Litcius/Paper detail

Development of a general-purpose machine-learning interatomic potential for aluminum by the physically informed neural network method

G. P. Purja Pun, V. Yamakov, J. Hickman, E. H. Glaessgen, Y. Mishin

2020Physical Review Materials38 citationsDOIOpen Access PDF

Abstract

Interatomic potentials are the key components of large-scale atomistic simulations of materials. The recently proposed physically-informed neural network (PINN) method combines a high-dimensional regression implemented by an artificial neural network with a physics-based bond-order interatomic potential. Here, the authors develop a highly accurate and transferable PINN potential that reproduces a broad spectrum of physical properties of Al, ranging from lattice dynamics and defect energies to liquid structure and dynamic and the solid-liquid interface tension. The potential enables atomistic simulations of Al with nearly first-principles accuracy while being orders of magnitudes faster.

Topics & Concepts

Interatomic potentialMaterials scienceArtificial neural networkMolecular dynamicsLattice (music)AluminiumStatistical physicsInterface (matter)RangingEmbedded atom modelNetwork structureChemical physicsDevelopment (topology)Key (lock)Machine Learning in Materials ScienceModel Reduction and Neural NetworksQuantum many-body systems
Development of a general-purpose machine-learning interatomic potential for aluminum by the physically informed neural network method | Litcius