Riemannian and sub-Riemannian geodesic flows (Q2012951)

Let \((M, \mathcal{H}, \mathbf{h})\) be a sub-Riemannian manifold, where \(\mathcal{H} \subset TM\) is the horizontal bundle and \(\mathbf{h}\) is a sub-Riemannian metric on \(\mathcal{H}\). A Riemannian metric \(\mathbf{g}\) on \(M\) is said to \textit{tame} \(\mathbf{h}\) if \(\mathbf{g}|_{\mathcal{H}} = \mathbf{h}\). Such \(\mathbf{g}\) is certainly not unique and it is of interest to understand ``good'' ways to choose \(\mathbf{g}\). In this paper, the authors consider when the corresponding sub-Riemannian and Riemannian geodesic flows commute. Let \(\exp^r : TM \to M\) be the Riemannian exponential map of \(\mathbf{g}\), so that \(\exp^r(x,v)\), \(x \in M\), \(v \in T_x M\), is the endpoint of a \(\mathbf{g}\)-geodesic starting at \(x\) with initial velocity \(v\). Likewise, let \(\exp^{sr} : T^* M \to M\) be the sub-Riemannian exponential map of \(\mathbf{h}\), so that \(\exp^{sr}(x,p)\), \(x \in M\), \(p \in T^*_x M\), is the endpoint of a sub-Riemannian normal geodesic starting at \(x\) with ``initial momentum'' \(p\). Let \(\mathcal{V} \subset TM\) be the \(\mathbf{g}\)-orthogonal complement of \(\mathcal{H}\) and assume that \(\mathcal{V}\) is integrable. Let \(\text{pr}_{\mathcal{V}}\) be orthogonal projection onto \(\mathcal{V}\), and let \(\sharp : T^* M \to TM\) be the isomorphism induced by \(\mathbf{g}\). For \(x \in M\) and \(p \in T^*_x M\), let \(P_t\) denote parallel transport along the curve \(\exp^r(x, t \sharp p)\) with respect to the Levi-Civita connection of \(\mathbf{g}\). The authors study the following commutation property: Under what conditions do we have \[ \exp^{sr}(x, tp) = \exp^r(\exp^r(x, t \sharp p), -t \text{pr}_{\mathcal{V}} P_t \sharp p) \tag{1} \] for all sufficiently small \(t\)? They consider an affine connection \({\nabla}\) previously introduced in [\textit{F. Baudoin} et al., Commun. Anal. Geom. 24, No. 5, 913--937 (2016; Zbl 1369.53020); the second author and \textit{A. Thalmaier}, Math. Z. 282, No. 1-2, 99--130 (2016; Zbl 1361.53028); Math. Z. 282, No. 1-2, 131--164 (2016; Zbl 1361.53029)] and show that a sufficient condition for (1) is to have \({\nabla} \mathbf{g} = 0\) (Theorem 2.4). Specifically, \({\nabla} \mathbf{g} = 0\) implies that the foliation \(\mathcal{F}\) of \(M\) with leaves tangent to the integrable bundle \(\mathcal{V}\) is a totally geodesic Riemannian foliation, which in turn implies (1). Now suppose that \((B, \check{\mathbf{g}})\) is another Riemannian manifold and \(\pi : M \to B\) is a surjective submersion, and take \(\mathcal{V} = \ker \pi_*\) with \(\mathcal{H}\) another bundle which is a linear complement of \(\mathcal{V}\) in \(TM\) (a so-called \textit{Ehresmann connection}). Then the pullback of \(\check{\mathbf{g}}\) under \(\pi\) gives a sub-Riemannian metric \(\mathbf{h}\) on \(\mathcal{H}\). If \(\mathbf{g}\) is a Riemannian metric which tames \(\mathbf{h}\), one might ask whether Riemannian and sub-Riemannian metrics are mapped by \(\pi\) to the same curve, i.e., \[ \pi(\exp^r(x, t \sharp p)) = \pi(\exp^{sr}(x, tp)) \tag{2} \] The authors show (Theorem 2.5) that (2) holds if and only if \(\mathcal{H}, \mathcal{V}\) are \(\mathbf{g}\)-orthogonal and the leaves of the foliation of \(\mathbf{V}\) are totally geodesic. Again, the latter of these conditions follows from \({\nabla} \mathbf{g}=0\).