The continuous flow Barbier reaction: An improved environmental alternative to the Grignard reaction?

Article abstract of DOI:10.1039/c2gc35050e

A key pharmaceutical intermediate (1) for production of edivoxetine·HCl was prepared in >99% ee via a continuous Barbier reaction, which improves the greenness of the process relative to a traditional Grignard batch process. The Barbier flow process was run optimally by Eli Lilly and Company in a series of continuous stirred tank reactors (CSTR) where residence times, solvent composition, stoichiometry, and operations temperature were optimized to produce 12 g h^-1 crude ketone 6 with 98% ee and 88% in situ yield for 47 hours total flow time. Continuous salt formation and isolation of intermediate 1 from the ketone solution was demonstrated at 89% yield, >99% purity, and 22 g h^-1 production rates using MSMPRs in series for 18 hours total flow time. Key benefits to this continuous approach include greater than 30% reduced process mass intensity and magnesium usage relative to a traditional batch process. In addition, the flow process imparts significant process safety benefits for Barbier/Grignard processes including >100× less excess magnesium to quench, >100× less diisobutylaluminum hydride to initiate, and in this system, maximum long-term scale is expected to be 50 L which replaces 4000-6000 L batch reactors.

Full text of DOI:10.1039/c2gc35050e

View Article Online / Journal Homepage / Table of Contents for this issue
Dynamic Article Links
Green Chemistry
Cite this: Green Chem., 2012, 14, 1524
www.rsc.org/greenchem
PAPER
The continuous flow Barbier reaction: an improved environmental alternative
to the Grignard reaction?
Michael E. Kopach,*a Dilwyn J. Roberts,a Martin D. Johnson,a Jennifer McClary Groh,a Jonathan J. Adler,b
John P. Schafer,b Michael E. Kobierskia and William G. Tranklea
Received 11th January 2012, Accepted 6th March 2012
DOI: 10.1039/c2gc35050e
A key pharmaceutical intermediate (1) for production of edivoxetine·HCl was prepared in >99% ee via a
continuous Barbier reaction, which improves the greenness of the process relative to a traditional Grignard
batch process. The Barbier flow process was run optimally by Eli Lilly and Company in a series of
continuous stirred tank reactors (CSTR) where residence times, solvent composition, stoichiometry, and
operations temperature were optimized to produce 12 g h−1 crude ketone 6 with 98% ee and 88% in situ
yield for 47 hours total flow time. Continuous salt formation and isolation of intermediate 1 from the
ketone solution was demonstrated at 89% yield, >99% purity, and 22 g h−1 production rates using
MSMPRs in series for 18 hours total flow time. Key benefits to this continuous approach include greater
than 30% reduced process mass intensity and magnesium usage relative to a traditional batch process. In
addition, the flow process imparts significant process safety benefits for Barbier/Grignard processes
including >100× less excess magnesium to quench, >100× less diisobutylaluminum hydride to initiate,
and in this system, maximum long-term scale is expected to be 50 L which replaces 4000–6000 L batch
reactors.
Flow technologies have recently received a significant amount
Introduction
of attention in organic synthesis.4 Established advantages of con-
At the start of the 20th century organomagnesium chemistry
emerged with the discovery of the Barbier and Grignard reac-
tions.1 Since inception, the Grignard reaction has been exten-
sively studied; it remains one of the most powerful synthetic
methodologies and has been widely used in manufacturing
within the fine chemicals, agricultural and pharmaceutical indus-
tries.2 The Grignard reaction first involves formation of the orga-
nomagnesium reagent by reaction of elemental magnesium with
an alkyl or aryl halide. Once formed, the nucleophilic Grignard
reagent is then reacted with an electrophile in a separate step to
produce an alcohol or ketonic product. To date the Grignard
reaction remains the best methodology to produce tertiary alco-
hols. Most commonly, a Grignard process is commercially prac-
tised in a three-vessel setup where the Grignard reagent
formation, reaction and quench are segregated and run sequen-
tially in separate vessels.3 In principle, a Barbier reaction
approach is operationally more simple than a conventional
Grignard process in that the organohalide and electrophile are
combined as a single stream, then added to activated magnesium.
However, for this approach to be successful, the key organomag-
nesium intermediate must be stable at the Grignard formation
temperature, which is often a limitation.
tinuous flow chemistry include the ability to precisely control
process variables such as residence time, temperature, pressure,
heat transfer, and operation at a different kinetic regime, e.g. con-
tinuous stirred tank reactor (CSTR) operates at end of reaction
conditions. All of these elements can ultimately lead to improved
yield and higher chemical purity for the target reaction. In fact,
select Grignard reactions with aldehydes and ketones have been
shown to be efficient in microreactors using commercially avail-
able Grignard reagents.5 However, in many instances, for practi-
cal targets, it is still necessary to prepare the Grignard reagent
prior to use due to stability, solubility, or commercial availability
issues. Continuous syntheses of select Grignard reagents have
been achieved via plug flow technology, where a solution of
organic halide is continuously passed through a fixed bed or
column of magnesium.6 While there are certain advantages to
the plug flow technology, the possibility exists, in certain
instances, for hot spots and precipitation of magnesium com-
plexes. CSTRs have been used efficiently for homogeneous pro-
cesses and recently a CSTR process train has been used to run a
heterogeneous hydrogenation process by Pfizer for preparation
of an intermediate for a smoking cessation drug.7 One of key
benefits to the CSTR approach is that any solid reagent used in a
stirred tank could in principle be used in the CSTR, and batch
kinetic data can be directly translated to a CSTR train via stan-
dard methods. In addition, up to 80% reduction in manufacturing
footprint is possible by adoption of the CSTR spproach. Despite
these advantages, along with those resulting from steady state
aChemical Product Research and Development, Eli Lilly and Company,
Indianapolis, IN 46285, USA. E-mail: kopach_michael@lilly.com
bD&M Continuous Solutions, Indianapolis, IN 46285, USA
1524 | Green Chem., 2012, 14, 1524–1536
This journal is © The Royal Society of Chemistry 2012

