Data-driven discovery of 2D materials by deep generative models
Abstract Efficient algorithms to generate candidate crystal structures with good stability properties can play a key role in data-driven materials discovery. Here, we show that a crystal diffusion variational autoencoder (CDVAE) is capable of generating two-dimensional (2D) materials of high chemica...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2022-11-01
|
| Series: | npj Computational Materials |
| Online Access: | https://doi.org/10.1038/s41524-022-00923-3 |
| _version_ | 1797988451373547520 |
|---|---|
| author | Peder Lyngby Kristian Sommer Thygesen |
| author_facet | Peder Lyngby Kristian Sommer Thygesen |
| author_sort | Peder Lyngby |
| collection | DOAJ |
| description | Abstract Efficient algorithms to generate candidate crystal structures with good stability properties can play a key role in data-driven materials discovery. Here, we show that a crystal diffusion variational autoencoder (CDVAE) is capable of generating two-dimensional (2D) materials of high chemical and structural diversity and formation energies mirroring the training structures. Specifically, we train the CDVAE on 2615 2D materials with energy above the convex hull ΔH hull < 0.3 eV/atom, and generate 5003 materials that we relax using density functional theory (DFT). We also generate 14192 new crystals by systematic element substitution of the training structures. We find that the generative model and lattice decoration approach are complementary and yield materials with similar stability properties but very different crystal structures and chemical compositions. In total we find 11630 predicted new 2D materials, where 8599 of these have ΔH hull < 0.3 eV/atom as the seed structures, while 2004 are within 50 meV of the convex hull and could potentially be synthesised. The relaxed atomic structures of all the materials are available in the open Computational 2D Materials Database (C2DB). Our work establishes the CDVAE as an efficient and reliable crystal generation machine, and significantly expands the space of 2D materials. |
| first_indexed | 2024-04-11T08:04:16Z |
| format | Article |
| id | doaj.art-ef7ab35130de4dd0b804976c4c3baaa2 |
| institution | Directory Open Access Journal |
| issn | 2057-3960 |
| language | English |
| last_indexed | 2024-04-11T08:04:16Z |
| publishDate | 2022-11-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | npj Computational Materials |
| spelling | doaj.art-ef7ab35130de4dd0b804976c4c3baaa22022-12-22T04:35:37ZengNature Portfolionpj Computational Materials2057-39602022-11-01811810.1038/s41524-022-00923-3Data-driven discovery of 2D materials by deep generative modelsPeder Lyngby0Kristian Sommer Thygesen1Computational Atomic-scale Materials Design (CAMD), Department of Physics, Technical University of DenmarkComputational Atomic-scale Materials Design (CAMD), Department of Physics, Technical University of DenmarkAbstract Efficient algorithms to generate candidate crystal structures with good stability properties can play a key role in data-driven materials discovery. Here, we show that a crystal diffusion variational autoencoder (CDVAE) is capable of generating two-dimensional (2D) materials of high chemical and structural diversity and formation energies mirroring the training structures. Specifically, we train the CDVAE on 2615 2D materials with energy above the convex hull ΔH hull < 0.3 eV/atom, and generate 5003 materials that we relax using density functional theory (DFT). We also generate 14192 new crystals by systematic element substitution of the training structures. We find that the generative model and lattice decoration approach are complementary and yield materials with similar stability properties but very different crystal structures and chemical compositions. In total we find 11630 predicted new 2D materials, where 8599 of these have ΔH hull < 0.3 eV/atom as the seed structures, while 2004 are within 50 meV of the convex hull and could potentially be synthesised. The relaxed atomic structures of all the materials are available in the open Computational 2D Materials Database (C2DB). Our work establishes the CDVAE as an efficient and reliable crystal generation machine, and significantly expands the space of 2D materials.https://doi.org/10.1038/s41524-022-00923-3 |
| spellingShingle | Peder Lyngby Kristian Sommer Thygesen Data-driven discovery of 2D materials by deep generative models npj Computational Materials |
| title | Data-driven discovery of 2D materials by deep generative models |
| title_full | Data-driven discovery of 2D materials by deep generative models |
| title_fullStr | Data-driven discovery of 2D materials by deep generative models |
| title_full_unstemmed | Data-driven discovery of 2D materials by deep generative models |
| title_short | Data-driven discovery of 2D materials by deep generative models |
| title_sort | data driven discovery of 2d materials by deep generative models |
| url | https://doi.org/10.1038/s41524-022-00923-3 |
| work_keys_str_mv | AT pederlyngby datadrivendiscoveryof2dmaterialsbydeepgenerativemodels AT kristiansommerthygesen datadrivendiscoveryof2dmaterialsbydeepgenerativemodels |