Peripheral separability and cusps of arithmetic hyperbolic orbifolds (Q1880649)

This paper deals with the following question. Given an almost flat manifold, is it the cusp cross-section of an arithmetic \(K\)-hyperbolic orbifold with finite volume? Here \(K\) denotes the (skew) field of real, complex, quaternionic or octonionic numbers. The real hyperbolic case has been studied by \textit{D.\ Long} and \textit{A.\ Reid} [Algebr. Geom. Topol. 2, 285-296 (2002; Zbl 0998.57038)]. A cusp cross-section of a complex hyperbolic orbifold of dimension \(n+1\) must be modelled on the \((2 n+1)\)-Heisenberg group. Among other results, the author shows that for any \(n\geq 2\), there are infinite families of closed manifolds modelled on the \((2 n+1)\)-Heisenberg group which are not cusp-cross sections of complex hyperbolic orbifolds. He also proves the analogue in the quaternionic setting. However, he shows that every nil 3-manifold is a cusp-cross section of an arithmetic complex 2-orbifold. The previous question is related to subgroup separability, which is an interesting question on its own. The author shows that for arithmetic groups in a \(k\)-hyperbolic space, peripheral subgroups are separable (as a particular case of a theorem on Borel subgroups of arithmetic lattices). Using this result, he establishes a nice criterion to decide whether a manifold is the cusp cross-section of an arithmetic orbifold or not.