View Article Online
control, very few examples of Grignard reactions run in a CSTR
process train exist. In addition, the Barbier reaction has been
much less studied than its Grignard counterpart, and to our
knowledge a continuous Barbier reaction where chirality is pre-
served is unprecedented. Recently we disclosed the details for
the synthesis of 1, an important intermediate for the synthesis of
edivoxetine·HCl, a highly selective norepinephrine reuptake
inhibitor under clinical evaluation for the treatment of depression
and ADHD (Scheme 1).8 Herein we report the development of a
continuous flow synthesis of 1 including a novel application of
the Barbier reaction which proceeds with preservation of
chirality.
4-chlorotetrahydropyran (4-Cl-THP) reagent, magnesium and
both processing and cleaning solvents. The key Grignard reagent
was formed continuously in the first tank of a 250 mL CSTR
series (CSTR #1) by concomitant feeding of a 4-Cl-THP and
THF into a suspension of activated magnesium (Fig. 1). Average
hydraulic residence time (τ) in the Grignard formation tank was
1 h. The 4-Cl-THP Grignard reagent, 5, flowed to the second
vessel in the CSTR series through a screen where the output
flow rate from the Grignard formation vessel was controlled by
an automated pressure swing cylinder which was filled by
trapped vacuum from CSTR #1 and pumped liquid forward
intermittently at a frequency of about once per minute to CSTR
#2 with nitrogen pressure. This transfer zone was designed to
prevent clogging by suspended solids, a common problem in
continuous flow applications. To keep the transfer zone clear and
prevent clogging of solid magnesium on the screen in the exit
tube from CSTR #1, after each bolus transfer was complete a
reverse-nitrogen purge was applied.
The freshly formed Grignard reagent THF solution flows into
a reaction tank (CSTR #2) concurrently with a toluene solution
of morpholine amide 3 free base.9 The hydraulic residence time
was 15 min for this second CSTR and the contents of the
Grignard reaction tank flowed continuously into a quench tank
(CSTR #3) with a concomitant acetic acid/water feed. Initial
development of the continuous approach focused on adaptation
of parameters developed for the batch process where the temp-
eratures for Grignard formation, reaction, and quench were main-
tained at 60, 23 and 0 °C, respectively. Unfortunately, when
using these conditions, the best conversions achieved for the
Grignard reaction were 80–90% and increasing residence time in
Results and discussion
Displacement of morpholine amides with strong nucleophiles is
well precedented and the Grignard approach was the most attrac-
tive for the production of 1 in that chirality of the S-morpholino
starting material could be preserved (Scheme 2).8 However,
during the history of this project, several contract manufacturing
organizations had difficulty with the Grignard process, and in
fact one supplier defaulted on delivery of material. The main
issue encountered was a challenging initiation which operated
close to the boiling point of the solvent with a potential 100 °C
adiabatic heat rise for the process. In addition, there were mul-
tiple process sensitivities to racemization, such as addition order,
amide feed rate, quench conditions and air/moisture contami-
nation. All of these issues led us to develop a continuous process
that operated where kinetic regime (or concentration of reagents
with respect to products) and heat evolution were controlled by
reactant feed rate and mean residence time at steady state.
Following completion of the most recent pilot plant campaign,
proof of concept R&D efforts ensued to transform the batch
process to continuous mode. Key objectives were improvements
to process safety and material usage reduction, including the
Scheme 1 Retrosynthesis of edivoxetine·HCl.
Fig. 1 Three vessel Grignard CSTR process train.
Scheme 2 Grignard synthesis of compound 1.
This journal is © The Royal Society of Chemistry 2012
Green Chem., 2012, 14, 1524–1536 | 1525

View Article Online
the Grignard formation and reaction vessels did not improve con-
version. However, a key visual observation was that the appear-
ance within the Grignard formation vessel was changing which
led to the hypothesis that the active Grignard complex, inter-
mediate 5, was precipitating from solution. This was confirmed
by the quenching of a Grignard reaction stream sample with a
surrogate aldehyde which revealed that the reaction vessel
(CSTR #2) was Grignard reagent starved.10 In order to address
the solubility issue of the chloro Grignard reagent, we decided to
evaluate the less frequently used Barbier approach. It was sur-
mised that rapid reaction of the Grignard complex as formed,
might alleviate the solubility issue. The key consideration for
this approach to be successful is that the Grignard reaction and
complex formation need to occur at the same temperature, and
an essential element in this system is preservation of chirality.
To assess the Barbier chemistry, a dry toluene solution con-
taining 1.5 equiv. of 4-Cl-THP and 1 equiv. of morpholine
amide free base 3 was prepared and fed to a THF suspension of
activated magnesium over 1 h at 55–60 °C.11 After completion
of the amide 3/4-Cl-THP stream feed, HPLC analysis revealed
98% conversion with 96% ee. Most importantly, aside from
residual magnesium there was no evidence of precipitated mag-
nesium complexes or other salts. With these results in hand, we
proceeded rapidly to evaluate the Barbier approach in continuous
flow mode. A simple two CSTR process train was used where
the Grignard reagent formation and reaction were performed in
CSTR #1, then an aqueous acetic acid quench was performed in
CSTR #2. A series of 7 experiments were performed over 7 days
consisting of 6–8 h continuous runs (Table 1). The run length
each day was based upon three volume turnovers of the CSTRs
so that the results obtained are near steady state. The average
hydraulic residence time for the reaction tank was set at 90 min
at 55–60 °C operational temperature and 30 min at 0 °C for the
acetic acid/water quench tank. This set of experiments evaluated
the impact of 4-Cl-THP and acid quench stoichiometries.
Overall, the best results balancing conversion and chiral purity
were achieved using 1.25–1.35 equiv. 4-Cl-THP and 3 equiv. of
aqueous HOAc for the quench operation. Under these con-
ditions, it was possible to achieve >97% conversion and >95%
ee consistently for preparation of crude ketone 6 (Table 1, entries
1–3). In all cases, analysis was performed on the final composite
crude solution which reflected closely steady state conversion
and ee levels.
Racemization sensitivity analysis
Since chiral control was a critical parameter, and sensitivities
were observed in the prior art batch process,8 we sought to
thoroughly investigate racemization sensitivity under Barbier
conditions. The first set of experiments involved aging the reac-
tion mixture at 60 °C and monitoring chiral erosion (Fig. 2). The
drop in ee of the reaction mixture at 60 °C was shown to be sig-
nificant with a ∼1.5% ee drop per hour. To further understand
the system, 0.5 equiv. of both the starting amide 3 and ketone 6
were added to the pre-formed tetrahedral intermediate stirring at
60 °C. As expected, the ketone rapidly racemized upon addition
to the tetrahedral intermediate, whereas the starting amide
showed reasonable chiral stability with a 1% drop in ee per hour
observed at 60 °C. Overall the results from the chiral stability
studies indicated that operating at longer than the 90 min resi-
dence time at 60 °C in the CSTR system would result in chiral
erosion. While it was known that ketone 6 with 90% ee could be
upgraded to >98% ee by conversion to its MSA salt 1, a more
robust system was desired. For example, any mechanical mal-
functions e.g. blockages, pump malfunction etc. could cause the
reaction mixture to reside in the reaction tank for an extended
period of time which would cause ee degradation, perhaps
beyond recovery. However, the benefit of CSTR operation is that
a small amount of material is at risk in the CSTRs at any one
time relative to campaign size.
Table 1 In situ synthesis of 1 via CSTR Barbier approach (HPLC
area%)
Entry
Conversion (%)
ee (%)
Equiv 4-Cl THP
Equiv acid
1
2
3
4
5
6
7
97.3
97.7
98.1
97.3
98.4
97.5
97.7
96.2
95.8
95.4
94.2
91.6
94.9
92.2
1.25
1.25
1.35
2.0
1.5
1.5
3.0
3.0
3.0
4.0
5.0
3.5
3.0
1.5
Fig. 2 (a) Drop in ee by HPLC area% as a function of reaction time at 60 °C, (b) drop in ee by HPLC area% aged to longer reaction time at 60 °C.
1526 | Green Chem., 2012, 14, 1524–1536 This journal is © The Royal Society of Chemistry 2012

