6
$\begingroup$

I'm trying to evaluate the following integral:

$$\mathcal{J}=\int_{0}^{\infty}\frac{\sin x}{x\left ( 1+x^2 \right )^2}\,{\rm d}x$$

Well there are $3$ poles , one lying on the real line the other on the upper half plane and the other on the lower half plane. The residue at $z=i$ is $\displaystyle -\frac{3}{4e}$ . I'm integrating on a contour that looks like a semicircle on the upper half plane and has a branch about the origin .

Well I considered the function $\displaystyle f(z)=\frac{e^{iz}+1}{z\left ( z+1 \right )^2}$ which is clearly analytic expect for the poles. Hence if $\gamma$ denotes the contour , then:

$$\oint_{\gamma}f(z)\,{\rm d}z=\oint_{\gamma}\frac{e^{iz}+1}{z\left ( z^2+1 \right )^2}\,{\rm d}z =\oint_{\gamma}\frac{e^{iz}}{z\left ( z^2+1 \right )^2}\,{\rm d}z+\oint_{\gamma}\frac{{\rm d}z}{z\left ( z^2+1 \right )^2}=-2\pi i \frac{3}{4e}+2\pi i = 2\pi i \left ( 1-\frac{3}{4e} \right )=i \left (2\pi - \frac{3\pi}{2e} \right )$$

Hmm... applying the classical method I get that:

$$\int_{-\infty}^{\infty}f(x)\,{\rm d}x =2\pi - \frac{3\pi}{2e}$$

which is almost correct except for that $2$ in front of $\pi$. Where I have gone wrong?

P.S: I used the very obvious that the integrand is even.

$\endgroup$
  • $\begingroup$ Is the denominator $x(1+x)^2$ or $x(1+x^2)$? $\endgroup$ – Empy2 Jan 31 '15 at 16:03
  • $\begingroup$ Oups.. Typo.. It is $\displaystyle \frac{\sin x}{x\left ( 1+x^2 \right )^2}$. Mea culpa. $\endgroup$ – Tolaso Jan 31 '15 at 16:05
4
$\begingroup$

Depict carefully the path of integration: it is a semicircle in the upper half plane with a bulge at $z=0$ and a keyhole around $z=i$. This gives that you have to compute the residues of $f(z)=\frac{e^{iz}}{z(z^2+1)^2}$ at $z=0$ and $z=i$, but to consider only half the residue at $z=0$:

$$\mathcal{J}=\frac{1}{2}\text{Im}\int_{-\infty}^{+\infty}\frac{e^{iz}}{z(z^2+1)^2}\,dz = \frac{1}{2}\text{Im}\left(2\pi i\operatorname{Res}(f(z),z=i)+\pi i\operatorname{Res}(f(z),z=0)\right)$$ so: $$\mathcal{J} = \frac{1}{2}\text{Im}\left(2\pi i\cdot \frac{-3}{4e}+\pi i\right)=\frac{1}{2}\left(\pi-\frac{3\pi}{2e}\right)=\color{red}{\frac{\pi}{2}\left(1-\frac{3}{2e}\right)}.$$

$\endgroup$
  • $\begingroup$ @Tolaso: fixed. Luckily the proof is almost the same, just the residue in $z=i$ is different. $\endgroup$ – Jack D'Aurizio Jan 31 '15 at 16:08
  • $\begingroup$ Unfortunately, I'm a litte lost. Can you explain me how you computed the residue at $z=0$ and $z=1$ ? I cannot seem to handle the residues when they come with branch... If I could see the details.. that would be fine... Also why did you take the half of the residue at $z=0$? $\endgroup$ – Tolaso Jan 31 '15 at 16:10
  • $\begingroup$ Branch? What branch? $z=0$ and $z=i$ are a simple pole and a double pole for $\frac{e^{iz}}{z(z^2+1)^2}$. $\endgroup$ – Jack D'Aurizio Jan 31 '15 at 16:16
  • 2
    $\begingroup$ Can't you accept that W|A could be wrong and Jack right ? I got the same result as Jack's numerically. $\endgroup$ – Claude Leibovici Jan 31 '15 at 16:57
  • 1
    $\begingroup$ Actually I was wrong.. Jack was right all along.. I ran the calculations now...and i got my mistake. Thanks Jack.. ! Got it. $\endgroup$ – Tolaso Jan 31 '15 at 18:08
3
$\begingroup$

