The definability criteria for convex projective polyhedral reflection groups (Q2339900)

Let \(V\) be an \(\mathbb{F}\)-vector space of dimension \(n + 1\), with \(\mathbb{F}\) a field of characteristic \(0\). Let \(SL^{\pm}(n + 1, \mathbb{F})\) be the group of linear automorphisms of \(\mathbb{F}\) of determinant \({\pm}1\). Let \(\Gamma \subset SL^{\pm}(n + 1, \mathbb{F})\) be a subgroup generated by reflections. The question that the paper under review examines is when the group \(\Gamma\) is definable over an integrally closed Noetherian subring \(\mathbb{A}\) of \(\mathbb{F}\). When \(\mathbb{F} = \mathbb{R}\), let \(\mathbb{S}^n\) be the \(n\)-dimensional real projective sphere. Then \(SL^{\pm}(n + 1, \mathbb{R})\) is the group of projective automorphisms of \(\mathbb{S}^n\). The automorphism of \(\mathbb{S}^n\) induced by a reflection on \(\mathbb{R}^{n+1}\) is called a projective reflection. The main result in this case is the following. Let \(\Gamma\) be a polyhedral reflection group dividing a strictly convex open domain \(\Omega \subset \mathbb{S}^n\). That means that \(\Gamma\) is generated by reflections about the faces of a polyhedron in \(\mathbb{S}^n\), \(\Omega\) is a subset of \(\mathbb{S}^n\) that is given by a linear inequality, its closure does not contain antipodal points and its boundary contains no nontrivial segments. The properly discontinuous action of \(\Gamma\) on \(\Omega\) is called dividing if the quotient space is compact and Hausdorff. Let \(\mathbb{A}\) be an integrally closed Noetherian subring of \(\mathbb{F}\). Let \(\Delta\) be a subgroup of \(\Gamma\) of finite index. Then the main result states that \(\Gamma\) is \(\mathbb{A}\)-definable if and only if the same is true for \(\Delta\). In this context, the group \(\Gamma\) is \(\mathbb{A}\)-definable if \(V\) contains an \(\mathbb{A}\)-lattice that is \(\Gamma\)-invariant. Vinberg already has proved that \(\Gamma\) is \(\mathbb{A}\)-definable if and only if all the cyclic products of the entries of the Cartan matrix of \(\Gamma\) are in \(\mathbb{A}\) [\textit{E. B. Vinberg}, Math. USSR, Izv. 5, 45--55 (1972; Zbl 0252.20043)]. The authors prove a more general result. Let \(\mathbb{F}\) be an algebraically closed field, \(\mathbb{A}\) an integrally closed Noetherian subring of \(\mathbb{F}\) and \(\mathbb{K}\) the field of fractions of \(\mathbb{A}\). Let \(\Gamma \subset SL^{\pm}(n + 1, \mathbb{F})\) be an \(\mathbb{F}\)-Zariski-dense subgroup generated by finitely many elements such a generator either has nonzero trace in \(\mathbb{K}\) or it has finite order prime to \(n + 1\). Let \(\Delta\) be a subgroup of \(\Gamma\) of finite index. Then \(\Gamma\) is \(\mathbb{A}\)-definable if and only if the same is true for \(\Delta\). The first result follows from the second because the Zariski closure \(\bar{\Gamma}\) over \(\mathbb{R}\) is either \(O^+(n, 1)\) or \(SL^{\pm}(n + 1, \mathbb{R})\). The authors apply their results to study when certain Coxeter groups that are determined by polytopes of small dimensions are \(\mathbb{Z}\)-definable. In particular they look at hyperbolic triangle Coxeter groups and certain tetrahedral and octahedral hyperbolic Coxeter groups.