View Article Online
increase the volume in the first CSTR or add a second CSTR to
achieve complete conversion. For illustrative purposes only, con-
sider the design equation for an ideal CSTR with irreversible
first order reaction and assuming that volumetric flow rate does
not change with reaction, where conversion of reagent to product
is equal to 1 − 1/(1 + Da). Da is the dimensionless Damkohler
number which is equal to reaction rate constant × mean residence
time for a first order reaction (kτ). If the dominating rate limiting
step of the Barbier process is pseudo first order, at 94% conver-
sion Da = 16. However, Da = 51 would be needed for 98% con-
version, therefore τ (and thus reaction volume) would need to be
increased by 3.3× to reach 98% conversion in the single CSTR.
Alternatively, if a second CSTR in series operating at the same
temperature with 1 h τ is used, then by the same design equation,
conversion to greater than 98% will be achieved overall. No
doubt, the kinetics of the Barbier Grignard are more complex,
but the purpose of the illustration is to explain that, for a positive
order reaction, 2 CSTRs in series will have much less combined
τ to achieve 98% conversion than a single CSTR. A second
CSTR was used with a 1 h residence time at 35–40 °C, which
increased conversion to 98%. Since the Barbier reaction went to
completion without additional metal added to CSTR #2, this is
indicative that Grignard reagent formation at steady state is not
rate limiting. The optimal 3 vessel 250 mL CSTR process train
for the Barbier process is shown in Fig. 3.
Reaction temperature and solvent composition
optimization
The Barbier reaction temperature was evaluated from 25 to
60 °C and ee degradation monitored under typical process con-
ditions. A key point of failure was observed at temperatures less
than 30 °C where productive conversion could not be achieved.
However, it was shown that at 40 °C there was minimal chiral
degradation with only a 1.3% drop in ee after 5 h. In compari-
son, for a 5 h age of the system at 60 °C, a ∼7.0% drop in ee
would be expected (Fig. 2b). In the standard lab model 5 vol of
THF was typically used for the Grignard initiation relative to
amide 3, which was fed into the activated magnesium as a 25 wt%
solution in toluene. Thus, the amount of THF–toluene present in
all lab models at steady state was a fixed ratio at 45 : 55 of THF–
toluene. It was found that increasing the ratio of THF to toluene
has a significant impact on chirality and the best results were
achieved increasing the THF composition to 63 vol% while
maintaining the same net reaction volume. Under these con-
ditions, the ee drop was reduced by 50% relative to the toluene
rich system, during extended stirs at 40 and 60 °C. Based on
these results, an operational temperature of 35–40 °C for the
Barbier reaction and a steady state solvent composition of 60–65
vol% THF were established as optimal for future experiments.
Development of CSTR Barbier reaction
Initiating the Barbier reaction
A CSTR series was set up evaluating the impact of temperature
and solvent composition, the two primary parameters found to
impact enantiomeric purity in the batch experiments. The con-
tinuous flow Barbier experiments were run on 5 consecutive
days with a minimum 4–6 h of steady state observation
(Table 2). A 90 min residence time was selected for CSTR #1
(τ = 60 min for CSTR #2) and the stoichiometry of 4-Cl-THP
was centered at 1.35 equiv, 10% lower than the traditional batch
process. The best conditions at temperature (35–40 °C) and
solvent composition (63% vol THF) matched well what was
achieved by the optimized batch Barbier process with crude con-
version and ee both at 98% or better (Table 2, entries 4 and 5).
For the 35–40 °C experiments the conversion in CSTR #1 was
94% using a 90 min residence time. This left the option to
One of the key advantages to the continuous flow approach for
the Barbier or Grignard reactions is that the possibility exists in
principle to activate the metal only once for an entire campaign,
which the typical length in the plant is several months. In fact,
the metal activation step can often be the most hazardous oper-
ation in a Grignard process, so maintaining an always active
metal heel is a highly attractive option from a commercial per-
spective. In principle there are two primary approaches that can
be utilized for metal activation which include: (1) use a large
excess of magnesium, activate then run for several turnovers
before cleaning the residue and starting over; (2) maintain a
Table 2 CSTR Barbier optimization experiments
Rxn temp. Vol. ratio THF–
Conversion
(%)
ee
(%)
Entry
(°C)
toluene (%)
1
2
3
4
55–60
55–60
35–40
35–40
35–40
35–40
44
63
45
63
63
61
98.2
97.9
96.3
97.9
98.6
98.0
95.5
97.0
93.7
98.3
98.0
98.7
5
6 (Batch
Barbier Model)
Fig. 3 3-Vessel 250 mL CSTR train for Barbier chemistry.
Green Chem., 2012, 14, 1524–1536 | 1527
This journal is © The Royal Society of Chemistry 2012

View Article Online
Fig. 4 Continuous crystallization of compound 1.
pseudo steady state by a periodic re-charge of metal to the
system without additional activation. Option 2 was our preferred
option, but this necessitated demonstration of effective stopping
and restarting the continuous system after metal recharge. To test
restart capability, a batch Barbier reaction was run with 3 equiv.
magnesium based on an 8 h processing day which delivered
98% crude ee and conversion. After completion of the reaction,
the mixture was siphoned from the heel leaving ∼5% of the reac-
tion solution on the metal heel (enough reaction material was left
so as to completely submerge the magnesium) then diluted with
THF under a nitrogen blanket. The next day the mixture was
heated to 55 °C, 0.1 equiv. of 4-Cl-THP was added to test if it
was still active and the reaction immediately re-initiated without
the addition of DIBAL-H and iodine. The temperature was then
dropped to 35 °C and a second Barbier was carried out using the
same magnesium supply, the material from the second run
afforded crude product with 97% ee and 98% conversion.12
In continuous flow mode (described in the Experimental
section) the same paradigm was tested with several metal
recharges accomplished with only temporary mild drop in ee
which quickly recovered after the first hour (<1τ). In these cases
∼75% of the magnesium sequestered in the CSTR was con-
sumed by reaction before magnesium add back. Unactivated
magnesium metal added during recharges is activated by
Grignard reagent an existing active metal present in the reactor.
In addition, magnesium is physically activated due to the mech-
anical collisions within the system which keep the fresh metal
surface exposed, which is necessary for productive reaction. In
principle, a large savings in use of metal is realized by the con-
tinuous approach relative to the batch process in that magnesium
usage over time theoretically approaches 1.0 equiv. in continuous
mode whereas excess magnesium is used for every run in the
batch process. This translates to a potential 33% reduction in
elemental magnesium usage for the continuous approach, yet
there is more magnesium in the Grignard formation tank at any
one point in time than batch mode until the very end of the cam-
paign. More importantly, in a production campaign >100× less
excess magnesium is used overall, therefore 100× less is dis-
posed of as waste.
Continuous reaction, extraction, and crystallization
for isolation of compound 1
Once the optimized Barbier reaction was developed, a two day
proof of concept continuous reaction experiment was under-
taken. On the basis of preliminary work that had been carried
out the flow rates were as follows; THF (0.896 mL min−1) and
amide 3/4-Cl-THP (0.771 mL min−1) which produces a solvent
composition at steady state consisting of 1.0 THF to 0.60
toluene. The residence time in CSTR #1 was set at 90 min at
35 °C, the residence times in CSTR #2 (reaction finishing
vessel) and CSTR #3 (quench vessel) were set at 60 and 30 min,
with 35 and 0 °C operation temperature, respectively. During the
course of the day, the conversion and ee of the quench samples
were as expected according to the lab model with 98% conver-
sion and >98% ee remained for the whole day. At the end of the
first processing day, enough reaction mixture was left in CSTR
#1 to cover the magnesium heel which was then diluted with
THF and left inerted overnight. The next day the reaction
mixture was heated to 55 °C, 0.1 equiv of 4-Cl-THP was added
to verify the mixture was still initiated and the temperature
lowered to 35 °C prior to commencing the Barbier stream. The
flow rates and residence times remained unchanged from the pre-
vious day. As with the day before, the conversion and % ee of
the quench samples during the course of the day were consistent
with expectation. Magnesium re-charge was also demonstrated
on the second day, which is essential if the process is going to
1528 | Green Chem., 2012, 14, 1524–1536
This journal is © The Royal Society of Chemistry 2012

