Fast and Efficient Acquisition of Kinetic Data in Microreactors Using In-Line Raman Analysis

Article abstract of DOI:10.1021/acs.oprd.5b00184

This study demonstrates that a microreactor setup with fast in-line reaction monitoring by Raman spectroscopy can be a highly efficient laboratory tool for kinetic studies and process development. Using a coaxial probe and commercial spectrometer to perform real-time measurements in the microchannel prevents the need for reaction quenching, sampling, and time-consuming off-line analysis methods such as GC or HPLC. A specially designed, temperature-controlled aluminum plate microreactor was developed and tested in the exothermic synthesis of 3-piperidino propionic acid ethyl ester by Michael addition. In-line measurements through a fused quartz screen in the reactor channel, which had an increasing cross-sectional area, allowed time-series kinetic data to be collected over nearly the full range of reaction conversions. An optimum flow rate range in which nearly ideal plug flow behavior can be assumed was identified. Furthermore, a time gradient was applied to the reactant flow rates, and the product concentration was simultaneously and repeatedly measured at various locations in the reactor channel. With this approach, the experiment duration and material consumption are significantly reduced relative to those of conventional steady-state experiments. Two hundred data points with residence times ranging from 0.3 to 49 s were collected in less than 1 h. Thus, this method can be used for the high-throughput screening of reaction parameters in a microreactor.

Full text of DOI:10.1021/acs.oprd.5b00184

Article
pubs.acs.org/OPRD
Fast and Efficient Acquisition of Kinetic Data in Microreactors Using
In-Line Raman Analysis
Sebastian Schwolow,†,⊥ Frank Braun, Matthias Rad
‡,⊥
le,‡,§ Norbert Kockmann, and Thorsten Roder*,†
∥
̈
̈
†
Mannheim University of Applied Sciences, Institute of Chemical Process Engineering, Paul-Wittsack-Str. 10, 68163 Mannheim,
Germany
‡
Mannheim University of Applied Sciences, Institute of Process Control and Innovative Energy Conversion, Paul-Wittsack-Str. 10,
6
8163 Mannheim, Germany
§
Heidelberg University and Mannheim University of Applied Sciences, Institute of Medical Technology, Paul-Wittsack-Str. 10, 68163
Mannheim, Germany
∥
TU Dortmund University, Biochemical and Chemical Engineering, Equipment Design, Emil-Figge-Straße 68, 44227 Dortmund,
Germany
*
S Supporting Information
ABSTRACT: This study demonstrates that a microreactor setup with fast in-line reaction monitoring by Raman spectroscopy
can be a highly efficient laboratory tool for kinetic studies and process development. Using a coaxial probe and commercial
spectrometer to perform real-time measurements in the microchannel prevents the need for reaction quenching, sampling, and
time-consuming off-line analysis methods such as GC or HPLC. A specially designed, temperature-controlled aluminum plate
microreactor was developed and tested in the exothermic synthesis of 3-piperidino propionic acid ethyl ester by Michael addition.
In-line measurements through a fused quartz screen in the reactor channel, which had an increasing cross-sectional area, allowed
time-series kinetic data to be collected over nearly the full range of reaction conversions. An optimum flow rate range in which
nearly ideal plug flow behavior can be assumed was identified. Furthermore, a time gradient was applied to the reactant flow rates,
and the product concentration was simultaneously and repeatedly measured at various locations in the reactor channel. With this
approach, the experiment duration and material consumption are significantly reduced relative to those of conventional steady-
state experiments. Two hundred data points with residence times ranging from 0.3 to 49 s were collected in less than 1 h. Thus,
this method can be used for the high-throughput screening of reaction parameters in a microreactor.
INTRODUCTION
integration into microreactor setups can be found in the
literature. In mid-IR spectroscopy, the fundamental vibrations
■
4
Microreactor systems have proven to be an effective tool for
process intensification and reaction development. The process
1
of structural groups are excited, and narrow bands are obtained,
providing highly specific information for the identification of
various molecular species. However, it is difficult to use this
technique to control reactions in a microreactor process
because thin optical layer thicknesses are required due to the
high molecular extinction coefficients in the mid-IR region.
