Limitations of a Fokker-Planck description of nucleation

In most nucleation theories, the dynamics of nucleation is characterized by the evolution in time of the mass of droplets, and this time evolution is described as a combination of drift and diffusion. This assumes that the mass fluctations are described by a Markovian, i.e., memoryless, stochastic process. This paper presents a method to assess in how far this assumption of Markovianity is valid. The method is employed in nucleation studies in a two-dimensional Ising model at temperature T=0.88T(c), both with spin flip dynamics and with local spin exchange dynamics. In the first case, it shows that the evolution of droplet masses might be effectively described by a Markov process on large time scales. In the latter case, however, the dynamics are far from Markovian. We argue that this is due to the presence of a locally conserved quantity.