Scaling-out pharmaceutical reactions in an automated stop-flow microwave reactor

Organic Process Research & Development 2008, 12, 967–981

Scaling-Out Pharmaceutical Reactions in an Automated Stop-Flow Microwave

Reactor

Jonathan D. Moseley* and Emily K. Woodman

AstraZeneca, Process R&D, AVlon Works, SeVern Road, Hallen, Bristol, BS10 7ZE, U.K.

Abstract:

However, the scale-up of microwave reactions in organic

synthesis has not been easy for a number of reasons, and so far

only a limited scale-up has been achieved.⁵The various

approaches to scale-up have followed a number of strategies.

Successful synthesis in small-scale sealed tubes (2-10 mL),

as exempliﬁed by the Biotage Initiator and CEM Discover

microwave reactors, has led to synthesis in larger sealed tubes

(20-35 mL), and thence to larger sealed vessels, for example

in the Milestone MicroSYNTH. The problem with these larger

vessels (100-3000 mL) is that a superheated solvent generates

considerable pressure at the temperatures typically required. To

contain these pressures, the vessel must be strong whilst being

microwave transparent. The materials typically chosen (ceramic,

glass, PTFE) are all thermal insulators, with the result that whilst

the heating time is fast, the cooling time is often poor, and

becomes worse as the vessel size grows. A way around this is

to use multiple small vessels in parallel, as for example in the

Anton Paar Synthos 3000, which achieves a larger volume,

whilst keeping the pressure in any individual vessel within

acceptable limits. However, for larger-scale synthesis, it would

be tedious to load multiple vessels with multiple reagents and

solvents, and so this approach also has a limited effective scale-

up capability.

Six pharmaceutically relevant reactions were assessed for scale-

out in an automated stop-ﬂow microwave reactor. Daily through-

puts of between 50 and 250 g were achieved at typical reaction

concentrations; for more concentrated reactions, or with 24 h

processing, productivity of 0.5-1.5 kg per day was achievable. This

study conﬁrms that the stop-ﬂow approach in combination with

rapid microwave heating can be equivalent to conventional

continuous ﬂow technology with comparable productivities.

Introduction

Microwave-assisted organic synthesis (MAOS)^1,2has be-

come hugely successful in the past few years in both academia

and especially industry, where it is now routinely used for initial

drug discovery synthesis.^1bMicrowave heating provides many

well-known beneﬁts at small scale, most of which are also

advantageous at larger scale. For example, increased reaction

rates, improved impurity proﬁles, reduced catalyst loadings,

access to superheated solvents (possibly leading to novel

chemistries such as near critical water chemistry for example),³

and possible energy savings,⁴are all as highly desirable at large

scale as at small scale.

An open vessel system using conventional laboratory

glassware as in the CEM MARS avoids this problem, and some

impressive synthetic reactions have been performed in this

apparatus on 5 L scale.⁶However, it is limited to the boiling

point of the solvent used for the reaction, which may limit the

chemistry compared to the original microwave tube-scale

reaction. And a further limiting factor for all open or closed

vessel systems is the penetration depth of microwaves,⁷which

is only a few cm in most solvents at 2450 MHz.⁸Therefore,

even the open vessel system cannot be scaled up beyond about

5 L. Consequently many have concluded that for effective scale-

up, a continuous ﬂow (CF) system will have to provide the

answer. Many prototypical or modiﬁed commercial CF micro-

wave reactors have been reported,⁹starting with the pioneering

example of Strauss in the mid 1990s,¹⁰and at least one viable

commercial instrument (the Milestone FlowSYNTH) is now

available. However, most if not all of these microwave reactors

suffer the same general limitation of CF reactors in requiring a

homogeneous solution both to enter and exit the reactor vessel.

* Author for correspondence. E-mail: jonathan.moseley@astrazeneca.com.

(1) (a) Loupy, A., Ed.; MicrowaVes in Organic Synthesis; Wiley-VCH:

Weinheim, 2006; (b) Kappe, O.; Stadler, A. MicrowaVes in Organic

and Medicinal Chemistry; Wiley-VCH: Weinheim, 2005. (c) Tierney,

J. P., Lidstro¨m, P., Eds.; MicrowaVe Assisted Organic Synthesis;

Blackwell: Oxford, 2005; (d) Hayes, B. L. MicrowaVe Synthesis:

Chemistry at the Speed of Light; CEM Publishing: Matthews, NC,

2002.

(2) (a) Kappe, C. O. Angew. Chem., Int. Ed. 2004, 43, 6250–6284. (b)

Hayes, B. L. Aldrichimica Acta 2004, 37, 66–76. (c) Nu¨chter, M.;

Ondruschka, B.; Bonrath, W.; Gum, A. Green Chem. 2004, 6, 128–

141. (d) Lidstro¨m, P.; Tierney, J.; Wathey, B.; Westman, J. Tetrahe-

dron 2001, 57, 9225–9283. (e) Perreux, L.; Loupy, A. Tetrahedron

2001, 57, 9199–9223.

(3) (a) Kremsner, J. M.; Kappe, C. O. Eur. J. Org. Chem. 2005, 3672–

3679. (b) Dallinger, D.; Kappe, C. O. Chem. ReV. 2007, 107, 2563–

2591.

(4) (a) Gronnow, M. J.; White, R. J.; Clark, J.; Macquarrie, D. Org.

Process Res. DeV. 2005, 9, 516–518 (Addition and Correction 2007,

11, 293). (b) Barnard, T. M.; Leadbeater, N. E.; Boucher, M. B.;

Stencel, L. M.; Wilhite, B. A. Energy Fuels 2007, 21, 1777–1781. (c)

Razzaq, T.; Kappe, C. O. Chem. Sus. Chem. 2008, 1, 123–132.

(5) For a more thorough treatment of this topic, including an overview of

current commercial microwave reactors, and a future perspective, see:

(a) Kremsner, J. M.; Stadler, A.; Kappe, C. O. Top. Curr. Chem. 2006,

266, 233–278, Our own recent report presents results for a single

reaction in all these commercially available microwave reactors: (b)

Moseley, J. D.; Lenden, P.; Lockwood, M.; Ruda, K.; Sherlock, J.-P.;

Thomson, A. D.; Gilday, J. P. Org. Process Res. DeV. 2008, 12, 31–

40; and the contemporaneous report from the Leadbeater group

discusses a number of reactions in several of these reactors: (c)

Bowman, M. D.; Holcomb, J. L.; Kormos, C. M.; Leadbeater, N. E.;

Williams, V. A. Org. Process Res. DeV. 2008, 12, 41–57.

(6) (a) Barnard, T. M.; Vanier, G. S.; Collins, M. J. Org. Process Res.

DeV. 2006, 10, 1233–1237. (b) Leadbeater, N. E.; Williams, V. A.;

Barnard, T. M.; Collins, M. J. Org. Process Res. DeV. 2006, 10, 833–

837.

(7) Gabriel, C.; Gabriel, S.; Grant, E. H.; Halstead, B. S. J.; Mingos,

D. M. P. Chem. Soc. ReV. 1998, 27, 213–223.

(8) Horikoshi, S.; Iida, S.; Kajitani, M.; Sato, S.; Serpone, N. Org. Process

Res. DeV. 2008, 12, 257–263.

10.1021/op800091p CCC: $40.75

Published on Web 08/23/2008

2008 American Chemical Society

Vol. 12, No. 5, 2008 / Organic Process Research & Development

•

967

Even so, a further hybrid microwave-CF reactor conﬁgura-

tion is also possible, that of an automated stop-ﬂow microwave

reactor, which exists in one form as the commercially available

CEM Voyager SF. This instrument relies on a small vessel (80

mL) to circumvent the microwave heating issues that limit larger

vessels as discussed above, and automation to facilitate repeti-

tious vessel charging. As with other CF reactors, homogeneous

or near homogeneous solutions are required, and it could be

argued that this system has the disadvantages of CF without

the beneﬁts of conventional batch reactors (i.e., simplicity, scale

and ubiquity). However, automated charging combined with

microwave heating on a small chemical inventory means that

this reactor effectively processes multiple small vessels in series,

rather than in parallel as for other microwave reactors; or by

comparison to conventional CF, it could be considered to

process segments of reaction material passing through a tube,

but in stop-start fashion, rather than in a continuous stream.

By retaining the small vessel size, the advantages of

microwave heating are easier to realise, such as for example

the overall rapid processing of batches, rapid heating and cooling

proﬁle, and lower mechanical pressures due to superheated

solvents. The combination with automated charging makes

processing many small batches practicable, where it would not

be so in other cases. In common with CF generally, the workup

procedure must be considered separately, but here also there

are several minor advantages; batches can be combined to

provide workup batches of whatever convenient size required

(to ﬁt vessels or for GMP considerations), and occasional out-

of-speciﬁcation batches result in a trivial loss of output

compared to the possible impact on small batch numbers from

large batch reactor manufacture.

We present here our results using the CEM Voyager in stop-

ﬂow mode for a number of pharmaceutically relevant reactions.

These reactions were chosen because they are typical of those

used within the pharmaceutical industry, and because they

would beneﬁt from microwave heating to increase the reaction

rate. They were also chosen because they all provided a

homogeneous solution which could easily be processed in the

Voyager. The Voyager is capable of processing some hetero-

geneous reaction mixtures, although the results have been

mixed.¹¹On the other hand, we wanted to test the instrument

on typical (albeit homogeneous) pharmaceutical reactions, and

at typical concentrations, by running many multiple batches; it

is only by running in CF mode for an extended period that the

true capability of the technology can be proven. In this case,

we were not disappointed.

Figure 1. CEM Voyager.

