A simple nonparametric procedure is devised for constructing Boltzmann entropy functions from statistically weighted entropy differences calculated from overlapping histograms. The method is noniterative, avoids numerical problems associated with large state densities, and accommodates variable bin widths for reducing systematic and statistical errors inherent to histogram techniques. Results show that the procedure can yield thermodynamic functions for an Ising spin lattice model that have average errors comparable to ones obtained from a conventional approach. Analysis of thermofunctions computed for a polyalanine peptide simulated by hybrid Monte Carlo replica exchange indicates that method performance can be enhanced through the use of nonuniform state space discretization schemes. An extension of the reweighting procedure for multidimensional applications is presented through calculations of vapor-liquid equilibrium densities of a model fluid simulated by grand canonical replica exchange.

You do not currently have access to this content.