Machine learning aids in simulating dynamics of interacting atoms

Machine learning aids in simulating dynamics of interacting atoms


IMAGE: Automated information set era offers a extremely various sampling of atomic positions for coaching an correct and normal machine studying mannequin.
view more 

Credit score: Los Alamos Nationwide Laboratory

LOS ALAMOS, N.M., February 23, 2021–A revolutionary machine-learning (ML) method to simulate the motions of atoms in supplies resembling aluminum is described on this week’s Nature Communications journal. This automated method to “interatomic potential growth” might remodel the sector of computational supplies discovery.

“This method guarantees to be an vital constructing block for the research of supplies injury and growing older from first rules,” stated undertaking lead Justin Smith of Los Alamos Nationwide Laboratory. “Simulating the dynamics of interacting atoms is a cornerstone of understanding and growing new supplies. Machine studying strategies are offering computational scientists new instruments to precisely and effectively conduct these atomistic simulations. Machine studying fashions like this are designed to emulate the outcomes of extremely correct quantum simulations, at a small fraction of the computational value.”

To maximise the overall accuracy of those machine studying fashions, he stated, it’s important to design a extremely various dataset from which to coach the mannequin. A problem is that it isn’t apparent, a priori, what coaching information will probably be most wanted by the ML mannequin. The group’s current work presents an automatic “energetic studying” methodology for iteratively constructing a coaching dataset.

At every iteration, the tactic makes use of the current-best machine studying mannequin to carry out atomistic simulations; when new bodily conditions are encountered which are past the ML mannequin’s data, new reference information is collected by way of costly quantum simulations, and the ML mannequin is retrained. By this course of, the energetic studying process collects information relating to many various kinds of atomic configurations, together with a wide range of crystal constructions, and a wide range of defect patterns showing inside crystals.


The paper: Automated discovery of a strong interatomic potential for aluminum, Nature Communications, DOI: 10.1038/s41467-Zero21-21376-Zero

The funding: This work was funded partly by the Los Alamos Nationwide Laboratory Superior Simulation and Computing (ASC) program and laptop time was supplied by the Lawrence Livermore Nationwide Laboratory Sierra Supercomputer throughout its open entry interval.

About Los Alamos National Laboratory

Los Alamos Nationwide Laboratory, a multidisciplinary analysis establishment engaged in strategic science on behalf of nationwide safety, is managed by Triad, a public service oriented, nationwide safety science group equally owned by its three founding members: Battelle Memorial Institute (Battelle), the Texas A&M College System (TAMUS), and the Regents of the College of California (UC) for the Division of Vitality’s Nationwide Nuclear Safety Administration.

Los Alamos enhances nationwide safety by making certain the protection and reliability of the U.S. nuclear stockpile, growing applied sciences to scale back threats from weapons of mass destruction, and fixing issues associated to vitality, surroundings, infrastructure, well being, and world safety issues.

Disclaimer: AAAS and EurekAlert! are usually not answerable for the accuracy of reports releases posted to EurekAlert! by contributing establishments or for the usage of any data by means of the EurekAlert system.

Leave a Reply

Your email address will not be published. Required fields are marked *