Machine-Learning Enabled Discovery of Advanced Low-Temperature Proton Conducting Perovskites Oxides

Hydrogen technologies need solid electrolytes that conduct protons efficiently at 300-400 degrees Celsius, resist carbon dioxide and steam, and show minimal electronic leakage. This project will discover and validate such perovskite oxides and demonstrate device-level performance. Objectives: (1) identify electrolyte compositions that achieve at least 10 mS cm-1 proton conductivity at 300-400 degrees Celsius, with stability for at least 100 hours in humid carbon dioxide and low electronic leakage; (2) prove performance in button-cell proton-conducting ceramic fuel cells with reproducible area-specific resistance and peak power; (3) release open, reusable datasets, analysis code, and standard operating procedures. Approach: a physics-informed machine learning model estimates mobile-proton population and hydration thermodynamics and maps them to proton conductivity; a parallel branch predicts electronic conduction to cap leakage. Active learning and a planned design-of-experiments schedule propose synthesizable batches. A robotic slurry-to-pellet line produces and gates single-phase materials by X-ray diffraction. Condition-matched measurements include impedance with hydrogen versus deuterium checks, thermogravimetry with van t Hoff fits, oxygen-pressure sweeps for leakage, and stability tests in humid carbon dioxide. Device tests verify ohmic consistency and reproducibility. Relevance: the work advances clean-energy materials while delivering excellent training in data-driven materials discovery, electrochemistry, open science, and research management, aligned with the MSCA Postdoctoral Fellowships work programme.