Machine Learning-Derived Correlations for Scale-Up and Technology Transfer of Primary Nucleation Kinetics.

Scaling up and technology transfer of crystallization processes have been and continue to be a challenge. This is often due to the stochastic nature of primary nucleation, various scale dependencies of nucleation mechanisms, and the multitude of scale-up approaches. To better understand these dependencies, a series of isothermal induction time studies were performed across a range of vessel volumes, impeller types, and impeller speeds. From these measurements, the nucleation rate and growth time were estimated as parameters of an induction time distribution model. Then using machine learning techniques, correlations between the vessel hydrodynamic features, calculated from computational flow dynamic simulations, and nucleation kinetic parameters were analyzed. Of the 18 machine learning models trained, two models for the nucleation rate were found to have the best performance (in terms of % of predictions within experimental variance): a nonlinear random Forest model and a nonlinear gradient boosting model. For growth time, a nonlinear gradient boosting model was found to outperform the other models tested. These models were then ensembled to directly predict the probability of nucleation, at a given time, solely from hydrodynamic features with an overall root mean square error of 0.16. This work shows how machine learning approaches can be used to analyze limited datasets of induction times to provide insights into what hydrodynamic parameters should be considered in the scale-up of an unseeded crystallization process.

Use of Wet Milling Combined with Temperature Cycling to Minimize Crystal Agglomeration in a Sequential Antisolvent-Cooling Crystallization.

The objective of the research was to improve the process design of a combined antisolvent-cooling crystallization to reduce the degree of agglomeration of a real active pharmaceutical ingredient product, which was manufactured using a crystallization stage employing a methanol/water solvent system. Knowledge was gained from the use of process analytical technology (PAT) tools to monitor the process variables, allowing particle size, degree of agglomeration, solute concentration, and supersaturation to be tracked throughout the process. Based on knowledge of the solubility behavior and interpretation of the PAT histories, changes were made to the sequences of antisolvent addition and cooling within the crystallization process to reduce agglomeration in the final product. Different seed loadings and seeding addition points were also investigated to maintain operation within lower supersaturation regions of the phase diagram to limit agglomeration and avoid an undesired polymorphic transformation to an unstable form. The improved sequences of operations and seeding conditions did not provide sufficient improvement in the product quality and so were augmented by applying wet milling for further deagglomeration followed by temperature cycling to remove fine particles generated during milling. Open-loop heating and cooling cycles produced some limited improvements, whereas closed-loop direct nucleation control methods using FBRM as a feedback sensor for particle counts per second were much more successful at producing high-quality crystals of the desired polymorphic form. The work shows that understanding the trajectory of the process through the phase diagram to follow appropriate supersaturation profiles gives improved control of the various kinetic mechanisms and can be used to improve the quality of the final product.

Templated nucleation of acetaminophen on spherical excipient agglomerates.

We investigated the effect of spherical agglomeration of heterogeneous crystalline substrates on the nucleation of acetaminophen (AAP). Optical and electron microscopy showed that the surface morphologies of single crystal triclinic lactose and D-mannitol differed significantly from their counterparts formed via spherical agglomeration. Spherical agglomerates of lactose were shown to enhance the nucleation rate of acetaminophen (AAP) by a factor of 11 compared to single crystal lactose; however, no such enhancement was observed for D-mannitol. X-ray powder diffraction identified the presence of new crystal faces of lactose present only in the spherical agglomerates However, D-mannitol did not show any significant change in crystal morphology. The new crystal faces of triclinic lactose were analyzed using geometric lattice matching software and molecular dynamics simulations to establish any new and significant epitaxial matches between lactose and AAP. A coincident lattice match and a large favorable energy interaction from hydrogen bonding were observed between the (141¯) and (001) crystal faces of lactose and AAP, respectively. The enhanced nucleation kinetics, X-ray data, and computational studies indicated that the spherical crystallization of lactose exposed the (141¯) face on the surface of the agglomerates, which subsequently enhanced the nucleation rate of AAP through geometric lattice matching and molecular functionality. This study highlights the importance of exploring different heterogeneous substrate morphologies for enhancing nucleation kinetics.