Article

pubs.acs.org/OPRD

Application-Oriented Analysis of Mixing Performance in

Microreactors
Sebastian Schwolow,† Jutta Hollmann,‡ Berthold Schenkel,‡ and Thorsten Röder*,†
†
Institut für Chemische Verfahrenstechnik, Hochschule Mannheim, Paul-Wittsack-Straße 10, 68163 Mannheim, Germany
‡
Novartis Pharma AG, Lichtstrasse 35, 4056 Basel, Switzerland

ABSTRACT: A methodology was developed to select adequate, commercially available micromixers for mixing sensitive
chemical reactions. The range of flow rates can be derived at which the selected micromixers have to be operated to ensure the
required mixing intensity. This methodology enables the selection of adequate micromixers for the scale up of the chemical
reactions to higher flow rates. Two chemical test reactions were used for an experimental approach to characterize the selected
microreactors. Both reactions are based on the effect of micromixing on the product distribution of competitive reaction systems.
Flow rates and pressure drop were determined at which the mixing times are short relative to the reaction times. In this case,
influences of mixing on the selectivity of the reference reaction can be neglected. Since two reference reactions with different time
scales for mixing and reaction were tested, it was possible to study the mixing performance of a variety of micromixers over a
wide range of flow rates. The investigated micromixers differ in their dimensions, internal geometry, and mixing principle. In the
present work, overview tables are provided as a tool to evaluate the commercially available micromixers for specific applications.
Further, the influence of mixing principle and pressure drop is discussed.

■ INTRODUCTION
In the past few years, the use of micromixers has gained
can have a strong influence on the conversion rate. In the case
of a reaction system with competitive parallel or consecutive
increasing significance for various applications in chemistry. reactions, mixing time can also affect the selectivity of this
The main advantage of microstructured reactors is the system, especially if parallel reactions proceed at a different time
enhancement of heat and mass transfer due to the high scale.
surface-to-volume ratio and the increased intensity of fluid This dependency can also be used for an application-oriented
mixing. characterization of the mixing time by defined reaction systems.
The fast development in microreaction technology has led to Several test reactions based on two competitive chemical
a high number of available micromixers, based on entirely reactions are described in the literature. They can be divided
different mixing principles.1 Thus, the choice of a commercially into two simplified schemes: The competitive parallel reaction
available mixer as well as the design of new devices for a system (Scheme 1) and the competitive consecutive reaction
particular application can be difficult. Complex interactions of system (Scheme 2).
fluid mechanics, mass transfer, and reactions have to be
considered. For an efficient reactor selection, optimized for Scheme 1. Competitive Parallel Reaction System
specific reactions, experimental investigations of mixing
processes are necessary.
The main objective of this work is the characterization of
several commercially available, passive micromixers to provide
assistance for the selection and scale-up of microreaction
devices. Due to the complex nature of mixing processes and the Scheme 2. Competitive Consecutive Reaction System
variety of applied mixing principles, an experimental approach
was taken. It is focusing on the interactions between mixing and
chemical reactions. The degree to which the chemical reactions
are determined by the mixing process is mainly influenced by
the ratio of the characteristic times for mixing and reaction. If
the mixing time is significantly shorter than the reaction time, it The influence of the two previously mentioned reaction-to-
can be assumed that the reaction takes place in a homogeneous mixing time ratios on the two reaction schemes is illustrated in
concentration field. In this case, the conversion rate cannot be Figures 1 and 2. The homogeneous condition (a) can be
increased by enhancing the mixing performance. In contrast, if achieved only if mixing time tm is short relative to the reaction
the reaction is fast compared to mixing, the mixing rate can time tR of the slower reaction. In this case, the selectivity is only
affect the conversion rate. The reaction takes place in a determined by the chemical kinetics. Since the rate constant k1
heterogeneous concentration field with multilamellae structures
which are determined by the mixing process. The length scales Received: April 30, 2012
of these structures as well as the time scale of their generation Published: August 15, 2012

© 2012 American Chemical Society 1513 dx.doi.org/10.1021/op300107z | Org. Process Res. Dev. 2012, 16, 1513−1522
Organic Process Research & Development Article

