Accelerated Discovery of Non-Stoichiometric Strontium Ferrite (SrFeO3-δ) Cathodes for Intermediate-Temperature Solid Oxide Fuel Cells via an Autonomous Materials AI Orchestrator

Developing highly efficient intermediate-temperature solid oxide fuel cells (IT-SOFCs) requires cathode materials with exceptional mixed ionic-electronic conductivity (MIEC). Strontium ferrite SrFeO3-x is a promising candidate, but its performance depends heavily on the precise control of non-stoichiometric oxygen vacancies (x), creating an immense configurational space that challenges traditional trial-and-error discovery. Here, we present an autonomous Materials AI Orchestrator that integrates machine learning screening, high-throughput density functional theory (DFT+U), and automated synthesis loops. The AI agent generated and screened over 50,000 defect configurations, identifying an optimal vacancy ordering phase (SrFeO2.75) with a predicted ionic migration barrier of <0.45 eV. Autonomous robotic synthesis and characterisation verified an ionic conductivity of 0.12 S/cm at 600°C. This active learning framework demonstrates a 15-fold acceleration in materials optimization compared to traditional methods, offering a paradigm shift for non-stoichiometric materials design.