Extremal values of the Epstein zeta functions (Q2397402): Difference between revisions

From the text: In 1928, \textit{E. C. Titchmarsh} proved the following fact (Titchmarsh's phenomenon, see ``The Theory of the Riemann zeta-function.'' 2nd ed. Oxford: Clarendon Press (1986; Zbl 0601.10026), Theorem 8.12). Let \(\sigma\) be a fixed number, \(1/2\leq \sigma < 1\). Then the inequality \[ |\zeta(\sigma + it)| > \exp(\log^\alpha t) \tag{1} \] holds for a certain sequence \(t = t_n\) such that \(\lim_{n\to\infty} t_n = \infty\), provided that \(\alpha< 1-\sigma\). The purpose of the present work is to obtain analogs of inequality (1) for the Epstein zeta functions for \(l \geq 2\). We are dealing with the extreme values on vertical lines in the strip \((l -1)/2 \leq \sigma \leq l/2\). Section 2 discusses the Epstein zeta functions \(\zeta_l(s)\) \((l \geq 4)\) and the possibility of obtaining Titchmarsh's phenomenon in the general case \(\zeta_Q(s)\) \((l \geq 4)\). In Section 3, the ternary case of \(\zeta_Q(s)\), \(l = 3\), is considered, with emphasis on the form \(Q(u_1, u_2, u_3)\) The Epstein zeta functions associated with binary forms are studied in Section 4. The author uses the works of \textit{K. Ramachandra} [J. Lond. Math. Soc., II. Ser. 8, 683--690 (1974; Zbl 0287.10023)] and of \textit{R. Balasubramanian} and \textit{K. Ramachandra} [Proc. Indian Acad. Sci., Sect. A 86, 341--351 (1977; Zbl 0373.10026)], in which Titchmarsh's method was generalized and refined. In deriving \(\Omega\)-results for the Epstein zeta functions by Titchmarsh's method, the classical question on representation of integers by positive definite quadratic forms plays the central part. Let \(Q(u_1, u_2, \dots, u_l)\) be a positive definite quadratic form in \(l\) \((\geq 2)\) variables with integer coefficients. Put \[ \zeta_Q(s) = \sum{}' (Q(u_1, u_2, \dots, u_l))^{-s}, \] where the prime on the summation sign indicates that summation is carried out over all integer \(l\)-tuples \((u_1, u_2, \dots, u_l)\) with the exception of \((0, 0, \dots, 0)\). It is known that \(\zeta_Q(s)(s-\tfrac{l}{2})\) is an entire function. The paper deals with \(\Omega\)-theorems for \(\zeta_Q(s)\) with \(l\geq 3\) and for certain \(\zeta_Q(s)\) with \(l = 2\). Let \(l\geq 4\) and let \(F_Q(s) = \zeta_Q(s) +\tfrac{l}{2}-1\). As \(t\) tends to infinity, we have \[ \log| F_Q(\tfrac12+it)| = \Omega_+\left(\left(\frac{\log t}{\log\log t}\right)^{1/2}\right) \] and, for fixed \(\sigma_0\in (\tfrac12,1)\), \[ \log| F_Q(\sigma_0+it)| = \Omega_+\left(\left(\frac{(\log t)^{1-\sigma_0}}{\log\log t}\right)^{1/2}\right) \]