Controlling the exothermicity of O-arylation by evaporative cooling during the process development of fluoxetine hydrochloride

Article abstract of DOI:10.1021/op400279n

This study illustrates the optimization of the O-arylation step of fluoxetine hydrochloride (1) synthesis. In the entire process, this is the most critical step that dictates the yield and quality of the product. The highlight of the process is the concept of evaporative cooling that was employed in manipulating the above highly exothermic reaction by introducing toluene as the cosolvent. The evaporative cooling not only aided in getting an efficient procedure but also increased the yield of 1 and simplified the work-up procedure. This was a protective approach adopted for process safety, considering the worst-case scenario in the plant.

Full text of DOI:10.1021/op400279n

Concept Article
pubs.acs.org/OPRD
Controlling the Exothermicity of O‑Arylation by Evaporative Cooling
during the Process Development of Fluoxetine Hydrochloride†
Sandeep Mohanty,*,‡ Amrendra Kumar Roy,‡ S. Phani Kiran,‡ G. Eduardo Rafael,§ K. P. Vinay Kumar,‡
and A. Chandra Karmakar‡
‡Process Research and Development, Dr. Reddy’s Laboratories Ltd., API Plant, Bollaram-III, Plot No.’s 116, 126C, Survey No.157,
S.V. Co-operative Industrial Estate, IDA Bollaram, Jinnaram Mandal, Medak District, Hyderabad 502325, Andra Pradesh India
§Dr. Reddy’s de Mexico, Km 4.5 Carr. Fed. Cuernavaca-Cuautla, 62578 CIVAC Jiutepec, Morelos, Mexico
S
* Supporting Information
ABSTRACT: This study illustrates the optimization of the O-arylation step of fluoxetine hydrochloride (1) synthesis. In the
entire process, this is the most critical step that dictates the yield and quality of the product. The highlight of the process is the
concept of evaporative cooling that was employed in manipulating the above highly exothermic reaction by introducing toluene
as the cosolvent. The evaporative cooling not only aided in getting an efficient procedure but also increased the yield of 1 and
simplified the work-up procedure. This was a protective approach adopted for process safety, considering the worst-case scenario
in the plant.
increase the chances of uncontrolled exothermicity and
explosion. (iii) Longer reaction time is required.
INTRODUCTION
■
Fluoxetine hydrochloride (1) was originally developed by Eli
Lilly1 and is available under the trade name of “Prozac”, which
was introduced into the market in 1987 for the treatment of
major depression. Fluoxetine is also used to treat trichotillo-
mania where cognitive behaviour therapy is unsuccessful.
Although the molecule has gone off patent in 2001, it still
has paramount commercial value, which is evident by the fact
that the price has increased by 10% (Table 1) even though the
sales and consumption across the different geographies are
going down during the same period. Although new agents have
later been introduced into the market, this molecule still has
remarkably consistent demand in the market. It is also
marketed under the brand name “Sarafem” for premenstrual
dysphoric disorder. “Symbyax” is a combination of 1 and
olanzapine and was developed to treat the depressive episodes
of bipolar I disorder as well as for treatment-resistant
depression.
Thirty two years since the discovery of fluoxetine, many
synthetic routes to it or its enantiomers have been published.3,4
The main steps that dictate the yield and quality of 1 are O-
alkylation step (2 → 4, Scheme 1) and O-arylation step (11 →
13, Scheme 2). Many base/solvent combinations have been
reported for the O-arylation step, e.g. KOH in NMP,5 C:18-
Crown-6 as the phase-transfer catalyst in sulpholane,6 NaOH in
DMSO along with TBAB as the phase-transfer catalyst,7 NaH
in anhydrous DMA,8a−k t-BuOK in THF,9 O-arylation of 1,2,3-
oxathiazolidine-2,2-dioxides10 with NaH in anhydrous
DMSO.11a−f The above methods suffer from one or more
drawbacks, but the major drawbacks are as follows: (i) The use
of NaH at a large scale has its own disadvantages such as
handling in commercial scale and the liberation of hydrogen
gas. (ii) The reaction of a strong base with DMSO generates
highly reactive sodium methylsulfinyl carbanion12,13 (the dimsyl
ion), which on decomposition at high temperatures could
Because O-arylation is the most critical step in the synthesis
of 1, we focused our efforts on optimizing this step. The first
strategy was to screen for an alternative, compatible
combination of base and solvent. In case we fail to find a
suitable solution, then the second strategy was to focus on
handling/controlling the exothermicity of the reaction by
utilizing the concept of evaporative cooling. This would involve a
cosolvent with a boiling point close to the reaction temperature.
The heat generated in the reaction would then be absorbed by
the cosolvent (as the heat of vaporization), thereby converting
it into its vapour phase that then transfers the heat to the
cooling fluid of the condenser (Figure 2). This concept would
enable the reaction to be performed in a safe mode. The two
prevailing schools of thought for providing the basis of process
safety are preventive and protective approaches. Evaporative
cooling is a protective approach that is used for handling the
worst-case scenario, which may occur during the execution of
an exothermic process.
