A number theoretic characterization of \(E\)-smooth and (FRS) morphisms: estimates on the number of \(\mathbb{Z}/p^k\mathbb{Z}\)-points (Q6142643)

Suppose that \(X\) is a scheme of finite type over \(\mathbb{Z}\). It is well-known that there are deep connections between the singularities of \(X\) and the asymptotic behaviour of the number \(\#X(\mathbb{Z}/p^k\mathbb{Z})\) of \(\mathbb{Z}/p^k\mathbb{Z}\)-rational points of \(X\), where \(p\) is a fixed prime and \(k\) tends to infinity. One result of this kind is obtained by Aizenbud and Avni [Duke Math. J. 167, No. 14, 2721--2743 (2018; Zbl 1436.14044)], which states that \(X_{\mathbb{Q}}\) has rational singularities if and only if \(\#X(\mathbb{Z}/p^k\mathbb{Z})/p^{k\cdot \dim X_{\mathbb Q}}\) is bounded independently of \(k\) and \(p\). The main result of the present paper (Theorem~A) is a family variant of that result: Suppose that we are given a morphism \(\phi\colon X \to Y\) between schemes \(X\) and \(Y\) of finite type over \(\mathbb{Z}\). Suppose moreover that \(X_{\mathbb Q}\) and \(Y_{\mathbb Q}\) are smooth and geometrically irreducible. Theorem~A states that the above bound holds uniformly for all fibers of \(\phi\) if and only if \(\phi_{\mathbb{Q}}\) has the (FRS) property, i.e., if each fiber of \(\phi_{\mathbb{Q}}\) has rational singularities and moreover \(\phi_{\mathbb{Q}}\) is flat. There are natural measures on \(X(\mathbb{Z}_p)\) and on \(Y(\mathbb{Z}_p)\). A third condition equivalent to the above ones is that the push-forward of the natural measure on \(X(\mathbb{Z}_p)\) along \(\phi\) has bounded (uniformly for all \(p\)) and continuous density with respect to the natural measure on \(Y(\mathbb{Z}_p)\). The paper also contains some variants of Theorem~A. In particular, there are two variants where the condition of having rational singularities is replaced by having terminal singularities or log-canonical singularities, respectively. As a key ingredient to the proof of Theorem~A, the authors prove a result (Theorem~B) about approximating suprema of certain formally nonnegative motivic functions. The motivic functions considered here are the constructible motivic functions appearing in Cluckers-Loeser style motivic integration. Given such a motivic function \(f(x,w)\) in two (tuples of) variables, we can make sense of a function \(g_F(x):= \sup_w f_F(x,w)\) over all local fields \(F\) of sufficiently big residue characteristic. This family of functions \(g_F\) is rarely given by a motivic function \(g\), but there are different results stating that \(g_F\) can be approximated by \(h_F\) for some motivic function \(h\). Theorem~B states that a particularly good approximation exists (namely, up to a factor bounded independently of \(x\) and \(F\)) if one assumes \(f\) to be formally nonnegative in a suitable sense.