Chemical Engineering and Processing 50 (2011) 1069–1075

Contents lists available at ScienceDirect

Chemical Engineering and Processing:

Process Intensification
journal homepage: www.elsevier.com/locate/cep

Pressure drop and mixing in single phase microreactors: Simplified designs of

micromixers
Craig P. Holvey a , Dominique M. Roberge b,∗ , Michael Gottsponer b , Norbert Kockmann c , Arturo Macchi a
a
Department of Chemical and Biological Engineering, University of Ottawa, Canada
b
Microreactor Technology, Lonza AG, 3930 Visp, Switzerland
c
Department of Biochemical and Chemical Engineering, TU Dortmund University, Germany

a r t i c l e i n f o a b s t r a c t

Article history: Continuous flow microreactors can greatly improve the safety and product yields of processes in the
Received 9 February 2011 pharmaceutical and fine chemical industry by overcoming many of the drawbacks of traditional batch
Received in revised form 22 May 2011 and semi-batch stirred reactors. This study compares on a common scale the pressure drop and mixing
Accepted 31 May 2011
performance of different size commercial microreactor plates composed of a tangential, SZ-shaped or
Available online 8 July 2011
caterpillar mixer followed by a rectangular serpentine main channel. The pressure drop was fitted to a
friction factor model, which suggests that the mixing zone had significant chaotic secondary flow patterns,
Keywords:
whereas the main channel did not. Moreover, the mixing zone was the main contributor to the overall
Microreactors
Pressure drop
pressure drop. Mixing performance was then characterized using competitive parallel reactions. Upon
Mixing the formation of chaotic secondary flows, typically due to the interactions of artificially induced vortices,
Competitive reactions the mixer performance was found to be independent of geometry for a given energy dissipation rate.
Chaotic advection However, the mixer geometry will affect the critical Reynolds number that induces chaotic advection
Scale-up and changes the mixing time scale.
© 2011 Elsevier B.V. All rights reserved.

1. Introduction greatest distinction between geometries appears to be the critical

Reynolds number for the onset of these unsteady circulation pat-
Process intensification is quickly becoming an important aspect terns, and the relative fraction of the energy dissipation rate, which
in the production of fine chemicals and active pharmaceutical contributes to chaotic mixing. This study will focus on the charac-
ingredients. Continuous flow microreactors are an integral part of terization of selected microreactors using semi-empirical pressure
process intensification as they overcome many of the heat transfer drop modeling and a competitive parallel reaction scheme, acetal
and safety drawbacks of traditional batch and semi-batch stirred cleavage, to compare their mixing performance on a common scale.
reactors. Microreactors with a wide variety of geometries have
been designed and several studies have attempted to provide tech-
niques to quantify their performance [1–5]. Quantification methods 2. Materials and methods
include the use of “chemical rulers” to portray the mixing perfor-
mance of a given geometry via the analysis of the yield or selectivity 2.1. Microreactor geometries
of a particular reaction scheme [6–9]. Moreover, several model-
ing techniques including semi-empirical dimensional analysis have The new generation of microreactors contains a series of reac-
also been used to characterize transport phenomena in these com- tor plates that are interchangeably placed between heat exchange
plex structures [10]. While geometry plays an important role in plates to allow high flexibility and versatility in the configuration
determining mixing efficiencies for a given pressure drop, design of a reactor for a specific reaction [11]. Each reactor plate consists
strategies to optimize these aspects and the distinct advantages of a channel with two or three components: a mixer (optional),
of each structure remain unclear. Geometry-induced, eddy-based a main channel and entrance/exit zones. Fig. 1 depicts a typical
secondary flows are known to benefit mixing either by chaotic arrangement (e.g., Plate 315) with these three components, where
advection [1] or turbulence at elevated Reynolds numbers. The the main channel is a serpentine arrangement with a rectangular
cross-section.
Three mixing structures that create secondary flows in different
ways were selected as shown in Fig. 2. The tangential mixer (TG)
∗ Corresponding author. is a convective type mixer that generates recirculation zones on
E-mail address: [email protected] (D.M. Roberge). each side of the structure with throughput primarily in the center

0255-2701/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.cep.2011.05.016
1070 C.P. Holvey et al. / Chemical Engineering and Processing 50 (2011) 1069–1075

for all the plates with 1/8 in. diameter tubing corresponding to an
internal diameter of 1.49 mm and a length of 35 mm. Although com-
mercially labeled as microreactors, the dimensions of these systems
are at the millimeter-scale [17], permitting sufficient throughput to
span production up to at least clinical trial phase III, and enabling
the creation of chaotic eddy-based flows while maintaining suitable
surface area/volume ratio for heat transfer.
The hydraulic diameter was used as the characteristic length
describing the individual components of the reactor plate, esti-
mated from the smallest cross-section within the structure, where
the highest Reynolds number occurred.