Results and Discussion

Description of the CEM Voyager. The CEM Voyager is

essentially a pump unit consisting of a peristaltic pump, two

valves and associated lines, which sits on top of the versatile

300 W Discover base unit.¹²The two units have integrated

software control and are connected by a single cable. The

Discover unit can be used separately from the Voyager unit,

and it takes only a couple minutes to switch between the two

functionalities. The combined Voyager unit will ﬁt comfortably

in a standard depth fume cupboard and occupies no larger

footprint than the Discover itself, except for the need for feed

and receiver vessels, and so occupies no more than half the

width of a standard fume cupboard in total (Figure 1). With a

small footprint, multiple units can be accommodated easily if

larger quantities are required (a scale-out option). Furthermore,

once the reaction conditions are optimised, the automation

reduces the interaction required by the chemist. A signiﬁcant

drawback is the need to have homogeneous or nearly homo-

geneous reaction solutions. Although the instrument uses a

peristaltic pump which can in principle transfer slurries, and is

ﬁtted with an anticlog device, in practice the lines (1.5 mm

i.d.), valves and pump are prone to blocking with slurries, as

noted by Lehmann.^11cHowever, it has been used successfully

with ﬁne slurries as reported by others.^11a,bAlthough continuous

addition and sampling whilst heating are not formally possible,

the multiple inlet and outlet lines could be used to mimic these

techniques if required, although with the potential to isolate

many small batches, this is probably of less interest. Finally,

the small reaction mass in a glass-walled reactor heats and cools

quickly and so the cycle time per batch is fast. Coupled with

genuine automation for rapid charging and discharging, this

small instrument has a lot to offer.

(9) (a) Bremner, W. S.; Organ, M. G. J. Comb. Chem. 2007, 9, 14–16.

(b) Baxendale, I. R.; Pitts, M. R. Chem. Today 2006, 24 (3), 41–45.

(c) Schwalbe, T.; Simons, K. Chem. Today 2006, 24 (2), 56–61. (d)

Jachuck, R. J. J.; Selvaraj, D. K.; Varma, R. S. Green Chem. 2006, 8,

29–33. (e) Bagley, M. C.; Jenkins, R. L.; Lubinu, M. C.; Mason, C.;

Wood, R. J. Org. Chem. 2005, 70, 7003–7006. (f) Comer, E.; Organ,

M. G. J. Am. Chem. Soc. 2005, 127, 8160–8167. (g) Wilson, N. S.;

Sarko, C. R.; Roth, G. P. Org. Process Res. DeV. 2004, 8, 535–538.

(h) Shieh, W.-C.; Lozanov, M.; Repic, O. Tetrahedron Lett. 2003,

44, 6943–6945. (i) Shieh, W.-C.; Dell, S; Repic, O. Tetrahedron Lett.

2002, 43, 5607–5609. (j) Khadilkar, B. M.; Madyar, V. R. Org.

Process Res. DeV. 2001, 5, 452–455. (k) Marquie, J.; Salmoria, G.;

Poux, M.; Laporterie, A.; Dubac, J.; Roques, N. Ind. Eng. Chem. Res.

2001, 40, 4485–4490. (l) Kabza, K. G.; Chapados, B. R.; Gestwicki,

J. E.; McGrath, J. L. J. Org. Chem. 2000, 65, 1210–1214.

In fuller detail, the Voyager SF has an 80 mL Pyrex glass

vessel with 50 mL operating capacity and a ﬁbre optic probe

to measure the temperature, sited in a sapphire thermowell

(10) (a) Cablewski, T.; Faux, A. F.; Strauss, C. R. J. Org. Chem. 1994,

59, 3408–3412. (b) Roberts, B. A.; Strauss, C. R. Acc. Chem. Res.

2005, 38, 653–661.

(11) (a) Pitts, M. R.; McCormack, P.; Whittall, J. Tetrahedron 2006, 62,

4705–4708. (b) Loones, K. T. J.; Maes, B. U. W.; Rombouts, G.;

Hostyn, S.; Diels, G. Tetrahedron 2005, 61, 10338–10348. (c)

Lehmann, H.; LaVecchia, L. JALA 2005, 10, 412–417.

(12) Ferguson, J. D. Mol. DiVersity 2003, 7, 281–286.

968

•

Vol. 12, No. 5, 2008 / Organic Process Research & Development

note the calibration constants, as different solvents and reaction

mixtures have markedly different viscosities. However, the

calibration constants stay with each add step and do not need

to be redone even if switching regularly between solvents. The

add step also blows the lines clear of reagent, and two solvent

charges can be added with further air blows to further ensure

the lines are clear for the next reagent or product. Finally, it is

worth noting that the pump speed can be adjusted down from

100% to a minimum of 20%, if particularly ﬁne control for a

small charge is required; or the charges can be controlled on

time alone if preferred.

The microwave step operates as expected, heating to a set

point temperature in a speciﬁed time, and then holding for a

speciﬁed time, before cooling to a set temperature. All the usual

parameters can be altered, including heat-up and hold times,

adjusting the wattage or using ﬁxed power, heating-with-

cooling, and designing multistage heat-up procedures. The

remove step simply empties the vessel through the appropriate

outlet line, and is best run on ﬁxed time, although there is also

a solvent sensor to detect when the vessel is empty. The clean

steps are designed to rinse the vessel and the lines, either to

ensure all product has been removed to the receiver vessel, or

in preparation for the next batch. They are basically simpliﬁed

add and remove steps combined in one operational step. Vessel

rinses can also be built into a method by using add and remove

steps separately, but we found in general that it is a more

efﬁcient to use the functionality built into the instrument

software.

Preparing the Chemistry. Chemistry for trial in the

Voyager should ﬁrst be assessed in a small-scale scientiﬁc

microwave reactor; indeed the chemistry may well have come

from a microwave reaction. If it is from conventional heating,

a tube-scale reaction should be tried. The trial should obviously

be at the same concentration as the planned scale-out reaction.

This will give an idea of the operating parameters required for

successful reaction (i.e., time and temperature) and also an early

indication of any potential hazards (e.g., unexpected pressure

build-up). However, this is unlikely as the maximum scale is

only 50 mL, so few reactions should be problematic. Even so,

stepwise scale-up (2 mL to 20 mL to 50 mL) when superheating

reaction mixtures is still wise (and recommended by the

manufacturers). A DSC test or Carius tube experiment will

provide sufﬁcient data in cases of concern. Furthermore, it is a

simple calculation to determine the pressure generated from a

particular solvent at a set temperature with a given vessel ﬁll

volume.

Figure 2. Voyager head unit, reaction vessel and ﬁbre optic

probe in thermowell.

(Figure 2). (There is also a true continuous ﬂow model (Voyager

CF), which has a 5 mL capacity coil in place of the 80 mL

glass vessel, and requires a different pump unit, but this has

been little used in comparison). Magnetic stirring provides the

agitation, and cooling is provided by compressed air which cools

the vessel and contents at the end of the heating cycle. The

vessel can also be cooled whilst heated if desired. The Voyager

will operate across the standard temperature and pressure range

for most scientiﬁc microwaves (i.e., up to ∼200 °C and 20 bar).

The vessel is charged through three inlet lines (made of

PTFE or stainless steel), two for starting material solutions (SM1

and SM2) and one for solvent; however, this could equally serve

as an SM3 inlet line also. And there are three outlet lines for

product, waste and air. The air line can also be used as an inlet

line for other gases, either for inertion or for other reaction gases.

There is also an overﬂow line which simply operates as a drain

if the vessel is overﬁlled. The lines can be blown clear with air

from valve 1 backwards, but because there is only one pump,

it is not possible to ﬂush the lines beyond this valve with solvent

without ﬁrst occupying the vessel and then blowing back (see

Figure 3). The line connecting valve 1, the pump, valve 2 and

the vessel is therefore common to all, which has to reach to

the bottom of the vessel if it is to discharge the contents

efﬁciently; the air line and overﬂow reach only to the top. The

Voyager therefore in essence provides automated charging and

discharging of the reaction vessel in a stop-ﬂow mode, so it

can be programmed to charge, heat, cool and discharge

repeatedly, for as many batches as are required. The total

volume to be processed is entered into the interface, and the

software calculates the number of batches required. Some idea

of the cycle time is a good idea however, if one wants to know

when it is likely to ﬁnish. The cycle can also be paused, halted

or canceled.

Once standard parameters for a tube-scale reaction have been

determined, and chemical and physical hazard issues addressed

if necessary, trials in the Voyager can begin. It is not worth

over-optimising the small-scale microwave reaction conditions

in our experience, as they will vary slightly when transferred

to the Voyager. This is because the vessel has a slightly different

geometry from a microwave tube, the stirring will be slightly

different, and the Voyager has the same size magnetron (300

W) as the smaller reactors to heat a much larger reaction mass.

As with other CF apparatus, considerable starting material is

needed for trials, depending on the concentrations to be run.

Presumably, this will be available if the aim is to process large

A sequence of batches uses a method which is made up of

individual steps (add, microwave, remove or clean), and each

of these steps has additional substeps. The add step adds a ﬁxed

volume from any one of the input lines into the reaction vessel.

It is wise to calibrate the pump ﬁrst for this procedure, and to

Vol. 12, No. 5, 2008 / Organic Process Research & Development

•

969

Figure 3. Schematic overlay of Voyager pumping operations for “add” (top) and “clean” (bottom) steps.

Scheme 1

quantities of material, and for many chemists, the initial product

from unoptimised batches will still be very useful for ongoing

studies. It also makes sense to ﬁll the vessel to capacity (50

mL) unless there is good reason to do otherwise, as this will

maximise the through-put. The Voyager will actually allow 60

mL into the vessel, so there is some scope for increasing

is required, other than to keep the instrument running, and of

productivity, although stirring and cooling might be less

efﬁcient. There is obviously a pay-off in terms of cycle time

for heating and cooling a larger batch (which takes slightly

longer) than in heating and cooling a smaller batch. Other factors

may determine what is the most efﬁcient protocol overall. What

the Voyager will not allow is more than 60 mL in the vessel if

it thinks this volume has been exceeded; this can happen if the

pump calibration volume deviates signiﬁcantly from the actual

input volume. Whilst in principle it is not essential to have these

ﬁgures close together, in practice it is more helpful to have the

calibration volume fairly close to the actual volume.

