Investigation of the moffatt#swern oxidation in a continuous flow microreactor system

Article abstract of DOI:10.1021/op700228e

The Moffatt-Swern oxidation of different alcohols is performed in a continuous flow microreactor system. The microreactor process offers significant advantages over the batch process. First, because of the small reactor volume, accumulation of the labile trifluoroacetoxydimethylsulfonium salt (3) and alkoxydimethylsulfonium salt (5) is minimized. Second, because of the short residence times, which can be applied in the microreactor, the exothermic Pummerer rearrangement of the unstable intermediate 3 is limited. Because of this, the process can be operated at remarkably high temperatures in comparison with a batch reaction, viz. 0-20°C instead of -70°C. In the present study, a continuous flow microreactor system was optimized using reactors of different volumes allowing modulation of the residence times of labile intermediates. The efficiency of mixing was studied using different mixing devices. It has been shown that the continuous flow microreactor is an ideal tool for rapid optimization of reaction parameters. Furthermore, the scaleability and reliability of the microreactor was tested by running the system for several hours. For testosterone, the system was in process for 1.5 h without any problems, resulting in an 4-androstene-3,17-dione production rate of 64 g·h^-1.

Full text of DOI:10.1021/op700228e

Organic Process Research & Development 2008, 12, 911–920
Full Papers
Investigation of the Moffatt-Swern Oxidation in a Continuous Flow Microreactor
System
Jacobus J. M. van der Linden, Peter W. Hilberink, Claudia M. P. Kronenburg, and Gerardus J. Kemperman*
N.V. Organon, Department of Process Chemistry, P.O. Box 20, 5340 BH Oss, The Netherlands
Abstract:
mixing and chemical reactions take place on a scale with
characteristic dimensions in the submillimeter range and with
reaction volumes in the nanoliter-to-milliliter range.5 As a result,
mixing and heat transfer in the microreactor are extremely fast.
These features are quite advantageous for conducting fast and
highly exothermic reactions. Additionally, the volume of a
microreactor is very small compared to that of a batch reactor,
which ensures that only small amounts of reactive, and
potentially dangerous, intermediates can accumulate. This
contributes to an increased safety.
The Moffatt-Swern oxidation of different alcohols is performed
in a continuous flow microreactor system. The microreactor
process offers significant advantages over the batch process. First,
because of the small reactor volume, accumulation of the labile
trifluoroacetoxydimethylsulfonium salt (3) and alkoxydimethyl-
sulfonium salt (5) is minimized. Second, because of the short
residence times, which can be applied in the microreactor, the
exothermic Pummerer rearrangement of the unstable intermediate
3 is limited. Because of this, the process can be operated at
remarkably high temperatures in comparison with a batch
reaction, viz. 0-20 °C instead of -70 °C. In the present study, a
continuous flow microreactor system was optimized using reactors
of different volumes allowing modulation of the residence times
of labile intermediates. The efficiency of mixing was studied using
different mixing devices. It has been shown that the continuous
flow microreactor is an ideal tool for rapid optimization of reaction
parameters. Furthermore, the scaleability and reliability of the
microreactor was tested by running the system for several hours.
For testosterone, the system was in process for 1.5 h without any
problems, resulting in an 4-androstene-3,17-dione production rate
In order to assess the advantages of microreactors for highly
exothermic reactions, the Moffat-Swern oxidation was inves-
tigated in a microreactor system from the company Ehrfeld
Mikrotechnik BTS (GmbH).6
The Moffatt-Swern oxidation is a multistep process by
which a number of side products can be formed (Scheme 1)
according to the mechanism as postulated by Omura et al.7,8
The Moffatt-Swern oxidation is a versatile metal-free oxidation
method that finds application in the transformation of primary
and secondary alcohols into aldehydes and ketones, respectively.
However, application of Moffatt-Swern oxidation in process
chemistry is hampered by the low temperature requirement, viz.
-70 °C, and the highly exothermic behavior, which makes
temperature control very difficult. In addition, the reaction
proceeds through labile reactive intermediates, viz. trifluoro-
acetoxydimethylsulfonium salt (3) and alkoxydimethylsulfo-
nium salt (5), which imposes the risk of a runaway when a
certain temperature is exceeded. The exotherm causing the
runaway can be attributed to the Pummerer rearrangement of
intermediates 3 and 5 which takes place above -30 °C and
leads to the formation of trifluoroacetoxymethyl methyl sulfide
(7) and methylthiomethyl ether (8), respectively. The trifluo-
roacetoxymethyl methyl sulfide (7) can subsequently react with
benzyl alcohol to benzyltrifluoroacetate (9) under the basic
conditions after the quench. Obviously the uncontrolled de-
composition of intermediates 3 and 5 is also detrimental to the
yield of the process.
of 64 g ·h-1
.
Introduction
In the pharmaceutical industry reactions are traditionally
performed in multipurpose batch reactors.1 However, for
extremely exothermic reactions or reactions that proceed through
highly reactive and unstable intermediates, the use of batch
reactors has major disadvantages that are related to temperature
control and safety problems.2,3 This is mainly attributed to a
diminishing cooling capacity and diminishing heat- and mass
transport through a batch reactor due to a limited stirring
capacity upon scale-up.4 In addition, in a batch reactor there is
the possibility of accumulation of reactive intermediates that
impose a risk for a runaway reaction.
Continuous flow microreactors could, in principle, provide
a solution to the above-mentioned problems. In a microreactor,
A continuous flow process would be an ideal setup to
consume the trifluoroacetoxydimethylsulfonium salt (3) and
* Corresponding author. E-mail: Gerjan.kemperman@organon.com. Tele-
phone: +31 412 669375. Fax: +31 412 663513.
(1) Roberge, D. M.; Dury, L.; Beiler, N. Chem. Eng. Technol. 2005, 28,
(5) Zhang, X.; Stefanick, S.; Villani, F. Org. Process Res. DeV 2004, 8,
318–322.
