Safe, convenient ortho-Claisen thermal rearrangement using a flow reactor

Article abstract of DOI:10.1021/op200162c

The [3,3] Claisen rearrangement is a well-known reaction that has been very useful for the synthesis of o-allyl phenols. The thermally induced rearrangement could present safety and operational issues at large batch scale. Herein, we report a process that

Full text of DOI:10.1021/op200162c

ARTICLE
pubs.acs.org/OPRD
Safe, Convenient ortho-Claisen Thermal Rearrangement
Using a Flow Reactor
,†
†
†,§
‡
‡
Juan A. Rinc ꢀo n,* Mario Barberis, María Gonz ꢀa lez-Esguevillas, Martin D. Johnson, Jeffry K. Niemeier,
‡
and Wei-Ming Sun
†
Centro de Investigaci ꢀo n Lilly S.A., Avda. de la Industria, 30, Alcobendas-Madrid 28108, Spain
‡
Eli Lilly & Company LTC-South, Indianapolis, Indiana 46285-4813, United States
S Supporting Information
b
ABSTRACT: The [3,3] Claisen rearrangement is a well-known reaction that has been very useful for the synthesis of o-allyl phenols.
The thermally induced rearrangement could present safety and operational issues at large batch scale. Herein, we report a process
that utilized a tube reactor to make 80 g of an early phase intermediate in a short time while mitigating the potential chemistry
hazards. Thus, both project material demands and flow chemistry proof of concept were achieved.
’
INTRODUCTION
as shown in Figure 1. The secondary reaction was not identified
but is believed to be due to decomposition of product.
With a reaction temperature of 220 °C a moderate exotherm
could cause the temperature to rise to the boiling point of
diphenyl ether (259 °C) and cause vaporization. If venting were
not sufficient the reactor could increase in temperature and
pressure and the secondary reaction could initiate. The heat of
reaction of 366 J/g for the desired reaction would be sufficient to
cause an adiabatic temperature rise of about 180 °C if the reaction
were run neat. These factors indicate a potentially hazardous or at
least difficult to control process in the case of a large-scale batch
campaign, so we considered these data significant enough to change
the process conditions.
During the development of one of our research projects,
phenol 1 was identified as a key piece for one of the most advanced
platforms. The 3-step sequence involved a [3,3] sigmatropic shift
of the allyl aryl ether (2) leading to the o-allyl phenol (3), which
was easily hydrogenated to the desired product (Scheme 1).
1
Although the original procedure described this rearrange-
ment using neat 4-(allyloxy)acetophenone (2) at 230 °C, our
medicinal chemists found that the use of diphenyl ether as solvent
facilitated the handling and the isolation of the product at 15ꢀ20 g
scale. However, the operational issues associated with the high
temperature of the process and the range of concentration used
(
(
>25% w/w) made us look for an alternative reactor technology
ART) that would provide a large amount of this material in a
Since the discovery of the Claisen rearrangement almost one
3
safe and timely fashion.
century ago, it has been focus of attention by many research
4
groups. Many synthetic methods involving Al(III) or bismuth(III)
5
6
7
derivatives, lanthanides triflates, clays or zeolites, or more
’
RESULTS AND DISCUSSION
8
recently, gold as catalysts for the sigmatropic rearrangements of
Batch Chemistry. The commercially available 1-(4-hydroxy-
phenyl)ethanone was transformed in high yield to the allyl ether
allyl aryl ethers have been reported.
We tried some of these additives to see if we could run the
reaction at lower temperatures. Unfortunately, none of the con-
ditions tested led to the selective formation of desired product,
and in many cases, degradation to byproduct were observed
2
(
2) in a similar manner as previously described in the literature.
Then, a mixture of the resulting oil crude and biphenyl ether was
heated to 250 °C for the thermal rearrangement, and the o-allyl
phenol (3) had to be isolated at a higher temperature than the
melting point of the solvent (25ꢀ27 °C). After some experi-
mentation, we also observed that the optimal temperature for the
(
(
Table 1). These results made us reconsider the process in thermal
noncatalytic) conditions but in a safer way.
