A versatile and cost-effective approach to automated laboratory organic synthesis

Article abstract of DOI:10.1021/op990005k

The application of a commercial laboratory automated synthesis system, the Anachem SK233 Workstation, is described for use in organic synthesis. A novel reactor design feature has been developed to enable the sampling of reactions under ambient to reflux temperatures while maintaining an effective inert atmosphere. Up to 10 reactions can be run simultaneously on the reactor block supplied. Examples are reported of multiple parallel reactions covering a range of chemistries encountered in synthesis encompassing heterogeneous and homogeneous reactions, air-sensitive and aggressive reagents, ambient to full reflux temperature, and full inert atmospheres with concomitant automated HPLC product analysis. The equipment and modifications described are of moderate cost, are robust in use, are of acceptable size for modern chemistry laboratories, and are readily acceptable to practising chemists.

Full text of DOI:10.1021/op990005k

Organic Process Research & Development 1999, 3, 189−195
A Versatile and Cost-Effective Approach to Automated Laboratory Organic
Synthesis
Mark A. Armitage,† Gillian E. Smith,‡ and Kenneth T. Veal*,‡
Synthetic Chemistry Department, SmithKline Beecham Pharmaceuticals, Old Powder Mills, Nr. Leigh,
Tonbridge, Kent, TN11 9AN, UK, and Synthetic Chemistry Department, SmithKline Beecham Pharmaceuticals,
New Frontiers Science Park, Third AVenue, Harlow, Essex, CM19 5AW, UK
Abstract:
Traditional combinatorial chemistry equipment was pre-
dominantly aimed at solid-phase synthesis under a single set
of reaction conditions, with no monitoring, and so was unable
to meet our needs.
The application of a commercial laboratory automated synthesis
system, the Anachem SK233 Workstation, is described for use
in organic synthesis. A novel reactor design feature has been
developed to enable the sampling of reactions under ambient
to reflux temperatures while maintaining an effective inert
atmosphere. Up to 10 reactions can be run simultaneously on
the reactor block supplied. Examples are reported of multiple
parallel reactions covering a range of chemistries encountered
in synthesis encompassing heterogeneous and homogeneous
reactions, air-sensitive and aggressive reagents, ambient to full
reflux temperature, and full inert atmospheres with concomitant
automated HPLC product analysis. The equipment and modi-
fications described are of moderate cost, are robust in use, are
of acceptable size for modern chemistry laboratories, and are
readily acceptable to practising chemists.
The literature reveals a number of attempts to automate
synthetic organic solution-phase chemistry. Despite these
ingenious5-8 and, in many cases, technically inspiring
efforts,9-13 they have not been widely adopted, possibly due
to their complexity and limited versatility. Furthermore, by
late 1996, very little commercial equipment14-16 was avail-
able that could be used or adapted for our purposes.
While looking for equipment, however, we did come
across a possible candidate for evaluation in the Anachem
SK233 Workstation.17 This system consisted of a 10-position
reactor block that could be sampled automatically via an XYZ
robotic arm with subsequent dilution and HPLC analysis.
Each reactor tube (ca. 5-25 mL reaction volume) could be
stirred magnetically and heated (ambient to 150 °C). A
separate reactor block could be used to cover the temperature
range from -30 to +70 °C. Control of solvent reflux was
accomplished by natural cooling of the exposed half of the
reactor tube. An optional water-fed or forced air-fed “reflux”
housing over the exposed section of the reactor tubes was
also available. It was suggested that an inert atmosphere
could be achieved by either capping purged tubes or
continuously purging the additional reflux housing. The
concept sounded fine, but could we do real chemistry in such
equipment? A purchase of one of these units for trial
purposes answered many of these questions very quickly.
Introduction
In recent years, emerging technologies have led to a
reappraisal of working practices within the chemical industry.
