3
$\begingroup$

I am trying to evaluate the integral $\int_0^\infty \frac {x^{1/2} dx}{x^2 + 1}$ using method of residues. I can solve this very easily without the $x^{1/2}$ on top, but I do not know what to do when the function on top is not a polynomial where the powers are integers. How do you compute this?

$\endgroup$
  • 1
    $\begingroup$ keyhole contour $\endgroup$ – tired Nov 22 '15 at 23:58
  • $\begingroup$ Use a keyhole contour as suggested and $$\sqrt{z} = \exp(1/2\log(z))$$ with the branch cut of the logarithm on the positive real axis.Compute the residues at the poles at $z=\pm i$ and observe that you get a multiple of the target integral just below the real axis. $\endgroup$ – Marko Riedel Nov 23 '15 at 1:35
  • 1
    $\begingroup$ Or substitute $x = t^2$ and notice that the resulting integrand is even. Exploiting this symmetry, we can write $$ \int_{0}^{\infty} \frac{x^{1/2}}{1+x^2} \, dx = \int_{-\infty}^{\infty} \frac{t^2}{t^4 + 1} \, dt. $$ Now you can use the method of residue by considering upper-semicircular contour. Geometrically, we are kind of folding the keyhole contour (for the original integral). $\endgroup$ – Sangchul Lee Mar 27 '16 at 7:37
2
$\begingroup$

Set $x=e^t$, we have $$ \int_0^\infty \frac {x^{1/2}}{x^2 + 1}dx=\int_{-\infty}^\infty \frac {e^{3/2t} }{e^{2t} + 1}dt=\int_{-\infty}^\infty \frac {e^{3/2x} }{e^{2x} + 1}dx $$ We consider the analytic function $$ F(z)=\frac {e^{3/2z} }{e^{2z} + 1} $$ on the rectangle contour of $C\cup C'\cup \gamma_1\cup \gamma_2$, where $C$ is $x$-axis from $-R$ to $R$, $C'$ is horizontal line from $R+\pi i$ to $-R+\pi i$, $\gamma_1$ is vertical line from $R$ to $R+\pi i$ and $\gamma_2$ is vertical line from $-R+\pi i$ to $-R$. By Cauchy's residue theorem, we have $$ \int_{-R}^RF(x)dx+\int_{C'}F(z)dz+\int_{\gamma_1}F(z)dz+\int_{\gamma_2}F(z)dz=2\pi iRes\left(F,\frac{\pi i}{2}\right)\tag1 $$ Note that only pole in the contour is $z=\frac{\pi i}{2}$. Since $$ Res\left(F,\frac{\pi i}{2}\right)=\lim_{z\to \frac{\pi i}{2}}(z-\frac{\pi i}{2})F(z)=\lim_{z\to \frac{\pi i}{2}}\frac{e^{3/2z}}{(e^{2z}+1)'}=-\frac1{2}e^{\frac{3\pi i}{4}}\tag2 $$ We have $$ \int_{C'}F(z)dz=\int_R^{-R}\frac{e^{3/2(x+\pi i)}}{e^{2(x+\pi i)}+1}\:dx=-e^{3/2\pi i}\int_{-R}^{R}\frac{e^{3/2x}}{e^{2x}+1}\:dx $$ Along with $$ \left|\int_{\gamma_1}F(z)dz\right|=\left|\int_{0}^{\pi}\frac{e^{3/2(R+iy )}}{e^{2(R+iy)}+1}idy\right|\leqslant \int_{0}^{\pi}\frac{|e^{3/2(R+iy )}|}{|e^{2(R+iy)}|-1}dy\leqslant \frac{\pi e^{3/2 R}}{e^{2R}-1}\to0 $$ as $R\to\infty$. And $$ \left|\int_{\gamma_2}F(z)dz\right|=\left|-\int_{0}^{\pi}\frac{e^{3/2(-R+iy )}}{e^{2(-R+iy)}+1}idy\right|\leqslant \int_{0}^{\pi}\frac{|e^{3/2(-R+iy )}|}{|e^{2(-R+iy)}|-1}dy\leqslant \frac{\pi e^{-3/2 R}}{1-e^{-2R}}\to0 $$ as $R\to\infty$. So by $(1),(2)$ $$ \lim_{R\to\infty}(1-e^{3/2\pi i})\int_{-R}^{R}\frac{e^{3/2x}}{e^{2x}+1}\:dx=-\pi ie^{\frac{3\pi i}{4}} $$ i.e. $$ \int_{-\infty}^{\infty}\frac{e^{3/2x}}{e^{2x}+1}\:dx=-\frac{\pi ie^{\frac{3\pi i}{4}}}{1-e^{3/2\pi i}}=\frac{\pi i}{e^{\frac{3\pi i}{4}}-e^{-\frac{3\pi i}{4}}}=\frac{\pi}{2\sin {\frac{3\pi }{4}}}=\frac{\pi}{\sqrt{2}} $$

$\endgroup$
0
$\begingroup$

Let

$$ I= \int_0^\infty \frac {x^{1/2}\, \mathrm{d}x}{x^2 + 1}. $$

By the Schwinger parametrization we have

$$ I=\int_0^\infty \mathrm{d}t\, \exp(-t)\int_0^\infty \mathrm{d}x\, \sqrt{x}\, \exp\left(-tx^2\right).$$

The last integral can be calculated along the lines of this answer. Using this result, one gets

$$I=\int_0^\infty \mathrm{d}t\,\frac{ \Gamma \left(\frac{3}{4}\right)\exp(-t)}{2 t^{3/4}}=\frac{1}{2}\Gamma \left(\frac{3}{4}\right)\Gamma \left(\frac{1}{4}\right)=\frac{\pi}{\sqrt{2}}$$

$\endgroup$

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.