Continuous Chemistry. Although the batch conditions for
reaction was 220 °C (with a ratio of 2 g Ph O/g substrate). At
2
the rearrangement reaction seemed to work well, we were
concerned about how well it would work on a larger scale. The
concerns were the following:
lower temperatures, the reaction times were too long (>48 h),
and many tar byproducts began to form. At higher temperatures
decomposition of the product was observed. Under desired
conditions, the product precipitated in high yield during the
cooling process and the biphenyl ether could be removed by
washing with hexane. Subsequent catalytic hydrogenation led to
the desired product in 60% of overall yield.
•
Standard large-scale reactors are not equipped to run at
temperatures above 200 °C due to limitations in heat
transfer systems.
An accelerating rate calorimeter (ARC) test on neat 2 showed
an exotherm onset of 172 °C, a maximum self-heat rate of 2.3 °C/min,
and a heat of reaction of 366 J/g (Φ = 2.0) The main reaction
went directly into an exothermic gassy secondary reaction at 265 °C
Special Issue: Safety of Chemical Processes 11
Received: June 20, 2011
Published: August 18, 2011
r 2011 American Chemical Society
1428
dx.doi.org/10.1021/op200162c
|Org. Process Res. Dev. 2011, 15, 1428–1432

Organic Process Research & Development
ARTICLE
Scheme 1. Synthetic Route
Figure 1. ARC results for neat 2.
Table 1. Catalyzed Sigmatropic Rearrangement
• A batch process may be prone to decompose the product
since the heat-up and cool-down times would be longer and
the product would be exposed to high temperature for a
longer time.
All of this data brought to our attention the possibility of
performing the reaction using alternative reactor technology
solvent
additive
temp
Results
toluene/hexane Et
2
2
AlCl
AlCl (1 M in hexanes) rt
rt to 70 °C dimerization
toluene/heptane Et
degradation
no reaction
messy reaction
no reaction
degradation
no reaction
no reaction
degradation
toluene
toluene
toluene
acetonitrile
toluene
DMF
Florisil (30ꢀ60 mesh)
p-TsOH/SiO2
p-TsOH
110 °C
110 °C
110 °C
reflux
(
ART) in a flow-chemistry fashion. Typically, the reaction would
take place on a minimum scale but in a continuous way, using a
long tube reactor and leading to the same result as in large batch
reactor while minimizing the risks of the process.
Yb(OTf)
Yb(OTf)
Yb(OTf)
3
3
3
reflux
The key point for this technique is that all the reagents must be
130 °C
reflux
in solution to avoid plugging the reaction tube. The use of Ph O
2
toluene
Bi(OTf)
3
as solvent could be a problem as both the starting material and
the final product only came into solution when heating the reaction.
We decided to use N-methylpyrrolidinone (NMP) as solvent
for the experiment because of its high boiling point and low vapor
pressure; it is a nonoxidizing agent at that temperature and
turned out to be a powerful solvent for our substrate (2 mL/g).
An accelerating rate calorimeter (ARC) test of a 50 wt %
sample of (2) in NMP gave an exotherm detection temperature
•
Although safety and temperature control could be improved
by running the reaction more dilute and utilizing a semibatch
process (by dosing a starting material solution) rather than a
batch process, calorimetry data indicated there could still be a
substantial temperature and pressure rise if the starting
material (2) was accidentally added all at once.
1
429
dx.doi.org/10.1021/op200162c |Org. Process Res. Dev. 2011, 15, 1428–1432

Organic Process Research & Development
ARTICLE
Figure 2. ARC self-heat rate data and model fit.
Table 2. Predicted Conversion in a Batch Reactor at Differ-
ent Temperatures
Figure 3. ISCO syringe pumps (0ꢀ510 bar, 0.001ꢀ107 mL/min).
temp (°C)
time (h)
conversion (%)
2
2
2
2
00
10
20
30
13
8
96
100
100
100
4.5
2.5
of 181 °C with a maximum self-heat rate of 0.23 °C/min and a
heat of reaction of 170 J/g total mixture (Φ factor = 1.8).
Although the exotherm observed was mild, the onset tempera-
ture was much lower than the desired process temperature and
the 30 min to maximum temperature rate was only 204 °C
(
2
corrected for Φ factor). A secondary exotherm was detected at
45 °C (80 J/g).
