Dynamic Stabilization of Ultrafine Pt Nanoparticles against Sintering: Insights from Machine Learning

Sintering of metals is a primary cause of catalyst deactivation. This work presents a dynamic confinement strategy to stabilize ultrafine Pt nanoparticles (<3 nm) on porous Fe₂O₃, achieving sinter-resistance up to 850 °C. In situ transmission electron microscopy and artificial neural network (ANN) analyses confirm that the nanoparticles retain mobility and catalytic activity while resisting coalescence and ripening. The ANN model quantitatively predicts size evolution across different surface environments. During exothermic reactions, localized heating can accelerate sintering, yet this system exhibits complete CO conversion at 150 °C even after aging at 500 °C, maintaining activity for over 600 h. Stability tests at elevated space velocities further demonstrate robustness. This work provides fundamental insights into antisintering mechanisms and guides the design of thermally stable nanocatalysts.