Zeolites are a class of regularly ordered, microporous crystalline aluminosilicates. Due to their high thermal stability and atomic-scale pores, they are ideal candidates for selective molecular separations, when used as the active layer in membranes. Atomistically-detailed simulations, such as molecular dynamics (MD), provide a first-principles model due to their high level of detail, but are computationally expensive. In contrast, models using a coarse-grained description of the molecules and zeolite pore topology, such as Dynamic Monte Carlo (DMC), can be used to study long-time diffusion and provide crucial insight in an efficient manner, especially for heterogeneous systems.

In the DMC modeling approach, molecules make random hops from one adsorption site to a neighboring adsorption site in a lattice representation of the porous topology. Hence one can study the role of static heterogeneity in the zeolite, which is the presence of adsorption sites with either short or long residence times, the latter resulting, e.g., from the presence of Al. Studies of the effect of the Si/Al ratio on diffusion are difficult to carry out with MD simulations.

In this study, we considered the diffusion of binary mixtures in ZSM-5 lattices corresponding to different Si/Al ratios. We compared the DMC derived self-diffusivities to those obtained using the Maxwell-Stefan (M-S) approach, which is frequently used in the zeolite research community. We found that the M-S approach worked well for homogeneous systems (with only weak residence times) but not always so well for systems of high static heterogeneity. The M-S approach does not explicitly account for the lattice structure. Correction factors do not completely remove this limitation. This discrepancy of the M-S approach motivated us to consider lattice-based theories, such as the Effective Medium Approximation and the Mean Field Theory, which show more consistent agreement with the DMC self-diffusivities.