(443f) Statistical-Learning-Assisted First-Principles Modeling of Catalysts Sintering

Supported metal nanoparticles on oxide supports are widely applied in numerous applications. The catalyst activity is often strongly affected by the nanoparticle size. However, predicting catalyst sintering has remained elusive in part due to the inability to observe the dynamics of the catalysts operando and in part because adsorbates affect catalyst dynamics and sintering [1]. Density functional theory (DFT) is a powerful tool to predict the energetics of metal-support and adsorbate-metal interactions. However, the computational time needed to describe the long-time scales for sintering makes direct first-principles calculations impractical. To address this multiscale problem, here we introduce statistical learning tools to rapidly predict the parameters of Pd sintering supported on CeO₂(111) using input from DFT calculations. Kinetic Monte Carlo (kMC) simulations are applied to model the catalyst sintering. The methodology is applicable to any metal/support system.

[1] Wang, Y., Mei, D., Glezakou, V., Li, J. & Rousseau, R. Nat. Commun. 6, 1–8 (2015).