455–460.
(2) Ainsworth, D.; Brocklebank, M. Chem. Eng. 2003, 110, 42–49.
(3) Anderson, N. Org. Process Res. DeV. 2001, 5, 613–621.
(4) Taghavi-Moghadam, S.; Kleemann, A.; Golbig, K. G. Org. Process
Res. DeV 2001, 5, 652–658.
(6) http://www.ehrfeld.com.
(7) Omura, K.; Sharma, A. K.; Swern, D. J. Org. Chem. 1976, 41, 957–
962.
(8) Omura, K.; Swern, D. Tetrahedron Lett. 1978, 34, 1651–1660.
10.1021/op700228e CCC: $40.75
Published on Web 04/30/2008
2008 American Chemical Society
Vol. 12, No. 5, 2008 / Organic Process Research & Development
•
911

Scheme 1. Moffatt-Swern oxidation
alkoxydimethylsulfonium salt (5) directly after their formation,
such that no accumulation takes place. Furthermore, the highly
efficient heat transfer in the microreactor should solve the
problem of limited cooling capacity experienced in a batch
reactor.
At the start of our study a paper from Yoshida et al.9
appeared, in which the Moffatt-Swern oxidation was carried
out at higher temperatures in a microreactor. However, whereas
Yoshida used syringe pumps to feed reagents into the microre-
actor, in our system HPLC pumps were employed, allowing
for a real continuous process.
The Moffatt-Swern oxidation was investigated in different
setups for the microreactor and by using a series of aromatic
alcohols as the substrate. We also report on our efforts to
perform reaction parameter screening in the microreactor. The
effects of mixing efficiency and residence time were investigated
in order to reveal the root cause for the beneficial effects of a
continuous flow microreactor on the Moffat-Swern oxidation.
Finally, the scaleability and consistency of the Moffatt-Swern
oxidation were tested using several more functionalized alcohols
such as a steroid and carbohydrates.
(R1, R2, and R3) of different lengths and diameters in order
to vary the residence times (see Experimental Section). The
micromixers were placed in a homemade double jacketed
stainless-steel vessel. The Teflon lid contains two holes for
cooling, four inlet holes for the reagents, one outlet hole and
one hole for a Pt-100 thermometer. The vessel is cooled or
heated by a Julabo LH85. The reagents were fed into the system
using Knauer K501 HPLC pumps (P1, P2, P3, and P4).
The reactor volumes of R1, R2, and R3, in the different
setups (I-IV) as referred to in the paper are detailed in the
Experimental Section. The reactor volumes of R1-R3 are in
the range of 2–253 µL. The volumes of the mixers are negligible
in comparison with those of the reactors. This implies that the
reactions predominantly take place in the reactors and not in
the mixers.
Mechanistic Considerations and Development of an
Appropriate Analytical Method. All HPLC analyses were
performed off-line. Due to the short reaction times, it was
possible to collect a large number of samples from the
microreactor in a relatively short time (up to 20 samples in 1 h).
Therefore, it was important that the collected samples were
stable before injection. The effluent of the microreactor contains,
besides dichloromethane, alcohol, the corresponding ketone or
aldehyde, DMSO, several side products such as the trifluoro-
acetate ester 9, methylthiomethyl ether 8, and the base. As
discussed in the Introduction, on the basis of the proposed
mechanism of the Moffat-Swern oxidation, the trifluoroacetate
ester 9 can be formed by reaction of Pummerer rearrangement
product 7 with the alcohol. In order to investigate whether
reactions between various species could take place in the
sample, alcohol was exposed to freshly prepared Pummerer
rearrangement product 7 under acidic conditions (TFA, 10 min
Results and Discussion
Microreactor Setup. A schematic drawing of the microre-
actor setup used is shown in Figure 1. The microreactor consists
of three stainless-steel LH2 stand-alone micromixers (M1, M2,
and M3).6 This type of micromixer is based on the principle of
multilamination, i.e. successive splitting and rearrangement. For
that purpose each micromixer contains a mixing plate with 2
× 10 slides and a slit diameter of 50/50 µm and an aperture
plate with a slide diameter of 50 µm and a length of 2 mm.
The mixers were connected with stainless-steel tube reactors
912
•
Vol. 12, No. 5, 2008 / Organic Process Research & Development

and 1 h) and under basic conditions (triethylamine, 10 min and
1 h). This revealed that under acidic conditions, compound 7
can react with the alcohol to give methylthiomethyl ether 8,
but under basic conditions, this reaction is not observed. As
the alcohol is only exposed to compound 7 in an acidic medium
during the short residence time in reactor R2, it can be
concluded that the pathway from compound 7 to methylthi-
omethyl ether 8 can be ignored in the microreactor setup. This
implies that the amount of methylthiomethyl ester 8 in the
sample is predominantly formed by Pummerer rearrangement
of intermediate 5. It was also found that under basic condi-
tions, the Pummerer rearrangement product 7 reacts with the
alcohol to give the trifluoroacetate ester 9 as is indicated in
Scheme 1. This reaction appears to be nearly completed in 10
min, while the quenched samples taken from the flow coming
out of R3 are stored for up to 16 h. According to the mechanism
shown in Scheme 1, trifluoroacetate ester 9 can also be formed
by solvolysis of intermediate 5. However, as this solvolysis can
only take place during the short residence time in R2, it is
conceivable that most of the trifluoroacetate ester 9 is formed
after quench and sample storage. The trifluoroacetate 9 can thus
be considered as “protected” alcohol formed by reaction of
alcohol with Pummerer rearrangement product 7 under the basic
conditions after quench. As such, there is a correlation between
the amount of Pummerer rearrangement taking place in R1 and
the amount of trifluoroacetate 9 that is formed. The more
Pummerer rearrangement takes place in R1, the more trifluo-
roacetate 9 is formed upon quench. A sample stability study
disclosed that, when a sample of effluent was diluted (1:200)
with THF, which was the preferred injection solvent for the
HPLC system used, unfortunately the trifluoroacetate ester 9
in the samples is instantaneously hydrolyzed to alcohol. Hence,
the amount of unreacted alcohol cannot be distinguished from
the amount of trifluoroacetate ester 9 formed by reaction of
alcohol with Pummerer rearrangement product 7. Also the
Pummerer rearrangement product 7 is hydrolyzed in the sample
matrix, resulting in small volatile molecules; thus, a direct
measure of the amount of Pummerer rearrangement taking place
in R1 was not possible. Considering the extremely fast rate of
the reaction between the alcohol and intermediate 3, it can be
safely assumed that any unreacted alcohol in the reaction
mixture results from an insufficient amount of intermediate 3
being fed into reactor R2 due to partial conversion of intermedi-
ate 3 into compound 7 via Pummerer rearrangement. Therefore,
the amount of alcohol in the sample is the most reliable indirect
measure for the amount of Pummerer rearrangement to com-
pound 7 taking place in R1. It was demonstrated that by using
this quench the methylthiomethyl ethers 8 were stable, and thus,
a direct measurement of the amount of Pummerer rearrangement
taking place in R2 was possible. In the graphs and tables
throughout this paper, the conversion of benzyl alcohol to
aldehyde (or ketone) and to the corresponding methyl thiom-
ethyl ether 8 is provided. The amount of alcohol in the reaction
mixture can be deduced from these numbers and is therefore
not explicitly presented.
