Kinetics for scale-up of a one-pot pathway to 5-(3-fluorophenyl)-2,4- dihydro-2,4-dimethyl-3H-1,2,4-triazole-3-thione using a hybrid model of parallel and consecutive reactions

Article abstract of DOI:10.1021/op700101g

Kinetic results are reported for the one-pot reaction of 2,4-dimethylthiosemicarbazide (1) and 3-fluorobenzoyl chloride (2) to the unisolated benzamide intermediate (3), and on to the ringclosed product, 5-(3-fluorophenyl)-2,4-dihydro-2,4-dimethyl-3H1,2,4-triazole-3-thione (4). The sensitivity of the synthetic route to competing side reactions was examined by studying the kinetics of each reaction step. A first-order kinetic Model of the hybrid consecutive and parallel reaction scheme was developed and fit to experimental data. The Model was extended to address potential mixing concerns resulting from the formation of the insoluble intermediate 3 upon scale-up where a more concentrated reaction would be used. Scale-up of the product drug substance from laboratory to 50-gal scale was successfully completed using this recently developed one-pot pathway.

Full text of DOI:10.1021/op700101g

Organic Process Research & Development 2007, 11, 877–884
Kinetics for Scale-Up of a One-Pot Pathway to
5
-(3-Fluorophenyl)-2,4-dihydro-2,4-dimethyl-3H-1,2,4-triazole-3-thione Using a Hybrid
Model of Parallel and Consecutive Reactions
David H. Louks and Sandra K. Stolz-Dunn*
Dowpharma, Designed Polymers and Latex, Dow Chemical Company, Midland, Michigan 48674, U.S.A.
Abstract:
Scheme 1
Kinetic results are reported for the one-pot reaction of 2,4-
dimethylthiosemicarbazide (1) and 3-fluorobenzoyl chloride (2) to
the unisolated benzamide intermediate (3), and on to the ring-
closed product, 5-(3-fluorophenyl)-2,4-dihydro-2,4-dimethyl-3H-
1,2,4-triazole-3-thione (4). The sensitivity of the synthetic route to
competing side reactions was examined by studying the kinetics
of each reaction step. A first-order kinetic Model of the hybrid
consecutive and parallel reaction scheme was developed and fit to
experimental data. The Model was extended to address potential
mixing concerns resulting from the formation of the insoluble
intermediate 3 upon scale-up where a more concentrated reaction
would be used. Scale-up of the product drug substance from
laboratory to 50-gal scale was successfully completed using this
recently developed one-pot pathway.
Introduction
The reaction product 5-(3-fluorophenyl)-2,4-dihydro-2,4-
dimethyl-3H-1,2,4-triazole-3-thione (4) was an early stage
pharmaceutical evaluated by Hoechst Marion Roussel for the
sometimes >30% during development of the one-pot process.
Also, intermediate 3 is nearly insoluble in this reaction system.
Previous routes that isolate 3 have produced very fine crystals.
Dilute concentrations were necessary to make a mobile slurry,
e.g., 1 g of 3 in 25 mL of toluene. To ensure an understanding
of the competing chemistry, the rates of these reactions, as well
as those of the principal reaction pathway, were sought as
illustrated in Scheme 1. Rather than a rigorous study to
determine rate constants, the desired outcome was to give a
reliable estimate of the relative reaction rates. Elaborate
sampling and quenching methods were not used. Moreover, the
two-phase system is complex and will likely be dependent upon
vessel geometry, impeller design, etc., which dictate mixing
1
treatment of depression and for memory enhancement in
2
treating Alzheimer’s disease. The reaction of 2,4-dimethylthio-
semicarbazide (1) and 3-fluorobenzoyl chloride (2) to ultimately
3
produce the drug product can be run in one reaction pot. As
shown in Scheme 1, this pathway involves two consecutive
reactions, first the coupling reaction to form the unisolated
benzamide (3) and then the ring-closure reaction to product.
Using aqueous caustic as base, 1 and 2 can be coupled by a
Schotten–Baumann procedure, and sodium hydroxide also
catalyzes the ring-closure reaction of 3 to 4. Toluene as the
solvent helps to prevent hydrolysis of acid chloride 2 and also
solubilizes the product for a simple isolation. Byproduct salt
and any hydrolysis product (sodium 3-fluorobenzoate) are
decanted in the aqueous layer. The product in toluene can be
crystallized directly or from other suitable solvents such as
4
flow, droplet size, and the like. Instead of configuring a small,
well-mixed apparatus, as is often used in a detailed kinetic study,
a conventional jacketed straight-wall reactor was used to
simulate the 50-gal scale-up vessel.
2-propanol.
