V. A. Steklov's problem of estimating the growth of orthogonal polynomials (Q280999)

From MaRDI portal
scientific article
Language Label Description Also known as
English
V. A. Steklov's problem of estimating the growth of orthogonal polynomials
scientific article

    Statements

    V. A. Steklov's problem of estimating the growth of orthogonal polynomials (English)
    0 references
    0 references
    0 references
    0 references
    10 May 2016
    0 references
    In 1921, Steklov conjectured that the sequence \(\{P_n(x)\}\) of orthonormal polynomials on \((-1,1)\) with respect to a strictly positive weight \(\rho(x)\geq\delta>0\) is bounded at the points \(x\in(-1,1)\), i.e. \(\limsup\limits_{n\to\infty}|P_n(x)|<\infty\). Let \(\{\psi_n\}\) be a sequence of polynomials in \(z=e^{i\theta}\) that are orthonormal on the unit circle with respect to a measure \(\sigma\in S_\delta\) and \(S_\delta\) is the Steklov class of probability measures \(\sigma\) on the unit circle such that \(\sigma'\geq \delta(\theta/(2\pi))>0\) at every Lebesgue point of \(\sigma\). Steklov's conjecture for \(\{\psi_n\}\) was disproved in 1979 by \textit{E. A. Rakhmanov}: \(\limsup\limits_{n\to\infty}\|\psi_n(z;\sigma)\|_{L^\infty(\mathbb T)}=\infty\) (see [Math. USSR, Sb. 36, 549--575 (1980); translation from Mat. Sb., Nov. Ser. 108(150), 581--608 (1979; Zbl 0452.33012)). The following extremal problem played an important role in Rakhmanov's construction: for a fixed \(n\in\mathbb N\), find \(M_{n,\delta}=\sup_{\sigma\in S_\delta}\|\psi_n\|_{L^\infty(\mathbb T)}\). There is an elementary estimate \(M_{n,\delta}\leq\sqrt{(n+1)/\delta}\), \(n\in\mathbb N\). In 1981, \textit{E. A. Rakhamanov} [Mat. Sb., Nov. Ser. 114(156), 269--298 (1981; Zbl 0458.42015)] proved that \(M_{n,\delta}\geq C\sqrt{(n+1)/(\delta\ln^3n)}\), \(C>0\), \(\delta\ll 1\). The main result of this paper is that \(M_n\lesssim\sqrt{n}\), i.e., the elementary estimate mentioned above is sharp. Theorem 1.1: Let \(\delta\in(0,1)\). Then, {\parindent=7mm \begin{itemize}\item[(1)] for sufficiently large \(n>n_0>0\), there exists a constant \(C(\delta)>0\) such that \(M_{n,\delta}>C(\delta)\sqrt{n}\); \item[(2)] for an arbitrary sequence \(\{\beta_n\}\), \(\beta_n\to0\), there exists an absolutely continuous probability measure \(\sigma^*:d\sigma^*={\sigma^*}'d\theta\), \(\sigma^*\in S_\delta\), such that \[ \|\psi_{k_n}(z;\sigma^*)\|_{L^\infty(\mathbb T)}>\beta_{k_n}(\delta)\sqrt{k_n},\quad \beta_{k_n}(\delta)>0, \] for some sequence \(\{k_n\}\subset\mathbb N\). \end{itemize}} The paper gives also a survey of the results on the solution of the extremal problem described above and on the general problem of Steklov in the theory of orthogonal polynomials. The paper analyzes the asymptotic of some trigonometric polynomials defined by the Fejér convolutions.
    0 references
    0 references
    orthogonal polynomials
    0 references
    weight function
    0 references
    Steklov conjecture
    0 references
    extremal measure
    0 references
    asymptotic
    0 references
    trigonometric polynomials
    0 references