To elucidate the kinetics underlying concomitant polymorphism, anti-solvent precipitation experiments were performed with L-histidine and anthranilic acid. Mixing effects may be neglected when the induction time is much larger than the mixing time. It was found that the polymorphic fraction of the crystalline product can be manipulated by altering the solvent composition, which is related to the interfacial energy, and the supersaturation.
Further experimental studies focused on simultaneous formation of two polymorphs of L-glutamic acid in pH-shift precipitation. It could be demonstrated that at moderate to low supersaturation the alpha and the beta polymorph nucleate simultaneously but through different mechanisms. At very high supersaturation the beta-polymorph forms initially by (heterogeneous) primary nucleation and appears to be preceded by liquid-liquid demixing or amorphous solidification. It has been proven that transformation of the metastable alpha to the stable beta polymorph does not play a role.
A new modeling & simulation methodology has been developed which eventually will allow for quantitative prediction of the polymorphic fractions, the crystal size distribution and the shape of the different polymorphs. The method can be applied to determine operational conditions at which concomitant polymorphism may occur but it can also be adopted for the screening of polymorphs.
With the help of 3D kinetic Monte Carlo simulations, following the growth probability method developed by Ter Horst and Kashchiev (J. Chem Phys, 2003), the size of the nucleus and the interfacial energy of the known polymorphs can be determined as a function of the supersaturation. Using 2D simulations, 2D nucleus size and step energy can be obtained for different crystal faces of each polymorph. This leads to the growth rates of the faces, assuming growth takes place through the birth and spread mechanism. The thus obtained kinetic data are implemented in dynamic multi-dimensional population balances which can be solved using our CrysCODE model-framework. Simulation results are available for several small organic compounds including L-glutamic acid and for this component the simulation results agree qualitatively well and provide explanations for the experimental observations.