absence of species A) the slower reaction can proceed. As a

result, a detectable yield of product S will be achieved,
depending on the degree of segregation.
A very commonly used competitive parallel reaction system
is the iodide iodate reaction method known as the Villermaux−
Figure 1. Competitive parallel reactions in a homogeneous (a) and a Dushman reaction.2 This reaction system is simple to
heterogeneous concentration field (b). implement in the laboratory, but it is also associated with
some difficulties, as reported by Falk3 and Kölbl.4 Several other
test reactions have been developed by Baldyga, Bourne, and co-
workers.5−7 Two of them were used for the investigations
described in the present paper: (1) neutralization and acetal
hydrolysis, a system of fast competitive parallel reactions;6 (2)
diazo coupling between 1-naphthol and diazotized sulphanilic
Figure 2. Competitive consecutive reactions in a homogeneous (a) acid, a system of competitive consecutive reactions.7
and a heterogeneous concentration field (b). In order to investigate micromixing phenomena at different
time scales, the main criterion for the choice of the test
is much larger than the rate constant k2, the yield of product S reactions was the significant difference between the character-
(YS) approaches zero. In contrast, if the mixing time tm is istic times for the slower reaction of each system. With suitable
comparable to or longer than the reaction time tR of the slower reactant concentrations, acetal hydrolysis takes place on the
reaction (b), the reaction takes place at the interfacial region of time scale of only a few milliseconds, while the time needed for
the segregated fluid lamellae. After the immediate reaction of A the secondary azo coupling is about 2 orders of magnitude
and B, this region now contains an excess of species C. longer. Furthermore, practical reasons led to the choice of the
Subsequently, reactant C continues to diffuse toward the region two systems: The reaction mechanisms are fully investigated,
with a local overconcentration of reactant B, where (in the and kinetic data are available in the literature.

Table 1. Investigated Micromixers

1514 dx.doi.org/10.1021/op300107z | Org. Process Res. Dev. 2012, 16, 1513−1522

Organic Process Research & Development Article

Different mixer types were investigated representing the 2cS 2cS

YS = =
most common mixing principles (Table 1). The different cB0 coR + cpR + 2cS (2)
mixing structures can be divided into four groups of commonly
used micromixers: (1) T-, Y-, and arrowhead configurations; where coR, cpR, and cS are the concentrations of the products o-
(2) flow obstacles and chicanes within microchannels; (3) R, p-R, and S (see Table 2). The initial solution containing the
repeated flow split and recombination structures; (4) multi-
lamination mixing principle. Table 2. Symbols and Names of Reagents in Eqs 1a−1d

■ EXPERIMENTAL SECTION
Experimental Setup. Since syringe pumps can provide
symbol
A
B
1-naphthol
name

diazotized sulphanilic acid

accurate flows with very low pulsation, they are well suited to o-R 2-[(4′-sulphophenyl)azo]-1-naphthol
employ micromixers at typical flow rates and moderate p-R 4-[(4′-sulphophenyl)azo]-1-naphthol
pressures. Two precision syringe dosers (SyrDos 2, HiTec S 2,4-bis[(4′-sulphophenyl)azo]-1-naphthol
Zang) were used to obtain continuous flows, each one
equipped with two pumps and 10 mL syringes. With an
automated sample collector (AutoSam, HiTec Zang), samples diazonium salt was prepared by diazotization of 3 mol/m3
were taken at different total flow rates in the range 0.5−50 mL/ sulphanilic acid with sodium nitrite and hydrochloric acid.
min. For pressure measurements, two piezoresistive pressure Excess nitrite was eliminated by adding sulphamic acid. 3.6
transmitters (model S-10, WIKA) were used. At the mixer mol/m3 of 1-naphthol was dissolved in water by stirring and
outlet, a tube with an inner diameter of 1.58 mm and a length heating the solution. This stream was buffered using 222.2 mol/
of 3 m was connected to ensure a sufficiently long residence m3 each of sodium carbonate and bicarbonate in order to
time. Otherwise, in the case of poor mixing performance, the obtain a pH of 9.9 and an ionic strength μ = 444.4 mol/m3 in
reaction may not be completed before the solution enters the the mixed product solution. Due to the limited stability of the
sample vial. reactants, both solutions were prepared directly before starting
Diazo Coupling. The principle of the competitive the experiments.
consecutive reaction system has been described previously in After collecting the samples, the product mixture was diluted
a simplified form (Scheme 2). The system developed by the with buffer solution. The absorption of each sample was
team of J. R. Bourne is more complex, since it contains two measured at a wavelength range of 400−700 nm. In order to
primary couplings (eqs 1a and 1b) in which isomeric monoazo determine the concentration of the different dyes, multilinear
dyes (p-R and o-R) are formed. Subsequently, their secondary regression analysis was performed for the measured absorption
couplings (eqs 1c and 1d) produce one bisazo dye (S). spectra. The calculations with multilinear spectra are based on
the molar extinction coefficients for each single substance
k1o
(1a) (Figure 3) given by Bourne and co-workers.7 Since the
A + B → o‐R
absorption spectra of the monoazo dyes are very similar, their
k1p concentrations were determined in sum (coR + cpR).
A + B → p‐R (1b)

k 2o
o‐R + B → S (1c)

k 2p
p‐R + B → S (1d)