Fiber optical materials, such as sapphire, chalcogenide glass, or
hollow glass waveguides, are expensive and difficult to assemble.
Furthermore, strong interference from water reduces the
sensitivity of mid-IR spectroscopy in aqueous systems.
Although near-IR spectroscopy is easier to implement for on-
line and in-line process monitoring, overtones and combina-
tions of fundamental vibrations result in broad bands, making
the spectra difficult to interpret. In principle, Raman spectros-
copy, which is often considered to be complementary to IR
spectroscopy, combines the main advantages of near- and mid-
IR technologies. This technique provides high chemical
specificity, and it can be easily integrated into continuous
processes by using common near-IR transmitting materials. In
addition, the weak Raman signal of water does not interfere
conditions of microreactors can be precisely defined due to
their well-defined flow patterns, high mixing performance, and
efficient heat removal. When these reactors are employed in
combination with laboratory automation systems, the reaction
conditions can be screened to obtain kinetic information using
a fully automated procedure, making the experiments material-
and time-efficient. In these types of studies, sampling and off-
line product analysis by HPLC or GC are often the time-
limiting steps. Furthermore, the reaction must be rapidly and
effectively quenched to determine the product compositions at
specific reaction times. The data acquisition time can be
reduced remarkably by integrating modern spectroscopic
detection techniques into the reactor system for in-line analysis.
Additional advantages of in-line analysis include the possible
detection of unstable intermediates, control of the reaction
progress by real-time monitoring, and inclusion of self-
2
optimizing algorithms in the laboratory automation system.
On-line reaction monitoring of continuous microreactor
systems has been accomplished using various spectroscopic
methods, such as fluorescence, ultraviolet−visible (UV/vis)3,
near- and mid-infrared (IR), and Raman spectroscopies.
Further details on the detection techniques and their
Received: June 8, 2015
©
XXXX American Chemical Society
A
DOI: 10.1021/acs.oprd.5b00184
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
with the signals of other species in aqueous systems. However,
the distinctive fluorescence background of some materials
overlaps with the Raman scattering peaks of interest.
Furthermore, comparatively high concentrations and high
laser power are required to overcome the low sensitivity.
Safety regulations for eyes and skin must be observed, and the
optical power has to be limited in hazardous areas. Along with
higher costs compared to other spectroscopic methods, this
prevents many researchers from using Raman spectroscopy.
This study investigated the potential of an automated
microreactor system with in-line Raman spectroscopic reaction
monitoring for use in kinetic studies. A microreactor with an
optimized channel design for fast and exothermic reactions was
developed. Because common flow cells can only cover a limited
range of reaction conversions at the reactor outlet, in-line
measurements throughout the entire channel length were
realized. The broad conversion range needed for reliable
conclusions on the reaction kinetics could be obtained by
switching between several measurement locations. A method
for collecting kinetic data under nonsteady-state conditions was
Figure 1. Channel design of the aluminum/glass reactor with the
numbered measurement locations 1 to 9 and a transition from 1 mm
channel width to 2 mm channel width at location 6 (a), milled reactor
channel in the mixing area (b), and CFD simulation of the flow
velocity distribution at one measurement location when the total flow
rate was 10 mL/min (Re = 164 with water, 20 °C) (c).
micromixer due to the high energy input resulting from the
5
recently published by Moore and Jensen. The combination of
high flow velocities in the small channels and to a sharp flow
this principle with multiple measurement locations was
investigated as a novel approach for fast and efficient
acquisition of kinetic data. Steady-state and nonsteady-state
methods were compared using the synthesis of 3-piperidino
propionic acid ethyl ester by Michael addition (Scheme 1) as a
7,8
redirection in the arrowhead mixer. The residence time
channel consisted of a 1 mm channel (total volume of 0.5 mL)
and a 2 mm channel (total volume of 1.5 mL). The 1 mm
channel width enabled good temperature control at the reactor
entrance, where the reaction is fast and hot spots might occur.
