Measure-valued discrete branching Markov processes (Q2790652)

Let \(E\) be a Lusin space and \(\widehat E\) be the set of all finite sums of Dirac measures on \(E\). Suppose to be given a standard Markov process (base process) on \(E\), a bounded killing density on \(E\) (stopping the base process for the purpose of branching), and a branching kernel with bounded first moment. Assuming sufficient separability, the authors construct from these data the transition semigroup of a standard Markov branching process in \(\widehat E\). Except for the more general space \(E\), the construction of the semigroup via solving the evolution equation of the (nonlinear) generating semigroup essentially follows [\textit{N. Ikeda} et al., J. Math. Kyoto Univ. 9, 95--160 (1969; Zbl 0233.60070)], see also [\textit{M. L. Silverstein}, Z. Wahrscheinlichkeitstheor. Verw. Geb. 9, 235--257 (1968; Zbl 0177.45801)]. The proof that the corresponding branching process has a standard version differs, however, from [\textit{N. Ikeda} et al., J. Math. Kyoto Univ. 8, 365--410 (1968); correction ibid. 11, 195--196 (1971; Zbl 0233.60069)], where regularity was obtained as part of the probabilistic construction of the process by piecing out from branching event to branching event. The authors here use potential theoretical methods based on the existence of a suitable harmonic function with compact level sets.NEWLINENEWLINE The formal framework of this paper is particularly well suited for taking advantage of the freedom in choosing the base process: the authors discuss an example in which the base process itself is a (continuous) branching process, so that the construction yields a superprocess, and they continue by taking this superprocess as base process.