Continuous-Flow Tubular Crystallization in Slugs Spontaneously
Induced by Hydrodynamics
Mo Jiang,
†
Zhilong Zhu,
†
Ernesto Jimenez,
†
Charles D. Papageorgiou,
‡
Josh Waetzig,
‡
Andrew Hardy,
‡
Marianne Langston,
‡
and Richard D. Braatz*
,†,‡
†
Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, Massachusetts, 02139, United States
‡
Millennium: The Takeda Oncology Company, Process Chemistry Research & Development, 35 Landsdowne Street, Cambridge,
Massachusetts 02139, United States
*
S
Supporting Information
ABSTRACT: A novel continuous crystallizer design is described with the
potential to provide improved control of crystal properties, improved process
reproducibility, and reduced scale-up risk. Liquid and gas are introduced into
one end of the tube at flow rates selected to spontaneously generate alternating
slugs of liquid and gas that remain stable while cooling crystallization occurs in
each liquid slug. Mixing within each stable self-circulating slug is maximized by
controlling the slug aspect ratio through specification of liquid and gas flow
rates. The crystallizer is designed so that nucleation and growth processes are
decoupled to enhance the individual control of each phenomenon. Coaxial or
radial mixers combine liquid streams to generate seed crystals immediately
upstream of the growth zone where nucleation is minimized, and crystal
growth is controlled by the varying temperature profile along the length of the
tube. The slug-flow crystallizer design is experimentally demonstrated to
generate large uniform crystals of
L-asparagine monohydrate in less than 5 min.
1. INTRODUCTION
Compared to batch and semibatch crystallizers, continuous-flow
crystallizers have the potential for higher reproducibility, process
efficiency, and flexibility, as well as lower capital and production
cost.
1−8
Many continuous-flow crystallizer designs have been
proposed in recent years with the goals of improving
pharmaceutical processing or control of crystal product pro-
perties. For example, a continuous oscillatory baffled crystallizer
has been proposed that employs a piston to agitate the crystal
slurry in a long pipe with baffles.
2
Such a crystallizer can typically
operate at lower shear rates than a single stirred-tank crystallizer
but is limited to lower solids densities and effective viscosities and
has limited degrees of freedom for feedback control. An alternative
design that employed laminar flow through a tube was
demonstrated for paracetamol crystallization, but it was warned
that inducing crystallization in a continuous slurry undergoing
laminar flow in a tube by cooling through its walls was very prone
to clogging for some solute−solvent systems.
6
With the notion
that the best control of crystal size distribution should be obtained
in a conti nu ou s-flow design that has negligible back-mixing with a
residence time distribution approaching that of an ideal plug-flow
crystallizer, several designs have tried to generate plug flow-like
conditions. One continuous crystallizer design consists of slurry
flow ing through a tube with Kenics-type static mixers through its
length to induce a more uniform velocity profile, with improved
controllability obtained by introducing antisolvent at multiple
points along the tube length.
1
A tubular crystallizer less prone to
clogging ensured plug-flow behavior by forcibly segmenting the
flow into liquid slugs separated by an immiscible fluid.
3−5,9
Because of recirculation, the slurry in each slug is mixed even in
the absence of any static mixer.
10,11
However, the separation of
immiscible solvent takes additional process time when multiple
liquids are used, and the slug production devices, usually with
more than three channels, were relatively complicated. A
significant advantage of all of these continuous-flow crystallizer
designs is the lack of a stirring blade, resulting in negligible attrition
compared to the mixed-suspension mixed-product-removal
crystallizers widely employed for inorganic crystallization.
12
Most existing continuous-flow crystallizers are not specifically
designed to provide many degrees of freedom for the control of
crystal shape and size distribution in the presence of process
disturbances and variations in crystallization kinetics associated
with changes in contaminant profile of the feed streams. An
improved continuous-flow crystallizer design would minimize
operational problems while retaining the advantages of the
best previous continuous crystallizers. At same time, there are
advantages to continuous-flow designs that enable the direct
application of best-practices operations developed for the optimal
control of batch and semibatch crystallization.
