Development of an efficient process for the preparation of Sch 39166: Aziridinium chemistry on scale

Article abstract of DOI:10.1021/op0402026

A large-scale synthesis of a tricyclic D1/D5 dopamine antagonist based on regio- and stereoselective ring opening of an aziridinium ion with a Grignard reagent was optimized and scaled up.

Full text of DOI:10.1021/op0402026

Organic Process Research & Development 2004, 8, 754−768
Development of an Efficient Process for the Preparation of Sch 39166:
Aziridinium Chemistry on Scale
Dinesh Gala,* Vilas H. Dahanukar,* Jeffery M. Eckert, Brian S. Lucas, Doris P. Schumacher, and Ilia A. Zavialov
Schering-Plough Research Institute, Department of Synthetic Chemistry, 1011 Morris AVenue,
Union, New Jersey 07083, U.S.A.
Patrik Buholzer, Peter Kubisch, Ingrid Mergelsberg, and Dominik Scherer
Chemical DeVelopment Department, Werthenstein Chemie AG, 6105 Schachen, Switzerland
Abstract:
hundred kilograms of Sch 39166. The aziridinium route was
demonstrated in the laboratory to provide a quick access to
1. However, the following challenges were identified with
this approach which required additional work to make it a
robust, plant-suitable process capable of making large
quantities of Sch 39166 needed to support clinical trials: (1)
An efficient and cheap synthesis of key starting material 2
was necessary. (2) N-Alkylation of 2 to 3 was slow and
required very long reaction times. (3) The yields obtained
in the aziridinium formation and opening step were highly
variable (45-60%), and an excess (2.1 equiv) of the
expensive Grignard reagent 4 was required to obtain these
moderate yields. The purity of isolated product varied from
85 to 95%, requiring a chromatographic purification of 5.17
Overall, the mechanistic aspects of the chemistry involved
in this transformation were poorly understood. (4) The
cyclization/reduction protocol for the conversion of 5 to 6
A large-scale synthesis of a tricyclic D1/D5 dopamine antagonist
based on regio- and stereoselective ring opening of an aziri-
dinium ion with a Grignard reagent was optimized and scaled
up.
Introduction
Sch 39166 (1) is a potent dopamine D-1 antagonist that
was initially developed to treat schizophrenia.1 Subsequent
in vivo animal autoradiography studies2 demonstrated that
1 was a highly selective D1/D5 antagonist. Since the
neurotransmitter dopamine is involved in the reward mech-
anism that induces craving, 1 was shown in animal models
to reduce cravings for addictive substances (cocaine, nicotine,
alcohol, etc.) and food.3 New phase II trials showed that 1
reduced the euphoric effects of cocaine,4 but was not effective
in completely treating the addiction.5 Additional clinical
studies were planned to evaluate 1 in reducing cravings for
food (to treat obesity) and nicotine. With renewed interest
in conducting phase II and phase III clinical trials of 1, a
commercial synthesis was urgently required to support these
activities and future marketing needs.
(8) Kalkote, U. R.; Choudhary, A. R.; Natu, A. A.; Ayyangar, N. R. Synth.
Commun. 1991, 21, 1129-1135.
(9) Purchase, C. F.; Goel, O. P. J. Org. Chem. 1991, 56, 457-459.
(10) (a) Leonard, N. J.; Paukstelis, J. V. J. Org. Chem. 1965, 30, 821. (b) Olah,
G. A.; Szilagyi, P. J. J. Am. Chem. Soc. 1969, 91, 2949-2955. (c) Golding,
B. T.; Kebbell, M. J.; Lockhart, I. M. J. Chem. Soc., Perkin Trans. 2 1987,
705-713.
(11) (a) Andrews, D. R.; Dahanukar, V. H.; Eckert, J. M.; Gala, D.; Lucas, B.
S.; Schumacher, D. P.; Zavialov I. A. Tetrahedron Lett. 2002, 43, 6121-
6125. (b) Dahanukar, V. H.; Zavialov, I. A. Curr. Opin. Drug DiscoVery
DeV. 2002, 5, 918-927.
Several different syntheses of 1 were explored,6,7 and the
approach based on aziridinium chemistry7 proved to be the
most attractive. This report deals with the chemistry aspects
of process development and optimization of the aziridinium
route (Scheme 1) that enabled us to manufacture several
(12) Other sulfonyl choride reagents investigated as activating reagents (ben-
zenesulfonyl chloride, 4-chlorobenzenesulfonyl chloride, 4-nitrobenzene-
sulfonyl chloride, and MsCl) showed no advantage over p-TsCl. Other
reagents (methanesulfonic anhydride, p-toluenesulfonyl imidazole, carbon-
yldimidazole, thionyl chloride, trifluoroacetic anhydride, and oxalyl chloride)
also proved to be inferior to p-TsCl.
(1) Karlsson, P.; Smith, L.; Farde, L.; Haernryd, C.; Sedvall, G.; Wiesel, F.-A.
Psychopharmacology 1995, 121, 309-316.
(2) Duffy, R. A.; Hunt, M. A.; Wamsley, J. K.; McQuade, R. D. J. Chem.
Neuroanat. 2000, 19, 41-46.
(3) Coffin, V.; Glue, P. W. Methods for reducing craving in mammals. WO
9921540-A2, 1999.
(4) Romach, M. K.; Glue, P.; Kampman, K.; Kaplan, H. L.; Somer, G. R.;
Poole, S.; Clarke, L.; Coffin, V.; Cornish, J. O’Brien, C. P.; Sellers, E. M.
Arch. Gen. Psychiatry 1999, 56, 1101-1106.
(13) 31P NMR (THF-d8, 161.98 MHz) spectra were recorded using H3PO4 as an
external reference standard.
(14) Erdik, E. Tetrahedron 1984, 40, 641-657.
(15) Several copper salts other than CuCl and CuCN were screened in catalytic
amounts (10 mol %): CuBr, CuI, CuSCN, CuSPh, Cu(OtBu), CuCl2,
Cu(OAc)2, Cu(BF4)2, Cu(OTf)2, copper (II) naphthenate, copper (II)
acetylacetonate, copper (II) tifluoroacetylacetonate, copper (II) benzoyl-
acetonate, lithium 2-thienylcyanocuprate, and Cu(I) phenylacetylide. Other
metal salts (Cr, Fe, Mn, and Ni) failed to show any improvement in the
reaction of 16 with 4.
(16) In the laboratory, after forming 15b at -20 °C, it was mixed with 4 at
various temperatures and held for several hours. Aliquots were taken and
analyzed periodically by HPLC to quantitate the amount of product 5. This
work showed that the reaction of 4 with 16 was sluggish below 0 °C and
proceeded faster at 25-35 °C.
(5) Nann-Vernotica, E.; Donny, E. C.; Bigelow, G. E.; Walsh, S. L. Psycho-
pharmacology 2001, 155, 338-347.
(6) (a) Wu, G.; Wong, Y.; Steinman, M.; Tormos, W.; Schumacher, D. P.;
Love, G. M.; Shutts, B. P.; Org. Process Res. DeV. 1997, 1, 359-364. (b)
Draper, R. W.; Hou, D.; Iyer, R.; Lee, G. M.; Liang, J. T.; Mas, J. L.;
Vater, E. J. V. Org. Process Res. DeV. 1998, 2, 186-193. (c) Hou, D.;
Schumacher, D. Curr. Opin. Drug DiscoVery DeV. 2001, 4, 792-799.
(7) Draper, R. W.; Hou, D.; Iyer, R.; Lee, G. M.; Liang, J. T.; Mas, J. L.;
Tormos, W.; Vater, E. J.; Gunter, F.; Mergelsberg, I.; Scherer, D. Org.
Process Res. DeV. 1998, 2, 175-185.
(17) This would be a large waste burden for a commercial process; hence, it
was to be avoided at any cost.
754
•
Vol. 8, No. 5, 2004 / Organic Process Research & Development
10.1021/op0402026 CCC: $27.50 © 2004 American Chemical Society
Published on Web 08/03/2004

Scheme 1. Laboratory synthesis of Sch 39166
Scheme 2. Commercial synthesis of 12
used methylene chloride as a solvent and then NaHCO3
neutralization (the literature7 does not have this procedure;
this is based on the optimized supply route chemistry
described in ref 6) of methanesulfonic acid before reduction
with tert-butylamine borane (TBAB). The cyclization time
was about 2-3 days, and neutralization of acid was difficult
on scale due to excessive foaming with CO2. Isolation and
purification of free base 6 was known only via crystallization
as a 1:1 mixture of (+)-di-p-toluoyl-D-tartaric acid (DTTA)
salt. The use of this expensive chiral acid in salt formation
of enantiopure 6 was undesirable. (5) Finally, the impurity
profile of Sch 39166 prepared via this route must meet or
exceed the specifications set by supply route as per ICH
guidelines. The last criterion was critical for this aziridinium-
route synthesis to be a viable option for providing drug
substance required for clinical studies. With these difficulties
and potential issues in mind, we embarked on the process
development of this route.
Results and Discussion
1. Synthesis of 12 (and 2). Naphthalene, tetralin, and
R-tetralone were identified as three readily available inex-
pensive potential starting materials to obtain (()-2 (Scheme
2). Due to scale-up issues related to handling liquid ammonia
and sodium metal in our facility, the Birch reduction of
naphthalene was not pursued. In the tetralin route, dibromi-
nation followed by substitution resulted in formation of
bromohydrin 11, which was converted to (()-2. Although
this synthesis was demonstrated to work in the laboratory,7
removal of other brominated impurities as well as constraints
Vol. 8, No. 5, 2004 / Organic Process Research & Development
•
755