For the reaction scheme and especially kinetic aspects, see

Baldyga5 and Bourne.7 The rate constants are given for a
temperature of 298 K, pH = 9.9, and an ionic strength μ =
444.4 mol/m3:
k1p = 12238 ± 446 m 3/(mol s)

k1o = 921 ± 31 m 3/(mol s)

k 2p = 22.25 ± 0.25 m 3/(mol s) Figure 3. Molar extinction coefficients for the bisazo dye (S) and both
monoazo dyes (o-R, p-R), for μ = 444.4 mol/m3 and T = 298 K.
k 2o = 1.835 ± 0.018 m 3/(mol s)
The limiting reagent in the system is the diazotized For each measurement, a mass balance check was carried out
sulphanilic acid (B), since it is consumed by all reactions. by comparing the initial amount of diazotized sulphanilic acid
When the reactions are completed, a determination of the with the amount of the azo dyes in the product solution. It was
product concentrations enables calculation of the selectivity of found that the mass balance closes very well (±5%) for low
the comparatively slower secondary couplings. Based on a values of YS. For higher yields (>20%), a significant amount of
fitting material balance and no further byproduct formation, the B is apparently not converted to any of the three dyes, which
yield of the bisazo dye S in relation to the initial concentration indicates the formation of unidentified byproducts.5 Since the
of B (cB0) can be defined as deviation in the mass balance is systematic and increases with
1515 dx.doi.org/10.1021/op300107z | Org. Process Res. Dev. 2012, 16, 1513−1522
Organic Process Research & Development Article

increasing values of YS, the qualitative conclusions about the The oven temperature of 60 °C was kept constant for 1 min
mixing performance are still valid. and then increased to 140 °C with a ramp of 10 °C/min.
DMP Hydrolysis. The competitive parallel reaction scheme,
containing a neutralization reaction and an acetal hydrolysis,
was studied by Baldyga, Bourne, and co-workers. A detailed
■ RESULTS AND DISCUSSION
Reaction Time and Mixing Time. In order to determine
thermochemical and kinetical characterization of this system the characteristic reaction times, the case of an approximately
can be found in publications of Baldyga et al.5,6 and homogeneous concentration field is assumed. Thus, the half-life
Lindenberg.8 In the experiments, a solution of sodium of species B in a second order reaction can be expressed as
hydroxide and 2,2-dimethoxypropane (DMP) is mixed with
an acid solution: 1
t1/2 =
k1
k 2cB0 (7)
H+ + OH− → H 2O (3)
With the chosen concentrations and kinetic data for a
+ temperature of 298 K (see Experimental Section), a half-life of
H + CH3C(OCH3)2 CH3 + H 2O
about 5 ms results for DMP hydrolysis. Due to the more
k2
→ H+ + CH3COCH3 + 2CH3OH (4) complex reaction scheme of the diazo coupling with several
parallel and consecutive reactions, further simplifications are
With a second-order rate constant k1 = 1.4 × 10 m mol−1 8 3
made for this system: Since k1o is small relative to k1p, the
−1
s , the neutralization of sodium hydroxide is very fast and can reaction path via the monoazo dye o-R (see eq 1) can be
be considered as quasi-instantaneous. By comparison, the neglected. The half-life is determined for a state of reaction
formation of acetone and methanol in the acid catalyzed when half of the diazotized sulphanilic acid already reacted with
hydrolysis of DMP is significantly slower. Due to the high the monoazo dye p-R. Therefore, the resulting time of 727 ms
excess of water, the reaction rate can be simplified to second can only be considered as a rough estimation for the time frame
order with a rate constant k2 given in m3 mol−1 s−1: of the reaction. Furthermore, if the chemical reaction proceeds
−5
cS)
in the time frame of mixing, it takes place in a heterogeneous
k 2 = 7.32 × 107 exp( − 5556/T )10(0.05434 + 7.07 × 10 (5) concentration field with unequal molar ratios. Due to the spatial
This relation was determined by Baldyga et al. for a 6 differences in concentration, a local chemical reaction rate exists
temperature range of 298−313 K, acid and DMP concen- for each region.
trations of 25−1333 mol/m3, and an ethanol concentration of In order to compare these reaction times with regard to the
25 wt %. Since increasing salt concentrations accelerate mixing process, diffusion time has to be considered. Molecular
hydrolysis, the expression additionally contains the sodium diffusion is always required as a final step when it comes to
chloride concentration cS and is valid for a range of 100−1200 mixing reactants on a molecular scale. According to Falk and
mol/m3. Commenge,9 diffusion time can be expressed as
As soon as all acid molecules are neutralized, the reaction R2
process is completed and the conversion XDMP of the slower tdiff = A
D (8)
reaction can be determined to characterize mixing:
c D is the diffusion coefficient while R denotes the half length
XDMP = 1 − DMP scale of the aggregate. The shape factor A integrates the shape
cDMP,0 (6) of the flow structure. It is defined by
Here, cDMP is the concentration of DMP in the final product 1
solution after mixing and reaction while cDMP,0 is the A=
(p + 1)(p + 3) (9)
concentration in the perfectly mixed product solution for the
theoretical case that no conversion of DMP occurred. The including a shape parameter p (slab, p = 0; cylinder, p = 1;
conversion of DMP can be considered as an average result of sphere, p = 2). In a plane multilamellae flow pattern, p = 0 can
fluid segregation during the reaction process. be used to calculate diffusion in micromixers. The shape
Both initial solutions contained 25 wt % ethanol in deionized parameter p = 1 can be used to describe tubes with concentric
water and 90 mmol/L NaCl. It must be noted that contrary to flow patterns (for a practical setup, see ref 10). In most cases,
the system described by Baldyga et al., nitric acid was used the structures are deformed, for example by laminar vortices, so
instead of hydrochloric acid. Thus, corrosion in the stainless that the shape parameter can only be used for a simplified
steel micromixers should be avoided. Due to the formation of evaluation of mixing time. The investigated micromixer with
NaNO3, deviations from the rate constant described by eq 5 are the finest structures is the slit interdigital micromixer (SIMM).
possible. However, a comparison between experiments with The inlet of this mixer generates a 40 μm multilamellae flow
both acids did not show a significant difference in the results. which is focused to 15.6 μm slab size in a following narrow
One solution with 600 mol/m3 HNO3 was prepared, and a bore.11 According to eq 8, for a slab in water (D ≈ 10−9 m2 s−1)
second solution contained 600 mol/m3 DMP and 630 mol/m3 a diffusion time of approximately 80 ms results. Additional
NaOH. The 5% excess of NaOH assures a pH larger than 8 in mixing effects due to the sudden changes of the diameter in the
the mixed solution in order to avoid further reaction of DMP. mixing element are neglected for this consideration. The
Within a time of 24 h after the experiments, no further comparison with the reaction time of the DMP hydrolysis
decomposition of DMP was observed. Anyway, all samples above shows that, even at this small length scale, mixing only
were analyzed immediately after collection on a HP 5890 gas based on diffusion is not significantly faster than the half-life of
chromatograph equipped with a 30 m Quadrex column (#007- the reactant. To achieve diffusion times even shorter than the
5-30-0.25F) and a flame ionization detector. A split injection time for DMP hydrolysis, slab sizes have to be reduced by at
with an injection volume of 0.2 μL provided the best results. least 1 order of magnitude. A possible way to increase the
1516 dx.doi.org/10.1021/op300107z | Org. Process Res. Dev. 2012, 16, 1513−1522
Organic Process Research & Development Article

