(376bt) Contacts between Molecular Surfaces Allow Quick First-Principle Prediction of Dielectric Constant in Pure Liquids and Mixtures

The relative permittivity ɛ_r (or dielectric constant) represents the decrease of an electric force experienced by a charge due to its surrounding medium. This property is intrinsic to the medium and thus, in principle, is completely determined by its microscopic structure and experimental conditions (temperature and pressure). ɛ_r is challenging to predict because small changes in the molecular structure of organic compounds can be related to large differences of ɛ_r between organic compounds. Since this property is related to the ability of molecules to have a polarizing effect on their surroundings, ɛ_r is particularly sensitive to hydrogen-bonding and 3D configuration of the molecules.

In the case of liquid mixtures, knowledge of ɛ_r has numerous applications. Because ɛ_r directly describes the ability of a medium to screen a charge from an electric field, many of its uses are related to ion solvation [1]. For example, rationalizing chemical reactions, which often involve ions dissolved in solvents, and their yields [2], is a first large field of applications of ion solvation in both research and industry. In analytical chemistry, the proper choice of solvents to analyze ions in High Performance Liquid Chromatography [3] or electroanalytical measurements [2] (like potentiometry or voltammetry) can be guided by ion solvation energy prediction. Knowledge of ɛ_r can also directly help in exclusively industrial applications, such as optimization of battery electrolytes [4], hydrometallurgy or nuclear waste disposal [2].

Quantitative Structure-Property Relationship (QSPR) [5] models and molecular dynamics [6] can be used to predict this property. However, as purely empirical methods, QSPR models are limited by the applicability domain spanned by the training set, and molecular dynamics is generally costly in terms of computational resources and requires a careful choice of force fields. In this contribution, an alternative approach that, in principle, does not suffer from both of these inherent limits is presented. The approach is based on the use of molecular contact probabilities from COnductor-like Screening MOdel for Realistic Solvation (COSMO-RS) [7] as hints of the mutual polarization of molecules in the liquid phase. This allowed to derive a model suitable for the prediction of the static dielectric constant from first principles. The model was tested on more than 400 pure liquids and 4000 binary mixtures, and allowed to reach a relative prediction error of about 20%, with calculation time ranging from a few seconds to ten minutes per studied system, provided a COSMO file, obtained from a standard quantum chemical calculation, is available for each distinct structure involved in the simulation. This success opens perspectives in terms of computational screening of appropriate liquids in all applications in which the dielectric constant is important to consider. Moreover, it encourages to explore the use of molecular surfaces and their contacts to predict other challenging properties, such as diffusion constant or viscosity.

References

[1] J. Chem. Eng. Data 2010 55(9): 2951
[2] Ions in Solution and Their Solvation. 2015, John Wiley & Sons
[3] Essentials in modern HPLC separations. 2013, Elsevier
[4] Environ. Sci., 2012 5(9): 8572
[5] J. Chem. Inf. Comput. Sci. 2002 42: 360
[6] J. Chem. Theory. Comput. 2012 8: 61
[7] COSMO-RS: From Quantum Chemistry to Fluid Phase Thermodynamics and Drug Design. 2005, Elsevier