How to convert a walk-in hood into a manufacturing facility: Demonstration of a continuous, high-temperature cyclization to process solids in flow

Article abstract of DOI:10.1021/op500239f

An intramolecular thermal cyclization protocol was developed in a flow reactor to take advantage of the high pressures and temperatures that are easily obtained in small scale autoclave reactors that have been modified to handle slurries. This reactor was equipped with a fill/empty pumping system to enable easy and nearly complete transfer of slurries. The reaction conditions were designed to take advantage of the insolubility of the product in order to separate it from residual starting material by filtration after short reaction times. Recycling of the filtrate maximized the yield and throughput while minimizing decomposition. Recycles were accomplished using a strip to dryness protocol that was easily performed in a rotary evaporator. This new equipment set was designed with lab-hood manufacturing in mind, a minimized footprint, and the system was completely automated for charging, emptying, rinsing, and reacting. Additional efforts for quick screening and alternate modes of addition were also investigated.

Full text of DOI:10.1021/op500239f

Article
pubs.acs.org/OPRD
How to Convert a Walk-in Hood into a Manufacturing Facility:
Demonstration of a Continuous, High-Temperature Cyclization to
Process Solids in Flow
Timothy D. White,* Charles A. Alt, Kevin P. Cole, Jennifer McClary Groh, Martin D. Johnson,
and Richard D. Miller
Small Molecule Design and Development, Eli Lilly and Company, Indianapolis, Indiana 46285, United States
S
* Supporting Information
ABSTRACT: An intramolecular thermal cyclization protocol was developed in a flow reactor to take advantage of the high
pressures and temperatures that are easily obtained in small scale autoclave reactors that have been modified to handle slurries.
This reactor was equipped with a fill/empty pumping system to enable easy and nearly complete transfer of slurries. The reaction
conditions were designed to take advantage of the insolubility of the product in order to separate it from residual starting material
by filtration after short reaction times. Recycling of the filtrate maximized the yield and throughput while minimizing
decomposition. Recycles were accomplished using a strip to dryness protocol that was easily performed in a rotary evaporator.
This new equipment set was designed with lab-hood manufacturing in mind, a minimized footprint, and the system was
completely automated for charging, emptying, rinsing, and reacting. Additional efforts for quick screening and alternate modes of
addition were also investigated.
technology to easily transfer slurries.14 This equipment has a
INTRODUCTION
■
small footprint and was aimed at becoming a tool for quick
Quinoline compounds play a key role as intermediates in the
reaction screening as well, obviating the tedious generation of
synthesis of a variety of potential drug therapies including
product or intermediate solubility data in order to expedite
Alzheimer’s,1 antimalarials,2 and antibiotics3 among others.
condition screening given that the reagent feed solutions were
While batch Gould−Jacobs thermal cyclizations and other high-
homogeneous.
temperature reactions are well-precedented, it was desirable to
develop an alternate method that would avoid some of the issues
observed in batch, such as long reaction times, difficulty in
EQUIPMENT
■
An automated fill-empty stirred tank reactor was envisioned as
an effective alternative for continuous reactions with solids in
achieving desired reaction temperatures, and undesirable
solvents such as diphenyl ether or Dowtherm.4,5 The temper-
flow. While the flow is intermittent rather than truly continuous,
atures necessary for these cyclizations require specialized batch
this reactor type has many of the characteristics of continuous
equipment that can heat to temperatures as high as 310 °C.
reactors. The reactor size is small compared to the volumetric
Recent large-scale requirements for this type of cyclization
throughput, as there are many reactor volume turnovers per day,
published in this journal have avoided this transformation
and the reactor remains at one temperature as material flows in
entirely by developing a new synthetic route6 or effecting the
and out (similar to a PFR). Reagents flowing in are heated
same transformation under acidic conditions with Eaton’s
quickly, the residence time in the reactor is precisely controlled,
reagent.7 There are examples of these cyclizations under
and the products flowing out are cooled quickly. The small
alternative conditions such as microwaves8 including examina-
reactor size also makes implementation of extreme temperatures
tion of microwaves on larger scale in flow,9 but mostly on
and pressures possible. All of these attributes are typical
characteristics of a continuous reactor, but this reactor type can
also handle solids in flow for an extended run without plugging
or clogging.
smaller scale in plug flow reactors (PFR) with solvent10 and
neat.11
Another aspect complicating the synthesis of these types of
molecules is the fact that the product readily precipitates from
The reactor that was chosen for this chemistry was a 25 mL
most solvent systems. Solids in continuous fine chemical
modified Parr autoclave reactor with overhead stirring as shown
processes have historically been viewed as problematic in
in Figure 1. The Parr ceramic heaters used to heat these reactors
small-scale PFR equipment,12 but improvements have been
are designed to readily achieve the temperatures necessary to
effect these cyclizations.15 The reactor itself is rated well above
the 600 psi necessary to keep the solvents of interest in the
made with an understanding of better equipment design and
improved reaction conditions to avoid fouling.13 Despite these
new methodologies, we hoped to isolate via a direct filtration
without performing a rework on the isolated solids which was a
requirement of some of the previous flow efforts mentioned
Special Issue: Continuous Processes 14
above. Finally, the key to streamlining the overall process was to
take advantage of our previously disclosed expansion zone
Received: July 21, 2014
© XXXX American Chemical Society
A
dx.doi.org/10.1021/op500239f | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Figure 1. High temperature, fill/empty reactor cart.
Figure 2. Pressure (orange line), balance trend (blue) and temperature (gray) profile of 5 min reaction cycle.