Overall, preparing the chemistry in practice from microwave

tube reactions to a suitable Voyager method can take typically

between 0.5 and 3 days, depending on familiarity with the

chemistry, the complexity of the method and problems encoun-

tered. Once the method is developed however, little further work

course to isolate the product.

ortho Claisen Rearrangement and Benzofuran Forma-

tion. The Claisen rearrangement is a valuable and venerable

reaction, ﬁrst reported in 1912, and has most recently been

reviewed by Martin-Castro.¹³We chose to focus on the ortho-

Claisen rearrangement and subsequent ring closure to the

benzofuran (Scheme 1) since this aligned with a recent

AstraZeneca project.

The readily prepared aryl-O-allylethers 1 were conveniently

used for scouting trials in small microwave tubes. Of these, we

chose to concentrate on the naphthyl compound 4 since

considerable quantities of this material were available from

investigating the preceding O-allylation reaction; and also

because it is converted to its Claisen product 5 under more facile

(13) Castro, A. M. M. Chem. ReV. 2004, 104, 2939–3002.

970

•

Vol. 12, No. 5, 2008 / Organic Process Research & Development

Scheme 2

number for a trial; the ﬁrst batch always behaves slightly

differently due to starting from cold or if the instrument has

been left idle for a period; the second batch should give a

reliable indication of performance; and the third batch provides

some measure of reproducibility. Once the conditions have been

ﬁne-tuned on the Voyager, a ﬁve-batch trial provides further

conﬁrmation just before starting a large cycle. In this case

however, the three-batch trial failed because the 300 W

magnetron, which had been more than adequate for a tube-

scale reaction, could not heat 25 mL of this very low polarity

reaction mixture beyond about 140 °C. Xylene was also

considered, but we have found 1,2-dichlorobenzene (DCB) to

be a good substitute for aromatic solvents such as toluene and

xylene in microwave reactions, aided by its higher bp (180 °C)

and dielectric constant (9.9).^1dThis necessitated another quick

check on tube scale, which heated considerably faster than the

comparable toluene solution, conﬁrming our choice of DCB.

Heating trials for the ﬁrst stage were now successful in the

Voyager, although the magnetron had to work hard to hold the

batch at 200 °C. By dropping the temperature back to 195 °C,

and heating for 2 extra minutes, the magnetron reached the set-

point temperature more quickly and held it at 195 °C more

comfortably on relatively low power. In fact, the reaction

mixture tended to overshoot the set-point temperature up to

∼200 °C once it reached ∼190 °C, which we assumed was

the heat of reaction. The volume of DCB used was also reduced

to half with respect to 4, which may also have aided the heating

time.

With the ﬁrst step ﬁxed, we attempted to ﬁne-tune the

second. The crude Claisen product 5 was cooled to 120 °C and

20 mL of formic acid charged through the SM2 line, followed

by a 3 mL line wash with fresh DCB through the solvent line.

This mixture was reheated to 150 °C for 5 min, and also tended

to overshoot the set-point considerably, up to 165 °C. This

caused a problem because enough formic acid had decomposed

to generate residual pressure greater than the normal 50 psi limit

for venting the vessel, although this limit can be adjusted

upwards. An alternative was to limit the microwave power

available in this step to 200 W, which then kept the temperature

somewhat closer to the desired set-point, and resulted in less

gaseous formic acid decomposition products. However, longer

heating was still required to get the pressure down to an

acceptable level, even when the temperature was already well

below 90 °C. The batch was removed from the vessel through

the product line, where it immediately became biphasic. Two

vessel rinses with fresh DCB of ∼10 mL each were used to

clean the vessel of residual formic acid through the waste line

in preparation for the next batch. The overall cycle time was

37 min for 10 g of starting material 4 (50 mmol) with ∼90%

conversion to naphthofuran 6.

conditions more readily within the scope of the Voyager

(Scheme 2). Typical conditions for the ortho-Claisen rearrange-

ment require several hours in high-boiling solvents such as DMF

or diethylaniline, or heated neat without solvent. Although we

tried a range of classic and nonclassic solvents for this

rearrangement, we found that conversions for all substrates were

cleanest when conducted neat. Compound 4 was also a light

oil, so could be ﬂowed into the Voyager. This reaction therefore

provided an example of a high-temperature, single-phase

reaction. The small-scale tests on 4 were conducted on 1.0 g

(4 mmol) scale in each tube, but because the reaction was neat,

we conducted a DSC test to determine that there were no

thermal stability issues before scaling up in the Voyager.

Initially we attempted to combine the Claisen rearrangement

and the acid-catalysed ring-closure step into a two-step sequence

to convert 4 into 6 (Scheme 2). Scouting studies in a tube-

scale microwave reactor indicated that the Claisen reaction

required heating at 200 °C for 10 min, followed by ∼140-150

°C for 5 min for the ring closure, depending on the quantity of

formic acid used. Conversions were good with only ∼2-3% 4

left and 5% 1-naphthol (7) formed (presumably by acid-

catalysed hydrolysis of unreacted 4); this could be washed out

of the product with dilute alkali solution. To simplify the

process, an all-in reaction with a roughly equal volume of formic

acid was attempted with an extended heating time, to convert

4 directly to 6. However, the level of 7 regenerated was now

13%, with an unknown impurity present at 10% (by HPLC

area), and although all starting material 4 was otherwise

converted to product 6, an overall conversion and purity of

∼75% was judged too low to justify this one-step procedure.

Concurrently, pumping trials in the Voyager had revealed

that although 4 appeared to be a mobile oil, it was viscous

enough to trigger the clog detector in the pump periodically,

causing it to reverse the ﬂow brieﬂy. This was not desirable

for the multiple-batch manufacture we were planning, so the

starting material was diluted with an inert solvent to reduce

the viscosity. This would also allow us to build in small solvent

washes to rinse the line of residual formic acid in-between

batches in the two-step procedure. Model studies in microwave

tubes worked well using between a quarter and one equal

volume of toluene to starting material 4, even though the formic

acid was biphasic with the toluene reaction mixture at RT. We

rationalised that at high temperature the reaction mixture would

be either monophasic or sufﬁciently well stirred to achieve good

mixing.

Unfortunately, analysis of both phases of these batches

indicated incomplete reaction in two out of three cases, even

though the small-scale tube reactions had been excellent. Since

there was no starting material 4 left, but some Claisen product

5 without much of the usual 1-naphthol (7), this suggested that

it was the second, ring-closing reaction that was the problem

(since conversion of 4 to 5 was complete, and 5 converts readily

to 6, incomplete reaction could only arise if the second reaction

Studies in the Voyager could now begin. A solution of 4 in

toluene was prepared suitable for a small number of batches

and a three-batch trial was run. This is the minimum realistic

Vol. 12, No. 5, 2008 / Organic Process Research & Development

•

971

Table 1. Comparison of reaction data

cycle time

(min)

temperature

hold time

(min)

conversion

(%)^a

relative

concentration

(g/batch)

through-put^c

g/h

reaction

(°C)

volume^b

Claisen (one step)

Heck #1 (10)

21

8.6

9.5

11

16

27

28

195

140

110

200

210

12

95

0.5

12

10

30

2

21

60

23

28

9.5

8.2

60

18

21

2.0

1.5

2.0

10

93-98^d

99

3.3

4.0

1.5

16

Heck #2 (13)

hydrolysis

hydrolysis + alkylation

NKR #1 (14a)

NKR #2 (14b)

NKR #3 (14c)

99

20

99

5

8.3

10

95

4

^aDetermined by HPLC. ^bRelative to starting material mass. ^cThrough-put of starting material. ^dTotal conversion to products 11 and 12.

Scheme 3

failed in some manner). It was thought likely that the explana-

tion was due to an intermittent mechanical problem with the

stirring on the instrument, which can sometimes be discerned

by the sound. Poor stirring of the probably biphasic reaction

mixture in the larger reaction vessel could account for the

incomplete reaction in some batches. At the time, we decided

to press on with the single-stage (Claisen) reaction to establish

However, it does demonstrate what one small instrument can

achieve for a suitable reaction operating in a type of CF mode,

albeit stop-ﬂow in this case. Furthermore, it also demonstrates

how microwave technology, in combination with a scale-up/-

out option, can deliver useful quantities of product that would

be difﬁcult to achieve on pilot-plant scale due the normal

temperature limits there. A further comparison with other

reactions considered here is given later in Table 2.

Heck Reaction #1. Although not as old as the Claisen

reaction, the Heck reaction has also established itself as an

invaluable synthetic transformation.¹⁴It also served our purposes

in being an example of a second order reaction. The requirement

for moderately high temperatures in combination with a metal

catalyst made it ideal for investigation under microwave heating.

Preliminary investigations reacting 4-bromoacetophenone (8)

with methyl acrylate (9) using Pd(OAc)₂as the catalyst had

established that a homogeneous reaction mixture could be

obtained. This is a standard model substrate for the Heck

reaction, and several microwave scale-up preparations have been

reported using low loading of catalyst (0.1 mol %).¹⁵Our

conditions¹⁶were slightly different from those reported, using

dimethylacetamide (DMA) as the solvent and methyldicyclo-

hexylamine (10) as the base, which has been found to work

well in more demanding cases (Scheme 3).¹⁷

the precedent of a 1 kg-scale manufacture in the Voyager.

However, we have discussed this attempt in some detail because

it illustrates a number of points in the use of this technology.

ortho Claisen Rearrangement. Having decided to fall-back

on the single-step Claisen reaction, virtually no additional

development work was required. The batch size was increased

to 30 mL to improve the productivity, since volume in the

reaction vessel was not needed for the formic acid. The full

capacity was not used since the cycle time increased signiﬁ-

cantly when trying to heat 50 mL to the temperature limit

around 200 °C. As noted, 195 °C for 12 min gave precisely

95% conversion with ∼5% starting material 4 left and a trace

of 1-naphthol. Reducing the heating to 10 min resulted in ∼94%

conversion, which we felt was just too low, whilst heating for

15 min gave only a 1% improvement to 96%, and was not worth

the gain in yield versus the cycle time. Some time was saved

by venting the product from the vessel at 100 °C which

shortened the cooling time, so that overall the planned cycle

time was 21.0 min for this simple Voyager three-step method

(full details of which can be found in the Supporting Informa-

tion). The three-batch proving trial worked exactly as planned,

so two 24-batch cycles using a total of 1006 g of 4 diluted

with 500 mL of DCB were processed over two 8.4 h days.