Optimization of Reaction Parameters. First a systematic
variation of process parameters such as temperature, residence
time, and equivalents of reagents was studied. Benzyl alcohol
(4a) was used as a test substrate in these experiments. As
discussed in the previous section, the amount of unconverted
benzylalcohol is related to the amount of Pummerer rearrange-
ment in reaction R1. The amount of methyl thiomethylether is
a direct measure for the amount of Pummerer rearrangement
in reactor R2.
The optimization of the microreactor was investigated with
three different types of setups as shown in Table 1. The error
in the measured conversions has been established to be
approximately 2%.10
Comparison of the Batch Reaction with the Continuous Flow
Reaction in Setups I, II, and III. Comparison of the temperature
profiles obtained from the different microreactor setups (Table
1) with the temperature profile of the batch reactions (Table 2)
shows that the conversion to benzaldehyde (6a) in the microre-
actor at the same temperature is higher than in the batch reactor.
It is notable that even at temperatures above -20 °C, the
conversion is still reasonable, while at these temperatures there
was no conversion at all in the batch reactions.
Figure 1. Microreactor setup.
Vol. 12, No. 5, 2008 / Organic Process Research & Development
•
913

Table 1. Conversiona to benzaldehyde (6a) and
finding is in accordance with literature data,11,12 indicating that
3 can crystallize at -30 °C. These initial microreactor results
were very promising for further studies and showed indeed that
the Moffatt-Swern oxidation is an almost instantaneous reac-
tion, taking place in the time frame of milliseconds.
methylthiomethyl ether (8a) in reactor setups I, II, and III
conversionb (mol/mol %)
benzaldehyde (6a)
methylthiomethyl ether (8a)
temp (°C) setup I setup II setup III setup I setup II setup III
It is notable that in the batch reactor at higher temperatures
no increased amounts of methylthiomethyl ether are observed.
This indicates that the Pummerer rearrangement of intermediate
5 does not play a major role, which is in contrast to conclusions
in prior literature.7
-20
-10
0
81
82
40
3.8
1.5
81
81
79
54
10
77
76
75
54
25
3.9
3.3
2.1
0.2
0.0
3.8
3.9
3.8
2.8
0.5
4.2
3.9
4.1
4.2
1.7
10
20
a Conditions: DMSO (4.0 M); TFAA (2.1 M); benzyl alcohol (2.0 M), TEA
(5.8 M). Flow rates of all the pumps were 5.0 mL/min-1 b Conversion determined
by HPLC. The residence times (ms) of the different reactors at a flow rate of 5.0
mL·min-1 setup I: R1 ) 760; R2 ) 507; setup II: R1 ) 12; R2 ) 507; setup III:
R1 ) 12; R2 ) 20.
InVestigation of the Stoichiometry of the Reaction. The
optimization of process parameters was continued by investigat-
ing the stoichiometry of the reaction. To increase the obtained
conversion to benzaldehyde of ∼78 mol/mol % the effect of
the amount of trifluoroacetoxydimethylsulfonium salt (3) relative
to benzylalcohol (4a) was studied using setup III at -20 °C.
This was done by changing the flow rate of the reagents in two
different types of experiments. It was ensured that in all cases
DMSO was present in excess to TFAA such that a competition
between TFAA and trifluoroacetoxydimethylsulfonium salt (3)
is avoided. In the first experiment the flow rates of DMSO and
TFAA were varied between 1.0 and 10.0 mL ·min-1, and the
flow rates of benzyl alcohol (4a) and TEA were maintained
constant at 5.0 mL ·min-1. By doing this, the residence times
in R1, R2, and R3 were variable. In the second experiment
only the flow rate of benzyl alcohol (4a) was varied between
1.0 and 8.0 mL ·min-1, and the flow rates of DMSO, TFAA,
and TEA were maintained constant at 5.0 mL·min-1. As a
consequence, the residence time in R1 was constant, and the
residence times in R2 and R3 were variable. According to the
results shown in Figure 2, the maximum conversion to ben-
zaldehyde is still around ∼80 mol/mol %.
.
Table 2. Temperature profile of the Moffatt-Swern
oxidation performed in a conventional batch reactor
conversion (mol/mol %)a
temp (°C) benzaldehyde (6a) methylthiomethyl ether (8a)
-70
-45
-30
-10
10b
88
1.6
1.5
1.2
1.4
0.0
0.0
58
28
2.1
0.0
0.0
20b
a Conversion determined by HPLC. b Isolation of only benzyl alcohol (4a).
