No label defined (Q4824399)

The authors study recurrence and transience of a random walk on \(\mathbb Z^d\) in a random environment which is constructed as follows: a site \(x\in\mathbb Z^d\) is ``special'' with probability \(p(x)\) independently of the other sites, otherwise it is ``ordinary''. At an ordinary site, the transition probabilities of the random walk are \(p_{x,y} = 1/2d\) and at every special site \(x\), the transition probabilities \(p_{x,y}\) verify: A) \(x\cdot \sum_y (y-x)p_{x,y} >0\) if \(d\leq 2\) and \(<0\) if \(d\geq 3\); B) for some constant \(D>0\), if \(\| x- y\| >D\), then \(p_{x,y}=0\). That is, at ordinary sites there is no mean drift while at special sites it is directed outwards the origin for \(d=1,2\) and towards the origin for \(d\geq 3\). The main result is that if there is \(\varepsilon >0\) such that for all \(x\) large enough \(p(x)\geq \| x\|^{ -1+\varepsilon}\) and if there is a constant \(a>0\) such that \(| x\cdot \sum_y (y-x)p_{x,y}| > a \| x\| \) at all special sites, then the walk is transient in \(\mathbb Z^d\) for \(d=2\) and it is recurrent in \(\mathbb Z^d\) for \(d\geq 3\). On the other hand, if there is \(\varepsilon >0\) such that for all \(x\) large enough \(p(x)\leq \| x\|^{ -1-\varepsilon}\), then the walk is recurrent in \(\mathbb Z^d\) for \(d=2\) and it is transient in \(\mathbb Z^d\) for \(d\geq 3\). The proofs are based on general criteria for countable Markov chains that can be found in Theorems 2.2.1 and 2.2.2 of the book by \textit{G. Fayolle, V. A. Malyshev} and \textit{M. V. Menshikov} [``Topics in the constructive theory of countable Markov chains'' (1995; Zbl 0823.60053)]. They require an explicit expression for the so-called Lyapunov function. For the random walks considered above, these functions are, respectively, \(f(x) = (\| x\| +1)^{-1}, f(x) = \| x\|^2, f(x)=\log\log\| x\|\) and \(f(x)= (\| x\| +1)^{1/2}\). In one dimension, it is possible to define the model in a more constructive way with a more precise classification and an explicit although more intricate Lyapunov function.