From discovery to production: Scale-out of continuous flow meso reactors

Article abstract of DOI:10.3762/bjoc.5.29

A continuous flow parallel reactor system has been developed to provide a rapid and seamless transition from the discovery phase and production phase of chemical synthesis, particularly in low volume-high value pharmaceuticals production. Using a single f

Full text of DOI:10.3762/bjoc.5.29

From discovery to production: Scale-out of
continuous flow meso reactors
Peter Styring* and Ana I. R. Parracho
Full Research Paper
Open Access
Address:
Beilstein Journal of Organic Chemistry 2009, 5, No. 29.
Department of Chemical & Process Engineering, The University of
Sheffield, Mappin Street, Sheffield S1 3JD, United Kingdom
doi:10.3762/bjoc.5.29
Received: 20 April 2009
Accepted: 04 June 2009
Published: 09 June 2009
Email:
Peter Styring* - p.styring@sheffield.ac.uk
*
Corresponding author
Guest Editor: A. Kirschning
Keywords:
© 2009 Styring and Parracho; licensee Beilstein-Institut.
License and terms: see end of document.
catalysis; continuous flow; Kumada reaction; parallel; scale-out
Abstract
A continuous flow parallel reactor system has been developed to provide a rapid and seamless transition from the discovery phase
and production phase of chemical synthesis, particularly in low volume-high value pharmaceuticals production. Using a single fixed
bed catalytic meso reactor, reactions can be screened on a small discovery scale over short time scales. The intensified process
produces sufficient material for a full analysis. By replication of the single reactor in parallel, the same chemistry can be achieved
on a larger scale, on a small footprint and without the mass and heat transport limitations of reactor scale-out in batch.
Introduction
Cross-coupling reactions are an essential tool in organic ible to move a reactor to the supply source rather than vice
synthesis, from pharmaceuticals through to functional materials. versa, particularly where hazardous chemical feedstocks are
In the majority of cases on the laboratory scale, coupling reac- involved [5]. Furthermore, small footprint processes would be
tions are performed in stirred batch reactors such as round- advantageous for safety reasons, particularly containment in the
bottom flasks using homogeneous catalysts and often activating case of a failure [2,3]. Scaling up in volume of a reactor has
agents such as substituted phosphines, or else catalysts implications on the mass and heat transfer within the system so
possessing ligands with complex and expensive motifs [1]. the reaction must be re-optimised for the prevailing conditions.
Often, the scale at which such reactions are performed is far This is a time-consuming process and often results in the
greater than that required for initial characterisation and prop- abandoning of the bench scale process in favour of conditions
erty screening [2,3]. Once a compound is identified for further more favourable in bulk. In particular, the use of homogeneous
study and even commercialisation it is then required to be catalysts in scaled-up processes is problematic. Homogeneous
produced on a larger scale. The problem is that to move from catalysts often decompose or dissociate under reaction condi-
small discovery to larger pilot or commercial scale production tions so that the true active catalyst is not in fact the compound
is not simply a case of scaling up the quantities of reagents and added at the onset of the reaction. This makes catalyst recovery
solvents [4]. There are also supply issues when dealing with and re-cycling problematic if not impossible. It also leads to the
large scale syntheses and at times it may be economically sens- possibility of metal contamination at trace levels in the final
Page 1 of 10
(page number not for citation purposes)

Beilstein Journal of Organic Chemistry 2009, 5, No. 29.
Scheme 1: Synthesis of 4-methoxybiphenyl (4-MeOBP) and by-products in the Kumada reaction.
product. Furthermore, added phosphines and other activating 12 tubes (45 ml) which are fitted with screw-on Teflon caps that
agents tend to be expensive and also difficult to recover, adding are equipped with valves for the introduction of inert or reactive
both an environmental and economic burden on the process.
gases and a septa for the introduction of reagents. The 12 reac-
tion tubes sit in two stacked aluminium blocks; the lower fits on
Previously, we have described a series of catalysts based on a hotplate-stirrer and can be maintained at a constant temper-
nickel(II) and palladium(II) that have functionalised salen-acac ature, while the upper block has circulating water which cools
(
salenac) ligands that were functionalised such that they could the top of tubes allowing reactions to be performed under reflux
be covalently bound to organic or inorganic polymer supports conditions if required.
