ADAMox delivered results that advance the state of the art in machine-learned modelling of dynamic metal oxide catalysts. The ATLAS framework demonstrated that carefully designed, diversity-aware training datasets can outperform substantially larger generic datasets, providing a more efficient and transferable strategy for developing accurate machine-learning potentials. This challenges the common assumption that model performance is primarily driven by database size and offers a practical methodology for the broader atomistic simulation community.
The Pt/CeO2₋ₓ study provided a new understanding of metal-support interactions by showing that polaron organisation, rather than oxygen-vacancy concentration alone, governs the structural and electronic properties of supported Pt clusters. This replaces a static descriptor with a dynamic electronic-structure perspective and establishes quantitative structure-property relationships relevant to reducible oxide catalysts.
The project also established a validated computational foundation for the study of amorphous IrO2 electrocatalysts and developed open access MLIP training software (ATLAS), including a specialised IrOx machine-learned potential, amorphous-structure generation workflows, and surface preparation methodologies. These results provide reusable tools for future investigations of disordered oxide materials and electrochemical energy-conversion catalysts.
Potential users include researchers in computational catalysis, machine-learning potentials, materials modelling, and electrocatalysis. The results have been disseminated through open-access publications, preprints, datasets, and software resources, facilitating their uptake by the scientific community.
Further uptake and exploitation will require continued research on explicit electrochemical environments, constant-potential simulations, reaction and degradation pathways, and microkinetic modelling. The methodological framework, datasets, and workflows developed in ADAMox provide the foundation for these future developments and for subsequent applications to catalyst design for sustainable energy technologies