New estimates for exponential sums over multiplicative subgroups and intervals in prime fields (Q784287)

Exponential sums over finite fields have been the subject of intense research in analytic number theory. Estimates of such sums almost always relate to the solutions of certain congruence Diophantine systems. In the paper under review, the authors established a new upper bound for the sum \[\mathcal{S}_a(H) := \sum_{x \in \mathcal{H}} e_p(ax)\] where \(p\) is a large prime number, \(a\) is a fixed integer such that \((a,p)=1\), \(e_p(t) := e^{2 \pi i t/p}\) as usual and \(\mathcal{H}\) is a multiplicative subgroup of \(\mathbb{F}_p^*\) of order \(H\). It has long been known since Gauss that \(\left| \mathcal{S}_a \left(\frac{p-1}{2} \right) \right| = p^{1/2}\). Using an explicit estimate on a trilinear exponential sum of the shape \(\sum_{x \in X}\sum_{y \in Y} \sum_{z \in Z} \alpha_x \beta_y \gamma_z e_p(xyz)\), where \(X,Y,Z\) are subgroups of \(\mathbb{F}_p^*\) and \(\alpha_x\), \(\beta_y\) and \(\gamma_z\) are any compex numbers of modulus \(\leqslant 1\), and new bounds for the number \(T_m(H)\) of the solutions of the congruence \[h_1+\dotsb+h_m \equiv h_{m+1} + \dotsb + h_{2m} \pmod p\] where \(h_i \in \mathcal{H}\), the authors improve an earlier result of \textit{J. Bourgain} and \textit{M. Z. Garaev} [Math. Proc. Camb. Philos. Soc. 146, No. 1, 1--21 (2009; Zbl 1194.11086)] by showing that, for all \(\varepsilon > 0\) \[\max_{(a,p)=1} \left| \mathcal{S}_a(H) \right| \ll_\varepsilon H^{\frac{2689}{2880}} p^{\frac{1}{72}+ \varepsilon}\] provided that \(p^{1/4} < H < p^{1/2}\). In particular, this yields \[\max_{(a,p)=1} \left| \mathcal{S}_a(H) \right| \ll_\varepsilon H^{1-\frac{31}{2880} + \varepsilon}\] for all \(H > p^{1/4}\), representing a saving of a little more than \(1\)\% in comparison to the bound given by Bourgain and Garaev. The authors next prove two results for the sum \[\mathcal{S}_a(N,H) := \sum_{L < n \leqslant L+N} \left| \sum_{x \in \mathcal{H}} e_p(anx) \right|\] for some large integer \(N \leqslant p\), assuming the interval \([L+1,L+N]\) avoids \(0 \pmod p\), and implying in particular that, for all \((a,p)=1\) and all \(\varepsilon > 0\) \[\mathcal{S}_a(p^{1/3+\varepsilon},p^{1/3+\varepsilon})\ll_\varepsilon N^{2-1/6+\varepsilon}\] improving in the case of subgroups a previous bound given in [\textit{M. Z. Garaev}, J. Number Theory 199, 377--388 (2019; Zbl 1458.11128)].