Improving predictions of solvation free energies from non-polarisable models by applying polarisation corrections

Rights statement
Awarding institution
  • University of Strathclyde.
Date of award
  • 2021
Thesis identifier
  • T15988
Person Identifier (Local)
  • 201779449
Qualification Level
Qualification Name
Department, School or Faculty
  • Classical non-polarisable models, normally based on a combination of Lennard-Jones (LJ) sites and point charges, are extensively used to model thermodynamic properties of fluids. An important shortcoming of this class of models is that they do not explicitly account for polarisation effects - i.e. a description of how the electron density responds to changes in the molecular environment. Instead, polarisation is implicitly included into the parameters of the model, usually by fitting to pure liquid properties (e.g. density). A problem arises when trying to describe thermodynamic properties that involve a change of phase (e.g. enthalpy of vaporisation), solutions/mixtures (e.g. solvation free energies), or properties that directly depend on the electronic response of the medium (e.g. dielectric constant). Fully polarisable models present a natural route for addressing these limitations, but at the price of a much higher computational cost. The main goal of this thesis is to obtain a non-polarisable force field for alcohols, amines and ketones able to predict both pure liquid properties and solvation free energies with a high degree of accuracy through the use of post-facto polarisation corrections. These corrections are applied to the properties computed using the non-polarisable force field in order to account for the effects of polarisation, and thus, increase the model's accuracy while maintaining its computational effciency. This work is part of a larger project which end goal is to predict the solubility of drug molecules (e.g. paracetamol). These molecules usually contain hydroxyl, amino and carbonyl groups, and thus, this thesis focuses on molecules with these functional groups. Aromatic rings are another functional group present in most drug molecules, however, they are not studied here due to time limitations. Alcohols and amines are interesting from a fundamental point of view as they are the simplest molecules that combine a hydrophobic moiety with a hydrogen-bonding functional group. Also, alcohols and ketones are widely used as solvents and amines are used in CO2 adsorption/desorption processes designed to decrease CO2 emissions. The model developed in this thesis is called PolCA, standing for 'Polarisation-Consistent Approach', and it is an extension of the modified TraPPE force field for hydrocarbons proposed by Jorge [1] that eliminates systematic deviations from experimental solvation free energies. The new amino, hydroxyl and carbonyl parameters were fitted to several pure-component experimental properties including the density and enthalpy of vaporisation, and in some cases also self-solvation free energies. The optimization was carried out using meta-models that predict how the simulated properties change with the input parameters, allowing for a better exploration of the force field parameters' space. The PolCA force field for alcohols can accurately predict methanol to decanol's densities, diffusion constants (except for methanol), enthalpies of vaporisation, free energies of self-solvation, dielectric constants and solvation free energies in hexadecane. PolCA also does a very good job at predicting the densities, enthalpies of vaporisation and free energies of self-solvation of linear and branched primary amines, and its predicted solvation free energies of linear primary amines in hexadecane are in very good agreement with experimental data. However, it greatly overpredicts the dielectric constant of methylamine and significantly overpredicts other linear and branched amines' dielectric constants. Furthermore, PolCA can accurately predict the densities, enthalpies of vaporisation, diffusion constants and self-solvation free energies of propanone to 2-decanone (except for butanone and 2-pentanone's densities which are underpredicted), however, it greatly overpredicts ketones' dielectric constants and sovation free energies in hexadecane (more negative values). Lastly, PolCA was used to calculate solvation free energies of amines and ketones in octanol and multifunctional compounds' densities and dielectric constants to test its transferability. Our force field was not able to simultaneously predict solvation free energies in hexadecane and octanol, and thus a re-parameterisation will be carried out in future work using polarisation corrections obtained with more accurate methods. Nonetheless, the results obtained in this work show that the approach proposed here is very promising since they significantly improve agreement with experiment for the dielectric constant and solvation free energies of alcohols and amines in hexadecane.
Advisor / supervisor
  • Jorge, Miguel
Resource Type