mixing performance in the SIMM after the establishment of the the DMP conversion XDMP, much shorter mixing times are
laminar fluid structure is to focus further the initial structure required. With a simple T-mixer, such rapid mixing is only
with a stronger narrowing mixing channel.11 However, in most possible at high flow rates above 20 mL/min. However, even at
micromixers, only one interface exists after contacting the a throughput of 48 mL/min, the mixing performance is still not
reactant flows. Mixing performance is here mainly based on the sufficient to reach DMP conversions close to zero.
formation of vortices (an example of which is illustrated in At flow rates below 15 mL/min, a slight increase in the value
Figure 4) which create fine structures with length scales of only of XDMP can be observed. A possible explanation is based on the
a few micrometers.12 Concentration differences on these scales change of the residence time in the short and narrow mixing
can be eliminated very fast by diffusion. channel. Both mixers were connected to a capillary with a larger
internal diameter of 1.58 mm. This fact is important if a strict
laminar flow without vortex formation is assumed for low
volume flows. In this case, molecular diffusion is the main
driving force for the mixing process. As can be seen from eq 8,
the mixing time for diffusion correlates with the characteristic
length scale by a quadratic dependence. Therefore, the sudden
enlargement of the channel diameter at the outlet of the T-
mixer has an important influence. Since diffusion time is
significantly longer in the connected capillary, mixing time is
mainly influenced by the actual narrow mixing channel of the
T-piece. Increasing the flow rate leads to a decrease in the
available time for diffusive mass transfer in this segment. Thus,
the mixing performance decreases with the residence time in
the T-mixer, until at higher flow rates vortex formation
predominates as a main driving force for mixing.
In order to prove the suitability of the chosen reaction
Figure 4. CFD simulation (COMSOL Multiphysics) of the vortex systems to classify different mixers by their mixing performance,
formation in a T-mixer. The planes depict the concentration profiles in T-mixers with different internal diameters were used for a
the mixing channel (size, 1 mm × 2 mm; total flow rate, 16.8 mL/ comparison. For equal volume flow rates, the flow velocity is
min).
higher in the smaller mixing channel. Therefore, more intense
vortex formation and thus shorter mixing times have to be
The influence of the different time scales of reaction in the
assumed. Vortex formation can take place in the T-piece as well
two test reaction systems is shown in Figure 5 using T-mixers
as in the transition to the connected tube due to the sudden
with different inner diameters as an example.
enlargement of the diameter. It has to be considered that the
In the experiments with the diazo coupling test system, the
yield of the bisazo dye YS decreases, indicating an enhancement two mixer set-ups are not geometrically similar, since the
of mixing performance in the range between 0 and 24 mL/min. relative increase of the internal diameter is much stronger for
The faster mixing occurs, the less time is available for the slower the T-mixer with an internal diameter of 0.5 mm.
formation of the bisazo dye. Due to the increasing flow rate, the For both reaction systems, the prediction of faster mixing at
improved vortex formation causes a reduction of mixing time. the smaller channel diameter can be confirmed. As can be seen
At a total flow rate above 20 mL/min, mixing time is short in Figure 5, the decrease of the values of YS and XDMP,
enough to approach fully homogeneous conditions for the respectively, starts at lower total flow rates for the mixer with an
reaction and the formation of the bisazo dye cannot take place. internal diameter of 0.5 mm compared to the mixer with an
Therefore, at these flow rates, the yield YS is close to zero. internal diameter of 1.25 mm. For the hydrolysis of DMP, at
As shown above, the reaction time of DMP hydrolysis is higher flow rates the DMP conversion is significantly lower
about 2 orders of magnitude shorter. Hence, for a decrease of when using the T-mixer with a channel diameter of 0.5 mm.

