Generation of global minimum energy structures of small molecular clusters using machine learning technique

To determine the ground state geometry of chemical systems, the minimization of an energy functional is carried out. Due to the large size of the search space and the probability to get stuck in nearby local minima in the potential energy surface, global optimization problems become very challenging. In the present chapter, we report the determination of global minimum energy structure using three different techniques, particle swarm optimization (PSO) combined with density functional theory, PSO with convolutional neural network, and density functional theory integrated firefly algorithm. We highlight their efficiency and accuracy by considering different metallic and nonmetallic clusters as prototype examples. These techniques can be used as efficient “global optimizer” tools. The proposed algorithms are interfaced with a computational chemistry software package, Gaussian, to calculate the single-point energies. The techniques discussed here turn out to be more efficient for smaller clusters with low computational cost compared to the typical large population. One can try with a large number of particles as well. At the same time, we have made a comparison of the efficiency of our proposed techniques with other evolutionary techniques.