Next Generation Models for Crystal Morphology Prediction – a Collaboration between ETC & Ucsb

Crystals almost always grow as faceted particles exposing a small number of chemically distinct faces on their surface. Since each face exposes different chemical groups, different faces interact differently with the solution environment, which is why crystals of the same solute with different morphologies behave differently (e.g., dissolve differently). Crystal shape is also important. Generally, needles and plates are problematic for processing and tableting. Therefore, crystal morphology and shape are important material characteristics and there is significant value for models that are capable of simultaneous prediction of crystal shape and morphology. Such models may be used to perform computer experiments that help guide experiments to identify preferred solvents, temperatures, supersaturations and other process conditions that lead to desired crystal forms. To be successful in a process development workflow, morphology prediction models must be fast (compute times measured in minutes not days or weeks) and yet remain faithful to the fundamental solid-state chemistry and surface physics governing the growth processes during evolution of particle morphology.

In recent years, a collaboration between our team at UCSB and the ETC Consortium has helped advance the model development to predict morphology and shape of organic crystals grown from solution. The models are based on surface integration kinetics as the rate determining step. The key features that have been or are continuing to be addressed include: (1) importance of optimization of the hydrogen positions in the crystal structure file, (2) use of the COSMO-SAC property model to estimate solvent-modified bond energies between growth units on the step edges that flow across crystal faces, (3) models for calculating the density of kink sites along step edges at high supersaturation (i.e., liberation from equilibrium Boltzmann statistics), (4) nonlinear step velocity models suitable for use at high supersaturation, (5) importance of validating the classical atom-atom force field from among those available (e.g., GAFF, CLP, Lifson, etc.). These advances either have been or are currently being incorporated into our digital design aid ADDICT. Sponsor testing and feedback has been a valuable source of inspiration.