London dispersion interactions play a critical role in condensed matter properties, protein folding, and lipid structure. Force fields for molecular simulations of these systems typically determine the parameters for these interactions empirically. These parameters are generally underdetermined and currently, there are no general safeguards to ensure they are physically realistic. The exchange-dipole model (XDM) from density functional theory can predict atomic London dispersion coefficients from first principles. This model predicts molecular dispersion coefficients in close agreement with experiment and has been effective for predicting molecular lattice energies and the stability of dispersion-complexes. We have used the XDM model to validate and parameterize the dispersion parameters for generalized molecular mechanical force fields. XDM was used to obtain the London dispersion coefficients for the full Virtual Chemistry test set, which contains over 300 organic molecules relevant to biochemistry and pharmaceutical chemistry. On average, the non-polarizable force field dispersion coefficients agree reasonably well with the XDM values, although some atom types underestimate or overestimate the strength of dispersion by as much as a factor of two. Simple water models, like TIP3P, have dispersion strengths that are similar to the experimental and XDM values, but others, like TIP4P-D, overestimate the strength of dispersion by a large margin. The Drude force field has many atomic dispersion coefficients that are inconsistent with XDM. We propose and demonstrate a method to define force field parameters that are consistent with the XDM values, which will enable the development of a dispersion-balanced force field.

Determining dispersion coefficients for polarizable force fields using density functional theory

254th National Meeting (2017)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.