Limit laws for transient random walks in random environment on \(\mathbb Z\) (Q848132): Difference between revisions

Let \(\omega:=(\omega_i,i\in\mathbb{Z})\) be a family of i.i.d. random variables (random environment) defined on a probability space \((\Omega,\mathcal{F},P)\) and taking values in \((0,1)\). The random walk \((X_n,n\geq 0)\) in random environment is a nearest-neighbor random walk on \(\mathbb{Z}\) with transition probabilities given by \(\omega\). Set \(\rho_i= (1-\omega_i)/\omega_i\) for \(i\in \mathbb{Z}\). Put \(\tau(x):= \inf\{n\geq 1: X_n=x\}\), \(x\in \mathbb{Z}\). The main result states that if there exists \(0<\kappa <1\) such that \(E(\rho_0^{\kappa})=1\), \(E(\rho_0^{\kappa}\log^+\rho_0)<\infty\) and the distribution of \(\log\rho_0\) is non-lattice then \(n^{-1/\kappa}\tau(n)\) and \(n^{-\kappa}X_n\) have the explicit limits (involving the \(\kappa\)-stable random variable, \(E(\rho_0^{\kappa}\log\rho_0)\) and \(\sin(\pi \kappa)/(2^{\kappa}\pi\kappa^2 C_K^2)\)) in law as \(n\to \infty\). Here \(C_K\) is the constant describing the tail of Kesten's renewal series \(R:=\sum_{k\geq 0} \rho_0\dots \rho_k\) [see \textit{H. Kesten}, Acta Math. 131, 207--248 (1973; Zbl 0291.60029)], namely, \(P(R>x) \sim C_K x^{-\kappa}\) as \(x\to\infty.\) Thus the generalization of a classical result by \textit{H. Kesten, M. V. Kozlov} and \textit{F. Spitzer} [Compos. Math. 30, 145--168 (1975; Zbl 0388.60069)] is provided (they proved that the hitting time of the level \(n\) converges in law after a proper normalization towards a positive stable law but did not indicate its parameters). The special case when the transition probabilities follow a Beta distribution (the Dirichlet enviroment) is considered as well. The technique developed allow to derive the convergence in law of the process \((n^{-\kappa}X_{[nt]}, t\geq 0)\) in the space of càdlàg functions.