Scientists from Skoltech and MIPT led by Prof. Artem R. Oganov have created a technique for reconstructing interatomic interaction patterns in crystals using machine learning approach.
Reconstructed and real potentials in AxBy Lennard-Jones system.
Pavel Dolgirev, first author: «We faced a very difficult challenge. It took a lot of time for people to learn how to predict structures by quantum mechanics. The progress was achieved just recently by the team of my supervisor, Artem R. Oganov. They have developed the USPEX algorithm, which does wonderful things by an evolutionary algorithm. Yet, it is still limited by speed of quantum calculations, for example. In the future it will probably be possible to combine it with machine learning and at the end get an algorithm with the accuracy of quantum calculations and speed of machine learning».
The input information for the machine learning is the atomic positions in the crystal structure. The output is the energy of such crystal structure. Scientists were interested in the energy, because the stability of the structure is determined by the minimum of energy. In the process of learning algorithm study how to find a correlation between the input and output signals.
The main problem was to interpret the results from the point of view of physics. Using the proposed model, scientists studied metals, noble gases and carbon. The suggested machine learning method allows restoring the picture of interatomic interactions without any assumptions. Researchers obtained the expected Lennard-Jones potentials for noble gases and, much more surprisingly, the expected oscillatory potentials for metals. Further work on this project will focus on the improvement of the method.
Further work on this project will focus on the improvement of the method.
The results of this project were published by American Institute of Physics (AIP).
*****
The Skolkovo Institute of Science and Technology (Skoltech) is a private graduate research university in Skolkovo, Russia, a suburb of Moscow. Established in 2011 in collaboration with MIT, Skoltech educates global leaders in innovation, advances scientific knowledge, and fosters new technologies to address critical issues facing Russia and the world. Applying international research and educational models, the university integrates the best Russian scientific traditions with twenty-first century entrepreneurship and innovation.: https://www.skoltech.ru/
Contact information:
Skoltech Communications
+7 (495) 280 14 81
