Chase-escape with death on trees (Q2057213)

This paper is a detailed study of the chase-escape process with death on infinite \(d\)-ary trees. In this model, each vertex has one of four states \((w, r, b, \dagger)\). When two adjacent vertices have states \((r,w)\), then with rate \(\lambda>0\) they become \((r,r)\); when two adjacent vertices have states \((b,r)\), then with rate 1 they become \((b,b)\). Finally, vertices which are on state \(r\) transfer to state \(\dagger\) at rate \(\rho >0\). Initially, only two adjacent vertices are in states \((r,b)\), respectively, whereas all other vertices are in state \(w\). Vertices in state \(r\) can be interpreted as prey-particles which escape to empty vertices in state \(w\) at rate \(\lambda\). Vertices in state \(b\) can be interpreted as predators who chase adjacent prey-particles at rate 1. Those vertices which are in state \(\dagger\) are considered to be \emph{dead} and cannot be re-occupied. This model is an extension of the chase-escape model (without death) which has its origins to the Richardson growth model for the spread of a single species. The object of study are the sets \(\mathcal{R}\) and \(\mathbb{B}\), which are the sets of those vertices which get into state \(r\) at some point and \(b\), respectively. If \(v\) is the vertex which was initially in state \(b\), then \(\mathcal{B} \setminus \{v\} \subseteq \mathcal{R}\). More specifically, the events that are studied in the paper are \(A = \{|\mathcal{R}| = \infty \}\) and \(B = \{ |\mathbb{B}| = \infty \}\). The above inclusion implies that \(B \subseteq A\). Depending on the survival of these states, there are three phases: \emph{coexistence} where \(\mathbb{P}(B)>0\); \emph{escape} where \(\mathbb{P}(A)>0\) but \(\mathbb{P}(B) =0\) and \emph{extinction} where \(\mathbb{P}(A) =0\). The paper gives a complete phase diagram in terms of the parameters \(\lambda\) and \(\rho\). It identifies the critical values for these parameters and gives asymptotics for large \(d\). The differentiability of one of the critical values for \(\rho\) is shown (that of the coexistence) and the existence of a (finite) algorithm which decides whether \(\rho\) is above or below this critical value is established (provided that \(\rho\) is not equal to this critical value).