2.2. Pressure drop model

The pressure drop experiments were performed at ambient

temperature using water as the process fluid. Pressures were mea-
Fig. 1. Microreactor Plate 315; dimensions are in mm.
sured at the reactor plate inlet and outlet. Flow rates were varied
using a gear pump that was controlled by a coriolis flow meter.
[12]. The SZ-shaped mixer (SZ) is also a convective type mixer that Both pressure and mass flow rate readings were logged using the
promotes radial mixing via the generation of Dean vortices [13]. online control system of Siemens PCS-7. Process fluid mass flow
The caterpillar mixer (CAT) creates a split and recombine flow pat- rates ranged from 20 to 450 g/min.
tern that reduces the molecular diffusion length by dividing and A generalized pressure drop model (Eq. (1)) for a given com-
stretching the lamellae [14,15]. The TG mixer was thus designed to ponent (i) in the reactor plate accounts for energy dissipated via
promote backmixing, whereas fluid travel in the SZ and CAT mix- laminar flow wall friction () and chaotic secondary flow patterns
ers is considered closer to plug flow. Backmixing has demonstrated (ω) [18].
advantages for auto-catalytic reactions, where some labile species li 
are formed to initiate reaction as for example in nitrations [16]. Pi = (i + ωi ) · u2 (1)
dh,i 2 i
Reactor plates with different channel sizes were chosen. The
dimensions are summarized in Table 1 and are based on a scaling 
factor, which is from smallest to largest: 300 series, 200 series and Ptotal = Pi = PMixing zone + PMain channel + PEntrance/exit
100 series. The size of 100 and 200 series plates corresponds to
(2)
the A5 European DIN/ISO norm while the 300 series to the A6 (half
of the area). The 100 series has a typical channel width of 10 mm
while the 200 and 300 series have a width of 5 mm. One individ- The laminar flow friction factor is inversely proportional to the
ual mixer, an IMM (Institute für Mikrotechnik Mainz) CAT mixer Reynolds number (i = Cf,i /Rei ). The length (li ) of each reactor plate
was also employed in this study. The inlet/outlet zones are similar component is the quotient of the internal channel volume to the

Fig. 2. Mixing structures: (a) tangential, (b) SZ-shaped, (c) caterpillar.

Table 1
Characteristic dimensions of selected reactor plate series and mixing structures.

Plate # Series Sizea Mixing zone Main channel zone

Length (mm) dh (mm) Length (mm) Width (mm) Depth (mm) dh (mm)

Plate 104 100 A5 – – 1858 10 0.9 1.65

Plate 102 (TG) 135 1.08 1710 10 0.5 0.95
Plate 110 (CAT) 60 1.0 2284 10 0.5 0.95
Plate 205 200 A5 – – 2757 5 1.4 2.19
Plate 215 (SZ) 288 0.95 2599 5 0.8 1.38
Plate 260 (TG) 113 1.38 1514 10 0.7 1.31
Plate 302 (TG) 300 A6 74 0.71 987 5 0.5 0.91
Plate 315 (SZ) 214 0.71 996 5 0.5 0.91
IMM (CAT) s.m. – 69 0.6 – – – –

s.m., single mixer (not a plate).

