Machine learning technique has helped to create exact models of interatomic interaction
Researchers from the Moscow Institute of Physics and Technology (MIPT), Skolkovo Institute of Science and Technology «Skoltech» and All-Russia Research Institute of Automatics (VNIIA) used the method of machine learning to model the substances properties at the atomic level in different conditions. The method with a high level of accuracy requires relatively few computational resources and may be used to create new materials.
The models make possible to study the behavior of substances under various conditions and at different temperatures before the experimental work beginning. Existing models either do not have sufficient accuracy or require excessive time and computational power. The machine learning technique allows one to use only a small series of results obtained with the help of quantum mechanical calculations to create the model. «These models can then be used instead of quantum mechanical calculations, since they have the same accuracy but require about a thousand times less computational resources,» the MIPT press service told.
The scientists have used the new method to create models of crystals and melts of aluminum and uranium. The study results are published in Scientific Reports.
Liked it? Share it! —