Batches were combined into groups of six or seven, and HPLC

showed conversions for all fell between 94.8 and 95.1%. Output

volumes were exact multiples of 30 mL (180 or 210 mL),

demonstrating very good control by the pump, and no operator

input was required during the day other than to supply the

receiver vessels.

All components of the reaction mixture could be made up

in a single solution at 10 volumes in DMA (relative to the

acetophenone 8) and stored for some time; there was no

background reaction at RT. The Pd(OAc)₂catalyst was loaded

at 0.1 mol % with tetrabutylammonium bromide (TBAB) at

0.4 mol %, and was completely soluble in DMA at this volume.

As a precaution, reaction mixtures for large sequences were

ﬁltered to remove any residual particles. Small-scale microwave

tube trials showed that the reaction was 95% complete after 5

Total reaction time for 48 batches was 16.8 h, or 21.0 min

per 21 g batch as planned. This averaged 1.000 g min^-1or

60.0 g h^-1comparative to CF mode, a good rate for a reaction

requiring 195 °C for 12 min, which is notably above what is

achievable by most standard pharmaceutical laboratory and pilot

plant equipment. Of the 21.0 min, the typical heating time was

3.5 min and cooling time was 3.3 min; the remaining nonheating

time of 2.2 min was taken up with pumping and valve

switching. Further details are given in Table 1. Overall,

productivity was very high because this reaction is almost neat.

(14) (a) Beletskaya, I.; Cheprakov, A. Chem. ReV. 2000, 100, 3009–3066.

(b) Farina, V. AdV. Synth. Catal. 2004, 346, 1553–1582.

(15) (a) Leadbeater, N. E.; Williams, V. A.; Barnard, T. M.; Collins, M. J.

Synlett 2006, 2953–2958. (b) Arvela, R. K.; Leadbeater, N. E.; Collins,

M. J. Tetrahedron 2005, 61, 9349–9355.

(16) Murray, P. M. Personal communication; manuscript in preparation.

(17) (a) Gurtler, C.; Buchwald, S. Chem. Eur. J. 1999, 5, 3107. (b) Littke,

A. F.; Fu, G. C. J. Am. Chem. Soc. 2001, 123, 6989.

972

•

Vol. 12, No. 5, 2008 / Organic Process Research & Development

min at 130 °C and 100% complete at 140 °C. These conditions

were transferred to the Voyager on 47 mL scale, but the reaction

mixture regularly overshot the 140 °C set point temperature up

to ∼150 °C. A DSC test established this was not due to a

reaction exotherm (and also established there were no other

thermal hazards), but was presumably due to the highly polar

reaction medium coupling well with microwave heating. Cutting

down both the microwave power available and heating time

did not minimise this temperature overshoot. Instead, we

decided to make use of it, and cut the heating time to 1 min to

reduce the cycle time. The heating proﬁle was therefore

effectively 150 °C for 1 min, although only 140 °C for 1 min

was programmed in the method. Conversions were >98%, with

<2% residual starting material; however, part of the conversion

(5-7%) was due to the formation of an impurity identiﬁed by

MS as the bis-coupled adduct 12 (and which is probably over-

reported by HPLC based on UV detection).

Figure 4. Pd-coated reaction vessel and stirrer bar.

min, and the reaction volume to 48 mL. To save wasting

material, and to avoid the slightly lower conversion of the ﬁrst

batch when the instrument is cold, the ﬁrst batch through the

cycle was simply blank solvent. The four following reaction

batches gave good and consistent conversions (97.5%) with

reliable output volumes (53 mL each) and a cycle time now of

9.2 min (due to the increased reaction time). None of the lines

blocked.

A 50-batch sequence was performed in a single day on 2650

mL of ﬁltered reaction mixture, heated to 140 °C for 2 min.

Total run time was 7.2 h, or 8.56 min per batch (the discrepancy

in cycle time is likely to be due to the instrument not being

fully hot during the short proving sequence). Total processed

mass of acetophenone 8 was therefore 163 g. Initial batches

were combined into groups of four, and showed conversions

of 89-94%, slightly low compared to the expected results.

Volumes for combined batches were around 212 mL (i.e., 4 ×

53 mL), and the wash volumes consumed were also exactly as

planned, demonstrating again very good control by the pump.

Further details are given in Tables 1 and 2.

After the ﬁrst 20 batches had been collected in groups of

four, the ﬁnal 30 were collected in one vessel. Surprisingly,

the conversion for this group had dropped to 80% of product

11, with still ∼5% of diadduct 12 but 15% of unreacted

acetophenone 8. However, when the vessel was opened and

examined, it became obvious what had changed; the vessel was

heavily plated with a Pd mirror (Figure 4). The ﬁbre optic

probe’s thermowell was also plated down most of its length

(not shown) and signiﬁcantly the window at the tip was

obscured. Operating above ∼110-120 °C was likely to produce

Pd nanoparticles,¹⁸and residual Pd-black could be seen in the

product reaction mixtures (Figure 5). However, this vessel had

been running batches more or less continually with trial batches

and it was not until more than 50 batches had been performed,

The overall cycle time was 7.5 min with cooling to 80 °C.

Three-batch sequences worked well for the ﬁrst batch, but the

subsequent batches had a tendency to block in the outlet lines

due to residues of the highly crystalline HBr salt of 10 seeding

the second and third batches. To overcome this, a small line

wash with fresh DMA was charged after the product was

removed from the vessel to rinse the outlet lines, and the outlet

temperature was raised to 90 °C. The reaction volume was also

cut down to 9.5 volumes to allow for the line wash. Unfortu-

nately, successful reaction batches were clearly on the limits

of solubility at these temperatures and concentrations, as the

lines blocked solid after the ﬁrst batch and had to be replaced.

Thus, the reaction volume was increased to 12 volumes of

DMA, and both a vessel wash and a line wash were built into

the Voyager method. The vessel wash of 10 mL of fresh DMA

from the solvent line cleared reaction residues from the lines

between valve 1, the pump and the vessel (cf. Figure 3), which

were the most important lines to keep clear, being common to

all pumping operations. Since the wash was relatively large

compared to the residues removed, this was rinsed to the waste

line and discarded. The second smaller line wash (5 mL) also

rewashed this section of line, making sure it was clear, but then

ﬂushed the product outlet line also, sweeping any remaining

reaction residues into the receiver vessel, and ensuring this line

was clear to accept the next batch. In this regard, running a

Voyager sequence was very much like running a pilot plant.

Lastly, an additional 20 s air blow was added to the method to

get the lines as clear as possible. The overall cycle time was

increased to 8.5 min, but the method was now robust, and

calculations predicted 20 g h^-1through-put.

(18) (a) Reetz, M.; Westermann, E. Angew. Chem., Int. Ed. 2000, 39, 165.

(b) de Vries, A. H. M.; Mulders, J. M. C. A.; Mommers, J. H. M.;

Henderickz, H. J. W.; de Vries, J. G. Org. Lett. 2003, 5, 3285–3288.

(c) Reetz, M.; de Vries, J. Chem. Commun. 2004, 1559.

During this work, the conversion appeared to have dropped

off slightly, so for the ﬁnal ﬁne-tuning of the reaction conditions

on a ﬁve-batch sequence, the heating time was increased to 2

Vol. 12, No. 5, 2008 / Organic Process Research & Development

•

973

Figure 5. Heck reaction product solutions. Left to right: (a) unreacted solution, (b) partially reacted solution, (c) fully reacted

solution, (d) partially reacted solution showing Pd nanoparticles before settling. Note: Pd residues have settled out in tubes

(b) and (c).

during the middle of this large-scale sequence, that performance

started to signiﬁcantly degrade. We speculate that the Pd mirror

will couple very well with the microwaves, thus perhaps giving

hot surfaces in contrast to the usual performance with micro-

wave vessels. Of more importance though is the likely increased

temperature reading on the inside of the Pd-coated window by

the ﬁbre optic probe. This will cause the probe to read a higher

temperature than the bulk reaction mixture, which will not then

receive the full microwave heating required for complete

reaction (and which is presumably not compensated by any

additional heating from hot Pd-coated walls).

Heck Reaction #2. The second Heck reaction attempted was

essentially identical, but with Hunig’s base (13) in place of the

much more expensive 10. Aside from being cheaper, especially

for the quantities required for this CF chemistry, its HBr salt

did not crystallise in the Voyager’s lines. This allowed the

Voyager method to be simpliﬁed, since the vessel and line

washes in the previous reaction had not been required for

the chemistry, but were only needed to ensure the lines were

free from blockage. This would save cycle time, but only if 13

was a suitable substitute for 10. Small-scale tube experiments

demonstrated equivalent performance for acetophenone 8 in

combination with acrylate 9; this is not to suggest that more

demanding coupling partners would react so readily, however.

Aside from this issue, the workup and isolation of the product

was relatively straightforward. The reaction liquors were diluted

into double their volume with water which precipitated a yellow

solid. This was isolated by ﬁltration and washed with water to

give the cinnamate product 11 as an off-white, slightly grey

solid (residual Pd levels will be reported separately). Quality

was excellent, and batches with conversions of >90% gave

isolated product 11 with typically 97-98% quality, the balance

being shared by residual acetophenone 8 and diadduct 12. Even

lower quality batches gave generally good quality product on

workup. However, residual amine 10 was seen at 33-50 mol

% in the ¹H NMR spectra. This was removed by reslurrying in

water and washing with 0.5-1.0 equiv of concentrated HCl

for several hours, re-isolating and washing with water, which

completely removed all amine residues. Quality was further

improved to >99%, and the recovery was typically 95%. The

HCl washing procedure was then built directly into the workup

for later batches which simpliﬁed the process, and gave an

overall yield of 80-90% with quality ∼97%.

