Invariant random subgroups of linear groups (Q2357006): Difference between revisions

Let \(\Gamma\) be a countable group. \(\text{Sub}(\Gamma)\), the collection of all subgroups of \(\Gamma\), is a compact metrizable space under the Chabauty topology. An invariant random subgroup (IRS) of \(\Gamma\) is by definition a Borel probability measure on \(\text{Sub}(\Gamma)\) that is invariant w.r.t.\ \(\Gamma\) acting by conjugation. For example, if \(\Gamma\) acts on a probability space by measure preserving transformations, then the distribution of the stabilizer of a random point defines an IRS. The main result of the paper under review is somewhat reminiscent of the Tits alternative. A simple version, given as Theorem 1.14, is concerned with the case in which \(\Gamma\) is non-amenable, and \(\chi\) is a linear representation of \(\Gamma\) whose image has simple, center free, Zariski closure. \(\chi\) gives rise to a subclass \(\text{IRS}^{\chi}(\Gamma)\), consisting of those \(\mu\) for which almost all samples have nontrivial image under \(\chi\). Then there is a topology on \(\Gamma\) such that \(\mu\)-almost every \(\Delta \in \text{Sub}(\Gamma)\) is open for every \(\mu \in \text{IRS}^{\chi}(\Gamma)\). Furthermore, \(\Gamma\) contains a proper free subgroup \(F\) such that for every \(\mu \in \text{IRS}^{\chi}(\Gamma)\) one has that \(F \cdot \Delta = \Gamma, F \cap \Delta \neq 1\) for \(\mu\)-almost all random subgroups \(\Delta\), and \(\Delta \mapsto \Delta \cap F\) is an isomorphism of the probability spaces \((\text{Sub}(\Gamma), \mu)\) and \((\text{Sub}(F), \mu_{|F})\).