The maximum conversion of ∼81 mol/mol % obtained at
-20 °C in the microreactor is for every setup the same and is
comparable with the maximum conversion obtained in the batch
reactor at much lower temperature (-70 °C). At higher
temperature the conversion in the microreactor also decreases.
In setup I, the conversion starts to decrease at -10 °C; for setups
II and III, the conversion is still good at -10 °C and starts to
decrease at 0 °C. This higher conversion in setups II and III is
related to the low thermal stability of trifluoroacetoxydi-
methylsulfonium salt (3) and its shorter residence time in setups
II and III compared to that in setup I. By going from setup I to
II and III the residence time of 3 in R1 decreases from 760 ms
to 12 ms. As a result of this shorter residence time, the
Pummerer rearrangement of 3 to 7 is suppressed, since 3 reacts
almost instantaneously with benzyl alcohol (4a). The difference
between the results in setup I and those in setups II and III is
even more pronounced at temperatures above 0 °C.
At temperatures above 0 °C the conversion in setup II and
III also decreases in spite of the short residence time in R1,
indicating that at these temperatures the Pummerer rearrange-
ment occurs even faster. Going from setup II to setup III the
residence time in R2 changed from 507 ms to 20 ms, and the
temperature profile gave a similar result, with a slightly higher
conversion for setup III at higher temperatures. In order to
increase the conversion even further, the temperature of the
cryovessel was lowered to -30 °C. However, at this temperature
the microstructures in the mixers were blocked due to the
precipitation of trifluoroacetoxydimethylsulfonium salt (3). This
The Moffatt-Swern oxidation is a very fast reaction, and
consequently, it takes a short time to reach steady state in the
microreactor. This feature offered the advantage that many
different settings could be tested in a short period of time (more
than 20 experiments in 1 h). This illustrates that the microreactor
is an ideal tool for continuous flow high throughput experi-
mentation. The standard deviation (typical in the order of <2%
of the average) calculated over three consecutive samples
showed that, when the system is in steady-state, the conversion
is very reproducible. Identical settings were repeated also at
different days, resulting in similar amounts of 6a and 8a. This
consistency is, of course, a vital prerequisite for successful
development and scale-up of a microreactor process.
InVestigation of the Residence Times at Different Temper-
atures. Subsequently, the residence time was investigated as a
reaction parameter (Figure 3). This was done by increasing the
flow rate of the four pumps from 2.0 mL ·min-1 up to 10
mL ·min-1 (the maximum flow of the pumps), while maintain-
ing the ratio of the flow rates, and thus the reagents, constant.
Decreasing the residence time appeared to have no effect
on the conversion at low temperature (-20 °C). At higher
temperature (10 °C) it is evident that at short residence times a
higher conversion to benzaldehyde (6a) is achieved. This effect
is again explained by the low thermal stability of the trifluo-
(9) Kawaguchi, T.; Miyata, H.; Ataka, K.; Mae, K.; Yoshida, J. Angew.
Chem., Int. Ed. 2005, 44, 2413–2416.
(11) Sharma, A. K.; Swern, D. Tetrahedron Lett. 1974, 16, 1503–1506.
(12) Sharma, A. K.; Ku, T.; Dawson, A. D.; Swern, D. J. Org. Chem. 1975,
40, 2758–2764.
(10) A table with all data points from which the errors can be calculated
is included in the Supporting Information.
914
•
Vol. 12, No. 5, 2008 / Organic Process Research & Development

Figure 2. Conversion of benzyl alcohol into benzaldehyde (6a) and methylthiomethyl ether (8a) as function of the equivalents
TFAA at –20 °C in microreactor setup III. Conditions: DMSO (4.0 M); TFAA (2.1 M); benzyl alcohol (2.0 M), TEA (5.8 M). Flow
rates of DMSO, TFAA, and benzyl alcohol were variable; flow rate of TEA was 5.0 mL/min-1. Conversion determined by HPLC.
Figure 3. Conversion of benzyl alcohol into benzaldehyde (6a) and methylthiomethyl ether (8a) as a function of the residence time
in microreactor setup III at two different temperatures. Conditions: DMSO (4.0 M); TFAA (2.1 M); benzyl alcohol (2.0 M), TEA
(5.8 M); flow rates 2.0 mL ·min-1 up to 10 mL ·min-1; conversion determined by HPLC.
roacetoxydimethylsulfonium salt (3) and alkoxydimethylsulfo-
nium salt (5). At higher temperature the decomposition of these
intermediates by the Pummerer rearrangement is faster which
can be counterbalanced by shorter residence times.
ylsulfonium salt (3) is formed in the presence of benzyl alcohol.
Thereby, 3 is immediately intercepted by benzyl alcohol leaving
virtually no time relative to the rate of the Pummerer rear-
rangement even at a temperature of 20 °C. In setup III,
intermediate 3 is first transported from M1 to M2 where it is
mixed with benzyl alcohol, resulting in longer residence time
during which the Pummerer rearrangement can take place.
Upon increasing the flow rate at 10 °C, the conversion
increased in setup III. The opposite effect was found for setup
IV (Figure 5); where at higher flow rates the conversion
decreased slightly from 73% to 68%. This indicates that at these
very short residence times, the residence time approaches the
required reaction time for the formation of alkoxysulfonium
salts (3 and 5). This appears to be the first instance in which
Figure 4 shows the temperature dependence of the Moffatt-
Swern oxidation when in M1 DMSO and benzylalcohol (4a)
were premixed followed by the addition of TFAA in M2 and
TEA in M3. In this microreactor setup (named setup IV) the
reaction volumes (R1-R3) are identical to those of setup III.
