In the book "Quantum Field Theory of Point Particles and Strings" by Hatfield, in pages 114-118 the author discusses the proper relation between interacting fields and in/out fields and the spectral representation of the two-point function. In particular \begin{eqnarray} \lim_{t\to -\infty}\varphi(x)&=&\sqrt{Z}\varphi_{\rm IN}(x)\\ \lim_{t\to \infty}\varphi(x)&=&\sqrt{Z}\varphi_{\rm OUT}(x)\tag{7.10} \end{eqnarray} and then studying the spectral representation one finds that $iZ$ is the residue of the propagator at the pole corresponding to the mass of the particle. Moreover the author shows that $0 < Z \leq 1$ and that $Z=1$ if and only if $\varphi(x)=\varphi_{\rm IN}(x)$. In page 118 he further comments that $Z=1$ implies $\varphi=\varphi_{\rm IN}$ and we end up with a free field theory.

Now consider the renormalized Lagrangian for the renormalized field $\varphi^R(x)$. In the on-shell scheme one sets the residue of the propagator at the mass pole to be $i$. In particular, comparing to the above discussion, this means that for the renormalized field $\varphi^R(x)$ we have one $Z^R=1$.

But now this is rather confusing because Hatfield says that a field for which the residue of the propagator at the mass pole is $i$ coincides with in/out fields and is a free field. But the renormalized field certainly can't be free (we are in the interacting theory after all!).

So what is reallly going on here? How can the renormalized field $\phi^R(x)$ have $Z^R=1$ and this still be an interacting theory? How to reconcile this renormalization condition with the remarks by Hatfield that a field whose corresponding $Z=1$ is free?

  • 1
    $\begingroup$ Who insists? Which page? $\endgroup$
    – Qmechanic
    Oct 12, 2021 at 14:51
  • $\begingroup$ For instance in the book "Quantum Field Theory and the Standard Model" by Schwartz. In page 332 he states the condition on the residue of the two-point function. The same is done in Peskin & Schroeder, pages 324 and 325. In particular equation 10.19 gives the condition on the residue of the two-point function at the mass pole. I mean, while identifying the physical mass with the pole of the two-point function seems justified, this condition on the residue seems quite arbitrary (why set it to $i$ and not some other arbitrary value?). $\endgroup$
    – Gold
    Oct 12, 2021 at 15:01
  • $\begingroup$ @Gold regarding your last sentence in your comment here, without looking into the precise details you quote, I think the reason is that you want a canonically-normalised kinetic term in your quantum effective action. that sign feeds through to that if I remember correctly. you don't need it to be canonically-normalised, but it is convenient when deriving the Feynman rules because it is the standard convention. $\endgroup$ Oct 13, 2021 at 8:24

1 Answer 1


You have pointed out that, in an interacting theory, it is standard to define the renormalized field as $Z^{-1/2}$ times a bare field. However, if you demand that the residue of the two-point function is $i$, you will certainly not get $Z = 1$. In all but very special finite theories, $Z$ has $\frac{1}{\epsilon}$ poles in dimensional regularization.

It is also important to remember that there is nothing special about wanting the residue at the particle mass to be $i$. This is called the on-shell scheme which is merely the first of many schemes that textbooks introduce. I discussed in a previous answer why it's also fine to use completely different conditions for making things finite. In short, QFTs make predictions about scheme independent quantities. Cross section as a function of energy, for example, is scheme independent. But cross section section as a function of coupling and coupling as a function of energy on their own are not.

  • $\begingroup$ Thanks for your answer. About your first paragraph I think that the way in which I stated that was confusing. You start with a bare field $\phi$ and through (1) in the question you get $Z$. Now you define $\phi^R = Z^{-1/2}\phi$. What I mean is that if you use (1) to relate $\phi^R$ to in/out fields, it will be a relation like $\phi^R(x)\to \sqrt{Z^R}\phi_{\rm in/out}(x)$ with $Z^R = 1$. The point is that $\phi^R(x) = Z^{-1/2}\phi(x)$ and so if $\phi(x)\to Z^{1/2}\phi_{\rm in/out}(x)$ we get $\phi^R(x)\to \phi_{\rm in/out}(x)$. I meant that the field $\phi^R$ has one $Z^R = 1$. $\endgroup$
    – Gold
    Oct 12, 2021 at 15:10

Your Answer

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

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