Figure 5. Comparison of the experimental results of T-mixers with different inner diameters for the azo coupling test reaction (a) and DMP
hydrolysis (b).

1517 dx.doi.org/10.1021/op300107z | Org. Process Res. Dev. 2012, 16, 1513−1522

Organic Process Research & Development Article

Figure 6. Experimental results of the bisazo dye yield YS as a function of the total flow rate for all investigated micromixers.

In conclusion, the combination of both test reactions allows dimensions of the mixing channels. In particular, the SIMM
an estimation of the mixing performance over a large shows excellent mixing performance for the competitive diazo
investigated flow range (0−48 mL/min). The example of the coupling in the entire throughput range. The value of YS tends
two T-mixers shows that mixing time can be shortened to zero at a flow rate of only 6 mL/min. In order to reach a
continuously by increasing the total flow rate. comparable mixing performance with the caterpillar mixer, a
Diazo Coupling Test Reaction. In Figure 6 the results for significantly higher flow rate of at least 16 mL/min is necessary.
the diazo coupling test reaction are shown sorted by vendor. The superior performance of the SIMM can be explained by
For some mixers, slightly negative values for the bisazo dye the fact that in the SIMM, after contacting the reactant
yield occur. They can be attributed to the error in conjunction solutions, instantaneously a multilamellae structure exists. In
with the multilinear regression analysis to determine the combination with focusing and the formation of vortices at
product concentrations out of the absorbance spectra. higher flow rates in the further mixing channel, mixing at a
Figure 6a shows the comparison between types of T-pieces molecular scale can be completed in a very short time. In
with alternative arrangements of the feed and mixing channels. contrast, the split and recombine mixing principle of the
For all flow rates, the lowest values of YS can be reached with caterpillar mixer is based on a successive halving of lamellae
the arrowhead geometry while the Y-mixer shows the worst width throughout the entire mixing channel. At the beginning
mixing performance. In the Y-mixer a lower energy dissipation of the mixing channel, the fluids are more segregated. Hence, in
is caused by the minor redirection of the flow. As a result, this region, the slower secondary azo coupling is able to
especially at high flow rates, the values for YS are considerably proceed.
higher compared with those for the T-mixer. In contrast, the In Figure 6c the three different glass mixers (Little Things
mixer with arrowhead geometry seems to provide slightly Factory GmbH) are compared. It is noticeable that, at flow
shorter mixing times. This is caused by a higher local energy rates up to about 10 mL/min, the slope of the curve is steeper
input due to the sharper angle between the feed and mixing than that at higher flow rates. The most effective glass mixer for
channels. this reaction system seems to be the HTM-STM-2-1 mixer; a
The investigation of the arrowhead mixer containing a frit in considerably poorer mixing performance can be observed for
the mixing channel (Figure 6b) showed considerable low bisazo the LTF-MS and LTF-MX. In these mixers, mixing takes place
dye yields. By comparing the results with an arrowhead mixer in a comparatively large reactor volume at mixing channel
of similar geometry, it is obvious that the frit with a pore size of widths of about 1 mm and a length of approximately 10 cm.
10 μm has an accelerating effect onto the mixing rate. Although The residence time in the mixer is significantly longer
an improved mixing performance was to be expected, the extent compared to those for the other investigated mixers. If the
of the increase in efficiency is remarkable. total flow rate is high enough to induce intensive vortex
The micromixers designed by the Institut für Mikrotechnik formation, it can be assumed that the mixing time is shorter
Mainz (results in Figure 6d) stand out due to their very small than the residence time in the mixer. Therefore, the mixing
1518 dx.doi.org/10.1021/op300107z | Org. Process Res. Dev. 2012, 16, 1513−1522
Organic Process Research & Development Article

