What is the Fourier transform of $f(x)e^{2\pi ix^2}$ with $f\in L^2({\mathbb{R}})$?


Suppose $f\in L^2$ and let $g(x):=e^{2\pi ix^2}$. Clearly, $fg$ is in $L^2(\mathbb R)$, so its Fourier transform is well defined and in $L^2(\mathbb R)$, but can it be written in terms of the Fourier transform of $f$?

  • 5
    $\begingroup$ This post is being discussed on the meta. $\endgroup$ – Stella Biderman Dec 8 '17 at 17:37
  • $\begingroup$ @user243301 What is your point exactly? What don't you like about the question? $\endgroup$ – amsmath Dec 11 '17 at 19:53
  • $\begingroup$ @user243301 You wrote "This is a ephemeral message, that I am going to delete in next minute. With my last message I have concluded all that I want to say, good luck." Seems a little childish to me, sorry. $\endgroup$ – amsmath Dec 11 '17 at 20:04

The Gaussian is its own Fourier transform $$ \int_{-\infty}^\infty e^{-\pi x^2} e^{-2i \pi \xi x}dx = e^{-\pi \xi^2}$$

(complete the square assuming $2i \pi \xi \in \mathbb{R}$ and extend the obtained result to $2i \pi \xi \in \mathbb{C}$ by analytic continuation)

Thus for $a > 0$

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

Again we can use analytic continuation to extend the result to $a \in \mathbb{C}, \Re(a) > 0$.

Therefore $$\mathcal{F}[f e^{- \pi a x^2}](\xi)=\widehat{f} \ast \frac{e^{-\pi \xi^2/a}}{a^{1/2}} $$ And by continuity in the $L^2$ sense $$\mathcal{F}[f e^{- \pi (-2i) x^2}]=\lim_{b \to 0^+}\mathcal{F}[f e^{- \pi (b-2i) x^2}] =\lim_{b \to 0^+} \widehat{f} \ast \frac{e^{-\pi \xi^2/(b-2i)}}{(b-2i)^{1/2}} $$ If $\widehat{f} \in L^1$ then the limits goes out and this is $$\mathcal{F}[f e^{- \pi (-2i) x^2}](\xi)=\widehat{f} \ast \frac{e^{-\pi \xi^2/(-2i)}}{(-2i)^{1/2}} $$

  • $\begingroup$ Dear reuns, thanks, very interesting. This confuses me a little. From the last term I can only see that it is in $L^\infty$, right? But actually, it should be in $L^2$. How can I see it from there? And what if $\hat f\notin L^1$? $\endgroup$ – amsmath Dec 9 '17 at 22:01
  • $\begingroup$ @amsmath If $\hat{f} \in L^2$ but $\not \in L^1$ then $\lim_{b \to 0^+} \widehat{f} \ast \frac{e^{-\pi \xi^2/(b-2i)}}{(b-2i)^{1/2}}$ converges in $L^2$ but not pointwise. To convince yourself it has good reasons to be $L^2$ beside Parseval's theorem, you can try with ... $\hat{f}(x) = e^{-c (x-d)^2}$, of course (and use that the linear combinations of Gaussians are dense in $L^2$). $\endgroup$ – reuns Dec 9 '17 at 22:10
  • $\begingroup$ So, if $f\notin L^1$, I can only write the Fourier transform as an $L^2$-limit and not explicitly. Thank you. $\endgroup$ – amsmath Dec 11 '17 at 1:24

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.