A Pinch of salt in ONETEP's solvent model
Lucian Anton, Scientific Computing Department, STFC, Daresbury
Jacek Dziedzic, Chris-Kriton Skylaris, School of Chemistry, University of Southampton
The inclusion of electrolytes (salt) in solvents such as water is crucial for biomolecular simulations, as most processes (e.g. protein-protein or protein-drug interactions or DNA mutations) happen in saline solutions. Density Functional Theory (DFT) calculations for biomolecules need to include the contributions from solvent and the salt ions to the electrostatic potential generated by the electronic density and nuclei. This can be done by implicit solvation models, the most accurate of which solve directly the Poisson-Boltzmann Equation (PBE) in 3D space, in a self-consistent fashion within the DFT algorithm and need to account for the polarisation of the electronic density by the solvent. To make such calculations practical a PBE solver based on a scalable and efficient multigrid algorithm is needed.
In the first part of this presentation we will describe the DL_MG package, which is a parallel multigrid PBE solver. DL_MG was originally developed with support form HECToR dCSE as a solver for the non-homogeneous Poisson equation and with the current ARCHER eCSE project it was extended to be able to solve the non-homogeneous PBE. We will touch on the selected numerical algorithms (including full exponential and linearized versions of the PBE), MPI-OpenMP hybrid parallelism, user interface and documentation.
In the second part of the presentation we will outline the developments to ONETEP linear-scaling DFT code that were necessary to include the effects of non-zero salt concentration in its solvation model, both in the implementation and in underlying theory (additional energy terms). We present results of validation tests, where the effect of the steric potential on free energies of solvation is compared between ONETEP and a classical force field approach (APBS with the PMG solver).