Predicting low-temperature free energy landscapes with flat-histogram Monte Carlo methods
Nathan Mahynski, Marco A. Blanco Medina, Jeffrey R. Errington, Vincent K. Shen
We present a method of predicting the free energy landscape of fluids at low temperatures from flat-histogram grand canonical Monte Carlo simulations performed at higher ones. We illustrate our approach for both pure and multicomponent systems using two different sampling methods as a demonstration. This allows us to predict the thermodynamic behavior of systems which undergo both first order and continuous phase transitions upon cooling using simulations performed only at higher temperatures. After surveying a variety of different systems, we identify a range of temperature differences over which the extrapolation of high temperature simulations tends to quantitatively predict the thermodynamic properties of fluids at lower ones. Beyond this range, extrapolation still provides a reasonably well-informed estimate of the free energy landscape; this prediction then requires less computational effort to refine with an additional simulation at the desired temperature than reconstruction of the surface without any initial estimate. In either case, this method significantly increases the computational efficiency of these flat-histogram methods when investigating thermodynamic properties of fluids over a wide range of temperatures. For example, we demonstrate how a binary fluid phase diagram may be quantitatively predicted for many temperatures using only information obtained from a single supercritical state.
, Blanco, M.
, Errington, J.
and Shen, V.
Predicting low-temperature free energy landscapes with flat-histogram Monte Carlo methods, The Journal of Chemical Physics, [online], https://doi.org/10.1063/1.4975331
(Accessed December 3, 2021)