Abstract
Classical simulations of real molecules require realistic nonbonded interactions between constituent atoms, which have traditionally been adjusted for good agreement with liquid properties and copied extensively among similar atom types. In this work, we propose ab initio methods to compute both the C12short-range repulsion and the C6dispersive attraction between atoms. We relate the repulsion to the distance at which the electron density near an atom falls below a certain threshold, chosen to match radii for atoms in the OPLS force field. We compute the dispersive attraction by applying the McLachlan integral formula to the polarizability contributions of each atom in a molecule as a function of imaginary frequency. These polarizability contributions can be computed by time-dependent Hartree–Fock methods in GAMESS, which conveniently partitions the total polarizability among bonds and lone pairs. Our method produces values for both C12and C6parameters in good agreement with existing OPLS values when applied to nearly 200 atom types in over 100 organic molecules from the virtualchemistry.org archive. We verify that different instances of OPLS atom types have nearly identical polarizabilities, lending credence both to our method and to atom types based on local chemical environments. We extend our frequency-integral method for computing dispersive interactions to atoms in molecules near metal surfaces, which screen nearby fluctuating fields, with a frequency response limited by the plasma frequency. The screening is equivalent to a fluctuating image dipole with which the atom interacts, giving rise to a 1/z3interaction with a metal half-space. This interaction can be conveniently represented as a conventional 1/r6interaction with each metal atom, summed over the half-space.
| Original language | English (US) |
|---|---|
| Pages (from-to) | 10413-10423 |
| Number of pages | 11 |
| Journal | Journal of Chemical Theory and Computation |
| Volume | 21 |
| Issue number | 20 |
| DOIs | |
| State | Published - Oct 28 2025 |
All Science Journal Classification (ASJC) codes
- Computer Science Applications
- Physical and Theoretical Chemistry