Equivariant bundles and isotropy representations (Q1049915)

This is a continuation of the authors' earlier work [Proc. Lond. Math. Soc., III. Ser. 86, No. 1, 250--272 (2003; Zbl 1032.55015)] on the classification of \(\text{SO}(n)\)-equivariant principal bundles over \(S^n\) in terms of their isotropy representations over the north and south poles. In this paper the authors present a classification theorem for \(\Gamma\)-equivariant \(G\)-bundles over a split \(\Gamma\)-CW-complex \(X\) by means of isotropy representations of the isotropy groupoid where \(G\) and \(\Gamma\) are compact Lie groups. Roughly speaking, if the projection \(X \to \Gamma\backslash X @>{\approx}>> A\) has a section \(\varphi : A \to X\), one calls the triple \((X, \pi, \varphi)\) a split \(\Gamma\)-space over \(A\). Then the isotropy groupoid is defined as the set \({\mathcal I}\) of all pairs \((\gamma, a)\) where \(a \in A\) and \(\gamma \in \Gamma_{\varphi(a)}\) with a natural multiplication. It is shown that such a split \(\Gamma\)-space is uniquely determined by \({\mathcal I}\) under reasonable assumptions (Proposition 2.1). Given a \(\Gamma\)-equivariant principal \(G\)-bundle \(p : E \to X\), for each \(x \in X\) there is a homomorphism \(\alpha_x : \Gamma_x \to G\) defined by \(\gamma\cdot\tilde{x}=\tilde{x}\cdot\alpha_x(\gamma)\) where \(\tilde{x}\in p^{-1}(x)\). In particular, when \(p : E \to X\) is split, i.e. the pull-back \(\varphi^*(p) : p^*(E) \to A\) is a trivial bundle, one finds that these homomorphisms produce a groupoid representation of \({\mathcal I}\) in \(G\). Denoting by \(\text{Rep}^G({\mathcal I})\) the space of conjugacy classes of such groupoid representations, the authors' classification theorem may be stated as follows: under suitable conditions on \(A\) and \({\mathcal I}\), the equivalence classes of split \(\Gamma\)-equivariant \(G\)-bundles over \(X\) are in bijection with \(\text{Rep}^G({\mathcal I})\) (Theorems 3.2 and 4.5). The latter part of this paper is devoted to reducing \(\text{Rep}^G({\mathcal I})\) to the corresponding set \(\text{Rep}^G_{\text{cell}}({\mathcal I})\) of cellular representations assuming \({\mathcal I}\) is cellular. This approach has the benefit to alleviate the computation of \(\text{Rep}^G({\mathcal I})\) by considering the restriction to the 0- or 1-skeleton (Proposition 5.1). Particular attention is paid here to the cases where \(G\) is abelian and where \(A\) is a graph. For abelian \(G\), as a corollary one obtains a homotopy-theoretic description of \(\text{Rep}^G({\mathcal I})\) by \textit{R. K. Lashof, J. P. May} and \textit{G. B. Segal} [Contemp. Math. 19, 167--176 (1983; Zbl 0526.55020)]. But here this result is strengthened in more detail, which states that \(\text{Rep}^G({\mathcal I}) \approx [B{\mathcal I}, BG]\) when assuming \(H^1(A; \pi_0(G))=H^2(A; \mathbb{Z})=0\) (Corollary 6.4).