Hydrogen diffusion in magnesium using machine learning potentials: a comparative study
Abstract Understanding and accurately predicting hydrogen diffusion in materials is challenging due to the complex interactions between hydrogen defects and the crystal lattice. These interactions span large length and time scales, making them difficult to address with standard ab-initio techniques....
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2025-03-01
|
| Series: | npj Computational Materials |
| Online Access: | https://doi.org/10.1038/s41524-025-01555-z |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850265455553413120 |
|---|---|
| author | Andrea Angeletti Luca Leoni Dario Massa Luca Pasquini Stefanos Papanikolaou Cesare Franchini |
| author_facet | Andrea Angeletti Luca Leoni Dario Massa Luca Pasquini Stefanos Papanikolaou Cesare Franchini |
| author_sort | Andrea Angeletti |
| collection | DOAJ |
| description | Abstract Understanding and accurately predicting hydrogen diffusion in materials is challenging due to the complex interactions between hydrogen defects and the crystal lattice. These interactions span large length and time scales, making them difficult to address with standard ab-initio techniques. This work addresses this challenge by employing accelerated machine learning (ML) molecular dynamics simulations through active learning. We conduct a comparative study of different ML-based interatomic potential schemes, including VASP, MACE, and CHGNet, utilizing various training strategies such as on-the-fly learning, pre-trained universal models, and fine-tuning. By considering different temperatures and concentration regimes, we obtain hydrogen diffusion coefficients and activation energy values which align remarkably well with experimental results, underlining the efficacy and accuracy of ML-assisted methodologies in the context of diffusive dynamics. Particularly, our procedure significantly reduces the computational effort associated with traditional transition state calculations or ad-hoc designed interatomic potentials. The results highlight the limitations of pre-trained universal solutions for defective materials and how they can be improved by fine-tuning. Specifically, fine-tuning the models on a database produced during on-the-fly training of VASP ML force-field allows the retrieval of DFT-level accuracy at a fraction of the computational cost. |
| format | Article |
| id | doaj-art-fe9fd8b607c04aa7bfbdd98d662aa442 |
| institution | OA Journals |
| issn | 2057-3960 |
| language | English |
| publishDate | 2025-03-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | npj Computational Materials |
| spelling | doaj-art-fe9fd8b607c04aa7bfbdd98d662aa4422025-08-20T01:54:25ZengNature Portfolionpj Computational Materials2057-39602025-03-011111810.1038/s41524-025-01555-zHydrogen diffusion in magnesium using machine learning potentials: a comparative studyAndrea Angeletti0Luca Leoni1Dario Massa2Luca Pasquini3Stefanos Papanikolaou4Cesare Franchini5Vienna Doctoral School in Physics, University of ViennaDepartment of Physics and Astronomy ’Augusto Righi’, Alma Mater Studiorum - Universitá di BolognaNOMATEN Centre of Excellence, National Center for Nuclear ResearchDepartment of Physics and Astronomy ’Augusto Righi’, Alma Mater Studiorum - Universitá di BolognaNOMATEN Centre of Excellence, National Center for Nuclear ResearchVienna Doctoral School in Physics, University of ViennaAbstract Understanding and accurately predicting hydrogen diffusion in materials is challenging due to the complex interactions between hydrogen defects and the crystal lattice. These interactions span large length and time scales, making them difficult to address with standard ab-initio techniques. This work addresses this challenge by employing accelerated machine learning (ML) molecular dynamics simulations through active learning. We conduct a comparative study of different ML-based interatomic potential schemes, including VASP, MACE, and CHGNet, utilizing various training strategies such as on-the-fly learning, pre-trained universal models, and fine-tuning. By considering different temperatures and concentration regimes, we obtain hydrogen diffusion coefficients and activation energy values which align remarkably well with experimental results, underlining the efficacy and accuracy of ML-assisted methodologies in the context of diffusive dynamics. Particularly, our procedure significantly reduces the computational effort associated with traditional transition state calculations or ad-hoc designed interatomic potentials. The results highlight the limitations of pre-trained universal solutions for defective materials and how they can be improved by fine-tuning. Specifically, fine-tuning the models on a database produced during on-the-fly training of VASP ML force-field allows the retrieval of DFT-level accuracy at a fraction of the computational cost.https://doi.org/10.1038/s41524-025-01555-z |
| spellingShingle | Andrea Angeletti Luca Leoni Dario Massa Luca Pasquini Stefanos Papanikolaou Cesare Franchini Hydrogen diffusion in magnesium using machine learning potentials: a comparative study npj Computational Materials |
| title | Hydrogen diffusion in magnesium using machine learning potentials: a comparative study |
| title_full | Hydrogen diffusion in magnesium using machine learning potentials: a comparative study |
| title_fullStr | Hydrogen diffusion in magnesium using machine learning potentials: a comparative study |
| title_full_unstemmed | Hydrogen diffusion in magnesium using machine learning potentials: a comparative study |
| title_short | Hydrogen diffusion in magnesium using machine learning potentials: a comparative study |
| title_sort | hydrogen diffusion in magnesium using machine learning potentials a comparative study |
| url | https://doi.org/10.1038/s41524-025-01555-z |
| work_keys_str_mv | AT andreaangeletti hydrogendiffusioninmagnesiumusingmachinelearningpotentialsacomparativestudy AT lucaleoni hydrogendiffusioninmagnesiumusingmachinelearningpotentialsacomparativestudy AT dariomassa hydrogendiffusioninmagnesiumusingmachinelearningpotentialsacomparativestudy AT lucapasquini hydrogendiffusioninmagnesiumusingmachinelearningpotentialsacomparativestudy AT stefanospapanikolaou hydrogendiffusioninmagnesiumusingmachinelearningpotentialsacomparativestudy AT cesarefranchini hydrogendiffusioninmagnesiumusingmachinelearningpotentialsacomparativestudy |