Online Classification of Mixed Co-Crystal and Solute Suspensions using Raman Spectroscopy

This work seeks to develop a calibration-free statistical approach based on Raman spectroscopy for online identification of impurities during co-crystallization. Caffeine-glutaric acid-acetonitrile was employed as the model system, which forms co-crystals from solution at appropriate conditions. Raman spectra were collected in three classes of suspensions with a solid mixture of caffeine crystals and co-crystals, pure co-crystals, and a mixture of glutaric acid crystals and co-crystals, respectively, at different temperatures. These suspensions were used to represent the possible products of co-crystallization processes during which single component could crystallize out concomitantly with the desired cocrystal. A statistical model combining principal component analysis (PCA) and discriminant analysis (DA) was developed to classify these suspensions. PCA was first performed for these spectra and the resulting first few principal components were subjected to DA. It was found that the three classes of suspensions can be distinguished clearly by DA.