Optimization and scale-up of a Pd-catalyzed aromatic C-N bond formation: A key step in the synthesis of a novel 5-HT1B receptor antagonist

Article abstract of DOI:10.1021/op8000146

Searching for the best synthetic route for a given target molecule is a complex task and, by the same token, a key deliverable from a process R&D department. In this vein the challenge for our group was to identify a sustainable manufacturing process for a chiral compound, AR-A2, to be developed for the treatment of certain neurological disorders. Besides designing a method for assembling the core (R)-2-aminotetralin nucleus, a key feature in the overall synthesis was to provide a robust procedure for creating a new C-N bond between an aromatic ring and a heterocyclic moiety. The methodology employed a Buchwald-Hartwig coupling, and a highly efficient catalytic process was developed using Pd(OAc)₂ as precatalyst, with loadings as low as 0.47 mol % (in laboratory trials one order of magnitude lower) together with (R)-BINAP as ligand. Optimizing the reaction conditions allowed a virtually quantitative conversion of the brominated aromatic substrate after heating to 110-115 °C in toluene for 4 h. Telescoping this step with a succeeding catalytic hydrogenation to effect an N-debenzylation, followed by precipitation of the benzoate salt offered an overall yield for the two consecutive steps of 88% at 125-kg batch size, combined with excellent stereochemical product purity of 98% ee.

Full text of DOI:10.1021/op8000146

Organic Process Research & Development 2008, 12, 512–521
Optimization and Scale-up of a Pd-Catalyzed Aromatic C-N Bond Formation: A Key
Step in the Synthesis of a Novel 5-HT1B Receptor Antagonist
Hans-Jürgen Federsel,*,† Martin Hedberg,*,‡ Fredrik R. Qvarnström,‡ and Wei Tian‡
Global Process R&D and Process Chemistry, Process R&D Södertälje, AstraZeneca, 151 85 Södertälje, Sweden
Abstract:
coupling step between an aromatic moiety and a nitrogen-
containing reactant to enable a commercially viable manufactur-
ing route for AR-A2 (1).
Searching for the best synthetic route for a given target molecule
is a complex task and, by the same token, a key deliverable from
a process R&D department. In this vein the challenge for our
group was to identify a sustainable manufacturing process for a
chiral compound, AR-A2, to be developed for the treatment of
certain neurological disorders. Besides designing a method for
assembling the core (R)-2-aminotetralin nucleus, a key feature in
the overall synthesis was to provide a robust procedure for creating
a new C-N bond between an aromatic ring and a heterocyclic
moiety. The methodology employed a Buchwald-Hartwig cou-
pling, and a highly efficient catalytic process was developed using
Pd(OAc)2 as precatalyst, with loadings as low as 0.47 mol % (in
laboratory trials one order of magnitude lower) together with (R)-
BINAP as ligand. Optimizing the reaction conditions allowed a
virtually quantitative conversion of the brominated aromatic
substrate after heating to 110-115 °C in toluene for 4 h.
Telescoping this step with a succeeding catalytic hydrogenation
to effect an N-debenzylation, followed by precipitation of the
benzoate salt offered an overall yield for the two consecutive steps
of 88% at 125-kg batch size, combined with excellent stereochem-
ical product purity of 98% ee.
This is a chiral drug that was being developed by Astra-
Zeneca to treat an array of illnesses in the central nervous system
(CNS), such as anxiety and depression. Unfortunately, during
phase II clinical studies the decision was taken to discontinue
further studies of this novel drug, notably due to lack of efficacy.
The Buchwald-Hartwig procedure to create a new C-N bond
in a catalytic mode has undergone a tremendous improvement
in efficiency and output and reached a level of sophistication
that looked promising for future commercial production. Failing
to deliver a new drug to the market should, however, in no
way blur the achievement that has been made to turn what at
the time was seen as a promising academic synthetic protocol
with interesting scope to carry out a wide variety of aromatic
substitutions into a proven and viable methodology for large-
scale production.
From Flask to Reactor: Reaching the Best Overall Route.
It is not uncommon in today’s pharmaceutical industry to rely
on relatively poor methods for synthesizing desired compounds.
The weaknesses can be exemplified as low yields, long and
complex workup procedures including expensive and tedious
techniques for product purification, demanding reaction condi-
tions (pressure/temperature) which, when combined, lead to an
unsustainable cost of the final product from a marketing point
of view. This is the scenario often confronted in the early stages
of R&D, typically around the point of candidate drug (CD)
nomination, when process R&D will be expected to deliver the
first batch of kg amounts to support wider toxicological
investigations, besides development of a formulation feasible
for early clinical testing and for use in first-time-in-man studies.
In this regard the current project was no exception, and all in
all three generations of synthetic routes were identified,
developed, and put into practical use during its lifetime.2 In order
to provide the necessary background to the strategy used for
the synthesis and to bring the crucial C-N bond-forming step
into a broader context, the different approaches are briefly
reviewed.
Introduction
Construction of complex molecules requires an arsenal of
methods and procedures. Depending on the purpose of the work,
the quality of the transformations applied will have to meet
varying standards. Thus, whilst small-scale bench chemistry in
the laboratory should aim at being intrinsically safe and reliable,
the mere fact that the amounts of material handledsintermediates,
building blocks, reagents, etc.sare limited (typically
<5-50 g) reduces the stringency in outcome and the perfor-
mance requirements needed. However, as soon as scale-up is
envisaged, the circumstances change dramatically. When aiming
at producing multikilogram quantities, the chemistry has to be
understood to a much higher degree in order to capture potential
risks in the form of exotherms and runaway events. Moreover,
it is essential to optimize yields and process throughput so that
the cost of goods for the active pharmaceutical ingredient (API),
a highly significant parameter, can reach acceptable levels.1 In
this case study,2 the prime focus is development of the crucial
* Corresponding authors. E-mail: hans-jurgen.federsel@astrazeneca.com (HJF);
martin.h.hedberg@astrazeneca.com (MH).
† Global Process R&D.
‡ Process Chemistry, Process R&D So¨derta¨lje.
(1) Federsel, H.-J. Nat. ReV. Drug DiscoVery 2003, 2 (8), 654–664.
(2) Federsel, H.-J.; Hedberg, M.; Qvarnstro¨m, F. R.; Sjo¨gren, M. P. T.;
Tian, W. Acc. Chem. Res. 2007, 40 (12), 1377–1384.