a
European DIN norm.
 C.P. Holvey et al. / Chemical Engineering and Processing 50 (2011) 1069–1075 1071

cross-sectional area, except for the TG mixer, where it is estimated

as the unit cell length. In complex winding and meandering struc-
tures with Re > 100, the flow is not laminar, but shows transitional
behavior to turbulent characteristics [19]. Under these conditions,
vortices appear in bend flow, which start to fluctuate with increas-
ing Re number and lead to chaotic flow patterns. Therefore, curved
geometries and mixers, which generate complex secondary flows
are known to yield a smooth transition from laminar to turbulent
conditions [20]. This is analogous to the flow field in a packed bed,
where laminar flow is gradually replaced by localized regions of tur-
bulence. Modeling the pressure drop thus requires contributions
from both a laminar friction factor and a chaotic secondary flow
friction factor as a means of interpolation within the transitional
regime.
The individual contributions of each component within the
reactor plate were experimentally determined through the Fig. 3. Experimental (points) and predicted (lines) pressure drops for the various
use of multiple plate designs. Configurations incorporating the reactor plates with and without mixers. The plates without mixers are essential to
entrance/exit effects and varying lengths of the main channel were decorrelate the entrace/exit and main channel effects. Thus, reliable pressure drop
values can be obtained for the mixing zone alone that are used to calculate the
utilized to determine friction factors for these two components.
energy dissipation rate (vide infra).
The different mixing zones were then incorporated into the flow
path and the pressure differential determined the friction factors
for each mixer geometry. All friction factors were determined via were ensured by submersing the reactor plate in a temperature
linear regression in MatLab, in conjunction with 95% confidence controlled thermal bath.
intervals and correlation coefficients. Batch tests in a stirred reactor with a HCl concentration of
2000 ppm were first performed in order to confirm the reaction
scheme as a chemical ruler for mixing efficiency. As expected, an
2.3. Mixing efficiency characterization
increase in the stirring rate from 250 to 750 rpm decreased the
methanol yield from 100 to 45% as acid–base neutralization is
The Bourne reactions, which consist of several competing reac-
favored by more intense mixing.
tion schemes with different kinetic rates, are extensively used in
both batch and continuous systems to analyze mixing efficiency
[21]. The 4th Bourne competitive parallel reaction tracks the rate of 3. Results and discussion
an acid–base neutralization (Eq. (3)) through the hydrolysis of 2,2-
dimethoxypropane (DMP) (Eq. (4)), and was recently demonstrated 3.1. Pressure drop
to be quite robust in characterizing mixing [9,22].
A pressure drop model with six friction factors presented an
NaOH + HCl → NaCl + H2 O (3) excellent fit to the experimental data for all plate series with
correlation coefficients exceeding 0.99, but there was a lack of
H+
(CH3 )2 C(OCH3 )2 + H2 O−→(CH3 )2 CO + 2CH3 OH (4) sensitivity indicating superfluous parameters. A 95% confidence
         interval was created for each friction factor to suggest model reduc-
DMP Acetone Methanol
tions. Laminar friction factors (i ) associated with the mixing and
When the characteristic mixing time is short relative to the char- entrance/exit zones had large confidence intervals and a minor
acteristic reaction time of the slower reaction, the system reaches effect on the model response. The removal of these terms suggests
a homogenous condition. Since the acid–base neutralization kinet- that frictional losses were dominated by chaotic eddy-based flow
ics are much faster than that of the DMP hydrolysis, HCl will be patterns for the range of Reynolds numbers explored (200–4000),
consumed too quickly for DMP to react and therefore acetone or consistent with the “transient laminar” flow regimes identified by
methanol will not be generated in detectable quantities. However, Dreher et al. [25] for T-mixers and Kraume [26] for static mixers as
when the characteristic mixing time is longer or equal to the char- well as discussed by Kockmann and Roberge [27]. The relatively
acteristic reaction time of the slower reaction, a local concentration small dimensions of the entrance/exit zones result in Reynolds
gradient exists at the fluids interface, where NaOH will react imme- numbers from 200 to 6000, thus corresponding to transitional and
diately with HCl creating a NaOH depleted zone. The interface turbulent flow regimes at higher flow rates. Chaotic secondary flow
becomes rich with DMP, allowing it to hydrolyze with HCl [23]. friction factors (ωi ) for the main channels were relatively small
Experimental conditions and feed solutions were prepared with large confidence intervals: consistent with the design basis
based on suggestions resulting from the batch experiments of of the main channels, inhibiting the formation of Dean vortices in
Baldyga et al. [21]. Solutions were prepared with a minimum con- the bends through the use of a large aspect ratio. Neglecting the
centration of 0.1 mol/L of NaCl to ensure consistent kinetics of the terms having confidence intervals that suggest insignificance and
neutralization and hydrolysis reactions [24]. The solution contain- then repeating regression of the experimental data, the final fric-
ing HCl was prepared with a concentration of 500–2000 ppm in tion factors and correlation coefficients were determined (Table 2).
a 25 wt.% aqueous ethanol solution. The alkali solution was pre- A comparison of the fitted model to experimental data is provided
pared according to a 4:1 volumetric ratio between the alkali and in Fig. 3.
acid solutions. The concentration of NaOH in the alkali solution The TG mixers present greater chaotic friction factors than the
yielded a 1.05 NaOH/HCl molar ratio to ensure that all the HCl was SZ and CAT mixers likely due to expansion/contraction and under-
consumed at the reactor outlet. The alkali solution also contained estimation of the true fluid travel length (li ) in a mixing unit. The
the DMP reagent, which is essential to react with local excess of entrance/exit zones have similar chaotic friction factors to the SZ
HCl. Similar to the acid solution, the alkali solution was diluted and CAT mixers, but are smaller in length and hence do not impact
with a 25 wt.% aqueous ethanol solution. The solvents were cho- the overall pressure drop as much. The theoretical Cf values for
sen to ensure that DMP remained soluble. Isothermal conditions steady laminar flow in a straight smooth rectangular channel with
1072 C.P. Holvey et al. / Chemical Engineering and Processing 50 (2011) 1069–1075

