I am having a problem finding the Fourier Transform of $$f(x) = \frac{\sin(\|x\|)}{\|x\|}$$

I am interested in the cases where $x \in \mathbb{R}^n, n= 2, 3$.

Any help much appreciated !

  • $\begingroup$ Does $n$ refer to the dimension of $x$ or to the norm? $\endgroup$ – Andy Walls Dec 24 '20 at 19:43
  • $\begingroup$ @AndyWalls It refers to the dimension of x, thanks for the observation, I have added it to the question. $\endgroup$ – kot Dec 24 '20 at 20:50
  • $\begingroup$ See the comments here. The function $(1 - \rho^2)^{-1/2} [\rho < 1]$ doesn't need to be regularized. $\endgroup$ – Maxim Dec 24 '20 at 21:06

From Bracewell, the n-dimensional Fourier Transform

$$F(s_1, \dots, s_n) = \int_{-\infty}^{\infty} \dots \int_{-\infty}^{\infty} f(x_1,\dots, x_n) e^{-2\pi i (x_1 s_1 + \dots + x_n s_n)} \space dx_1 \dots dx_n $$

when there is n-dimensional symmetry, can be manipulated into

$$F(q) = \dfrac{2\pi}{q^{\frac{1}{2}n-1}} \int_{0}^{\infty} f(r) J_{\frac{1}{2}n-1}(2\pi qr)\space r^{\frac{1}{2}n} \space dr$$

Where $r$ and $q$ are the radii from the origin in the original domain and transform domain respectively.

For the case of $n=2$, you need to compute

$$F(q) = 2\pi \int_0^\infty \dfrac{\sin(r)}{r}J_0(2\pi qr)\,r\, dr = 2\pi \int_0^\infty \sin(r)J_0(2\pi qr) \,dr$$

For the case of $n=3$ note that

$$J_{\frac{1}{2}}(x) = \sqrt{\dfrac{2}{\pi x}}\sin(x)$$

so you need to compute

$$F(q) = \dfrac{2\pi}{\sqrt{q}} \int_0^\infty \dfrac{\sin(r)}{r}\sqrt{\dfrac{2}{\pi (2\pi qr)}}\sin(2\pi qr)\,r\sqrt{r}\, dr = \dfrac{2}{q}\int_0^\infty \sin(r)\sin(2\pi qr)\, dr$$


It is worth noting that the n-spherically symmetric Fourier transform (aka Hankel Transform) is its own inverse. Also for the $n=3$ case

$$\begin{align*} f(r) &= \dfrac{2\pi}{\sqrt{r}} \int_0^\infty \dfrac{\delta\left(q-\frac{c}{2\pi}\right)}{2q}\sqrt{\dfrac{2}{\pi(2\pi rq)}}\sin(2\pi rq)q\sqrt{q} \, dq\\ \\ &= \dfrac{1}{r}\int_0^\infty \delta\left(q-\frac{c}{2\pi}\right)\sin(2\pi rq) \, dq\\ \\ &= \dfrac{\sin(cr)}{r}\\ \end{align*}$$

  • $\begingroup$ Many thanks Andy ! $\endgroup$ – kot Dec 24 '20 at 21:16
  • $\begingroup$ You're welcome. Also note that the case for $n=3$ looks like it may be spherical delta function, something like $k\dfrac{\delta\left(q-\frac{1}{2\pi}\right)}{q}$, since it looks like the integral may be 0 for $q\ne\frac{1}{2\pi}$. $\endgroup$ – Andy Walls Dec 24 '20 at 22:17
  • $\begingroup$ Can I do the above analysis also for $\frac{sin(c||x||)}{||x||}$, $c$ a real constant ? $\endgroup$ – kot Dec 24 '20 at 22:38
  • $\begingroup$ Sure. See my update for the $n=3$ transform pair $\dfrac{\delta\left(q-\frac{c}{2\pi}\right)}{2q} \leftrightarrow \dfrac{\sin(cr)}{r}$ $\endgroup$ – Andy Walls Dec 24 '20 at 22:47

Formally $$\int_{\mathbb{R}}\int_{\mathbb{R}}\frac{\sin\sqrt{x^2+y^2}}{\sqrt{x^2+y^2}}e^{-isx}e^{-ity}\,dx\,dy =\int_{-\pi}^{\pi}\int_{0}^{+\infty}\sin(\rho)e^{-is\rho\cos\theta}e^{-it\rho\sin\theta}\,d\rho\,d\theta $$ equals $$ \int_{-\pi}^{\pi}\frac{d\theta}{(s\cos\theta+t\sin\theta)^2-1}=\int_{-\pi}^{\pi}\frac{d\phi}{(s^2+t^2)\sin^2\phi-1}=\frac{-2\pi}{\sqrt{1-(s^2+t^2)}} $$ and the Fourier transform of a radial function is a radial function, but there are issues in the application of Fubini's theorem since $\frac{\sin\|x\|}{\|x\|}$ does not belong to $L^1(\mathbb{R}^2)$ or $L^1(\mathbb{R}^3)$.


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.