efficiency in only the front part of the mixing channel would be this procedure cause a good mixing quality at the end of the
determined. This assumption is confirmed by the view of the mixer, but due to the high inner volume of the mixer, it
visible mixing state in the glass mixers. As an example, in Figure processes too slowly to avoid the secondary azo coupling
7 the progressive formation of the azo dyes in the LTF-MS completely.
mixing channel is shown for a total flow rate of (a) 8 mL/min DMP Hydrolysis Test Reaction. The comparison between
and (b) 16 mL/min. the T-mixer, the Y-mixer, and the arrowhead mixer (Figure 8a)
shows effects that are similar to those at the azo coupling test
reaction. However, although the performance of the Y-mixer
again is poorer compared to the case of the T-mixer at high
flow rates, no observable difference can be seen between the T-
and the arrowhead geometry. Here again, the frit as an
additional mixing element causes a significant enhancement of
the mixing performance (see Figure 8b). It must be noted that
Figure 7. Pictures of the fluid dyeing in the mixing channel of the the curves in Figure 8 are shown for a different range of flow
LTF-MS mixer for a total flow rate of 8 mL/min (a) and 16 mL/min rates compared to the case of Figure 6. A decrease of the DMP
(b). conversion occurs at significantly higher flow rates.
The fast decrease of the DMP conversion in the curve of the
At the higher flow rate, a uniform coloring of the fluid is SIMM (Figure 8d) indicates a superior mixing performance of
observed shortly after the entrance into the mixing channel this mixer, which is consistent with the results of the diazo
even though the residence time in the whole mixing channel is coupling experiments. It is the only investigated micromixer in
shorter. This indicates that the improvement of mixing by which DMP conversion could be avoided completely. However,
increasing the flow rate outweighs the reduction of the a minimum total flow rate of about 8 mL/min is necessary to
residence time. reach this performance. At lower flow rates, approximately pure
However, the HTM-ST-2-1 mixer performed very well laminar flow resulting in a bad mixing performance can be
compared to the other mixers of Little Things Factory GmbH. assumed.
The LTF-MX, on the other hand, was the only mixer unable to The above-mentioned disadvantages of the caterpillar mixer
reach yields lower than 5% even at higher flow rates. A possible compared to the SIMM can also be observed using the DMP
explanation is the mixing principle, which is based on the hydrolysis reaction system. As a result of the successive scaling
repeated change of the two levels of the mixer. The mixing down of lamellae width throughout the mixing channel, a
quality increases with the number of mixing elements. Hence, homogeneous concentration field cannot be created as fast as in

Figure 8. Experimental results of the DMP conversion XDMP as a function of the total flow rate for all investigated micromixers.

1519 dx.doi.org/10.1021/op300107z | Org. Process Res. Dev. 2012, 16, 1513−1522

Organic Process Research & Development Article

Figure 9. Results of the DMP hydrolysis test system (a) and diazo coupling test system (b) with regard to the measured pressure drop for
comparable flow rates. For each micromixer, the pressure drop was determined at the same flow rates that are shown in Figures 6 and 8.

