Novel innovation systems for a cellular approach to continuous process chemistry from discovery to market

Article abstract of DOI:10.1021/op049970n

Continuous processing of liquid/liquid synthesis and microreaction technology are shown to reduce the cost of process development and manufacturing of active pharmaceutical ingredients and other functional molecules on a commercial scale. Combinatorial synthesis systems for continuous chemistry are introduced, and their applications are described. Reactions within these systems scale seamlessly in standardized commercial continuous synthesis equipment allowing rapid access to kilogram quantities of advanced intermediates. Chemical and process development within such systems are illustrated by a case study of a continuous multistep process. Additionally, another case study shows the benefit of microreaction technology in the manufacture of high value added functional chemicals.

Full text of DOI:10.1021/op049970n

Organic Process Research & Development 2004, 8, 440−454
Technology Reports
Novel Innovation Systems for a Cellular Approach to Continuous Process
Chemistry from Discovery to Market
Thomas Schwalbe,*,† Volker Autze,‡ Michael Hohmann,‡ and Wolfgang Stirner‡
Cellular Process Chemistry, Inc., One Broadway Street, Suite 600, Cambridge, Massachusetts 02142, U.S.A.,
and CPC - Cellular Process Chemistry Systems GmbH, Hanauer Landstrasse 526, 60343 Frankfurt/Main, Germany
Abstract:
Continuous processing of liquid/liquid synthesis and micro-
reaction technology are shown to reduce the cost of process
development and manufacturing of active pharmaceutical
ingredients and other functional molecules on a commercial
scale. Combinatorial synthesis systems for continuous chemistry
are introduced, and their applications are described. Reactions
within these systems scale seamlessly in standardized com-
mercial continuous synthesis equipment allowing rapid access
to kilogram quantities of advanced intermediates. Chemical and
process development within such systems are illustrated by a
case study of a continuous multistep process. Additionally,
another case study shows the benefit of microreaction technol-
ogy in the manufacture of high value added functional chemi-
cals.
Figure 1. A microreactor integrating thermal conditioning,
mixing, and thermal reaction control into one module.
significant boundary condition for a continuous manufactur-
ing strategy for compounds demands that a manufacturing
process can be set up before quality specification such as
those required in the drug master file. In these stages of
synthesis development, process materials are scarce, time is
of the essence, and hence extensive sourcing of process
development material is counterproductive. Consistently
implemented microreaction2 technology3 enables a solution
to these challenges: A broad synthetic applicability of
standard microreactors4 (Figure 1) permits even difficult
reactions.5 Low internal volumes and optimized flow profiles
enable fast process development at small compound con-
sumption.
Parallel reactor arrays can give access to commercial scale
manufacturing quantities. In summary, microreaction tech-
nology enables continuous processing for high value added
chemicals. In this contribution we wish to give an overview
of the basics of the technology and their utilization in
synthesis. Experimental procedures are given for new ap-
plications.6
1. Introduction
Recent technological innovation in the field of micro-
reaction technology or process intensification has combined
with an economically driven search for the benefit of
continuous chemical processing in the field of high value
added compounds, in particular active pharmaceutical in-
gredients (API). Continuous processing predominates in
many industries, where it is perceived to generate lower
operating cost and enhanced process safety.
Moreover there is now general consensus that continuous
processes are operationally more stable and reproducible and
hence yield more consistent product quality. However, in
chemicals manufacturing, continuous processes have mostly
been applied in dedicated production facilities. In such a
highly inflexible setup capital investment is comparatively
large. Further, a significant adaptation to varying output rates
is difficult. By contrast, traditional batch production plants
have been flexible towards product changes, requiring lower
capital investment and utilizing small discrete steps to the
changing of manufacturing quantity schemes. Therefore,
from a manufacturing perspective there is a challenge to
evolve with a new continuous processing answer1 to the
target of uniting the traditional logistic convenience of batch
production with the advantages of continuous processing. A
(2) Floyd, T. M.; Kopp, M. W.; Firebaugh, S. L.; Jensen, K. F.; Schmidt, M.
A. In Proceedings of the 3rd International Conference on Microreaction
Technology (IMRET 3), Frankfurt/Main; Ehrfeld, W., Ed.; Springer: Berlin,
1999; p 171.
(3) (a) Ja¨hnisch, K.; Hessel, V.; Lo¨we, H.; Baerns, M. Angew. Chem., Int. Ed.
2004, 43, 406. (b) Schwalbe, T.; Autze, V.; Wille, G. Chimia 2002, 56,
636.
(4) The broad applicability of such systems results from an integration of a
thermal conditioner for starting materials, a mixer for these, and an
integrated heat exchanger for thermal reaction control. This patented
integration is commercially available as CYTOS.
* Author for correspondence. E-mail: schwalbe@cpc-net.com
† Cellular Process Chemistry, Inc.
‡ CPC - Cellular Process Chemistry Systems GmbH.
(1) (a) Behr, A.; Brehme, V. A.; Ewers, C. L. J.; Gro¨n, H.; Kimmel, T.;
Ku¨ppers, S.; Symietz, I. Chem. -Ing. -Tech. 2003, 75, 417. (b) Ewers, C.
L. J.; Ku¨ppers, S. Nachr. Chem., Tech. Lab. 2002, 50, 1255.
440
•
Vol. 8, No. 3, 2004 / Organic Process Research & Development
10.1021/op049970n CCC: $27.50 © 2004 American Chemical Society
Published on Web 05/01/2004

Figure 2. Potential energy profile: kinetic reaction vs thermodynamic reaction, catalysis.
2. Generic Benefits from Integrated Microreactors in
Synthesis
susceptible to macroscopic flow tubes. Finally, practically
any reaction operated in a microreactor of suitable dimension
can be immediately operated to yield kilogram quantities of
synthetic material at a day’s notice. However, one problem-
atic aspect connected with the handling of the systems should
be mentioned. In some cases, preceding optimization is
essential before submitting a reaction from the conventional
mode to the microreactor. Since the microchannels are
restricted to reactions in a liquid/liquid, liquid/gas, or gas/
gas phase system, solid reagents have to be replaced by
soluble alternatives. This process may require additional time
for the optimization of conditions. In addition, the time
saving effect of continuously operating synthesis may be
compensated by standard workup protocols that are mostly
still based on conventional methods. Due to the efficient
mixing in a microreactor, for the rest of the discussion we
can assume mixing is nearly ideal. Under these conditions
it is appropriate to focus the discussion of reactions on the
potential energy profile as related to reaction temperature
along the reaction coordinate and in passing through transi-
tion states (Figure 2, lit.9). On the left side of Figure 2, a
reaction is shown to proceed under kinetic control, and the
formation of product C is favored if the activation energy
provided by the reaction is limited. These reactions are most
favorably run in MRs. A reversible reaction path allows for
the predominant formation of the more stable product through
reversible reaction conduct (thermodynamic control, right).
Whilst such reactions can in principle be run in a MR, there
are few cases where we would expect incrementally bene-
ficial results.
Microreactors (MRs) possess certain advantageous engi-
neering features,7 for example, fast diffusive mixing8 and
high heat transfer rates. Further, a tightened residence time
distribution can lead to elution profiles that more closely
approximate plug flow than, for instance, the elution profiles
(5) The breadth of applicability of microreactors was reported: (a) Chambers,
R. D.; Spink, R. C. H. Chem. Commun. 1999, 883. (b) DeWitt, S. H.; Curr.
Opin. Chem. Biol. 1999, 3, 350. (c) Okamoto, H. J. Synth. Org. Chem.
Jpn. 1999, 57, 805. (d) Krummradt, H.; Kopp, U.; Stoldt, J. In Proceedings
of the 3rd International Conference on Microreaction Technology (IMRET
3), Frankfurt/Main; Ehrfeld, W., Ed.; Springer: Berlin, 1999; p 181. (e)
Autze, V.; Kleemann, A.; Golbig, K. Nachr. Chem., Tech. Lab. 2000, 48,
683. (f) Example reactions are listed at: www.cpc-net.com. (g) Fo¨rster,
A. Chem. -Ing. -Tech. plus 2001, 4, 28. (h) Hessel, V.; Lo¨we, H.; Chem.
-Ing. Tech. 2002, 74, 186. (i) Wo¨rz, O.; Ja¨ckel, K. P.; Richter, K. P.; Wolf,
A. Chem. Eng. Technol. 2001, 24, 138. (k) Haswell, S. J.; Middleton, R.
J.; O’Sullivan, B.; Skelton, V.; Watts, P.; Styring, P. Chem. Commun. 2001,
391. (l) Watts, P.; Wiles, Ch.; Haswell, S. J.; Pombo-Villar, E. Tetrahedron
2002, 58, 5427. (m) Fletcher, P. D. I.; Haswell, S. J.; E. Pombo-Villar, E.;
Warrington, B. H.; Watts, P.; Wong, S. Y. F.; Zhang, X. Tetrahedron 2002,
58, 4735.
(6) All experiments described in the CYTOS Lab System, unless otherwise
indicated, were carried out in the CYTOS microreactor developed by CPC
- Cellular Process Chemistry Systems GmbH, Germany. Schwalbe, Th.;
Golbig, K.; Hohmann, M.; Georg, P.; Oberbeck, A.; Dittmann, B.; Stasna,
J.; Oberbeck, S. Eur. Pat. Appl. EP 1 123 734, 2001; Chem. Abstr. 2001,
135, 154468b. The effective reaction volume of the CYTOS microreactor
is 1.2 mL; however, along with the feed lines inside the cell the total volume
results in 2.0 mL. Additional reaction volume is provided by equipping
the microreactor with up to three residence time units with a 15 mL volume
each in the standard configuration. Further information is available on the
Internet: www.cpc-net.com. CYTOS is registered by CPC-Systems GmbH.
(7) Taghavi-Moghadam, S.; Kleemann, A.; Golbig, K. G. Org. Process Res.
DeV. 2001, 5, 652.
(8) To practically benefit from the use of microreactors in synthesis, it is
worthwhile to conceptually separate the effects of mass transfer from
thermal effects onto a reaction. For practical synthetic considerations,
diffusive mixing in microreactors is significantly faster than reacting. This
has been amply discussed in the literature, and mixing in microreactors is
commonly characterized by the Villermaux/Dushman (Fournier, M. C.;
Falk, L.; Villermaux, K. J. Chem. Eng. Sci. 1996, 51, 5187) reaction. Further
well-designed microreactors demonstrate a stable mixing performance over
a wide range of flows. By and large limitations to the comparative value
of this characterization only surface with the occurrence of multiple phases
as well as with widely differing viscosities, hence mixing energy require-
ments in the starting media or changes of viscosity after mixing. Hence,
alternative characterization reactions have been proposed (Panic, S.; Antes,
J.; Richert, B.; Boskovic, D.; Tuercke, T.; Schnuerer, F.; Marioth, E.;
Loebbecke S. In Abstracts of Papers, 7th International Conference on
Microreaction Technology (IMRET 7); Lausanne, Switzerland; DECHEMA-
Gesellschaft fu¨r in Chemische Technik und Biotechnologie: Frankfurt/
Main, Germany, 2003; p 236.
From an experimental operating perspective the continu-
ous conduct of microreactor experimentation avoids dosing
times, yet it is due to the integration of conditioning heat
exchangers into microreactors that undesired side reactions
such as thermal decomposition or reactive aggregation such
as polymerization of starting materials can be largely
avoided. In the conceptual view of a reaction profile, this
conditioning works in a direction as if the reaction compo-
(9) Morrison, R. T.; Boyd, R. N. Organic Chemistry, 4th ed.; Allyn and
Bacon: Boston, MA, 1983.
Vol. 8, No. 3, 2004 / Organic Process Research & Development
•
441

