20

What are the reasons why we usually treat Quantum Field Theory in momentum space instead of position space? Are the computations (e.g. of Feynman diagrams) generally easier and are there other advantages of this formulation?

Qmechanic
  • 201,751
Thomas
  • 1,773
  • 2
    I believe it's just because we can take the $\partial_\mu$ of spacetime and replace it with a momentum $p_\mu$, no? In other words it's the same reason that anyone does anything in $k$-space. – CR Drost Dec 20 '16 at 15:18
  • There are plenty of resources at least briefly discussing Feynman diagrams in momentum space. Have you tried googling for e.g. "position space Feynman diagram"? – ACuriousMind Dec 20 '16 at 15:18
  • I think it should be made clearer by specific examples, because I saw about four different QFT lecture styles and all of them accented the use of momentum space at different points and for different purposes. – Void Dec 20 '16 at 15:32
  • You should still be mindful of position space diagrams. Think about boundaries etc ways translation gets destroyed. – AHusain Dec 21 '16 at 12:08

2 Answers2

29

The most important reasons we use momentum space Feynman rules are:

  • In position space, the Feynman rules generate convolutions of propagators. Because of the convolution theorem, the momentum space rules generate products of propagators, which are clearly easier to handle.

  • Moreover, in position space you have an integral for each vertex, while on momentum space you have one integral per loop, and in a general diagram there are many more vertices than loops, thus making the momentum space rules easier to use.

  • What's more, the LSZ theorem in momentum space is trivial to implement: we just drop the propagators on the external lines; in position you'd have to evaluate some exponential integrals (which are straightforward, but cumbersome).

  • Finally, the renormalisation conditions are naturally imposed in momentum space, and therefore you want the diagrams in momentum space.

AccidentalFourierTransform
  • 53,248
  • 20
  • 131
  • 253
5

I may add that the expressions for propagators $G(x,x^{\prime})\propto \int \frac{\mathrm{d}^D k}{k^2+m^2} \mathrm{e}^{-\mathrm{i}k (x-x^{\prime})}$ are quite cumbersome in the position space, and have a plenty of singularities. See, for example, this Wiki article.

Andrey Feldman
  • 1,986