Scheme 3. Asymmetric epoxidation route
Table 1. N-Alkylation of the amino alcohol 2 with BADMA
entry
conditions
% yield
1
2
3
4
5
6
7
8
9
K2CO3 (2.0 equiv), acetal (1.5 equiv), MeCN, reflux, 48 h
55
NaI (0.1 equiv), K2CO3 (2.0 equiv), BADMA (1.5), acetone, reflux, 18 h
KI (0.1 equiv), K2CO3 (1.5), BADMA (1.5 equiv), MeCN, reflux, 48 h
50% NaOH, nBu4NHSO4 (0.1 equiv), BADMA (1.1 equiv), PhCH3, rt, 48 h
NaOH (2.0 equiv), K2CO3 (1.0 equiv), BADMA(1.1 equiv), PhCH3, reflux 6 h
K2CO3 (1.5 equiv), BADMA (1.2 equiv), DMF, reflux, 14 h
K2CO3 (1.5 equiv), BADMA (1.2 equiv), NMP, 100 °C, 14 h
K2CO3 (1.5 equiv), BADMA (1.2 equiv), diglyme, 130 °C 18 h
Na2CO3 (0.8 equiv), BADM (1.2 equiv), diglyme, 130 °C 18 h
NRa
50
NRa
Decompb
80
52
90c
80d
a NR ) no reaction. b Decomp ) decomposition. c Yield corrected for HPLC purity vs standard. d Reaction showed other side products (5-10 area %, HPLC).
associated with handling bromine were significant; hence,
this process was not scaled up. The first three steps in the
R-tetralone route, reduction to 7 followed by dehydration to
8 and bromohydrin formation, could be telescoped in 85%
overall yield. Since purification of 8 was difficult, procedures
were developed to partially purify 8 and react it with
dibromohydantoin in wet acetone to form the desired
bromohydrin 9 which can be isolated. Treatment of 9 with
an excess of 40% aqueous methylamine solution resulted in
formation of (()-2. This reaction proceeds via an epoxide
intermediate (()-13. Compared to direct epoxidation of 8,
the bromohydrin sequence gave higher yields of (()-2. Since
(()-2 has high water solubility, saturated NaCl solution use
before extraction with MTBE was incorporated in the workup
(see the Experimental Section). L-Tartaric acid proved to be
the best among other chiral acids (camphorsulfonic acid,
malic acid, mandelic acid, glutamic acid, and aspartic acid)
screened for the resolution of (()-2. The addition of small
amounts (0.05 equiv) of achiral acids such as HCl, HOAc,
and H2SO4 in tartaric acid crystallization failed to improve
the yield or ee of 12. In the early stages of this work the
addition of L-(+)-tartaric acid (0.25 equiv) to a solution of
(()-2 in MeOH was used to obtain 12 in 78-85% de, which
was upgraded to 98-99% de by two additional crystalliza-
tions from MeOH. Additional development led to an even
more efficient purification by slurrying the crude product in
iPrOH-H2O (95:5) to afford 12 of the desired chemical and
chiral (98-99% de) purity. Thus, the R-tetralone-based route
was scaleable, and several hundred kilograms of 12 was
prepared in an overall yield of 32%.
major problems that could not be overcome even after
extensive screening of reaction conditions.
2. N-Alkylation of 2 (Derived from 12). The original
reaction conditions for the conversion of 12 to 3 involved
the removal of tartaric acid from 12 to form 2 followed by
N-alkylation with bromoacetaldehyde dimethyl acetal (BAD-
MA) in acetonitrile as a solvent [MeCN/K2CO3 (2 equiv)/
BADMA (1.5 equiv)]. 7 Under these conditions, the reaction
required 6 days for completion and afforded 78% isolated
yield after workup. Thus, further development of this
chemistry was required to improve the step yield as well as
shorten the reaction time. Some of this work is summarized
in Table 1.
Conversion of 12 to its free base form 2 was a necessary
operation as direct N-alkylation of 12 with BADMA in the
presence of a base afforded low yields under a variety of
conditions. Our initial conditions for preparing 2 were
addition of 30% NH4OH to a suspension of 12 in MTBE.
Due to significant loss of 2 to the aqueous layers under these
conditions, the aqueous layer had to be saturated with NaCl
for an efficient extraction of 2. To avoid use of NaCl, MTBE,
and NH4OH an improved procedure was developed, wherein
a limited amount of water and 25% NaOH were added to a
suspension of 12 in toluene. Sodium tartrate formed during
neutralization, saturated the water layer, and after extraction
with toluene, less than 2% of 2 was lost in the aqueous phase.
Atmospheric distillation of the batch provided low levels of
residual water which was necessary (see below) before the
start of N-alkylation step.
Lowering the reaction time of the conditions described
above to 2 days resulted in a 55% yield (Table 1, entry 1),
with unreacted 2 washed out in the aqueous layer during
the original workup. An alternate reaction condition that was
also included in the earlier report7 that involved the use of
KF adsorbed over alumina (3.3 equiv) provided a higher yield
(94%) with a relatively short reaction time (2 days). This
option was unattractive because of the unavailability of KF
An alternate more direct approach to 3 involved the
opening of chiral epoxide 13 with N-methylaminoacetalde-
hyde dimethylacetal (Scheme 3). This approach was inves-
tigated as well. Although the feasibility of this approach had
been established earlier, low and irreproducible enantiose-
lectivity (55-88%), poor yields (20-40%) of chiral epoxi-
dation and over-oxidation to naphthalene in our hands were
756
•
Vol. 8, No. 5, 2004 / Organic Process Research & Development

adsorbed over alumina on large scale and the waste-handling
issues.
Addition of KI to induce in situ generation of the more
reactive iodo acetal from the BADMA failed to improve the
conversion (entries 2,3). Phase transfer conditions reported
for the alkylation of amides and anilines8 were also ineffec-
tive (entries 4 and 5). Screening experiments revealed that
anhydrous K2CO3 was the best base compared to other
inorganic bases such as Na2CO3, NaHCO3, and KHCO3. Use
of organic bases led to complex reaction mixtures and the
bases proved difficult to remove from the product (compound
3 is a viscous liquid). It was thought that the cause of slow
reaction was the reaction temperature, which was limited by
the boiling point of the solvent. Hence, alternate polar
solvents and bases were screened. Although, polar aprotic
solvents such as DMF and NMP (entries 6,7) appeared
promising in the screening experiments, they were avoided
as compound 3 could not be isolated free of these solvents
and these solvents are incompatible with organometallic
reagents used in the subsequent aziridinium chemistry step.
Diglyme provided an excellent conversion (97-99%) in 18-
20 h at 130-135 °C (entries 8,9). Use of these reaction
conditions for N-alkylation of secondary amines has been
reported.9 After further optimization the amount of BADMA
required was lowered to 1.2 equiv.
Alkylation of 2 in diglyme is a heterogeneous reaction.
Laboratory studies on this reaction indicated that the type
of K2CO3 used, the amount of water, the agitation rate, and
the rate of heating of the reaction mixture were important
factors influencing the reaction outcome. In general, small
particles of anhydrous grade K2CO3 were more preferable
than large, spherical hydrated K2CO3. The former allowed
for rapid trapping of HBr and did not add water burden to
the reaction mixture. Laboratory studies also showed that
presence of 5% (v/v compared to diglyme) water led to a
slower and inefficient reaction (50-60% conversion vs 98%
conversion without the added water). A slow, programmed
heating of the reaction mixture over a period of 3 h from
room temperature to 128 °C was recommended so that the
rate of HBr generation would not exceed the rate of its
neutralization by K2CO3 under these heterogeneous condi-
tions. Control experiments showed that rapid heating and/or
inefficient trapping of the HBr formed during the reaction
led to the cyclic acetal byproducts.18 Other expected side
products such as tetra-N-alkylated 2, O-alkylation product,
and N-demethylation impurities were not detected under
these stressed conditions. Finally, a rapid agitation of the
reaction mixture allowed for improved suspension of the
K2CO3, resulting in efficient removal of HBr.
Figure 1. Progress of reaction of 2 with BADMA in the
presence of milled K2CO3.
of reaction required about 10 h for completion (see Figure
1). On the basis of reaction calorimetery studies, the addition
of BADMA to the diglyme solution of 2 was shown to be
mildly exothermic (adiabatic temperature rise of 2.9 °C, and
heat of reaction -1.1cal/g) and in the subsequent heat up
an adiabatic temperature rise of 75 °C was detected. The
water content of 2 was controlled (0.1% by Karl-Fisher),
and water formed during the reaction via neutralization of
HBr with potassium carbonate was absorbed by anhydrous
potassium carbonate. It is worth noting that under these
conditions, distillation of a water-diglyme azeotrope (lit-
erature bp 100 °C at 760 mm, 78% water) was not seen.
The agitation rate was equipment dependent and a minimum
agitation was determined prior to the use of equipment as
this reaction was scaled up in three plants with about 10-
fold variation in the batch size.
Since compound 3 is a viscous liquid, we decided to use
it in solution for the subsequent steps. Purification via
vacuum distillation was avoidable, and we chose that path.
In the workup step, inorganic solids were filtered and washed
with toluene. Then 3 was extracted with dilute sulfuric acid.
The aqueous acidic layer was saturated with sodium chloride,
neutralized with NH4OH to pH 9-10 (use of NaOH in this
step resulted in emulsion and pH control proved difficult)
and extracted with toluene. Because of complete water
miscibility of diglyme, saturation of the aqueous layers with
NaCl was necessary to minimize product loss to the aqueous
layer. Removal of toluene from the organic layer via
distillation provided 3 as a 45-50% w/w solution in diglyme,
suitable for the next reactions and which could be stored
under nitrogen atmosphere for about 12 months without any
decomposition or color change. This process was scaled up
to 160 kg to produce product of consistent quality in about
90-95% molar yield.
To avoid lengthy acid-base workup an alternate distil-
lative procedure was also developed, wherein after reaction
completion the solids were filtered and washed with toluene
as above. Unreacted BADMA was removed by vacuum
distillation (20 mm, 120 °C). The reaction mixture was
cooled to room temperature and filtered to remove any
precipitated solids. Although the yield was comparable, this
approach was abandoned because pilot-plant isolated 3 had
an intense dark-brown color (light-colored solutions of 3
facilitate the observation of color change during the formation
of lithium alkoxide 14 upon treatment with n-HexLi in the
subsequent steps). Also, higher levels of impurities were
The above laboratory findings led to the following plant
recommendations for efficient scale-up of this reaction.
Commercially available milled anhydrous K2CO3 (screened
through a #23 or #28 screen) was used in the plant. The rate
of heating was limited to about 0.5-1 °C/min to avoid the
rapid heat up. The rate of reaction under these conditions
was rapid during the first 5 h and then slowed. The last 10%
(18) During stress studies, this cyclic acetal was isolated and fully characterized
(NMR, MS, IR).
Vol. 8, No. 5, 2004 / Organic Process Research & Development
•
757

