Intermolecular Potentials for Water: Molecular Simulation of Bulk Properties, Thermodynamic Properties and Phase Behavior

Water plays a near ubiquitous role in biological, chemical and industrial processes. Historically, predicting the properties of water involved either empirical correlations or equation of state modelling [1], [2], whereas more recently molecular simulation [3] has become he method of choice because of the nexus between underlying intermolecular interactions and observable macroscopic properties. The availability of an intermolecular potential to evaluate inter-particle forces or energies is the key to accurate predictions. There are many alternative intermolecular potentials for water [4], although the basis of many water potentials is at best semi-empirical. The most widely used models are rigid and variants of either the four-site transferable interaction potential (TIP4P) or the three-site simple point charge (SPC, SPC/E) models. The appeal of such potentials is computational expedience and in many cases they have provided worthwhile predictions. However, comparisons with experiment are often focused at relatively low temperatures and pressures, with a temperature of 25 ºC and a pressure of 1 atmosphere being a popular choice.

The accuracy for 20 atomistic models for water is comprehensively evaluated [5] with particular attention to thermodynamic properties such as the thermal pressure coefficient, thermal expansion coefficient, isothermal and adiabatic compressibilities, isobaric and isochoric heat capacities, Joule-Thomson coefficient, speed of sound and vapor-liquid equlibria. The addition of bond flexibility, while improving agreement with experiment for such properties as phase coexistence, dielectric constants, viscosity and diffusion, does not appear to significantly improve the prediction of thermodynamic properties in general. For the supercritical heat capacities and thermal expansion coefficient the flexible TIP4P/2005f model yields less accurate values than its rigid TIP4P/2005 counterpart. In contrast, accounting for polarizability consistently results in improved agreement with experiment. For properties such as the isochoric heat capacity and thermal expansion coefficient, the polarizable MCYna model yields values that are in very close agreement with experimental data in the temperature range of 300 – 600 K.. At or near ambient conditions, the overall ranking of models investigated is (iAMOEBA, MYCYna, BKd3, GCPM) > (BK3, TIP4P/FQ, SPC/Fw) > (NvdE, TIP4P/2005, SPC/E) > (SPC, TIP4P/2005f, TIP5P) > (TIP3P, TIP4P) > (MCY, MCYL).

[1] Y. Wei and R. J. Sadus, AIChE J. 2000, 46, 169-196 ; N. G. Stetenskaja, R. J. Sadus and E. U. Franck, J. Phys. Chem. 1995, 99, 4273 -4277 ; A. E. Mather, R. J. Sadus and E. U. Franck, J. Chem. Thermodyn. 1993, 25, 771-779.

[2] W. Wagner and A. Pruß, J. Phys. Chem. Ref. Data. 2002, 31, 387-535.

[3] R. J. Sadus, Molecular Simulation of Fluids: Theory, Algorithms and Object-Orientation, Elsevier, Amsterdam, 1999.

[4] C. Vega and J. L. F. Abascal, Phys. Chem. Chem. Phys. 2011, 13, 19663.

[5] I. Shvab and R. J. Sadus, Fluid Phase Equilib. 2016, 407, 7.