2021 at 9:00
  • 1
    $\begingroup$ @Mio Yeah it connects the two independent twist-3 two-particle (pseudoscalar) distribution amplitudes, the pseudoscalar and pseudotensor amplitudes. The overall normalization constants for these distributions are the same because of this identity - supposedly, since I haven't seen the explicit justification of this statment. See page 4 of this article: PhysRevD.71.014015 $\endgroup$ Jun 15, 2021 at 1:47

1 Answer 1



  • The non-commuting nature of the operators is not relevant here, because the points are split and all terms have the same operators in the same order.

  • Start with the Dirac equation: $\big\langle\bar\psi(0)\gamma_\mu \gamma_5 (i\gamma^\nu\partial_\nu+m)\psi(x)\big\rangle=0$.

From here, it's just $\gamma$-matrix identities.

  • $\begingroup$ Thanks I forgot about using the quantum-version of the equations of motion - that $$\left< \cdots \frac{\delta S}{\delta \psi (x)}\right>=0$$ where $\cdots$ represent other $\bar\psi, \psi$ fields. In the ultimate equation we define $\sigma_{\mu\nu}\equiv \frac{i}{2}\left[ \gamma_\mu,\gamma_\nu \right]$. $\endgroup$ Jun 14, 2021 at 7:11
  • $\begingroup$ Correction, above that should be $$\left<\cdots \frac{\delta S}{\delta \bar \psi (x)}\right>=0$$ $\endgroup$ Jun 14, 2021 at 7:23

Your Answer

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

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