Double sums of Kloosterman sums in finite fields (Q2007705)

For a nontrivial additive character \(\psi\) of \(\mathbb F_q\) and subsets \(\mathcal{U},\mathcal{V}\subseteq\mathbb F_q\) the authors study the sums of weighted Kloosterman sums \[\mathcal{S}_\psi(\boldsymbol{\alpha};\mathcal{U},\mathcal{V})=\sum_{u\in \mathcal{U}}\sum_{v\in \mathcal{V}} \alpha_v \sum_{x\in \mathbb F_q^*} \psi(ux+vx^{-1})\] with a sequence of complex weights \(\boldsymbol{\alpha}=(\alpha_v)_{v\in \mathcal{ V}}\). The Weil bound immediately gives \[|\mathcal{S}_\psi(\boldsymbol{\alpha};\mathcal{U},\mathcal{V})|\le 2|\mathcal{U}|q^{1/2}\sum_{v\in \mathcal{V}}|\alpha_v|.\] The authors improve this bound in two cases \begin{itemize} \item \(\alpha\) is the all \(1\)'s vector and \(\mathcal{U}\) and \(\mathcal{V}\) are both affine spaces, \item \(\mathcal{U}\) is an affine space but \(\alpha\) and \(\mathcal{V}\) are arbitrary. \end{itemize} The proofs are based on new results from additive combinatorics of \textit{A. Mohammadi} [``Szemerédi-Trotter type results in arbitrary finite fields'', Preprint, \url{arXiv:1808.05543}]. More precisely, in Theorem 1.1 \(q=p^{2k+1}\) is an odd power of a prime \(p\), \(\mathcal{U}\) and~\(\mathcal{V}\) affine subspaces of \(\mathbb F_q\) with \(|\mathcal{U}|\le |\mathcal{V}|\) and \(\alpha\) the all \(1\) vector. Then \[S_\psi(\boldsymbol{\alpha}; \mathcal{U},\mathcal{V})\ll |\mathcal{U}||\mathcal{V}|\max\{q^{52/153},(q/|\mathcal{U}|)^{831/832},(q/|\mathcal{U}|)^{761/760}q^{-1/760}\}.\] This is an improvement for \(|\mathcal{U}|\ge q^{415/831}\). In Theorem 1.2 again \(q=p^{2k+1}\) and \(\mathcal{U}\) is an affine subspace of \(\mathbb F_q\) but \(\boldsymbol{\alpha}\) and \(\mathcal{V}\) are arbitrary. Then \[S_\psi(\boldsymbol{\alpha}; \mathcal{U},\mathcal{V})\ll c(\boldsymbol{\alpha}) \max\{|\mathcal{U}|q^{103/204},|\mathcal{U}|^{313/1248}q^{1247/1248},|\mathcal{U}|^{71/285}q\},\] where \[c(\boldsymbol{\alpha})=\left(\sum_{v\in \mathcal{V}} |\alpha_v|\right)^{1/2}\left(\sum_{v\in \mathcal{V}} |\alpha_v|^2\right)^{1/4}.\] The condition that \(q\) is an odd power of \(p\) is needed to avoid the existence of large subfields and it is an interesting open problem to drop this condition.