Mott law as lower bound for a random walk in a random environment (Q2456756): Difference between revisions

The authors consider a random walk on the support of a simple point process \(\widehat \xi=\sum_ j\delta_ {x_ j}\) in \(\mathbb R^ d\). It is assumed that this point process is shift-invariant and ergodic, and some moment of \(\widehat \xi([-\frac 12,\frac 12]^ d)\) is assumed finite. Given \(\widehat \xi\), each point \(x_ j\) of the cloud posseses an independent mark \(E_{x_ j}\in[-1,1]\). Let \(\mathcal P_ 0\) denote the Palm distribution of the marked random field \(\xi=\sum_ j\delta_ {(x_ j, E_{x_ j})}\), i.e., its conditional distribution given that \(0\) is in the support of \(\widehat\xi\). This process \(\xi\) serves as the random environment for a continuous-time Markov chain \((X_t^\xi)_{t\in[0,\infty}\) on the support of \(\widehat\xi\), which starts from \(0\) and has transition rates from \(x\) to \(y\) given by \[ c_{x,y}(\xi)=\exp\Big(-| x-y| -\beta(| E_ x-E_ y| +E_ x| +| E_ y| )\Big),\qquad x\not=y, \] where \(\beta\in(0,\infty)\) is the inverse temperature. The main interest of the paper is in the annealed long-time asymptotics of the Markov chain in random environment and the limiting diffusion matrix \(D\) given by \[ (a\cdot Da)=\lim_ {t\to\infty}\frac 1t E[(X_t^\xi,a)^ 2],\qquad a\in\mathbb R^ d, \] where \(E\) denotes expectation over both randomnesses and \((\cdot,\cdot)\) the standard inner product on \(\mathbb R^ d\). The main results of the paper are a proof of the existence of this limit and, in addition, the proof of an invariance principle for the accordingly rescaled Markov chain towards a Brownian motion with diffusion matrix \(D\). Furthermore, a quantitative lower bound on \(D\) in \(d\geq 2\) is given under some technical assumptions (which still admit practically all interesting examples). An intensive discussion of the underlying physics, including an explanation of the relation to Mott's law, is given.