Branching geodesics in sub-Riemannian geometry (Q2216461)

Geodesics in sub-Riemannian manifolds are not yet well understood. For example, it is not known whether or not sub-Riemannian geodesics must even be \(C^1\). A geodesic \(\gamma:[0,1] \to M\) in a manifold \(M\) is said to branch at time \(t \in (0,1)\) if there is another geodesic \(\gamma'\) with \(\gamma|_{[0,t]} = \gamma'|_{[0,t]}\) but \(\gamma|_{[0,s]} \neq \gamma'|_{[0,s]}\) for any \(s>t\). In a Riemannian manifold with a lower curvature bound, it is known that all geodesics are non-branching. In contrast, it is shown in the paper under review that normal geodesics in sub-Riemannian manifolds may indeed branch. Recall that a sub-Riemannian geodesic is normal if its associated control is not a critical point of the end-point map. The authors observe in Theorem 5 that a normal geodesic \(\gamma\) branches at time \(t\) if and only if the corank function is discontinuous at \(t\). The corank function assigns to any time \(t\) the corank of \(D_{u_t}E_x\) where \(u_t\) is the minimal control of \(\gamma|_{[0,t]}\) and \(E_x\) is the endpoint map. Further, the authors prove in Theorem 13 that any non-increasing, left continuous function \(f:[0,1] \to \mathbb{N}\) is the corank function of a normal sub-Riemannian geodesic, and thus this geodesic must branch at the discontinuities of \(f\). They also show in Theorem~7 that the size of the time interval around \(t\) over which some compact family of branching paths is length minimizing is uniformly controlled. Finally, in Section 4, the authors provide an example of a sub-Riemannian structure on \(\mathbb{R}^3\) for which the path \(t \mapsto (0,t,0)\) branches at \(t=0\).