Non-residually finite extensions of arithmetic groups (Q2316315)

The author shows that there are many arithmetic groups admitting finite extensions which are not residually finite. Inspired by an old result of Deligne's he shows that if \(\Gamma\) is an arithmetic group in an algebraic \(\mathbb Q\)-group \(G\) which is simple and satisfies strong approximation as well as the congruence subgroup property, satisfies \[\overline H^2(\mathbb Z/n\mathbb Z) := \lim_\to H^2(\Delta, \mathbb Z/n\mathbb Z)\neq\{0\}\] for every natural \(n,\) the limit being taken over all congruence subgroups. One therefore has a finite index subgroup \(\Delta\) of \(\Gamma\) such that the Shapiro-Lemma gives an extension \(\tilde \Gamma\) of \(\Gamma\) which contains a finite index subgroup projecting to a non-split extension \(\tilde \Delta\) of \(\Delta\) by \( \mathbb Z/n\mathbb Z,\) contradicting residual finiteness. The author then proceeds to study for a fixed prime \(\ell\) the limit of \(\overline H^2(\mathbb Z/\ell^k\mathbb Z).\) If \(\widehat{G(\mathbb Q)}\) denotes the arithmetic completion of \(G(\mathbb Q)\) he shows that the \(\widehat{G(\mathbb Q)}\)-invariant elements in this limit are a finitely generated free \(\mathbb Z_\ell\)-module and can calculate its rank in interesting special cases. This nicely written and interesting paper gives a much more complete account on this facet of the cohomology of arithmetic groups than the results outlined above. It makes use of continuous cohomology throughout and ends with an appendix on the cohomology of compact symmetric spaces.