Open Access

ARTICLE

JavaGenes: Evolving Molecular Force Field Parameters with Genetic Algorithm

Al Globus¹, Madhu Menon², Deepak Srivastava¹

1 NASA Ames Research Center, CSC/NAS, Moffett Field, CA 94035
2 Department of Physics and Center for Computational Sciences, University of Kentucky, Lexington, KY 40506-0045

Computer Modeling in Engineering & Sciences 2002, 3(5), 557-574. https://doi.org/10.3970/cmes.2002.003.557

Abstract

A genetic algorithm procedure has been developed for fitting parameters for many-body interatomic force field functions. Given a physics or chemistry based analytic form for the force field function, parameters are typically chosen to fit a range of structural and physical properties given either by experiments and/or by higher accuracy tight-binding or ab-initio simulations. The method involves using both near equilibrium and far from equilibrium configurations in the fitting procedure, and is unlikely to be trapped in local minima in the complex many-dimensional parameter space. As a proof of concept, we demonstrate the procedure for Stillinger-Weber (S-W) potential by (a) reproducing the published parameters for Si by using S-W energetics in the fitness function, and (b) evolving a "new'' set of parameters, with a fitness function based on a non-orthogonal tight-binding method, which are better suited for Si cluster energetics as compared to the published S-W potential. Evolution is driven by a fitness function based on the energies and forces calculated for Si_n clusters (n < 7), and is able to predict accurate energies for minimum energy and deformed configurations of Si_n (n = 7, 8, 33) clusters, which were not used in the fitness function.

---