B
dx.doi.org/10.1021/op500239f | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
liquid phase at 265 °C (for example, the boiling point
temperature for toluene at 600 psig is 325 °C).16 Also, because
more than 100 volume turnovers could be possible in a day, the
reactor is more than 100 times smaller than a batch reactor for
the same throughput. Reagents are heated from 20 to 265 °C in
approximately 2 min when introduced into the reactor, and they
are cooled from 265 °C to near 40 °C in less than 1 min when
removed from the reactor while the residence time is precisely
controlled. None of this is feasible in conventional batch
equipment at any scale. The product slurry flowed out of the
reactor system through expansion chambers in series and a
hydrocyclone. Product slurry was pushed out of the reactor
intermittently through a 1/4″ OD and 0.18″ ID stainless steel
diptube into expansion chambers in series separated by
automated block valves. This design allows a process to run
for extended periods without solids plugging at the outlet
because of turbulent intermittent flow through the 1/4″
automated block valves. The flow was zero when the valves
are closed followed by very high velocity flow out when the
valves are open due to the 550−600 psig pressure difference
from one chamber to the next. There are no restricting orifices
like in most conventional back pressure regulators, because the
restricting orifices clog with solids. Instead, the flow path is
greater than 3 mm in diameter when each of the block valves are
opened in sequence with pressure driving the force to achieve
turbulent flow, which prohibits clogging with solids.
achieve the desired 265 °C reaction temperature. The high
thermal mass of the reactor on 25 mL scale was adequate for this
fast heat up from room temperature solution. After feed
addition, it was anticipated that some amount of the solution
was pushed into the dip tube. In order to push this material back
into the reactor and avoid solids buildup in the dip tube, the
reactor was repressurized to 600 psig through valve C. Toluene
wash cycles were also introduced through the diptube to rinse
the tube and reactor every third cycle from a second feed pump
in order to prevent solids build up. The automated cycle
repeated for the duration of the planned run time depending on
the amount of feed material to be reacted.
DISCUSSION
■
The cyclization of interest was the formation of hydroxyquino-
line 2 (Scheme 1). Batch methodology to generate this molecule
Scheme 1. Thermal cyclization to form 2
has been well-documented but only reported on a small scale in
diphenyl ether.18 In this case, a preferred solvent replacement
for high boiling solvents was 2-methyltetrahydrofuran (2-
MeTHF) because of its utility as a green solvent.19 An early
proof of concept for using this solvent for the formation of 2 was
previously published using the equipment set described above
but was not fully optimized.20 As efforts to optimize this process
were pursued, the shortcomings of this solvent under these
reaction conditions came to light. 2-MeTHF suffers from high
thermal expansion: when 2-MeTHF is heated from 20 to 260
°C, its volume increases 2.13 times. While acceptable reaction
profiles were observed, thermal expansion began to impede
throughput since less starting solution could be charged in each
cycle. Alternatively, when toluene is heated from 20 to 260 °C,
its volume only increases 1.48 times, so toluene was chosen for
additional studies.21 To exemplify the throughput issue, the
maximum reaction charge to a 25 mL reactor was 15 mL of
toluene feed but only 10 mL of 2-MeTHF feed.
Safety and yield were significant concerns with the existing
batch process. When 1 was processed at 250 °C, a 55.0% yield
was isolated as a brown solid with 96.2−99.0% purity. As
mentioned previously, the quinoline product has very low
solubility in most room temperature solvents, but 1 has very
high solubility (32 wt %) as well as excellent stability in toluene,
reinforcing the solvent choice mentioned previously. Due to the
high solubility of the starting material in toluene, a starting feed
consisting of 1 in only 4.2 L/kg of toluene was investigated.22
Batch screening studies were initiated in port connector reactors
to determine the optimal reaction time and temperature.23 The
reactors enabled rapid screening of time, temperature, and
concentration to build a data package to inform initial
continuous processing. These runs suggested that high
conversion and good yield could be obtained at 250 °C with a
2 h residence time; so these conditions were selected for the first
continuous run.
AUTOMATION
■
The computerized automated valve sequence which repeated
more than 700 times and was vital to successful operation was as
follows: when valve D opened (Figure 1), the high pressure of
the reactor system immediately pushed the product slurry
(pushout) into a 40 mL stainless steel tube which was initially at
0 psig pressure. The reactor pressure only dropped by 10%
when pressure equalized with the 40 mL zone because the total
interconnected reactor headspace was about 400 mL. As seen in
Figure 2, the reactor pressure dropped from approximately 600
psig to about 545 psig. This sudden drop in pressure represents
the speed that the slurry was pushed out of the reactor through
the 4.5 mm ID tubing, which minimized the potential for solids
clogging. The product slurry settled to the bottom of this 40 mL
zone due to gravity, and the headspace high pressure nitrogen
was trapped when valve D closed. Subsequently, automated
valve E opened, and the product slurry was immediately pushed
to a hydrocyclone, which separated vapor from the slurry as it
depressurized. The slurry gradually accumulated in a product
vessel over time as each cycle flowed out of the reactor. This
pushout sequence was repeated two additional times to ensure
the contents of the reactor had been removed prior to charging
the next starting material cycle. The reactor pressure dropped
about 10% each time valves D and E were cycled. The lowest
pressure during this operation was about 430 psig, which is
significantly higher than the vapor pressure of toluene at 265 °C
(285 psig). The next section of feed was immediately charged to
the hot reactor by the automated intermittent flow feed pump.
