Introduction 0.1. Let X be a scheme, reduced and separated, of finite type over Z. For p a prime number, one may consider the set X(Z p ) of its Z p -rational points. For every n in N, there is a natural map π n : X(Z p ) → X(Z/p n+1 ) assigning to a Z p -rational point its class modulo p n+1 . The image Y n,p of X(Z p ) by π n is exactly the set of Z/p n+1 -rational points which can be lifted to Z p -rational points. Denote by N n,p the cardinality of the finite set Y n,p . By a result of the first author [7], the Poincaré seriesis a rational function of T . Later Macintyre [24], Pas [26] and Denef [10] proved that the degrees of the numerator and denominator of the rational function P p (T ) are bounded independently of p. One task of the present paper is to prove a much stronger uniformity result by constructing a canonical rational function P ar (T ) which specializes to P p (T ) for almost all p. It follows in particular from our results that there exist, for every n in N, varieties Z n,i and rational numbers r n,i in Q, 1 ≤ i ≤ m n , such that, for almost all p and every n,Hence a natural idea would be to try to construct P ar (T ) as a series with coefficients in K 0 (Sch Q ) ⊗ Q, with K 0 (Sch Q ) the "Grothendieck ring of algebraic varieties over Q", defined in 1.2. However, since different varieties over a number field may have the same L-function, and we want the function P ar (T ) to be canonical, we have to replace the varieties Z n,i by Chow motives and the naive Grothendieck ringin the Grothendieck ring of Chow motives with coefficients inQ, as defined in 1.3.We can now state our uniformity result on the series P p (T ) as follows. In fact, one constructs in 9.2, for X a variety over any field k of characteristic zero, a canonical series P ar (T ) with coefficients in the Grothendieck ring K v 0 (Mot k,Q )⊗Q -defined in 1.3 -which is equal to the former one when k = Q and, furthermore, the series P ar (T ) is rational in a precise sense (Theorem 9.2.1).
Now assume X is a subscheme of the affine spaceThe starting point in the proof of the rationality of the serieswhere W p is defined asx ≡ y mod w, and f i (y) = 0, for i = 1, . . . , r . In particular, W p is defined by a formula in the first order language of valued fields independently of the prime p.More generally, let k be a finite extension of Q with ring of integers O andwith N a nonzero multiple of the discriminant. For x a closed point of Spec R, we denote by O x the completion of the localization of R at x, by K x its fraction field, and by F x the residue field at x, a finite field of cardinality q x . Let f (x) be a polynomial in k[x 1 , . . . , x m ] (or more generally a definable function in the first order language of valued fields with variables and values taking place in the valued field and with coefficients in k) and let ϕ be a formula in the language of valued fields with coefficients in k, free variables x 1 , . . . , x m running over the valued field and no other free variables. Now set W x := {y ∈ O m x |ϕ(y) holds in K x } a...