Dependent random graphs and spatial epidemics (Q1296730)

The authors study a model of spatial epidemics in \(\mathbb{Z}^d\), where each site is in one of 3 possible states: dead, susceptible or ill. Taking the death rate to be 1, it is characterized by the infection rate \(\alpha\) with \(0<\alpha<\infty\) and the rate \(\beta\) of the birth of new susceptibles with \(0\leq\beta\leq\infty\). \textit{R. Durrett} and \textit{C. Neuhauser} [Ann. Appl. Probab. 1, No. 2, 189-206 (1991; Zbl 0733.92022)] and \textit{E. Andjel} and \textit{R. Schinazi} [J. Appl. Probab. 33, No. 3, 741-748 (1996; Zbl 0859.92020)] showed coexistence for certain regions in the \((\alpha,\beta)\) phase space. In this paper the phase diagram is extended. It is proved that for \(\alpha\) lying between two critical values, which are shown to be distinct, no coexistence is possible for sufficiently small \(\beta>0\). An important part of the proof are exponential decay estimates of subcritical percolations for a class of locally dependent random graphs, introduced by \textit{K. Kuulasmaa} [ibid. 19, 745-758 (1982; Zbl 0509.60094)] for an epidemic model without recovery \((\beta=0)\). These results are of interest in their own.