Combining synchrotron-based spectroscopy with machine-learning-based data-driven approaches has the potential to accelerate catalyst discovery for fuel cells, but approaches capable of handling complex feature spaces remain underexplored. Here, the authors propose a reverse-engineering framework for electrocatalyst discovery that incorporates multimodal spectral descriptors derived from synchrotron-based X-ray spectroscopic techniques, demonstrating its application to the prediction of catalytic performance metrics such as mass and specific activity and electrochemical surface area.
- Ankur Baliyan
- Sarthak Verma
- Hideto Imai
