On a class of transient random walks in random environment (Q1872205)

The author considers the following model, the random walk in random environment. For any \(k\in\mathbb Z^d\), there is a random vector \(p_k=(p_k(e))_{|e|=1}\) of probabilities summing up to one. The family \(p=(p_k)_{k\in\mathbb Z^d}\) is the random environment. Given this family, a random walker \(X=(X_n)_{n\in\mathbb N}\) on \(\mathbb Z^d\) jumps, when present at site \(k\), to the neighbor \(k+e\) with probability \(p_k(e)\). Hence, conditioned on the (highly disordered) environment \(p\), \(X\) is a spatially inhomogeneous Markov chain, a setting which one calls the quenched setting. The Markov property is lost under the so-called annealed law, \(P_0\), which averages over the walk and the environment. The one-dimensional case has been studied intensively since the mid-seventies and is basically well understood. However, the higher-dimensional case (on which a huge research activity has been started at the end of the nineties) is still largely open and is a source of many heuristic errors. The environment \(p\) is assumed to be i.i.d., and the variables \(p_k(e)\) are uniformly bounded away from 0. The author discusses various asymptotic results in the quenched setting, like a strong law of large numbers, a central limit theorem and large-deviation estimates, and their connection to various conditions on the environment, like three nestling conditions (about the position of the origin with respect to the convex hull of the support of the local drift \(\sum_{|e|=1}e p_k(e)\)), and two transience conditions (a.s. convergence into a certain direction, \(l\), and finiteness of exponential moments of the walker up to time \(\tau_1\); here \(\tau_1\) is, roughly speaking, the first time \(X_n\cdot l\) goes above an amount \(a\) beyond its previous local maxima and then never backtracks). Furthermore, the relations between the respective conditions and the Kalikov condition are discussed and their relation to the notion of traps (pockets where the walker loses much time with high probability), which is conveniently measured in terms of the spectral radius of a certain operator. The present paper develops new methods and improves earlier results of the author. One of the main novelties are the annealed upper tails of the random variable \(\tau_1\) mentioned above, under the above mentioned transience condition. More precisely, it is proved that \[ \limsup_{u\to\infty}(\log u) ^{-\alpha}\log P_0(\tau_1 >u)<0. \] for any \(\alpha<\frac{2d}{d+1}\) in \(d\geq 2\).