(371d) Modeling Bioethanol Enrichment Using Hydrophobic and Hydrophilic Zeolite Membranes

Mittal, N. - Presenter, Exxonmobil Chemical Co
Tsapatsis, M., University of Minnesota
Bai, P., University of Minnesota
Siepmann, J., University of Minnesota-Twin Cities
Daoutidis, P., University of Minnesota-Twin Cities
Zeolite membrane separation is considered to be a promising alternative to the traditional energy-intensive industrial separation techniques such as distillation. We describe the development of a detailed mathematical model based on the real adsorption solution theory and the Maxwell-Stefan formulation for transport to describe permeation through zeolite membranes. We also discuss the application of the model for bioethanol (obtained at ~ 5 wt. % ethanol) enrichment using hydrophobic all-silica MFI and hydrophilic NaA zeolite membranes.

Although significant improvements have been realized in the separation performance of zeolite membranes for water-ethanol separation, rigorous models, which can accurately predict the membrane performance are lacking. MFI zeolite membranes separate ethanol based on its greater adsorption affinity over water. However, unavoidable water co-adsorption due to hydrogen-bonding effects (i.e., water is brought in by adsorbed ethanol molecules) and due to favorable interaction of water molecules with silanol defects adversely affects the separation performance. As a result of these non-ideal intermolecular interactions, using ideal theories based on single-component adsorption and diffusion leads to over-estimation of their separation capability.

In this work, mixture adsorption is modeled using real adsorption solution theory. The co-adsorption of water molecules due to hydrogen bonding is accounted by using activity coefficients based on the two-constant Margules equation. The multi-component diffusion is modeled using the Maxwell-Stefan formulation with non-zero correlation coefficients. Furthermore, the effects due to support layer and hydrophilic defects are also incorporated. Ethanol permeance predicted by the model is found to be lower while the selectivity is found to be higher as compared to those determined in experiments. These results suggest that the zeolite layer in the simulated membrane is more permeable than the real membranes, and thus the overall permeation in the model is governed predominantly by the support layer. This discrepancy has been attributed to the reasoning that the diffusion coefficients used in the model are obtained using molecular simulations and reflect the behavior in crystals with ideal pore structure, while the presence of structural non-idealities, compressive stresses, thermodynamic and physical surface barriers in the real membranes results in reduced permeation. Hence, the diffusion coefficients in the model were reduced by a factor of 120 to obtain an experimentally validated model.

This experimentally-validated model is further used in flowsheeting thermally-integrated conceptual process designs to assess the viability of zeolite membranes over the current distillation technology for bioethanol enrichment. It is found that while hydrophobic zeolite membranes have potential for significant energy savings but lack in separation performance and hydrophilic zeolite membranes can achieve the separation target but result in no energy savings, a process with a combination of hydrophobic and hydrophilic membranes can achieve the separation target at the current performance of zeolite membranes with 10 % energy savings over distillation. However, techno-economic analysis suggests that 5- to 10-fold improvements in permeation or equivalent cost reductions are required for economic feasibility.