Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
119 tokens/sec
GPT-4o
56 tokens/sec
Gemini 2.5 Pro Pro
43 tokens/sec
o3 Pro
6 tokens/sec
GPT-4.1 Pro
47 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Overcoming systematic softening in universal machine learning interatomic potentials by fine-tuning (2405.07105v1)

Published 11 May 2024 in cond-mat.mtrl-sci, cs.AI, and cs.LG

Abstract: Machine learning interatomic potentials (MLIPs) have introduced a new paradigm for atomic simulations. Recent advancements have seen the emergence of universal MLIPs (uMLIPs) that are pre-trained on diverse materials datasets, providing opportunities for both ready-to-use universal force fields and robust foundations for downstream machine learning refinements. However, their performance in extrapolating to out-of-distribution complex atomic environments remains unclear. In this study, we highlight a consistent potential energy surface (PES) softening effect in three uMLIPs: M3GNet, CHGNet, and MACE-MP-0, which is characterized by energy and force under-prediction in a series of atomic-modeling benchmarks including surfaces, defects, solid-solution energetics, phonon vibration modes, ion migration barriers, and general high-energy states. We find that the PES softening behavior originates from a systematic underprediction error of the PES curvature, which derives from the biased sampling of near-equilibrium atomic arrangements in uMLIP pre-training datasets. We demonstrate that the PES softening issue can be effectively rectified by fine-tuning with a single additional data point. Our findings suggest that a considerable fraction of uMLIP errors are highly systematic, and can therefore be efficiently corrected. This result rationalizes the data-efficient fine-tuning performance boost commonly observed with foundational MLIPs. We argue for the importance of a comprehensive materials dataset with improved PES sampling for next-generation foundational MLIPs.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (28)
  1. J. Gasteiger, J. Groß, and S. Günnemann, Directional message passing for molecular graphs, ICLR 10.48550/arxiv.2003.03123 (2020).
  2. C. Chen and S. P. Ong, A universal graph deep learning interatomic potential for the periodic table, Nature Computational Science 2, 718–728 (2022).
  3. B. Focassio, L. P. M. Freitas, and G. R. Schleder, Performance assessment of universal machine learning interatomic potentials: Challenges and directions for materials’ surfaces, arXiv preprint arXiv:2403.04217  (2024).
  4. T. Xie and J. C. Grossman, Crystal graph convolutional neural networks for an accurate and interpretable prediction of material properties, Physical Review Letters 120, 145301 (2018).
  5. S. Luo, T. Chen, and A. S. Krishnapriyan, Enabling efficient equivariant operations in the fourier basis via gaunt tensor products, arXiv preprint arXiv:2401.10216  (2024).
  6. B. Cheng, Cartesian atomic cluster expansion for machine learning interatomic potentials, arXiv preprint arXiv:2402.07472  (2024).
  7. I. Obot, D. Macdonald, and Z. Gasem, Density functional theory (dft) as a powerful tool for designing new organic corrosion inhibitors. part 1: An overview, Corrosion Science 99, 1–30 (2015).
  8. K. A. Fichthorn and M. Scheffler, Island nucleation in thin-film epitaxy: A first-principles investigation, Physical Review Letters 84, 5371–5374 (2000).
  9. V. Gurylev and T. P. Perng, Defect engineering of zno: Review on oxygen and zinc vacancies, Journal of the European Ceramic Society 41, 4977–4996 (2021).
  10. K. Kang and G. Ceder, Factors that affect li mobility in layered lithium transition metal oxides, Physical Review B 74, 094105 (2006).
  11. A. van de Walle, Multicomponent multisublattice alloys, nonconfigurational entropy and other additions to the Alloy Theoretic Automated Toolkit, Calphad 33, 266 (2009).
  12. Cation-disordered rocksalt-type high-entropy cathodes for Li-ion batteries, Nature Materials 20, 214 (2021).
  13. G. Ceder, A derivation of the ising model for the computation of phase diagrams, Computational Materials Science 1, 144–150 (1993).
  14. I.-H. Jung, S. A. Decterov, and A. D. Pelton, Critical thermodynamic evaluation and optimization of the cao–mgo–sio2 system, Journal of the European Ceramic Society 25, 313–333 (2005).
  15. A. v. d. Walle and G. Ceder, The effect of lattice vibrations on substitutional alloy thermodynamics, Reviews of Modern Physics 74, 11–45 (2002).
  16. A. Togo and I. Tanaka, First principles phonon calculations in materials science, Scripta Materialia 108, 1–5 (2015).
  17. K. Parlinski, Z. Q. Li, and Y. Kawazoe, First-Principles Determination of the Soft Mode in Cubic ZrO2, Physical Review Letters 78, 4063–4066 (1997).
  18. A. Togo, First-principles Phonon Calculations with Phonopy and Phono3py, Journal of the Physical Society of Japan 92, 012001.
  19. Z. Deng, B. Radhakrishnan, and S. P. Ong, Rational Composition Optimization of the Lithium-Rich Li3OCl1–x Br x Anti-Perovskite Superionic Conductors, Chemistry of Materials 27, 3749–3755 (2015).
  20. J.-X. Shen, M. Horton, and K. A. Persson, A charge-density-based general cation insertion algorithm for generating new li-ion cathode materials, npj Computational Materials 6, 161 (2020).
  21. H.-C. Wang, S. Botti, and M. A. L. Marques, Predicting stable crystalline compounds using chemical similarity, npj Computational Materials 7, 12 (2021).
  22. C. J. Bartel, Data-centric approach to improve machine learning models for inorganic materials, Patterns 2, 100382 (2021).
  23. T. W. Ko and S. P. Ong, Recent advances and outstanding challenges for machine learning interatomic potentials, Nature Computational Science , 1–3 (2023).
  24. W. Sun and G. Ceder, Efficient creation and convergence of surface slabs, Surface Science 617, 53–59 (2013).
  25. G. Henkelman, B. P. Uberuaga, and H. Jónsson, A climbing image nudged elastic band method for finding saddle points and minimum energy paths, The Journal of Chemical Physics 113, 9901–9904 (2000).
  26. D. P. Kingma and J. Ba, Adam: A method for stochastic optimization, arXiv 10.48550/arxiv.1412.6980 (2014).
  27. G. Kresse and J. Furthmüller, Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set, Computational Materials Science 6, 15 (1996).
  28. G. Kresse and D. Joubert, From ultrasoft pseudopotentials to the projector augmented-wave method, Physical Review B 59, 1758 (1999).
User Edit Pencil Streamline Icon: https://streamlinehq.com
Authors (9)
  1. Bowen Deng (30 papers)
  2. Yunyeong Choi (2 papers)
  3. Peichen Zhong (16 papers)
  4. Janosh Riebesell (10 papers)
  5. Shashwat Anand (7 papers)
  6. Zhuohan Li (29 papers)
  7. KyuJung Jun (8 papers)
  8. Kristin A. Persson (49 papers)
  9. Gerbrand Ceder (72 papers)
Citations (10)
View on Semantic Scholar

