Wulff droplets and the metastable relaxation of kinetic Ising models (Q1270185)

This is an important paper devoted to the systematic study of metastable relaxation of the kinetic Ising model in two dimensions. The main result of the paper describes the asymptotics of the relaxation time of the system evolving with a reversible spin-flip dynamics in a neighbourhood of the phase-coexistence line. The authors show that if such a dynamics starts from a distribution that is ``opposite'' to a small positive external magnetic field \(h\), then the relaxation time blows up, as \(h\to 0\), like \(\exp\{\lambda_c(T)/h\}\), where \(\lambda_c(T)\) is a certain constant. This, in particular, fully confirms the conjecture of \textit{M. Aizenman} and \textit{J. L. Lebowitz} [J. Phys. A 21, No. 19, 3801-3813 (1988; Zbl 0656.60106)]. The heuristics behind this phenomenon is that for a spin at the origin to feel itself as being in the \((+)\)-phase, the system waits until the first critical nucleating droplet of the \((+)\)-phase (that has a shape prescribed by the Wulff construction) is created somewhere in a neighbourhood of the origin and then this critical lattice droplet grows to eventually include the origin. It is interesting to note that the critical value \(\lambda_c(T)\) that separates the metastable behaviour of the system from the stable one is determined by purely static considerations (indeed, it is defined in terms of certain characteristics of the Wulff shape); this seems to be the first rigorous example of a relation between the equilibrium Wulff shape and the time evolution of kinetic Ising models. The authors compare the metastable state and the \((-)\)-state by presenting the difference in terms of asymptotic expansion in powers of the external field \(h\); this expansion can be interpreted as \(C^\infty\)-continuations in \(h\) of the families of Gibbs distributions with \(h<0\) to the region of positive fields. The results described here strengthen the known facts obtained earlier by the first author at low temperatures, but in arbitrary dimension \(d\geq 2\) [Commun. Math. Phys. 161, No. 1, 1-49 (1994; Zbl 0796.60103)]; the main improvements are as follows: 1) existence of a single constant separating the metastable behaviour from the stable one is shown; 2) the result is valid for all subcritical temperatures \(T\); 3) arbitrary initial conditions that are stochastically lower than the \((-)\)-phase can be chosen.