Towards an Effective Evolutionary Approach for Binary Lennard-Jones Clusters

We present a hybrid approach for the optimization of the geometry of heterogeneous chemical aggregates. The method combines a steady-state evolutionary algorithm with a local search procedure and is able to deal with an optimization situation where both the composition of the aggregate and the spatial distribution of the particles must be determined. Specifically, we address the problem of discovering the optimal configuration for several binary Lennard-Jones instances with different atomic sizes. Results show that the hybrid approach is effective in determining the putative global optima of clusters up to 55 atoms.