In this case there were several safety advantages to running the
reaction in a tube reactor. Heat transfer surface area per unit
volume of the flow tube reactor is much higher than for a batch
2
3
reactor. The 220 mL flow tube reactor had A/V = 2050 m /m .
Figure 4. Oven, rt to 350 °C, 27 cm ꢁ 30 cm ꢁ 16 cm.
In comparison, if a 1-L batch reactor were used for producing
2
3
8
0 g compound instead, it would have A/V about 45 m /m .
Therefore, heat transfer surface area per unit volume is about
The ARC heat flow data was used to develop a kinetic model
with DynoChem software. As we have observed in the batch
9
4
0ꢀ50 times higher for the flow tube reactor than a batch reactor
with comparable throughput. In runaway reactions, the concept
of critical temperature is important. This is the temperature
where the heat generation rate equals heat removal rate. At
higher temperatures a runaway reaction will occur unless extra
cooling is added. Increasing heat removal rate increases critical
temperature.
experiments, the optimal temperature for the reaction was 220 °C,
and it matched very well that predicted in the ARC experiment
when using the Dynochem model. Figure 2 shows that the model
fits the data very well, so the confidence in the model should
be high.
With a working kinetic model in hand we could predict the
required reaction times at various temperatures. This is helpful
when setting flow rates (and hence residence times) in flow
reactors. In Table 2 we can see the model-predicted conversion
and time of reaction for different temperatures below and above
220 °C.
A plug flow tube reactor will exhibit the same conversion
versus time as a batch reactor (residence time in the case of
the flow reactor). Past experience had indicated the small tube
reactors behave much like a plug flow reactor with negligible
axial dispersion (sometimes called backmixing). Therefore, the
Furthermore, standard stainless steel tubing reactors have
pressure rating about 100ꢀ300 bar, which is much higher than
standard batch reactors. This is important because critical temp-
erature is based solely on heat transfer and does not address
pressure build-up. Hazardous pressure buildup could occur in a
closed system at temperatures below the critical temperature. For
this reason, the high pressure rating of the tube reactor is a
backup line of defense against mechanical failure and rupture in
the event of thermal decomposition (assuming the maximum
pressure during a runaway is considered).
1
430
dx.doi.org/10.1021/op200162c |Org. Process Res. Dev. 2011, 15, 1428–1432

Organic Process Research & Development
ARTICLE
0
0
Figure 5. Dispersion data for 1/8 tube reactor.
Table 3. Comparison of Batch and Continuous Processes
batch process
continuous process
solvent
diphenyl ether
33 wt %
N-methyl pyrrolidinone (NMP)
concentration
50 wt %
230 °C
202ꢀ204 °C
1.8 bar
chosen reaction temperature
solvent boiling point
solvent vapor pressure at
reaction temperature
operating pressure
workup
220 °C
259 °C
0.41 bar
1 bar
15 bar
crystallization on cooling (potential to freeze
solvent if temperature is less than 27 °C)
5 h
aqueous drown out and extraction
reaction time
safety
4 h
potential for runaway reaction causing solvent
boiling and/or gas generation especially in large-scale manufacturing
improved heat transfer provides better temperature
control; contained system; potential to run
reaction more concentrated without safety hazards
230 °C easily reached in ovens
reaction temperature profile can be reproduced
with little variation
manufacturability
robustness
220 °C not easily reached by typical batch reactor heat transfer systems
reaction temperature profile will vary with scale due to changes
in heating and cooling capabilities, which can lead to
variation in yield and purity
predicted conversion versus time for a batch reactor (as shown in
Table 2) should be close to that observed in the tube reactor.
Armed with this information, plans were made to produce material
The actual axial dispersion of the tube reactor was measured
after the campaign was completed when time permitted. A sur-
rogate molecule (Benzocaine) was used for the study. The reactor
was filled with a 10 wt % solution of Benzocaine in NMP, and the
feed was quickly changed to pure NMP to initiate a concentra-
tion step change. The dispersion was monitored by tracking the
outlet concentration of Benzocaine. An IR flow cell was installed
near the outlet of the reactor to compliment data collected by
HPLC grab samples (Figure 5). For a plug flow reactor the
concentration change at the outlet of the reactor would also be a
step change from 10 to 0 wt %. Some dispersion will occur in any
real reactor, especially those using laminar flow conditions.