The pressures of these changes have been felt acutely,
particularly within the pharmaceutical sector, where the
spiralling costs of modern innovative drug development,
together with a desire to shorten the time to market, has led
to a number of novel scientific approaches for the discovery
of new drug candidates.1 For example, the impact of
massively automated high-throughput screening and com-
binatorial chemistry approaches has been well documented.2
While the outcome of these dramatic changes in drug
discovery has yet to be fully established, the drive for
speeding up the development of many more target structures
is very real.3 Automation of development chemistry4 could
contribute to a potential solution to this anticipated problem.
To successfully automate our work, a system to handle
solution-phase chemistry under a variety of reaction condi-
tions, with on-line analytical monitoring, was required.
(5) Winicov, H.; Schainbaum, J.; Buckley, J.; Langino, G.; Hill, J.; Berkoff,
C. E. Anal. Chim. Acta 1978, 103, 469-476.
(6) Chodosh, D. F.; Wdzieckowski, F. E.; Schainbaum, J.; Berkoff, C. E. J.
Autom. Chem. 1983, 5 (2), 99-102.
(7) Chodosh, D. F.; Levinson, S. H.; Weber, J. L.; Kamholz, K.; Berkoff, C.
E. J. Autom. Chem. 1983, 5 (2), 103-107.
(8) Chodosh, D. F.; Kamholz, K.; Levinson, S. H.; Rhinesmith, R. J. Autom.
Chem. 986, 8 (3), 106-121.
(9) Frisbee, A. R.; Nantz, M. H.; Kramer, G. W.; Fuchs, P. L. J. Am. Chem.
Soc. 1984, 106, 7143-7145.
(10) Cork, D. G.; Kato, S.; Sugawara, T. Lab. Rob. Autom. 1996, 7, 301-308.
(11) Sugawara, T.; Cork, D. G. Lab. Rob. Autom. 1996, 8, 221-230.
(12) Zso´te´r, A.; Tse, R. S. Lab. Rob. Autom. 1996, 8, 329-337.
(13) Josses, P., Joux, B., Barrier, R., Desmurs, J. R., Bulliot, H., Ploquin, Y.,
Me´tivier, P., Karjalainen, E. J., Eds. Scientific Computing and Automation
(Europe); Elsevier-Science Publishers: Amsterdam, The Netherlands, 1990,
pp 249-258.
(14) Hilberink, P.; Aelst, S. V.; Vink, T. Q.; Gool, E. V.; Kasperson, F. ISLAR
Proc. (Boston) 1996, 93-106.
(15) Main, B. G.; Rudge, D. A. ISLAR Proc. (Boston) 1994, 425-434.
(16) Ciaramella, B. M. ISLAR Proc. (Boston) 1996, 314-323.
(17) The SK233 Workstation was derived from a collaboration between Anachem
and Glaxo-Wellcome.
* To whom correspondence should be addressed. Telephone: +44 (0)1279
622269. Fax: +44 (0)1279 622348. E-mail: Kenneth_T_Veal@sbphrd.com.
† SmithKline Beecham, Kent, UK.
‡ SmithKline Beecham, Essex, UK.
(1) Balkenhohl, F.; von dem Bussche-Hu¨nnefeld, C.; Lansky, A.; Zechel, C.
Angew. Chem., Int. Ed. Engl. 1996, 35, 2288-2337.
(2) Tapolczay, D. J.; Kobylecki, R. J.; Payne, L. J.; Hall, B. Chem. Ind. 1998,
19, 772-775.
(3) Persidis, A. Chem. Ind. 1998, 19, 782-784.
(4) Guette, J. P.; Crenne, N.; Bulliot, H.; Desmurs, J. R.; Igersheim, F. Pure
Appl. Chem. 1988, 60 (11), 1669-1678.
10.1021/op990005k CCC: $18.00 © 1999 American Chemical Society and Royal Society of Chemistry
Published on Web 04/24/1999
Vol. 3, No. 3, 1999 / Organic Process Research & Development
•
189

Figure 1. Tube reactors and coldfinger condensers.