6-10]. Such materials show the beneficial properties of homo-
[
geneous catalysts (activity and selectivity) as well as those of a Single Channel Mini Reactor
heterogeneous catalyst (robust and re-usable). These materials A stainless steel reactor was constructed using Swagelok
have been termed ‘androgynous’ due to these dichotomous components as shown in Figure 1. This has a bed size of 30 mm
properties [8]. The palladium catalysts are active in the Suzuki × 3 mm. The stainless steel column is packed with the catalyst
and Heck coupling reactions at elevated temperatures while the particles; these are held in place by 40 μm woven stainless steel
nickel catalyst is active in the Kumada coupling [11,12] reac- wire mesh with a 25 μm wire integrated in to the reduction
tion at room temperature in batch (Scheme 1) and continuous unions so that it connects the column to the fluidic delivery
flow reactors. In the latter, the catalyst is packed into a 3 mm system at the reactor entrance and exit in order to prevent loss
diameter glass reactor tube of length 25 mm and the reagent of the resin catalyst. The reduction unions allowed one end of
solution flowed using a syringe pump [8,9,13]. This is termed the reactor to be connected to a disposable solvent-resistant
as meso reactor as it shows many of the benefits of a micro syringe, while the other end was attached to a syringe needle
reactor but at a slightly larger volume. Using a mesoporous leading to a vessel containing quenching solution. The design of
version of the catalyst the rate of reaction was enhanced over the reactor system makes catalyst filling and removal an
3
000 times which meant that useful quantities of material could extremely easy and quick process. This channel was packed
be produced within minutes rather than overnight [13].
with nickel immobilised onto Merrifield resin (mean average
particles size of 45 μm when dry and 72 μm when wet in THF),
In this paper, we report on the fabrication and testing of a new with approximately 3% nickel loading. A syringe pump
parallel reactor capable of scaling out chemistries using the (RAZAL A-99) was used to pump a pre-determined volume of
same chemistries developed on the discovery scale, while the solution of known concentrations at different flow rates.
retaining a small area footprint that sits on a standard stirrer hot-
plate and is capable of being employed in a standard laboratory
fume cupboard. Despite a number of publications describing the
theory and practicalities of scaled-out micro and meso reactors
[
14,15], no practical examples of large-scale production have
been described.
Results and Discussion
Batch Reactor
Batch reactions were performed using a Radley’s Carousel
Parallel Synthesiser fitted with a fuzzy logic controller unit
giving temperature control of ±0.1 °C. The systems consists of
Figure 1: Stainless steel single column flow reactor for discovery scale
synthesis.
Page 2 of 10
(page number not for citation purposes)

Beilstein Journal of Organic Chemistry 2009, 5, No. 29.
Figure 2: (a): Schematic diagram of the parallel reactor housing (dimensions in mm); (b) the stainless steel parallel flow reactor with the reactor
housing sat between the flow inlet manifold (bottom) and headspace with gas and liquid outlets (top).
Parallel Capillary Reactor
top of the headspace unit in order to relieve any pressure build
The parallel capillary reactor consisted of a capillary block, up. The capillary block is the body of the reactor, which
head space, bottom space inlet manifold, spacer plate, PTFE contains 120 capillary tubes of 3 mm diameter and 30 mm
seal and a packing spacer. Figure 2a shows the schematic of the length. An access point was also drilled in the middle of the
reactor while Figure 2b is an external photograph of the actual reactor, between the reactor tubes, in order to monitor reactor
reactor giving an indication of scale. Uniform packing was temperature. The inserted temperature probe can be used to
achieved using an array of stainless steel spacer pins that were control the temperature through feedback with a stirrer hot plate
inserted into the base of the reactor body while the catalyst was using an IKA Fuzzy Logic controller. Reagents of known
added from the top. The stainless steel mesh was used to hold concentrations were pumped into the reactor using a pneumatic-
the catalyst in place at top and bottom of the reactor. Reagents ally driven calibrated system with nitrogen as the carrier gas.
flowed against gravity along the capillary tubes to ensure The pre-mixed reagents were stored in a glass bottle (vessel 1)
uniform distribution. The bottom space helps uniform distribu- as shown in Figure 3. In order to maintain a constant flow from
tion of the reactant solution to be achieved and the reaction the vessel it was necessary to maintain a constant pressure and
products are collected from the head space through a side this was achieved using a second vessel filled with water, which
mounted outlet valve. A gas release valve was included at the was filled at the same rate as reagents were exiting vessel 1.
Figure 3: Schematic diagram of the pneumatic pumping system.
Page 3 of 10
(page number not for citation purposes)

Beilstein Journal of Organic Chemistry 2009, 5, No. 29.
