Abstract
Proton exchange membrane fuel cells (PEMFCs) are central to hydrogen-based energy systems, yet their large-scale commercial deployment remains constrained by cost, durability, and multi-scale transport limitations within the membrane electrode assembly. Although previous reviews have addressed catalysts, membranes, flow fields, or modelling approaches individually, few have integrated quantitative performance metrics, physics-based models, readiness assessment, and artificial-intelligence-assisted decision frameworks across the recent 2020–2025 literature. This review addresses that gap by developing a quantitative, multi-scale framework for evaluating next-generation PEMFC materials, components, and system-level design strategies. A dedicated comparison with five representative prior reviews is provided to clarify the novelty and scope of the proposed framework. Recent advances show that nanostructured catalysts, including PtNi nanowires and PtPd nanodendrites, can achieve mass activities of approximately 0.85–1.12 A mgPt⁻¹, corresponding to a three- to five-fold improvement over commercial Pt/C catalysts, while reducing electrochemically active surface area loss to approximately 14–18% after 30,000 accelerated stress cycles. Porous metal foam flow fields demonstrate oxygen transport efficiencies approaching 0.90 and reduce pressure drop by 45–60% compared with conventional serpentine designs under representative operating conditions. Short-side-chain perfluorosulfonic acid membranes further show promising proton conductivity of approximately 0.1 S cm⁻¹ at 120 °C and 50% relative humidity. The main contributions of this review are fivefold: first, a Combined Performance Index is introduced using peak power density under standardised operating conditions of 80 °C, 100% relative humidity, H₂/air operation, and 200 kPa absolute pressure; second, a Scalability Index is proposed for industry-oriented technology prioritisation; third, a dual Technology Readiness Level and Manufacturing Readiness Level framework is developed; fourth, physics-based mathematical models are synthesised for catalyst kinetics, membrane transport, and flow-field fluid dynamics; and fifth, a physics-informed neural network digital-twin framework is proposed with defined input/output variables, embedded physical constraints, and benchmarking against conventional machine-learning methods. Overall, this review provides a quantitative and decision-oriented pathway for accelerating scalable, durable, and commercially viable PEMFC deployment.