For the most part, the equipment did as it claimed to do, but
for a very restricted range of our chemistry and in a very
restricted manner. As well as the usual minor mechanical
problems encountered with new and innovative equipment,
we identified several key areas that needed to be addressed
in order for us to exploit the equipment to our satisfaction.
The most immediate and pressing issue was the need for
the equipment to allow us to carry out reactions under inert
atmosphere conditions using either subambient cooling or
heating under reflux, as the inerting and refluxing capabilities
of the system supplied proved to be limited. At the same
time, we needed to be able to add reagents to the reactor
vessels and take out samples for analysis at any time during
the course of the reactions.
After numerous abortive trials, we designed18 and manu-
factured what proved to be a very satisfactory solution to
this specific problem. The tube reactors supplied were
shortened slightly and provided with a female quickfit socket
joint. Into this was inserted a coldfinger type condenser with
a hollow centre. At the top of the condenser was constructed
an inert gas purging arrangement. Above this, the vessel was
capped with a septum.19 Using this condenser/inerting/
sampling arrangement (see Figure 1), we were able to fit
the SK233 reactor block with the ability to allow us to boil
under reflux, under an inert atmosphere, and simultaneously
take an analytical sample via an XYZ robotic arm using an
automated syringe. With a minor modification to the con-
figuration of the condensers, we were able to use subambient
cooling or a reaction sequence of both cooling and heating
cycles.
The merits of the new condenser became immediately
obvious:
(i) The SK233 needed no modification for its use.
(ii) The condensers proved to be easy to fit and remove
and were very easy to clean.
(iii) The condensers being of all glass construction, we
could easily see what was going on in the reactions.
(iv) The inert atmosphere arrangement proved more than
adequate for our purposes.
(v) The operation of the reactor vessels was easy to
demonstrate and easy for our chemists to follow.
(vi) The equipment proved to be reliable and robust in
operation.
(vii) The new modified reactors could be used for most
of our chemical applications, whether with aggressive
reagents or with air-/moisture-sensitive compounds and
catalysts.
(viii) The condensing capacity of the device (using mains
water for cooling) proved outstanding and well up to our
requirements (see Table 1).
The individual condenser units were connected together
in series for use with the isolated reactor block or in
conjunction with the SK233. This multiple condenser ar-
rangement is now known as the REACTarray.20 This
arrangement has proved to be so acceptable to our chemists
(18) Furlong, B. H.; Smith, C. A. (SmithKline Beecham Plc, UK). Patent WO
9908767, 1999.
(19) An alternative arrangement is described in ref 13.
(20) The REACTarray is available under licence from Anachem Ltd., 20 Charles
St., Luton, Bedfordshire, LU2 0EB, UK.
190
•
Vol. 3, No. 3, 1999 / Organic Process Research & Development

Table 1. REACTarray condensing efficiency, demonstrated
by solvent weight loss after 2 h at or above reflux
temperature
Table 2. Design and results of fractional factorial Suzuki
experiment
response,
% conversion
block temp,
Ar flow,
% loss
by wt
factor A, factor B, factor C, factor D,
solvent
°C
mL min-1
total
water
level
boronic catalyst supplied REACT-
run solvent
acid
loading equipment
array
CH2Cl2
MeOH
MeOH
EtOH
2-propanol
toluene
45
68
120
120
120
120
10
20
20
20
20
20
∼1.3
∼0.8
∼3.1
∼1.4
∼1.3
∼0.7
1
2
3
4
5
6
7
8
9
1
1
-1
-1
-1
0
1
0
1
-1
-1
-1
-1
-1
1
0
1
0
1
1
-1
1
-1
1
-1
0
-1
0
1
-1
-1
1
1
0
-1
0
1
-1
23
4
17.5
98
98
51
18
98.5
99
12.5
91
76
84
100
95
100
100
100
100
94
Scheme 1
1
1
10
Scheme 2
that it is now used routinely for all chemistries carried out
on the reactor block or the SK233.
