I want to find the integral

$$I=\int_0^\infty \frac{dx}{(x^2+2x+12)^2}$$

using contour integration; I am familiar with the trigonometric substitution in real analysis.

There are no branch cuts, there are two second order poles at $z=-1\pm\sqrt{11}i$. I cannot use $\int_{-\infty}^\infty=2\int_0^\infty$ since the function isn't even.

Clearly, the choice of contour is important here (compare with this question). I tried half-circle and quarter-circle contours, but they just give some weird rationals. Is there some better contour here? Am I missing something?

  • 1
    $\begingroup$ I think you can do this integral by calculating instead $\int_\Gamma \frac{\log z}{(z^2+2x+12)^2}\,dz$ where $\Gamma$ is a curve something like: from origin out to $(R,\epsilon)$, then a circle to $(R,-\epsilon)$ and then back to the origin. Let $R\to+\infty$ and $\epsilon\to 0+$. Also, you use the $\log$ with a branch cut along the positive real axis. The reason this works (if I don't make a mistake) is that the logarithms cancel... $\endgroup$ – mickep Jan 11 '15 at 19:00

Yes. There is a standard trick for integrating functions over $[0,\infty)$, and that is to exploit the multi-valuedness of the logarithm about a branch cut. Thus, consider

$$\oint_C dz \frac{\log{z}}{(z^2+2 z+12)^2} $$

where $C$ is a keyhole contour as pictured below:

enter image description here

The radius of the small circle is $\epsilon$ and that of the bigger circle is $R$. The contour integral over $C$ is then

$$\int_{\epsilon}^R dx \frac{\log{x}}{(x^2+2 x+12)^2} + i R \int_0^{2 \pi} d\theta \, e^{i \theta}\frac{\log{\left ( R e^{i \theta} \right )}}{\left (R^2 e^{i 2 \theta} + 2 R e^{i \theta}+12 \right )^2} \\ + \int_R^{\epsilon} dx \frac{\log{x} + i 2 \pi}{(x^2+2 x+12)^2} + i \epsilon \int_{2 \pi}^0 d\phi \, e^{i \phi} \frac{\log{\left ( \epsilon e^{i \phi} \right )}}{\left (\epsilon^2 e^{i 2 \phi}+ 2 \epsilon e^{i \phi}+12 \right )^2}$$

It should be evident that, as $R \to \infty$, the second integral vanishes as $\log{R}/R^3$, and that as $\epsilon \to 0$, the fourth integral vanishes as $\epsilon \log{\epsilon}$. Thus, in this limit, the contour integral is

$$-i 2 \pi \int_0^{\infty} \frac{dx}{(x^2+2 x+12)^2} $$

By the residue theorem, this contour integral is also equal to $i 2 \pi$ times the sum of the residues at the poles $z_{\pm} = -1 \pm i \sqrt{11}$. Therefore we have

$$\begin{align}\int_0^{\infty} \frac{dx}{(x^2+2 x+12)^2} &= - \left [\frac{d}{dz} \frac{\log{z}}{(z-z_-)^2} \right ]_{z=z_+} - \left [\frac{d}{dz} \frac{\log{z}}{(z-z_+)^2} \right ]_{z=z_-}\\ &= -\frac1{z_+ (z_+-z_-)^2} + \frac{2 \log{z_+}}{ (z_+-z_-)^3}\\ &-\frac1{z_- (z_--z_+)^2} + \frac{2 \log{z_-}}{(z_--z_+)^3} \\ &= -\frac{z_++z_-}{z_- z_+ (z_+-z_-)^2} + 2 \frac{\log{z_+}-\log{z_-}}{(z_+-z_-)^3} \end{align}$$

To get this right, it is crucial that we understand how to compute the logs. Understand that, in defining the contour, we demanded that $\arg{z} \in [0,2 \pi)$. Thus,

$$\log{z_+} = \frac12 \log{12} + i \left (\pi- \arctan{\sqrt{11}}\right )$$ $$\log{z_-} = \frac12 \log{12} +i \left (\pi+ \arctan{\sqrt{11}}\right )$$

Putting this altogether, we find, after some arithmetic, that

$$\int_0^{\infty} \frac{dx}{(x^2+2 x+12)^2} = \frac{1}{22 \sqrt{11}} \arctan{\sqrt{11}} - \frac1{264} $$


Another way of evaluating this integral is to observe that

$$\int_0^{\infty} \frac{dx}{(x^2+2 x+b)^2} = -\frac{\partial}{\partial b} \int_0^{\infty} \frac{dx}{x^2+2 x +b} $$

then consider

$$\oint_C dz \frac{\log{z}}{z^2+2 z+b} $$

The analysis is the same as above; the result is that

$$-\int_0^{\infty} \frac{dx}{x^2+2 x +b} = \frac{\log{z_+}-\log{z_-}}{z_+-z_-} = -\frac{\arctan{\sqrt{b-1}}}{\sqrt{b-1}}$$

Differentiating, we get

$$\int_0^{\infty} \frac{dx}{(x^2+2 x+b)^2} = \frac{\arctan{\sqrt{b-1}}}{2 (b-1)^{3/2}}-\frac{1}{2 b (b-1)}$$

Plugging in $b=12$ produces the above result.

  • $\begingroup$ (+1) I'm now happy I did not write some half-crappy solution to this :) $\endgroup$ – mickep Jan 11 '15 at 19:11
  • $\begingroup$ Wonderful, thank you! $\endgroup$ – Minethlos Jan 11 '15 at 19:39

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.