Weak extinction versus global exponential growth of total mass for superdiffusions

Abstract

Consider a superdiffusion $X$ on $\mathbb R^d$ corresponding to the semilinear operator $\mathcal{A}(u)=Lu+\beta u-ku^2,$ where $L$ is a second order elliptic operator, $\beta(\cdot)$ is in the Kato class and bounded from above, and $k(\cdot)\ge 0$ is bounded on compact subsets of $\R^d$ and is positive on a set of positive Lebesgue measure. The main purpose of this paper is to complement the results obtained in \cite{Englander:2004}, in the following sense. Let $\lambda_\infty $ be the $L^\infty$-growth bound of the semigroup corresponding to the Schr\"odinger operator $L+\beta $. If $\lambda_\infty \neq0$, then we prove that, in some sense, the exponential growth/decay rate of $\|X_t\|$, the total mass of $X_t$, is $\lambda_\infty $. We also describe the limiting behavior of $\exp(-\lambda_\infty t)\|X_t\|$ in these cases. This should be compared to the result in \cite{Englander:2004}, which says that the generalized principal eigenvalue $\lambda_2$ of the operator gives the rate of {\it local} growth when it is positive, and implies local extinction otherwise. It is easy to show that $\lambda_{\infty}\ge \lambda_2$, and we discuss cases when $\lambda_{\infty}> \lambda_2$ and when $\lambda_{\infty}= \lambda_2$. When $\lambda_\infty =0$, and under some conditions on $\beta$, we give a sufficient and necessary condition for the superdiffusion $X$ to exhibit weak extinction. We show that the branching intensity $k$ affects weak extinction; this should be compared to the known result that $k$ does not affect weak {\it local} extinction (which only depends on the sign of $\lambda_2$, and which turns out to be equivalent to local extinction) of $X$.