Near-optimal mean value estimates for multidimensional Weyl sums (Q2636915): Difference between revisions

Let \(s,k,d\) be positive integers, \(X\) a positive number. The authors count the solutions of the following system of Diophantine equations: \[ \sum_{j=1}^s x_{j1}^{i_1} \cdots x_{jd}^{i_d} = \sum_{j=1}^s y_{j1}^{i_1} \cdots y_{jd}^{i_d}, \] where the \(i_m\) are non-negative integers and \( 1\leq i_1 + \cdots +i_d \leq k\). Let \(r=\binom{k+d}{d}-1\) and \(K_{d,k}=\frac{d}{d+1}(r+1)k\). Let \(J_{s,k,d}(X)\) be the number of integral solutions of the system with \(1\leq x_{jm},y_{jm}\leq X\), where \(1\leq j \leq s, 1\leq m\leq d\). The authors prove: (1) If \(k\geq 2, s\geq r(k+1)\), then for each \(\varepsilon>0\), one has \(J_{s,k,d}(X)\ll X^{2sd-K+\varepsilon}\). This extends work of Wooley, who has treated the case \(d=1\). There is a corresponding lower bound (2): \(J_{s,k,d}(X)\gg X^{sd}+\sum_{j=1}^d X^{(2s-1)j+d-K_{j,k}}\). Problems of this type have been studied earlier, e.g. by \textit{G. I. Arkhipov} et al. [Trigonometric sums in number theory and analysis. Transl. from the Russian. Berlin: Walter de Gruyter (2004; Zbl 1074.11043)] and \textit{S. T. Parsell} [Proc. Lond. Math. Soc. (3) 91, No. 1, 1--32 (2005; Zbl 1119.11024)]. In the current paper, the number \(s\) of variables in (1) may be within a factor of \(2+2/d+O(d/k)\) of the least value of \(s\) for which the upper bound could possibly hold. This means that the number of variables required for this result is reduced by a factor of order \(\log(rk)\), compared to previous work, now to the correct order. Quite naturally, this considerable improvement has a wide range of applications, and the authors apply this e.g. to asymptotic counting of the solutions, questions on the validity of the Hasse-principle, or to counting solutions from a thin subset of the integers. The key method is the efficient congruence method of \textit{T. D. Wooley} [Ann. Math. (2) 175, No. 3, 1575--1627 (2012; Zbl 1267.11105)].