On P.217 of Quantum Field Theory by Peskin and Schroeder, it is stated that the Eq.(7.16) is infrared divergent and therefore a small photon mass $\mu$ is added to the photon propagator. The equation in question reads

$$-i\Sigma_2(p)=(ie)^2\int \frac{d^4k}{(2\pi)^4}\gamma^\mu\frac{i(k\!\!\!\big /+m_0)}{k^2-m_0^2+i\epsilon}\gamma_\mu\frac{-i}{(p-k)^2-\mu^2+i\epsilon}$$

It seems to me that the integral might be infrared divergent at $\vec{k}\rightarrow 0$, if one exchanges the momentum of the virtual eletron $k$ with that of the photon $(p-k)$, the above expression is then rewritten as

$$-i\Sigma_2(p)=(ie)^2\int \frac{d^4k}{(2\pi)^4}\gamma^\mu\frac{i(p\!\!\!\big /-k\!\!\!\big /+m_0)}{(p-k)^2-m_0^2+i\epsilon}\gamma_\mu\frac{-i}{k^2+i\epsilon}$$

First, I tried to carry out the integral in $k_0$. By choosing the contour which goes around the upper half of the complex plane, one picks the poles at $k^0=-|k|+i\epsilon$ and $k^0=k^0_{-}=p^0-\sqrt{p^{02}+|\vec{k}|^2-2\vec{p}\cdot\vec{k}}+i\epsilon$ (the latter corresponds to one of the two poles $k^0_{\pm}=p^0\pm\sqrt{p^{02}+|\vec{k}|^2-2\vec{p}\cdot\vec{k}}\mp i\epsilon$ at $k^2-2p\cdot k=0$). The resulting expression turns out to be the summation of two terms (considering that $p$ is on the mass shell)

$$-i\Sigma_2(p)\\=-ie^2\int \frac{d^3k}{(2\pi)^3}\left[\gamma^\mu\frac{i(p\!\!\!\big /-k\!\!\!\big /+m_0)}{2(p^0+|p|)|k|}\gamma_\mu\frac{-i}{-2|k|}|_{k^0=-|k|}+\gamma^\mu\frac{i(p\!\!\!\big /-k\!\!\!\big /+m_0)}{-2\sqrt{p^{02}+|\vec{k}|^2-2\vec{p}\cdot\vec{k}}}\gamma_\mu\frac{-i}{2p\cdot k}|_{k^0=k^0_-}\right]$$

So that the resulting integral of $\vec{k}$ goes like $\sim k^2dk(1/k^2+\text{no singularity})$, which does not diverge at $|k|\rightarrow 0$. What am I missing?

Also, if one evaluates the integral of $k^0$ by choosing the lower half of the complex plane (which seems also valid to me, since the module of $k^0$ becomes infinity as well and the integral at infinity approaches $2\pi k^{0}/k^{04} \rightarrow 0$), different poles will be picked but the resulting expression seems to be the same.

Edit: user110373's answer provided below explains that the divergence is infrared as it is due to the pole at $k^2=0$. I thought about it but did not understand well. It seems to me on P.199, when it discusses the infrared divergence of vertex function, the divergence disappears when $\mu$ stay finite (its origin connected to $k^2=0$ is not straightforward there). However, in the present case, the divergence is always there (for $k\rightarrow +\infty$) even when $\mu$ is finite.

  • 1
    $\begingroup$ Just count powers. :) $\endgroup$
    – user122066
    Jul 8, 2016 at 4:37
  • $\begingroup$ Thanks for the hint, but I still didn't get it, would you please be more specific? $\endgroup$
    – gamebm
    Jul 8, 2016 at 9:11

1 Answer 1


If you look at the first part in your expression after integrating out $k^0$, you will find it divergent, i.e.

$$\lim_{L\to\infty}\int_{L^3} \frac{d^3k}{(2\pi)^3}\left[\gamma^\mu\frac{i(p\!\!\!\big /-k\!\!\!\big /+m_0)}{2(p^0+|p|)|k|}\gamma_\mu\frac{-i}{-2|k|}|_{k^0=-|k|}\right]=\infty $$

as one can direct count the power of $|k|$ in the integrand.

This is exactly your infrared divergence. Because the condition $k^2=0$ actually gives you a unbounded hyper surface instead of a point, the behaviour of your integral around $k^2=0$ should really be probed by the above term you wrote, which is obviously divergent.

  • $\begingroup$ Many thanks for the explanation. I still have some doubts. My question is about how can I tell the origin of the divergence is NOT ultraviolet? In particular, if one inserts the (regulating) photo mass term $\mu$ into the propagator, then for the first term one has $2E_k=2\sqrt{\mu^2+|k|^2}$ instead of $2|k|$, however it seems that the divergence for large $|k|$ (you pointed out) persists(?) $\endgroup$
    – gamebm
    Jul 9, 2016 at 5:39
  • $\begingroup$ The key point is that after introducing the fictitious photon mass we no longer have the divergence from $k^2=0$. The divergence caused by $k^2=0$(IR divergence) behaves differently than the divergence caused by the usual on shell condition(UV divergence):for $k^2=0$, the Lehmann-Kallen form no longer has isolated pole and that makes the residue ill-defined. In fact, a short calculation can show that if one does not use the fictitious mass for photon, the usual procedure for 'on shell scheme' does not yield a finite result. $\endgroup$
    – user110373
    Jul 9, 2016 at 6:38
  • $\begingroup$ Thanks again for the comments. I understood the "Lehmann-Kallen form" part. But for this specific example, following the arguments of the textbook until Eq.(7.29), it seems to me when one assumes $\mu=0$ in Eq.(7.19) or (7.28), one still gets (7.29) which is UV divergent (only depends on $\Lambda$ but not on $\mu$). Does this imply that the infrared divergence in this particular case, if any, does not really affect the discussion? $\endgroup$
    – gamebm
    Jul 9, 2016 at 7:03

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.