Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2509.26022 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908780331008000 |
|---|---|
| author | Takigawa, Atsushi Kiyohara, Shin Kumagai, Yu |
| author_facet | Takigawa, Atsushi Kiyohara, Shin Kumagai, Yu |
| contents | Considerable effort continues to be devoted to the exploration of next-generation high-\k{appa} materials that combine a high dielectric constant with a wide band gap. However, machine learning (ML)-based virtual screening has remained challenging, primarily due to the low accuracy in predicting the ionic contribution to the dielectric tensor, which dominates the dielectric performance of high-\k{appa} materials. We here propose a joint ML model that predicts Born effective charges using an equivariant graph neural network, and phonon properties using a highly accurate pretrained ML potential. The ionic dielectric tensor is then computed analytically from these quantities. This approach significantly improves the accuracy of ionic contribution. Using the proposed model, we successfully identified 38 novel high-\k{appa} oxides from a screening pool of over 8,000 candidates. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_26022 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Physics-Based Factorized Machine Learning for Predicting Ionic Dielectric Tensors Takigawa, Atsushi Kiyohara, Shin Kumagai, Yu Materials Science Considerable effort continues to be devoted to the exploration of next-generation high-\k{appa} materials that combine a high dielectric constant with a wide band gap. However, machine learning (ML)-based virtual screening has remained challenging, primarily due to the low accuracy in predicting the ionic contribution to the dielectric tensor, which dominates the dielectric performance of high-\k{appa} materials. We here propose a joint ML model that predicts Born effective charges using an equivariant graph neural network, and phonon properties using a highly accurate pretrained ML potential. The ionic dielectric tensor is then computed analytically from these quantities. This approach significantly improves the accuracy of ionic contribution. Using the proposed model, we successfully identified 38 novel high-\k{appa} oxides from a screening pool of over 8,000 candidates. |
| title | Physics-Based Factorized Machine Learning for Predicting Ionic Dielectric Tensors |
| topic | Materials Science |
| url | https://arxiv.org/abs/2509.26022 |