Data for "How accurate are foundational machine learning interatomic potentials for heterogeneous catalysis?"

Dataset containing all data created for the paper "How accurate are foundational machine learning interatomic potentials for heterogeneous catalysis?". This dataset was created with and can be analyzed with the MLIP Cats package, which can be found on GitLab.

The structural data in the ASE databases in the datasets folder was originally generated in the following publications:

Taylor, N.T., Morgan, M.T., Hepplestone, S.P. Instability of ABO₃ perovskite surfaces induced by vacancy formation (A = Ca, Sr, Ba; B = Ti, Zr, Sn). Physical Review B 112, 045308 (2025)

Cheula, R., Tran, T.A.M.Q., Andersen, M. Unraveling the Effect of Dopants in Zirconia-Based Catalysts for CO₂ Hydrogenation to Methanol. ACS Catalysis 14, 17, 13126–13135 (2024)

Cheula, R., Andersen, M. Transition States Energies from Machine Learning: An Application to Reverse Water–Gas Shift on Single-Atom Alloys. ACS Catalysis 15, 13, 11377–11388 (2025)

Kempen, L.H.E., Andersen, M. Inverse catalysts: tuning the composition and structure of oxide clusters through the metal support. npj Computational Materials 11, 8 (2025)

Nielsen, M.J., et al. Interpretable machine learned predictions of adsorption energies at the metal–oxide interface. The Journal of Chemical Physics 163, 4, 044708 (2025)

Kempen, L.H.E., Nielsen, M.J., Andersen, M. Breaking Scaling Relations with Inverse Catalysts: A Machine Learning Exploration of Trends in CO₂-to-Formate Energy Barriers. ACS Catalysis 15, 21, 17635–17644 (2025)