Analysis of Crossover Operators for Cluster Geometry Optimization

Abstract

We study the effectiveness of different crossover operators in the global optimization of atomic clusters. Hybrid approaches combining a steady-state evolutionary algorithm and a local search procedure are state-of-the-art methods for this problem. In this paper we describe several crossover operators usually adopted for cluster geometry optimization tasks. Results show that operators that are sensitive to the phenotypical properties of the solutions help to enhance the performance of the optimization algorithm. They are able to identify and recombine useful building blocks and, therefore, increase the likelihood of performing a meaningful exploration of the search space.