Large deviation probabilities for some rescaled superprocesses (Q1341863)

This paper deals with super-\(\alpha\)-stable processes in \(\mathbb{R}^d\) with Lebesgue initial measure. In supercritical dimensions these superprocesses, under a critical space-time-mass rescaling, satisfy a ``law of large numbers'' in which the limit is described by an \(\alpha\)- stable flow. The results of this paper are concerned with large deviations away from this limit. The main result is to establish that the system satisfies the large deviation principle with a ``good'' convex rate functional which is determined by a variational formula. In the special case of a vanishing spatial diffusion the variational problem is explicitly solved and demonstrates that as a rule the rate functional is not strongly convex and not continuous. The proofs employ the general methodology for large deviations together with a careful analysis of the log-Laplace functionals of the superprocess. In order to carry this out a number of supplementary results are obtained which are of independent interest. In particular, there is a very detailed analysis of the cumulant reaction diffusion equation which arises in this setting as the cumulant equation.