Another way to evaluate this integral is to use Parseval's theorem, which states that for functions $f$ and $g$ with respective Fourier transforms $F$ and $G$, we have

$$\int_{-\infty}^{\infty} dx \, f(x) g^*(x) = \frac1{2 \pi} \int_{-\infty}^{\infty} dk \, F(k) G^*(k) $$

Here $f(x) = \sin{x}/x$ and $g(x) = (1+x^2)^{-2}$. Thus, $F(k) = \pi$ when $k \in [-1.1]$ and $0$ otherwise and $G(k) = (\pi/2) (|k|+1) e^{-|k|}$. The integral is then

$$\frac{\pi}{8} \int_{-1}^1 dk \, (|k|+1) e^{-|k|} = \frac{\pi}{4} \int_0^1 dk (k+1) e^{-k} $$

$$\int_0^1 dk \, e^{-k} = 1-e^{-1}$$ $$\int_0^1 dk \,k \, e^{-k} = -e^{-1}+ \int_0^1 dk \, e^{-k} = 1-2 e^{-1}$$

Therefore, the integral is equal to

$$\frac{\pi}{4} \left (2-\frac{3}{e} \right ) = \frac{\pi}{2} \left (1-\frac{3}{2 e} \right )$$

$\endgroup$
  • $\begingroup$ Very nice.. I realy like the Fourier approach. $\endgroup$ – Tolaso Jan 31 '15 at 18:09
3
$\begingroup$

Another approach: Parameterize the integral as $$I(a)=\int_{0}^{\infty}\frac{\sin ax}{x\left ( 1+x^2 \right )^2}\,{\rm d}x$$ Take the Laplace transform, find your partial fractions, and take the inverse transform: $$\begin{align*}\mathcal{L}_s\{I(a)\}&=\int_0^\infty\int_{0}^{\infty}\frac{\sin ax}{x\left ( 1+x^2 \right )^2}e^{-as}\,{\rm d}a\,{\rm d}x\\ &=\int_0^\infty\frac{\mathcal{L}_s\{\sin ax\}}{x\left ( 1+x^2 \right )^2}\,{\rm d}x\\ &=\int_0^\infty\frac{x}{x\left ( 1+x^2 \right )^2\left(s^2+x^2\right)}\,{\rm d}x\\ &=\int_0^\infty\frac{{\rm d}x}{\left ( 1+x^2 \right )^2\left(s^2+x^2\right)}\\ &=-\frac{1}{(s^2-1)^2}\int_0^\infty\left(\frac{1}{1+x^2}-\frac{s^2-1}{(1+x^2)^2}-\frac{1}{s^2+x^2}\right)\,{\rm d}x\\ &=-\frac{1}{(s^2-1)^2}\left(\frac{\pi}{2}-\frac{\pi(s^2-1)}{4}-\frac{\pi}{2s}\right)\\ &=\frac{\pi}{4}\left(\frac{2}{s}-\frac{2}{s+1}-\frac{1}{(s+1)^2}\right)\\ I(a)&=\frac{\pi}{4}{\mathcal{L}^{-1}}_a\left\{\frac{2}{s}-\frac{2}{s+1}-\frac{1}{(s+1)^2}\right\}\\ &=\frac{\pi}{4}(2+-2e^{-a}-ae^{-a})\end{align*}$$ Finally, since $\mathcal{J}=I(1)$, you have $$\mathcal{J}=\frac{\pi}{2}-\frac{3\pi}{4e}$$

$\endgroup$
  • $\begingroup$ Very nich approach using Laplace Transformation. $\endgroup$ – Tolaso Feb 1 '15 at 18:13

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.