On Benjamini-Schramm limits of congruence subgroups (Q2204400)

Benjamini-Schramm convergence of Riemannian manifolds was introduced by \textit{M. Abert} et al. [Ann. Math. (2) 185, No. 3, 711--790 (2017; Zbl 1379.22006)] to have a general concept for studying convergence of invariants, generalizing e.g.\ the study of finite covering spaces for a given manifold as in the Lück approximation theorem. One of the main results was that for Benjamini-Schramm convergent, uniformly discrete sequences \(M_n=\Gamma_n\backslash{\widetilde M}\) one has convergence of normalized Betti numbers \(\frac{b_i(M_n)}{\mathrm{vol}(M_n)}\) towards the \(L^2\)-Betti numbers of \(\widetilde{M}\). A semisimple analytic group is a product \(G=\Pi_{i\in I}G_i\) of connected, simply connected, \(k_i\)-isotropic, almost \(k_i\)-simple, linear groups \(G_i\) over local fields \(k_i\), e.g. \(k_i=\mathbb{R}, \mathbb{C}\) or \(\mathbb{Q}_p\). It is said to have higher rank if either \(\vert I\vert\ge 2\) or \(\vert I\vert=1\) and \(rank(G_1)\ge 2\). In the latter case, the group has property T and thus satisfies the Stuck-Zimmer theorem which says that every nontrivial invariant random subgroup is given by a lattice. This was used by Abert et al. [loc. cit.] to show that for every sequence of pairwise non-conjugate lattices \(\Gamma_n\subset G\) in a higher rank, simple Lie group \(G\) the sequence \(\Gamma_n\backslash X\) Benjamini-Schramm converges towards the symmetric space \(X=G/K\). \textit{T. Gelander} and \textit{A. Levit} [Math. Ann. 372, No. 3--4, 1503--1544 (2018; Zbl 1404.22025)] showed that this Benjamini-Schramm convergence is equivalent to the condition that the sequence \(\Gamma_n\) is weakly central, that is, the \(G\)-invariant probability measure \(\eta_n\) on \(\Gamma_n\backslash G\) satisfies \(\lim_{n\to\infty}\eta_n(\left\{\Gamma_ng\colon g\Gamma_n g^{-1}\cap Q\subset Z(G)\right\})=1\) for every compact subset \(Q\subset G\). For \(\vert I\vert=1\) and \(\mathrm{rank}(G)\ge 2\) they proved this condition for general local fields. The paper under review handles the case of congruence lattices for \(\vert I\vert\ge 2\), thus extending (for congruence lattices) the result from the simple to the semisimple higher rank case. In this case, property T does in general not hold and the Stuck-Zimmer theorem is not known. Instead the author uses property \((\tau)\), which holds for the family of congruence lattices. Property \((\tau)\) is a relative version of property T, one of its equivalent formulations says that there is a uniform lower bound for the smallest positive eigenvalue of the Laplace-Beltrami operator. The proof then proceeds as follows: the sequence of invariant random subgroups \(\mu_{\Gamma_n}\) has a uniform spectral gap, denoting its weak-*-limit by \(\mu\) the author then shows that the \(G\)-action on \((\mathrm{Sub}(G),\mu)\) is irreducible (that is, each factor \(G_i\) acts ergodically) and essentially transitive, from which he concludes that \(\mu\) is the point mass on some central subgroup. From this the weakly centrality condition follows.