# A contour integral frequenctly encountered

1. Aug 29, 2014

### hneder

1. The problem statement, all variables and given/known data

Need some help here on a frequently encountered integral in Green's function formalism.

2. Relevant equations

I have an integral/summation as a product of a retarded and advanced Green's functions, looks simply like
$\sum_{p^{\prime}}\frac{1}{p^{2}-{p^{\prime}}^{2}-i\epsilon}\frac{1}{p^{2}-{p^{\prime}}^{2}+i\epsilon}$
where I have omitted the mass $m$ to make the notation simple. $\epsilon$ is a positive infinitesimal.
3. The attempt at a solution
I can convert the summation to an integral over momentum $p$, this is standard. Then I follow the complex analysis and identify 4 poles and further apply the residue theorem, what I get is an expression proportional to $\frac{1}{\epsilon}$. This means it is divergent.

Can any one point out to me what have I missed in this calculation? What mistakes did I make? Many many thanks

2. Aug 29, 2014

### nrqed

Hi.

Can you provide more details on your calculation? I am assuming you are doing the four-dimensional integral over $d^4 p$. Then did you do the $p_0$ integral using contour integration and got the $1/ \epsilon$ term before doing the remaining 3-dimensional integral?

3. Aug 29, 2014

### hneder

Thanks for the reply. This is simply a product of two Green's functions, one retarded and one advanced. No, it is a regular 3D integral (summation) and momentum $p$ and $p^{\prime}$ here are both 3D vectors. Somehow I know there is a contour integral to be done. But it is not clear to me how should I do it.

What I did was the following. If I perform the $p$ integration using residue theorem, as I mentioned in the post, I end up with four poles $p\pm i\epsilon$, $-p\pm i\epsilon$ and the final result $\sim \frac{1}{\epsilon}$. The same result comes out if I convert the $p$ integration into one with energy using $E=p^{2}/{2m}$.

Last edited: Aug 29, 2014
4. Aug 30, 2014

### vanhees71

It doesn't make physical sense if the vectors $p$ and $p'$ are Euclidean.

Anyway, as a purely mathematical exercise, it's an interesting problem, because it demonstrates the problem of Pinch singularities that sometimes occur in real-time many-body QFT if one is not careful enough, because, as you seem to have realized, the expressions do not make sense in the weak limit $\epsilon \rightarrow 0^+$. Fortunately, one can prove rigorously that such pinch singularities are absent when the correct Schwinger-Keldysh Contour techniques are applied. The details are a bit tedious, and one has to use some care to correctly make sense of the distributions involved in the calculation. See my lecture notes on relativistic many-body theory:

http://fias.uni-frankfurt.de/~hees/publ/off-eq-qft.pdf

(the long Sect. 2.2).

5. Aug 30, 2014

### hneder

Thanks, vanhees71, for your explanation. Your notes in the link are a bit too advanced for a junior student. Hopefully I will get there later. Is there any chance to explain in a simplified language? To be more specific, my problem arises from calculating, for example, the inner product of two scattered wave functions built up with Lippmann-Schwinger equation in momentum space (see, for example, in the Sakurai book), where you immediately encounter the product of two propagators described above.