Summary

Overcoming Systematic Softening in Machine Learning Interatomic Potentials

Introduction

In the field of materials science, creating accurate simulations at the atomic level is a complex but essential task. Recently, Machine Learning Interatomic Potentials (MLIPs) have gained momentum as a means to bridge the gap between precise quantum mechanical simulations and feasible computational costs. However, despite advancements, a critical challenge remains: how well these models extrapolate to complex atomic environments that were not part of their training data.

This paper evaluates how three specific universal MLIPs (uMLIPs)—M3GNet, CHGNet, and MACE-MP-0—perform on challenging cases, what systematic errors are common, and how simple fine-tuning methods can significantly improve their predictions.

MLIP Framework and Common Challenges

First, let's unpack how these MLIPs work. The total energy EE of a system in MLIPs is approximated as a sum of atomic contributions. Forces on each atom are derived as gradients of this energy:

1
2

E = \sum_i^n E_i(\{\vec{r}_i\}, \{C_i\}),
f_i = - \frac{\partial E}{\partial \vec{r}_i}
Here, each EiE_i is a function that predicts the energy contribution of atom ii based on its position ({r⃗i}\{\vec{r}_i\}) and chemical identity ({Ci}\{C_i\}).

However, the Potential Energy Surface (PES) shows systematic "softening"—basically, the models underpredict the energies and forces in high-energy configurations. This softening is traceable to biased training datasets, which mostly include near-equilibrium atomic arrangements, limiting how well MLIPs can generalize to unseen environments.

Results Across Different Modeling Tasks

The authors extensively tested the uMLIPs on several tasks:

Surface Energies

For predicting surface energies—a vital factor in applications like catalysis and thin film growth—the uMLIPs consistently underestimated the surface energies compared to Density Functional Theory (DFT) calculations.

Defect Energies

Similarly, for modeling point defects, which impact a material's properties significantly, uMLIPs again tended to underpredict the defect energies compared to DFT norms.

Solid-Solution Energetics

Evaluating the mixing energetics in systems like the Cax_xMg2−x_{2-x}O2_2 solid solution, uMLIPs underpredicted the interaction energies, leading to inaccuracies in modeling phases and solubility behaviors.

Phonon Properties

Phonon frequencies, which dictate vibrational modes in materials, were another area of underprediction, especially for the highest-energy vibrational modes, suggesting a softer PES in high-energy regions.

Ion Migration Barriers

Ion migration, critical for understanding ionic conductors and battery materials, showed the same softening trend, with barriers systematically underestimated by all models.

Quantifying the Softening

To quantify the degree of PES softening, the authors examined the relationship between predicted forces and corresponding DFT calculations. They introduced a "softening scale," derived from the slope of the force prediction plot. A scale of less than 1 indicates softening.

Notably, over 90% of 1,000 compounds tested exhibited this softening across all three uMLIPs.

Data-Efficient Fine-Tuning

Here's the good news: a significant portion of these errors can be corrected with minimal additional training data. By fine-tuning the models with even a single new data point in high-energy regions, the softening can be mitigated effectively.

For instance, fine-tuning the CHGNet model on a single high-energy data point rotated the force prediction distribution back towards the diagonal, vastly improving prediction accuracy. Extending fine-tuning to ten new data points resulted in even further error reduction.

Implications and Future Directions

The paper underscores two major points:

  1. Systematic Softening: The PES softening is a prevalent issue across current uMLIPs, affecting predictions on various complex tasks.
  2. Fine-Tuning: Addressing these errors is feasible and efficient, with even minimal additional data points significantly improving model performance.

Future efforts should focus on generating more comprehensive datasets that cover a broader range of atomic configurations, especially high-energy states. Moreover, scaling up the model size could help mitigate the softening effect, as larger models like MACE tended to perform better.

In conclusion, while current uMLIPs exhibit PES softening limiting their accuracy in high-energy atomic configurations, this is not an insurmountable obstacle. Data-efficient fine-tuning presents a practical solution, making these models robust tools for atomistic simulations and materials discovery.

File Document Download Save Streamline Icon: https://streamlinehq.com PDF File Document Download Save Streamline Icon: https://streamlinehq.com Markdown