Fast polarizable force field computation in biomolecular simulations
Polarizable force fields are considered to be the single most significant development in the next-generation force fields used in biomolecular simulations. The self-consistent computation of induced atomic dipoles in a polarizable force field is expensive due to the cost of solving a large dense linear system at each timestep in molecular dynamics simulations. Methods are developed that reduce the cost of computing the electrostatic energy and force of a polarizable model from about 7.5 times the cost of computing those of a non-polarizable model to less than twice the cost. The reduction is achieved by an efficient implementation of the particle-mesh Ewald method, an accurate and robust predictor based on least squares fitting, and two non-stationary iterative methods whose fast convergence is empowered by a simple preconditioner. Furthermore, with these methods, we show that the self-consistent approach with a larger timestep is faster than the extended Lagrangian approach. The use of dipole moments from previous timesteps to calculate an accurate initial guess for iterative methods leads to an energy drift and compromises the volume-preserving property of the integration. Iterative methods with zero initial guess do not lead to perceptible energy drift if a reasonably strict convergence criterion for the iteration is imposed and the numerical integrator is volume-preserving. The approximate solution computed by an iterative method ruins the symplectic property of the integrator. To address this problem, a non-iterative method has been developed based on an approximation to the electrostatic potential energy and has been efficiently implemented. The method preserves the symplecticness of the integrator and is suitable for long time simulations. The research will help polarizable force fields modeling and computation to become a routine part of molecular dynamics simulations for biomolecular systems.