The temperature of the liquid phase was constantly measured,
but the temperature control was based on the reactor wall
temperature.17 Typical temperature, pressure and balance
profiles for a complete fill-empty sequence are shown in Figure
2. This highlights the sequence of events showing the three
pushout sequences resulting in a lower temperature and
pressure, followed by the addition of the next segment of feed
requiring approximately 2 min to reach 260 °C and 3 min to
Using the reactor described in Figure 1 at 250 °C and
residence time of 2 h, the reactor plugged after less than 36 h,
which was detected from the pressure transmitter shown in
C
dx.doi.org/10.1021/op500239f | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Figure 1. After cooling and depressurizing the reactor the
contents appeared as dark waxy solids (Figure 3). With high wt
(Scheme 2). Once intermediate 3 is generated, the desired
product can form, starting material can be regenerated, or
decomposition can occur. To prove that the presence of ethanol
was detrimental to conversion, an equivalent of ethanol was
added to the starting feed and half the conversion was observed
compared to the case without ethanol present for a 1 h reaction
at 250 °C. This also helps explain the observation that a more
dilute reaction (10 L/kg) results in higher conversion as a larger
ethanol “sink” was available. Traditional batch conditions have
the option of removing the ethanol via distillation, but since the
continuous reactor was a closed system at high pressure,
distillation was not an option.
Temperature was investigated as a handle to drive the
conversion further. As seen in Figure 5, improved conversion is
indeed observed until no starting material remains after 10 min
at 300 °C. Unfortunately, disappearance of starting material in
this case did not translate to higher yield, as the decomposition
pathways became predominant. It was already well-documented
that cyclizations such as these are temperature-dependent with
regard to decomposition.24 A longer term temperature study for
this substrate was initiated at 240 °C, and both product and
starting material were monitored. As seen in Figure 6, product
formation levels off quickly, and the decomposition pathway
takes over.
Figure 3. Twenty-five mL reactor after fouling, running with 4.2 L/kg
of toluene.
While increasing the toluene to 15 L/kg seems counter-
intuitive to the goal of a greener process, the additional yield and
shorter reaction times that could be achieved may outweigh the
additional solvent requirements. Upon reconsidering the data in
Figure 5, after only 10 min at 260 °C, 40% conversion was
observed. While this yield is not close to the desired target, it
does mitigate the decomposition issues. The product cleanly
precipitates and due to its insolubility, no significant levels of 2
are observed in the mother liquor. Also, due to the exceptional
solubility of the starting material, only low levels were trapped in
the resulting solids. However, while the mother liquor contains
clean starting material, the presence of ethanol would continue
to hamper conversion to the desired product. A rotovap strip to
dryness operation has previously been disclosed as a viable unit
operation for continuous processing25 and allowed removal of
% solutions in toluene, even the use of wash cycles was not
enough to efficiently empty the thick slurry from the reactor,
which gradually became packed with solids. The particles were
very fine and at this dilution were not well behaved in terms of
mobility. During this phase of development, longer reaction
times and lower temperatures were also investigated, but these
conditions did not alleviate reactor fouling.
OPTIMIZATION
■
At this point a better understanding behind the behavior of the
reaction was required for improvements to be possible. An
investigation of solvent volumes showed better conversion over
time with more volumes of toluene (Figure 4). Upon
considering the mechanism of the reaction, the equivalent of
ethanol generated during the reaction could be problematic
Figure 4. Solvent volume comparison for conversion of 1.
D
dx.doi.org/10.1021/op500239f | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Scheme 2. Mode of decomposition
efficiency. Listed below are general drivers for consideration of
a recycle process (i.e., continuous reaction with continuous
recycle):
1. Late forming impurities in the process;
2. Removal of a reaction byproduct enables higher conversion
or purity;
3. For competing reactions, an intermediate conversion point
exists where there is a maximum ratio of desired product to
impurity;
4. The process benefits from a very high stoichiometric ratio
between two reagents;
5. The process has a solubility issue or fluid mechanics and
mixing difficulty at low conversion;
6. Catalyst recycle achieves a lower overall catalyst loading but
higher instantaneous catalyst loading to the reactor;
7. Solvent recycle results in a lower PMI;
Figure 5. Temperature dependence on conversion.
8. Product is lost to the filtrate which could be recovered
during recycle.
the ethanol. Subsequent reconstitution of the resulting solids in
15 L/kg of toluene afforded feed ready for resubmission to the
reaction conditions. The recycled feed resulted in similar
conversion to product because the ethanol had been removed
from the system.26 Here, the operating premise of accepting a
low single pass conversion with a subsequent recycle was
selected because continuous reactions with recycle of unreacted
reagents can improve the overall yield, throughput, and
A separation step is typically employed to remove byproduct
and solvents from the recycled material so that it enters the
reactor at the same concentration as the fresh feed. For this
cyclization a separations step to reconstitute feed was easily
accomplished. This process was viewed as a good candidate for
recycle in a continuous reactor with an integrated downstream
Figure 6. Decomposition of 2 over time at 240 °C.
E
dx.doi.org/10.1021/op500239f | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Table 1. Data from the 300 g run
yield
mass balance
product balance
HPLC area %
isolated 2 (g)
potency
fraction 1
25.6%
49.2%
48.6%
46.4%
47.0%
43.4%
49.5%
65.1%
52.1%
76.2%29
64.0%
110.7%
87.6%
92.0%
93.4%
89.6%
93.4%
90.8%
64.2%
86.9%
52.2%
105.1%
104.3%
100.6%
93.9%
91.4%
86.2%
89.8%
92.5%
95.4%
95.5
96.3
97.0
97.2
97.6
12.47
25.14
23.87
23.92
19.57
76.7%
94.5%
94.0%
97.3%
94.2%
fraction 2
fraction 3
fraction 4
fraction 5
first pass cumulative
recycle 1
96.4
93.7
52.48
94.5%
85.5%
second pass cumulative
recycle 2
27.04
process cumulative
169.46
continuous separation step (rotovap strip to dryness) because
drivers 1−3 applied.
percentage of the flow combined with clean feed and a
percentage of mother liquor going to waste. Determining the
percentage of mother liquor going to waste (recycle ratio) is
defined as the filtrate mass recycling back to the reactor divided
by the filtrate mass flowing to waste and would be 66% for this
process going forward.
