The speed of biased random walk on percolation clusters (Q1404208): Difference between revisions

The authors consider a biased random walk on a two-dimensional bond-percolation cluster. Fix the percolation probability parameter above the critical value (which is \(\frac 12\)), and condition, for convenience, on the event that the origin belongs to the infinite cluster. Given the infinite cluster, a random walk on \(\mathbb Z^2\) starts at the origin and jumps along open bonds only. With a parameter \(\beta>1\), the probability of a jump in the East-direction (if possible) is \(\beta/(\beta+l_n-1)\), where \(l_n\) is the number of neighbors that can be reached along open edges, and the probability for an other nearest-neighbor jump along an open edge is \(1/(\beta+l_n-1)\). That is, the random walker is biased in the East-direction with strength \(\beta\). This model has been studied for the first time some 20 years ago in the Physics literature. There are only few mathematical results available yet. The result of the paper is that, for \(\beta\) sufficiently small, the walker's East-drift (averaged over the medium and the walk, i.e., in the annealed setting) is positive, while for sufficiently large \(\beta\), the drift is zero. The conjecture is that there is no gap between these two regimes, but the proof of that is open. The lower bound for the upper regime given in the paper coincides with the conjectured critical value of \(\beta\). The result sounds counterintuitive at first. An explanation is that, for large drift parameter \(\beta\), the walker gets more often and for longer time stuck in dead ends of the percolation cluster and loses therefore more time. Recently Sznitman obtained rather similar results (even in general dimension greater than or equal to two), but his methods are different.