HPLC area, formed especially in the presence of p-toluene-
sulfonyl chloride) with 16, whereas 23 probably resulted from
hydrolysis of 5. Lab work also indicated that a large excess
of 4 typically resulted in higher amounts of 21. The
mechanism for formation of 19 and 20 is unclear. In earlier
studies it was proposed that aziridinium chloride 16 was the
reactive intermediate. However, we subsequently identified
that chloroamines 17 and 18 are formed during the prepara-
tion of 16 and influenced the course of reaction. A detailed
mechanistic investigation of this reaction has been published
recently.11
Thus, though small discrete pieces of the conversion of
3 to 5 were understood, the overall understanding remained
inadequate after the initial work. Hence, more detailed studies
were undertaken. The conversion of 3 to 5 involves several
major distinct steps as depicted in Scheme 4: (a) deproton-
ation of 3 to form the lithium alkoxide 14, (b) reaction of
the alkoxide with an activating reagent to form intermediate
15a, (c) in situ generation of aziridinium ion 16, which exists
in equilibrium with chloroamines 17 and 18, and (d) opening
of the aziridinium ion with Grignard reagent 4 to form
product 5. The discussion herein is focused on the salient
aspects of process development, optimization, and scale-up
of the above steps of this reaction.
Figure 2. Impurities isolated in aziridinium ring-opening
reaction.
observed due to the prolonged vacuum distillation operation
in the pilot plant.
3. Synthesis of 5. This is a critical multistep conversion
with minimal prior mechanistic understanding. The labora-
tory procedure identified earlier gave variable yields (45-
60%) in the key reaction sequence from 3 to 5. Our initial
efforts to improve the reaction yield by varying standard
reaction parameters (concentration, mode of addition, reac-
tion time/temperature) led to only minimal improvement in
the outcome. During this initial laboratory work some major
byproducts identified (LC-MS, or via isolation) from this
reaction sequence were R-tetralone and compounds 19-23
(Figure 2). Additionally, the presence of piperazinium
salts10,11 resulting from the dimerization of aziridinium ion
were also confirmed by LC-MS. Compound 22 originated
from reaction of dimeric Grignard reagent (up to 15% by
A. Deprotonation of 3. Of several different bases
n
screened for deprotonation, initially BuLi (2.5 M solution
in hexanes) was the reagent of choice in the alkoxide-
generation step because of quantitative conversion of 3 to
14 at low temperature and formation of a homogeneous
reaction mixture. Butane gas formed at low temperature (-10
to -30 °C) remained dissolved in the reaction solvents and
was liberated during subsequent reaction steps, which
Scheme 4. Aziridinium ion formation
758
•
Vol. 8, No. 5, 2004 / Organic Process Research & Development

Table 2. In-solution yields for conversion of 3 to 5 with
a pentavalent phosphorus. POCl3 on the other hand has
“three” leaving groups, and this probably led to other
reactions resulting in very low yield. In the DPPC reaction,
an insoluble precipitate of diphenyl phosphinic acid formed
at the end of reaction, which could be conveniently removed
by filtration. This, however, did not lower the workup burden
for this process. Because of the higher cost of the reagent
and the need for disposal of diphenyl phosphinic acid solid
waste, the use of DPPC was not pursued further. Potential
recycling of this reagent offered only minimal cost savings.
C. Generation and Stability of 16. The formation of 16
phosphorus activating reagentsa
1
was extensively studied by H, 13C, 15N (with 15N-labeled
a Experiments were conducted in the presence of CuCl‚2LiCl (5 mol %) for
reasons discussed later in this manuscript.
3), and 31P NMR (for DPCP) as well as by ReactIR. Due to
scant literature information on 15N NMR chemical shifts and
IR wavenumbers for aziridinium ions, an indisputable
interpretation of experimental data from these experiments
alone was difficult. These data will be published elsewhere.
However, on the basis of the combination of all spectroscopic
data above, HPLC chromatograms (of in-process and isolated
materials) and isolated yield the following conclusions were
drawn. These allowed for an identification of the preferred
process conditions described in the Experimental Section.
The conversion of 14 to 15a/15b was rapid at -20 °C, and
no byproducts from unreacted alkoxide 14 with 15a/15b were
observed in the reaction mixture. In the case of DPCP, 31P
NMR showed that 15b (δ -7.6 ppm, doublet, J ) 6.5 Hz)13
formed and stored at -20 °C was stable for 6-8 h. Only
upon warming to room temperature was the formation of
lithium diphenyl phosphate (δ -6.8 ppm) observed. On the
necessitated appropriate controls throughout this procedure.
Due to safety concerns on large scale, and due to the
precautions required in the usage of this reagent, nBuLi was
substituted with n-hexyllithium (n-HexLi, 2.5 M in hexanes)
as the byproduct hexane from this reagent posed significantly
reduced safety concerns. 1,10-Phenanthroline was added as
an indicator to the reaction mixture to facilitate observation
of the endpoint of n-HexLi addition to a solution of 3 in
THF (reaction mixture color change from pale yellow to deep
red). On-scale appropriate equipment modifications were put
in place for a facile monitoring of this change (site glass
with a recycling loop). Laboratory stress studies had estab-
lished that addition of up to 10% of excess n-hexyllithium
did not affect the reaction yield and purity. However, an
undercharge of n-hexyllithium led to unreacted 3, which
interfered with the isolation of 5. Thus, slight overcharging
was practiced for this reaction.
B. Activating Reagents. In published work p-TsCl was
used as an activating reagent. Among other sulfonyl chloride
activating reagents12 screened in our laboratory, p-TsCl
remained the reagent of choice. However, variable yields
and higher amounts of dimer 25 were obtained, especially
in the presence of the copper catalyst used to improve the
reactivity of the Grignard reagent (vide infra). A variety of
other (nonsulfonyl chloride) activating reagents were inves-
tigated of which phosphorus-based activating reagents gave
the most promising results (Table 2). From commercially
available phosphorus reagents diphenyl chlorophosphate
(DPCP) was selected for scale-up for the following reasons:
(a) a complete consumption of 3 (unlike with p-TsCl, where
about 2-4% of unreacted 3 was always detected); (b) overall
highest in-solution yield; (c) about 33% less dichloro dimer
25 formed (more details later) as the side product from 4
which lowered the usage of expensive 4; (d) a minimal
formation of 22.
1
basis of H NMR and 13C NMR studies it was established
that 16 in the reaction mixture existed in equilibrium with
chloroamines 17 and 18. Compound 18 was isolated and
characterized. In contrast, 15a was stable at -25 °C for about
1 h, and upon warming to -5 °C, 16, 17, and 18 were
formed. Compounds 17 and 18 are converted in situ to
aziridinium ion 16, which reacts with 4 to form 5. Between
17 and 18, the latter is about an order of magnitude less
reactive than the former towards formation of 5.11 In fact,
the existence of such a “chloramine sink” was beneficial as
it prevented the buildup of aziridinium ion in solution and
helped to reduce the formation of dimeric piperazinium
byproducts.
Finally, proof that the conversion of 3 to 16 was near
quantitative was obtained via the following control experi-
ments (in triplicate). Reaction of deschloro-4, i.e., 4-meth-
oxyphenylmagnesium bromide with 16 formed under the
optimum conditions described in the Experimental Section
afforded the corresponding product (deschloro-5) in overall
95% isolated yield. These results assured us that with DPCP
the yield for the aziridinium formation step (3 to 16) was
95% or higher.
Although other chlorophosphates were equally effective
as activating reagents, DPCP was preferred due to its lower
toxicity and moderate cost. Diethyl chlorophoshate (DECP)
is highly neurotoxic, and hence it was not explored further
despite obvious advantages of atom economy and the absence
of phenol in waste streams. DPPC was found to be
comparable to DPCP, whereas DPC and POCl3 were less
effective. DPC failed to provide any product, presumably
reflecting the poor leaving-group ability of the trivalent verses
D. Grignard Reagent (4). This was prepared from
5-bromo-2-chloroanisole (26) using the standard procedure
(addition of 5-bromo-2-chloroanisole to a well-agitated
suspension of magnesium metal turnings in THF). A flat-
bottom reactor with efficient agitation provided a large
surface area for this heterogeneous reaction. With conical-
bottom reactors Mg turnings accumulated at the bottom
Vol. 8, No. 5, 2004 / Organic Process Research & Development
•
759

