I am trying to evaluate the following integral \begin{equation} I = \int \frac{d^d p_\text{E}}{(2 \pi)^d} \frac{1}{(p_\text{E}^2+m^2)((q_\text{E}-p_\text{E})^2 + m^2)} \tag{1} \end{equation} where $p_\text{E}$ is the momentum in Euclidean space (i.e. after having performed a Wick rotation). Using Feynman's parametrization trick, and some effort, equation $(1)$ can be written as: \begin{equation} I = \int\limits^1_0 \mathrm{d} x \int \frac{d^d p'_\text{E}}{(2 \pi)^d} \frac{1}{\left[p_\text{E}'^2 + q_\text{E}^2 x(1-x) + m^2 \right]^2} \tag{2} \end{equation} where $p'_\text{E} = p_\text{E} - q(1-x)$. I know that equation $(2)$ is correct as I have verified by checking 2 different sources. Now, according to the two aforementioned sources, equation $(2)$ can be written evaluated as: \begin{equation} I = \frac{\pi^{d/2}}{(2 \pi)^d} \Gamma(2-d/2) \int\limits^1_0 \mathrm{d} x \; (q_\text{E}^2 x(1-x) + m^2)^{d/2-2} \tag{3} \end{equation} where $\Gamma(2-d/2)$ is the gamma function. However, I don't know where this comes from.

So far, the closest I have come is by writing equation $(2)$ as: \begin{equation} I = \int\limits^1_0 \mathrm{d} x \int \frac{d^d p'_\text{E}}{(2 \pi)^d} \int\limits^\infty_0 \mathrm{d}u \int\limits^\infty_0 \mathrm{d}u \; e^{-u(p_\text{E}'^2 + q_\text{E}^2 x(1-x) + m^2 )} \tag{4} \end{equation} We can perform the integration over $p'_\text{E}$ by using Gaussian integration: \begin{equation} \int \mathrm{d}^d p_{\text{E}} \; e^{-up'^2_{\text{E}}} = \left(\frac{\pi}{u}\right)^{d/2} \end{equation} Thus equation $(4)$ becomes: \begin{equation} I = \frac{\pi^{d/2}}{(2 \pi)^d} \int\limits^1_0 \mathrm{d} x \int\limits^\infty_0 \mathrm{d}u \int\limits^\infty_0 \mathrm{d}u \; u^{-d/2}e^{-u(q_\text{E}^2 x(1-x) + m^2 )} \tag{5} \end{equation} Now, letting: \begin{equation} w = \left(q_\text{E}^2 x(1-x) + m^2 \right) u \Rightarrow \mathrm{d} u = \left(q_\text{E}^2 x(1-x) + m^2 \right)^{-1} \; \mathrm{d} w \end{equation} and substituting the above into equation $(5)$ yields: \begin{equation} \begin{aligned} I & = \frac{\pi^{d/2}}{(2 \pi)^d} \int\limits^1_0 \mathrm{d} x \int\limits^\infty_0 \mathrm{d}w \int\limits^\infty_0 \mathrm{d}w \; w^{-d/2}e^{-w} \left(q_\text{E}^2 x(1-x) + m^2 \right)^{d/2-2} \\& = \frac{\pi^{d/2}}{(2 \pi)^d} \int\limits^1_0 \mathrm{d} x \int\limits^\infty_0 \mathrm{d}w \; \Gamma(1-d/2) \left(q_\text{E}^2 x(1-x) + m^2 \right)^{d/2-2} \tag{6} \end{aligned} \end{equation} At this point I'm stuck. I have no idea if the steps taken to get to equation $(6)$ are correct, but equation $(6)$ does kind of look like equation $(3)$ so I hope they are. Could anybody help me and tell me how I can derive equation $(3)$?

  • $\begingroup$ The expression you mentioned is known as "the master formula" when doing loop calculations. QFT books probably have a derivation of it. I am pretty sure Peskin and Schroeder do, when they introduce these techniques. $\endgroup$ – JeffDror Feb 12 '14 at 12:17
  • $\begingroup$ @JeffDror I have looked at P&S, but couldn't find anything. Maybe I have missed it; I will check again. $\endgroup$ – Hunter Feb 12 '14 at 12:20
  • $\begingroup$ @JeffDror what expression in my original post is known as "the master formula". I can't find anything in P&S which could resolve my problem. $\endgroup$ – Hunter Feb 12 '14 at 12:58
  • 1
    $\begingroup$ I was referring to equation 3. This equation and its modifications are derived on pg. 249-251. The set of master formula in Minkowski space (i.e., you avoid the hassle of Wick rotation) are just quoted in the appendix, p.g. 807. $\endgroup$ – JeffDror Feb 12 '14 at 13:01
  • $\begingroup$ Also, I think you have a typo on your equation 6. I don't think it needs the extra $w$ integration. $\endgroup$ – JeffDror Feb 12 '14 at 13:05

You can prove a general formula $$ \int\frac{d^dp_{\mathrm{E}}}{(2\pi)^d} \frac{(p_{\mathrm{E}}^2)^m}{(p_{\mathrm{E}}^2+\Delta)^n}= \frac{1}{(4\pi)^{d/2}}\frac{\Gamma(m+d/2)\Gamma(n-m-d/2)}{\Gamma(d/2)\Gamma(n)} \left(\frac{1}{\Delta}\right)^{n-m-d/2},\quad n>m+d/2 $$ by using Gaussian integral and Euler integral of the first kind.