The study of the temperature dependence in setups III and
IV shows that at 20 °C the conversion in setup IV is 70%,
whereas in setup III the conversion is only 25%. This influence
on the maximum temperature at which the process can be
operated, can be explained as follows: In microreactor setup
IV, upon addition of TFAA in 2, the trifluoroacetoxydimeth-
Vol. 12, No. 5, 2008 / Organic Process Research & Development
•
915

Figure 4. Conversion of benzyl alcohol into benzaldehyde (6a) and methylthiomethyl ether (8a) as a function of the temperature
for setups III and IV. Conditions: DMSO (4.0 M); TFAA (2.1 M); benzyl alcohol (2.0 M), TEA (5.8 M); flow rates of all pumps
were 5.0 mL ·min-1; conversion determined by HPLC.
Figure 5. Residence-time dependence, in microreactor setups III and IV. Conditions: DMSO (4.0 M); TFAA (2.1 M); benzyl alcohol
(2.0 M), TEA (5.8 M); reaction temperature 10 °C; flow rates 2.0 mL ·min-1 up to 10 mL ·min-1; conversion determined by HPLC.
Table 3. Effect of amine bases on the Moffatt-Swern
oxidationa of benzyl alcohol (4a) at different temperatures in
microreactor setup III
not well understood mechanistically but has been noted by
others as well.8
Effects of Micromixing. The effect of micromixing was
investigated by measuring the temperature dependence in two
different types of mixers in setup III. The results are presented
in Figure 6 showing that there is not an evident difference in
the conversion to benzaldehyde (6a). These experiments clearly
show that for fast exothermic Moffatt-Swern oxidations, it is
not necessary to mix the reagents on microscale. Simple
T-shaped joints, in which mixing took place on milliscale,
worked well and gave, in some cases, even better results. This
outcome is in contrast with the findings of Yoshida et al.,9 who
concluded that the high selectivity at elevated temperatures can
be attributed to both the short residence times and the extremely
fast and efficient mixing. The above results indicate that the
high selectivity is merely an effect of the short residence times,
and is not influenced by the use of a micromixer.
TEA (5.8 M)
DIPEA (5.8 M)
benzaldehyde (6a)
benzaldehyde (6a)
temp (°C)
(mol/mol%)
(mol/mol%)
-20
-10
0
77
76
75
54
25
88
87
87
86
84
10
20
a Conditions: DMSO (4.0 M); TFAA (2.1 M); benzyl alcohol (2.0 M), base (5.8
M). Flow rates of all the pumps were 5.0 mL/min-1. Conversion to benzaldehyde
determined by HPLC.
the reaction rates of steps 1 and 2 in scheme 1 become a limiting
factor in this microreactor.
It is known from literature that, besides triethylamine (TEA),
other tertiary amines could also be used.8 When diisopropyl-
ethylamine (DIPEA) was used in the oxidation of benzyl alcohol
(4a), higher conversions were obtained, and also the amount
of methylthiomethyl ether (8a) decreased (Table 3). The
difference between triethylamine and diisopropylethylamine is
Scope and Limitations. This section describes the results
of the Moffat-Swern oxidation of a variety of alcohols carried
(Figure 7) out in a conventional batch reactor and in the
microreactor with setups III and IV. With both setups, the
916
•
Vol. 12, No. 5, 2008 / Organic Process Research & Development

Figure 6. Effect of a micromixer in comparison with a T-shaped joint, measured in setup III. Data plotted as an average (n ) 3)
and 2× the standard deviation. Conditions: DMSO (4.4 M); TFAA (2.4 M); benzyl alcohol (2.0 M), DIPEA (5.8 M). Flow rates of
all the pumps were 5.0 mL/min-1. Conversion determined by HPLC.
Figure 7. Different types of alcohols used in the Moffatt-Swern oxidation, either in batch or the microreactor.
Table 4. Aldehyde/ketone conversion of the Moffatt-Swern
oxidation of different aromatic alcohols in the microreactor
using setups III and IV and a batch reactor
temperature was varied between -20 and 30 °C at a constant
flow rate of 5 mL ·min-1.
Comparison of the temperature profiles of the Moffat-Swern
oxidation of alcohols 4b-4f in batch reactions with those in
the microreactor with setups III and IV (Table 4) reveals the
same general trends as observed for benzyl alcohol (4a). In the
batch reactor, the conversion of the aldehydes or ketones sharply
decreases above -45 °C, while in the microreactor the
conversions remain the same for alcohols 4b, 4c, 4e, 4f, 4g,
4h, and 4i at temperatures up to 10 °C. Setup IV allows for
even higher reaction temperatures than setup III. However, in
setup III the conversion at temperatures below 10 °C is for all
alcohols higher than for setup IV. By using the in situ generation
(setup IV) of trifluoroacetoxydimethylsulfonium salt (3), con-
versions up 93 mol/mol % can be obtained evan at 30 °C, viz.
1-phenyl-2-propanol (4e), phenylethylalcohol (4f), and test-
osterone (4 h).
The remarkably high conversions obtained for the nonben-
zylic alcohols (4e-4i) in contrast to the benzylic alcohols (4a,
4b, 4d), can be explained in terms of the lower susceptibility
of the alkoxysulfonium salts (5) to solvolytic attack by TFA.7,8
Oxidation of the alcohols 4g, 4 h, and 4i in batch experiments
showed that addition of DIPEA at 20 °C results in a similar
conversion when DIPEA was added at -70 °C, indicating that
the alkoxysulfonium salts (5g-5i) were stable at 20 °C. For
conversion to aldehyde or ketone
(mol/mol %)
temp (°C) 6ba,b 6ca,b 6da,b 6ea,b 6fa,b 6gb,c 6hc,d 6ic,d
setup III
setup IV
batch
-20
-10
0
83 62 54 96 97 97 99 91
82 61 53 96 97 97 98 92
82 60 47 94 97 96 98 92
80 61 36 94 97 94 98 79
76 61 20 89 93 86 18 1.7
62 50 10 74 77 71 6.4 n.d.
