Joint universality of Hurwitz zeta-functions and nontrivial zeros of the Riemann zeta-function. II (Q2231362)

Let \(s\) be a complex variable, \(s=\sigma+it\). Let \(\zeta(s)\) denote the Riemann zeta function, so \(\zeta(s)=\sum_{n\ge 1} n^{-s}\) for \(\sigma>1\). Let \(\zeta(s,\alpha)\) denote the Hurwitz zeta function, where \(\alpha\in (0, 1]\), so \(\zeta(s,\alpha)=\sum_{m\ge 0} (m+\alpha)^{-s}\) for \(\sigma>1\) and for instance \(\zeta(s,1)=\zeta(s)\). Lastly, let \(0<\gamma_1 < \cdots \le \gamma_k \le \cdots\) denote the ordinates of the nontrivial zeros of \(\zeta(s)\), so \(\gamma_1 = 14.134725\cdots\). The paper under review is a followup to previous work by the same authors [Part I, Lith. Math. J. 59, No. 1, 81--95 (2019; Zbl 1486.11110)] in which they established a joint universality theorem to approximate a large class of analytic functions by the shifted functions \(\zeta(s+i h \gamma_k,\alpha_1),\ldots, \zeta(s+i h \gamma_k,\alpha_r)\), where \(h>0\) and \(\alpha_1,\ldots,\alpha_r\in (0,1]\), and such that the \(\alpha_j\) satisfy a natural linear independence requirement over \(\mathbb{Q}\). This previous result was conditional on the estimate, \(\# \{\gamma_k,\gamma_l \le T : |\gamma_k-\gamma_l|< c/\log T\} \ll T\log T\) where \(c>0\). For example, when \(r=1\), the authors' result was that for all positive \(\varepsilon\) and any compact subset \(K\) of the vertical strip \(1/2<\sigma <1\) such that the complement of \(K\) is connected, \(\liminf_{N\to \infty} \#\{1\le k\le N : \sup_{s\in K} |\zeta(s+ih \gamma_k,\alpha)-f(s)|<\varepsilon\} >0\), where \(f\) is any function continuous on \(K\) and analytic in the interior of \(K\). In this followup work, and under similar assumptions as above, the authors generalize their universality theorem so that (continuing to take \(r=1\) for simplicity) \(\zeta(s+ih\gamma_k, \alpha)\) may be replaced with \(F(\zeta(s+ih\gamma_k,\alpha))\) where \(F\) belongs to a suitable class of operators; e.g., \(F(g)=g^k\), where \(k\in \mathbb{N}\) and \(F(g) = e^g+e^{-g}\) are both admissible. The proof of these generalizations use a theorem by \textit{S. N. Mergeljan} [Usp. Mat. Nauk 7, No. 2(48), 31--122 (1952; Zbl 0049.32701)] to approximate analytic functions by polynomials, as well as a probabilistic model for a randomized Hurwitz zeta function which the authors developed in their previous work.