Simultaneous zero-free approximation and universal optimal polynomial approximants (Q2182913)

Let \(H^2\) be the Hardy space of the unit disk \(\mathbb{D}\) and \(f\in H^2\setminus\{0\}\), then \(Q_n(1/f)\) is an optimal polynomial approximation (o.p.a.) for \(1/f\) if it minimizes \(\|qf-1\|_2\) over all polynomials \(q\) of degree at most \(n\). \(C(E)\) is the space of continuous functions on a closed subset \(E\) of the unit circle \(\mathbb{T}\) of measure zero, and \(\mathcal{U}_E\) the class of functions \(f\) with universal o.p.a.\ on \(E\), meaning that there is a subsequence of o.p.a.'s \((Q_n(1/f))\) that converges to \(g\) for any \(g\in C(E)\). The main theorem of this paper states that \(\mathcal{U}_E\) is \(G_\delta\) dense in \(H^2\). This is a generalization of a theorem proved in [\textit{H.-P. Beise} and \textit{J. Müller}, Math. Z. 284, No. 3--4, 1185--1197 (2016; Zbl 1360.30003)] for truncated Taylor series of \(1/f\) instead of o.p.a.'s. The proof is based on a result stating that for \((f,g)\in H^2\times C(E)\), with \(f(z)\ne0\) in \(\mathbb{D}\), there is a sequence of polynomials \(P_n\) not vanishing in \(\mathbb{D}\cup\mathbb{T}\) that converges simultaneously to \(f\) and to \(g\) in the associated topologies. To prove these, Baire's category theorem is used to obtain an explicit form for \(\mathcal{U}_E\). It is also shown that the truncated Taylor series result for the Dirichlet space of [\textit{J. Müller}, Anal. Math. Phys. 9, No. 2, 875--885 (2019; Zbl 1429.30057)] can be generalized to the o.p.a.\ setting for nowhere vanishing functions.