Results and Discussion
Suzuki Coupling Reaction. To confirm the improved
performance of the REACTarray over the reaction tubes
supplied, the Suzuki reaction, shown in Scheme 1, was
investigated manually on a reactor block. This reaction
requires both reflux conditions and a good inert atmosphere.
A fractional factorial experiment was designed (using
Design Expert) to look at the effects of concentration, water
level, catalyst loading, and boronic acid stoichiometry. Two
centrepoints were included in the design to give 10 reactions
in all and to ensure that two identical reactions were carried
out. The experiment was conducted both in the reaction tubes
supplied with the reactor block, fitted with the air-fed
“reflux” unit, and in our glassware. The results are shown
in Table 2.
It can be seen that when the supplied equipment is used,
the results vary widely. In particular, the two centrepoints
(entries 6 and 8), which should be identical, give conversions
of 51% and 98.5%. In contrast, using our device, the results
are spread over a much narrower range, as expected for this
reaction, and the centrepoints give results which are in
agreement (complete conversion). The results from the
REACTarray experiment were analysed using Design Expert
to show that, at high concentration (desirable for scale-up),
a robust and reliable process can be achieved with a high
catalyst loading.
elevated temperatures, using air- and moisture-sensitive
reagents, under inert atmospheres.
Solvent Screening of a Michael Addition Reaction. The
Michael addition of homoveratrylamine (1) to acrylonitrile21
(Scheme 2) is the first stage in the synthesis of one of our
development compounds. It is a straightforward homoge-
neous reaction, requiring a good inert atmosphere to prevent
formation of the CO2 adduct of the amine, which occurs
readily on exposure to air. It can be run neat at ambient
temperature; however, complete reaction is slow, taking at
least 2 days. An impurity (3) due to double addition of
acrylonitrile is formed in the reaction at variable levels (up
to 3%) and increases with time.
Running the reaction in a solvent was known to result in
reduced reaction times. The SK233 was used to screen a
range of solvents for the reaction simultaneously in order to
select the best one, in terms of reaction time, amount of
impurity formed, and suitability for combining with a second
synthetic step. Each reaction was set up manually, with
automatic sampling, sample preparation, and HPLC analysis
of each vessel at preset intervals. This allowed unattended
monitoring of the reactions with collection of data outside
normal working hours.
This confirmed the suitability of our device for performing
reactions under reflux in an inert atmosphere.
The glassware was then fitted to the SK233, and a wide
range of reactions were carried out with automated sampling
and on-line HPLC monitoring. This allowed unattended
operation of reactions with monitoring over extended time
periods, including overnight and over weekends. Reaction
scouting, process screening, and process optimisation inves-
tigations have all been addressed. Examples below cover
homogeneous and heterogeneous reactions, at ambient and
Initially, 10 solvents (MeOH, EtOH, IPA, H2O, DMF,
THF, 10% aqueous MeOH, 10% aqueous EtOH, 10%
aqueous IPA, and 10% aqueous THF) were screened at
ambient temperature, with sampling of each reaction 10 times
over a 24-h period (0.5, 2, 3.5, 5, 6.5, 8, 12, 16, 20, and 24
(21) Yamazaki, T. Yakugaku Zasshi 1959, 79, 1003-1008; Chem. Abstr. 1960,
54, 5678.
Vol. 3, No. 3, 1999 / Organic Process Research & Development
•
191

Figure 4. Michael addition reaction, conditions as for Figure
3, automated addition of solvents and acrylonitrile. Products:
2, water; 9, 2-propanol; [, MeOH. Impurities: 4, water; 0,
2-propanol; ], MeOH.
Figure 2. Michael addition reaction at 60 °C, 1.05 equiv of
acrylonitrile. Reaction profiles for product (2) and impurity (3)
formation. Products: 2, water; 9, 2-propanol; [, MeOH.
