Inhomogeneous phase separation kinetics in liquid binary mixtures: Sensitivity to initial local composition †

Abstract

The kinetics of phase separation subsequent to a finite temperature quench is assumed to be driven by diffusion on the altered free energy surface and is generally assumed to be slow. The situation can be different in phase separating liquid binary mixtures, especially for systems characterized by the large difference in mutual interactions between solute and solvent molecules. In such cases, the phase separation kinetics could be fast and may get completed within a short time (ns) scale. As a result, in these systems, one may observe diverse dynamical features arising out of local heterogeneity leading to the onset of phase separation through pattern formation, spinodal decomposition, nucleation, and growth. By using a coarse-grained analysis, we examine phase separation kinetics in each spatial grid and indeed observe important effects of initial heterogeneity on the subsequent evolution. Interestingly, we observe slower separation kinetics for those regions that correspond to the composition at the minimum of the high-temperature surface. The heterogeneous dynamics has been captured here through the non-linear susceptibility function, which shows a pattern similar to what is observed in the supercooled liquid. Each grid shows somewhat different dynamics in the three-stage (exponential, power-law, and logarithmic regime) phase separation dynamics. The late stage of phase separation kinetics is usually attributed to the coarsening of the phase-separated domains. However, in a liquid binary mixture, the late-stage power-law decay undergoes a further change. A new dynamical regime arises characterized by a logarithmic time dependence, which is due to the “smoothening” of the rough interface of already well-separated phases. This can also be described as opposite to the roughening transition described by Chui and Weeks [Phys. Rev. Lett. 40, 733 (1978)]. This reverse roughening transition can explain the logarithmic time dependence observed in the simulation.