Degenerate stochastic differential equations and super-Markov chains (Q1849350): Difference between revisions

Let \(\gamma_i\), \(b_i :\mathbb R^d_+\to\mathbb R\) be continuous functions and each \(\gamma_i\) be strictly positive. Consider the operator \(\mathcal L\) on \(C^2(\mathbb R^d_+)\) defined by \[ \mathcal Lf(x)=\sum_{i=1}^dx_i\gamma_i(x)\frac{\partial^2f}{\partial x_i^2}(x)+ b_i(x)\frac{\partial f}{\partial x_i}(x),\quad x\in\mathbb R^d_+, \] and the diffusion \(X_t\) associated to \(\mathcal L\) is the process on \(\mathbb R^d_+\) that solves the stochastic differential equation \[ dX_t^i = \sqrt{2X_t^i\gamma_i(X_t)} dB_t^i + b_i(X_t) dt,\quad X_t^i\geq 0,\quad i =1,\dots,d,\tag{1} \] where \(B_t\) is a standard \(d\)-dimensional Brownian motion. The purpose of this paper is to prove uniqueness of the martingale problem for the operator \(\mathcal L\). As is well known, this is equivalent to prove weak uniqueness (i.e., uniqueness in law) to the solution of (1). Assume that for all \(i=1,\dots,d,\) \[ b_i(x)>0,\;x\in\partial\mathbb R^d_+, \quad |b_i(x)|\leq C(1+\|x\|),\;x\in\mathbb R^d_+,\quad \|x\|=\max_{i=1,\dots,d}|x_i|. \] Then for any initial law \(\nu\), there exists a unique solution to the martingale problem for \(\mathcal L\). If \(\mathbb P^x\) is the solution with initial law \(\delta_x\), then \((\mathbb P^x, X_t)\) is a strong Markov process, and for any bounded measurable function \(f\) on \(\mathbb R^d_+\), its resolvent \(S_\lambda f(x)=\mathbb E^x(\int_0^\infty e^{-\lambda t}f(X_t) dt)\) is a continuous function of \(x\). As a special case, the uniqueness in law for some classes of super-Markov chains with state dependent branching rates and spatial motions is established. The one-dimensional counterexample is given which shows that the condition \(b_i(x)>0\) on \(x\in\partial\mathbb R^d_+\) cannot be weakend. Examples of super-Markov chains, generalized mutually catalytic branching and stepping stone models, are considered.