With scale-up, heat-up and cool-down cycles and reactant
Results and Discussion
addition times often increase as a result of jacket and mechanical
limitations in larger equipment. Amine 1 was shown to
decompose in aqueous caustic, and hydrolysis of 2 was
1. Initial Conditions for the One-Pot Reaction. Explor-
atory reactions found that 4 could be produced in good yield
by the following procedure: charge 1, toluene, and aqueous
caustic and heat to about 80 °C; continuously add 2 over 30–60
min; after complete addition, hold at 80 °C for 60 min before
*
To whom correspondence should be addressed. E-mail: stolzdunn@
dow.com.
(
(
(
1) Kane, J. M.; Miller, F.P. U.S. Patent 4,912,095, March 27, 1990.
2) Miller, J. A. U.S. Patent 5,100,906, March 31, 1992.
3) Stolz-Dunn, S. K.; Louks, D. H.; Puga, Y. M.; Goralski, C. T. U.S.
Patent 5,723,624, March 3, 1998.
(4) Zhang, J.; Atherton, J. H.; Unwin, P. R. Acid chloride hydrolysis
kinetics are valid only for the specific agitation conditons used.
Langmuir 2004, 20, 1864.
1
0.1021/op700101g CCC: $37.00

2007 American Chemical Society
Vol. 11, No. 5, 2007 / Organic Process Research & Development
•
877
Published on Web 09/06/2007

Table 1. Typical laboratory reaction recipe to prepare 4 by
the one-pot procedure
component
amount
molar ratio
0.96
1
27.8 g
toluene
0% NaOH
NaOH)
O)
465 mL
114 mL
(27.8 g)
(111.1 g)
38.4 g
2
(
(
2.87
25.5
≡1
H
2
2
cooling to about 50 °C to begin the phase separation. A typical
recipe is shown in Table 1.
Figure 1. Hydrolysis of 2, at 41 °C in aqueous caustic/toluene
by two analytical procedures.
The decomposition of 1, the hydrolysis of 2, and ring-closure
reaction of 3 to 4 were evaluated in separate experiments. More
dilute conditions were used to control reaction exotherms, and
lower temperatures were used to slow the reactions to a
measurable rate. In order to determine rates for the overall
reaction scheme, all components were charged and the reaction
was started. The rate-limited reaction, by continuously adding
acid chloride 2, was examined later.
Results show that the reactions are generally first-order. Of
most interest is the rate of acid chloride 2 hydrolysis (k
the desired route to intermediate 3 and product 4 (k
This part thus represents the hybrid of parallel reactions (2 f
and 2 f 5) with the consecutive reaction of one of the
products to another (3 f 4). Although the individual first-order
3
) versus
1
and k ).
2
3
5–7
reaction paths are both commonly found in the literature, the
hybrid” is not. A reaction model was derived that solves for
Figure 2. Determination of reaction order for the hydrolysis
reaction of 2, from inorganic chloride analyses in Figure 1 (after
“
7
Holleran ).
all four component concentrations (2, 5, 3, and 4) versus time,
given the three rate constants and the initial concentration of 2.
2. Hydrolysis of 3-Fluorobenzoyl Chloride (2). A warm
toluene solution of acid chloride 2 was added to 1 N NaOH
and agitated to measure the rate of hydrolysis at 41 °C. The
molar ratios of (2:NaOH:H O) were (1:2.96:165). By analyzing
2
for the amount of 2 remaining in the toluene layer (LC as the
methyl ester), the disappearance over time was easily found.
Because salt is a byproduct, this conveniently allowed another
way to follow the reaction. The inorganic chloride concentration
in the aqueous layer was measured by ion chromatography in
mg/mL solution. By assuming a constant volume of the aqueous
layer, the total amount of chloride produced can be determined.
Subtracting the initial molar amount of 2 charged from the salt
produced during the course of the reaction gives the amount of
unreacted 2. This should closely resemble the profile found
previously by direct LC analysis. The two reaction profiles are
very similar as shown in Figure 1.
Figure 3. Determination of first-order rate constant for the
hydrolysis of 2, at 41 °c in aqueous caustic/toluene, using
inorganic chloride data.
a sharp curvature late in the reaction and will converge to (0,1).
As shown in Figure 2, the shape was indicative of a first-order
reaction. The contrasting shapes of curves, representative of
other reaction orders, are also shown.
The classical method of plotting ln(mol 2 remaining) versus
3
time and evaluating the linear slope ) –k , can certainly be
Because the kinetic order was unknown, the method of
8
Holleran was used, whereby a graphic plot of y ) x/t versus
x ) 1 – (c/c
order (x is the fraction of reactant consumed, where c is the
concentration at time t and c is the initial concentration). The
0
) gives a shape that is characteristic of the reaction
0
x,y plot for a first-order reaction will be linear and sloping
employed since the order n ) 1 is established. This was done
downward for the first third of the reaction. Then, it will exhibit
for the more frequent analyses from inorganic chloride (Figure
1
), with the result in Figure 3. Samples taken late in the reaction
are expected to be more prone to error: the small molar amount
of unreacted 2 was calculated as c (initial moles of 2 charged)
(
(
(
(
5) Laidler, K. J. Chemical Kinetics, 3rd ed.; Harper and Roe: New York,
1
987.