Figure 3. Temperature and activation energy, temperature quality, and activation energy.
Scheme 1. Paal-Knorr synthesis of pyrrole 3
Scheme 2. Discrimination between the double-nitrated
products 6 and 7 according to Loebbecke et al.a
Table 1. Preconditioning reagents for accelerated reaction
times in the Paal-Knorr synthesis
a Ratio (batch): 6/7 ) 1:3.6. Ratio (microreactor): 6/7 ) 1:2.8.
An application example for this consideration in this
discussion is given for increasing the second step selectivity
of the nitration of naphthalene (4) by Loebbecke et al.15a
(Scheme 2). The authors mentioned that “the isomeric ratio”
of 1,5-dinitronaphthalene (6) and 1,8-dinitronaphthalene (7)
was found to be almost constant in macroscopic batch
reactions at different process conditions: 1,5-dinitro/1,8-
dinitro ≈ 1:3.6. By applying microreactors the amount of
the unfavored 1,5-dinitro isomer could be significantly
increased resulting in an isomeric ratio of 1,5-dinitro/1,8-
dinitro ≈ 1:2.8.
residence time
total flow rate
reaction temperature
throughput of product
yield [%]
5.2 min
6.1 mL/min
65 °C
260 g/h
91
nents were entering the reaction coordinate and contacting
just at the transition structure energy level.
One example for the preconditioning of reaction inter-
mediates is the Paal-Knorr condensation10 of ethanolamine
(1) and acetonylacetone (2, Scheme 1; Table 1; for experi-
mental details see section 8).
Selectivity gains or potential increases of reaction tem-
peratures are amplified by simultaneous gains in control of
temperature and concentration. High concentration and tem-
perature gradients in microstructure reactors are again drivers
In the exothermic synthesis of pyrrole 3, the decelerated
dosing of the starting material to the reaction mixture adds
significantly to the total process time. The design of the
microreactor enables the synthesis of pyrrole 3 directly from
the neat material in a chemical yield of 91% (process yield11
87%) saving significant amounts of solvent with high
throughput rates of 260 g/h.
In Figure 3, a normal distribution of particles at certain
temperatures is vertically turned and imposed upon the
reaction profile (taken from Figure 2) of competing reactions
hence transition structures. Here it is illustrated that increas-
ing temperature quality can be used to increase selectivity
under the kinetic control of reactions. Alternatively, or in
combination, the increase in temperature control can be used
to increase the technical reaction target temperature level
and hence increase time/space yield.
(12) Fukuyama, T.; Nishitani, S.; Yamaura, R.; Sato, M.; Ryu, I. In Abstracts
of Papers, 7th International Conference on Microreaction Technology
(IMRET 7); Lausanne, Switzerland; DECHEMA-Gesellschaft fu¨r Che-
mische Technik und Biotechnologie; Frankfurt/Main: Germany, 2003; p
236.
(13) Panke, G.; Schwalbe, T.; Stirner, W.; Taghavi-Moghadam, S.; Wille, G.
Synthesis 2003, 2827.
(14) Dale, J. D.; Dunn, P. J.; Golightly, C.; Hughes, M. L.; Levett, P. C.; Pearce,
A. K.; Searle, P. M.; Ward, G.; Wood, A. S. Org. Process Res. DeV. 2000,
4, 17.
(15) For related investigations into nitration reactions, see also: (a) Antes, J.;
Tuerecke, T.; Marioth, E.; Schmid, K.; Krause, H.; Loebbecke, S. In Topical
Conference Proceedings, 4th International Conference on Microreaction
Technology (IMRET 4); Rinard, I., Ed.; AIChE Spring National Meeting,
Atlanta, GA, 2000; p 194. (b) Burns, J. R.; Ramshaw, C. In Topical
Conference Proceedings, 4th International Conference on Microreaction
Technology (IMRET 4); Rinard, I., Ed.; AIChE Spring National Meeting,
Atlanta, GA, 2000; p 133. (c) Burns, J. R.; Ramshaw, C. Trans. Inst. Chem.
Eng. 1999, 77, 206.
(10) (a) Buu-Hoi, Ng. P. J. Org. Chem. 1955, 20, 639. (b) Bishop, W. S. J. Am.
Chem. Soc. 1945, 67, 2261.
(11) The chemical yield refers to the material exclusively isolated from steady-
state conditions. In contrast to this, the overall amount of the consumed
material is taken into account in the process yield which includes also the
prerun. Upon enhanced operating periods of the microreaction system, the
lack of yield in the process yield is increasingly negligible.
(16) Schmalz, H. G.; Schwalbe, Th.; Sakamoto, Y.; Matsumoto, K.; Goto, S.
Germ. Patent DE 103 33 174.3, 2003.
(17) (a) Baldyga, J.; Bourne, J. R. Turbulent Mixing and Chemical Reactions,
1st ed.; John Wiley & Sons: New York, NY, 1999. (b) Borune, J. R. Org.
Process Res. DeV. 2003, 7, 471.
442
•
Vol. 8, No. 3, 2004 / Organic Process Research & Development

