General framework for E(3)-equivariant neural network representation of density functional theory Hamiltonian
Abstract The combination of deep learning and ab initio calculation has shown great promise in revolutionizing future scientific research, but how to design neural network models incorporating a priori knowledge and symmetry requirements is a key challenging subject. Here we propose an E(3)-equivari...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2023-05-01
|
| Series: | Nature Communications |
| Online Access: | https://doi.org/10.1038/s41467-023-38468-8 |
| _version_ | 1797822817405763584 |
|---|---|
| author | Xiaoxun Gong He Li Nianlong Zou Runzhang Xu Wenhui Duan Yong Xu |
| author_facet | Xiaoxun Gong He Li Nianlong Zou Runzhang Xu Wenhui Duan Yong Xu |
| author_sort | Xiaoxun Gong |
| collection | DOAJ |
| description | Abstract The combination of deep learning and ab initio calculation has shown great promise in revolutionizing future scientific research, but how to design neural network models incorporating a priori knowledge and symmetry requirements is a key challenging subject. Here we propose an E(3)-equivariant deep-learning framework to represent density functional theory (DFT) Hamiltonian as a function of material structure, which can naturally preserve the Euclidean symmetry even in the presence of spin–orbit coupling. Our DeepH-E3 method enables efficient electronic structure calculation at ab initio accuracy by learning from DFT data of small-sized structures, making the routine study of large-scale supercells (>104 atoms) feasible. The method can reach sub-meV prediction accuracy at high training efficiency, showing state-of-the-art performance in our experiments. The work is not only of general significance to deep-learning method development but also creates opportunities for materials research, such as building a Moiré-twisted material database. |
| first_indexed | 2024-03-13T10:13:44Z |
| format | Article |
| id | doaj.art-9047c48016244dba9494cae0bcd4136a |
| institution | Directory Open Access Journal |
| issn | 2041-1723 |
| language | English |
| last_indexed | 2024-03-13T10:13:44Z |
| publishDate | 2023-05-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | Nature Communications |
| spelling | doaj.art-9047c48016244dba9494cae0bcd4136a2023-05-21T11:20:35ZengNature PortfolioNature Communications2041-17232023-05-0114111010.1038/s41467-023-38468-8General framework for E(3)-equivariant neural network representation of density functional theory HamiltonianXiaoxun Gong0He Li1Nianlong Zou2Runzhang Xu3Wenhui Duan4Yong Xu5State Key Laboratory of Low Dimensional Quantum Physics and Department of Physics, Tsinghua UniversityState Key Laboratory of Low Dimensional Quantum Physics and Department of Physics, Tsinghua UniversityState Key Laboratory of Low Dimensional Quantum Physics and Department of Physics, Tsinghua UniversityState Key Laboratory of Low Dimensional Quantum Physics and Department of Physics, Tsinghua UniversityState Key Laboratory of Low Dimensional Quantum Physics and Department of Physics, Tsinghua UniversityState Key Laboratory of Low Dimensional Quantum Physics and Department of Physics, Tsinghua UniversityAbstract The combination of deep learning and ab initio calculation has shown great promise in revolutionizing future scientific research, but how to design neural network models incorporating a priori knowledge and symmetry requirements is a key challenging subject. Here we propose an E(3)-equivariant deep-learning framework to represent density functional theory (DFT) Hamiltonian as a function of material structure, which can naturally preserve the Euclidean symmetry even in the presence of spin–orbit coupling. Our DeepH-E3 method enables efficient electronic structure calculation at ab initio accuracy by learning from DFT data of small-sized structures, making the routine study of large-scale supercells (>104 atoms) feasible. The method can reach sub-meV prediction accuracy at high training efficiency, showing state-of-the-art performance in our experiments. The work is not only of general significance to deep-learning method development but also creates opportunities for materials research, such as building a Moiré-twisted material database.https://doi.org/10.1038/s41467-023-38468-8 |
| spellingShingle | Xiaoxun Gong He Li Nianlong Zou Runzhang Xu Wenhui Duan Yong Xu General framework for E(3)-equivariant neural network representation of density functional theory Hamiltonian Nature Communications |
| title | General framework for E(3)-equivariant neural network representation of density functional theory Hamiltonian |
| title_full | General framework for E(3)-equivariant neural network representation of density functional theory Hamiltonian |
| title_fullStr | General framework for E(3)-equivariant neural network representation of density functional theory Hamiltonian |
| title_full_unstemmed | General framework for E(3)-equivariant neural network representation of density functional theory Hamiltonian |
| title_short | General framework for E(3)-equivariant neural network representation of density functional theory Hamiltonian |
| title_sort | general framework for e 3 equivariant neural network representation of density functional theory hamiltonian |
| url | https://doi.org/10.1038/s41467-023-38468-8 |
| work_keys_str_mv | AT xiaoxungong generalframeworkfore3equivariantneuralnetworkrepresentationofdensityfunctionaltheoryhamiltonian AT heli generalframeworkfore3equivariantneuralnetworkrepresentationofdensityfunctionaltheoryhamiltonian AT nianlongzou generalframeworkfore3equivariantneuralnetworkrepresentationofdensityfunctionaltheoryhamiltonian AT runzhangxu generalframeworkfore3equivariantneuralnetworkrepresentationofdensityfunctionaltheoryhamiltonian AT wenhuiduan generalframeworkfore3equivariantneuralnetworkrepresentationofdensityfunctionaltheoryhamiltonian AT yongxu generalframeworkfore3equivariantneuralnetworkrepresentationofdensityfunctionaltheoryhamiltonian |