View Article Online
be run continuously for any prolonged period of time; no
changes in % conversion or % ee were observed after return to
steady state conditions. The final organic composite sample
showed 98.8% conversion, 98.9% ee of ketone 6 and an overall
yield of 88% for both days.
(potency corrected) over the two days was 89% and the ee of
isolated compound 1 was 99.4%.
Impurities
The resulting organic solutions from the two days of continu-
ous reaction experiment were washed in a two-stage continuous
cross-flow extraction. Aqueous sodium carbonate was added in
stage one to neutralize and the second stage received a water
wash to remove residual salts prior to solvent exchange. Solvent
exchange was performed to place ketone 6 in a 20 wt% IPA feed
stream with precise concentration. This stream was pumped into
a continuous crystallization (Fig. 4). The flow rates were
designed with the aim to have equi-molar feed streams and a τ of
1 h in each mixed suspension mixed product removal (MSMPR)
with a given throughput of material. The 20 wt% ketone 6 IPA
solution and a 7.7 wt% methanesulfonic acid IPA solution were
pumped into the first MSMPR crystallizer at 23 °C with flow
rates of 1.33 and 1.19 mL min−1 respectively, which afforded
equi-molar addition; the residence time was set for 1 h and the
compound 1 slurry was continuously transferred to the second
CSTR crystallizer at an average flowrate of 2.52 mL min−1. The
hydraulic residence time was 1 h in MSMPR #2. This was
designed to allow further desupersaturation of the slurry.
Aside from the R-enantiomer, the principle impurity observed in
the batch process was isobutyl ketone 7 which was observed at
levels as high as 0.5% in MSA salt 1. The source of this impur-
ity was determined to be from diisobutylaluminum hydride that
was used during the Grignard initiation.8 In addition, minor
amounts of alcohol 9 are also observed which is well rejected in
the crystallization of MSA salt 1. For the continuous Barbier
chemistry, since it is only necessary to activate the metal once
and the system quickly reaches steady state, only trace amounts
of impurity 7 which only forms during startup transition are
observed in the crude product. However, the Barbier approach
does generate impurity 8, which is observed at minor levels
when the standard Grignard reaction is employed (Scheme 3).
Benzyl ketone 8 is the primary impurity which results from reac-
tion of benzyl magnesium chloride and morpholine amide 3.
Continuous flow demo campaign
The compound 1 slurry was continuously transferred to a 1 L
glass pressure filter using another 4 valve transfer zone pump at
a flowrate of 2.52 mL min−1. Liquid was removed from the
slurry via filtration collecting the mother liquor in a tared filtrate
receiver. Once the crystallizers were empty, 2.5 volumes of IPA
was added to the first MSMPR crystallizer as a cake rinse. The
rinse solution was transferred through the transfer zones as well
as the second MSMPR crystallizer, and finally the rinse was
transferred to the filter where the cake was rinsed and the filtrate
collected in a separate tared filtrate receiver. The wet cake in the
filter was transferred to a tared drying dish and dried to a con-
stant weight in a vacuum dryer at 45 °C. The overall yield
A 47 h flow campaign using 649g amide 3 was run over four
days with a 4 vessel CSTR sequence which included a primary
Grignard reagent formation reaction vessel (CSTR #1), reaction
Scheme 3 Barbier process impurities.
Fig. 5 Flow diagram for the whole continuous process from amide 3 to product 1.
This journal is © The Royal Society of Chemistry 2012
Green Chem., 2012, 14, 1524–1536 | 1529

View Article Online
Table 4 Continuous/batch crystallization comparison for isolated
solid 1
Entry 1
Batch
Entry 2 Entry 3 Entry 4 Entry 5
Batch CSTR CSTR CSTR
Chemistry
(batch/CSTR)
Crystallization
(batch/MSMPR)
ee (%)
% 2
% 7
Grignard Barbier Barbier Barbier Barbier
Batch
Batch
Batch
Batch
MSMPR
99.9
0.15
0.32
<0.05
<0.05
5.38
99.9
0.18
<0.05
<0.05
0.08
4.39
<0.05
99.7
0.17
<0.05
0.09
0.13
3.68
99.8
0.16
<0.05
0.10
0.07
3.51
1.05
99.7
0.19
<0.05
0.27
<0.05
2.53
0.33
Fig. 6 47 h continuous campaign in situ data.
% 8
% 9a
% IPA
% Tol
Table 3 Purity data for compound
1
produced by continuous
<0.05
<0.05
crystallization
a Mixture of diastereomers.
Despite the high conversion observed in CSTR #1 the second
CSTR reaction finishing vessel was still operated to provide
additional conversion. The final datapoint for the continuous
demo campaign was collected at 47 h and revealed 97.5% con-
version and 99.7% ee. For this campaign a starting charge of
25 g magnesium was used and two 25 g charges of magnesium
were accomplished without loss of activation. At the 47 h time
point, 2.7 g magnesium remained (96% consumption) and all
fractions collected up to this point were acceptable for further
processing including the start-up fraction generated during the
first hour. Overall 592 g (potency corrected) of compound 1 was
isolated. Not all of the material from continuous reaction was
isolated, a portion of this campaign was crystallized continu-
ously in the MSMPRs, and a portion was crystallized batch. Iso-
lated compound 1 had about 99.7% ee whether it was isolated
by batch or continuous crystallization. Subsequently, two
additional continuous Barbier reaction campaigns were run, one
using 639 g amide 3 and running for a total of 47 hours flow
time, and the other using 1013 g amide 3 and running for
76 hours total flow time. A continuous crystallization demon-
stration was done to isolate 404 g compound 1 using the ketone
6 solution produced in the second 47 hour continuous Barbier
campaign.
Wt
ee
(%)
Cpd 2
(%)
Cpd 8a
(%)
IPA
(%)
Toluene
(%)
Fraction (g)
1
2
3
4
5
6
Avg/
total
48.2 99.7
115.8 99.8
56.8 99.7
40.2 99.8
55.7 99.7
87.8 99.4
404.4 99.7
0.22
0.17
0.17
0.17
0.20
0.20
0.19
0.19
0.22
0.27
0.35
0.31
0.28
0.27
2.89
2.76
2.65
2.63
2.21
2.18
2.55
0.40
0.36
0.30
0.30
0.30
0.31
0.33
a cpd 8 observed as MSA salt.
The continuous crystallization was improved by adding IPA
solutions containing 0.5 equiv. of MSA (1.14 mL min−1) con-
current with 1.0 equiv. of ketone 6 (2.22 mL min−1 in MSMPR
#1 with a 30 min residence time at 20 °C. The resulting thin
slurry containing product 1 continuously flowed to a second
crystallization MSMPR where another 0.5 equiv. of 7.8 wt%
MSAwas added at 1.14 mL min−1 with a 30 min residence time,
and pumped to MSMPR #3 with 30 min τ. The compound 1
slurry was continuously transferred to a 1 liter glass pressure
filter using another 4 valve transfer zone pump at an average
overall flow rate of 4.5 mL min−1. Two filters were interchanged
throughout the run with each filter cake washed with 3 volumes
of isopropanol based on the amount of product entering the
system while the filter was online. Six product fractions were
collected over 18 h of continuous operation and the resulting
finishing vessel (CSTR #2); quench vessel (CSTR #3) and a con-
tinuous sodium carbonate extraction (CSTR #4). The flow
diagram for the entire continuous process including the continu-
ous crystallization is shown in Fig. 5. For this campaign an
adjustment was made to the free base procedure where a silica
treatment was added to the organic solution in order to remove a
dark coloration present in the amide 3 feedstock. The silica treat-
ment effectively removed the dark coloration from the source
amide, but notably had a positive impact on the Barbier reaction
chemistry where 98% conversion was observed in the first CSTR
vessel at the 35 °C operation temperature. Conversion and ee
were monitored hourly through the duration of the campaign
(Fig. 6).
1530 | Green Chem., 2012, 14, 1524–1536
This journal is © The Royal Society of Chemistry 2012

