Physically informed artificial neural networks for atomistic modeling of materials

By G. P. P. Pun, Rohit Batra, Rampi Ramprasad, Yuri Mishin. Nature Communications, 2018

333 citations recorded by the source corpus.

Large-scale atomistic computer simulations of materials heavily rely on interatomic potentials predicting the energy and Newtonian forces on atoms. Traditional interatomic potentials are based on physical intuition but contain few adjustable parameters and are usually not accurate. The emerging machine-learning (ML) potentials achieve highly accurate interpolation within a large DFT database but, being purely mathematical constructions, suffer from poor transferability to unknown structures. We propose a new approach that can drastically improve the transferability of ML potentials by informing them of the physical nature of interatomic bonding. This is achieved by combining a rather general physics-based model (analytical bond-order potential) with a neural-network regression. This approach, called the physically informed neural network (PINN) potential, is demonstrated by developing a general-purpose PINN potential for Al. We suggest that the development of physics-based ML potentials is the most effective way forward in the field of atomistic simulations. Traditional machine learning potentials suffer from poor transferability to unknown structures. Here the authors present an approach to improve the transferability of machine-learning potentials by including information on the physical nature of interatomic bonding.

Subfields: Atomic and Molecular Physics, and Optics, Materials Chemistry, Statistical and Nonlinear Physics

View Paper: https://www.semanticscholar.org/paper/9275d72d6f08c5b05a91088df9fc2e76fa505361

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Atomic and Molecular Physics, and Optics

1. Yuri Mishin
2. Evgeny V. Podryabinkin
3. Shyue Ping Ong
4. Chi Chen
5. Rohit Batra
6. Alexander V. Shapeev
7. Konstantin Gubaev
8. Rampi Ramprasad

In Materials Chemistry

1. Yuri Mishin
2. Shyue Ping Ong
3. Chi Chen
4. Evgeny V. Podryabinkin
5. Alexander V. Shapeev
6. Rohit Batra
7. Konstantin Gubaev
8. Ivan S. Novikov

In Statistical and Nonlinear Physics

1. Yuri Mishin
2. Simon J. L. Billinge
3. Alexander V. Shapeev
4. Rampi Ramprasad

Citations in the corpus

• Machine-Learning Interatomic Potentials for Materials Science (2021, methodology): “General-purpose PINN potentials have been constructed for Al [85, 86] and Ta [87], with several other potentials being currently developed.”
• Emerging materials intelligence ecosystems propelled by machine learning (2020, background): “Thus, in analogy with how the last sigmoid layer of a NN classifier constrains its output to class probabilities, the last physics-based layer ensures that the output energy and force predictions follow a physically meaningful trend, as dictated by its functional form 221 .”
• The MLIP package: moment tensor potentials with MPI and active learning (2020, background): “Neural network has been the first and are probably the most popular form of regressor for machine-learning potentials [1,4,17,20,60,69,79].”
• Recent advances and applications of deep learning methods in materials science (2021, methodology): “In addition to development of pure NN based FFs, there have also been recent developments of combining traditional FFs such as bondorder potentials with NNs and ReaxFF with message passing neural network (MPNN) that can help mitigate the NNs issue for extrapolation [119, 120].”