Predicting Solvate Formation in Complex Multicomponent Solvent Mixtures from Binary Measurements and Thermodynamics Modeling

Abstract

Mapping crystallization outcomes in ternary solvent mixtures for systems able to solvate remains experimentally intensive due to the significant compositional phase space that needs to be explored to derive complete phase diagrams. Here, we show that these phase diagrams can be generated readily by combining rapid mechanochemical determination of critical solvent activities using binary solvent mixtures with thermodynamic activity mapping such as NRTL modeling. This approach enables quantitative prediction of ternary phase diagrams without direct experimental mapping of the full compositional space. Using nitrofurantoin as a model system, which forms an anhydrous phase, a monohydrate, and a dmf solvate, we determine the critical solvent activities from binary mixtures and use these together with predicted activity maps to generate ternary phase diagrams. The resulting framework accurately predicts phase boundaries and solid-form outcomes across two ternary solvent systems. Comparison with experimental data across 108 controlled solvent activity liquid assisted grinding (CSA-LAG) outcomes shows excellent agreement, with predictive accuracy exceeding 90% and correct reproduction of phase diagram topology. This strategy reduces the required experimental effort by more than 70% while providing mechanistic insight into competing solvation equilibria through solvent activity and activity-ratio landscapes. Since multicomponent solvate phase behavior is governed by solvent activity rather than composition, determination of critical activities in low dimensional space with activities modeling allow for a shift from empirical screening to thermodynamically guided prediction in crystallization design.