Predicting challenging phase transitions with Bayesian active learning
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866915964397813760 |
|---|---|
| author | Bastonero, Lorenzo Joalland, Gabriel Cignarella, Chiara Monacelli, Lorenzo Marzari, Nicola |
| author_facet | Bastonero, Lorenzo Joalland, Gabriel Cignarella, Chiara Monacelli, Lorenzo Marzari, Nicola |
| contents | Materials underpin modern technologies, from energy harvesting, storage, and conversion to information and communication technologies. Their functionality is often governed by the interplay between competing phases, as thermodynamic behavior shapes microscopic properties and ultimately determines technological performance; for instance, the light absorption of inorganic metal-halide perovskites in solar cells. Accurately predicting crystal thermodynamics, however, remains a major challenge for computational approaches because strong anharmonic effects require extensive sampling of the potential energy surface. Here, we present an on-the-fly Bayesian framework, combined with the stochastic self-consistent harmonic approximation, for learning first-principles interatomic potentials. This approach enables the prediction of thermodynamic properties over a broad temperature range with first-principles accuracy while requiring training on only a few tens to a few hundreds of atomic configurations. To demonstrate its power, we investigate the thermodynamic and dynamical properties of Li$_2$O, $α$-CsPbI$_3$, and $δ$-CsPbI$_3$, requiring only 44, 256, and 50 total-energy calculations, respectively. Notably, we show that this framework accurately captures the phase diagram of CsPbI$_3$, which explains its spontaneous degradation into the non-absorbing yellow phase, predicting the transition temperature with remarkable accuracy and efficiency. More broadly, the method presented opens a novel route toward accelerated materials engineering under realistic conditions for a wide range of technologically relevant applications, including solid-state batteries, optoelectronic devices, and memristors. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_25756 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Predicting challenging phase transitions with Bayesian active learning Bastonero, Lorenzo Joalland, Gabriel Cignarella, Chiara Monacelli, Lorenzo Marzari, Nicola Materials Science Disordered Systems and Neural Networks Computational Physics Materials underpin modern technologies, from energy harvesting, storage, and conversion to information and communication technologies. Their functionality is often governed by the interplay between competing phases, as thermodynamic behavior shapes microscopic properties and ultimately determines technological performance; for instance, the light absorption of inorganic metal-halide perovskites in solar cells. Accurately predicting crystal thermodynamics, however, remains a major challenge for computational approaches because strong anharmonic effects require extensive sampling of the potential energy surface. Here, we present an on-the-fly Bayesian framework, combined with the stochastic self-consistent harmonic approximation, for learning first-principles interatomic potentials. This approach enables the prediction of thermodynamic properties over a broad temperature range with first-principles accuracy while requiring training on only a few tens to a few hundreds of atomic configurations. To demonstrate its power, we investigate the thermodynamic and dynamical properties of Li$_2$O, $α$-CsPbI$_3$, and $δ$-CsPbI$_3$, requiring only 44, 256, and 50 total-energy calculations, respectively. Notably, we show that this framework accurately captures the phase diagram of CsPbI$_3$, which explains its spontaneous degradation into the non-absorbing yellow phase, predicting the transition temperature with remarkable accuracy and efficiency. More broadly, the method presented opens a novel route toward accelerated materials engineering under realistic conditions for a wide range of technologically relevant applications, including solid-state batteries, optoelectronic devices, and memristors. |
| title | Predicting challenging phase transitions with Bayesian active learning |
| topic | Materials Science Disordered Systems and Neural Networks Computational Physics |
| url | https://arxiv.org/abs/2604.25756 |