6) Moore, J. W.; Pearson R. G. Kinetics and Mechanism, 3rd ed.; John
Wiley and Sons: New York, 1981.
7) Zuman P.; Patel R. C. Techniques in Organic Reaction Kinetics; John
Wiley and Sons: New York, 1984.
0
– cCl (moles of chloride at time t). By ion chromatography, this
difference gives large percentage errors with small fluctuations
in cCl (e.g., at t ) “∞”, by the experimental analysis, cCl may
8) Holleran, E. M. CHEMTECH 1981, 500–503.
8
78
•
Vol. 11, No. 5, 2007 / Organic Process Research & Development

Figure 4. Ring-closure reaction of 3 to 4, at 41 °C in aqueous
caustic/toluene.
Figure 5. Determination of the first-order rate constant for the
ring-closure reaction of 3 to 4, at 41 °C in aqueous caustic/
toluene.
0
not exactly equal c ). Therefore, in Figure 3, the rate constant
-1
k
0
3
) 0.0334 min was determined using data from samples at
–90 min reaction time. Alternately, using the direct LC analysis
for the amount of 2 remaining, the rate constant at 0–180 min
-1
was k ) 0.0356 min .
3
Finding the hydrolysis reaction to be pseudo-first-order is
not surprising, as reactions of this type in a large molar excess
of water are often so. The effect of NaOH and H O concentra-
2
tions upon the hydrolysis rate could be determined later, if
needed.
3. Ring-Closure Reaction of Intermediate Benzamide 3
to Product 4. The reaction was run at approximately the
concentration that would be expected midway through the
coupling reaction: 2.5 equiv of NaOH and 0.5 equiv of
byproduct salt in the aqueous/toluene solvent system. Benz-
amide 3 is a fine solid that is nearly insoluble in both phases,
so its dissolution may limit the kinetic rate. The reaction began
as a stirred slurry at 41 °C using the molar ratios (3:NaOH:
Figure 6. Decomposition of 1, at 50 and 80 °C in aqueous
caustic/toluene.
H
2
O:NaCl) ) (1:2.32:166:0.48). Obtaining a respresentative
219). Samples of each phase were taken at timed intervals, and
analysis by LC quantified the amount of 1 remaining. By
summing the individual layer analyses at a given time, the total
amount of decomposition was found. Figure 6 shows the total
decomposition rate at both temperatures. After an initial
“induction period”, the rate increased, but then slowed sometime
after 480 min. The decomposition products are unknown but
could reach some meta-stable equilibrium as the end point. Over
the 120- to 480-min interval, the decomposition rates at 50 and
80 °C were 4.7% and 6.1% per hour. Thus, the rate at 80 °C
was only 1.3 times the rate at 50 °C. Although the data do not
fit well to a first-order reaction, the decomposition at 50 °C,
sample of the multiphase reaction mixture to determine the
amount of nearly insoluble 3 remaining in the reaction vessel
by LC is a difficult task. Instead, the reaction was followed by
analyzing for 4 formed in the toluene layer, since the product
is essentially insoluble in the aqueous layer and yields are nearly
quantitative. The reaction profile is shown in Figure 4. Both
phases became clear of solids after 30 min, visually indicating
high conversion of the insoluble reactant 3. By LC, the amount
of 4 formed after complete reaction (at 180 min) was 37.0
mmol, whereas the total amount of 3 charged to the reaction
was 36.5 mmol. The reaction fits first-order kinetics, as a plot
of ln(mmol 3 unreacted) versus time, derived from the formation
rate of product, gave a straight line for the majority of the
reaction.
-1
4
about 0.001 min (k ), is at least an order of magnitude slower
than the others that were determined in this study. Still, care
must be taken to prevent prolonged heating of 1 in aqueous
caustic before starting the coupling reaction.
Figure 5 shows the linear least-squares line, giving k )
2
-1
0
.255 min .
. Decomposition of 2,4-Dimethylthiosemicarbazide (1).
A 0.76 wt % solution of 1 in toluene was stable when held at
0 °C for 48 h. Rotary evaporation yielded the near quantitative
5. First-Order Kinetic Model for the Hybrid Parallel and
Consecutive Reaction Scheme. Of the principal and competing
reactions studied so far, the decomposition of 1 is nearly
insignificant in comparison to the other rates. This makes the
coupling and competing hydrolysis of acid chloride 2 and
the ring-closure reaction of 3 to 4 the primary interest in Scheme
1 for these first-order or pseudo-first-order reactions. This route
is the combination of parallel reactions, with consecutive
reaction of one of the products to a new product. For this Model,
the concentration versus time profile is derived for the four
components of Scheme 1.
4
8
recovery of solid, and LC analysis showed high assay. In the
presence of aqueous caustic, however, decomposition occurred.
Amine 1 was heated and stirred under nitrogen, in a two-phase
system of 2.93% aqueous caustic and a half-volume of toluene.
