Machine-learned interatomic potentials: Recent developments and prospective applications

High-throughput generation of large and consistent ab initio data combined with advanced machine-learning techniques are enabling the creation of interatomic potentials of near ab initio quality. This capability has the potential of dramatically impacting materials research: (i) while classical interatomic potentials have become indispensable in atomistic simulations, such potentials are typically restricted to certain classes of materials. Machine-learned potentials (MLPs) are applicable to all classes of materials individually and, importantly, to any combinations of them; (ii) MLPs are by design reactive force fields. This Focus Issue provides an overview of the state of the art of MLPs by presenting a range of impressive applications including metallurgy, photovoltaics, proton transport, nanoparticles for catalysis, ionic conductors for solid state batteries, and crystal structure predictions. These investigations provide insight into the current challenges, and they present pathways for their solutions, thus setting the stage for exciting perspectives in computational materials research.