Counting arithmetic subgroups and subgroup growth of virtually free groups (Q2344502)

This is a clearly written paper proving some interesting results. We first state the main result. Let \(K\) be a \(p\)-adic field and let \(H = \mathrm{PSL}_2(K)\). Fixing a Haar measure on \(H\) with respect to which, a fixed maximal compact subgroup has unit covolume, one considers the number \( \mathrm{AL}_H(x)\) of conjugacy classes of arithmetic lattices in \(H\) with covolume at the most \(x\). If \(H\) does not contain an element of order \(p\) (equivalently, the field \(K\) does not contain \(\zeta + \zeta^{-1}\) for a \(p\)-th root of unity \(\zeta \neq 1\)), then \[ \lim_{x \rightarrow \infty} \frac{\log \mathrm{AL}_H(x)}{x \log x} = q-1 \] where \(q\) is the cardinality of the residue field of \(K\). The ``\textit{real}'' counterpart of this theorem is due to Belolipetsky, Gelander, Lubotzky and Shalev (see [\textit{M. Belolipetsky} et al., Ann. Math. (2) 172, No. 3, 2197--2221 (2010; Zbl 1214.22002)]) where the constant \(q-1\) has the counterpart \(\frac{1}{2 \pi}\). The author uses the results and methods of this last-mentioned paper as well. The upper bound for \( \mathrm{AL}_H(x)\) is the more difficult part of the theorem. Note that we are counting only arithmetic lattices because the nonarithmetic ones are uncountably many. In the proof, the author uses the method of proof for the ``real'' analogue as mentioned above but there are points of departure. The part which carries over with minimal effort is (as the author shows): Let \(H\) be a finite product of \( \mathrm{PGL}_2(K)\)'s where each \(K\) is a local field of characteristic \(0\). Let \( \mathrm{MAL}_H(x)\) be the number of conjugacy classes of maximal arithmetic lattices in \(H\) with covolume at the most \(x\). Then, for every \(\epsilon > 0\), there is a constant \(\beta(\epsilon) > 0\) such that \[ \mathrm{MAL}_H(x) \leq x^{\beta(\epsilon) (\log x)^{\epsilon}} \] for \(x \gg 0.\) The proof in the earlier paper did not cover nonarchimedean fields and the author essentially checks that the subsidiary results used along that proof remained valid when there are nonarchimedean factors as above. This has not been checked as yet for local fields of positive characteristic which is the principal reason the results of the present paper are stated only for \(p\)-adic fields and not for all cocompact arithmetic lattices over nonarchimedean local fields of any characteristic. The part of the author's proof which requires much more work addresses and proves the following assertions: Let \(K\) be a \(p\)-adic field such that \(H = \mathrm{PSL}_2(K)\) has no elements of order \(p\). Then, there exists a constant \(c>0\) depending only on \(K\) such that for every lattice \(\Gamma\) in \(H\) and, for every \(n\), we have \[ s_n(\Gamma) \leq (cn)^{-\chi(\Gamma)m} \] where \(s_n\) counts the number of subgroups of index \(\leq n\) in \(\Gamma\) and \(\chi(\Gamma)\) denotes the Euler characteristic of \(\Gamma\). The author remarks that this result carries over to cocompact lattices in \( \mathrm{PSL}_2(K)\) for nonarchimedean local fields \(K\) of positive characteristic also. The basic idea behind the proof of the above result is to view a lattice in \(H\) as the fundamental group of a graph of groups. These lattices are necessarily virtually free and a recent, general method of Schlage-Puchta shows how to compute the number \[ \nu(\Gamma) = \limsup_{n \rightarrow \infty} \frac{\log a_n(\Gamma)}{n \log n} \] for any finitely generated virtually free \(\Gamma\) as the solution of a linear optimization problem, where \(a_n\) counts the number of subgroups of index equal to \(n\). The bulk of the work in the present paper is to show that when \( \mathrm{PSL}_2(K)\) does not have elements of order \(p\) (where \(K\) is a \(p\)-adic field), then for any lattice \(\Gamma\) in \(H\), the number \(\nu(\Gamma)\) coincides with \(-\chi(\Gamma)\). In this quest, the author needs to look at vertex stabilizers of the lattices when they act on the Bruhat-Tits tree; these are finite groups isomorphic to \(D_{2n},C_n,A_4,A_5\) or \(S_4\). This is due to the fact that the subgroups are supposed to be without elements of order \(p\), and so, they do not intersect the pro-\(p\) group which is the kernel of \( \mathrm{PSL}_2(O_K) \rightarrow \mathrm{PSL}_2(F)\) nontrivially, where \(F\) is the residue field of \(K\). Each of these finite subgroups is dealt with case by case.