Skip to content
DraconiticResearch
Search Sources API
194,161 studies31 sourcesUpdated 37 min ago
Peer-reviewedCC BY-NC-NDnpj Computational Materials

Fine-tuning universal machine learning potentials for transition state search in surface catalysis

  • Raffaele Cheula1,2
  • Mie Andersen1
  • John R. Kitchin2
  1. 1Center for Interstellar Catalysis, Department of Physics and Astronomy, Aarhus University, Aarhus C, Denmark
  2. 2Department of Chemical Engineering, Carnegie Mellon University, Pittsburgh, PA, USA
Open at sourcePDFDOICite

Abstract

Determining transition states (TSs) of surface reactions is central to understanding and designing heterogeneous catalysts but remains computationally prohibitive with density functional theory (DFT). While machine learning potentials (MLPs) offer significant speedups, task-specific models have limited transferability across catalytic systems, and universal MLPs (uMLPs) lack the accuracy needed for reactive configurations. Here, we present a workflow based on active learning to iteratively fine-tune uMLPs for DFT-quality TS search. Using 250 TSs from the CO 2 hydrogenation reaction network on metal and single-atom alloy surfaces, we first benchmark TS search algorithms, identifying the Sella algorithm as most robust, and propose a modification (Bond-Aware Sella) that substantially improves its success rate. We then explore sequential and batch active-learning strategies for fine-tuning and show that DFT-quality TS structures can be found using only 8 DFT single-point calculations on average per structure. This demonstrates the viability of fine-tuned uMLPs for high-throughput catalyst screening.

Related

  1. Can Universal Machine-Learning Interatomic Potentials Capture Small Polarons? A Geometry-Based Benchmark in Transition-Metal Oxides
    Vladislav Korostelev, Pjotrs Žguns, Konstantin Klyukin
    arXivPreprintarXiv licenceFull textPhysical chemistryComputational physicsarXiv
  2. How accurate are foundational machine learning interatomic potentials for heterogeneous catalysis?
    Luuk H. E. Kempen, Raffaele Cheula, Mie Andersen
    The Journal of Chemical PhysicsPeer-reviewedLicence not statedFull text9 citedMaterials & polymersOpenAlex
  3. Simulating enzyme catalysis with electrostatically embedded machine learning potentials
    Valentin Gradisteanu, Elliot W Chan, Lester Hedges +7
    Chemical SciencePeer-reviewedCC BY 4.0Full text15 citedMaterials & polymersHAL
  4. Correcting the Liquid Densities of Universal Machine-Learning Interatomic Potentials Using Local-Number-Density Repulsion
    Shunsuke Mieda
    ChemRxivPreprintLicence not statedGeneral chemistryChemRxiv
  5. Applicability Domain of Universal Machine Learning Interatomic Potentials at Solid-State Battery Interfaces
    Umang Agarwal, Mengke Li, Gopalakrishnan Sai Gautam +1
    ChemRxivPreprintLicence not statedGeneral chemistryChemRxiv

Cite

BibTeXRIS
BibTeX
@article{cheula2026finetuning,
  title = {{Fine-tuning universal machine learning potentials for transition state search in surface catalysis}},
  author = {Raffaele Cheula and Mie Andersen and John R. Kitchin},
  journal = {npj Computational Materials},
  year = {2026},
  month = jul,
  doi = {10.1038/s41524-026-02228-1},
  url = {https://doi.org/10.1038/s41524-026-02228-1},
  note = {CC BY-NC-ND},
}
RIS
TY  - JOUR
TI  - Fine-tuning universal machine learning potentials for transition state search in surface catalysis
AU  - Raffaele Cheula
AU  - Mie Andersen
AU  - John R. Kitchin
PY  - 2026
DA  - 2026/07/09
JO  - npj Computational Materials
DO  - 10.1038/s41524-026-02228-1
UR  - https://doi.org/10.1038/s41524-026-02228-1
AB  - Determining transition states (TSs) of surface reactions is central to understanding and designing heterogeneous catalysts but remains computationally prohibitive with density functional theory (DFT). While machine learning potentials (MLPs) offer significant speedups, task-specific models have limited transferability across catalytic systems, and universal MLPs (uMLPs) lack the accuracy needed for reactive configurations. Here, we present a workflow based on active learning to iteratively fine-tune uMLPs for DFT-quality TS search. Using 250 TSs from the CO 2 hydrogenation reaction network on metal and single-atom alloy surfaces, we first benchmark TS search algorithms, identifying the Sella algorithm as most robust, and propose a modification (Bond-Aware Sella) that substantially improves its success rate. We then explore sequential and batch active-learning strategies for fine-tuning and show that DFT-quality TS structures can be found using only 8 DFT single-point calculations on average per structure. This demonstrates the viability of fine-tuned uMLPs for high-throughput catalyst screening.
SN  - 2057-3960
ER  -