the SIMM. Since DMP conversion proceeds in only a few mixing channel of the arrowhead mixer (i.d. = 0.5 mm) is
milliseconds, a considerable influence on XDMP results. comparatively moderate, the 150 μm structured mixing channel
According to the curve in Figure 8d, a DMP conversion of of the caterpillar mixer gives a very high loss of pressure even at
15% seems to be a minimum value in the investigated volume low flow rates. Thus, for both reaction systems, the caterpillar
flow range. mixer seems to be less energy efficient.
The curves in Figure 8c suggest that the investigated glass Due to the fine structures of the SIMM inlet, the pressure
mixers do not mix fast enough to suppress the DMP hydrolysis. drop in the upper range of flow rate is even higher in the SIMM
This observation makes sense, since even at the highest flow compared to the caterpillar mixer. Since the SIMM performs
rate, the residence time would be longer than 200 ms. For a low very well even at low flow rates, nevertheless, short mixing
DMP conversion, mixing on the molecular scale would have to times can be achieved with moderate pressure drop at limited
be completed after a few milliseconds, in the first millimeters of throughput.
the mixing channel. Further mixing elements do not have any In both parts a and b of Figure 9, similarities between the
influence on the DMP conversion. Therefore, mixers with curves of the T-mixer and the LTF-MS mixer can be identified.
dimensions in the range of 1 mm are inefficient if mixing times Only in these mixers, no splitting of the flow by the mixer
in the range of DMP hydrolysis are required. structure takes place. Thus, fast mixing can only be reached by
Mixing Performance and Pressure Drop. For the the formation of vortices in the mixing channel. It can be
evaluation of the mixing efficiency of a micromixer, the assumed that the extent of the vortex formation correlates with
required energy input of the mixing process has to be taken the energy input and the pressure drop, respectively.
into account. The pressure drop of a micromixer is directly As indicated by the significant differences of the curves in
related to the energy input. Ideally, most of the energy input is both figures, not only the energy input but also the dimensions
used effectively for mixing. In contrast, a poorly constructed and the mixing principle seem to have a great influence on the
micromixer shows a high pressure drop without using this mixing performance. In some mixers, for certain volume flows,
energy input for the mixing process. Therefore, the results of mixing may be already completed in the front part of the
both test reaction systems have to be discussed in terms of the mixing channel (e.g., LTF-MS). Therefore, only part of the
measured pressure drops (Figure 9). An efficient mixer energy input is effectively used for mixing. The measured
provides a certain mixing time short enough for a specific pressure drop is higher than the actual pressure drop needed to
application by causing a low pressure drop. To give an example, achieve a certain mixing level. Especially for the glass mixers
the results of the caterpillar mixer and the arrowhead mixer with comparably high inner volumes, this effect may lead to an
with a frit in the mixing channel indicate similar mixing underestimation of the mixing efficiency.
performance for both test reactions (see Figures 6 and 8). Micromixer Selection and Scale-up. The investigations
However, with regard to the pressure drop, the values differ by presented in this paper focus on the efficiency of micromixers.
nearly 2 orders of magnitude, as shown in Figure 9. While the In order to select a commercially available micromixer for a
pressure drop caused by the short frit (10 μm pore size) in the mixing sensitive reaction, the following procedure is proposed:
1520 dx.doi.org/10.1021/op300107z | Org. Process Res. Dev. 2012, 16, 1513−1522
Organic Process Research & Development Article

Table 3. Colored Representation of the Mixing Performance Based on the Yield of Bisazo Dye YS Using the Diazo Coupling
Test Reaction (t1/2 ≈ 730 ms)a

a
Green, best performance; red, poorest performance. Negative values can be attributed to the error in conjunction with the multilinear regression
analysis to determine the product concentrations out of the absorbance spectra (see Experimental Section).

Table 4. Colored Representation of the Mixing Performance Based on the DMP Conversion XDMP Using the DMP Hydrolysis
Test Reaction (t1/2 ≈ 5 ms)a

a
Green, best performance; red, poorest performance.

Table 5. Colored Representation of the Pressure Drop Δp of the Investigated Micromixers (Values in bar)a

a
Green, lowest pressure drop; red, highest pressure drop.

First, it has to be considered if the reaction system is mixing- system. The process has been carried out successfully in the
sensitive. As an experimental approach, the comparison of the laboratory scale with a slit interdigital micromixer at a total flow
product distribution resulting from mixing with two micro- rate of 10 mL/min. It is to be scaled-up to a significantly higher
mixers which strongly differ in their mixing performance could throughput of 50 mL/min. As can be seen in Table 5, the
show this mixing sensitivity. Furthermore, an approximate time SIMM is not suitable for such a high flow rate, since a pressure
range that is required for mixing should be known. On that drop higher than 7 bar would result. It is apparent from Table 4
basis, either Table 3 (t1/2 ≈ 730 ms) or Table 4 (t1/2 ≈ 5 ms)
that the arrowhead mixer including the short frit in the mixing
can be used to find an appropriate mixer type. In combination
channel is a good alternative in this range of mixing time.
with Table 5 (pressure drop), an overview is given to provide a
scale-up strategy for the use of different micromixing devices: A Furthermore, with a pressure drop of less than 2 bar at 48 mL/
scale-up to another micromixer can be considered to be assured min, this mixer can be used with a moderate energy input.
if both mixers show similar behavior for the test reaction. For the case that a mixing time of about 1 s is sufficient for
As an example, we assume a process that requires a mixing the process, Table 3 can be used to choose an appropriate
time in the range of only a few milliseconds. Therefore, the mixer. At 50 mL/min, most of the investigated micromixers
decision for the scale-up has to be based on the DMP reaction meet the requirements of the mixing performance. In this case,
1521 dx.doi.org/10.1021/op300107z | Org. Process Res. Dev. 2012, 16, 1513−1522
Organic Process Research & Development Article

