1268 Research Article

Lukas Hohmann1
Design of a Continuous Tubular Cooling
Ramona Gorny2
Oliver Klaas1 Crystallizer for Process Development
Jonas Ahlert1
on Lab-Scale
Kerstin Wohlgemuth3
Norbert Kockmann1
Continuous manufacturing of fine chemical, life science, and pharmaceutical
1 products is under recent investigation in R&D. As cooling crystallization is an
TU Dortmund University,
important unit operation for purification of products, continuously operated, scal-
Department of Biochemical
and Chemical Engineering,
able devices are required for process development on lab-scale. A tubular crystal-
Laboratory of Equipment lizer was developed, based on the coiled flow inverter design. Experimental char-
Design, Dortmund, Germany. acterization proved a narrow residence time distribution of the liquid phase close
2
to ideal plug flow. Counter-current air cooling allows for adjusting linear and
TU Dortmund University, curved temperature profiles. Unseeded operation with the L-alanine (water)
Department of Biochemical
system demonstrated that nucleation has to be actively controlled to successfully
and Chemical Engineering,
apply intensified continuous cooling crystallization processes.
Laboratory of Solids Process
Engineering, Dortmund,
Keywords: Coiled flow inverter, Continuous cooling crystallization, Product purification,
Germany.
Tubular device
3
TU Dortmund University, Received: January 29, 2016; revised: May 02, 2016; accepted: May 09, 2016
Department of Biochemical
and Chemical Engineering, DOI: 10.1002/ceat.201600072
Laboratory of Plant and
Process Design, Dortmund,
Germany.

1 Introduction ious single and multiphase applications [15]. For downstream

operation, less toolbox or platform devices have been reported,
Concepts for continuous production of small lot size products, yet. Typically, syntheses of fine chemical, life science, and phar-
e.g., fine chemicals, life science products, and pharmaceuticals, maceutical products are carried out in the liquid phase [16].
recently rise interest [1]. While the manufacturing of those The products are often characterized by larger molecular
products is conventionally carried out in batch mode, continu- weight, a diversity of functional groups, and increased thermal
ous processing offers multiple advantages to improve product sensitivity compared to bulk chemicals [17]. Thus, crystalliza-
quality [2], cost efficiency [3], safety [4, 5], and to reduce time- tion from solution is an important downstream operation for
to-market [6]. In renowned research projects with the collabo- purification and product isolation, when, e.g., distillation is not
ration of both industry and academia innovative continuous applicable.
production concepts have been successfully demonstrated The aim of this work is to design a continuously operated
[7–9]. Nevertheless, product and process development are usu- tubular cooling crystallizer for lab-scale application with a total
ally carried out in batch mode which hinders a later shift to throughput of Fm1) ~ 10 g min–1. Hence, continuous crystalliza-
continuous operation [2]. Thus, well-characterized, scalable, tion processes can be demonstrated in a benchtop-size plant
continuous lab-scale equipment is required to demonstrate and with a total amount of product solution of a few kilograms.
optimize continuous production concepts early in process Furthermore, important characteristics of the crystallizer, such
development [2, 10]. as residence time distribution (RTD), particle fluidization, and
For upstream operation, flexible equipment toolboxes and temperature profile have to be analyzed to use experimental
platform solutions [11, 12] are available for process develop- results from lab-scale for process development and scale-up.
ment in laboratory and scale-up to production capacity The project focuses on the crystallization of small organic com-
[13, 14]. Due to their small dimensions and high specific sur- pounds from aqueous solution, in particular amino acids. A
face area, microreactors proved superior performance for var- binary mixture of L-alanine and water serves as a test system.
In this case, L-alanine crystallizes exclusively in an orthorhom-
bic morphology [18, 19]. Nevertheless, the system tends to
– agglomeration during batch cooling crystallization [20].
Correspondence: Prof. Norbert Kockmann (norbert.kockmann@bci.
tu-dortmund.de), TU Dortmund University, Department of Biochemi-
cal and Chemical Engineering, Laboratory of Equipment Design, Emil- –
Figge-Straße 68, 44227 Dortmund, Germany. 1) List of symbols at the end of the paper.

www.cet-journal.com ª 2016 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Chem. Eng. Technol. 2016, 39, No. 7, 1268–1280
 Research Article 1269

2 Design of the Tubular Crystallizer more robust operation without encrustation of the walls [26].
Ultrasound was successfully applied in a COBC setup to induce
Chen et al. reviewed the application of continuous crystalliza- primary nucleation [25]. A flexible adjustment of the axial tem-
tion in the pharmaceutical industry. They point out the advan- perature profiles, e.g., a linear profile, can be achieved by a
tages of continuous crystallization compared to batch crystalli- sophisticated control system [27] or multiple thermostats, e.g.,
zation, i.e., smaller equipment size, constant product quality 13 heating/cooling thermostats [26]. The COBC technology is
due to steady-state operation, and increasing yield by applying promising for pilot and production scale. However, scale-down
recycling strategies [21]. to smaller dimensions for process development has not been
Considering cooling crystallization, heat transfer is more and reported, yet. Batch oscillatory baffled crystallizers are used to
more limited by the reduced specific surface area when scaling- decrease material consumption during process development
up conventional batch cooling crystallizers [22]. This can limit [23, 25, 26].
the efficiency, when a certain crystallizing system and the Méndez del Rı́o et al. describe a milli-structured tube-in-tube
desired product quality would allow for faster cooling. Com- heat exchanger for cooling crystallization purposes. Rapid
paring continuous plug-flow (PF) crystallizers to continuous unseeded natural cooling of the binary systems paracetamol
mixed-suspension mixed-product removal crystallizers, PF (methanol) and paracetamol (ethanol) were carried out. A clear
equipment will obtain higher space-time yield and is therefore solution was gained at the end of the tubular section due to
extremely promising, if a high supersaturation level is tolerable. expansion of the metastable zone width (MSZW). Primary
However, temperature control is more difficult in tubular devi- nucleation was detected in the receiving vessel, leading to sig-
ces and settling/clogging of particles is an issue [21]. nificantly smaller crystals compared to batch experiments.
Some experimental parameter sets led to a rapid clogging of
the tubes [28].
2.1 State of the Art A continuously operated, seeded crystallizer based on heli-
cally coiled tubes was reported by Eder et al. for the acetylsali-
Various designs of continuous tubular PF equipment are cylic acid (ethanol) system. Seeding was achieved by injection
reported in literature. Design parameters, process conditions, of a seed suspension into the main flow from a second storage
and investigated test systems are summarized in Tab. 1. Further vessel. The seed suspension was prepared in a batch process
details are given in the following section. and the mean particle size was x50,3 ~ 100 mm [29]. Cooling of
A renowned example for PF crystallization is the continuous the tubular section was realized by one gas-cooled enclosure
oscillatory baffled crystallizer (COBC). In this macro-struc- box [29] or four cooling thermostats [30]. Unseeded operation
tured tubular device, particle fluidization and a narrow RTD led to blockage of the tube, whereas seeding and sufficiently
are achieved by a pulsed flow through baffles in the pipe, high flow velocity led to robust operation up to 15 min [29].
whereas the pulsation is achieved by a piston movement Wong et al. propose a method for continuous seed genera-
[23–25]. Seeded and unseeded operations [23] of various crys- tion via continuous contact secondary nucleation. Therefore, a
tallizing systems were reported. However, seeding enabled a parent crystal or tablet is fixed on a rotating shaft. Seed crystals

Table 1. Design parameters and test systems of various continuous tubular crystallizers from literature.

