A normal subgroup theorem for commensurators of lattices (Q473144)

Margulis's celebrated super-rigidity and normal subgroup theorems in the 1970's led to a deluge of sorts. The rigidity phenomenon has been generalized vastly to realms more general than lattices in higher rank semi-simple Lie groups. One theme that has evolved is that several such results hold for lattices with dense commensurators. Many rigidity phenomena have been proved -- notably, see two of the the second author's works [Invent. Math. 141, No. 1, 1--54 (2000; Zbl 0978.22010); Ann. Math. (2) 152, No. 1, 113--182 (2000; Zbl 0970.22011)], and the paper [Comment. Math. Helv. 69, No. 4, 523--548 (1994; Zbl 0839.22011)] by \textit{A. Lubotzky} et al.. However, the normal subgroup theorem itself has resisted generalizations. In 2005, U. Bader and the second author proved a normal subgroup theorem for irreducible lattices in products of at least two locally compact, compactly generated groups [\textit{U. Bader} and \textit{Y. Shalom}, Invent. Math. 163, No. 2, 415--454 (2006; Zbl 1085.22005)]. The important paper under review here demonstrates an analogous theorem for lattices in a single group which admit dense commensurators and the result is free of rank conditions. The approach is similar in spirit to Margulis's original one and has two parts -- one involving property T and the other involving amenability. These two ``mutually opposing'' properties imply the result. The steps involve deep analytic techniques. The main result stated below can be regarded as the commensurator analogue of the normal subgroup theorem for lattices. A more precise statement of the main result is: Let \(G\) be a locally compact, second countable, compactly generated group that is not a compact extension of an abelian group. Let \(\Gamma < G\) be a finitely generated square-integrable lattice, which is contained in and, is commensurated by, a dense subgroup \(\Lambda\) of \(G\). If every closed, normal, non-cocompact subgroup of \(G\) has finite intersection with \(\Lambda\), then every infinite normal subgroup \(N\) of \(\Lambda\) contains a subgroup of finite index in \(\Gamma\). Consequently, if \(\phi : \Lambda \rightarrow H\) is a dense homomorphism into a locally compact, totally disconnected group \(H\) such that \(K := \overline{\phi(\Gamma)}\) is compact open and \(\Gamma = \phi^{-1}(K)\), then there is a natural bijection between commensurability classes of infinite normal subgroups of \(\Lambda\) and commensurability classes of open, normal subgroups of \(H\). There exists subgroups \(H\) having the property mentioned above; a minimal such group is the so-called relative profinite completion \(\Lambda//\Gamma\) of \(\Lambda\) with respect to \(\Gamma\). The notion of square-integrable lattices referred to in the above theorem generalizes co-compact lattices. The theorem also shows that for any \(H\) as above, the property that all infinite, normal subgroups of \(\Lambda\) have finite index holds good if, and only if, the analogous property holds for all open, normal subgroups of \(H\). The condition that \(G\) not be a compact extension of an abelian group is necessary as the authors point out. In the example \(\Gamma = \mathbb{Z}^n < \Lambda =\mathrm{SO}(n,\mathbb{Q}) \ltimes (\mathbb{Q}(\sqrt{2}))^n < G =\mathrm{SO}(n,\mathbb{R}) \ltimes \mathbb{R}^n\), the normal subgroup \(N := \sqrt{2} \mathbb{Q}^n < \Lambda\) intersects \(\Gamma\) trivially. The authors point out that the main theorem could motivate a study of the exotic locally compact group \(\Lambda//\Gamma\).