Prediction of the lowest energy structure of clusters using a genetic algorithm

An optimization approach using a genetic algorithm in the search for a global minimum is described. The method is based on the use of control variables to form the genotypes in each generation. This procedure allows an accurate representation of the control variables, and consequently, a high resolution determination of the optimum solution. A set of genetic operators, appropriate for the operation on genes represented by real numbers, is introduced. The method is used to predict the lowest energy structures of Arₙ{H}₂$ microclusters with n=4, 5, 6, 7, and 12. Comparison between the performance of this optimization approach and the well established simulated annealing method clearly demonstrates the superiority of the genetic algorithm based search.