Aging in two-dimensional Bouchaud's model (Q816983): Difference between revisions

This paper studies (sub)-aging properties of the two-dimensional Bouchaud's trap model. The model is defined as follows: to every site \(x\) of \(\mathbb Z^2\) is assigned a (quenched) random ``energy'' variable \(E_x\), distributed exponentially with mean \(1\). Variables at different sites are independent. For \(\beta>0\), one interprets \(\tau_x:=\exp(\beta E_x)\) as the depth of the trap at \(x\). In other words, one considers a random walk \(X(t)\) on \(\mathbb Z^2\) which, once it is at some \(x\), leaves this point for one of his \(4\) nearest neighbors (chosen uniformly) after an exponential time of average \(\tau_x\). The picture is that, if \(\beta>1\) (low temperature regime, where the disorder-average of \(\tau_x\) is infinite), the random walk stays a long time in a deep trap, until it finds an even deeper one. More precisely, after a ``waiting time'' \(t_w\), the walk is (roughly speaking) in a trap with mean waiting time \(t_w/\log t_w\). Formally, what is proven is the following: let the random walk start at zero, \(X(0)=0\), and let (for a given environment \(\{E_x\}_x\)) the two-point function \(R\) be defined as \[ R(t_w, t_w+t)=\mathbb P(X(t_w+t)=X(t_w)), \] where \(\mathbb P\) denotes the law of the random walk, while \[ \Pi(t_w,t_w+t)=\mathbb P(X(t')=X(t_w)\qquad \forall t'\in [t_w,t_w+t]). \] Then (Theorem 1.1) one has aging for \(R\), i.e., for almost every disorder realization one has \[ \lim_{t_w\to\infty}R(t_w, (1+\theta)t_w)=R(\theta) \] for an explicitly given, non-trivial, disorder-independent function \(R(\theta)\) (see Propositions 7.1). As for \(\Pi\), on has sub-aging (Theorem 1.2): \[ \lim_{t_w\to\infty}\Pi(t_w, t_w+\theta\frac{t_w}{\log t_w})=\Pi(\theta) \] (see Proposition 8.2 for the definition of \(\Pi(\theta)\)). Both results hold as soon as \(\beta>1\). If \(\mathbb Z^2\) were replaced by \(\mathbb Z^d\) with \(d\geq3\), the picture would remain qualitatively the same except that one would have aging also for \(\Pi\). On the other hand, the situation is quite different in \(d=1\). This paper in any case discusses in detail only \(d=2\) (the case \(d=1\) being treated in [\textit{L. R. G. Fontes, M. Isopi, C. M. Newman}, Ann. Probab. 30, No. 2, 579--604 (2002; Zbl 1015.60099)], and \(d\geq3\) in [\textit{J. Černý}, On two properties of strongly disordered systems, aging and critical path analysis. Ph.D. thesis, EPF Lausanne, 2003]).