Impurities: 4, water; 0, 2-propanol; ], MeOH.
Table 3. Michael addition reaction: comparison of Figure 3
and Figure 4 final time point values, HPLC % PAR
Figure 3
Figure 4
2, water product
59
95
77
30
1
66
96
83
30
2
9, 2-propanol product
[, MeOH product
4, water impurity
0, 2-propanol impurity
], MeOH impurity
19
15
Scheme 3
Figure 3. Michael addition reaction at 60 °C, 1.5 equiv of
acrylonitrile. Reaction profiles for product (2) and impurity (3)
formation. Products: 2, water; 9, 2-propanol; [, MeOH.
Impurities: 4, water; 0, 2-propanol; ], MeOH.
h). From this experiment, five solvents (MeOH, EtOH, IPA,
H2O, and 10% aqueous THF) were selected for further
evaluation. The second set of reactions was run at 60 °C,
with stoichiometries of 1.05 and 1.5 equiv of acrylonitrile.
As the reactions were expected to be much faster at elevated
temperature, sampling was more frequent in the early stages,
with each reaction being sampled after 5, 20, 35, and 50
min, and then after 2.5, 4.25, 7.5, and 15 h.
The large amount of data collected in these experiments
(180 chromatograms in total) was analysed using Excel, to
allow easier interpretation of the results. Reaction profiles
for appearance of product (2) and impurity (3) were
generated, and selected examples are shown in Figures 2
and 3. It can be seen that 2-propanol is an excellent solvent
for the reaction, with complete conversion achieved within
2.5 h at 60 °C and little formation of impurity (3), even in
the presence of excess acrylonitrile.
acid-catalyzed addition of silylenol ether (4) to acetoxyaze-
tidinone (5) (Scheme 3). Initial investigations in traditional
glassware indicated that using boron trifluoride etherate in
dimethoxyethane gave a reasonable yield (ca. 50%) of desired
product ketone (6). We wished to investigate the effect of
different Lewis acids and solvents on this chemistry to see
whether we could obtain an improved yield and a more
robust reaction (the product ketone (6) is unstable to boron
trifluoride etherate under prolonged reaction times). Using
the SK233, we were able to screen, in a short period of time,
all the combinations of the 20 Lewis acids and 7 solvents
(Scheme 3). As expected, boron trifluoride etherate gave
good results, although a higher yield was obtained using
tetrahydrofuran as the solvent. A number of other effective
catalysts were identified, notably, zinc chloride in all solvents
(except toluene), magnesium bromide in dichloromethane,
and copper(I) iodide in dichloromethane. These results were
subsequently verified in traditional glassware.
To check reproducibility of the system, the second set of
experiments was repeated. In this instance, the SK233 was
additionally used to automatically dispense solvents and
acrylonitrile. The resulting data and graphs were almost
identical to those acquired previously, as illustrated in Figures
3 and 4 and Table 3.
Rapid Screening of Reagents and Solvents. A key step
in the preparation of a novel oral carbapenem is the Lewis
192
•
Vol. 3, No. 3, 1999 / Organic Process Research & Development

Scheme 4
N
Even the notoriously air- and moisture-sensitive reagents
such as titanium tetrachloride, boron trichloride, and boron
tribromide were handled with ease by the equipment. No
white HCl vapour could be seen in the reactors during the
reagent addition step, indicating an effective moisture-free
environment. The large amount of data generated (some 420
chromatograms in total) was again handled using Excel.
Fluoride Desilylation of a Protected Alcohol. As part
of the same carbapenem synthesis, we wished to optimise
the fluoride-mediated desilylation of the protected alcohol
(7) (Scheme 4). Preliminary labwork indicated that a solution
yield of 69% could be obtained using 1.1 molar equivalents
of fluoride reagent in 10 volumes of N-methylpyrrolidinone
solvent at 50 °C under nitrogen. However, this reaction was
not particularly robust; if it was left too long, the yield of 8
decreased dramatically. Furthermore, increasing the temper-
ature increased the reaction rate but resulted in a lower yield.