Equipment type di [mm] V [mL] Tube material Fm [g min–1] FV [mL min–1] Product (solvent) Ref.

COBC DN25 12000 Glass PVDF – 1000 API (solvent) [23]

COBC DN15 1250 Glass – 100–400 RTD study [24]

COBC DN65 ~ 12 000 Steel FEP – 50 RTD study; a-lactose [25]

monohydrate (water)

COBC DN15 5100 Glass 50 – L-glutamic acid (water) [26]

Tube-in-tube heat 1.6 15.3 PTFE – 10.8–47.2 Paracetamol (methanol), [28]

exchanger (ethanol)

Helically coiled tubes 2.0 47.0 Polysiloxane – 12.4–25.2 Acetyl salicylic acid (eth- [29]
anol)

Tubular growth section 3.2 49.0 Chem-Durance Bio – 11.3–14.2 Paracetamol (water) [31, 32]

Gas-liquid slug flow 2.0 85 Polysiloxane – 15.0 (L) Acetyl salicylic acid (eth- [33]
15.0 (G) anol, air)

Gas-liquid slug flow 3.1 115 Polysiloxane – 7.0–23.5 (L) L-asparagine monohy- [34, 35]
7.0–23.5 (G) drate (water, air)

Liquid-liquid slug flow 1.0 0.1–2.3 PFA – 0.4 (L1) Adipic acid (water, [36]
0.4 (L2) hexane)

Chem. Eng. Technol. 2016, 39, No. 7, 1268–1280 ª 2016 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.cet-journal.com
1270 Research Article

are abraded from the parent crystal and become entrained by – the temperature profile needs to be measurable and control-
supersaturated solution. The process was demonstrated for the lable with a reasonable equipment effort, especially in terms
glycine (water) and the paracetamol (water) systems. Seed crys- of thermostats
tals were obtained in the range of x50,0 ~ 4–8 mm for glycine and – the pressure drop needs to be less than 1 bar to enable robust
x50,0 ~ 6–13 mm for paracetamol. Seed crystals of paracetamol operation of peristaltic pumps.
were subsequently grown in a tubular growth section [31, 32]. For lab-scale operation with the test system L-alanine
In order to minimize axial dispersion [33, 34] and to prevent (water), the desired total mass flow rate for the crystallizer was
clogging of tubular devices [34], the application of a gas-liquid set to Fm = 5–10 g min–1 of solution or suspension. The design
slug flow is under investigation for continuous cooling crystal- case was further based on unseeded batch cooling crystalliza-
lization. Jiang et al. reported on a slug-flow device which was tion processes with linear cooling profiles [37, 38]. The corre-
used as a growth section for the L-asparagine monohydrate sponding physical properties and process conditions used for
(water) system. Primary nucleation was induced by rapid cool- the design are given in Tab. 2 and Tab. 3, respectively.
ing of a saturated solution in a tubular heat exchanger or by
mixing of two saturated solutions of different temperatures in Table 2. Physical properties of the aqueous L-alanine solution
various micromixers. Multiple combinations of heating to in the crystallizer.
achieve fines dissolution, natural cooling at room temperature,
and controlled cooling in cooling bathes were investigated in Property Value Ref.
the growth section [34]. –1
Mala [g mol ] 89.09
Another option for inducing nucleation in continuous oper-
–1
ation is the application of ultrasound [33, 35]. Eder et al. fed wala,sat(50 C) [galagsol ] 0.177 [37]
supersaturated solution of acetylsalicylic acid (ethanol) into a wala,sat(25 C) [galagsol ] –1
0.141 [37]
tubular coil and placed it in an ultrasonic bath. The seed crys- –3
tals of x50,3 ~ 20 mm were subsequently grown in gas-liquid rw(25 C) [kg m ] 997.05 [50]
slug flow by means of five thermostatic bathes to construct the rala,S(23 C) [kg m ] –3
1375 [51]
cooling section [33]. Jiang et al. applied an ultrasonic probe –3
rsol(25 C) [kg m ] 1054.43 [52]
which was pressed directly on the process tube, followed by a
tubular gas-liquid slug-flow growth section with natural cool- hw(25 C) [mPa s] 0.890 [50]
ing at room temperature [35]. An example of continuous cool-
hsol(25 C) [mPa s] 1.462 [53]
ing crystallization in small-scale tubular devices, utilizing a liq-
–1 –1
uid-liquid slug flow, was reported by Rossi et al. They cp,w(25 C) [J kg K ] 4182 [50]
employed the adipic acid (water) system with n-hexane as a
second liquid phase, investigating the primary nucleation The fluid properties at the outlet temperature of the crystallizer
mechanism [36]. were estimated from literature data for the inlet mass fraction
of L-alanine. As a worst-case assumption, the heat capacity of
The application of slug flow in continuous cooling crystalli-
pure water at outlet temperature was applied.
zation provides a narrow RTD and can assist particle transport.
Nevertheless, the tube volume has to be significantly increased
Table 3. Desired process conditions for the continuous cooling
to maintain a desired mean residence time as the second phase crystallization of L-alanine from water.
increases the total volumetric flow rate. Scale-up of slug flow is
an issue as known from microreactors [14]. Furthermore, the Fm [g min–1] Tin [C] Tout [C]
generation and splitting of the two fluid phases increases
equipment effort, and the influences of the second fluid phase 5.0 50.0 25.0
on solubility, crystallization kinetics, and product quality have 10.0 50.0 25.0
not been addressed, yet. These aspects need to be investigated
when comparing continuous processes to conventional small-
scale batch processes without another fluid phase. In order to design the tubular crystallizer, an equipment con-
cept is to choose, which fulfills the demands stated above. Var-
ious concepts for continuous seeding and induced primary nu-
2.2 Equipment Design Procedure cleation have been reviewed above, which can be connected to
a tubular device. Thus, the aspect of controlling nucleation can
Concluding the state of the art on continuous tubular cooling be located in a separate device. This work focuses on designing
crystallization, various demands on a tubular cooling crystal- a tubular cooling section for particle growth. Due to the low
lizer for process development on lab-scale can be stated: flow rates being desired for lab-scale application (see Tab. 3)
– a mean residence time of several minutes needs to be ob- and the solution properties (see Tab. 2), a turbulent flow profile
tained is hardly achievable. Klutz et al. recently reviewed tubular
– a narrow RTD close to ideal plug flow is required to enable equipment concepts for application in laminar flow with
batch-to-conti approaches respect to achieving a narrow RTD. Considering a low manu-
– particle fluidization and transport have to be facilitated, facturing effort and a flexible, segmented design, helically
while the application of a second fluid phase is avoided in coiled tubes, especially in the coiled flow inverter (CFI) design,
this project appeared most promising [39].

www.cet-journal.com ª 2016 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Chem. Eng. Technol. 2016, 39, No. 7, 1268–1280
 Research Article 1271

