The process development of ravuconazole: An efficient multikilogram scale preparation of an antifungal agent

Article abstract of DOI:10.1021/op900065c

The development of a safe, robust process for the preparation of ravuconazole (1), an antifungal agent, is described. The discovery and development of procedures enabling the efficient synthesis of multikilogram quantities of 1 and the process demonstration through plant scale preparations are presented. A controlled means to prepare a Grignard reagent and utilization of Fourier Transform Infrared spectroscopy (FTIR) monitoring to safely conduct the reaction is featured.

Full text of DOI:10.1021/op900065c

Organic Process Research & Development 2009, 13, 716–728
The Process Development of Ravuconazole: An Efficient Multikilogram Scale
Preparation of an Antifungal Agent1
Jaan Pesti,* Chien-Kuang Chen,* Lori Spangler, Albert J. DelMonte, Serge Benoit, Derek Berglund, Jeffrey Bien,
Paul Brodfuehrer, Yeung Chan, Elisabeth Corbett, Carrie Costello, Paul DeMena, Robert P. Discordia, Wendel Doubleday,
Zhinong Gao, Stephane Gingras, John Grosso, Oscar Haas, David Kacsur, Chiajen Lai, Simon Leung, Melanie Miller,
Jale Muslehiddinoglu, Nina Nguyen, Jun Qiu, Martina Olzog, Emily Reiff, Dominique Thoraval, Michael Totleben,
Dale Vanyo, Purushotham Vemishetti, John Wasylak, and Chenkou Wei
Process Research and DeVelopment, Bristol-Myers Squibb Pharmaceutical Co., One Squibb DriVe, P.O. Box 191,
New Brunswick, New Jersey, 08903-0191, U.S.A.
Abstract:
The development of a safe, robust process for the preparation of
ravuconazole (1), an antifungal agent, is described. The discovery
and development of procedures enabling the efficient synthesis of
multikilogram quantities of 1 and the process demonstration
through plant scale preparations are presented. A controlled means
to prepare a Grignard reagent and utilization of Fourier Trans-
form Infrared spectroscopy (FTIR) monitoring to safely conduct
In addition to communicating an informative description of
the scale-up of 1 from the laboratory to the plant, intermediates
8, 10, and 14 are precursors of other antifungal agents.3,4,6-13
This description of their large-scale preparations should facilitate
future syntheses of these antimycotic agents and their related
analogues.14 We also describe the use of FTIR spectroscopy
for the control of the initiation, preparation, and reaction of a
reactive Grignard reagent at pilot-plant scale.15
the reaction is featured.
Introduction
Ravuconazole (1) is a potent and orally active broad
spectrum antifungal agent formerly in development at Bristol-
Myers Squibb.2 This novel thiazole-containing antifungal has
the longest half-life of all of the known azole antifungals, and
its activity against Cryptococcus neoformans, Candida albicans,
and Aspergillus species is particularly well suited to combat
opportunistic fungal infections in immunocompromised patients.
The continued emergence of resistance against existing anti-
mycotic drugs requires that our industry continue to discover
new remedies for these diseases.
This report describes the development of a process for the
large-scale preparation of 1 from commercially available
methyl-(R)-lactate, bromo-difluorobenzene, triazole, and 2-bromo-
4′-cyanoacetophenone in a safe, robust, and cost-effective
manner. The evolution of the process encompasses three
campaigns of increasing sophistication and scale. Although there
are several citations regarding the gram-scale preparations of 1
and/or its precursors,3-5 to our knowledge there are no published
accounts of the large-scale preparation of ravuconazole or its
intermediates. Our work presents one method for the multiki-
logram preparation of 1.
(3) Naito, T.; Hata, K.; Kaku, Y.; Tsuruoka, A.; Tsukada, I.; Yanagisawa,
M.; Toyosawa, T.; Nara, K. U.S. Patent 5789429, Antifungal agents,
processes for the preparation thereof, and intermediates, Aug 4, 1998;
Chem. Abstr. 1995, 123, 340144.
(4) Tsuruoka, A.; Kaku, Y.; Kakinuma, H.; Tsukada, I.; Yanagisawa, M.;
Nara, K.; Naito, T. Chem. Pharm. Bull. 1998, 46, 623–630.
(5) (a) Soukup, M. Patent WO 03/002498 A1, Intermediate halophenyl
deriVatiVes and their use in a process for preparing azole deriVatiVes,
Jan 9, 2003; Chem. Abstr. 2003, 138, 89814. (b) Xu, L.; Muller, M. R.;
Yu, X.; Zhu, B.-Q. Synth. Commun. 2009, 39, 1611–1625.
(6) Saji, I.; Tamoto, K.; Tanaka, Y.; Miyauchi, H.; Fujimoto, K.; Ohashi,
N. Bull. Chem. Soc. Jpn. 1994, 67, 1427–1433.
(7) Tasaka, A.; Tamura, N.; Matsushita, Y.; Teranishi, K.; Hayashi, R.;
Okonogi, K.; Itoh, K. Chem. Pharm. Bull. 1993, 41, 1035–1042.
(8) Konosu, T.; Tajima, Y.; Takeda, N.; Miyaoka, T.; Kasahara, M.;
Yasuda, H.; Oida, S. Chem. Pharm. Bull. 1990, 38, 2476–2486.
(9) (a) Tasaka, A.; Tamura, N.; Matsushita, Y.; Kitazaki, T.; Hayashi,
R.; Okonogi, K.; Itoh, K. Chem. Pharm. Bull. 1995, 43, 432–440. (b)
Tasaka, A.; Tsuchimori, N.; Kitazaki, T.; Hiroe, K.; Hayashi, R.;
Okonogi, K.; Itoh, K. Chem. Pharm. Bull. 1995, 43, 441–449. (c) Oida,
S.; Tajima, Y.; Konosu, T.; Nakamura, Y.; Somada, A.; Tanaka, T.;
Habuki, S.; Harasaki, T.; Kamai, Y.; Fukuoka, T.; Ohya, S.; Yasuda,
H. Chem. Pharm. Bull. 2000, 48, 694–707. (d) Kitazaki, T.; Ichikawa,
T.; Tasaka, A.; Hosono, H.; Matsushita, Y.; Hayashi, R.; Okonogi,
K.; Itoh, K. Chem. Pharm. Bull. 2000, 48, 1935–1946.
* To whom correspondence should be addressed. E-mail: jaan.pesti@
bms.com.
(10) Kitazaki, T.; Tamura, N.; Tasaka, A.; Matsushita, Y.; Hayashi, R.;
Okonogi, K.; Itoh, K. Chem. Pharm. Bull. 1996, 44, 314–327.
(11) Tsuruoka, A.; Kaku, Y.; Kakinuma, H.; Tsukada, I.; Yanagisawa, M.;
Naito, T. Chem. Pharm. Bull. 1997, 45, 1169–1176.
(1) Aspects of this work were also carried out at Bristol-Myers Squibb’s
facilities in Princeton, NJ; Candiac, Canada; and Syracuse, NY.
(2) Ravuconazole was licensed from Eisai Co. Ltd. References to earlier
studies include: (a) Hata, K.; Kimura, J.; Mike, H.; Toyosawa, T.;
Nakamura, T.; Katsu, K. Antimicrob. Agents Chemother. 1996, 40,
2237–2242. (b) Arikan, S.; Rex, J. H. Curr. Opin. InVest. Drugs 2002,
3, 555–561. (c) Fung-Tomc, J. C.; Huczko, E.; Minassian, B.; Bonner,
D. Antimicrob. Agents Chemother. 1998, 42, 313–318, and references
within. (d) Naito, T.; Hata, K.; Tsuruoka, A. Drugs Future 1996, 21,
20. (e) Hoffman, H. L.; Ernst, E. J.; Klepser, M. E. Expert Opin. InVest.
Drugs 2000, 9, 593–605.
(12) Kawanishi, H.; Morimoto, H.; Nakano, T.; Miyajima, T.; Oda, K.;
Takeda, K.; Yano, S.; Hirano, N.; Tsujihara, K. Heterocycles 1998,
49, 169–180.
(13) Bartroli, J.; Turmo, E.; Alguero´, M.; Boncompte, E.; Vericat, M. L.;
Conte, L.; Ramis, J.; Merlos, M.; Garc´ıa-Rafanell, J.; Forn, J. J. Med.
Chem. 1998, 41, 1855–1868.
(14) Chen, A.; Sobel, J. D. Expert Opin. Emerging Drugs 2005, 10 (1),
21–33.
716
•
Vol. 13, No. 4, 2009 / Organic Process Research & Development
10.1021/op900065c CCC: $40.75  2009 American Chemical Society
Published on Web 07/17/2009

Scheme 1. Process R&D synthesis of ravuconazole
Results and Discussion
cyanohydrin). The early goals of the developmental program
were to circumvent the most undesirable operations and
reagents, develop means to safely conduct others, and consider
alternative protecting groups or salt forms to facilitate process-
ing. The path taken to address these issues is described below.
Early Process Development Leading to the First Ravu-
conazole Campaign. Preparation of the Arylpropanone 7. At
the initiation of process development, the immediate goal was
the rapid production of 1-2 kg of 1 for toxicology studies and
a phase I clinical trial. The focus of the development team was
to quickly enable the current synthesis to satisfy API material
demands.
The morpholine amide 4, or a close analogue, was targeted
as the first isolated intermediate. Flexibility in the choice of
the intermediate existed, as neither the amide functionality nor
the alcohol protecting group would be retained in the final
product. The amide functionality would only need to serve as
a good electrophile for the Grignard coupling to prepare 7.
Methyl (R)-lactate (2) was inexpensive ($12/kg) and, thus, was
a clear choice as starting material. Early experiments focused
on finding a crystalline analogue of the oil 4 by varying the
alcohol’s protecting group. We believed that the diastereomeric
nature of 4 led to the noncrystallinity18 and substitution of a
nonchiral protecting group would remedy this. This expectation
was realized as trityl, adamantane carboxylic ester, and tert-
butyldiphenylsilane provided crystalline intermediates; however
The optimized multikilogram route for the preparation of 1
is shown in Scheme 1. The existing preparative route was first
evaluated to provide a basis for a synthetic development plan.16
As such, the sequence of intermediates in Scheme 1 was
determined to be an expeditious starting point for small scale
preparative work and process development. While this estab-
lished chemistry satisfied API (active pharmaceutical ingredi-
ents) material demands for early development activities, alternate
approaches to 1 were explored to select the most advantageous
and scalable process. While the tertiary alcohol of 1 allows
installation of the three branches of the nascent drug in differing
sequential order,17 all of the alternative routes to 1 investigated
proved inferior to the established approach. Consequently,
instead of focusing on these unsatisfactory synthetic approaches,
this report will discuss the process development of the route
described in Scheme 1.
Upon evaluation of the initial route, some elements required
for large-scale preparations were already in place, including
readily available starting materials and reagents. Some of the
challenges included the following: recurring column chroma-
tography, lack of crystalline intermediates, the preparation of a
reactive Grignard reagent, malodorous reagents and side
products (diethyl dithiophosphate and hydrogen sulfide), and
the use of dangerous reagents and solvents (lithium hydride,
sodium hydride, diisopropyl ether, diethyl ether, and acetone
(15) For other examples, see: (a) am Ende, D. J.; Clifford, P. J.; DeAntonis,
D. M.; SantaMaria, C.; Brenek, S. J. Org. Process Res. DeV. 1999, 3,
319–329. (b) Stares, K. E.; am Ende, D. Pilot Plants and Scale-up of
Chemical Processes II; The safe scale-up of a Grignard reaction using
an in-situ reaction monitoring technique; Hoyle, W., Ed.; Special
Publication - Royal Society of Chemistry: 1999; pp 62-80; Chem.
Abstr. 130, 313820. (c) Wiss, J.; Lanzlinger, M.; Wermuth, M. Org.
Process Res. DeV. 2005, 9, 365–371.
(17) For example, flutriafol has been prepared by different sequences of
appending the various pieces: Worthington, P. A. Recent DeVelopments
in the Chemistry of Azole Fungicides; Proceedings of 1984 British
Crop Protection Conference, Pests and Diseases; BCPC Publications:
Croydon, U.K., 1984; Vol. 3, pp 955-962.
(18) As support for this hypothesis, note that all of the THP-protected
intermediates in a penem synthesis are noncrystalline while the
remaining molecules are all crystalline. The authors thank a referee
for this reference: Gala, D.; Steinman, M.; Jaret, R. S. J. Org. Chem.
1986, 51, 4488–4490.
(16) Privileged technology transfer documentation from Eisai Co., Ltd.;
Sept 26, 1996.
Vol. 13, No. 4, 2009 / Organic Process Research & Development
•
717

