Identifying stable Nb-O clusters using evolutionary algorithm and DFT: A foundation for machine learning potentials

Abstract

Crystal structure of the bulk NbO can be described as the NaCl (B1) lattice with equimolar 25 % content of ordered vacancies in both sublattices. While numerous studies have explored the phase stability ranges and crystallography of niobium oxides under various temperatures and pressures, the atomic structure of these compounds as small clusters remains unsolved. Understanding this structure is crucial for investigating the formation and growth of niobium oxide nanoparticles and thin films. In this work, the evolutionary algorithms guided by DFT calculations were employed to identify the most viable structures of Nb_nO_m clusters with indices 1 ≤ n ≤ 6, 0 ≤ m ≤ 6. The indices of clusters with enhanced stability and higher probabilities of formation during stochastic synthesis processes were proposed. Additionally, a machine learning potential for the Nb-O system was derived from the accumulated set of DFT calculations of Nb_nO_m clusters.