Refinement of the theoretical solubility model and prediction of solute solubility in mixed solvent systems

- This STMIPE-AC model can distinguish between isomers.
- We have refined the non-polar and hydrogen bonding interaction term of model.
- This refined model is reliable for prediction of drug solubility in mixed solvents.

We have refined the non-polar (van der Waalls) and hydrogen bonding interaction term of STMIPE-AC (statistical thermodynamic molar intermolecular potential energy activity coefficient) model, which reduces the average error of the model in correlating the molar internal energy change on vaporization of pure liquids (62 solvents, 10 temperature points for each solvent) from 6.6% to 3.1%. The binary interaction parameters kij are fitted to the solubility data of solid solutes in pure solvents or estimated from a semi-empirical equation using a similar local composition method, and then they are used to predict the solute solubility in mixed solvents. The overall average relative error (ARE) in predicted solubility (15 solutes, 64 systems, 1091 data points) is 11.24%, which makes this model a practical approach for crystallization process development.