View Article Online
Fig. 7 Microscopy of compound 1 obtained via different crystallization processes; (a) batch high temperature crystallization, (b) continuous crystalli-
zation at room temperature.
crystallization is that racemization is minimized by operating
at 23 °C.
Table 5 Batch/continuous flow comparison for production of 625 kg
compound 1 at 125 kg per week throughput
Single
batch
Total 5
batches
Continuous
Barbier
Environmental metrics comparison
THF (L)
DIBAL (L)
Mg (kg)
Initiated Cl-THP (kg)
Total reaction vol. (L)
Vessel required (L)
Total time (h)
884
17
16
4400
85
80
395
1400
2000
840
625
3480
0.07
53
0.084
10.4
15
A useful comparative metric is to review the amount and type of
materials required to produce a given amount of compound 1
produced by both the batch and continuous approaches. A com-
parison was performed between batch and continuous
approaches evaluating production of 625 kg of compound 1
based on traditional batch Grignard process (Table 5).
79
1400
2000
168
125
840
625
Compound 1 kg per week
THF usage was directly reduced by 21% by the continuous
approach which is due primarily to elimination of azeotropic
drying steps necessary in the batch process. The estimate here is
in fact conservative since in practice it may be necessary to use
significantly higher quantities of THF to obtain dry reaction
vessels (<500 ppm water) suitable for Grignard reagent for-
mation. From a materials usage perspective, not only is 10% less
4-Cl-THP used for the continuous approach, but a single
initiation event is required, whereas in batch several metal acti-
vations are required. Other hazardous materials such as diisobu-
tylaluminum hydride are also reduced by >99% due to the
necessity of only a single metal activation. Another significant
upside to the continuous approach is the reaction scale has con-
siderable intrinsic process safety advantages with a maximum
operation scale of 15 L whereas, in batch, 2000 L scale reactions
would need to be run to achieve the same endpoint of production
of 125 kg of 1 in a week. For this project it is expected that ∼50
L would be the maximum scale required to produce ∼15 MT of
compound 1 per year, which would replace 4000–6000 L batch
reactors or several 2000 L reactors in series.
product wet cakes were dried to constant weight over a minimum
of 12 h to produce 404.4 g of compound 1 with exceptional
purity (Table 3). The rest of the 6 solution was crystallized batch
for comparison.
Continuous/batch crystallization comparison
Compound 1 produced via the fully continuous process (Table 4,
entry 5) was compared with material made continuously, but
crystallized in batch mode (Table 4, entries 3 and 4) along with
the batch Barbier process (Table 4, entry 2), as well as material
made entirely in batch mode by the Grignard process (Table 4,
entry 1).
Overall the impurity compositions, between batch/continuous
and Barbier/Grignard approaches were similar except impurity 8
was slightly enriched in the continuous crystallization system
(Table 4). For example, entries 4 and 5 in Table 4 directly
compare the purity profile of product produced by the batch and
continuous crystallizations using the same feedstock from the
continuous reaction. One difference between the approaches is
that large plate like crystals of 1 are formed when the crystalliza-
tion is carried out in batch with a slow cooldown from 70 °C. In
comparison, much finer crystals of 1 are obtained by the room
temperature continuous crystallization and isolation (Fig. 7).
Since compound 1 exists as a channel solvate, the finer crystals
produced in continuous mode also predictably occluded less
residual solvent. However, a significant benefit to the continuous
Conclusions
A novel Barbier process was run in continuous mode to produce
ketone 1, a useful pharmaceutical intermediate for production of
edivoxetine·HCl. Excellent conversion (98%) and enantiomeric
purity were achieved (>99% ee) in the crude ketone 6 via a four
vessel CSTR series for the chemistry and aqueous work-up
steps. An efficient continuous reactive crystallization was also
developed using a cascade of three 250 mL MSMPRs in series
This journal is © The Royal Society of Chemistry 2012
Green Chem., 2012, 14, 1524–1536 | 1531

View Article Online
in which a production rate of 0.4 kg of MSA salt 1 per 18 hours
was demonstrated. Overall, the greenness of the process was
demonstrated to be significantly improved with reduced process
mass intensity and improved intrinsic process safety relative to a
traditional batch approach.
hydride 1 M toluene solution (3.0 mL, 3.0 mmol) was added fol-
lowed by iodine (0.75 g; 2.95 mmol) and the dark suspension
was heated to 55 °C. Dry 4-chlorotetrahydropyran (0.10 equiv.)
was added. After 10 min of stirring, a 3 °C exotherm was
observed. After stirring for an additional 20 min, the mixture
was cooled to 35 °C and the toluene Barbier solution (1.0 equiv
amide 3 and 1.35 equiv 4-Cl-THP, preparation above) was added
dropwise over at least 1 h (Note: during the Barbier solution
feed, a steady 2–3 °C temperature rise was observed). After the
feed was complete the mixture was stirred for an additional 1 h
at 35–40 °C. The reaction was considered complete when <2%
starting material remained by HPLC analysis. The supernatant
was then cannulated to an inerted addition funnel and the con-
tents were then added dropwise over at least 1 h to a stirred 0 °C
inerted 17 wt% acetic acid solution (3 equiv.). The residual mag-
nesium heel was then rinsed with toluene (2 × 30 mL), and the
rinses were cannulated into the work-up vessel through an
addition funnel (Note: the residual magnesium heel can be
reused for subsequent runs provided it is kept inert). After the
rinse was complete the mixture was warmed to room temperature
and stirred for 30 min, then the layers were separated. The
aqueous layer was removed and the organic layer washed with
10 wt% Na2CO3 (100 mL) followed by water (100 mL). The
organic layer was twice concentrated to an oil, diluted twice with
IPA (250 mL), and re-concentrated to an oil. IPA (170 mL) was
added to produce a final solution volume of ∼190 mL. The
crude ketone 6 isopropanol solution was taken directly into the
methansulfonic acid salt formation step.
Experimental
(S)-(4-Benzylmorpholin-2-yl)(morpholino)methanone
methanesulfonate, 2 (batch synthesis)
A 25 wt% solution of morpholine amide free base 3 in toluene
(preparation, ref. 8) was concentrated to an oil, then the toluene
displaced with ethyl acetate by vacuum distillation to produce a
25 wt% solution of 3 (14.7 kg). The 25 wt% solution of amide 3
(14.7 kg, 0.735 kg active) was charged into a 3 necked 22 L RB
flask under nitrogen. Methanesulfonic acid (246.5 g; 2.55 moles)
was charged slowly, at a temperature between 15–25 °C allowing
the product to precipitate slowly. Once addition was complete,
the mixture was cooled to 0–15 °C and stirred at this temperature
for 1 h. The slurry was filtered and washed with 1.5 L of cold
ethyl acetate and dried in vacuo at 36 °C overnight to produce
1
0.92 kg of compound 2 as a white powder in 93.4% yield. H
NMR (500 MHz, CDCl3) δ = 2.81 (3H, s), 2.90 (1H, m), 3.20
(1H, m), 3.4–3.7 (10H, m), 4.00 (1H, dd, J = 12.9, 3.0 Hz), 4.25
(2H, m), 4.35 (1H, dt, J = 12.9, 1.6 Hz), 7.44 (3H, m), 7.55 (2H,
m). 13C NMR (125 MHz, CDCl3) δ = 39.6 (CH3), 42.4 (CH2),
46.2 (CH2), 50.9 (CH2), 52.2 (CH2), 61.7 (CH2), 63.5 (CH2),
66.5 (CH2), 66.7 (CH2), 69.7 (CH), 127.2 (C), 129.5 (2 × CH),
130.5 (CH), 131.4 (2 × CH), 164.6 (C). HRMS m/z calcd
for C16H23N203 (M + H)+: 291.1703. Found: m/z = 291.1703
(M + H)+.
(S)-(4-Benzylmorpholin-2-yl)(tetrahydro-2H-pyran-4-yl)-
methanone methanesulfonate, 1 (batch synthesis)
Compound 6 (21.3 g active; 1.00 equiv.; 73.75 mmol) isopropa-
nol solution (190 mL) from the Barbier reaction was transferred
to a 500 mL 3 necked round-bottomed flask and stirred under
nitrogen. The solution was heated to 68 °C with mechanical stir-
ring then methanesulfonic acid (4.85 mL; 74.36 mmoles) was
added in one portion. The solution was cooled to 60 °C, seeded
(1 wt% compound 1 seeds), then stirred for 1 h at 60 °C. The
mixture was then cooled to 23 °C, stirred for 1 h and then
filtered. The resulting wet cake was washed with IPA (60 mL)
and dried overnight at 45 °C to produce 28.4 g of 1 as a white
solid in 86% yield.
(S)-(4-Benzylmorpholin-2-yl)(morpholino)methanone, 3 and
4-chlorotetrahydropyran, 4 Barbier solution (batch synthesis)
Sodium carbonate (47.0 g, 439 mmol) was dissolved in water
(425 mL) then charged, along with toluene (850 mL), into a 5 L
jacketed vessel. Morpholine amide mesylate
2 (150 g,
365 mmol) was slowly added then the mixture stirred for
60 min. The layers were separated and the lower aqueous layer
discarded. The organic layer was washed with water (140 mL),
then the aqueous layer was removed and discarded. The organic
layer was concentrated with a maximum jacket temperature of
60 °C to a volume of 200 mL. Toluene (600 mL) was then
charged followed by silica gel (37.5 g). The mixture was stirred
for 1 h then filtered. The waste silica cake was rinsed with
toluene (150 mL) and the combined filtrates were re-concen-
trated to a 200 mL volume. Toluene (485 mL) was charged and
water content measured by Karl Fischer titration to verify
<500 ppm. The wt% of 3 was determined quantitatively using
HPLC analysis (20–25 wt% target). Once the precise wt% of 3
was determined, 1.35 equiv. of dry 4-chlorotetrahydropyran, 4
was charged to complete Barbier solution preparation.
(S)-(4-Benzylmorpholin-2-yl)(tetrahydro-2H-pyran-4-yl)-
methanone, 6 (flow synthesis). Initiation
Magnesium (25 g, 10 equiv) was suspended in THF (115 mL) in
a 250 mL Argonaut flask (CSTR #1) equipped with baffle cage
and 0.5 inch outside diameter, 0.376 inch inside diameter draft
tube which contained a 0.25 inch outside diameter, 0.125 inch
inside diameter, capped transfer line with 1 mm diameter holes
drilled in concentric circles on the sides. This transfer line served
to pump the product solution out but keep magnesium seques-
tered in CSTR #1. The resulting slurry was heated to 55 °C. Di-
isobutylaluminum hydride 1 M toluene solution (1.35 mL;
9.49 mmol) was added followed by a THF solution (5 mL) con-
taining iodine (0.4 g; 1.58 mmol) producing a total volume of
(S)-(4-Benzylmorpholin-2-yl)(tetrahydro-2H-pyran-4-yl)-
methanone, 6 (batch synthesis)
Magnesium (17.95 g; 738.5 mmoles) was suspended in THF
(100 mL, 3.3 vol) in a 500 mL flask. Diisobutylaluminum
1532 | Green Chem., 2012, 14, 1524–1536
This journal is © The Royal Society of Chemistry 2012