As the conversion increases, the third-order Michael addition
reaction rate decreases significantly; therefore, a 2 mm channel
was employed to achieve high conversions at longer residence
times. Measurement locations with diameters of 2 mm were
integrated into the smaller channel (Figure 1b) to facilitate
focusing the laser beam. The design of the measurement
locations was optimized by CFD simulations (Figure 1c). At a
total flow rate of 10 mL/min, the Reynolds number (Re = ud/
ν) was 164 at the measurement location inlet. Due to the
Scheme 1. Synthesis of 3-piperidino propionic acid ethyl
ester
model reaction. The reaction behavior and a kinetic model of
6
9
this reaction were described in a previous publication. The
induced secondary flow, the concentration distribution in the
kinetic model (see Supporting Information), which is based on
off-line GC analyses, was used as a reference to evaluate the
new results obtained by more time- and material-efficient
analysis methods. Because the Michael addition reaction is
significantly accelerated by the addition of water, the weak
spectroscopic signal of water is a crucial advantage of Raman
analysis over near- and mid-IR spectroscopies.
region of the focused laser beam could be assumed to be
uniform. Furthermore, the secondary flow caused by repeatedly
redirecting the flow by 90° (creating a zigzag pattern) made the
radial concentration profiles more uniform during laminar flow.
Measurements show that narrow residence time distributions
and reactor behavior close to that of an ideal plug flow reactor
can be obtained with this channel design. Experimental data on
the residence time distribution for a similar reactor channel (2
MATERIALS AND METHODS
6
■
mm channel width) is available in a recent publication. Flow
Reactor Design. For this study, a plate microreactor was
designed and manufactured by mechanical precision milling of
aluminum. The entire reactor channel was covered with a fused
quartz plate, which was evenly pressed onto the reactor plate to
avoid significant bypass flow. For a suitable reactor design, a
first rough estimation of the reaction time scale and the heat of
reaction should be available. In case of the Michael addition
with a reaction time below 1 min and an adiabatic temperature
increase of approximately 190 K, the applied reactor was
optimized for a fast and exothermic reaction. The main factors
considered in the reactor channel design (Figure 1a) were the
mixing performance, residence time distribution, integration of
the optical measurements, and heat transfer.
channels for cooling water or a heat carrier (5 mm × 6 mm)
were milled on the opposite side of the reaction plate. Thus, the
reactor temperature could be uniformly maintained at a
specified value, and the reaction heat of the Michael addition
could be efficiently removed. Compared to common glass
reactors, the aluminum/glass combination allows for signifi-
cantly improved temperature control for exothermic reactions
due to the high thermal conductivity of aluminum, which is
over 100 times higher than that of glass (AlZnMgCu , λ = 140
1
.5
W/(m·K)). The results of an infrared camera temperature
measurement and measurements of the channel surface
roughness by white light interferometry can be found in the
Supporting Information.
The reactor had channels of various sizes, and each channel
had a quadratic cross-sectional area. The reactants were
preheated in 1 mm channels and mixed in 0.5 mm channels,
which were arranged in an arrow-shaped geometry (Figure 1b).
A high mixing performance could be obtained by this type of
Experimental Setup. 3-Piperidino propionic acid ethyl
ester was synthesized without any solvent. However, because
10
water acts as a catalyst in this reaction, piperidine was
premixed with 10.2 wt % of water. Neat ethyl acrylate was used.
For all of the experiments, the reactant feeds were mixed at
B
DOI: 10.1021/acs.oprd.5b00184
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
equal flow rates. Hence, the molar ratio was 1.01 equiv of
piperidine per equivalent of ethyl acrylate. Both feed streams
were supplied by continuous syringe pumps (Syrdos 2, HiTec
Zang GmbH, Germany) with 10 mL glass syringes. By using a
laboratory automation system (LabBox and LabVision
Software, HiTec Zang GmbH), the flow rate settings for both
the steady-state experiments and experiments with a con-
tinuously decreasing flow velocity could be controlled. Cooling
water was flowed through the channels on the other side of the
reactor plate to maintain the reactor temperature at 25 °C.