Table 2
Friction factors and correlation coefficients of pressure drop model.

Series 100 series 200 series 300 series

Cf ω Re Cf ω Re Cf ω Re

Main channel 90.9 – 54–622 79.6 – 61–1127 49.3 0.03 99–2421

TG mixer – 0.77 430–2579 – 0.04 226–4079 – 0.68 430–2579
SZ mixer N/A N/A N/A – 0.31 297–2971 – 0.33 373–2241
CAT mixer – 0.20 327–3268 N/A N/A N/A N/A N/A N/A
Entrance/exit – 0.19 234–2811 – 0.26 234–2388 – 0.06 281–6326
R2 0.997 0.999 0.999

the dimensions of the 100 and 200 series reactor plates are respec- 0.25
tively 87.2 and 76.9 [10], which compare well (within 5%) to those
extracted by the model and hence confirm the flow regime in the 0.2
main channel as laminar and without significant formation of Dean

Methanol Yield
vortices.
0.15
Fig. 4 shows the percentage of the overall pressure drop associ-
ated to the TG mixer and main channel for the three reactor plate
series. The mixing zone is generally the main contributor with its 0.1
percentage rising with increasing flow rate or decreasing channel
size. This is due to fact that the pressure drop in the main chain 0.05
increased linearly with velocity, whereas the pressure drop in the
mixing zone exhibited a quadratic relationship due to the forma- 0
tion of chaotic secondary flow patterns. The plate series 100 has a 0 500 1000 1500 2000 2500
hydraulic diameter of 1.08 mm, but a quite narrow main channel
HCl concentration [ppm]
with 10 × 0.5 mm2 , which leads to quite high pressure loss therein.
Hence, the pressure drop in the mixing zone is larger than that in the Fig. 5. Methanol yield at different HCl concentrations with Plate 110 (CAT). Flow
main channel for relatively high flow rates higher than 80 mL/min. rate of 80 g/min and temperature of 25 ◦ C.
Although not shown in Fig. 4, the main channel is the dominant
contributor to the overall pressure drop at all flow rates when the
ing hydrodynamics. From the Arrhenius relationship, an increase
CAT mixer is used, which is likely due to a reduced mixer length
in temperature will increase both reaction rates but since the DMP
(see Table 1).
hydrolysis is in the range of the characteristic time of mixing it will
be more sensitive to the temperature rise [21]. From Fig. 6, a rise in
3.2. Mixing efficiency flow rate creates more intense vortices and increases the interac-
tion between NaOH and HCl molecules, which reduces the extent
3.2.1. Effect of acid concentration and system temperature of DMP hydrolysis as observed by the decrease in methanol yield.
Experiments with a CAT mixer were first conducted in order All future experiments were thus conducted at a temperature of
to evaluate the effect of HCl concentration and temperature on 25 ◦ C and a HCl concentration of 2000 ppm.
methanol yield (Figs. 5 and 6). All other system parameters, includ-
ing molar and volumetric flow rate ratios, are kept constant. The
3.2.2. Effect of flow rate
rate of neutralization will rise due to increased concentrations of
Figs. 7–9 present the methanol yield as function of flow rate,
HCl and NaOH. However, the constant stoichiometric ratio also
Reynolds number and pressure drop for different mixers and reac-
increases the concentration of DMP at the fluid interface allow-
tor plate series. Sufficient spreading of methanol yields in Fig. 7
ing more DMP to be hydrolyzed by HCl yielding more methanol.
validates the chosen operating conditions to distinguish trends
The effect of system temperature on the methanol yield is likely
exhibited by the different mixing structures and sizes. In general,
due to changes in the relative kinetics of the neutralization and
at a given flow rate and plate series, CAT and SZ mixers had the
hydrolysis reactions rather than a change in viscosity and result-