View Article Online
120 mL in CSTR #1. The suspension was stirred at 55 °C for
15 min then 4-Cl-THP (1.62 g; 13.44 mmol; 0.1 equiv.) was
added and a mild exotherm to 59 °C was quickly observed.
After stirring for 10 min at 55 °C the temperature of the initiated
suspension was lowered to 35 °C. Magnesium was charged two
more times through the experiment, each time the same amount
of magnesium was charged as initially, 25 g each time after
>70% metal consumption. Note: The amount of magnesium
(75.04 g; 3.08 mol, 1.35 equiv.) used was calculated based on
total magnesium needed for a 47 h run in order to generate 1.35
equiv. at all times in CSTR #1 throughout the entire experiment,
with the goal of 1 mol of magnesium for 1 mol of 4-Cl-THP.
The 1.35 equiv. stoichiometry translates to 2.5 g magnesium
consumption per 90 min residence time.
Separation of the liquid layers after the carbonate wash was per-
formed batch, for simplicity of this demonstration, however, on
scale it would be a continuous gravity decanter similar to the
separator following CSTR# 3. The bottom aqueous layer was
removed and the upper organic layer was carried on to the final
water wash. The organic layer obtained from the carbonate wash
was washed as one batch with water in the ratio of half the
volume of water to the volume of organic. The bottom aqueous
layer was removed and the resulting organic layer was carried
into the solvent exchange.
Solvent exchange
Solvent (toluene–THF) in the organic layer was obtained after
the final water wash was removed by distillation using reduced
pressure in a rotary evaporator (distillation conditions: 100 mm
Hg with a water bath of 60 °C). The distillation was continued
until an amber oil remained. The oil was assumed to be quanti-
tatively ketone 6 and QS IPA was added to the oil obtained from
the vacuum distillation to prepare a 20 wt% solution of ketone 6
in IPA.
Note: This solvent exchange procedure is not intended for tra-
ditional plant modules with batch stirred tanks. It is not a vessel
ready process for batch, because minimum stir volume prohibits
distillation to an oil. However, the parameters have been demon-
strated at a 200 L per day scale in-house by running the solvent
exchange automated batch in a rotary evaporator distillation
system, where the cycle time of the charges is set to match the
throughput on the preceding and trailing unit operations. All the
intermittent flows and pressure sequences are fully automated
and precisely controlled by automated valves DeltaV control
system: reaction/extraction product solution in, distillate out,
solvent add-back, and dissolved product out. If the workup and
isolation must be done in batch due to equipment availability,
then the procedure would need to be modified so that it can
work in traditional batch distillation tanks. This also could be
worse environmentally as there are numerous solvent charge and
distillations that would be required to achieve the same consist-
ent crystallization feed stream.
Barbier reaction
After the initiation the reaction mixture was adjusted to 35 °C in
CSTR #1, the pumps were started connecting the CSTRs. A
22 wt% solution of S-amide 3 in toluene containing 1.35 equiv.
of 4-chlorotetrahydropyran (Barbier stream, preparation above)
was fed continuously with a flowrate of 1.09 mL min−1 to CSTR
#1, containing the initiated Grignard, with dual module syringe
pumps. Average hydraulic residence time in the CSTR #1 was
90 min governed by the diptube height and mixing in flask, and
the total volumetric flow rate. THF was also added continuously
via a separate syringe pump with a flowrate of 1.35 mL min−1 to
maintain a 2 : 1 THF–toluene ratio, the pump flow rates were
verified by feed cans on balances. During the course of the feed
a ΔT (Tjacket − Tprocess) of 1.5 °C was maintained. Due to the
intermittent flow with the use of the transfer zones, the range of
τCSTR#1 was from 90 to 93 min. The reaction solution was trans-
ferred to a second CSTR using a 4 valve transfer zone pump at
an average flowrate of 2.44 mL min−1; τCSTR#2 was 60 min. The
temperature of CSTR #2 was also kept at 35 °C. As with CSTR
#1, the intermittent flow kept the τCSTR#1 to range from 60 to
63 min.
Quench and aqueous work-up
A 17.3 wt% acetic acid in water solution was prepared and trans-
ferred continuously via syringe pump at a flowrate of 1.02 mL
min−1 to the third CSTR, whose jacket temperature was con-
trolled at 5 °C. Simultaneously, the reaction solution in the
second CSTR was transferred to the third CSTR using a second
4 valve transfer zone pump at an average flowrate of 2.44 mL
min−1 and the mixture stirred in the CSTR. τCSTR#3 was 30 min.
The biphasic solution in the third CSTR was transferred to a
settler/separator tank using a third 4 valve transfer zone pump at
an average flowrate of 3.46 mL min−1 and the layers were
allowed to separate continuously. Tsettler = 30 min. The lower
aqueous layer was continuously overflowed to waste using a
hydraulic decanter. The upper organic layer from the settler/
separator above continuously flowed to a fourth CSTR and a
5 wt% sodium carbonate in water solution was also being con-
tinuously added via syringe pump at a flowrate of 1.78 mL
min−1; τCSTR#4 was 30 min. This bi-phasic mixture was trans-
ferred via a 4 valve transfer zone pump to a collection bottle.
Magnesium re-initiation and re-charge
When the same heel of magnesium was to be used over a week,
the solutions in CSTRs #1 and #2 were held overnight when
Barbier production was stopped for the day. CSTR #3, the acid
quench tank, was emptied nightly to minimize racemization. The
quench tank is forward processed through the separator until the
sodium carbonate wash. After the quench solution has emptied
into the carbonate wash, the process is held overnight in CSTRs
#1, #2 and #4 with all the tanks inerted. The next day a residence
volume of acetic acid is added to CSTR #3, magnesium is re-
charged to CSTR #1 if needed, then the reactors are adjusted to
the correct temperatures, and Barbier mixture and THF streams
are started at the desired flow rates. Once CSTR #3 reaches the
residence volume, the transfer zone pump from CSTR #3 is
started up along with the acetic acid flow. Once the separator
begins to gravity drain into the carbonate wash tank, the fourth
transfer zone and sodium carbonate stream are started up again,
This journal is © The Royal Society of Chemistry 2012
Green Chem., 2012, 14, 1524–1536 | 1533