  • $\begingroup$ Thanks, however I am not sure how to apply it to the above problem. According to your identity, $n=2$ and $m=0$, thus: $$I = \frac{1}{(4 \pi)^{d/2}} \frac{\Gamma(d/2)\Gamma(2-d/2)}{\Gamma(d/2)\Gamma(2)} (\Delta)^{2-d/2}$$ which is not equal to equation $(3)$. $\endgroup$ – Hunter Feb 12 '14 at 13:14
  • $\begingroup$ And $\Delta=q_{\mathrm{E}}^2x(1-x)+m^2$, it is independent of $p_{\mathrm{E}}$. $\endgroup$ – soliton Feb 12 '14 at 13:17
  • $\begingroup$ Ahhh ok, so it does look very similar. What has happened to the $\int\limits^1_0 \mathrm{d} x$ part that should go in front of $(\Delta)^{2-d/2}$ (see equation $(3)$)? $\endgroup$ – Hunter Feb 12 '14 at 13:20
  • $\begingroup$ We cannot obtain an analytical expression of the integral (3). For $d=4$, (1) is log divergent. $\endgroup$ – soliton Feb 12 '14 at 13:29
  • 1
    $\begingroup$ The $\int_0^1d x$ is copyed from (2). The two $I$s are not the same in your comments. $\endgroup$ – soliton Feb 12 '14 at 13:43

Although the answer given by soliton is sufficient, I've found a way to explicitly evaluate this integral (in case anybody might be interested). Let us start from equation $(2)$ in the original message: \begin{equation} I = \int\limits^1_0 \mathrm{d} x \int \frac{d^d p'_\text{E}}{(2 \pi)^d} \frac{1}{\left[p_\text{E}'^2 + q_\text{E}^2 x(1-x) + m^2 \right]^2} \tag{2} \end{equation} This can be written as: \begin{equation} I = \int\limits_0^1 \mathrm{d}x \int \frac{\mathrm{d}^d p_{\text{E}}'}{(2\pi)^d} \int\limits^\infty_0 \mathrm{d} u_1 \; e^{- u_1 \left( p_{\text{E}}'^2 + q_\text{E}^2 x(1-x) + m^2 \right)} \int\limits^\infty_0 \mathrm{d} u_2 \; e^{- u_2 \left( p_{\text{E}}'^2 +q_\text{E}^2 x(1-x) + m^2 \right)} \end{equation} The integration over $p_{\text{E}}'$ becomes an ordinary Gaussian integral: \begin{equation} \int \mathrm{d}^d p_{\text{E}}' \; e^{-p_{\text{E}}'^2(u_1+u_2)} = \left( \frac{\pi}{u_1+u_2} \right)^{d/2} \end{equation} and so: \begin{equation} I = \frac{1}{(4 \pi)^{d/2}} \int\limits_0^1 \mathrm{d}x \int\limits^\infty_0 \mathrm{d} u_1 \int\limits^\infty_0 \mathrm{d} u_2 \; \left(\frac{1}{u_1+u_2}\right)^{d/2} e^{- \left( q_\text{E}^2 x(1-x) + m^2 \right)\left(u_1+u_2 \right)} \tag{3} \end{equation} We can make the following substitution: \begin{equation} \begin{array}{cc} u_1 = sx \; ,& u_2 = s(1-x) \end{array} \end{equation} where $0 < s < \infty$ (and of course $0 < x < 1$), such that: \begin{equation} u_1+u_2=s \; \; \; \text{and} \; \; \; J = \left|\frac{\partial(u_1,u_2)}{\partial(x,s)} \right| = s \end{equation} Thus, equation $(3)$ becomes: \begin{equation} I = \frac{1}{(4 \pi)^{d/2}} \int\limits_0^1 \mathrm{d}x \int\limits^\infty_0 \mathrm{d} s \; s^{1-d/2} e^{- \left( q_\text{E}^2 x(1-x) + m^2 \right)s} \end{equation} Now, we make the substitution: \begin{equation} w = \left( q_\text{E}^2 x(1-x) + m^2 \right)s \Rightarrow \mathrm{d} s = \left( q_\text{E}^2 x(1-x) + m^2 \right)^{-1} \mathrm{d} w \end{equation} such that: \begin{equation} \begin{aligned} I & = \frac{1}{(4 \pi)^{d/2}} \int\limits_0^1 \mathrm{d}x \int\limits^\infty_0 \mathrm{d} w \; w^{1-d/2} e^{- w} \left( q_\text{E}^2 x(1-x) + m^2 \right)^{d/2-2} \\& = \frac{1}{(4 \pi)^{d/2}} \Gamma(2-d/2) \int\limits_0^1 \mathrm{d}x \; \left( q_\text{E}^2 x(1-x) + m^2 \right)^{d/2-2} \end{aligned} \end{equation}


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.