Helically coiled tubes can be manufactured by coiling flexible tical straight tube which is flown through contrarily to the
tubes, e.g., polymeric tubes, on a supporting structure. Due to direction of the gravitational force by a suspension. This
the flow on the circular pathway, centrifugal forces act on the reflects the upward flow in a tubular helix, whereas the coil is
fluid, shifting the maximum flow velocity of the laminar flow directed horizontally. Particles being larger than the calculated
profile to the outer side of the wall. Thereby, Dean vortices are cut-size diameter xcut will sediment in the flow, whereas smaller
induced, which can significantly intensify radial mixing [40]. particles will be entrained. The model takes the tube geometry,
This effect can be further enhanced using the CFI design. By the operating conditions, and physical properties into account.
bending the tubular helix after a certain number of turns with The gradients of the axial flow velocity in radial direction in
a 90 angle, Dean vortices have to be reformed and change the laminar flow are neglected. Furthermore, the model is desig-
direction [40]. The Reynolds number (Eq. (1)), the Dean num- nated for spherical particles and particle Reynolds numbers
ber (Eq. (2)), the modified torsion parameter (Eq. (3)), and fur- ReP < 1. The effect of Dean vortices is not incorporated in the
ther design guidelines taking the bend number and the number model but there is evidence that Dean vortices can significantly
of turns per coil into consideration can be applied to achieve a enhance particle fluidization compared to devices without a
narrow RTD in laminar flow [39, 41]. Here, di, hsol, dc, and p Dean vortices flow [45]. Despite the limitations, the model can
denote the inner diameter of the tube, the viscosity of the solu- represent a short-cut method to estimate the allowable particle
tion, the coil diameter, and the coil pitch distance, respectively. size of a tubular crystallizer to avoid sedimentation and clog-
4 Fm ging from a fluid-dynamic perspective.
Re ¼ (1) For dimensioning the tubular cooling crystallizer, the inter-
p di hsol
sffiffiffiffiffi nal volume and the mean residence time are directly associated
di with the desired cooling rate of the crystallization process. If
Dn ¼ Re (2) the temperature profile of a batch cooling crystallization pro-
dc
cess is linear, the cooling rate rb can be defined by Eq. (5),
pdc where T and t denote the temperature and the time, respective-
T* ¼ Re (3) ly. For a continuous tubular crystallizer with a linear axial tem-
p
perature profile, the cooling rate rc can be defined analogously
The fluidization of crystals is crucial for suspension crystalli- by Eq. (6), where t, u, l, and rsol denote the mean residence
zation purposes but the fundamentals of solid-liquid flow in time, the mean flow velocity, the tube length, and the density of
small-scale tubular devices are still poorly understood. Consid- the solution, respectively.
ering a crystal suspension with significantly lower crystal sizes
compared to the tube diameter of the crystallizer, various phe- T0  T
rb ¼ (5)
nomena can lead to settling of particles or scaling of the tube t  t0
surface which can further lead to unsteady operation or even
Tin  Tout u F 4
clogging of the device. rc ¼ ¼ ðTin  Tout Þ ¼ ðTin  Tout Þ m 2 (6)
Chen et al. recently reviewed scaling mechanisms in micro- t l rsol di l p
reactors and commercial tubing with respect to precipitation of
by-products during chemical syntheses. Some combinations of Heat transfer is often limited in stirred-tank batch crystalli-
particles and wall materials might be generally unfavorable due zers. Thus, batch cooling rates in the range of rb £ 0.45 K min–1
to strong particle-to-surface attraction, i.e., adsorption, adhe- have been investigated in reference studies for the L-alanine
sion due to surface roughness or heterogeneous nucleation of (water) system [20, 37, 38]. As process intensification and
crystals on the wall surface. If this is not the case, mechanical increased efficiency are main advantages of continuous crystal-
entrapment of particles due to inertial impaction was identified lization processes in tubular devices [21, 22] the desired cooling
to be a major reason for scaling. Unsteady changes in the tube rate is set to rc ~ 1.5–3.0 K min–1 for the prototype being intro-
diameter or constrictions are prone to this particle deposition duced in this work. When the dimensions of the tubular crys-
mechanism [42]. tallizer are once fixed, the cooling rate can be varied by the
Moreover, settling of particles in the gravitational field is a mass flow rate Fm (see Eq. (6)), providing that a linear cooling
well-known phenomenon in macroscale heat exchangers when profile and the desired outlet temperature can be maintained
being operated at nearly stagnant conditions [43]. A simplified by adjusting the cooling agent parameters. Thus, the mean resi-
Stokes flow model was proposed to estimate particle sizes dence time for the prototype is set to t ~ 17.0 min for a mass
which are not affected by gravitational settling and remain flu- flow rate of Fm = 5 g min–1 and t ~ 8.5 min for a mass flow rate
idized in a helically coiled tubular device with Dean vortices of Fm = 10 g min–1, respectively. This results in an internal vol-
flow [44]; see Eq. (4). ume of V ~ 81 mL. In order to achieve a lower linear cooling
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rate at constant mass flow rate, the mean residence time has to
72 hsol Fm 4 Fm xcut be increased, e.g., by expanding the tube length at constant
xcut ¼ with ReP ¼ <1
p ðrS  rsol Þg rsol di2 p hsol di di tube diameter by adding helical segments.
(4) Combining these information, a CFI design-space diagram
according to Klutz et al. [39] can be elaborated (see Fig. 1). This
Therein, xcut, rS, g, and ReP denote the cut-size diameter, the diagram spans an area of feasible designs, i.e., inner tube dia-
density of the particles, the gravitational constant, and the par- meter di and diameter of the supporting structure dct, which
ticle Reynolds number, respectively. The model describes a ver- provide sufficiently high Dean numbers, modified torsion

Chem. Eng. Technol. 2016, 39, No. 7, 1268–1280 ª 2016 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.cet-journal.com
1272 Research Article

parameters, and a sufficiently low pressure drop on the inner

tube side. For realization, standard tube sizes have to be consid-
ered, e.g., for polymeric tube material. The design space dia-
gram is applicable for larger flow rates as well. Therefore, scale-
up to pilot or production scale can be facilitated while flow
characteristics, e.g., RTD, can be maintained [12, 39].
A compact design with nseg = 9 helical tube segments,
nbends = 8 bends with an angle of 90, and a calculated volume
of V ~ 82 mL was developed and constructed. The resulting
design parameters are summarized in Tab. 4. The zick-zack
design [41] was used as illustrated in Fig. 2. For an unrestricted
flow of the cooling agent around the process tube, the support-
ing structure was realized by plates with milled holders for a
reproducible coiling of the process tube. Each plate carries a
complete helix with nturns = 5 turns and forms a repeating ele-
ment or segment. Therefore, the mean residence time of the
crystallizer can be varied at constant flow conditions by adding
or removing segments. The inlet part of the crystallizer was
made from a straight tube of the length of one turn. The transi-
tion part between two helical coils was made from a straight
Figure 1. CFI design space diagram according to [39] for tube of the length of half a turn. The outlet part of the crystalli-
Fm = 5 g min–1 and Fm = 10 g min–1 and a mean residence time of zer is another straight tube with a length of 250 mm.
t = 17.0 and t = 8.5 min, respectively. The pressure drop in the The calculated cut size diameter of the L-alanine (water) sys-
CFI is estimated by use of the correlation of Kumar et al. with a
tem in the tubular crystallizer ranges from xcut = 229–324 mm
friction factor of fCFI = 64Re–1(1 + 0.0456Dn0.603) [54].

