Localization of diffusion processes in one-dimensional random environment (Q1201782)

The authors consider the asymptotic behavior of a one-dimensional diffusion process in a random environment, that is, given a stochastic process \(\{W(x), x\in\mathbb{R}\}\) as an environment, they consider the limit theorem of a diffusion process \(X_ t=X(t,W)\), which is formally characterized by the following stochastic differential equation \[ dX_ t=\text{Brownian differential}-{1\over 2} W'(X_ t)dt. \] When \(\{W(x), x\in\mathbb{R}\}\) is a Brownian motion, \textit{Th. Brox} [Ann. Probab. 14, 1206- 1218 (1986; Zbl 0608.60072)] showed that \((\log t)^{-2} X(t,\cdot)- b(t,\cdot)\) tends to 0 in probability as \(t\to\infty\) where \(b(t,W)\) is a suitable function only depending on \(t\) and the environment. This results correspond to those by \textit{Ya. G. Sinaj} [Theory Probab. Appl. 27, 256- 268 (1982); translation from Teor. Veroyatn. Primen. 27, 247-258 (1982; Zbl 0497.60065)] for random walk in random environment. A more refined result (localization) is obtained by \textit{A. O. Golosov} [Commun. Math. Phys. 92, 491-506 (1984; Zbl 0534.60065)] in the case of Sinaj's reflecting random walk in random environment. The authors consider the localization property of diffusions in a wider class of environment. Their main result is stated as follows: The process \(\{X(e^ \lambda+t,\cdot)-b_ \lambda(\cdot), t\geq 0\}\) converges as \(\lambda\to\infty\) to a stationary process \(\{\widetilde\omega(t), t\geq 0\}\) in the sense of weak convergence. --- They characterize completely the limit process \(\{\widetilde\omega(t), t\geq 0\}\). In order to obtain the result, they study certain time reversals of random walks, conditioned random walks and their relations for the study of the shape of random environment processes.