In-line Raman reaction monitoring was performed using a
highly sensitive Raman system (MultiSpec Raman, tec5 AG,
Germany) equipped with a coaxial probe (InPhotonics, Inc.,
USA). This Raman system is employed in process environ-
ments with a thermoelectrically cooled CCD array. Raman
scattering was excited by a temperature-stabilized semi-
conductor laser source at 785 nm with a 500 mW output
power (approximately 260 mW after losses from optical
components) and then coupled into the optical fiber. The
The second data acquisition method was based on a new
approach developed by Moore and Jensen for rapidly
5
generating kinetic data using in-line analysis. These researchers
applied residence time gradients by manipulating the flow rate
and thus generated conversion/residence time profiles at
different temperatures. Further details on this approach can
be found in their publication.
In this study, all of the measurements performed with a
residence time gradient were started under steady-state
conditions with a total flow rate Q of 20 mL/min, resulting
0
in a short initial residence time τ = V /Q . From this starting
0
r
0
point (t = 0), a constant gradient α was applied to the
instantaneous residence time τ , which is calculated at each
ins
experiment time t using the equation
τins = τ + αt
(1)
0
The time profile of the flow rate Q(t) is obtained from the
following equation:
Q(t) = V
=
Vr
−1
r
Stokes Raman spectra (300−3100 cm ) were collected with an
−1
τ
τ0 + αt
optical wavelength resolution of 5 cm . Figure 2 shows a
ins
(2)
For each measurement location, the effective reactor volume
Vr was defined as the channel volume between the mixer and
the measurement location. The gradient coefficient α is given
by
⎛
end ⎝
⎞
0 ⎠
Vr
1
Q end
− Q1
α = t
⎜
⎜
⎟
⎟
(3)
where Qend is the lowest flow rate at the end of the
measurement (t = tend). Figure 3 shows the residence time
and flow rate as a function of time at measurement location 8
when the gradient α was 0.057.
Figure 2. Schematic of the experimental setup for the Raman
spectroscopic measurements in the microreactor channel.
schematic of the entire setup for performing measurements in
the microreactor channel. The probe was connected to a Z-
positioner, which allowed for tracking in the vertical direction.
For lateral positioning, the reactor was firmly attached to an XY
cross table. Prior to each measurement series, the focal point of
the probe (approximately 150 μm spot diameter) was adjusted
to the center of the measurement location by aiming it at the
maximum signal intensity.
Figure 3. Time profiles of the total flow rate Q, instantaneous
residence time τ , and fluid element residence time τ in the
ins
nonsteady-state experiments at measurement location 8.
Data Acquisition Methods. Two different data acquisition
methods were employed to obtain kinetic information from the
spectroscopic measurements performed in the reactor channel.
Raman spectra were recorded under (1) steady-state conditions
and (2) nonsteady-state conditions with a flow rate gradient. All
of the steady-state measurements were performed after running
the reaction at a constant flow rate for at least ten times the
hydrodynamic residence time. Because the Raman signal
intensity is sensitive to small deviations in the probe position,
the spectra of both reactant streams were recorded at all of the
measurement locations for each vertical and lateral probe
position. Thus, normalized results could be obtained by
calculating the chemical conversion. Steady-state measurements
were performed in the reaction mixture using an integration
time of 6.5 s and were repeated for each series of flow rates at
different measurement locations.
