3
$\begingroup$

Let us define the following conformal integral:

$$X_{1234} = \int \frac{d^4 x_5}{(2\pi)^8} \frac{1}{x_{15}^2 x_{25}^2 x_{35}^2 x_{45}^2}\tag{1}$$

This is the box integral in position space, and it is finite when the $x_i$'s aren't equal.

I am interested in knowing if the integral $X_{1123}$ is divergent or not. In this paper, they say in Appendix A.$2$ (p.$25$) that $X_{1234}$ diverges logarithmically in the limit $x_1 \rightarrow x_2$. However, since I am only interested in the divergence occurring at $x_5 \sim x_1$, I can set:

$$\left. X_{1123} \right|_\text{div} \sim \frac{1}{(2\pi)^4 x_{12}^2 x_{13}^2} \int \frac{d^4 x_5}{(2\pi)^4} \frac{1}{x_{15}^4}, \tag{2}$$

and the remaining integral is known to be $0$, as shown in several sources such as the QFT book by Schwartz (eq. (B$.48$), p. 829), since the IR and UV divergences cancel each other.

So did the authors of the paper forget that this integral vanishes? Or is $X_{1123}$ indeed divergent, and if yes how can I extract the corresponding divergent term?

$\endgroup$

1 Answer 1

1
$\begingroup$

I suspect the integral is in Minkowski signature which requires more care, but let me take the simpler Euclidean case. If the points $x_1,x_2,x_3,x_4$ in $\mathbb{R}^4$ are distinct, then the integral $$ \int_{\mathbb{R}^4\backslash\{x_1,x_2,x_3,x_4\}}\frac{d^4 x_5}{(2\pi)^8}\ \frac{1}{(x_1-x_5)^2(x_2-x_5)^2(x_3-x_5)^2(x_4-x_5)^2} $$ converges rigorously in the sense of Lebesgue theory of integration. In particular there is no divergence at $x_5\sim x_1$ (divergence in the sense that the integral is not well defined). The above integral defines a function $X_{1234}(x_1,x_2,x_3,x_4)$ of the distinct points $x_1,x_2,x_3,x_4$.

Now if you take $x_2,x_3,x_4$ fixed and distinct, and then consider the limit $$ \lim_{x_1\rightarrow x_2} X_{1234}(x_1,x_2,x_3,x_4) $$ you will find infinity or more precisely a behavior of the form $$ X_{1234}(x_1,x_2,x_3,x_4)\sim C\log|x_1-x_2| $$ for some constant $C=C(x_2,x_3,x_4)$ when $x_1\rightarrow x_2$.

$\endgroup$
6
  • $\begingroup$ Thank you for your answer. The integral is indeed in Euclidean space, sorry for forgetting to mention that. You don’t really explain why the limit is logarithmically divergent, do you? In particular, you don’t address the point that I raised with eq. (2). $\endgroup$
    – Pxx
    Commented Jan 22, 2020 at 8:00
  • $\begingroup$ Sorry my answer is as complete and detailed as you would like it to be. I would need more time (which I don't have right now) to elaborate. Hopefully, I should be able to comlete it with: 1) a proof of the asymptotic statement I gave for $X_{1234}$, 2) the consideration of a different question (more important to you I presume) which is: take $x_1=x_2$ which make the integral ill defined, put a UV cutoff $\epsilon>0$ by restricting the domain of integration to the complement of a ball ball of radius $$ around $x_1$, and show that this diverges logarithmicaly when $\epsilon\rightarrow 0$... $\endgroup$ Commented Jan 22, 2020 at 14:59
  • $\begingroup$ ...This in fact illustrates a way of understanding the OPE that I mentioned on page 6 of my article "QFT, RG, and all that, for mathematicians, in eleven pages", arxiv.org/abs/1311.4897 Namely, showing the OPE is proving a commutation of limits: $x_1\rightarrow x_2$ then $\epsilon\rightarrow 0$ (renormalization of composite operators) versus $\epsilon\rightarrow 0$ then $x_1\rightarrow x_2$ (point splitting construction of composite operators). $\endgroup$ Commented Jan 22, 2020 at 15:15
  • $\begingroup$ Thank you for the additional details and for the paper! In response to your first comment, what about an IR divergence? The point of eq. $(2)$ in the OP is that the IR and UV divergences cancel in integrals of that form. I do not doubt you are right with saying that the integral diverges, but I am yet to see a compelling argument for why there should be no IR divergence cancelling the UV one in that case. Sorry if you felt that this was clear in your answer. $\endgroup$
    – Pxx
    Commented Jan 22, 2020 at 16:13
  • 1
    $\begingroup$ Okay, that makes sense. Then I have no problem seeing the logarithmic divergence anymore, thanks a lot! $\endgroup$
    – Pxx
    Commented Jan 22, 2020 at 18:35

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.