- 1:30 PM

Comparison of the Use of Different Solubility Models for the Cooling Crystallization of Acetaminophen

David Widenski, Cain Department of Chemical Engineering, Louisiana State University, 110 Chemical Engineering, South Stadium Road, Baton Rouge, LA 70803, Ali Abbas, School of Chemical and Biomolecular Engineering, The University of Sydney, Sydney, NSW 2006, Australia, Sydney, Australia, and Jose A. Romagnoli, Chemical Engineering, Louisiana State University, Cain Department of Chemical Engineering, South Stadium Road, Baton Rouge, LA 70803.

The production of pharmaceuticals is a multibillion dollar industry. These pharmaceuticals are primarily produced and purified by crystallization. Optimal performance of these pharmaceuticals is often accompanied with a particular crystal size and shape. The size and shape of these crystals are directly affected by the supersaturation conditions present in the solution. These conditions affect both nucleation and growth of the crystals which will affect the final crystal size distribution (CSD). In order to obtain the desired CSD, it is important to control both nucleation and growth at specified values by controlling the degree of supersaturation. In order to control supersaturation, a solubility model for the solute's equilibrium solubility in the solvent must be known. Currently, the most commonly used solubility models are either empirical or correlative thermodynamic models that require experimental data to generate the model. Recently, the development of predictive thermodynamic models such as MOSCED may lessen the need for experimental data in the early stages of crystallizer design.

In this paper the sensitivity of different solubility models on the predicted CSD is evaluated for the cooling crystallization of acetaminophen in ethanol. Empirical, correlative thermodynamic, such as NRTL, and predictive thermodynamic models such as MOSCED are considered. Since crystallization is driven by supersaturation, the equilibrium solubility prediction of the solubility model will determine the predicted supersaturation profile. If the supersaturation profile is predicted incorrectly then the predicted CSD will also be incorrect. After the crystallization model's solubility sensitivity is evaluated, two methods are proposed and implemented to strengthen the crystallization model's prediction against solubility model errors.