Figure 2. Design details of the tubular cooling crystallizer. (a) Complete setup with nine segments in the zick-zack design [41], straight
tube at the inlet (left upper side), straight tube transition parts, straight tube at the outlet (right lower side). (b) Sketch of a single plate of
the supporting structure with milled holders for the tube. (c) Supporting structure with coiled tube and plug-in connection for baffles in
the bend region. (d) Fixation of thermocouple (Type K) on the outer side of the tube and PE foam insulation covering the thermocouple.

www.cet-journal.com ª 2016 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Chem. Eng. Technol. 2016, 39, No. 7, 1268–1280
 Research Article 1273

Table 4. CFI design parameters for the tubular cooling crystalli- tion under the given conditions (see Tab. 3) and at maximum
zer. yield would not exceed some mW and can be neglected.

Parameter Value

di [mm] 4.0 3 Materials and Experimental Methods

do [mm] 6.0
3.1 Materials
p [mm] 8.0
For cooling crystallization experiments, L-alanine ( ‡ 99 %,
dct [mm] 35.0
VWR Chemicals) was used. Deionized water (0.34 MW cm)
dc [mm] 41.0 served as solvent and pressurized air (3 barg) as cooling gas.
Both of them were obtained on-site. Sodium thiosulfate
nturns [–] 5
(Na2S2O3, ‡ 99 %, AppliChem) was used as a tracer substance
nseg [–] 9 for RTD characterization.
nbends [–] 8

l [m] 6.5 3.2 Experimental Setup

The complete experimental setup was placed on a mobile lab

according to Eq. (4) for the mass flow rates ranging from bench (see Fig. 3). A flow diagram is provided in Fig. 4. The
Fm = 5–10 g min–1. tubular crystallizer was made from transparent polyvinyl
The cooling strategy for the tubular cooling crystallizer is chloride (PVC) tubing (Nalgene, food grade, di = 4 mm,
based on the principle of a counter-current heat exchanger do = 6 mm, Thermo Scientific Inc.). The supporting structure
with a gaseous cooling agent. The CFI structure is placed in a was made from PVC and the jacket pipe is made from
jacket tube with Di = 100 mm, Do = 110 mm and l = 0.7 m. The poly(methyl methacrylate) (Mechanical Workshops of the BCI
gas flow is directed along the coils by means of baffles at each Department, TU Dortmund University).
bend. The baffles are removable with a plug-in connection in Four resistance thermometers for gas temperature measure-
order to simplify coiling of the tubes during construction and ment (Pt100, d = 1.0 mm, ± 0.1 K, Rössel Messtechnik GmbH)
reconfiguration of the device (see Fig. 2). Similar heat capacity were placed directly into the gas flow in the bend region and
flows FV r cp are applied both on the product side and the utili- sealed with the jacket pipe. One resistance thermometer was
ty side. Therefore, the cooling agent heats up in a similar range used to measure the temperature of the ambience. Five thermo-
as the product solution/suspension is cooled down, which couples for measuring the solution temperature (Type K,
allows for approximately linear cooling. Due to the coil struc- d = 0.5 mm, ± 0.4 %, OMEGA Engineering GmbH) were fixed
ture of the CFI, the outer volume is
significantly larger than the inter-
nal volume of the tube. Utilizing a
liquid cooling agent, e.g., an aque-
ous solution, would result in a long
start-up procedure for the utility
side when the heat capacity flow is
maintained. Precooled air at atmo-
spheric pressure provides signifi-
cantly lower density and lower heat
capacity compared to a liquid cool-
ing agent. Hence, a higher volumet-
ric flow rate of the gaseous cooling
agent is required to achieve the
desired heat capacity flow, which
drastically decreases the start-up
time of the utility side.
The total heat flux that has to be
removed from the product at
Fm = 10 g min–1 is approximately
Q = 17.4 W (see Tab. 2 and Tab. 3).
The crystallization of L-alanine
from saturated aqueous solution is
slightly exothermal, as the enthalpy
of solution is DhSL = + 0.75 kJ kg–1
[46]. Thus, the heat of crystalliza- Figure 3. Experimental setup on mobile lab bench.

Chem. Eng. Technol. 2016, 39, No. 7, 1268–1280 ª 2016 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.cet-journal.com
1274 Research Article

multiple times along an experiment for

time intervals of 5 min, each. The volumet-
ric flow rate of the cooling gas was con-
trolled manually by means of a rotameter
(102-16-N, FV = 0–10 L min–1, ± 0.1 L min–1,
Aalborg Instruments and Controls, Inc.)
with needle valve (CVTM, Aalborg Instru-
ments and Controls, Inc.).
The tracer concentration in RTD experi-
ments was measured via UV/VIS spectros-
copy (Evolution 201, Thermo Scientific
Inc.) equipped with a flow-through cell
(QS170, V = 12.4 mL, d = 0.2 mm, Hellma
GmbH). For MSZW experiments a batch
vessel with jacket (V = 100 mL, glass), one
of the magnetic stirrer plates, the heating/
cooling thermostat, and one of the resis-
tance thermometer was used. Particles
from continuous experiments were ana-
lyzed in suspension immediately (t < 30 s)
by means of a light microscope (Advance
ICD, Bresser GmbH) equipped with a digi-
tal reflex camera (D7000, Nikon GmbH)
after samples (m ~ 1 g of suspension) were
drawn.

3.3 Residence Time Distribution

The residence time distribution (RTD) of

the tubular crystallizer was characterized
via step response experiments according to
the method described by Kurt et al. [41] at
room temperature. The RTD studies were
carried out in single-phase operation. The
mass fraction of the tracer substance was
Figure 4. Flow diagram of experimental setup. wT = 475 ·10–6 gTgsol–1 in aqueous solution.
A wavelength of 219 nm was applied.
on the outside of the tube and insulated by means of conven- The step response F(q) was characterized for both the low
tional PE polymeric foam (Climapor, s = 13 mm, Saarpor pulsating peristaltic pump at Fm = 3 and 10 g min–1 and the
Klaus Eckhardt GmbH) for tube insulation (see Fig. 2). This less pulsating syringe pump at Fm = 1, 3, 5, and 10 g min–1 as a
allows for a non-invasive measurement of the solution temper- benchmark. Each experiment was reproduced three times. The
ature with an error < 30 % compared to the temperature differ- volume V of the tubular crystallizer was determined gravimetri-
ence between solution and surrounding cooling gas. The inlet cally by means of the peristaltic pump, replacing water (density
temperature of the cooling gas was kept constant by means of a rw, see Tab. 2) with air. Therefore, the mean residence time t and
stainless-steel heat exchanger, equipped with a heating/cooling the dimensionless time q can be calculated by Eq. (7) and Eq. (8),
thermostat (Ministat 125, Huber GmbH) and feedback control. respectively. The dispersion model was fitted to experimental da-
Two storage vessels with jacket (V = 1000 mL, glass), equipped ta [39, 41, 47] in order to determine the Bodenstein number Bo,
with a magnetic stirrer (MR 3001, Heidolph Instruments GmbH which is defined by Eq. (9). Here, Dax denotes the axial disper-
& Co. KG), and heating thermostat (304 CC, Huber GmbH) were sion coefficient. A Bodenstein number Bo > 100 indicates an
used to prepare and supply the product solution and solvent. A RTD close to ideal plug flow [47].
low pulsating peristaltic pump (Ismatec REGLO Digital Vrw
MS-4/12, 12 rollers, Cole-Parmer GmbH) equipped with t¼ (7)
Fm
Tygon 3607 tubing (di = 2.06 mm, do = 3.78 mm, Compagnie
de Saint Gobain) served as feed pump. A less pulsating syringe t
pump (SyrDos2, HiTec Zang GmbH) was taken for RTD char- q¼ (8)
t
acterization. The mass flow rates of both pumps were measured
gravimetrically (NewClassic MF, ± 1 mg, Mettler-Toledo Int. ul
Bo ¼ (9)
Inc.). Therefore, the product solution was captured in a beaker Dax

