Simulated moving bed starting conditions using an empirical model for S-shaped adsorption isotherms.

Simulated moving bed (SMB) chromatography is a complicated process that generally requires modeling to determine starting conditions. Typically one uses the triangle theory to arrive at these starting conditions. The most common adsorption isotherm model used to construct the triangle is the Langmuir isotherm or an isotherm derived from the Langmuir isotherm. Often, modeling software supplied by the SMB manufacturer is used to determine the Langmuirian isotherm parameters. This proprietary approach, while successful in most cases, gives inaccurate results when the adsorption of one of the components is dominated by an S-shaped isotherm. Such failures require lengthy and expensive trial-and-error procedures to optimize the separation. In this paper, we apply an empirical model for S-shaped isotherms to the problem. With this isotherm model a triangle was constructed using the equilibrium dispersive model to simulate the SMB process. The starting conditions predicted by this approach were far more accurate than those determined by the proprietary approach.

Resolution of a racemic pharmaceutical intermediate. A comparison of preparative HPLC, steady state recycling, and simulated moving bed.

Preparative HPLC and simulated moving bed (SMB) chromatography were used to resolve significant quantities of a racemic pharmaceutical intermediate. In addition, smaller scale studies using closed-loop recycling and steady state recycling (SSR) were performed so that a meaningful comparison of all these techniques could be made using the same real world separation. A highly optimized, six-column SMB process was clearly the superior technique and was used for the process-scale (247 kg of racemate) resolution. At the more moderate lab-scale (33 kg of racemate and 19 kg of racemate), a frequently used but less optimized eight-column SMB process was used. It was found that SSR was comparable to the lab-scale SMB process in productivity and solvent consumption. Thus, it appears that SSR can be a useful choice at such moderate scales. Finally, at moderate scales when neither SSR nor SMB is available, it was found that acceptable results were obtained with both closed-loop recycling and with a two-step preparative process.