Hydrogenation of a pharmaceutical intermediate by a continuous stirred tank reactor system

Article abstract of DOI:10.1021/op800170r

A proof of concept study on the continuous hydrogenation of a pharmaceutical intermediate is presented. A slurry feed of CP- 548495, a dintro intermediate in a smoking cessation drug, was reduced in a two-reactor continuous stirred tank train to the diami

Full text of DOI:10.1021/op800170r

Organic Process Research & Development 2009, 13, 629–633
Technical Notes
Hydrogenation of a Pharmaceutical Intermediate by a Continuous Stirred Tank
Reactor System
John G. Van Alsten,* Matthew L. Jorgensen, and David J. am Ende
Engineering Technologies Laboratory, Pfizer Global R&D, Groton, Connecticut 06340
Abstract:
commercial scale, including continuous stirred tank reactors
(CSTR), venturi, fluidized bed, loop, and trickle bed reactors.
Excellent reviews of the relative merits of these have been
A proof of concept study on the continuous hydrogenation of a
pharmaceutical intermediate is presented. A slurry feed of CP-
2
548495, a dintro intermediate in a smoking cessation drug, was
published. Because a pharmaceutical processing facility must
reduced in a two-reactor continuous stirred tank train to the
diamine product. The reactors were found to operate within
different kinetic regimes, with the upstream reactor exhibiting
hydrogen mass transfer limited behavior, and the downstream
reactor, substrate concentration limited behavior. A longer-term
catalyst degradation study was conducted to explore the potential
for extended operation.
be capable of processing a wide variety of feeds under widely
varying conditions, we contend that the CSTR or loop reactor
is the most practical option. Since the smallest loop configu-
ration requires a minimum of 15 L, its use for early phase
3
process development is quite limited. Hence, the CSTR is the
most useful reactor configuration for relatively small scale
development studies. Any catalyst currently in use in a stirred
tank can be utilized in the CSTR, and batch kinetic data may
4
be translated to a CSTR train with standard methods. Wang
Introduction
has demonstrated a CSTR system in the homogeneous reduction
A significant amount of work has transpired regarding the
potential of supplanting batch reactions and processes with
continuous ones. One of the largest reaction classes is hydro-
genations, which by one estimate constitute ∼20% of all fine
5
of an intermediate in the synthesis of mibefradil. We report
here on extending this approach to heterogeneous (substrate
slurry and solid catalyst) hydrogenation of an intermediate used
for the manufacture of a drug for smoking cessation.
Chemistry. The reaction under consideration is the reduction
of the aromatic nitro groups of CP-548495, an intermediate in
the synthesis of varenicline, the active ingredient of Chantix
1
chemical synthetic reactions. These steps appear particularly
ripe for improvement by continuous technology for numerous
reasons:
*
Safety: Minimization of hydrogen inventory and smaller-diameter
vessels for increased mechanical integrity under high
pressure. A decreased need for manual catalyst handling,
often considered the most hazardous of operations. More
stable reaction control by operating under steady state.
(U.S.A) or Champix (EU). The commercial process for this
material is a telescoped reduction of the dinitro intermediate to
the diamine followed by cyclization with glyoxal and hydrolysis
6
to form the API (Scheme 1). Since considerable effort had
already been expended in the optimization of the reaction
conditions and because we sought to explore a continuous
“drop-in” alternative for the current batch process, the study
was constrained to conditions similar to those used com-
mercially, which involve contacting hydrogen and about
3 wt % of 5% Pd/C with a 5% solution of substrate in a mixture
of isopropanol/water until the reaction is complete. The product
*
*
Quality: With smaller volumes, hydrogen mass transfer rates may be
more tightly controlled, allowing the potential to minimize
impurities due to either hydrogen starvation or over-
reduction.
Cost:
The potential to increase catalyst utility from the current
typical range of 20-100 kg product/kg catalyst to perhaps
∼
1000 kg/kg. The reduction of cycle times inherent in
charging/purging/filtering operations. Reduced waste
streams from catalysts and washes.