Other minor changes were also made. The methyl acrylate

charge was reduced to 1.05 equiv in the starting material

solution. The DMA volume was reduced back to 10 volumes

relative to acetophenone 8, and the batch size was increased

from 47 to 50 mL, in both cases to increase productivity. A

small line wash was added, initially of 3 mL but 1 mL was

found to be adequate. This was added to wash traces of residual

starting material solution out of the SM1 line and into the vessel,

so that they did not contaminate the product on the way out.

This translated into a simple three-step Voyager method of add,

microwave and remove (full details in the Supporting Informa-

tion), with a cycle time of 8.6 min. Proving trials on a three-

974

•

Vol. 12, No. 5, 2008 / Organic Process Research & Development

batch sequence then worked as planned, with the 1.0 mL line

wash proving to be reliable and accurate, and conversions of

93% product 11, <2% starting material 8, and 6% diadduct

12, in agreement with the small-scale studies.

However, when a 50-batch sequence was attempted, all the

initial batches showed only a trace of conversion which

gradually increased up to ∼10%. Running the reaction mixture

in a small-scale microwave reactor proved that the reaction

material solution was good. Examination of the reaction vessel,

which was new, showed slight blackening from Pd residues

which were also present on the ﬁbre optic probe. The ﬁbre optic

probe was thoroughly recleaned, and a clean reaction vessel

was used which had previously been used for some base

hydrolyses. The next batches then worked successfully. We

speculate that the surface of this vessel may have been either

cleaned or deactivated by the base in some manner, thus

allowing successful reaction.

It is not known what caused this failure, whether some

unseen contamination on the new reaction vessel, or residues

left on the ﬁbre optic probe that had not been cleaned

completely. With only ∼4 mg of catalyst entering the vessel

with every batch, poisoning of the catalyst by a small amount

of contaminant could have been signiﬁcant. This might also

explain why the reaction was very slowly improving over the

course of the ﬁrst 20 batches. In either case, the Voyager itself

was clearly not at fault. Changing it over to Discover mode

(small-scale tube) gave complete conversion as expected, and

the reaction mixture was clearly being heated sufﬁciently in

Voyager mode because the metallic Pd nanoparticles could be

seen in the reaction product (Figure 5). Further studies are

ongoing.

Once this problem had been avoided (if not solved), a further

sequence of 20 batches was prepared which were grouped into

batches of four for analysis and which showed conversions

>90% as expected. Workup was as described before, but adding

acid during the drown-out provided essentially quantitative yield

of the product cinnamate 11 of 98.6% quality by HPLC, with

0.6% of 8 and 0.8% of diadduct 12. Further details are included

in Table 1.

The reactor vessel showed some level of Pd plating after

another 25 batches of this Heck chemistry and so was removed

and cleaned along with the probe. Fifty-batch sequences were

now set in train, using the method described above with 13 as

the base, to determine how many batches could be run before

the vessel and probe needed to be cleaned. The results for the

conversions of 8 to 11 and 12 are collected in Figure 6 and

showed some surprising trends. For example, the ﬁrst few

batches of each sequence tended to perform particularly poorly,

whether starting from a clean vessel (batches 1-5) or restarting

after an overnight break (batches 51-55; the vessel was not

cleaned in between). This trend also got worse, so that after

the second overnight break at batch 100, it took more than 15

batches to get the conversion back to a barely acceptable ∼80%.

After the ﬁrst 50 batches, the upper part of the vessel was plated

with Pd, but the lower part and the probe were relatively clean;

after the second group of 50, the vessel and especially the probe

were heavily coated in Pd residues. Cleaning the probe, but

not the vessel itself, did not help, as batches 126-130 show,

Figure 6. Vessel performance showing conversions of 8 to 11

and 12 over multiple batches (note irregular batch scale).

but recleaning the probe and installing a clean vessel in mid-

sequence did restore very good performance (batches 131-140).

The cycle time also improved to 7.5 min for these latter batches,

compared to 8.6 min for the previous ones with incomplete

conversions.

In summary, it appears that about 50 batches taking typically

7 h can be conducted before the performance starts to deteriorate

signiﬁcantly below the 90% conversion level (it should be noted

that production of diadduct 12 actually represents part of the

forward reaction conversion, albeit undesired over-reaction). It

also appears that performance might be better if run continually,

rather than leaving the reaction vessel empty overnight.

However, in the worst case, the probe may need to be cleaned

and the reaction vessel swapped for a clean one once a day,

which is the work of a few minutes.

The workup for these batches was as before, and could

conveniently be conducted on 20-batch scale (1050 mL). The

reaction liquors could be decanted from settled Pd residues (cf.

Figure 5) and any crystallised Hunig’s base HBr salts, and an

equal volume of warm water added with stirring, which gave

a better form than previously of the product as a yellow

precipitate. Adding acid was not necessary to remove residual

Hunig’s base as this washed out easily, so that typical isolated

yields were in the range 81-87% with no base or residual

1

solvent as determined by H NMR. Quality by HPLC was

excellent, showing 95-96% product 11 and 4-5% diadduct

12 with no residual starting material 8 in these cases.

Newman-Kwart Rearrangement. Like the Claisen rear-

rangement, the Newman-Kwart rearrangement (NKR)¹⁹pre-

sents another example of a single-phase unimolecular rear-

rangement which requires high temperatures and has been

reviewed recently (Scheme 4).²⁰Hence, high-boiling solvents

are often used, or it can be performed as a melt. Since we have

reported on this reaction previously^5band several aspects are

similar to the Claisen rearrangement, only brief comments

speciﬁc to the Voyager work will be made here.

(19) (a) Newman, M. S.; Karnes, H. A. J. Org. Chem. 1966, 31, 3980–

3984. (b) Kwart, H.; Evans, E. R. J. Org. Chem. 1966, 31, 410–412.

(20) Lloyd-Jones, G. C.; Moseley, J. D.; Renny, J. S. Synthesis 2008, 661–

689.

Vol. 12, No. 5, 2008 / Organic Process Research & Development

•

975

Scheme 4

Small-scale tube studies were conducted as for other reac-

tions, generally at 4 volumes in DMA (or NMP). Reaction

conditions were then readily transferred to the Voyager, using

a simple three-step method similar to that used for the Claisen

rearrangement (i.e., add, microwave and remove). Short auto-

mated sequences were run for all three O-thiocarbamates 14a-

c. Under standard conditions for 14a (200 °C for 10 min in 4

volumes of DMA), a ﬁve-batch run of 14a gave >99%

conversion for each batch of 15a, with a cycle time of 16 min

per batch, and an unoptimised yield of 77%. A three-batch run

of 14a under similar but more concentrated conditions (2

volumes) gave complete conversion with a 93% isolated yield

in similar cycle time. This would equate to 460 g per 8 h day

(30 batches).

Compound 14b required a slightly longer time and was run

in a six-batch sequence to give again complete conversion with

an unoptimised yield of 79% with a cycle time of 27 min. In

this case, the discharge from the vessel was slowed to the

maximum possible (∼3 min) to mimic a direct aqueous drown-

out of the product into 12 volumes of water. This naturally

lengthened the cycle time by bringing part of the workup within

the Voyager sequencing. However, it did establish that a slow

pump-out could be achieved, if for example the unquenched

reaction mixture was unstable, and required immediate neu-

tralisation. Such a quench could also be achieved by neutralising

inside the vessel, but this would give issues with removal of

precipitated solids, and also residual water in the vessel for the

next batch. Overall, there was no beneﬁt to the form of the

product isolated in this case, and given the additional time in-

volved, it was not as efﬁcient as the previous example. The

isolated yields from the individual batches were in the range

76-84%, which was felt to be more a test of the reproducibility

of the workup than of the Voyager.

Lastly, compound 14c was heated at 4 volumes of DMA to

210 °C for 20 min in a ﬁve-batch cycle to give ∼95%

conversion for each batch with an overall isolated yield of 82%.

At this concentration with this yield and a cycle time of 28

min, 160 g per day could be produced. Running the reactions

below complete conversion proved beneﬁcial in this case as

the unreacted O-thiocarbamate 14c was more readily removed

on aqueous drown-out than some very minor impurities if the

reaction was forced to completion. The aqueous drown-out was

performed off-line from the Voyager runs so that both conver-

sion and yield data could be obtained for each batch as a

measure of reproducibility. Figure 7 shows excellent reproduc-

ibility for both aspects, which was also typical of the related

NKR examples. Furthermore, in every case above, the quality

of the products was excellent (>98%) with only aqueous drown-

outs for puriﬁcations. Comparison data with other reactions

performed here are included in Tables 1 and 2.

Figure 7. Conversions and yields for individual batches of 15c.

These were the ﬁrst examples performed in the Voyager,

and larger runs could not be performed at the time due to lack

of materials. This is why individual batches were isolated and

analysed, to give an idea of the batch-to-batch reproducibility.

In later work, batches were combined into larger groups (5-20

batches) for the Claisen, Heck and hydrolysis reactions, when

we knew the reproducibility was good. Note that slightly higher

temperatures could be achieved for these reactions than for the

Claisen rearrangement (210 °C compared to 195 °C); this was

possible because both the solvent (DMA) and the substrates

(14a-c) heated much better than 4 and DCB, and even 20 min

at 210 °C did not require the magnetron to work at full power.

The limit for the Voyager on a 30-50 mL sample appears to

be about 220-230 °C in a favourable case. Finally, it should

be noted that these compounds (14a-c) have also been converted

to their S-thiocarbamate products 15a-c in a Milestone Flow-

SYNTH,²¹which is a true continuous ﬂow, large-scale micro-

wave reactor.