Toward these goals, this paper describes the design and
implementation of a slug-flow crystallizer in which a multiphase
mixture of liquid and gas in a tube spontaneously separates into
Received: November 17, 2013
Revised: January 2, 2014
Published: January 7, 2014
Article
pubs.acs.org/crystal
© 2014 American Chemical Society 851 dx.doi.org/10.1021/cg401715e | Cryst. Growth Des. 2014, 14, 851−860
slugs of liquid or slurry separated by slugs of gas.
13−19
The slug-
flow crystallizer has all the advantages of a segmented-flow
crystallizer but does not require any specialized equipment
to induce slug formation and avoids the use of liquid/liquid
separators. In the proposed design, hydrodynamically stable slugs
form spontaneously immediately upon the contact of the liquid/
slurry and the gas. This paper demonstrates the crystallizer
design for a cooling crystallization, which is inherently more
challenging than antisolvent crystallization due to the latter’s
easier nucleation to generate seed crystals.
Figure 1. Schematic of slug-flow cooling crystallization. Crystal nucleation is induced by combining hot and cold LAM-aqueous solutions
(photographs in Figure 2). Slug formation spontaneously occurred by combining LAM slurry and air streams (Figure 3a). Crystal growth occurs in
the slurry in each slug going through the tube for a specified residence time (Figure 3b).
Figure 2. Photographs and schematics of setup for nucleation induced by cooling: (a) laminar flow tube; (b) coaxial mixer, the inner diameters of the
two inlets are 3.1 mm (hot) and 0.26 mm (cold), respectively; (c) radial mixer, the inner diameters of the two inlets are 2 mm (hot) and 0.3 mm
(cold), respectively.
Crystal Growth & Design Article
dx.doi.org/10.1021/cg401715e | Cryst. Growth Des. 2014, 14, 851−860852
2. EXPERIMENTAL METHODS AND EQUIPMENT SETUP
For better control of product crystal size distribution (CSD), the
cooling crystallization process is decoupled into three parts: crystal
nucleation, slug formation, and crystal growth (Figure 1). Nuclei were
generated within laminar flow by direct cooling, or within a coaxial or
radial jet mixer by mixing hot and cold solutions. The slurry stream
containing nuclei acting as seed crystals are combined with an air
stream at flow rates selected to spontaneously form stable slugs. The
motion of the slug through the tube mixes the solution without
requiring any of the stirring blades that cause attrition in mixed-tank
crystallizers. With each slug going through the tube with the same
residence time, and with no stirring blades to cause attrition, the
crystals in the slugs would be expected to grow to a large uniform size.
The following experimental methods and equipment setup were
selected to demonstrate the feasibility of this process design.
The solute is
L-asparagine monohydrate (LAM, purity ≥99%
(TLC), from Sigma Aldrich), and the solvent is deionized (DI) water,
which were selected because this solute−solvent combination is a
well-studied model system that is challenging due to the very strong
tendency of LAM crystals (product form, Figure A1, Supporting
Information) to aggregate in this solvent.
20
2.1. Nuclei Generation. In a preliminary experiment, a high solids
amount (high initial concentration, 0.16 g of LAM/g of DI, as shown
in Table 1) in each slug was used to demonstrate slug stability and
CSD improvement in the growth stage. In another preliminary experi-
ment, a low solids amount (low initial concentration, 0.09 g of LAM/g
of DI, as shown in Table 1) was applied to facilitate in-line and o ff-line
imaging and to compare the nucleation obtained from coaxial and
radial mixers, to identify the most promising mixer type for subsequent
experiments.
The nucleation experimental conditions are listed in Table 1.
