\(\mathbb N\)-measure for continuous state branching processes and its application (Q1946943)

The authors give a direct construction of the \(\mathbb{N}\)-measure of continuous-time, continuous-state branching processes. In particular, their approach avoids the more general model of superprocesses in the context of which the existence of the \(\mathbb{N}\)-measure has been established before. To be more precise, let \(X = (X_t: t \geq 0)\) be a continuous-time, continuous-state branching process and denote by \(\operatorname{P}_x\) the law of the process on the canonical path space \((\Omega,\mathcal{F})\) (of right-continuous functions with left limits) when started at \(x \in (0,\infty)\). The main result of the paper is the following. Suppose that \(\operatorname{P}_x\) satisfies \(\operatorname{P}_x(X_t = 0) > 0\) for all \(t > 0\). Then there exists a unique measure \(\mathbb{N}_x\) on \((\Omega,\mathcal{F})\) such that (i) for any \(n \in \mathbb{N}\), and \(t_i, \lambda_i \geq 0\), \(i=1,\dots,n\), \[ \mathbb{N}_x \bigg( 1-\exp\bigg(\sum_{i=1}^n \lambda_i X_{t_i} \bigg)\bigg) = -\log \operatorname{P}_x \bigg[\exp \bigg(-\sum_{i=1}^n \lambda_i X_{t_i}\bigg) \bigg], \] where \(X\) is assumed to be canonical, i.e., \(X_t(\omega) = \omega_t\); (ii) \(\mathbb{N}(\cap_{t \geq 0} \{X_t=0\}) = 0\). \(\mathbb{N}_x\) can be thought of as the Lévy measure of \(\operatorname{P}_x\). As applications of the existence of the unique \(\mathbb{N}\)-measure, the authors (a) prove necessity and sufficiency of a condition for a continuous-state branching process with immigration to converge a.s. to some finite constant -- a result that has first been formulated by \textit{M. A. Pinsky} [Bull. Am. Math. Soc. 78, 242--244 (1972; Zbl 0252.60031)]; (b) give a conceptual proof of an \((L \log L)\)-criterion for the martingale associated with a continuous-state branching process to converge in \(\mathcal{L}^1\).