512
•
Vol. 12, No. 3, 2008 / Organic Process Research & Development
10.1021/op8000146 CCC: $40.75
2008 American Chemical Society
Published on Web 04/26/2008

Scheme 1. First-generation route to AR-A2 (1) applied by medicinal chemistry.a
a Abbreviations: dba ) 1,3-benzylideneacetone; CDI ) 1,1′-carbonyl-di-imidazole; DMF ) N,N-dimethylformamide.
Initially, the synthesis adopted by Medicinal Chemistry was
seen as the best option. Their approach relied on the availability
of chiral aminotetralin 2 (see Scheme 1) in high stereochemical
purity, which had the advantage that there was no need for a
de noVo construction of the core tetrahydronaphthalene part nor
for generating the desired configuration (R). A major downside,
however, was the limited availability and cost ($50,000-100,000/
kg) of the des-methyl-tetralin starting material. This starting
material required that the methyl group be introduced by using
a problematic low-temperature lithiation/methylation procedure.
So, in spite of a reasonable 21-38% overall yield, this route
(see Scheme 1) was abandoned.
In the second-generation route, the synthesis was further
back-integrated such that the tetralin moiety was created from
3-methylphenyl acetic acid (3). A key step in this sequence was
the formation of tetralone 5 from precursor 4a3 using ethene as
an ideally suited C2-building block under Friedel-Crafts-like
conditions.4 This transformation proceeded surprisingly well,
and yields of 60-70% were generally obtained. The main
bottleneck in this considerably more scale-up friendly synthetic
route was the reductive amination of ketone 5 [using NaB-
H(OAc)3]5 followed by a one-pot resolution with unnatural
tartaric acid leading to 6 (see Scheme 2). Even after investing
a large resource the total yield in this two-step process could
not be increased above 25% which, eventually, disqualified this
option for further development.
Nonetheless, inspired by the improvements seen when
running the manufacture on pilot-plant scale, it was felt that
the basic route was correct. The intrinsic weakness was the
resolution, and this eventually led to the use of (S)-1-phenyl-
ethylamine (8) to induce asymmetry in the tetralin. This
approach has been shown to work when performing the imine
reduction under appropriate conditions.6 Thus,when applying
a two-stage procedure consisting of imine formation using 8
as chiral handle, followed by reduction (NaBH4 in MeOH/i-
PrOH), resulted in a 4:1 mixture of substituted tetralin 9 in
favour of the desired (R,S) diastereomer. Interestingly, carrying
(3) Lewis, E. E.; Elderfield, R. C. J. Org. Chem. 1940, 5 (3), 290–299.
(4) (a) Sims, J. J.; Selman, L. H.; Cadogan, M. Org. Synth. 1971, 51,
109–112. (b) Hunden, D. C. Org. Prep. Proced. Int. 1984, 16 (3-4),
294–297.
(5) Abel-Magid, A. F.; Carson, K. G.; Harris, B. D.; Maryanoff, C. A.;
Shah, R. D. J. Org. Chem. 1996, 61 (11), 3849–3862.
(6) Kanai, Y.; Saitou, K.; Koki, U. Japan Kokai Koho 10,072,411, 1998.
Vol. 12, No. 3, 2008 / Organic Process Research & Development
•
513

Scheme 2. Second-generation route to AR-A2: an improved, scale-up friendly synthesis of 1 that requires resolution of a
racemic intermediate.
Scheme 3. Third-generation route: a diastereoselective synthesis of AR-A2 (1) using (S)-1-phenylethylamine (8) as a crucial,
multi-tasking component.
it out as a one-pot reaction led to a 3:2 (R,S)/(S,S) diastereomer
composition. Formation of an HCl salt followed by a simple
polishing crystallization gave intermediate 10 with 96% de in
a 55% yield from tetralone 5. Running the reaction under a
slightly different experimental protocol (NaBH3CN) and solvent
system (MeOH/THF, 1:1) failed to give any asymmetric
induction at C-2.7 The chiral induction offered by (S)-phenyl-
ethylamine proved to be successful enough so that no resolution
would become necessary, which for obvious reasons was seen
as a decisive step-change in reaching better process efficiency.
This sequence, which also became the last one to be developed
and operated on pilot scale before closing the project, is shown
in Scheme 3.
Common to all routes described above was the need for
conducting a suitable coupling between the aromatic nucleus
of related but varying tetralin species and a defined piperazine
side chain somewhere along the synthetic sequence. The
protocol that was immediately identified as being capable of
(7) Romero, A. G.; Leiby, J. A.; McCall, R. B.; Piercey, M. F.; Smith,
M. W.; Han, F. J. Med. Chem. 1993, 36 (15), 2066–2074.
514
•
Vol. 12, No. 3, 2008 / Organic Process Research & Development

responding to this task and generating the required C-N bond
was the Buchwald-Hartwig procedure. From its invention in
the late 1990s it has been demonstrated to perform with a wide
variety of substrates and reactants;8,9 however, reports of scale-
up and bulk production are still scarce. The specific application
of this chemical transformation on our case and the tremendous
development that the method has undergone, together with
experience from upscaling gathered over the course of the
project life cycle, are herewith disseminated and explored in
detail.
Mastering C-N Bond Formation: Development and
Optimization. In order to put the versatility of this aromatic
substitution reaction and its relative user-friendliness into a
broader context, suffice it to say that in a very early and
premature version of our synthetic effort to create Ar-piperidino
motifs, other well-documented standard procedures were ap-
plied. Thus, one approach starting from a phenol precursor made
use of the Smiles rearrangement10 as a key reaction step, a
methodology emanating from the 1930s requiring stongly basic
conditions (NaH) and elevated temperatures (130 °C). This
created an anilino-derivative that could be transformed to the
piperazine-substituted product using a suitable reactant under
reductive conditions. In the other route, the heterocycle was
again constructed from an aniline by conducting a double
alkylation using an N,N-bis-2-chloroethylamine as building
block, a renowned carcinogen with pronounced toxic pro-
perties.11–13 These protocols were deemed unsuitable for large-
scale production by virtue of the many steps and hazardous
chemistry involved.
The need for alternative approaches is, thus, easily empha-
sized, and as it turned out, help was not too far away in the
form of a homogeneous, metal-catalyzed procedure pioneered
independently by Buchwald14 and Hartwig,15 respectively.