The reaction studied was the Kumada reaction [11,12] which product. This was indeed found to be the case. Without any pre-
involves the coupling of an aryl halide and a Grignard reagent, wash or with a 4-bromoanisole pre-wash there was no differ-
in this case 4-bromoanisole and phenylmagnesium chloride, to ence in conversion with a maximum yield of 44% being
produce 4-methoxybiphenyl (4-MeOBP, R-Ar). It was also observed by GC. However, with a pre-wash using phenylmag-
observed that anisole, 4′,4-dimethoxybyphenyl (4,4′-di-MeOBP, nesium chloride, the yield increased to 52%. The revised
Ar-Ar) and biphenyl (BP, R-R) were obtained as reaction scheme is shown (Scheme 2) and includes the original cycle,
by-products. The catalytic cycle that we originally proposed [6] shown as cycles A and B. Two mechanisms were originally
neglects the formation of the other by-products so we investig- thought to be important: (a) the activation of the catalyst
ated the whole scheme in detail. The chloro rather than the through sacrificial loss of the Grignard reagent to form the
bromo Grignard reagent was used, even though it was less homo-coupled product biphenyl then subsequent cross-coup-
reactive, as the waste magnesium chloride or mixed halide was ling to give the desired product (Cycle A); (b) continued
more soluble in THF under the reaction conditions than production of the cross-coupled product without the need to
magnesium bromide and this prevented significant build up of reactivate the catalyst (Cycle B).
the salt in the reactor while in use, which would otherwise lead
to blockage. Batch reactions were performed in order to probe The complex in the free and immobilised form possesses a hard
the mechanism for the reaction further. If the activation of the central Ni2+ ion chelated by both hard (anionic oxygen) and soft
catalyst was important then a pre-wash with the Grignard (neutral nitrogen) donor atoms from the ligand [6,7]. Cycle A
reagent should improve the yield of the desired cross-coupled involves pre-activation of the catalyst by reduction to Ni(0),
Scheme 2: Proposed catalytic cycles for the transformation of 4-haloanisole (Ar-X) and Grignard reagent (RMgX) into 4-methoxybiphenyl (Ar-R) with
by-products anisole (Ar-H), biphenyl (R-R) and 4,4′-dimethoxy-biphenyl (Ar-Ar) using the polymer supported nickel catalyst. Ligands are simplified for
clarity.
Page 4 of 10
(page number not for citation purposes)

Beilstein Journal of Organic Chemistry 2009, 5, No. 29.
chelated by soft nitrogen donors, with the production of the reactor. In further studies [6-8], a similar nickel resin catalyst
homo-coupled organic product through sacrifice of some of the [6] and a silica supported nickel catalyst were used [7,8]. The
Grignard reagent. This is an essential step as Ni2+ is inactive Kumada reaction was carried out in a glass mini flow reactor
towards the oxidative addition of the organobromide. In general (Omnifit) and good yields were achieved in a matter of minutes,
the yield of this homo-coupled product is not proportional to the compared to the conversion obtained in batch over a period of
principal cross-coupled product so the possibility arises that the 24 h, albeit on a much smaller scale. The optimum conversion
catalyst takes an alternative pathway (route B) once the catalyst and selectivity was achieved using a flow rate of 13 μl min−1,
initiation has occurred. As the cross-coupled product is with a 67% yield of 4-methoxybiphenyl being achieved using
observed as the highest yielding product, it was proposed that phenylmagnesium bromide.
the rate constant for route A is smaller than for route B.
As the residence time has an important effect on the product
The proposed catalytic cycle therefore serves as a working yield, and with the aim of scaling out the same reaction in the
hypothesis for the formation of the principal cross-coupling parallel capillary reactor, the reaction was carried out in the
product, 4-methoxybiphenyl. A study of bromo-Grignard continuous flow reactor at different flow rates. The motivation
reagent to organo bromide ratio necessary to optimise the yield was to investigate the effect of the residence time of the
of cross-coupling product by Phan et al. [7,8] showed that the reagents within the catalyst bed on the coupling process. The
reaction was best performed using a 1:1 substrate ratio with mean residence time in a continuous flow reactor is also
6
7% conversion to the desired 4-methoxybiphenyl. The produc- assumed to be the reaction time. Therefore the reaction was
tion of biphenyl as a homo-coupling product was also observed performed at different flow rates in a Swagelok stainless steel
in a significant amount of approximately 20% of the final column, of the same size of a single channel in the capillary
mixture. Indeed, the problem of homo-coupling was reactor. The reaction was carried out using an equimolar solu-
encountered in earlier publications involving the coupling reac- tion of 0.5 M of 4-bromoanisole and phenylmagnesium
tion under discussion. This by-product can be formed by sacrifi- chloride. The results obtained are shown in Figure 4.
cial loss of the Grignard reagent in the reduction of Ni(II) to the
real catalytically active Ni(0) form as the first step in the
proposed catalytic cycle mentioned above. However, the
composition of the final reaction mixture also included anisole
and 4,4′-dimethoxybiphenyl and the level of biphenyl itself
indicates that it is also forming by an alternative pathway.
