L-functions of certain exponential sums over finite fields (Q2069656)

Let \(\mathbb F_q\) be the finite field of \(q\) elements with characteristic \(p\), \(\mathbb F_{q^k}\) be the finite extension of \(\mathbb F_q\) of degree \(k\), and \(\mathbb F_{q^k} = \mathbb F_{q^k} \backslash \{0\}\). Let \(\Psi: \mathbb F_p \to \mathbb C^*\) be a fixed nontrivial additive character, and let \(\mathrm{Tr}_k: \mathbb F_{q^k} \to \mathbb F_p\) be the trace map. For \(\vec{a} = (a_1, \dots, a_6) \in (\mathbb F_q^*)^6\), let \[ S_k(\vec{a}) = \sum_{\substack{\frac{a_5}{x_1x_2} + \frac{a_6}{x_3x_4} = 1 \\ x_i \in \mathbb F_{q^k}^*}} \Psi(\mathrm{Tr}_k(a_1x_1 + a_2x_2 + a_3x_3 + a_4x_4)), \] and let \[ L(\vec{a},T) = \exp \left( \sum_{k=1}^\infty S_k(\vec{a}) \frac{T^k}{k} \right). \] The main result of this paper is a complete determination of the weights and slopes of the \(L\)-function attached to the sequence \(\{S_k(\vec{a})\}_{k \ge 1}\) of exponential sums, whose proof relies on the Adolphson-Sperber's work on toric exponential sums [\textit{A. Adolphson} and \textit{S. Sperber}, Ann. Math. (2) 130, No. 2, 367--406 (1989; Zbl 0723.14017)] and on the Wan's decomposition theorems [\textit{D. Wan}, Ann. Math. (2) 137, No. 2, 249--293 (1993; Zbl 0799.11058)]. Theorem 1.1. The generating \(L\)-function of \(S_k(\vec{a})\) is a polynomial of degree \(8\), i.e., \[ L(\vec{a},T) = (1-T)(1-qT)\prod_{i=1}^6 (1 - \alpha_iT), \] where \(\alpha_i \in \overline{\mathbb Q}\) satisfies \(|\alpha_i| = q^{3/2}\). Seeing the reciprocal roots \(\alpha_i\) as \(p\)-adic numbers and enumerating them with respect to the \(q\)-adic slopes, their \(p\)-adic norms are given by \[ |\alpha_i|_p = \begin{cases} 1 & \text{ if } i=1, \\ q^{-1} & \text{ if } i=2,3, \\ q^{-2} & \text{ if } i=4,5, \\ q^{-3} & \text{ if } i=6. \end{cases} \] \textit{T. H. M. Smits} [Indag. Math. 47, 345--350 (1985; Zbl 0587.12011)] proved that there exist positive constants \(c_0, c_1\) such that if \(p \ge c_0\), then \(c_1q^{3k/2}\). This result has applications in Heath-Brown's and in Friedlander and Iwaniec's works on the divisor function \(d_3(n)\) [\textit{D. R. Heath-Brown}, Acta Arith. 47, 29--56 (1986; Zbl 0549.10034); \textit{J. B. Friedlander} and \textit{H. Iwaniec}, Ann. Math. (2) 121, 319--350 (1985; Zbl 0572.10029)] as well as in Zhang's work on twin prime conjecture [\textit{Y. Zhang}, Ann. Math. (2) 179, No. 3, 1121--1174 (2014; Zbl 1290.11128)]. As a consequence of Theorem 1.1, the authors prove the following sharp estimate. Corollary 1.2. For all \(p\), \(k\) and \(\vec{a}\) as before, it holds \(|S_k(\vec{a})| \le 6q^{3k/2} + q^k + 1\). Another consequence of Theorem 1.1 is to obtain an explicit counterexample of Adolphson-Sperber's conjecture (Corollary 3.9) on weights of toric exponential sums. In fact, \textit{A. Adolphson} and \textit{S. Sperber} [Invent. Math. 101, No. 1, 63--79 (1990; Zbl 0764.11037)] conjectured a simple weight formula which is true for dimensions up to \(4\). \textit{J. Denef} and \textit{F. Loeser} [Invent. Math. 106, No. 2, 275--294 (1991; Zbl 0763.14025)] showed that such conjecture is false for some \(5\)-dimensional simplicial Newton polyhedron. They proved the existence of a counterexample, but it was not explicit. This paper provides the first concrete counterexample.