Anomalous diffusion in zeolites

Abstract Anomalous diffusion plays an important role in many pivotal chemical engineering processes involving zeolites. However, the structure–property relationships of anomalous diffusion remain unclear, and fast prediction of anomalous diffusion properties is still challenging. Herein, the anomalous diffusion behaviors of light alkanes (methane, ethane and propane) in zeolites are investigated by combining molecular dynamics (MD) simulations with machine learning (ML) method. The Gradient Boosted Regression Trees (GBRT) algorithm is utilized to construct the structure–property relationship from 2200 groups of anomalous diffusion exponent α and anomalous diffusion coefficient D α calculated by MD simulations. Furthermore, the structural parameters are ranked in order of importance and it is identified that the largest free sphere is the key factor governing anomalous diffusion phenomena. Finally, the method is employed to predict the diffusion behaviors of 200,000 hypothetical zeolites, which provides in-depth understanding of the anomalous diffusion trends in porous materials.