(2) (a) Edvardsson, J.; Irandoust, S. J. Am. Oil Chem. Soc. 1994, 71 (3),
2
35. (b) Jenk, J. F. In Heterogeneous Catalysis and Fine Chemicals II;
Guisnet, M., Ed.; Elsevier Science Publishers: Dordrecht, The Neth-
erlands, 1991. (c) Stitt, E. H. Chem. Eng. J. 2002, (90), 47. (d) Patterson,
H. B. W. Hydrogenation of Fats and Oils: Theory and Practice;
American Oil Chemists’ Society (AOCS Press): Champaign, IL, 1994.
While the benefits appear significant, continuous operation
of a three-phase (solid-liquid-gas) reactor system requires
increased operational sophistication compared to the traditional
batch reactor. A variety of configurations are available at
(
e) Puri, P. S. J. Am. Oil Chem. Soc. 1980, (November), 850A.
(
3) Product literature and personal communications, BUSS ChemTech AG,
Switzerland. .
(
4) (a) Levenspiel, O. Chemical Reaction Engineering; John Wiley & Sons:
New York, 1999. (b) Murthy, A. K. S. Chem. Eng. 1999, 106 (10), 94.
*
Author to whom correspondence may be sent. E-mail: john.g.vanalsten@
(5) Wang, S.; Kienzle, F. Org. Process Res. DeV. 1998, 2 (4), 226.
(6) Busch, F. R.; Hawkins, J. M.; Mustakis, L. G.; Sinay, T. G.; Watson,
T. J. N. Preparation of High Purity Substituted Quinoxaline. WO/2006/
090236 A1, 2006.
pfizer.com.
(
1) Doku, N.; Verboom, W.; Reinhoudt, D. N.; van den Berg, A.
Tetrahedron 2005, 61 (11), 2733.
1
0.1021/op800170r CCC: $40.75  2009 American Chemical Society
Vol. 13, No. 3, 2009 / Organic Process Research & Development
•
629
Published on Web 02/04/2009

Scheme 1. Synthetic sequence to varenicline
Figure 1. CSTR hydrogenation system. Substrate slurry and
hydrogen are fed to reactors with fixed catalyst beds. Level is
maintained by dip tube. Slug flow of product liquid and excess
gas is reduced to near atmospheric pressure over a control
valve.
of the reduction is not isolated from solution but is taken as is
to the cyclization. The yield over these three steps is ∼80%.
The hydrogenation of aromatic nitro groups is the subject
7
of an enormous body of literature. The particular reaction
pathway appears to be dependent on the precise reaction
conditions, e.g., catalyst, temperature, and substrate concentra-
tion. In the hydrogenation of neat aniline, for example,
appreciable quantities of azobenzene and azoxybenzene are
Scheme 2. Reduction of the CP-548495 showing identified
intermediates
8
formed. Kinetic experiments clearly show that in our system
the material undergoes at least three separate adsorption/
reduction steps on the path to complete reaction, as the nitroso/
nitro, nitroso/hydroxylamine, and diamine are all identified in
the reactor supernatants (Scheme 2). Analysis of process streams
allows us to quantify conversion in terms of the number of
equivalents of hydrogen by:
Conversion )
(
Laboratory Alliance Series III) or as a slurry via diaphragm
[
NO ⁄ NO ] + 3[NH ⁄ NO ] + 6[NH ⁄ NH ]
2 2 2 2 2
(1)
pump (Prominent gamma2L), with a check valve to restrict
backflow. Hydrogen gas is introduced into reactor #1 via mass
flow controller (Brooks) and serves as both reactant and source
of hydrostatic head to drive the product solution downstream.