Table 3. Role of Cu(I) catalyst in the formation of 5 from 3
demonstrated that Cu significantly enhances the nucleophi-
licity of 4 and hence favors the formation of 5 compared to
that of deschloro-5.
Grignard reagent
catalysta
product (% yield)b
4
none
none
CuCl‚2LiCl
CuCl‚2LiCl
5 (60)
deschloro-5 (95)
5 (80)
Other observations of interest are the following: Cu(II)
salts gave about 5-10% lower yield than Cu(I) salts.
Compared to CuCl alone, a soluble form of copper reagent
in THF formed by the addition of LiCl to CuCl gave
reproducible yields and was more amenable to scale-up.
Hence, this was used for the production process. Although
CuCN‚2LiCl gave about 3-5% more in-solution yield than
CuCl‚2LiCl, the use of CuCN on scale was avoided due to
toxicity and waste disposal issues. Addition of more than 5
mol % CuCl‚2LiCl did not further improve the yield and
also made workup difficult due to emulsion formation.
When the Grignard reagent was mixed with activating
reagents to simulate stressed conditions in the absence of
air and in the presence of copper salts (CuCl and its THF
soluble LiCl complex), only small amounts of phenol 29 and
dichlorodimer 25 formed when DPCP was used as the
activating reagent. However, when p-TsCl was used as the
activating reagent under similar conditions, dichlorodimer
25 was formed as the major product along with magnesium
p-toluene sulfinate as a byproduct. A very small amount of
tosylated product 32 was obtained. In contrast, DPCP did
not enhance reductive dimerization, but it slowly reacted with
4 to form the corresponding phosphate adduct 31. This
difference in reactivity of DPCP and p-TsCl towards 4 was
one more factor that led us to select DPCP on scale. Thus,
with DPCP the amount of 4 was reduced to 1.2 equiv from
approximately 2 equiv needed with p-TsCl for an optimum
yield of 5. Under the optimum conditions, the former offered
higher yields of 5.
deschloro-4
4
deschloro-4
deschloro-5 (94)
Grignard reagent
catalysta
ratio 5:deschloro-5
4 + deschloro-4 (1:1)
4 + deschloro-4 (1:1)
none
CuCl‚2LiCl
1:5.4
1:1.7
a 5 mol % catalyst. b Solution yields via HPLC.
valve, resulting in a slow or an incomplete reaction. The
presence of air during the reaction led to the formation of
phenol 29 and other dimeric biphenyl impurities (24 and 25).
Hence, before charging 26 the reactor was evacuated and
flushed with nitrogen. The reaction was reliably initiated by
adding about 7% of the total charge of 26 (as a solution in
THF) to a suspension of Mg chips in THF. To check for
reaction completion, a portion of the reaction mixture was
quenched in methanol and assayed for 2-chloroanisole (27)
content by GC versus unreacted 26. Cooling this solution to
0-5 °C resulted in a gel that dissolved upon warming to
30-35 °C. Once formed, the reagent (ca, 1.2 M in THF)
could be stored at room temperature under nitrogen atmo-
sphere for 4 months without loss of titer.
E. Effect of Copper Salts on the Reactivity of 4 with
16.19 Even under the optimized aziridinium formation
conditions the yield for the conversion of 3 to 5 remained
around 60-65% with the remaining mass in many byprod-
ucts. The control experiments reported above clearly indi-
cated that the formation of 3 to 16 was high yielding (>95%).
Thus, the loss of yield was in the conversion of 16 to 5. The
control experiments also suggested that with a more reactive
nucleophile (i.e., deschloro-4) overall conversion of 3 to the
corresponding product (deschloro-5) can be high (95%).
Thus, it was concluded that enhancement of the reactivity
of 4 was needed, and research efforts were directed towards
that end. Copper is known to influence the reactivity of
Grignard reagents with different electrophiles.14 Of the
several copper salts screened,15 addition of a catalytic amount
(5 mol %) of CuCN and CuCl (either to 4 or to a solution of
16, or to a mixture of both) increased the nucleophilicity of
the Grignard reagent and improved the solution yields of
the aziridinium ring-opening reaction by 20-25% (Table 3).
It is plausible that a more reactive cuprate might be generated
in situ by transmetalation of 4 with Cu(I) salts. Competitive
kinetic experiments (last two entries in Table 3) clearly
F. Order of Mixing. Once the aziridinium intermediate
was generated, it could be reacted with 4 in three different
ways: (a) direct addition of solution of 4 to a solution of
16; (b) inverse addition of 16 to 4; and (c) mixing streams
of 4 and 16 in a continuous mixing reactor. All three of these
modes of addition were studied for improving the yields and
processability of this reaction. These studies also showed
that the copper catalyst could be added to a solution of 4 or
16 before mixing the solutions of 4 and 16 with practically
no impact on the outcome.
(a) Direct Addition. For the first-generation p-TsCl
process, the direct-addition mode gave better yields. To avoid
formation of the relatively less reactive chloroamine 18,11
the solution of 15a was held at -20 °C, and the copper
catalyst was added prior to adding 4. Since the reaction of
4 with 16 and chloroamines was relatively slow at -20 °C,
the reaction mixture was rapidly warmed to 30-35 °C to
(19) Organolithium and organozinc reagents were evaluated in place of Grignard
reagents. The former two gave poor yields and/or many byproducts; hence,
the decision was made to optimize the Grignard reagent process.
760
•
Vol. 8, No. 5, 2004 / Organic Process Research & Development

speed up the desired reaction.16 If the reaction mixture was
not warmed promptly after the addition of 4, higher amounts
of 19, 20, 25, and piperazinium dimers were formed. Under
ideal conditions about 65-75% solution yield of 5 was
obtained via the direct-addition process. The adiabatic
temperature rise for the reaction of 4 with 16 was 50 °C.
On-large-scale heat-transfer manipulations were tedious and
demanding due to the initial need of keeping the reaction
mixture cooler during addition of 4 and then the need to
heat to 30-35 °C. When 15a was subjected to inverse
addition in the lab, about 10% lower solution yield of 5 was
obtained. Here, large amounts of byproducts from 4 were
generated, necessitating an excess of 4 for consumption of
16. This in turn led to workup difficulties, leading to lower
isolated yields.
(b) InVerse Addition. This mode proved to be best when
DPCP was used as an activating reagent. The conversion of
phosphate 15b to 16 and further to chlooramines (17,18) was
very slow at -20 °C. Thus, the undesired formation of
chloramine 18 was minimized when cold 15b was added to
a solution of 4 (containing the catalyst) maintained at 20-
25 °C. Unlike the direct addition mode, an excess of 4 is
available for reaction with 15b in the initial phase of the
reaction. Typical solution yields of 5 obtained in this
procedure were 80-85%. Heat exchange in the inverse
addition mode was well controlled, resulting in reproducible
yields. This process was readily scaled up in three plants
without any significant heat transfer issues.
(c) Continuous Mixing. Mixing streams of the Grignard
reagent containing the catalyst and the aziridinium reaction
mixture in a static mixer equipped with a heat exchanger
also gave 80-85% yield. The proof of principle was
demonstrated on a 200-g-scale reaction in the laboratory. This
process was not scaled up in the pilot plant as the inverse
addition process described above allowed for a rapid scale-
up with the available plant equipment and utilities. Note that
this continuous mixing procedure worked well for p-TsCl
as well as for DPCP processes.
G. Workup. This is a multistep transformation with
several reagents, byproducts, and side products. An immense
cost of raw materials as well as processing is involved at
this stage of the process. Thus, it was very important to
identify a sleek and reproducible isolation procedure for
good-quality 5. After completion of reaction, aqueous
ammonium chloride solution was added to quench the
unreacted Grignard reagent. Mg and Cu salts did not
precipitate with the NH4Cl workup (aqueous layer, pH 8).
Copper chloride formed the blue [Cu(NH3)4]+2 complex with
ammonium chloride that remained in the aqueous layer, thus
efficiently removing it from 5 in the organic layer. Other
quenching reagents such as phosphate buffer, sodium acetate,
acetic or dilute sulfuric acid were unsatisfactory due to
precipitation of salts and emulsion formation.
remained dissolved in the water, and the pure product was
extracted into MTBE. This tedious acid-base workup was
unnecessary with DPCP because the solution yield of 5 was
high and the amount of neutral dimeric and other side
products was low. This was another criterion that further
supported the selection of the DPCP process for the com-
mercial process. The procedure used for large-scale work is
available in the Experimental Section.
Environmental aspects can play in important role in the
commercial development of a process. For the DPCP process
the amount of phenol generated during the workup was
closely monitored as the plan was to remedy the aqueous
waste by treating it with microorganisms. These organisms
are affected by the antimicrobial phenol generated in the
process via hydrolysis of DPCP. The amount of phenol in
the aqueous waste streams was <0.5%, which was an
important consideration for waste management.
Aqueous sodium hydroxide could be used to remove the
small amount of phenol present in organic extracts, but
emulsions were formed and the color of organic extracts
darkened under basic conditions. This was avoided by a
controlled NH4Cl quench that minimized their formation.
Also, residual inorganic salts were sometimes carried into
the isolation of 5 as a salt and gave product with acceptable
impurities but lower wt % purity. To overcome this, an
aqueous NH4Cl quench followed by aqueous sodium acetate
washing of the organic extract was developed that prevented
emulsions and facilitated the removal of inorganic salts. A
final 5% aqueous sodium chloride wash of the organic extract
gave a stream suitable for direct crystallization of 5.
H. Salt Formation of 5. In the early phase of this work
5 obtained as an oil was used for further reactions. However,
this practice left only one isolation via crystallization
throughout the process, i.e. the isolation of penultimate
intermediate 6. Although this was practiced for the initial
amounts of API, in the long range it was desirable to have
a process that would allow isolation of 5 as a solid of known
quality. This would also allow for purification as well as
extended storage. Thus, a tremendous amount of effort was
exerted in identifying an efficient purification/isolation of
5. Of dozens of organic and inorganic acids screened in a
number of solvents, only oxalic acid gave a crystalline salt
in high yield and purity.
Isolation of the oxalic acid salt of 5 was unique. For
example, various combinations of alcoholic solvents (MeOH,
i
n
i
t
EtOH, PrOH, BuOH, BuOH, BuOH) or etheral solvents
(THF, MTBE, DME, diglyme) or esters (MeOAc, EtOAc,
iPrOAc) and hydrocarbons (hexanes, heptane, toluene) were
unsuitable for direct crystallization of 5A. Precipitation of
oxalic acid salt from ethers and hydrocarbon solvents was
also difficult. On the other hand, a combination of MTBE
i
and PrOH (1:1) was found to be the best system for
crystallization. Via this process neutral dimeric impurities
as well as polar impurities were purged efficiently. Seeding
of the crystallization mixture at room temperature and then
gradual cooling to 0-5 °C and a hold time of about 18 h
were critical to get the maximum isolated yield of quality
product. The presence of excess water and inadequate
For the p-TsCl process an acid-base workup was needed
to isolate 5. The organic extract was treated with 1 N H2-
SO4, and then the acidic extract was neutralized with 10%
NH4OH to pH 5.0-5.5. Since 5 is less basic (pKa of 5 was
determined to be about 6.0), other more basic impurities
Vol. 8, No. 5, 2004 / Organic Process Research & Development
•
761