they were poor substrates for the subsequent Grignard reaction.
The triethylsilane, benzyl, 4-bromobenzyl, 4-nitrobenzyl, MOM,
and MOP protecting analogues were not crystalline and were
also inferior to 4 in regards to yield. The diastereomeric mixture
of morpholine amide 4, in spite of its lack of crystallinity, was
retained for further development.
The initial preparations of 3 were adapted without notable
deviation from the published process and consisted of the
reaction of excess morpholine with methyl lactate at 85 °C for
30-60 h.7 Workup and concentration yielded the amide 3 in
79-82% yields at up to 750 g scale. The product contained up
to 17 wt % of morpholine by weight, which tended to be
difficult to remove by further distillation. However its presence
was inconsequential for further processing. The subsequent use
of excess dihydropyran with catalytic toluenesulfonic acid or
methanesulfonic acid in methylene chloride produced a dia-
stereomeric mixture of 4 in 92-100% unpurified yields at up
to 1140 g scale.
Several approaches to the preparation of intermediate 7 (or
closely related structures) are known, including the Friedel-
Crafts acylation of 1,3-difluorobenzene,19,20 enzymatic resolution
of a racemic acetate ester,21,22 the opening of a chiral aryl
epoxide,23 lyase-catalyzed carboligation of difluorobenzalde-
hyde,24 and R-hydroxylation using camphorsulfonyl oxaziri-
dine,25 but the organomagnesium route to 7 illustrated in Scheme
1 had been most thoroughly established.26 It proceeded from
the preparation of the Grignard reagent of bromo-2,4-difluo-
robenzene (6) and its subsequent coupling with 4.7,27,28 In our
laboratories, the formation of the Grignard reagent initiated
reproducibly, and the generation of 6 only required 1 h at 40-45
°C. Subsequently, the solution was cooled to -20 °C and 4
was added. A workup hydrolyzed any remaining Grignard
reagent and converted the tetrahedral Grignard intermediate29
to the ketone 7 in >95% crude yield as a light brown oil. The
oil could be purified by distillation, but this led to significant
losses (62-67%) and was not retained. This procedure per-
formed adequately in the laboratory at up to 870 g scale;
however significant changes in operational procedure were
required before multikilogram scale could be considered.
Preparation of the Diol 10. The second phase of the
synthesis of 1 required the introduction of the triazole with
concomitant formation of the tertiary chiral alcohol and
subsequent deprotection of the secondary alcohol. This strategy
utilized epoxidation of 73,4,7,19,21,30 as a convenient means to
introduce the methylene fragment of 1 and to provide a substrate
for the nucleophilic attack by the triazole ring. This sequence
also sets the second chiral center, effectively originating all
stereogenic centers from methyl (R)-lactate.
The epoxide 8 was formed from ketone 7 via treatment with
the Corey-Chaykovsky reagent.31 Trimethylsulfoxonium iodide
was added in portions to a suspension of sodium hydride (NaH)
in DMSO and THF to form the sulfur ylide, followed by the
addition of the substrate 7.7 The adjacent chiral center induced
an ∼4:1 stereoselectivity for the formation of the oxirane. The
diastereomeric ratio was improved later by crystallization in
the downstream chemistry. Upon reaction completion, the
mixture was quenched into water and extracted with ethyl
acetate. The triazole anion formed by reaction with NaH in
DMF was introduced, opening the epoxide to yield the
monoprotected diol 9. After the THP group was cleaved by
pyridinium p-toluenesulfonate, the diol 10 was obtained as a
orange solid.7,8,19b,21,23,27 Recrystallization from diethyl ether
reduced impurities and increased the de to >99%.7 The purging
of impurities and diastereomers which had accumulated over
six telescoped steps largely accounted for the low overall yields
(38-41%). Following this protocol, five batches totaling 1125 g
of 10 were completed in a rapid and reproducible manner. This
completed the construction of the western half of ravuconazole
and left only introduction of the cyanobenzene substituted
thiazole side chain to complete the synthesis.
Preparation of RaVuconazole. Appending the side chain
required sequential epoxide formation, ring opening with
cyanide, conversion to the thioamide, and ultimately condensa-
tion with a bromo-cyano-acetophenone to form the thiazole ring.
This sequence commenced with formation of the mesylate 16,
itself the final intermediate to another antifungal: SM 8668/
Sch 42427.19b As an alternative to isolation, the mesylation
reaction mixture could be immediately treated with sodium
hydroxide to form the epoxide 12 in good yield and purity.
This synthetic sequence was demonstrated to be superior to
previous routes to 12, which required chromatographic
purification.6,7,10,19a,32
(19) (a) Konosu, T.; Miyaoka, T.; Yaima, Y.; Oida, S. Chem. Pharm. Bull.
1991, 39, 2241–2246. (b) Girijavallabhan, V. M.; Ganguly, A. K.;
Pinto, P. A.; Sarre, O. Z. Bioorg. Med. Chem. Lett. 1991, 1, 349–
352.
(20) Konosu, T.; Tajima, Y.; Miyaoka, T.; Oida, S. Tetrahedron Lett. 1991,
32, 7545–7548.
(21) Gala, D.; DiBenedetto, D. J.; Clark, J. E.; Murphy, B. L.; Schumacher,
D. P.; Steinman, M. Tetrahedron Lett. 1996, 37, 611–614.
(22) Demir, A. S.; Hamamci, H.; Sesenoglu, O.; Aydogan, F.; Capanoglu,
D.; Neslihanoglu, R. Tetrahedron: Asymmetry 2001, 12, 1953–1956.
(23) Gala, D.; DiBenedetto, D. J. Tetrahedron Lett. 1994, 35, 8299–8302.
¨
(24) Demir, A. S.; S¸es¸enoglu, O.; Eren, E.; Hosrik, B.; Pohl, M.; Janzen,
E.; Kolter, D.; Feldmann, R.; Du¨nkelmann, P.; Mu¨ller, M. AdV. Synth.
Catal. 2002, 344, 96–103.
This reaction sequence continued with epoxide ring opening
with a choice between lithium cyanide (LiCN)33,34 or diethy-
(25) Gala, D.; DiBenedetto, D. J.; Mergelsberg, I.; Kugelman, M. Tetra-
hedron Lett. 1996, 37, 8117–8120.
(26) Also, the trityl analogue of 7 has been made by Grignard addition to
2-pyridylthioester; see ref 30.
(27) Singh, I. P.; Sidhu, I.; Palak, B.; Micetich, R. G. U.S. Patent 6133485,
Asymmetric synthesis of 2-aryl-1-substituted butane-2,3-diols, Oct 17,
2000; Chem. Abstr. 1999, 131, 286254.
(28) The analogous reaction but using benzyl as a protecting group: Crosby,
J.; Bailey, K. D.; Monteith, M. J. U.S. Patent 6362376; Process for
the preparation of 2-hydroxyalkyl halophenones; March 26, 2002;
Chem. Abstr. 1999, 131, 310447.
(29) (a) Peters, R.; Waldmeier, P.; Joncour, A. Org. Process Res. DeV.
2005, 9, 508–512. (b) Olah, G. A.; Surya Prakash, G. K.; Arvanaghi,
M. Synthesis 1984, 228–230. (c) Katritzky, A. R.; Le, K. N. B.;
Khelashvili, L.; Mohapatra, P. P. J. Org. Chem. 2006, 71, 9861–9864.
(d) Larcheveque, M.; Petit, Y. Synthesis 1986, 60–64.
(30) Kaku, Y.; Tsuruoka, A.; Kakinuma, H.; Tsukada, I.; Yanagisawa, M.;
Naito, T. Chem. Pharm. Bull. 1998, 46, 1125–1129.
(31) Corey, E. J.; Chaykovsky, M. J. Am. Chem. Soc. 1965, 87, 1353–
1364.
(32) For a different route to 12 via a Payne rearrangement, see: Konosu,
T.; Miyaoka, T.; Tajima, Y.; Oida, S. Chem. Pharm. Bull. 1992, 40,
562–564.
(33) Tsuruoka, A.; Negi, S.; Yanagisawa, M.; Nara, K.; Naito, T.; Minami,
N. Synth. Commun. 1997, 27, 3547–3557.
(34) Naito, T.; Hata, K.; Kaku, Y.; Tsuruoka, A.; Tsukada, I.; Yanagisawa,
M.; Toyosawa, T.; Nara, K. U.S. Patent 5648372, Antifungal Agents,
and Compositions, July 15, 1997; Chem. Abstr. 1995, 123, 340144.
718
•
Vol. 13, No. 4, 2009 / Organic Process Research & Development

