Trigonometric series with a given spectrum (Q2307482)

This paper is a long essay on \(p\)-coherent sets. If \(\Lambda\in \mathbb{R}^{n}\) is a closed and discrete set, denote by \(T_{\Lambda}\) the vector space of all trigonometric sums whose frequencies belong to \(\Lambda\). Given \(1\le p \le +\infty\), we say that \(\Lambda\) is \(p\)-coherent if there exist a compact set \(K\subset \mathbb{R}^{n}\) and a continuous function \(\omega:\mathbb{R}^{n}\to [1, +\infty)\), such that for each \(P\in T_{\Lambda}\) and each \(y\in \mathbb{R}^{n}\), it holds \(\left (\int_{|x-y|\le 1}|P(x)|^{p}\right )^{1/p}\le \omega(y)\left (\int_{K}|P(x)|^{p}d x\right )^{1/p}\). \par Several properties of \(p\)-coherent sets are proved in this essay. Thus, Section 2 presents the \(L^{2}\) theory and an important result of \textit{S. Jaffard} et al. [J. Fourier Anal. Appl. 3, No. 5, 577--582 (1997; Zbl 0939.42004)] on weighted \(L^{2}\) estimates is proved in Section 3. Examples of sets which are not \(p\)-coherent sets are given in the next five sections. In Section 4 is proved that a \(p\)-coherent set \(\Lambda\) has a finite Beurling and Malliavin density. \par The role of Section 4 is to bridge the gap between the problems raised in Section 1 and growth estimates satisfied by mean periodic functions with a given spectrum. When this growth cannot be controlled by a weight \(\omega\), then we say that \(\Lambda\) is a wild set. In other words, a wild set is a set which is not \(\infty\)-coherent. Theorem 5.4 proves that the digital cone \(\Lambda\in \mathbb{R}^{3}\) is not \(p\)-coherent if \(2 \le p \le \infty\). However the digital cone is \(2\)-coherent. A more involved example is the Pisot set \(\Lambda_{\theta}\),which is \(2\)-coherent. If \(\theta\) is a Pisot number, then the Pisot set is contained in a quasicrystal. Therefore it is \(p\)-coherent for \(1\le p \le +\infty\). When \(\theta\) is not a Pisot number, then the Pisot set is a wild set. Unfortunately we do not know whether the Pisot set is \(p\)-coherent or not when \(\theta\) is not a Pisot number and when \(2 < p < +\infty\). A third example of a wild set is given by Theorem 6.1. This wild set has a finite Beurling and Malliavin density but is not \(p\)-coherent when \(1\le p\le \infty\). For Theorem 6.1 is given another proof in Section 7, where it is also proved in full generality that a \(p\)-coherent set has a finite upper uniform density. Finally, in Section 8 sufficient conditions implying that a set \(\Lambda\) is \(p\)-coherent and relate these \(L^{p}\) estimates to the spectral properties or to the additive properties of \(\Lambda\) are provided.