Initial samples from two experiments, at 50 and 80 °C, showed
that about 92% of 1 resided in the aqueous layer. The molar
ratios used in both experiments were (1:NaOH:H O) ) (1:2.94:
2
Vol. 11, No. 5, 2007 / Organic Process Research & Development
•
879

The time-dependent solution for the concentrations of 2 and
k2
ln
4
are readily obtained from the differential equations of the
(
)
k1 + k3
6
parallel reaction. (To avoid confusion with numerals, com-
pounds 2, 3, 4, and 5 in Scheme 1 are designated as A, B, C,
and D, respectively.)
tmax[B]
)
(10)
(11)
k - (k + k )
2
1
3
if (2k
1
+ k
3
) > k , then [A] ) [B] at
2
The time-dependent concentrations of A and D are
) A0 e-(k1+k3)t
(1)
(2)
-k1
[
A
]
ln
(
)
k - 2k - k
3
2
1
t[A])[B]
)
k3A0
k - (k + k )
-(k +k )t
1
3
2
1
3
D
[ ]
) D +
(1 - e
)
0
k1 + k3
6
. Kinetics of the One-Pot Pathway to 4. The one-pot
process was run by adding a solution of 2 in toluene to a slurry
of 1 in 1 N NaOH at about 41 °C. See Table 2 for the amounts
added. By analyzing both the aqueous and toluene layers, the
amounts of 2, 4, and 5 were determined. The product layer was
quite clean by LC, with only three small unknown peaks totaling
about 1 area %. The amount of insoluble intermediate 3 was
determined indirectly by [3] ) [total inorganic chloride] – [5]
The time-dependent concentration of B is
dB
dt
)
k A - k B
(3)
(4)
1
2
dB
dt
1 3
-(k +k )t
)
k1A0
e
- k2B
–
[4]. From the rate of disappearance of acid chloride 2, the
first-order rate constant k ) (k + k ) could be determined as
k ) 0.129 min , shown in Figure 7. Alternately, the formation
1
3
This equation was solved by introducing the integrating factor,
-1
-1
rate of 5 gave k ) 0.119 min by evaluating the slope of ln(2final
– 2) versus time over the 0–21 min interval. Salt as the
unreactive byproduct of both coupling and hydrolysis of 2
allows one to determine the formation rate of 3, irrespective of
the ring-closure reaction. Because 5 was determined and total
salt was known, the total amount of 3 produced by coupling is
[3] ) [total inorganic chloride] – [5]. Plotting ln(3final – 3) versus
∫Pdt
F ) e
,
P ) k2
(5)
By multiplying each side of the dB/dt equation by F, the solution
can be found as
k1A0
-(k +k )t
1
3
-k2t
B
[ ]
)
e
+ K e
(6)
-1
time at 0–21 min gave k ) 0.134 min . The variation in k
k - (k + k )
2
1
3
values are likely a reflection of the different analytical
determinations.
where K is the constant of integration. This constant can be
defined by specifying the initial conditions, at t ) 0, B ) 0
Since the ring-closure reaction went cleanly to completion,
the final ratio of (4/5) gives the ratio of (k
1
/k
3
) ) (22.9 mmol/
(
i.e., B
0
) 0). This yields the solution as:
9
-1
1
)
0.7 mmol) ) 2.14. From (k
1
+ k
3
) ) 0.129 min and (k
1
/k
) 0.0411
min . Using the previously determined value of k ) 0.26
3
)
-
1
2.14, solving yields k
1
) 0.0879 min and k
3
k1A
0
-1
-(k +k )t
1
3
- e-k2t)
2
B
[ ]
)
(
e
(7)
-1
k - (k + k )
min , the reaction profile by the kinetic Model can be overlaid
onto the actual data. Figure 8 shows a reasonable fit. A better
2
1
3
-1
By the conservation of mass, the concentration of C can be
fit was obtained with k , k , k ) 0.11, 0.26, 0.05 min ,
1
2
3
found by difference:
respectively. Inferences from the Model aid our understanding
of the one-pot pathway:
[
C
]
) C + A + D -
[
A
]
-
[
B
]
-
[
D
]
(8)
• The reaction is essentially complete after 60 min at 41 °C.
The common practice of adding 2 over 30–60 min and holding
for another 60 min at 80 °C certainly ensures completion.
Reactions at >40 °C are only necessary in more concentrated
systems, where higher temperatures can ensure solubility of the
product in the toluene layer and byproducts (especially 5) in
the aqueous layer. Fortunately, little product decomposition has
been observed at 80 °C, and decomposition of 1 is manageable.