campaign. The new goal was to translate this 5-12 L reactor
experience into a pilot plant campaign to satisfy larger require-
ments for additional clinical studies and pharmaceutical
development.
ImproVement of the Preparation of Amide 3. The kilolab
preparation of 3 utilized the reaction of excess morpholine (3
equiv) with methyl lactate at 85 °C for 30-60 h to produce
∼90% yields of product which still contained morpholine.7 A
purer product and a reduction of the morpholine charge would
be desirable improvements to the process. The addition of 15
mol % of sodium methoxide catalyst37 permitted the morpholine
charge to be reduced nearly 3-fold to 1.03 equiv. Under these
modified conditions, the reaction proceeded to completion at
ambient temperature in only 4-6 h instead of 30-60 h. After
dilution with THF, the bases were directly sequestered by a
slurry of Amberlyst-15 sulfonic acid resin. After the resin was
filtered, the filtrate was concentrated to remove most of the
methanol. The typical yield range was 89-92% of the amide
solution at up to 2 kg scale.
This modified procedure and all other process knowledge
up to diol 10 were transferred to a vendor who used this
information to prepare 30 kg of diol 10. By contracting with a
vendor the project team was able to focus its resources
exclusively on the more challenging chemistry involved in
handling cyanide, forming the thiazole ring, and controlling the
purity and particle size of the API.
Preparation of Thioamide. During the early phases of
development for the conversion of the diol 10 into epoxide 12,
either THF or MTBE was used as solvent for the mesylation
and epoxide ring formation. Both solvents were effective, but
MTBE systems produced thick slurries and THF systems
required repeated extractions during the aqueous workup to
achieve high recoveries. Mixing the two solvents 1:1 retained
the advantages of each without incorporating the disadvantages.
The product 12 could be precipitated from solution upon
addition of hydrocarbon antisolvents; however the yield never
exceeded 75%. A higher yielding approach used an aqueous
ethanol crystallization. The optimized recovery of 12 consisted
of a solvent exchange of the worked-up reaction mixture to
ethanol followed by the addition of three volumes of water.
This process was transferred into the plant with no further
modifications and produced 78-86% yields of 12 in 98 HPLC
area % (99.8-99.9% ee) in batches ranging from 55 to 71 kg.
For the purposes of preparing 13 during the first campaign,
lithium cyanide (LiCN) was generated in situ from lithium
hydride (LiH) and acetone cyanohydrin.38 Unfortunately, so-
dium or potassium cyanide, whether in organic solvents (slow
reaction rates) or with phase transfer catalysis in biphasic
systems (emulsions during workup), did not perform effectively
when substituted for LiCN. However, lithium hexamethyldisi-
lazane (LiHMDS) may be purchased in solution, and this more
convenient and safer base was substituted for LiH. A minimum
laluminum cyanide.3,4,34,35 In consideration of a future pilot plant
campaign, the handling of LiCN was considered a less hazard-
ous operation of the two;33 however it would still require in
situ generation, as LiCN was not commercially available. The
optimized reaction conditions required the reaction of 3 equiv
of LiH with a slight excess of acetone cyanohydrin. After the
in situ generation of LiCN, the epoxide was charged and the
reaction was held at 65 °C until the generation of cyano alcohol
13 was complete (∼17 h). White crystals of high purity 13 (>99
HPLC area %) in 87-89% yield were obtained after water
quench, extraction, and recrystallization from isopropanol. While
this procedure provided pure material in good yield, it was
apparent that significant development would be required for safe
production on multikilogram scale.
The conversion of a nitrile group into a thioamide appeared
to be straightforward,36 but the majority of literature procedures
necessitated the use of hydrogen sulfide (H2S). The preparative
synthesis of 14 required treatment of 13 in refluxing aqueous
isopropanol with diethyl dithiophosphate, followed by a workup
and finally a recrystallization from diisopropyl ether.4 A similar
preparation had also been reported in which water was the only
solvent.3 Our first preparations of 14 were made in aqueous
isopropanol, and two lots totaling 570 g were produced in >99%
HPLC area % purity in 76-77% yield. This process was
adequate for the production of moderate amounts of material,
in spite of the inherent safety limitations to this route.
Reflux of the thioamide 14 with 2-bromo-4′-cyanoacetophe-
none in ethanol rapidly formed the thiazole ring to produce
ravuconazole 1.3,4 Workup involved the addition of saturated
sodium bicarbonate solution followed by collection of crystalline
1 by filtration. To achieve high API purity and the desired
polymorph, recrystallizaion of 1 from aqueous ethanol was
required. Yields of 78-81% were obtained over four batches,
the largest producing 440 g (99 HPLC area %).
This completed the first ravuconazole campaign. The ma-
terial satisfied clinical requirements and allowed development
of analytical methods for future preparations. Perhaps most
importantly, the first campaign provided familiarity with the
chemistry and identified critical areas which required improve-
ment for further scale-up. Subsequent work would address
concerns related to safety, cost, equipment limitations, and
elimination of impractical operations (i.e., stripping to dryness)
with the objective of transforming an expedient laboratory
process into a practical one for plant operation.
Second Campaign, First Pilot Plant Experience. The first
campaign made extensive use of kilolab facilities, which would
not suffice for the 10-100 kg batches anticipated in the next
(37) (a) Baltzly, R. B.; Berger, I. M.; Rothstein, A. A. J. Am. Chem. Soc.
1950, 72, 4149–4152. (b) For a similar but noncatalytic ester to
morpholine conversion: Al-Masoudi, N. A.; Al-Soud, Y. A. Nucleo-
sides, Nucleotides Nucleic Acids 2002, 21 (4-5), 361–375.
(38) (a) Livinghouse, T. Org. Syntheses, Coll. Vol. 7, 1990, 517-522; Vol.
60, 1981, 126-131. (b) For a recent and convenient means to prepare
LiCN, see: Ciaccio, J. A.; Smrtka, M.; Maio, W. A.; Rucando, D.
Tetrahedron Lett. 2004, 45, 7201–7204.
(35) The cyano-alcohol 13 may also be prepared Via treatment of the
corresponding aldehyde with hydroxylamine-O-sulfonic acid.3,30
(36) Walter, W.; Bode, K.-D. Angew. Chem., Int. Ed. Engl. 1966, 5 (5),
447–461.
Vol. 13, No. 4, 2009 / Organic Process Research & Development
•
719

of 2 equiv of base was required to avoid formation of
unsaturated nitrile impurities.
(EtO)2POSH, as no gas was produced if the reaction was
conducted in neat isopropanol. Unfortunately, isopropanol was
not a viable reaction solvent, since the rate of reaction became
much slower and the formation of side products interfered with
the subsequent crystallization of 14. To address the H2S
associated safety issues, its rate of formation was measured to
better understand and control this reaction. It was determined
that if diethyl dithiophosphate was added to an aqueous
isopropanol slurry of 13 preheated to 80 °C, as compared to
heating the reaction mixture subsequent to the addition of diethyl
dithiophosphate, the maximum rate of H2S off-gassing dropped
by ∼70%. The rate further decreased as the reaction proceeded
to completion.39 Both reactions had similar profiles and required
7-8 h to reach completion. One additional advantage to
maintaining a high addition temperature was that the production
of H2S was addition controlled. By modulating the diethyl
dithiophosphate charge, a greater level of control over potential
spikes in off-gassing was instituted. This offered a safer
operation with the same overall reaction time as the existing
process.
Upon reaction completion, the cooled mixture was diluted
with water and ethyl acetate before it was rendered basic. If
the organic phase was concentrated to an oil and triturated with
diisopropyl ether, crystalline free base gradually formed. As
diisopropyl ether could not be used on scale and no other useful
solvent combinations were discovered, salts of 14 were con-
sidered for the purpose of ease of isolation. Sulfuric acid cleanly
formed either the sulfate or hydrogen sulfate salt, depending
upon the number of equivalents charged. The partial hydrate
of the latter salt (15) was selected for scale-up.
Once 1 equiv of sulfuric acid had been charged, workup
proceeded by concentration of the slurry by azeotropic distil-
lation. Isolation was concluded by the addition of an antisolvent,
MTBE, and cooling to 5 °C to crystallize 15. While the initially
isolated crystal was a hydrate, the drying conditions ultimately
determined the final water content. Water content typically was
reduced to <1 wt % if active drying protocols were present (i.e.,
azeotropic distillation, oven drying); however exposure to
atmospheric moisture for more than 2 h would allow 15 to
recapture sufficient water to form a monohydrate. In general,
the oven-dried solids were analyzed for weight percent and then
shielded from moisture to avoid further hydration. The varying
quantity of water would not be a critical process parameter, as
the next step was conducted in aqueous ethanol.
At multikilogram scale, the lab procedures scaled up as
predicted. A nitrogen sweep above the reactor contents was
adjusted to carry the H2S headspace contents directly to
scrubbers charged with bleach and then to a thermal oxidizer.
No H2S was detected by the monitors in the processing area
except momentarily during sampling operations (15 ppm). The
yields were 87-91% (g99.9 HPLC area %) at up to 58 kg of
15 scale.
The concentration of the reaction mixture was a critical
parameter for reaction rate and completion. Commercially
available 1.65 M LiHMDS solutions required further concentra-
tion to ∼2.5 M to attain reaction completion once all the
components were combined. Following the addition of acetone
cyanohydrin, the epoxide 12 was charged as a 1 g/mL THF
solution and the reaction was complete after 16 h at reflux. The
reaction could be completed in only 5-6 h by further
concentration of this reaction mixture, but the resulting thick
slurry presented problems for efficient scale-up. Once the
reaction was complete, the mixture was cooled to 0 °C and the
pH was adjusted to 9-10 by the subsurface addition of HCl
over 3 h, during which time the product crystallized. We assume
any hydrogen cyanide (HCN) within the reactor headspace thus
had time to redissolve during the slow addition of acid and
subsequently reform the lithium salt. Attempts to directly
precipitate the product by the addition of water led to the
appearance of a second phase. If aqueous isopropanol was
charged instead, the reaction mixture was homogenized but
losses to the mother liquor did not appreciably increase. The
yield was further optimized to 88-92% by removal of THF
by distillation and cooling the mixture to -15 °C for at least
1 h.
While this chemistry functioned well in the laboratory, a
reconsideration and review of safe techniques for the large-
scale handling of cyanide was essential, as extractions, washes,
solvent exchanges, and other operations allowed numerous
opportunities for exposure to cyanide-containing streams. In
addition to personal protective equipment and the design of
operations that sequestered the mixtures within reactors as much
as possible, the HCl charge was preweighed to eliminate the
possibility of an excessive charge that would raise HCN to
dangerous levels. Further control of HCN was accomplished
by monitoring the pH for all reaction streams to maintain
basicity (pH >8) at all times. These safety precautions ensured
the safe execution of this step in the plant. Upon isolation, the
purity of 13 was typically high (>99 HPLC area %). This
process scaled up well, and 44 kg were made in 90% yield and
99.6 HPLC wt %.
This modified workup, consisting of pH adjustment with
HCl, isopropanol addition, solvent exchange, and cooling,
replaced the potentially hazardous extractions, washes, and
drying operations of the original process and reduced the
hazardous waste produced by at least 75%. The revised protocol
also provided safe handling procedures for acetone cyanohydrin
and cyanide-containing reaction streams on a multikilogram
scale.
Examination of alternative methods for transforming nitriles
to thioamides, such as those utilizing sodium sulfide or P10S4,
showed them to be low yielding and impurity-laden for the
preparation of thioamide 14. So while diethyl dithiophosphate
was unpleasant and difficult to handle, the existing chemistry
was clearly the best choice among several unattractive options.36
The use of diethyl dithiophosphate also produced H2S, which
was flammable, toxic, and even more malodorous. H2S origi-
nated from the reaction of water with the side-product
(39) If diethyl dithiophosphate and 100 g of 13 in aqueous isopropanol
were heated to 85 °C over 25 min, off-gassing of H2S reached a peak
level of 550 mL/min (6 L total H2S off-gassing), after which the rate
decreased as the reaction proceeded to completion. If diethyl dithio-
phosphate was added instead at 2.6 g/min over 2 h to an aqueous
isopropanol slurry of 13 heated to 80 °C, the rate of off-gassing was
reduced, peaking at only 180 mL/min after ∼60% had been added.
720
•
Vol. 13, No. 4, 2009 / Organic Process Research & Development