Scheme 5. Cyclization and reduction of 5A and 6A
crystallization time led to lower recovery of the product. Use
of anhydrous oxalic acid improved the yield by 2-3% at
the expense of purity. The amount of water present in oxalic
acid dihydrate was optimal and gave 5A in consistently good
yield (85-90% recovery of 5) and purity. On average, about
8-12% product was lost in the mother liquor.
Residual isopropyl alcohol in 5A forms an isopropylated
impurity (37) in the cyclization/elimination reduction se-
quence, which is difficult to purge and which contaminates
the final product. Hence, the amount of residual isopropyl
alcohol trapped in 5A was controlled to less than 0.5% w/w
by vacuum-drying at 50-60 °C.
Finally, isolation of 5 as an oxalate salt (5A) led to an
unexpected result. It was found that starting with 12 of 94%
ee, 5A with 98-99% ee was isolated after crystallization.
Thus, formation of 5A not only removed other chemical
impurities but also purged diastereomeric impurities and
significantly improved the enantiomeric purity of 5A without
the use of an expensive resolving reagent. Since 12 was
always obtained in high ees, this finding was not put to use
for commercial purposes.
I. Scale-Up of Aziridinium Chemistry Experience. Our
first-generation process involved nBuLi/p-TsCl (1.2 equiv)/4
(1.3 equiv)/CuCN‚2LiCl (10 mol %) and a direct addition
mode that gave an average solution yield of about 65%. Upon
acid-base workup followed by oxalate salt formation, 5A
was isolated in about 50-60% yield. This process was scaled
up to 50 kg. A robust second-generation process was sought
because of irreproducible yields, scale-up difficulties resulting
from rapid warming requirements, acid-base workup, and
issues with cyanide waste from CuCN.
to 34, derived from the supply-route plant conditions where
racemic 6 was made via an alternate route, used CH2Cl2 as
the solvent with MeSO3H or sulfuric acid as the reagents
that caused cyclization/elimination. The reaction sequence
for Friedel-Crafts cyclization of 5 proceeded via intermedi-
ate 33 (structure confirmed by LC-MS) which eliminated
methanol to form enamine 34. The reaction mixture was
partially neutralized by adding a slurry of NaHCO3 in CH2-
Cl2, and then 34 was reduced in situ with tert-butylamine
borane complex (TBAB) to amine 6 (Scheme 5).
The reaction of free base 5 with methanesulfonic acid to
give enamine 34 in CH2Cl2 took about 3 days for completion
(<2% 33) and required 16 volumes of CH2Cl2. The higher
dilution was needed to minimize the dimeric/polymeric
byproducts. Also the neutralization of the acidic reaction
mixture with NaHCO3 before the addition of TBAB led to
severe frothing which was difficult to control on scale. The
controlled addition of NaHCO3 as a CH2Cl2 slurry was time
consuming and equipment constraining. The prolonged
reaction and processing times seriously limited the through-
put for this step. Finally, due to environmental concerns with
the use of CH2Cl2 on scale, an alternate nonhalogenated
solvent was required for the reaction sequence. In fact this
provided an opportunity to arrive at a better process that
eliminated all these restraints.
Several acids such as sulfuric acid, BF3‚AcOH, AcOH,
formic acid (which in principle can serve as a Brønsted acid
as well as a hydride source), P2O5-methanesulfonic acid
(Eaton’s reagent), p-toluenesulfonic acid, and camphorsul-
fonic acid were evaluated for the cyclization/elimination
reaction sequence in a variety of solvents such as sulfolane,
i
In our second-generation process, n-HexLi/DPCP (1.1
equiv)/4 (1.2 equiv)/CuCl‚2LiCl (5mol %), with inverse
addition mode gave solution yields of about 85%. After NH4-
Cl quench, followed by MTBE extraction and 5% NaOAc
and NaCl washes, the product was directly crystallized from
MTBE-IPA (1:1) in ∼70% yield. This process was scaled
up to 100+ kg of 3 per batch.
diglyme, toluene, NMP, MeCN, and PrOAc. Interesting
results from these screening experiments are summarized in
Table 4. Although Eaton’s reagent, and the sulfuric acid/
sulfolane system provided desirable conversions of the free
base 5 to 34, they were not pursued as more efficient
conditions were discovered with neat methanesulfonic acid
(entry 6). Sulfolane with about 12 equiv of sulfuric acid at
70-75 °C for 3 h gave a clean conversion of 5 to 34. This
reaction mixture was then subjected to typical transfer
4. Formation of 6A. The laboratory conditions for the
cyclization/elimination steps, i.e., conversion of free base 5
762
•
Vol. 8, No. 5, 2004 / Organic Process Research & Development

Table 4. Friedel-Crafts cyclization/elimination and reduction of 5 or 5a to 6
entry
conditions
results/comments
A. Cyclization/Elimination
camphorsulfonic acid (1.8 equiv)/PhCH3, reflux, 24 h
formic acid (excess), reflux, 24 h
Eaton’s reagent (excess), rt, 5 h
diglyme/ H2SO4 (12 equiv), rt, 4 h
sulfolane/H2SO4 (12 equiv), 70-75 °C, 3 h
MeSO3H (12 equiv), 65-70 °C, 4 ha
B. Reductionb
% Conversion to 34
1
2
3
4
5
6
5
0
95
46
98
99
% Conversion to 6
7
8
9
10
11
12
polymethylhydrosilane, rt, 24 h
Et3SiH, rt, 24 h
0
0
60
53
80
99
HCO2NH4/Pd/C, H2O, rt, 12 h
Na2HPO4/NaBH4 (5 equiv)/iPrOH-H2O, rt, 6 h
NH4OH/NaBH4 (5 equiv)/iPrOH-H2O, rt, 6 h
TBAB (2.0 equiv)/MTBE, 5-10 °C, 2 h
a 5A was used instead of free base 5. b Reaction conducted with mixture from entry 5.
Figure 3. Impurities identified in 6.
hydrogenation and hydride reduction conditions. The main
challenge associated with using hydride reagents was the high
acidity of the reaction medium which had to be moderated
by addition of a base prior to the addition of the hydride
reagent to avoid a violent reaction with hydrogen evolution.
Although sodium borohydride showed good results, a
complete conversion of 34 to 6 could not be accomplished
by adding excess reagent. With TBAB a complete conversion
was obtained (<1.0% of 32).
Some impurities formed in trace amounts were identified
on the basis of LC-MS studies and are shown in Figure 3.
Their genesis was traced to the following. When lumps
formed during the charge of 5A to MeSO3H (e.g., inefficient
agitation, rapid addition, cooler temperature), unreacted 5
was carried over in isolated 6. Analyses of several samples
of isolated 6 showed that the des methyl impurity 35 and
the des-chloro impurity 36 originated in the aziridinium
chemistry step as well as in the cyclization/reduction
sequence. Typically impurities 35-37 formed in small
(<0.5% amounts) and were removed during the purification
of 6 and/or 1. The isopropyl derivative of 6 (37) was
Finally, we discovered that when 5A itself was added
directly to MeSO3H (12 equiv) and heated to 55-65 °C for
2 h, a near quantitative conversion to 34 was observed. Since
MeSO3H is a stronger acid than oxalic acid, 5A is converted
to the MeSO3H salt which then undergoes cyclization. This
procedure precluded the need of any solvent and the necessity
of converting solid 5A into a free base. Furthermore, the
reaction time was shortened to a few hours. As dissolution
of solid 5A in methanesulfonic acid was rather slow, efficient
agitation and warming of the reaction mixture was necessary
for complete reaction. The viscous reaction mixture was
cooled to 0-5 °C and diluted with MTBE. A suspension of
TBAB (1.1 equiv) in MTBE was added slowly to the reaction
mixture to reduce 34 to 6. Here too, of various reducing
agents evaluated, TBAB provided a fast and complete
reduction. The acidity moderation of the reaction mixture
by the tert-butylamine in the TBAB complex, and its
solubility in organic solvent made it an ideal reagent for the
reduction step. After dilution with water (exothermic) the
reaction mixture was quenched with 25% aqueous KOH
solution. Use of NaOH led to formation of less soluble
sodium oxalate and emulsions. As potassium oxalate has
higher water solubility, KOH was the base of choice which
enabled an efficient removal of oxalic acid.
i
correlated with residual PrOH content in 5A. Impurity 37
when formed in >0.5% could not be easily purged from 6
or 1; hence, it was controlled by limiting the iPrOH content
to e0.5% in 5A. Dimeric impurities 38 were also identified,
and their formation could be lowered by using a higher
temperature (70-80 °C) during the conversion of 5A to 34,
but the higher temperature led to more 37. Since 38 could
be purged more easily than 37 in the subsequent purifications,
the higher temperature conditions were not used.
Crystallization of 6. Since penultimate intermediate
amine 6 is an oil, a crystalline salt was desirable to aid its
purification. Of several inorganic and organic acids screened
for salt formation with 6 under a variety of conditions, (+)-
DTTA and malonic acid gave the best yield and purity. Since
compound 6 was of high chiral purity to begin with, the use
of expensive chiral (+)-DTTA was undesirable for a non-
chiral purification. Hence, readily available, inexpensive
malonic acid was chosen for optimization/scale-up. Our
criteria for salt selection were yield and quality (to meet or
exceed 6 made via the supply route where it was isolated as
(+)-DTTA salt). Specific optimization was needed to ef-
Vol. 8, No. 5, 2004 / Organic Process Research & Development
•
763