www.cet-journal.com ª 2016 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Chem. Eng. Technol. 2016, 39, No. 7, 1268–1280
 Research Article 1275

3.4 Metastable Zone Width T0 = 50.0 C before start-up to ensure operation of the crystalli-
zer in the supersaturated regime. In the latter experiments, the
The metastable zone width (MSZW) of the system L-alanine temperature in the storage vessels was set to T = T0 = 72.0 C
(water) was characterized in batch experiments for linear cool- for both dissolution and continuous operation in order to com-
ing rates of rb = 0.5 and 0.9 K min–1, using the polythermal pensate heat losses in the peristaltic pump, valves, and connect-
method [38, 48]. The revolution speed of the magnetic stirrer ing tubes, while shifting the crossing of the solubility line into
was kept constant at n = 660 min–1. The cooling rate was the device.
adjusted using the feedback control mode of the heating/cool- For the start-up of the device, cooling gas was expanded to
ing thermostat. Each MSZW measurement was reproduced atmospheric pressure and room temperature by means of the
three times. needle valve. A constant volumetric flow rate of the cooling gas
For each experiment, a solution of msol = 48.023 ± 0.003 g was was adjusted as it was desired for the experiment. The gas was
prepared with a mass fraction of wsat(50.0 C) = 0.177 galagsol–1. cooled afterwards in order to maintain the desired inlet tem-
The solution was heated up to 55.0 C and kept at this tempera- perature. Preheated water from the second storage vessel was
ture for at least 1 h. Thereafter, the linear cooling rate was flushed through the tubular crystallizer at a constant mass flow
initialized. The appearance of the first crystal at the tempera- rate to elaborate the axial temperature profile. The elaboration
ture Tnuc was detected with the naked eye. Thus, the metastable of the solution and the cooling gas temperature profiles were
zone width DTMSZW was calculated by Eq. (10). observed via thermocouples and resistance thermometers,
respectively. When a steady-state temperature profile had been
DTMSZW ¼ Tnuc  Tsat (10) established, the inlet of the tubular crystallizer (valve V1, see
Fig. 4) was changed to aqueous L-alanine solution, while the
mass flow rate was kept constant. The product solution/sus-
pension was analyzed via light microscopy. ImageJ (National
3.5 Continuous Cooling Crystallization Institute of Health) was employed for image analysis. For the
shut-down procedure, the flow of the cooling gas was stopped
Three experiments on continuous cooling crystallization of and water was flushed through the crystallizer at an elevated
L-alanine from water were carried out under different condi- flow rate.
tions (see Tab. 5). For preparing the crystallization process,
both of the storage vessels were filled with deionized water.
Crystalline L-alanine was added to one of them, thus a mass 4 Results and Discussion
fraction of wsat(50.0 C) = 0.177 galagsol–1 was achieved. In total,
a volume of 1 L of aqueous solution was prepared. In the first 4.1 Residence Time Distribution
experiment, the stirred storage vessels were heated up to
T = 55.0 C by means of the heating thermostat in order to dis- A volume V = 78.7 ± 1.4 mL of the crystallizer was determined
solve all crystalline matter in the product supply vessel and to by a gravimetric method. The results of the RTD characteriza-
heat up water in the solvent supply vessel for the start-up pro- tion are summarized in Tab. 6. A Bodenstein number Bo > 100
cedure. After dissolving the crystals the temperature was main- was achieved in all of the experiments in good agreement with
tained for at least 1 h. The temperature was then reduced to the dispersion model (see Fig. 5). Therefore, an ideal plug flow

Table 5. Experimental conditions for the continuous cooling crystallization experiments.

Exp. Fm [g min–1] T0 [C] t [min] Re [–] Dn [–] T* [–] xcut [mm] FV,G [L min–1]

1 8.80 ± 0.14 50.0 8.93 ± 0.14 31.92 ± 0.51 9.97 ± 0.16 513.9 ± 8.2 304 10.0

2 4.95 ± 0.04 72.0 15.87 ± 0.12 17.95 ± 0.14 5.61 ± 0.04 289.0 ± 2.2 228 3.0

3 2.34 ± 0.01 72.0 33.49 ± 0.11 8.51 ± 0.03 2.66 ± 0.01 137.0 ± 0.5 157 10.0

Table 6. Experimental conditions and results for RTD characterization of the tubular crystallizer.

Pump type Fm [g min–1] t [min] Re [–] Dn [–] T* [–] Bo [–]

Syringe 0.98 ± 0.02 79.75 ± 1.34 5.87 ± 0.10 1.83 ± 0.03 94.5 ± 1.6 170.5 ± 2.3

Syringe 3.01 ± 0.02 26.09 ± 0.17 17.94 ± 0.12 5.60 ± 0.04 288.9 ± 1.9 165.6 ± 2.2

Syringe 4.92 ± 0.03 15.98 ± 0.10 29.30 ± 0.18 9.15 ± 0.06 471.7 ± 2.9 147.5 ± 4.3

Syringe 9.94 ± 0.08 7.91 ± 0.03 55.00 ± 0.45 17.18 ± 0.14 885.6 ± 7.2 113.1 ± 2.6

Peristaltic 3.07 ± 0.04 25.56 ± 0.37 18.31 ± 0.27 5.72 ± 0.08 294.8 ± 4.3 145.0 ± 10.6

Peristaltic 9.76 ± 0.10 8.06 ± 0.09 54.02 ± 0.57 16.87 ± 0.18 869.7 ± 1.2 110.2 ± 2.4

Chem. Eng. Technol. 2016, 39, No. 7, 1268–1280 ª 2016 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.cet-journal.com
1276 Research Article

Figure 5. Results of the RTD characterization experiments. (a) Experimental step response curves for the syringe pump at different Rey-
nolds numbers. (b) Comparison of experimental step response curves with fitted dispersion model for both the syringe pump and the
peristaltic pump at comparable Reynolds numbers.

