20

I'm curious as to how the Fourier transform of the various types of Bessel functions would be calculated. The Wikipedia page on the Fourier transform gives the transform of $J_o(x)$ as being $\frac{2rect(\pi\zeta)}{\sqrt{1-4\pi^2\zeta^2}}$. I've searched the web some but I seem to be unable to find a derivation for that value. Since $J_o$ only "damps out" to zero at infinity I'm guessing the transform must be computed using some kind of generalized function representation of the Bessel function. Any references would be much appreciated.

MattyZ
  • 2,313

2 Answers2

26

Although this question was asked and answered quite a while ago, I thought that it might be useful to see an alternative development, one that uses classical analysis only and forgoes the use of Generalized Functions.

Here, we will find the Fourier Transform of $J_0(x)$ by first finding the Fourier Transform representation of $J_0(x)$ and subsequently invoking the Fourier Inversion Theorem.

We begin, as @Sasha began, with the integral representation

$$\begin{align} J_0(x)&=\frac{1}{2\pi}\int_{-\pi}^{\pi}e^{ix\sin \phi}\,d\phi\\\\ &=\frac{1}{\pi}\int_{0}^{\pi}\cos (x\sin \phi)\,d\phi \tag 1 \end{align}$$

where we exploited the fact that real part of the integrand is an even function of $k$ while the imaginary part is odd. Making the substitutions

$$\phi= \begin{cases} \arcsin (k),&\text{for}\,\,0\le\phi\le\pi/2\\\\ \pi-\arcsin (k),&\text{for}\,\,\pi/2\le\phi\le\pi \end{cases}$$

into $(1)$ yields

$$\begin{align} J_0(x)&=\frac{2}{\pi}\int_{0}^{1}\frac{1}{\sqrt{1-k^2}}\cos (kx)\,dk \tag 2\\\\ &=\frac{1}{\pi}\int_{-1}^{1}\frac{1}{\sqrt{1-k^2}}\cos (kx)\,dk \tag 3\\\\ &=\frac{1}{\pi}\int_{-1}^{1}\frac{1}{\sqrt{1-k^2}}e^{ikx}\,dk \tag 4\\\\ &=\frac{1}{2\pi}\int_{-1}^{1}\frac{2}{\sqrt{1-k^2}}e^{ikx}\,dk \tag 5\\\\ &=\frac{1}{2\pi}\int_{-\infty}^{\infty}\left(\text{rect}\left(\frac{k}{2}\right)\frac{2}{\sqrt{1-k^2}}\right)e^{ikx}\,dk \tag 6\\\\ \end{align}$$

where $\text{rect}$ is the Rectangular Function. Finally, using the Fourier Inversion Theorem, we have

$$\bbox[5px,border:2px solid #C0A000]{\mathscr{F}\left(J_0(x)\right)(k)=\text{rect}\left(\frac{k}{2}\right)\frac{2}{\sqrt{1-k^2}}}$$

recovering the well-known result.


NOTES:

In arriving at $(2)$, we wrote $\int_{0}^{\pi}\cos (x\sin(\phi))\,d\phi=\int_{0}^{\pi/2}\cos (x\sin(\phi))\,d\phi+\int_{\pi/2}^{\pi}\cos (x\sin(\phi))\,d\phi$, enforced the substitutions, and combined the resulting integrals.

In going from $(2)$ to $(3)$, we exploited the even property of the cosine.

In going from $(3)$ to $(4)$, we exploited the odd property of the sine.

In going from $(4)$ to $(5)$, we placed a factor of $1/2$ outside the integral and a factor of $2$ in the integrand.

In going from $(5)$ to $(6)$, we multiplied the integrand by the rectangle function and extended the limits.

