Continuous crystallisation using oscillatory baffled plug flow
crystalliser
Professor Xiong-Wei Ni, BSc, PhD, CEng, CSci, FIChemENiTech Solutions Ltd
Scottish Enterprise Technology ParkEast Kilbride, Glasgow G75 0QF, UK
www.nitechsolutions.co.uk
1818thth June 2008June 2008
© NiTech Solutions, 2008 Confidential
Structure
1. Crystallisation science
2. Challenges in industrial crystallisation
3. Continuous oscillatory baffled crystalliser
4. Case studies
5. Forward remarks
© NiTech Solutions, 2008 Confidential
• Crystallisation is a relatively simple process
• But the underlying science and its control are very complex!
• In industrial scales, operators’experience still plays a major part
1. Crystallisation Science
Supersolubility or metastable limit is thermodynamically not found and kinetically not well defined, depending on temperature, rate of generating supersaturation, solution history, impurities, fluid dynamics, scale and etc.
Under-saturated
Solubility
Metastablezone
Temperature
Concentration
Supersolubility
Labile
Crystallisation “pathway”
© NiTech Solutions, 2008 Confidential
Science
Nucleation mechanism
Nucleation rate
Growth mechanism
Growth rate
Supersaturation
Concentration
Temperature
Cooling profile
Mixing
Solvent/additives
Seeding
Crystal size and shape
Morphology
PurityProcess
Science
ProductProduct
Functionality (materials)
Bio-availability (drugs)
Surface activity (catalysts)
Texture (foods)
© NiTech Solutions, 2008 Confidential
• How does it happen?• Primary nucleation – without the presence of any crystalline
matter
• Secondary nucleation – collision breeding, seeding
Nucleation• What initiates?
• not well understood
• 3-D assembly of molecular clusters on nm scale
• very fast kinetics
• not easy to detect (soft X-ray absorption spectroscopy)
• modelling available, lack of validation
© NiTech Solutions, 2008 Confidential
Determination of metastable zone width (MSZW) using a turbidity probe
Tsat
On-set dissolution
On-set nucleation
Tcry
crysat TTMSZW −=
Detection of post-nucleation
© NiTech Solutions, 2008 Confidential
MetastableMetastable zone width (MSZW) zone width (MSZW)
concentration = 45 g/L, f = 2 Hz, xo = 10 mm
MSZW depends on reactor, scale and operating conditions
• Dissolution – a thermodynamic process
• Crystallisation –a kinetic process
© NiTech Solutions, 2008 Confidential
• Mass transfer phenomenon
• Crystal growth phase must take place within the MSZW
Crystal growth
• Nucleation dominates small crystals
• Growth dominates large crystals
the degree of supersaturation is changed by crystal growth, simultaneously MSZW is altered due to different temperature levels and changes in impurity concentration
Under-saturated
Solubility
Metastablezone
Temperature
Concentration
Supersolubility
Labile
Crystallisation “pathway”
© NiTech Solutions, 2008 Confidential
Crystal Size
PolymorphicForm
Supersaturation
ProcessConditions
Shape
MSZW
© NiTech Solutions, 2008 Confidential
Crystal Size
PolymorphicForm
Supersaturation
ProcessConditions
Shape
MSZW
Growth Kinetics Nucleation Kinetics
Hydrodynamics Heat Transfer
© NiTech Solutions, 2008 Confidential
Chemicals Behaving Badly (CBB)Chemicals Behaving Badly (CBB)
Crystal Size
PolymorphicForm
Supersaturation
ProcessConditions
Shape
MSZW
Growth Kinetics Nucleation Kinetics
Hydrodynamics Heat Transfer
FTIR
USS
VideoMicroscopy
PIV/LDA ReactionCalorimetry
Flow thruXRD
UV Vis
CFD
Courtesy of Dr White, Heriot-Watt University
© NiTech Solutions, 2008 Confidential
• Simultaneous multi-technique measurements during a crystallisation process
• Have promoted better understanding of crystallisation in lab scale
• Promoted better operation of crystallisation in small scale
• Linear cooling identified as one of the key operational parameters
Lab scale batch crystallisation
• Data rich
• Lack of correlations between data
• Disturbing flow conditions
© NiTech Solutions, 2008 Confidential
• Nucleation via collision breeding:
