Reports: AC9

Back to Table of Contents

43145-AC9
Molecular Simulation of Liquid-Phase Adsorption of Chain Molecules in Zeolites

Randall Q. Snurr, Northwestern University

Nanoporous adsorbents such as zeolites possess outstanding properties for use as heterogeneous catalysts or adsorbents due to their ability to discriminate between molecules based on their size and shape. One important example is zeolite LTA 5A. The pore structure of this zeolite consists of nearly spherical supercages with diameter of 1.14 nm connected via 0.5 nm windows. Due to the narrow windows, this zeolite can adsorb linear hydrocarbons such as n-alkanes while completely excluding branched hydrocarbons.

In the first year of the grant, we had studied adsorption of single-component linear alkanes from C5 to C24 in LTA-5A using atomistic Monte Carlo simulations, and we compared the results with experiments performed by our collaborators Dr. Joeri Denayer and Prof. Gino Baron in Belgium. In the second year, we have extended our study of liquid-phase adsorption to binary mixtures of n-alkanes. The simulations were performed in the grand canonical ensemble using configurational bias techniques. These simulations are very challenging because under liquid-phase conditions, the zeolite pores are almost completely full and this results in extremely low acceptance rates for insertion and deletion moves.

Our studies showed that simulation of adsorption from binary mixtures is not a trivial extension of single-component simulations. Due to the low acceptance rates for insertion and deletion moves, equilibration of the systems to the correct concentration is achieved mainly by identity change moves. In LTA-5A, the long n-alkanes chains prefer to remain coiled inside the cages and rarely straddle across adjacent cages. This results in a highly constrained system with very low diffusion rates. Under such conditions, equilibration through identity change moves only works when the molecules of both components pack inside the pores in a similar manner. For example, under liquid phase conditions, the packing of both n-hexane and n-heptane consists of three molecules per supercage. On the other hand, the packing of n-octane consists of two molecules per supercage. Hence adsorption from a liquid mixture of n-hexane and n-heptane gives satisfactory results. However, a GCMC simulation of adsorption from a liquid mixture of n-heptane and n-octane fails to achieve proper equilibration and the results differ considerably from experiments.

To overcome these difficulties, a parallel tempering scheme was implemented. Under such a scheme, a set of simulations at different temperatures are performed in parallel. At regular intervals, the configurations from simulations having adjacent temperatures are exchanged. At higher temperatures, the acceptance rates for insertion and deletion moves increase due to increased thermal energy and lower loading. The lowest temperature is the temperature of interest, and the highest temperature is chosen to be one where sufficient insertion and deletion moves are accepted. The intermediate temperatures are chosen such that the energy histogram of each simulation overlaps with those of adjacent temperatures.

Our simulations show that the parallel tempering scheme results in improved sampling of the phase space for binary mixtures where the components pack differently and consequently gives improved results. For mixtures where the components pack similarly, the parallel tempering simulation gives the same result as obtained from normal GCMC simulations. Implementing the parallel tempering scheme has allowed us to simulate adsorption in zeolite systems that would be impossible with conventional Monte Carlo methods.

We have also continued our experiments to measure liquid-phase adsorption of small organic acids onto zeolites from aqueous solution. These studies examine the feasibility of using adsorption as an alternative to distillation for recovery of organic acids from dilute aqueous solutions. The experimental data generated will also serve as a test for future molecular simulations. The organic acids investigated have alkyl groups of increasing length: formic acid, acetic acid and propionic acid. The experiments were performed in the zeolites silicalite, H-ZSM-5, Na-ZSM-5, and Li-ZSM-5 using a simple batch adsorption technique.

Our results show that the hydrophobicity of the zeolites is the main factor that determines the amount of acid that is adsorbed. Zeolites with higher Si/Al ratio are more hydrophobic and hence we see that among the zeolites used, silicalite adsorbs the maximum amount of acid from aqueous solution. Also the amount adsorbed is higher for those acids with larger alkyl groups. Thus, among the acids, formic acid is adsorbed the least and propionic acid is adsorbed the most. We also find that the nature of the cations in the zeolites has little effect on the amount of acid adsorbed. Thus H-ZSM-5, Na-ZSM-5, and Li-ZSM-5, all adsorb nearly the same amount of acid.

Back to top