Rigidity for Markovian maximal couplings of elliptic diffusions (Q2359738): Difference between revisions

It is shown that Markovian maximal couplings of elliptic diffusions are a rare phenomenon whose existence imposes severe geometric constraints on the underlying manifold. Recall that a maximal coupling is one that attains Aldous' bound \(\mathbb P (\tau > t) \geq \|\mu_{1,t}-\mu_{2,t}\|_{TV}\) for the coupling time \(\tau=\inf\{s>0: \forall t>s,X_t=Y_t\}\). A Markovian coupling is one defined by strategies where neither process is allowed to look into the future of the other before making the next transition. The local perturbation condition (LPC) means that the coupling exists for the diffusions starting from any pair of points \(\mathbf x\) and \(\mathbf y\) in some neigbourhood of the initial points \({\mathbf x}_0\) and \({\mathbf y}_0\). The authors prove that the only complete connected Riemannian manifolds on which there exist Brownian motions with drifts that admit a Markovian maximal coupling with LPC are (up to a scaling) Euclidean space \(\mathbb R^d\), the sphere \(\mathbb S^d\) and the hyperbolic space \(\mathbb H^d\). Furthermore, the only vector fields which can yield Markovian maximal couplings are given by the Killing vector fields, i.e., the infinitesimal generators of the rigid motion group of the manifold. For curvature \(K=0\) (Euclidean space) the drift can be described by \({\mathbf b}({\mathbf x}) = \lambda {\mathbf x} + T {\mathbf x} + {\mathbf c}\) with a scalar \(\lambda\), a skew-symmetric matrix \(T\) and a vector \({\mathbf c}\).