Dynamics of viscous liquids and the random barrier model

This paper combines the particle-swap Monte Carlo algorithm with long GPU molecular dynamics simulations to analyze the dynamics of a ternary Lennard-Jones glass-forming liquid in the extremely viscous regime. The focus is on the inherent dynamics, obtained by quenching configurations along the configuration-space trajectory into their inherent state. We compare how two functional forms, the von Schweidler law and the prediction of the random barrier model (RBM) in the extreme disorder limit, fit data for the inherent mean-square displacement as a function of time. We find that the RBM, which has no dimensionless free parameters, generally fits the data better than the von Schweidler law, despite the latter's one dimensionless free parameter. In particular, this implies that the RBM predicts the value of the diffusion coefficient from short-time simulation data more accurately than does the von Schweidler expression. Furthermore, we find that the RBM also fits very well the inherent mean-square displacement of a polydisperse Lennard-Jones mixture. It remains an open question why the RBM reproduces well the data despite this model's (unrealistic) assumption of identical energy minima.