One of the cornerstones of point splitting technique of calculating chiral anomaly (Peskin and Schroeder 19.1, p.655) is a symmetric limit $\epsilon \rightarrow 0$. And this is the point that I don't get. Is it really possible to take such a limit? For example, consider the expression $$ \text{symm}\,\text{lim}_{\epsilon \rightarrow 0} \Bigl\{\frac{\epsilon^{\mu}\epsilon^{\nu}}{\epsilon^2}\Bigr\} = \frac{g^{\mu \nu}}{d} \tag{19.23} $$ in $d=2$ spacetime dimensions. Let $\mu = \nu = 0$. Then $$ \frac{\epsilon^{0}\epsilon^{0}}{\epsilon^2} = \frac{1}{1-(\frac{\epsilon^1}{\epsilon^0})^2}. $$ But the latter expression either greater than 1 or less than 0 and so can't be equal to $$\frac{g^{00}}{2} = \frac{1}{2}.$$


The symmetric limit (19.23)

$$S^{\mu\nu}~:=~ \text{symm}\,\lim_{\epsilon \rightarrow 0} \left\{\frac{\epsilon^{\mu}\epsilon^{\nu}}{\epsilon^2}\right\}, \qquad \epsilon^2~:=~\epsilon^{\mu}g_{\mu\nu}\epsilon^{\nu},$$ should be thought of as a regularization prescription. It is part of a symmetric point splitting regularization scheme, cf. Ref. 1. The traditional notion of limit $$\lim_{\epsilon \rightarrow 0} \left\{\frac{\epsilon^{\mu}\epsilon^{\nu}}{\epsilon^2}\right\}$$ does not exist. The prescription $$S^{\mu\nu}~:=~\frac{g^{\mu\nu}}{d}$$ is motivated by three things:

  1. The prescription $S^{\mu\nu}$ can only depend on the metric.

  2. $S^{\mu\nu}$ should be a (2,0) tensor wrt. Lorentz transformations.

  3. If we contract $S^{\mu\nu}$ with $g_{\mu\nu}$, the result should be $1$.


  1. M.E. Peskin & D.V. Schroeder, An Intro to QFT; Section 19.1, p. 655.
  • 1
    $\begingroup$ But since such a limit does not exist, this regularization scheme is not mathematically correct, is it? How can we trust the results obtained with the help of it? $\endgroup$ – user43283 May 1 '15 at 16:03
  • 2
    $\begingroup$ 1. Regularization assigns a value to a mathematically ill-defined expression. Often this is done with the help of a regulator, but not always, cf. e.g. zeta function regularization. 2. In a physically meaningful theory, the result should not depend on regularization scheme. $\endgroup$ – Qmechanic May 1 '15 at 16:43
  • $\begingroup$ Ok, then, it turns out that the only criterion that this particular regularization scheme is valid is agreement of theory with experiment? $\endgroup$ – user43283 May 1 '15 at 18:28

Same logic used in (7.87) in Peskin itself: $l^{\mu}l^{\nu} \rightarrow \frac{1}{d}l^2 g^{\mu\nu}$.

Ref. Peskin & Schroeder, Section 7.5, p. 251.

  • $\begingroup$ Sorry, but this is wrong. Eq. 7.87 uses dimensional regularisation. Here OP is asking about point-splitting regularisation. Those are different methods. $\endgroup$ – AccidentalFourierTransform Jul 21 '18 at 17:57
  • $\begingroup$ I mentioned the logic, not the method. The symmetry property of Lorentz transformation is used to reach (19.23) and (7.87). I don't see the difference caused by the two "different methods". $\endgroup$ – Luxtau Jul 21 '18 at 18:18

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.