A local limit theorem for random walks in random scenery and on randomly oriented lattices (Q653299): Difference between revisions

Random walks in random scenery are processes defined by \[ Z_{n} := \sum_{k=1}^n \xi_{X_1+\dots+X_k}, \] where \((X_{k},k\geq1)\) and \((\xi_{y},y\in\mathbb{Z})\) are two independent sequences of i.i.d. random variables. The distribution of \(\xi_1\) belongs to the domain of normal attraction of a strictly stable distribution law \(S_{\beta}\) with index \(\beta\in(0,2]\). Concerning the random walk \(Z_n\), the distribution of \(X_1\) belongs to the domain of normal attraction of a strictly stable distribution \(S_{\alpha}\) with \(\alpha\in(0,2]\). The case \(\alpha=1\) is not considered here. If \(\alpha\in(1,2]\), the random walk \(S_n=\sum_{k=1}^nX_l\), \(n\geq1\), is recurrent and the limiting process \(U\), \[ (n^{-\frac{1}{2}}S_{[nt]})_{t\geq0}\underset{n\rightarrow \infty }{\overset{\alpha }{\Rightarrow }}(U(t))_{t\geq0}, \] admits a jointly continuous version of the local time process \(\{L_t(x),t\in\mathbb{R}^+,x\in \mathbb{R}\}\). Let \(\delta :=1-\frac{1}{\alpha}+\frac{1}{\alpha\beta}\). If \(\xi_1\) is a \(\mathbb{Z}\)-valued random variable, then, for its characteristic function \(\varphi_{\xi}(u)\), there exists an integer \(d\geq 1\) such that \[ \{u:|\varphi_{\xi}(u)|=1\}=\frac{2\pi}{d}\mathbb{Z}. \] The main results are as follows. For \(\alpha>1\), \(\beta\in(0,2]\), let \(C(x)=\mathrm{E}\left(|L|^{-1}_{\beta}f_{\beta}(|L|^{-1}_{\beta}x)\right)\), \(x\in \mathbb{R}\), where \[ |L|_{\beta}=\left(\int_{\mathbb{R}}L^{\beta}_1(y)dy\right)^{\frac{1}{\beta}}. \] Theorem 1 states that, if \(\xi_y\) are \(\mathbb{Z}\)-valued and \(\alpha>1\), \(\beta\in(0,2)\), then for every \(x\in \mathbb{R}\), \(0<C(x)<\infty\) and {\parindent=7mm \begin{itemize}\item[(i)]if the relation \(n\xi_1-[n^{\delta}x]\notin d\mathbb{Z}\) holds with probability 1, then \(\mathrm{P}\{Z_n=[n^{\delta}x]\}=0\); \item[(ii)]if it belongs to \(d \mathbb{Z}\) with probability 1, then \[ \mathrm{P}\{Z_n=[n^{\delta}x]\}=d\frac{C(x)}{n^{\delta}}+o(n^{-\delta}) \text{ as } n\rightarrow\infty. \] \end{itemize}} In the lattice case and for \(\alpha<1\), a similar statement holds with \(n^{\frac{1}{\beta}}\) instead of \(n^{\delta}\) and \(D(x)=rf_{\beta}(rx)\) instead of \(C(x)\) for certain \(r\) depending on the total time spent in 0 by the two-sided random walk \((S_k, k\in\mathbb{Z})\). Theorem 3 deals with the strongly non-lattice case. Then the following holds: {\parindent=7mm \begin{itemize}\item[(i)]if \(\alpha>1\), \(\beta\in(0,2]\), then, for all \(a,b\in \mathbb{R}\), \(a<b\), \[ \lim_{n\rightarrow\infty}n^{\delta}\mathrm{P}\{Z_n\in[n^{\delta}x+a;n^{\delta}x+b]\}=C(x)(b-a); \] \item[(ii)]if \(\alpha<1\), \(\beta\in[0,2]\), a similar limit with \(n^{\frac{1}{\beta}}\) instead of \(n^{\delta}\) equals to \(D(x)(b-a)\). \end{itemize}} Random walks on randomly oriented lattices are considered in Section 5. Let \(\mu_x\) be a distribution on \(\mathbb{Z}\) in the domain of normal attraction of a centered stable distribution with \(1<\alpha\leq2\). Let \(\xi:=(\xi_y,y\in\mathbb{Z})\) be a sequence of independent centered \(\mathbb{Z}\)-valued random variables with distribution belonging to the domain of normal attraction of a stable distribution with index \(1<\beta\leq2\). For each \(y\in\mathbb{Z}\), \(\xi_y\) is the only horizontal displacement allowed on the line \(y\). Let \(0<p<1\). Given \(\xi\), the random walk \((M_n=M^{(1)}_n,M^{(2)}_n)\), \(n\geq0\), is a Markov chain starting at \(M_0:=(0,0)\) such that, at time \(n+1\), it moves either horizontally of \(\xi_{M^{2}_n}\) (with probability \(p\)) or makes a vertical jump with distribution \(\mu_x\) (with probability \((1-p)\)). The asymptotics of the probability \( \mathrm{P}\{M_n=(0,0)\}\) is considered as \(n\rightarrow\infty\) in Theorem 20. The paper is organized as follows. The introduction contains the problem, short remarks on references of related papers, the results and the outline of the proof. Sections 2, 3 and 4 contain the proofs of Theorems 1, 2 and 3, respectively. The local limit theorem for random walks on randomly oriented lattices is proved in Section 5. Finally, some auxiliary statements on the range of the random walk \(S_n\), \(n\geq1\), are obtained.