Invariant distributions and scaling limits for some diffusions in time-varying random environments (Q2447279): Difference between revisions

The author studies a class of stochastic processes that includes the process \(Y\) with random environment specified by the equation \[ dY_{t}=dB_{t}-\frac{1}{2}\frac{W^{\prime}(Y_{t})}{t^{\beta}}dt, \] where \(B_{t}\) is a Brownian motion with \(t\geq 0\) and \(W(x)\) is a Brownian motion with \(x\in \mathbb{R}\). In fact, the extension he studies is of the form \[ dZ_{t}= dB_{t}-\frac{1}{2}\partial_{x} V(t, Z_{t})dt, \] where \(V(t,x)\) is a stochastic process with \((t,x)\in [0,\infty)\times \mathbb{R}\), which is specified using transformations and a scaling of \(W\). These two previous equations are not well-posed since \(W^{\prime}\) and \(\partial_{x}V\) do not exist. Thus, the author first makes sense of the solutions, \(Y\) and \(Z\), of this type of equations using the associated martingale problem posed with the corresponding generator. This procedure is possible to do after fixing a trajectory of \(W\), which is called the quenched case. Part of the results in the paper is that there is a semigroup associated to \(Z\) that satisfies certain properties, which helps to construct a random dynamical system acting on signed measures. In turn, this is used to establish that there is a measure that satisfies an invariant property for the semigroup. Moreover, the author shows the existence of measures that are stationary in some specific sense, for the quenched case, but also for the annealed case, i.e., without fixing the environment \(V\). In the quenched case, the author is also able to give a bound for the probability density of \(Z_{t}\), called the Aronson estimate. As a corollary of the results, he obtains that, for any \(\beta>1/4\) and for almost every trajectory \(W\), the random variable \(\lim_{t\to\infty}Y_{t}/\sqrt{t}\) is normal distributed.