Poisson limits of sums of point processes and an particle-survivor model (Q2467119)

\textit{R. Durrett} and \textit{V. Limic}, [Stochastic Process. Appl., 102, 301--309 (2002; Zbl 1075.60582)] gave a formal treatment of species competition processes. They studied a system in which particles arrive to the interval \((0,1)\) according to a Poisson process, and the location of each particle in the interval has the uniform distribution, independent of everything else. The location of a particle represents the particle's rank, which affects how long it will survive. Whenever a new particle arrives, each particle currently in the interval with a rank below that of the arrival is independently deleted with a fixed probability. The evolution over time of the point process of particles in the interval is represented by a Markov jump process on a space of counting measures. Durrett and Limit [loc. cit.] proved that this Markov process converges to a Poisson process on \((0,1)\). In trying to determine why Poisson processes arise as limits in such applications, the authors of the present paper studied several types of sums of point processes with dependencies that are abstractions of those in the particle process. This led to a general Poisson convergence theorem for sums of dependent point processes (Theorem 1 (p. 267) is a natural framework for establishing Poisson limits of particle-survivor processes). The contents of the remaining sections are as follows. Section 2 contains the main result for the sums of point processes and several corollaries for special cases. Section 3 describes the particle-survivor model and its Poisson limit. Sojourn times of particles in this system and properties of departures are the topics of Section 4. Section 5 gives necessary and sufficient conditions for the stationary distribution of a Markovian particle-survivor process to be that of a Poisson process.