Stationary distributions for diffusions with inert drift (Q843701)

Let \(D\subset\mathbb R^d\) be a bounded \(C^2\)-domain, \(n\) the inward unit normal vector field on \(\partial D\) and \(B\) a \(d\)-dimensional Brownian motion. The authors study the existence and uniqueness of the solution \((X,K)\) of the system \(dX_t=dB_t+n(X_t)dL_t^X+K_tdt,\) \(dK_t=n(X_t)dL_t^X,\, t\geq 0.\) Here the process \(X\), called the reflecting Brownian motion, is \(\overline{D}\)-valued, and \(L^X\) denotes the local time of \(X\) on the boundary \(\partial D\) of \(D\). Computer simulations made by \textit{K. Burdzy, R. Holyst} and \textit{L. Pruski } [Brownian motion with inert drift, but without flux: a model. Physica A 384, 278--284 (2007; \url{doi:10.1016/j.physa.2007.05.032})], done for a flat two-dimensional torus with a closed subset removed from its center, have suggested that a stationary distribution \((X,K)\) exists, has a product form and \(X_t\) is uniformly distributed in \(D\) under the stationary distribution. The authors of the present paper confirm this conjecture for the above general setting and they show that the second component of the stationary distribution is Gaussian. Moreover, they also consider a larger class of reflecting diffusions, including the distorted reflected Brownian motion.