We decided to use a design of experiments approach to
optimise the reaction. Three factors were considered to be
important: temperature, concentration, and stoichiometry of
reagent. Using Design Expert software, a three-factor, two-
level full factorial design was drawn up, resulting in eight
experiments. Four identical centrepoint experiments were
included, making 12 reactions in total.
Since temperature was a variable and the reactor block
can only be run at one temperature at a time, we performed
the 12 experiments in three blocks of four experiments. Each
reaction was monitored hourly by HPLC, and solution yields
were calculated for each time point by reference to an internal
standard, 4-nitroaniline. Design Expert software indicated
that a yield of 83% could be expected at 30 °C with 1.25
equiv of fluoride reagent in 4-7 volumes of solvent.
This result was verified by experiment in a traditional
round-bottom flask. The optimised conditions have subse-
quently been used at plant scale with good success.
Figure 5. Desilylation of TBS alcohol (7), reaction profiles for
formation of product (8). 4, HHH; 0, HLL.
Scheme 5
Scheme 6
Homogeneous Starting Material and Product: Het-
erogeneous Base. To investigate the viability of handling
heterogeneous reaction mixtures with the SK233, the alkyl-
ation of isatin (9) with 1-bromo-3-chloropropane in DMF
with solid potassium carbonate22 was carried out (Scheme
5). Three identical reactions were set up by dispensing a
solution of isatin in DMF to stirred reaction vessels which
had been precharged with the base. A visual check showed
that efficient mixing was obtained. 1-Bromo-3-chloropropane
was added at varying dispense rates, and each reaction was
sampled five times over a 10-h period, with duplicate samples
being taken at each time point. Analysis of the HPLC results
found that reasonable product peak reproducibility was
obtained quantitatively ((6%) between duplicate samples
and from vessel to vessel. Excellent reproducibility of
impurity profile data was observed (main peak area ratio
(0.04%).
Heterogeneous Starting Material and Heterogeneous
Product. Following the success of the above example, the
ability of the SK233 to handle more difficult heterogeneous
reaction mixtures was explored. The chemistry shown in
Scheme 6 presented a particular challenge in that the starting
material slowly dissolved and the product crystallised out
as the reaction proceeded. As a consequence, the reaction
was both heterogeneous and viscous, making it difficult to
sample. To investigate the effectiveness of the machine to
sample efficiently, we performed two sets (0.5 and 1.0 molar
equivalents of NBS) of three identical reactions. Each
reaction was sampled at three time points (5 min, 45 min,
The 12 experiments were performed in a 3-day period,
with HPLC data being transferred to Excel for manipulation.
Good reproducibility was seen, with the four centrepoints
indicating good sampling by the robot.
Excellent reaction profile data were also obtained. It can
be seen from Figure 5 why higher temperatures or prolonged
stirring at 50 °C were detrimental to the process. The reaction
where all the levels were high (depicted HHH in the figure),
i.e., high temperature, high amount of reagent, and high
concentration, clearly showed a rapid reaction, reaching a
maximum concentration of product after 4 h and then a rapid
decomposition of the product. In the experiment where the
temperature and amount of reagent were low (HLL), a much
slower reaction resulted, but a higher yield was obtained,
presumably due to little or no decomposition of the product.
(22) Radul, O. M.; Zhungietu, G. I.; Rekhter, M. A.; Bukhanyuk, S. M. Chem.
Heterocycl. Compd. 1983, 19 (3), 286-288.
Vol. 3, No. 3, 1999 / Organic Process Research & Development
•
193

and 1.5 h), with duplicate samples taken at the final time
point.
Dimethoxyethane (1.55-7 mL) and water (0.9-2.5 mL), as
designated by the experimental design, were added to the
vessels, and the reactions were stirred under argon for 5 min.