EXTENDED RUN
■
After minimal optimization of temperature and residence time, a
300 g long duration run was planned at 265 °C with a 5 min
reaction time utilizing a feed of 1 in 15 L/kg of toluene. Due to
concerns about fouling of the diptube and solid buildup in the
reactor, a 15 mL toluene wash after every three reaction cycles
(15 mL of reaction feed per cycle) was planned. For this run
material was collected as five bulk fractions and filtered offline.
As seen in Table 1, very comparable solid isolated yields of 2
were collected for each of these fractions in good purity.27 The
mass balance28 and product balance were monitored throughout
the run and were good with the exception of the first section.
The first section was less potent and exhibited inferior product
and mass balances, possibly suggesting decomposition. The
yield of the first section was correspondingly lower, but the
remaining sections were well understood and as expected. At
this time, a root cause for the low mass balance observed in the
first section has not been identified. The mother liquor from the
first five sections was combined and concentrated to a solid.
After dissolving in 15 L/kg of toluene, it was resubjected to the
reaction conditions. After filtration, the resulting mother liquor
was again concentrated, redissolved, and submitted to the
reaction for a third pass. As seen in Table 2, the potency of the
Next, a shorter run to process 60.2 g of 1 for over 100 reactor
turnovers without toluene washes30 was completed. No fouling
or impact on product purity was observed over the operating
time of 14 h. It is also important to note that the issues observed
during startup as discussed above for the first section of the
extended run were not observed here. Therefore, there is no
reason to believe the startup will be problematic for every new
run. The yield was 71.2%, and the purity of the product was 95.0
HPLC area %. After opening the reactor, no evidence of solid
buildup was observed, and the inside of the reactor was visually
similar to the run that had incorporated washes (Figure 7),
implying that the toluene wash was not necessary. Elimination
of the toluene wash would have a significant impact on
throughput and decrease toluene usage by 25%. As seen in Table
3, a readily available one gallon autoclave with no wash cycle
would be capable of producing over 80 kg of product per week
in a lab hood.
TESTING OF ADDITIONAL SUBSTRATES
■
To test the scope of this equipment set, a quick screen of other
thermal cyclizations that are known in the literature were chosen
to determine if improvement was possible for these substrates as
well (Table 4). Both trifluoromethyl isomers 4 and 6 were
examined, and while the cyclization to 4 was not as high yielding
in this equipment set, the cyclization to form 7 appeared to be a
significant improvement when a recycle was employed. Even
more impressive was the yield increase for the methoxy
substrate 8, which saw a significant improvement to 52%.
While all the substrates suffer from some degree of
decarboxylation, xylene 10 afforded the desired Gould−Jacobs
product along with significant levels of 12 resulting from ester
hydrolysis and decarboxylation. When the reaction was run
under the continuous conditions used to generate 2 (265 °C for
5 min), 20% of this impurity was observed in the isolated solids.
Lowering the temperature to 250 °C with the same 5 min
residence time resulted in a significant improvement (only 6.9%
of the decarboxylated impurity). Changes in reaction temper-
ature could increase the levels of either of the two products, but
without further optimization neither pure 11 or 12 was
generated. Based on the improvements resulting from minor
changes for compound 2, optimization could lead to a protocol
for generating any of these molecules in synthetically useful
yields and purity.
Table 2. Potency of 1 in mother liquor
g/sol of
reconstituted feed
wt of stripped
solids (g)
mother liquor
potency
end of first pass lot 5
24.10
21.10
25.48
23.87
94.58%
88.39%
section of first recycle
lot 6
second recycle
28.77
39.26
73.27%
solids after removal of toluene decreased with each pass through
the reactor, so only minimal benefit from a fourth pass was
expected, and this material was discarded. No evidence of solids
buildup or fouling was observed during this run. Figure 7 shows
the reactor was very clean and would have been expected to
continue running without solids fouling for a considerable
length of time.
When the recycles were factored in, the total volume charged
to the reactor was over 10 L, including washes, equating to a
total number of reactor turnovers for the run including the
recycle fractions of over 700. The equipment ran successfully for
a total operating time during this 300 g campaign of 3.9 days
without fouling. While for this campaign, the recycles were
discrete separate runs, in the future this recycle will be a
F
dx.doi.org/10.1021/op500239f | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Figure 7. Reactor after processing over 300 g of 1.