Therefore, the catalytic cycle is more complex than originally
proposed. However, using the results from this study together
with previously obtained data we have been able to construct a
viable alternative route that explains the observed by-products.
Product scrambling occurs through trans-metallation between
the intermediate nickel complex 3 formed by oxidative addition
of the aryl bromide (Ar-Br), and the Grignard reagent. This
results in the formation of the homo-coupled 4,4′-dimethoxybi-
phenyl by-product. Anisole (Ar-H) is formed by hydrodebrom-
ination of the aryl bromide via the same oxidative addition
intermediate complex 3 due to the presence of traces of water in
Figure 4: Comparative study of flow rates in a single channel meso
flow reactor. The output was sampled hourly and analysed by GC
during the period of extended flow.
the system. This occurs both in the batch system, where there is It was observed that as the flow rate increases the yield of
only residual water in the solvent, and in the flow system. Even 4-methoxybiphenyl decreases, as expected. At higher flow rates
though water is used to equalise the pressure in the flow system, of 1.2 and 2.5 ml h−1, the yields of 4-methoxybiphenyl were
the reactor is protected from the pumping system using a self- between 0 and 2%. Therefore, it was decided to carry out the
indicating silica gel drying tube that is regularly recharged.
Kumada reaction in the capillary parallel reactor at a flow rate
of 0.79 ml h−1 (13 μl min−1) as in previous studies. All condi-
tions were kept the same as in the study of Phan et al. [8],
Mini-Continuous Flow Reactor
Previous work carried out by Styring et al. [13] has described however, it was observed that a lower product yield was
the use of a salenac nickel complex immobilised on a polymer obtained. This is due to the lower reactivity of the chloro
support for the Kumada reaction in pressure driven mini flow Grignard reagent compared to the bromo derivative used in
Page 5 of 10
(page number not for citation purposes)

Beilstein Journal of Organic Chemistry 2009, 5, No. 29.
previous studies. No deterioration of the catalyst was detected dimethoxybiphenyl was monitored. The nickel complex cata-
over the course of the cycles and ICP-ES studies showed the lyst instigates the homo-coupling of the Grignard reagent, in
nickel content in the resin was the same as when the catalyst this case to produce biphenyl. The yield of 4-methoxybiphenyl
was initially prepared.
was increased by over 20% as seen in Table 1. However, while
the yield of 4,4′-dimethoxybiphenyl almost doubled, the yield
An important point of concern when using heterogeneous cata- of biphenyl increased three-fold. If there was simple trans-
lyst is its lifetime, particularly for industrial and pharmaceutical metallation occurring then there should be equity in the yields.
applications. In an ideal case [16] the catalyst can be recovered However, the elevated value for biphenyl production together
and reused several times before it eventually deactivates. At the with the effectiveness of the Grignard pre-wash supports our
same time, catalyst recovery can also reduce the environmental theory that the activation of the catalyst through Cycle A is
pollution caused by heavy metals used in the catalyst systems. essential for the reaction to proceed effectively. While the
When using a supported metal catalyst, another critical issue is comparisons indicates that a stoichiometry of 1:2 yielded better
the possibility that some active metal migrates from the solid to results than a 1:1 stoichiometry, it should be noted that this is
the liquid phase and this leached metal would become respons- both an economic and environmental burden as the excess
ible for a significant part of the catalytic activity which would Grignard reagent is quenched at the end of the process. This
then occur homogeneously. When the solid catalyst was could be overcome by building a recycle into the process. In
removed from a batch reactor no reaction occurred until the fact, when a mass balance was carried out over the reaction it
catalyst was replaced, showing leaching is not a problem. The was found that less Grignard reagent was actually consumed
polymer supported catalyst was then tested for its reusability. when the higher stoichiometry was used. It should also be noted
The catalyst was packed in the Swagelok column and the that the increased stoichiometric ratio has little effect on the
Kumada reaction was carried out three times without unpacking turnover frequency (TOF). Although a slight increase was
and without any catalyst regeneration. The conditions were observed, this is within 5% experimental error.
assumed to be the same as in the parallel capillary reactor. In
between the reactions the column was flushed with dried THF
Table 1: Comparative study of aryl bromide to Grignard reagent ratio
to remove any excess reagents and salts deposited on the
surface of the catalyst. Results showed that the nickel resin
catalyst can indeed be re-used in further reactions without a
significant degradation in activity. Although there was a slight
decrease in activity over subsequent runs, the average yield of
on product and by-product yields.
