(667g) Computational Fluid Dynamics Boosted Stochastic Modelling for Integrated Quantitative Understanding of API Crystalline Product Manufacturing Process

A large number of APIs are marketed in dried crystalline form and have to comply with tight specifications on crystal morphology and crystal size distribution. Manufacture of these APIs involves several processing steps which include reaction, crystallization (anti-solvent and/or cooling), filtration and drying. Each operation significantly modifies crystal morphology. Considerable efforts are needed towards meeting the market and regulatory specifications on a commercial scale which is often a resource intensive task. However, resources expended, especially time and cost, can be greatly reduced by employing simulation tools which help in the tuning of production parameters to meet the final specification in an optimized manner.

Simple mathematical models provide information on gross behavior (e.g. rates of precipitation) but may not be effective towards generating more detailed and granular information, for example, distribution of morphology or particle size. The availability of inexpensive computing power opens up the opportunities to efficiently execute stochastic models. Such models, using “kinetic Monte-Carlo” methods, were used to simulate diffusive processes such as growth of crystal defects, dissolution of photolithography polymers, etc. A further advantage entailed by this approach lies in identifying a full crystal size distribution using a minimal set of parameters describing nucleation, growth, combination and break up.

Our approach used data from various sources: crystal size distributions from dynamic light scattering analyzer; phase diagram from gravimetric experiments; nucleation and growth kinetics from particle track technologies; and crystal breakup from shear contours of crystallizer using computational fluid dynamics. Similar models were deployed to understand crystal breakup and/or lump formation in filtration, drying and milling equipment.

This approach demonstrates an end-to-end quantitative model for API processes using quality-by-design (QbD) framework allowing engineers and scientists to develop robust API commercial manufacturing processes.