buildup of solids in the reactor when a toluene rinse was
introduced after every three reaction cycles. The total run time
of the automated reactor system was 94 h with no solids
plugging or clogging. The yield was increased and similar purity
delivered compared to the batch conditions. A shorter duration
run was demonstrated without a toluene rinse between reaction
cycles, which would result in a greener process and increase
throughput.35
Table 3. Comparison of first pass throughput in different
reactors with reaction parameters
25 mL
reactor
1000 mL
a
1 gal/no
a
reactor
wash
residence time (sec)
additional operation time (sec)
time for 4 cycles (3 rxn, 1 wash)
wt % feed
300
167
1868
7.2
300
167
1868
7.2
300
167
NA
7.2
This reactor type is a legitimate alternative for high-
temperature, high-pressure continuous reaction with solids in
flow. The value of running this setup with the potential for
manufacturing in a lab hood setting would result in a significant
reduction in capital investment for new equipment, utilize
modified equipment that was readily available, and increase
temperature and pressure capabilities compared to standard
batch reactors. Furthermore, scale up in a batch autoclave is not
capable of achieving 3 min heat up to the 265 °C reaction
temperature, precisely controlled 5 min reaction time, and less
than 1 min cool down to room temperature after the desired
reaction time. Therefore, scale up in a batch reactor would not
achieve the same in situ yield and impurity profile as the
automated intermittent flow reactor. Additional examples were
quickly screened ,and while optimization would be required if
higher purity material was necessary, these reactions showed
great potential for an improved process in this equipment set for
these substrates as well. Investigations are ongoing to determine
the root cause of poor results during the first section of the long
run to generate 2. Larger scale efforts are being explored in order
to examine energy consumption and toluene recycling as
additional potential green benefits to the process.
density feed
0.88
15
0.88
600
49
0.88
2300
249
mL per cycle
2 (g) per 4 cycles (first pass 50%
yield)
1.23
2 (g) per day
57
2280
11500
81000
2 (g) per week
399
16000
a
Projected throughput.
An important point to remember is that previously reported
small scale literature yields may not be repeatable in large scale
batch equipment and route redesign may be required.6,7 While
these reactions utilized a single reagent stream, slow addition of
one reagent to another or true continuous stirred tank reactor
conditions could be explored if there was a benefit to operating
at end of reaction conditions in this equipment. It is worth
noting that the proposed scale up to generate 80 kg per week in
a one gallon reactor represents a 151× scale up over the 25 mL
reactor demonstrated here. For a similar batch scale up to a 500
L reactor from a 25 mL batch demonstration, this is a 20 000×
scale up that would require good understanding of heating and
cooling times and require demonstration on intermediary scales.
Furthermore, if this chemistry was scaled to a 500 L batch
reactor, toluene would not be used due to the high pressures
required. The reaction would instead be run in a higher boiling
solvent which could complicate product isolation.
EXPERIMENTAL SECTION
■
General. Starting materials and reagents were purchased
commercially and used without further purification. The
reaction feed and toluene rinse were delivered by 1000D
ISCO pumps36 to accurately supply feeds under high pressure
with automation control. Reagent feeds flowed into the reactor
through 1/8″ OD, 0.069″ ID stainless steel tubing from the
automated intermittent flow high pressure feed pumps. The
diptube was designed with minimal clearance to almost
completely remove the reaction slurry from the reactor, and
the reactor itself was machined to allow clearance of the diptube
with the pitched blade impeller resulting in approximately one
mL of slurry heel remaining in the reactor after emptying.
DeltaV37 was used to control all automation for all pumps and
CONCLUSIONS
■
An improved protocol for the thermal cyclization to form 2 was
developed in a high temperature/high pressure, fill/empty
reactor that readily handled solids in flow. The reactor utilized
expansion chambers in series with sequenced automated block
valves and a hydrocyclone to control the slurry flowing out of
the reactor, maintain reactor back-pressure on reactor at all
times, depressurize the product slurry to atmospheric pressure,
separate slurry from vapor, and transport slurry to a product
collection tank. The process was demonstrated in flow for 3.9 d
equating to over 700 turnovers without fouling or significant
G
dx.doi.org/10.1021/op500239f | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
a
Table 4. Screening of additional substrates
a
b
c
All reactions run at 600 psig, 265 °C with a 5 min tau with the exception of 10 which was run at 250 °C. First recycle. Second recycle.
valve sequences and to trend temperatures and the product
balance. Initially Teflon gaskets that are rated to 350 °C were
used, but the changes in pressure caused the slightly melted
PTFE to flow into the reactor.38 The better option was
inexpensive Grafoil gaskets which are replaced every time the
reactor is taken apart. Additional details about the equipment
utilized has been well-described in the body of this paper with
further details in the Supporting Information. The reported
yields are corrected for potency and reactions were monitored
by HPLC and purities and potencies quoted herein refer to
HPLC area % at 220 nm. HPLC method: Zorbax SB-C8, 3.5 μm
4.6 mm × 75 mm column; flow rate 2 mL/min; mobile phase A:
0.1% trifluoroacetic acid in water; mobile phase B: 0.1%
trifluoroacetic acid in acetonitrile; gradient: 96:4 A/B to 10:90
A/B over 12 min, 3 min at 96:4 A/B. Temperature: 30 °C.
Injection volume: 2 μL. Retention times are expressed in
minutes. Cyclization substrates 1, 4, 6, 8, and 10 were all
generated via known literature procedures.39,31−34 Melting
points were collected on a Buchi Melt Point B-540 and are
uncorrected. NMR data was collected in TFA-d or solids were
dissolved in minimal TFA-d and diluted with acetonitrile-d3.
Ethyl 6-bromo-4-hydroxyquinoline-3-carboxylate (2).
Diethyl 2-(((4-bromophenyl)amino)methylene)malonate 1
(300.0 g, 876.7 mmol) was dissolved in toluene (4.5 L), and a
portion was transferred to a 1000D ISCO pump, while a second
1000D ISCO pump was filled with toluene. The pump then
feeds 15 mL of this solution to a 25 mL reactor at 265 °C and
600 psi nitrogen to remain above the vapor pressure of the
reaction solution. After 5 min in the reactor, the resulting slurry
exits the reactor through a valve via a diptube to near the bottom
of the reactor to a 40 mL depressurization and cooling zone.
