A computer simulation based screening method for crystallization processes

Crystallization of organic solids from solutions is often a difficult task because phase transitions, formation of solvates or complexation could occur. Therefore it is important to find out solvent properties which have a positive influence on the crystallization of desired polymorphs. As it is material and time consuming to determine solvent properties by experimental methods we used a SGI Octane workstation equipped with the program Cerius 4.2 (Accelrys) for calculating solvent by using Force Field (cff 91_950_1.01) and semiempirical methods (WinMopac). Moreover, a screening plan for the crystallization of nitramines was developed on the basis of 40 solvents, representing twelve different functional groups and the four influence factors: dipole moment, nonbond energy, molar volume and solvent polarity. It's known that the electrostatic potential of a molecular surface is related to solvent properties like the H-bond donating parameter alpha and the H-bond acceptor parameter beta . It seems therefore necessary to include the electrostatic potential as an important influence factor. Crystallization experiments were carried out with epsilon-CL-20 in a automated crystallization device Quest from Argonaut. The recrystallized samples were analyzed by X-ray diffraction (XRD), REM and DSC to characterize occurring polymorphs, donor-acceptor complexes and morphologies of the crystals. We found high correlations between polymorphic phases and solvent properties. Furthermore models and concepts for solution mechanism, complexation or decompositions of CL-20 were developed. The results and systematic evaluation allowed us to predict further suitable solvents for the crystallization of epsilon-CL-20.