Figure 5. Avoiding follow-up reactions.
The selective monoalkylation of a N-acyliminiumion 11
with an electron-rich aryltriether 12 in an academic prototype
mixer as a case of potential subsequent reactions with kinetics
in the range of mixing times was reported by Yoshida and
co-workers (Scheme 4).18 In this case efficient mixing
ensured that each aryl compound instantly interacts with an
electrophilic reaction partner and that each electrophile is
quenched by an unreacted aryl compound. This reaction
conduct resulted in a 96:4 selectivity in favor of the
monoalkylated product 13. It was benchmarked against a
batch as well as a T-piece flow experiment, both of which
led to the formation of approximately similar amounts of
mono- and dialkylated product 14.
Figure 4. Suppression of follow-up reactions by exact adjust-
ment of the inner temperature in a two-stage process.
Scheme 3. Nitration of 8
In another example, the nucleophilic addition of C2F5-
MgCl (16) to benzophenone (17)19 serves as a case of a
discriminated somewhat slower subsequent reaction (Scheme
5; for experimental details see section 8).20
In the conventional reaction sequence, intermediate 16
was formed by a metal-halide exchange reaction at -78
°C. The instability due to the elimination of MgXF or LiF
from such kind of perfluorinated organometal-reagents is
known to occur even at low temperatures.21 As a consequence
of the limited half-life of these intermediates (even at -78
°C in minute range) and extended dosing times, the scale-
up of this reaction is very difficult.
Upon operating the reaction in a two-stage microreaction
system, both the temperature of the formation of 16 and the
temperature of the nucleophilic addition reaction could be
increased by approximately 70 K. Since the sensitive
organometallic intermediate C2F5MgCl was submitted di-
rectly to the second stage, presuming full conversion of 15,
its decomposition was suppressed significantly. In addition
to this, the competing addition reaction of R-M (used in
of tight control. Since tighter concentration and temperature
control allows a closer tune of reactions to conversion or
selectivity requirements, the reaction control works similar
to (and combined synergistic with) catalysis.12
In a recently published example13 for the practical
applicability of this reaction design benefit, the pyrrazole
nitration reaction 8, a precursor in the Sildenafil synthesis
(Scheme 3), is the crucial reaction. This highly exothermal
reaction has proved difficult to scale;14 upon addition of the
nitration mixture to the starting material, it was difficult to
maintain a target reaction temperature of 70 °C, since the
reaction heats the mixture to above 100 °C, a temperature
at which the product 9 decarboxylates under the reaction
conditions forming the undesired follow-up product 10. The
reaction could be converted into a microreactor and safely
run at a higher temperature of 90 °C, thus avoiding
decarboxylation. The effect of the energy distribution inside
the microsystem on the product profile is outlined in Figure
4.15
Shortened mixing times and tightened residence time
distribution, plug flow, beneficially discriminate against
follow-up reactions in certain practical cases. They both work
as if the reaction coordinate of two subsequent reactions
(Figure 5) could be intersected right after the formation of
the desired product.
For fast reaction sequences wherein the reaction rate of
both the first and the second reaction are of the magnitude
of the mixing time of traditional mixers,16 faster mixing in
microreactors diminishes the impact of mass transfer con-
straints.17 For sequences of a fast and a considerably slower
reaction, the sharpened concentration profile of plug flow
helps in avoiding, for example, decomposition or over-
reaction.
(18) (a) Suga, S., Nagaki, A.; Yoshida, J. Chem. Commun. 2003, 354.
(b) Yoshida, J.; Nagaki, A.; Suga, S. In Abstracts of Papers, 7th
International Conference on Microreaction Technology (IMRET 7); Lau-
sanne, Switzerland; DECHEMA-Gesellschaft fu¨r Chemische Technik und
Biotechnologie: Frankfurt/Main, Germany, 2003; p 1.
(19) In previous papers, the target compound Ph2(C2F5)COH 18a was synthesized
from flourinated propionic esters and 2 equiv of Ph-M (M ) Li, MgX).
See: (a) Kaluszyner, A. J. Am. Chem. Soc. 1955, 77, 4164. (b) Chen, L.
S.; Chen, G. J.; Tamborski, C. J. Flourine Chem. 1981, 18, 117.
(20) To the best of our knowledge, the transfer of perfluorinated ethyl groups
using C2F5MgX has not been reported yet. Some applications for C2F5Li
have been published for nucleophilic additions to carbonyl groups: (a)
Kuroboshi, K.; Hiyama, T. Chem. Lett. 1990, 9, 1607. (b) Nelson, D. W.;
O’Reilly, N. J.; Speier, J.; Gassman, P. G. J. Org. Chem. 1994, 59, 8157.
(21) Examples for the decomposition rate of n-C3H7MgBr have been pub-
lished: McBee, E. T. Proc. Indiana Acad. Sci. 1954/55, 64, 108. Addition
of n-C3H7MgBr to a macrocyclic ketone was found to yield 60% at -78
to -20 °C: Bartsch, R. A.; Bitalac, L. P.; Cowey, C. L.; Elshani, S.; Goo,
M.-J.; Huber, V. J.; Ivy, S. N.; Jang, Y.; Johnson, R. J.; Kim, J. S.; Luboch,
E. J. Heterocycl. Chem. 2000, 37, 1337.
Vol. 8, No. 3, 2004 / Organic Process Research & Development
•
443

Scheme 4. Selective electrophilic monosubstitution as reported by Yoshida et al.a
a Reaction conditions: -78 °C, CH2Cl2.
Scheme 5. Formation of a short-living organometallic compound 16 (R-M ) BuLi, MeMgCl) and its addition to
benzphenone (17) in a two-stage microreaction system
Table 2. Reaction conditions for the two-step addition of C2F5-M to 17
t
t
reagent
R-M
T (°C)
stage 1
T (°C)
stage 2
ratio
R-M/C2F5I/17
stage 1
(min)
stage 2
(min)
amount
17 [%]
amount
18a [%]
amount
18b [%]
entry
1
n-BuLi
-78
0
2.8:3.1:1
10
>10
86
14
batch
2
3
n-BuLi
MeMgCl
-61
-30
-10
2.8:3.1:1
2.5:2.8:1
4
10
17
>60
14
84
51
15
35
1
0
batch
4
5
6
7
8
MeMgCl
MeMgCl
MeMgCl
MeMgCl
MeMgCl
2
1
-6
-8
-6
-4
-4
-4
-6
-4
3.9:4.3:1
3.9:2.7:1
3.9:2.7:1
5.4:2.7:1
7.8:7.8:1
0.9
0.9
0.9
0.7
0.8
<10
<10
8
<10
<10
65
13
9
7
28
25
80
86
82
72
10
7
5
11
*Residence time in the batch mode describes the reaction time of the lithium-halide exchange (stage 1, formation of C2F5-M) followed by the nucleophilic
reaction after the addition of 17 (stage 2, formation of 18a).
excess) to the carbonyl (formation of 18b) was reduced to a
minimum (Table 2, entry 6). This demonstrates the advan-
tages of a reaction conducted in a two-stage system that
enables the generation of C2F5MgCl from C2F5I and R-M
in the absence of the competing electrophile 17. The results
from Table 2 will be discussed in more detail.
Treatment of 15 with BuLi for a total reaction time of 20
min at -78 °C (entry 1) in the batch mode yielded a very
low conversion. In the microreactor at -61 °C, within 4 min
of residence time in the first stage the desired product 18a
was formed in 51% yield (entry 2) alongside 18b (RdBu)
that was formed in 35%. In this experiment, the halogen-
lithium exchange did not proceed completely and significant
amounts of BuLi were left unconsumed (side reaction
generating 18b).
The benchmark experiment for MeMgCl (entry 3) at -30
°C (stage 1) in the conventional mode gave almost the same
results that were obtained with n-BuLi at -78 °C (entry 1).
The initial experiment under “typical” reaction conditions
in the microreactor (residence time in stage 1 of less than 1
min in combination with an elevated temperature) gave little
improvement to a 25% yield (entry 4). However, upon
changing the MeMgCl/C2F5I ratio to an excess of MeMgCl,
the yield was improved to 80% (entry 5). Before performing
all these experiments, an organometallic reagent was used
in sub-stoichiometric amounts in order to suppress the
formation of the addition product 18b. After complete
consumption of C2F5I in the first stage of experiment 5,
obviously the addition rate of C2F5MgCl towards 17 is much
higher than that of MeMgCl. The optimum with 86% yield
(entry 6) was found by decreasing the reaction temperature
in the first stage to -6 °C (residence time and molar excess
left unchanged). The methylation of the benzophenone was
minimized to 5%. Further enhancement of the stoichiometric
amounts of MeMgCl and C2F5I even had negative effects
on the yield and the product profile (entries 7 and 8).22
The latter reaction is an example of the capability of
today’s commercial microreactor systems to be operated in
a serial sequence of up to 5 reactor units that are suitable
for multistep synthesis. The lithium-halide exchange reac-
tion of 3-bromo-anisole (19) (Scheme 6; Table 3; for
experimental details see section 8) is another example for
the direct connection of two microreaction systems.
Due to the known instability of 20 at elevated tempera-
tures,23 the microsystem appeared to be an ideal tool for
improving its performance. In a preceding experimental
(22) The optimization of the throughput (around 10 g/h for experiment 6) was
not part of this project. However, since the required residence times have
been realized using low flow rates and small reaction volumes, there is
potential for improvement.
444
•
Vol. 8, No. 3, 2004 / Organic Process Research & Development