0.7
Temperature=15°C
0.6
Temperature=25°C
0.5
Methanol Yield

Temperature= 35°C
0.4

0.3

0.2

0.1

0
30 50 70 90 110 130
Flow rate [g/min]

Fig. 4. Fraction of total pressure drop associated to the TG mixer and associated Fig. 6. Methanol yield at different temperatures and flow rates with Plate 110 (CAT).
main channel (MC) for all three plate series. HCl concentration of 2000 ppm.
 C.P. Holvey et al. / Chemical Engineering and Processing 50 (2011) 1069–1075 1073

0.8 0.8
Plate 102 (TG) Plate 215 (SZ) Plate 102 (TG) Plate 215 (SZ)
0.7 Plate 260 (TG) Plate 315 (SZ) 0.7
Plate 260 (TG) Plate 315 (SZ)
Plate 110 (CAT) IMM (CAT)
0.6 0.6 Plate 110 (CAT) IMM (CAT)

Methanol Yield
Methanol Yield

0.5 0.5
0.4 0.4
0.3 0.3
0.2
0.2
0.1
0.1
0
0
30 50 70 90 110 130 1 10 100 1000 10000 100000
Flow rate [g/min] Energy Dissipation [W/L]

Fig. 7. Methanol yield versus flow rate for various mixing structures and reactor Fig. 10. Methanol yield versus energy dissipation rate for various mixing structures
plate series. and reactor plate series.

0.8 sented by the pressure drop along the reactor plate, with greater
Plate 102 (TG) Plate 215 (SZ)
0.7 values leading to smaller methanol yields. The effect of geometry
Plate 260 (TG) Plate 315 (SZ) and selection of characteristic length is now less apparent, espe-
0.6
Methanol Yield

Plate 110 (CAT) IMM (CAT) cially when comparing the SZ and TG mixers. CAT mixers again
0.5 provide the lowest methanol yields.
Another measure is the energy dissipation rate, which, in units
0.4
of power per volume, is defined as the pressure drop multiplied by
0.3 the ratio of superficial velocity to channel length.

0.2 P · u
Energy dissipation rate = (5)
V/A
0.1
If the structures operate at the same mixing time scale, then
0.0
0 250 500 750 1000 1250 1500 1750
the energy dissipation rate can be used to compare them on
a common basis. Note that from the work of Falk and Com-
Reynolds Number
menge [3], micromixers typically operate under a mesomixing time
scale, where both turbulent eddies and molecular diffusion play a
Fig. 8. Methanol yield versus Reynolds number for various mixing structures and
reactor plate series. role.
The methanol yield as a function of the energy dissipation
rate is presented in Fig. 10. Increasing the energy dissipation rate
lowest methanol yield with the TG mixers grouping towards the increases the mixing efficiency as the energy input into the system
upper limits of the methanol yield range. The smallest size mixers and its subsequent conversion to vortices is increased. For these
(300 series) presented the lowest methanol yields due to the higher flow conditions, there is no obvious difference in the methanol
velocities producing stronger secondary unsteady flows and better yield of the various mixers at a given energy dissipation rate,
mixing. This effect is apparent in Fig. 8, where the methanol yield which also confirms the initial assumption that the structures mix
is presented as a function of the Reynolds number. Increasing the at the same time scale. The energy dissipation rate can then be
Reynolds number increases the intensity and shedding frequency of considered a good tool for scale-up [11]. However, the different
the vortices in the flow field and results in lower methanol yields. types of flow patterns created by the mixers do present advan-
The intensity of the vortices leading to frictional losses is repre- tages relative to each other. As will be shown in Fig. 11, different