RESULTS AND DISCUSSION
■
Fluoxetine Manufacturing Process (Scheme 2).14
Benzaldehyde (7) was coupled with monomethylamine to
afford benzylidene methylamine (8) followed by the reduction
of imine to afford benzylmethylamine (9). The secondary
amine (9) was then subjected to the classical Mannich reaction
in the presence of formaldehyde and acetophenone in aqueous
HCl to afford 3-(benzylmethylamino)-1-phenylpropane-1-one
(10) as the HCl salt. N-debenzylation followed by the
reduction of the keto group of compound 10 in one pot
afforded 3-methylamino-1-phenylpropan-1-ol (11). The O-
Received: October 4, 2013
© XXXX American Chemical Society
A
dx.doi.org/10.1021/op400279n | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Concept Article
Table 1. Consumption and sales data of fluoxetine (irrespective of salt form)
sales (in MUSD)
2012−13 2011−12
259.6 300.3
consumption in kg
price USD/kg
price USD/kg
change %
2012−13
2011−12
change %
2012−13
2011−12
price change %
U.S.A.
−13.6
−14.2
−16.1
−7.7
−5.8
−11.2
32,472.7
15,280.6
5437
33,013.5
14,899.5
5397.7
5835
−1.6
2.6
0.7
−0.9
7
125.09
174.84
103.96
47.82
109.9
146.2
86.6
44.5
82.3
95.7
13.8
19.6
20
EU top 5
87.4
52.3
101.9
62.3
131
rest of Europe
Latin America
rest of world
worldwide
120.9
151.7
671.8
5781.9
7.4
161
14,182.1
73,154.4
13,250
72,395.7
93.49
13.6
13.8
756.5
1
108.89
Scheme 1. Original synthetic route by Molloy et al. (1982)2
Scheme 2. Synthetic route for the manufacturing of fluoxetine hydrochloride (1)
arylation of (11) with 1-chloro-4-(trifluoromethyl)benzene
(12) in the presence of NaOH in DMA at 90 °C for 20 h
afforded crude fluoxetine free base (13), which was then
converted to its HCl salt (1) in methanolic HCl. Recrystalliza-
tion from acetone and two recrystallizations from acetonitrile
afforded pure 1 with 51% overall yield.
As stated earlier, the conversion of the O-arylation step was
low along with the formation of impurities. On the basis of this
information, we envisioned that resolving the above issue would
result in a cleaner reaction to afford 13, which in turn would
reduce the number of recrystallizations at the final stage,
thereby increasing the yield and efficiency of the process. First,
B
dx.doi.org/10.1021/op400279n | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Concept Article
importantly, its safe toxicology profile.16,17 The reactions in
ACN and toluene remained incomplete, probably because of
the heterogeneous nature of the reaction mass.
Screening for a Suitable HCl Scavenger for O-
Arylation. After selecting DMSO as the solvent, we screened
for a suitable base for the reaction in order to increase the
conversion beyond 72% (entry 3, Table 2). The results are
summarized in Table 3.
Figure 1. Fluoxetine hydrochloride (1).
Table 3. Screening of HCl scavenger for O-arylation (all the
reactions were carried out at ∼100−110 °C)
HCl scavenger
product
(2)
entry solvent dilution (×)
base
equiv time (h)
1
2
3
4
DMSO
DMSO
DMSO
DMSO
2
2
5
2
Na2CO3
K2CO3
NaOH
KOH
5
5
5
5
3
3
3
3
2.77
4.35
82.03
91.36
The reaction was found to be sluggish with Na2CO3 and
K2CO3, whereas KOH afforded the best conversion at 110 °C
(entry 4, Table 3) in DMSO. Therefore, the temperature range
110−115 °C was the optimal reaction temperature for the best
conversion. It is important to mention here that the reaction
initiates at 85−90 °C and completes at the temperature range
110−115 °C. Therefore, the reaction mass was required to be
heated to 85−90 °C using an external heat source prior to the
addition of 12. The addition of 12 results in uncontrolled
exothermicity, and the temperature of the reaction mass rapidly
increased to ∼135−145 °C in the absence of a controlled
addition of 12 (entry 1, Table 5). The reactions shown in Table
3 were performed at 5 g scale, and even at this small scale, it
was highly exothermic. This prompted us for the safety
evaluation of this reaction.
To check the exothermicity, a test reaction was performed in
a reaction calorimeter by employing all safety measures. KOH
was added to the solution of 11 in DMSO under nitrogen
atmosphere, and the reaction mass was heated to 85−90 °C
(reaction initiation temperature). Later, external heating was
stopped, and 12 was added slowly over a period of 45 min (a
slower addition slows down the reaction as the temperature of
the reaction mass falls below 85 °C). During this addition
period, the temperature of the reaction mass increased up to
>135 °C. The amount of heat generated from the calorimeter
was calculated and extrapolated for a 70 kg batch (Table 4).
This exothermicity study was also attempted using differential
scanning calorimetry (DSC); however, in the absence of mixing in a
crucible, no data could be obtained (Supporting InformationSI).
