Local trapping for elliptic random walks in random environments in \(\mathbb {Z}^d\) (Q737318): Difference between revisions

The paper under review studies under which ellipticity conditions local trapping occurs. \textit{A.-S. Sznitman} [ICTP Lecture Notes 17, 203--266 (2004; Zbl 1060.60102)] expected that if the transition probabilities are uniformly elliptic, then the random walks in i.i.d. random environments on \(\mathbb{Z}^d\) are ballistic. \textit{C. Sabot} [Ann. Probab. 41, No. 2, 722--743 (2013; Zbl 1269.60077)] and \textit{E. Bouchet} [Electron. J. Probab. 18, Paper No. 58, 25 p. (2013; Zbl 1296.60267)] explained that the random walks become sub-ballistic even if the transition probability is elliptic, but not uniformly elliptic. Sznitman [loc. cit.] conjectured that under the uniformly elliptic i.i.d. environment the transience of the walk in direction \(l\in S^{d-1}\) for \(d\geq 2\) implies the ballisticity of the walk in direction \(l\). The main difficulty in proving this conjecture is to understand how the walk explores in the space. Section 2 introduces a central tool on regeneration times for determining whether or not a walk is ballistic. The random walk in the environment \(\omega\) started from \(x\) is the Markov chain \((X_n)_{n\geq 0}\) in \(\mathbb{Z}^d\) with the (quenched) law \(P_x^{\omega} (P_x^{\omega}[X_0=x]=1, P_x^{\omega}[X_{n+1}=x+e|X_n=x]=p^{\omega}(x, e))\) for any \(x\in \mathbb{Z}^d\) and \(e\in U\). The basic regeneration time is \(\tau_1=S_K\) with \(K=\inf \{k\geq 1: S_k<\infty, R_k=\infty\}\), where \(S_0=0\), \(M_0=X_0\cdot l\), \(S_{k+1} =T_{H^+(M_k+a)}\leq \infty\), \(R_{k+1} =D \circ \theta_{S_{k+1}} + S_{k+1}\leq \infty\) (\(\theta\) is the shift operator) with \(D=\inf\{n\geq 0: X_n\cdot l < X_0 \cdot l\}\), \(M_{k+1}=\sup_{n\leq R_{k+1}}X_n\cdot l \leq \infty\), and \[ \tau_{k+1} = \tau_1 + \tau_k(X_{\tau_1^+} - X_{\tau_1}, \omega (\cdot + X_{\tau_1})), \;\;k\geq 0. \] \textit{A.-S. Sznitman} and \textit{M. Zerner} [Ann. Probab. 27, No.4, 1851--1869 (1999; Zbl 0965.60100)] showed that if the random walk is transient in the direction \(l\), then for all \(x\in \mathbb{Z}^d\), \(\tau_k < \infty\) almost surely in the quenched law for a RWRE (random walk in random environment) in an elliptic i.i.d. environment, where a RWRE is transient in the direction \(l\) if \({\mathbf P}[A_l]=1\) with \(A_l=\{\lim_{n\to \infty} X_n\cdot l\}\) and the average annealed law \({\mathbf P}_x =\int P^{\omega}_x dP\). They further showed the fundamental renewal property. Section 3 first exhibits a criterion for zero-speed (Theorem 3.1). The conditions \((K)_1\) (see Definition 3.1) and \((E)_0\) (defined in (1.6)) imply positive speed under sufficiently strong transience conditions (Theorem 3.2), and the annealed central limit theorem under some stronger assumptions (Theorem 3.3). There are technical criterions like \((K)_{\alpha}\) and \((H)_{\alpha}\) as follows. Define the unit hypercube located at \(x\) to be the set \(h_x=\{x+\sum_{i=1}^d\varepsilon_i e_i: \varepsilon \in [0, 1], i=1, 2, \cdots, d\}\). The \((H)_{\alpha}\) criterion is the condition \(\max_{y\in h_x} E_y[(T^{\mathrm{ex}}_{h_x})^{\alpha}]=\infty\), and the criterion \((K)_{\alpha}\) consists of the hypotheses (1) there exists \(\gamma_y \in \mathbb{R}^+\) such that \(E[(Q_y^{h_x})^{- \gamma_y}]< \infty\) for all \(y\in h_x\), (2) there exists a marked Markovian hypercube \((h(\omega), (\alpha_y(\omega))_{y\in h_x})\) such that \(E[\Pi_{y\in h_x}(\tilde{Q}^{h(\omega)_{0, x_0(\omega)+x}})^{-\alpha_y (\omega)} ]<\infty\), and (3) there exists \(\varepsilon > 0\) such that \(\sum_{y\in h_x}(\gamma_y \wedge \alpha_y (\omega)) \geq \alpha + \varepsilon\) almost surely. A RWRE satisfies condition \((E)_0\) if for all \(e\in U\) there exists \(\eta_e > 0\) such that \(E[(p^{\omega (0, e)})^{-\eta_e}] <\infty\). Theorem 3.1 basically states that \((H)_1\) implies that the RWRE has zero speed under the elliptic i.i.d. environment and transience in direction \(l\). Theorem 3.2 states that a RWRE in an elliptic i.i.d. environment with conditions \((E)_0, (K)_1\) and \((P)_M^l\) for some \(M\geq 15 d +5\) is ballistic in the direction \(l\), where the RWRE is of type \((P)_M^l\) if for all \(L\geq 2\cdot 3^{29d-1}\) there exists \(L^{'} \leq \frac{5L}{4}\) and \(\tilde{L}\leq 72 L^3\) such that \({\mathbf P}[lX_{T^{\mathrm{ex}}_{B^l_{L, L', \tilde{L}}}}< L] \leq L^{-M}\). Theorem 3.3 shows that if a RWRE satisfies \((P)_M^l, (E)_0\) and \((K)_2\), then an annealed central limit theorem holds: \(\varepsilon^{1/2}(X_{[\varepsilon^{-1}n]} - [\varepsilon^{-1}n]v)\) converges in law under \({\mathbf P}\) as \(\varepsilon \to 0\) to a Brownian motion with non-degenerate covariance matrix. The quenched central limit theorem holds under the same conditions. Section 4 is devoted to the intuitive interpretations of these conditions and reasonings. The condition \((K)_1\) is typically verified in the positive speed regime, is close to complementary with \((H)_1\), and also related to exit times of hypercubes. The reasons why one can apply these criterions and why they are general are discussed in subsection 4.2, the hypercubes occur in \S 4.3 in Proposition 4,4. Section 5 is devoted to the proofs of Theorem 3.2 and Theorem 3.3. The proof of Theorem 3.2 follows from Proposition 5.1 using earlier results collected in Theorem 2.2. On the other hand, the proof of Theorem 3.3 follows from Theorem 2.2 and the main result of \textit{E. Bouchet} et al. [Stochastic Processes Appl. 126, No. 4, 1206--1225 (2016; Zbl 1335.60049)] for the quenched case. Section 6 is devoted to the proof of Theorem 3.1 in the zero-speed regime. The proof follows from the lower bounds on the regeneration times, from the tail estimates on the exit times of hypercubes (Lemma 6.1), and from verifying \((E)\) and \((H)_{\alpha}\) for \(\alpha > 0\) and using Theorem 2.2. It would be better to have simpler and sharper criterions for the ballistic behavior.