Accurate and scalable graph neural network force field and molecular dynamics with direct force architecture
Abstract Recently, machine learning (ML) has been used to address the computational cost that has been limiting ab initio molecular dynamics (AIMD). Here, we present GNNFF, a graph neural network framework to directly predict atomic forces from automatically extracted features of the local atomic en...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2021-05-01
|
| Series: | npj Computational Materials |
| Online Access: | https://doi.org/10.1038/s41524-021-00543-3 |
| _version_ | 1818432538517962752 |
|---|---|
| author | Cheol Woo Park Mordechai Kornbluth Jonathan Vandermause Chris Wolverton Boris Kozinsky Jonathan P. Mailoa |
| author_facet | Cheol Woo Park Mordechai Kornbluth Jonathan Vandermause Chris Wolverton Boris Kozinsky Jonathan P. Mailoa |
| author_sort | Cheol Woo Park |
| collection | DOAJ |
| description | Abstract Recently, machine learning (ML) has been used to address the computational cost that has been limiting ab initio molecular dynamics (AIMD). Here, we present GNNFF, a graph neural network framework to directly predict atomic forces from automatically extracted features of the local atomic environment that are translationally-invariant, but rotationally-covariant to the coordinate of the atoms. We demonstrate that GNNFF not only achieves high performance in terms of force prediction accuracy and computational speed on various materials systems, but also accurately predicts the forces of a large MD system after being trained on forces obtained from a smaller system. Finally, we use our framework to perform an MD simulation of Li7P3S11, a superionic conductor, and show that resulting Li diffusion coefficient is within 14% of that obtained directly from AIMD. The high performance exhibited by GNNFF can be easily generalized to study atomistic level dynamics of other material systems. |
| first_indexed | 2024-12-14T16:06:48Z |
| format | Article |
| id | doaj.art-69d11dd21aef4871997919972e929e7b |
| institution | Directory Open Access Journal |
| issn | 2057-3960 |
| language | English |
| last_indexed | 2024-12-14T16:06:48Z |
| publishDate | 2021-05-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | npj Computational Materials |
| spelling | doaj.art-69d11dd21aef4871997919972e929e7b2022-12-21T22:55:04ZengNature Portfolionpj Computational Materials2057-39602021-05-01711910.1038/s41524-021-00543-3Accurate and scalable graph neural network force field and molecular dynamics with direct force architectureCheol Woo Park0Mordechai Kornbluth1Jonathan Vandermause2Chris Wolverton3Boris Kozinsky4Jonathan P. Mailoa5Robert Bosch Research and Technology CenterRobert Bosch Research and Technology CenterHarvard School of Engineering and Applied SciencesNorthwestern UniversityRobert Bosch Research and Technology CenterRobert Bosch Research and Technology CenterAbstract Recently, machine learning (ML) has been used to address the computational cost that has been limiting ab initio molecular dynamics (AIMD). Here, we present GNNFF, a graph neural network framework to directly predict atomic forces from automatically extracted features of the local atomic environment that are translationally-invariant, but rotationally-covariant to the coordinate of the atoms. We demonstrate that GNNFF not only achieves high performance in terms of force prediction accuracy and computational speed on various materials systems, but also accurately predicts the forces of a large MD system after being trained on forces obtained from a smaller system. Finally, we use our framework to perform an MD simulation of Li7P3S11, a superionic conductor, and show that resulting Li diffusion coefficient is within 14% of that obtained directly from AIMD. The high performance exhibited by GNNFF can be easily generalized to study atomistic level dynamics of other material systems.https://doi.org/10.1038/s41524-021-00543-3 |
| spellingShingle | Cheol Woo Park Mordechai Kornbluth Jonathan Vandermause Chris Wolverton Boris Kozinsky Jonathan P. Mailoa Accurate and scalable graph neural network force field and molecular dynamics with direct force architecture npj Computational Materials |
| title | Accurate and scalable graph neural network force field and molecular dynamics with direct force architecture |
| title_full | Accurate and scalable graph neural network force field and molecular dynamics with direct force architecture |
| title_fullStr | Accurate and scalable graph neural network force field and molecular dynamics with direct force architecture |
| title_full_unstemmed | Accurate and scalable graph neural network force field and molecular dynamics with direct force architecture |
| title_short | Accurate and scalable graph neural network force field and molecular dynamics with direct force architecture |
| title_sort | accurate and scalable graph neural network force field and molecular dynamics with direct force architecture |
| url | https://doi.org/10.1038/s41524-021-00543-3 |
| work_keys_str_mv | AT cheolwoopark accurateandscalablegraphneuralnetworkforcefieldandmoleculardynamicswithdirectforcearchitecture AT mordechaikornbluth accurateandscalablegraphneuralnetworkforcefieldandmoleculardynamicswithdirectforcearchitecture AT jonathanvandermause accurateandscalablegraphneuralnetworkforcefieldandmoleculardynamicswithdirectforcearchitecture AT chriswolverton accurateandscalablegraphneuralnetworkforcefieldandmoleculardynamicswithdirectforcearchitecture AT boriskozinsky accurateandscalablegraphneuralnetworkforcefieldandmoleculardynamicswithdirectforcearchitecture AT jonathanpmailoa accurateandscalablegraphneuralnetworkforcefieldandmoleculardynamicswithdirectforcearchitecture |