From reactor #2 the process stream flows across a control valve
6
CP-548495 is a compound with a high thermal potential but
a reasonably high onset temperature near 230 °C. The adiabatic
temperature rise of the hydrogenation reaction is 57 °C so that
even under these conditions there is little decomposition hazard
from the starting material. The properties of the intermediates
are unknown, however, although hydroxylamines are known
to be relatively unstable materials. One of the advantages to a
continuous reactor is that it operates at steady state, with reaction
rate and heat evolution controlled by the rate at which reactants
are fed to the reactor. Hence, there is no exotherm per se in an
ideal continuous reactor. It will also be demonstrated that the
system operates in a regime where the concentration of
intermediates is low, further reducing the risk of an undesirable
decomposition event.
Reactor System. The reaction system (Figure 1) consists
of two 0.1 L stirred, jacketed vessels with inlets and outlets for
feeds, products, and sampling. The reactors are constructed of
Hastelloy C and are 8.9 cm tall and 4.1 cm in diameter. Small
removable baffles may be inserted into the vessels, onto which
may be attached fixed catalyst beds constructed of fine wire
mesh. The impeller is a six-bladed vortexing system that is
designed to push fluid radially outward through the fixed catalyst
beds. Substrate may be fed either as a solution via HPLC pump
(Badger Meter, Inc., Tulsa, OK, U.S.A.) and into a nitrogen-
purged receiver vessel where excess gas is vented. While
operationally simple, the use of hydrogen as the fluid “pump”
prevents accurate quantification of gas uptake. The system was
constructed by Pressure Products Industries (Warminster, PA,
U.S.A.) and instrumented by Crest Integrators (Erie, PA,
U.S.A.). The reaction residence time is controlled by manipulat-
ing liquid feed rate and the reaction volume, the latter
determined by level control of the reactors and set by the height
of a stand pipe. The most significant challenge is the need to
confine the catalyst within the reactor volume while removing
the product-containing liquid stream. In larger-scale commercial
operations this is often handled in a dedicated external crossflow
9
filtration unit, which unfortunately, is not practical on the small
laboratory demonstration scale. Instead, we have chosen to either
confine the catalysts to a flow-through bed located annularly
within the reactor or utilize in situ crossflow filters. Neither
approach is without compromise, i.e., reduced catalyst activity
for the former or eventual filter blinding by the latter. In this
contribution we present results only from the use of fixed flow-
through beds.
(
7) (a) Rylander, P. N. Hydrogenation Methods; Academic Press: New
York, 1985. (b) Visentin, F.; Puxty, G.; Kut, O. M.; Hungerbuhler, K.
Ind. Eng. Chem. Res. 2006, 45, 4544.
(9) Catalyst RecoVery from Continuous Flow Reactors with Mott Hypulse
LSM Filters, technical publication. Mott Corporation: Connecticutt,
U.S.A., 1997.
(
8) Wisniak, J.; Klein, M. Ind. Eng. Prod. Res. DeV. 1984, 23, 44.
6
30
•
Vol. 13, No. 3, 2009 / Organic Process Research & Development

Figure 2. Distribution of intermediates during batch reduction.
Granular catalyst.
Figure 3. Batch kinetic experiments contrasting performance
of two morphologies of 5% Pd/C.
We have found that one of the highest barriers to effective
laboratory scale down is in handling slurries, where the low
momentum inherent with small flow rates makes solids
fluidization difficult at best. The study presented here required
a feed of reactant slurry to a reactor under a modest (5.2 barg)
head of hydrogen. To effect this we first wet milled the reactant
to reduce its particle size from ∼250 to 50 µm. This allowed
particle fluidization in a stirred feed tank sufficient to be
successfully injected into the reactor using the high-pressure
diaphragm pump (Prominent gamma/2 L).