Scheme 6. Lithium-halide exchange of 19 and formylation
with DMF in a two-stage microreaction system
Table 3. Dual-step synthesis in microreactor systems:
lithium-halide exchange of 19 followed by DMF quench
residence time, stage 1
total flow rate, stage 1
reaction temperature, stage 1
residence time, stage 2
total flow rate, stage 2
reaction temperature, stage 2
total throughput of product
0.19 min
10.8 mL/min
0 °C
0.15 min
13.2 mL/min
0 °C
Figure 6. General simplified scheme for combinatorial experi-
ments in an MR.
59 g/h
the scope and limitation of the reaction, the flow rate was
increased by a factor of 2 in all three pumps resulting in a
doubled throughput rate and halved residence times. Apart
from a minor drop of the yield (83%), product quality
remained constant and adaptations at the microsystem to the
set parameters were not required. In conclusion, the straight-
forward handling of the lithium-halide exchange reaction
plus an electrophilic addition was demonstrated without any
scale-up risks and at convenient temperatures.
Table 4. Previous investigation for the kilogram synthesis of
3-methoxybenzaldehydea
entry
T [°C]
scale [mol]
yield [%]
1
2
3
4
<-65
<-60
-50
0.04
0.8
4.8
quant
85/76
60
-40
4.8
24
3. Sequential Development of Chemistry Choices in
Continuously Conducted Organic Synthesis
a Yields were determined by GC except isolated material.
The comparative ease of scaling output quantity from
synthesis in continuously operated microreactor systems has
created the opportunity to apply the same principle to the
combinatorial synthesis of discovery compounds as well as
to combinatorial process development.
investigation in the batch mode, we examined the effect of
the temperature on the formation of 3-methoxy-phenyllithium
(20). An additional focus of this series of experiments was
on scaling up. This was expected to cause problems, due to
the temperature sensitivity of 20 along with its exothermic
formation and accordingly extended dosing times.
The initial experiment (Table 4, entry 1) at a 40 mmol
scale (T < -65°C) provided 3-methoxybenzaldehyde 2124
in almost quantitative yield (in process control, IPC, by GC
analysis). Upon scale enhancement to 0.8 mol, the IPC yield
decreased to 85% (76% isolated). Finally in a 4.8 mol scale
(entry 4), the temperature was kept below -40°C but dosing
times of 4 h for the n-BuLi and 2.5 h for the DMF,
respectively, were required. However, in the end of the
conversion, the IPC yield was poor23b (24%) and the product
was accompanied by considerable impurities.
To compliment these findings, both exothermic reactions,
the lithium-halide exchange and the formylation, were
performed in a two-stage microreaction system. A throughput
of 59 g/h was established for approximately 24 h under nearly
isothermal conditions. The product 21 was isolated in 88%
yield accompanied only by small amounts of side products.
The contact time was reduced to a minimum of 11 s (stage
1), which helps to overcome the problem with the intermedi-
ates’ lifetime, followed by the addition to the carbonyl group
within a residence time of 9 s (stage 2). To learn more about
Recently this novel concept25 for combinatorial experi-
mentation has been commercially introduced (Figure 6). It
aims at the synthesis of focused compound libraries and at
selecting the right reagent choices for that synthesis. It is
organized around the same scaleable process used throughout
the chemical synthesis process. Within this concept, varia-
tions on starting materials Ax and Bx are sequentially reacted
in a microreactor process plant to yield a variety of Cx. The
concept enables the unrestricted combination of Ax and Bx.
This new regimen developed by our group lends itself in
particular to temperature sensitive reactions, moisture sensi-
tive reactions, kinetically controlled C-C bond formation
processes, and homogeneous catalysis26 in the field of
synthesis of targeted (i.e., hypothesis) driven libraries.
Further, it is of similar importance to the optimization of
solvent and reagent choices from the chemical diversity
space. Finally, it allows to scale synthesis output within these
experimental phases to any quantity requirement from
functional screening methods and thereby allows compounds
to be manufactured under the same conditions when desired.
From an operative perspective, experiments are entered
into the system in pulse sequence and their elution profile
from the reactor is optically analyzed. Different operating
and analysis modes have been developed to allow for the
(23) (a) Zhang, X.; Stefanick, S.; Villani, F. J. Org. Process Res. DeV. 2004, 8,
455-460. (b) The authors in the latter publication report the conversion at
-10 °C in a gram scale flask experiment to provide only 31% yield.
(24) (a) Saboureau, C.; Troupel, M.; Sibille, S.; d’Incan, E.; Pe´richon, J. J. Chem.
Soc., Chem. Commun. 1989, 895, (b) Hajipour, A. R.; Mallakpour, S. E.;
Samimi, H. A. Synth. Lett. 2001, 11, 1735.
(26) The formation of carbocations in a sequenced manner in microflow systems
has been reported: (a) Suga, S.; Okajima, M.; Fujiwara, K.; Yoshida, J. J.
Am. Chem. Soc. 2001, 123, 7941. (b) Yoshida, J.; Suga, S. Chem.sEur. J.
2002, 8, 2651. Also see: de Bellefon, S.; Tanchoux, N.; Caravieihes, S.;
Grenouillet, P.; Hessel V. Angew. Chem., Int. Ed. 2000, 39, 3342.
(25) Schwalbe, Th.; Oberbeck, A.; Taghavi-Moghadam, S.; Golbig, K.; Hohm-
ann, M. U.S. Patent 9,617,068, 2000 and Eur. Patent EP 1 174 184, 2001.
Vol. 8, No. 3, 2004 / Organic Process Research & Development
•
445

Scheme 7. Heck reaction
Table 5. Heck reactions performed in the sequential
microreaction system
catalyst con-
amount versiona
T
entry reagent product
catalyst
[%]
[%]
[°C]
1
2
3
4
23
23
23
23
26
26
26
26
Pd(OAc)2
Pd(OAc)2 /P(t-Bu)3 2.0/4.0
Pd[(PPh3)]4
Pd[(PPh3)2Cl2]
2.0
32
53
76
76
105
105
105
105
1.2
1.2
fractioned collection of crude products. Individual reaction
plugs are separated by a spacer solvent. To ensure the
reliability of each experiment, a clear separation of each plug
is essential. The overall close to plug flow behaviour is the
result of a careful system design.27
catalyst
amount
[%]
T
entry reagent product
catalyst
yield [%] [°C]
5
23
26
Pd[(PPh3)2Cl2]
0.6
86
125
(E/Z 4.3:1)
57
94
In the chemical development of applications, the ultimate
objective is to capture representative samples for the overall
stable process. Hence, steady state fractions are collected
according to the selected fractioning conditions. Again
careful system design minimizes waste under these condi-
tions. In this field the principle has been demonstrated in an
investigation aimed at systematically exploring alternative
solvents for diverse organolithium reactions as well as in
the continuous process screen of palladium catalyzed reac-
tions.28,29 In a sequential screening, four palladium complexes
were investigated with respect to their conversion rate for
the coupling of three arylhalogens (Scheme 7), amongst them
iodobenzene and three olefins at three different catalyst
concentrations at three different temperatures.30 In the
experiment, olefins and arylhalogens were premixed as one
reactor feed and the catalyst solution at preset concentrations
was fed as the second feed. The different reaction temper-
atures were applied after completing each full chemistry
cycle.
Table 5 demonstrates a selection of results from this
experimental run (for experimental details, see section 8).
In agreement with results from other groups,31 lower
concentrations of palladium catalyst were found to give much
better results in terms of yield and product profile. Since
synthesis in microstructures is bound on particle free
solutions, it will prove beneficial to find the lowest permis-
sible concentration of reactive palladium concentration in
the continuous microsystem. The full investigation covered
a variety of independent experiments (dealing with temper-
ature and several catalyst concentrations which will be
published in a later paper).
6
7
24
25
28a
27
Pd[(PPh3)2Cl2]
Pd[(PPh3)2Cl2]
0.6
0.6
125
125
(E/Z > 98:1)
a The product in entry 6 undergoes an internal rearrangement.
4. Approach to Maximize Throughput in Continuous
Integrated Microreactor Systems
With increasing quantity requirements in the development
process of chemicals, there is a requirement to maximize
time/space yield in a continuously operated microreactor to
access commercial scale manufacture. Recently, an instru-
ment became commercially available that couples a micro-
reactor plant to in-line analytics (Figure 7). Hardware and
experimental procedures have been developed to enable a
rapid kinetic screening of a reaction and to continuously
enhance reagent concentration and temperature. This resulted
in a new high-throughput tool for the optimization of
continuous reactions. The transient nature of experiments in
these setups allows for a rapid adaptation of synthesis
towards its optimal conditions.
As an example for a rapid screening of a reaction kinetic,
the acylation (Boc protection, Scheme 8; for experimental
details see section 8) of benzylamine (29) is illustrated. The
reaction starts from a residence time τ ) 0.5 min and is
conducted at 22 °C in 0.8 M (each reactant) concentration,
respectively. At a combined flow rate of 4 mL/min (2 × 2
mL/min), the starting materials are reacted and passed
through a continuous microreactor system before the flow
reaches a switching valve that directs the material to an in-
line infrared analyzer flow cell. Changes in the product
formation are monitored directly by comparison of the signal
intensity at 1711 cm-1 (product carbonyl absorbance). After
a stable spectrum at a residence time of 0.5 min has been
collected, the valve is switched to allow a flow path through
an additional reaction time volume container of another 15
mL. After this residence time unit (RTU), the flow again
passes through a switching valve to the analyzer flow cell.
The flow through that cell is again monitored until a stable
spectrum in representation of a 4.3 min reaction time is
collected. The cycle of switching additional residence time
is repeated until the full set of incremental residence time
units (RTUs) has been incorporated as reaction chambers or
vice versa either quantitative conversion or maximal product
formation has been observed (Table 6). When either or both
(27) Golbig, K.; Kursawe, A.; Hohmann, M.; Taghavi-Moghadam, S.; Schwalbe,
Th. Chem. Eng. Commun. 2004, in press.
(28) (a) Fukuyama, T.; Shinmen, M.; Nishitani, S.; Sato, M.; Ryu, I. Org. Lett.
2002, 4, 1691. (b) Liu, S.; Fukuyama, T.; Sato, M.; Ryu, I. Org. Process
Res. DeV. Submitted for publication.
(29) Beller, M.; Riermeier, T. H.; Stark, G. In Transition Metals for Organic
Synthesis; Building Blocks and Fine Chemicals; Beller, M., Bohm C., Eds.;
Wiley-VCH: Weinheim, Germany, 1998; Vol 1, p 208.
(30) In this application, different olefins, arylhalogens, catalysts, catalyst
concentrations, and reaction temperatures were screened. A suitable setup
to investigate all chemistry choices in one sequence at a given temperature
prior to changing the temperature was chosen. The primary motive was to
investigate the substrate/catalyst suitability. In the event these chemistry
choices can be made aforehand, continuous parameter optimization
regarding concentration and temperature with in-line analysis is advisable
as opposed to discrete experimentation.
(31) (a) de Vries, A. H. M.; Mulders, J. M. C. A.; Mommers, J. H. M.;
Henderickx, H. J. Org. Lett. 2003, 5, 3285. (b) Also see: Laird, T.;
Hermitage, S.; Tilstam, U. Org. Process Res. DeV. 2004, 8, 2.
446
•
Vol. 8, No. 3, 2004 / Organic Process Research & Development

