Properly convex bending of hyperbolic manifolds (Q784900)

Let \(\mathbb{RP}^d\) denote the \(d\)-dimensional real projective space and \(\mathrm{PGL}_{d+1}(\mathbb{R})\) denote the projective general linear group. A subset \(\Omega\) of \(\mathbb{RP}^d\) is called \textit{properly convex} if its closure is a convex set that is disjoint from some projective hyperplane. A \textit{properly convex} set \(\Omega\) is called \textit{strictly convex} if \(\partial\Omega\) is contains no non-trivial line segments. Given two \textit{properly convex} domains \(\Omega_1\) and \(\Omega_2\) it is possible to construct a new \textit{properly convex} domain \(\Omega_1\otimes\Omega_2\) via an obvious product construction. A \textit{properly convex} domain \(\Omega\) is called irreducible if the only way \(\Omega\) can be written as such a product is if one of the factors is trivial. To each \textit{properly convex} \(\Omega\) we can associate an automorphism group \[ \mathrm{PGL}(\Omega) = \{ A\in \mathrm{PGL}_{d+1}(\mathbb{R})\mid A(\Omega) = \Omega\} \] and we say that \(\Omega\) is homogeneous if \(\mathrm{PGL}(\Omega)\) acts transitively on \(\Omega\). The Klein model, \(\mathbb{H}^d,\) of hyperbolic space provides the quintessential example of a homogenous, irreducible, strictly convex domain, with automorphism group equal to the group Isom\((\mathbb{H}^d)\) of hyperbolic isometries. If \(\Gamma\subset \mathrm{PGL}(\Omega)\) is a discrete group then \(\Omega/\Gamma\) is a properly convex orbifold. A domain \(\Omega\) is called divisible (resp. quasi-dvisible) if there is a discrete group \(\Gamma\subset \mathrm{PGL}(\Omega\)) such that \(\Omega/\Gamma\) is compact (resp. finite volume). The main result of the paper under review is the following Theorem. For each \(d\geq 3\) there exists an irreducible, non-homogenous, quasi-divisible \(\Omega\subset\mathbb{RP}^d\). Furthermore, the domain \(\Omega\) can be chosen to be either strictly convex or non-strictly convex. The proof of the above theorem relies on deformation theoretic arguments. It starts with a finite volume hyperbolic \(d\)-manifold \(M\) that contains an embedded finite volume totally geodesic hypersurface \(\Sigma\). Then the authors realize \(\mathbb{H}^d\) as a strictly convex subset of \(\mathbb{RP}^d\) and thus they realize \(M\) as \(\mathbb{H}^d/\Gamma\) where \(\Gamma\) is a discrete subgroup of \(\mathrm{PSO}(d,1)\subset \mathrm{PGL}_{d+1}(\mathbb{R})\). Next using the bending construction of \textit{D. Johnson} and \textit{J. J. Millson} [Prog. Math. 67, 48--106 (1987; Zbl 0664.53023)], they produce a family \(\Gamma_t\subset \mathrm{PGL}_{d+1}(\mathbb{R})\) of subgroups such that \(\Gamma_0 =\Gamma\). Next, they apply arguments of the second author's paper [Enseign. Math. (2) 58, No. 1--2, 3--47 (2012; Zbl 1284.57021)] to conclude that for each \(t\) the group \(\Gamma_t\) preserves a properly convex domain \(\Omega_t\). Finally, a detailed analysis of the geometry of the cusps of \(\Omega_t/\Gamma_t\) allows them to conclude that \(\Gamma_t\) quasi-divides \(\Omega_t\) and can be either strictly convex or non-strictly convex (for different choices of \(M\) and \(\Sigma\)). The paper under review is in some sense a ``correction'' of the last cited paper [Zbl 1284.57021], where the first author has found some incorrectness. See Remark 1.2.