Table 5. Salt formation Screening of 6a
an overall 34% via a commercially feasible route. Next, the
N-alkylation step was optimized wherein the reaction time
was reduced and 90% yield was realized. The aziridinium
ion based synthesis was successfully scaled in three different
plants to produce several hundred kilograms of Sch 39166.
This synthesis highlights the formation and regio- and
stereoselective opening of an aziridinium ion with a Grignard
reagent. The stability and reactivity of the aziridinium ion
was studied to improve the reaction yield. A catalytic amount
(5 mol %) of CuCl‚2LiCl complex improved the nucleophi-
licity of the Grignard reagent. In this sequence of four
reactions, an overall isolated yield of 70% was achieved
without any need of chromatographic purification. Another
key step where two reactions were telescoped was the
conversion of 5A to 6A. Here, by careful selection of reaction
parameters the processing time was lowered to 2 days from
7 days, throughput was improved 4-fold, and the reaction
yield was improved by 20%. Furthermore, the use of
environmentally unacceptable CH2Cl2 and expensive (+)-
DTTA was eliminated. After final demethylation followed
by recrystallization, the hydrochloride salt was isolated from
R-tetralone in an overall yield of about 20%.
%
LC
35
salt/conditions
recovery area % purity
%
(+)-DTTA (nBuOH:H2O, 9:1)
HCl (MeCN)
80
56
89
76
99.5
98.4
93.8
99.6
0.16
0.11
0.38
0.10
HCl (PhCH3)
malonate (iPrOH:H2O, 95:5)
a Compound 6 used for these experiments had LC area % purity of 93.0%
and 0.67% of 35.
ficiently purge desmethyl impurity 35 (Table 5) via the
malonate salt process. Among several solvents evaluated,
aqueous iPrOH was the solvent of choice. A higher amount
of water (>5%) led to high-quality product with a corre-
sponding increased loss of product to the mother liquor.
Impurities resulting from dimerization of 5A were removed
i
by crystallization in PrOH-H2O (95:5). However, since
these impurities were controlled by process conditions, the
efficiency for their removal during the salt-formation step
was not critical. For purging impurities for a typical 6
generated via the chosen process ∼97:3 PrOH:H2O was
i
ideal. Via this procedure (described in the Experimental
Section) the yield for the isolation of 6A was 90%. It is
interesting that the stressed conditions in the laboratory
showed that the isopropylated impurities in >0.5% could
not be removed by IPA-water crystallization. They could
be removed only after treating isolated 6A in THF, albeit
with a large loss of yield. These isopropylated impurities 37
cannot be purged from 1, even after recrystallization, whereas
dimeric impurities 38, if present, were purged out in the final
recrystallization step. Both the reaction and the isolation were
successfully implemented in three plants where the overall
yields for the conversion of 5A to 6A were 80-85%.
Final Demethylation Step. Conversion of 6 to 1. The
chemistry used in this step, BCl3-mediated selective O-
demethylation of 6, was kept the same as in the supply route
to maintain the process lock and match the purity as well as
the polymorphic profile of the drug substance. Advantages
of BCl3 were a clean reaction and direct isolation of the
desired hydrochloride salt in high yield. Salt 6A was
converted to 6 by treatment with 25% NaOH at 40-50 °C
in toluene. and the toluene extract was distilled to remove
water as an azeotrope. The reactor used for the BCl3 reaction
was equipped with a condenser maintained at -20 °C to
condense BCl3 (bp 12 °C) but not MeCl (bp -24 °C). BCl3
was charged as a liquid above the batch surface, and as the
addition was exothermic, the batch temperature was main-
tained below 50 °C. The reaction mixture was heated at 60-
70 °C for 18 h to obtain g99.5% conversion. The batch was
cooled to room temperature, and excess BCl3 as well as the
borate complex was quenched by adding in portions to excess
MeOH. After distillation of B(OMe)3, MTBE was added to
precipitate 1. A final crystallization of 1 from a MeOH/
MTBE mixture gave Sch 39166 in 99+% chemical and
>99% chiral purity.
Experimental Section
Melting points are uncorrected. 1H NMR (400 MHz) and
13C NMR (100 MHz) were recorded in CDCl3 or (CD3)2SO
solutions. Unless specified otherwise, all reagents and
solvents were used as supplied by the manufacturer. All
reactions were conducted under an inert N2 atmosphere.
trans-(()-2-Bromo-1-hydroxytetralin (9). To a glass-
lined reactor were charged ethanol (400L) and sodium
borohydride (7.20 kg, 190 mol) at ambient temperature. The
reactor was evacuated and purged twice with nitrogen to
reduce the oxygen concentration. Slowly, R-tetralone (80.4
kg, 550 mol) was added to the NaBH4 suspension over a
period of about 30 min (Caution! Hydrogen evolution!)
without cooling. The batch temperature increased to about
45 °C. After the completion of addition, the solution was
heated to 50-55 °C for 3 h and sampled for reaction
completion. Another reactor was charged with water (240
L), evacuated and purged twice with nitrogen. The reduction
reaction mixture (precooled to 20-25 °C) was added over a
period of about 20 min to water without cooling. Ethanol
was distilled under reduced pressure (45-50 °C at 120-
160 mbar pressure) to a residual volume of about 120 L.
The batch was cooled to 20-25° C, diluted with toluene
(240 L) and agitated, and the water layer was separated. The
aqueous layer was extracted twice with toluene (2 × 80 L),
and the combined toluene phases were washed with water
(80 L). This solution of 7 (98-100% in-solution yield) was
used directly for the next stage.
The toluene solution of 7 was atmospherically concen-
trated to a volume of about 305 L at normal pressure. The
water content of the batch was determined by KF titration
(0.05%). The solution was cooled to 60-70 °C, and
Summary
The process for the key starting material 12 from readily
available R-tetralone was developed and optimized to obtain
764
•
Vol. 8, No. 5, 2004 / Organic Process Research & Development