65 59 46 93 94 94 n.d n.d
67 58 51 93 95 94 n.d n.d
68 57 49 93 95 93 n.d n.d
67 56 40 94 95 92 n.d n.d
67 55 26 94 95 91 n.d n.d
64 55 12 93 94 91 n.d n.d
88 86 98 95 98 98 99 95
87 78 24 89 95 90 n.d. 92
10
20
30
-20
-10
0
10
20
30
-70
-45
-30
-10
10
20 41
0.5 32 43 45 60 32
0
0
0
4.1 0.4 4.0 3.3 2.8 n.d. 3.3
0
0
0.4
0.1
0
0
0.8 0.5 n.d. n.d.
0.2 0
20
0
0
a Conditions: DMSO (4.0 M); TFAA (2.1 M); alcohol (2.0 M), DIPEA (5.8 M);
flow rates of all the pumps were 5.0 mL ·min-1 b Conversion determined by
.
HPLC. c Conditions: DMSO (4.4 M); TFAA (2.2 M); alcohol (0.5 M), DIPEA (5.8
M); flow rates of alcohol pump was 8.0 mL ·min-1, flow rates of the other pumps
were 5.0 mL/min-1 d Conversion determined by 1H NMR.
.
Vol. 12, No. 5, 2008 / Organic Process Research & Development
•
917

Table 5. Scaleability of the continuous flow Moffatt-Swern
oxidation for testosterone at 0 °C using setup IIIa
Conclusions
It has been demonstrated that the Moffatt-Swern oxidation
conversionb (mol/mol %)
can be performed in the Ehrfeld microreactor system. The
microreactor process offers significant advantages over the batch
process. First, because of the small reactor volume, the
accumulation of the labile trifluoroacetoxydimethylsulfonium
salt (3) and alkoxydimethylsulfonium salt (5) is avoided, which
increases the safety of the Mofatt-Swern oxidation drastically.
Second, because of the short residence times (millisecond),
which can be applied in the continuous flow microreactor, the
exothermic Pummerer rearrangement of these unstable inter-
mediates is avoided. As a result, the process can be operated at
remarkably higher temperatures, thus avoiding the need for
cryogenic temperatures.
testosterone 4-androstene-3, methylthiomethyl
time (min)
(4g)
17-dione (6g)
ether (8g)
2
1.9
2.1
1.9
0.5
3.3
0.5
2.1
1.2
1.7
0.9
0.9
98
98
98
99
96
99
97
98
98
0.8
99
0.4
0.4
0.4
0.6
0.3
0.6
0.5
0.9
0.5
0.2
0.5
6
15
30
50
60
75
90
average
stdev
after workup
The present study has also shown that optimization of a
microreactor process can be executed in a very efficient manner
by sequential variation of the process parameters followed by
sampling when a steady state has been reached. Clearly,
parameter screening can be performed much faster in continuous
mode than in batch mode. Therefore, the microreactor could
serve as an ideal tool for high throughput experimentation.
However, in order to speed up the process optimization, fast
online analysis tools must be implemented in the microreactor
system.
For this fast exothermic Moffatt-Swern oxidation, the use
of a micromixer instead of T-shaped joints does not enhance
the conversion of benzylalcohol to benzaldehyde. Neither do
the micromixers influence the temperature at which the continu-
ous flow Moffat-Swern reaction can be performed. The use of
T-shaped joints give the same results as obtained with the
micromixers.
a Conditions: DMSO (1.1 M); TFAA (0.6 M); testosterone (0.5 M), DIPEA
(1.45 M); flow rate of all the pumps was 7.5 mL/min-1; reaction temperature
-20 °C. b Conversion determined by HPLC.
both setups benzhydrol (4d) showed the most significant
decrease of product with increasing temperatures probably by
solvolysis of the corresponding alkoxydimethylsulfonium salt
(5d) for which this bis-benzylic species is more susceptible than
the other alcohols.
Scalability and Consistency. Scale-up of a continuous flow
microreactor process involves running the process for a longer
period of time. The scaleability and consistency of the microre-
actor over a longer period of time has been demonstrated using
the more functionalized alcohols 4g, 4 h, and 4i.
Table 5 shows that the microreactor can be operated for over
1.5 h, with a high conversion and consistent product selectivity.
The isolated yield of 4-androstene-3,17-dione (6g) was 95 mol/
mol %, with a production rate of 60 g ·h-1 and a throughput of
23.6 g ·L-1 ·h-1.
It was attempted to increase the throughput by increasing
the reagent concentrations and keeping the testosterone con-
centration constant. In order to use the same equivalents of
reagents, the flow rate of testosterone was increased to 8.0
mL·min-1 and that of the reagents was decreased to 2.0
mL·min-1. By doing this, the total amount of effluent is
decreased. However, at a lower flow rate the residence time of
the intermediates is longer. As a consequence, with these
settings the optimum temperature for the process was -20 °C
instead of 0 °C.
Using these optimized conditions the reaction was run for
40 min again with a highly consistent product quality. The
isolated yield of 4-androstene-3,17-dione (6g) was 93 mol/mol
%, with a production rate of 64 g ·h-1 and a throughput of 117
g·L-1 ·h-1, an improvement with a factor of 5, compared to
the previous experiment.
The generality of the process developed for testosterone, was
also tested in the scaleability experiments for the two carbo-
hydrates (Tables 6, and 7). The isolated yield of ketone 6 h
after 20 min was 81 mol/mol %, with a production rate of 91.2
g·h-1 and a throughput of 350 g ·L-1 ·h-1; for aldehyde 6i after
running the reaction for 25 min these figures were respectively
72 mol/mol %; 79.7 g ·h-1 and 257 g ·L-1 ·h-1. The above
presented result shows that with these two substrates the process
can be operated for more than 20 min without any problems.