For the radial mixing experiment (LAM nuclei were generated by
combining hot and cold saturation solutions (66 °C, 0.16 g of LAM/g
of DI; and 22 °C, 0.02 g of LAM/g of DI, respectively) through
a coaxial or radial jet mixer (Figure 2b,c). This approach, first
demonstrated in ref 20, contrasts with other continuous tubular
cooling crystallizations that require a separate batch crystallizer to pro-
duce seed crystals to feed into the continuous crystallizer. Peristaltic
pumps enable simple operation of tubing at high temperature, while
a syringe pump provides smooth flow rates. A peristaltic pump
(Masterflex pump drive 7521-40, Easy Load II pump head with model
#77200-50) and a silicone tube (Masterflex BioPharm Plus platinum-
cured silicone tubing, 3.1 mm inner diameter) were used to transfer
hot solution to the mixer at a rate of about 3.7 mL/min. The tubing
choice was based on the following criteria: low extractable and
spallation, high temperature endurance, surface hydrophobicity
(a more convex water slug shape is obtained by using a more hydro-
phobic surface),
21
and long life of the pump tubing. Around the pump
head, a silicone tube with smaller inner diameter (2 mm) was used
to increase the pump rotation rate for the same flow rate, to reduce
the amplitude of flow pulsations. A syringe pump (model NE-4000,
from New Era Pump Systems, Inc.) was used to transfer cold solution
to the mixer at a rate of 3.3 mL/min. The operational details of the
other experiments in Table 1 are similar to the radial mixing
experiment but with the different values given in the associated column
of the table.
2.2. Slug Generation. A key element in the proposed crystallizer
design is to exploit the hydrodynamically stable spontaneous
generation of slugs. Many hydrodynamically stable flow regimes can
occur when a liquid and gas are combined in a tube, as shown in
Figure 3,
15,16,22
with the stability of the flow regimes determined by
the ranges of inlet gas and liquid flow rates, tubing diameter, and fluids
properties. For crystallization from liquid solution, most of the flow
regimes have relatively poor spatial mixing of the liquid and allow
liquid solution to remain close to the tube walls for long periods of
time, which would encourage crystallization on the surface of the inner
Table 1. Main Experimental Conditions for Cooling
Nucleation
a
experiment preliminary coaxial radial
concentration of hot stream
(g of LAM/g of DI)
0.16 or
0.09
0.16 0.16
concentration of cold stream
(g of LAM/g of DI)
N/A 0.02 0.02
temperature of hot stream (°C) 65 or 44 65 66
temperature of cold stream (°C) N/A 23 22
pump for hot stream peristaltic peristaltic peristaltic
pump for cold stream N/A peristaltic syringe
tubing for hot and cold streams Tygon Tygon BioPharm
volumetric flow rate of hot stream
(mL/min)
23.5 7.2 3.7
volumetric flow rate of cold stream
(mL/min)
N/A 6.8 3.3
average velocity of hot stream (m/s) 0.05 0.02 0.02
average velocity of cold stream (m/s) N/A 2.14 0.78
nucleation site (mixer) Figure 2a Figure 2b Figure 2c
mixer inner diameter of hot inlet
(mm)
3.1 3.1 2
mixer inner diameter of cold inlet
(mm)
N/A 0.26 0.3
a
The preliminary experiment did not have a cold stream (the hot
stream was directly cooled in an ice bath; see Figure 2a). “Peristaltic”
refers to a peristaltic pump (Masterflex pump drive 7521-40, Easy
Load II pump head with model # 77200-50), and “Syringe” refers to a
syringe pump (model NE-4000, from New Era Pump Systems, Inc.).
“Tygon” refers to Masterflex Tygon tubing, and “BioPharm” refers to
Masterflex BioPharm Plus platinum-cured silicone tubing. The
volumetric flow rate reported for the syringe pump was precalibrated,
and the volumetric flow rate reported for the peristaltic pump was
estimated from the mass flow rate and temperature measured right
before or after the experiment. The average velocities are calculated
from the volumetric flow rates and mixer inner diameters.
Figure 3. Schematics of hydrodynamically stable flow patterns of a gas
(white color) and liquid (blue color) mixture in a horizontal round
tubing. The notations and images of the flow were based on ref 22.
