Title: Ray-Knight compactification of birth and death processes

Abstract: A birth and death process is a continuous time Markov chain with the minimal state space $\mathbb{N}$, whose transition matrix is standard and whose density matrix is the given birth-death matrix. Birth and death processes are unique if and only if $\infty$ is entrance or natural. When $\infty$ is neither entrance nor natural, there have appeared two methods in history to obtain all birth and death processes. The first one is an analytic treatment put forward by Feller in 1959, and the second is a probabilistic construction completed by Wang in 1958.
In this paper we will raise a third way by virtue of the Ray-Knight compactification. This new way has benefits of both analytic and probabilistic treatments. By Ray-Knight compactification, every birth and death process admits a modification regarded as a Ray process on $\mathbb{N}_\infty=\mathbb{N}\cup \{\infty\}$. All of them are further classified into two classes: One is the class of celebrated Doob processes and the other one is the class of Feller $Q$-processes. Every birth and death process in the second class is actually a Feller process on $\mathbb{N}_\infty$. We will derive the expression of its infinitesimal operator, which indicates probabilistic heuristics for its boundary behaviours at $\infty$. Furthermore, utilizing transformations of killing and Ikeda-Nagasawa-Watanabe's piecing out, we will also give a probabilistic construction for all birth and death processes.