The microreactor could also serve as a tool for quick scale-
up of Moffatt-Swern oxidations. This was demonstrated during
the scaleability experiments. Throughput rates for testosterone
at consistent product quality up to 117 g ·L-1 ·h-1 have been
achieved for periods up to 1.5 h.
Experimental Section
General Information. NMR spectra were recorded on a
Bruker advance DPX 400. Chemical shifts are reported in parts
per million (ppm). 1H NMR chemical shifts are referenced to
TMS as internal standard (abbreviation s singlet; d doublet; dd
double doublet; dt double triplet; m multiplet). For the carbo-
hydrates, the conversion was determined by measuring the
integrals of the methoxy group of the starting material and the
products. The conversion of the other reaction mixtures were
determined by HPLC (Gilson system consists of a 305 and a
306 pump, an 811C gradient mixer, an 805 manometeric
module, and a 155 UV/vis dual wave detector) with acetonitrile
and water as eluent. The HPLC system consists of the following
chromatographic conditions: chromolith Performance RP18e
column with 100 mm × 4.6 mm dimensions, flow rate 2.0
mL ·min-1; water/acetonitrile: initial (5/95); 2.0 min (5/95); 7.5
min (95/5); 9.7 min (95/5); 9.8 min (5/95); 15 min (5/95).
Analysis channels 210 and 240 nm. Integration was done on
the 210 nm channel, except that for the steroids the 240 nm
channel was used.
All commercially available chemicals were used as received.
Testosterone and the two carbohydrates were obtained from the
918
•
Vol. 12, No. 5, 2008 / Organic Process Research & Development

Table 6. Scalability of the continuous flow Moffatt-Swern oxidation for carbohydrates (4h and 4i) using setup III
4h, conversion
6h, conversion
4i, conversion
6i, conversion
time (min)
(mol/mol %)a
(mol/mol %)a
time (min)
(mol/mol %)a
(mol/mol %)a
2
8.8
6.8
7.5
7.7
1.0
7.5
91
93
92
92
1.0
92
2.0
1.5
1.5
1.7
1.6
0.1
1.6
98
98
98
98
0.1
98
10
14.0
20
average
stdev
25.0
average
stdev
after workup
after workup
a Conditions: DMSO (4.4 M); TFAA (2.4 M); carbohydrates (0.5 M), DIPEA (5.8 M); flow-rate for DMSO, TFAA, DIPEA was 2.0 mL ·min-1; carbohydrates, 8.0
mL·min-1; reaction temperature primary alcohol (4h) and secondary alcohol (4i) respectively 0 and 10 °C. b Conversion determined via 1H NMR.
Table 7. Different microreactor setups used
was checked by a balance (obtained from Satorius BP 1200)
using the density of the stock solutions. No feedback loop was
used to control and adjust the flow rates. The cryovessel is a
homemade design and was made in-house. It consists of double-
jacketed vessel and a Teflon lid with two holes for cooling,
four inlet holes for the reagents, one outlet hole, and one hole
for the Pt-thermometer. The cryo-vessel is cooled or heated by
a Julabo LH85. The cryovessel contains the three stand-alone
LH-2 mixers (Ehrfeld mikrotechnik BTS (GmbH6). These
mixers are based on the principle of multilamination, successive
splitting, and rearrangement. Streams of two different fluids
are fanned out in a large number of small streams which are
arranged alternately in an interdigital configuration. The LH-2
mixer consists of a mixing plate with 2 × 10 slides and a slide
diameter of 50/50 µm and an aperture plate with a slide diameter
of 50 µm and a length of 2 mm.
internal
volume
(µL)
residence
setup I length (cm) diameter (mm)
timea (ms)
R1
R2
R3
10
10
20
1.27
1.27
1.27
127
127
253
760
507
760
setup II
0.254
1.27
R1
R2
R3
4
10
20
2
127
253
12
507
760
1.27
setup III
0.254
0.24
R1
R2
R3
4
10
20
2
5.1
253
12
20
760
1.27
setup IV
0.254
0.24
R1
R2
R3
4
10
20
2
5.1
253
12
20
760
1.27
All reagents were precooled in a tube with a volume of 1
mL prior to feeding into the microreactor.
a The residence time in the different reactors, calculated with a flow rate of all
the pumps fixed to 5.0 mL ·min-1
.
Batch Reactions. To a cooled solution of DMSO (4.0 M)
in dichloromethane (2.0 mL) was added slowly a solution of
TFAA (2.4 M) in dichloromethane (2.0 mL) under magnetic
stirring. After the addition was completed, the resulting mixture
was stirred for 10 min at -70, -45, -30, -10, 10, or 20 °C.
Then a solution of alcohol (2.0 M) in dichloromethane (2.0 mL)
was added. After the addition was completed, the mixture was
stirred at the same temperature for 10 min, and TEA or DIPEA
(5.8 M) in dichloromethane (2.0 mL) was added. After stirring
for 10 min, the temperature was allowed to warm to room
temperature. Analyses: 100 µL of the reaction mixture was
quenched in a mixture of THF/TEA (1 mL) and analyzed by
in-house stock supply and used without purification. The reagent
solutions were made accurately by weighing the reagent into
preweighed volumetric flasks. After dilution with dichlo-
romethane, the flasks were again weighed, so that the density
of the stock solution was known. The density is used for the
conversion of the flow rates from gr ·min-1 measured by the
balances to mL ·min-1.
Microreactor. A schematic drawing of the microreactor
setup was given in Figure 1. The reagents were fed into the
microreactor with K 501 HPLC pumps purchased from Knauer.
The pump chambers (10 mL) were made of ceramics; all the
seals used were made of FFKM or Teflon. In order for the
pumps to work accurately, back-pressure modules (obtained
form Alltech) of 33 bar were placed behind the pump outlets.
It was not possible to measure the pressure within the microre-
actor system. The pressure measured over the total microreactor
setup was achieved with the pressure sensors that were
integrated in the pumps. Normally the pressure was 35-38 bar.
