Twisted limit formula for torsion and cyclic base change (Q2011986)

Let \(G\) be a semisimple Liegroup and \((\Gamma_n)\) a sequence of normal finite index subgroups inside a cocompact torsion-free lattice \(\Gamma\) in \(G\) which intersect in \(\{1_G\}.\) The authors of the paper under review study the growth of torsion in certain cohomology groups \(H^j(\Gamma_n, {\mathcal O})\) as \(n\) goes to infinity. Corresponding questions for the growth of Betti-numbers with trivial complex coefficients have been studied by various authors and have been related to \(L^2\)-Betti-numbers, which provides asymptotic information if the fundamental rank \(\delta(G) = \text{rank}_{\mathbb C}(G) - \text{rank}_{\mathbb C}(K)\) is zero, where \(K\) denotes a maximal compact subgroup of \(G.\) Only in this situation the \(L^2\)-Betti-numbers can be nonzero. Turning to torsion, let \(\rho : G\to \text{Aut}(F)\) be a finite-dimensional representation on a real vector space and assume that \(F\) contains a \(\Gamma\)-invariant lattice \({\mathcal O}.\) Then under a technical assumption (strong acyclicity of \(\rho\)), the first author and \textit{A. Venkatesh} showed in [J. Inst. Math. Jussieu 12, No. 2, 391--447 (2013; Zbl 1266.22013)] that \(\sum_j \frac{(-1)^j \log|H^j(\Gamma_n, {\mathcal O})_{\text{tors}}|} {[\Gamma:\Gamma_n]} \to c(G,\rho)\) as \(n\) goes to \(\infty,\) where \(c(G,\rho)\) is a constant independent of \(\Gamma\) and \({\mathcal O}\). This constant is nonzero iff \(\delta(G) = 1,\) and in that case the size of the torsion grows exponentially with the covolume of \(\Gamma_n.\) The present authors go one step further, considering the situation that \(G\) is the group of real points in \(\text{Res}_{\mathbb E / \mathbb R} ({\mathbf G}),\) where \({\mathbf G}\) is a connected, semi-simple, quasisplit real algebraic group and \(\mathbb E\) a cyclic étale extension of \(\mathbb R\) with a fixed generator \(\sigma\) of the Galois group. They need to assume that \(\sigma\) has prime order and \(H^1(\langle \sigma\rangle, G) = \{1\}.\) Of course, \(\Gamma_n\) has to be \(\sigma\)-invariant for every \(n\) and \(\rho\) has to extent to a representation of the semi-direct product \(G\rtimes \langle \sigma\rangle\) and again has to be strongly acyclic. They then proof a general result relating the size of the torsion with the covolume of the \(\sigma\)-fixed part \(\Gamma^\sigma\) inside \(G^\sigma\) and then, for example, produce results of towers of congruence subgroups \(\Gamma_n\) in certain Lie-groups, where the size of the torsion grows exponentially with the square-root of the covolume. One of the main ingredients in the proofs is an equivariant Cheeger-Müller-theorem proved by \textit{M. Lipnowski}, [J. Inst. Math. Jussieu 15, No. 1, 165--202 (2016; Zbl 1335.58021)], allowing to perform a kind of base-change.