Attractive features of this route were the catalytic protocol
allowing low catalyst loading, and the high yields that had been
reported by the inventors. A further benefit was the conciseness
of the methodology that enabled abundantly available aromatic
chlorides or bromides to be directly coupled with the appropriate
amine moiety in just one step, totally avoiding the multistage
procedures described above. Hence, with this seemingly power-
ful tool at our disposal we focused all our attention to this novel
reaction type.
Starting with the first-generation synthesis of the target
molecule AR-A2, 1 (see Scheme 1) the Buchwald-Hartwig
reaction had already been successfully applied with 84-88%
yields in large laboratory-scale runs. Using Pd(dba)2 as the
precatalyst, initially at 3.1 mol %, but later demonstrating that
this could be more than halved (to 1.5 mol %) at 200-g scale,
with (R)-BINAP as ligand in the presence of NaO-t-Bu in
toluene, a method was developed that provided a straightforward
access to the desired piperazinyl-substituted tetralin motif.
When proceeding to the next synthetic sequence for 1, the
overall procedure for making the final product was considerably
improved as there were fewer steps and better prospects for
successful scale-up. In this case the coupling reaction with the
piperazine moiety was conducted on a mono-(6) rather than
dibenzyl-amino structure (13). An obvious risk was the compet-
ing intermolecular dimerization using a secondary amine. Much
to our surprise, no formation of this dimer could be detected,
and the yield of compound 7 remained >80%. The only
noticeable byproduct formed was the 8-H derivative 15, albeit
in sufficiently small amounts (<2%) not to cause any problems
downstream.
Closer examination of the reaction rate provided the op-
portunity to reduce the catalyst loading from 1.5 mol %. Indeed,
the chemistry was found to work at a catalyst loading as low
as 0.1 mol %, but reaction time increased from 4 to 20 h. A
process compromise was reached at 0.4 mol % catalyst added
in 2 portions. Encouraged by these findings, this step was scaled
up for pilot plant production where in total about 30 kg of the
(R)-8-piperazinyl-N-benzyl-2-aminotetralin 7was obtained in
excellent quality. This was ultimately converted to one batch
of 26 kg of 1.
Operating in a partially asymmetric fashion the third and
final route has the potential to provide the most efficient process,
and accordingly, a programme was devised to test for good
process understanding and to conduct optimization work along
the entire sequence. Focussing specifically on the Buchwald-
Hartwig step, the feeling was that the previous studies had
developed the method to a relatively advanced level, especially
considering the first pilot run. Unfortunately it turned out to be
the opposite, as the established experimental conditions on free
base substrate 10 (Scheme 3) gave disappointing results in the
initial pilot trials showing a pronounced increase in the 8-H
analogue 15. Widely differing results gave a lack of robust-
nesssin the worst case up to as much as 11% impurity 8-H
(15) on 23-kg scale in 600 L reactor size. Another problem
observed was the precipitation of the Pd catalyst from the
reaction mixture in the form of a solid, brownish material. The
existing protocol was obviously insufficiently robust to accom-
modate a new substrate (10), irrespective of the close resem-
blance to the structure of compound (6) that it had originally
been developed for. Thus, a thorough investigation into the
reasons and mechanisms behind the formation of the undesired
H-analogue 15 was launched.
(8) Old, D. W.; Wolfe, J. P.; Buchwald, S. L. J. Am. Chem. Soc. 1998,
120 (37), 9722–9723.
(9) Hartwig, J. F.; Kawatsura, M.; Hauck, S. I.; Shaughnessy, K. H.;
Alcazar-Roman, L. M. J. Org. Chem. 1999, 64 (15), 5575–5580.
(10) (a) Levy, A. A.; Rains, H. C.; Smiles, S. J. Chem. Soc. 1931, 3264–
3269. (b) Evans, W. J.; Smiles, S. J. Chem. Soc. 1935, 181–188.
(11) Halazy, S.; Jorand, C.; Perez, M. PCT Appl. WO 96/12713, 1996.
(12) Sommers, A. H.; Barnes, J. D. J. Am. Chem. Soc. 1955, 77 (15), 4171–
4172.
Based on on literature precedence,16 it was thought that the
amine reactant (1-methylpiperazine) was the hydride donor by
(13) Su¨sse, M.; Skubatz, R.; Demus, D.; Zaschke, H. J. Prakt. Chem. 1986,
328 (3), 349–358.
(14) Guram, A. S.; Rennels, R. A.; Buchwald, S. L. Angew. Chem., Int.
Ed. Engl. 1995, 34 (12), 1348–1350.
(16) Hartwig, J. F.; Richards, S.; Baran˜ano, D.; Paul, F. J. Am. Chem. Soc.
1996, 118 (15), 3626–3633.
(15) Louie, J.; Hartwig, J. F. Tetrahedron Lett. 1995, 36 (21), 3609–3612.
Vol. 12, No. 3, 2008 / Organic Process Research & Development
•
515

Table 1. Key experiments for the Buchwald-Hartwig coupling
product 11 after
8-H analogue 15 after
entry
altered conditionsa
3 h(% GC)
3 h(% GC)
remarks
1
2
3
4
5
6
7
8
rac-BINAP as ligand
Xantphos as ligand
(o-tol)3P as ligand
99
62
18
10
0
99
99
99
0.3
7
14
6
0
0.7
0.7
1.2
xylene 130 °C
NaOMe as base
NaOEt as base
xylene, 130 °C
0.1 equiv () 0.06% v/v) H2O
under noninerted conditions
<30 min to full conv.
slight increase in 15 was stillwithin
the normal variation
9
10
11
1.4 equiv BINAP to Pd(OAc)2
2 equiv BINAP to Pd(OAc)2
5 equiv BINAP to Pd(OAc)2
88
96
99
11
3.1
0.3
a Normal conditions are 1.0 equiv of 10 free base, 2.0 equiv of N-methylpiperazine, 0.0047 equiv of Pd(OAc)2, 0.02 equiv of (R)-BINAP, 1.4 equiv of NaO-t-Bu, in
toluene at 110-115 °C for 4 h and operating under an inert atmosphere of N2. Thus, the crucial BINAP/Pd(OAc)2-ratio is set at 4.25:1. Listed in the table are deviations from
this experimental protocol.
