Highlighting the origins and consequences of thermodynamic non-idealities in mixture separations using zeolites and metal-organic frameworks

The Ideal Adsorbed Solution Theory (LAST) is widely used for the estimation of the mixture adsorption equilibrium, and for quantitative modeling of separations using microporous adsorbents and membranes. With the aid of Configurational-Bias Monte Carlo (CBMC) simulations, the accuracy of the LAST estimations of the component loadings for mixture adsorption equilibrium is investigated for a wide variety of mixtures in zeolites, and metal organic frameworks (MOFs). The LAST estimations are found to be of inadequate accuracy under two different scenarios: (1) when there is molecular clustering, caused by strong hydrogen bonding between the adsorbates, as is the case for water/alcohol, alcohol/alcohol, and alcohol/aromatic mixtures, and (2) there is inhomogeneous, segregated, distribution of adsorbates within the pore network, caused by preferential siting and locations of guest molecules. For both these scenarios, quantitative agreement with CBMC simulations of mixture adsorption is realized by application of the Real Adsorbed Solution Theory (RAST) by incorporation of activity coefficients, suitably parameterized by the Wilson model for the excess Gibbs free energy of adsorption.

The important consequences of thermodynamic non-idealities are underscored for transient operations of fixed bed adsorbers for which the LAST and RAST may anticipate opposite sequences of component breakthroughs. For water/alcohol separations in membrane pervaporation processes, the permeation selectivities predicted by the LAST and RAST may differ by an order of magnitude.