Path-valued branching processes and nonlocal branching superprocesses (Q2438745): Difference between revisions

This paper treats a continuous-state branching process (CBI-process) with immigration which is constructed by a stochastic equation driven by time-space noises and can also be regarded as an inhomogeneous path-valued branching process. More precisely, let \(( \Omega, {\mathcal F}, {\mathcal F}_t, {\mathbb P})\) be a filtered probability space, and \(W(t, \cdot)\), \(t \geq 0\), be an \(( {\mathcal F}_t)\)-white noise on \(( 0, \infty)\). The processes \(p_0(t)\) and \(p_1(t)\) \(( t \geq 0)\) denote \(( {\mathcal F}_t)\)-Poisson point precesses on \(( 0, \infty)^2\) with characteristic measures \(m(d z) du\) and \(n(dz) du\), respectively, while \(N_0( ds, dz, du)\) and \(N_1( ds, dz, du)\) are the Poisson random measures on \(( 0, \infty)^3\) associated with \(p_0(t)\) and \(p_1(t)\), respectively. In addition, let \(\tilde{N}_0( ds, dz, du)\) denote the compensated random measure associated with \(p_0(t)\). When \(z \cdot n(dz)\) is a finite measure on \((0, \infty)\), an immigration mechanism \(\psi\) is given by \[ \psi(\lambda) = h \lambda + \int_0^{\infty} ( 1 - e^{- z \lambda} ) n( dz) \tag{1} \] for \(\lambda \geq 0\) and \(h \geq 0\). First of all, the author shows that a family of CBI-processes is constructed as the solution flow of a stochastic equation system driven by time-space noises. As a matter of fact, the statement is as follows: if \(\rho(t)\), \(t \geq 0\), is a deterministic locally bounded positive Borel function, then the solution \(Y_t\), \(t \geq 0\), of the stochastic equation \[ \begin{aligned} Y_t &= Y_0 + \sigma \int_0^t \int_0^{ Y_{s-}} W( ds, du) + \int_0^t \int_0^{\infty} \int_0^{ Y_{s-}} z \cdot \tilde{N}_0 ( ds, dz, du) \tag{2} \\ &+ \int_0^t ( h \rho(s) - b Y_{s-} ) ds + \int_0^t \int_0^{\infty} \int_0^{\rho(s)} z \cdot N_1(ds, dz, du), \quad \sigma \geq 0,\, b \in {\mathbb R}, \end{aligned} \] is an inhomogeneous CBI-process with transition semigroup \(\{ P_{r,t}^{\rho}\): \(t \geq r \geq 0 \}\) defined by \[ \int_{ {\mathbb R}_+} e^{- \lambda y} P_{r,t}^{\rho}(x, dy) = \exp \left\{ - x \cdot v_{t-r}(\lambda) - \int_r^t \psi( v_{t-s}(\lambda)) \rho(s) ds \right\} \tag{3} \] for \(\lambda \geq 0\), where \(v_{t-r}\) satisfies the so-called log-Laplace equation. Here, a positive Markov process \(Y_t\) with inhomogeneous transition semigroup \(( P_{r,t}^{\rho} )\), \(t \geq r \geq 0\), is called an inhomogeneous CBI-process with immigration rate \(\rho\) \(=\) \(\{ \rho(t)\) : \(t \geq 0 \}\). Moreover, he shows that the above constructed family of CBI-processes can be regarded as an inhomogeneous increasing path-valued branching process with immigration. Furthermore, two nonlocal branching immigration superprocesses can be defined from the solution flow of a similar type of stochastic equations as (1), and the properties of the process under an excursion law are studied as well. The results provide new perspectives into the tree-valued Markov processes of \textit{D. Aldous} and \textit{J. Pitman} [Ann. Inst. Henri Poincaré, Probab. Stat. 34, No. 5, 637--686 (1998; Zbl 0917.60082)]. This paper is technically due to the author's monograph [Measure-valued branching Markov processes. Berlin: Springer (2011; Zbl 1235.60003)], for instance, in derivation of Markov processes with transition semigroups, he studies properties of stochastic equations driven by Poisson random measures, and he verifies the existence of the corresponding excursion laws, and so on. As to other related works (see, e.g. [\textit{R. Abraham} and \textit{J.-F. Delmas}, Ann. Probab. 40, No. 3, 1167--1211 (2012; Zbl 1252.60072)]), where the tree-valued processes associated with general continuous-state branching processes (CB-processes) were studied; and see also [\textit{D. A. Dawson} and \textit{Z. Li}, Ann. Probab. 40, No. 2, 813--857 (2012; Zbl 1254.60088)] for construction of measure-valued processes from flows of a certain class of stochastic equations.