Splitting of Gysin extensions (Q5949799)

Let \(S^n\to X\to B\) be an orientable sphere bundle. Let \(H^*\) denote the usual cohomology theory where its coefficient ring is commutative and hereditary. This paper is concerned with a description of the \(H^*(B)\)-module structure of \(H^*(X)\); more precisely, the authors introduce a canonical Hochschild 3-cohomology class \([\theta(B)]\) which can be interpreted as the class of the Gysin extension of this sphere bundle \[ 0\to H^*(B)/ \bigl(cH^* (B) [n+1] \bigr)\to H^*(X)\to \text{Ann}_{H^*(B)} (c)[n]\to 0 \] in the extension group \(\text{Ext}^1_{H^*(B)} (\text{Ann}_{H^*(B)} (c)[n]\), \(H^*(B)/(cH^* (B) [n+1]))\). Here \(c\in H^{n+1}(B)\) is the characteristic class and \(\text{Ann}_{H^* (B)}(c)= \{x\in H^*(B) \mid cx=0\}\) is the right annihilator of \(c\) and \(R[k]\) denotes the shifted one of a graded module \(R\) such that \((R[k])^m =R^{m-k}\). It seems that the authors could obtain the idea how to define such a new cohomology class through their considerable study of the connections between Hochschild cohomology and \(A_\infty\)-structures on cohomology. In Sections 2 and 3 they describe a self-contained construction of this Hochschild 3-cocycle for the cohomology of a differential graded algebra and give an explicit correspondence of this to the relevant Gysin extension. The final section deals with two types of examples where \([\theta(B)]\) vanishes, so that the Gysin extension is split. One of them, for cohomology with rational coefficients, is that of sphere bundles where \(B\) is a space called formal and one more type, for cohomology with integral coefficients, is that of those where \(B\) is a torus. It is noted that there are many examples of formal spaces as simply-connected compact Kähler manifolds, complex smooth algebraic varieties, \((k-1)\)-connected compact manifolds of dimension less than or equal to \(4k-2\) \((k\geq 2)\), some flag varieties and some loop spaces. This paper is not only exactly but also readably written.