Experimental and model-based approach to evaluate solvent effects on the solubility of the pharmaceutical artemisinin

The separation and purification of plant-based Active Pharmaceutical Ingredients (API) from extracts is a crucial part in pharmaceutical process development. For the purification of the antimalarial drug component artemisinin (ARTE) from an Artemisia anna L. toluene extract, antisolvent crystallization is considered. Solubilities of ARTE in binary solvent mixtures of toluene and two potential antisolvents, n-heptane and ethanol, were determined at temperatures from 278.15 K to 313.15 K. The experimental work was supported by the application of various models, utilizing varying amounts of experimental input data. The goal was the identification of models that are able to predict solubilities in binary solvent mixtures sufficiently accurate and, thus, can help to reduce the experimental effort for future solvent screenings. In this study, we applied the PC-SAFT model both with and without fitting the binary interaction parameter k_ij between ARTE and the respective solvent, as well as the empirical Jouyban-Acree model. From the experiments, n-heptane demonstrated to be a promising antisolvent, while ethanol acted more as a cosolvent. All models tested were capable of distinguishing between effective and ineffective antisolvents. The purely predictive PC-SAFT model applied with k_ij = 0 exhibited the largest deviation from the experimental data. This was followed by the PC-SAFT model including fitted k_ij values, based on at least four experimental data points. The Jouyban-Acree model fitted the data most accurately. Its parametrization required a minimum of ten experimental data points.