Acceleration of convergence in Monte Carlo simulations of aqueous solutions using the metropolis algorithm. Hydrophobic hydration of methane

Convergence problems encountered in the computer simulations of aqueous solutions are discussed. Solute–solvent radial distribution functions are shown to converge very poorly when the standard Metropolis Monte Carlo procedure is applied. To overcome this difficulty, several modifications are made in the Metropolis method. Optimum maximum step sizes are determined for simulations of liquid water. A scheme is employed for preferential sampling of both the solvent and the solute molecules. To test these modifications, a simulation is carried out for pure liquid water, treating a single water molecule as a “solute.” The convergence of the radial distribution functions is found to be accelerated significantly. A further test is made by simulating an aqueous solution of methane, consisting of one methane molecule (using the EPEN /2 potential for methane–water interactions) and 124 water molecules (using the MCY potential for water–water interactions). Again, the convergence of solute–solvent radial distribution functions is found to be accelerated. The computation of partial molar thermodynamic quantities, however, still suffers from convergence difficulties. This problem is discussed in detail. The EPEN /2 potential is found to yield structural and thermodynamic features of hydrophobic hydration that are consistent with available experimental and theoretical results for aqueous solutions of methane.