On explicit estimates for \(S(t)\), \(S_1(t)\), and \(\zeta ( 1 / 2 + \operatorname{i} t )\) under the Riemann hypothesis (Q2239166)

Let \(\zeta(s)\) be the Riemann zeta function. Let \(S(t)=\frac{1}{\pi}\arg\zeta(1/2+it)\), where \(t\) does not coincide with a zeta zero and the argument is defined by a continuous variation along the horizontal line from \(+\infty+it\) to \(1/2+it\) starting with the value \(0\). So \(S(t)\) is the ``fluctuating part'' of the zeros-counting function \(N(t)\) for zeta. Let \(S_1(t)\) be the integral \(S_1(t)=\int_0^t S(u)\,du\). Littlewood proved bounds on \(S(t)\), \(S_1(t)\), and \(|\zeta(1/2+it)|\) under the assumption of the Riemann hypothesis. Using the author's notation, the Littlewood result is there exist positive constants \(C_0\), \(C_1\), and \(C_{00}\) such that as \(t\to \infty\) one has \(|S(t)| \le (C_0+o(1))\log t/\log\log t\), \(|S_1(t)|\le (C_1+o(1))\log t/(\log\log t)^2\), and \(|\zeta(1/2+it)|\le \exp((C_{00}+o(1))\log t/\log\log t)\). Although the order of magnitude in these conditional bounds has never been improved, the admissible values of the constants \(C_0\), \(C_1\), and \(C_{00}\) have gone through improvements. The author gives a concise survey of results on \(C_0\), \(C_1\), and \(C_{00}\). Specifically, \textit{K. Ramachandra} and \textit{A. Sankaranarayanan} [J. Number Theory 44, No. 3, 281--291 (1993; Zbl 0778.11047)], using a method of Selberg, obtained that \(1.12\) is an admissible value of \(C_0\). This has been improved by \textit{A. Fujii} [Comment. Math. Univ. St. Pauli 55, No. 2, 135--147 (2006; Zbl 1157.11035)], \textit{D. A. Goldston} and \textit{S. M. Gonek} [Bull. Lond. Math. Soc. 39, No. 3, 482--486 (2007; Zbl 1127.11058)], and \textit{E. Carneiro} et al. [Math. Ann. 356, No. 3, 939--968 (2013; Zbl 1325.11084)] who obtained that \(1/4\) is admissible. As for \(C_1\), \textit{A. A. Karatsuba} and \textit{M. A. Korolev} [Russ. Math. Surv. 60, No. 3, 433--488 (2005; Zbl 1116.11070); translation from Usp. Mat. Nauk 60, No. 3, 41--96 (2005)] calculated an admissible value of \(C_1\), which has been improved by Fuji and Goldston-Gonek who obtained \(\pi/24=0.1308\ldots\) is admissible. Lastly, an admissible value of \(C_{00}\) was calculated by Ramachandra-Sankaranarayanan, and has since been improved by \textit{K. Soundararajan} [Ann. Math. (2) 170, No. 2, 981--993 (2009; Zbl 1251.11058)] and \textit{V. Chandee} and \textit{K. Soundararajan} [Bull. Lond. Math. Soc. 43, No. 2, 243--250 (2011; Zbl 1238.11078)] who obtained that \(\log \sqrt{2}=0.3465\ldots\) is admissible. The author points out that throughout this progress on improving the constants \(C_0\), \(C_1\), and \(C_{00}\), the implicit \(o\)-terms have not been made explicit. The author presents a complete and detailed derivation to make these \(o\)-term explicit. The author elaborates on Fuji's work and employs a classical formula by Selberg to approximate \(S(t)\) and \(S_1(t)\) by short Dirichlet series. As for approximating \(\log|\zeta(1/2+it)|\) by a short Dirichlet series, the author relies on the Soundararajan refinement of the Selberg formula. The main result by the author is that the implicit \(o\)-terms may all take the form \(\frac{b}{(\log t)^c\log\log t}\) for some positive constants \(b\) and \(c\). The author calculates explicit values of \(b\) and \(c\) for each of \(S(t)\), \(S_1(t)\) and \(|\zeta(1/2+it)|\). The resulting bounds are applicable for very large \(t\ge t_0\) where \(t_0\) is given explicitly in each case. The author's method yields admissible values of \(C_0\), \(C_1\), and \(C_{00}\), though these are weaker than the best known ones. For example, in the case of \(S(t)\), the author proves that one may take \(b=20.1911\), \(c= 0.285\), and \(t_0=10^{2465}\), and obtains \[|S(t)|\le \left(0.759282+\frac{20.1911}{(\log t)^{0.285}\log\log t}\right)\frac{\log t}{\log\log t},\qquad\qquad (t\ge 10^{2465}).\] The value \(0.759282\) is larger than the best known admissible value of \(C_0\), namely \(1/4\). Finally, as an application of these bounds, the author gives an explicit bound on gaps between ordinates of consecutive nontrivial zeros of zeta in Corollary 2.