Neural Network Interatomic Potentials

Abstract

Modern large-scale atomistic simulations of materials are enabled by classical interatomic potentials. Such potentials directly relate atomic coordinates to the energy and forces acting on atoms without performing first-principles calculations. Traditional potentialsare based on physical models of interatomic bonding with free parameters fit to a limited set of experimental and first-principles data. In recent years, a new approach has emerged in which the potentials are developed by numerical interpolation between thepoints of a large first-principles database by applying machine-learning methods. In particular, artificial neural networks have been utilized to construct several machine-learning potentials. Being purely mathematical, such potentials enable extremely accurate predictionsof energy and forces but suffer from slow speed and poor transferability outside the training dataset. This proposal will explore a different direction in which neural networks will be combined with physics-based models of interatomic bonding. By contrast to the mathematical potentials, the proposed physical neural- network potentials are expected to be much faster and demonstrate a better transferability while drastically improving the accuracy of the traditional potentials. Copper, aluminum and tungsten are proposed asmodel materials to demonstrate the basic principles and prepare the ground for extension to binary and multicomponent systems in the future.

Document Details

Document Type
DoD Grant Award
Publication Date
May 05, 2017
Source ID
N000141712444

Entities

People

  • Y. Mishin

Organizations

  • George Mason University
  • Office of Naval Research
  • United States Navy

Tags

Readers

  • Computational Fluid Dynamics (CFD)
  • Neural Network Machine Learning.
  • Quantum spin resonance or Electron Paramagnetic Resonance spectroscopy.

Technology Areas

  • AI & ML
  • AI & ML - Neural Networks