Surface reconstruction of Si (001) by genetic algorithm and simulated annealing method

The Genetic Algorithm (GA) is one of the most recently developed techniques to find the "Global" minimum of an energy functional. This technique combined with conjugated gradient or molecular dynamics has been demonstrated to be efficient for the ground-state configuration search in materials research, e.g. fullerene formation. In this paper, based on the generalized tight-binding molecular dynamics, we apply the GA to study the surface reconstruction of Silicon (001) for the first time. Up to 65 generations, the "Global" minimum or the ground-state configuration for the surface reconstruction of Si (001) was detected efficiently in our GA simulation. In our tight-binding model, a perfect symmetry-dimer structure was found to be the most energetic while some defect asymmetry-dimer structure could coexist in the lists of candidate structures due to the thermal defect or charge transfer which was described with the smearing parameter empirically. We also perform the more traditional Simulated Annealing (SA) technique to deal with the same problem. The results in terms of efficiency, accuracy of the ground-state reconstructed surface and CPU time are compared.