Atkinson's formula for the mean square of \(\zeta (s)\) with an explicit error term (Q2112778)

The `Atkinson's formula' of the title of the paper under review refers to a result of Atkinson dealing with the error term \(E(T)\) of the asymptotic formula for the mean square of the Riemann zeta function on the line \(\sigma = \frac{1}{2}\). \textit{F. V. Atkinson} [Acta Math. 81, 353--376 (1949; Zbl 0036.18603)] proved that \(E(T) = \Sigma_1(T) + \Sigma_2(T) + \mathfrak{E}(T)\), where \(\Sigma_1(T)\) and \(\Sigma_2(T)\) are explicit sums of length \(\asymp T\) involving coefficients \(d(n)\), and \(\mathfrak{E}(T)\) is an error term. See \textit{A. Ivić}'s book [The Riemann zeta-function. Theory and applications. Reprint of the 1985 original. Mineola, NY: Dover Publications (2003; Zbl 1034.11046)] for a complete exposition of Atkinson's formula. Atkinson showed that \(\mathfrak{E}(T) \ll (\log T)^2\), and subsequently several experts used Atkinson's formula to get nontrivial results for \(E(T)\), see for instance the works of \textit{D. R. Heath-Brown} [Mathematika 25, 177--184 (1979; Zbl 0387.10023)] and \textit{M. Jutila} [Ark. Mat. 21, 75--96 (1983; Zbl 0513.10040)], from which it can be deduced that \(E(T) \ll T^{1/3} (\log T)^3\). In this article, the authors establish a totally explicit upper bound for \(|\mathfrak{E}(T)|\) for large \(T\). As a corollary, they get the estimate \[ \left| E(T) \right| \leqslant 2.856 \, T^{1/3} (\log T)^{5/3} \] provided that \(T \geqslant 1.1 \times 10^{100}\). One of the main tools of the proof is an explicit version of the saddle point method in estimating integrals of the shape \[ \int_a^b g(x) \, e \left( f(x) + kx \right)\,\mathrm{d}x \] where \(f(z)\) and \(g(z)\) are complex functions satisfying certain properties and \(a < b\) and \(k\) are real numbers.