Hydrolysis of S-Thiocarbamates. Another second-order

reaction worth trying in the Voyager was the hydrolysis of the

S-thiocarbamates produced earlier. Furthermore, hydrolysis of

the S-thiocarbamate group, which behaves somewhat like an

amide, would complete the sequence from phenol to thiophenol,

the usual rationale for the NKR.²⁰We were also hopeful this

would provide another homogeneous reaction mixture. Unfor-

tunately, the S-thiocarbamates generally had low solubility in

methanol; the 2-nitro compound 15a would not dissolve below

30 volumes, and whilst the 4-nitro compound 15b did dissolve,

it hydrolysed too quickly to be worth testing under microwave

conditions. The 3-methyl-4-nitro compound 15c however would

dissolve in about 25-30 volumes of methanol, and we decided

this would provide a good test of productivity for a dilute

reaction mixture on the limit of acceptable volumes for process

scale-up (Scheme 5).

Small-scale screening reactions quickly established that

complete hydrolysis could be achieved after ∼2 min at 110 °C

with 1.5 equiv of 1.0 M aqueous NaOH. The stoichiometry

was also tested, and at this temperature fast and complete

conversions were still obtained down to about 1.1 equiv (Figure

8). However, some reaction tubes showed slight variability at

lower stoichiometries, and so 1.3 equiv was chosen as a robust

(21) Moseley, J. D.; Lawton, S, J. Chem. Today 2007, 25 (2), 16–19.

976

•

Vol. 12, No. 5, 2008 / Organic Process Research & Development

Scheme 5

the need to cool the vessel contents to 65 °C; above this

temperature, the methanol was superheated, and hot vapours

condensed in the air line as a bright orange liquid as soon as the

valve opened. Although this appeared to have no practical

drawback (most of the liquid was blown into the vessel when

it was emptied), it seemed a less than ideal way to run the

process. If a higher temperature could be accepted at this point

or a higher-boiling alcohol was used, then less cooling would

be required before emptying the vessel.

A solution of 101 g of 15c in 3050 mL (30 volumes) of

methanol was prepared and ﬁltered to remove residual ﬁnes

and charged through the SM1 line, whilst the NaOH was

charged from a measuring cylinder through the SM2 line. Fresh

methanol was fed from a vessel through the solvent line. The

three-batch proving trial worked exactly as planned, so a single

64-batch sequence using a total of 96 g of 15c in 2880 mL of

methanol was processed over one 10.1 h period. Batches were

combined into groups of three initially and then 10, and HPLC

showed conversions for all were >99%. Output volumes were

exactly 58 mL, once again demonstrating very good control

and reproducibility by the pump for repeated multiple charges.

Total reaction time for 64 batches was 10.1 h, or 9.45 min

per 1.5 g batch. Further details are given in Table 1. Overall,

productivity was low because, although the reaction time and

temperature were low and the cycle time was fast, the reaction

was very dilute. However, it does demonstrate that worthwhile

laboratory-scale quantities of product (50-100 g) can be

obtained per day using this instrument even in an unfavourable

case. Kilo-scale manufacture would be possible with two such

instruments operating for a week. Productivity would be further

enhanced if improvements to the cycle time or concentration

could be made. Further data and comparisons are given in

Table 2.

level for reaction. Having performed this brief survey of reaction

parameters, ﬁne-tuning of conditions began on the Voyager.

Since there was a high background rate of reaction at RT, the

S-thiocarbamate and the NaOH were not premixed. Although

additional (non-microwaved) reaction in the feed vessel is in

principle desirable, premixing the reagents would result in

considerable precipitation of products in the feed vessel, which

would block the SM1 line. We also wanted to model a situation

where premixing was undesirable due to potentially deleterious

side reactions, so the reagents were segregated. This also

provided the opportunity to use the SM2 line for a second

reagent, and the requirement to use the solvent line for a wash

(to avoid precipitate in the lines).

Furthermore, the reagent charge of NaOH was going to be

relatively small compared to the starting material solution, which

would test the pump calibration and reproducibility more

thoroughly than previously, since small errors in a small reagent

charge could signiﬁcantly affect the stoichiometry in each batch

(cf. Figure 8). A concentration of 2.0 M had been chosen as a

compromise between not adding too much water to the reaction

mixture compared to the methanol volume (in case of precipi-

tating either the starting material or the product), but not too

small such that errors in the charging became too critical for

the stoichiometry.

The workup was developed ﬁrst on preliminary batches, then

on groups of 10, and ﬁnally on the remaining bulk reaction

mixture, 44 batches at one time. The reaction liquors were

reduced to about one-third volume, water was added, and then

concentrated HCl was added (2.5 equiv) to neutralise the

remaining base, which resulted in precipitation of the product

17. This was isolated by ﬁltration and washed with more water

to remove inorganic salts. The overall quality of 17 was 93%

The Voyager sequence was therefore as follows: charge 47

mL of starting reaction mixture (15c) through SM1; charge 4.6

mL of 2.0 M NaOH (1.3 equiv) through SM2; wash the lines

twice with 3 mL of fresh methanol through the solvent line

(i.e., 6 mL total); heat to 110 °C and hold for 1.5 min; cool to

65 °C and remove the product from the vessel (full details are

given in the Supporting Information). The overall cycle time

was relatively long at 9.45 min, and one reason for this was

1

by HPLC (>99% by H NMR), with 6% of the disulﬁde

impurity 18 (which may be over-reported by HPLC), unchanged

from the level in the crude reaction mixture. The ﬁnal isolated

yield was 62 g (91%).

Combined Hydrolysis and Alkylation Reaction. Having

performed the hydrolysis reaction successfully, we now deter-

mined to combine this with an alkylation reaction, taking

advantage of the intermediate thiophenolate 16 generated in situ

(Scheme 6). Although in principle a trivial reaction, this

provided an opportunity to use all three input lines for the

starting material and two reagents. This would require foregoing

the solvent wash since there was no spare solvent line. It also

Figure 8. Affect of NaOH stoichiometry on reaction conversion

for hydrolysis of 15c to 17.

Vol. 12, No. 5, 2008 / Organic Process Research & Development

•

977

Table 2. Overall comparison of productivities

calculated mass processed^b

total

batches

cycle

time (min)

total run

time (h)

relative

actual mass

in 8 h

(g)

in 24 h

(g)

reaction

volume^a

processed^b(g)

Claisen (one step)

Heck #1 (10)

48

50

21

8.6

9.5

11

16

27

28

16.8

7.2

0.5

12

10

30

2

1006

163

>1000

96

480

184

222

76

1440

552

666

228

197

1440

432

504

Heck #2 (13)

>300

∼7^c

10.0

5.5

hydrolysis

64

hydrolysis + alkylation

NKR #1 (14a)

NKR #2 (14b)

NKR #3 (14c)

30

45

66

5

n/a

80

480

144

168

6

n/a

5

50

5

n/a

4

50

^aRelative to starting material mass. ^bBased on input starting material. ^cPer 50 batches.

Scheme 6

so residues from the previous batch were unlikely to be critical.

With these changes, the overall cycle time was now 11.0 min.

The three-batch proving trial worked exactly as planned with

no residual 15c or 16 left. From this a 30-batch sequence was

then prepared. Unfortunately, the only remaining large batch

of 15c left in hand was highly coloured, and although it looked

identical to other materials by HPLC and ¹H NMR, ∼10% of

its mass did not dissolve in methanol when the reaction solution

was ﬁltered. This mass has been accounted for in subsequent

calculations presented here, but the immediate consequence was

that the stoichiometric charges of NaOH and methyl iodide were

slightly higher than those planned. This did not affect the

chemistry adversely and, if anything, improved it. In all other

respects, the reaction proceeded as expected in the Voyager

method, which now involved three separate charges and no

solvent line washes. The chemistry was unaffected by the lack

of solvent washes, which improved the cycle time for this more

complex sequence. Conversions were 99% for all batch groups

assayed, and the total mass of 15c processed was 43 g in 5.5 h.

The reaction product was initially isolated by extraction to give

analytically pure product, but an aqueous drown-out was

developed which was simpler and provided good quality product

without chromatography. Finally, is should be noted that we

had been concerned that running multiple batches of hot alkaline

solutions might damage the integrity of the glass vessel.

However, there was no visible sign of deterioration, and its mass

was unchanged after both long hydrolysis sequences.

Summary of Continuous Flow. In regard to the continuous

ﬂow aspects of this study, the Voyager SF microwave reactor

has much in common with conventionally heated continuous

ﬂow instruments and the learning was similar. The ﬁrst priority

is to have a homogeneous reaction mixture. CF studies can use

a lot of material/volume to develop the process, but the

development time required should be relatively short (0.5-3

days in our experience) and easily developed from tube-scale

microwave reaction conditions in a largely scale-up free manner.

Once developed, minor modiﬁcations are easily incorporated,

and large quantities of product are manufactured with relative

ease. The ﬁrst batch (and possibly the second and third) through

the system will be different from the standard later batches.

This will be the same as starting up a conventional CF

instrument, and heating blank reaction solvent in the instrument

will warm it up before starting the sequence properly, which

will save some material or avoid a possible out-of-speciﬁcation

batch at the start. One ﬁnal signiﬁcant point of learning was

involved two small reagent charges (NaOH, then methyl iodide)

which would further test the pump accuracy on a combined

reaction sequence. To reduce the criticality of the charge of

methyl iodide, which is very dense (d ) 2.28), it was diluted

with methanol, although perhaps this was overcautious, since

the small line wash of 1.0 mL in Heck reaction 2 had worked

very reliably.

Small-scale scouting trials in the small microwave reactors

with fresh reaction liquors of 16 quickly showed that the

alkylation reaction would go to completion at modest temper-

atures with only a modest excess of methyl iodide (1.2 equiv).

From this screen, 1.5 equiv at 40 °C gave almost instantaneous

reaction and were chosen as the reaction conditions. Running

the reaction “all-in” was not possible since about 10% of

unreacted starting material 15c was left, even though all 16 had

been converted to 19; therefore the two-step reaction was

required as planned. Although microwave heating was not

required for the second step, the reaction mixture still needed

to be cooled to below 65 °C to stop intermediate product getting

into the air line when the vessel was opened. There was no

point heating again, so the methyl iodide solution (3.2 M in

methanol) was added through the solvent line, and cooling was

continued for a ﬁxed period of 30 s. This was achieved by using

a microwave step with the wattage set to 0 W, and the heating-

with-cooling function used, simply to cool the vessel further

(bp of methyl iodide is 41-43 °C). Once this step was complete,

the reaction mixture was already well below the bp of methanol

(65 °C) and so could be discharged from the vessel.

