Utilization of OLI Thermodynamic Model in Reactive Crystallization - - PowerPoint PPT Presentation

Utilization of OLI Thermodynamic Model in Reactive Crystallization Modeling

Zhilong Zhu, You Peng, Richard D. Braatz, Allan S. Myerson Department of Chemical Engineering, MIT OLI Conference Presentation, Oct 21st, 2014

Outline

• Introduction to crystallization

– Quality attributes: CSD, purity, polymorph, yield – Modeling: population balance model and thermodynamic model

• Why OLI thermodynamic model?

– Mixed Solvent Electrolyte (MSE) model

• Application in reactive crystallization

– OLI databank construction and result – Integration with Matlab – Simulation result

Page 2

Introduction to crystallization

• Crystallization is an important separation and

purification process

• Key attributes in crystallization

– Crystal size distribution (CSD) – Crystal purity – Crystal polymorphs – Crystal yield

Page 3

( ) f x

# density Crystal size a b x

# p articles/vo l ( )

b a f

x d x  

Crystallization methods

• Driving force: supersaturation
• Methods:

– Cooling – Antisolvent addition – Evaporation – Chemical reaction

Page 4

sat sat

( ) ( ) c c T S c T  

Metastable zone

Temperature

Solubility curve Metastable limit Labile zone Nucleation Growth Undersaturated

A.S. Myerson, Handbook of Industrial Crystallization, 1993

Population Balance Model (PBM)

Conservation equation for number density, f

– spatially homogenous – one independent crystal internal coordinate, x – constant volume – batch process

Page 5

 

G f f B t x      

1st order partial integro-differential equation Nucleation rate Crystal growth rate Total mass = solute mass + crystal mass PBM:

Crystals

Thermodynamic model in crystallization

Page 6

Nucleation rate Crystal growth rate Supersaturation ~ driving force For most simple organic system Concentration Temperature

sat sat

( ) ( ) c c T S c T  

For multi-species, ionic, concentrated system

i i sp

a S k







,w h ere

i i i

a m  

Need a rigorous electrolyte model to predict i



Solubility / equilibrium

sat (

) c T

OLI Thermodynamic models

• We chose to use OLI Mixed Solvent Electrolyte

(MSE) model because

– First-principles model – Agrees better with our test cases

• The MSE models the excess Gibbs free energy

Page 7

ex ex ex ex L R S R II

G G G G   

Long Range Short Range Ionic Interaction

,

, ,

ln

j j k

ex k k T P n

G n R T 



        

Integration with Matlab via Excel

Page 8

Solution composition Speciation

Supersaturation (scaling tendency)

PBM

• de45 solver

(iterative) Matlab GUI Initial values model file from OLI Analyzer Update solution composition via Excel Macro CSD plots Excel data storage

speed: ~25 seconds for 300 OLI calls

OLI application in reactive crystallization

Page 9

OLI Analyzer # of solids: 18 # of liquids: 27 # of parameters: 38 Existing OLI databank does not have all the parameters in the MSE model for our system We carried out solubility experiments at various conditions We did regression using OLIREG for missing parameters Scaling Tendency for CaSO4.2H2O in OLI

Parity plot for CaSO4

.2H2O solubility

Page 10

0.00 0.05 0.10 0.15 0.20 0.00 0.05 0.10 0.15 0.20 Model predicted solubility (wt%) Solubility from literature data (wt%)

(1) Taperova, A. A. Zh. Prikl. Khim 18 (1940): 643. (2) Taperova, A. A., and N. M. Shulgiva. Zh. Prikl. Khim 18 (1945): 521. (3) Kurteva, O. I., and E. B. Brustkus. Zh. Prikl. Khim 34 (1961): 1714.

Batch reactive crystallization simulation

Provide C0 , T , and initial CSD Transform the PDE into ODEs

( , ) d x G x t d t 

 

( , ) ( , ) ( , ), , d f G x t f x t B f x t x t d t x     

Solve the ODEs using Matlab

• calculate drop in concentration
• update Excel
• trigger OLI Engine
• return scaling tendency
• next time step

Page 11

Reactive crystallization simulation result

Page 12

Matlab GUI

Summary

• Model for reactive crystallization needs accurate

thermodynamics

• OLI MSE model provides a good supersaturation

prediction for our electrolyte system

• OLI Engine was coupled with Matlab via Excel to

simulate batch reactive crystallization

Page 13