On the topology and the geometry of SO(3)-manifolds (Q540464)

The authors consider the non-standard embedding of \(\text{SO}(3)\) into \(\text{SO}(5)\) given by the five-dimensional irreducible representation of \(\text{SO}(3)\), called \(\text{SO}(3)_{\text{ir}}\). The authors study the topology and differential geometry of five-dimensional Riemannian manifolds carrying such an \(\text{SO}(3)_{\text{ir}}\) structure, i.e., with a reduction of the frame bundle to \(\text{SO}(3)_{\text{ir}}\). These spaces are the non-integrable analogues of the symmetric space \(\text{SU}(3)/\text{SO}(3)\) and its non-compact dual \(\text{SL}(3,\mathbb{R})/\text{SO}(3)\). While the general framework for the investigation of such structures was outlined by \textit{B. Alexandrov, T. Friedrich} and \textit{N. Schoemann} [J. Geom. Phys. 53, No.~1, 1--30 (2005; Zbl 1075.53036)], first concrete results were obtained by \textit{M. Bobieński} and \textit{P. Nurowski} [J. Reine Angew. Math. 605, 51--93 (2007; Zbl 1128.53017)] (general theory) as well as \textit{S. G. Chiossi} and \textit{A. Fino} [J. Lie Theory 17, No.~3, 539--562 (2007; Zbl 1141.53018)]. The Lie group \(\text{SO}(3)\) admits two inequivalent embeddings into \(\text{SO}(5)\). The standard embedding is given by upper diagonal matrices, \[ \text{SO}(3)_{\text{st}}\subset \text{SO}(5),\quad A\rightarrow \text{diag}(A,1,1), \] while the second embedding corresponds to the unique faithful irreducible five-dimensional representation of \(\text{SO}(3)\), and is denoted by \(\text{SO}(3)_{\text{ir}}\subset \text{SO}(5)\). In the first part of the article, the authors describe the topological properties of the two different types of \(\text{SO}(3)\) structures. While classical results by Thomas and Atiyah are available for the standard diagonal embedding \(\text{SO}(3)_{\text{st}}\rightarrow \text{SO}(5)\), the case \(\text{SO}(3)_{\text{ir}}\) is first investigated here. The authors show that the symmetric space \(\text{SU}(3)/\text{SO}(3)\) admits an \(\text{SO}(3)_{\text{ir}}\) structure, but not an \(\text{SO}(3)_{\text{st}}\) structure. The authors prove necessary relations for the characteristic classes of a five-manifold with a topological \(\text{SO}(3)_{\text{ir}}\) structure: its first Pontryagin class \(p_1(M)\) has to be divisible by five, the Stiefel-Whitney classes \(w_1(M),w_4(M),w_5(M)\) must vanish, etc. Moreover, a simply-connected \(\text{SO}(3)_{\text{ir}}\)-manifold that is spin is automatically parallelizable. The authors construct explicit examples of \(S^1\)-fibrations over a four-dimensional base that admit a topological \(\text{SO}(3)_{\text{ir}}\) structure. In the second part, the differential geometry of some homogeneous examples is studied in detail. The authors focus on a `twisted' Stiefel manifold \(V^{\text{ir}}_{2,4}:=\text{SO}(3)\times \text{SO}(3)/\text{SO}(2)_{\text{ir}}\), its non-compact partner \(\widetilde{V}^{\text{ir}}_{2,4}:=\text{SO}(2,1)\times \text{SO}(3)/\text{SO}(2)_{\text{ir}}\) and the space \(W^{\text{ir}}:=\mathbb{R}\times (\text{SL}(2,\mathbb{R})\times \mathbb{R}^2)/\text{SO}(2)_{\text{ir}}\). On each of these, a family of metrics depending of three deformation parameters \(\alpha,\beta,\gamma\) is considered. In the case \(W^{\text{ir}}\), it is in addition necessary to consider a family of possible embeddings of \(\text{SO}(2)_{\text{ir}}\) into \(\mathbb{R}\times (\text{SL}(2,\mathbb{R})\times \mathbb{R}^2)\), as the ones leading to \(\text{SO}(3)_{\text{ir}}\) structures are far from being trivial. The standard Stiefel manifold is known to admit an Einstein-Sasaki metric that was crucial for the understanding of Riemannian Killing spinors. In contrast, the authors show that the twisted Stiefel manifold admits a nearly integrable \(\text{SO}(3)_{\text{ir}}\) structure with parallel torsion, a compatible Sasaki structure whose contact connection coincides with the \(\text{SO}(3)_{\text{ir}}\) connection, but none of the Sasaki structures is Einstein (however, an Einstein metric is shown to exist). All in all, the twisted Stiefel manifold is an example of a rather well-behaved \(\text{SO}(3)_{\text{ir}}\)-manifold. The manifold \(W^{\text{ir}}\) carries an \(\text{SO}(3)_{\text{ir}}\) structure that disproves several conjectures on \(\text{SO}(3)_{\text{ir}}\)-manifolds which one might be tempted to conclude from the previous example. It carries an \(\text{SO}(3)_{\text{ir}}\) structure only for two possible embeddings of \(\text{SO}(2)_{\text{ir}}\) depending on the parameters \(\alpha,\beta,\gamma\) of the metric. The torsion of the \(\text{SO}(3)_{\text{ir}}\) connection turns out to be non-parallel, the space is neither Einstein nor naturally reductive, and there exists no compatible contact structure whose contact connection would coincide with the \(\text{SO}(3)_{\text{ir}}\) connection. In particular, this shows that an \(\text{SO}(3)_{\text{ir}}\) structure is conceptionally really different from a contact structure; it thus defines a new type of geometry in five dimensions.