virtue of a ꢀ-elimination mechanism. This hypothesis was found
not to be entirely correct, as excluding the piperazine moiety
from the reaction mixture still produced 15 (29% after 30 h
reaction time; two other impurities were generated alongside
in a total amount of about 5%, but these were not pursued
further). Our conclusion was that other hydride sources were
involved, foremostly the R-methylbenzylamine moiety, which
by virtue of its secondary acyclic amine structure with R-branch-
ing shows much higher propensity for hydride donation.17
Furthermore, the base (NaO-t-Bu) might exert an influence and
contribute to this side reaction. The problem of uncontrolled
formation of 15 was systematically approached by scanning a
range of potential factors: both the individual constituents
present in the system as well as different aspects of the reaction
conditions. Broad examination covered parameters such as the
H2O content, the relative amounts of 1-methylpiperazine and
tert-butoxide, variations in the charging order of reactants, the
absolute concentration of the reaction mixture, benefits from
operating under inert atmosphere, temperature, time, amount
of ligand and metal and the ratio between them. From the vast
amount of data generated, some parameters were linked more
strongly to the formation of 15. (a) In spite of some inconsisten-
cies there is a clear relationship to the amount of 1-methylpip-
erazine, although differences are small in the range scrutinized
most carefully (1.6-2.2 equiv); (b) butoxide at nearly equimolar
amounts (1.1-1.2 equiv) showed a strong effect (up to 50%
15), while in the normal operating range (1.4-1.6 equiv) the
content was reduced to 0.3-14%, a trend that was further
augmented when running at >2 equiv (1.5-1.9%); (c) the
relationship between ligand (BINAP) and metal precatalyst (Pd-
salt) has the most pronounced and unambiguous effect with a
tendency that a ligand/Pd ratio <2 generated 15 as high as 10%
or above, while a ratio of >2.5 guaranteed that the amount
would be lowered to <1.6%. In this context, a comprehensive
factorial design study was performed to gain further insight into
the relation between various parameters, and to establish a
quantitative basis for better process optimization. The results
clearly revealed that the main factor was the ligand/Pd ratio
which had to be kept >2.5 in order to minimize formation of
15. It was also evident that the base (NaO-t-Bu) played some
role and should, therefore, be used at a higher level (1.4 equiv)
compared with the level used in the “standard” procedure (1.1
equiv). It had been established early on that switching the
precatalyst from Pd(dba)2 to Pd(OAc)2 was beneficial,18 not only
as a consequence of a lower cost per mole but also from the
point of view of reaction performance. Finding the optimal
ligand, however, was still an issue of highest priority and the
initial choice of (R)-BINAP had largely been steered by what
was reported in the literature at the time. This was now
challenged in a search for competitive alternatives. The main
focus of our screening studies and evaluation became (1) the
cost of the ligand hoping that it might be less than BINAP; (2)
the temperature for conducting the reaction, where the aspiration
was to operate at a lower level than before (110-115 °C). In
connection with this we also evaluated the base, by comparing
the results achieved when using various alkoxides and carbon-
ates, as well as the influence of water and solvent type. The
outcome of these investigations are summarized in Table 1, from
which the following conclusions can be drawn: (i) BINAP
constitutes the best ligand (entries 2 and 3); (ii) racemic versus
enantiopure BINAP is of no importance (entries 1 and 11); (iii)
the choice of base is crucial (entries 4 and 5); (iv) the reaction
tolerates up to 0.06% v/v H2O (entry 7); (v) changing of solvent
from toluene to xylene has only a marginal effect on the yield,
but significantly accelerates the reaction rate since it allows the
reaction to be performed at higher temperatures (entry 6); (vi)
shifting from inert to noninert atmosphere during the reaction
has an almost negligible effect (entry 8). It is worth mentioning
that besides the ligands reported in Table 1, representing both
the mono- and bidentate classes, we also evaluated some other
alternatives belonging to the former group. Of those tested, tri-
tert-butylphosphine stood out as the best, albeit with a slow
reaction rate and a ratio between product and H-analogue of
6:1, which was felt to be not good enough to be taken forward.
Furthermore, from a performance aspect, it was found that the
only base competitive with sodium tert-butoxide, was the
homologous sodium tert-pentoxide. The latter has the advantage
(17) (a) Wolfe, J. P.; Wagaw, S.; Buchwald, S. L. J. Am. Chem. Soc. 1996,
118 (30), 7215–7216. (b) Driver, M. S.; Hartwig, J. F. J. Am. Chem.
Soc. 1996, 118 (30), 7217–7218. (c) Wagaw, S.; Rennels, R. A.;
Buchwald, S. L. J. Am. Chem. Soc. 1997, 119 (36), 8451–8458.
(18) Wolfe, J. P.; Buchwald, S. L. J. Org. Chem. 2000, 65 (4), 1144–
1157.
516
•
Vol. 12, No. 3, 2008 / Organic Process Research & Development

charging to the substrate (after liberating the free base) in
the main reaction vessel. This procedure guaranteed the
formation of an active catalytic complex that was found,
fortuitously, to be very stable and, hence, user-friendly, as
demonstrated by its tolerance towards water and normal
atmospheric conditions. In spite of laboratory evidence
showing that the reaction could be performed in a validated
manner at a catalyst loading as low as 0.065 mol % Pd and
yet obtain full conversion after 2 h, it was decided for
practical reasons to stay with the already established level
of 0.47 mol % with the main reason being the increased risk
of facing a charging error in case the catalyst amount was
taken too far down; running on a 100 kg batch size under
such conditions the total amount of Pd(OAc)2 to be added
to a 2000 L reactor would be merely 38 g, and it was felt
that full repeatability of this operation would be difficult to
achieve. After a 4-h reaction time in refluxing toluene, the
process stream was submitted to an extractive workup as
described above before directly taking the product-rich H2O
phase to the next stage. Here a “classical” catalytic deben-
zylation was conducted with the help of Pd on charcoal and
H2, before precipitating the desired product as the benzoate
salt 12 in 88% overall yield from 10 and in stereochemical
purity of 98% ee.2 Another important quality aspect relevant
to patient safety, is the amount of residual heavy metal
remaining in the isolated product. This concern is especially
pronounced when conducting organometallic reactions, such
as the Buchwald-Hartwig,19 in homogeneous phase, and
reaching the low ppm levels requested by the authorities can
often be quite challenging. Fortunately, in our case virtually
all Pd disappears in the toluene stream after the acidic
extractive workup at pH 4-6 (flowchart in Figure 2), leading
to the final intermediate 12 showing a Pd content below the
regulatory acceptance limit for APIs.