0.8 1
Plate 102 (TG) Plate 215 (SZ)
0.7 Plate 260 (TG) Plate 315 (SZ) 0.9
0.6 Plate 110 (CAT) IMM (CAT) 0.8
Methanol Yield

0.5 0.7
Methanol Yield

0.4 0.6
0.5
0.3
0.4
0.2 Plate 315 (SZ) Standard
0.3 Plate 315 (SZ) PEG 8000
0.1 Plate 260 (TG) Standard
0.2 Plate 260 (TG) PEG 8000
0.0 Plate 110 (CAT) Standard
0.1 Plate 110 (CAT) PEG 8000
1 10 100 1000 Plate 302 (TG) PEG 8000
0
Pressure drop [kPa] 1 10 100 1000 10000 100000
Energy Dissipation [W/L]
Fig. 9. Methanol yield versus pressure drop for various mixing structures and reac-
tor plate series. It is important to mention that the pressure drop indicated in this Fig. 11. Methanol yield versus energy dissipation rate for various mixing structures
figure is the intrinsic pressure of the mixing zone decoupled from the entrance/exit and reactor plate series. “Standard” represents low viscosity data, whereas “PEG
and main channel effects (vide supra, Fig. 3). 8000” represents high viscosity data.
1074 C.P. Holvey et al. / Chemical Engineering and Processing 50 (2011) 1069–1075

1000 2300 for smooth straight pipes but the various flow bifurcations
Plate 315 (SZ) PEG 8000 of the studied mixers increase significantly their friction. This leads
Plate 302 (TG) PEG 8000 to chaotic eddy-based secondary flow patterns at lower Reynolds.
800
The energy dissipation rate is a scale-up criterion under turbulent
Pressure drop [kPa]

Plate 110 (CAT) PEG 8000

flow conditions, and it was also applicable to the present study
600 as demonstrated in Fig. 10. Thus, by combining Eqs. (1) and (5)
it is possible to obtain following equation for the scale-up of the
mixers:
400
 Q 3/7
2
dh,2 = dh,1 (6)
200 Q1

