Firefly Algorithm Applied to Non-collinear Magnetic Materials Prediction

Computational materials prediction has become an important technique for the experimental realization of material systems. These predictive methods are based on searching over possible configurations in parameter space. While the optimization of structural parameters has been successful [1], optimization of magnetic parameters has proven to be more challenging, as magnetic systems support a large number of metastable states [2], so calculations are prone to falling into one of these states. In this talk the generalization of the population-based metaheuristic firefly algorithm (FA) [1] to the problem of magnetic ground state prediction in non-collinear magnets will be presented. We extend the different steps of FA to this problem by using polarized density functional theory calculations with the use of Lagrange multipliers to fix the directions of the atomic magnetic moments. This allows for the exploration of the entire Born-Oppenheimer energy surface. Through applications to molecular magnets and magnetic crystals, we demonstrate that the number of minima can be large, which restrains the use of exhaustive searches.

[1] G. Avadaño-Franco and A. Romero, J. Chem. Theory Comput., 2016, 12, 3416–3428
[2] J. Allen and G. Watson, Phys. Chem. Chem. Phys., 2014, 16, 21016-21031