Mergelyan's approximation theorem with nonvanishing polynomials and universality of zeta-functions (Q1944315)

The following universality theorem generalizes a result by S. M. Voronin (1975) and was proved by \textit{B. Bagchi} [Math. Z. 181, 319--334 (1982; Zbl 0479.10028)]. Theorem. Let \(K\) be any compact set with connected complement lying entirely within \(1/2<\operatorname{Re}(s)<1\), and suppose that \(f\) is any continuous nonvanishing function on \(K\) that is analytic in the interior of \(K\). Then, for any \(\varepsilon>0\), the following condition holds true \[ \liminf_{T\to \infty}\frac{1}{T}\operatorname{meas }\{t\in [0,T]: \max_{z\in K}|\zeta(z+it)-f(z)|<\varepsilon\}>0, \] where \(\zeta\) stands for the Riemann zeta-function. The above result is technically related to the classical theorem of Mergelyan which guarantees the polynomial approximation to analytic functions. The present paper considers the problem of removing the condition that \(f(z)\) is nonvanishing on the boundary of \(K\) in the above theorem. These ideas lead to Conjecture 1. Conjecture 2 is a variant of Mergelyan's theorem which assumes that \(f\) is nonvanishing in the interior of \(K\). Then, it is expected that \(f\) can be approximated by a polynomial that is nonzero at any point of \(K\). The author proves that these two conjectures are equivalent and that they are true for special cases. One case is that when the interior of \(K\) is a Jordan domain. Another case considers that \(K\) has finitely many maximally connected open subsets \(O\) which are Jordan domains, and furthermore if \(O_1\neq O_2\) then \(f(z)\) is nonvanishing on \(\overline{O_1}\cap \overline{O_2}\). Actually, these two particular cases give some strength to the previous conjecture although, according to the remarks made in the paper, the author does not appear too convinced about its validity.