The RC1e data shown in Figure 3 indicate that the heat
release was instantaneous; however, the reaction could also be
Figure 2. Concept of evaporative cooling.
a trial reaction was performed using NaOH as the base, and the
reaction was maintained at 95 °C for 20 h in DMA. The HPLC
analysis of the reaction mass showed ∼86% conversion to the
product (13), along with ∼5% of an undesired N-acyl derivative
(14).15 Although the acetylated impurity (14) could be
removed into the mother liquor during multiple recrystalliza-
tions, it resulted in an overall yield loss at the final stage. The
process development was challenging because of the inex-
pensive raw materials (12, NaOH, and HCl) and solvents
(MeOH, acetone, IPA, and ACN) used in the process, with
little scope for reducing the cost further. The only way to
increase the efficiency and reduce the cost was to increase the
conversion at the O-arylation step with a cleaner profile of 13.
Herein, we report the development of an overall efficient and
safe process with high conversion at the O-arylation step, thus
avoiding the multiple recrystallizations at the final stage.
Stage 1 of the Process Development: Screening for
an Alternative Solvent for O-Arylation. Various solvents
were screened for studying the conversion at the O-arylation
step with 10 equiv of NaOH. The reaction was monitored by
HPLC, and the % conversions to (1) are shown in Table 2.
Evidently, a polar aprotic solvent (except DMF) was required
for good conversion (Table 2). Although the conversion in
DMA was the best, DMSO appeared to be a good alternative as
there is no possibility for the formation of impurity 14 and also
because of less reaction time (entry 3, Table 2). Other
advantages of using DMSO are its ability to enhance the rate of
Williamson synthesis with better solvation effect,13 and most
Table 2. Product conversion in various solvents
a
entry
solvent
polarity index (water 10.2)
bp (°C)
time (h)
temp (°C)
% product (1)
% N-acylated impurity
b
1
2
3
4
5
6
DMA
6.5
6.5
7.2
6.4
5.8
2.4
165
20
7
90
86.44
85.39
72.53
3.31
5
4.5
b
b
DMA
165
165
DMSO
DMF
189
3
90−110
153
ND
ND
ND
ND
154
8
ACN
81.6
110.6
20
12
80
3.08
toluene
90−110
9.42
a
b
Reaction monitoring by HPLC. N-acyl impurity (14) formation in DMA due to the hydrolysis of DMA, ND = not detected.
C
dx.doi.org/10.1021/op400279n | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Concept Article
Arylation Step. Many excellent articles are available on
controlling the exothermicity of chemical reactions.20 The
concept of evaporative cooling prevails in the polymer industry,
where it is used extensively for controlling the heat of the
reaction.21 However, the most common use of evaporative
cooling is evident when any reaction is performed at the boiling
point of the solvent. Thus, the heat of the reaction is
immediately transferred to the heat exchanger/condenser by
the vaporized solvent. However, the above method is only
suitable for the ‘low-energy release potential’ reactions.18 In this
case, the rate of energy release was 23431.58 kcal/h (Table 4)
that was too high to be handled by the above methods.
Therefore, to overcome this problem, we thought of employing
a cosolvent that will serve as the internal coolant, which would
immediately transfer the heat of the reaction to the condenser
as shown in Figure 2. We selected toluene as the cosolvent
because its boiling point is 110 °C, and our desired reaction
temperature was also ∼110 °C. If this thought process works,
then it could also act as a protective safety mechanism in worst-
case scenarios in the plant. The other advantage of using
toluene in this case was to simplify the downstream process
after the aqueous work-up. The experiments with different
volumes of toluene as the cosolvent were carried out at 5 g
scale (Table 5).22
Table 4. Heat of reaction extrapolated for 70.0 kg DMSO
process
heat input
remarks
A
total enthalpy observed during the
addition in RC1e
52.47 kg
50.00 G
from Figure2
B
batch size in RC1e
C
heat of reaction per kg of KSM as per 1049.40
= (A × B) ×
the RC1e data
pilot run batch details
batch size
kJ/kg
1000
D
E
F
70.00 kg
total heat liberated (kJ)
73458.00 kJ C × D
17574.00
kcal
= E/4.18
a
G
H
time duration for heat release
rate of heat release
45.00 min
23431.58
kcal/h
= (D × 60)/E
a
Addition time of 12. Heat release was found to be the function of
addition/dosing rate.
controlled by a slow dosing of compound 12. During the
addition of the remaining compound 12, the heat release was
not instantaneous, indicating delayed exothermicity.
The data in Table 4 show that the rate of heat release was too
high to be handled by any normal jacketed reactor cooling
system, posing a threat of a runaway reaction.18 Because the
total energy release occurs in 45 min, this reaction is classified
as a reaction with ‘significant energy release potential’ or an
extremely high hazardous reaction19 and is not suitable for
scale-up as such.
The experimental data shown in Table 5 proved our
assumption that toluene could control the exothermicity of
the reaction by evaporative cooling as the temperature of
reaction mass did not exceed 110 °C. Notably, the volume of
toluene required for best heat transfer and conversion at 5 g
scale was more than ∼3 volumes (entries 4, 5, and 6, Table 5).
