journal article·2025-12-01

Development and transferability of neural-network models for plasma-surface interactions

J. S. Draney, A. Z. Panagiotopoulos, D. B. Graves

Publisher page

Abstract

Plasma-surface interactions are increasingly critical to modern technologies; yet, accurate molecular dynamics simulations remain limited by the capabilities of interatomic potentials. Deep Potentials (DPs) promise to revolutionize the field by providing a systematic method for producing accurate interatomic potentials. The primary challenge of DP development is selecting a dataset, which efficiently spans the set of atomic environments one expects to encounter in the subsequent molecular dynamics simulations. The computational cost of density functional theory calculations, which are the typical basis for DP development, makes it impossible to directly verify the quality of a given DP. To address this challenge, we explore the development of a deep-learned interatomic potential, “DeepREBO,” trained to reproduce the behavior of the REBO2 empirical potential, enabling direct validation of training methodology and transferability. Using an active learning framework, we begin with a minimal dataset and iteratively expand it to train a Deep Potential-Smooth Edition model that faithfully reproduces REBO2 results for 25 eV hydrogen bombardment of diamond (001), a particularly challenging case. We show that small, carefully curated datasets can outperform large, unguided ones, with effective models requiring fewer than 15 000 snapshots. Subsequent transferability tests demonstrate that while DeepREBO generalizes well to diamond (111) surfaces, performance degrades for amorphous carbon or higher-energy impacts, highlighting the need for use-case-specific training data. We also evaluate methods to improve short-range repulsion. This study outlines best practices for training robust deep potentials and underscores the importance of dataset design for predictive plasma simulations.
Source: crossref · Quality: baseline

Research context

OpenAlex topics: Machine Learning in Materials Science · Electronic and Structural Properties of Oxides

SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.

Affiliations

Chemical and Biological Engineering, Princeton University 1 , Princeton, New Jersey 08540

Princeton Plasma Physics Laboratory 2 , Princeton, New Jersey 08540

Funding

U.S. Department of Energy

References (30)

Show source-provided references
1. J. Vac. Sci. Technol. B · 2024 · 42 · 04150110.1116/6.0003579 ↗
2. Phys. Rev. X · 2019 · 9 · 03105210.1103/PhysRevX.9.031052 ↗
3. Diamond Relat. Mater. · 2020 · 108 · 10783910.1016/j.diamond.2020.107839 ↗
4. J. Vac. Sci. Technol. A · 2007 · 25 · 152910.1116/1.2787713 ↗
5. J. Phys. D: Appl. Phys. · 2009 · 42 · 22200110.1088/0022-3727/42/22/222001 ↗
6. Plasma Sources Sci. Technol. · 2024 · 33 · 07500910.1088/1361-6595/ad5d12 ↗
7. J. Vac. Sci. Technol. A · 2024 · 42 · 04260410.1116/6.0003651 ↗
8. J. Nucl. Mater. · 2010 · 401 · 110.1016/j.jnucmat.2010.03.012 ↗
9. J. Phys. Chem. A · 2001 · 105 · 939610.1021/jp004368u ↗
10. Phys. Rev. Lett. · 2010 · 104 · 13640310.1103/PhysRevLett.104.136403 ↗
11. Nat. Commun. · 2021 · 12 · 727310.1038/s41467-021-27504-0 ↗
12. M. Neumann , J.Gin, B.Rhodes, S.Bennett, Z.Li, H.Choubisa, A.Hussey, and J.Godwin, “Orb: A fast, scalable neural network potential,” arXiv:2410.22570 (2024).
13. J. Chem. Phys. · 2023 · 159 · 05480110.1063/5.0155600 ↗
14. Comput. Phys. Commun. · 2022 · 271 · 10817110.1016/j.cpc.2021.108171 ↗
15. X. Fu , Z.Wu, W.Wang, T.Xie, S.Keten, R.Gomez-Bombarelli, and TommiJaakkola, “Forces are not enough: Benchmark and critical evaluation for machine learning force fields with molecular simulations,” Trans. Mach. Learn. Res. (published online 2023), https://openreview.net/forum?id=A8pqQipwkt.
16. Appl. Phys. Lett. · 2019 · 114 · 24410110.1063/1.5098061 ↗
17. Phys. Rev. B · 1989 · 40 · 9310.1103/PhysRevB.40.93 ↗
18. J. Vac. Sci. Technol. A · 2025 · 43 · 01260310.1116/6.0004027 ↗
19. J. Phys.: Condens. Matter · 2002 · 14 · 78310.1088/0953-8984/14/4/312 ↗
20. J. Phys. D: Appl. Phys. · 2025 · 58 · 02520610.1088/1361-6463/ad78e6 ↗
21. Phys. Rev. B · 2008 · 78 · 16140210.1103/PhysRevB.78.161402 ↗
22. Comput. Phys. Commun. · 2020 · 253 · 10720610.1016/j.cpc.2020.107206 ↗
23. J. Phys.: Condens. Matter · 2014 · 26 · 18300110.1088/0953-8984/26/18/183001 ↗
24. Phys. Rev. Mater. · 2019 · 3 · 02380410.1103/PhysRevMaterials.3.023804 ↗
25. J. Chem. Theory Comput. · 2022 · 18 · 555910.1021/acs.jctc.2c00102 ↗
26. J. Chem. Eng. Data · 2024 · 69 · 20410.1021/acs.jced.3c00580 ↗
27. J. Chem. Phys. · 2024 · 161 · 13450610.1063/5.0230440 ↗
28. Z. Naturforsch. A · 1947 · 2 · 13310.1515/zna-1947-0302 ↗
29. ACS Appl. Mater. Interfaces · 2024 · 16 · 4845710.1021/acsami.4c07949 ↗
30. J. S. Draney , A.Panagiotopoulos, and D.Graves, “Dataset, code, and models for training deep learning potentials for low temperature plasma-surface interactions,” data set, version 1 (Princeton University, 2025), https://doi.org/10.34770/wq4t-wa25.