376470 The Use of CFD Modeling to Understand Variability in Particle Agglomeration in an API Crystallization

Wednesday, November 19, 2014: 8:30 AM
313 (Hilton Atlanta)
Chester E. Markwalter1, Brenda Remy1, Jason Sweeney1, Shawn Pack1, Tarang Bulchandani2, Gopal Kasat2 and Damodaran Vedapuri2, (1)Chemical Development, Bristol-Myers Squibb Co., New Brunswick, NJ, (2)Tridiagonal Solutions Inc., San Antonio, TX

Powder property control for Active Pharmaceutical Ingredients (API) is important to ensure efficacy and to enable consistent manufacturing of the drug product.  These properties can be controlled through crystallization design, targeting conditions that favor particle growth relative to nucleation and agglomeration.  During API crystallization development, the process parameters critical to ensure particle growth are identified, addressing the challenges in scaling from the laboratory to the plant.  Here, an API crystallization is discussed where no single mode – nucleation, growth, or agglomeration – is dominant.  Initial process scale-up identified a mixing dependence resulting in variable agglomerate formation.  Standard mixing correlations (tip speed, energy input) did not provide predictive agglomerate control. 

These scale-up results prompted the development team to seek a more predictive mixing model to provide greater insight into agglomerate formation and control.  MixIT, a software developed by Tridiagonal Solutions, was used to perform computational fluid dynamics (CFD) analysis of mixing conditions across the pilot scale reactors.  For each reactor under consideration, a mesh was generated and the 3D, steady-state Navier-Stokes equations were iteratively solved.  As an output of the analysis, important mixing parameters such as tip speed, average strain rate, and impeller power number were predicted.  Further, distribution curves (e.g. volume percent vs. strain rate) provided a richer understanding of the strain rate experienced by particles in different pilot scale reactors.  These curves provided insight into the different mixing conditions which may influence agglomerate formation.  The mixing simulations were then applied to the proposed manufacturing scale crystallizer and process parameters were selected to match the desired strain rate distribution.    The resulting batches demonstrated consistent powder property control in the larger scale API crystallizer.

Extended Abstract: File Not Uploaded