Ratio of 4-BA to PhMgCl
1:1
1:2
4-Methoxybiphenyl (mol%)
57
23
7
80
1
4
-methoxybiphenyl still remained greater than 40%.
4-Bromoanisole (mol%)
Anisole (mol%)
9
4
,4′-Dimethoxybiphenyl (mol%) 13
-Methoxybiphenyl (mmol) 0.57
23
0.8
0.09
719
One of the by products of the Kumada reaction is biphenyl,
which is the result of the self-coupling of the phenylmagnesium
chloride, as explained in Scheme 2. As the yield of product
obtained in the Kumada reaction in this work was comparat-
ively low compared to that of Phan et al. [8], the stoichiometry
of the reagents was varied accordingly. Two different studies
4
Biphenyl (mmol)
TOF (h−1)
0.03
683
were carried out, in a continuous reaction and the other in a Parallel Flow Reactor
batch process. In the continuous reaction studies, the column Having established the reaction protocol from the mini-
was packed with catalyst and pre-washed with a solution of continuous flow reactor runs, and having checked for the
phenylmagnesium chloride in THF (0.5 M) and then rinsed with reusability of the catalyst, the Kumada reaction was performed
THF and the experiment was carried out as described previ- in the scaled-out system using the parallel capillary reactor. The
ously. Two different stoichiometric ratios (1:1 and 1:1.5 of stoichiometry used remained constant in all studies, with 1 equi-
4
-bromoanisole and phenylmagnesium chloride respectively) valent of 4-bromoanisole to 1.5 equivalent of phenylmag-
were used. There an increase in the 4-methoxybiphenyl yield nesium chloride. This would enhance the yield of product while
from 28 to 45% at the higher ratio when a pre-wash was carried minimising the Grignard reagent waste in the absence of a
out. At the same time the amount of biphenyl produced almost recycle stream. Equimolar (0.5 M) solutions were used in all
doubled compared to the smaller increase in 4-methoxybi- cases.
phenyl yield. A comparative study of 1:1 and 1:2 of 4-bromoan-
isole to phenylmagnesium chloride was then undertaken in a In the first study, the reaction was carried out for 6 h at a flow
batch process and the production of biphenyl as well as 4,4′- rate of 95 ml h−1, which equates to 0.79 ml h−1 per channel, as
Page 6 of 10
(page number not for citation purposes)

Beilstein Journal of Organic Chemistry 2009, 5, No. 29.
used in the single channel studies. The graph of conversion consequence of the salts being washed off in the continuous
against time is shown in Figure 5. Samples were taken at hourly flow and adsorption of the organobromide on to the newly freed
intervals by diverting the flow away from the quenching vessel reactive sites.
to a separate small volume of quenching solution. The ether
layer was decanted off, dried and then analysed by GC. It was
observed that the yield increased steadily for the first 3 h and
reached a peak of 74% where it remained constant within exper-
imental error. The yield was consistent for a 1:1.5 stoi-
chiometry and higher than the yields originally reported for
single channel meso flow reactors. It is proposed that during the
first hours of the reaction there is period of induction that condi-
tions the reactor and activates the catalyst. Steady state is then
achieved after this induction period, which occurs irrespective
of whether or not a pre-wash is used. The induction is therefore
most likely to be limited by adsorption of the aryl bromide onto
the catalyst surface once catalyst activation is achieved. This is
consistent with the Langmuir–Hinshelwood mechanism [17] for
surface kinetics which depends on adsorption of both species on
to the catalyst surface before reaction can proceed.
Figure 6: Kumada reaction carried out in a parallel channel meso
reactor over a 31-hour period at a flow rate of 95 ml h−1 to show cata-
lyst stability and lifetime.
It is seen from Figure 6 hat the production of the by-product
anisole remains constant throughout the reaction while produc-
tion of 4-4′-dimethoxybiphenyl increased in a similar way to the
main product. The production of 4-methoxybiphenyl during this
3
1-hour period was found to give an average yield of 58%. This
gives a production rate using the parallel capillary reactor of
.07 g h−1 which gives 122 g of 4-methoxybiphenyl per day. If
5
the average at steady state (65%) is taken however, this equates
to 5.70 g h−1 or 137 g day−1. In order to increase production
there are two options. One is to replicate further channels in
parallel, relying on the fact that the chemistry in each reactor
will be identical. The dimensions of the parallel reactor used
were such that the design would fit easily onto a standard
hotplate-stirrer. However, by making the reactor blocks rectan-
gular they could be readily stacked in three dimensions. The
Figure 5: Kumada reaction carried out in a parallel channel meso
reactor at a flow rate of 95 ml h−1.