Figure 1. Area % of 11 and 15, respectively, in the toluene
phase as a function of pH in the H2O phase used in the
extraction.
of being considerably more soluble in the preferred solvent
toluene, in which the rate of reaction remained roughly constant,
albeit that the H-analogue 15 was formed to a somewhat larger
extent. As there was no cost benefit using pentoxide, this option
was not pursued.
Designing the Final Process. Having optimized the catalyst
loading in the Buchwald-Hartwig reaction to ensure a quantita-
tive conversion of starting material resulting in high productivity,
design of an effective workup and isolation remained. The key
development target was to reduce or eliminate the amount of
the 8-H analogue 15, normally 1.8-6% using the optimized
protocol, as early in the workup as possible to ensure that it
would not be carried along further downstream. The reason was
that if the purification was postponed to the final stage, previous
work had shown separation of the 8-H congener from the final
product AR-A2 was inefficient and would cause a loss of
valuable product. This approach was, therefore, suspended as
not sustainable for reaching the specification limit of <0.5%
byproduct. Instead it was felt that utilizing the difference in
product and impurity pKa by careful adjustment of the pH
should be successful. The impurity (15) has pKa 9.29, and the
product (11) with three basic nitrogens has pKas of 9.48, 8.89,
and 3.05. Evaluating a wide pH range from 4 to 13 showed
good separation with a maximum of 83 area % (GC) of 15 in
the toluene phase at pH 5, whilst 11 was at a level of only 8
area %. Fine-tuning revealed that the optimum pH value was
between 5.1-5.3, where the carry-over of 15 into the H2O phase
would be minimized, at the same time achieving a minimal
loss of 11 in the organic layer. The pH dependence of
compounds 11 and 15expressed as their relative presence
in the toluene phase is shown in Figure 1. Switching from
the previous method where the extraction was carried out
at pH e 2 and instead operating around pH 5, gave a purer
process stream for further work-up and, moreover, an
efficient separation of 15 at minimum loss of 11. Further-
more, this change enabled the process to be carried out in
standard stainless steel reactors, lessening the risk of reactor
corrosion. Summarizing the experimental studies, optimiza-
tions, and pilot runs into a fine-tuned process description,
the C-N bond-forming Buchwald-Hartwig step was de-
signed as described in the flowchart in Figure 2. It had
become clear that the best way to ensure a successful
transformation from 10 free base to 11 was to conduct a
premixing operation of Pd(OAc)2, BINAP, and methylpip-
erazine in toluene while heating the mixture to 45 °C before
Conclusions
In this contribution, which represents excerpts from a major
drug project undertaken in our Process R&D organization from
the late 1990s into 2000, we have focused on one crucial step
in the synthetic sequence: the formation of a C-N bond
between an aromatic moiety and a heterocyclic building block.
During the course of the development, three different routes
for making the target molecule AR-A2 (1) were designed and
evaluated; in each, creation of the C-N bond was a key
component. The Buchwald-Hartwig coupling procedure en-
ables synthesis of the C-N bond in a catalytic manner using
Pd(OAc)2 and (()-BINAP. We have developed, refined, and
optimized the protocol tailored to our specific bromo-aminote-
tralin and N-methylpiperazine. The final process (Scheme 3;
flowchart in Figure 2) operates at a high capacity with only
about 0.5 mol % Pd and produces the desired product (12) in
a telescoped fashion, where the coupling is combined with a
catalytic hydrogenation to effect a debenzylation. Overall
performance has been validated in the pilot plant, where batches
(19) An approach to the Buchwald-Harttwig amination making use of the
catalyst in heterogenous form on a fibre-support has recently been
published. Christensen, H.; Kiil, S.; Dam-Johansen, K.; Nielsen, O.
Org. Process Res. DeV. 2007, 11 (6), 956–965.
Vol. 12, No. 3, 2008 / Organic Process Research & Development
•
517

Figure 2. Fully integrated process of converting the bromo-aromatic intermediate 10 (mBBMT×HCl) to the piperazine-substituted
product 12 in two stages; a homogenous Pd/BINAP-catalyzed Buchwald-Hartwig coupling followed by a heterogeneous catalytic
hydrogenation effecting debenzylation.
of 125 kg material (88% yield over two stages) has been
manufactured at an impeccable quality (98% ee).
a mixture of H2O (400 L) and 33% HCl (aq) (196 L, 2.06
kmol) at 15 °C. The first reactor and the associated tubings
were rinsed with H2O (50 L), and this aqueous portion was
added to the main mixture. The resulting suspension was
stirred for 1.5 h at 15 °C before the solids were isolated by
centrifugation. The three centrifuged portions were rinsed
with 200 L of H2O each, which gave 570 kg of crude solid
material as a wet cake, containing 20% w/w of H2O, and
the isomer distribution was 55/34/8 for isomers 4a/4b/4c,
respectively, according to HPLC. This crude product also
contained about 1% of dibrominated analogue 4d. This
material was purified by crystallization as described below.
The crude product prepared as described above (958 kg) was
dissolved in a mixture of H2O (892 L), i-PrOH (843 L) and
HOAc (153 L) in a 3000 L glass-lined reactor at 60 °C. The
obtained solution was cooled to 0 °C over 6 h, during which,
first, an emulsion was formed, followed by a white suspension.
After stirring at 0 °C for 2.5 h, the solids were isolated by
centrifugation. This gave 384 kg of a wet product consisting
of a mixture of 4a, 4b, and 4c with a regioisomer distribution
of 80/14/4, respectively, according to HPLC and a loss on
drying of 15%. This material was dissolved in a mixture of
H2O (576 L) and i-PrOH (334 L) at 60 °C, after which the
resulting solution was cooled to 0 °C over 4 h. During this time
a white solid precipitated, generating a suspension, which was
agitated at 0 °C for another 2 h. The solid was isolated by
centrifugation in two portions, and each portion was rinsed with
H2O (200 L). Drying at 60 °C and <2 kPa gave 247 kg (32%
Experimental Section
General. NMR spectra were taken on a Bruker 400 Ultra
Shield instrument operating at 400 MHz, and chemical shifts
are reported in parts per million (ppm) with the respective
deuterated solvent peak as an internal standard. HRMS data
were recorded using ESI positive or negative ionization mode
depending on the compound. Reagents and solvents were
obtained from commercial sources and used without further
purification. Optical rotation measurements were performed
using an Optical Activity AA-5 automatic polarimeter, a 200
mm cell, and the sodium D-line as wavelength at room
temperature.