behavior can be assumed for the tubular crystallizer setup, 4.2 Metastable Zone Width
which indicates a homogeneous, radial concentration and tem-
perature distribution in the liquid phase. The results of the batch investigation indicate an increase of
By increasing the mass flow rate, Reynolds number, Dean the MSZW at elevated linear cooling rates (see Tab. 7). Even
number, and modified torsion parameter, the Bodenstein num- though the MSZW increased, the time interval tMSZW–tsat was
ber was slightly decreased and the RTD of the liquid phase was decreased due to faster cooling. These trends are in good agree-
broadened (see Tab. 6). This result can be attributed to higher ment with measurements of Wohlgemuth at lower linear cool-
axial dispersion due to laminar convection [47]. Comparing ing rate [38]. Higher cooling rates could not be investigated in
results of both the syringe pump and peristaltic pump batch experiments due to a limited cooling capacity of the uti-
(see Fig. 5), a narrower RTD was reached by the syringe pump. lized heating/cooling thermostat.
Even though the pulsation of the peristaltic pump with its
12 rollers was low, it slightly increased axial back-mixing.
For a crystallization application it has to be taken into 4.3 Continuous Cooling Crystallization
account that the RTD of the liquid phase might be affected by
the suspended crystals. Moreover, the RTD of the solid phase The steady-state temperature profiles for all experiments are
might be broader, when large crystals tend to deposit and move given in Fig. 6. An approximately linear axial temperature
slower through the tube than smaller crystals and the liquid profile was reached for the product mass flow rates of the solu-
phase. tion of Fm = 8.80 and 4.95 g min–1. For the lower flow rate of

Table 7. Results of the determination of the metastable zone width (MSZW) of the binary system L-alanine (water) from batch experi-
ments.

rb [K min–1] n [min–1] T0 [C] Tsat [C] DTMSZW [K] tMSZW–tsat [min] Ref.

0.15 200 60 50 6.89 45.9 [38]

0.30 200 60 50 7.70 ± 0.59 25.7 ± 2.0 [38]

0.45 200 60 50 7.86 17.5 [38]

0.50 660 55 50 9.79 ± 0.48 19.6 ± 1.0 This work

0.90 660 55 50 11.36 ± 0.39 12.6 ± 0.4 This work

www.cet-journal.com ª 2016 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Chem. Eng. Technol. 2016, 39, No. 7, 1268–1280
 Research Article 1277

Fm = 2.34 g min–1 a curved convex temperature profile was

achieved. The resulting linearized cooling rates for continuous
operation are summarized in Tab. 8.
The first experiment was operated for more than 1 h up to
complete consumption of the storage volume. The resulting
product was a homogeneous product solution. No particles
were detected in the outlet via light microscopy. Nucleation
occurred afterwards in the product beaker, where temperature
and residence time were not defined. In the second experiment
the flow rates of both the solution and the cooling gas were
reduced, whereas the solution inlet temperature was increased.
The cooling rate and the outlet temperature were similar to the
first experiment. Again a homogeneous solution was obtained
in the outlet up to complete consumption of the storage vol-
ume after more than 2 h. In both continuous experiments the
outlet temperatures of the tubular crystallizer were lower com-
pared to the cloud point temperatures measured in the batch
experiments, while higher cooling rates rc were applied,
see Tab. 7 and Tab. 8. Thus, the MSZW was further increased
in the continuous experiment.
This result might be attributed to the dependency of the
MSZW on the cooling rate, which was observed in the batch
experiments, see Sect. 4.2. Nevertheless, further parameters can
influence primary nucleation, such as mixing intensity, shear
rate, temperature, and holding time in the undersaturated
regime, concentration of soluble or insoluble impurities, prop-
erties of the wall material [49], and the absence of a gas-liquid
surface in the continuous experiments in contrast to the batch
experiments [38]. This highlights the fact that primary nuclea-
tion processes are difficult to transfer from a stirred batch ves-
sel to continuous PF operation.
In the third experiment, the mass flow rate of the solution
was further reduced to increase the mean residence time. Fur-
thermore, the volumetric flow rate and the inlet temperature of
the cooling gas were adjusted in order to minimize the outlet
temperature of the solution. Approximately two mean resi-
dence times after start-up, small crystals with a narrow size dis-
tribution were obtained in the product stream (see Fig. 7).
Subsequently, agglomerates were observed in the product sus-
pension. As the cut size diameter was estimated xcut = 157 mm
for this experiment (see Tab. 5), large agglomerates could no
longer be fluidized at the low flow velocity and sedimented in
the tube. This led to clogging of the tube 2 h after switching
from water to L-alanine solution.
These results indicate that varying the mean residence time
by changing the mass flow rate of the solution to shift the cloud
Figure 6. Solution (L) and cooling gas (G) temperature profiles point into the tubular section is limited, as important perfor-
during continuous cooling crystallization experiments; error mance indicators, e.g., particle transport, are directly depend-
bars show uncertainties due to non-invasive temperature mea- ing on the mass flow rate, as well.
surement on the liquid side. (a) Experiment 1; (b) experiment 2;
(c) experiment 3.

Table 8. Resulting cooling rates in the continuous cooling crystallization experiments.

Exp. Fm [g min–1] u [m min–1] rax [K m–1] rc [K min–1] FV,G [L min–1]

1 8.80 ± 0.14 0.664 ± 0.011 2.12 1.409 ± 0.022 10.0

2 4.95 ± 0.04 0.373 ± 0.003 4.05 1.673 ± 0.013 3.0

3 2.34 ± 0.01 0.177 ± 0.001 5.87 1.039 ± 0.003 10.0

Chem. Eng. Technol. 2016, 39, No. 7, 1268–1280 ª 2016 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.cet-journal.com
1278 Research Article

degree of freedom for choosing operating condi-

tions, compared to conventional batch setups.
Future work focuses on continuous seed genera-
tion in a device separated from the tubular growth
section and on determining the operating range of
the continuously operated tubular crystallizer
according to particle fluidization and clogging. Fur-
thermore, the heat transfer characteristics of the
device will be investigated by means of a simulation
study basing on experimental data.

Acknowledgment
The authors gratefully acknowledge Andreas Grun-
wald for manufacturing the tubular crystallizer.
Carsten Schrömges and Martin Matuschek are
acknowledged for their technical and experimental
support. The German Federal Ministry of Eco-
nomic Affairs and Energy (BMWi) is acknowledged
for financial support of the research leading to this
Figure 7. Particle characteristics in the product crystal suspension of experi- work as part of the ENPRO initiative, project num-
ment 3. (a) Light microscopy image of product crystals 1 h after switching the
ber: 03ET1254D.
feed from water to aqueous L-alanine solution. (b) Corresponding particle size
distribution to (a). (c) Light microscopy image of product crystals 2 h after
switching the feed from water to aqueous L-alanine solution. The authors have declared no conflict of interest.