The hydraulic diameter of a mixer will be increased by a power

0 factor 3/7 of the ratio of the flow rates to ensure constant mixing
0 50 100 150 200 250 characteristics at higher flow rates.
Reynolds Number Finally it is worth to observe in Fig. 9 that at comparable pres-
sure drop, the mixing efficiency of Plate 110-CAT and IMM-CAT
Fig. 12. Pressure drop versus Reynolds number for three reactor plates. Solutions
contain PEG 8000 to increase its viscosity. SZ, TG and CAT represent SZ-shaped is notably better. This outcome can be understood by the smaller
tangential and caterpillar mixers, respectively. length of these mixers compared to their homologue tangential and
SZ-shaped mixers (see Table 1). Thus, when comparing various mix-
geometries induce chaotic secondary flow patterns and change the ers, the energy dissipation rate is fundamental but does not account
dominant mixing mechanism at different critical Reynolds num- for the various residence times in the mixing zone. The residence
bers. times in the mixing zones of the tangential and SZ-shaped mixers
are too long or, in other words, the length of the mixers is exagger-
3.2.3. Effect of viscosity ated. Thus, for low viscosity systems just a few number of mixing
Polyethylene Glycol (PEG) with an average molecular weight elements is required. However, at higher viscosity our experimen-
of 8000 was added at 15 wt.% to the standard solution in order tal data show that more mixing elements are required and will be
to increase the viscosity from 0.002 to 0.01 Pa s and dampen the subject of forthcoming work.
the vortices. Reagent concentrations remained constant and the
PEG concentration employed conserved the DMP solubility in the 4. Conclusions
system. Fig. 11 presents the methanol yield versus the energy dis-
sipation rate at the two different viscosities over the same range Several micro-structured plates with various mixing struc-
of flow rates as with the standard solution. As opposed to the low tures were selected in order to characterize their pressure drop
viscosity system, the methanol yields in the high viscosity system and mixing efficiency. The pressure drop was successfully mod-
do not converge to a single trend line. In general, yields remain rel- eled via friction factors fitted to the experimental data over a
atively constant with an increase in energy dissipation rate until a Reynolds number range of around 50–2400 in the main channel and
critical point, which varies for the different mixers, where they then 200–4000 in the mixers. From a sensitivity analysis of the friction
rapidly decrease. The abrupt change in the methanol yield slope factors, it was shown that the flow regime in the mixing zone pre-
is likely due to the formation of chaotic secondary flow patterns. sented chaotic secondary flow patterns, whereas in the serpentine
Rosaguti et al. [20] and Dreher et al. [25] observed a transition from main channel the flow was demonstrated to be laminar with negli-
stable to unstable vortex flow in respectively serpentine channels gible formation of vortices. Friction factors for the tangential mixers
and T-junctions at Re values around 200. The CAT mixers do not were slightly greater than for the SZ-shaped and caterpillar mixers.
exhibit a change in methanol yield slope suggesting that the mixing Furthermore, over the range of flow rates, the main contributor to
mechanism remained primarily via diffusion. The methanol yield the pressure drop was the mixing zone.
is high (80%) as HCl consumption is slowed by the reduced mixing The mixer performance was determined by the yield of compet-
rate and will therefore be available to catalyze the DMP hydrolysis itive parallel reactions. When frictional pressure losses are caused
reaction. primarily by chaotic advection, all tested mixing structures gave
Fig. 12 illustrates the relationship between the pressure drop similar performances for a given energy dissipation rate. Tests were
and Reynolds number for selected reactor plates. Linear (Plate also conducted with a more viscous solution in order to dampen
110-CAT) and non-linear (Plates 315-SZ and 302-TG) relationships the chaotic nature of the flow and translate the mixing mecha-
respectively suggest the absence and presence of significant chaotic nism primarily towards molecular diffusion. This exercise allowed
secondary flows patterns, which corroborates with their mixing to differentiate the performances of the various microreactors, with
mechanism/performance in Fig. 11. Note that Plate 110-CAT may the tangential and SZ-shaped convective mixers inducing vortices
still exhibit secondary flow patterns at the present range of Re num- and chaotic mixing at lower critical Reynolds numbers and hence
bers (20–200), as reported by Schönfeld et al. [15], but they do not energy dissipation rates. For the operating conditions explored, it is
significantly contribute to the pressure loss. In addition, the CAT therefore beneficial to maximize the pressure drop due to chaotic
mixer length is shorter than the other plates, which contributes to advection in a given mixing structure geometry in order to increase
a smaller pressure drop. its mixing efficiency.

3.3. Scale-up concept and discussion

Acknowledgements
Section 3.1 allows decoupling the various pressure drops asso-
ciated to the entrance/exit, main channel, and mixing zones. Thus The authors are grateful to Adam Donaldson, Bertin Zim-
reliable estimates of the mixing zone pressure drop can be used mermann, Aswin Yellepeddi, and Markus Eyholzer for valuable
to calculate the intrinsic energy dissipation rate. Classical turbu- discussions and to the Natural Sciences and Engineering Research
lent flow regime usually is initiated at Reynolds number above Council of Canada and Lonza AG for financial assistance.
 C.P. Holvey et al. / Chemical Engineering and Processing 50 (2011) 1069–1075 1075

Appendix A. Nomenclature [8] P. Guichardon, L. Falk, J. Villermaux, Characterisation of micromixing efficiency