• Walls of vessel
• free surface of liquid
• internals, e.g. stirrer, baffle
• impurities, e.g. particulates, seeds
• crystal/crystal & crystal/vessel collision
• inhomogeneities due to mixing
Mixing affects
0
5
10
15
20
25
30
35
0 5000 10000 15000
Reynolds Number, Re o [-]
MSZ
W [
o C]
30g/L 0.5°C/min
35g/L 0.5°C/min
30g/L 2.0°C/min
35g/L 2.0°C/min
0
20
40
60
80
100
120
140
160
180
200
0 5000 10000 15000 20000 25000 30000
Reynolds Number, Re o [-]
Cry
stal
Siz
e [ µ
m]
30g/L 2.0°C/min
35g/L 2.0°C/min• MSZW
• Crystal size distribution
© NiTech Solutions, 2008 Confidential
• Our understanding in scaling up STR is limited
• There is no agreed rule or parameter on scaling
• Velocity gradient or non-homogeneity increases with scale
• Mixing gradient leads to temperature and mass gradient
2. Challenges in industrial crystallisation
© NiTech Solutions, 2008 Confidential
ScaleScale--up of stirred tanksup of stirred tanks
Mixing cannot linearly be scaled in STR
Laboratory
Industrial scale
Kinetic control
Mass/heat transfer control
© NiTech Solutions, 2008 Confidential
Heat Transfer on ScaleHeat Transfer on Scale--UpUp
Specific heat transfer area decreases with scale
2.315.86.82Typical 6500 litre STR
100.90.09Typical 90 litre STR
Area / unit Volume (m2/m3)
Area (m2)Volume (m3)
© NiTech Solutions, 2008 Confidential
Industrial batch crystallisation
Temperature
gradient
MSZW and
Crystallisation path
nucleation
& growth
Polymorph
& size
Inconsistent morphology
Wide size distribution
Difficult to filter
Mixing
gradient
Concentration
gradient
supersaturtion
gradientpolymorph
MSZW
Temperature
Concentration
∆C
∆T
© NiTech Solutions, 2008 Confidential
Practical issues
• Linear cooling is difficult to achieve in large STR
• Implementation of lab multi-measurement techniques is practically impossible in industrial scale
© NiTech Solutions, 2008 Confidential
i) consistent product quality can only be achieved in plug flows;
ii) it is very rare to obtain plug flow conditions in batch STR!
iii) Plug flow conditions can only be attained in continuous operation.
From the viewpoint of fluid mechanics
3. Continuous crystallisation
© NiTech Solutions, 2008 Confidential
The conventional ways to achieve plug flow include
a) Using a series of CSTRs
b) Employing a tubular reactor
© NiTech Solutions, 2008 Confidential
a) Using a series of a) Using a series of CSTRsCSTRs
Conclusions: a) significant increase in inventory, running and capital costs
b) far from plug flow
…..
Plug flow is achieved when the number of CSTRsgoes to infinite
© NiTech Solutions, 2008 Confidential
b) Employ a tubular reactor
Conclusion: a) significant high flow rates, leading to very long reactor and large capital costs
b) very short residence time
Operating a tubular reactor at turbulent flow regime in order to obtain near to plug flow
© NiTech Solutions, 2008 Confidential
Reo = 1250 (xo = 4 mm, f = 1 Hz)
Real system
Demonstration of radial mixingDemonstration of radial mixingwith good dispersionwith good dispersion
3-D CFD simulation
© NiTech Solutions, 2008 Confidential
NiTechNiTech’’ss Continuous OBR Continuous OBR (COBR(COBRTMTM))
Oscillation
Out
Tracer
liquid
Conductivity probes
© NiTech Solutions, 2008 Confidential
Residence time distribution (RTD) in a COBR
Probe 13.7 meters away from injectionProbe 27.9 meters away from injectionProbe 310.1 meters away from injection
Mixing controlled by the combination of oscillation and bafflesEach baffled cell is a CTSR Plug flow at laminar flowsHandles particulatesLong residence times
D = 0.00007~0.0003 m2/s
© NiTech Solutions, 2008 Confidential
20
0
0.1
0.2
0.3
0.4
0.5
0 500 1000 1500 2000Drop diameter (µm)
Dro
p N
umbe
r Fr
actio
n
15mm, 1Hz,Ren = 50015mm, 2Hz,Ren = 50015mm, 3Hz,Ren = 500
Drop/Particle/crystal size distribution obtained in a COBR
© NiTech Solutions, 2008 Confidential
• Mixing is not controlled by net flow
• Plug flow achieved at laminar flows
• Excellent heat & mass transfers
