Green hydrogen production via ammonia electrolysis represents a promising low-energy pathway; however, its practical viability is currently constrained by sluggish ammonia oxidation reaction (AOR) kinetics, catalyst deactivation, and suboptimal mass transport. In this work, we engineered a flow-cell configured anion-exchange membrane (AEM) ammonia electrolyzer, systematically optimizing the electrode microenvironment by elevating local interfacial temperatures and increasing reactant alkalinity. These adjustments significantly enhanced reaction kinetics, enabling the system to sustain a high current density of 1.0 A cm −2 at a low cell voltage of 0.8 V. Despite these performance gains, continuous operation under such demanding conditions revealed a critical stability bottleneck: rapid catalyst poisoning limited the operational lifetime to merely 15 h under constant current electrolysis modes. To unravel the underlying deactivation mechanism, in situ Raman spectroscopy was employed, which identified the accumulation of adsorbed nitrogenous intermediates, specifically NH 2 and NH species, as the primary cause of active site blockage. Leveraging this mechanistic insight, we devised an optimized pulsed reduction strategy designed to periodically regenerate the catalyst surface by electrochemically stripping these poisoning intermediates. This dynamic modulation effectively mitigated deactivation, extending the single-run operational stability from 15 h to 65 h. By elucidating the critical role of surface intermediates and demonstrating the efficacy of microenvironment engineering coupled with dynamic potential control, this study provides a robust framework for developing durable, high-rate ammonia electrolyzers for sustainable hydrogen economies. • A membrane electrode assembly-based ammonia electrolyzer was developed. • A current density of 1.0 A cm −2 at a voltage of 0.8 V was achieved. • The key toxic intermediate in ammonia oxidation has been identified. • An optimized pulse reduction strategy was employed.
Liu et al. (2026) studied this question.