View Article Online
Fig. 8 Distillation and continuous crystallization of compound 1.
An observed ΔT (Tmax − Tinitial) of 1.5 °C will return quickly to
CSTR #1 once the streams begin and no DIBAL-H or I2 was
required for re-initiation on subsequent days. This ΔT was not
adiabatic as the reactor was being controlled with a jacket at
35 °C. It was also demonstrated that fresh untreated magnesium
could be charged into the first CSTR during the course of the
reaction, with no loss in conversion nor ee of ketone 6. Also
during the re-charge no 4-Cl-THP, DIBAL-H nor I2 was
required, the reaction mixture remained initiated and the exo-
therm seen as the amide/4-Cl-THP was fed into the first CSTR
was maintained throughout the re-charge. The demonstration run
went for 47 h, cumulative flow time not including overnight
shutdowns, put 2.199 mol into the system and came out with
1.929 mol in the oil, giving us a yield of 88%. The oil was used
directly in the continuous MSA salt formation/crystallization
process (Fig. 8).
7.82% w/w methane sulfonic acid in isopropanol was prepared
and continuously added via syringe pump at a flowrate of
1.14 mL min−1, while simultaneously adding the 22 wt% ketone
6 isopropanol solution using another syringe pump at a flowrate
of 2.22 mL min−1 to afford 0.5 mol MSA : 1 mol ketone
additions, τMSMPR#1 = 30 min in the first MSMPR crystallizer.
The compound 1 slurry was continuously transferred to the
second MSMPR crystallizer using a 4 valve transfer zone pump
at an average overall flow rate of 3.36 mL min−1; τMSMPR#2
=
30 min. The temperature of the second MSMPR was also kept
ambient. Another 0.5 mol of MSA from the 7.82% w/w metha-
nesulfonic acid in isopropanol was added at 1.14 mL min−1 into
tank 2. The compound 1 slurry was continuously transferred to
MSMPR # 3 with τ = 30 minutes, and pumped intermittently to
a 1 liter glass pressure filter using another 4 valve transfer zone
pump at an average overall flow rate of 4.5 mL min−1. Liquid
was removed from the slurry via filtration collecting mother
liquor in a tared filtrate receiver.
Two filters were interchanged throughout the run with each
filter being washed with 3 vol of IPA off line based on the
amount of ketone put through the system during the time the
filter was online. The continuous crystallization progressed until
the supply of 22 wt% ketone 6 IPA solution was exhausted at
which point the crystallizers were emptied at the same rate as
(S)-(4-Benzylmorpholin-2-yl)(tetrahydro-2H-pyran-4-yl)-
methanone methanesulfonate, 1 (flow synthesis)
A seed bed of compound 1 in isopropanol was prepared by char-
ging 2 g of cpd 1 and 50 mL of IPA to the first MSMPR crystal-
lizer and allowed to stir at ambient temperature. A solution of
1534 | Green Chem., 2012, 14, 1524–1536
This journal is © The Royal Society of Chemistry 2012