Catalyst Selection. All reactions described herein were
performed under a hydrogen pressure of 5 barg and a temper-
ature of 50 °C. The two catalysts screened were 5% Pd on
carbon support, and were obtained in very different morphol-
ogies. One support was extruded carbon pellets of dimensions
homogenizing for ∼2 h. To begin a run, 1.3 g of catalyst was
loaded into each of two screened containers which were attached
to the removable baffle for each vessel. The reactors were then
installed in the system. Reactor #1 was isolated from down-
stream by closing a valve, and then 50 mL of reactant slurry
was injected into it; 50 mL of solvent was charged by syringe
to reactor #2. The system was purged of air by five pressuriza-
tion/vent cycles with 0.7 barg nitrogen, and heated to 50 °C.
Five pressurization/vent cycles of hydrogen at 5.2 barg were
then performed, and agitation commenced. The reaction was
sampled for completion after ∼1 h and the feed of CP-548495
slurry begun at a rate of 0.36 mL/min. (A residence time of
140 min.) The isolation valve was opened to allow material
flow to reactor #2. After about ten more minutes, the outlet
valve was opened to enable product and excess hydrogen to
exit into the receiver, which was mounted on a weigh scale.
Samples of the reactors’ contents were collected periodically
via needle valves at the bottom of each vessel. The system was
operated in continuous mode for about 10 h. Material from the
reduction was not taken on to an isolated product.
Physically, the reaction proceeds from a 20 L/kg slurry in a
mixture of 80/20 isopropanol/water and becomes a solution at
a conversion of CP-548495 of about 70%. As our small flow
rates preclude transport of slurry from reactor to reactor, this
fixes the minimum residence time in the vessel to about 50
min, or an overall residence time of 100 min.
Samples were analyzed by HPLC (Agilent) using a C18
column and an acetonitrile/water (0.1% formic acid) gradient,
with detection at 230 nm. HPLC/MS (Agilent) was performed
on selected samples for intermediate identification. Assignments
were made by the presence of the molecular ion in each case.
In the case of the nitroamine, which has a mass identical to
that of the nitroso/hydroxlamine, identification was further
refined by the absence of an M-30 (NO) peak that was observed
in high intensity in the spectrum of the nitroso/nitro species.
∼
2 mm in diameter and 5 mm in length (Englehard Corp.,
SOC10161). The second was a granular carbon with randomly
shaped particles ranging from 5 to 30 µm in size (Englehard
Corp., CD04063). These were retained within the reactors by
sealing within metal mesh, which was then fixed to the
thermowells and baffles within the reactors. Smaller catalyst
particles common to batch processing were not investigated
because of our inability to retain them in a mesh large enough
to allow unrestricted liquid flow. Batch screening tests dem-
onstrated a profound difference in activity of the two materials,
with the granular material being ∼3-4× more active than the
extrudate. In each case two reaction intermediates are observable
by HPLC. (Figure 2) The granular material exhibits kinetics
indicative of both hydrogen-limited and substrate-limited re-
gimes with conversion essentially linear with time up to ∼90%,
indicative of a regime limited by the mass transfer of hydrogen
into the solution. (Figure 3) Beyond this level, conversion slows
as substrate concentration diminishes. All further experimenta-
tion was conducted exclusively with the granular catalyst.
Experimental Section
Hydrogen (Airgas UHP, 99.999%) was supplied by tank.
CP548495 was supplied by Pfizer Global Manufacturing and
used as received. Isopropanol was laboratory grade, and water
was from the tap.
About 1.25 L of feed was prepared by charging 60 g of
CP548495 and 1.2 L of 80/20 isopropanol/water mixture to a
vessel equipped with an overhead homogenizer (IKA T-25) and
Results
Measured over time on stream, the concentrations of the
intermediates are sensitive indicators of the attainment of steady
state and the natural fluctuations inherent in the system. Figure
4 illustrates that the relative product distribution in reactor #1
fluctuates by as by as much as 7% in terms of residual dinitro
material. Extension of the batch reactor behavior shows that
Vol. 13, No. 3, 2009 / Organic Process Research & Development
•
631

Figure 6. Correlation of reaction rate between reactors,
showing how reactor #2 compensates for fluctuation in conver-
sion from reactor #1.