Slight changes had been made to the previous hydrolysis

method, reducing the initial input volume of 15c to 45 mL for

example to account for the small (3.3 mL) methyl iodide

solution charge. The solvent washes had to be removed, and

only the air blows could be used to ensure the lines were clear.

There was also no interbatch clean-out, but this was unlikely

to present a problem; the “all-in” reaction was nearly successful,

978

•

Vol. 12, No. 5, 2008 / Organic Process Research & Development

the need to run long sequences of many batches to really test

the reliability of the instrument, as noted for the Heck reaction;

only by running multiple batches over long periods can the true

reliability and out-put of the technology be fairly evaluated.

Summary of the Voyager. In regard to the Voyager, the

overall reliability and reproducibility, batch after batch in a long

sequence, was excellent. Once the pump calibration had been

set, the volumes pumped were reliable down to small values

(1 mL), even when changing between multiple solvents of

varying viscosities. It was worth the effort invested to get the

pump calibration and volumes correct at the start of develop-

ment for each reaction. The heating proﬁle was also very

reproducible between batches, except when the chemistry

affected the process as in the Pd-plating seen in the Heck

reaction. Running short trial sequences was essential before

committing larger quantities of material; this was useful to

conﬁrm both the chemistry and sequencing operations were

performing as planned, and also to produce initial supplies of

material for downstream studies. It was not worth overdevelop-

ing the chemistry beforehand, however, as subtle changes from

small-scale microwave tube reactions were inevitable; ﬁne-

tuning of conditions should be completed on the Voyager.

Overall this development time took typically 0.5 to 3 days per

reaction, but the ease of subsequent manufacture was well worth

the effort invested.

Two-step and more complex procedures were successfully

accommodated on the Voyager, even to using the third inlet

line as a reagent line. As with other CF instruments, line

blockages were a potential issue, either on entry or exit.

However, line washes and vessel rinses were useful to avoid

or ameliorate this problem, and the Voyager functionality made

it easy to incorporate these. In fact it was best to use the

functionality of the instrument as it was generally more efﬁcient

than designing other protocols. One interesting observation that

we made was the need to consider the Voyager like a mini-

pilot plant when sequencing reactions; and once a long sequence

was running, it behaved somewhat like a pilot plant in

manufacturing product at a greater rate than could comfortably

be accommodated in a research laboratory setting! Furthermore,

for certain concentrated high-temperature reactions in particular,

the combination of CF with microwave heating means that it

can out-perform most standard pilot-plant reactions in reaching

>200 °C to produce kg-scale quantities of product per day

(Table 2).

operation would bring even the dilute hydrolysis reaction

discussed above within the kg-scale range in less than one week.

Our experience strongly suggests that the Voyager is fully

capable of performing reliably over continuous 24-h periods,

and therefore of kg-scale production for homogeneous reactions.

Experimental Section

HPLC Methods. Reaction mixtures and products were

analysed by reverse phase HPLC on an Agilent 1100 series

instrument according to the following conditions: method 1;

column, Waters Symmetry C18 3.5 µm, 50 mm × 3.0 mm

i.d.; eluent A, puriﬁed water with 0.1% v/v formic acid; eluent

B, acetonitrile with 0.1% v/v formic acid; ﬂow rate 1.25 mL/

min.; wavelength 230 nm; temperature 45 °C; injection volume

2 µL; at t ) 0 min, 5% eluent B; at t ) 6 min, 95% eluent B;

at t ) 7 min, 95% eluent B; post time 1.5 min; method 2;

column, HiChrom ACE Phenyl 3.0 µm, 50 mm × 3.0 mm i.d.;

eluent A, puriﬁed water with 0.03% v/v formic acid; eluent B,

methanol with 0.03% v/v formic acid; ﬂow rate 1.25 mL/min.;

wavelength 220 nm; temperature 45 °C; injection volume 2 µL;

at t ) 0 min, 5% eluent B; at t ) 6 min, 95% eluent B; at t )

7.5 min, 95% eluent B; post time 3 min; method 3; column,

Genesis C18 3 µm, 100 mm × 3.0 mm i.d.; eluent A, 95%

puriﬁed water, 5% acetonitrile, 0.1% v/v formic acid; eluent

B, 95% acetonitrile, 5% puriﬁed water, 0.1% v/v formic acid;

ﬂow rate 0.75 mL/min.; wavelength 254 nm; temperature 35

°C; injection volume 5 µL; at t ) 0 min, 40% eluent B; at t )

5 min, 70% eluent B; at t ) 7 min, 70% eluent B; post time 3

min. Typical retention times (RT) are noted in each case.

General Procedures. Melting points were determined using

a Grifﬁn melting point apparatus (aluminium heating block)

1

and are uncorrected. H and ¹³C NMR spectra were recorded

on a Varian Inova 400 spectrometer at 400 and 100.6 MHz

respectively with chemical shifts given in ppm relative to TMS

at δ ) 0. Electrospray (ES+) mass spectra were performed on

Micromass ZQ or a Micromass Platform LC. Analytical TLC

was carried out on commercially prepared plates coated with

0.25 mm of self-indicating Merck Kieselgel 60 F₂₅₄and

visualised by UV light at 254 nm. Preparative scale silica gel

ﬂash chromatography (for puriﬁcation of analytical samples

only) was carried out by standard procedures using Merck

Kieselgel 60 (230-400 mesh). Where not stated otherwise,

assume standard practices have been applied.

Typical Small-Scale Microwave Procedures. Small-scale

microwave reactions were performed in thick-walled glass

sealed tubes in Biotage Initiator or CEM DiscoVer focused 300

W microwave reactors with IR temperature monitoring and

noninvasive pressure transducer. Procedures were typically

performed on 1-2 mmol scale exactly as described for the

large-scale microwave procedures. However, analytically pure

samples, if required, were more often puriﬁed by ﬂash silica

chromatography than the aqueous drown-out procedures used

on the larger scales. The heating time to reach the set

temperature was typically 30-90 s, depending on the scale,

the maximum wattage supplied (100-300 W) and the temper-

ature required (100-250 °C). The heating time is not included

in the quoted hold time for any given procedure.

Conclusions

In summary, we have presented results from scaling-out six

different reactions used in the pharmaceutical industry which

can beneﬁt from microwave heating. For these typical phar-

maceutical reactions, the Voyager stop-ﬂow microwave can

produce daily through-puts of between 50 and 250 g. For more

concentrated reactions, such as the Claisen and NKR reactions

discussed above, up to 0.5 kg is possible in a normal 8 h day,

or more if run continuously. Others have also demonstrated kg-

scale capability on a daily basis for concentrated reactions.^11a,22

The scale-out option of using multiple Voyagers, and/or 24-h

(22) Leadbeater, N. E.; Smith, R. J.; Barnard, T. M. Org. Biomol. Chem.

2007, 5, 822–825.

Vol. 12, No. 5, 2008 / Organic Process Research & Development

•

979

Preparation of Reaction Mixture for Claisen Rearrange-

ment of 4 to 5. 1-(2-Methyl-2-propenyl)oxynaphthalene (4)

(1006 g, 5.07 mol) was diluted with DCB (500 mL) to give a

dark-coloured solution (1500 mL) which was used in the CEM

Voyager microwave reactor as described in the Discussion

section, and according to the detailed protocol which can be

found in the Supporting Information. An analytically pure

sample could be obtained by ﬂash silica gel chromatography

in 9:1 isohexane/ethyl acetate, to yield 2-(2-methyl-2-propenyl)-

1-naphthol (5) as a colourless oil; HPLC (method 1, RT 4.89

min); ¹H NMR (400 MHz, CDCl₃) δ 8.19 (1H, m), 7.76 (1H,

m), 7.36-7.48 (3H, m), 7.18 (1H, m), 5.76 (1H, s), 4.98 (2H,

m), 3.53 (2H, s), 1.75 (3H, s); ¹³C NMR (100.6 MHz, CDCl₃)

δ 150.1, 144.7, 133.8, 128.9, 127.4, 125.7, 125.2, 124.9, 124.5,

120.1, 117.6, 112.7, 40.6, 22.0.

a solution of concentrated HCl (11.6M, 18.2 mL, 21.2 g, 212

mmol, 0.50 equiv) and water (1050 mL) over 1 h with good

agitation. Considerable heat of mixing is generated and a dense

white or pale yellow precipitate forms. Once the mixture has

cooled back to RT (with cooling water if required), the

precipitate is isolated by ﬁltration and the product cake slurry

washed twice with water (250 mL). The solid is dried in a

vacuum oven at 40 °C with an air bleed to yield methyl 3-(4-

acetylphenyl)acrylate (11) as an off-white solid (70-75 g,

81-87%). HPLC (method 1, RT 3.63, 97-99%); mp 98-99

°C; ¹H NMR (400 MHz, CDCl₃) δ 7.97 (2H, dt, J ) 8.8, 1.8

Hz), 7.71 (1H, d, J ) 16.0 Hz), 7.61 (2H, dd, 6.8, 1.6 Hz),

6.53 (1H, d, J ) 16.0 Hz), 3.83 (3H, s), 2.62 (3H, s); ¹³C NMR

(100.6 MHz, CDCl₃) δ 197.2, 166.9, 143.2, 138.66, 138.0,

128.8, 128.1, 120.3, 51.8, 26.6.

Preparation of Reaction Mixture for Heck Reaction of

8 and 9 Using Dicyclohexylmethylamine (10). Methyl acrylate

(79 mL, 880 mmol, 1.10 equiv) and dicyclohexylmethylamine

(10) (257 mL, 1200 mmol, 1.50 equiv) were added to a solution

of 4-bromoacetophenone (159 g, 800 mmol) dissolved in DMA

(1700 mL) and thoroughly mixed. In a separate ﬂask, tetrabu-

tylammonium bromide (0.89 g, 3.2 mmol, 0.4 mol %) was

added to a solution of Pd(OAc)₂(180 mg, 0.8 mmol, 0.1 mol

%) dissolved in DMA (210 mL) to give a dark orange solution.

