Forcefield Optimizer

Forcefield based simulations allow you to study systems with thousands of atoms and millions of configurations using molecular dynamics and Monte Carlo methods. For many systems such as organics which are accurately described by PCFF+, carefully validated forcefield parameters are available. When forcefield parameters are lacking for a given system, the Forcefield Optimizer allows you to generate forcefield parameters based on first-principles simulations.

The Forcefield Optimizer takes as input structures, energies, forces, and stresses obtained from first-principles simulations using VASP, and employs least-squares and evolutionary algorithms to optimize the forcefield based on the comparison between forcefield and first-principles derived properties. The Forcefield Optimizer provides a user friendly environment for the selection of refinement variables, setting parameter bounds, and analyzing the results of the refinement calculation.

The handling of first-principles training set is flexible. You can employ molecular dynamics trajectories, individual structures, periodic, and cluster calculations, in any combination, in order to develop a representative training set for a given system. You can also separate first-principles information into training and validation sets, in order to provide cross validation of the parameters obtained by the Forcefield Optimizer.

In addition to the automatic and flexible interface provided by the Forcefield Optimizer a range of analyses are automatically provided including graphical comparisons of forcefield energies, forces, and stresses with first-principles data, estimated parameter errors, and information on the statistics of the correlations obtained during fitting. The key output of the Forcefield Optimizer is a new forcefield .frc file, which can be directly used in subsequent forcefield calculations with MedeA.

Forcefield Optimizer

The Forcefield Optimizer supports EAM, Class II (such as PCFF+), Buckingham, Stillinger-Weber, Tersoff, and COMB3 forcefields. The Forcefield Optimizer allows the user to select from the forcefield terms necessary to describe a given set of first-principles results, select fitting weights, and set parameters to control the fitting algorithms employed in refining parameters. Within the context of supported forcefields, the openness of the MedeA environment allows users to add atom types to support entirely new chemical and physical simulations.

Forcefield Optimizer Features:

  • Determine forcefield parameters based on VASP structures and trajectories: energies, forces, stresses
  • Flexible interactive control of parameters to be optimized, including the use of limits
  • Genetic algorithm and least squares based methods to optimize parameters
  • Freedom in construction of training sets allowing flexible combination of structures and trajectories
  • Extensive reporting on fitting results, including regression statistics, graphing of target and fitted properties, and error estimation
  • Automated construction of an optimized forcefield file for subsequent use in simulation studies
  • Support for EAM, Buckingham, PCFF+, Stillinger-Weber, Tersoff, and COMB3 forcefield forms

    Required MedeA Modules:

  • Core MedeA Environment

  • MedeA JobServer and TaskServer
  • MedeA-VASP
  • MedeA-LAMMPS

The methodologies of the Forcefield Optimizer have been described in the following publications:

  1. Thermal expansion, diffusion and melting of Li₂O using a compact forcefield derived from ab initio molecular dynamics, Ryoji Asahi, Clive M. Freeman, Paul Saxe and Erich Wimmer, Modelling Simul. Mater. Sci. Eng. 22 (2014)

  2. Diffusion of point defects, nucleation of dislocation loops, and effect of hydrogen in hcp-Zr: Ab initio and classical simulations, M. Christensen, W. Wolf, C. Freeman, E. Wimmer, R.B. Adamson, L. Hallstadius, P.E. Cantonwine, E.V. Mader, Journal of Nuclear Materials 460, 82–96 (2015)