This corresponds to a pressure inside the microreactor system
of 2-5 bar. The pumps were also equipped with one-way valves
(obtained from Swagelok) to ensure that no reaction products
could get back into the pump chambers. The pump of TFAA
was equipped with a cooled pump head (obtained from Knauer)
because of the low boiling point of the TFAA solution (32 °C).
The pump head was cooled with a Huber polystat CC3 at
5 °C. In order to protect the pump against clogging, the reagents
were filtered trough 10 µm HPLC stainless steel filters (obtained
from Alltech). The flow rate (0.1-10 mL ·min-1) of every pump
1
HPLC. For the carbohydrates the analyses were done via H
NMR after concentration of the reaction mixtures in vacuum.
Continuous Flow Microreactor. After the temperature of
the cooling bath had reached its set-point, the pumps were
switched on, and after a steady state was reached (typically this
takes ∼20 mL of effluent), three 10-mL samples of each effluent
were taken, of which 100 µL was quenched in a mixture of
THF/TEA (1 mL) and analyzed off-line by HPLC. For the
carbohydrates the analyses were performed using 1H NMR after
concentration of 10 mL effluent in vacuum.
Scalability and Consistency Experiments. After the temper-
ature of the cooling bath had reached its set-point, the pumps
were switched on. At regular time intervals a sample was
analyzed. Workup procedure: testosterone (4g): In total the
reaction was run for 90 min. The effluent (∼2600 mL) was
washed with water (2 × 2 L) and dichloromethane. The organic
phase was removed in vacuo. Toluene was added (2 L), and
Vol. 12, No. 5, 2008 / Organic Process Research & Development
•
919

the reaction mixture was washed with 1 M HCl (2 × 2 L) and
water (2 × 2 L). The water phases were extracted with toluene
(150 mL). The combined toluene phases were dried on MgSO4
and concentrated in vacuum to yield 92.1 g of 4-androstene-
3,17-dione (6g) as a white solid.
worked up identically as was described for carbohydrate (4 h).
This gave 6i in a yield of 30.4 g (81 mol/mol %) as colorless
oil.
1H NMR data (measured at 399.87 MHZ in CDCl3 relative
to TMS): δ (ppm): 3.38 (s, 3H, OMe); 3.50 (1H, dd, 2Hꢀ),
3.58 (1H, dd, 4Hꢀ), 4.09 (1H, t, 3HR), 4.18 (1H, d, 5HR),
4.61(1H, m,1Hꢀ), 4.63 (2H, d, benzyl CH2), 4.78 to 5.0 (4H,
m, 2× benzyl CH2), 7.31 to 7.65 (12H, m. benzyl), 9.67 (s,
1H, CHO).
1H NMR data (CDCl3): δ (ppm): 0.94 (s, 3H, 18-CH3), 1.00
(dt, 1H, 9R), 1,13 (m. 1H, 7R), 1,22 (s, 3H, 19-CH3), 1,30 (m,
2H, 12R, 14 R), 1,48 (m, 1H, H11), 1,58 (m, 1H, 15ꢀ), 1,70
(m, 3H, 1R, 8ꢀ, H11), 1.90 (m, 1H, 12ꢀ), 2,00 (m, 3H, 1ꢀ, 7ꢀ,
15R), 2,11 (t, 1H, H16), 2,35 (m, 1H, 6R), 2,38 (m, 1H, 2R),
2,45 (m, 2H, 2ꢀ, 6ꢀ), 2,49 (1H, d, H16), 5,75 (s, 1H, H4).
In total 675 mL of the testosterone solution was pumped
into the microreactor, and on this basis, the isolated yield was
95 mol/mol %, with a HPLC purity of 99 mol/mol %.
Workup procedure: carbohydrate (4h): In total the reaction
was run for 25 min. The collected effluent (∼310 mL) was
concentrated under vacuum. To the obtained oil were added
THF (300 mL), MeOH (100 mL), and 0.1 M NaOH (100 mL);
these mixtures were stirred for 30 min at room temperature.
The pH was set to ∼7 with 1 M HCl, solution and the volatiles
were removed in vacuo. The aqueous layer was extracted with
dichloromethane (2 × 150 mL). The dichloromethane layer was
washed with sat. NaCl (2 × 100 mL) and dried on MgSO4.
The solvent was removed under vacuum to yield 33.2 g (72
mol/mol %) of 6 h as a yellowish oil.
Acknowledgment
We thank Dr. Frans Kaspersen for his valuable comments
in preparation of the manuscript. Prof. Floris Rutjes and Pieter
Nieuwland from the Institute for Molecules and Materials,
Radboud University of Nijmegen, The Netherlands, are kindly
acknowledged for the pleasant collaboration on the Swern
oxidation in the microreactor developed at the University.
Note Added after ASAP: The paragraph which states
that Supporting Information is available was missing from
the version published April 30, 2008. This paragraph is
present in the printed version as well as the Internet version
published September 19, 2008.
Supporting Information Available
A table with all data points from which the conversion errors
can be calculated for optimization of the microreactor. This
material is available free of charge via the Internet at
http://pubs.acs.org.
1H NMR data (CDCl3): δ (ppm): 3.48 (s, 3H, OMe); 3.68
(1H, dd, H6); 3.80 (1H, dd, 2Hꢀ), 3.92 (1H, dd, H6′), 4.29
(1H, q, 5HR), 4.43 (1H, d, 3HR), 4.59 (2H, q, benzyl CH2),
4.70 to 4.92 (4H, m, 2× benzyl CH2), 4.80 (1H, d, 1Hꢀ), 7.12
to 7.45 (12H, m, benzyl).
Workup procedure: carbohydrate (4i): In total the reaction
was run for 20 min. The collected effluent (∼260 mL) was
Received for review October 11, 2007.
OP700228E
920
•
Vol. 12, No. 5, 2008 / Organic Process Research & Development