Hyperkloosterman sums revisited (Q2109409)

Set \(K^{(n)}(\Lambda ,x):=x_1+\cdots +x_n+\Lambda /(x_1\cdots x_n)\in \mathbb{F}_q[\Lambda ,x_1^{\pm 1},\ldots ,x_n^{\pm 1}]\), where \(\mathbb{F}_q\) is a finite field of charecteristic \(p\) and \(q=p^a\). For any \(\bar{\lambda}\in \bar{\mathbb{F}}_q^*\), the hyperkloosterman sum is given by \(\operatorname{Kl}_{n+1}(\bar{\lambda}):=\sum_{\bar{x}\in (\mathbb{F}_q(\bar{\lambda})^*)^n} \Theta_{\bar{\lambda}}(K^{(n)}(\bar{\lambda} ,\bar{x})\), where \(\Theta_{\bar{\lambda}}\) is a non-trivial additive character on \(\mathbb{F}_q(\bar{\lambda})\). The authors return to some past studies of hyperkloosterman sums via \(p\)-adic cohomology with the aim to improve earlier results. They work with Dwork's \(\theta_{\infty}\)-splitting function and a better choice of basis for cohomology. This choice is motivated by the authors' recent work on the \(p\)-integrality of coefficients of \(A\)-hypergeometric series. They manage to remove all characteristic restrictions from earlier results.