2-Bromo-5-methylphenylacetic Acid (4a).3 K2CO3
(257 kg, 1.86 kmol) and H2O (999 L) were mixed under
nitrogen in a 2500 L glass-lined reactor. 3-Methylphenylacetic
acid (3) (300 kg, 2.00 kmol) was added, and the resulting
mixture was agitated with release of carbon dioxide, giving a
solution of the potassium salt of 3, which was cooled to 10 °C.
Br2 (l) (325.8 kg, 2.05 kmol) was charged over 3 h via a dip
pipe under vigorous agitation, maintaining the temperature
<15 °C. During the bromine addition more carbon dioxide was
released. After the addition was complete, the resulting solution
was stirred for 1 h at 10 °C. Excess bromine was destroyed by
adding sodium sulfite (3.0 kg), and the resulting mixture was
allowed to reach 15 °C. The solution obtained was then added
over 1.5 h to another 2500 L glass-lined reactor containing
518
•
Vol. 12, No. 3, 2008 / Organic Process Research & Development

assay-corrected yield) of 4a with 95% purity and an isomer
composition of 95/2.5/1.5 for 4a/4b/4c, respectively, according
to HPLC.
Melting point 119-122 °C (uncorrected) [lit.3 122-123 °C];
1H NMR (400 MHz, DMSO-d6) δ 12.77 (s, 1 H), 7.78 (d, J )
8 Hz, 1 H), 7.52 (app d, J ) 2 Hz, 1 H), 7.34 (app dd, J1 ) 2
Hz, J2 ) 8 Hz, 1 H), 3.98 (s, 2 H), 2.57 (s, 3 H); 13C NMR
(100 MHz, DMSO-d6) δ 171.8, 137.4, 135.0, 133.1, 132.2,
129.8, 121.5, 41.3, 20.6; HRMS C9H8O2Br (M - H)+ calcd
226.9708; found 226.9719.
been removed. A second portion of methylcyclohexane
(160 L) was added, and the vacuum distillation was repeated
until 190 kg of volatiles had been removed. A third portion of
methylcyclohexane (160 L) was added, and the vacuum
distillation was repeated until 119 kg of volatiles had been
removed. Subsequently, the residual amount of 1,2-dichloro-
ethane was only 0.4% as determined by GC. Di-isopropyl ether
(70.2 kg) was added to the residue at 30 °C followed by
methylcyclohexane (325 L) allowing the temperature to reach
25 °C. The resulting suspension was agitated at this temperature
for 0.5 h before cooling over 3 h to -5 °C. Solid material was
isolated by centrifugation giving 139.3 kg of 5 (52.2% yield,
corrected for assay and purity, from compound 4a) having a
loss on drying value of 7.6% w/w and a GC purity of, in this
case, 86.4 area %. The purity of compound 5 has varied between
83-93% comparing eight batches on similar scale. This material
was taken forward to the reductive amination step without
further purification.
(R)-8-Bromo-5-methyl-1,2,3,4-tetrahydronaphthalen-
2-yl)-(S)-1-(phenylethyl)amine Hydrochloride (10). Com-
pound 5 (170 kg, 90% purity by GC, 711 mol) was
charged to a 2500 L glass-lined reactor under nitrogen
atmosphere. Toluene (731 L) was added, the resulting
solution was mixed with p-toluenesulphonic acid mono-
hydrate (493 g, 2.59 mol), and the resulting solution was
agitated at 30 °C. (S)-Phenylethylamine (8) (86.2 kg,
711 mol) was added, and the mixture obtained was
concentrated by vacuum distillation at 35-50 °C and 4-30
kPa until approximately 459 kg of volatiles had been
removed. At this point, a >98% conversion to the imine/
enamine had been obtained. The residue was mixed with
MeOH (680 L) and i-PrOH (901 L), and the resulting
solution was cooled to -3 °C. NaBH4 (50.0 kg, 1.32 kmol)
was added in 5-kg portions in such a way that the inner
temperature did not exceed 10 °C. After complete addition,
the mixture was stirred for 7 h at 10 °C. Since the conversion
was not >96% at this point, two additional portions of
NaBH4 (5.0 kg, 132 mol each) were added, and agitation
was continued at 10 °C for another 3 h. The mixture was
then heated to 35 °C for 30 min to destroy residual NaBH4
(this is important to avoid foaming in the distillation
described below). The resulting mixture was concentrated
by vacuum distillation at 35-45 °C and 10-30 kPa until
approximately 917 kg of volatiles had been removed.
Toluene (690 L) and H2O (646 L) were added to the residue,
and this mixture was agitated for 1 h at 47 °C. The aqueous
layer was removed, and the organic phase was concentrated
by vacuum distillation at 35-55 °C and 3-25 kPa until
approximately 818 kg of volatiles had been removed. EtOAc
(629 L) was added to the residue, and the resulting mixture
was transferred to a second 2500 L glass-lined reactor via a
GAF filter. The first reactor and the filter were rinsed with
EtOAc (100 L). The solution obtained was diluted with
absolute EtOH (313 L) and heated to 58 °C. HCl (g) (39.1
kg, 1.07 kmol) was added via a dip pipe at 60 °C over 3 h.
