Obstruction theory on 8-manifolds (Q958178)

This paper gives a uniform approach to obstruction-theoretic problems on 8-manifolds, namely the problem of reducing of the structure group of 8-vector bundles and the problem of obtaining necessary and sufficient cohomological criteria for the existence of complex and quaternionic structures in such bundles. Let \(\xi\) be an 8-dimensional oriented real vector bundle over a closed connected smooth \(\text{spin}^c\)-manifold \(M\) of the same dimension. The authors study the reduction of the structure group \(\text{SO(8)}\) of \(\xi\) to \(G\) via \(\rho\) for the following compact Lie groups \(G\) and homomorphisms \(\rho : G \to \text{SO(8)}\): (i) the standard inclusion \(\text{U(4)} \subset \text{SO(8)}\); (ii) the homomorphism \(\text{U(3)} \to \text{SO(8)}\) determined by the adjoint representation of \(\text{PSU(3)}\); (iii) the composition of the standard inclusions \(\text{Sp(2)}\cdot\text{U(1)} \subset \text{U(4)} \subset \text{SO(8)}\); (iv) the standard inclusion \(\text{Sp(2)}\cdot\text{Sp(1)} \subset \text{SO(8)}\); (v) the composition of the canonical projection and the standard inclusion \(\text{Spin}^c(k) \to \text{SO(\textit{k})} \subset \text{SO(8)}\) for \(k=6, 5, 4, 3\). There are five papers by the first and third author concerning the reduction to these groups listed in the references, but this work covers a broader class of vector bundles than is discussed there. The main result of this paper consists of a number of propositions which, as is observed below, are obtained in terms of cohomology characteristic classes of \(\xi\) and \(M\). For example the first of these propositions (Proposition 4.1) considers an 8-vector bundle \(\xi\) admitting a \(\text{spin}^c\)-structure. But for the sake of simplicity assume here that \(w_2(\xi)=0\) and \(w_2(M)=0\). Then this proposition states that the structure group \(\text{SO}(8)\) of \(\xi\) admits a reduction to \(\text{U}(4)\) if and only if the relations hold that \(1/2(p_2(\xi)+q_1(\xi)q_1(\tau M))[M]\equiv q_1^2(\xi) \;(\text{mod} \;2)\) and \(p_2(\xi)-q^2_1(\xi)=2e(\xi)\). Here \(q_1(\xi)\) denotes a spin characteristic class of \(\xi\) induced by a universal class in \(H^4(B\text{Spin}(\infty); \mathbb{Z})\) corresponding to \(-c_2\in H^4(B\text{SU}(\infty); \mathbb{Z})\). The basic idea is to use the \(\text{spin}^c\) index \(K^0(M) \to \mathbb{Z}\). It gives rise to a splitting of \(K^0(M)\) as a direct sum \(K^0(M-B)\oplus \mathbb{Z}\), where \(B\subset M\) denotes an open cell of dimension 8, from which the corresponding one of \(KO^0(M)\) also can be obtained. This means that \(\xi\) is determined stably by its restriction to \(M-B\) and the image of \([\xi]\in KO^0(M)\) under the \(\text{spin}^c\) index. The complex type splitting allows one to extend a fundamental description of the Chern classes of complex bundles over \(M-B\) to bundles over \(M\). The real one is used to distingush oriented real vector bundles over \(M\). The above sample proposition is in fact an immediate consequence of necessary and sufficient conditions for the existence of a complex structure that can be derived from such observation. The corresponding arguments for the almost quaternionic structure groups \(\text{Sp(2)}\cdot\text{U(1)}\) and \(\text{Sp(2)}\cdot\text{Sp(1)}\) are given in Sections 7, 8 and 10, and their results are summarized in three propositions (Propositions 8.2, 10.1 and 10.4). The remaning cases (ii) and (v) are discussed in Sections 5, 6 and 11.