• No concentration gradients
• Good with solids
Key differences from other tubular devices on the market
© NiTech Solutions, 2008 Confidential
Heat Transfer on Scale-Up
COBC™ has larger specific surface area for HT
4022.60.56Typical 72m long 100mm
diameter COBCTM
2.315.86.82Typical 6500 litre STR
100.90.09Typical 90 litre STR
Area / unit Volume (m2/m3)
Area (m2)Volume (m3)
© NiTech Solutions, 2008 Confidential
Tube side heat transfer coefficient (w/m2K)
Plug flow delivers excellent heat transfer
47 - 1965,209 - 23,730232,744 - 1,067,5341500 kg/m3
21 - 822,327 - 10,575133,315 - 611,4471000 kg/m3
7 - 28817 - 3,69461,654 - 282,717600 kg/m3
100cp10cp1cp
© NiTech Solutions, 2008 Confidential
Control over cooling profile
90-10 oC Flow Rate 0.3 L/min Cooling Rate 2.0 oC/min
y = -1.7118x + 81.031R2 = 0.9934
y = -1.7103x + 80.986R2 = 0.9935
y = -1.7106x + 80.97R2 = 0.9933
y = -1.7112x + 80.972R2 = 0.9933
y = -1.7121x + 81.032R2 = 0.9936
0
10
20
30
40
50
60
70
80
90
0 10 20 30 40 50
Time (min)
Tem
p (o
C)
80-10oC Flow Rate 0.945 L/min Cooling Rate 6.5oC/min
y = -6.5517x + 83.727R2 = 0.9918
y = -6.5524x + 83.696R2 = 0.9916
y = -6.5657x + 83.8R2 = 0.9919
y = -6.5541x + 83.668R2 = 0.9917
y = -6.5629x + 83.699R2 = 0.9919
0
10
20
30
40
50
60
70
80
90
0 2 4 6 8 10 12
Time (min)
Tem
p (o
C)
Any cooling profile, e.g. linear, parabolic, non-continuous, step-wise and etc, can be obtained along COBCTM
Multi-monitoring techniques in lab scale can be implemented along COBCTM
© NiTech Solutions, 2008 Confidential
4. Case studies
COBCTM enables greater control over crystallisation path
Able to be more selective on crystal morphology β crystals
α crystals
α + β crystals
a) Morphology of a pharma crystal
Seed α crystals
Seed β crystals
© NiTech Solutions, 2008 Confidential
Change on crystal sizeImpact on filtration
The residence time in the COBCtranslates to 12 minutes, compared to a batch cycle time of 8 hours, demonstrating a significant potential improvement in throughput relative to a batch manufacturing facility.
b) Pharma API - size on filtration
© NiTech Solutions, 2008 Confidential
c) Filtration index for a pharma product
COBCTM affects outcomes achievedProduct characteristics/performanceConsistent filtration index of 25
STR COBCTM
© NiTech Solutions, 2008 Confidential
d) Crystal size distribution of a fine chemical product
More uniform size can be achieved
Greater control on outcome
µ
0
5
10
15
20
25
0 250 500
Particle Diameter (um)
Vol
umn
(%)
STR
OBCEXP 1OBCEXP 2
0
5
10
15
20
25
0 250 500
Particle Diameter (um)
Vol
umn
(%)
STR
OBCEXP 1OBCEXP 2
0
5
10
15
20
25
0 250 500
Particle Diameter (µm)
Vol
umn
(%)
STR
OBCEXP 1OBCEXP 2
© NiTech Solutions, 2008 Confidential
e) Fractionation of edible oils
Two stage cooling (17 and 0.3 oC/min)
Consistent IV achieved0
5
10
15
20
25
0 50 100 150 200
Process Time (min)Fi
ltrat
ion
time
(min
)
Benchmark
T05C
T08C
T10C
T11C
T12C
T13C
T14C
T16C
• Significant reduction of crystallisation time
• Improved filtration time
© NiTech Solutions, 2008 Confidential
f) Combined reaction & crystallisationCrash cooling Variable mean sizes from crystallisation after a reaction Crystals stuck to the walls of the vessel and surfaces of the impeller
Consistent mean crystal sizeLinear cooling profiles in COBCTM eliminated the events of crystal-stickingMinimise the down-time of washing/cleaning
0
10
20
30
40
50
60
70
80
90
0 50 100 150
Oscillation velocity (mm/s)
d 90 (
µm
)
2 oC/minUncontrolled cooling
0
20
40
60
80
100
120
140
0 1 2 3 4 5 6
Cooling rate (oC/min)
d 90 (
µm
)
© NiTech Solutions, 2008 Confidential
• Plug flow tubular crystallisation technology, such as COBCTM, can be used to bridge the gap between lab and industrial scale crystallisation operations
• Multi-monitoring techniques from lab scale crystallisation can directly be fitted along the COBCTM
• COBCTM can be incorporated into existing plants
• Continuous filtration could also be implemented in COBCTM
4. Forward remarks