On the Sato-Tate conjecture for Kloosterman sums. II. (Q1974799)

Part I, cf. Invent. Math. 121, 61-78 (1975; Zbl 0844.11055). The Sato-Tate conjecture for Kloosterman sums predicts that if \(K(1,m;q)= 2^{\omega(q)} q^{1/2} \cos (\theta_{m,q})\), with \(0\leq \theta_{m,q}\leq \pi\), then \(\theta_{m,q}\) should be equi-distributed with respect to the measure \(\frac 2\pi \sin^2 \theta d\theta\). The aim of the paper is to investigate this, with varying \(q\), and \(m\) as restricted as possible. Ideally one would want to respect \(q\) to be prime. From the author's first result one deduces the following. Let \(\alpha\) and \(\beta\) be constants satisfying \(16/17< \beta< 1\) and \((2\beta)^{-1}< \alpha< \{32(1- \beta)\}^{-1}\), and let \({\mathcal M}\subset \mathbb{N}\) have \(M(x)\gg x^\beta\) elements up to \(x\), for every \(x\geq 1\). Then \[ \sum_{m\in{\mathcal M}, m\leq P^\alpha} \Biggl|\sum_{p\leq P, p\nmid m} \frac{K(1,m;p)} {2\sqrt{p}} \Bigg|\ll M(P^\alpha) P^{1-\delta} \] for some small constant \(\delta> 0\). Here \(p\) runs over primes, so that one sees that the first moment of \(K(1,m;p)\) contains some cancellation for `most' values of \(m\). For higher moments the author estimates \[ \sum_{m\leq M} \Biggl|\sum_{r\leq R} \sum_{s\leq S, (m,r_s)=1} \mu^2(rs) \biggl\{ \frac{K(1,m;rs)} {2^{\omega(rs)} \sqrt{rs}} \biggr\}^j \Biggr| \] for suitable ranges of \(M\), \(R\), \(s\). The proofs require bounds for averages of Kloosterman sums due to \textit{J. M. Deshouillers} and \textit{H. Iwaniec} [Invent. Math. 70, 219-288 (1982; Zbl 0502.10021)], and a generalization of an estimate for individual exponential sums, due to \textit{E. Fouvry, H. Iwaniec} and \textit{N. Katz} [Acta Arith. 61, 271-287 (1992; Zbl 0764.11040)].