General applicability and quantitative predictability of complex population-based crystallization models

Modeling of crystallization processes offer several advantages. Accurate and predictive mathematical models have been and can be used for understanding the underlying physics, for process design, optimization, and control. Within the domain of crystallization, these models are composed of population balance equations (PBE) for the solid phase and the mass conservation constraint for the liquid phase. The former enables one to model the evolution of the particle size and shape distribution of ensembles of crystals. The latter enables one to model the evolution of the liquid phase concentration/supersaturation. Often, these models are developed in an academic setting at scales and conditions that are far from a realistic industrial scenario. Additionally, even though all the models developed in the literature account for the evolution of "size", majority of them do not account for the evolution of the shape of crystals (e.g., needles, plates, etc.). These factors can have potential implications on the general applicability and quantitative predictability of crystallization process models across different scales (lab to industrial scale).

The overarching goal of the proposed collaborative project between Bayer and University of Manchester is two-fold. These are
G1. To quantitatively evaluate the limits of the predictive capability of lab-scale models.
G2. To propose a sound experimentally validated framework that facilitates developing predictive crystallization process models applicable over different scales.

These goals will be addressed using four interconnected computational and experimental work packages (WPs). The reference case will be a non-proprietary binary system (one solute and one solvent) under pure growth conditions in lab-scale (500 mL to 1 L) at low suspension densities (0.1-1 w/w%). Subsequently, additional phenomena (dissolution, secondary nucleation), complex solvent systems, higher suspension densities (4-10 w/w%), and scales (100 mL to 20 L) will be considered.

It is anticipated that this project will help Bayer and the broader crystallization community to judge whether pursuing significant efforts in developing process models for process design and operation is worthwhile in terms of its predictability at different scales and the time and resources invested.