Chaotic time dependence in a disordered spin system (Q1960927)

A Markov process is said to exhibit chaotic time dependence (CTD) if its distribution at time \(t\) has multiple subsequence limits as \(t\to\infty\). Such a phenomenon can be produced by a proper choice of initial state in some interacting particle systems, such as stochastic Ising or voter models. The aim of the current paper is to give an example of a system for which the CTD occurs when the initial condition is taken at random. Consider the \(d\)-dimensional voter model whose initial state \(\sigma(0)\) is chosen uniformly in \(\{-1,+1\}^{\mathbb Z^d}\). If the rates of exponential updatings are random (say, uniformly positive i.i.d. r.v.), the model is called a voter model with random rates (VMRR). The main results of the paper are as follows: In dimension \(1\), let the common distribution of the mean update time \(\tau_0\) have heavy tails; then for almost all realizations of the jump rates and the initial condition the distribution of \(\sigma(t)\) does not converge as \(t\to\infty\). In other words, the VMRR exhibits the CTD in a strong sense. On the other hand, if \(\tau_0\) is not heavy-tailed or if the dimension \(d\) is at least \(2\), then the distribution of \(\sigma(t)\) has a weak limit as \(t\to\infty\) thus implying absence of the CTD. The proofs are based on an analysis of the localization properties of random walks with random rates appearing in the dual representation of linear voter models.