The reaction was then carried out for a period of 31 h using the second approach is to increase the flow rate into the reactor,
same flow rate as above, giving a residence time on each of the however as this will reduce the residence time it will therefore
capillary reactors of 11 min. A linear relationship in the 4-meth- reduce the effective reaction time with the consequence that the
oxybiphenyl yield with respect to time in the initial hours of the conversion may fall. Therefore, a study in which the flow rate
reaction is seen from Figure 6. This induction period is longer was doubled was undertaken.
than in the six-hour study, however the average yield of 4-meth-
oxybiphenyl after this period was found to be 66% which is A flow rate of 190 ml h−1 (1.6 ml h−1 per single channel) was
consistent with previous studies. As the catalyst used was that used, giving a residence time of 5 min 30 s. The reactor was run
used in the six-hour study the possibility of catalyst deactiva- for a period of 9.5 h. A decrease in the production yield (%) of
tion was considered. However, the fact that steady state produc- 4-methoxybiphenyl with time was observed (Figure 7). The
tion of around 66% over the study up to 31 h would dispel this maximum yield was observed after 4 h, however this was only
theory. Indeed ICP-AES studies showed no significant loss of 21%. Yield gradually decreased after this, averaging around
nickel from the catalyst. The likely scenario is that magnesium 12% although production did not cease. Production of anisole
salts had become deposited on the catalyst beads on standing and 4,4′-dimethoxybiphenyl was unaffected however, aver-
between studies and that the initial linear increase in yield was a aging 6 and 1% respectively, so selectivity decreases with
Page 7 of 10
(page number not for citation purposes)

Beilstein Journal of Organic Chemistry 2009, 5, No. 29.
increased flow. Likewise, there was very little change in the 30 mm for the parallel reactor. We have not looked at longer
turnover frequency (0.35 h−1) based on 4-bromoanisole. discovery reactors because for the Kumada reaction the
However, the TOF was significantly reduced from the reaction selectivity is decreased. At equivalent flow rates on longer tubes
carried out at half the flow rate (1.5 h−1). It should be noted that there is a large increase in the production of biphenyl with
the TOFs in flow reactors are lower than in batch as we are increased residence time on the reactor.
dealing with small flows through a packed bed reactor and
hence the catalyst concentration is necessarily high.
Conclusions
The Kumada reaction was carried out at room temperature,
using a nickel catalyst supported on Merrifield resin beads in a
Swagelok meso reactor column, which is of the same dimen-
sions as a single channel in a scaled-out parallel capillary
reactor. The objective of this research was to demonstrate that it
is possible practically to scale-out a reaction carried out on a
single channel reactor on a pilot scale/small production reactor
by simple replication of the original channel geometry in
parallel. This was clearly demonstrated. Reactions optimised in
a single channel are simply replicated in a 120-channel reactor.
In the case of the Kumada reaction, the yield is up to 137 g over
a 24-hour period on a reactor that sits on a standard hot-plate
stirrer. Increased yields can be achieved simply by replicating
further in two- or even three-dimensions through more sophist-
icated manifolding.
Figure 7: Kumada reaction carried out in a parallel channel meso
reactor at a flow rate of 190 ml h−1.
The catalytic mechanism was studied using batch reactions, and
different pre-washes with the starting reagents were carried out.
From the results obtained it is proposed that for the nickel
If the rate limiting step in the reaction is the adsorption of the salenac complex, the reaction followed a Langmuir–Hinshel-
organobromide onto the surface of the catalyst, as we have wood mechanism, and that a pre-wash with Grignard reagent
proposed according to the Langmuir–Hinchelwood mechanism activates the many reactive sites present on the catalyst.
[
17], then increasing the flow rate does not give sufficient time However there can be a probability of other mechanisms occur-
for adsorption of the aryl bromide and reaction with the ring and hence further studies are being undertaken.
Grignard reagent to occur and hence the conversion will
decrease, as was observed. It appears from the data that while In the continuous meso flow column, recycling studies of the
adsorption of the aryl bromide can occur there is low availab- catalyst were carried out in order to simulate the re-use of the
ility of the Grignard reagent at high flow rates and hence the catalyst bed within the capillary parallel reactor. Due to the
homo-coupled product and anisole have the opportunity to volume of catalyst needed to pack the reactor and the down-
form. The product formation decreases as at the start of the time requirements for repacking it is essential that the catalyst is
reaction there is an induced high concentration of the Grignard stable over prolonged periods of operation. This has been
reagent on the beads as a result of the pre-wash to activate the shown to be the case and the reactor ran continuously over a
catalyst. As the flow proceeds, the concentration is reduced and 31-hour period with steady state product synthesis without loss
if it is difficult for more reagent to access the surface then the of performance.
production slows. This is an interesting observation and we are
Experimental
undertaking detailed mechanistic studies in an attempt to clarify
the situation.
