Artificial Neural Network Potential for Encapsulated Platinum Clusters in MOF-808
Yangyang Yu, Weiwei Zhang, Donghai Mei
Abstract
Metal-organic frameworks (MOFs) have been recognized as one of the ideal supporting sintering-resistant catalyst materials because of their high specific surface area and unique nano-porous structure capable of manipulating the sizes and shapes of encapsulated metal clusters. To explore the binding sites, the stability, and migration mechanisms of encapsulated metal clusters in MOF materials, a robust potential model that accurately describes the interaction between metal clusters and MOF materials is highly desired for large-scale atomic simulations. Herein, as a demonstration case, an artificial neural network potential for encapsulated platinum (Pt) clusters in MOF-808 was developed using the machine learning-based global neural network (G-NN) technique. The artificial G-NN potential was tested and validated against a series of density functional theory calculation data, including structure optimization, adsorption energies, and the migration energy barrier of Ptn (n = 1–13) clusters in MOF-808. The newly developed Pt-MOF G-NN potential was further used to predict the adsorption and migration behaviors of Ptn clusters in MOF-808. It is found that the most stable adsorption site varies with the Ptn cluster size. The migration possibility of the Ptn cluster is strongly correlated with the adsorption energies of the Ptn clusters. Finally, the CO adsorption on the single Pt atom would effectively promote the aggregation of Ptn clusters via the Ostwald ripening mechanism.