A phase diagram for a stochastic reaction diffusion system (Q639877)

The paper is devoted to the following stochastic reaction diffusion equation on \(\mathbb R^d\) (\(d\geq 3\)), \[ \partial_tu =\Delta u +\beta uv-\gamma u+\sqrt{u}\,\dot{W}, \] where \(u\geq 0\) describes the distribution of a population on \(\mathbb R^d\) at time \(t\geq 0\) and \(v\geq 0\) is the density of a nutrient gradually used up by the population following the rule \(\partial_tv =-uv\) with initial density \(v_0 =1\). Here, \(\dot{W}\) denotes a space-time white noise. The main concern lies in the long-time behaviour of solutions. Certain death will refer to the situation when a.s. \(u_t = 0\) for large \(t\), and possible life will refer to \(\text{P}[u_t > 0\;\forall t] > 0\) for any non-zero initial condition \(u_0\). The main result is that, when \(d = 1\), certain death occurs for any \(\gamma,\beta\geq 0\), whereas, when \(d = 2\) or \(d = 3\), there exists a non-decreasing function \(\beta\rightarrow \Psi(\beta)\in [0,\beta)\) such that, when \(0\leq \gamma<\Psi(\beta)\), possible life occurs and, when \(\gamma>\Psi(\beta)\), certain death occurs. The first task is to construct solutions to the equation. This is done by a change of measure starting from the case \(\beta=0\) which corresponds to Dawson-Watanabe processes. The results on possible life use a construction of a supercritical oriented percolation process coupled to the solution. To set up the percolation comparison, solutions to equation are constructed on larger and larger domains with Dirichlet boundary conditions under appropriate monotony properties on the associated exit measures. The results on certain death are simpler, and the underlying idea is a comparison with a subcritical branching process.