Counting lattices in simple Lie groups: the positive characteristic case (Q408111)

Let \(G = \mathbb{G}_0(K)\), where \(K\) is a local field of characteristic \(p \geq 5\) and \(\mathbb{G}_0\) is a simply connected, absolutely almost simple \(K\)-group of \(K\)-rank at least \(2\). Fix a Haar measure \(\mu\) on \(G\) and let \(\rho_x(G)\), resp. \(m_x(G)\), denote the number of lattices, resp. maximal lattices, in \(G\) of covolume at most \(x\) counted up to the action of the group of automorphisms of \(G\). Let also \(c_x(\Gamma)\) denote the number of congruence subgroups of a lattice \(\Gamma\) of index at most \(x\). Assume that the congruence subgroup property, the Margulis-Platonov conjecture, and the Weil conjecture about the Tamagawa number hold for any group with the same absolute type as \(\mathbb{G}_0\). These conjectures are known in many special cases and widely believed to be true in general. Their precise formulation and a short review of the related results are given in the paper. Some of the results cited below do not require all of the conjectures but we will not discuss these differences in the review. The main results of the paper are the following. {Theorem 1.} (Theorem 2 in the paper.) There are positive numbers \(A\) and \(B\) which depend only on \(G\) such that \[ x^A \leq m_x(G) \leq x^{B\log\log x} \] for all sufficiently large \(x\). {Theorem 2.} (Corollary 4 in the paper.) There are positive numbers \(C\), \(D\), and \(x_0\) depending only on \(G\) such that for any lattice \(\Gamma\) in \(G\) we have \[ x^{C\log x/\log(vol(G/\Gamma))} \leq c_x(\Gamma) \leq (vol(G/\Gamma)x)^{D\log(vol(G/\Gamma)x)} \] for any positive number \(x \geq x_0\). {Theorem 3.} (Theorem 1 in the paper.) There exist positive numbers \(C\) and \(D\) depending only on \(G\) such that \[ x^{C\log x} \leq \rho_x(G) \leq x^{D\log x} \] for all sufficiently large \(x\). Moreover, the lower bound is unconditional. Analogous results were previously known in the characteristic zero case. Thus, Theorem~1 gives a positive characteristic analogue of a result of the reviewer [Duke Math. J. 140, No. 1, 1--33 (2007; Zbl 1131.22008)], Theorem~2 corresponds to the work of \textit{D. Goldfeld, A. Lubotzky} and \textit{L. Pyber} [Acta Math. 193, No. 1, 73--104 (2004; Zbl 1159.20329)] and \textit{A. Lubotzky} and \textit{N. Nikolov} [Acta Math. 193, No. 1, 105--139 (2004; Zbl 1145.22004)], and the third result is parallel to a recent work of \textit{A. Lubotzky} and the reviewer [Adv. Math. 229, No. 6, 3123--3146 (2012; Zbl 1241.22014)]. In previous papers Lubotzky conjectured that the rate of growth of \(\rho_x(G)\) might be the same as the growth of the number of subgroups of index at most \(x\) in any fixed lattice in \(G\). Theorem~3 proves this conjecture in the positive characteristic case, which is in sharp contrast with the characteristic zero case where we now know that the conjecture is not true. The general framework of the proofs of the results in the positive characteristic case is similar to that in the characteristic zero case but there are many technical differences. The first difference appears already in the definition of the growth function \(\rho_x(G)\): in the positive characteristic the group of outer automorphisms of \(G\) is an infinite compact group while in the characteristic zero it is always finite. This implies that one cannot just count the conjugacy classes of lattices as it is done in the latter case. The next essential differences arise in the proof of Theorem~1 where the author uses the Riemann-Roch theorem, Weil's theorem on the Riemann hypothesis for curves over finite fields and a result of de Jong and Katz on the number of curves. The proofs of the other two theorems are based on the study of graded Lie algebras associated with parahoric subgroups. Here the author develops and applies the previous work of \textit{G. Prasad} and \textit{M. S. Raghunathan} [Ann. Math. (2) 119, 143--201 (1984; Zbl 0552.20025)], which was the first one where these methods came up in connection with subgroups counting. The novelty of the proof of Theorem~2 is that in order to obtain the congruence subgroup growth of a given lattice the author first considers a maximal lattice that contains it and then employs the structural information about the maximal lattices. This idea may be useful in the characteristic zero case as well. There are many important technical results in the paper which may find further applications in the study of lattices in Lie groups over fields of positive characteristic.