Achieving ergodic sampling using replica-exchange free-energy calculations

Seamless combination of multiple-walker strategies with importance-sampling free-energy methods relying upon perturbation theory and the application of time-dependent biases on collective variables is explored as an alternate avenue to achieve ergodic sampling in molecular simulations. Free-energy perturbation is associated with a parallel-tempering scheme, wherein both temperatures and the Hamiltonians representative of the various strata of a stratification scheme are swapped sequentially. In addition, adaptive biasing force simulations are performed at different temperatures in a replica-exchange context, updating continuously across the different replicas the gradient of the free energy along the relevant collective variable. Performance of the proposed methodology is evaluated in two distinct applications, namely the computation of the entropy of hydration of ethanol, following either an alchemical or a geometrical route, and the entropy that underlies the reversible folding of a short peptide.