Boronic acid (lithium salt; 1.0-1.3 equiv) was added to the
reaction mixture, followed after a further 2 min by tetrakis-
(triphenylphosphine)-palladium(0) (0.5-3.0 mol %). The
reactions were heated to reflux and stirred under argon at
this temperature for 15 h. Each reaction was analysed by
HPLC (Zorbax Eclipse XDB-C8 3.5-µm, 4.6- × 75-mm
column; 60% MeCN/H2O; 1.5 mL/min; 265 nm), and the
conversion of aryl bromide to the coupled product was
calculated.
Typical Procedure for Michael Addition Reaction.
Homoveratrylamine (1) (1 g) was charged to each reaction
vessel, and the vessels were stirred at ambient temperature
under argon. Solvent (10 mL) was added to each vessel, and
the Stem block was heated to 60 °C. The reactions were
then treated with acrylonitrile (1.05 or 1.5 equiv) and were
monitored automatically by HPLC over 15 h (Merck RPSelB
4.6- × 125-mm column; 30% MeCN/0.05 M aqueousTFA;
1.5 mL/min; 275 nm).
Typical Procedure for Lewis Acid Screen. A stock
solution of the silylenol ether (4) (20.9 mmol) and the
acetoxy azetidinone (5) (17.4 mmol) in 100 mL of the
reaction solvent was prepared. An aliquot (7 mL) of this
solution was charged by the SK233 to each reaction vessel.
An aliquot (1.39 mL) of the selected Lewis acid as a 1.0 M
solution in dichloromethane (except for ZnCl2 and FeCl3,
which were 1.0 M solutions in diethyl ether) was charged
to the specific reaction vessel over 1.4 min. The reactions
were stirred at room temperature and automatically sampled
(30 µL) at 5 min, 30 min, and 6 h. The samples were diluted
with acetonitrile (1.5 mL), mixed, and analysed by LC
(Perkin-Elmer C18 HS 3-µm, 3.3-mm × 4.6-cm column;
gradient MeCN/water 30:70 to 100:0 over 3 min, 100%
MeCN for 1 min, back to 30:70 over 0.03 min, and 30%
MeCN for 2 min; 2 mL/min; 246 nm).
Typical Procedure for Desilylation. The solid-protected
alcohol 7 (1.0 g) was manually preweighed into each reaction
vessel. The appropriate amounts of N-methylpyrrolidinone
solvent, as designated by the experimental design, containing
a precise amount of 4-nitroaniline as internal standard, was
dispensed by the SK233, followed by the triethylamine
trihydrofluoride complex. The reactions were stirred at the
appropriate temperature and sampled hourly (30 µL). The
samples were diluted with acetonitrile (1.5 mL), mixed, and
analysed by LC (Perkin-Elmer C18 HS 3-µm, 3.3-mm ×
4.6-cm column; gradient MeCN/water 30:70 to 100:0 over
3 min, 100% MeCN for 1 min, back to 30:70 over 0.03 min,
and 30% MeCN for 2 min; 2 mL/min; 246 nm).
HPLC analysis of the samples showed that no meaningful
quantitative data could be obtained, because, as expected,
the equipment could not accurately aspirate precise amounts
of the viscous heterogeneous reaction mixture.23 However,
qualitatively very good reaction profile data were obtained,
with good agreement between the replicate reactions and
duplicate samples (main peak area ratio (1%). To verify
these results, we dissolved the final reaction mixture in
solvent to give an homogeneous mixture. HPLC analysis of
this gave a profile comparable to that of the on-line sample.
This illustrates clearly that the equipment can be utilised to
sample an heterogeneous reaction mixture representatively,
though equally clearly there are practical limits to what can
be handled.
Conclusion
We have shown that parallel synthesis and automation
can play a valuable role in a process chemistry environment.
It can assist in increasing throughput and accumulating more
high-quality data, while releasing chemists to do other tasks.