Stage 2 of the Process Development: Controlling the
Exothermicity by Evaporative Cooling at the O-
Figure 3. RC1e graph for DMSO/KOH process.
D
dx.doi.org/10.1021/op400279n | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Concept Article
Table 5. Effect of toluene volume (toluene as the co-solvent)
temperature of reaction mass (°C)
a
b
c
entry
solvent
DMSO
dilution (×)
step 1
step 2
step 3
time (min)
% conversion in HPLC product (2)
1
2
3
4
5
6
2
80
80
80
80
80
80
100
90
90
90
90
90
>135
110
110
110
110
110
45
90
90
90
90
90
91.36
88.31
91.6
DMSO/toluene
DMSO/toluene
DMSO/toluene
DMSO/toluene
DMSO/toluene
2:1
2:2
2:3
2:4
2:5
91.80
93.90
94.20
a
b
c
Temperature of the reaction mass before the addition of 12. External heating to the reaction mass was stopped. Final temperature of the reaction
mass.
To understand the effect of various reaction components
Table 8. ANOVA for factorial model for % conversion (after
removing the centre points) at α = 0.05
(DMSO, toluene, and KOH) on the conversion and unknown
impurity, we planned a 23 full factorial design of experiments
(DoE) with three centre points as shown in Table 6, and results
of the experiments are shown in Table 7.23
sum of
squares
mean
p-value
source
model
df
square F value prob > F
247.19
10.58
3.00
1.00
82.40
10.58
6.85
0.88
0.0173
0.3796
significant
B-
toluene
Table 6. Variables used for DoE studies and their ranges
C-KOH
BC
136.79
99.83
84.24
52.27
1.00
1.00
7.00
5.00
136.79
99.83
12.03
10.45
11.37
8.30
0.0119
0.0236
low level
high level
(+)
variables
units
(−)
residual
lack of fit
a
1
2
3
DMSO volumes wrt the weight
mL/g
1
1
2
3
6
5
0.65
0.6967
not
significant
of (4)
a
toluene volumes wrt the weight
mL/g
equiv
pure
error
31.97
2.00
15.98
of (4)
a
KOH equiv wrt (4)
cor total
331.43
10.00
a
wrt = with respect to.
Table 7. A 23 full factorial experiment (all the experiments
were conducted with 5 g batch size)
case of highly exothermic reactions, determine the batch size that
can be handled safely in the lab before conducting any experiments.
%conversion = 91.2 − 3.76 × toluene − 0.54 × KOH
Reaction variables
B:
Responses
(1)
+ 0.94 × toluene × KOH
A:
C:
%
%
entry DMSO Toluene KOH conversion impurity
Observations on % Impurity Formation from DoE
Studies. The ANOVA analysis (Table 9) showed that the %
impurity is affected by all three reaction parameters, and their
relationship is expressed by eq 2. Notably, the % impurity
increased with increasing DMSO quantity, whereas it decreased
with increasing toluene quantity. The same has been shown in
the contour plot of Figure 5.
factorial
points
1
1
3
1
3
1
3
1
3
2
2
2
1
2
87.97
88.53
76.07
81.7
0.98
2.36
0.24
0.61
3.4
2
1
2
3
6
2
4
6
2
5
1
5
93.5
6
1
5
85.41
92.89
95.55
91.29
83.39
88.41
8.01
0.84
1.90
1.38
1.42
1.43
7
6
5
8
6
5
In(%impurity) = −0.97 + 0.44 × DMSO
centre points
9
3.5
3.5
3.5
3.5
3.5
3.5
(2)
− 0.28 × toluene + 0.4 × KOH
10
11
Optimisation Plan. From the above discussion, it is clear
that increasing the toluene volume not only controlled the
exothermicity but also reduced the formation of the impurity.
Virtual Optimisation. At this point of development, it
became important for us to validate the model experimentally.
Observations from DoE Studies on % Conversion. The
initial ANOVA analysis including the centre points showed that
the curvature was not significant for % conversion. Therefore,
the ANOVA was once again calculated by ignoring the centre
points, and the results are shown in Table 8. The % conversion
was influenced by the equivalents of KOH (p-value 0.0119)
used, and it was also affected by the two-factor interaction of
toluene and KOH (p-value 0.0236); i.e., toluene itself has no
role in increasing the % conversion, but along with KOH, it
affected the conversion. The same has been expressed in the
form of regression eq 1 and as a contour plot in Figure 4.
Surprisingly, the ANOVA table does not contain the DMSO,
indicating that the 1−3 volumes DMSO do not affect the
conversion. It is evident from Figure 4 that the optimum
condition for obtaining >90% conversion requires >4.5 equiv of
KOH and >3 volumes of toluene with respect to 11. Caution: in
We decided to work with 2−2.5 volumes of DMSO,24
3
volumes of toluene, and 4.5 equiv of KOH (Table 10). As per
eq 1, the predicted conversion was >91%, and the predicted
undesired impurity was <2% at 95% confidence level. On the
basis of the above-defined constraints, an experiment was
conducted at lab scale, where 93% conversion was observed
with <2% impurity. The predicted and observed values (Table
10) are close enough to say that this model holds good.
