Globally-stable and metastable crystal structure enumeration using polynomial machine learning potentials in elemental As, Bi, Ga, In, La, P, Sb, Sn, and Te

Machine learning potentials (MLPs) have become essential tools for accelerating accurate large-scale atomistic simulations and crystal structure predictions. The polynomial MLPs, which are described by polynomial rotational invariants, have been systematically developed for many elemental and alloy systems. This study presents an efficient procedure for enumerating global minimum and metastable structures using robust polynomial MLPs. To develop MLPs that can handle the structure enumeration, performing random structure searches and updating MLPs are iteratively repeated. The current procedure is systematically applied to the elemental systems of As, Bi, Ga, In, La, P, Sb, Sn, and Te, with many local minimum structures competing with the global minimum structure in terms of energy. This procedure will significantly accelerate global structure searches and expand their search space.