Using limiting activity coefficients to efficiently evaluate the ability of fixed-charge force fields to model miscible water plus cosolvent mixtures

ABSTRACT

Fixed-charge force fields are a common choice for modelling solvents, many of which have been optimised around pure-component property predictions. Their use for modelling solvent mixtures, however, is questionable as the charge distribution of a molecule in the pure component and infinite dilution limits are expected to be different. Due to the small system sizes and simulation times typically used, it is not possible to observe phase-splitting in a molecular simulation. Here, we propose and demonstrate the efficient use of limiting activity coefficients computed using the conventional solvation free energy calculations in the infinite dilution and pure-component limits to evaluate whether or not the resulting solution is miscible. We demonstrate the application for binary aqueous systems with the following cosolvents: methanol, ethanol, 2-propanol, acetone and 1,4-dioxane. We find that the binary mixture of TIP4P water and TraPPE-UA 2-propanol is only partially miscible, as confirmed from detailed free energy analysis. One needs to take care when taking force fields off the shelf and applying them to model mixtures. The proposed strategy is general and not limited to molecular simulations with fixed-charge force fields, and we demonstrate its use with electronic structure calculations with the solvation model based on density continuum solvation model.

KEYWORDS:

• Free energy calculations
• activity coefficients
• miscibility
• water

Acknowledgments

ASP would like to thank Carl Lira (Michigan State University) for fruitful discussion at the 2014 Midwest Thermodynamics and Statistical Mechanics Conference.

Disclosure statement

No potential conflict of interest was reported by the authors.

ORCID

Andrew S. Paluch http://orcid.org/0000-0002-2748-0783

Additional information

Funding

The authors acknowledge the donors of the American Chemical Society Petroleum Research Fund (56896-UNI6) for support of this research. GEL and BNR are additionally thankful for financial support from the Miami University College of Engineering and Computing. Computing support was provided by the Ohio Supercomputer Center.