Non-negatively curved GKM orbifolds (Q2069663)

In this paper the authors study isometric torus actions of \(\mathrm{GKM}\) type on Riemannian manifolds and orbifolds of non-negative sectional curvature. The main characteristic of this kind of actions is that one can construct a graph that encodes many interesting informations, called the \(\mathrm{GKM}\) graph. The main results of the paper give restrictions about the combinatorial type of \(\mathrm{GKM}\) graphs of orbifolds (and manifolds in particular) admitting a metric with non-negative sectional curvature and a \(\mathrm{GKM}_4\) isometric action. Precisely, as proved in one of the main theorems of the paper, the \( \mathrm{GKM} \) graphs of these orbifolds are finitely covered by the vertex-edge graph of a finite product of very simple combinatorial objects, (see Theorem 1.1 of the paper for more details). Recall that, for \(k \ge 2\), a \(\mathrm{GKM}_k\) action on a closed oriented manifold \(M\) is an effective action of a torus \(T=T^m=(\mathbb{S}^1)^m\) on \(M\) such that: \begin{itemize} \item the fixed points are a finite set \item for all \(p < k\), the \(p\)-skeleton of the action \[ M_p=\{x \in M \mid \dim T(x) \le p \} \] can be expressed as the union of \(2p\)-dimensional invariant submanifolds. \end{itemize} If we take \(p=1\), this is saying that the \(1\)-skeleton \(M_1\) is a union of \(2\)-dimensional invariant submanifolds. From this, one can construct the so called \(\mathrm{GKM}\) graph of the action: the vertices are the fixed points of the action, and an edge is added for each \(2\)-sphere in \(M_1\) which contains two fixed points. The edges are then labelled with the weights of the isotropy representation in the tangent space to the sphere at any of the two fixed points. The relevance of this kind of actions comes from the fact that, under the assumption that the odd-dimensional rational cohomology of \(M\) vanishes, one can recover the equivariant cohomology \(H^*_T(M)\) of \(M\) from the \(\mathrm{GKM}\) graph. In addition, under this assumption, one can also recover tha rational cohomology \(H^*(M)\) from \(H^*_T(M)\). Hence, if one is able to find restrictions on the combinatorial structure of the \(\mathrm{GKM}\) graph, one obtains as a corollary restrictions about the rational cohomology of \(M\). For instance, the authors obtain the following corollary about the Betti numbers: \textbf{Corollary:} Let \(O\) be a \(2n\)-dimensional closed and oriented Riemmannian orbifold with non-negative curvature. Assume that it admits a \(\mathrm{GKM}_4 \) action. Then the total Betti number \(b(O)=\sum_i b_i(O)\) is either \(\le 3 \cdot 2^{n-2}\) or equal to \(2^n\).