Parameter optimization in molecular dynamics simulations using a genetic algorithm

In this work, we introduce a genetic algorithm for the parameterization of the reactive force field developed by Kieffer . This potential includes directional covalent bonds and dispersion terms. Important features of this force field for simulating systems that undergo significant structural reorganization are (i) the ability to account for the redistribution of electron density upon ionization, formation, or breaking of bonds, through a charge transfer term, and (ii) the fact that the angular constraints dynamically adjust when a change in the coordination number of an atom occurs. In this paper, we present the implementation of the genetic algorithm into the existing code as well as the algorithm efficiency and preliminary results on Si-Si force field optimization. The parameters obtained by this method will be compared to existing parameter sets obtained by a trial-and-error process.