nucleation and growth kinetics for combined cooling and … · 2018-10-15 · jennifer m. schall....
TRANSCRIPT
Jennifer M. SchallCo-authors: Dr. Jasdeep Mandur, Prof. Richard D. Braatz, Prof. Allan S. MyersonOctober 3rd, 2018
Nucleation and Growth Kinetics for Combined Cooling and Antisolvent Crystallizationin an MSMPR System: Estimating Solvent Dependency
Pharmaceutical companies are evaluating continuous processing to decrease costs and increase control.
SafetyEconomics Control
API C
once
ntra
tion
Antisolvent Fraction
Outlet
InletHow do we design robust continuous crystallization processes quickly, using minimal API?
2
MSMPR Advantages:
Robust, steady-state operation
Consistent final product and quality
Take advantage of operating conditions with beneficial kinetics
Can select configuration to control polymorphism, purity, morphology
MSMPR crystallizers offer many advantages over batch crystallizers.
MSMPR ≡ Mixed-suspension, mixed-product reactor3
API C
once
ntra
tion
Temperature
Outlet
Inlet
Model assumptions:
1. Well-mixed
2. Negligible agglomeration and breakage
3. No growth dispersion
4. Nucleate from size zero
Model equations:
1. Material balance
2. Population balance
3. Growth expression
4. Nucleation expression
SS MSMPR models are readily expanded for multi-stage SS MSMPR cascade design.
4
1−=+ iii
ii nndLdnGτ
( )∫ −== − iiivsiT CCdLnLkM 13ρ
,,
lnig
ii g i
s i
CG kC
=
2/3,
,
lnib
ii b i T
s i
CB k MC
=
Growth
𝐺𝐺 = 𝑘𝑘𝑔𝑔 𝑙𝑙𝑙𝑙𝐶𝐶𝐶𝐶𝑠𝑠
𝑔𝑔
Nucleation
𝐵𝐵 = 𝑘𝑘𝑏𝑏 𝑙𝑙𝑙𝑙𝐶𝐶𝐶𝐶𝑠𝑠
𝑏𝑏
𝑀𝑀𝑇𝑇2/3
Growth and nucleation parameters can be regressed simultaneously from SS MSMPR experimental data.
Parameter Regression
min𝜃𝜃
Φ 𝜃𝜃 = �0
𝐿𝐿𝑚𝑚𝑚𝑚𝑚𝑚
𝑙𝑙𝑒𝑒𝑒𝑒𝑒𝑒 − 𝑙𝑙𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐(𝐿𝐿) 2
subject to SS MSMPR Model equations
where 𝜃𝜃 = 𝑘𝑘𝑔𝑔,𝑔𝑔,𝑘𝑘𝑏𝑏 , 𝑏𝑏
5
Crystallization kinetic parameters can have both temperature and solvent dependence…
… so models of antisolvent crystallization processes must account for the role of solvent composition.
1. Improve protocol for designing continuous antisolvent crystallizations.2. Decrease the number of experiments required for process design.
KineticsThermodynamics
𝐺𝐺 = 𝑘𝑘𝑔𝑔(𝑇𝑇, 𝑥𝑥𝑠𝑠)𝜎𝜎𝑔𝑔(𝑒𝑒𝑠𝑠)
𝐵𝐵 = 𝑘𝑘𝑏𝑏(𝑇𝑇, 𝑥𝑥𝑠𝑠)𝜎𝜎𝑏𝑏(𝑒𝑒𝑠𝑠)𝑀𝑀𝑇𝑇2/3
API C
once
ntra
tion
Solvent Volume Fraction
6
𝐶𝐶 = 𝑓𝑓(𝑇𝑇, 𝑥𝑥𝑠𝑠)
We use a four-step process to evaluate solvent-dependent kinetics for continuous crystallizer design.
Obtain single-stage MSMPR kinetic data.Experiment
Determine the nucleation and growth parameters through regression.Regress
Evaluate changes in kinetic parameters as functions of solvent composition.Evaluate
Construct and validate continuous crystallization model using kinetic expressions
for nucleation and growth. Model
7
Our ultimate goal is to develop a continuous crystallization process for a commercial API.
Process Constraints
• < 8 stages• Combined cooling
+ antisolvent
Crystallization System
• Confidential API• Solvent mixture:
92 v% EtOH / 8 v% THF
• Antisolvent: Water
Final Product Characteristics
• Maximize yield (>90%)
• x50 > 40 μm• x90 < 250 μm
8
We use a four-step process to evaluate solvent-dependent kinetics for continuous crystallizer design.
