enter image description here

I am trying to figure out the symmetry factor of the above diagram in the $\phi^4$ theory. The correct answer is $S=4$ (in that the Feynman diagram gets a multiplicative factor of $\frac{1}{4}$) but I get $S=8$ with the following reasoning.

For each closed loop, we can exchange the lines that start/end in the loop and belong to it. So, each loop gives a factor of $2$. Moreover, we can exchange the two vertices and simultaneously exchange the propagators of each loop. This gives an extra factor of $2$. I think it's the last bit that I'm getting wrong. Peskin and Schroeder write (p.93) that "A third possible type of symmetry is the equivalence of two vertices" but I don't think I understand precisely what this statement means. For example, why would the two vertices of this particular Feynman diagram not be equivalent? We can simply rename the integration variables and interchange them.

  • 1
    $\begingroup$ I think, exchanging lines of second loop should not be count as an independent symmetry, because you can replace it with exchanging two loops, exchanging lines in first loop, and then exchanging loops back — the result of such combination will be exchanging lines of second loop. $\endgroup$
    – warlock
    Jan 10, 2021 at 7:56
  • $\begingroup$ Oh I see. So, generally, the echange of vertices could be consider a symmetry as long it does not produce this kind of "degeneracy" $\endgroup$ Jan 10, 2021 at 8:00
  • $\begingroup$ I am not 100% sure, but I think — yes, it is true. $\endgroup$
    – warlock
    Jan 10, 2021 at 8:04
  • 1
    $\begingroup$ When you expand the exponential of the interaction to second order to get the two vertices, it comes with a factor if 1/2. This cancels the 2 from the possibility of exchanging the two vertices. Is that the factor you need? $\endgroup$
    – kaylimekay
    Jan 10, 2021 at 8:37
  • $\begingroup$ Well, the symmetry factor is the inverse of the number you take when you include the $1/n!$ from the exponential expandion, the $1/N!$ that's in front of the interaction parameter and the number of equivalent diagrams. But practically, it's the number of ways you can change stuff on the diagram while keeping it the same. $\endgroup$ Jan 10, 2021 at 8:44

2 Answers 2


Personally, I've always found the concept of symmetry factors more confusing than they're worth. I find it much easier and faster to simply count the number of contractions. The diagram you are looking at includes the fields $$\phi_1 \phi_2 \phi_x\phi_x\phi_x\phi_x \phi_y\phi_y\phi_y\phi_y$$ $\phi_1$ can contract with 8 possible fields (any of the $\phi_x$ or $\phi_y$). Without loss of generality, we can call the field it contracts with $\phi_x$. $\phi_2$ must then contract with $\phi_y$ in 4 different ways. One of the remaining 3 $\phi_x$ (chosen in 3 ways) contracts with one of the remaining 3 $\phi_y$ (chosen in 3 different ways). All other contractions are then uniquely determined. The number of contractions is then $$8\times 4\times 3\times 3$$ Putting in the factors of $\frac{1}{4!}$ for each vertex and a factor of $\frac{1}{2!}$ from the exponentiation, the overall factor in the Feynman diagram is $$ \frac{1}{2!} \frac{1}{4! }\frac{1}{4!} 8\times 4\times 3\times 3 = \frac{1}{4} . $$

  • $\begingroup$ Thanks for the answer. I struggle a bit with the specifics of symmetry factors and while I would rather use contractions to figure it out, simply calculating the symmetry factor directly seems to be much more efficient. $\endgroup$ Jan 10, 2021 at 11:55
  • $\begingroup$ @TheQuantumMan - To each his own. The symmetry factor in the diagram you have is not that hard to work out. It's just that once the diagrams become more complicated like 4 or 5 loop diagrams in theories with 3-pt and 4-pt vertices (eg. QCD), I get completely lost, and even if I have the right answer for the symmetry factor, I'm never confident in what I have unless I verify it by doing the contractions. $\endgroup$
    – Prahar
    Jan 10, 2021 at 11:59
  • $\begingroup$ BTW, to add to the @Qmechanic's answer, if $\phi_1$ and $\phi_2$ were indistinguishable, then I would have to divide by another factor of 2 so the overall factor would be $\frac{1}{8}$. $\endgroup$
    – Prahar
    Jan 10, 2021 at 12:01
  • $\begingroup$ I see. I suppose I'm also in the same situation as you where I feel much more confident with contractions rather than directly calculating symmetry factors. Again, thanks for the answer. $\endgroup$ Jan 10, 2021 at 12:38
  • $\begingroup$ @Prahar you contracted one of the 3 remaining \phi_x with one of the remaining \phi_y, but wouldn’t that correspond to a diagram without bubbles? Bubbled have contractions at the same point, so shouldn’t you contract them \phi_x with \phi_x $\endgroup$
    – user310742
    Jun 26, 2023 at 14:57

OP's main issue seems to be that the symmetry factor $S$ depends on whether the external legs are distinguishable ($S=4$) or indistinguishable ($S=8$), respectively.

  1. On one hand, P&S on p. 93 consider the distinguishable situation$^1$ where there are not integrated over the external positions/momenta.

  2. On the other hand, OP apparently considers the indistinguishable situation where there are attached sources $J$ to each external legs and integrated over the external positions/momenta.

P&S sentence "A third possible type of symmetry is the equivalence of two vertices" seems to refer to the $\frac{1}{n!}$ in Taylor expansion of the exponential function of an interaction term.


$^1$ Although P&S on p. 93 are not considering amputated diagrams it might be helpful to mention that the external legs are treated as distinguishable in an amputated diagram.

  • $\begingroup$ So, when the momentum attached to a vertex is fixed (so the corresponding leg is distinguished), we do not consider the vertex as part of the possible permutations of vertices that might constitue symmetries of the diagram? $\endgroup$ Jan 10, 2021 at 11:58
  • $\begingroup$ I updated the answer. $\endgroup$
    – Qmechanic
    Jan 10, 2021 at 12:22

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.