This solution was added to the ﬁrst solution to give the reaction

mixture as an overall light orange solution. (Note: if any

particulates were visible, the combined solution was ﬁltered

through a grade 3 sinter at this point.) The reaction mixture

could be used immediately, or stored in a sealed vessel for some

time if required (at least one week if volatile components could

not evaporate). The reaction mixture was used in the CEM

Voyager microwave reactor as described in the Discussion

section, and according to the detailed protocol which can be

found in the Supporting Information. A typical workup proce-

dure follows.

Preparation of Reaction Mixture for Heck Reaction of

8 and 9 using Hunig’s Base (13). Methyl acrylate (95 mL,

1050 mmol, 1.05 equiv) and di-iso-propylethylamine (13) (262

mL, 1500 mmol, 1.50 equiv) were added to a solution of

4-bromoacetophenone (199 g, 1000 mmol) dissolved in DMA

(1700 mL) and thoroughly mixed. In a separate ﬂask, tetrabu-

tylammonium bromide (1.11 g, 4.0 mmol, 0.4 mol %) was

added to a solution of Pd(OAc)₂(225 mg, 1.0 mmol, 0.1 mol

%) dissolved in DMA (290 mL) to give a dark orange solution.

This solution was added to the ﬁrst solution to give the reaction

mixture as an overall light orange solution. (Note: if any

particulates were visible, the combined solution was ﬁltered

through a grade 3 sinter at this point.) The reaction mixture

could be used immediately, or stored in a sealed vessel for some

time if required (at least one week if volatile components could

not evaporate). The reaction mixture was used in the CEM

Voyager microwave reactor as described in the Discussion

section, and according to the detailed protocol which can be

found in the Supporting Information. A typical workup proce-

dure follows.

NKR Reaction of 14b to 15b with Aqueous Drown-Out.

Six batches of a warm solution of O-thiocarbamate 14b in DMA

held at 60 °C were sequentially charged by automation through

a CEM Voyager equipped with a ﬁbre optic probe and magnetic

stirrer bar (each batch contained 10.0 g of 14b (41.6 mmol) in

40 mL DMA (4 vols)). Each batch was heated with magnetic

stirring to 210 °C over 3.5 min with 300 W available power,

held at 210 °C for 20 min, then cooled by compressed air to

70 °C over 5 min. The individual batches were collected

separately and drowned out with varying quantities of water

from which it was determined that 12 volumes of water was

most efﬁcient to precipitate the product. The products were

isolated by ﬁltration, washed with more water and dried in a

vacuum oven at 50 °C to give 4-nitrophenyl-S-thiocarbamate

(15b) as a pale yellow or buff solid (combined yield 41.2 g,

82%). R_f0.33 (2:1 isohexane/ethyl acetate); HPLC (RT 3.01,

99.7%); mp 118-120 °C (lit.^19a122-124 °C); ¹H NMR (400

MHz, CDCl₃) δ 8.22 (2H, d, J ) 8.0 Hz), 7.68 (2H, d, J ) 8.0

Hz), 3.11 (3H, bs), 3.06 (3H, bs); ¹³C NMR (100.6 MHz,

CDCl₃) δ 164.58, 147.88, 137.65, 135.57, 123.54, 36.89 (2C);

MS (ES+) 227 (M + 1, 5%), 142 (60%), 101 (100%).

NKR Reaction of 14c to 15c (Aqueous Drown-Out

Separate). Five batches of a warm solution of O-thiocarbamate

14c in DMA held at 60 °C were sequentially charged by

automation through a CEM Voyager equipped with a ﬁbre optic

probe and magnetic stirrer bar (each batch contained 10.0 g of

15c (41.6 mmol) in 40 mL DMA (4 vols)). Each batch was

heated with magnetic stirring to 210 °C over 3.5 min with 300

W available power, held at 210 °C for 20 min, then cooled by

compressed air to 70 °C over 5 min. The individual batches

were collected separately and drowned-out with varying quanti-

ties of water from which it was determined that 12 volumes of

water was most efﬁcient to precipitate the product. The products

were isolated by ﬁltration, washed with more water and dried

in a vacuum oven at 50 °C to give 3-methyl-4-nitrophenyl-S-

thiocarbamate (15c) as a pale orange solid (combined yield

41.2 g, 82%). HPLC (RT 3.8, 96%); mp 74-75 °C; ¹H NMR

(400 MHz, CDCl₃) δ 7.97 (1H, d, J ) 8.5 Hz), 7.48 (2H, m),

3.10 (3H, bs), 3.05 (3H, bs), 2.60 (3H, s); ¹³C NMR (100.6

MHz, CDCl₃) δ 164.96, 149.03, 139.02, 135.30, 133.90, 133.31,

124.81, 36.97, 20.38; MS (ZQ) (ES+) 241 (M + 1, 100%).

Typical Workup Procedure of the Heck Reaction using

Hunig’s Base. The combined dark orange reaction liquors from

20 successful batches (1050 mL, 424 mmol) were poured into

980

•

Vol. 12, No. 5, 2008 / Organic Process Research & Development

Preparation of the Reaction Mixture for Hydrolysis of

15c to 17. 3-Methyl-4-nitrophenyl-S-thiocarbamate (15c) (100.6

g, 418 mmol) was dissolved in methanol (3050 mL) and the

solution ﬁltered through a grade 3 sinter to remove any ﬁnes.

This solution was added to the CEM Voyager through the SM1

line, sodium hydroxide (2.0M) through the SM2 line, and fresh

methanol through the solvent line. These solutions were charged

to the CEM Voyager microwave reactor as described in the

Discussion section, and according to the detailed protocol which

can be found in the Supporting Information. A typical workup

procedure follows.

Typical Workup Procedure for the Hydrolysis Reaction

of 15c to 17. The combined dark red reaction liquors of 10

successful batches (580 mL, 30 vols) were reduced to 10 vols

under reduced pressure. Water (200 mL, 13.5 vols) was added

to the solution with stirring followed by hydrochloric acid

(11.6M, 13.0 mL, 156 mmol) which precipitated a pale yellow

solid. The precipitate was isolated by ﬁltration and the product

cake slurry washed with water (10 mL). The solid was dried in

a vacuum oven at RT with an air bleed to yield the desired

product, 3-methyl-4-nitrothiophenol (17) (10.7 g, 101%). This

procedure was repeated with another 10 successful batches (10.8

g) and then repeated on a larger scale using the remaining 44

batches to yield a pale orange solid (40.1 g) giving a total overall

yield for 64 batches of 61.7 g (91%). HPLC (method 2, RT

3.81, 92.7% with 6.1% disulﬁde 18); mp 59-62 °C; ¹H NMR

(400 MHz, CDCl₃) δ 7.91 (1H, d, J ) 10.0 Hz), 7.18 (1H, s),

7.16 (1H, m), 3.66 (1H, s), 2.58 (3H, s); ¹³C NMR (100.6 MHz,

CDCl₃) δ 139.7, 135.0, 131.6, 129.6, 126.3, 125.7, 20.8.

Preparation of the Reaction Mixture for the Combined

Hydrolysis/Alkylation Reaction of 15c to 19. 3-Methyl-4-

nitrophenyl-S-thiocarbamate (15c) (50.1 g, 208 mmol) was

dissolved in methanol (1500 mL) and the solution ﬁltered

through a grade 3 sinter to remove undissolved dark material

(3.3 g). This solution was added to the CEM Voyager through

the SM1 line, sodium hydroxide (2.0 M) through the SM2 line,

and a methyl iodide/methanol solution (containing 20.0 mL

methyl iodide (321 mmol) made up to 100 mL with methanol)

through the solvent line. These solutions were charged to the

CEM Voyager microwave reactor as described in the Discussion

section, and according to the detailed protocol which can be

found in the Supporting Information. A typical workup proce-

dure follows.

Typical Workup Procedure for the Combined Hydroly-

sis/Alkylation Reaction of 15c to 19. The combined reaction

liquors of two successful batches (104 mL, 30 vols) were

reduced to 10 vols under reduced pressure. Water (150 mL, 50

vols) was added to the solution with stirring to form a brown

precipitate. The precipitate was isolated by ﬁltration and the

product cake slurry washed with water (3 mL, 1 vol). The solid

was dried on the ﬁlter to yield S-methyl-3-methyl-4-nitrophe-

nylsulﬁde (19) as a brown solid (1.92 g, 84%). This procedure

was repeated in ﬁve-batch portions for the 30 batches generated

in the Voyager trial. HPLC (method 3, RT 5.08, 98.4%); mp

1

40-42 °C; H NMR (400 MHz, CDCl₃) δ 7.98 (1H, d, J )

8.4 Hz), 7.12 (2H, m), 2.62 (3H, s), 2.53 (3H, s); ¹³C NMR

(100.6 MHz, CDCl₃) δ 146.7, 145.5, 134.7, 128.5, 125.4, 122.9,

21.2, 14.8.

Acknowledgment

We thank Ros Sankey and Anthony Thomson for some

preparatory studies; Philip Lenden for the NKR results reported

previously;^5bPaul Murray for technical advice on the Heck

reactions; Neil Weston for DSC data (all of AstraZeneca,

Avlon); and CEM for technical advice on the use of the CEM

Voyager, and Paul Greenwood in particular.

Supporting Information Available

Detailed protocol. This material is available free of charge

via the Internet at http://pubs.acs.org.

Received for review April 12, 2008.

OP800091P

Vol. 12, No. 5, 2008 / Organic Process Research & Development

•

981

Article Doi

Scaling-out pharmaceutical reactions in an automated stop-flow microwave reactor

DOI: 10.1021/op800091p

Source and publish data:

Authors:

Article abstract of DOI:10.1021/op800091p

Full text of DOI:10.1021/op800091p

Products guided by the article

R&D Labs maybe for 83086-55-9

Relevant to this article

Hot Product