Module derivations and cohomological splitting of adjoint bundles (Q4811320)

Let \(G\) be a connected finite loop space, \(P\to M\) a principal \(G\)-bundle over a connected space \(M\) and \(P\times_{ad}G\to M\) be the associated adjoint bundle. In this well written article the authors study systematically under which conditions this adjoint bundle admits a \(\text{mod}-p\) cohomological splitting and to what extend this splitting is compatible with the action of the \(\text{mod}-p\) Steenrod algebra. They show that in the case that \(p= 0\) or that \(H^*(G,\mathbb{Z})\) has no \(p\)-torsion, the adjoint bundle always admits a \(\text{mod}-p\) cohomological splitting. If \(p= 2\) and the cohomology \(H^*(BG,\mathbb{F}_2)\) is a polynomial algebra with odd degree generators, they show that the obstruction for a \(\text{mod}-p\) cohomological splitting of the adjoint bundle is an element of the Hochschild homology \(HH_*(H^*(BG,\mathbb{F}_2)\), \(H^*(M,\mathbb{F}_2))\). Using the notion of module derivation introduced by \textit{K. Kuribayashi} [J. Math. Kyoto Univ. 39, No. 1, 67--85 (1999; Zbl 0938.55014)] it is then shown that these splittings are in general not compatible with the action of the \(\text{mod}-p\) Steenrod algebra. Several interesting examples are discussed in detail, including the classifying space of the loop group \(LG\) and \(\text{SU}(n)\)-adjoint bundles over a 4- and 5-dimensional CW complex.