0
0
0
k3
k1 + k
(
1 - e k1+k3)t) 1 -
-
(
-
C
[ ]
) C + A0
0
{
(
)
3
k1
(
e-(k1+K3)t - e-k2t)
(9)
(
)}
k - (k + k )
2
1
3
The four concentrations versus time can easily be plotted in
(
9) A similar reaction at 41 °C (sampled less frequently) gave [4/5]240nm
(27.9 mmol/5.62 mmol) ) 4.96. From the disappearance of 2 over
a personal computer’s spreadsheet by specifying initial con-
centrations and the three rate constants, and choosing appropriate
time intervals. By changing any of the rate constants in the
Model, instant visual feedback is obtained by graphing the
predicted reaction profile versus the experimental data. Other
useful equations from the Model include:
)
-
1
15–45 m, the approximate rate constants were k ) 0.11 min , k1 )
-
1
-1
0
.092 min , and k3 ) 0.018 min . This lower amount of hydrolysis
was more typical of exploratory reactions. Experiments in a round-
bottomed flask at 70–80 °C gave yields of 74–83% of 4 to the toluene
layer when using 3 equiv of 1 N NaOH. This implies (k1/k3) ) 2.9–
4
.9, assuming all yield loss was from hydrolysis of 2. When using 3
equiv of 20% NaOH, yields were 94–95%, suggesting (k1/k3) ) 16–
time at maximum [B]:
19.
8
80
•
Vol. 11, No. 5, 2007 / Organic Process Research & Development

Table 3. Laboratory reaction recipe for the concentrated,
rate-limited one-pot procedure to prepare 4
component
amount
molar ratio
0.98
1
28.7 g
toluene
0% NaOH
NaOH)
O)
465 mL
138.8 g
(27.8 g)
(111.0 g)
38.8g
2
(
(
2.84
25.2
≡1
H
2
2
in the Model, the maximum concentration of the insoluble
intermediate occurs after about half of 2 has been reacted. With
the maximum of 25% of 3 formed (solids), this again implies
that a heavy slurry is not likely, but instead the reaction will
become more dilute in solids as it proceeds.
Figure 7. Determination of the first-order rate constant k )
+ k ) from the one-pot reaction at 41 °C in aqueous caustic/
toluene.
(k
1
3
7. Rate-Limited Kinetics (Continuous Addition of 2) for
the One-Pot Pathway to 4. In practice, the one-pot pathway
is run more concentrated at a higher temperature, with continu-
ous addition of 2 over about 30 min. A compromise temperature
of about 60 °C was chosen, half-way between the kinetic model
experiment and the typical 80 °C reaction temperature. Table
3
shows the reaction charge. The liquids (except 2) were
preheated and poured to the jacketed straight wall reactor at 60
C. 1 was added, and the contents were stirred for 20 min before
°
continuously adding 2 over 27 min to initiate reaction. At 400
rpm, large aqueous layer droplets and undissolved solids of 1
were dispersed throughout the toluene phase. The reaction
11
temperature was about 60–61 °C, although the initial exotherm
after 10 min raised the temperature to 62.5 °C, and overcom-
pensation of cooling gave a minimum temperature of 57.5 °C,
Figure 8. One-Pot reaction to 4, at 41 °C in aqueous caustic/
20 min later. By sampling at 10 min intervals, the combined
toluene: overlay of kinetic model (lines) and experimental data
aqueous/toluene layer mass balance gives the reaction profile
in Figure 9a and b. Note that the reaction proceeded in high
yield to 4, with much less hydrolysis of 2. The ratio of 4/5 at
(
symbols).
Table 2. Laboratory reaction recipe for the kinetic
evaluation of the one-pot procedure to prepare 4
85 min was (227 mmol/10.9 mmol) ) 20.8 and was typical
component
amount
molar ratio
1.00
for reactions charged with 3 equiv of 20% NaOH (see ref 9).
Acid chloride 2 appeared to react as fast as it was added,
because only a trace (about 0.1 mmol) was detected by LC in
the 30- and 40-min samples. By changing to a linear addition
rate of 2, formation of products from 2 now follow a pseudo-
zero-order mode. Over the first half of the reaction, however,
a first-order plot of ln(mmol unreacted 2) versus time can be
approximated by a straight-line fit. Thus, reacting at the linear
feed rate over the first half of the reaction would suggest that
1
4.02 g
toluene
400 mL
100 mL
(4.0 g)
(100.1 g)
5.35 g
1
N NaOH
NaOH)
O)
(
(
2.96
165
≡1
H
2
2
•
The ring-closure reaction is over twice as fast as the
coupling reaction, so substantial amounts of the insoluble
intermediate 3 never build up. The Model shows that the
maximum concentration of 3 is about 25% of the eventual total
of product 4. This assures that a heavy slurry in the two-phase
liquid will never be present which allows the use of more
concentrated reaction mixtures. The original process, employ-
ing two consecutive reactions, was necessarily more dilute
because a stoichiometric amount of fine, insoluble 3, with its
associated agitation problems, is present at the end of the first
reaction.
-1
the apparent first-order rate constant is about 0.05 min . See
Figure 10.
