Localization for constrained martingale problems and optimal conditions for uniqueness of reflecting diffusions in 2-dimensional domains
arXiv:2206.05621 · doi:10.1016/j.spa.2024.104295
Abstract
We prove existence and uniqueness for semimartingale reflecting diffusions in 2-dimensional piecewise smooth domains with varying, oblique directions of reflection on each "side", under geometric, easily verifiable conditions. Our conditions are optimal in the sense that, in the case of a convex polygon with constant direction of reflection on each side, they reduce to the conditions of Dai and Williams (1996), which are necessary for existence of Reflecting Brownian Motion. Moreover our conditions allow for cusps. Our argument is based on a new localization result for constrained martingale problems which holds quite generally: as an additional example, we show that it holds for diffusions with jump boundary conditions.
References in corpus (5)
- Viscosity methods giving uniqueness for martingale problems
- Diffusion with nonlocal boundary conditions
- Existence and uniqueness of reflecting diffusions in cusps
- Markov selection for constrained martingale problems
- A reverse ergodic theorem for inhomogeneous killed Markov chains and application to a new uniqueness result for reflecting diffusions