Unveiling the Structural Factors Governing the Diffusion of Ethene in Small-Pore Zeolites through Machine Learning

Investigating diffusion dynamics in nanoscale confined
spaces of
zeolites has long been of significant interest. Although zeolite structures
are known to significantly influence their diffusion properties, determining
how a specific structure impacts diffusion performance is challenging.
This difficulty arises from a lack of methodologies for quantifying
structural factors and linking correlated variables to the diffusion
performance. To address this challenge, this study utilizes a data-driven
approach to systematically assess diffusion behaviors in small-pore
zeolites. Machine learning models were trained and achieved high diffusion
predictive performance through newly devised descriptors. Model interpretation
and further molecular dynamics simulations highlighted the significant
role of the cage architecture in influencing diffusion. Moreover,
we discovered that small-pore zeolites with specific stacking sequences,
the parallel stacking of 6-rings, acquired enhanced diffusion properties
synergistically as the proportion of double-six-ring units increased.
This study advances the understanding of diffusion in micropores and
offers insights into zeolite design for promoting molecular transport.