Modeling Sorption and Diffusion of Alkanes, Alkenes, and their Mixtures in Silicalite: From MD and GCMC Molecular Simulations to Artificial Neural Networks

Molecular dynamics (MD) in canonical (NVT) statistical ensemble and grand canonical-Monte Carlo (GCMC) simulations along with artificial neural network (ANN) techniques are used for the study of diffusion and sorption characteristics of small alkanes, alkenes, and their mixtures in silicalite. In particular, sorption isotherms and self-diffusion coefficients of alkanes (ethane to hexane), alkenes (ethene to hexene), and the respective alkane-alkene mixtures (consisting of the same number of carbon atoms) in silicalite are studied. The findings are directly compared with recent magic-angle spinning pulsed field-gradient nuclear magnetic resonance experimental diffusivity measurements and are in close agreement. Furthermore, new results are provided for the alkane-alkene systems. The sorption data from GCMC simulations, the self-diffusivity calculations from the MD simulations along with available experimental data are used for the development of ANN predictive modeling procedures in order to give generic sorption and diffusion predictions for pure alkanes, alkenes in different input values of fugacity, temperature, and sorbate loadings at the minimum computational resources and time. Finally, structural characteristics for pure alkane, alkenes, and alkane-alkene mixtures when confined in the silicalite framework are computed revealing sorption domains and siting preferences.

Automated summary

This study utilizes molecular dynamics, artificial neural networks, and simulation methods to investigate the diffusion and sorption characteristics of small alkanes, alkenes, and their mixtures in silicalite. Comparisons with experimental data show agreement in results, providing new insights into the alkane-alkene systems and developing predictive modeling procedures for sorption and diffusion. Structural characteristics of pure alkane, alkenes, and their mixtures in the silicalite framework are computed, revealing sorption domains and preferences.