Equivariant cohomology of certain moduli of weighted pointed rational curves (Q276054)

Let \(M_{0,n}\) be the moduli space of \(n\)-pointed smooth curves of genus zero and \(\bar{M}_{0,n}\) its Deligne-Mumford compactification. \textit{B. Hassett} [Adv. Math. 173, No. 2, 316--352 (2003; Zbl 1072.14014)] introduced the notion of moduli space of \textit{weighted} stable pointed curves: this is a family of compactifications of \(M_{g,n}\), parametrized by choices of weights \(a_1,\ldots,a_n \in [0,1]\), such that a subset of the markings may coincide if and only if the total weight of these markings is at most \(1\). When all weights equal \(1\) one recovers the Deligne-Mumford compactification. The paper under review considers the ``weighted'' compactification of \(M_{0,m+n}\) in the case that the first \(m\) markings have weight \(1\) and the last \(n\) markings have weight \(\epsilon \ll 1\), so that any subset of the last \(n\) markings may coincide. This space is denoted he \(\bar M_{0,m| n}\). To get a Deligne-Mumford stack one assumes that \(m \geq 2\) and \(m+n \geq 3\). The main result is the calculation of the \(S_m \times S_n\)-equivariant Poincaré polynomial of \(\bar M_{0,m| n}\). When \(n=0\) this recovers a result of \textit{E. Getzler} [Prog. Math. 129, 199--230 (1995; Zbl 0851.18005)] and the method of proof is a generalization of Getzler's. Let me briefly summarize the idea of Getzler. First he calculates the \(S_n\)-equivariant cohomology of \(M_{0,n}\). Then he uses that \(\bar M_{0,n}\) has a stratification in which all strata are products of smaller moduli spaces \(M_{0,n_i}\), so that one basically needs to sum the contributions of all strata. He does this by encoding the data in terms of symmetric functions, using the bijection between symmetric functions and virtual representations of the symmetric group. Since the strata of \(\bar M_{0,n}\) correspond to trees, he needs to compute a sum over trees, which can then be encoded in terms of an operation on symmetric functions which is an analogue of the classical Legendre transform. In the situation of the paper under review, the author instead begins by computing the \(S_m \times S_n\)-equivariant Poincaré polynomial of \(M_{0,m | n}\), which denotes the Zariski open subset of \(\bar M_{0,m | n}\) parametrizing smooth curves. The equivariant Poincaré polynomials of \(M_{0,m | n}\) are now indexed instead by a bisymmetric function. Then \(\bar M_{0,m| n}\) has a stratification whose strata are all products of smaller spaces \(M_{0,m_i | n_i}\), and the strata are now indexed by \textit{bicolored} trees -- more specifically, trees whose legs and internal edges and legs can have two colors (corresponding to markings of weight \(1\) resp. \(\epsilon\)), where all internal edges correspond to weight \(1\). The operation of summing over such bicolored trees is interpreted by an operation that the author calls the ``partial Legendre transform'' of bisymmetric functions.