I want to compute $$\int_{0}^{+\infty} \frac{\sin x}{\sqrt{x}}\ dx$$ using Gamma function.

I know that by change of variable, $y=\sqrt{x}$, one gets $$\int_{0}^{+\infty} \frac{\sin x}{\sqrt{x}}\ dx=2\int_{0}^{+\infty}\sin y^2\ dy=\frac{\sqrt{2\pi}}{2}$$ by Fresnel's integral.

I try it by considering this: $$\int_{0}^{+\infty}x^{-\frac{1}{2}}e^{ix}\ dx$$ It converges for both real and imaginary part using Dirichlet test, and $0$ is not a problem here. Let the square root take the pricipal branch where $\sqrt{1}=1$. Let $y=-ix$, then $$\int_{0}^{+\infty}x^{-\frac{1}{2}}e^{ix}\ dx=\sqrt{i}\int_{0}^{-i\infty}y^{-\frac{1}{2}}e^{-y}\ dy=(\frac{\sqrt{2}}{2} + \frac{\sqrt{2}}{2}i)\Gamma(\frac{1}{2})=\frac{\sqrt{2\pi}}{2} + \frac{\sqrt{2\pi}}{2}i$$ And it coincides with the final answer!

My problem is, suppose $L$ is a ray starting from $0$ and has an angle $\phi$ with the $x$-axis, and let $\phi\in(0,2\pi)$. I want to argue that (maybe it is incorrect though) $$\Gamma(z)=\int_{L}t^{z-1}e^{-t}\ dt$$

I firstly know that it converges when $\Re(z)>0$. Choose a contour like sector, let $L_1=\{z=x+iy:y=0, r<x<R\}$, $L_2=\{z=Re^{i\theta}:0<\theta<\phi\}$, $L_3=\{z=xe^{i\phi}:r<x<R\}$ and $L_4=\{z=re^{i\theta}:0<\theta<\phi\}$, where $r<R$ and the contour is counterclockwise. By Cauchy theorem we have the contour integral should be $0$. Easy to see that (let $z=x+iy$) $$ \lim_{r\to0+, R\to+\infty}\int_{L_1}t^{z-1}e^{-t}\ dt=\Gamma(z) $$ $$ |\int_{L_4}t^{z-1}e^{-t}\ dt|=|\int_{\phi}^{0} e^{-re^{i\theta}} (re^{i\theta})^{z-1}ire^{i\theta}\ d\theta| \leq \int_{0}^{\phi} e^{-r\cos\theta} |r^{z}e^{i\theta(z-1)}|\ d\theta=\int_{0}^{\phi} e^{-r\cos\theta} r^{x}e^{-\theta y}\ d\theta \to 0, r \to 0+ $$ But when considering $L_2$: $$ |\int_{L_2}t^{z-1}e^{-t}\ dt|\leq\int_{0}^{\phi} e^{-R\cos\theta} R^{x}e^{-\theta y}\ d\theta $$ when for example, $\frac{3\pi}{2}>\phi>\frac{\pi}{2}$, we have $\cos\theta<0$ and I failed to prove the above integral goes to zero when $R\to +\infty$.

Is my usage of Gamma function to compute the original integral a coincidence to get the correct result, or there is a way to prove my argument?

Thank you so much!

  • 1
    $\begingroup$ $\int_0^\infty \sin y^2dy=\int_0^\infty\sin(y^2)dy\ne\int_0^\infty\sin^2ydy$. $\endgroup$ – Kemono Chen Jan 1 at 8:07

There's a bit of a problem because square roots are ambiguous. While $\int_0^\infty x^{-1/2}\exp -zxdx=\sqrt{\pi}z^{-1/2}$ is easily proven on $\Bbb R^+$, if you try an analytic continuation (whether your proof approach is as detailed as you've attempted or otherwise) the intended theorem isn't even clear. Which kind of $z^{-1/2}$ should be used for the desired case $z=-i$?

Funnily enough, there's a completely different approach using the gamma function: $$\int_0^\infty\frac{\sin x dx}{\sqrt{x}}=\frac{1}{\sqrt{\pi}}\Im\int_0^\infty dx\int_0^\infty y^{-1/2}\exp -x(y-i)dy=\frac{1}{\sqrt{\pi}}\Im\int_0^\infty dy\frac{y^{-1/2}}{y-i}\\=\frac{1}{\sqrt{\pi}}\int_0^\infty\frac{y^{-1/2}dy}{y^2+1}=\frac{1}{\sqrt{\pi}}\int_0^{\pi/2}\tan^{-1/2}\theta d\theta\\=\frac{1}{2\sqrt{\pi}}\operatorname{B}\left(\frac{1}{4},\,\frac{3}{4}\right)=\frac{1}{2\sqrt{\pi}}\Gamma\left(\frac{1}{4}\right)\Gamma\left(\frac{3}{4}\right)=\frac{\sqrt{\pi}}{2}\csc\frac{\pi}{4}=\sqrt{\frac{\pi}{2}}.$$


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.