Figure 7. General system setup for an optimization system with in-line analyzer (left) and switchable residence time units (right).
Scheme 8. Boc protection of benzylamine (29)
Table 6. Conversion rates of the Boc-protection of
benzylamine (29)
residence time τ
conversion
[%]
[min]
0.5
4.3
8.0
70
93
97
98
11.8
Figure 8. Inline IR analysis of the Boc protection (initial
experiment). Carbonyl absorbances: 30, 1711 cm-1; Boc2O,
1750, 1800 cm-1
.
of these criteria are fulfilled, both flow rates are enhanced
by 20% and the eluant is allowed to stabilize as analyzed
by the sequence of spectra observed within approximately
1.4 times the new resulting residence time. With a stable
composition of the eluant from the maximal residence time,
the switching valves are operated in reverse sequence and
product formation is observed after 11.8, 8.0, 4.3, and 0.5
min respectively. Thus, an eight-point isothermal kinetic
curve is experimentally determined. In cases where the initial
flow rate and the total reaction volume were chosen such
that upon switching the maximum reaction volume complete
conversion could not be obtained, the flow rate would be
lowered in 20% steps to a minimal combined flow rate of 1
mL and the analysis completed as described before.
This methodology allows the continuous manipulation of
reagent concentration and reaction temperature in a similar
manner. Further applications using these manipulations will
be reported elsewhere. A screenshot of the in-line IR spectra
series for the acylation of 29 is given in Figure 8.32 The
diagram outlines the first set of residence time screenings
before intensification of the flow rates. The gaps between
the individual measurements originate from solvent leftovers
inside the RTUs, which were removed prior to product
sampling and gave no IR absorbance in the carbonyl area.
Unconsumed amounts of the reagent Boc2O are visible at
1800 and 1750 cm-1. The smooth conversion was verified
by GC. Apart from small impurities, only starting material
and product were found in the individual samples. The yield
of the optimized experiment (residence time 11.8 min) was
determined to be 92% from product crystallization of a
defined sample.33
5. A Case Study on Optimising Processes with an
Emphasis on Minimal Compound Consumption and
Shortened Learning Cycles
A recent case study conducted in collaboration between
GlaxoSmithKline and CPC-Systems aimed at establishing a
continuous multistep process route for the synthesis of an
advanced pharmaceutical intermediate. The objective of the
study was to demonstrate the feasibility of continuous
processing in this context with a particular emphasis on a
(32) Theoretical considerations suggested and benchmark experiments verified
1.4 residence times as an sufficient forerun for synthesis under steady-
state conditions. From this point, switching to the next residence time unit
would be advisable to make the parameter screening after that period as
efficient as possible. However, each measurement in Figure 8 was kept
for an enhanced period in order to demonstrate the constant level of the
product formation clearly.
(33) Yield in the conventional experiment: 96% (dioxane, 90 °C, 4 h,
chromatography). Padwa, A.; Dean, D. C.; Fairfax, D. J.; Simon, L. X. J.
Org. Chem. 1993, 58, 4646.
Vol. 8, No. 3, 2004 / Organic Process Research & Development
•
447

Figure 9. Synthesis scheme of the three-stage microreaction plant.
Figure 10. Process flow diagram. Abbreviations: AUO, additional unit operation; MR, microreactor; P, pump; PS, pressure
sensor; RTU, residence time unit; T, tank; TE, thermo element, TS, thermo sensor; V, valve.
short process optimization time. An intent to minimize
material consumption was pragmatically balanced by a target
process output of a medium-sized double digit gram per hour
production rate. The underlying synthesis was comprised of
three steps and included two purification steps (Figure 9).
The first reaction of starting materials A and B was a
nucleophilic substitution reaction to form crude C. The batch
protocol used a high excess of the nucleophilic reagent which
required a significant dilution and hence a rather long reaction
time. Prior to building the integrated process development
plant, the reaction was transferred to a microreactor under
reduction of the nucleophile’s excess and the dilution. The
temperature of the reaction conduct was enhanced slightly,
and the reaction time was cut to one-third of the batch
reaction time. Eventually, in the pilot plant the crude material
C was continuously purified. This reaction step was operated
in a coupled manner with the preceding reaction as well as
the two subsequent reactions and the second downstream
purification. The next reaction step was an intermolecular
condensation34 reaction to yield crude E. Again it required
long reaction times due to the azeotropic removal of water
in conventional batch processing. In an initial investigative
step, the reaction was converted to a continuous microreactor
operation through a change of solvent, and the reaction time
was reduced to a few minutes. Crude E was again subjected
to a purifying unit operation. Purified E was reacted in a
nucleophilic reaction to yield G, a reaction that imposed
neither any extraordinary inherent conceptual challenge nor
any issues of conversion to a continuous microreactor system.
Nonetheless, the yield was improved. The entire synthetic
work was subordinated to a quantitative benchmark of yield
and throughput as well as a quantified qualification based
(34) We would have generally expected difficulties in running condensation
reactions in continuous microreactor systems. According to several textbook
rules, the product composition is thermodynamically controlled and the
equilibrium favors the starting materials over the desired product. Azeotropic
water removal is, prior to the use of a dehydrating reagent, a paramount
physicochemical means of stirring the equilibrium towards product forma-
tion. To our surprise, we found that many condensation reactions work
well in microreactors. In many instances (Paal-Knorr synthesis, Guaresky-
Thorpe synthesis, generally cyclocondensations yielding conjugated olefins),
the equilibrium may well lie by far on the product side. Additionally, in
many practical cases, running condensation reactions above the boiling point
of water (even when the reaction becomes pressurized to avoid evaporation
of lower boiling solvents) ensures close to complete conversion. In this
case, the interaction of the solvent with water served the desired role of
controlling conversion.
448
•
Vol. 8, No. 3, 2004 / Organic Process Research & Development

