Conditional \(L^1\)-convergence for the martingale of a critical branching process in random environment (Q2135130): Difference between revisions

Let \(\{Z_n,\mathbf{P};n\ge 0\}\) be a critical Galton-Watson process in a random environment, the latter given as a sequence of i.i.d. random variables \(\{\xi_n;n\ge 0\}\) with values in a probability space \((\Theta,\mathcal{B})\), see \textit{W. L. Smith} and \textit{W. E. Wilkinson} [Ann. Math. Stat. 40, 814--827 (1969; Zbl 0184.21103)] and \textit{K. B. Athreya} and \textit{S. Karlin} [Ann. Math. Stat. 42, 1499--1520 (1971; Zbl 0228.60032); ibid. 42, 1843--1858 (1971; Zbl 0228.60033)]. The process may be represented in the form \(Z_n= \sum^{Z_{n-1}}_{i=1} Y_{i,n-1}\), with \(\{Y_{i,n-1}; i\ge 1\}\) a sequence of i.i.d. random variables with generating function \(f_{n-1}(s)= \sum^\infty_{k=0} p_k(\xi_{n-1})s^k\), where \(p(\theta)= \{p_k(\theta); k\ge 0\}\) is a measurable map defined on \(\Theta\) with values in the set of probability laws on \(\mathbb{N}_0\). The martingale of interest is \(W_n= Z_n/\exp S_n\), \(n\ge 0\), where \(S_0=0\), \(S_n= \sum^n_{i=1} \log f_{i-1}'(1)\). It converges a.s. to \(W\). Define \(L(n):= \min\{S_k; 0\le k\le n\}\), \(\tau(0):= 0\), \(\tau_{k+1}:= \min\{n\ge\tau(k): S_n< S_{\tau(k)}\}\), \(k\ge 0\), and \(R(x):= \sum^\infty_{k=0} \mathbf{P}(S_{\tau(k)}\ge-x)\), \(x\in\mathbb{R}\). Then \((R(S_n)\cdot 1_{\{L(n)\ge 0\}}, n\ge 0)\) is a martingale with respect to \(F(n):= \sigma(Z_1,\dots,Z_n, f_0,\dots,f_{n-1})\), leading to a probability \(\mathbf{P}^+\) such that \(d\mathbf{P}^+|_{F(n)}= R(S_n)\cdot 1_{\{L(n)\ge 0\}}d\mathbf{P}|_{F(n)}\). Assuming \(\mathbf{E}\log f_0'(1)= 0\) and \(\mathbf{E}\log^2f_0'(1)\in (0,\infty)\), the authors provide sufficient conditions for \(W_n\) to converge in \(L^1\) under \(\mathbf{P}^+\) to \(W\) with \(\mathbf{E}_\xi W=1\) \(\mathbf{P}^+\)-a.s., respectively \(\mathbf{E}_\xi W= 0\) \(\mathbf{P}^+\)-a.s., where \(\mathbf{E}_\xi\) is the expectation respective the (quenched) conditional probability \(\mathbf{P}_\xi\) given the random environment \(\xi\).