Scheme 2. Competing routes from methyl lactate to 4
The last step was conducted in the same manner as that for
the previous campaign: thioamide 15 and bromocyanoacetophe-
none were heated in 95% ethanol for 2 h to complete the
cyclization. The use of hydroxide and bicarbonate bases for
the removal of the acids in the mixture produced insoluble
hydrogen sulfate salts and contaminated the isolated product.
However, if triethylamine (TEA) was used to adjust the pH to
4, the TEA salts remained soluble and no further purification
was required. The pH adjustment required the solution tem-
perature to remain at >55 °C to maintain 1 in solution. A cooling
ramp to 20 °C allowed 1 to crystallize smoothly to afford a
86-90% yield (>99.5 HPLC area %). In this manner, satisfac-
tory potency, purity, and impurity benchmarks were realized,
but particle size range control derived from either traditional
crystallization approaches or milling failed to attain the require-
ments for formulation: D90 ) e10 µm. This motivated the
development of sonic impinging jet crystallization,40 now an
established technique for small particle formation.41
The ability of two impinging liquid jets of solution to
produce high intensity micromixing and supersaturation leads
to rapid crystallization within the small mixing volume at the
intersection of the jets.42 Since new nuclei are continuously
forming in this area, small particle sizes are achieved. When
solutions of 1 were so treated, the majority of the crystals formed
by the impinging jet process were in the 3-20 µm range, still
above the desired range. This crystallization method was further
refined by placing a sonicator probe in the same plane as the
impinging jets. After experimentation with factors such as
sonication intensity, the angle between the jets, temperature,
and solution concentration, the method was successfully tuned
to produce particles in the 2-5 µm D90 range in 97-98%
recovery. The impinging jet crystallization also improved the
purity of 1. For instance, multihundred gram batches of 99.4
HPLC area % material were raised to 99.7-100.0%.
Effective control of the particle size was demonstrated
reproducibly in the plant. Typically 1 was redissolved into hot
aqueous ethanol. While a single operation was possible by
charging the solution derived from the workup directly into the
sonication apparatus, it was still beneficial from an impurity
perspective to proceed in a stepwise fashion and conduct a
traditional crystallization prior to impinging jet recrystallization.
Ultimately, a total of seven 20-41 kg batches of 1 were made
in 85-91% yields with high quality (2-8 µm particle size,
99.75-99.80 HPLC area %, 99.9% de, and 99.97% ee).
Development of a Commercial Process; The Third
Campaign. Promising clinical results at this stage of pharma-
ceutical development mandated a third campaign with much
higher supply needs. This plant campaign was approached with
the goal of developing a robust and durable43 commercial
process to enable phase III operations and so that multiple
vendors could be utilized for future campaigns.
ImproVed Workup and Isolation of Functionalized Lactate.
Amide 4 had been prepared7,27 from methyl (R)-lactate over
two steps: amide formation and THP protection, with flexibility
in the order of the steps44 (Scheme 2). Both sequences were
re-examined for comparison of process attributes such as yield,
quality (both chemical and chiral), and crystallinity. When the
amide formation was completed first, 3 was converted to 4 by
reaction with dihydropyran Via methanesulfonic acid (MSA)
catalysis in an 80% overall yield. In comparison, the THP ester
17 was the intermediate if the steps were reversed but forming
4 in an identical 80% overall yield. However, the amidation
rate of 17 was slow, and the use of sodium methoxide as an
amidation catalyst led to a slow epimerization of 17. Once the
amidation was complete, the product failed to meet the
acceptance criterion of g98% ee. In comparison, since inter-
mediate 3 was stable (loss of only 1.3% ee on scale compared
to the ee of the methyl lactate charge) under the optimized
reaction conditions, this pathway was selected.
A reconsideration of the acid quench workup eliminated the
use of the ion-exchange resin as a base scavenger. While the
resin had simplified the large-scale processing, it also presented
several disadvantages, including swelling of the resin, a vigorous
exotherm, and difficulty in removing the exhausted resin beads
from the reactor. An improved process resulted when the
reaction was instead quenched with methanolic hydrogen
chloride, generated in situ from acetyl chloride and methanol.
Following this quench, the volatiles were exchanged for THF
by distillation, and the resulting solution of alcohol 3 was
directly converted to 4. The MSA-catalyzed ketalization with
dihydropyran was g95% complete within 12-24 h. Unfortu-
nately 4 was exceedingly water-soluble, and it partitioned into
both wet THF and the aqueous phase during workup. This issue
could be minimized by the use of a wash solution which
contained sufficient sodium bicarbonate to simultaneously
quench the MSA and form salts, leading to an increase in
polarity and pH that allowed an acceptable partition without
requiring a cosolvent. The THF was subsequently exchanged
for MTBE to dry the solution (water content <0.1%), and then
dry THF was recharged for improved reactivity in the subse-
quent coupling to 7. If desired, 70-80% of 4 that remained in
the aqueous phase could be recovered by methylene chloride
extraction. This optimized process was transferred to the plant
and proceeded as expected producing 187 kg of 4 in 78%
overall yield from methyl lactate in 98% ee.
(40) Lindrud, M. D.; Kim, S.; Wei, C. U.S. Patent 6302958; Sonic
impinging jet crystallization apparatus and process, Oct 16, 2001;
Chem. Abstr. 2000, 133, 140214.
(41) Gleeson, M.; Kim, S.; Kientzler, D. Kiang, S. WO Patent 041970 A1,
Process for making sterile aripiprazole of desired mean particle size,
May 12, 2005; Chem. Abstr. 2005, 142, 451843.
(42) Midler, M.; Paul, E. L.; Whittington, E. F.; Futran, M.; Liu, P. D.;
Hsu, J.; Pan, S.-H. U.S. Patent 5314506, Crystallization method to
improVe crystal structure and size, Nov 18, 1991; Chem. Abstr. 1994,
121, 117827.
(44) Mizuno, Y.; Yamane, T.; Ikeuchi, M. WO Patent 98/39305, Crystalline
form of a bis 1,2,4-triazole compound, Sept 11, 1998; Chem. Abstr.
1999, 129, 230729.
(43) Zhang, T. Y. Chem. ReV. 2006, 106, 2583–2595.
Vol. 13, No. 4, 2009 / Organic Process Research & Development
•
721

Scheme 3. Synthesis of 6
These modifications removed the need to isolate two
noncrystalline intermediates, avoided a problematic workup, and
retained the advantages established in the earlier process work
generating a safe, convenient, and scalable process to 4.
Grignard Preparation at Plant Scale Facilitated by Use of
FTIR Monitoring. Although the use of Grignard reagents are a
well-known synthetic methodology,45-47 problems in large-scale
preparation, such as irreproducible or difficult to detect initia-
tions, difficulty in monitoring the rate of its formation,48
limitations in cooling highly exothermic reactions, and the
dangers of handling pyrophoric metals pose limitations. All
these challenges were successfully managed in the plant-scale
preparation of arylpropanone 7.
Laboratory examination established that when the solvent
was dry (<800 ppm water) the formation of the Grignard reagent
could be initiated quickly at 10-15% concentrations of the aryl
bromide. However the temperature window for this reaction
was narrow. If the formation was conducted at >40 °C or if the
resulting solution was stored for extended periods at >0 °C,
the impurity content would become unacceptable with the main
constituent being the tetrafluorobiphenyl 19.49,50 In consideration
of the inherent exothermicity of the metal insertion (380 kJ/
mol Mg), the difficulty in charging the aryl bromide safely,
and the need to utilize the unstable solution of 6 immediately,
we decided to explore alternatives to the classical Grignard
reaction.
distillation had been ineffective as a drying protocol. The
subsequent addition of 10-15% of 5 at 30 °C led to inconsistent
initiation of the metal insertion,53 and the subsequent reactions
to form 7 were typically incomplete and unsatisfactory. This
route was abandoned.
Instead, the traditional two-step process was re-examined.
Once 0.9 equiv of magnesium turnings (relative to 5) were
slurried in THF and the water content was ascertained, the ethyl
Grignard drying protocol described above was conducted. If
10% of the total aryl bromide charge was now added at 30 °C,
the insertion reaction reproducibly initiated in <15 min. Trace
ethyl Grignard remaining from the water quench step may be
responsible for the dependable initiations. The Process Safety
and Evaluation group determined that the aryl bromide level
should not exceed 15% of the total charge at any point of the
addition period to limit the potential adiabatic temperature rise
to 33 °C. This still allowed a reasonable charging period and
fully met our requirements for a safe reaction.
The excess of aryl bromide led to more consistent reactions
and ensured the consumption of all of the metal, minimizing
the risk of residual highly pyrophoric magnesium remaining in
the workup. These reaction conditions led to a reproducible 90%
conversion (>99% theoretical yield) of aryl bromide into
Grignard reagent 6. Unreacted aryl bromide 5 was innocuous
during the subsequent chemistry and was removed during the
solvent exchange to DMF.
A promising approach appeared to be the infrequently
utilized Barbier modification, in which the Grignard reagent 6
is prepared in the presence of the electrophile 4.51 This would
effectively circumvent the stability issues of 6 as its existence
would be transitory. Magnesium metal and 4 were mixed, and
the solution was dried via a charge of 1.05 equiv (relative to
residual water) of ethyl magnesium bromide.52 Azeotropic
Real-time monitoring54 of the reactions (initiation, reaction
of 6, and formation of the coupled intermediate 18) comprising
this step would be critical to process safety. Sampling and
external analysis is quantitative but can be slow. We were
pleased to find that the aryl bromide 5, Grignard reagent 6,
starting material 4, and the unquenched intermediate 18 could
all be quantified by FTIR,15,55 allowing real time data collection
and process control for initiation of Grignard formation,
Grignard formation rate, and the coupling reaction with the
amide. In particular, initiation would be difficult to ascertain in
(45) Grignard Reagents, New DeVelopments; Richey, H. G., Jr., Ed.; J.
Wiley & Sons: Chichester, 2000.
(46) Handbook of Grignard Reagents; Silverman, G. S., Ed.; Marcel
Dekker: New York, 1996.
(47) Wakefield, B. J. Organomagnesium Methods in Organic Synthesis;
Academic Press: London, 1995.
(51) (a) Blomberg, C.; Hartog, F. A. Synthesis 1977, 18–30, and ref 32
therein for an example of an amide as electrophile. (b) March, J.
AdVanced Organic Chemistry, 4th ed.; John Wiley & Sons: New York,
1992; pp 921-922. (c) A Pfizer group has employed a similar Barbier
modification to prepare plant-scale quantities of an aspartyl protease
inhibitor precursor: Urban, F. J.; Jasy, V. J. Org. Process Res. DeV.
2004, 8, 169–175.
(48) Pesti, J. A.; Downard, J. A.; Lauritsen, M. D.; Kauffman, G. S.; Bryant,
W. M., III; Huhn, G. F.; Arnett, J. F.; Yule, R. E.; Segretario, J.;
Nelson, K. A.; Gorko, E. F.; Page, G.; Lloyd, L. M.; Olson, R. E.;
Barnum, C. S.; Mrowca, J. J. Heterocycl. Chem. 1998, 35, 249–255.
(49) Stability studies in THF indicated a loss of 2.l wt % of Grignard
intermediate per day at 0 °C and 11.1 wt %/day at ambient
temperatures. This may be partially due to elimination of the ortho
fluorine and magnesium halide to lead to reactive benzyne species.63
(50) Dickerson, D. R.; Finger, G. C.; Shiley, R. H. J. Fluorine Chem. 1973/
1974, 3, 113–116.
(52) The amide insertion side product 20 arising from ethyl Grignard
addition to 4 was not significant under these conditions.
(53) There are other established means to accomplish plant-scale initiations
of Grignard reagent preparations: Tilstam, U.; Weinmann, H. Org.
Process Res. DeV. 2002, 6, 906–910.
722
•
Vol. 13, No. 4, 2009 / Organic Process Research & Development