Process Scale-up. It was felt necessary to evaluate the
condenser design and cooling fluid before the process was
scaled up. A shell and tube type condenser was selected, and
the various scenarios evaluated (using ASPEN) for selecting a
suitable cooling fluid for the condenser are shown in Table 11.
E
dx.doi.org/10.1021/op400279n | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Concept Article
Figure 4. Contour plot for % conversion with two volumes of DMSO in reaction mass.
Table 9. ANOVA for factorial model for % impurity
Table 10. Validating the model
sum of
squares
mean
p-value
reaction variables
response
source
model
df
square
F-value
Prob > F
% conversion
% impurity
8.380
1.518
3
1
2.79
1.52
3124.49
1697.98
<0.0001 significant
A:
B:
C:
A-
DMSO
<0.0001
a
a
DMSO toluene KOH observed predicted observed predicted
2.5
2.5
5
3
4.5
4.5
93.45
91.5
87−94
87−93.4
1.4
1.6−1.7
2.9−3.1
B-
3.916
1
3.92
4379.86
3295.63
<0.0001
<0.0001
toluene
1.95
a
C-KOH
residual
2.946
0.006
0.004
1
7
5
2.95
0.00
0.00
95% confidence interval.
lack of fit
0.87
0.6120 not
The basis of selecting a proper cooling fluid was to make the
significant
evaporative cooling more effective, i.e., the reaction mass
temperature should not fall below 110 °C by refluxing toluene.
Therefore, the toluene needed to be condensed at ∼110 °C
itself, and the easiest way to achieve this was to use cooling
pure
error
0.002
8.386
2
0.00
cor total
10
Figure 5. % Impurity formation.
F
dx.doi.org/10.1021/op400279n | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Concept Article
Table 11. Selection of the type of cooling fluid for the condenser
batch size, kg
70
225
UOM
I
II
III
IV
hot fluid from the reactor
cold fluid in the condenser
boil up
kg/h
220
110
110
220
110
110
700
700
110
110
inlet T
C
110
110
outlet T
C
type
−
kg/h
chilled brine
CT water
5257
30
chilled brine
18 230
−8
CT water
16 271
30
flow rate
inlet T
5745
C
−8
outlet T
C
−4
34
−4
34
condenser details
type
shell and tube
a
heat exchanged/h
calculated Area
considered area
kcal/h
m2
m2
19 471
3.5
19 471
3.5
61 952
3.5
61 952
3.5
5
5
5
5
a
Calculations are shown in Table 12
tower (CT) water in the condenser with different flow rates for
different combinations of batch size and addition time (Table
11). Also CT water was more economical than chilled brine on
a commercial scale. However, a provision was made for a
secondary condenser with chilled brine for avoiding the toluene
loss from the system.
(already present in the reaction as the cosolvent) containing the
product was separated and washed with water to remove the
inorganic impurities. This toluene layer was directly used for
HCl salt formation using IPA·HCl, resulting in crude 1, which
was then recrystallized from acetonitrile to afford pure 1. The
process is summarized in Scheme 3 and Table 13, resulting in
an increased yield (from 55% to 71%) with reduced number of
isolation and recrystallization steps.
After selecting the condenser and cooling fluid, a pilot trial of
70 kg was planned. As per calculations, 2.6 L of toluene was
used per kg of compound 11 (Tables 10 and 12). Some part of
the total heat liberated by the reaction (17574 kcal, Table 4)
would be used to raise the temperature of reaction mass from
90 to 110 °C (3644 kcal, Table 12), and the rest of the excess
heat would be absorbed by the toluene at 110 °C for its
vaporization (13930 kcal, Table 12), thereby controlling the
exothermicity. The reflux rate of toluene required for this heat
removal was ∼216 kg/h. To maintain this rate, a calculated
condenser area of 3.5 m2 for CT water with a flow rate of 5257
kg was required (Table 12). The condenser area was 5 m2, and
the CT water flow rate was maintained as per the calculation.
As stated above that 2.6 L of toluene was to be used as per
calculation, however, a slightly higher quantity of toluene (3 L
per kg of 11) was used as per DoE optimization.
The temperature profile of 70 kg batch is shown in Figure 6,
and it is evident that the above concept of evaporative cooling
worked well in controlling the exothermicity at 70 kg scale. The
temperature trend of 70 kg batch without toluene is also shown
in Figure 7, indicating that the exothermicity could also be
controlled by controlling the dosing rate of compound 12.25
This information proved very useful during the planning for the
225 kg batch.
Other Advantages of Using Toluene As the Cosol-
vent. Ease of Work-up. As per the reported procedure,
fluoxetine base (13) obtained after the O-arylation was
extracted with EtOAc, followed by three water washings and
distillation of EtOAc under vacuum below 90 °C. Prolonged
distillation of EtOAc also results in the formation of impurity
14, which further contributes towards the yield drop.
Fluoxetine base 13 was converted to its HCl salt with
methanolic HCl in methanol. The distillation of methanol
and recrystallization from acetone afforded crude 1. Crude HCl
salt 1 required two recrystallizations from acetonitrile/IPA
mixture to afford pure 1.
