An improved genetic algorithm for global optimization and its application to sodium chloride clusters

We have developed an improved genetic algorithm with a self-guiding search strategy, using a combination of "traditional" and geometric genetic operators, whose relative weights are determined by the algorithm itself. Explicit mutation becomes unnecessary, because randomness is included in some of the geometric operators. This still quite simple algorithm is applied to the search for global minima on the potential energy hyper-surface of the most widely studied ionic clusters, sodium chloride clusters, with up to 100 ions. We present 32 previously unpublished apparent global minima of the Coulomb + Born-Meyer potential in the Tosi Fumi parametrization. Comparison with already published minima of the shell model due to Ritmer shows no structural differences. We are able to reproduce the experimental sodium chloride bulk energy quite closely. Therefore, the Coulomb + Born-Meyer potential appears to be a good approximation for neutral clusters of all sizes.