In the last decades the salting-out technique has drawn more attention and has been more frequently seen as a valid alternative to cooling and evaporation. The antisolvent addition reduces the solubility of the product compound in the original solvent thereby facilitating supersaturation generation. This method is attractive since it can lead to significant savings in energy consumption and operation costs in comparison to the conventional techniques. Furthermore it can be seen as an alternative methodology whereby the limited temperature stability of the solid product precludes the use of evaporation, as in the case of pharmaceuticals and biochemicals, or when, because of the weak temperature dependence of the solute solubility, it is not possible to use the cooling techniques, such as for sodium chloride in water.
Currently most of the industrial crystallization processes are still designed and tested by means of time consuming and expensive trial-and-error experimentations. The use of model based techniques is extremely useful in improving the implementation of these technologies and in understanding the complex mechanism involved in the process at different scale. The main control objectives in a crystallization process are usually defined in terms of crystal characteristics, such as size, morphology and purity. In the case of antisolvent crystallization, the antisolvent feedrate trajectory is crucial to control the supersaturation level and therefore to achieve the desired crystal size distribution (CSD). Thus, optimization of feed flow rate profiles is of great importance, though one of the major obstacles for designing the optimal feed profiles is the accuracy of the mathematical models.
In this work a new approach to predict solubility is applied to the antisolvent crystallization modeling. The thermodynamic model proposed by Kolker et al. [1] is used here to describe the ternary equilibrium and to calculate the supersaturation for the water-NaCl-ethanol system. In contrast to previous empirical approaches, in which experimental data need to be derived and validated, this model is more general and requires only standard thermodynamic properties for the pure components and for the solutes at infinite dilution in each solvent. The thermodynamic model is then blended into an overall mechanistic crystallization model. The complete mathematical model for the crystallization process is developed based on the conservation of mass, energy and population moments. In particular, population balance equations are numerically solved to predict end product properties expressed in terms of particles' mean size and CSD.
Alternative kinetic models for growth and nucleation are also investigated and compared. The kinetic parameter estimation is performed experimentally by means of maximum likelihood method. In this step data collected from a number of experiments under different antisolvent feeding conditions are used. Once the kinetic model is identified, model validation is carried out using experimental data obtained for new antisolvent feed profiles. The results are in accordance with the experimental data, and show that the dynamic behavior of the antisolvent crystallization process can be successfully modeled and predicted.
The overall model described here is proposed as a useful tool to understand the basic phenomena involved in the process and is used to find optimal operational procedures under different objectives. In particular, a dynamic optimization was conducted in order to find the optimal time-varying trajectory for the antisolvent flow rate. The results presented here have been obtained from different final size constraints. These results show that model-based approaches are powerful tools which can be implemented for the design of optimal operation policies in the area of particle size control.
[1] Kolker, A., De Pablo, J. Thermodynamic modeling of the Solubility of salt in mixed aqueous-organic solvents. Ind. Eng. Chem. Res. 1996, 35, 228-233