Beyond potentials: Integrated machine learning models for materials

Abstract Over the past decade, interatomic potentials based on machine learning (ML) techniques have become an indispensable tool in atomic-scale modeling of materials. Trained energies and forces obtained from electronic-structure calculations, they inherit their predictive accuracy, extend greatly length time scales that are accessible to explicit atomistic simulations. Inexpensive predictions energetics individual configurations facilitated calculation thermodynamics materials, including finite-temperature effects disorder. More recently, ML models been closing gap with first-principles calculations another area: prediction arbitrarily complicated functional properties, vibrational optical spectroscopies electronic excitations. The implementation integrated combine energetic statistical dynamical sampling properties is bringing promise predictive, uncompromising simulations existing novel materials closer its full realization. Graphical abstract