Nonnegative trigonometric polynomials and a zero-free region for the Riemann zeta-function (Q2517141)

Let \(\zeta(s)\) denote the Riemann zeta-function where \(s=\sigma+it\) is a complex number. Determining zero-free regions for \(\zeta(s)\) is an old and important problem in number theory. In 1899, as part of his proof of the prime number theorem (with error term), de la Vallée Poussin proved that \(\zeta(s)\) had no zeros in the region \[ \sigma>1 -\frac{1}{R_0 \log|t|} \] where \(R_0\) is a particular constant and \(|t|\) is sufficiently large. This is now referred to as the classical zero-free region for \(\zeta(s)\). A great deal of work has gone into establishing admissible values of \(R_0\) in the classical zero-free region. Until recently, the state of the art was a method of \textit{H. Kadiri} [Acta Arith. 117, No. 4, 303--339 (2005; Zbl 1101.11029)] which proved that \(R_0 =5.69693\) is admissible if \(|t|\geq 2\). Using Kadiri's method, \textit{W.-J. Jang} and \textit{S.-H. Kwon} [J. Korean Math. Soc. 51, No. 6, 1291--1304 (2014; Zbl 1315.11072)] improved this slightly by proving that \(R_0=5.68371\) is admissible for \(|t|\geq 2\). In the paper under review, the authors prove that \[ \sigma>1 -\frac{1}{5.573412 \log|t|} \quad \text{for }|t| \geq 2. \] Their improvements come from two factors: determining trigonometric polynomials having some favorable properties and from analyzing the error term in Kadiri's method. The authors also improve an upper bound in a question of Landau regarding nonnegative trigonometric polynomials. Let \(P_n\) denote the set of even trigonometric polynomials \[ f_n(\varphi)=\sum_{k=0}^n a_k \cos(k\varphi) \] of degree \(n\) having each \(a_k\geq 0\) and \(a_1>a_0\), and let \[ V = \lim_{n\to\infty} \inf_{f_n\in P_n} \frac{f_n(0)-a_0}{(\sqrt{a_1}-\sqrt{a_0})^2}. \] Then \(34.468305 <V<34.4889920009.\) This improves on some previous work of \textit{V. V. Arestov} and \textit{V. P. Kondrat'ev} [Mat. Zametki 47, No. 1, 15--28 (1990; Zbl 0695.42001)].