We re-emphasize that k ) 0.05 min- is a fictitious rate
constant, which allows the first-order kinetic model to ap-
proximate the zero-order linear addition rate of 2 over the first
half of the reaction. It is dependent upon the feed rate. The
1
10
-1
constant (k ) 0.05 min ) shows that the addition rate at 61 °C
is much slower than the intrinsic first-order rate, estimated at k
-1
≈
0.52 min (by the rule-of-thumb doubling of the reaction
•
When the one-pot pathway is run at higher reactant loading,
(
11) The heat of reaction for the coupling and neutralization (Scheme 11)
was estimated as moderately exothermic (about -60 kcal/mol). The
ring-closure reaction was estimated as mildly exothermic (about -2
kcal/mol). Calculated results were obtained using the computer
program CHETAH, which predicts heats of reaction in the gaseous
phase, and by data from analogous reactions.
the reaction starts with a slurry of undissolved 1. From tmax[B]
(
10) As mentioned, the relative kinetic rates are dependent upon agitation
droplet size, etc.) and mass transfer rates (particle size, dissolution
rates, etc.).
(
Vol. 11, No. 5, 2007 / Organic Process Research & Development
•
881

Figure 9. (a) Rate-limited one-pot reaction at about 61 °C in
aqueous caustic/toluene. Overlay of kinetic model (solid lines
estimating the first half of the reaction, where compounds 2, 3,
4
and 5 are designated as A, B, C and D, respectively) and
experimental data (dashed lines and symbols). (b) Expanded
scale.
Figure 11. (a) Rate-limited one-pot reaction with poor agitation,
at about 60–65 °C in aqueous caustic/toluene: combined
aqueous/toluene layer analyses (neglects solids at wall, 10–70
min). (b) Expanded scale.
reaction and lag behind for the remainder. This determination
(the model assumes that all of 2 was initially added and reacts
at the given rate) is shown in Figure 9a. One point that the
model clearly illustrates is that by continuously adding 2, the
maximum amount of intermediate 3 formed is substantially less,
shown in Figure 9b, and occurs after only about 15% of 1 has
reacted. The coupling reaction produces 3 at a slow rate (by
the limited addition rate of 2), while the ring-closure reaction
is relatively fast. The maximum intermediate amount is only
4% of the eventual total of 4 formed.
8. Rate-Limited Kinetics (Continuous Addition of 2) for
the One-Pot Pathway to 4 under Conditions of Poor Mixing.
The reaction in Part 7 (same feed amounts as Table 3) was
repeated, but intentionally with poor agitation by reducing
the stirring rate from 400 to 200 rpm. Before adding 2, the
interface between the swirling bottom aqueous layer and the
top toluene layer was evident. Most of the undissolved 1 lies
just above the liquid interface, although some fines were
suspended to make a weak slurry throughout the toluene layer.
Six minutes after starting the addition of 2 (dropwise addition
to the toluene layer), the temperature had increased to 65 °C
and solids coated the vessel wall of the toluene layer. After
about 20 min, the acid chloride 2 feed pump malfunctioned
after adding three-fourths of the charge. The remaining one-
fourth was added all at once at about 23 min. The vessel was
sampled at 10 min intervals for 70 min. Because the solids still
coated the wall in the toluene layer after 90 min, the contents
were then well mixed at 400 rpm before ending the experiment
5 min later. Figure 11a and b shows the reaction profile. A
Figure 10. Deviation from first-order kinetics for the one-pot
reaction with a linear feed rate of 2.
rate for each 10 °C increase, k(61°C) ≈ 4 × k(41°C) ) 4 × 0.13 )
-1
0
.52 min ). Zero-order kinetics will govern, and 2 will react
nearly as fast as it is added. Conversely, if 2 had been added in
only 2 min instead of 27 min, the fictitious first-order rate
-1
constant would be calculated as k ) 0.65 min , which suggests
that first-order kinetics may be dominating (i.e., 2 is being added
faster than it can react).
With these reservations, the first-order kinetic model can still
be used as a predictive tool in estimating component concentra-
tions at 0–50% conversion. The kinetic rate of the ring-closure
reaction is unaffected by the addition rate of 2 and should be
faster than the 0.26 min- rate at 41 °C (e.g., the rule-of-thumb
doubling for each 10 °C increase). By approximating these rate
1
-
1
-1
constants as k
2
) 4 × 0.26 ) 1.04 min , k ) 0.05 min , and
k
1
/k ) 20.8, the model should approximate the first half of the
3
8
82
•
Vol. 11, No. 5, 2007 / Organic Process Research & Development

Table 4. Comparison of well-mixed and poorly mixed
was vacuum distilled to remove most of the solvent. 2-Propanol
was added, and the product was crystallized, yielding 24.5 kg
from three batches in 80.2% overall isolated yield from 1, the
limiting reagent. See Table 6. The low isolated yield of batch
1 was attributed to poor execution of the phase separation,
wherein some of the product toluene layer was discarded with
the aqueous layer.
reactions for the one-pot pathway to 4
from Figure 9
from Figure 10
mixing
good
good
85 min
94.3
poor
70 min
90.0
3.8
good
95 min
92.9
sample after
60 min
93.9
4.0
a
%
%
yield to 4
5 formed
b
4.5
3.0
ratio [4/5]
23.2
20.8
23.6
30.7
a
Mol % based on 1 charged. b Mol % based on 2 charged.