The air and liquid in slug flow can swap places depending on the
affinity of the liquid for the material on the inner surface of the tube;
for example, a water slug with a strongly hydrophobic surface would
appear as shown, with convex/roundish water slugs moving through
the tube. For a strongly hydrophilic surface with water as the liquid,
the air slugs would be convex. Regardless of surface affinities, the slugs
become asymmetric at sufficiently high velocities, as observed in
Figure 10.
Crystal Growth & Design Article
dx.doi.org/10.1021/cg401715e | Cryst. Growth Des. 2014, 14, 851−860853
tube (aka fouling), eventually clogging the tube. In contrast, slug
flow has an internal circulation of liquid
11
(see video in Supporting
Information) that limits the time that liquid stays close to the tube
wall, which results in the fluid dynamics within each slug operating like
a small well-mixed crystallizer, with limited fouling.
Figure 4a shows an experimental demonstration of the spontaneous
formation of hydrodynamically stable slugs after streams of slurry and
air are passed through two branches of a T mixer that had a similar
inner diameter as the tubes. Stable slugs were maintained throughout
crystallization within each slug as it transports through the tubing to its
exit. Air was transferred to the mixer through the same peristaltic pump
with dual pump heads, with flow oscillation suppressed by offsetting
their rollers, and a silicone tube with filters (pore sizes 3 and 0.2 μm) to
prevent dust particles from getting into the tubes at the pump heads,
and from contacting slurry at the mixer, respectively. Within a large
hydrodynamically stable regime
16
for slugs in a 3.1 mm (inner
diameter) tube, the length ratio of liquid and gas slugs was adjustable by
specifying the flow rates (or pump rotation rates) between slurry and
air. A pump rpm ratio of 1:1 between hot solution and air was used
to generate stable liquid slugs of similar sizes with all aspect ratios of
about 1 (Figure 5a,b). A horizontal wrapping of tubing around a bucket
produced hydrodynamically stable slugs with no combination or
partitioning (Figures 4b and 5a,c).
2.3. Crystal Growth. The growth experimental conditions were
reported in Table 2. Again the radial mixing experiment was used to
justify the optimized design. The residence time after slug formation
was on the order of 5 min (to prevent generation of a large amount of
waste materials) in 15.2 m of transparent hydrophobic silicone tube
(Dow Corning Pharma-80 tubing, 3.1 mm inner diameter). While the
slugs were moving inside the silicone tube toward the exit, videos of
them were recorded under an in-line trinocular stereomicroscope
(microscope model #XV331AC20C, from Cyber Scientific Corp.;
camera model #DFK 22BUC03, from The Imaging Source, LLC) at
1 m before the exit. After the slurry slugs exited the end of tube, they
were collected one by one into polystyrene wells (1.5 cm diameter,
with the aqueous slugs covered with corn oil after collection to
suppress evaporation) for off-line imaging under the stereomicroscope.
Crystallizations are generally more poorly mixed for higher solids
densities. Since each slug operates as a small well-mixed batch
Figure 5. (a) Photograph of stable water slugs (aspect ratio about 1)
separated by air slugs (aspect ratio about 4) in packed silicone tube,
with black background to improve contrast. (b) Microscope image of
water slug (slug in the center) and parts of air slugs (dark regions at
both edges) inside a silicone tube. The flow direction is from left to
right. The front of the water slug is less curved toward the flow
direction than the back,
14
as indicated from the thickness of black
shades at water/air interfaces. At higher flow rates, the shape of the
water slug would be dominated by the velocity (e.g., the drag from
wall) rather than the hydrophobicity of the tubing material. (c)
Horizontal wrapping of silicone tube around a cylinder of the same
diameter. The photograph (a) is a close-up of (c) indicated by the
edge of the black background.
Figure 4. (a) Slug formation from streams of LAM slurry and air
through a T mixer. The white slugs contain LAM-aqueous slurry, and
the transparent slugs are air. (b) Slurry-containing slugs in the growth
stage. The white slugs have high solids densities in this photograph.
(c) Crystals in the funnel after filtration obtained under operations of
high solids density in each slug.
