Journal article · 2025 · Open access

Machine-learning interatomic potentials from a users perspective: a comparison of accuracy, speed and data efficiency

Niklas Leimeroth, Linus C. Erhard, Karsten Albe, Jochen Rohrer

Modelling and Simulation in Materials Science and Engineering, 33(6), 065012 · Published 29 July 2025 · DOI 10.1088/1361-651x/adf56d

Open full textCite this paper

In brief

“Machine-learning interatomic potentials from a users perspective: a comparison of accuracy, speed and data efficiency” is a 2025 journal article by Niklas Leimeroth and 3 co-authors, published in Modelling and Simulation in Materials Science and Engineering. OpenAlex counts 61 citations as of 19 November 2026, which puts it in the top 1% of works published in 2025 by citations. It is open access under a CC BY licence: the full text is free to read at Modelling and Simulation in Materials Science and Engineering.

Citations
61
Most cited in
2026 (50)
Field-weighted impact
9.39
References
66
Authors
4

Abstract

Machine learning interatomic potentials (MLIPs) have massively changed the field of atomistic modeling. They enable the accuracy of density functional theory in large-scale simulations while being nearly as fast as classical interatomic potentials (IPs). Over the last few years, a wide range of different types of MLIPs have been developed, but it is often difficult to judge which approach is the best for a given problem setting. For the case of structurally and chemically complex solids, namely Al–Cu–Zr and Si–O, we benchmark a range of MLIP approaches, in particular, the Gaussian approximation potential, high-dimensional neural network potentials, moment tensor potentials, the atomic cluster expansion (ACE) in its linear and nonlinear version, neural equivariant interatomic potentials (NequIP), Allegro, and MACE. We find that nonlinear ACE and the equivariant, message-passing graph neural networks NequIP and MACE form the Pareto front in the accuracy vs. computational cost trade-off. In case of the Al–Cu–Zr system we find that MACE and Allegro offer the highest accuracy, while NequIP outperforms them for Si–O. Furthermore, GPUs can massively accelerate the MLIPs, bringing them on par with and even ahead of non-accelerated classical IPs with regards to accessible timescales. Finally, we explore the extrapolation behavior of the corresponding potentials, probe the smoothness of the potential energy surfaces, and estimate the user friendliness of the corresponding fitting codes and molecular dynamics interfaces.

Cite this paper

Formatted from the catalogue record. Check names, capitals and page numbers against the paper before you submit.

APA 7

Leimeroth, N., Erhard, L. C., Albe, K., & Rohrer, J. (2025). Machine-learning interatomic potentials from a users perspective: a comparison of accuracy, speed and data efficiency. Modelling and Simulation in Materials Science and Engineering, 33(6), 065012. https://doi.org/10.1088/1361-651x/adf56d

MLA 9

Leimeroth, Niklas, et al. “Machine-learning interatomic potentials from a users perspective: a comparison of accuracy, speed and data efficiency.” Modelling and Simulation in Materials Science and Engineering, vol. 33, no. 6, 2025, p. 065012. https://doi.org/10.1088/1361-651x/adf56d.

Chicago

Leimeroth, Niklas, Linus C. Erhard, Karsten Albe, and Jochen Rohrer. 2025. “Machine-learning interatomic potentials from a users perspective: a comparison of accuracy, speed and data efficiency.” Modelling and Simulation in Materials Science and Engineering 33 (6): 065012. https://doi.org/10.1088/1361-651x/adf56d.

Harvard

Leimeroth, N. et al. (2025) ‘Machine-learning interatomic potentials from a users perspective: a comparison of accuracy, speed and data efficiency’, Modelling and Simulation in Materials Science and Engineering, 33(6), p. 065012. doi:10.1088/1361-651x/adf56d.

IEEE

N. Leimeroth, L. C. Erhard, K. Albe, and J. Rohrer, “Machine-learning interatomic potentials from a users perspective: a comparison of accuracy, speed and data efficiency,” Modelling and Simulation in Materials Science and Engineering, vol. 33, no. 6, p. 065012, 2025, doi: 10.1088/1361-651x/adf56d.

BibTeX

@article{leimeroth2025machine,
  title     = {{Machine-learning interatomic potentials from a users perspective: a comparison of accuracy, speed and data efficiency}},
  author    = {Leimeroth, Niklas and Erhard, Linus C. and Albe, Karsten and Rohrer, Jochen},
  journal   = {Modelling and Simulation in Materials Science and Engineering},
  year      = {2025},
  volume    = {33},
  number    = {6},
  pages     = {065012},
  doi       = {10.1088/1361-651x/adf56d},
  url       = {https://doi.org/10.1088/1361-651x/adf56d}
}

Download for a reference manager: BibTeX (.bib) · RIS for Zotero, Mendeley and EndNote (.ris)

Topic and keywords

Physical Sciences › Materials Science › Materials Chemistry › Machine Learning in Materials Science

  • machine learning interatomic potentials
  • benchmarking
  • Cu-Zr-Al alloys
  • Gaussian approximation potential
  • neural network potentials
  • moment tensor potential
  • atomic cluster expansion
  • NequIP

Work with this paper

Most-cited papers in Materials Science

All highly cited Materials Science papers

Questions about this paper

How do I cite this paper in APA style?

Leimeroth, N., Erhard, L. C., Albe, K., & Rohrer, J. (2025). Machine-learning interatomic potentials from a users perspective: a comparison of accuracy, speed and data efficiency. Modelling and Simulation in Materials Science and Engineering, 33(6), 065012. https://doi.org/10.1088/1361-651x/adf56d

How many times has this paper been cited?

61 times, according to OpenAlex on 19 November 2026. Google Scholar and Scopus count different sets of sources, so their numbers differ.

Is this paper free to read?

Yes. It is open access at Modelling and Simulation in Materials Science and Engineering: sign in and open the full text from this page.

Who wrote this paper?

Niklas Leimeroth, Linus C. Erhard, Karsten Albe, Jochen Rohrer.

What is the DOI of this paper?

10.1088/1361-651x/adf56d, permanently at https://doi.org/10.1088/1361-651x/adf56d. Sign in and open it from this page.

Bibliographic data from OpenAlex (CC0), last checked 19 November 2026. ResearchGuru does not host this paper; once you are signed in, the links open it at the publisher or the repository that does.