Experimental Section
The following raw materials were used in the kinetic studies:
amine 1 was supplied by Hoechst Marion Roussel and was the
same material used in the Pilot Plant campaign. However, the
decomposition study of 1 used freshly synthesized material.
Acid chloride 2 was also material used in the Pilot Plant
campaign, manufactured by Hoechst. Benzamide 3 was pro-
duced in-house, and purity by LC was >99.5 area %, with 0.1
area % 4. For molar calculations, a 100% assay was assumed
for all materials. Other common reagents included toluene
Table 5. Pilot Plant reaction recipe for the concentrated,
rate-limited one-pot procedure to prepare 4
component
amount
molar ratio
0.99
1
5.44 kg
91.6 L
Toluene
8.6% NaOH
NaOH)
O)
1
(
(
(5.44 kg)
(23.9 kg)
7.35 kg
2.94
28.6
≡1
H
2
2
(Fisher Scientific, ACS grade), 0.995–1.005 N NaOH (Fisher
Scientific, certified), and 20% NaOH (Red Bird Service, reagent
grade).
good yield of 4 was still produced, 90.0% of the theoretical
amount (at 70 min). Note also over the 30- to 70-min interval
the slow disappearance of 2 and formation of 4 and 5. Solids
taken from the wall at the 10-min sample were dried and
indicated a composition of about 24 wt % 1, 31 wt % 3, 7.3 wt
General Kinetic Study Procedure. A nominal 1-L jacketed
straight-wall reactor with a bottom drain valve was most often
used for the kinetic experiments. The vessel i.d. was about 3.9
in., and the height from the bottom to the ground glass joint
was about 5.5 in. The openings in the vessel head, near the
perimeter and 120° apart, were fitted with a thermowell, a 0.5-
in.-wide baffle, and a Claisen adapter to a nitrogen purge and
a condenser. The agitator consisted of a three-tiered configu-
ration of four-bladed impellers, 2.5 in. in diameter. The bottom
two were pitched 45° to pump downward, while the top impeller
blades were mounted flush for radial flow. The unagitated
volume required to successively submerge each impeller was
%
4, 6 wt % unknowns, and perhaps 32 wt % salt (by difference
from 100% LC accountability).
The following scenario may explain the results. Coupling
of 1 and 2 occurred in the upper toluene layer to form insoluble
intermediate 3 and byproduct salt or perhaps also hydrochloride
salts. Amine 1 is apparently basic enough itself such that caustic
is not even needed to cause coupling. The reaction products
and some soluble 1 in the toluene layer then formed a rind on
the cold vessel wall as the reaction exotherm warmed the
toluene layer. Contact of the solid intermediate 3 with the
aqueous layer allowed the ring-closure reaction to proceed,
except for intermediate coating the wall. The last 12% of
product (over 30–70 min) was formed slowly, perhaps limited
by dissolution of 1 off the vessel wall or by transfer across the
aqueous/toluene layer interface. By increasing agitation at the
end, contact of the rind with the aqueous layer allowed
dissolution of the salt and contact of 3 with caustic for the
transformation to 4. Table 4 compares this reaction to the
previous well-mixed one.
50, 330, and 500 mL. For experiments with a constant addition
of 2, a piston pump with 0.125-in. o.d. poly(tetrafluoroethylene)
tubing was used. The tubing was routed through the condenser
opening and gave a dropwise addition to the liquid surface. An
immersion heater with a built-in pump circulated “constant-
temperature” water from a 7-L insulated reservoir, through the
vessel jacket, and back to the reservoir.
The most common sampling procedure involved removing
the thermowell and withdrawing about 5–8 mL of the agitated
reaction mixture with a pipet to a sample vial. Vials and pipets
were taken from an oven at about the temperature of the study.
Upon momentary phase separation in the vial, portions of each
layer were removed to separate vials. Methanol was im-
mediately added to the toluene layer sample to convert unreacted
The uncatalyzed coupling reaction (without NaOH added)
was verified by reacting 1 and 2 in toluene/water at about 80
12
°
C. The solids from the reaction were filtered and found to
be composed of 99.0% of 3, 0.50% of 1, and 0.46% of 4. The
overall yield to 3 was 78%.
2
to its methyl ester for LC analysis. Further dilution with
mobile phase (acetonitrile/water/2 M sodium dihydrogen phos-
phate ) 25.0/72.5/2.5, v/v/v) and LC injection were done as
soon as practical for analysis on a Supelcosil LC-18-DP column
9. Scale-Up of the One-Pot Pathway to 4 in 50-gal
Equipment. The one-pot process was run in our Pilot Plant, in
a 50-gal vessel using the concentrated recipe in Table 5.