Figure 11. Serial construction of the multistep plant.
Figure 12. Photograph of the pilot plant.
Table 7. Hold-up and start-up/response time
on the final product’s conventional processing impurity
profile. The synthetic work in the fully coupled completed
pilot plant passed all of these criteria.
After the initial reaction transfer from batch to continuous
microreactor, a process flow diagram was established (Figure
10).
Hold-up [mL]
stage purification stage buffer purification buffer stage
1
C
2
1
E
2
3
total
180
20
180 540
1000
360
90 1470
(2370)a
Three requirements determined the further steps in setting
up the pilot plant.35 First, it was desirable to evolve with an
overall plant layout that would enable a rapid response to
the parameter changes under investigation throughout the
process. Second, the overall process time/yield was set by
the pilot plant specification. Third, at this stage it was
considered desirable for the general pilot plant start-up to
include operating buffer vessels at key points. For further
optimization work, these buffer vessels would be bypassed.
Thus, the process layout translated into a hold-up scheme
that would contain 2.37 L of process material including the
buffer and 1.47 L of process material excluding the buffer.
The corresponding start-up time of the plant describes both
the time it would take from front to end until a stable final
product solution would be eluded from the final conversion
reaction stage to yield G and the front to end time constant
for a process parameter change. This start-up time amounted
to 13.7 h including the buffer vessels and 7.7 h excluding
the buffer vessels (Table 7, Figure 11).
Start-up Time [min]
35 140 300
84
30
180
10 7.7 h
(13.7 h)a
a Including buffer elements.
reactor in the direction of theoretical plug flow. Whilst a
tube reactor would require 2.5 τ until a satisfactory elution
consistency is achieved, the commercially available micro-
reactor systems used in this case (Figure 12) have been
systematically optimized to accomplish a similar consistency
within 1.4 τ.
In such a system, material not produced under steady-
state conditions does not meet the quality specifications and
does not provide information to the process. Assuming
similar flow propagation velocity in a microreactor and in a
macroscopic tube of 10 mm internal diameter, the compound
consumption until a steady state is reached would be an order
of magnitude higher for the macroscopic tube. At a mini-
mum, ignoring flow velocity considerations, it would be 40%
higher based on the elution profile comparison.
It is noteworthy that of the latter time requirement 5 h
are held within one of the purification steps. The rather low
time constants for the reaction step are driven by the extent
to which hydrodynamic pulse shaping in the laminar flow
regime of microreactors sharpen the elution profile of a tube
Through detective work in the analytical procedures and
protocols that accompanied this work, it was found that the
yield of the final product G could be tracked back through
the ratio of its starting materials E to F after the purification
step to the yield of precursor E. The extent to which this
(35) Hasebe, S. In Proceedings, International Workshop on Micro Chemical
Plants; Kaikan, H., Ed.; Micro Chemical Process Technology Research
Association:Kyoto, Japan, 2003; p 33.
Vol. 8, No. 3, 2004 / Organic Process Research & Development
•
449

Scheme 9. Diazopigment synthesis
Third, a microreactor plant investment is economically
valid. The investment generates value to the investor. This
value can be modelled. Increased sales expectations com-
pared to the baseline assumption for the noninvestment case
and improved cost elements drive such a model. Improved
cost may stem from four areas. Lower direct labor cost is
commonly associated with continuous manufacturing, lower
starting material consumption is expected with enhanced
yields, or reduced application cost is expected from higher
performance components and lower cost of quality assurance.
A few decades ago, a reduction of rejected batches or the
cost of blending operations to accomplish commercial
specifications might have been involved in assessing such
cost reductions.
Figure 13. The yield of E is given along with the purification
ratios E/F and yield G on a normalized scale.
correlation holds is immaterial for the point to be made about
the pilot plant. Rather it is essential that the time constants
of this tracking phenomenon and the sharp responses coincide
with the design (Figure 13).
6. A Case Study on Integrated Microreactor Systems in
Multiple Annual Tonnage Production
A few cases of applications of microreactors in higher
output pilot manufacturing have been referenced,36 yet
detailed reports on their synthesitic objective, quantified
performance have not been disclosed thus far.37 It is therefore
appropriate to discuss considerations associated with the
building of an advanced pilot or production plant using the
example of a three-stage pilot plant built for the synthesis
of diazopigments. The synthesis was adapted to continuous
microreactor conduct by CPC-Systems for the Clariant
company, Switzerland. CPC-Systems built and commissioned
the pilot plant under contract.
The important difference that a third party “turn key
contract” has compared to a research investigation (aimed
at unravelling novel synthesis concepts internally) lies in the
financial criteria important to the decision makers. For a
manufacturing investment into microreactor plants, such
decisions have been based on a priori or smart abductive
judgment that confirms three hypotheses.
First, a microreactor plant has satisfactory operating
performance. It solves a relevant chemical process challenge
reliably. The investigational results are transferable to
production and onto a relevant set of chemical products.
Roadblocks to plant operating performance and technical
availability are removed, or their removal is foreseeable. The
plant can be operated safely, conceivably often with a lesser
risk exposure over traditional plants.
Second, output and throughput targets are matched. Novel
processes can be established with reasonable investigational
effort and seamlessly scaled. Time/space yields can be
attuned and, in conjunction with throughput enhancing
constructive means, brought to manufacturing outputs.
Understandably, many of the facts that affect decisions
in these considerations are considered business sensitive and
are inappropriate to be reported here.
CPC-Systems and Clariant reported the conversion of the
three synthesis steps for diazopigments into a microreactor
(Scheme 9).38 Notably, in the first step where two solutions
were fed into the microreactor, a suspension of the diazonium
salt 32 was formed. This suspension could in turn be fed
into a second microreactor for the diazocoupling reaction39
to again form the deeply coloured crude pigment suspension
containing 33. Finally, this suspension was subjected to a
final pigmenting step again in a microreactor.
The chemical process challenge in these investigations
resided in the aggregate characteristics of the diazopigment
products. When manufactured in batches, application char-
acteristics such as color strength, brightness, and transparency
would depend on a set of more readily quantifiable proper-
ties, such as, for instance, the volume density distribution
of particles. When manufactured in batch processes, this
property would be widely dispersed and vary considerably
from batch to batch (Figure 14).
By contrast pigments made in microreactors contained a
far tighter distribution of crystal sizes centered around much
smaller mean particle sizes. The investigation was extended
to cover another pigment, and again application character-
istics were improved (Table 8).
A laboratory prepilot system was built and has since been
used for further explorative work in the field. Encouraging
results from the laboratory process screen and the economic
evaluation40 of the project led to the decision to build a pilot
(36) (a) Schirrmeister, S.; Markowz, G. In Abstracts of Papers, 7th Inter-
national Conference on Microreaction Technology (IMRET 7); Lausanne,
Switzerland; DECHEMA-Gesellschaft fu¨r Chemische Technik und Bio-
technologie: Frankfurt/Main, Germany, 2003; p 49. (b) Schu¨tte, R. Germ.
Patent DE 100 02 514 A1.
(38) Wille, Ch.; Autze, V.; Kim, H.; Nickel, U.; Oberbeck, S.; Schwalbe, Th.;
Unverdorben L. In Conference Proceedings, 6th International Conference
on Microreaction Technology (IMRET 6); New Orleans, LA, 2002; Baselt,
P., Eul, U., Wegeng, R. S., Eds.; AIChE: New York, NY, 2002; p 7.
(39) Wooton, R. C. R.; Fortt, R.; de Mello, A. J. Lab Chip 2002, 2, 5.
(37) Matlosz, M.; Jenck, J.; Bayer, Th. Integrated multiscale process units with
locally structured elements (Impulse), synopsis, version 3.6, Oct 22, 2002.
450
•
Vol. 8, No. 3, 2004 / Organic Process Research & Development

The authors would like to thank the CPC-Systems’
scientific and engineering team, Klaus Bednarz, Peter Born,
Daniel Kadzimirsz, Ansgar Kursawe, Andreas Oberbeck,
Sebastian Oberbeck, Klaus Po¨lderl, Joachim Sommer, Dunja
Wallner, and Gregor Wille and all those co-workers quoted
in the references.
We would like to acknowledge the GlaxoSmithKline staff
for their contribution and partnership.
We thank Max Linder, John Ryan, and Jeffrey Sherman
of Mettler Toledo/ASI for their continued support in building
the continuous optimization system utilizing in-line analytics
and Uwe Nickel, Frank Schmidt, and Leonhard Unverdorben
of Clariant for their collaboration.
We gratefully acknowledge grant support from the Ger-
man Federal Environmental Foundation (DBU, AZ: 19348),
the Ministry of Economics of the State of Rhineland
Palatinate (AZ: 3.3.04-1140), the Federal Ministry of
Education and Research (BMBF, AZ: 03C0247B), and the
Federal Ministry of Economics and Labour (BMWA, AZ:
KU0487901KAS3).
Figure 14. Comparison of batch and microreactor pigments’
crystal size distribution.
Table 8. Comparative application characteristics of
microreactor manufactured pigments
microreactor
pigment 1
microreactor
pigment 2
color strength [%]
brightness
transparency
119
139
8. Experimental Section
5 steps glossier
5 steps more
transparent
6 steps glossier
6 steps more
transparent
General. The CYTOS Lab System (CLS) was filled with
solvent by activating the two pumps simultaneously while
the temperature was adjusted using an external thermostat
(Huber Tango). Prior to synthesis, the pumps A and B of
the microreaction system (SEQUOS or standard CLS) were
calibrated to the desired flow rates. The residence time τ
was calculated according to the equation: τ [min] ) system
volume [mL]/total flow rate [mL/min]. The temperature
inside the microreactor (volume 2 mL) and the attached
residence time units (volume 15 mL each) were adjusted.
The system was cleaned with the 2-fold installed volume of
solvent. A sample of 200 mL of ethanol was evaporated,
and the amount of residue was determined (<1 mg) and
analyzed by GC. Inorganic impurities from organometallic
reactions were removed by an extra run with water. Gas
chromatography was performed in a Hewlett-Packard 6890
equipped with an HP 1 column (cross-linked methyl siloxane,
30 m × 0.32 mm × 0.25 mm). Operation method 50 °C, 1
min; 20 °C/min f 250 °C; 250 °C, 10 min. Samples were
dissolved in dichloromethane. NMR spectra were recorded
on a Bruker AC 300 spectrometer.
production plant comprised of all three process steps. The
operating safety of this plant was improved by the compara-
tively small internal volume of a few liters for a 10 tpa output
capacity and was shown to significantly enhance the handling
of components.
Adequate output of the plant was enabled through
numbering up the reaction chamber within the pilot reactor
as well as numbering up the reactors by a factor of 3.
7. Summary and Outlook
Recent progress in continuous chemical process and
chemistry research has led to versatile microreactor system
evolution. Continuous microreactor equipment is now com-
mercially available to cover all aspects of research, develop-
ment, and manufacturing. More importantly, it is widely
accomplishing the objectives of scientists and managers
working in these fields.
In the past, many applications for complex organic
synthesis in microreactors have been described for the
laboratory scale. Since our results described here show the
adaptability of microreaction technology in the field of
kilogram synthesis (coming from bench scale), it is now
clearly the time to apply this technology to its benefit (e.g.,
transferring the numbering-up concept into a practical
approach). In a joint project focusing on the continuous
manufacturing of small and large molecules, CPC-Systems
along with partners41 will be operating in the field of small
molecule manufacturing in early 2005.
2-(2,5-Dimethyl-pyrrol-1-yl)ethanol (3).
reaction conditions
flow rate A
neat ethanolamine
flow rate B
2.1 mL/min (34.8 mmol/min)
4.0 mL/min (34.2 mmol/min)
neat acetonylacetone
reaction volume
residence time
reaction temperature
throughput of product
32 mL
5.2 min
65 °C
260 g/h
(40) Wille, Ch.; Haller, Th.; Gabski, H.-P.; Kim, H.; Unverdorben, L.; Winter,
R. In Abstracts of Papers, 7th International Conference on Microreaction
Technology (IMRET 7); Lausanne, Switzerland; DECHEMA-Gesellschaft
fu¨r Chemische Technik und Biotechnologie: Frankfurt/Main, Germany,
2003; p 219.
System Configuration: The synthesis was performed in a
standard CLS equipped with one microreactor6 and two
Vol. 8, No. 3, 2004 / Organic Process Research & Development
•
451

