Accelerating renewable ammonia production via machine learning-driven photocatalysis

Ammonia (NH3) produced from renewable sources is a promising carbon-free fuel and H2 carrier. However, photocatalytic NH3 synthesis via nitrogen reduction remains inefficient due to poor N2 activation and charge carrier recombination. These challenges can be effectively addressed through advanced data science–based modeling and optimization strategies. In this study, a machine learning (ML) framework was developed to predict and optimize NH3 yields based on catalyst properties and reaction conditions. Among them, the Hist Gradient Boosting model achieved the best performance (R2 = 0.90). Validation with two independent experimental datasets confirmed robust generalization (R2 > 0.78). SHAP analysis revealed that light irradiation time, light wavelength lower band, and catalyst bandgap energy were the most influential factors. Using multi-objective optimization, NH3 yields were enhanced to 2221.45 μmol/gcat while minimizing catalyst loading and irradiation time. TiO2-based photocatalysts exhibited superior performance due to improved charge separation and N2 activation. This data-driven approach offers a scalable pathway for efficient NH3 production, supporting sustainable hydrogen storage and green chemical manufacturing.