5 Conclusions Symbols used

A versatile milli-structured tubular device for developing con- Bo [–] Bodenstein number
tinuous cooling crystallization processes on lab-scale was cp [J kg–1K–1] specific heat capacity
designed, constructed, and characterized. The design was d [m] process tube diameter
based on bended helically coiled tubes, following the design dc [m] coil diameter
concept of a coiled flow inverter (CFI). Due to established dct [m] diameter of a supporting structure
design guidelines for CFI devices, a scale-up from lab to pilot D [m] jacket tube diameter
or production scale can be facilitated while important flow Dax [m2s–1] axial dispersion coefficient
characteristics, e.g., a narrow RTD, can be maintained. The Dn [–] Dean number
CFI tubular crystallizer provides a narrow RTD close to ideal fCFI [–] tube-side friction factor of CFI
plug flow. Linear axial cooling profiles can be obtained in Fm [kg s–1] mass flow rate
steady-state operation by means of counter-current cooling FV [m3s–1] volumetric flow rate
with a gaseous cooling agent. The cooling gas is supplied by F(q) [–] step response
means of a single heating/cooling thermostat with feedback g [m s–2] gravitational constant
control. Therefore, the equipment effort is comparable to DhSL [kJ kg–1] enthalpy of solution
typical batch setups on lab-scale. l [m] tube length
Cooling crystallization experiments with the system L-ala- m [kg] mass
nine (water) were carried out both in batch and in continuous M [kg mol–1] molar mass
mode at elevated linear cooling rates. While the metastable n [s–1] revolution speed
zone width (MSZW) was characterized in the batch mode, no nbends [–] number of 90 bends in a CFI
primary nucleation was detected within the continuous crystal- nseg [–] number of helical segments in a CFI
lizer. The system was driven into primary nucleation at con- nturns [–] number for turns per helical coil
stant tube volume by reducing the feed flow rate and thus p [m] coil pitch distance
increasing the mean residence time. Unstable operation was Dp [bar] tube side pressure drop
achieved, which led to formation of agglomerates and subse- r [K s–1] process cooling rate
quently to clogging of the device. This underlines that impor- rax [K m–1] axial cooling rate
tant operation characteristics of the continuously operated Re [–] Reynolds number
tubular crystallizer, such as mean residence time, RTD, and ReP [–] particle Reynolds number
particle fluidization are directly depending on the mean flow t [s] time
rate of the product solution/suspension, resulting in a reduced T [C] temperature

www.cet-journal.com ª 2016 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Chem. Eng. Technol. 2016, 39, No. 7, 1268–1280
 Research Article 1279

T* [–] modified torsion parameter [3] T. Seifert, S. Sievers, C. Bramsiepe, G. Schembecker, Chem.
DTMSZW [K] metastable zone width Eng. Process. 2012, 52, 140. DOI: 10.1016/j.cep.2011.10.007
u [m s–1] mean flow velocity [4] N. Kockmann, Chem. Ing. Tech. 2012, 84 (5), 715. DOI:
V [m3] volume 10.1002/cite.201100222
wk [kgkkgsol–1] mass fraction of component k [5] C. Fleischer, J. Wittmann, N. Kockmann, T. Bieringer,
xcut [m] cut size diameter C. Bramsiepe, Chem. Ing. Tech. 2015, 87 (9), 1258. DOI:
x50,j [m] mean particle diameter 10.1002/cite.201400188
(j = 0: number-base, j = 2: area-base, [6] L. Grundemann, M. Schoenitz, S. Scholl, Chem. Ing. Tech.
j = 3: volume-base) 2012, 84 (5), 685. DOI: 10.1002/cite.201100238
xF [m] maximum Ferret’s diameter [7] Bayer Technology Services GmbH, F3 Factory: Creating the
Future of Production, http://www.f3factory.com/scripts/
Greek letters pages/en/home.php, accessed 22 December 2015.
[8] S. Mascia, P. L. Heider, H. Zhang, R. Lakerveld, B. Benyahia,
h [Pa s] dynamic viscosity P. I. Barton, R. D. Braatz, C. L. Cooney, J. M. B. Evans, T. F.
p [–] pi number Jamison, K. F. Jensen, A. S. Myerson, B. L. Trout, Angew.
q [–] dimensionless time Chem. Int. Ed. 2013, 52 (47), 12359. DOI: 10.1002/anie.
r [kg m–3] density 201305429
t [s] mean residence time [9] M. D. Johnson, S. A. May, J. R. Calvin, J. Remacle, J. R. Stout,
W. D. Diseroad, N. Zaborenko, B. D. Haeberle, W.-M. Sun,
Sub- and superscripts M. T. Miller, J. Brennan, Org. Process Res. Dev. 2012, 16 (5),
0 initial condition 1017. DOI: 10.1021/op200362h
[10] C. Bramsiepe, N. Krasberg, C. Fleischer, L. Hohmann,
ala L-alanine
N. Kockmann, G. Schembecker, Chem. Ing. Tech. 2014,
b batch process
86 (7), 966. DOI: 10.1002/cite.201400029
c continuous process
[11] P. Plouffe, A. Macchi, D. M. Roberge, Org. Process Res. Dev.
G gaseous phase
2014, 18 (11), 1286. DOI: 10.1021/op5001918
i inner tube side
[12] N. Kockmann, ChemBioEng Rev. 2016, 3 (1), 5. DOI:
in inlet
10.1002/cben.201500025
L liquid phase [13] N. Kockmann, M. Gottsponer, D. M. Roberge, Chem. Eng. J.
nuc primary nucleation 2011, 167 (2–3), 718. DOI: 10.1016/j.cej.2010.08.089
o outer tube side [14] N. Kockmann, D. M. Roberge, Chem. Eng. Process. 2011,
out outlet 50 (10), 1017. DOI: 10.1016/j.cep.2011.05.021
sat property of the saturated solution [15] V. Hessel, D. Kralisch, N. Kockmann, T. Noël, Q. Wang,
sol solution ChemSusChem 2013, 6 (5), 746. DOI: 10.1002/cssc.2012
S solid phase 00766
T tracer substance [16] a) D. M. Roberge, L. Ducry, N. Bieler, P. Cretton, B. Zimmer-
w water mann, Chem. Eng. Technol. 2005, 28 (3), 318. DOI: 10.1002/
ceat.200407128; b) D. M. Roberge, Org. Proc. Res. Dev. 2004,
Abbreviations 8 (6), 1049. DOI: 10.1021/op.0400160
API active pharmaceutical ingredient [17] D. J. Kirwan, C. J. Orella, in Handbook of Industrial
CFI coiled flow inverter Crystallization, 2nd ed. (Ed: A. S. Myerson), Butterworth-
COBC continuous oscillatory baffled crystallizer Heinemann, Woburn, MA 2002, 249.
FEP fluorinated ethylene propylene [18] N. A. Tumanov, E. V. Boldyreva, B. A. Kolesov, A. V. Kurno-
MSZW metastable zone width sov, R. Quesada Cabrera, Acta Crystallogr., Sect. B: Struct.
PF continuous plug-flow crystallizer Sci., Cryst. Eng. Mater. 2010, 66 (4), 458. DOI: 10.1107/
PFA perfluoroalkoxy alkanes S010876811001983X
[19] I. Azuri, E. Meirzadeh, D. Ehre, S. R. Cohen, A. M. Rappe,
PTFE polytetrafluoroethylene
M. Lahav, I. Lubomirsky, L. Kronik, Angew. Chem. Int. Ed.
PVC polyvinyl chloride
2015, 54 (46), 13566. DOI: 10.1002/anie.201505813
PVDF polyvinylidene fluoride
[20] L.-M. Terdenge, S. Heisel, G. Schembecker, K. Wohlgemuth,
RTD residence time distribution
Chem. Eng. Sci. 2015, 133, 157. DOI: 10.1016/j.ces.2015.
03.002
[21] J. Chen, B. Sarma, J. M. B. Evans, A. S. Myerson, Cryst.
References Growth Des. 2011, 11 (4), 887. DOI: 10.1021/cg101556s
[22] X.-W. Ni, ACHEMA 2015, Frankfurt a. Main, June 2015.
[1] T. Bieringer, S. Buchholz, N. Kockmann, Chem. Eng. [23] S. Lawton, G. Steele, P. Shering, L. Zhao, I. Laird, X.-W. Ni,
Technol. 2013, 36 (6), 900. DOI: 10.1002/ceat.201200631 Org. Process Res. Dev. 2009, 13 (6), 1357. DOI: 10.1021/
[2] K. Plumb, Chem. Eng. Res. Des. 2005, 83 (6), 730. DOI: op900237x
10.1205/cherd.04359 [24] R. Kacker, S. I. Regensburg, H. J. M. Kramer, BIWIC 2015,
Deajeon, Korea, September 2015.