Figure 4. Product distribution in reactor #1 measured during
time of operation.
Figure 5. Differential conversion in each reactor over time of
operation.
Figure 7. Outlet product compositions of both reactors mea-
sured over time of operation.
reactor #1 operates exclusively within the kinetic regime limited
by hydrogen mass transfer. Overall conversion also fluctuates
in reactor #1 by about 6% beyond startup (Figure 5). While
we have not isolated the cause of these fluctuations, two sources
appear likely. The first might be due to slight variations in the
charge of slurry delivered by the diaphragm pump. The second
might be caused by the variations in the reactor residence time
caused by the slug flow nature of our system, which forces the
reactor level to alternate between the level of the dip pipe to
slightly overfilled. Regardless of these fluctuations, comparison
with Figure 3 indicates that this reactor is operating in the kinetic
regime limited by hydrogen mass transfer.
The main function of the downstream reactor is to carry the
reaction to completion. In contrast to reactor #1, reactor #2
operates in the kinetic regime limited by substrate concentration.
Figure 5 shows that the rate of reaction is on average only 18%
of that in the first reactor. There is also an inverse cor-
respondence in reaction rate between reactors # 1 and #2: when
conversion in #1 is highest, the reaction rate in #2 is relatively
low, and when conversion in #1 dips, the rate in #2 increases.
Figure 8. Measurement of catalyst activity as a function of the
mass of CP-548495 processed.
operation that might be on stream for days, weeks, or months. In
most real catalysis the temporal degradation of catalyst activity
makes establishment of a truly steady state impossible and
eventually limits the total quantity of material processable. Due to
time limitations, the total catalyst utility of the work presented here
is still rather low at 7 kg substrate/kg catalyst, significantly lower
than the 20-30 kg/kg typical of batch hydrogenations in our
industry. To investigate further the possible long-term impact of
activity degradation, a series of batch experiments was performed
in which we recycled catalyst to fresh feeds and deliberately ran
to incomplete conversions. Figure 8 shows that catalyst activity
undergoes a marked decrease in the early cycles before plateauing
(or with a mild rate of decrease) near about 30 kg/kg. For long-
(Figure 6) Thus, reactor #2 automatically compensates for
fluctuations in the performance of the first reactor. As a result,
the total variation in outlet composition of the reactor train at
steady state is quite small, ranging from 97.4 to 98.6% CP-
548496 (Figure 7).
Two of the most difficult questions to address for bench-scale
laboratory evaluation of continuous processes are (1) achievement
of steady state in the system and (2) evaluating viability of extended
6
32
•
Vol. 13, No. 3, 2009 / Organic Process Research & Development

term operation it is apparent that the processing rate would thus
be about 50-60% of that demonstrated in our short duration run.
Conversely, this could be ameliorated by approximately doubling
the catalyst charge.
Conclusions
We have presented a proof of concept study on the potential
for continuous hydrogenation of pharmaceutical intermediates. A
laboratory CSTR system utilizing a slurry feed of substrate in
solvent was run for ∼10 h to demonstrate the feasibility of this
approach. While significant work is still required for com-
mercialization, we believe that many of the engineering difficulties
inherent in small-scale operation will diminish on scale-
up.
The work presented here represents the initial steps on a long
path to commercialization. A significant body of engineering work
on catalyst handling, long-term stability, startup, shutdown, and a
myriad of other issues remains to be addressed. These issues are
balanced somewhat by the fact that many of the problems inherent
in laboratory scale down are more readily resolved when moving
to larger scale facilities. For example, operational difficulties with
transport of slurries, feed control, and level control are much less
problematic at large scale. We thus submit that the CSTR platform
represents an attractive and potentially quite practical commercial
alternative to batch processing.
Received for review July 21, 2008.
OP800170R
Vol. 13, No. 3, 2009 / Organic Process Research & Development
•
633