I recently needed this simple fact of harmonic analysis:

Let $G$ be a discrete abelian group and $X\subset G$ be a finite subset. Then for any $\epsilon>0$ the set $\Gamma_\epsilon=\{ \gamma\in\widehat{G} : \lvert \widehat{X}(\gamma)\rvert \geq (1-\epsilon)\lvert X\rvert\}$ is a closed subset of $\widehat{G}$ with strictly positive Haar measure; $\mu(\Gamma_\epsilon)>0$.

The point is the positivity of the Haar measure; this fails for $\epsilon=0$ or $X$ infinite, for example.

The proof is simple: observe that this set contains $\Gamma_\epsilon'-\Gamma_\epsilon'$ if $\Gamma_\epsilon'$ has the shape $$ \{ \gamma : \lvert \theta_x - \gamma(x) \rvert \leq \epsilon/2\textrm{ for all }x\in X\}$$ for any $\theta_x$, and then note that $\widehat{G}$ is covered by at most $(2\epsilon^{-1})^{\lvert X\rvert}$ many such sets.

I was surprised that I couldn't find a reference or more discussion in Rudin's `Fourier analysis on groups', and Google didn't turn up anything either (at least, nothing I could recognise as this fact). It feels like this fact is a simple corollary of some deeper fact of harmonic analysis.

Hence my questions are:

What is a reference for this fact? What is the correct generalisation to general locally compact abelian groups?

An even vaguer question: what is the iceberg of which this is a tip?

  • 2
    $\begingroup$ Does $\widehat{X}$ for a subset $X$ mean the Fourier transform of the characteristic function of that subset? If so, the inequality defining your $\Gamma_\varepsilon$ appears in the proof of Corollary 4 of math.uconn.edu/~kconrad/blurbs/gradnumthy/loccptascoli.pdf with $G$ being any locally compact abelian group and $X$ being any compact subset of $G$ (denoted as $K$ rather than $X$). Maybe Theorem 2 or Corollary 4 there is the iceberg you're looking for. The positivity of Haar measure of a measurable subset of a locally compact abelian group is true if it contains an open set. $\endgroup$ – KConrad Jun 4 '14 at 12:17
  • $\begingroup$ Yes, it does. Thanks! That's the kind of statement I was looking for. $\endgroup$ – Thomas Bloom Jun 4 '14 at 12:30

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Browse other questions tagged or ask your own question.