Finally, a second valve in series opens and the trapped nitrogen
pressure sends the slurry to a 300 mL depressurization
hydrocyclone zone. This sequence to empty the reactor was
repeated 2 additional times as described above to ensure the
residual slurry was removed to a minimal volume. Then, the
reactor was refilled with reagent solutions and repressurized
from approximately 430 psi to 600 psi through the diptube.
After an additional 2 cycles were performed in this fashion,
toluene (15 mL) was sent to the reactor, held for 5 min, and sent
to the collection bottle in the same fashion as reaction slurry. A
total of 5 reaction sections were collected, each was filtered,
washed with toluene (75 mL), and dried under vacuum at 50
°C. Each section was approximately 100 reactor turnovers. The
mother liquors were combined and concentrated on a rotary
evaporator at 60 °C bath temperature. The resulting solid
(147.0 g) was dissolved in toluene (2.2 L). This recycle was
submitted to the same procedure as the first pass material.
Finally, the mother liquor from the recycle was concentrated as
above, and the resulting solid (63.13 g) was dissolved in toluene
(950 mL). This material was subjected to the same conditions as
above. After drying the 6-bromo-4-hydroxyquinoline-3-carbox-
ylic acid ethyl ester was collected as 7 lots as beige solids (12.47
g (76.7% potency), 25.14 g (94.5% potency), 23.87 g (94.0%
potency), 23.92 g (97.3% potency), 19.57 g (94.2% potency),
52.48 g (recycle lot, 94.5% potency), 27.04 g (2nd recycle lot,
H
dx.doi.org/10.1021/op500239f | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
85.5% potency), 170.18 g (potency corrected), 575.0 mmol) for
a 65.6% yield. Mp = 327.8 to 328.9 °C; HPLC: tR = 4.63 min. 1H
NMR: (400 MHz, TFA-d), 1.52 (3H, t, J = 7.04 Hz), 4.67 (2H,
q, J = 7.03 Hz), 8.03 (1H, d, J = 8.79 Hz), 8.28 (1H, d, J = 8.79
Hz), 8.80 (1H, s), 9.32 (1H, s); 13C NMR (100 MHz, TFA-d),
11.9, 64.7, 105.3, 121.0, 121.2, 124.9, 126.9, 137.9, 141.1, 145.0,
167.2, 172.4.
Ethyl 4-hydroxy-6-(trifluoromethyl)quinoline-3-car-
boxylate (5). Diethyl 2-(((4-(trifluoromethyl)phenyl)amino)-
methylene)malonate 4 (30.09 g, 90.8 mmol) was dissolved in
toluene (450 mL) and reacted in the same manner as 1. After
drying, 7.22 g (96.7% potency) of 5 were collected as a tan solid
for a 27.0% yield. Mp = 324.2 to 333.4 °C; HPLC: tR = 5.10 min.
1H NMR (499 MHz, acetonitrile-d3) δ 1.43 (t, J = 7.09 Hz, 3H),
4.57 (q, J = 7.09 Hz, 2H), 8.31 (d, J = 9.10 Hz, 1H), 8.35 (dd, J =
9.10, 1.90 Hz, 1H), 8.81−8.84 (m, 1H), 9.37 (s, 1H); 13C NMR
(126 MHz, acetonitrile-d3) δ 14.4, 66.1, 108.1, 121.5, 123.9,
124.3 (q, J = 4.8 Hz), 124.7 (q, J = 271.8 Hz), 133.0 (q, J = 34.3
Hz), 134.7 (q, J = 2.9 Hz), 142.7, 149.4, 169.0, 175.1.
Ethyl 4-hydroxy-8-(trifluoromethyl)quinoline-3-car-
boxylate (7). Diethyl 2-(((2-(trifluoromethyl)phenyl)amino)-
methylene)malonate 6 (30.00 g, 90.6 mmol) was dissolved in
toluene (450 mL) and reacted in the same manner as 1. After
the resulting amber solution stood at room temperature solids
precipitated and the solution was filtered. After drying, 13.85 g
(94.8% potency) of 7 was collected as a tan solid for 50.8% yield.
The filtrate was concentrated on a rotary evaporator and
reconstituted in toluene and resubjected to the reaction
conditions. An additional 3.01 g (98.5% potency) was collected
for an additional 11.5% yield. The overall yield for the process
was 62.3%. Mp = 219.2 to 220.2 °C; HPLC: tR = 3.95 min. 1H
NMR (499 MHz, acetonitrile-d3) δ 1.43 (t, J = 7.09 Hz, 3H),
4.58 (q, J = 7.10 Hz, 2H), 8.04 (dd, J = 8.30, 7.80 Hz, 1H), 8.52
(d, J = 7.83 Hz, 1H), 8.82 (d, J = 8.31 Hz, 1H), 9.18 (s, 1H); 13C
NMR (126 MHz, acetonitrile-d3) δ 14.5, 66.4, 108.2, 122.1 (q, J
= 33.5 Hz), 123.2, 124.4 (q, J = 272.8 Hz), 131.1, 131.4, 137.1,
137.5 (q, J = 5.2 Hz), 148.9, 168.9, 175.4.
2.53 (s, 3H), 2.67 (s, 3H), 4.55 (q, J = 6.85 Hz, 2H), 7.83 (s,
1H), 8.16 (s, 1H), 9.05 (s, 1H); 13C NMR (126 MHz,
acetonitrile-d3) δ 14.5, 17.7, 22.0, 65.8, 106.8, 122.2, 123.0,
131.1, 138.7, 142.0, 143.1, 145.4, 169.5, 174.3.
ASSOCIATED CONTENT
■
S
* Supporting Information
Details of the equipment used. This material is available free of
charge via the Internet at http://pubs.acs.org.
