In view of the dearth of experimental data on all but the most common polymer-polymer and polymer-small molecule mixtures, there is strong interest in determining the enthalpy of mixing using simulations. The degree of sophistication of simulation methods applied to the prediction of mixing enthalpies varies considerably - ranging from application of Hildebrand solubility parameter theory, through measurement of pairwise interactions between fragments, to full atomistic modeling of bulk mixtures using molecular dynamics simulation, with the latter being most appealing in view of the fact that the method can in principle be used to predict effects of concentration and temperature, and is not in any way restricted to mixtures for which the sign of the enthalpy of mixing is positive (as is the case with solubility parameter theory).
Assuming an accurate force field is available, the main limitation of direct calculations of mixing energies arises from the fact that the total potential energy of pure components and mixtures is sizable when compared with the mixing enthalpy itself, and accordingly the precision of the simulation measurements is of paramount importance, together with the possibility of errors associated with other simulation characteristics, although these factors are not often considered when using modeling to predict compatibility in polymer mixtures. Accordingly, we have performed a series of systematic investigations of the effect of various factors including the size of the model system, use of nonbond cutoffs when generating a series of structures to be analyzed, and the total number of independent mixture configurations generated, using polystyrene-polybutadiene systems for which considerable experimental phase diagram data are available. In addition, we report on two other factors which often affect computed mixing energies, namely the frequent need to determine component densities using constant pressure simulations, which itself leads to small errors in density and associated errors in system potential energy, and the effect of the limited ability to sample polymer conformations using molecular dynamics simulations. These results are then combined to define the nature and scale of simulations required to yield predictions of mixing energies with an acceptable precision.