Improved screening of doped ZrO2 catalysts for CO2 hydrogenation from graph-based machine learning
Abstract
Dopants can tune the activity of zirconia-based CO2 hydrogenation catalysts through changes in adsorbate and transition-state (TS) energetics. Computing these quantities with density functional theory (DFT) becomes expensive when several dopants, reaction intermediates, and TSs must be screened. Here, we further develop the graph-based machine learning method Wasserstein Weisfeiler-Lehman Gaussian Process Regression (WWL-GPR): we introduce additional electronic-structure and local-environment features relevant for oxide surfaces and use the resulting model to predict adsorption and TS energies on ZrO2(101) surfaces doped with Al3+, Cd2+, Ga3+, In3+, and Zn2+. Adsorbate-surface structures are represented as attributed graphs that combine bonds connectivity and features. We benchmark WWL-GPR against thermodynamic scaling relations (TSR) for reaction intermediates and Bronsted-Evans-Polanyi relations (BEP) for TSs, and then propagate the predicted energies to microkinetic CH3OH-yield estimates. WWL-GPR consistently improves the prediction of reaction intermediate energetics relative to linear scaling approaches and provides TS estimates that are competitive with BEP-based models. These improvements become especially relevant when the predicted energetics are propagated through the reaction network. Microkinetic calculations show that WWL-GPR-derived energies give CH3OH yields closer to DFT-derived yields than the linear-relation workflow. Degree-of-rate-control analysis links the activity errors to dopant-dependent rate-controlling steps, including HCOO* formation, H2COOH* dissociation, and H3CO* formation. These results support WWL-GPR as a data-efficient surrogate for connecting atomistic energetics with catalyst-screening observables on doped oxide surfaces.
Related
- Machine Learning for High-Entropy Catalysts: Methods and Applications
- Sustainability-Aware Discovery of Dual-Atom Hydrogen Evolution Catalysts via a Unified Machine Learning Framework for Activity and Carbon Cost
- Going Beyond the d-band Center to Design Intermetallic Catalysts for Nitrogen Reduction: A High-Throughput DFT and Machine Learning Study
- Thermal properties of the ρ meson from a machine-learning-improved holographic QCD model
- Indirect prediction and screening of frustrated magnetic materials using machine learning
BibTeX
@misc{jensen2026improved,
title = {{Improved screening of doped ZrO$_2$ catalysts for CO$_2$ hydrogenation from graph-based machine learning}},
author = {Karsten Walz Jensen and Mie Andersen and Raffaele Cheula},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.13795},
eprint = {2610.13795},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.13795},
}
RIS
TY - UNPB
TI - Improved screening of doped ZrO_2 catalysts for CO_2 hydrogenation from graph-based machine learning
AU - Karsten Walz Jensen
AU - Mie Andersen
AU - Raffaele Cheula
PY - 2026
DA - 2026/10/09
PB - arXiv
DO - 10.48550/arxiv.2610.13795
UR - https://arxiv.org/abs/2610.13795
AB - Dopants can tune the activity of zirconia-based CO_2 hydrogenation catalysts through changes in adsorbate and transition-state (TS) energetics. Computing these quantities with density functional theory (DFT) becomes expensive when several dopants, reaction intermediates, and TSs must be screened. Here, we further develop the graph-based machine learning method Wasserstein Weisfeiler-Lehman Gaussian Process Regression (WWL-GPR): we introduce additional electronic-structure and local-environment features relevant for oxide surfaces and use the resulting model to predict adsorption and TS energies on ZrO_2(101) surfaces doped with Al^3+, Cd^2+, Ga^3+, In^3+, and Zn^2+. Adsorbate-surface structures are represented as attributed graphs that combine bonds connectivity and features. We benchmark WWL-GPR against thermodynamic scaling relations (TSR) for reaction intermediates and Bronsted-Evans-Polanyi relations (BEP) for TSs, and then propagate the predicted energies to microkinetic CH_3OH-yield estimates. WWL-GPR consistently improves the prediction of reaction intermediate energetics relative to linear scaling approaches and provides TS estimates that are competitive with BEP-based models. These improvements become especially relevant when the predicted energetics are propagated through the reaction network. Microkinetic calculations show that WWL-GPR-derived energies give CH_3OH yields closer to DFT-derived yields than the linear-relation workflow. Degree-of-rate-control analysis links the activity errors to dopant-dependent rate-controlling steps, including HCOO* formation, H_2COOH* dissociation, and H_3CO* formation. These results support WWL-GPR as a data-efficient surrogate for connecting atomistic energetics with catalyst-screening observables on doped oxide surfaces.
ER -