Quenched invariance principle for multidimensional ballistic random walk in a random environment with a forbidden direction (Q879246)

The authors consider a random walk in random environment (RWRE) on \(\mathbb Z^d\), \(d\geq 2\), defined as follows. Let \(\omega =(\omega_x)_{x\in \mathbb Z^d}\), the random environment, be an i.i.d. set of probability vectors on \(\mathbb Z^d\); the law of \(\omega \) is denoted by \(\mathbb P\). Given \(\omega \), a random walk, \(X=(X_n)_{n \geq 0}\), starts at the origin, and, being located at \(x\), it jumps to \(y\) with probability \(\pi_{x,y}(\omega )\) which is defined by \(\omega_x=(\pi_{x,x+y})_{y\in \mathbb Z^d}\). The probability distribution \(P_0^\omega\) of \(X\) with fixed \(\omega \) is called quenched distribution. If the average over \(\omega \) is performed, the distribution is called annealed and denoted by \(P_0\). In the present paper the authors assume that the random walk is not allowed to retreat in a fixed direction \(\hat u\), that is \(\pi_{x,y}=0\) if \((y-x)\cdot \hat u<0\). This creates a drift in a certain direction and there is a law of large numbers with a non-zero speed \(v\). Moreover, it is assumed that the environment is non-nestling, elliptic, acts on more than two spatial dimensions and the step of the walk has \(2+\epsilon\) moment uniformly in the environment. The main result of the paper is a quenched invariance principle which states that the processes \( B_n=n^{-1/2}\{X_{[nt]}-[nt]v\}\) and \( \tilde B_n=n^{-1/2}\{X_{[nt]}-E_0^\omega(X[nt])\}\) converge to a Brownian motion with a certain diffusion matrix. The proof is based on previous papers of both authors [ALEA, Lat. Am. J. Probab. Math. Stat. 1, 111--147, electronic only (2006; Zbl 1115.60106)], where the annealed invariance principle was proved under similar conditions, and [Probab. Theory Relat. Fields 133, No. 3, 299--314 (2005; Zbl 1088.60094)], where a theorem allowing the proof of the quenched invariance principle under certain conditions was derived. The main difficulty in the paper is then verifying these conditions, in particular bounding the variance of the quenched mean: \(\mathbb E[| E^\omega_0(X_n)-E_0(X_n)| ^2]\leq C n^\gamma\), where \(\gamma \) depends on order of the moment that is assumed for the step distribution of \(X\).