The reaction conversion in a fluid element passing the
Raman probe depends on the residence time τ, which is the
amount of time that the fluid element spends in the effective
reactor volume from an initial time t to the time of
i
measurement tf :
τ = t − ti
(4)
f
For a fluid element traveling through the reactor volume with
this residence time, the effective reactor volume is
tf
tf
Vr
τ0 + αt
Vr =
Q(t) dt =
dt
∫t
∫t
i
i
(5)
Integrating and rearranging eq 5 yields t as a function of t :
i
f
C
DOI: 10.1021/acs.oprd.5b00184
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
−
α
τ (1 − e )
−
α
0
t = t e
−
i
f
(6)
α
Finally, substituting eq 6 into eq 4 yields
⎛
⎜
τ ⎞
0
−
α
τ = (1 − e )
tf +
⎟
⎝
⎠
(7)
α
Hence, the residence time τ of fluid elements at the
−α
measurement location increases with the gradient (1 − e ). In
the experiments, when the residence time started to increase,
the Raman spectra measurements were initiated and collected
with integration times of 6.5 s at intervals of 13 s.
Measurements were repeated with identical flow rate gradients
(
Q = 20 mL/min; Q
= 6 mL/min; tend = 150 s) at
0
end
measurement locations 1 to 9 (Figure 1a). With the effective
reactor volume V , the gradient coefficient α for each
r
measurement location results from eq 3. The applied flow
rate gradient was chosen to ensure that the measurement time
scale was small compared to the time scale of the residence
time gradient. For example, at the first measurement location
where the reaction was the fastest, the residence time changed
by only 0.03 s during the integration time.
Raman Spectra. Figure 5 shows the typical Raman spectra
obtained from measurements collected when a flow rate
gradient was applied. They consist of spectra arising from
piperidine, ethyl acrylate, and 3-piperidino propionic acid ethyl
ester. The spectra of the pure compounds are shown in Figure
Figure 5. Raman spectra collected in the reactor channel during cyclic
measurements as the residence time increased from 0.9 s (blue line) to
2
.6 s (red line).
quantified by calculating the difference between the intensities
−1
at 1550 and 1637 cm .
4
for comparison. The reaction mixture also included water,
RESULTS AND DISCUSSION
■
Steady-State Experiments. To obtain reliable experimen-
tal kinetic data, the residence time must be clearly defined, and
the reaction must start almost instantaneously. Thus, the
following conditions must be met in the microreactor: (1) the
residence time distribution in the reactor must be close to that
in a plug flow reactor and (2) the chemical conversion rate
cannot be limited by poor mixing. Therefore, the radial
concentration profiles must become uniform quickly. This
process can be significantly enhanced by the formation of
12
secondary vortices, which is mainly determined by the flow
velocities in the micromixer and in the zigzag channel.
Conversely, at low flow rates, the flow behavior increasingly
deviates from the ideal behavior, making the experimental
results unreliable.
In Figure 6, the steady-state conversions measured at
different total flow rates are compared to those obtained by a
kinetic model of the reaction based on previous experiments
4
analyzed by off-line methods. The reaction volumes of all data
series collected at the different measurement locations range
from 0.1 to 1.6 mL. The ranges of the residence times overlap
due to the variation in the flow rate at each location. Thus,
nonkinetic limitations on the conversion, which depend on the
flow velocity, can be identified. At high flow rates, the various
data series are in good agreement with each other and with the
kinetic model. The conversions decrease slightly at a flow rate
of 4 mL/min, as shown in Figure 6, whereas the conversions
deviate significantly from the kinetic model at the lowest flow
rate of 2 mL/min (Re ≈ 30). These deviations can be partially
explained by the reduced mixing efficiency because our results
are in good agreement with those from other experimental
Figure 4. Raman spectra of piperidine (a), ethyl acrylate (b), and 3-
piperidino propionic acid ethyl ester (c).
which has no significant Raman bands in the measurement
range. The spectra are dominated by overlapping signals from
−1
the fused quartz cover up to 600 cm . Piperidine can be
−1
identified by a sharp Raman band at 813 cm , whereas the
band at 760 cm− is characteristic of the product. Ethyl acrylate
1
−1
exhibits a band at 1637 cm , which can be assigned to its
11
characteristic vinyl stretching mode. Because this band is well-
defined and does not overlap with the bands of other
compounds, it was used to determine the reaction conversion.
