SO(3)-structures on 8-manifolds (Q1941073)

The authors aim to characterize \(\mathrm{SO}(3)\)-structures on 8-dimensional Riemannian manifolds, for which each tangent space splits into two summands of dimensions three and five, irreducible under \(\mathrm{SO}(3)\). The main tool is a fixed representation of \(\mathrm{SO}(3)\) on \(\mathbb{R}^8\), with image \(\mathcal{G}\) contained in the Lie subgroups \(\mathrm{SO}(3)\times\mathrm{SO}(5)\), \(\mathrm{PSU}(3)\), \(\mathrm{Sp}(2)\mathrm{Sp}(1)\) of \(\mathrm{SO}(8)\). The paper reveals the relationship between \(\mathcal{G}\) and almost quaternion-Hermitian, almost product, and \(\mathrm{PSU}(3)\)-geometry, and proves how the underlying \(\mathcal G\)-structures determine a \(\mathcal{G}\)-action. Proposition 2.1 shows that a certain \(\mathcal{G}\)-structure on an 8-dimensional Riemannian manifold induces an almost product structure, a \(\mathrm{PSU}(3)\)-structure and an almost quaternion-Hermitian structure. According to Theorem 3.1, \(\mathcal{G}\) can be recovered by intersecting only two of the three subgroups of \(\mathrm{SO}(8)\) mentioned above. An equivalent result in terms of Lie algebras is Theorem 3.5. This asserts that the sum of any two quotients of the type \(\mathfrak g_i/\mathfrak{g}\), \(i=1,\ldots,3\), is isomorphic to the orthogonal complement of the third Lie algebra, and \(\mathfrak{g}=\bigoplus_{i=1}^3\mathfrak g_i/\mathfrak{g}\), where \(\mathfrak g_i\), \(i=1,\ldots,3\), denote the Lie algebras of the aforementioned subgroups, and \(\mathfrak{g}\) the Lie algebra of \(\mathcal{G}=\mathrm{SO}(3)\). Proposition 3.6 proves that the triality automorphism rotates the Lie algebras \(\mathfrak{so}(3)\oplus\mathfrak{so}(5)\) and \(\mathfrak{sp}(2)\oplus \mathfrak{sp}(1)\), and keeps fixed \(\mathfrak{psu}(3)\). Then, the authors obtain some topological properties of the 8-dimensional \(\mathcal{G}\)-manifolds, related e.g. to Stiefel-Whitney, Chern, Pontryagin classes and Euler characteristic, useful for providing obstructions to the existence of \(\mathcal{G}\)-actions. Proposition 5.1 gives the splitting into irreducible \(\mathcal{G}\)-modules of the intrinsic-torsion space (of dimension 200) for \(\mathcal{G}\). This space contains a 3-dimensional subspace of \(\mathcal{G}\)-invariant tensors. Proposition 5.4 illustrates that the \(\mathcal{G}\)-intrinsic torsion and the relative \(\mathcal{G}\)-torsion of one among the aforementioned subgroups is determined by the relative \(\mathcal{G}\)-torsion of the other two subgroups of \(\mathrm{SO}(8)\). Theorem 6.2 provides the conditions which must be satisfied by the differential forms of the six \(\mathcal{G}\)-invariant forms, when the intrinsic torsion is \(\mathcal{G}\)-invariant. Some examples of \(\mathrm{SO}(3)\)-manifolds with invariant torsion are given at the end of the paper.