On a problem by Steklov

Abstract:

Given any $\delta \in (0,1)$, we define the Steklov class $S_\delta$ to be the set of probability measures $\sigma$ on the unit circle $\mathbb {T}$, such that $\sigma ’(\theta )\geqslant \delta /(2\pi )>0$ for Lebesgue almost every $\theta \in [0,2\pi )$. One can define the orthonormal polynomials $\phi _n(z)$ with respect to $\sigma \in S_\delta$. In this paper, we obtain the sharp estimates on the uniform norms $\|\phi _n\|_{L^\infty (\mathbb T)}$ as $n\to \infty$ which settles a question asked by Steklov in 1921. As an important intermediate step, we consider the following variational problem. Fix $n\in \mathbb N$ and define $M_{n,\delta }=\sup \limits _{\sigma \in S_\delta }\|\phi _n\|_{L^\infty (\mathbb T)}$. Then, we prove \[ C(\delta )\sqrt n < M_{n,\delta }\leqslant \sqrt {\frac {n+1}\delta } . \] A new method is developed that can be used to study other important variational problems. For instance, we prove the sharp estimates for the polynomial entropy in the Steklov class.