The inset in Figure 5 shows the decrease in this band with
increasing residence time in the nonsteady-state experiments.
To perform a simple baseline correction, this peak was
7
investigations that showed that the mixing performance of a
similar arrowhead micromixer decreased considerably when the
flow rate was reduced from 8 to 2 mL/min. Additionally, the
lower conversions obtained at low flow rates can be attributed
to less secondary flow in the 90° redirections of the zigzag
D
DOI: 10.1021/acs.oprd.5b00184
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Figure 6. Comparison of the ethyl acrylate conversions determined by in-line Raman spectroscopy under steady-state conditions to those obtained
6
by the kinetic model of Schwolow et al.
Figure 7. Comparison of the ethyl acrylate conversions determined by in-line Raman spectroscopy under nonsteady-state conditions to those
6
obtained by the kinetic model of Schwolow et al.
channel and in the measurement locations. As a result, the
residence time distributions are broader, leading to deviations
from ideal plug flow behavior in the reactor. The reproducibility
of the results was investigated by repeated measurements
including repeated adjustments of the Raman probe position.
For measurements at flow rates above 4 mL/min, standard
deviations in the reaction conversion were below 2%. Higher
standard deviations were found for low flow rates (approx-
imately 4% at 2 mL/min).
Hence, the flow rates in this microreactor should be limited
to the range of 6−20 mL/min for a reaction mixture with the
same viscosity as that used in this study (Re = 100−330).
Higher flow rates are possible in microreactors, but they are not
beneficial because they require higher material consumption
and lead to larger pressure drops in the reactor.
Nonsteady-State Experiments. Equation 7, which is used
to calculate the residence time of a fluid element from the
gradient α, is only valid for nearly ideal plug flow behavior.
Thus, the aforementioned requirements are even more
important for the nonsteady-state experiments. Hence, the
flow rates were limited to those at which the mixing conditions
did not affect the reactor performance in the steady-state
experiments (Figure 6).
All of the data obtained from the experiments in which
residence time gradients were applied at the eight different
measurement locations in the reaction channel are shown in
Figure 7. Again, the results do not systematically deviate from
those obtained by the kinetic model based on off-line GC
analysis. The standard deviation between the model and
experimental nonsteady-state conversions is 3.1%, which is even
lower than that between the model and experimental steady-
state conversions (3.6% for the results acquired in the same
flow rate range (6−20 mL/min)).
Method Efficiency. Previously published kinetic inves-
tigations of the Michael addition reaction that utilized off-line
analysis techniques were performed with an automated
microreactor system, continuous quench system, autosampler,
6
and gas chromatography. The completion of a time series with
one set of parameters and ten data points, including the sample
preparation and GC analyses, required approximately 3 h. Large
amounts of solvent with a flow rate ratio of 10:1 were required
to effectively quench the reaction.
In this study, the experiments required to obtain the same
amount of data were performed in only 20 min under steady-
state conditions by employing a microreactor setup with in-line
Raman spectroscopic reaction monitoring. By increasing the
residence time at a constant gradient, the reaction time was
further reduced to less than 3 min while simultaneously
increasing the number of data points. Because the reaction
could be monitored at different locations in the reactor channel,
the residence time could be varied by a factor of 163 between
0
.3 and 49 s. In contrast, when the reaction was monitored by a
flow cell at the reactor outlet, the range of possible residence
times was limited by the range of suitable flow rates for the
microreactor system. Therefore, several microreactors with
different inner volumes were required to study the second-
order reaction at both low and high conversions, which added
to the time required to change the setup and clean the reactor.