After complete addition, the resulting suspension was agitated
at 60 °C for 2 h and then cooled to 15 °C over 1.5 h. After
the solution stirred at 15 °C for 1 h and the pH of the mother
8-Bromo-5-methyl-3,4-dihydro-1H-naphthalen-2-one (5).4
Compound 4a (204.0 kg, 891 mol) was mixed with 1,2-
dichloroethane (727 L) and N,N-dimethylformamide (0.14 kg,
1.9 mol) in a 2500 L glass-lined reactor. The obtained mixture
was heated to reflux (82 °C jacket temperature) before addition
of thionyl chloride (128.2 kg, 1078 mol) over 2.5 h. After
completed addition, the mixture was refluxed for another 2 h
and cooled to a temperature of 58 °C; at this temperature a
sample showed complete conversion to the acid chloride. This
mixture was cooled to 26 °C and was then concentrated by
vacuum distillation at 26-33 °C and 9.5-10.2 kPa until
approximately 612 kg of volatiles had been removed. 1,2-
Dichloroethane (500 L) was added to the residue, and the
resulting solution was cooled to -2 °C before addition of AlCl3
(133.0 kg, 997.5 mol) over 0.5 h under efficient stirring, while
keeping the temperature between -4 to +2 °C. A further
portion of 1,2-dichloroethane (55 L) was used to rinse the
reactor walls from solid AlCl3, and this portion was pooled with
the main suspension. Ethene (g) (60.6 kg, 2.16 kmol) was added
through a dip pipe over 6 h 20 min, which gave a steady inner
temperature of between -1 to +1 °C for this strongly
exothermic reaction. The mixture was stirred at 0 °C for an
additional 30 min, before analysis showed complete conversion.
Transfer of the mixture to a second 2500 L glass-lined reactor
containing H2O (710 L) was performed over 4 h at temperatures
between 10-24 °C. The first vessel and the connected tubings
were rinsed with 1,2-dichloroethane (50 L) and this portion was
added to the two-phase system thus obtained, which was heated
to 25 °C and stirred at this temperature for 30 min before
agitation was stopped and the phases were allowed to separate.
The organic layer still contained some aqueous residues, and it
was therefore heated to 28 °C and left for another 45 min before
conducting the second phase separation. KHCO3 (85.0 kg) and
H2O (760 L) were added to a second 2500 L glass-lined reactor,
and the resulting solution was heated to 25 °C before mixing it
with the organic phase above. This two-phase mixture was
stirred, maintaining the temperature at this level for 30 min,
before being allowed to settle for 1 h. The organic layer was
then washed with H2O (750 L) at 28 °C before a concentration
was conducted by vacuum distillation at 20-25 °C and around
10 kPa until approximately 231 kg of volatiles had been
removed. This gave 1008 kg of a 14% w/w solution of crude
tetralone 5 to be purified by crystallization as described below.
The solution obtained above (1008 kg) was concentrated by
vacuum distillation until approximately 662 kg of volatiles had
been removed. Methylcyclohexane (240 L) was added to the
residue, and the resulting mixture was further concentrated by
vacuum distillation until approximately 212 kg of volatiles had
Vol. 12, No. 3, 2008 / Organic Process Research & Development
•
519

liquor was checked to make sure it was <2, the white solid
was isolated by centrifugation. Each of two loadings on the
centrifuge was washed with EtOAc (120 L per portion)
giving a total of 256 kg of wet crude 10 (HCl salt) having
94% diastereomeric excess (de), 30% loss on drying, and a
chromatographic purity of 99% by GC. This material was
subjected to the recystallization procedure below.
system, which was then agitated at room temperature for 15
min. The pH of the aqueous layer at this point is about 5, and
the two phases were allowed to settle for 1 h. The aqueous
layer was then transferred to a hydrogenator. Pd/C (5% Pd, 57%
H2O w/w, 77 g) was added, and the reactor was purged with
nitrogen and then evacuated. H2 was added until reaching a
relative pressure of 350 kPa, and the temperature was raised to
95 °C. The reaction was complete after 5 h, and the mixture
was then cooled to 30 °C before removing the catalyst by
filtration. Reactor and filter were rinsed with H2O (0.48 L), and
the rinsing water was then pooled with the reaction mixture.
Toluene (3.5 kg) was added followed by the addition of 50%
NaOH (aq) (760 g), and the obtained two-phase system was
agitated at ambient temperature. The aqueous layer was
separated and washed with toluene (2.7 kg), and this second
toluene phase was pooled with the first one. A solution of
benzoic acid (385 g, 3.15 mol) in a mixture of i-PrOH (130 g)
and toluene (2.7 kg) was prepared and heated to 45 °C. This
solution was added to the solution of the crude, free amine 12
at a temperature of 80 °C during 2 h under agitation.
Subsequently, the jacket of the vessel was cooled to 10 °C over
2 h, and the slurry obtained was agitated for another 2 h, keeping
the temperature constant at this level before filtration. After
filtration the solid was rinsed with toluene (3.1 kg) and dried
to give 760 g (76% yield) of pure compound 12 (benzoate salt).
Crude 10 obtained as described above (256 kg) was charged
to a 2500 L glass-lined reactor under nitrogen atmosphere.
Absolute EtOH (376 L) and EtOAc (877 L) were added, and
the resulting suspension was stirred at 70 °C for 4 h. Cooling
to 5 °C over a period of 2 h was followed by agitation at this
temperature for another 1.5 h. Centrifugation and washing with
EtOAc (100 L), followed by drying of the wet solid at 50 °C
and 2 kPa gave 151 kg (55.8% yield) of 10 having a purity
and assay >99% and a stereochemical purity of 97% de.
Melting point 286 °C dec; 1H NMR (400 MHz, MeOH-d4)
δ 7.59-7.65 (m, 2 H), 7.42-7.56 (m, 3 H), 7.26 (d, J ) 8 Hz,
1 H), 6.90 (d, J ) 8 Hz, 1 H), 4.74 (q, J ) 7 Hz, 1 H),
3.15-3.28 (m, 2 H), 2.85-2.97 (m, 1 H), 2.68-2.80 (m, 1 H),
2.43-2.62 (m, 2 H), 2.15 (s, 3 H), 1.80-1.92 (m, 1 H), 1.77
(d, J ) 7 Hz, 3 H); 13C NMR (100 MHz, MeOH-d4) δ 137.7,
137.3, 137.2, 132.4, 131.1, 130.9, 130.8, 130.6, 128.7, 123.3,
57.1, 53.8, 35.3, 26.8, 25.9, 20.4, 19.4; HRMS C19H23NBr (M
+ H)+ calcd 344.1014; found 344.1018; [R]22D ) +18 (c 3.0,
CH3OH).
