Developing optimal Wertheim-like models of water for use in Statistical Associating Fluid Theory (SAFT) and related approaches

Abstract The statistical associating fluid theory (SAFT) is now well established as an approach for the description of the thermodynamics and phase equilibria of a wide variety of fluid systems. Numerous SAFT studies of the fluid phase equilibria of pure water and aqueous mixtures have been made with the various incarnations of the theory, yet there is no consensus on what the ‘optimal’ values of the intermolecular parameters are for water, or what the association scheme should be (two-, three-, or four-site models). We show that the conventional use of vapour-pressure and saturated-liquid-density data on their own leads to a degeneracy in the values of the parameters, particularly in the relative values of the dispersion and hydrogen-bonding energies. A discretized energy-parameter space is examined and a long valley of ‘optimal’ parameter sets for the vapour–liquid equilibria is found for the various models, ranging from low dispersion (high hydrogen-bonding) to high dispersion (low hydrogen-bonding) energies. Other thermodynamic information such as the heat of vaporization does not allow one to discriminate between the values of the parameters or the association scheme. An examination of the degree of association (hydrogen bonding) allows such a discrimination to be made: the four-site model is found to provide the best overall description of the thermodynamics, fluid phase equilibria, and degree of association. The set of parameters obtained in this way also provides the best description of the fluid-phase equilibria for a mixture of water and methanol. This degeneracy of parameters could also be important when models of water are refined to vapour–liquid equilibria in simulation studies of aqueous systems, and for other systems of associating molecules. Keywords: Statistical mechanicsPerturbation theoryMolecular simulationsParameter estimationFluid phase equilibriaHydrogen bondingAqueous mixtures Acknowledgements We are very grateful to Steve Rogers and Andrew Burgess of the Strategic Technology Group of ICI plc for very useful discussions. G.N.I.C. would like to thank ICI plc and the Engineering and Physical Sciences Research Council (EPSRC) of the UK for funding an Industrial Case studentship. A.J.H. thanks the Department of Trade and Industry (DTI) of the UK for a research fellowship (TP/2/5C/6/1/10244). We acknowledge further support from the EPSRC (GR/N03358, GR/N35991, and GR/R09497), the Joint Research Equipment Initiative (JREI) for computer hardware (GR/M94427), and the Royal Society-Wolfson Foundation for the award of a refurbishment grant.