DOE OSTI · 1543873
Optimized symmetry functions for machine-learning interatomic potentials of multicomponent systems
Abstract
Current machine-learning methods to reproduce ab initio potential energy landscapes suffer from an unfavorable computational scaling with respect to the number of chemical species. In this work, we propose a new approach by using optimized symmetry functions to explore similarities of structures in multicomponent systems in order to yield linear complexity. Here, we combine these symmetry functions with the charge equilibration via neural network technique, a reliable artificial neural network potential for ionic materials, and apply this method to study alkali-halide materials MX with 6 chemical species (M = {Li, Na, K} and X = {F, Cl, Br}). Our results show that our approach provides good agreement both with experimental and DFT reference data of many physical and structural properties for any chemical combination.
Explore related subjects
Keep this discovery
Explore connections, maps & timelines
Rostami, Samare, Amsler, Maximilian, Ghasemi, S. Alireza. 2018-09-25. Optimized symmetry functions for machine-learning interatomic potentials of multicomponent systems. https://doi.org/10.1063/1.5040005. https://searchnasa.com/record/osti%3A1543873/optimized-symmetry-functions-for-machine-learning-interatomic-potentials-of-multicomponent-systems
Citations retain the original authors and DOI and include this site’s record URL. If this discovery service contributed to your research, you can also cite Search NASA by Indiciae: BibTeX · RIS.