Heat kernel estimates for strongly recurrent random walk on random media (Q960181)

Consider a random walk on a random graph. The goal of the paper is to show that if the volume of the ball of radius \(R\) on the random graph is of order \(R^D\) with high probability and the resistance between the center and the outside of the ball is of order \(R^\alpha\) with high probability, then one can establish both a quenched and annealed estimates for the exit time from the ball, the on-diagonal heat kernel and the mean displacement. In particular, the spectral dimension of the graph is given by \[ d_s(G) = -2\lim_n\frac{\log p_{2n}(x,x)}{\log n}= \frac{2D}{D+\alpha}. \] Although the calculations are more involved, the principal ideas of the proofs are similar to the ones of \textit{M. T. Barlow, A. A. Járai, T. Kumagai, G. Slade} [Commun. Phys. 278, 385--431 (2008; Zbl 1144.82030)]. Results from \textit{M. T. Barlow, T. Coulhon} and \textit{T. Kumagai} [Commun. Pure Appl. Math. 58, No. 12, 1642--1677 (2005; Zbl 1083.60060)] are also used. The results are new for the random walk on the incipient infinite cluster of \(\mathbb Z^d\) for high dimensions and for long-range percolation on \(\mathbb Z\). For the latter model, consider a sequence \((p_n)\) such that \(0\leq p_n\leq 1\) for all \(n\geq 1\). Then a one dimensional long-range percolation model is a random graph constructed by connecting two vertices \(x\) and \(y\) of \(\mathbb Z\) with probability \(p_n\) where \(| x-y| = n\geq 1\). \textit{N. Berger} [Commun. Math. Phys. 226, No. 3, 531--558 (2002; Zbl 0991.82017)] showed that if \(0\leq p_n<1 \) for \(n\geq 1\) and if \[ \lim_{n\to \infty} \frac{p_n}{\beta n^{-s}}=1 \] for some \(1<s<2\) and \(\beta>0\), then the random walks are transient whilst they are recurrent if (1) holds for \(s=2\) and \(\beta>0\). If \((p_n)\) verifies (1) for some \(s>2\) then there is no infinite cluster. The main results of this paper apply to models where \(p_1 =1\) and (1) holds for some \(s>2\) and \(\beta>0\). In particular, it is proved that \(d_s(G) = 1\) and \[ \lim_{R\to\infty}\frac{E_{x,\omega}\tau_R}{\log R} = 2 \] where \(\tau_R\) is the exit time of a ball of radius \(R>0\).