Chem. Eng. Technol. 2016, 39, No. 7, 1268–1280 ª 2016 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.cet-journal.com
1280 Research Article

[25] H. Siddique, C. J. Brown, I. Houson, A. J. Florence, Org. [39] S. Klutz, S. K. Kurt, M. Lobedann, N. Kockmann, Chem.
Process Res. Dev. 2015, 19 (12), 1871. DOI: 10.1021/acs.oprd. Eng. Res. Des. 2015, 95, 22. DOI: 10.1016/j.cherd.2015.
5b00127 01.003
[26] N. E. B. Briggs, U. Schacht, V. Raval, T. McGlone, J. Sefcik, [40] A. K. Saxena, K. D. P. Nigam, AIChE J. 1984, 30 (3), 363.
A. J. Florence, Org. Process Res. Dev. 2015, 19 (12), 1903. DOI: 10.1002/aic.690300303
DOI: 10.1021/acs.oprd.5b00206 [41] S. K. Kurt, M. G. Gelhausen, N. Kockmann, Chem. Eng.
[27] X.-W. Ni, I. Laird, A.-T. Liao, US Patent 2009/0304890 A1, Technol. 2015, 38 (7), 1122. DOI: 10.1002/ceat.201400515
2005. [42] Y. Chen, J. C. Sabio, R. L. Hartman, J. Flow Chem. 2015,
[28] J. R. Méndez del Rı́o, R. W. Rousseau, Cryst. Growth Des. 5 (3), 166. DOI: 10.1556/1846.2015.00001
2006, 6 (6), 1407. DOI: 10.1021/cg060025v [43] T. R. Bott, Fouling of Heat Exchangers, Elsevier, Amsterdam
[29] R. J. P. Eder, S. Radl, E. K. Schmitt, S. Innerhofer, M. Maier, 1995.
H. Gruber-Woelfler, J. G. Khinast, Cryst. Growth Des. 2010, [44] A. Van de Runstraat, P. Geerdink, E. L. V. Goetheer, WO
10 (5), 2247. DOI: 10.1021/cg9015788 2009/151322 A1, 2008.
[30] R. J. P. Eder, E. K. Schmitt, J. Grill, S. Radl, H. Gruber-Woelf- [45] H. Mallubhotla, G. Belfort, J. Membr. Sci. 1997, 125 (1), 75.
ler, J. G. Khinast, Cryst. Res. Technol. 2011, 46 (3), 227. DOI: DOI: 10.1016/S0376-7388(96)00125-1
10.1002/crat.201000634 [46] J. O. Hutchens, in Handbook of Biochemistry: Selected Data
[31] S. Y. Wong, Y. Cui, A. S. Myerson, Cryst. Growth Des. 2013, for Molecular Biology, 2nd ed.(Ed: H. A. Sober), The Chemi-
13 (6), 2514. DOI: 10.1021/cg4002303 cal Rubber Co., Cleveland, OH 1970, B67.
[32] A. S. Myerson, S. Y. Wong, US Patent 2014/0256984 A1, [47] O. Levenspiel, Chemical Reaction Engineering, 3rd ed., Wiley,
2014. New York 1999.
[33] R. J. P. Eder, S. Schrank, M. O. Besenhard, E. Roblegg, [48] J. Nývlt, The Kinetics of Industrial Crystallization, Elsevier,
H. Gruber-Woelfler, J. G. Khinast, Cryst. Growth Des. 2012, Amsterdam 1985.
12 (10), 4733. DOI: 10.1021/cg201567y [49] W. Beckmann, in Crystallization: Basic Concepts and Indus-
[34] M. Jiang, Z. Zhu, E. Jimenez, C. D. Papageorgiou, J. Waetzig, trial Applications (Ed: W. Beckmann), Wiley-VCH, Wein-
A. Hardy, M. Langston, R. D. Braatz, Cryst. Growth Des. heim 2013, 7.
2014, 14 (2), 851. DOI: 10.1021/cg401715e [50] W. Wagner, H.-J. Kretzschmar, in VDI-Wärmeatlas, Springer,
[35] M. Jiang, C. D. Papageorgiou, J. Waetzig, A. Hardy, M. Lang- Berlin 2013, 175.
ston, R. D. Braatz, Cryst. Growth Des. 2015, 15 (5), 2486. [51] R. Sullivan, M. Pyda, J. Pak, B. Wunderlich, J. R. Thompson,
DOI: 10.1021/acs.cgd.5b00263 R. Pagni, H. Pan, C. Barnes, P. Schwerdtfeger, R. Compton,
[36] D. Rossi, A. Gavriilidis, S. Kuhn, M. A. Candel, A. G. Jones, J. Phys. Chem. A 2003, 107 (34), 6674. DOI: 10.1021/
C. Price, L. Mazzei, Cryst. Growth Des. 2015, 15 (4), 1784. jp0225673
DOI: 10.1021/cg501836e [52] K. Zhuo, Q. Liu, Y. Wang, Q. Ren, J. Wang, J. Chem. Eng.
[37] K. Wohlgemuth, G. Schembecker, Comput. Chem. Eng. Data 2006, 51 (3), 919. DOI: 10.1021/je050412t
2013, 52, 216. DOI: 10.1016/j.compchemeng.2012.12.001 [53] T. Ogawa, K. Mizutani, M. Yasuda, Bull. Chem. Soc. Jpn.
[38] K. Wohlgemuth, Ph.D. Thesis, Technical University Dort- 1984, 57 (8), 2064. DOI: 10.1246/bcsj.57.2064
mund 2012. [54] V. Kumar, M. Mridha, A. K. Gupta, K. D. P. Nigam, Chem.
Eng. Sci. 2007, 62 (9), 2386. DOI: 10.1016/j.ces.2007.01.032

www.cet-journal.com ª 2016 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Chem. Eng. Technol. 2016, 39, No. 7, 1268–1280