13
$\begingroup$

We know that the Fourier transform of a Gaussian function is Gaussian function itself. Can anyone give one or more functions which have themselves as Fourier transform?

$\endgroup$
5
  • $\begingroup$ Yes we can. But why do you need? $\endgroup$
    – user17762
    Mar 9, 2012 at 2:42
  • 5
    $\begingroup$ The key here was the phrase "fixed point" or "eigenfunction", then Google performed the rest of the work. mathoverflow.net/questions/12045/… and en.wikipedia.org/wiki/Fourier_transform#Eigenfunctions $\endgroup$
    – dls
    Mar 9, 2012 at 2:49
  • $\begingroup$ Also, have a look at the answers here. $\endgroup$
    – kuch nahi
    Mar 9, 2012 at 2:59
  • 2
    $\begingroup$ In the old days the "fixed-point functions" were called self-reciprocal functions and investigated by the likes of Hardy and Titchmarsh. See "On self-reciprocal functions under a class of integral transforms" by Kurt Wolf, fis.unam.mx/~bwolf/Articles/28.pdf . You might follow the references in this paper, particularly those of Titchmarsh– Tom Copeland 1 min ago $\endgroup$ Apr 7, 2012 at 22:30
  • $\begingroup$ See this MO post. $\endgroup$
    – Nemo
    Nov 24, 2017 at 10:02

2 Answers 2

6
$\begingroup$

Pick a function $f$ that is reasonable enough for the inversion formula to hold (e.g. take $f$ in Schwartz space, which contains the Gaussian among other functions). If $\mathcal{F}$ denotes the linear transformation which takes $f$ to its Fourier transform, then it's easy to check that $\mathcal{F}^{4}$ is the identity map. In particular, by playing some games you find that $$g \ = \ f + \mathcal{F}(f) + \mathcal{F}^{2}(f) + \mathcal{F}^{3}(f) $$ is fixed by $\mathcal{F}$. So $g$ is its own Fourier transform.

This argument doesn't produce a concrete function, but it at least shows you that the Gaussian is far from the only function that is equal to its own Fourier transform. If you want a more specific example, you can show that $(\cosh \pi x)^{-1}$ is its own Fourier transform (use contour integration and the residue theorem).

$\endgroup$
4
$\begingroup$

Well, the Dirac delta function, strictly speaking, is not a 'function' but it serves your purpose.

The Fourier transform of a Dirac comb (or an impulse train, as it is called in Electrical Engineering) is another Dirac comb.

$$ \sum_{n=-\infty}^{\infty} \delta(t-nT_o) = \frac {2\pi} {T_o} \sum_{m=-\infty}^{\infty} \delta(\omega - m\omega_o) $$

For the proof, refer http://nptel.ac.in/courses/IIT-MADRAS/Principles_Of_Communication/pdf/Lecture07_FTPeriodic.pdf

$\endgroup$
2
  • $\begingroup$ Welcome to Maths.SE ! Thanks for your answer. Please note that on this site it is recomended to include excerpts of your work, providing a link is not enough to reach this site's standards. $\endgroup$
    – Tom-Tom
    Sep 29, 2015 at 19:38
  • $\begingroup$ @Tom-Tom I'm a greenhorn here. Thanks for the suggestion. Will keep this in mind for all my future answers :) $\endgroup$ Jul 21, 2016 at 5:42

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.