Vector fields, torus actions and equivariant cohomology (Q952987): Difference between revisions

Suppose that \(X\) is a compact Kähler manifold and \(V\) is a holomorphic vector field on \(X\) whose zero set \(Z\) is nonempty. A result of Carrell and Lieberman shows that \(H^p(X, \Omega^q)=0\) if \(|p-q|>\dim Z\) where \(\Omega^q_X\) is the sheaf of holomorphic \(q\)-forms on \(X\), and there is a filtration of \(H^*(Z, \mathbb C)\) such that \(\text{Gr}H^*(Z, \mathbb C)\cong H^*(X, \mathbb C)\) as graded algebras. But this result does not give any insight into how the filtration can be described. The main goal of this paper is to give an explicit connection between this filtration and the \(\mathbb C^*\)-equivariant cohomology of \(X\) provided \(X\) is a compact Kähler manifold with a \(\mathbb C^*\)-action and \(V\) is a holomorphic vector field generated by this \(\mathbb C^*\)-action. More explicitly, the contraction operator \(i_V\) defines a complex of sheaves \[ 0\rightarrow \Omega^n_X \rightarrow \Omega^{n-1}_X \rightarrow \cdots \rightarrow \Omega^1_X \rightarrow \Omega^0_X \rightarrow 0 \] where \(n=\dim X\). Define \(\mathcal{H}^s(X)=\bigoplus_{q-p=s}H^{p, q}(X)\), and \(K^s_X=\bigoplus_{q-p=s}A^{p, q}\) for \(s\in \mathbb Z\) where \(A^{p, q}\) is the collection of smooth forms of \(X\) of type \((p, q)\). Then 1. \(\dim H^s(K^*_X)=\dim \mathcal{H}^s(X);\) 2. there exists a \(\mathbb C\)-algebra isomorphism \[ H^s(K^*_X)\cong \mathcal{H}^s(Z); \] 3. \[ \sum_{q-p=s}\dim H^{p, q}(X)=\sum_{q-p=s}\dim H^{p, q}(Z); \] 4. there exists a filtration of \(\mathcal{H}^*(Z)\) that yields an isomorphism of graded rings \[ \bigoplus_s\mathcal{H}^s(X)\cong \text{Gr}\mathcal{H}^*(Z) \] The proof goes as following. Let \(T\) denote the compact torus in \(\mathbb C^*\) and suppose that \(H^*_T(X, \mathbb C)\) denotes the \(T\)-equivariant cohomology of \(X\) over \(\mathbb C\). Then as a \(\mathbb C[t]\)-module, \(H^*_T(X, \mathbb C)\cong \mathbb C[t]\otimes H^*(X, \mathbb C)\). Teleman and Lillywhite defined Dolbeault equivariant cohomology groups \(H^{p, q}_{T, \partial}(X)\) and showed that \(H^*_T(X, \mathbb C)\) admits the usual Hodge decomposition \[ H^k_T(X, \mathbb C)=\bigoplus_{p+q=k}H^{p, q}_{T, \partial}(X) \] Then we can define the groups \(\mathcal{H}^s_T(X)\) analogous to the groups \(\mathcal{H}^s(X)\) defined above. They showed that evaluation polynomials at \(t=1\) gives a map of \(\mathbb C\)-algebras \(\mathcal{H}^*_T(X, \mathbb C) \rightarrow H^*(K^*_X)\). The filtration of \(H^*(K^*_X)\) is the image of a canonical filtration of \(\mathcal{H}^*_T(Z, \mathbb C) \rightarrow H^*(K^*_Z)=\mathcal{H}^*(Z)\) via the evaluating map. This nice result reduces the complexity of calculating the cohomology of \(X\) to the cohomology of \(Z\), and the filtration gives more Hodge theoretic-information about the cohomology of \(X\).