Figure 2. Waterfall plot of 6 preparation and reaction with 4
to form 18.
Figure 1. FTIR results for conversion of 5 into 6.
and the assurance of rapid hydrolysis of any remaining Grignard
reagent or magnesium metal. The quench was addition con-
trolled. Not surprisingly, 2-7% of the product consisted of the
double-addition impurity 21, which is known to be a traditional
problem for Grignard reactions with carboxylic acid deriva-
tives.56 However, this was still preferable to charging the quench
solution into the reaction mixture as the content of 21 then rose
to 10-15%. This is explained as any unreacted Grignard reagent
would have had a higher probability of reacting further with
any freshly formed ketone 7 under those conditions.
The workup required the presence of ethyl acetate to produce
an adequate phase separation. Sufficient acetic acid was present
in the quench solution to produce a final pH of 5-6, sufficiently
low to both minimize base-catalyzed epimerization and avoid
the precipitation of voluminous magnesium salts, but high
enough to retain the THP group. The subsequent water wash
required the addition of heptane for efficient phase separation
since the magnesium salts, which had assisted the separation
of the first phase separation from the THF-rich organic phase,
were no longer present. The resulting solution was azeotropi-
cally dried by distillation with ethyl acetate to lower the water
content from 3-4% to e0.1%.
the plant by other means. Similarly, the level of aryl bromide
5 could be monitored and thus maintained at nonhazardous
levels by prior calibration of the signal generated by the presence
of 10% of the aryl bromide charge, the preagreed safety limit
to minimize chances for a runaway reaction. FTIR utilization
scaled up exactly as expected in the plant. Once the addition
of 5 began, the 1582 cm-1 signal which had been correlated to
the Grignard intermediate grew within 5 min, concomitantly
with a slow rise in temperature. The aryl bromide was then
metered in while the IR frequencies representing 5 and 6 were
monitored. The aryl bromide addition rate was decreased
whenever 5 approached the 10% threshold or if the temperature
approached 50 °C. In practice, no batch ever exceeded 45 °C
if the addition was metered over 2-2.5 h. The use of FTIR
monitoring permitted the safe and efficient completion of five
batches of Grignard reagent formation and subsequent coupling
(Figures 1 and 2).
Once the Grignard reagent formation was complete, amide
4 was charged directly into the Grignard solution to form the
putative aminol intermediate 18. As calorimetry had established
the heat of reaction as 47 kJ/mol with a adiabatic temperature
rise of only 20 °C, the reaction temperature was easily controlled
at <35 °C over the 15 min charging period. The conversion of
amide 4 into 18 was completed during an aging period at 40
°C. FTIR also effectively detected the reaction end point by
marking when the Grignard signal no longer decreased. For a
quantitative confirmation, a GC sample demonstrated that the
product/starting material ratio was >40:1.
Once the solution was dry, a final solvent exchange to DMF
permitted the formation of a 28-56 wt % solution of 7, optimal
for the subsequent epoxidation reaction. The preparations were
carried out reproducibly and safely in the plant on a 390 mol
scale to produce 355 kg of 7 in 79% overall yield and
98.0-98.5% ee over five batches. The consideration of safety
factors and use of FTIR allowed an efficient implementation
of a potentially dangerous reaction into the plant.
The quench of the aminol alkoxide 18 to form 7 was
accomplished by the transfer of the reaction mixture into cold
aqueous acetic acid (<25 °C) which permitted both control of
a very exothermic quench (adiabatic temperature rise of 67 °C)
Optimization of Reaction Conditions To Allow Plant Pro-
duction of Diol 11. The use of sodium hydride in DMSO and
a poor 4:1 diastereomeric selectivity during the epoxide forma-
tion of 8 required remediation before it could be considered an
efficient process.57 To avoid hydride bases, oxirane formation
using phase transfer catalysts was explored first, as milder bases
are usually used in combination with such catalysts.58 Modestly
higher diastereomeric selectivity (up to 7:1) was achieved at
(54) (a) Guidance for industry PAT - A framework for innoVatiVe
pharmaceutical deVelopment, manufacturing and quality assurance:
US Food and Drug Administration, Rockville, MD, USA (2004), http://
www.fda.gov/cder/guidance/6419fnl.pdf. (b) Process Analytical Tech-
nology (PAT) InitiatiVe: U.S. Food and Drug Administration, Rock-
ville, MD, USA (2005). http://www.fda.gov/cder/OPS/PAT.htm. (c)
Tummala, S.; Shabaker, J. W.; Leung, S. S. W. Curr. Opin. Drug
DiscoVery 2005, 8, 789–797.
(56) Huston, R. C.; Bailey, D. L. J. Am. Chem. Soc. 1946, 68, 1382–1383.
(57) Although others report higher (>7:1) ratios of the desired threo to
erythro isomers; see ref 21.
(58) Phase-Transfer Catalysis ; Halpern, M. E., Ed.; ACS Symposium
Series 659; American Chemical Society: Washington, DC, 1996.
(55) A detailed description of this use of PAT will appear separately: Leung,
S. Manuscript in preparation. For additional discussion of Grignard
reaction safety using real time monitoring, see ref 62.
Vol. 13, No. 4, 2009 / Organic Process Research & Development
•
723

The hydrochloride salt was prepared by adding 6 N HCl in
isopropanol to a dried and concentrated toluene solution of 9;
however the crystalline salt was problematic to isolate and
unstable to drying. In comparison, the MSA salt 11 could be
easily crystallized in high recovery from isopropyl acetate,
isopropanol, or a mixture of the two solvents and was a stable
salt after isolation. When MSA was charged slowly to a
isopropyl acetate/isopropanol solution of 9, the precipitation of
the crystalline salt signaled that the deprotection was complete.
If further purification was required following workup, isopro-
panol/heptane recrystallization would raise the purity from 96
to 99.7+ HPLC area % in 86-88% recovery while retaining
only 0.2% of the symmetrical triazole regioisomer correspond-
ing to 23. Although the properties of both the HCl and the MSA
salts were comparable in terms of overall yield and purity, the
ready isolation and robustness of the MSA salt led to its further
development.
In the plant, trimethylsulfoxonium iodide, 1,2,4-triazole,
DMF, and NaOtBu were stirred for 30 min, and a solution of
7 in DMF was charged. The epoxidation reaction was complete
within 30 min, and reaction mixture was heated to 90 °C to
initiate triazole addition to the epoxide. Complete conversion
to 9 required 9-12 h and typically formed 11-12% of 22 and
6-7% of regioisomer 23, among additional smaller impurities.
Following workup, the subsequent addition of MSA cleaved
the THP group. The resulting slurry was cooled, and filtration
yielded 254.3 kg of 11 over five batches. The overall yield was
49%61 (95 HPLC area % and 98.5% ee). The salt formation
was not only an effective means for isolation of 11 but also
efficient in rejecting the high levels of diastereomeric and
regioisomeric impurities formed in this process, as well as those
impurities that had been carried along during the several steps
telescoped into this reaction.
Figure 3. Impurities produced during triazole opening of 8.
the cost of reaction rate (∼24 h for completion) and difficult
workups. A variety of other bases (alkali HMDS, sodium tert-
butoxide and n-butyl lithium) and solvents were screened. The
best combination of diastereoselectivity and yield resulted from
the use of sodium tert-butoxide (NaOtBu) in DMF, producing
an improved yield (88%) and diastereomeric ratio (8.6:1).
The original conditions for the conversion of 8 to the
monoprotected diol 9 required triazole and sodium hydride.7,21
As in the previous step, less hazardous bases (potassium
carbonate, lithium hydroxide, NaOtBu, and sodium triazole)
were screened in DMF.59 All provided yields of >88% if heated
at 80-90 °C for 6-8 h but still created the impurity 22 (arising
from the undesired epoxide diastereomer of 8) and the sym-
metrical triazole regioisomer impurity (23), both identified by
independent synthesis. The symmetrical triazole regioisomer 23
is an ubiquitous impurity to this class of reactions due to the
well-known propensity of triazole to produce regioisomeric
products when employed as nucleophiles.48 All efforts to limit
the formation of 23 to <4% were unsuccessful, but this level
still compared favorably with triazole openings of similar
epoxides.60 Presumably an impurity incorporating both of these
undesired moieties also formed during this reaction but at
undetectable levels. The majority of these impurities were
removed by the subsequent workup.
Since the epoxide opening with 1,2,4-triazole could also be
carried out in DMF using a NaOtBu solution in THF, it
appeared that telescoping the epoxidation and triazole opening
into a single process operation would be possible. It was possible
to combine all reagents for both steps and select the desired
transformation by varying the temperature. Once 1.9 equiv of
a solution of 30 wt % NaOtBu in THF were added to
trimethylsulfoxonium iodide and triazole in DMF, the ylide
formed within 30 min and the subsequent addition of 7 led to
a clean conversion to 8. Once the epoxidation was complete,
heating the reaction mixture to 80-90 °C now activated the
triazole to regioselectively attack the oxirane to form alcohol
9. Following workup, the overall isolated yields for these two
steps were 80-85%.
The campaign and further project development was termi-
nated upon the completion of this step. While it was disap-
pointing to interrupt a successful and interesting campaign, the
remaining steps had been adequately demonstrated at pilot plant
scale in the previous campaign; thus a completed preparation
of ravuconazole at multikilo scale had been established.
Conclusions
At the onset of this work, there existed numerous challenges
to a large-scale preparation of ravuconazole. Practical solutions,
such as identifying salts to permit isolation of crystalline
intermediates, eliminating numerous purifications/isolations by
telescoping steps, developing safe means to handle potentially
hazardous chemicals or operations, and utilizing a novel
crystallization technology to produce small particles of bulk
drug, were put in place to provide a viable process. In particular,
the use of in-line FTIR monitoring allowed the safe formation
of a Grignard reagent and its subsequent reaction. This body
Removal of the THP protecting group by the standard means
of pyridinium p-toluenesulfonate yielded a hydrophilic brown
oil that was difficult to isolate upon aqueous workup. Purifica-
tion required trituration in diethyl ether and usually silica gel
chromatography to obtain acceptable product quality. We found
that hydrochloric acid, methanesulfonic acid, sulfuric acid, or
p-toluenesulfonic acid would also cleave the protecting group
cleanly, but isolations remained inefficient and impurities
remained entrained. A salt was sought to improve purity and
to streamline the workup and isolation.
(61) The last batch is not included in the yield range reported, as an
incomplete epoxidation lowered the purity.
(62) (a) Kryk, H.; Hessel, G.; Schmitt, W. Org. Process Res. DeV. 2007,
11, 1135–1140. (b) Wiss, J.; Ermini, G. Org. Process Res. DeV. 2006,
10, 1282–1286.
(59) The hazards of combining DMF and NaH are well established:
Handbook of ReactiVe Chemical Hazards; Bretherick, L., Ed.; But-
terworths: London, 1985; p 1133.
(63) (a) Mazza, D. D.; Reinecke, M. G. J. Org. Chem. 1988, 53, 5799–
5806. (b) Newman, M. S.; Dali, H. M.; Hung, W. M. J. Org. Chem.
1975, 40, 262–264. (c) Gribble, G. W.; Saulnier, M. G.; Sibi, M. P.;
Obaza-Nutaitis, J. A. J. Org. Chem. 1984, 49, 4518–4523.
(60) See ref 48 and refs 17-19 therein.
724
•
Vol. 13, No. 4, 2009 / Organic Process Research & Development

