When are two Coxeter orbifolds diffeomorphic? (Q2512541): Difference between revisions

An \(n\)-orbifold \(\mathcal{Q}\) is \textit{of reflection type} if it is locally modeled on quotients of \(\mathbb{R}^n\) by finite linear reflection groups. The underlying space of a smooth orbifold of reflection type has naturally the structure of a smooth manifold with corners, which is also \textit{nice}, i.e. each stratum of codimension two is contained in the closure of exactly two strata of codimension one. Suppose that each codimension \(k\) stratum of \(\mathcal{Q}\) is the intersection of \(k\) codimension one strata. The author defines a \textit{Coxeter orbifold} to be an orbifold of reflection type \(\mathcal{Q}\) for which one of the following successively stronger conditions hold: (I) Each stratum of \(\mathcal{Q}\) is a compact contractible manifold. (II) Each stratum of \(\mathcal{Q}\) is homeomorphic to a disk. (III) As a manifold with corners, \(\mathcal{Q}\) is isomorphic to a simple convex polygon. A Coxeter orbifold \(\mathcal{Q}\) is said to be of type (II) or type (III) if it satisfies the corresponding condition. The orbifold structure on \(\mathcal{Q}\) induces a labeling of each codimension two stratum by an integer \(m \geq 2\) indicating the order \(2m\) of the local dihedral group along the stratum, and such that it determines a finite Coxeter group of rank \(k\) on each codimension \(k\) stratum. The labeling on the codimension two strata of \(\mathcal{Q}\) becomes a \textit{proper} labeling on the edges of the \textit{nerve} \(N(\mathcal{Q})\), the dual simplicial complex to \(\mathcal{Q}\). A \textit{combinatorial equivalence} between two Coxeter orbifolds is a label-preserving simplicial isomorphism between their nerves. In the paper under review the author answers the question \textit{``When are two combinatorially equivalent Coxeter orbifolds diffeomorphic?'' } For a general Coxeter \(n\)-orbifold \(\mathcal{Q}\) of type (I), the nerve \(N(\mathcal{Q})\) is a \(\text{GHS}^{n-1}\) (generalized homology \((n-1)\)-sphere), so the question can be reformulated as follows: given a \(\text{GHS}^{n-1}\) simplicial complex \(N\) with a proper labeling on its edges, does there exist a unique (up to isotopy) Coxeter orbifold \(\mathcal{Q}\) such that \(N(\mathcal{Q})=N\)? In the topological category, the existence part follows from the author's work in [Ann. Math. (2) 117, No. 2, 293--324 (1983; Zbl 0531.57041)] and the uniqueness is a consequence of the fact that the Poincaré conjecture and the topological \(h\)-cobordism theorem hold in all dimensions. In the smooth category, the existence part holds, again as shown by the author in [loc. cit.], provided each three-dimensional link in \(N\) bounds a smooth contractible \(4\)-manifold. In the paper under review, the author proves that two such smooth Coxeter orbifolds are diffeomorphic if each of their four-dimensional faces are diffeomorphic rel boundary. Furthermore, using a result of \textit{S. Akbulut} [J. Differ. Geom. 33, No. 2, 335--356 (1991; Zbl 0839.57015)], the author shows that in each dimension \(\geq 4\) there are smooth Coxeter orbifolds which are combinatorially equivalent but not diffeomorphic, so the hypothesis of diffeomorphic four-dimensional faces is necessary. As a consequence the author proves that combinatorially equivalent Coxeter orbifolds of type (III) are necessarily diffeomorphic (the nerve of a type (III) Coxeter orbifold \(\mathcal{Q}\) is the boundary complex of the simplicial polytope that is dual to \(\mathcal{Q}\)). As pointed out in the paper, the fact that combinatorially equivalent polytopes are diffeomorphic as manifolds with corners was also independently established by \textit{M. Wiemeler} [Math. Z. 273, No. 3--4, 1063--1084 (2013; Zbl 1269.57014)]. As applications, the author obtains smooth equivariant rigidity results for reflection groups and locally standard torus actions.