standard capillary residence time modules (volume 15 mL
each). Due to the higher viscosity of the starting materials,
the inlet valves were bypassed and the reactants were directly
pumped via 1/8” tubing.
System Configuration: The reaction was performed in a
two-stage CPC microreaction system consisting of the
CYTOS microreactor in both stages equipped with individu-
ally designed residence time units of 3 mL (stage 1) and 50
mL (stage 2) of volume, respectively. The temperature inside
each stage was controlled independently using two Huber
Tango thermostates. All solutions were pumped through the
system using rotary piston pumps from graduated cylinders,
enabling flow control in regular periods. The storage
container of the Grignard reagent was kept in an argon
atmosphere, whereas the solution of the C2F5I was cooled
to 0 °C in order to prevent spontaneous evaporation.
Reaction Procedure: The inlet tubes of stage 1 were
changed from solvent to reactants, and the reaction was
started by activation of pump A and B (pump C running on
solvent). After approximately 2 τ (2 min) elapsed in stage
1, the inlet tube for stage 2 was switched to reactant. After
additional 2 τ (18 min) in stage 2, sampling of the final
product was initiated. The crude product mixture was
collected in a flask under instant quenching with a stirred
saturated NaCl solution. During the reaction sequence, both
reaction temperature and inside pressure were monitored for
each stage. For yield estimation, a defined sample (5 mL)
was collected in a separate flask. After finishing the reaction,
the entire system was flushed with THF and cleaned
according to the general cleaning procedure.
Workup and Yield Estimation: The organic layer was
separated after adjusting to a pH of 6 (HOAc), and the
aqueous layer was extracted with MTBE (2 × 30 mL). The
combined organic layers were dried (MgSO4), and the solvent
was evaporated to provide 1.63 g of solid raw material. The
yield of 86% for compound 18a was calculated by quantita-
tive GC analysis based on the isolated material from exper-
iment 1 (Table 2). 1H NMR (300 MHz, CDCl3): δ 2.88 (1H,
s, COH), 7.32 (3H, m, J ) 2.1; 7.5 Hz, CHAr), 7.54 (2H, br
d, J ) 2.0; 7.5 Hz, CHAr). 13C NMR (75 MHz, CDCl3): δ
78.84 (COH), 115.12 (CFx), 117.21 (CFx), 127.07 (CAr, CH),
128.24 (CAr, CH), 128.45 (CAr, CH), 139.77 (CAr, quart). CH
assignation in 13C NMR according to DEPT measurements.
3-Methoxybenzaldehyde (21).
Reaction Procedure: After the system was filled with
ethanol and the temperature equilibration, the feed vessels
(0.5 litre graduated cylinders) were changed from solvents
to reactants and the reaction was started. Product sampling
began two residence times (10 min) later, and the previously
produced material was rejected. The consistency of the flow
rate was verified by measuring the consumed volume of
starting material in a certain period. IPC was performed every
20 min by GC to monitor constancy of the reaction.
Temperature and pressure monitoring verified parameter
consistency throughout the entire reaction. After a collection
time of 160 min, the inlets were changed back to ethanol.
The product was collected for approximately one more
residence time (5 min), and then the product valve was
switched to waste. The system was cleaned according to the
general cleaning procedure.
Workup and Purification: The water formed in the
condensation reaction was removed by evaporation in a
vacuum to provide 772.0 g of crude product (content >96%
according to GC analysis). A sample of 150 g was purified
by distillation (bp 74-76 °C, 0.74-0.84 mbar, lit.10 106 °C,
5 mbar) to provide 138.8 g of the pyrrole 3 (yield 92.5%,
GC purity >99%, residue from the distillation 3.9 g). The
process yield was calculated based on extrapolation of the
complete run assuming a total mass of 714.2 g. Yield: 91%
based on 165 min collecting time (chemical yield) and 86%
including 10 min for the prerun (process yield). Total
operation time: 185 min, including 10 min start-up time and
1
10 min for cleaning. H NMR (300 MHz, CDCl3): δ 2.23
(6H, s, CH3), 3.75 (2H, br t, J ) 7.1 Hz, N-CH2), 3.91
(2H, q, J ) 8.3 Hz, O-CH2), 5.77 (2H, s, CHAr).
2,2,3,3,3-Pentafluoro-1,1-diphenylpropan-1-ol (18a).
reaction conditions
Stage 1
flow rate A
2.6 mL/min (2.1 mmol/min)
0.82 M MeMgCl in THF
flow rate B
0.75 M C2F5I in CH2Cl2
reaction volume
residence time
2.0 mL/min (2.3 mmol/min)
reaction conditions
Stage 1
5 mL
0.9 min
-6 °C
flow rate A
6.0 mL/min (9.6 mmol/min)
reaction temperature
1.6 M BuLi in hexane
flow rate B
1.9 M 3-bromoanisol
4.7 mL/min (8.9 mmol/min)
Stage 2
flow rate C
0.62 M benzophenone (17)
1.0 mL/min (0.6 mmol/min)
in THF
reaction volume
residence time
reaction temperature
2 mL
0.19 min
0 °C
in CH2Cl2
reaction volume
residence time
reaction temperature
total throughput of product
52 mL
8.0 min
-4 °C
9.7 g/h
Stage 2
flow rate C
2.5 mL/min (12.5 mmol/min)
5 M DMF in THF
reaction volume
residence time
reaction temperature
total throughput of product
2 mL
0.15 min
0 °C
(41) One of our partners, ProBioGen, is a leader in mammalian cell line design
and a technology promotor for continuous production of monoclonal
antibodies based on novel proprietary technology. The company founded,
Synthacon GmbH, is located in Leuna, Germany.
59 g/h
452
•
Vol. 8, No. 3, 2004 / Organic Process Research & Development