a simple T-mixer would be the cheapest and most efficient financial support within the framework of the “MINT Projekt
choice. 2010”.
It has to be emphasized that, in both cases, the overview
Tables 3 and 4 only provide information about the ratio
between mixing time and the particular reaction time. If mixing
■ REFERENCES
(1) For reviews on mixing principles, see: (a) Hessel, V.; Löwe, H.;
time does not affect the selectivity of a reaction system, slower Schönfeld, F. Chem. Eng. Sci. 2005, 60, 2479−2501. (b) Nguyen, N.;
mixing does not have to be a disadvantage as long as a Wu, Z. J. Micromech. Microeng. 2005, 15, R1−R6.
homogeneous concentration field exists at the end of the (2) Guichardon, P.; Falk, L. Chem. Eng. Sci. 2000, 55, 4233−4243.
mixing channel. In order to achieve this state, the mixing time (3) Falk, L.; Commenge, J.-M. Characterization of Mixing and
has to be shorter than the residence time in the mixing device. Segregation in Homogeneous Flow Systems. In Micro Process
Engineering: A Comprehensive Handbook; Hessel, V., Renken, A.,
Nevertheless, information about the range of mixing time is Schouten, J. C., Yoshida, J., Eds.; WILEY-VCH: 2009; Vol. 1, pp 162−
important for the choice of an effective reactor. As shown in the 165.
sections above, in the mixers with a large internal volume (e.g., (4) Kölbl, A. Chem. Eng. J. 2008, 145, 176−177.
LTF-MS), effective mixing takes place in only a small part of (5) Baldyga, J.; Bourne, J. R. Turbulent Mixing and Chemical
the mixing channel. Therefore, part of the energy input is used Reactions; John Wiley & Sons: Chichester, 1999.
to mix a fluid with an already homogeneous concentration field. (6) Baldyga, J.; Bourne, J. R.; Walker, B. Can. J. Chem. Eng. 1998, 76,
In principle, the same scale-up concept as shown in this 641−649.
paper can be used for a scale-up to full-scale production. The (7) Bourne, J. R.; Kut, O. M.; Lenzner, J.; Maire, H. Ind. Eng. Chem.
mixing behavior at these significantly higher flow rates is an Res. 1990, 29, 1761−1765.
(8) Lindenberg, C.; Schöll, J.; Vicum, L.; Mazzotti, M.; Brozio, J.
interesting topic for further investigations. A possible setup for
Chem. Eng. Sci. 2008, 63, 4135−4149.
this purpose could include static mixers,13 the star laminator (9) Falk, L.; Commenge, J.-M. Chem. Eng. Sci. 2010, 65, 405−411.
from the IMM,14 or microreactors from Lonza.15

■
(10) Hassan, A. A.; Sandre, O.; Cabuil, V.; Tabeling, P. Chem.
Commun. 2008, 15, 1783−1785.
CONCLUSIONS (11) Hessel, V.; Hardt, S.; Löwe, H.; Schönfeld, F. AIChE J. 2003, 49,
The mixing behavior of selected micromixers was experimen- 566−577.
tally characterized with two different competitive reaction (12) Kockmann, N. Transport Phenomena in Micro Process Engineer-
ing; Springer-Verlag: 2007.
systems. The systems differ in their reaction rate, which made it (13) Brechtelsbauer, C.; Ricard, F. Org. Process Res. Dev. 2001, 5,
possible to evaluate the mixing performance of the different 646−651.
mixer types at two different time scales (∼5 ms and ∼730 ms). (14) Werner, B.; Hessel, V.; Löb, P. Chem. Eng. Technol. 2005, 28,
It has been shown that mixing times of only a few milliseconds 401−407.
can be achieved by using microstructured mixing channels. The (15) Kockmann, N.; Gottsponer, M.; Roberge, D. M. Chem. Eng. J.
investigated slit interdigital mixer exhibits superior mixing 2011, 167, 718−726.
performance even at low flow rates. It was found that an
arrowhead-mixer containing a short frit in the mixing channel
shows a very effective mixing behavior. Mixing devices with
structures in the range of a millimeter are suitable to achieve
mixing times in the time range of the slower reaction system
(>700 ms). Furthermore, it was shown that simple T-mixers
can be used for very fast mixing if the energy input is
sufficiently high. The influence of the energy input was studied
by measuring the pressure drop of the mixers. Although
pressure drop is a key parameter for mixing performance, the
results indicate that mixing time is not exclusively determined
by the overall energy input. It can be concluded that the mixing
principle as well as the mixer design and geometry have an
important influence on the effectiveness of a mixer.
The results that can be obtained by using test reaction
systems always must be considered in context with the
characteristic reaction time. An example scale-up consideration
for the selection of an appropriate micromixer for mixing
sensitive reactions based on the presented results is shown.

■ AUTHOR INFORMATION
Corresponding Author
*E-mail: [email protected]. Telephone: +49 621 292
6800.
Notes
The authors declare no competing financial interest.

■ ACKNOWLEDGMENTS
The authors would like to thank Novartis Pharma AG (CH-
4002 Basel, Switzerland) and Land Baden-Württemberg for the
1522 dx.doi.org/10.1021/op300107z | Org. Process Res. Dev. 2012, 16, 1513−1522