However, the dispersion was relatively low in this case (dispersion
number (D/μL) = 0.00024). For kinetic modeling purposes this
amount of dispersion can be ignored, and the reactor can be
assumed to be plug flow. Additional background on the effect of
dispersion on kinetics can be found in the Supporting Information.
00
in a 1/8 tube reactor.
The dissolved reagent was pumped continuously (using ISCO
00
Syringe pumps, Figure 3) through the 1/8 stainless steel tube,
which was heated inside a GC oven. The product solution
continuously flowed out the end of the tube into a product col-
lection vessel.
Some experimentation was needed to optimize the tempera-
ture and the residence time of the process. The best conditions
10
found were 230 °C, pressure about 15 bar, in a 73.5-m long
tube, and residence time of 4 h (Figure 4). The linear velocity of
liquid pumping through the tube was about 30 cm/min; the tube
volume was about 222 mL, and the volumetric flow rate was
0.92 mL/min. At the end of the process the entire product was
pushed out at the end with fresh NMP at the same flow rate.
1
431
dx.doi.org/10.1021/op200162c |Org. Process Res. Dev. 2011, 15, 1428–1432

Organic Process Research & Development
ARTICLE
Another observation is that the HPLC data indicated more
dispersion data than the IR data. This makes sense because the
HPLC grab samples were collected after the stream passed
through a series of small pressure vessels. These vessels would
create some additional dispersion.
The IR flow cell is a valuable option for collecting concentra-
tion data in flow reactors. It has the advantages of providing real-
time data, greater sample frequency, no requirement for sample
preparation, and no data interference due to additional back-
mixing. However, it does have some requirements: the flow cell
must be capable of withstanding the system pressures; the flow
cell material of construction must be suitable for the mixture; the
molecule of interest must have a spectrum sufficiently different
from the rest of the mixture to allow concentration tracking; and
an IR calibration curve must be created.
regulator into the collection vessel at room temperature and
atmospheric pressure. Average hydraulic residence time in the
hot zone (230 °C) of the reactor tube was 4 h. When the reaction
solution ran out, the high-pressure syringe pump was switched
from pumping reaction solution to pumping NMP only without
stopping flow for the last 4 h. Therefore, all reagent solution had
precisely 4 h average hydraulic residence time in the thermal tube
reactor to complete the reaction. Flow rate, reactor pressure, and
reactor temperature remained the same during the continuous
NMP pushout of the reactor. The reactor tube contained NMP
solvent only at the end of the experiment. The product solution
from the collection tank was worked up and isolated in batch
mode. Thus, the mixture was transferred to a 2 L RBF, and 1 L of
water was added and extracted with MTBE (700 mL). The
mixture was filtered through a short pad of Celite to give a clear
phase cut. The organic layer was dried over MgSO , filtered, and
4
concentrated to dryness. The residue was stirred in 1000 mL of
hexane/MTBE (10:1), filtered, and dried to give 3-allyl-4-
hydroxyacetophenone (80.8 g, 73.45% yield, 93% pure).
’
CONCLUSIONS
In summary, scaling this thermal rearrangement in a flow tube
reactor provided several advantages:
•
High temperature was easily obtained compared to a batch
reactor.
Safety was enhanced.
The timeꢀtemperature profile can be easily reproduced
regardless of material requirements. The reaction mixture
quickly reaches temperature upon entering the high tem-
perature zone and quickly cools after leaving the high tem-
perature zone. This allows better control over side reac-
tions (such as product decomposition in this case).
’
ASSOCIATED CONTENT
Supporting Information. H NMR spectra for 1, 2, and 3.
•
•
1
S
b
ARC plots for neat 2 and in NMP solution. Rationale about the
effect of flow reactor dispersion on reaction kinetics. This material
is available free of charge via the Internet at http://pubs.acs.org.
’
AUTHOR INFORMATION
This process allowed us to make 80 g of the material in a short
time while mitigating the potential chemistry hazards. A compar-
ison of the batch and continuous processes are shown in Table 3.
Corresponding Author
*E-mail: rinconja@lilly.com.
Present Address
Departamento de Química Org ꢀa nica, Facultad de Ciencias, Uni-
§
’
EXPERIMENTAL SECTION
versidad Aut ꢀo noma de Madrid, Cantoblanco, 28049 Madrid, Spain.