Obtain single-stage MSMPR kinetic data.Experiment
Determine the nucleation and growth parameters through regression.Regress
Evaluate changes in kinetic parameters as functions of solvent composition.Evaluate
Construct and validate continuous crystallization model using kinetic expressions
for nucleation and growth. Model
9
Single-stage MSMPR experiments are used to obtain kinetic data.
Cryst. Growth Des. 18, 3, 1560-157010
Single-stage MSMPR experiments are used to obtain kinetic data.
MSMPR Experimental Operating Conditions
Crystallizer Volume 80 mL
Feed Temp. 55°C
Crystallizer Temp. 10 - 30°C
Residence Time (RT) 1 - 3 hours
Solvent Volume Fractions 44 – 66%
Product Withdrawal Intermittent (1/10 RT)
11
Online, FBRM CLDs are used to track the transition to steady state. Steady-state operation is confirmed offline using HPLC and IR data.
0.3
0.5
0.7
0 300 600 900 1200 1500
Solvent VolumeFraction, xs(unitless)
Time, t (min)
0
5
10
15
20
25
0 300 600 900 1200 1500
Crystallizer Temperature, T
(°C)
Time, t (min)
0.0
0.5
1.0
1.5
2.0
2.5
0 300 600 900 1200 1500
Mother Liquor Concentration, C
(g API/kg solution)
Time, t (min)
0100020003000400050006000700080009000
10000
0 300 600 900 1200 1500
Particle Count Frequency
(#/s)
Time, t (min)
< 10 micron10 - 50 micron50 - 150 micron150 - 300 micron
Cryst. Growth Des. 18, 3, 1560-157012
Six steady-state continuous MSMPR experiments were used to map the operating space.
Experimental Conditions for Single-Stage MSMPR Kinetic Experiments
Steady-State Average
RunMSMPR Exp. #
TemperatureResidence
Time
Solvent Volume Fraction
Feed Concentration*
°C min unitless g API / kg solution
11 10 90 0.44 17.4312 20 90 0.44 17.431
23 10 180 0.66 27.1834 30 180 0.66 27.183
35 10 90 0.48 19.2206 30 90 0.48 19.220
4 7 20 90 0.47 20.154
* Adjusted for antisolvent addition
13
Both cooling and antisolvent addition affect solubility and crystallization kinetics, which affect yield.
Steady-State Average
RunMSMPR Exp. #
TemperatureResidence
Time
Solvent Volume Fraction
Supersaturation Concentration Yield
°C min unitless unitless g API / kg solutionmass
%% of equil.
11 10 90 0.44 1.19 1.17 93.29 95.22 20 90 0.44 1.05 1.34 92.31 94.9
23 10 180 0.66 1.26 10.24 62.34 69.84 30 180 0.66 1.05 15.11 44.40 55.1
35 10 90 0.48 1.75 2.92 84.80 87.16 30 90 0.48 1.27 3.23 83.19 87.3
4 7 20 90 0.47 1.32 2.29 88.64 91.4
Experimental Results for Single-Stage MSMPR Kinetic Experiments
14
We use a four-step process to evaluate solvent-dependent kinetics for continuous crystallizer design.
Obtain single-stage MSMPR kinetic data.Experiment
Determine the nucleation and growth parameters through regression.Regress
Evaluate changes in kinetic parameters as functions of solvent composition.Evaluate
Construct and validate continuous crystallization model using kinetic expressions
for nucleation and growth. Model
15
Before regressing kinetic parameters, we must first establish the solubility and approximate the CSD.
Equations for Regression
Operating Conditions
Outlet API concentration
Solubility
CLD
Supersaturation
CSD
Material balance
Population balance
Growth
Nucleation
Experimental Data
16
Solubility is required to predict supersaturation, and kinetics are very sensitive to errors in thermodynamic estimates.
03100.5
3000.6
10
Temperature, T (K)
Solvent Fraction, Xs
2900.7
2800.8
20
Sol
ubili
ty, S
(g/k
g of
sol
utio
n)
2700.9
30
40
Se x p
Sp re d
𝐺𝐺 = 𝑘𝑘𝑔𝑔 ln𝐶𝐶𝐶𝐶𝑠𝑠𝑠𝑠𝑐𝑐
𝑔𝑔
𝐵𝐵 = 𝑘𝑘𝑏𝑏 ln𝐶𝐶𝐶𝐶𝑠𝑠𝑠𝑠𝑐𝑐
𝑏𝑏
𝜌𝜌𝑘𝑘𝑣𝑣𝜇𝜇323
17
We evaluate solubility as a function of temperature and solvent composition.