Reagents
Chemicals were obtained from Sigma–Aldrich and Fisher.
Another approach to increasing yield would be to increase the Diethyl ether (99.8%, Fisher), anisole (99%, Aldrich), biphenyl
length of the reactor tubes. We standardised on 30 mm channel (97%, Aldrich), 4-bromoanisole (99%, Aldrich), phenylmag-
lengths as we were targeting a generic reactor system that could nesium chloride (2M in THF, Aldrich), mesitylene (98%,
be used for many different reactions without the need to change Aldrich), 4-methoxybiphenyl (97%, Aldrich), 4,4′-dimethoxybi-
components. All our discovery studies were performed in 25 phenyl (99%, Aldrich), were used as received. THF (99%,
mm glass and 30 mm stainless steel tubes, hence the choice of Aldrich) was dried over 4 Å molecular sieves.
Page 8 of 10
(page number not for citation purposes)

Beilstein Journal of Organic Chemistry 2009, 5, No. 29.
Batch Reaction Procedure
rate of 95 ml h−1 and the reactor top removed. The reactor was
Unless otherwise stated, a solution of 4-bromoanisole (0.187 g, inspected visually for the emergence of the blue solution. This
1
mmol) in dry THF (1 ml) was added to a Radley’s Carousel was found to occur simultaneously over all 120 channel within
reaction tube containing the required amount of the nickel cata- a 1 s time range, confirming equal packing over all channels.
lyst. The Grignard reagent, phenylmagnesium chloride (1 M, 1 When all the beads were removed from the reactor, homogen-
ml, 1 mmol) in THF was transferred via syringe under a eous staining was observed.
nitrogen atmosphere and added directly into the solution of the
organobromide. The mixture was stirred at room temperature Parallel Capillary Reactor Procedure
for 24 h under a nitrogen atmosphere. To work-up the reaction, The Kumada reaction between 4-bromoanisole and phenylmag-
saturated aqueous NaCl solution (2 ml) was added to quench nesium chloride in THF was scaled out in the parallel capillary
excess Grignard reagent. The organic components were reactor. The reactor was packed with approximately 8.7 g of
extracted into diethyl ether (2 × 3 ml) which was then dried nickel complex immobilised on Merrifield resin. This gives an
over anhydrous MgSO4 and the resulting solution analysed by average of 75.3 mg of resin per channel. A mixture of
GC and GC-MS with reference to standard solutions of 4-meth- 4-bromoanisole and PhMgCl was placed in a sealed glass bottle
oxybiphenyl, anisole and 4,4′-dimethoxybiphenyl.
in THF under an atmosphere of nitrogen. In each case
-bromoanisole was the limiting reagent. The mixture was
pumped continuously through the nickel resin bed packed in the
4
Single Channel Mini Reactor Procedure
The reaction of 4-bromoanisole with phenylmagnesium chloride reactor, using the pressure driven pumping system described
in THF was carried out in the pressure driven mini flow stain- previously, at room temperature. The flow rates used were of 95
less steel reactor. The reactor was filled with the resin beads ml h−1 and 190 ml h−1, with samples collected at 1-hour inter-
containing the immobilised nickel catalyst (72 mg). A syringe vals during a period of 7 h or more. The organic components
pump (RAZAL A-99) was used to drive a pre-determined were extracted into diethyl ether and analysed by GC with refer-
volume of a solution containing equimolar amounts (0.5 M) of ence to standard solutions of 4-bromoanisole and mesitylene as
the reagents in THF through the reactor at the different flow an internal standard.
rates for at least 5 h. No reaction was observed in the mixed
solution in the absence of the catalyst. The organic components Analysis Methods
were extracted into diethyl ether and analysed by GC with refer- A Varian-GC 3900 gas chromatograph fitted with a 15 m
ence to standard solutions of 4-bromoanisole.
column of 0.25 millimetre diameter and 0.25 μm thickness
CP-SIL5CB coating. The temperature program was an initial
Parallel Capillary Reactor Packing Procedure hold at 50 °C for 30 seconds followed by a ramp from 40–110
Each channel of the parallel reactor was filled simultaneously °C at 40 °C min−1 followed by a ramp of 110–250 °C at 20 °C
with resin to a preset volume. Uniformity was achieved using a min−1. GC-MS analyses were performed using a Perkin Elmer
jig constructed of a series of pillars with spacings identical to GC-MS with a 30 m × 0.25 mm × 0.25 μm Phenomenex-2B5
the layout of the channel positions, each pillar having the same column. The temperature program was 60–260 °C at 10 °C
diameter as the internal diameter of the channel. A series of jigs min−1 with a final temperature isothermal hold for 10 min. The
were constructed with different pillar heights to allow different MS limit was set between 50 and 450 Da.