1
Melting point 214-217 °C dec; H NMR (400 MHz,
(R)-5-Methyl-8-(N4-methyl-piperazin-1-yl)-1,2,3,4-tetrahy-
dronaphthalen-2-ylamine Benzoate (12). Compound 10
(1.00 kg, 2.63 mol) was mixed with H2O (1.60 kg) and 50%
NaOH (aq) (0.60 kg, 7.5 mol) in a reactor at 25 °C. Toluene
(7.50 kg) was added, and the resulting mixture was heated to
60 °C under stirring until all material had dissolved. The
aqueous layer was removed, and the organic phase was
concentrated by vacuum distillation using a jacket temperature
<60 °C to a total volume of 2.2 L to remove residual water
from the system. (R)-BINAP (32.7 g, 0.0525 mol) was charged
to a second reactor, followed by Pd(OAc)2 (2.75 g, 0.0123 mol)
and toluene (2.3 kg), and this mixture was heated to 40 °C under
stirring, which gave a solution. 1-Methylpiperazine (0.53 kg,
5.3 mol) was added to the catalyst mixture which caused a color
change from orange to deep wine-red. The resulting mixture
was transferred to the previously prepared toluene solution of
compound 10 under stirring. In a second reactor, sodium tert-
butoxide (0.35 kg, 3.7 mol) was mixed with toluene (1.3 kg),
and the resulting suspension was vigorously agitated before
transfer to the solution containing 10, 1-methylpiperazine, and
catalyst. The mixture obtained was then heated to 100 °C and
kept there for 4 h, before GC analysis showed complete
conversion. At this point, the reaction mixture was cooled to
20 °C, and H2O (2.5 kg) was added. Agitation for 15 min at
room (ambient) temperature was followed by phase separation
and addition of a further portion of H2O (4.0 kg) to the organic
layer. HOAc (0.48 kg, 7.9 mol) was added to the two-phase
MeOH-d4) δ 7.92-8.00 (m, 2 H), 7.31-7.45 (m, 3 H), 7.02
(d, J ) 8 Hz, 1 H), 6.91 (d, J ) 8 Hz, 1 H), 3.36-3.49 (m, 2
H), 2.52-3.06 (m, 11 H), 2.39 (s, 3 H), 2.22-2.31 (m, 1 H),
2.18 (s, 3 H), 1.79-1.93 (m, 1 H); 13C NMR (100 MHz,
MeOH-d4) δ 175.3, 150.2, 138.7, 135.3, 133.3, 131.5, 130.3,
129.5, 129.2, 128.9, 118.5, 56.4, 52.4, 45.8, 30.9, 28.4, 26.5,
19.4; HRMS C16H26N3 (M + H)+ calcd 260.2127; found
260.2119; [R]22 ) +13 (c 3.0, CH3OH).
D
((R)-5-Methyl-1,2,3,4-tetrahydro-naphthalen-2-yl)-(S)-1-
(phenylethyl)amine Hydrochloride (15). 20 Compound 15 was
isolated and purified as the hydrochloride salt from the reaction
mixture of a worst case run, in which it was formed in 11 area
% as mentioned in the main text. Thus, the reaction mixture
(100 mL, toluene solution) was extracted with H2O, and the
organic layer was filtered to remove solid impurities and then
dried azeotropically. The dried solution containing the desired
product 11, the H-analoque 15, and unreacted starting material
10 was allowed to react with (R)-BINAP (199 mg, 0.321 mmol),
Pd(OAc)2 (57 mg, 0.252 mmol), 1-methylpiperazine (5.54 g,
55.4 mmol), and tert-butoxide (3.59 g, 41 mmol) at 80 °C for
2 h. On completion, the reaction mixture was treated with H2O
(25 mL), and the organic layer was extracted with a mixture of
H2O (100 mL) and HOAc (100%, 7.3 g). A pH measurement
of the aqueous layer showed 4-5. The organic layer was
evaporated to dryness and the residue was dissolved in EtOAc
(40 mL). To the solution was added HCl in i-PrOH (6 M,
3 mL) to precipitate the salt. The white crystalline salt was
collected by filtration and recrystallized from a mixture of
i-PrOH (40 mL) and H2O (7 mL) to obtain 3 g of 15 as the
HCl salt and with a purity >98%.
(20) The experimental procedure given is not directed at providing the most
efficient way of synthesizing compound 15. Instead, the protocol
presents an attempt to improve the outcome of an authentic pilot trial,
in which 15 had been generated to an extent of about 11%, by taking
a sample and resubmitting this to Buchwald-Hartwig conditions in
order to increase the yield of desired product 11. As a spin-off exercise
from this, 15 was able to be isolated and characterized.
Melting point 300 °C dec, 1H NMR (400 MHz, CDCl3) δ
10.36 (broad, 1H), 10.08 (broad, 1H), 7.71-7.74 (m, 2H), 7.43
520
•
Vol. 12, No. 3, 2008 / Organic Process Research & Development

(t, J ) 7 Hz, 2H), 7.34-7.38 (m, 1H), 6.94-6.95 (m, 2H),
6.78-6.81 (m, 1H), 4.49-4.54 (m, 1H), 3.35-3.51 (m, 2H),
3.03 (broad, 1H), 2.85-2.91 (m, 1H), 2.40-2.50 (m, 2H),
forming the basis for this article, we need to specifically
highlight a few names. Thus, we express our deep gratitude to
Drs. Vern Delisser and Magnus Sjo¨gren in Process Chemistry,
Andreas Lindgren in Process Engineering and Go¨ran Lundin
in Analytical Chemistry (HRMS measurements), all residing
at the So¨derta¨lje site. Furthermore, the great support provided
by the team in Development Manufacture at Macclesfield led
by Frank Harkness producing material on pilot scale is
thankfully acknowledged.
2.17-2.30 (m, 1H), 2.14 (s, 3H), 2.03 (d, J ) 7 Hz, 3H); 13
C
NMR (100 MHz, CDCl3) δ 136.7, 136.5, 133.2, 133.1, 129.9,
129.6, 128.3, 128.1, 127.3, 126.7, 56.6, 52.8, 34.29, 25.8, 25.1,
21.5, 19.8; HRMS C19H23N (M + H)+ as the free base calcd
266.1009; found 266.1902; [R]22 ) -11 (c 1.0, CH3OH).
D
Acknowledgment
The achievements in the entire AR-A2 project within Process
R&D, of which the current case study presents a subsection,
are the fruits of many co-workers’ dedication, enthusiasm, and
efforts. However, in the context of the particular piece of work
Received for review January 23, 2008.
OP8000146
Vol. 12, No. 3, 2008 / Organic Process Research & Development
•
521