Exit probability of two-dimensional random walk from the quadrant (Q1302098)

Let \(Z_0=(0,0)\), \(Z_1=(X_1,Y_1)\), \(Z_2=(X_2,Y_2)\),... be a random walk in the two-dimensional integer lattice \({\mathbb{Z}}^2\). The random walk is a sum of i.i.d. random vectors with common distribution \(F\): \(Z_n=\xi_1+...+\xi_n, n=1,2,...\) Let \(D=\{(x,y)\in {\mathbb{R}}^2: x>0,y>0\}\) be the first quadrant in \({\mathbb{R}}^2\). Assumption 1.1. For every \(\theta=(\theta_1,\theta_2)\) in \({\mathbb{R}}^2\), \(\lambda:={\mathbb{E}}e^{\langle\theta Z_1\rangle} <\infty.\) Assumption 1.2. \(\mu:={\mathbb{E}}Z_1\in D\), and \({\mathbb{P}}(X_n>0,Y_n<0)>0\) for some positive integer \(n\). Assumption 1.3. The \(y\)-coordinate of the random walk is left-continuous, that is, \({\mathbb{P}}(Y_1 \{-1,0,1,2,...\})=1\). Let \(a\) and \(b\) be positive integers. The author takes \(a\) arbitrarily fixed. Set \(T_b:=\inf\{n\geq 1: (a,b)\not \in D\}, r_b={\mathbb{P}}(T_b<\infty, a+X_{T_b}>0).\) Since \(Z_n\sim n\mu\) a.s. (\(n\to \infty\)), we have \(r_b\to 0\) (\(b\to \infty\)) from Assumption 1.2. The purpose of the paper is to study the decay of \(r_b\) to \(0\). Denote \[ \tilde{\theta}_2:=\inf\{\theta_2\in {\mathbb{R}}^1:\lambda((0,\theta_2))=1\},\quad \lambda_1= \left.\frac{\partial \lambda(\theta)}{\partial \theta_1}\right |_{\theta=(0,\tilde{\theta}_2)}. \] Let us state the main result of the paper. If \(\lambda_1>0\), then (Theorem 2.1) \(r_b\sim K_1\exp\{\tilde{\theta}_2 b\}\) \((b\to \infty);\) if \(\lambda_1=0\), then (Theorem 2.2) \(r_b\sim K_2b^{-1/2}\exp\{\tilde{\theta}_2 b\}\) \((b\to \infty);\) if \(\lambda_1<0\), then (Theorem 2.3) \(r_b=O(b^{-3/2}\exp\{\overline{\theta}_2 b\})\) \((b\to \infty),\) where \(K_1\), \(K_2\) are two positive constants depending only on \(F\) and \(a\), \(\overline{\theta}_2:=\inf\{\theta_2\in {\mathbb{R}}^1: \lambda((\theta_1,\theta_2))=1\}\) and \(\overline{\theta}_2>-\infty\) (Assumption 2.1).