Based on these considerations, the amount of time required
to conduct the experiments to determine the rate constant for
one set of reaction conditions was estimated (Table 1). In-line
analysis reduces the experimental time because the effective
reaction volume can be rapidly changed by changing the probe
position. Additionally, time-consuming sample preparation
procedures and GC analyses are not required before the
results can be evaluated. In the nonsteady-state experiments,
the experimental time, which mainly depends on the residence
E
DOI: 10.1021/acs.oprd.5b00184
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Table 1. Estimated time requirements to determine the
conversion−time relation experimentally using at least 30
data points
Brief description of the used kinetic model; IR camera
temperature measurement; measurement of the surface
roughness (PDF)
Experiment
Analysis
6 h
AUTHOR INFORMATION
Off-line analysis (GC), steady state
In-line analysis (Raman), steady state
In-line analysis (Raman), nonsteady state
4 h
3 h
1 h
■
Instantaneous
Instantaneous
Corresponding Author
*E-mail: t.roeder@hs-mannheim.de. Telephone: +49 621 292
800.
6
Author Contributions
⊥
S.S. and F.B. contributed equally to the experimental work.
time gradient α, can be further reduced, making the nonsteady-
state technique a highly efficient method for kinetic studies. For
example, the 200 data points shown in Figure 7 were measured
at eight different locations in the reactor channel in less than 1
h. The efficiency would be further improved by the use of a
motorized cross table in the experimental setup. Hence, the
probe position could be changed by an automated procedure to
investigate the full residence time range, which could then be
repeated for reaction parameter screening. Alternatively, the
realization of multipoint measurements with a fiber multiplexer
would enable simultaneous data acquisition at several measure-
ment locations during the residence time gradient program.
Funding
This work was funded by the German Federal Ministry of
Education and Research (BMBF, Funding Code 03FH012I2)
and the German Federation of Industrial Research Associations
(
AiF Project GmbH, Funding Code 2035756LW3).
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
(
The authors would like to thank Andreas Neumu
Hochschule Mannheim, Germany) for constructing and
̈
ller
manufacturing the plate microreactor, Dr. Hanns Simon
Eckhardt (tec5 AG, Germany), for providing the Raman
probe and technical support, and Michael Ruland (Hochschule
Mannheim, Germany) for measurements of the surface
roughness.
CONCLUSIONS
■
In this study, an approach for combining microreactor
technology with fiber-optic Raman spectroscopy was described
and optimized for kinetic studies of rapid exothermic reactions.
This reactor setup is advantageous compared to the use of flow
cells for spectroscopic analysis because measurements can be
performed at different locations throughout the channel. Thus,
a large range of residence times could be explored while
keeping the flow rate within a relatively small range. This
advantage is crucial because the optimal flow rate range is
limited by the dependence of the reactor behavior on the flow
velocity. In these experiments, nonkinetic limitations due to
poor mixing or dispersion effects are observed at flow rates
below 6 mL/min. Instead of employing several reactors of
different sizes, channels with different cross-sectional areas were
combined. Hence, reaction data could be collected at both high
and low conversions by varying the measurement location.
Our results are in excellent agreement with those obtained by
a kinetic model based on experiments that utilized capillary
reactors and off-line GC analysis for total flow rates between 6
and 20 mL/min. Thus, it can be assumed that this analytical
method is accurate and that ideal plug flow behavior occurs in
the microreactor.
Performing real-time measurements at various locations in
the reactor results in a considerable increase in the experimental
throughput. The direct measurements in the reaction channel
were combined with a decreasing flow rate to dynamically
measure the conversion. As a result, time-series reaction data
over nearly the full range of reaction conversions were collected
in approximately one-third of the time required for steady-state
experiments. Combining this reactor system with a laboratory
automation system would enable the rapid screening of reaction
parameters to find the optimal conditions in a minimal amount
of time with minimal reactant consumption.
REFERENCES
1) (a) Geyer, K.; Codee
■
(
2
́
, J. D. C.; Seeberger, P. H. Chem. - Eur. J.
006, 12, 8434−8442. (b) McMullen, J. P.; Jensen, K. F. Annu. Rev.
Anal. Chem. 2010, 3, 19−42. (c) Hessel, V.; Kralisch, D.; Kockmann,
N.; Noel, T.; Wang, Q. ChemSusChem 2013, 6, 746−789.
̈
(2) (a) McMullen, J. P.; Stone, M. T.; Buchwald, S. L.; Jensen, K. F.
