Toward Predicting Nucleation Difficulty of Organic Compounds Crystallized from Different Solvents

Abstract

A relationship between the nucleation difficulty and solvent–solute interaction strength has been reported in recent years for various organic molecules, strongly suggesting that the desolvation of a solute molecule is one of the key factors governing crystal nucleation kinetics. We demonstrate for the first time that activity coefficients (computed with a COSMO-RS thermodynamic model) can be used to estimate the strength of the solvent–solute interaction in solution to serve as in silico descriptors, enabling prediction of the relative nucleation ease/difficulty of a solute crystallized out of a set of different solvents. In this work, we provide an experimental validation of the above hypothesis. The validation study encompasses drug-like organic solute molecules of different polarities: menadione, propyl paraben, and paracetamol, along with solvents typically used in pharmaceutical process development. In line with our hypothesis, the experimental order of nucleation difficulty established using two orthogonal methods, induction time experiments and metastable zone width determination, was found to correlate with the order of activity coefficients calculated for the respective systems. In contrast to current computational approaches that require both specialized knowledge and supercomputing capabilities, the proposed predictive model can be easily employed by crystallization practitioners in industry to quickly assess a broad range of solvent possibilities to select the most suitable solvent candidates for experimental interrogation.