AUTHOR INFORMATION
■
Corresponding Author
*E-mail: whiteti@lilly.com.
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
The authors wish to thank John Rizzo and Scott Bradley for
helpful discussions and Ed DeWeese and Paul Millenbaugh of
D&M Continuous Solutions for their support of this work and
John Howell for preparing starting materials. The authors also
thank Bret Huff for initiating and sponsoring the flow chemistry
efforts in development at Eli Lilly.
REFERENCES
■
(1) Mistry, S. N.; Valant, C.; Sexton, P. M.; Capuano, B.;
Christopoulos, A.; Scammells, P. J. J. Med. Chem. 2013, 56, 5151−5172.
(2) Zhang, Y.; Clark, J. A.; Connelly, M. C.; Zhu, F.; Min, J.;
Guiguemde, W. A.; Pradhan, A.; Iyer, L.; Furimsky, A.; Gow, J.;
Parman, T.; Mazouni, F. E.; Phillips, M. A.; Kyle, D. E.; Mirsalis, J.;
Guy, R. K. J. Med. Chem. 2012, 55, 4205−4219.
(3) Milata, V.; Claramunt, R. M.; Elguero, J.; Zalupsky, P. Targets in
heterocyclic systems: chemistry and properties; Royal Society of
Chemistry: London, 2000; Vol. 4, pp 167−203.
(4) Diphenyl ether melts at 25 °C, and both solvents can be difficult to
remove from isolated solids.
Ethyl 4-hydroxy-6-methoxyquinoline-3-carboxylate
(9). Diethyl 2-(((4-methoxyphenyl)amino)methylene)malonate
8 (30.00 g, 102.3 mmol) was dissolved in toluene (450 mL) and
reacted in the same manner as 1. After drying 8.85 g (91.3%
potency) of 9 was collected as a tan solid for a 32.0% yield. The
filtrate was concentrated and resubjected to the reaction
conditions. An additional 3.88 g (92.1% potency) was collected
for an additional 14.1% yield. This filtrate was concentrated and
resubjected to the reaction conditions a second time. An
additional 1.63 g (91.8% potency) was collected for an
additional 5.9% yield. The overall yield for this process was
52.0%. Mp = 292.2 to 293.4 °C; HPLC: tR = 5.18 min. 1H NMR
(499 MHz, acetonitrile-d3) δ 1.39 (t, J = 7.09 Hz, 3H), 3.95 (s,
3H), 4.52 (q, J = 7.09 Hz, 2H), 7.71 (dd, J = 9.30, 2.90 Hz, 1H),
7.74 (d, J = 2.90 Hz, 1H), 7.98 (d, J = 9.29 Hz, 1H), 9.03 (s,
1H); 13C NMR (126 MHz, acetonitrile-d3) δ 14.6, 57.6, 66.0,
104.3, 106.9, 123.7, 123.7, 131.4, 136.8, 144.4, 162.7, 169.7,
173.3
Ethyl 4-hydroxy-6,8-dimethylquinoline-3-carboxylate
(11). Diethyl 2-(((2,4-dimethylphenyl)amino)methylene)-
malonate 10 (30.19 g, 103.6 mmol) was dissolved in toluene
(465 mL) and reacted in the same manner as 1 but with a
reactor setpoint temperature of 250 °C. After drying, 10.06 g
(90.0% potency) of 11 were collected as a tan powder for a
35.6% yield. mp =282.8 to 285.2 °C; HPLC: tR = 4.71 min. 1H
NMR (499 MHz, acetonitrile-d3) δ 1.42 (t, J = 7.09 Hz, 3H),
(5) Gould, G.; Jacobs, W. A. J. Am. Chem. Soc. 1939, 61, 2890.
(6) Atkins, R. J.; Breen, G. F.; Crawford, L. P.; Grinter, T. J.; Harris,
M. A.; Hayes, J. F.; Moores, C. J.; Saunders, R. N.; Share, A. C.;
Walsgrove, T. C.; Wicks, C. Org. Process Res. Dev. 1997, 1, 185−197.
(7) Dorow, R. L.; Herrinton, P. M.; Hohler, R. A.; Maloney, M. T.;
Mauragis, M. A.; McGhee, W. E.; Moeslein, J. A.; Strohbach, J. W.;
Veley, M. F. Org. Process Res. Dev. 2006, 10, 493−499.
(8) Romek, A.; Opatz, T. Eur. J. Org. Chem. 2010, 30, 5841−5849.
(9) (a) Bergamelli, F.; Iannelli, M.; Marafie, J. A.; Moseley, J. D. Org.
Process Res. Dev. 2010, 14, 926−930. (b) Nishioka, M.; Miyakawa, M.;
Daino, Y.; Kataoka, H.; Koda, H.; Sato, K.; Suzuki, T. M. Ind. Eng.
Chem. Res. 2013, 52, 4683−4687.
(10) Lengyel, L.; Nagy, T. Z.; Sipos, G.; Jones, R.; Dorman, G.; Urge,
L.; Darvas, F. Tetrahedron Lett. 2012, 53, 738−743.
(11) (a) Cablewski, T.; Gurr, P. A.; Pajalic, P. J.; Strauss, C. R. Green
Chem. 2000, 2, 25−28. (b) Dave, C. G.; Joshipura, H. M. Indian J.
Chem., Sect B 2002, 41, 650−652. (c) Cernuchova, P.; Vo-Thanh, G.;
Milata, V.; Loupy, A. Heterocycles 2004, 64, 177.
