3
$\begingroup$

According to Mathematica with Fourier transform convention

$$\widehat{f}(\xi)=(2\pi)^{-1/2}\int_{-\infty}^{\infty}f(x)e^{i\pi x}dx$$

The Fourier transform of the function $f(x):=|x|^{-1/2}e^{-|x|}$ is given by

$$\widehat{f}(\xi)=\left(\frac{1}{1+|\xi|^{2}}+\frac{1}{\sqrt{1+|\xi|^{2}}}\right)^{1/2}$$

Mathematica spitted out this result very quickly. This is not a function I recognize as having a well-known Fourier transform, but using Fourier analysis, I believe the Fourier transform should be something like (up to a multiplicative constant) the convolution

$$\int_{\mathbb{R}}|y|^{-1/2}\frac{1}{1+|x-y|^{2}}dy$$

But Mathematica does not give a nice expression for this integral. I imagine one can evaluate this integral using contour integration and Cauchy residue theorem, but does anyone see a simpler argument, say using Fourier transforms of known functions?

$\endgroup$
5
$\begingroup$

It is easier to do the direct computation.

Forgive me for using my own notation for the time being. Let

$$F(k) = \int_{-\infty}^{\infty} dx \, f(x) e^{i k x}$$

where

$$f(x) = (2 \pi |x|)^{-1/2} \, e^{-|x|}$$

Then

$$F(k) = \sqrt{\frac{2}{\pi}} \int_0^{\infty} dx \, x^{-1/2} \, e^{-x} \cos{k x} = \sqrt{\frac{8}{\pi}} \int_0^{\infty} dx \, e^{-x^2} \cos{k x^2}$$

The integral is the real part of a complex Gaussian integral:

$$\begin{align}F(k) &= \sqrt{\frac{2}{\pi}} \operatorname{Re}{\left [\int_{-\infty}^{\infty} dx \, e^{-(1-i k) x^2} \right ]} \\ &= \operatorname{Re}{\left (\sqrt{\frac{2}{1-i k}}\right )}\\ &= \sqrt{\frac{2}{\sqrt{1+k^2}}} \cos{\left (\frac12 \arctan{k} \right )}\\ &= \sqrt{\frac{2}{\sqrt{1+k^2}}} \sqrt{\frac{1+\frac1{\sqrt{1+k^2}}}{2}}\end{align}$$

Thus

$$F(k) = \sqrt{\frac1{1+k^2}+\frac1{\sqrt{1+k^2}}} $$

as was to be shown.

$\endgroup$
  • $\begingroup$ I hope you don't mind I posted an answer that is pretty much the same with added details. If you want to add the details to your answer, I will delete my answer. $\endgroup$ – robjohn Jan 29 '16 at 8:50
  • $\begingroup$ @robjohn: I wouldn't do that. Thanks for explaining what you have done. $\endgroup$ – Ron Gordon Jan 29 '16 at 10:03
2
$\begingroup$

This answer simply adds some details to Ron Gordon's answer.


Use symmetry to get rid of the absolute values. $$ \begin{align} \int_{-\infty}^\infty\left|x\right|^{-1/2}e^{-\left|x\right|}e^{ixy}\,\mathrm{d}x &=2\operatorname{Re}\left(\int_0^\infty x^{-1/2}e^{-x}e^{ixy}\,\mathrm{d}x\right)\\ &=2\operatorname{Re}\left(\int_0^\infty x^{-1/2}e^{-(1-iy)x}\,\mathrm{d}x\right)\\ \end{align} $$ Since $z^{-1/2}e^{-z}$ has no singularities in the right half-plane, we can use Cauchy's Theorem to get $$ \begin{align} &\int_0^Rx^{-1/2}e^{-(1-iy)x}\,\mathrm{d}x\\ &=\frac1{\sqrt{1-iy}}\int_0^{R(1-iy)}x^{-1/2}e^{-x}\,\mathrm{d}x\\ &=\frac1{\sqrt{1-iy}}\int_0^Rx^{-1/2}e^{-x}\,\mathrm{d}x +\frac1{\sqrt{1-iy}}\int_R^{R(1-iy)}x^{-1/2}e^{-x}\,\mathrm{d}x\\ &\stackrel{R\to\infty}{\to}\frac1{\sqrt{1-iy}}\int_0^\infty x^{-1/2}e^{-x}\,\mathrm{d}x\\ &=\sqrt{\frac\pi{1-iy}} \end{align} $$ since $\Gamma\left(\frac12\right)=\sqrt\pi$ and $$ \int_R^{R(1-iy)}x^{-1/2}e^{-x}\,\mathrm{d}x=O\!\left(R^{1/2}e^{-R}\right) $$ Therefore, $$ \begin{align} \frac1{\sqrt{2\pi}}\int_{-\infty}^\infty\left|x\right|^{-1/2}e^{-\left|x\right|}e^{ixy}\,\mathrm{d}x &=\operatorname{Re}\left(\sqrt{\frac2{1-iy}}\right)\\ &=\bbox[5px,border:2px solid #C0A000]{\sqrt{\frac{1+\sqrt{1+y^2}}{1+y^2}}} \end{align} $$ by solving $(a+ib)^2=\dfrac2{1+y^2}(1+iy)$ for $a$.

$\endgroup$

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.