Fast hybrid methods for the simulation of dielectric constants of liquids

The distance-dependent Kirkwood factors and dielectric constants of the SPC/E and TIP3P models of water, the H1 model of methanol, and the OPLS model of acetonitrile are investigated with the help of long molecular dynamics (MD) simulation runs and the hypernetted chain (HNC) approximation of the molecular Omstein-Zernike (MOZ) theory. The MOZ theory predicts dielectric constants which show the same dependence on the liquid model as their simulated counterparts, are very accurate in the case of acetonitrile, but are too low by about 20% for the H-bonding liquids, because the HNC closure moderately accounts for the short-range orientational order of their molecules. In the case of too short simulation runs, some pitfalls concerning the accuracy of the dielectric properties and their errors are emphasized. For a given liquid model, if the distance-dependent Kirkwood factor of the MOZ theory has a long-range profile similar to the converged one of an accurate simulation, rapid and reliable hybrid methods based on the combination of the MOZ and simulation results can be proposed to estimate the dielectric constant. Test calculations on the above-mentioned models are carried out to assess the validity of the approach. Hybrid methods are shown to give dielectric constants of similar accuracy as those derived from long simulation runs of one million of steps or more, but with a CPU time demand at least 10 times shorter. (C) 2004 Elsevier B.V. All rights reserved.