Anomalous behaviour for the random corrections to the cumulants of random walks in fluctuating random media (Q5943925)

For a certain class of random walks in time-dependent random environment in dimensions one and two, an almost sure central limit theorem is established for the deviation from the average random walk, with identification of the rate of the norming constants. The model is as follows. Let \(\xi=\{\xi(t,x): t\in\mathbb N,x\in\mathbb Z^d\}\) be a collection of i.i.d. random variables taking values in a finite set \(\mathcal S\). With some compactly supported distribution \(P_0\) on \(\mathbb Z^d\) and some compactly supported function \(c: \mathbb Z^d\times\mathcal S\to\mathbb R\), define a random walk \((X_t)_{t\in\mathbb N_0}\) in the time-dependent random environment \(\xi\) as the walk that makes at time \(t\in \mathbb N\) a step from the point \(x\in\mathbb Z^d\) to \(y\in\mathbb Z^d\) with probability \(P_0(y-x)+\varepsilon c(y-x,\xi(t,x))\), where \(\varepsilon>0\) is a small parameter. Certain assumptions on \(P_0\) and \(c\) ensure that the perturbative term \(c\) does not influence the drift. Let \(b\in \mathbb R^d\) and \(C\in \mathbb R^{d\times d}\) denote the drift vector and covariance matrix of the average random walk, which is the walk obtained when putting \(\varepsilon =0\). The walk starts at \(X_0=0\). We assume that \(d=1\) or \(d=2\). The cases of higher dimensions have been studied by the author and coauthors elsewhere. The author derives, for sufficiently small values of the parameter \(\varepsilon\), central limit theorems for the random corrections \(\mathcal {E}_T(\xi)= E(X_T\mid \xi)-b T\) and \(\mathcal{C}_T(\xi)=E((X_T-b t)(X_T-bT) ^{\text{ T}}\mid \xi)-CT\) to the first and second cumulants as \(T\to\infty\). More precisely, there are norming constants \(s_T\) and \(\hat s_T\) such that, almost surely, \(\mathcal {E}_T(\xi)/s_T\) and \(\mathcal{C}_T(\xi)/\hat s_T\) converge in distribution to standard normal variables. The rates of the norming constants are in dimension one \(s_T\approx T^{1/4}\) and \(\hat s_T\approx T^{3/4}\), and in dimension two \(s_T\approx (\ln T)^{1/2}\) and \(\hat s_T\approx T^{1/2}\). The proof uses martingale arguments, weak dependence conditions and a cluster-expansion technique.