Equivariant cohomology for Hamiltonian torus actions on symplectic orbifolds (Q1928557): Difference between revisions

Given a group action of a compact torus \(T\) on a manifold \(M\), and a subtorus \(S\subset T\) acting locally freely, we obtain an induced \(R:=T/S\)-action on the orbifold \([M/S]\). Global quotients of this kind and their equivariant cohomological invariants are the objects studied in this paper. It is additionally assumed that \([M/S]\) is a symplectic orbifold in the sense of \textit{E. Lerman} and \textit{S. Tolman} [Trans. Am. Math. Soc. 349, No. 10, 4201--4230 (1997; Zbl 0897.58016)], and that the \(R\)-action is Hamiltonian. The authors consider \(R\)-equivariant cohomology \(H^*_R([M/S])\) of \([M/S]\), which they define as the \(T\)-equivariant cohomology of \(M\) (with \({\mathbb Z}\)- or \({\mathbb Q}\)-coefficients), and an \(R\)-equivariant version of Chen-Ruan orbifold cohomology. Concerning the first one, they prove that the inclusion of the \(R\)-fixed points into \([M/S]\) induces an injection in equivariant cohomology, and identify its image as the image of the equivariant cohomology of the \(1\)-skeleton, i.e., the union of one-dimensional \(R\)-orbits. (For integer coefficients, they assume connectedness of isotropy groups and primitivity of the isotropy weights.) The authors refer to this as a version of the GKM theorem, but historically it would be more appropriate to call it a generalization of a classical lemma proven by \textit{T. Chang} and \textit{T. Skjelbred} [Ann. Math. (2) 100, 307--321 (1974; Zbl 0249.57023), Lemma 2.3]. Concerning equivariant injectivity, they follow classical arguments using that a generic component of the momentum map is a Morse-Bott function on \(M\). The main difference is that the \(T\)-action on \(M\) does not have any fixed points, and that the critical set of the Morse-Bott function is given by lowest-dimensional orbits instead of fixed points. This situation was considered previously by the reviewer and \textit{D. Töben} [J. Topol. 3, No. 4, 819--846 (2010; Zbl 1208.55005)], where it is proven that the equivariant cohomology is, under the assumption of the existence of such a Morse-Bott function, a Cohen-Macaulay module, and that the Cohen-Macaulay property (everything for rational coefficients) is equivalent to the exactness of an Atiyah-Bredon-Sequence. This includes equivariant injectivity and the validity of a Chang-Skjelbred Lemma. The arguments in [loc. cit.] follow \textit{M. Franz} and \textit{V. Puppe} [C. R. Math. Acad. Sci., Soc. R. Can. 33, No. 1, 1--10 (2011; Zbl 1223.55003)], and Franz and Puppe on the other hand observe that a generalization of these statements holds for integer coefficients. In the second part of the paper the authors define an \(R\)-equivariant version of Chen-Ruan orbifold cohomology: \(M\) decomposes as a disjoint union of submanifolds of the form \(M^g\), where for \(g\in S\), \(M^g\) denotes the set of fixed points of \(g\). As \(T\) is commutative, \(M^g\) is invariant under the \(T\)-action, and one obtains an \(R\)-action on the orbifold given by the union of all \([M^g/S]\), called the inertia orbifold. As a vector space, the \(R\)-equivariant Chen-Ruan cohomology of \([M/S]\) is defined as the sum \(\bigoplus H^*_R([M^g/S])\), which in turn is defined as \(\bigoplus H^*_T(M^g)\), and the authors introduce on this space the structure of a rationally graded, associative \(H^*_T({\mathrm{pt}})\)-algebra. The authors want to prove a generalization of the equivariant injectivity theorem for \(R\)-equivariant Chen-Ruan cohomology, but as this cohomology theory is not functorial, this is not possible. Instead, they define a new ring \({\mathcal{NH}}^*_R(\nu[F/S])\) depending on the \(R\)-fixed point orbifold \([F/S]\) and the isotopy data at the fixed points, and prove an injectivity result using this ring. Finally, some computations for compact symplectic toric orbifolds are presented.