Limit theorems for a random walk in a random environment (Q5903014)

Let \(\{X(k))_{k\in \mathbb N}\), \(P_ x^{\pi}\}\) be a Markov chain on \(\mathbb Z^ d\), starting from \(x\in \mathbb Z^ d\), and characterized by the family \(\pi =(\pi (x,y)\), \(x,y\in \mathbb Z^ d)\) of transition probabilities. Suppose there exists a probability distribution \(\mu\) on the set of all possible \(\pi\), and put \(P_ x(\cdot)=\int \mu (d\pi)P_ x^{\pi}(\cdot).\) The process \((X(k),P_ x)\) is called a random walk in a random environment (\(\pi\),\(\mu)\). A very interesting and yet not completely solved problem is to study the asymptotic behaviour of the process \(\{X([nt])-a_{[nt]})/b_ n,\quad t\geq 0\}\); here the normalizing constants \(a_ n\) and \(b_ n\) are nonrandom, i.e. they do not depend neither on \(\pi\) nor on the path \(X(\cdot)\). The purpose of the present paper is to give an answer to this problem in the case \(\sum_{y\in \mathbb Z^ d}(y-x)\pi (x,y)=0\), i.e. in the case the process \((X(k),P_ x^{\pi})\) is a martingale for almost all \(\pi\). Under this assumption and some other (technical) assumptions the authors are able to prove that the process \(\{X([nt])/n^{1/2},t\geq 0\}\) converges weakly to the Wiener process (of null expectation) on \(\mathbb R^ d\). This result comes out to be a generalization of a previous result by \textit{G. F. Lawler} [Commun. Math. Phys. 87, 81--87 (1982; Zbl 0502.60056)].