However, these benefits can only be realised if the equipment
can carry out a wide range of chemistry and is acceptable to
chemists. By designing our own reactors (by chemists for
chemists), we have achieved this.
The SK233 and REACTarray combination has proven to
be a versatile tool, capable of performing reactions at reflux,
in an inert atmosphere, with aggressive and air-/moisture-
sensitive reagents. Reaction monitoring is integral to the
system. It is useful for route scouting, process screening,
and process optimisation. The system is reliable, reproduc-
ible, and convenient to use. Along with these advantages,
the equipment described is compact, fitting standard labora-
tory facilities, and, more importantly, is relatively affordable,
being of comparable cost to a modern HPLC system.
Other commercial systems targeted at process develop-
ment have now become available,24,25 and others are no doubt
“in the pipeline”, which may cover some aspects of the
equipment described here. We intend to evaluate these
alternatives, whilst exploiting the capabilities of the SK233
and REACTarray to the fullest extent.
Experimental Section
The reaction block used was a 10-position RS1000 reacto-
station from Stem Corp.,26 capable of heating from ambient
temperature to 150 °C with variable-speed magnetic stirring.
HPLCs were run on either Gilson or Merck-Hitachi instru-
ments, and the data were collected using Gilson Unipoint
software.
Typical Procedure for Isatin Alkylation. Potassium
carbonate (1.2 g) was charged to each reaction vessel, and
the stirring was started. A 10% (w/v) solution of isatin (9)
in DMF was prepared, and 4.9 mL of the solution was
dispensed to each reaction. 1-Bromo-3-chloropropane (3.35
mL) was then added at varying dispense rates (2.5-10 mL/
min). The reactions were stirred at ambient temperature and
monitored overnight by HPLC (Merck RPSelB 4.6- × 125-
Typical Procedure for Suzuki Coupling Reaction. Aryl
bromide (250 mg) was charged to the reaction vessels. 1,2-
(23) These experiments were carried out using the 0.4-mm-i.d. needle supplied.
Improved sampling has been demonstrated using a recently fitted 0.8-mm-
i.d. needle.
(24) Studt, T. R&D Mag. 1997, 39 (12), 38-42.
(25) Harness, J.; Tedesco, J. R&D Mag. Suppl. 1998 (April), 20-25.
(26) Stem Corp., Woodrolfe Road, Tollesbury, CM9 8SJ, UK.
194
•
Vol. 3, No. 3, 1999 / Organic Process Research & Development

mm column; 35% MeCN/0.05 M NH4OAc pH 4.5; 1 mL/
min; 245 nm).
Acknowledgment
We thank Adrian Bateman and Kevin Sutcliffe for
analytical support, and Bryan Davies, Tom Ramsay, Robin
Attrill, Simon Howie, Richard Anderson, and John Hayler
for input to the chemistry. Grateful thanks go to Clive Smith
for the condenser design and Brian Furlong for developing
the idea into real equipment. We also acknowledge Anachem
Ltd., Glaxo-Wellcome, and Stem Corp. for the development
of the SK233 hardware and Aitken Scientific for writing the
SK233 software.
Typical Procedure for Indoline Bromination. N-Acetyl-
5-hydroxyindoline (350 mg) was charged to the reaction
vessels. Acetic acid (5.8 mL) was dispensed at ambient
temperature to the stirred vessels, and then N-bromosuccin-
imide (0.5 or 1 equiv) was added in three portions to each
reaction. The reactions were monitored automatically by
HPLC over 1.5 h, with repeat samples being taken at the
final time point. DMSO was added to the reactions to give
complete solution, and a further HPLC analysis was carried
out (HPLC conditions: Merck RPSelB 4.6- × 125-mm
column; 40% MeCN/0.1% aqueous TFA; 1 mL/min; 256
nm).
Received for review January 11, 1999.
OP990005K
Vol. 3, No. 3, 1999 / Organic Process Research & Development
•
195