Let $\hat f$ denote the Fourier transform of $f$ (non-unitary, angular frequency). I'm interested in a class of $\hat f$ that is characterized by the differential equation $$\hat f ' (t) = - i \alpha e^{-\beta t^2} \hat f(t) , \quad \alpha,\beta >0.$$

In particular, I am interested in finding the class of $\hat f$ that satisfies the differential equation and further satisfies the following properties on $f$:

  • $f(x)\in (0,1)$ for all $x \in \mathbb{R}$
  • $f'(x)<0$ for all $x \in \mathbb{R}$
  • $\lim_{x\to-\infty} f(x) = 1$ and $\lim_{x\to\infty} f(x) = 0$

Unfortunately, I don't know how (or if) I can translate those properties to $\hat f$. For instance, from the properties I can rule out the trivial solution to the differential equation $\hat f = 0$. But I don't know how to translate them in a more structured way.

More generally, the question is, how can I use the properties to refine the possible solutions to the differential equation? E.g., can I pin down a (possibly unique?) starting value for $\hat f (t)$ for some $t$?

Any help is much appreciated!

A first attempt

Suppose $\hat f(t) = c\cdot e^{h(t)}$ and substitute into the differential equation to conclude that $$ h(t) = \int -i \alpha e^{-\beta t^2} d t . $$ For any solution $c$, we can now compute the inverse Fourier transform, $$ f(x) = \frac{c}{2 \pi} g(x)$$ with $$g(x) = \int_{-\infty}^\infty e^{h(t) + i t x} d t, $$ and check whether it satisfies the required properties on $f$.

For now I haven't made much further progress. But it seems to me that, if any solution satisfies the properties, it has to be exactly one. Specifically, as the limit for $x\to -\infty$ requires $c = 2\pi / g(-\infty)$, we either have that for such $c$ all the other properties hold or not.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.