of work has created a means to prepare multikilogram batches
of ravuconazole and advanced intermediates for related com-
pounds with a potential for further scale-up to commercial
quantities
(750 kg). The combined organic phases were distilled down to
∼250 L at 50-55 °C jacket temperature to reduce the water
level to e0.5%. THF (400 kg) was added and distilled at 55-65
°C until the water level was e0.05%. THF was added to
produce a total of 274.8 kg of a 35 wt % solution of 4 for a
77% recovery of 4 from the aqueous washes. No detectable
methylene chloride remained. An analytical sample was purified
by silica gel chromatography (10f50% ethyl acetate/heptane)
to yield a colorless oil (>98% ee, >99 GC area %). HRMS calcd
M + H: 244.1524; obsd. M + H: 244.1544. IR (neat, thin film)
Experimental Section
(R)-1-Morpholino-2-(tetrahydro-2H-pyran-2-yloxy)pro-
pan-1-one (3). Methyl (R)-lactate (100 kg, 961 mol) and MTBE
(300 kg) were heated to 60 °C and distilled until all of the
MTBE had been removed (final volume ∼109 L). The solution
was cooled to 10 °C and treated with morpholine (86.2 kg, 1000
mol) at 10-20 °C. The batch was cooled to 5 °C, and a 25 wt
% solution of sodium methoxide in methanol (31.2 kg, 149
mol) was added at <5 °C. The mixture was warmed to 20 °C
and sampled for reaction completion by GC after 16 h (residual
lactate: 1.18 HPLC area %). The batch was titrated to a pH of
6.8 using a solution of methanolic hydrogen chloride which
had been previously prepared by mixing methanol (84.7 kg) at
5 °C with acetyl chloride (14.6 kg, 186 mol). The neutralized
reaction mixture was stirred 30 min and filtered through a 10
µ polypropylene cloth bag to remove salts. The reactor was
rinsed with THF (50 kg) to facilitate removal of the solids and
charged to the reaction mixture. The filtrate was heated to
60-65 °C, and the methanol was exchanged for THF (total of
880 kg of THF added) while maintaining a constant volume
(residual methanol relative to 3 was 3.9 GC A%). This solution
was carried into the next step without further isolation. An
analytical sample of 3 was purified by chromatography on silica
gel (20f50% ethyl acetate/heptane) to yield a colorless oil:
HRMS calcd M - H: 158.0832; obsd. M - H: 158.0824. IR
1
2941, 1655, 1116, 1032 cm-1. H NMR (CDCl3, 400 MHz)
mixture of diastereomers, 4.51-4.69 (m, 2Η), 3.86-3.88 (m,
1H), 3.61-3.73 (m, 8H), 3.47-3.50 (m, 1H), 1.70-1.84 (m,
3H), 1.52-1.56 (m, 3H), 1.41 (dd, J ) 7.0, 18.2 Hz, 3H). 13
C
NMR (CDCl3, 100 MHz) mixture of diastereomers, 171.5,
171.1, 99.0, 97.7, 72.3, 70.4, 67.4, 67.3, 67.2, 64.2, 63.0, 46.3,
46.2, 42.9, 42.8, 31.3, 31.0, 25.7, 25.6, 20.7, 19.7, 19.0, 17.6.
See ref 7 for a laboratory scale preparation and further spectral
data.
2R-1-1-(2,4-Difluorophenyl)-2-(tetrahydro-pyran-2-yloxy)-
propan-1-one (7). A stainless steel reactor with a recirculation
loop was assembled. The loop inlet was equipped with a
protective 40 mesh screen, and a Mettler FTIR instrument with
a 1” diameter flowcell equipped with a DeComp IR sensor
installed on a recirculation loop and running on ProcessIR was
installed.55 THF (303 kg) was charged followed by magnesium
turnings (9.30 kg, 383 mol) Via a solids addition funnel. A 1
M solution of ethyl magnesium bromide (2.25 kg, 1.16 eq
versus water in solution) was charged followed by a THF rinse
(10 kg) of the line. After 1 h, bromo-2,4-difluorobenzene (9.15
kg, 47.4 mol) was charged at 29-32 °C over 5 min, followed
by a THF rinse (1 kg). Initiation was observed by FTIR before
the entire charge was complete. Additional bromo-2,4-difluo-
robenzene (69.7 kg, 361.2 mol) was charged at <32 °C over
2.5 h, so that the FTIR-monitored aryl bromide levels never
exceeded 10% of the total aryl bromide charge. After a THF
rinse (1 kg) of the charging container, the reaction mixture was
aged at 25-27 °C until the Grignard level was no longer
increasing as monitored by FTIR (30 min). A solution of 4 (30.0
wt %, 257.3 kg, 317.3 mol) was charged over 20 min at <35
°C, and the reaction aged at 35 °C for 18 min. FTIR monitoring
indicated no further decrease of the Grignard content at this
point, and a quenched sample analyzed by HPLC displayed
1.0 area % of 4 relative to 7. The reaction mixture was
transferred over 40 min into a glass-lined reactor containing
acetic acid (41.5 kg, 691 mol) in water (207 kg) while
maintaining the temperature at <18 °C. The lines were rinsed
with THF (35 kg), and the hydrolysis was completed by
warming the mixture to 25 °C over 30 min. Ethyl acetate (207
kg) was charged, and the mixture was agitated and allowed to
settle for 30 min. The phases were separated, the organic phase
was diluted with heptane (85 kg), and the solution was washed
with water (206 kg). The solution was vacuum distillated at
0.1 bar and 15-20 °C to dry the solution to <0.1% by KFT.
Additional ethyl acetate (398 kg) was charged, and the mixture
was distilled down to a volume of ∼125 L (final water content
) 0.08%). Dimethylformamide (90 kg) was charged, and
volatiles were removed by distillation at 22-50 °C at 0.5 bar
1
(neat, thin film) 3419 (broad), 1644, 1114 cm-1. H NMR
(CDCl3, 400 MHz) 4.43-4.48 (m, 1H), 3.61-3.83 (m, 7H),
3.42-3.44 (m, 2H), 1.33 (d, 3H, J ) 6.8 Hz). 13C NMR (CDCl3,
100 MHz) 175.1, 67.5, 67.1, 64.8, 46.0, 43.3, 21.8. See ref 7
for a laboratory scale preparation and further spectral data.
(2R)-1-Morpholino-2-(tetrahydro-2H-pyran-2-yloxy)pro-
pan-1-one (4). A mixture of the solution containing 3 prepared
above and 3,4-dihydro-2H-pyran (122 kg, 1450 mol) at 5 °C
was treated with MSA (4.6 kg, 48 mol) in two portions while
maintaining the temperature at <10 °C. The reaction mixture
was warmed to 20-25 °C and held for 12 h. GC at this point
indicated 4.5 A% of 3 remaining. The reaction was washed
sequentially with the following solutions: 4% aqueous sodium
bicarbonate solution (150 kg), water (75 kg), and twice with
17% aqueous sodium chloride solution (220 kg, 440 kg). The
aqueous streams were held for later rework. The organic phase
was diluted with MTBE (400 kg), and water as a distilled
mixture of MTBE, THF, and water was removed at 65 °C. This
protocol was repeated until a total of 1000 kg of MTBE had
been charged (water content at end <0.1%). THF (400 kg) was
added, and the distillation continued until the MTBE was
removed as noted by the attainment of a batch temperature of
65-70 °C. The reaction mixture was sampled for wt % 4 (75
GC wt %), THF (170 kg) was charged, and the solution was
sent through a 5 µ filter to produce 407.2 kg (78.4% yield) of
a 45 wt % solution. The purity of 4 relative to the non-THF
components was 82.5 HPLC A % (ee ) 98%). Additional 4
was recovered by extraction of the aqueous washes. Two
batches were individually extracted with methylene chloride
Vol. 13, No. 4, 2009 / Organic Process Research & Development
•
725