View Article Online
they were filled. Once the crystallizers were empty, 2.5 volumes
(compound 6 oil basis) of IPA was added to the first MSMPR
crystallizer as a cake rinse. The rinse solution was transferred
through the transfer zones as well as the second and third
MSMPR crystallizer and finally the rinse was transferred to the
filter where the cake was rinsed and the filtrate collected in a sep-
arate tared filtrate receiver. The wet cake in the filter was trans-
ferred to a tared drying dish and dried to a constant weight in a
vacuum dryer at 45 °C. The latest run went for 18 h and
1.156 mol of ketone was put into the system, with 1.027 mol of
salt collected (corrected for potency), giving a yield of 89%.
and cooled to 0 °C in an ice-bath. To this mixture was added
compound 6 (5.17 g; 1.00 equiv.; 17.87 mmol) dissolved in
ethanol (15 mL; 257.6 mmol) which was added dropwise at
0 °C over 3 min while keeping the temperature below 3 °C. The
mixture was allowed to stir at 0 °C for 15 min before it was
warmed to 23 °C then aged for 30 min. The organic layer was
concentrated to an oil then dichloromethane (20 mL) was added.
The mixture was placed in an ice-bath then 1 N hydrogen chlor-
ide (14.3 mL; 14.30 mmol) was added and the mixture stirred
for 20 min, then allowed to warm to room temperature. The
layers were separated then aqueous Na2CO3 was added to the
aqueous layer until basic. The aqueous layer was extracted with
ethyl acetate (50 mL) then dichloromethane (50 mL). The com-
bined organic layers were concentrated to an oil to produce 4.5 g
(88% yield) of compound 9 as a 61 : 39 ratio of diastereomers.
(S)-1-(4-Benzylmorpholin-2-yl)-2-phenylethanone methane
sulfonate, 8
1
Amide 3 (10 g; 1.00 equiv.; 34.44 mmol) was dissolved in THF
(70 mL) then cooled to 0 °C. Benzylmagnesium chloride
(16.4 mL; 0.95 equiv.; 32.80 mmol) was then added over
10 min. Once the addition was complete, the cooling was
removed and the reaction mixture warmed to 40 °C and stirred
for 1 h. The reaction mixture was re-cooled to 0 °C then was
quenched by addition of an acetic acid (2 mL, 34.9 mmol) sol-
ution in water (20 mL). Toluene (70 mL) was charged and the
mixture warmed to room temperature. The layers were separated,
then the aqueous layer was extracted with toluene (100 mL). The
organic layer was washed with 10 wt% aqueous sodium carbon-
ate (70 mL) then with water (70 mL). The combined organic
extracts were concentrated to an oil (17.46 g). The crude oil was
purified by silica gel chromatography to produce 6.48 g of
purified oil which was dissolved in isopropanol (33 mL) and
transferred to a 100 mL, 3-necked round bottom jacketed flask.
The solution was heated to 68 °C with mechanical stirring, then
methanesulfonic acid (1.45 mL; 22.12 mmol) was added in a
single portion, resulting in an exotherm to 80 °C. The solution
was cooled to 20 °C to crystallize product 8. The compound 8
slurry was stirred at room temperature for 1 h, cooled to 0 °C,
then stirred for an additional hour. The crystals were filtered, the
cake, washed with cold IPA (30 mL). The resulting wet cake was
dried in vacuo at 45 °C overnight to produce compound 8 as
Diastereomer #1: H NMR (500 MHz, CDCl3) δ = 1.29 (1H,
m), 1.33 (1H, m), 1.45 (1H, m), 1.68 (2H, m), 2.80 (4H, m),
3.09 (1H, dd, J = 6.9, 3.4 Hz), 3.27 (2H, m), 3.67 (5H, m), 3.87
(2H, m), 7.33 (1H, m), 7.36 (2H, m), 7.39 (2H, m). 13C NMR
(125 MHz, CDCl3) δ = 29.6 (CH2), 29.9 (CH2), 38.2 (CH), 53.6
(CH2), 55.5 (CH2), 63.4 (CH2), 66.9 (CH2), 68.2 (2 × CH2),
75.7 (CH), 76.3 (CH), 129.1 (CH), 129.4 (2 × CH), 130.6 (2 ×
1
CH). Diastereomer #2: H NMR (500 MHz, CDCl3) δ = 1.33
(1H, m), 1.36 (1H, m), 1.48 (1H, m), 1.55 (1H, m), 1.72 (1H,
m), 2.66 (2H, d J = 10.4 Hz), 2.93 (2H. D J = 10.4 Hz), 3.26
(1H, m), 3.29 (1H, m), 3.44 (1H, m), 3.58 (4H, m), 3.80 (2H,
m), 3.87 (1H, m), 7.28 (1H, m), 7.33 (2H, m), 7.34 (2H, m). 13
C
NMR (125 MHz, CDCl3) δ = 27.6 (CH2), 30.0 (CH2), 37.3
(CH), 53.6 (CH2), 55.5 (CH2), 63.4 (CH2), 66.9 (CH2), 68.2 (2
× CH2), 75.7 (CH), 76.1 (CH), 128.1 (CH), 129.1 (2 × CH),
130.6 (2 × CH). HRMS m/z calcd for C19H22NO2 [M + H]+:
292.1907. Found: m/z = 292.1912 [M + H]+.
Acknowledgements
The authors are grateful for the work of Brad Tuck, Bill Dise-
road, Dave Smith and Shane Glassburn for providing operational
support. The authors would also like to thank D&M Continuous
Solutions for equipment construction and John Howell for pro-
viding starting material process streams. In addition, the authors
gratefully acknowledge the contributions of Brian Scherer, Mary
K. McCauley and Mike Miller for providing analytical support
for the project.
1
white solid (5.08 g, 46% yield). H NMR (500 MHz, CDCl3) δ
= 2.65 (1H, t, J = 11.8 Hz), 2.80 (1H, dd, J = 12.9, 8.4 Hz), 3.39
(1H, d, J = 11.8 Hz), 3.63 (1H, d, J = 12.2 Hz), 3.88 (1H, d, J =
16.2 Hz), 3.94 (1H, d, J = 16.2 Hz), 4.14 (1H, dd, J = 13.2, 3.8
Hz), 4.19 (2H, m), 4.46 (1H, t, J = 12.3 Hz), 4.85 (d, 1H, J =
9.9Hz), 7.15 (2H, d J = 7.0 Hz), 7.27 (1H, m), 7.31 (2H, m),
7.45 (2H, m), 7.48 (1H, m), 7.55 (2H, m). 13C NMR (125 MHz,
CDCl3) δ = 45.3 (CH2), 50.5 (CH2), 51.4 (CH2), 61.5 (CH2),
63.7 (CH2), 76.0 (CH), 126.8 (C), 127.4 (CH), 128.6 (2 × CH),
129.4 (2 × CH), 129.6 (2 × CH), 130.6 (CH), 131.4 (2 × CH),
132.2 (C), 202.8 (C). Elemental analysis calcd for C19H22NO2.
Theory 61.36% C; 6.44% H; 3.54% N; % Found 61.36% C;
6.29% H. 3.54% N. HRMS m/z calcd for C19H22NO2 (M + H)+:
296.1645. Found: m/z = 296.1642 [M + H]+.
Notes and references
1 (a) V. Grignard, Compt. Rend., 1900, 130, 1322; (b) V. Grignard, Ann.
Chim., 1901, 24(7), 433; (c) P. Barbier, Compt Rend., 1899, 128, 110.
2 (a) H. Schickaneder, R. Loser and H. Grill, US Patent # 5,047,431, 1991,
Kilnge Pharma GmbH & Co; (b) R. McCague, J. Chem. Soc., Perkin
Trans. 1, 1987, 1011–1015, and references cited therein (c) P. C. Ruenitz,
J. R. Bagley and C. M. Mockler, J. Med. Chem., 1982, 25, 1056;
(d) A. B. Foster, M. Jatman, O. T. Leung, R. McCague, G. Leclercq and
N. Devleeschouwer, J. Med. Chem., 1985, 28, 1491; (e) P. G. Holton,
U.S. Patent #4,515,811, 1985, Syntex; (f) P. J. Harrington and
E. Lodewijk, Org. Process Res. Dev., 1997, 1, 72; (g) A. Pohland, US
Pat., #2,728,779, 1955, Eli Lilly and Company; (h) A. Pohland,
L. R. Peters and H. R. Sullivan, J. Org. Chem., 1963, 28, 2483.
(S)-4-Benzylmorpholin-2-yl)(tetrahydro-2H-pyran-4-yl)-
methanol, 9
3 G. S. Silverman and P. E. Rakita, Handbook of Grignard Reagents,
Marcel Dekker, NewYork, NY, 1996, pp. 79–87.
4 (a) Z. Ye, M. D. Johnson, T. Diao, M. H. Yates and S. S. Stahl, Green
Chem., 2010, 12, 1180–1186; (b) I. R. Baxendale, J. J. Hayward,
Sodium tetrahydroborate (0.71 g; 18.77 mmol) was dissolved in
ethanol (25 mL; 429.41 mmol) in a 50 mL round-bottom flask
This journal is © The Royal Society of Chemistry 2012
Green Chem., 2012, 14, 1524–1536 | 1535

View Article Online
S. Lanners, S. V. Lye and C. D. Smith, in Microreactors in Organic Syn-
thesis and Catalysis, ed. T. Wirth, Wiley-VCH, Weinheim, 2008, ch. 4.2,
pp. 84–122; (c) X. Zhang, S. Stefanich and F. J. Vilani, Org. Process Res.
Dev., 2004, 8, 455.
5 E. Riva, S. Galiardi, M. Martinelli, D. Passarella, D. Vigo and Rencurosi,
Tetrahedron, 2010, 66, 3242.
6 (a) B. A. Klokov, Org. Process Res. Dev., 2001, 5, 234;
(b) J. Kollonitsch, Ann. N. Y. Acad. Sci., 1965, 125, 161.
7 J. G. Van Alsten, M. L. Jorgensen and D. J. Am Ende, Org. Process Res.
Dev., 2009, 13, 629.
8 M. E. Kopach, U. K. Singh, M. E. Kobierski, W. G. Trankle, M.
M. Murray, M. A. Pietz, M. B. Forst, G. A. Stephenson, V. Mancuso,
T. Giard, M. Vanmarsenille and T. DeFrance, Org. Process Res. Dev.,
2009, 13, 209.
9 (S)-Morpholine amide mesylate 2, was selected as an API starting
material for preparation of LY2216684.HCl. The reason for the starting
material selection was that the mesylate salt is a stable crystalline solid
which is easily campaigned, whereas the freebase form 3 does not have
favorable handling properties. The preparation of morpholine amide
mesylate2 is described in the experimental along with an efficient pro-
cedure for production of free base 3.
10 5-Fluoro-2-methoxy benzaldehyde was used as the Grignard reagent deri-
vatization agent.
11 The Grignard initiation was performed with a 0.1 equiv. initiation charge
of 4-chloro tetrahydropyran using diisobutylaluminum hydride and iodine
as activators, as per ref. 8.
12 It was not necessary to re-initiate the magnesium, addition of 4-Cl-THP
resulted in an immediate exotherm indicating productive initiation.
1536 | Green Chem., 2012, 14, 1524–1536
This journal is © The Royal Society of Chemistry 2012