Negative diffusion coefficient in a two-dimensional lattice-gas system with attractive nearest-neighbor interactions

Collective diffusion in a two-dimensional lattice-gas system undergoing first-order phase transition is studied both theoretically and by means of Monte Carlo (MC) simulations. The nearest-neighbor attractive interactions result in the formation of a two-phase mixture in which the characteristic size of the dense phase grows with time as t(1/3). It is shown analytically that the evolution of large-scale coverage inhomogeneities is governed by the diffusion equation with a negative diffusion coefficient. Similar to the phenomenon of Ostwald ripening, the Gibbs-Thompson effect is responsible for this abnormal diffusion. MC simulations of random jumps of individual particles also show the presence of negative diffusion caused by the macroscopically inhomogeneous distribution of particle density. The collective diffusion coefficients obtained both theoretically and by means of MC simulations are in satisfactory agreement.