Superprocesses with spatial interactions in a random medium (Q1045788)

The author considers the following stochastic equation: \[ Z_t = Z_0 + \int_0^t \sigma_1 ( X_s, Z_s ) d y(s) + \int_0^t \int_{ {\mathbb R}^m} \sigma_2( X_s, Z_s, \xi ) d W(s, \xi) + \int_0^t b(X_s, Z_s) ds,\tag{1} \] \[ X_t(A) = \int 1 ( Z_t(y) \in A ) K_t( dy), \quad \forall A \in {\mathcal B}({\mathbb R}^d), \tag{2} \] where \(\sigma_1 : M_F( {\mathbb R}^d) \times {\mathbb R}^d \to {\mathbb R}^{d \times d}\), \(\sigma_2\): \(M_F( {\mathbb R}^d)\) \(\times\) \({\mathbb R}^d \times\) \({\mathbb R}^m\) \(\to\) \({\mathbb R}^{d \times d}\), \(b :\) \(M_F( {\mathbb R}^d) \times {\mathbb R}^d\) \(\to\) \({\mathbb R}^d\), \(M_F({\mathbb R}^d)\) denotes the space of all non-negative finite measures on \({\mathbb R}^d\), \(K\) is a historical super-Brownian motion, \(y\) is a path in \({\mathbb R}^d\), and \(W\) is a \(d\)-dimensional white noise on \({\mathbb R}_+ \times {\mathbb R}^m\) that is independent of \(K\). Note that \(Z\) takes values in the space of paths in \({\mathbb R}^d\), while \(X\) takes values in the space of paths in \(M_F({\mathbb R}^d)\). In connection with plankton dynamics, \textit{G. Skoulakis} and \textit{R. J. Adler} [Ann.\ Appl.\ Probab.\ 11, No. 2, 488--543 (2001; Zbl 1018.60052)] studied the special case of the above model with a Brownian motion \(B\) instead of the white noise \(W\). While \textit{D. Dawson, Z. Li} and \textit{H. Wang} [Electron.\ J.\ Probab.\ 6, Paper No. 25, 33p. electronic only (2001; Zbl 1008.60093)] studied the interacting particle systems whose particle moves according to a stochastic equation similar to (1), in order to construct a measure-valued diffusion. The purpose of this paper is to construct a new class of interactive measure-valued diffusions driven by a historial super-Brownian motion \(K\) and an independent white noise \(W\) by solving the stochastic equation (1)--(2). The author points out in the process of proceeding the study that the approach of \textit{E. Perkins} [Lect.\ Notes Math.\ 1781, 125--329 (2002; Zbl 1020.60075)] is still valid for the study of the problem examined by Dawson et al. [op.\ cit.\ (2001; Zbl 1008.60093)], and the method adopted in this paper is a kind of unification and extension of both studies by Dawson et al. and Perkins. Under the usual Lipschitz conditions on the terms \(\sigma_1\), \(\sigma_2\) and \(b\), the author proves the existence of a pathwise unique solution \((X, Z)\) to the stochastic equation (1)--(2), such that the map \(t \mapsto X_t\) is a.s. continuous in \(t\). In addition, it is shown that the solutions satisfy the natural martingale problem. More precisely, when for a function \(f =\) \(\otimes_{i=1}^n \phi_i\): \({\mathbb R}^{nd} \to {\mathbb R}\) with \(\phi_i \in C_b^2({\mathbb R}^d)\), \(F_{n,f}(\mu)\): \(M_F({\mathbb R}^d) \to {\mathbb R}\) is given by \(F_{n,f}(\mu)\) \(=\) \(\int f d \mu^n\), then \[ N_t(n,f) = F_{n,f}(X_t) - F_{n,f}(X_0) - \int_0^t \hat{L} F_{n,f}(X_s) ds \] is a martingale, where \[ \begin{multlined} \hat{L} F_{n,f}(\mu) = \sum_{i=1}^n \mu( L \phi_i) ( \prod_{ j \not= i} \mu ( \phi_j)) + \frac{1}{2} \sum_{ {\scriptstyle i,j = 1 \atop {\scriptstyle i \not= j}} }^n ( \prod_{ k \not= i, j} \mu( \phi_k)) \times \\ \times [ \mu( \phi_i \phi_j) + \int_{ {\mathbb R}^m } \mu( \nabla \phi_i \sigma_2 (\xi)) \cdot \mu( \nabla \phi_j \sigma_2 (\xi)) d \xi ] \end{multlined} \] with some suitable generator \(L\). Finally, the strong Markov property and the compact support property for the solutions of (1)--(2) are proved as well.