I would like to understand a bit more the aspects of the gauge theory in statistical field theory. In particular, I would like to understand how the replacement $\tau \rightarrow it/\hbar$ is performed in a mathematically proper way, when $\tau$ is an inverse temperature. This replacement comes from the resemblance between the evolution operator $e^{-iHt/\hbar}$ in quantum field theory, and the statistical weight $e^{-H\tau}$ in statistical physics (one sees then that $\tau=\left(k_{B}T\right)^{-1}$ in case you wonder :-).

In principle it leads to a compact momentum space, when the frequencies become discrete, and called Matsubara frequencies $\omega_{n}=2\pi k_B T \left(n+1/2\right)$ and $\omega_{n}=2\pi n k_{B}T$ with $n$ an integer for fermions and bosons. I'm perfectly aware of the classic book

Methods of quantum field theory in statistical physics by Abrikosov, Gor'kov and Dzyalochinski - Dover Books on Physics

but I'm stuck on the gauge formalism. Can we do a gauge transformation in imaginary time-$\tau$ as we do with real time-$t$ ? Does the imaginary-time-covariant derivative $\partial_{\tau}+A_{\tau}$ make any sense? Are there some precautions to take?

Any comment, answer, indications or even good reference (or even just keywords) about this topic is warm welcome. I precise I'm a condensed matter physicist, so if you could adapt your parlance to me (for instance, please talk slowly and loudly), I would greatly appreciate :-)

EDIT: Clearly the keyword is thermal quantum field theory and there is an associated Wikipedia page with plenty good references. Anyway, any comment are still welcome, since I progress really slowly understanding this, especially what a gauge choice mean? Thanks in advance.


Under a Wick rotation, which is what you do in order to go from Minkowski to Euclidean space, both the partial derivative $\partial_0$ with respect to time and the zero-component of the gauge field transform as

$$\partial_0\rightarrow i\partial_\tau$$

$$A_0\rightarrow iA_0.$$

This defines the covariant derivative for statistical field theory.


The general gauge transformation behaviour of the theory does not change, apart from the sign changes due to imaginary time.

In a theory that exhibits confinement, Wilson loops wrapped around the compact time direction, i.e. Polyakov loops, tell us something about the deconfinement phase transition of the theory.

For a pedagocial introduction to thermal/statistical field theory, you might consider these lecture notes.

  • $\begingroup$ Thanks for your answer. The compactness of the reciprocal space (the fact that the Matsubara frequencies are discrete) is not a problem ? What is the gauge transform in that case ? Thanks in advance for more details. $\endgroup$ – FraSchelle Mar 11 '14 at 10:07
  • $\begingroup$ @FraSchelle I have edited my answer and added some points. $\endgroup$ – Frederic Brünner Mar 17 '14 at 23:15
  • 1
    $\begingroup$ @FraSchelle: There are subtleties with the Matsubara frequencies and Ward Identities. For instance, one has to be careful when formally deriving an expression with respect to $\omega_n$ in an expression. $\endgroup$ – Adam Mar 18 '14 at 13:54
  • $\begingroup$ @FredericBrünner Thanks a lot for the complement and the reference as well. I'll have a check as soon as possible. $\endgroup$ – FraSchelle Mar 19 '14 at 10:06
  • 1
    $\begingroup$ @FraSchelle: they are useful! Ward identities impose symmetries on the propagators and other correlation functions, that you can use to prove powerful results, for example. $\endgroup$ – Adam Mar 19 '14 at 13:27

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.