by the Iodide–Iodate reaction system. Part II. Kinetic study, Chem. Eng. Sci. 55
(2000) 4245–4253.
[9] C. Lindenberg, J. Schöll, L. Vicum, M. Mazotti, J. Brozio, Experimental character-
A cross-sectional area (m2 ) ization and multi-scale modeling of mixing in static mixers, Chem. Eng. Sci. 63
(2008) 4135–4149.
Cf friction factor constant in laminar flow
[10] N. Kockmann, Transport Phenomena in Micro Process Engineering, Springer,
d diameter (m) 2008.
dh hydraulic diameter (m) [11] D.M. Roberge, M. Gottsponer, M. Eyholzer, N. Kockmann, Industrial design,
l length (m) scale-up, and use of microreactors, Chem. Today 27 (2009) 8–11.
[12] L. Jongkwang, S. Kwon, Mixing efficiency of a multilamination micromixer with
Q volumetric flow rate (m3 s−1 ) consecutive recirculation, Chem. Eng. Sci. 64 (2009) 1223–1231.
Re Reynolds number (ud/) [13] D.M. Fries, S. Waelchli, P.R. von Rohr, Gas–liquid two-phase flow in meandering
u superficial velocity (m s−1 ) microchannels, Chem. Eng. J. 135 (2008) S37–S45.
[14] S.W. Lee, S.S. Lee, Rotation effect in split and recombination, Sens. Actuators B
V volume (m3 ) 129 (2008) 364–371.
[15] F. Schönfeld, V. Hessel, C. Hofmann, An optimised split-and-recombine micro-
mixer with uniform ‘chaotic’ mixing, Lab Chip 4 (2004) 65–69.
Greek letters
[16] L. Ducry, D.M. Roberge, Controlled autocatalytic nitration of phenol in a
P pressure drop (Pa) microreactor, Angew. Chem. Int. Ed. 44 (2005) 7972–7975.
 laminar friction factor [17] S.G. Kandlikar, W.J. Grande, Evolution of microchannel flow
 viscosity (Pa s) passages–thermohydraulic performance and fabrication technology, in:
Proc. of the International Mechanical Engineering Conference and Exposition,
 density (kg m−3 ) New Orleans, LA, November 17–21, 2002, Paper No. IMECE2002-32043.
ω chaotic secondary flow friction factor [18] N. Kockmann, T. Kiefer, M. Engler, P. Woias, Convective mixing and chemical
reactions in microchannels with high flow rates, Sens. Actuators B 117 (2006)
495–508.
References [19] N. Kockmann, M. Gottsponer, D.M. Roberge, Scale-up concept of single-channel
microreactors from process development to industrial production, Chem. Eng.
[1] V. Hessel, H. Löwe, F. Schönfeld, Micromixers – a review on passive and active J. 167 (2011) 718–726.
mixing principles, Chem. Eng. Sci. 60 (2005) 2479–2501. [20] N.R. Rosaguti, D.F. Fletcher, B.S. Haynes, Laminar flow and heat transfer in a
[2] J.-M. Commenge, L. Falk, J.-P. Corriou, M. Matlosz, Analysis of microstructured periodic serpentine channel, Chem. Eng. Technol. 28 (2005) 353–361.
reactor characteristics for process miniaturization and intensification, Chem. [21] J. Baldyga, J.R. Bourne, B. Walker, Non-isothermal micromixing in turbulent
Eng. Technol. 28 (2005) 446–458. liquids: theory and experiment, Can. J. Chem. Eng. 76 (1998) 641–649.
[3] L. Falk, J.-M. Commenge, Performance comparison of micromixers, Chem. Eng. [22] A. Kölbl, S. Schmidt-Lehr, The Iodide Iodate reaction method: the choice of the
Sci. 65 (2010) 405–411. acid, Chem. Eng. Sci. 65 (2010) 1897–1901.
[4] A. Kukukovaa, J. Aubin, S. Kresta, New definition of mixing and segregation: [23] B.K. Johnson, R.K. Prud’homme, Chemical processing and micromixing in con-
three dimensions of a key process variable, Chem. Eng. Res. Des. 87 (2009) fined impinging jets, AIChE J. 49 (2003) 2264–2282.
633–647. [24] D.B. Dennison, G.A. Gettys, D.G. Kubler, D. Shepard, Kinetic salt effects on the
[5] M. Kashid, A. Renken, L. Kiwi-Minsker, Mixing efficiency and energy con- hydrolysis of benzaldehyde dimethyl acetal, J. Org. Chem. 41 (1976) 2344–2348.
sumption for five generic microchannel designs, Chem. Eng. J. 167 (2011) [25] S. Dreher, N. Kockmann, P. Woias, Characterization of laminar transient flow
436–443. regimes and mixing in T-shaped micromixers, Heat Transfer Eng. 30 (2009)
[6] J. Baldyga, J.R. Bourne, Turbulent Mixing and Chemical Reactions, Wiley, New 91–100.
York, 1999. [26] M. Kraume, Mischen und Rühren – Grundlagen und moderne Verfahren, Wiley-
[7] P. Guichardon, L. Falk, Characterisation of micromixing efficiency by the VCH, Weinheim, 2003.
Iodide–Iodate reaction system. Part I. Experimental procedure, Chem. Eng. Sci. [27] N. Kockmann, D.M. Roberge, Transitional flow and related transport phenom-
55 (2000) 4233–4243. ena in curved microchannels, Heat Transfer Eng. 32 (2011) 595–608.