CONCLUSION
■
This article reports a protective approach for process safety,
wherein both evaporative cooling and dosing control worked
together in controlling the exothermicity of the reaction. This
process also eliminates the need for additional recrystallization
at the final stage, thereby increasing the yield. Later, the DoE
and lab optimisation at 5 g scale complemented the findings
from the energy calculations and helped us in developing a
robust process, which was scaled up in-plant without any safety
or quality issues.
EXPERIMENTAL SECTION
■
Preparation of Fluoxetine Hydrochloride (1): 70 kg
Batch. DMSO (140−155 L) and toluene (210−220 L) were
charged into a 2 kL SS reactor equipped with a propeller-type
agitator. This was followed by the slow addition of KOH flakes
(140 kg, 2.5 kmol) under stirring, and finally compound 11
(69.7 kg, 0.42 kmol) was added to reactor slowly. Once the
addition was completed, hot water (90−95 °C) was circulated
in the jacket, and the reaction mass was heated to 85−90 °C.
After reaching this temperature, the reactor jacket was emptied.
To this reaction mass, compound 12 (83.7 kg, 0.46 kmol) was
added slowly in ∼30 min; the reaction mass temperature was
increased to ∼110 °C. At this point, hot water was recirculated
in the jacket to control the temperature at ∼100−110 °C for
another 45 min. The reaction was completed in 30−45 min.
The reaction mass was then cooled to 60 °C and quenched
with water (400−420 L), and another lot of toluene (420 L)
was added. The reaction mass was stirred for another 30 min.
After settling, the lower aqueous layer was discarded, and the
toluene layer containing the product (13) was washed with
water (3 × 100 L). The toluene layer was separated, and the
moisture in the organic mass was removed by azeotropic
distillation in the same reactor. The reaction mass was then
cooled to 0−5 °C using chilled brine, and the pH of the
In this process, the reaction mass was quenched with water
after the completion of the reaction. The upper toluene layer
G
dx.doi.org/10.1021/op400279n | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Concept Article
H
dx.doi.org/10.1021/op400279n | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Concept Article
Figure 6. Detail of reaction temperature profile of 70 kg batch.
Figure 7. Reaction temperature trend during addition (with and without toluene). Note: In both DMSO/KOH and DMSO/KOH/toluene systems,
the addition time was the same (150 min).
Scheme 3. Improved process for 1
reaction mass was adjusted to 2 to 3 using IPA/HCl (200 L).
The resulting suspension was stirred for another 2 h, followed
by the distillation of 20% (∼124 L) of the toluene below 75 °C
under vacuum. The concentrated reaction mass was cooled to
0−5 °C, and the precipitated solid was filtered off using an
ANF and washed with toluene (310 L) to afford crude 1. The
crude product 1 was recrystallized from acetonitrile (700 L) in
a 1 kL SS reactor to obtain 46.5 kg (71.7%) of 1 as a white
powder with a HPLC purity of 99.94%.
1H NMR (200 MHz, DMSO-d6): δ ppm 2.54 (s, 3H), 9.32
(s, NH), 3.01 (t, J = 7.16 Hz, 3H), 2.24−2.29 (m, 2H), 5.74 (d,
J = 4.9, 2.8, 1H), 7.1 (d, J = 8.4 Hz, 2H), 7.58 (d, J = 8.4 Hz,
2H), 7.2−7.5 (5H); 13C NMR (125 MHz, DMSO-d6): δ ppm
32.4, 34.2, 45.23, 76.55, 116.37, 126.02, 121.57, 126.02, 116.37,
124.39, 140.07, 126.67, 128.72, 128; MS: calcd for
C17H18F3NO 309.1340 (M+), found 310.1354 (M + H+); IR
(KBr) 2961, 2783, 2732 cm−1 (aliphatic CH), 2490, 2451, 2361
cm−1 (N+H), 1616, 1518 cm−1 (aromatic CC), 1476, 1429
I
dx.doi.org/10.1021/op400279n | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Concept Article
HCl
HPLC
IPA
kcal
kJ
hydrochloride
high-performance/pressure liquid chromatography
isopropanol
kilocalorie
kilojoules
Table 13. Yield comparison between the previous and
current processes
process 1:
AcNMe2−
NaOH
process 2: DMSO−toluene−KOH
kL
kilolitre
kilomoles
potassium hydroxide
key starting material (compound 11)
methanol
mass curve
million U.S. dollars
sodium hydride
sodium hydroxide
not detected
methylpyrrolidone
heat flow curve
reaction calorimetry
room temperature (25−30 °C)
stainless steel
input output yield crop-1
crop-1
crop-2
(g)
crop-2
kmoles
KOH
KSM
MeOH
Mr
MUSD
NaH
NaOH
ND
NMP
Qr
RC1e
RT
SS
entry
(g)
(g)
(%)
(g)
yield (%)
yield (%)
1
3
4
100
100
100
115
115
116
55
150
151
149
71.77
72.2
71.2
24
23
25
7.00
6.6
55
55.5
7.2
cm−1 (aliphatic CH), 1330, 1244, 1109 cm−1 (C−O), 1164
cm−1 (C−N), 844, 699 (CH aromatic).