Angew. Chem., Int. Ed. 2010, 49, 7076−7080. (b) Rasheed, M.; Wirth,
T. Angew. Chem., Int. Ed. 2011, 50, 357−358.
(
̈
3) (a) Fraulin, C.; Rinke, G.; Dittmeyer, R.; Waters, G.; Nilles, M.
Chem. Ing. Tech. 2014, 86, 524−528. (b) Yadav, A. K.; Krell, M.;
Hergeth, W.-D.; de la Cal, J. C.; Barandiaran, M. J. Macromol. React.
Eng. 2014, 8, 543−549. (c) Rinke, G.; Ewinger, A.; Kerschbaum, S.;
Rinke, M. Microfluid. Nanofluid. 2011, 10, 145−153. (d) Mozharov, S.;
Nordon, A.; Littlejohn, D.; Wiles, C.; Watts, P.; Dallin, P.; Girkin, J. M.
J. Am. Chem. Soc. 2011, 133, 3601−3608.
(4) (a) Yue, J.; Schouten, J. C.; Nijhuis, T. A. Ind. Eng. Chem. Res.
2012, 51, 14583−14609. (b) Kessler, R. W. Prozessanalytik: Strategien
und Fallbeispiele aus der industriellen Praxis; Wiley-VCH: Weinheim,
2006.
(
5) Moore, J. S.; Jensen, K. F. Angew. Chem. 2014, 126, 480−483.
6) Schwolow, S.; Heikenwalder, B.; Abahmane, L.; Kockmann, N.;
der, T. Org. Process Res. Dev. 2014, 18, 1535−1544.
7) Schwolow, S.; Hollmann, J.; Schenkel, B.; Roder, T. Org. Process
Res. Dev. 2012, 16, 1513−1522.
8) (a) Hessel, V.; Lowe, H.; Scho
479−2501. (b) Soleymani, A.; Kolehmainen, E.; Turunen, I. Chem.
(
Ro
(
̈
̈
̈
(
̈
̈
nfeld, F. Chem. Eng. Sci. 2005, 60,
2
Eng. J. 2008, 135, S219−S228. (c) Kockmann, N. Proc. Inst. Mech. Eng.,
Part C 2008, 222, 807−816. (d) Wong, S.; Ward, M.; Wharton, C.
Sens. Actuators, B 2004, 100, 359−379.
(9) (a) Holvey, C. P.; Roberge, D. M.; Gottsponer, M.; Kockmann,
N.; Macchi, A. Chem. Eng. Process. 2011, 50, 1069−1075. (b) Jiang, F.;
̈
̈
Drese, K. S.; Hardt, S.; Kupper, M.; Schonfeld, F. AIChE J. 2004, 50,
2
297−2305. (c) Howell, P. B., Jr.; Mott, D. R.; Golden, J. P.; Ligler, F.
ASSOCIATED CONTENT
Supporting Information
The Supporting Information is available free of charge on the
ACS Publications website at DOI: 10.1021/acs.oprd.5b00184.
■
S. Lab Chip 2004, 4, 663−669.
(
(11) Pouchert, C. J. The Aldrich Library of Infrared Spectra; Sigma
*
S
10) Ranu, B. C.; Banerjee, S. Tetrahedron Lett. 2007, 48, 141−143.
Aldrich: Milwankee, WI, 1981.
F
DOI: 10.1021/acs.oprd.5b00184
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
(
12) (a) Saxena, A. K.; Nigam, K. D. P. AIChE J. 1984, 30, 363−368.
(
b) Klutz, S.; Kurt, S. K.; Lobedann, M.; Kockmann, N. Chem. Eng.
Res. Des. 2015, 95, 22−33. (c) Castelain, C.; Mokrani, A.;
Legentilhomme, P.; Peerhossaini, H. Exp. Fluids 1997, 22, 359−368.
G
DOI: 10.1021/acs.oprd.5b00184
Org. Process Res. Dev. XXXX, XXX, XXX−XXX