p-toluenesulfonic acid monohydrate (400 g, 2.1 mol) was
added. The solution was heated at reflux (80-110 °C) for
about 4 h, and water generated in the reaction was distilled
off azeotropically (approximately 9.9 L water separated).
When no more water was separated, the batch was sampled
for reaction completion by HPLC. The solution of dihy-
dronaphthalene (8, about 98-100% in-solution yield) was
cooled to 20-25 °C and used directly in the next step.
To a glass-lined reactor were charged 1,3-dibromo-5,5-
dimethylhydantoin (78.6 kg, 275 mol, 0.5 equiv), acetone
(720 L), and water (80 L). The mixture was agitated at 20-
25 °C to get a clear solution and then cooled to 0-5 °C. A
solution of 8 in toluene (as obtained above), was added over
about 60 min while maintaining the batch temperature
between 0 and 10 °C (exothermic reaction). If at the end of
addition any solids precipitated, more acetone was added to
get a clear solution. The reaction mixture was warmed to
20-25 °C and agitated for 4 h and sampled for reaction
completion. After complete reaction, add 320 L of water was
added, and the reaction mixture was concentrated under
vacuum (45-50 °C at 200-300 mbar pressure) to a volume
of 400 L. Toluene (80 L) was added to the suspension, and
the batch was cooled to 0-5 °C. The precipitated product
was filtered, washed well with water (6 × 80 L), and dried
in a vacuum tray dryer at 45-50 °C for 12-24 h to obtain
106.4 kg of 9 (overall 84.2% yield from R-tetralone).
trans-(1R,2R)-2-Hydroxy-1-methylaminotetralin, Tar-
trate (12). Compound 9 (107.5 kg, 473 mol) was charged
to a glass-lined reactor, and then water (215 L) and
methylamine (40% solution in water, 424 L, 4 mol, 5.0
equiv) were added. The suspension turned into a solution
after stirring for 5 h at 35-40 °C. The reaction mixture was
sampled for reaction completion, and sodium chloride (53
kg) was charged to saturate the aqueous layer. The batch
was extracted with tert-butylmethyl ether (MTBE, 3 × 107
L), and the combined MTBE phases were washed with water
(28 L). The MTBE solution was concentrated as much as
possible under reduced pressure (40-45°C at 200-400 mbar
pressure), methanol (53 L) was added to the residue and
distilled again under vacuum at 40-45°C to remove meth-
ylamine, MTBE, and water. More methanol (53 L) was
charged, and the distillation procedure was repeated. The
resulting residue of (()-2 was dissolved in methanol (242
L) at 40-45 °C. In a separate reactor, L-(+)-tartaric acid
(17.7 kg, 118 mol, 0.25 equiv) was dissolved in methanol
(35 L) at 20-40 °C. This solution was added to the racemic
solution of (()-2 in MeOH over a period of about 20 min,
and the temperature was held at 50-55°C for 1 h. The
suspension was cooled to 0-5 °C and agitated for 1 h. The
precipitated solid was filtered, and the mother liquor was
removed as much as possible under vacuum and washed with
cold (0-5°C) methanol (41 L). The product was dried in a
vacuum tray dryer at 50-55 °C for about 12 h to afford
58.5 kg of 12 (overall 49% yield from 9, 78-83% ee).
The enantiomeric purity of the product was upgraded by
recrystallization from aqueous 2-propanol. Compound 12
(obtained above, 58.5 kg) was charged to a reactor, and
2-propanol (175 L) and water (52 L) were added. The
suspension was heated to reflux (80 °C), and at reflux
temperature water (10 L) was added to get a clear solution.
The batch was held at reflux temperature, and additional
2-propanol (293 L) was added over 40 min without further
heating. The temperature dropped to about 50 °C, and the
suspension was cooled to 0-5 °C and agitated for 1 h. The
product was filtered, washed with cold (0-5°C) 2-propanol
(75 L), and dried at 50-55 °C for 18 h in a vacuum tray
dryer to provide 43.0 kg of 2 (78% yield, 99.0% ee).
trans-(+)-(1R,2R)-trans-1,2,3,4-Tetrahydro-1-[(2,2-di-
methoxyethyl)methylamino]-2-naphthalenol (3). A. Acid-
Base Workup Procedure. Methyl tert-butyl ether (MTBE,
200 mL) and water (200 mL) were added to a mixture of
sodium chloride (70 g) and (+)-(1R,2R)-trans-1,2,3,4-
tetrahydro-1-(methylamino)-2-naphthalenol hemitartrate (100
g, 0.198 mol). Ammonium hydroxide (100 mL) was added
to the suspension, and the mixture was stirred at room
temperature for 30 min to get a clear solution. The aqueous
layer was separated and extracted with MTBE (200 mL),
and the combined MTBE extracts were distilled atmospheri-
cally to a volume of 200 mL. Diglyme (200 mL) was then
added to the free base and again distilled atmospherically
under nitrogen atmosphere to a remove MTBE. The solution
was further heated to 105-115 °C, and vacuum was applied
carefully until the temperature dropped to 70-75 °C. This
ensured removal of MTBE as well as water (the water
content of the batch by Karl-Fisher titration was <0.1%)
and cooling of the batch needed for subsequent operation.
After cooling to room temperature, freshly powdered anhy-
drous potassium carbonate (93 g, 0.67 mol) and bromoac-
etaldehyde dimethyl acetal (57 mL, 0.48 mol) were added
to the free base solution in diglyme. The suspension was
agitated and heated gradually over a period of 3 h to 125-
130 °C and then held at 130 °C for 18 h. HPLC analysis of
the reaction mixture indicated less than 2% of starting
material. The suspension was cooled to room temperature
and filtered to remove the solids. Solids and the reaction
flask were washed with two portions of MTBE (200 mL
and 100 mL). To the combined filtrate was added 1 N
sulfuric acid (2 × 200 mL), and the mixture was stirred at
room temperature for 10 min. The lower aqueous layer was
separated, and the organic layer was again extracted with 1
N sulfuric acid (200 mL). Sodium chloride (120 g) was added
to the acidic aqueous layer and stirred until the solids almost
dissolved. MTBE (200 mL) was added to the mixture, and
then pH was adjusted to 10 using ammonium hydroxide. The
organic layer was separated, and the aqueous layer was
extracted again with MTBE (200 mL). The combined organic
extracts were distilled atmospherically under nitrogen atmo-
sphere to remove MTBE and concentrated to a volume of
200 mL. By HPLC analysis, the in-solution yield of the
product in diglyme and its purity were determined to be 93-
97 g (0.351-0.366 mol, 90-92% molar yield) and >95%,
respectively. EIMS: 267 (M + 2, 24), 266 (M + 1, 100),
235 (20), 234 (95), 202 (35), 184 (7), 171 (8), 155 (15).
B. Vacuum Distillation Procedure. Toluene (200 mL)
and water (81 mL) were added to (+)-(1R,2R)-trans-1,2,3,4-
tetrahydro-1-(methylamino)-2-naphthalenol hemitartrate (100
Vol. 8, No. 5, 2004 / Organic Process Research & Development
•
765

g, 0.198 mol). The suspension was stirred at room temper-
ature, and then 25% aqueous sodium hydroxide (68 mL) was
added in one portion. The temperature of the reaction mixture
increased to 35 °C, and the suspension upon further stirring
for 30 min turned into a clear solution. The lower aqueous
layer was separated and discarded. 2-Methoxyethyl ether
(diglyme, 220 mL) was added to the pale-yellow toluene
extract and then heated under nitrogen atmosphere to 120-
125 °C to distill toluene. The solution was cooled to 70-80
°C, and vacuum (10 mmHg) was applied to concentrate the
solution to a volume of 250 mL. The water content of the
batch by Karl-Fisher titration was found to be 0.1%. Freshly
powdered anhydrous potassium carbonate (93 g, 0.67 mol)
and bromoacetaldehyde dimethyl acetal (57 mL, 0.48 mol)
were added to the free base solution in diglyme. The
suspension was agitated with a mechanical stirrer and
gradually heated over a period of 3 h to 125-130 °C under
a nitrogen atmosphere. The heating was continued for 18 h
after which the pale red-brown reaction mixture was cooled
to room temperature and filtered. The reaction flask and
solids were rinsed with two portions of toluene (200 and
100 mL each), and the filtrate was concentrated under
vacuum at 65-70 °C to a volume of 150 mL. By HPLC
analysis, the in-solution yield of the product in diglyme and
its purity were determined to be 93-96 g (0.351-0.362 mol,
90-91.2% molar yield) and >95%, respectively.
stirring the reaction mixture at -30 °C for 10 min, a solution
of p-toluenesulfonyl chloride (79 g, 0.414 mol) in THF (200
mL) was added dropwise. During this exothermic addition
the reaction temperature was maintained between -20 to
-30 °C. The reaction mixture was held at -20 °C for 2 h.
Separately, a slurry of copper (I) cyanide (3.37 g, 0.037 mol)
in THF (50 mL) was cannulated to the aziridinium solution
held at -25 °C. The mixture was stirred for 5 min, and then
the Grignard solution was added to the aziridinium ion
solution held at -25 °C via a cannula. The addition took
about 25 min, and during the addition the temperature of
the reaction mixture solution was allowed to rise from -25
to -5 °C. After the addition, the reaction mixture was quickly
heated to 40-45 °C and held at that temperature for 1.5 h.
The reaction mixture was cooled to 0 °C and quenched with
7% aqueous ammonium chloride solution (500 mL). The
reaction mixture was warmed to room temperature and stirred
for 30 min. The organic layer was separated and stirred with
7% aqueous ammonium chloride solution (700 mL) for 15
min, and the aqueous layer was separated. The aqueous layers
were extracted with MTBE (200 mL). The combined organic
layers were stirred with 1 N sulfuric acid (220 mL) for 15
min, the lower aqueous layer was separated, and then the
organic layer was stirred with 1 N sulfuric acid (220 mL)
for 15 min. MTBE (200 mL) was added to the combined
aqueous layer, and the pH was adjusted to 5.5-6.0 using
ammonium hydroxide. The organic layer was separated, and
to the aqueous layer was added MTBE (200 mL). The pH
was again adjusted to 5.5-6.0 with ammonium hydroxide,
and the organic layer was separated. The organic layer was
distilled atmospherically to a volume of 200 mL. 2-Propanol
(200 mL) was added to the dark-brown viscous solution, and
the solution was filtered through a sintered glass funnel. The
funnel was rinsed with 2-propanol (100 mL), and the
combined filtrate was warmed to 40 °C. Separately, a
solution of oxalic acid dihydrate (47 g, 0.37 mol) in
2-propanol (200 mL) was prepared by warming to 45 °C
and added to the free base solution. The mixture was stirred,
gradually cooled to room temperature, and seeded with a
slurry of product (0.5 g) in 2-propanol (5 mL). After 2-4 h
the product crystallized out, and the slurry was stirred at room
temperature for an additional 8 h. The slurry was cooled and
held at 0 °C for 4 h and then filtered. The cake was washed
with ice-cold 2-propanol (200 mL), then with ice-cold MTBE
(200 mL), and finally again with ice-cold MTBE (100 mL).
The solid was dried under vacuum for 4 h and then in a
vacuum oven at 45 °C for 12 h to give an off-white solid
(109-118 g, 60-65% molar yield). FABMS: 235, 271, 314,
358, 390.
(+)-(1R,2R)-trans-1-(4-Chloro-3-methoxyphenyl)-N-
(2,2-dimethoxyethyl)-1,2,3,4-tetrahydro-N-methyl-2-naph-
thaleneamine Oxalate (1:1) 5A: 3-Methoxy-4-chloro-
phenylmagnesium Bromide. Magnesium metal turnings
(11.3 g, 0.47 mol) were charged to a nitrogen flushed flask,
and then anhydrous THF (100 mL) was added. A solution
of 5-bromo-2-chloroanisole (100 g, 0.45 mol) in anhydrous
THF (100 mL) was prepared, and 5 mL of this solution was
added to a vigorously stirred magnesium suspension in THF.
Initiation of the reaction was observed by a temperature
increase to 40-45 °C. The reaction mixture was held in a
water bath, and the remaining solution of 5-bromo-2-
chloroanisole was added at such a rate so as to maintain the
reaction temperature between 40 and 50 °C. After completion
of addition, the reaction mixture was maintained at 40-45
°C for 3 h. HPLC analysis of a MeOH-quenched sample of
the reaction mixture indicated >95% of 2-chloroanisole.
A. p-Toluenesulfonyl Chloride/CuCN Procedure. To
a solution of (+)-(1R,2R)-trans-1,2,3,4-tetrahydro-1-[(2,2-
dimethoxyethyl) methylamino]-2-naphthalenol in diglyme
(100 g, 0.377 mol) was added 1,10-phenanthroline (50 mg)
and toluene (200 mL). The solution was heated to 120 °C at
atmospheric pressure under nitrogen to distill toluene (200
mL). The pale red-brown colored solution was cooled to
room temperature, and a Karl Fisher titration of the sample
indicated a water content of 0.02%. Anhydrous THF (200
mL) was added to the above obtained solution and then
B. Diphenyl Chlorophosphate/CuCl‚2LiCl Procedure.
To a solution of (+)-(1R,2R)-trans-1,2,3,4-tetrahydro-1-[(2,2-
dimethoxyethyl) methylamino]-2-naphthalenol in diglyme
(100 g, 0.377 mol) was added 1,10-phenanthroline (50 mg)
and toluene (200 mL). The solution was heated to 120 °C at
atmospheric pressure under nitrogen to distill toluene (200
mL). The pale red-brown solution was cooled to room
temperature, and a Karl Fisher titration of the sample
indicated a water content of 0.02%. Anhydrous THF (300
n
cooled to -30 °C. Butyllithium (160 mL, 2.5 M solution
in hexanes, 0.40 mol) was added dropwise so as to maintain
the reaction temperature between -15 to -30 °C during the
exothermic addition. Towards the end of addition, the red-
brown colored reaction mixture turnd dark red in color. After
766
•
Vol. 8, No. 5, 2004 / Organic Process Research & Development