Sampling of the toluene layer revealed that yields to 4 were
(Supelco, Inc.). A gradient elution profile was used, and known
compounds were standardized by an external standard method.
Aqueous layers were analyzed by LC and ion chromatography.
The study for the decomposition of 1 used a 500-mL flask,
heated with a mantle and mixed with a magnetic stirring bar.
The flask was equipped with a thermometer, condenser, and
nitrogen bubbler.
1 3
about 92–95% and indicated k /k ) 16–25. The toluene layer
(
12) To a mixture of 3.70 g of 1, 62 mL of toluene, and 15 mL of deionized
water in a stirred, 100-mL flask was added 5.1 g of 2 over 10 min at
0–88 °C. After 2.2 h at about 80 °C, the mixture was cooled and
filtered. The solids were washed with 2-propanol and dried to give
.9 g of fine, white crystals.
8
5
Vol. 11, No. 5, 2007 / Organic Process Research & Development
•
883

Table 6. Scale-up of the one-pot procedure in a 50-gal vessel
4
molar ratio [4/5]
after reaction
% yield of 4 (from 1)
isolated
(dry product), kg
% isolated yield
(from 1) of 4
batch
in toluene layer
1
2
3
24.6
15.6
23.0
95.0
91.8
94.1
7.38
8.66
8.48
72.4a
85.0
83.2
a
Approximately 82%, when corrected for product loss during phase cut.
General Scale-Up Procedure. The reaction and crystal-
2
ring-closure reaction (k ) is more than twice as fast as the
lization were conducted in a jacketed 50-gal glass-lined vessel
using reactant loading as in Table 5. The vessel was baffled
with a radial-flow (retreat-curve) impeller located near the
bottom and had demonstrated good mixing of toluene/water
phases. The general procedure involved adding toluene and 1,
heating to 50 °C, followed by the addition of the caustic solution
and heating to 80 °C. Acid chloride 2 was added over about
coupling reaction (k ), and this prevents significant quantities
1
of the insoluble intermediate 3 from being present. The
maximum amount of 3, about 25% in a batch reaction, occurs
after about half of 2 has been reacted. In practice, the reaction
is run more concentrated at 80 °C, with the continuous addition
of 2 to control the reaction exotherm as it reacts almost
instantaneously. The added benefit of this procedure is that
45 min at 78–84 °C, and the contents were held at about 80 °C
1 2
whereas k is kinetically limited by the reactant feed rate, k is
for another 60 min. The contents were cooled to 50 °C and
sampled. The bottom aqueous layer was decanted, and the
remaining organic layer was washed with water (40 kg) and
decanted. The washed toluene layer was vacuum distilled to
remove most of the solvent, 2-propanol was added, and the
product was dissolved and polish filtered at 70–75 °C. Product
not. Though 3 is not easily quantified under reaction conditions,
the model predicts that this reduces the amount of insoluble
benzamide 3 even more and belies concern for poor agitation
from thick slurries. Under conditions of poor mixing, the
reaction products can form a rind on the vessel wall, causing
potential loss of cooling capacity. The one-pot procedure was
successfully scaled up in a 50-gal vessel to yield 4 in >92%
yield. After crystallization, three batches produced 24.5 kg of
product 4 in 80% overall yield from 1.
4
was crystallized by cooling, centrifuged, washed, and dried,
producing a total of 24.5 kg from three batches in 80.2% overall
isolated yield from 1. See Table 6.
Conclusion
Acknowledgment
A kinetic model for the combination of consecutive and
parallel first-order reactions was developed to provide a
fundamental understanding of component concentrations with
time. Experimental data were generated to investigate the
relative importance of each reaction pathway in Scheme 1.
Because the system is multiphase, the kinetic model can not
be extrapolated to systems where the agitation results in
defferent interfactial areas from those used in this study.
This work was done under contract for Hoechst Marion
Roussel (now Aventis), whose kind permission for publication
is appreciated. Yolanda Puga provided engineering support for
the 50-gal scale-up. Some of the analytical results were provided
by V. T. Turkelson and S. R. Erskine, Analytical Sciences
Department, Dow Chemical Co., Midland, MI. Heat of reaction
calculations (ref 11) were provided by R. J. Zondlak and D. J.
Frurip, Analytical Sciences Thermal Group, Dow Chemical Co.,
Midland, MI.
The k
smaller than k
great concern. The ratio of the undesired hydrolysis of 2 to
coupling (k /k ) was affected by caustic strength. The amount
4
decomposition rate is at least an order of magnitude
1
and k , so loss of 1 during reaction is not a
3
3
1
Received for review May 11, 2007.
OP700101G
of 5 was reduced from 17% to 32% (based upon 2 fed) to 4–5%
by using 20% NaOH instead of 3.8% NaOH. The rate of the
8
84
•
Vol. 11, No. 5, 2007 / Organic Process Research & Development