Bounded exponential sums (Q913846): Difference between revisions

Answering a question of Kamae, the authors give a proof that there exist no sequence \(a_ k\in \{-1,1\},\) such that \[ | \sum^{m+n}_{k=m}a_ ke^{-ik\theta}| \leq c(\theta)<\infty \text{ for all } m,n\in {\mathbb{Z}},\quad n\geq 0,\text{ and } all\quad \theta \in [-\pi,\pi]. \] In general they prove that such sequences \(a=(a_ k)\) are ``essentially'' zero. Considering the closure X of the orbit of a in \(I^{{\mathbb{Z}}}\) (I a compact interval) under the shift transformation, and the according operator T on \(L^ 2(X,\mu)\) (\(\mu\) an invariant measure) the problem is reduced to the following Theorem: Let T be a unitary operator on a Hilbert space H and \(f\in H\) such that \(\sup_{n}\| \sum^{n-1}_{k=0}e^{-ik\theta}T^ kf\| <\infty\forall \theta \in [-\pi,\pi]\) then \(f=0\). This follows from the spectral theorem. As consequences (also of the proof) one obtains several corollaries, e.g. the boundedness assumption (\(\leq c(\theta))\) can be weakened to \(\sigma\sqrt{n}\) (but clearly not to O(\(\sqrt{n})\) in general) or results on the space of ``\(\theta\)-twisted coboundaries'' etc.