N-Methyl-N-(3-phenyl-3-(4-(trifluoromethyl)-
phenoxy)propyl)acetamide (14): 1H NMR (400 MHz,
DMSO-d6): δ ppm 2.1 (t, J = 8.4 Hz, 2H), 2.8 (s, 3H), 2.9 (s,
3H), 3.5 (m, 2H), 5.5 (m, 1H), 7.1 (d, J = 7.2, 2H), 7.25 (t, J =
9.2 Hz, 2H), 7.4 (d, J = 7.2 Hz, 2H), 7.45 (t, J = 9.2 Hz, 1H),
7.56 (d, J = 6.8 Hz, 2H); 13C NMR (400 MHz, DMSO-d6): δ
ppm 43.7, 46.49, 76.5, 77.3, 116.1, 126.8, 127.7, 128.6, 140.7,
160.4, 169.6; MS: calcd for C19H20F3NO2 351.1446 (M+),
found 352.1450 (M + H+); IR (KBr) 3063 (aromatic CH),
1650, 1643 (CO), 2928.68 cm−1 (aliphatic CH), 2490, 2451,
2361 cm−1 (N+H), 1614, 1516 cm−1 (aromatic CC), 1454,
1425 cm−1 (aliphatic CH), 1326, 1240, 1050 cm−1 (C−O),
836, 702 (CH aromatic).
TBAB
tetrabutylammonium bromide
t-BuOK Potassium tert-butoxide
Temp
THF
Tr
wrt
UOM
USD
temperature
tetrahydrofuran
reaction temperature curve
with respect to
unit of measurement
U.S. dollars
ASSOCIATED CONTENT
* Supporting Information
This material is available free of charge via the Internet at
http://pubs.acs.org.
■
S
REFERENCES
■
(1) Wong, D. T.; Horng, J. S.; Bymaster, F. P.; Hauser, K. L.; Molloy,
B. B. Life Sci. 1974, 15, 471−479.
AUTHOR INFORMATION
Corresponding Author
*sandeepmohonty@drreddys.com.
Author Contributions
Amrendra Kumar Roy, S. Phani Kiran, G. Eduardo Rafael, K. P.
Vinay Kumar, and A. Chandra Karmakar contributed equally.
(2) Molloy, B. B.; Schmiegel, K. K. U.S. Patent 4,314,081, 1982.
(3) (a) Theriot, K. J. U.S. Patent 5,760,243, 1998. (b) David, D. W.;
Marybeth, S. M.; Satish, K. B.; Thomas, M. K. Org. Process. Res. Dev.
2000, 4, 513−519.
■
(4) Theriot, K. J. PCT Int. Appl., 9856753, 1998.
(5) Marti, H. R.; Weber, B. PCT Int. Appl. 2000037425, 2000.
(6) Kumar, N.;, Ramana, K. V.;, Sathyanarayana, S.; Vijayaraghavan,
B. PCT Int. Appl., 2001044166, 2001.
(7) Againe, C. E.; Aracsne, T. Z.; Drexler, F.; Harsanyi, K.; Ujvari, B.;
Vargane, G. G. WO/1994/00416, 1994.
Notes
†Dr. Reddy’s Communication no: IPDO IPM-00379
The authors declare no competing financial interest.
(8) (a) Bhandari, K.; Srivastava, S.; Shanker, G.; Nath, C. Bioorg. Med.
Chem. 2005, 13, 1739−1747. (b) Khatik, G. L.; Kumar, V.; A. Nair, V.
A. Org. Lett. 2012, 14, 2442−2245. (c) Gao, Y.; Sharpless, K. B. J. Org.
ACKNOWLEDGMENTS
■
We are thankful to the management of Dr. Reddy’s laboratories
limited for giving support to carry out this work. We are also
thankful to the colleagues of analytical and process develop-
ment departments for their cooperation especially B. Srinivas
Chakravarthy and Chandra Mohan for energy calculations.
Chem. 1988, 53, 4081−4084. (d) Chenevert, R.; Fortier, G.; Rhlid, R.
̂
B. Tetrahedron 1992, 48, 6769−6776. (e) Ratovelomanana-Vidal, V.;
̂
Girard, C.; Touati, R.; Tranchier, J. P.; Hassine, B. B.; Genet, J. P. Adv.
Synth. Catal. 2003, 345, 261−274. (f) Panunzio, M.; Rossi, K.;
Tamanini, E.; Campana, E.; Martelli, G. Tetrahedron Asym. 2004, 15,
3489−3493. (g) Corey, E. J.; Reichard, G. A. Tetrahedron Lett. 1989,
30, 5207−5210. (h) Hammadi, A.; Crouzel, C. J. Labelled Compd.
Radiat. 1993, 33, 703−710. (i) Sakuraba, S.; Achiwa, K. Chem. Pharm.
Bull. 1995, 43, 748. (j) Ali, L. S.; Sudalai, A. Tetrahedron Lett. 2002, 43,
5435−5436. (k) Kumar, A.; Ner, D. H.; Dike, S. Y. Tetrahedron Lett.
1991, 32, 1901−1904.
