The congruence topology, Grothendieck duality and thin groups (Q2324668): Difference between revisions

Let \(\Gamma\) be a finitely generated subgroup of \(\mathrm{GL}_n(\mathbb{Z})\). For a representation \(\rho: \Gamma\to \mathrm{GL}(V)\) on a lattice \(V\), let \(\hat{\rho}:\hat{V}\to \mathrm{GL}(\hat{V})\) be the natural continuous extension to the profinite completions. Let \(\mathrm{Cl}_{\rho}(\Gamma)=\{g\in\hat{\Gamma}\,:\,\hat{\rho}(g)(V)\subseteq V\}\). In [Am. J. Math. 102, 663--689 (1980; Zbl 0449.20051)] the first author proved that \(\mathrm{Cl}(\Gamma):=\bigcap_{\rho}\mathrm{Cl}_{\rho}(\Gamma)\) coincides with the group \(\mathrm{Cl}_{\mathbb{Z}}(\Gamma)\), which is essentially defined in [\textit{A. Grothendieck}, Manuscr. Math. 2, 375--396 (1970; Zbl 0239.20065)] and called the \textit{Grothendieck closure} of \(\Gamma\). The main theorem of the paper under review asserts that the equality \(\Gamma=\mathrm{Cl}(\Gamma)\) holds if and only if \(\Gamma\) has the so-called \textit{congruence subgroup property}. As a corollary, Conjecture A of [Lubotzky, loc. cit.] is confirmed. That is, if \(\Gamma\) is a congruence subgroup of \(G(\mathbb{Q})\), where \(G\) is a simply connected semisimple algebraic group over \(\mathbb{Q}\), then \(\mathrm{Cl}(\Gamma)=\Gamma\) happens if and only if \(\Gamma\) has the strict congruence subgroup property. In particular, if \(\Gamma=\mathrm{SL}_2(\mathbb{Z})\) or \(\Gamma\) is a finitely generated free group on at least two generators, then \(\mathrm{Cl}(\Gamma)\neq\Gamma\). A classical theorem of Chevalley says that a homomorphism of complex algebraic groups sends a closed subgroup to a closed subgroup. In the course of proving their main theorem, the authors obtained an analogue of Chevalley's theorem for congruence subgroups. The techniques in this paper also have a very interesting application to thin subgroups of \(\mathrm{GL}_n(\mathbb{Z})\). Recall that \(\Gamma\) is called a thin subgroup of \(\mathrm{GL}_n(\mathbb{Z})\) if it is of infinite index in \(G\cap\mathrm{GL}_n(\mathbb{Z})\), where \(G\) is the Zariski closure of \(\Gamma\) in \(\mathrm{GL}_n\). Theorem\;0.10 of the present paper provides a purely group theoretic criterion for the thinness: \(\Gamma\) is thin if and only if \(\Gamma\) does not have a finite index subgroup with an infinite abelianization or \(\mathrm{Cl}(\Gamma)/\Gamma\) is not compact.