System Configuration: Synthesis was performed in an
assembly of two standard CYTOS Lab Systems connected
using a thermocontrolled metal tube (1/16”). Both stage 1 and
stage 2 were equipped only with the microreactor. The
temperatures of both microreaction systems were controlled
independently with individual Huber Tango thermostates. For
test purposes special newly developed pressure sensors were
used. Due to the higher viscosity of the starting materials,
the inlet valves were bypassed and the reactants were directly
product flow plug has passed the system (indicated by
approximation of the UV absorption to that of the spacer
solvent), the collection is stopped, and the next sequence is
automatically started. The product solutions were directly
quenched in 50 mL septum capped bottles containing 20 mL
of 1 M aq HCl each.
Solution A: Phenyliodide 22 (7,65 g, 37.5 mmol), acrylic
nitrile 23 (2.19 g, 41.3 mmol), tributylamine (10.43 g, 56.3
mmol) were filled up to 100 mL with dry DMF
Solution B: Pd[(PPh3)2Cl2] (320 mg, 0.46 mmol) dis-
solved in dry degassed DMF (100 mL).
1
pumped via /16” tubing.
Reaction Procedure: The inlet tubes of stage 1 were
changed from solvent to reactants, and the reaction was
started by activation of pump A and B (pump C running on
solvent). After 1 min elapsed in stage 1, the inlet for stage
2 was changed to reactant. After an additional minute in stage
2, the crude product mixture was collected in a flask under
instant quenching by stirring with an aqueous solution of 3
M HCl (540 mL of aqueous solution per 90 min of
collection). The flasks were exchanged every 90 min, and
the quenched material was worked up. The consistency of
the flow rate was verified by measuring the consumed
volume of starting material during a certain period. During
the reaction sequence, both reaction temperature and inside
pressure were monitored for each stage. After a collection
time of 24 h, the stage 1 inlets were changed back to THF
and the product tube was changed to waste. After an
additional minute, stage 2 was also changed to THF. The
system was cleaned according to the general cleaning
procedure.
reaction conditions B reaction conditions A
flow rate channel A
flow rate channel B
reaction volume
residence time
1.0 mL/min
1.0 mL/min
47 mL
23.5 min
105 °C
1.25 mL/min
0.75 mL/min
47 mL
23.5 min
125 °C
temperature
Entry 1 (compound 26): Channel 1, solution A. Channel
2, Pd(OAc)2 (168 mg, 0.75 mmol) in dry degassed DMF
(100 mL total vol), reaction conditions A.
Entry 2 (compound 26): Channel 1, solution A. Channel
2, Pd(OAc)2 (168 mg, 0.75 mmol), P(t-Bu)3 (278 mg, 1.5
mmol) in dry degassed DMF (100 mL total vol), reaction
conditions A.
Entry 3 (compound 26): Channel 1, solution A. Channel
2, Pd(PPh3)4 (527 mg, 0.46 mmol) in dry degassed DMF
(100 mL total vol), reaction conditions A.
Workup and Purification: Each fraction was filled directly
after finishing collection into a separation funnel, and the
layers were separated. The aqueous layer (pH ) 0-1) was
discarded, and the organic layer was concentrated (50 mbar,
40 °C). The concentrated organic layers were combined (2.81
kg) and purified by distillation (100-109 °C, 13 mbar; lit.24
101-103 °C, 10 Torr) to obtain 1.4 kg (88%) of 3-methoxy-
benzaldehyde (GC-purity > 96%). Total operation time: 24
h 35 min, including 2 min start-up time and 33 min for
Entry 4 (compound 26): Channel 1, solution A. Channel
2, Pd[(PPh3)2Cl2] (320 mg, 0.46 mmol) in dry degassed DMF
(100 mL total vol), reaction conditions A.
Entry 5 (compound 26): Channel 1, phenyliodide 22 (7,65
g, 37.5 mmol), acrylic nitrile 23 (2.19 g, 41.3 mmol), and
tributylamine (10.43 g, 56.3 mmol) in dry DMF (100 mL
total vol). Channel 2, solution B. Reaction conditions B.
Entry 6 (compound 28): Channel 1, phenyliodide 22 (7,65
g, 37.5 mmol), 3-buten-2-ol 24 (2.97 g, 41.3 mmol), and
tributylamine (10.43 g, 56.3 mmol) in dry DMF (100 mL
total vol). Channel 2, solution B. Reaction conditions B.
Entry 7 (compound 27): Channel 1, phenyliodide 22 (7,65
g, 37.5 mmol), acrylic ester 25 (4.13 g, 41.3 mmol), and
tributylamine (10.43 g, 56.3 mmol) in dry DMF (100 mL
total vol). Channel 2, solution B. Reaction conditions B.
Workup and Yield Estimation: MTBE (50 mL) was added
to the quenched product solution, and the mixture was stirred
for 5 min. The phases were separated, and the aqueous layer
was extracted with MTBE (2 × 50 mL). The combined
organic layers were washed with water and dried (MgSO4),
and the solvent was evaporated in a vacuum. For entries 1-4
(compound 26 each), the conversion rate was estimated by
GC analysis. The product was identified by coinjection using
authentic material from commercial suppliers (cinnamonic
nitrile 26, Fluka catalogue nr. 96415). The yields for entries
5-7 were calculated based on quantitative GC analysis using
commercially available material as external standard.
Cinnamonic nitrile 26 (isolated raw material 496 mg, GC
purity 80%, yield 86%). The E/Z ratio of 4.3:1 was
1
cleaning. H NMR spectroscopy: 1H NMR (300 MHz,
CDCl3) δ 3.87 (3H, s, OCH3), 7.18 (1H, m, CHAr), 7.38 (1H,
br d, J ) 3.3 Hz, CHAr), 7.44 (1H, m, CHAr), 7.46 (1H, m,
CHAr), 10.00 (1H, s, CHO). The isolated material exhibited
an identical GC retention time to that of commercially
available material of 21 (Fluka catalogue nr. 64780).
Catalyst Screening of the Heck Reaction (Compounds
26-28). General description of sequential synthesis in
SEQUOS microreaction system: The racks of the feed
sampler were loaded with 100 mL septum capped bottles
containing the starting material and catalyst solutions.
Feeding was performed automatically by switching the
reactant valves in a predefined time elapsed mode (volume
dependent feeding). After the desired volume of reactant
solution had been fed into the system, the inlet valve
automatically switched to the spacer solvent (DMF). Frac-
tioning was performed automatically by switching the outlet
valve from waste to product after a predefined deviation
between the actual and reference spectrum has been exceeded
(mode: UV dependent spacing and collecting). After the
Vol. 8, No. 3, 2004 / Organic Process Research & Development
•
453

determined by 1H NMR spectroscopy comparing the integrals
of the olefinic signals. Cinnamonic ethyl ester (27, Fluka
catalogue nr. 96 350, isolated raw material 750 mg, GC purity
83%, yield 94%). Only one single isomer was found
according to GC analysis and H NMR. 4-Phenylbutan-2-
on (28, Fluka catalogue nr. 13150, isolated raw material 501
mg, GC purity 63%, yield 57%).
changed from solvents to reactants and the reaction started.
Recording of the IR spectrum was initiated simultaneously.
The solution was passed through the microreactor (experi-
ment 1, τ ) 0.5 min) before stepwise hooking up the other
RTUs (expt 2, τ ) 4.3 min; expt 3, τ ) 8.0 min; expt 4, τ
) 11.8 min). For a reliable signal estimation, the IR spectrum
was recorded for two residence time units plus at least 5
min. The consistency of the flow rate was verified by
measuring the consumed volume of starting material during
a certain period. After finishing the cycle, the material was
removed with THF. The system was cleaned according to
the general cleaning procedure.
Workup and Purification: The reaction mixture of experi-
ment 1-4 was quenched instantly in aq HCl (0.05 M, 20
mL) after the IR detection cell. After decomposition of the
Boc2O (stirring for 30 min), the aqueous layer was adjusted
to a pH of 8-9 (NaOH, 0.1 M) and extracted with ether (3
× 50 mL). The product profile of the reaction was estimated
by GC analysis from the combined organic layers. A sample
of 21.7 mL from experiment 4 was collected for quantitative
analysis. The aqueous layer (without neutralization, pH 1-2)
was extracted with ether (3 × 80 mL), and the combined
organic layers were dried (MgSO4). The solvent was
evaporated, and the product 30 was crystallized from hexane
at -20 °C in colorless prisms (1.74 g plus small amounts in
the mother liquid, yield 92%). Mp 55-57 °C (lit.33 53-55
°C). 1H NMR (300 MHz, CDCl3): δ 1.47 (9H, s, CH3), 4.31
(2H, d, J ) 8.0 Hz, CH2), 4.84 (1H, br s, NH), 7.25-7.35
(5H, m, CHAr).
1
N-Boc-benzylamine (30).
reaction conditions
channel A
benzylamine 29 (21.41 g, 21.80 mL,
200 mmol), triethylamine (2.02 g,
2.88 mL, 20 mmol) in dry THF
(250 mL total vol)
flow rate channel A
channel B
2.0 mL/min
Boc2O (43.64 g, 200 mmol) in dry THF
(250 mL total vol)
flow rate channel B
reaction volume
residence time
temperature
2.0 mL/min
2, 17, 32 and 47 mL
0.5, 4.3, 8.0, and 11.8 min
22 °C
System Configuration: The synthesis was performed in
a standard CLS with one microreactor and three standard
capillary residence time modules (volume 15 mL each). The
microreaction block was equipped with switchable outlets
for compound extraction and direct submission of the
solution to a Mettler Toldeo React IR 4000 (sample rate 32
scans/30 s). The reaction mixture was quenched instantly in
aq HCl (0.05 M, 20 mL).
Received for review January 19, 2004.
OP049970N
Reaction Procedure: After the system was filled with THF
and temperature equilibration, the reactant valves were
454
•
Vol. 8, No. 3, 2004 / Organic Process Research & Development