Mark Viola
  • 179,405
  • Thank you! I receive your notification. I'm not able to use this site as much as I'd like these days unfortunately, but your answer is great. – MattyZ Jul 17 '15 at 01:24
  • 1
    You're welcome! My pleasure. It gives another way to see the transform. – Mark Viola Jul 17 '15 at 01:49
  • I was trying to compute the Fourier transform of the zeroth order Bessel function of second kind and stumbled on this. Your solution is really elegant and adaptable to my problem as well. Thank You! One small question, I don't see how your second substitution works for going from equation (1) to (2). So instead I chose $$\phi= \begin{cases} \arcsin (k),&\text{for},,0\le\phi\le\pi/2\\ \pi/2+\arccos (k),&\text{for},,\pi/2\le\phi\le\pi \end{cases}$$ Great approach nevertheless! – Mandar Kulkarni Jun 30 '17 at 02:11
  • @MandarKulkarni Thank you for the nice note. And you're welcome. Pleased to hear that this was useful. There was a typo in that substitution; I've edited. It should be $$\phi= \begin{cases} \arcsin (k),&\text{for},,0\le\phi\le\pi/2\\ \pi-\arcsin (k),&\text{for},,\pi/2\le\phi\le\pi \end{cases}$$ – Mark Viola Jun 30 '17 at 02:33
  • Yes the edited form works. – Mandar Kulkarni Jun 30 '17 at 02:38
  • What about the Fourier transform of $n^{th}$ order Bessel functions of the first kind? – exp ikx Jun 18 '19 at 13:54
  • Use the recurrence relationships for Bessel functions. – Mark Viola Jun 18 '19 at 15:00
  • Very clever @MarkViola. Your answers always lead me to a new way of thinking about something. – Cameron Williams Oct 09 '19 at 18:21
  • @cameronwilliams Hi Cameron. Much appreciated! Are you still in Houston? I live in The Woodlands. – Mark Viola Oct 10 '19 at 00:23
  • Unfortunately not. I moved a while back. Probably should have updated my bio in this time.. Forcibly left academia, too. – Cameron Williams Oct 10 '19 at 02:54
  • @CameronWilliams "Forcibly?" What do you mean? – Mark Viola Oct 10 '19 at 03:38
  • Department politics left me not getting a tenure track after working for a year as a visiting assistant professor. – Cameron Williams Oct 10 '19 at 07:25
  • @CameronWilliams I'm sorry to hear that. But there are other universities that have tenure track openings. – Mark Viola Oct 10 '19 at 13:21
22

Using integral representation: $$ J_0(x) = \frac{1}{2 \pi} \int_{-\pi}^\pi \mathrm{e}^{i x \sin \tau} \mathrm{d} \tau $$ Thus the Fourier transform: $$ \begin{eqnarray} \mathcal{F}_x(J_0(x))(\omega) &=& \int_{-\infty}^\infty J_0(x) \mathrm{e}^{i \omega x} \mathrm{d} x \\ &=& \frac{1}{2 \pi} \int_{-\pi}^\pi \mathrm{d} \tau \, \mathcal{F}_x(\mathrm{e}^{i x \sin \tau})(\omega) \\ &=& \frac{1}{2\pi} \int_{-\pi}^\pi \mathrm{d} \tau \, \left( 2 \pi \right) \delta\left( \omega + \sin(\tau) \right) \\ &=& \int_{-\pi}^\pi \mathbf{1}_{-1 \le \omega \le 1} \delta\left( \omega + \sin(\tau) \right) \,\, \mathrm{d} \tau \\ &=& \int_{-\pi}^\pi \mathbf{1}_{-1 \le \omega \le 1} \frac{1}{\vert \cos(\tau) \vert} \left( \delta\left( \arcsin \omega + \tau \right) + \delta\left( \arcsin \omega - \operatorname{sign}(\omega) \pi + \tau \right) \right)\,\, \mathrm{d} \tau \\ &=& \mathbf{1}_{-1 \le \omega \le 1} \frac{2}{\sqrt{1-\omega^2}} \end{eqnarray} $$

Sasha
  • 70,631
  • 5
    Hi, I am trying to do this problem, but instead I am trying $$ \int_{-\infty}^\infty J^3_0(x)e^{i\omega x} dx. $$ Any idea how I would go about this? THanks... – Jeff Faraci Mar 12 '14 at 04:45
  • @JeffFaraci From the convolution theorem, we know that the Fourier Transform of a product of functions is equal the convolution of the Fourier Transforms. Apply this first to find the Fourier Transform of $J_0^2$. Then apply it once more to find the Fourier Transform of $J_0^3$. – Mark Viola Jun 27 '23 at 13:13