Application of a genetic algorithm for selecting the energies of paired interatomic interactions that provide the minimum energy of a given binary ordered alloy of the AnBm composition

Abstract

In this work, we present a methodology for selecting pairwise interatomic potentials for binary metallic alloys AₙBₘ with an fcc lattice that accounts for interactions up to the fifth coordination shell. Extending the interaction range is essential for accurately describing Friedel oscillations, which influence point-defect energies, the stability of ordered phases, the kinetics of plastic deformation, and dislocation mobility. A direct parameter sweep leads to a geometric growth in the number of independent variables and an exponential rise in computational cost (on the order of 1045 variants at four-digit precision), rendering such an approach impractical. To achieve efficient optimization, we employed a genetic algorithm, which performs a global search in multidimensional space and reduces optimization time to seconds while preserving accuracy and reproducibility. The resulting potentials were validated by simulating vacancy diffusion via the Monte Carlo Metropolis algorithm on a two-dimensional 15×40 lattice at 100 °C, yielding eutectic, checkerboard, and striped ordering patterns. The combination of a physically grounded multichannel model with an evolutionary optimizer opens new opportunities for detailed large-scale studies of ordering mechanisms and phase dynamics in multicomponent systems. The findings demonstrate the high accuracy and adaptability of the developed approach.