to attain the following levels of volatiles: ethyl acetate (<5 A
%), heptane (<1 A %), THF (<1 A %), bromodifluorobenzene
(<2 A %), and KF < 0.5%. After 4 h, the final ethyl acetate
content by GC was 0.004%. This produced 122.7 kg of solution
(56.7 wt % 7, 81.1% yield, 98.0% ee). An analytical sample
was prepared by silica gel chromatography using ethyl acetate/
heptane. HRMS calcd M + H: 271.1146; obsd. M + H:
271.1145. 1H NMR (CDCl3, 500 MHz) mixture of diastereo-
mers, 7.85-7.92 (m, 1H), 6.96 (dd, J ) 8.6, 7.8 Hz; 1H),
6.79-6.90 (m, 1H), 5.08 (q, J ) 7.1, 2.4 Hz; S diast. of THP),
4.86 (qd, J ) 6.8, 1.2 Hz; R diast; 1H if combined with previous
signal), 4.75 (t, J ) 3.6 Hz; R diast.), 4.65 (t, J ) 3.6 Hz; S
diast.; 1H if combined with previous signal), 3.82-3.91 (m),
3.66-3.75 (m, 1H if combined with previous signal), 3.45-3.54
(m), 3.28-3.36 (m, 1H if combined with precious signal),
1.30-1.90 (m, 6H), 1.47 (dd, J ) 6.8, 1.2 Hz; S diast.), 1.42
(dd, J ) 6.8, 1.2 Hz; R diast.; 3H if combined with previous
signal). 13C NMR (CDCl3, 100 MHz) mixture of diastereomers,
198.2 (S diast.), 197.9 (R diast.), 165.9 (dd, J ) 258, 12 Hz; S
diast.), 165.7 (dd, J ) 257, 12 Hz; R diast.), 163.0 (dd, J )
257, 13 Hz; S diast.), 162.0 (dd, J ) 257, 13 Hz; R diast.),
133.1 (dd, J ) 11, 10.8 Hz; S diast.), 133.0 (dd, J ) 11, 10.8
Hz; R diast.), 121.4 (dd, J ) 42.1, 14.0 Hz; R diast.), 121.2
(dd, J ) 42.1, 14 Hz; S diast.), 112.5 (dd, J ) 21.6, 3.4 Hz; R
diast.), 112.3 (dd, J ) 21.6, 3.4 Hz; S diast.), 104.8 (dd, J )
27.8, 25.4 Hz; R diast.), 104.7 (dd, J ) 27.8, 25.4 Hz; S diast.),
98.7 (R diast.), 98.3 (S diast.), 77.4 (R diast.), 76.7 (S diast.),
62.8 (S diast.), 62.6 (R diast.), 30.7 (assign. uncer.), 30.6 (assign.
uncer.), 25.6 (R diast.), 25.4 (S diast.), 19.5, 18.4 (S diast.), 17.5
(R diast.). Anal. Calcd for C14H16F2O3: C, 62.21; H, 5.96. Found:
C, 62.37; H, 5.81. See ref 7 for a laboratory scale preparation
and further spectral data.
Methanesulfonate 1-[(2R,3R)-2-(2,4-difluoro-phenyl)-2,3-
dihydroxy-butyl]-1H-[1,2,4]triazol-4-ium (11). A solution of
30 wt % sodium tert-butoxide in THF (197.3 kg, 615.9 mol)
was added to a solution of trimethylsulfoxonium iodide (70.8
kg, 321.7 mol), 1,2,4-triazole (22.2 kg, 321.4 mol), and DMF
(498 kg) at <35 °C. The addition line was rinsed with DMF (5
kg), and the reaction mixture was stirred at 20-25 °C for 30
min. A solution of 53.1 wt % 7 in DMF (142.4 kg, 279.8 mol)
was added at <30 °C followed by a DMF (26 kg) rinse. The
reaction mixture was stirred for 30 min at which point an HPLC
sample determined complete reaction (<0.1 HPLC area % of 7
relative to 8; target was <5%). The mixture was heated to
85-90 °C for 12 h at which point an HPLC sample determined
a complete reaction (2.9 HPLC area % 8 relative to 9; target
was <5%). The mixture was cooled to 20 °C and quenched
with water (381 kg). The solution was extracted with isopropyl
acetate (2 × 393 kg), and the combined extracts were washed
first with 12 wt % NaCl solution (378 kg) followed by water
(2 × 378 kg). The organic solution was distilled to 150 L at
0.08 bar at 55 °C. Isopropyl acetate (298 kg) was added and
again distilled to achieve the end points of <0.3% water and of
isopropyl acetate at e15 relative GC area % versus all other
solvents. The solution was concentrated to 150 L and cooled
to 25 °C, and MSA (34.9 kg, 363.1 mol) was added at <35 °C.
The mixture was stirred at 20-25 °C for 30 min during which
time 11 started to crystallize. Sampling indicated complete
reaction (1.1 HPLC area % 9 relative to 10; target <5%). The
slurry was stirred for an additional 12 h at 20-25 °C, cooled
to 0-5 °C, and stirred for 3 h. The solids were collected by
filtration and washed sequentially with isopropyl acetate (2 ×
90 kg) and n-heptane (77 kg). The cake was dried under vacuum
at 45-50 °C to yield 11 (52.48 kg, 143.6 mol) in 51.3% yield
as an off-white crystalline solid (mp ) 173.5-174.5 °C).
HRMS calcd M + H for C12H14F2N3O2: 270.1054; found M
+ H: 270.1054. 1H NMR (DMSO-d6, 400 MHz) 8.92 (s, 1H),
8.09 (s, 1H), 7.25-7.19 (m, 1H), 7.18-7.11 (m, 1H), 6.87 (td,
J ) 8.46, 2.53 Hz, 1H), 4.80 (d, J ) 14.40 Hz, 1H), 4.72 (d,
J ) 14.40 Hz, 1H), 4.19 (qd, J ) 6.31, 2.78 Hz, 1H), 2.40 (s,
3H), 0.80 (d, J ) 6.31 Hz, 3H); exchangeable protons not
detected. 13C NMR (DMSO-d6, 126 MHz) 161.70 (dd, JC-F
)
246.05, 12.62 Hz), 158.56 (dd, JC-F ) 246.83, 11.83 Hz),
146.45, 143.45, 129.79 (t, JC-F ) 8.68 Hz), 124.01 (dd, JC-F
) 12.62, 3.15 Hz), 110.94 (dm, JC-F ) 20.50 Hz), 103.81 (t,
JC-F ) 27.61 Hz), 76.73 (d, JC-F ) 6.31 Hz), 68.14 (d, JC-F
)
4.73 Hz), 56.20 (d, JC-F ) 4.73 Hz), 39.68, 17.45. Anal. Calcd
for C13H17F2N3O5S: C, 42.73; H, 4.69; F, 10.40; N, 11.50; S,
8.77. Found: C, 42.63; H, 4.55; F, 10.47; N, 11.51; S, 9.16.
See ref 7 for a laboratory scale preparation of the free base and
spectral data.
2-((R)-1-((R)-2-(2,4-Difluorophenyl)oxiran-2-yl)ethoxy)tet-
rahydro-2H-pyran (8). An analytical sample was prepared by
flash chromatography on silica gel using ethyl acetate/heptane
to produce a colorless oil. HRMS calcd M + H for C15H19F2O3:
1
285.1302; found M + H: 285.1302. H NMR (CDCl3, 400
MHz) mixture of diastereomers, 7.54-7.48 (m, 1H), 7.40-7.35
(m, 1H), 6.89-6.83 (m, 2H), 6.81-6.75 (m, 2H), 4.92-4.90
(m, 1H), 4.75-4.74 (m, 1H), 4.07-4.00 (m, 1H), 3.99-3.93
(m, 1H), 3.89-3.79 (m, 2H), 3.53-3.46 (m, 2H), 3.31 (d, J )
5.31 Hz, 1H), 3.03 (d, J ) 5.06 Hz, 1H), 2.84-2.80 (m, 2H),
1.86-1.46 (m, 12H), 1.21 (dd, J ) 6.82, 1.51 Hz, 3H), 1.12
(dd, J ) 6.44, 1.14 Hz, 3H). 13C NMR (CDCl3, 126 MHz)
mixture of diastereomers, 162.76 (dd, JC-F ) 249.20, 4.73 Hz),
162.66 (dd, JC-F ) 249.20, 4.73 Hz), 160.56 (dd, JC-F
)
249.20, 3.15 Hz), 160.46 (dd, JC-F ) 249.20, 3.15 Hz), 131.70
(t, JC-F ) 4.73 Hz), 131.58 (dd, JC-F ) 9.47, 6.31 Hz), 121.40
(dd, JC-F ) 14.99, 3.95 Hz), 120.76 (dd, JC-F ) 14.99, 3.95
Hz), 111.16 (dd, JC-F ) 20.51, 3.16 Hz), 111.11 (dd, JC-F
)
22.08, 3.15 Hz), 103.54 (t, JC-F ) 26.03 Hz), 103.41 (t, JC-F
) 26.03 Hz), 98.61, 94.59, 74.45, 71.08, 62.58, 61.60, 59.77,
59.73, 51.18, 50.77, 30.59, 30.49, 25.41, 19.74, 19.48, 19.00,
17.48, 14.55. Anal. Calcd for C15H18F2O3: C, 63.37; H, 6.25;
F, 13.36. Found: C, 63.29; H, 6.25; F, 13.40. See ref 7 for a
laboratory scale preparation and further spectral data.
(3R)-2-(2,4-Difluoro-phenyl)-3-(tetrahydro-pyran-2-yloxy)-
1-[1,2,4]triazol-1-yl-butan-2-ol (9). An analytical sample was
prepared by flash chromatography on silica gel using ethyl
acetate/heptane to isolate 9 as a colorless oil. HRMS calcd M
+ H for C17H22F2N3O3: 354.1629; found: 354.1629. 1H NMR
(CDCl3, 400 MHz) mixture of diastereomers, 7.94 (s, 1H), 7.91
(s, 1H), 7.70 (s, 1H), 7.69 (s, 1H), 7.43-7.37 (m, 2H),
6.75-6.68 (m, 4H), 4.90-4.79 (m, 3H), 4.69-4.67 (m, 1H),
4.65-4.57 (m, 2H), 4.40-4.34 (m, 1H), 4.34-4.29 (m, 1H),
3.99-3.93 (m, 1H), 3.92-3.86 (m, 1H), 3.56-3.49 (m, 2H),
1.89-1.70 (m, 4H), 1.65-1.48 (m, 8H), 1.09 (d, J ) 6.32 Hz,
726
•
Vol. 13, No. 4, 2009 / Organic Process Research & Development