Crystal Growth & Design Article
dx.doi.org/10.1021/cg401715e | Cryst. Growth Des. 2014, 14, 851−860854
crystallizer, each slug will not be as well mixed for a high solids density,
in which case a fines dissolution step can be inserted into the crystal
growth stage to enlarge the final product crystals and reduce aggregation.
The end stage of the preliminary experiment had fines dissolution
implemented as a fast heating of the slurry slugs to 50 °C, followed by
going through 2 m of tubing in each of the two temperature baths at
39 and 22 °C, respectively, and then to room temperature (Figure 6).
The temperature in each bath was held constant using peristaltic pumps
(same model as before) and hot and cold water reservoirs (Figure 6),
controlled with Proportional-Integral controller tuned as in past
studies.
23
The temperature baths were not used in experiments with
coaxial and radial mixing, as the product crystals from those experiments
were uniformly large without fines dissolution.
3. RESULTS AND DISCUSSION
Below are results and discussion from the experimental
demonstrations of the slug-flow crystallizer that include (1) a
nonmonotonic spatial temperature profile to reduce aggrega-
tion and promote growth, (2) a comparison of coaxial and
radial mixers for in situ generation of seed crystals, and (3)
the improvement of hydrodynamics for reducing variations in
product quality.
3.1. Preliminary Experiment with Temperature Baths
to Reduce Aggregation and Promote Growth. After
crystal nucleation in a tube under laminar flow (no mixer, Re
number of about 150; see Figure 2a) and slug formation
(Figure 4a), the solids amount increases in the slugs as they
move along the tube length, indicated by the whiter color of
slurry slugs (Figure 4b, top) observed visually after a short
period of growth. Under high initial solute concentrations,
the slugs were able to carry 12.3 wt % crystals in the slurry,
with crystals nearly occupying the whole slug (Figure 4b) at a
reasonably fast flow rate (Table 1). The total yield (mass of
product crystals/total mass of solute) from this preliminary
experiment was 87%, close to the theoretical yield of a batch
cooling experiment from the same hot solution (87.5%).
The final product crystals after growth from laminar flow
nucleation were small and aggregated (Figure 7a) (LAM has a
strong tendency toward aggregation
20
). Heating followed by
Figure 7. Microscope (with polarizers) images of product crystals
from the preliminary experiment (after laminar flow nucleation and
growth in slugs) with (a) no heating; (b) heating to 50 °C followed by
natural cooling in the tubing; (c) heating to 50 °C followed by two
temperature baths of set temperatures (39 and 22 °C).
Table 2. Experimental Conditions for Slug Formation and
Growth by Cooling
a
experiment preliminary coaxial radial
pump for air stream peristaltic peristaltic peristaltic
pump rotation rate for air stream
(rpm)
66 78 53.4 (dual
head)
tubing for slug movement Tygon Tygon Pharma-80
tubing length (m) 15.2 15.2 15.2
residence time (min) 1.7 2.9 5.0
tube inner diameter (mm) 3.1 3.1 3.1
a
“Peristaltic” and “Tygon” are the same terms as in Table 1, and
“Pharma-80” refers to Dow Corning Pharma-80 tubing. Residence
time counts the time between when the hot solution reaches the mixer
and the exit of the tubing.
Figure 6. Photograph of the experimental setup with two temperature
baths. The hot reservoir is in a temperature-controlled bath (lower
right), and the cold reservoir is in an ice bucket at the lower left (not
shown in the picture). Each of the two temperature baths at the top of
the photograph is controlled with only one dual-head peristaltic pump
using Proportional-Integral controllers tuned by Internal Model
Control.
30,31
The dual head transfers both inlet and outlet water,
keeping the water level of the reservoir constant. The temperature of
the slugs in the tube increases or decreases relatively slowly due to the
low thermal conductivity of the tube wall. Only one temperature bath
is needed for coarse temperature control; additional temperature baths
enable finer tuning of the spatial temperature profile of the slugs as
they move through the tube.
Crystal Growth & Design Article
dx.doi.org/10.1021/cg401715e | Cryst. Growth Des. 2014, 14, 851−860855