I played a little with convolutions of the form $$g(\xi) = \int_{-\infty}^\infty f(x) f(\xi - x) \mathrm{d}x,$$ where $f(x)$ should be a rapidly decaying function (not necessarily in a Schwartz sense, it should be enough that $f(x)$ is integrable).

Concretely, I looked at these examples

\begin{align*} f_1(x) &= e^{-|x|} & g_1(\xi) &= (1+|\xi|)e^{-|\xi|} \\ f_2(x) &= e^{-x^2} & g_2(\xi) &= \sqrt{\frac\pi 2}e^{-\xi^2/2} \\ f_3(x) &= \frac{1}{x^2+1} & g_3(\xi) &= \frac\pi 2 \frac{1}{1+(\xi/2)^2} \\ f_3(x) &= \frac{1}{x^4+1} & g_3(\xi) &= \frac{\pi}{8\sqrt{2}} \frac{12+\xi^2}{2+\xi^2} \frac{1}{1+(\xi/2)^4} \\ f_4(x) &= Ai(x) & g_4(\xi) &= Ai(\xi/\sqrt[3]{2}), \end{align*} where $Ai(x)$ is the Airy function.

Noticeably, it seems in every case that $g(\xi) = h(\xi) f(\alpha\xi)$, where $\alpha \in \mathbb{R}$ and $h(\xi)$ is some function. This is a statement, which I very much like to proof, or, at least, give some conditions, for when this holds. However, I wasn't very successful so far.

Obviously, for rational functions like $f_3$ and $f_4$ the convolution theorem provides an easy answer, yet this seems to be true for even a wider class of functions.

Side Note Actually, this topic sounds so simple that I assume that this must have been asked somewhere already, yet I did not manage to find anything on this problem


Obviously you could always simply choose $h(\xi) = g(\xi)/f(\alpha\xi)$, so the question is in its current form not well posed.

Moreover note that for example for $f(x) = |x|e^{-|x|}$ one has $g(\xi) =\tfrac 16 (|\xi|^3 + 3|\xi| + 3)e^{-|\xi|}$, which does not exhibit the pattern you observed without resorting to the 'trivial' choice of $h$ pointed out above.


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.