(148e) Force-Field Based Quasi-Chemical Method for Rapid Evaluation of Binary Phase Diagrams

For many years the cornerstone of chemical engineering thermodynamics has been the evaluation of segment based activity coefficients through some surface interaction site model; of which we address two well known variants:  UNIFAC (group interactions) and COSMO-RS or COSMO-SAC (charge panel interactions). In the case of UNIFAC one relies heavily on (a) group definitions, that more than often are based on history or chemical intuition than a sound principle, and (b) parameterization through fitting on experimental phase diagrams. In COSMO variants, one does not use groups per se, but instead surface charge panels; with a parameter space that is only a fraction of that in  UNIFAC. In quite a different corner one has a collection of force fields,  that have been and will remain popular in  molecular dynamics; in this case with full atomistic detail, and potentially a smaller parameter space (more ‘ab initio’) than either UNIFAC or COSMO, but unfortunately too slow and unreliable to be of practical use to the engineer. What now, if we could combine the power of the quasi-chemical derived activity coefficients directly with some force field, without the need for a condensed simulation? We developed such method: fast (no condensed simulation needed), also no ad hoc groups, and no surface charges needed, and in principle a relatively small parameter set. As we will demonstrate, by showing results against experimental data from the NIST databank, sometimes the predictions are of almost unbelievable accuracy. Not always though, sometimes the method fails badly, that we can attribute to the absence of correlations in molecular packing. We also compare the method with other popular approaches, for example Blanco’s method for prediction of Flory-Huggins parameters, where we do much better. In the new method, from a practical point of view, we have removed the burden of group assignment, and introduced the difficulty of atom typing. Provided such atom types can be and will be defined by the biomolecular simulation community, this does open the door to application in a wide range of problems in the engineering of specialty chemicals, polymers, and pharmaceuticals.