Research work · 2022
CASTEP solvation forces: eCSE01-9
Durham, Ben, Jacek Dziedzic, Philip J. Hasnip, Matt I. J. Probert
Zenodo (CERN European Organization for Nuclear Research)
Abstract
The aim of this eCSE was to add new functionality to the Density Functional Theory (DFT) program CASTEP to enable it to perform structural relaxation, transition state and/or molecular dynamics simulations in Open Boundary Condtions (OBCs) and in the presence of a solvent. Most CASTEP calculations use periodic boundary conditions and so are best suited to studying extended systems (i.e. solids), but sometimes it is helpful to have open boundary conditions instead, such as when studying a molecular system, or a system in the presence of a solvent. Whilst the solvent could be represented explicitly, this can be very computationally expensive, and so in a previous eCSE project (eCSE07-6), the “minimal-parameter solvent model” was added to CASTEP. In this model the solvent is treated implicitly via solving the generalised Poisson equation, by treating the dielectric permittivity of the system as a continuum and a function of the electronic density. This original implementation was limited to calculating the free energy of solvation only, and could not relax the atomic positions for system under study in OBCs or in the presence of a solvent. Furthermore, the dielectric permittivity, and therefore the implicit model of the solvent, was only computed once at the beginning of each CASTEP execution, and was held fixed for the duration of the calculation. This meant that the model of the solvent was not updated to respond to changes in the electronic density; thus, even if the force terms had been implemented, once the atoms had moved significantly over the course of a Geometry Optimisation, or Molecular Dynamics simulation, then the implicit solvent model would have become obsolete and the results non-physical. An additional limitation was the most computationally intensive step in this model, the solving of the generalised Poisson equation, was not optimally parallelised. Hence, in this project, we added the calculation of implicit solvation forces, as well as a model for implicit solvent that allows the dielectric permittivity to be updated in a self-consistent manner with the electronic density and improved the parallelisation of these calculations. This significantly increases the usefulness of the OBCs/implicit solvation functionality and enables qualitatively new calculations.