De novo exploration and self-guided learning of potential-energy surfaces

Abstract

Interatomic potential models based on machine learning (ML) are rapidly developing as tools for materials simulations. However, because of their flexibility, they require large fitting databases that are normally created with substantial manual selection and tuning of reference configurations. Here, we show that ML potentials can be built in a largely automated fashion, exploring and fitting potential-energy surfaces from the beginning (de novo) within one and the same protocol. The key enabling step is the use of a configuration-averaged kernel metric that allows one to select the few most relevant structures at each step. The resulting potentials are accurate and robust for the wide range of configurations that occur during structure searching, despite only requiring a relatively small number of single-point DFT calculations on small unit cells. We apply the method to materials with diverse chemical nature and coordination environments, marking a milestone toward the more routine application of ML potentials in physics, chemistry, and materials science.

The ANI-1ccx and ANI-1x data sets, coupled-cluster and density functional theory properties for molecules

“…Robotics has spearheaded the efforts to build such data sets through the use of active learning 2 : building data sets by asking ML models to choose what data needs to be added to a training set to perform better next time. Although the concept of active learning originates from robotics, it has recently grown into an extremely important tool for collecting quantum chemistry data sets for use in ML applications [3][4][5][6][7][8][9][10][11] .…”

Machine-learned interatomic potentials by active learning: amorphous and liquid hafnium dioxide

“…The full details of the AL scheme are discussed in the "Methods" section. We also refer the reader to the recent success in the applications of AL [24][25][26] .…”

Understanding the geometric diversity of inorganic and hybrid frameworks through structural coarse-graining

“…Structures were visualised using VESTA. 29 distance in any given structure is the same (here, 1.0Å) 24 an idea that originated in the eld of chemical topology. 25 This is a step of key importance, because otherwise the overlap of neighbour densities will be necessarily diminished as soon as there are different A-B distances ( Fig.…”