1. Model temperature dependence using Apelblat equation.
2. Expand parameters to account for solvent composition.
ln 𝐶𝐶𝑠𝑠𝑠𝑠𝑐𝑐,𝑖𝑖 = 𝛽𝛽1𝑖𝑖 +𝛽𝛽2𝑖𝑖𝑇𝑇 + 𝛽𝛽3𝑖𝑖 ln 𝑇𝑇
𝛽𝛽𝑘𝑘 = 𝛼𝛼𝑘𝑘1 + 𝛼𝛼𝑘𝑘2𝑥𝑥𝑠𝑠 +𝛼𝛼𝑘𝑘3𝑥𝑥𝑠𝑠
+ 𝛼𝛼𝑘𝑘4 ln 𝑥𝑥𝑠𝑠
275 280 285 290 295 300 305 310
Temperature, K
0
5
10
15
20
25
30
35
Sol
ubili
ty, g
/kg
of s
olut
ion
0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9
Solvent Composition, (v/v)
0
5
10
15
20
25
30
35
Sol
ubili
ty, g
/kg
of s
olut
ion
Decreasing solvent composition
Decreasing temperature
Cryst. Growth Des. 18, 3, 1560-157018
The volume-weighted CSD is approximated by the L4-weighted CLD from FBRM.
19
0
0.01
0.02
0.03
0.04
0.05
0.06
0 200 400 600
Volu
me-
base
d CS
D
Length, L (microns)
normalized CLD, L3normalized cubed PSD
0
0.01
0.02
0.03
0.04
0.05
0.06
0 200 400 600
Volu
me-
base
d CS
D
Length, L (microns)
normalized CLD, L4normalized cubed PSD
Cryst. Growth Des. 18, 3, 1560-1570
Kinetic parameters are determined through least-squares regression.
Steady-state parameter estimation:
min𝑘𝑘𝑔𝑔 ,𝑘𝑘𝑏𝑏
𝑤𝑤1 𝐶𝐶𝑒𝑒𝑝𝑝𝑒𝑒𝑝𝑝 − 𝐶𝐶𝑒𝑒𝑒𝑒𝑒𝑒2 + 𝑤𝑤2�
𝐿𝐿
𝑉𝑉𝑉𝑉𝑙𝑙𝐶𝐶𝐶𝐶𝐷𝐷𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝(𝐿𝐿)− 𝑉𝑉𝑉𝑉𝑙𝑙𝐶𝐶𝐶𝐶𝐷𝐷𝑝𝑝𝑚𝑚𝑝𝑝(𝐿𝐿)2
s.t.
𝑙𝑙 𝐿𝐿 =𝐵𝐵𝐺𝐺 exp −
𝐿𝐿𝜏𝜏𝐺𝐺
𝑉𝑉𝑉𝑉𝑙𝑙𝐶𝐶𝐶𝐶𝐷𝐷 =∫𝐿𝐿𝑖𝑖𝐿𝐿𝑖𝑖+1 𝑙𝑙 𝐿𝐿 𝐿𝐿3𝑑𝑑𝐿𝐿
∫𝐿𝐿0𝐿𝐿𝑓𝑓 𝑙𝑙 𝐿𝐿 𝐿𝐿3𝑑𝑑𝐿𝐿
𝐶𝐶𝑖𝑖𝑖𝑖 − 𝐶𝐶 − 6𝜌𝜌𝑘𝑘𝑣𝑣𝐵𝐵𝐺𝐺3𝜏𝜏4 = 0𝜇𝜇3 − 6𝐵𝐵𝐺𝐺3𝜏𝜏4 = 0
𝐺𝐺 = 𝑘𝑘𝑔𝑔 ln𝐶𝐶𝐶𝐶𝑠𝑠𝑠𝑠𝑐𝑐
𝑔𝑔
𝐵𝐵 = 𝑘𝑘𝑏𝑏 ln𝐶𝐶𝐶𝐶𝑠𝑠𝑠𝑠𝑐𝑐
𝑏𝑏 𝜇𝜇33𝐺𝐺𝜏𝜏
𝑔𝑔 = 1𝑏𝑏 = 2
20
We use a four-step process to evaluate solvent-dependent kinetics for continuous crystallizer design.