ABBREVIATIONS
■
ACN
ANF
acetonitrile
agitated Nutsche filter
ANOVA analysis of variance
CT water cooling tower water
(9) Ljubic, G.; Cepanec, I.; Melnjak, A.; Koretic, S.; Litvi, M.;
Mikotic-Mihun, Z. Croat. Pat. Appl., 9800495, 2000.
d
day
(10) Allegrini, P.; Razzetti, G.; Rossi, R. Ital. Appl. MI0859, 2007.
(11) (a) Schulze, M. Synth. Commun. 2010, 40, 3415−3422.
(b) Wang, G.; Liu, X.; Zhao, G. Tetrahedron: Asymmetry 2005, 16,
1873−1879. (c) Xu, C.; Yuan, C. Tetrahedron 2005, 61, 2169−2186.
(d) Li, Y. Z.; Li, Z. M.; Li, P.; Wang, Q. R.; Tao, F. G. Org. Biomol.
Chem. 2005, 3, 2513−2518. (e) Fatima, A.; Lapis, A. A. M.; Pilli, R. A.
J. Braz. Chem. Soc. 2005, 16 (3a), 495−499. (f) Xu, C.; Yuan, C.
Tetrahedron 2005, 61, 2169−2186.
DMA
DMF
DMSO
DoE
equiv
EtOAc
h
N,N-dimethylacetamide
dimethylformamide
dimethyl sulfoxide
design of experiment
equivalent
ethyl acetate
hour
J
dx.doi.org/10.1021/op400279n | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Concept Article
(12) Corey, E. J.; Chaykovsky, M. J. Am. Chem. Soc. 1965, 87, 1345−
1353.
(13) Gaylord Chemical Corporation, Dimethyl Sulfoxide (DMSO)
Reaction Solvent Technical Bulletin 105, Feb. 1992. http://www.
gaylordchemical.com/index.php?page=105b-reaction-solvent-
dimethyl-sulfoxide.
(14) Current process at Dr. Reddy’s Laboratories Ltd.
(15) (a) Daisuke, W.; Hiro-o, H. J. Phys. Chem. C 2009, 113 (27),
11662−11666. (b) Constantin, I. C.; Fulga, T.; Marioara, N. Rev.
Roumaine Chim. 2007, 52 (10), 949−952.
(16) Richard, D. B. The Human Toxicology of Dimethyl Sulfoxide;
Bainbridge Medical Centre Bainbridge Island, WA. http://www.dmso.
org/articles/information/pbrobyn.html.
(17) Robert, V. DMSO, A New Clean, Unique, Superior Solvent; ACS,
2000, Annual Meeting.
(18) (a) Etchells, J. C. Org. Process Res. Dev. 1997, 1, 435−437.
(b) Barton, J. A.; Nolan, P. F. Runaway Reactions in Batch Reactors;
IChemE Symposium Series, Vol. 85; Institution of Chemical
Engineers: Rugby, UK., 1984.
(19) David, J. F. Org. Process Res. Dev. 2008, 12, 1287−1292.
(20) (a) Control of Exothermic Chemical Reactions; IChemE Training
Package; Institution of Chemical Engineers: Rugby, UK, 1993. (b)
Laird, T. SPECIAL FEATURE SECTION: Safety of Chemical
Processes. Org. Process Res. Dev. 2002, 6, 876; and Laird, T. SPECIAL
FEATURE SECTION: Safety of Chemical Processes. Org. Process Res.
Dev. 2002, 9, 951 (c) Barton, J.; Rogers, R. Chemical Reaction
Hazards, 2nd ed.; Institute of Chemical Engineers: Rugby, U.K., 1996.
(d) Skelton, R. Process Safety Analysis, An Introduction; Institute of
Chemical Engineers: Rugby, U.K., 1997. (e) Marshall, V.; Ruhemann,
S. Fundamental Process Safety; Institute of Chemical Engineers: Rugby,
U.K., 2001. (f) Gillett, J. E. Hazard Study and Risk Assessment in the
Pharmaceutical Industry; Interpharm Press: Buffalo Grove, IL, 1997.
(g) Steinbach, J. Safety Assessment for Chemical Processes; Wiley: New
York, 1998.
(21) Solor
9025−9039.
́
zano, M. Z.; Ray, W. H. Ind. Eng. Chem. Res. 2008, 47,
(22) Etchells, J. C. Org. Process. Res. Dev. 1997, 1, 435−437.
(23) Lazic, Z. R.; Design of Experiments in Chemical Engineering, A
Practical Guide; Wiley-VCH Verlag GmbH & Co: Weinheim,
Germany, 2004.
(24) Although equation 2 indicates to reduce the DMSO volumes in
order to minimize the impurity, we proposed 2−2.5 volumes for the
reaction to make sure that in the worst-case scenario (evaporation of
all toluene) some solvent is present in the system.
(25) Figure 7 shows that the maximum heat is liberated during the
addition of compound 12. Therefore, a slow addition helped in
controlling the exothermicity of the reaction.
K
dx.doi.org/10.1021/op400279n | Org. Process Res. Dev. XXXX, XXX, XXX−XXX