Batch Reaction. A mixture of 1-(4-(allyloxy)phenyl)ethanone
96.47 mmol; 17.0 g) and diphenyl ether (34.0 g) was placed into
(
’ REFERENCES
a open vial and heated to 220 °C under N for 5 h. The mixture
2
(1) Marshall, W. S.; Goodson, T.; Cullinan, G. J.; Swanson-Bean, D.;
Haisch, K. D.; Rinkema, L. E.; Fleisch, J. H. J. Med. Chem. 1987, 30 (4), 682.
(2) U.S. Patent Application 20100048610.
was then cooled to 40 °C, 100 mL of hexanes was added, and the
mixture was cooled further to 0 °C. The solid was filtered off and
washed with 100 mL of hexanes. The brown solid was stirred in
(3) Claisen, L. Chem. Ber. 1912, 45, 3157.
(
4) For a recent review see: Martín Castro, A. M. Chem. Rev. 2004,
04, 2939.
5) Aluminum(III) derivatives: (a) Wipf, P.; Rodriguez, S. Adv.
1
60 mL of MTBE/hexane (2:1) overnight. The solid was rinsed
1
with 20 mL of hexanes to give a first crop of 3-allyl-4-hydro-
xyacetophenone (4.3 g, >98% pure).
(
Synth. Catal. 2002, 344, 434. (b) Marcune, B. F.; Hillier; Michael, C.;
Marcoux, J-F; Humprey, G. R. Tetrahedron Lett. 2005, 46, 7823. (c)
Maruoka, K.; Sato, J.; Banno, H.; Yamamoto, H. Tetrahedron Lett. 1990,
Liquors were concentrated in vacuo, and the solid was stirred
in 120 mL of hexane/MTBE (5:1). The solid was filtered to give
a second crop (7.68 g, 91%). Both solids were mixed to give a
single lot of 3-allyl-4-hydroxyacetophenone (11.98 g, 70.47%
yield, 94% pure.
31, 377. Bismuth(III) derivatives:(a) Ollevier, T.; Mwene-Mbeja, T. M.
Synthesis 2006, 23, 3963. (b) Ollevier, T.; Mwene-Mbeja, T. M. Tetra-
hedron Lett. 2006, 47, 4051.
Flow Reaction. A solution of 82 g of 1-(4-(allyloxy)-
phenyl)ethanone in 174 mL of NMP was pumped through a
(6) Sharma, G. V. M.; Ilangovan, A.; Sreenivas, P.; Mahalingam, A. K.
Synlett 2000, 615.
(7) Clays: (a) Dauben, W. G.; Cogen, J. M.; Behar, V. Tetrahedron
7
2
3.5-m long continuous thermal tube reactor (GC oven, V =
21 mL, o.d. = 3.175 mm, i.d. = 2.0 mm) heated at 230 °C and
Lett. 1990, 31, 3241. (b) Dintzner, M. R.; Morse, K. M.; McClelland,
K. M.; Coligado, D. M. Tetrahedron Lett. 2004, 45, 79. Zeolites:
Ganapati, D. Y.; Sharad, V. L. Org. Process Res. Dev. 2009, 9, 547.
pressurized to 15 bar. At the beginning of the run, before reagents
were pumped through the reactor tube, the tube was prepressur-
ized with 15 bar nitrogen and preheated. High pressure ISCO
syringe pumps were used to continuously push reagent solution
through the tube and maintain constant liquid flow rate of
(
8) Karageorge, G. N.; Macor; John, E. Tetrahedron Lett. 2011,
2, 1011.
9) DynoChem software is a product of Scale-up Systems Ltd., Dublin,
5
(
Ireland. The kinetics were modeled using the first order reaction (2) f (3)
(10) Backpressure of 15 bar was utilized to prevent the solvent from
boiling at the reaction temperature.
0
.92 mL/min for the entire run. The product solution continu-
ously flowed out the end of the reactor tube, through a cooling
tube heat exchanger, and through a back pressure regulating
device. Product continuously dripped out the back pressure
1
432
dx.doi.org/10.1021/op200162c |Org. Process Res. Dev. 2011, 15, 1428–1432