A phase transition for measure-valued SIR epidemic processes (Q2438750): Difference between revisions

In epidemic models, the acronym SIR stands for suspectible-infected-removed or suspectible-infected-recovered. In these models, an infected individual remains infected for a certain period of time, during which it can transmit the disease to suspectible individuals, and then recovers, after which it is immune to further infection. In the present paper, the authors investigate measure-valued processes \(X_t\) that solve the martingale problem \[ X_t(\varphi)=X_0(\varphi)+\frac12\int_0^t X_s(\Delta\varphi)ds+\theta\int_0^t X_s(\varphi)ds - \int_0^t X_s(L_s\varphi)ds+M_t(\varphi), \] for a given initial measure \(X_0\) and for all compactly supported smooth test functions \(\varphi\). Here \(L_s(x)\) is the Sugitani local time density process associated with \(X\), and \(M_t(\varphi)\) is a martingale with quadratic variation \([M(\varphi)]_t=\int_0^t X_s(\varphi^2)\,ds\). In SIR models, \(X_t\) represents the distribution of the infected set in space, \(R_t=\int_0^t X_s\,ds\) the recovered set, \(\int_0^t X_s(L_s\varphi)\,ds\) is the resource depletion term and \(\theta\) the transmission rate. The main result of the paper is the following: In dimensions \(d=2,3\), there exist critical values \(\theta_c(d)\in(0,\infty)\), such that the solution survives forever with positive probability if \(\theta>\theta_c(d)\), and the solution dies out in finite time almost surely if \(\theta<\theta_c(d)\). For \(d=1\), the solution dies out almost surely for all values of \(\theta\). The authors also prove that, for \(d=2,3\), the process dies out locally almost surely for all values of \(\theta\).