Exponential sums over definable subsets of finite fields (Q2480574)

The paper gives some general estimates for exponential sums over subsets of finite fields which are definable in the language of rings. It combines in this way classical interests in analytic number theory and ideas and methods from model theory. In fact, it generalizes both the exponential sum estimates of varieties of finite fields due to Weil, Deligne and others, and the results of Chatzidakis, van den Dries and Macintyre on the number of points of those definable sets. The main theorem refers to a formula \(\varphi(x, y)\) of the language of rings and the set it defines in a finite field \({\mathbb F}_p\) (for \(p\) a prime) with respect to a given choice of a tuple \(y\) of parameters; also \(x\) denotes here a tuple of variables of length possibly \(> 1\). Let \(\psi_p : {\mathbb F}_p \to {\mathbb C}^\times\) be a non-trivial additive character and \(\chi_p : {\mathbb F}_p^\times \to {\mathbb C}^\times\) be a multiplicative character. Also, let \(f\) and \(g\) be two rational functions of \(x\) over the integers. Then one can form the exponential sum \[ S(y, \varphi, p) = \sum_{x \in \varphi({\mathbb F}_p, y)} \psi_p (f(x)) \chi_p (g(x)) \] for all primes \(p\) for which all the involved data make sense. The paper shows that there are constants \(C \geq 0\) and \(\eta >0\), depending only on \(\varphi\) and the degrees of the numerators and denominators of \(f\) and \(g\), such that, for any prime \(p\) and any tuple of parameters \(y\) in \({\mathbb F}_p\), \[ | S(y, \varphi, p)| \leq C p^{-{1 \over 2}} \sum_{x \in \varphi ({\mathbb F}_p, y)} 1 \] unless there is \(c \in {\mathbb F}_p\) for which \[ | \{ x \in \varphi({\mathbb F}_p, y) : f(x) = c \}| \geq \eta | \varphi({\mathbb F}_p, y)|. \] As a consequence (with a model-theoretic flavour), it is observed that there is no formula in the language of rings that defines for infinitely many primes \(p\) an ``interval'' in \({\mathbb Z}/p{\mathbb Z}\) that is neither bounded nor with bounded complement.