A proprietary nucleophilic aromatic substitution reaction generates numerous impurities at an intermediate step in the synthesis of an active pharmaceutical ingredient. The corresponding crystallization process must purge these impurities—most of which are structurally similar to the desired product—to isolate the product at the desired quality and yield. We find a ternary solvent system comprised of tetrahydrofuran, acetonitrile, and water effectively purges these impurities. However, the potential variability of solvent composition in-process and slow desaturation/metastability of the impurities posed a significant risk to the reproducibility of this methodology from lab to industrial scales. We address these challenges with a combination of experimental and computational methods. A distillation model was developed to evaluate various distillation procedures by comparing predicted solution compositions with experimental data. The distillation model enabled rapid and accurate modification of the solvent composition leading into the crystallization step. Subsequently, semi-empirical ternary solubility models of the desired product and the most prevalent impurity were developed from in-process samples. We mapped the crystallization process by targeting desired product/impurity forms, and balancing impurity rejection with mother liquor losses. We find these calculations are concordant with experimental results. The model development provided a reliable method for tuning solvent composition and optimizing crystallization processes for optimal product purity and yield from lab to industrial scales.