Obtain single-stage MSMPR kinetic data.Experiment
Determine the nucleation and growth parameters through regression.Regress
Evaluate changes in kinetic parameters as functions of solvent composition.Evaluate
Construct and validate continuous crystallization model using kinetic expressions
for nucleation and growth. Model
21
Growth and nucleation rate coefficients increase with antisolvent fraction.
kbkg
0
0.1
0.2
0.3
0.4
0.5
0.4 0.5 0.6 0.7
Grow
th ra
te co
effic
ient
[kg]
(μm
/min
x 10
6 )
Solvent volume fraction [xs] (unitless)
0
1
2
3
4
5
0.4 0.5 0.6 0.7
Nuc
leat
ion
rate
coef
ficie
nt[k
b] (#
/min
/kg
solu
tion
x 10-6
)
Solvent volume fraction [xs] (unitless)
Cryst. Growth Des. 18, 3, 1560-157022
Growth coefficients follow an Arrhenius temperature relationship, while nucleation coefficients show mild temperature sensitivity.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.4 0.5 0.6 0.7
Grow
th ra
te co
effic
ient
[k
g] (μ
m/m
in x
106 )
Solvent volume fraction [xs](unitless)
0
1
2
3
4
5
0.4 0.5 0.6 0.7
Aver
age
nucl
eatio
n ra
te co
effic
ient
[k
b] (#
/min
/kg
solu
tion
x 10-6
)
Solvent volume fraction [xs] (unitless)
Cryst. Growth Des. 18, 3, 1560-157023
kbkg
Regressed kinetic parameters are used to reconstruct steady-state PSDs and concentrations.
MSMPR Exp. # Temperature (°C) Residence Time (min) Solvent Fraction (unitless)3 10 180 0.664 30 180 0.66
0 500 1000 1500 2000 2500 3000 3500
Time, (min)
0.002
0.004
0.006
0.008
0.01
0.012
0.014
0.016
0.018
AP
I Con
cent
ratio
n (g
/g)
Measured conc; Xs=66%; T=10C
Measured conc; Xs=66%; T=30C
predicted conc; Xs=66%; T=10C
predicted conc; Xs=66%; T=30C
0 100 200 300 400 500 600 700 800 900 1000
Length, L (microns)
0
0.01
0.02
0.03
0.04
0.05
0.06
Vol
ume
base
d di
strib
utio
n
L 4 weighted CLD; Xs=66%; T=10C
Predicted vol PSD; Xs=66%; T=10C
L 4 weighted CLD; Xs=66%; T=30C
Predicted vol PSD; Xs=66%; T=30C
Cryst. Growth Des. 18, 3, 1560-157024
We use a four-step process to evaluate solvent-dependent kinetics for continuous crystallizer design.
Obtain single-stage MSMPR kinetic data.Experiment
Determine the nucleation and growth parameters through regression.Regress
Evaluate changes in kinetic parameters as functions of solvent composition.Evaluate
Construct and validate continuous crystallization model using kinetic expressions
for nucleation and growth. Model
25
Data from MSMPR Experiment 7 were used to validate the kinetic parameter model.
2000 2500 3000 3500
Time, (min)
1.4
1.6
1.8
2
2.2
2.4
2.6
2.8
AP
I Con
cent
ratio
n (g
/g)
10-3
Measured conc; Xs=47%; T=20C
predicted conc; Xs=47%; T=20C
0 200 400 600 800 1000
Length, L (microns)
0
0.01
0.02
0.03
0.04
0.05
0.06
Vol
ume
base
d di
strib
utio
n
L 4 weighted CLD; Xs=47%; T=20C
Predicted vol PSD; Xs=47%; T=20C
Cryst. Growth Des. 18, 3, 1560-157026
If we neglect solvent dependence in kinetics, crystallizer performance will not be acceptably predicted.
Kinetic parameters regressed from Experiments 5 & 6 (xs = 0.48) were used to predict data from Experiments 3 & 4 (xs = 0.66).
0 500 1000 1500 2000 2500 3000 35000.002
0.004
0.006
0.008
0.01
0.012
0.014
0.016
0.018
Time (min)
AP
I Con
cent
ratio
n (g
AP
I/g s
olve
nt)
Measured conc; xs=66%; T=10°C
Measured conc; xs=66%; T=30°C
predicted conc; xs=66%; T=10°C
predicted conc; xs=66%; T=30°C
0 200 400 600 800 10000
0.01
0.02
0.03
0.04
0.05
0.06
Length, L (microns)
Vol
ume-
base
d di
strib
utio
n fra
ctio
n (u
nitle
ss)
L4 weighted CLD; xs=66%; T=10°C
Predicted vol PSD; xs=66%; T=10°C
L4 weighted CLD; xs=66%; T=30°C
Predicted vol PSD; xs=66%; T=30°C
Cryst. Growth Des. 18, 3, 1560-157027
SS prediction with solvent-dependent kinetics
Crystallizer performance may not be acceptably predicted even at similar solvent compositions.
Kinetic parameters regressed from Experiments 5 & 6 (xs = 0.48) were used to predict data from Experiments 1 & 2 (xs = 0.44).