mL) was added to the above solution and then cooled to
-30 °C. n-Hexyllithium (160 mL, 2.5 M solution in hexanes,
0.40 mol) was added dropwise so as to maintain the reaction
temperature between -15 to -30 °C during the exothermic
addition. Towards the end of addition, the red-brown reaction
mixture turned dark red in color. After stirring the reaction
mixture at -30 °C for 10 min, diphenyl chlorophosphate
(86 mL, 0.415 mol) was added dropwise. During this
exothermic addition the reaction temperature was maintained
between -20 to -30 °C. The reaction mixture was held at
-20 °C for 2 h. Separately, a solution of CuCl‚2LiCl was
prepared by adding anhydrous THF (50 mL) to a mixture of
copper (I) chloride (1.9 g, 0.019 mol) and lithium chloride
(1.6 g, 0.038 mol). The mixture was stirred for 15 min at
room temperature to give a golden-yellow solution. This
solution was cannulated to the Grignard solution held at 25
°C. The mixture was stirred for 5 min, and then the
aziridinium ion solution held at -20 °C was added to the
Grignard solution via a cannula. The addition took about 45
min, and during the addition the temperature of the Grignard
solution was in the range of 30-35 °C. After the addition,
the reaction mixture was heated to 40-45 °C and held at
that temperature for 1.5 h. The reaction mixture was cooled
to 0 °C and quenched with 7% aqueous ammonium chloride
solution (500 mL). The reaction mixture was warmed to
room temperature and stirred for 30 min. The organic layer
was separated and stirred with 7% aqueous ammonium
chloride solution (500 mL) for 15 min, and the aqueous layer
was separated. The combined aqueous layers were extracted
with MTBE (200 mL). The combined organic layers were
stirred with 5% NaOH (200 mL), and the organic layer was
separated and then stirred with 5% brine solution. The
organic layer was separated and then distilled atmospherically
to a volume of 200 mL. MTBE (300 mL) was added to the
dark-brown viscous solution and filtered through a sintered
glass funnel. The funnel was rinsed with 2-propanol (100
mL), and the combined filtrate was warmed to 40 °C.
Separately, a solution of oxalic acid dihydrate (47 g, 0.37
mol) in 2-propanol (200 mL) was prepared by warming to
45 °C and added to the free base solution. The mixture was
stirred and gradually cooled to room temperature. After 12-
14 h the product crystallized out, and the suspension was
cooled to 0 °C. The suspension was held at 0 °C for 4 h and
then filtered. The cake was washed with ice-cold 2-propanol
(200 mL), then with ice-cold MTBE (200 mL), and finally
again with ice-cold MTBE (100 mL). The solid was dried
under vacuum for 4 h and then in a vacuum oven at 45 °C
for 12 h to give an off-white solid (125-135 g, 69-75%
molar yield). FABMS: 235, 271, 314, 358, 390.
mL). A slurry of tert-butylamine borane complex (16 g, 0.813
mol) in MTBE (80 mL) was prepared and added slowly to
the reaction mixture at such a rate that the temperature was
maintained between 10 and 15 °C. The reaction mixture was
cooled to 0 °C, and water (320 mL) was added slowly,
followed by 25% potassium hydroxide (640 mL) solution.
The organic layer was separated, and the basic aqueous layer
was extracted with MTBE (200 mL). The combined MTBE
extracts were washed with water (160 mL), and the organic
layer was distilled atmospherically under a nitrogen atmo-
sphere to a volume of 200 mL. 2-Propanol (320 mL) was
charged, and the batch was concentrated to 300 mL.
Separately, a solution of malonic acid (23 g, 0.22 mol) in
2-propanol (200 mL) was prepared, warmed to 45 °C, and
added to the stirred free base solution. In a few minutes the
product crashed out of solution, and the slurry was cooled
gradually to room temperature and then held at 0 °C for 1
h. The slurry was filtered, and the cake was washed with
ice-cold 2-propanol (120 mL) and then with ice-cold MTBE
(160 mL). The cake was dried under vacuum and then in a
vacuum oven at 50 °C for 12 h to give an off white solid
(83 g, 0.192 mol, 92% molar yield). FABMS: 331, 330,
329, 328, 327, 289, 232, 181. Anal. Calcd for C23H25-
ClO5N: C, 63.96; H, 6.07%; N, 3.24. Found: C, 64.30; H,
6.04; N, 3.38.
trans-(-)-(6aS,13bR)-11-Chloro-6,6a,7,8,9,13b-hexahy-
dro-7-methyl-5H-benzo[d]naphth[2,1-b]azepine-12-ol, Hy-
drochloride (1:1): 1‚HCl. To 6A (100 g, 0.232 mol) was
added toluene (300 mL) and water (300 mL) followed by
25% NaOH (80 mL) solution. The suspension was stirred at
40-50 °C for 30 min to give a biphasic solution which was
cooled to room temperature. The toluene layer was separated,
and the remaining aqueous layer was extracted with toluene
(200 mL). The combined toluene layers were washed with
water (50 mL), and the water layer was discarded. The pale-
yellow organic layer was concentrated atmospherically to a
volume of 270 mL. The water content of the free base in
toluene was found to be <0.1%. Boron trichloride (68 g,
0.58 mol) was charged to the stirred free base solution at
room temperature, and then the mixture was heated to about
70 °C for 18 h. The reaction mixture was dropped carefully
into stirred methanol (180 mL) held in a 25 °C water bath.
After the exothermic quench, the reaction mixture was
concentrated to about 180 mL. The suspension was cooled
to room temperature, MTBE (90 mL) was added, and the
mixture was cooled to 0-5 °C for 1 h. The product was
filtered and washed with an ice-cold MeOH-MTBE (1:2,
75 mL) mixture and then with ice-cold MTBE (50 mL). The
product was dried in a vacuum oven at 50-60 °C for 12 h
to afford a colorless to pale-yellow solid (75 g, 0.214 mol,
92% molar yield).
trans-(-)-(6aS,13bR)-11-Chloro-6,6a,7,8,9,13b-hexahy-
dro-12-methoxy-7-methyl-5H-benzo[d]naphth[2,1-b]aze-
pine, Malonate (1:1) 6A. 5A (100 g, 0.208 mol) was charged
in portions to stirred methanesulfonic acid (160 mL, 2.47
mol) at room-temperature maintaining a 45-50 °C reaction
temperature. After the exothermic addition the suspension
was heated to 55-60 °C for 2 h. The dark red-brown reaction
mixture was cooled to 0 °C and diluted with MTBE (160
The product was recrystallized as follows: 100 g of 1.HCl
was heated to reflux in MeOH (1.2 L), and then the solution
was cooled to room temperature and filtered through a 1-µ
filter. The solution was concentrated by atmospheric distil-
lation to 200 mL and then cooled to room temperature.
MTBE (300 mL) was added to the suspension and cooled
Vol. 8, No. 5, 2004 / Organic Process Research & Development
•
767

to 0-5 °C for 1 h. The product was filtered and washed
with an ice-cold MeOH-MTBE (1:2, 120 mL) mixture and
then with ice-cold MTBE (200 mL). The product was dried
in a vacuum oven at 50-60 °C for 12 h to give a colorless
solid (87 g, 87% recovery).
lowed by cooling to -20 to -30 °C led to 95% recovery
after washing the cake with only ice-cold MTBE (2 × 200
mL).
Received for review January 16, 2004.
OP0402026
Alternatively, addition of MTBE (700 mL) to the con-
centrated solution obtained after distillation (200 mL) fol-
768
•
Vol. 8, No. 5, 2004 / Organic Process Research & Development