Molecular simulation of multicomponent adsorption and diffusion in zeolites

Adsorption and diffusion of guest molecules in zeolite channels was studied with molecular simulations. Using improved algorithms we have calculated adsorption equilibria and diffusion coefficients at conditions that could not be studied previously.; Two new algorithms, H-GCMG and MD-GCMC, are introduced here to study adsorption of flexible molecules. In these algorithms, molecular dynamics (MD) is used as a Monte Carlo perturbation move which explores the intramolecular degrees of freedom of the adsorbed molecules. Simulations with these algorithms were used to demonstrate that model chiral molecules show enantioselectivity in chiral pores of zeolite-like materials.; Various biasing techniques were developed and implemented to improve the convergence of grand canonical Monte Carlo (GCMC) simulations when the zeolite pores are saturated with sorbate molecules. Cavity biasing, orientational biasing, configurational biasing, and identity-swapping were used in the study of alkane mixtures, aromatic mixtures, and hexene and various solvents in zeolite silicalite. Liquid-phase adsorption equilibria of these systems were calculated for the first time with reasonable computational time. For the case of alkane mixtures and aromatic mixtures, very good quantitative agreement was observed between the experimental data and our simulations using force fields originally developed for single-component gas-phase adsorption.; Nonequilibrium MD (NEMD), where an external force is applied on the particles, was tested for multicomponent diffusion in silicalite. The error bars on diffusion coefficients obtained from NEMD were smaller compared to the error bars from equilibrium MD (EMD) simulations. NEMD was used to calculate binary diffusion coefficients for mixtures of methane, ethane, propane, butane and CF4 in zeolite faujasite. NEMD was able to give converged values of diffusion coefficients that could not be calculated using EMD. The validity of a scheme based on the Maxwell-Stefan (MS) formulation of diffusion was tested for predicting multicomponent diffusion based on single component diffusion data of the individual components. The MS estimation scheme was found to work accurately for the systems studied here.