Exponential sums with coefficients 0 or 1 and concentrated \(L^p\) norms (Q2461171)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Exponential sums with coefficients 0 or 1 and concentrated \(L^p\) norms
scientific article

    Statements

    Exponential sums with coefficients 0 or 1 and concentrated \(L^p\) norms (English)
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    27 November 2007
    0 references
    Let \(e(x):=\exp(2\pi i x)\). A sum of exponentials of the form \(\displaystyle f(x)=\sum_{k=1}^{m} e(N_{k}\,x),\) \((x\in \mathbb{R})\), where \(N_{k}\) are distinct integers is called an idempotent trigonometric polynomial (because the convolution of \(f\) with itself is \(f\)) or, simply an idempotent. Let \(\mathcal{P}\) be the set of all such idempotents. For each \(p\in [1,\,\infty[\), define \(C_{p}\) as the largest number such that for every set \(E\), \(E\subset \mathbb{T}\) with \(| E| >0\), the inequality \[ \sup_{f \in \mathcal{P}}\left(\int_{E} | f| ^{p}dx \Big/\int_{\mathbb{T}} | f| ^{p}dx \right)^{1/p} \geqslant C_{p} \] holds. \(| E| \) denotes the Lebesgue measure of \(E\) and the definition of \(C_{p}\) is extended to the limit case \(p=\infty\) in the usual way. In this paper the authors prove a theorem that provides estimates for the constant \(C_{p}\), when \(1<p<\infty\). More specifically, the authors define the constants \(c_{p}\) and \(c^{*}_{p}\) as follows \[ c_{p}:=\sup_{0<\omega<1/2}\,\frac{\sin(\pi \omega )/(\pi \omega)}{2^{1+1/p}\, \left(\lfloor 1/\omega \rfloor+1 +\frac{1}{p-1}\,(\frac{3}{8})^{p}\,\lfloor1/\omega\rfloor\right)^{1/p}}, \] where \(\lfloor r \rfloor\) denotes the largest integer less than or equal to the real number \(r\); and \[ c_{p}^{*}:= \Big( \frac{2}{\pi^{p+1}}\int_{0}^{\infty}\Big| \frac{\sin x}{x}\Big| ^{p}\,dx\Big)^{1/p}\,.\,\max_{0 \leqslant \omega \leqslant 1}\, \frac{\sin(\pi \omega)}{\omega^{1-1/p}}\,. \] Then the authors prove the following result: Whenever \(1<p<\infty\), we have the estimate \[ C_{p} \geqslant \begin{cases} c_{p} & {\text{if }} 1<p\leqslant 2\\ c_{p}^{*} & {\text{if }} 2 \leqslant p<\infty \end{cases}. \] In other words, if \(p>1\) and \(\varepsilon >0\) are given, then for each set \(E\subset \mathbb{T}\) with \(| E| >0\), there is a finite set of integers \(S=S(E, p, \varepsilon)\) such that \[ \int_{E}\Big| \sum_{n\in S} e(nx)\Big | ^{p}\,dx \Big/\int_{\mathbb{T}}\Big| \sum_{n\in S} e(nx)\Big | ^{p}\,dx \geqslant \begin{cases} c_{p}^{p}-\varepsilon & {\text{if }} 1<p\leqslant 2\\ c_{p}^{*p}-\varepsilon & {\text{if }} 2 \leqslant p<\infty\end{cases} . \] Furthermore, \(c_{p}^{*}\) (and \textit{a fortiori} \(C_{p}\)) tends to \(1\) as \(p\) tends to infinity. Several conjectures and open problems are also stated.
    0 references
    idempotent trigonometric polynomials
    0 references
    \(L^{p}\) norms
    0 references
    Dirichlet kernel
    0 references
    concentrating norms
    0 references
    sums of exponentials
    0 references
    \(L^{1}\) concentration conjecture
    0 references
    weak restricted operators
    0 references

    Identifiers

    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references