(203ag) Describing Crystal Morphology By 2D/3D Molecular Descriptors

Authors: 
Eden, M. R., Auburn University
Haser, J. C., Auburn University
Herring, R. III, Auburn University



Crystallization is one of the least-studied separation techniques compared to distillation and extraction processes, especially in computer modeling and prediction. There are many more variables in determining final quality (crystal morphology, crystal purity, potential recovery) in crystallization processes than distillation and extraction [1]. Crystallization process products are typically produced in small quantities and their properties can have a large effect on the downstream processes. Therefore it is important that crystals meet strict guidelines in final quality.

Studies have been done to predict ibuprofen crystal morphologies using linear models based on hydrogen bonding properties of several different pure solvents [2]. The goal of this work is to develop a predictive model of crystal morphology based on solvent choice. The model will be developed using Principal Component Analysis and multivariable linear regression. Using a combination of 2D & 3D molecular descriptors is ideal for crystallization solvent design because the structure of the solvent is critical to the resulting morphology of the crystal [2]. With this model, the relationship between solvent molecular geometry and crystal aspect ratio can be determined. The first work will be focused on the aspect ratio of ibuprofen crystals and then further expanded to account for any solute molecule. It is advantageous to be able to determine what solvent characterization will produce crystals of the desired quality, and this method will allow solvents to be chosen without using time- and money-consuming experimental measurements.

[1] Karunanithi A.T., Achenie, L.E.K, and Gani, R. (2006) A computer-aided molecular design framework for crystallization solvent design. Chem. Eng. Sci., 61, pp. 1247–1260.

[2] Acquah, C., Karunanithi, A.T., Cagnetta, M., Achenie, L.E.K., and Suib, S.L. (2009) Linear models for prediction of ibuprofen crystal morphology based on hydrogen bonding propensities. Fluid Phase Equilibria, 277, pp. 73-80.