MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

This question was suggested when trying to find an explicit example of a continuous function with compact support in $\mathbb{R}$ whose Fourier transform is not integrable. The existence of such a function was proved by an abstract argument in this MO question, but no explicit example was given. It is clear that such a function cannot be differentiable everywhere.

The functions $(1-x^2)^{\alpha}\_{+}$ for $0<\alpha<1$ look like reasonable candidates, since they have a singularity at $\xi=\pm1$ and converge as $\alpha\to0$ to the characteristic function of $[-1,1]$, whose Fourier transform is not integrable. However $\hat f_{\alpha}(\xi)=O(|\xi|^{-(1+\alpha)})$ as $|\xi|\to\infty$, and hence is integrable.

For any $\alpha,\beta>0$, the function $$g_{\alpha,\beta}(x)=(1-\beta\log(1-x^2))^{-\alpha}\hbox{ if }|x|<1,\quad 0 \hbox{ if } |x|\ge1$$ is more singular than any $f_{\alpha}$. So my questions are:

  • Is $\hat g_{\alpha,\beta}$ is integrable?
  • What is the asymptotic behaviour of $\hat g_{\alpha,\beta}$ at infinity?

I suspect, based on numerical calculations, that the answer to the first question is no, at least for sufficiently small $\alpha$ and $\beta$.

share|cite|improve this question
up vote 11 down vote accepted

If one performs a smooth dyadic decomposition of $g_{\alpha,\beta}$ around the singularities $x = \pm 1$ (i.e. using smooth partitions of unity to decompose $g_{\alpha,\beta}$ into pieces that are localised in the region $1-|x| \sim 2^{-n}$ for $n \geq 0$), and then takes the Fourier transform of these pieces, one soon arrives at the conclusion that $\hat g_{\alpha,\beta}$ decays like $\frac{1}{|\xi| \log^\alpha |\xi|}$ (times something like $\sin \xi$) as $\xi \to \infty$ (and one can then back up this intuition with stationary phase, or use some localised form of the Plancherel theorem for an $L^2$ averaged result, which should be enough for the application at hand). So absolute integrability should fail for $\alpha \leq 1$.

share|cite|improve this answer
There are many details to fill in, but I see now how to prove the result. – Julián Aguirre Oct 26 '10 at 14:01

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.