packing volumes to be achieved. The reactor body was placed
over the jig and resin beads added to the top of the reactor and Acknowledgments
spread so that they evenly occupied the channels. The stainless We thank GlaxoSmithKline (GSK) for financial support in the
steel mesh was then located on top of the channel assembly and form of a studentship to AIRP and EPSRC for support of PS.
the reactor lid bolted in place. The reactor was then inverted and We also thank the Analytical Services department in the
the jig carefully removed. The second stainless steel mesh was Department of Chemistry at the University of Sheffield for
placed over the reactor assembly and the reactor base bolted in access to analytical procedures.
place. The reactor was then ready for connection to the
pumping and collection pipe work. An average packing of 75.3 References
g per channel was calculated. It is not possible to isolate indi-
vidual channels to inspect their contents due to the large number
of capillaries present. However, in order to ensure that the
packing procedure was consistent we ran tests on a reactor
packed with Merrifield resin beads without catalyst attached by
pumping a solution of guaiazulene dye in THF (0.5 M) at a flow
1. Diedrich, F.; Strang, P. J. Metal-catalysed Cross-coupling Reactions;
Wiley-VCH: Weinheim, New York, 1998.
2
.
.
Haswell, S. J.; Watts, P. Green Chem. 2003, 5, 240–249.
doi:10.1039/b210539j
3
Watts, P.; Haswell, S. J. Chem. Soc. Rev. 2005, 34, 235–246.
doi:10.1039/b313866f
Page 9 of 10
(page number not for citation purposes)

Beilstein Journal of Organic Chemistry 2009, 5, No. 29.
4.
Lee, S.; Robinson, G. Process Development: Fine Chemicals from
Grams to Kilograms; Oxford Chemistry Primer No. 30; Oxford
University Press: Oxford, 1995.
5
6
7
8
9
1
1
1
1
1
1
1
1
.
.
.
.
.
Garcia-Egido, E.; Spikmans, V.; Wong, S. Y. F.; Warrington, B. H.
Lab Chip 2003, 3, 73–76. doi:10.1039/b302381h
Styring, P.; Grindon, C.; Fisher, C. M. Catal. Lett. 2001, 77, 219–225.
doi:10.1023/A:1013209202418
Phan, N. T. S.; Brown, D. H.; Adams, D. H.; Spey, S. E.; Styring, P.
Dalton Trans. 2004, 1348–1357. doi:10.1039/b316553c
Phan, N. T. S.; Brown, D. H.; Styring, P.
Green Chem. 2004, 6, 526–532. doi:10.1039/b405203j
Phan, N. T. S.; Khan, J.; Styring, P.
Tetrahedron 2005, 61, 12065–12072. doi:10.1016/j.tet.2005.07.109
0.Styring, P.; Phan, N. T. S. Green Chem. 2008, 10, 1055–1060.
doi:10.1039/b805290e
1.Tamao, K.; Sumitani, K.; Kumada, M.
J. Am. Chem. Soc. 1972, 94, 4374–4376. doi:10.1021/ja00767a075
2.Kumada, M. Pure Appl. Chem. 1980, 52, 669–679.
doi:10.1351/pac198052030669
3.Haswell, S. J.; O’Sullivan, B.; Styring, P. Lab Chip 2001, 1, 164–166.
doi:10.1039/b104035a
4.Amador, C.; Gavriilidis, A.; Angeli, P.
Chem. Eng. J. 2004, 101, 379–390. doi:10.1016/j.cej.2003.11.031
5.Löb, P.; Löwe, H.; Hessel, V. J. Fluorine Chem. 2004, 125, 1677–1694.
doi:10.1016/j.jfluchem.2004.09.006
6.Djakovitch, L.; Koehler, K. J. Mol. Catal. A: Chem. 1999, 142, 275–284.
doi:10.1016/S1381-1169(98)00292-1
7.Thomas, J. M.; Thomas, W. J. Principles and Practice of
Heterogeneous Catalysis; VCH: New York, 1997.
License and Terms
This is an Open Access article under the terms of the
Creative Commons Attribution License
(http://creativecommons.org/licenses/by/2.0), which
permits unrestricted use, distribution, and reproduction in
any medium, provided the original work is properly cited.
The license is subject to the Beilstein Journal of Organic
Chemistry terms and conditions:
(http://www.beilstein-journals.org/bjoc)
The definitive version of this article is the electronic one
which can be found at:
doi:10.3762/bjoc.5.29
Page 10 of 10
(page number not for citation purposes)