Crystallization solvent design based on a new quantitative prediction model of crystal morphology

Solution crystallization is an important separation unit operation in active pharmaceutical ingredient production. Solvent is one of the important factors affecting crystal morphology. How to select/design suitable solvents is still one of the most urgent problems in the crystallization field. In this article, a framework for crystallization solvent design based on the developed quantitative prediction model of crystal morphology is proposed. First, molecular dynamics is used to predict the crystal morphology in solvents. Next, solvent descriptors are selected by stepwise regression method. Then, the quantitative relationship between crystal aspect ratio and solvent descriptors is developed. Subsequently, computer-aided molecular design method is integrated with the developed quantitative prediction model. The crystallization solvent design problem is expressed as a mixed-integer nonlinear programming model with maximum yield as the objective function, which is solved by decomposition algorithm. Finally, the crystallization solvent design framework is applied to two cases and experimental verification is implemented.

Automated summary

This article proposes a crystallization solvent design framework based on a developed quantitative prediction model of crystal morphology. Molecular dynamics is utilized to predict crystal morphology in solvents and solvent descriptors are selected through stepwise regression. A quantitative relationship between crystal aspect ratio and solvent descriptors is established, and a mixed-integer nonlinear programming model with maximum yield as the objective function is used to solve the crystallization solvent design problem.