(12) Roberge, D. M.; Ducry, L.; Bieler, N.; Cretton, P.; Zimmermann,
B. Chem. Eng. Technol. 2005, 28, 318−323.
(13) Hartman, R. L. Org. Process Res. Dev. 2012, 16, 870−887.
(14) Braden, T. M.; Gonzalez, M. A.; Jines, A. M.; Johnson, M. D.;
Sun, W.-M. Reactors and methods for processing reactants therein.
WO/2009/023515, CAN 150:216733.2009.
(15) Temperature control was performed by the controller for the
reactor, and the thermocouple was in contact with the wall of the
reactor to avoid overheating while the reactor was empty between
cycles.
I
dx.doi.org/10.1021/op500239f | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
(16) Parr reactors have a rating of 3000 psi, and this system was set up
with a pressure relief valve set at 1400 psi and pumps set to stop at 1000
psi.
(39) Liu, P.; Niu, Y.; Wang, C.; Sun, Q.; Zhai, Y.; Yu, J.; Sun, J.; Xu, F.;
Yan, G.; Huang, W.; Liang, L.; Xu, P. Eur. J. Med. Chem. 2014, 79, 413−
421.
(17) A great deal of effort was put into this as the high temperatures
for successful cyclization are also close to decomposition temperatures
for some molecules. When temperature control was performed off the
liquid temperature, the empty reactor would overheat prior to the next
charge of material, and tars would form. Once tars formed, even the
most granular of solids could get trapped and have difficulty exiting the
reactor. No heat exchanger was necessary on this scale, but the larger
scale equipment set discussed later in this paper would require a heat
exchanger on the inlet and outlet.
(18) Lager, E.; Andersson, P.; Nilsson, J.; Pettersson, I.; Nielsen, E.
Ø.; Nielsen, M.; Sterner, O.; Liljefors, T. J. Med. Chem. 2006, 49, 2526−
2533.
(19) Pace, V.; Hoyos, P.; Fernandez, M.; Sinisterra, J. V.; Alcantara, A.
R. Green Chem. 2010, 12, 1380−1382.
(20) Barda, D. A.; Mader, M. M. Imidazo [4, 5-C] Quinolin-2-one
Compound and its Use as PI3Kinase/MTOR Dual Inhibitor. WO
2012/097039.
(21) Thermal expansions are based on NRTL model simulation using
the NIST database in Aspen V 7.1
(22) Even more desirable was direct telescoping of the reaction of 4-
bromoaniline and diethyl ethoxymethylenemalonate followed by
cyclization to 2. Reduced reactivity and lower purity of the reaction
resulted in the decision to run the process stepwise.
(23) See Supporting Information for some details around these
reactors and a paper regarding this equipment set and their safe use will
be published in due course.
(24) Strauss, C. R. Aust. J. Chem. 1999, 52, 83−96.
(25) (a) White, T. D.; Berglund, K. D.; Groh, J. M.; Johnson, M. D.;
Miller, R. D.; Yates, M. H. Org. Process Res. Dev. 2012, 16, 939−957.
(b) Johnson, M. D.; May, S. A.; Calvin, J. R.; Remacle, J.; Stout, J. R.;
Diseroad, W. D.; Zaborenko, N.; Haeberle, B. D.; Sun, W. M.; Miller,
M. T.; Brennan, J. Org. Process Res. Dev. 2012, 16, 1017−1038.
(26) Scavenging reagents such as acetic anhydride were considered to
remove ethanol from the system but were not explored.
(27) Typical HPLC area % impurity levels were 2% of the
decarboxylated product of 2, 1% toluene, and 0.6% 1.
(28) Mass balance was expected to be low due to solvent loss during
expansion upon exiting the system via the hydrocyclone. The mass
deficiency is a close approximation of the solvent loss during this
expansion.
(29) This yield number is higher than reported in the experimental
because it is based on the potency corrected 1 placed back into recycle
rather than a dry in/dry out calculation, and the product dry weight is
not corrected for potency.
(30) This run utilized a first and second recycle, and a rinse was added
to the reactor to hold at temperature as the slurry was filtered and the
mother liquor concentrated and redissolved for reprocessing.
(31) Niedermeier, S.; Singethan, K.; Rohrer, S. G.; Matz, M.; Kossner,
M.; Diederich, S.; Maisner, A.; Schmitz, J.; Hiltensperger, G.; Baumann,
K.; Holzgrabe, U.; Schneider-Schaulies, J. J. Med. Chem. 2009, 52,
4257−4265.
(32) Thomas, K. D.; Adhikari, A. V.; Telkar, S.; Chowdhury, I. H.;
Mahmood, R.; Pal, N. K.; Row, G.; Sumesh, E. Eur. J. Med. Chem. 2011,
46, 5283−5292.
(33) Goncalves, V.; Brannigan, J. A.; Whalley, D.; Ansell, K. H.; Saxty,
B.; Holder, A. A.; Wilkinson, A. J.; Tate, E. W.; Leatherbrow, R. J. J.
Med. Chem. 2012, 55, 3578−3582.
(34) Moerdyk, J. P.; Speelman, A. L.; Kuper, K. E.; Heiberger, B. R.;
TerLouw, R. P.; Zeller, D. J.; Radler, A. J.; Gillmore, J. G. J. Photochem.
Photobiol., A 2009, 205, 84−92.
(35) Toluene recovery and recycle could also be envisioned providing
an additional benefit.
(36) http://www.isco.com/ (accessed Sept 2014).
(37) http://www.emerson.com/en-US/Pages/Default.aspx (accessed
Sept 2014).
(38) See Supporting Information for a photo of one of these gaskets.
J
dx.doi.org/10.1021/op500239f | Org. Process Res. Dev. XXXX, XXX, XXX−XXX