0 200 400 600 800 10000
0.01
0.02
0.03
0.04
0.05
0.06
Length, L (microns)
Vol
ume-
base
d di
strib
utio
n fra
ctio
n (u
nitle
ss)
L4 weighted CLD; xs=44%; T=10°C
Predicted vol PSD; xs=44%; T=10°C
L4 weighted CLD; xs=44%; T=20°C
Predicted vol PSD; xs=44%; T=20°C
0 500 1000 1500
0.5
1
1.5
2
x 10-3
Time (min)
API C
once
ntra
tion
(g A
PI/g
sol
vent
)
Measured conc; xs=44%; T=10°C
Measured conc; xs=44%; T=20°C
predicted conc; xs=44%; T=10°C
predicted conc; xs=44%; T=20°C
Cryst. Growth Des. 18, 3, 1560-157028
SS prediction with solvent-dependent kinetics
Small changes in solvent composition can have large yield impacts, arising from changing kinetics.
Steady-State Average
RunMSMPR Exp. #
TemperatureResidence
TimeSolvent Volume
FractionYield
°C min unitless mass % % of equil.
35 10 90 0.48 84.80 87.16 30 90 0.48 83.19 87.3
4 7 20 90 0.47 88.64 91.4
Results for 90 min RT Single-Stage MSMPR Kinetic Experiments
29
We use a four-step process to evaluate solvent-dependent kinetics for continuous crystallizer design.
Obtain single-stage MSMPR kinetic data.Experiment
Determine the nucleation and growth parameters through regression.Regress
Evaluate changes in kinetic parameters as functions of solvent composition.Evaluate
Construct and validate continuous crystallization model using kinetic expressions
for nucleation and growth. Model
30
Once we know the functionality of kinetic parameters, we can predict crystallizer performance in MSMPR cascades.
31
Thermodynamics & kinetics depend on operating conditions in each stage
Ti, xs,i kg,i, gi, kb,i, bi
Can now estimate attainable region & optimize cascade performance
AS
…AS ASAS
T1, xs,1 T2, xs,2 Ti-1, xs,i-1 Ti, xs,i
Product
In the simplest case, we can predict performance in a single crystallizer.
32
ASFeed
Ti, xs,i
Product
Optimization ProblemMaximize yield (Y) subject to:
60min 180mintotτ< <
10 55iC T C° ≤ ≤ °
,0.44 0.90s ix≤ ≤
50 40d mµ>
90 250d mµ<
70 g/kgfeedC =
Optimized Operating Conditions Product Constraints Met?
Case Feed Concentration
Residence Time
Temp. Solvent Fraction
Yield Yield > 90%? x50 > 40 μm x90 < 240 μm
g API / kg soln min °C unitless mass%Neglecting solvent effects on kinetics. 70 71.4 10.0 0.55 79.5% X
Kinetics are solvent-dependent. 70 104.9 10.0 0.47 95.4%
Final product specifications can be met in a single MSMPR crystallizer!
In the future, we will predict crystallization performance in multi-stage antisolvent MSMPR cascades.
This method requires 6+ steady-state experiments; fewer for transient experiments.Method
Growth and nucleation kinetic parameters are functions of solvent composition.Functionality
To acceptably model crystallizer performance, we should include solvent composition effects in kinetic expressions.
Modeling
Simulate and validate a multi-stage MSMPR crystallization cascade.Future Work
33
Acknowledgements
Prof. Allan Myerson
Prof. Bernhardt Trout
Prof. Richard Braatz
Myerson & Trout group members
Dr. Jasdeep Mandur
UROPs Tony Elian & Zach Schmitz
34
Supplemental Slides
At steady-state, particles are not appreciably agglomerated.
Experimental Conditions (6): 48% (EtOH/THF)/52% Water (v/v), 90 min RT, 30°C
36
Heavy fouling prevented online IR analysis and influenced start-up procedure.
37
Model for Multistage MSMPR
( )∫ −== − iiivsiT CCdLnLkM 13ρ
g
s
sg C
CCkG
−=
2/3b
sb T
s
C CB k MC
−=
Population Balance: Conservation equation for the number of crystals in a population
Mass Balance
Crystal Growth
Nucleation
ni: population density at stage iτi: residence time at stage iL: crystal sizeGi: crystal growth rate at stage iB: nucleation rate
MT i: Suspension density at stage iC: steady state solute concentrationρs: crystal densitykv : volume shape factor
Cs: equilibrium concentrationkg, g, kb, b: model parameters to
be estimated
1−=+ iii
ii nndLdnGτ (1)
(2)
(3)
(4)