Random-walk in beta-distributed random environment (Q525105)

The paper under review introduces the beta RWRE (random walk environment) model in space-time i.i.d. random walk on \(\mathbb Z\) which performs nearest neighbour jumps with transition probabilities following the beta distribution and to prove second-order cubic-root scale corrections to the large deviation principle with limiting theorem in terms of the maximum of strongly correlated random variables. The definitions and main results are stated in Section 2. Let \(B_{i, j}\) be independent random variables distributed by the beta distribution with parameters \(\mu\) and \(\nu - \mu\), where \(0< \mu < \nu\), \[ P(B_{i, j} \leq r) =\int_0^r x^{\mu} (1-x)^{\nu - \mu -1} \frac{\Gamma (\nu)}{\Gamma (\mu) \Gamma (\nu - \mu)} dx, \] and a point-to-point beta polymer is a measure \(Q_{t, n}\) on lattice paths \(\pi\) between \((0, 1)\) and \((t, n)\) with \[ Q_{t, n}(\pi) = \frac{\prod_{e\in \pi} w_e}{Z(t, n)}, \] and the partition function \(Z(t, n)=Z(t-1, n)B_{t, n} + Z(t-1, n-1)(1-B_{t, n})\) if \(t\geq n >1\); \(Z(t, t+1)=Z(t-1, t)\) if \(t>0\) and \(Z(t, 1) = Z(t-1, 1)B_{t, 1}\) for \(t>0\). Note that the free energy of the beta polymer is \(\log Z(t, n)\) with \(Z(0, 1)=1\). Section 3 devotes to explain how the beta-RWRE and the beta polymer arise as limits of the \(q\)-Hahn TASEP in [\textit{A. M. Povolotsky}, J. Phys. A, Math. Theor. 46, No. 46, Article ID 465205, 25 p. (2013; Zbl 1290.82022)]. The \(N\)-particle \(q\)-Hahn TASEP is a discrete time Markov chain \(\{x_n(t)\}_{0\leq n \leq N}\in X_{N}\), where each coordinate \(x_n(t)\) is updated independently and \(x_n(t+1)=x_n(t)+j_n\) for \(0\leq j_n \leq x_{n-1}(t) - x_n(t) -1\) with \(q\)-Hahn probability distribution \(\phi_{q, \tilde{\mu}, \tilde{\nu}}(j| m)\). With \(q=e^{-\varepsilon}\), \(\tilde{\mu}=q^{\mu}\), \(\tilde{\nu} =q^{\nu}\), Proposition 3,1 shows that the rescaled quantity \(Z^{\varepsilon}(t, n) =q^{x_n(t)+n}\) converges weakly as processes to the partition function \(Z(t, n)\) of the beta polymer with parameters \(\mu\) and \(\nu - \mu\) where \(0< \mu <\nu\). Proposition 3.3 gives the Fredholm determinant formula for the Laplace transform of \(Z(t, n)\) as a limit version of Theorem 1.10 of the second author [Int. Math. Res. Not. 2015, No. 14, 5577--5603 (2015; Zbl 1335.82018)], \[ E[e^{uZ(t, n)}] = \det (I + K_u^{BP})_{L^2(C_0)}, \] where \(K_u^{BP}: L^2(C_0) \to L^2(C_0)\) is an integral kernel for a small positively oriented circle \(C_0\) containing 0 but not \(- \nu\) nor \(-1\). Proposition 2.6 shows the equivalence of the beta RWRE and the beta polymer for \(Z(t, n) = P(t, t-2n+2)=P(X_t \geq t-2n+2)\) for the beta RWRE, which is proved in Subsection 3.3. Section 4 illustrates the exact moment formula first in Proposition 4.4 for various contours, in Proposition 4.6 for all same contours as a suitable limit of Theorem 1.8 of the second author [loc. cit.]. The proof of Theorem 2.12, as well as Proposition 3.3, is given in Subsection 4.2 by verifying the power series of \(e^{u Z(t, n)}\) with each term in the Fredholm determinant expression. In Section 5, Proposition 2.10 is proved stating that the Bernoulli-exponential first passage percolation model is the zero-temperature limit of the beta-RWRE using change of variable and limits, and the proof of Theorem 2.18 \[ P(T(n, m)>r) =\det (I + K_u^{\mathrm{FPP}})_{L^2(C_0)}, \] where \(T(n, m)\) is the first passage time to the set \(D_{n, m}\) in the Bernoulli-exponential FPP model by splitting into two pieces from the convergence of \(E[e^{uP_{\varepsilon}(n+m, n-m)}]\) to \(P(T(n, m)\geq r)\) on the left hand and from the convergence of \(E[e^{uP_{\varepsilon}(n+m, n-m)}]\) to \(\det (I + K_u^{\mathrm{FPP}})_{L^2(C_0)}\) on the right hand with various technical estimates and contour integrals. Section 6 mainly deals with the proof of Theorem 6.2 that \[ \lim_{t\to \infty} P(\frac{\log (P(t, x(\theta)t)) + I (x(\theta)t)}{t^{1/3}\sigma (\theta) }\leq y) = F_{\mathrm{GUE}}(y), \] where \(x(\theta) = (1+2\theta)/(2 \theta^2+2\theta +1), I(x(\theta))= 1/(2 \theta^2+2\theta +1)\) for \(\alpha = \beta\) and \(\sigma^3 (\theta) = 1/(\theta + 3\theta^2 + 4\theta^3 + 2 \theta^4)\) for \(\alpha = \beta = 1\) and \(0<\theta <1/2\), \(F_{\mathrm{GUE}}(y) = \det (I - K_{Ai})_{L^2(y, \infty)}\) is the GUE Tracy-Widom distribution with the Airy kernel \(K_{Ai}\). This is an application of Theorem 2.13 for the Fredholm determinant as \(t\to \infty\). \textit{A. Borodin} and the second author [Probab. Theory Relat. Fields 158, No. 1--2, 225--400 (2014; Zbl 1291.82077)] proved that the continuous probability distribution function converges in Lemma 4.1.39 for the left hand side of the moment function \(E[e^{uP(t, x(\theta)t)}]\), the other side follows from the asymptotic analysis of the Fredholm determinant by steepest descent analysis from various previous works. Subsection 6.2 gives precise estimates and steepest-descent properties and relation to extreme value theory. Section 7 goes further to analyze the behaviour of the first passage time \(T(n, kn)\) for \(n\to \infty\). Theorem 7.1 states that \[ \lim_{n\to \infty} P (\frac{T(n, k(\theta)n) - \tau (\theta)n}{\rho (\theta ) n^{1/3}} \leq y) = F_{\mathrm{GUE}}(y), \] by Theorem 2.18 with \(r= \tau (\theta)n+ \rho (\theta ) n^{1/3}y\) and the limit \(\lim_{n\to \infty}\det (I - K_r^{\mathrm{FPP}})_{L^2(C_0^{'})}\). Then the proof is completed by first finding steepest-descent contours for the kernel \(K_r^{\mathrm{FPP}}\) in several lemmas and Proposition 7.5 with the required properties for the contour path, and by the limit similar to Proposition 6.7 in Proposition 7.6 to get into the Airy kernel.