3H), 0.96 (d, J ) 6.31 Hz, 3H); alcohol proton not detected.
13C NMR (CDCl3, 126 MHz) mixture of diastereomers 162.64
(dd, JC-F ) 249.99, 12.61 Hz, 2C), 158.54 (dd, JC-F ) 246.05,
12.62 Hz, 2C), 151.14, 151.11, 143.99, 143.82, 130.46-130.25
(m, 2C), 122.91 (dd, JC-F ) 12.62, 4.73 Hz), 122.83 (dd, JC-F
) 12.61, 3.15 Hz), 111.57 (dd, JC-F ) 15.77, 3.15 Hz), 111.40
(dd, JC-F ) 14.20, 3.16 Hz), 103.77 (t, JC-F ) 26.82 Hz, 2C),
a summary of the safety precautions our plant operators used,
and the authors do not recommend this procedure be used at
any other facility without prior evaluation by a safety officer
familiar with cyanide chemistry.
Reaction Protocol. A solution of 1.65 M lithium bis(trim-
ethylsilyl)amide in THF (190.6 kg, 352.5 mol) was distilled
under vacuum at <30 °C to a volume of 140 L over 1 h and
held at 0 °C for 10.5 h. Acetone cyanohydrin (37.9 kg, 441.1
mol) was charged over 45 min, raising the temperature to 22
°C. In another vessel, 12 (98.1 wt %, 44.2 kg, 173 mol) and
THF (40 kg) were heated to 39 °C and held at that temperature
until dissolution was complete. This solution was transferred
to the reactor containing the mixture of lithium bis(trimethyl-
silyl)amide and acetone cyanohydrin over 10 min. The reactor
and transfer line were washed over with THF (8 kg). The
reaction mixture was heated to 65 °C and held at reflux for
12 h at which time the reaction was determined to be complete
by HPLC (0.2% 12 remaining in relation to 13). After 24 h
(the reaction was held an extra 12 h only to accommodate a
new shift; this had been established as an acceptable hold point),
the reaction was cooled to 1 °C and the pH was adjusted to 9
-10 with a 13.7 wt % solution of aqueous hydrochloric acid
(243 kg). During the neutralization, the temperature rose to 7
°C and the product crystallized. Water (51 kg) and isopropanol
(166 kg) were charged over 0.5 h. The solution was heated to
40 °C and concentrated under vacuum to 550 L. The resulting
slurry was cooled to -11 °C and held for 15 h. The product
was isolated in a Hastelloy centrifuge and was washed with
25% aqueous isopropanol solution (3 × 15 kg) at 15 °C. The
cake was dried at 45 °C under vacuum over 20 h to yield 43.5
kg of an off-white crystalline solid (99.6 HPLC area %, 90.2%
yield). An analytical sample was recrystallized from isopro-
panol: mp 178-182 °C (lit.4 179-182 °C). HRMS calcd M +
H for C13H13F2N4O: 279.1052; found M + H: 279.1059.
1HNMR (CDCl3, 300 MHz) 7.98 (d, J ) 9.0 Hz, 1H), 7.84 (d,
J ) 9.0 Hz, 1H), 7.41 (m, 1H), 6.78 (m, 2H), 5.57 (s, 1H),
4.97 (d, J ) 14.6 Hz, 1H), 4.84 (d, J ) 14.6 Hz, 1H), 3.31 (q,
J ) 9.0 Hz, 1H), 1.18 (d, J ) 9.0 Hz, 3H); 13C NMR (CDCl3,
126 MHz) 163.25 (dd, J ) 12.7, 251.8 Hz), 157.89 (dd, J )
12.7, 256.9 Hz), 152.37, 143.98, 130.97 (d, J ) 5.1 Hz), 121.28
(d, J ) 12.7 Hz), 119.65, 112.20 (d, J ) 17.8 Hz), 104.34 (t,
J ) 27 Hz), 75.83 (d, J ) 5.1 Hz), 55.76 (d, J ) 7.6 Hz),
33.57, 12.86. Anal. Calcd for C13H12F2N4O: C, 56.11; H, 4.35;
F, 13.66; N, 20.13. Found: C, 56.10; H, 4.23; F, 13.35; N, 20.09.
See ref 33 for a laboratory scale preparation and further spectral
data.
100.41, 97.25, 78.15 (d, JC-F ) 4.73 Hz), 77.90 (d, JC-F
4.73 Hz), 77.72 (d, JC-F ) 4.73 Hz), 73.57 (d, JC-F ) 4.73
Hz), 64.46, 63.64, 55.97 (d, JC-F ) 6.31 Hz), 55.72 (d, JC-F
)
)
6.31 Hz), 31.23, 30.99, 25.18, 25.12, 20.74, 20.18, 15.61, 13.62.
Anal. Calcd for C17H21F2N3O3: C, 57.78; H, 5.98; N, 11.89; F,
10.75. Found: C, 56.95; H, 5.78; N, 11.58; F, 10.61. See ref 7
for a laboratory scale preparation and further spectral data.
1-(((2R,3S)-2-(2,4-Difluorophenyl)-3-methyloxiran-2-yl)-
methyl)-1H-1,2,4-triazole (12). Triethylamine (109.3 kg, 1080
mol) was charged to a solution of 11 (130 kg, 356 mol) in THF
(520 kg) and MTBE (443 kg) held at -10 to -5 °C.
Methanesulfonyl chloride (61.4 kg, 536 mol) was charged at
<10 °C. The solution was cooled to 2 °C and held for 1 h.
Sampling at this point indicated a complete reaction by HPLC.
A 3.75 M solution of sodium hydroxide (394 kg) was charged
at <5 °C and held until reaction completion by HPLC. Agitation
was stopped, and the lower aqueous phase was discarded. The
temperature was raised to 20 °C, and the organic layer was
washed with a 20 wt % NaCl solution (624 kg). The solution
was distilled at 0.4 bar and <55 °C to remove MTBE and THF
to a volume of 50 L, and 95% ethanol (434 kg) was charged.
The solution was again distilled at 0.4 bar and <60 °C until
GC indicated <2 relative GC area % MTBE remained (final
volume ∼275 L). Water (812 kg) was charged over 1 h. The
slurry was cooled to 2 °C and held 8 h, and the product was
collected by filtration. The cake was washed with 10 °C water
(2 × 400 kg) and dried under vacuum at 50 °C to produce
71.4 kg of an off-white crystalline solid in 78% yield (96.9
HPLC area %, 97.8 HPLC wt %; 99.8% ee). The physical data
for this compound have been published.7
(2S,3R)-3-(2,4-Difluorophenyl)-3-hydroxy-2-methyl-4-
(1H-1,2,4-triazol-1-yl)butanenitrile (13). Safety Requirements
for Handling a Process InVolVing Cyanide. PPE: chemical
resistant gloves, safety goggles, a supplied air respirator, and
chemical resistant suit, such as Saranax, for any charging or
sampling operations. All personnel were equipped with cyanide
monitors set to 10 ppm, and amyl nitrate capsules were supplied
near but outside the bay. The bay was barricaded with cyanide
warning signs during operations. An observer outside the bay
watched during all operations involving potential exposure. All
reaction streams were kept caustic (pH >8) to minimize the
possibility of formation of HCN. Quench solutions of 3% bleach
were kept immediately available for decontamination purposes.
The reactors were cleaned by filling them with a 5% bleach
solution and agitating for 30 min. The risers and lines were
similarly treated. The waste, such as the mother liquor and
washes, contained 3-4% lithium cyanide and was not quenched
with bleach due to the possibility of forming peroxides from
the THF present. It was safer to drum up the waste, label as
containing cyanide, and ship to a waste disposal facility
equipped to handle cyanide. This description is meant only as
(2R,3R)-3-(2,4-Difluorophenyl)-3-hydroxy-2-methyl-4-
(1H-1,2,4-triazol-1-yl)butanethioamide Hydrate Hydrosul-
fate (15). Safety Requirements for Handling a Process InVolVing
Hydrogen Sulfide. PPE: same as for compound 13 except that
H2S monitors were used for all operations. Cleaning and waste
treatment are the same except the waste was labeled as
containing hydrogen sulfide and other stench producing chemi-
cals prior to shipment to waste disposal.
Reaction Protocol. A solution of 13 (43.0 kg, 154.5 mol),
water (17.2 L), and isopropyl alcohol (54.0 kg) was heated to
78 °C over 80 min at which point a solution had occurred.
Diethyl dithiophosphate (86.2 kg, 462.9 mol) was added over
Vol. 13, No. 4, 2009 / Organic Process Research & Development
•
727

1.5 h while maintaining the temperature at 75-80 °C. (CAU-
TION: diethyl dithiophosphate is highly toxic, corrosive, and
combustible and produces a stench. This reaction also leads to
off-gasing of hydrogen sulfide, which is highly flammable, toxic,
and irritating and produces a stench. Proper precautions are
required prior to conducting this reaction.) Additional isopro-
panol (2 kg) was used to rinse the line into the reactor. The
heating was maintained for 7.5 h, until reaction completion (13
) 0.93 HPLC area %). The mixture was cooled to 20 °C over
1 h. Ethyl acetate (194 kg) and water (215 kg) were charged to
the mixture over 50 min, and the mixture cooled to 2 °C over
80 min. Sodium hydroxide solution (3N, 284 kg) was charged
over 75 min while maintaining the resulting temperature at
10-20 °C. The final pH was 8.1, and the final temperature 15
°C. The phases were separated, and the organic phase was
treated with 20 wt % sodium hydroxide (259 kg). The mixture
was stirred for 30 min at ambient temperature (23 °C), and the
layers were separated. The transfer lines were rinsed with ethyl
acetate (2 kg). The combined organic phases were then treated
with 98 wt % sulfuric acid (15.3 kg, 152.9 mol) over 3 min at
a final temperature of 40 °C. The lines were rinsed with ethyl
acetate (2 kg). The solution was concentrated to 87 L by vacuum
distillation at 50-60 °C over 6 h. The slurry was cooled to 26
°C over 70 min, and MTBE (128 kg) was charged over 9 min.
The slurry was further cooled to 5 °C over 1.5 h and held at
0.2-5 °C over 1 h. The crystals were collected by filtration
and washed with MTBE (136 kg). The salt was dried at 60 °C
under vacuum for 34 h to achieve an LOD of 0.35%. A total
of 58.2 kg were isolated in 91.2% yield (99.9 HPLC area %).
An analytical sample was recrystallized from ethyl acetate/
MTBE. Mp ) 152 °C (dec.). Anal. Calcd for C13H18F2N4O6S2:
C, 36.44; H, 4.23; N, 13.08; F, 8.76; S, 14.97. Found: C, 36.41;
H, 4.07; N, 13.13; F, 9.14; S, 15.02. 1H NMR (DMSO-d6, 300
MHz) (not all exchangeable protons detected) 10.19 (d, J )
46.1 Hz, 2H), 8.85 (s, 1H), 7.98 (s, 1H), 7.28 (m, 2H), 6.92
(m, 1H), 6.70 (br s, 1H), 4.84 (d, J ) 14.1 Hz, 1H), 4.57 (d, J
) 14.1 Hz, 1H), 3.60 (q, J ) 6.6 Hz, 1H), 0.92 (d, J ) 6.8 Hz,
3H). 13C NMR (DMSO-d6, 76 MHz) 208.3, 162.7 (d, J ) 250
Hz), 157.1 (d, J ) 264 Hz), 146.7, 143.9, 130.7, 123.3, 111.6,
104.2, 75.2, 57.3, 49.6, 16.3. Spectral data for the free base 14
are found in ref 4.
at 55-70 °C. The pH of the solution was adjusted with
triethylamine (15.6 kg, 154.2 mol) to pH 3.5-4.0. The solution
was stirred at 55-65 °C for 20 min followed by cooling to 20
°C over 2 h. The slurry was stirred for 10 h at 20-25 °C, and
the crystals were isolated in a Hastelloy filter dryer. The cake
was washed twice with a 50 v/v% ethanol/water solution (98
kg ethanol, 123 kg water) and vacuum-dried at 50-55 °C for
8 h to yield 22.7 kg of a white, crystalline solid (99.7 HPLC
area %, LOD 0.12%, 85.8% yield).
Sonic Impinging Jet Crystallization Protocol. A solution of
1 (2 kg) in 25 L of 95% ethanol at 70-79 °C was filtered
through a 0.2 µm cartridge filter into a holding vessel and
maintained at 70-79 °C. Water (55 L) was filtered in the same
manner and transferred to another holding vessel at 0-5 °C.
Seed crystals (∼0.25 g) of 1 were suspended in ∼600 mL of
1:2 v/v 95% ethanol/water in the impingment vessel. The
suspension was sonicated for 1 min by a 0.5” probe with a 120
W power output. The solution containing 1 was pumped through
a 0.02” diameter nozzle at 0.18 kg/min and was impinged with
a water jet being pumped at 0.37 kg/min through a 0.03”
diameter nozzle. The newly formed crystal slurry was sonicated
in the impingment vessel which was maintained at a constant
volume by continuous transfer of slurry to a receiver (Residence
time of the crystal slurry in the impingment vessel was ∼1 min).
After the entire solution was crystallized and transferred to the
receiver, the slurry was cooled to 0-5 °C and filtered. The solid
was vacuum-dried at 55 °C until the LOD was e1 w/w% to
afford 1.95 kg (97.5% recovery) of 1 with a mean particle size
of 7.5 µm. The physical and spectral data for the pure compound
is in ref 4.
Acknowledgment
We gratefully acknowledge David Kronenthal for reviewing
this manuscript; Tammy Baker, Lewis Bernstein, Melissa Chau,
Scott Chadwick, Doris Chen, Masaki Endo, Margaret Gleeson,
Jacques Jerome, San Kiang, Donald Kientzler, Jean Lajeunesse,
Cary McConlogue, Brian Siesel, and Pierre Sirard for labora-
tory, kilo laboratory, and pilot plant research or operation
contributions; George B. Crull for conducting NMR experi-
ments; Francis Okuniewicz and Steve Chan for hazard evalu-
ations; Mary Talmadge-Grebenar and Jeannette Manello for
literature searches; and the pilot plant supervisors/operators for
their skilled technical assistance.
4-(2-((2R,3R)-3-(2,4-Difluorophenyl)-3-hydroxy-4-(1H-
1,2,4-triazol-1-yl)butan-2-yl)thiazol-4-yl)benzonitrile (1). A
mixture of 15 (25.8 kg, 99.4 area % purity, 60.2 mol adjusted
for water content), 2-bromo-4′-cyanoacetophenone (hazard, this
is a powerful lachrymator and requires PPE as described for
the previous step) (13.8 kg, 61.9 mol), and 95% ethanol (104
kg) was heated to 60-70 °C and held for 2 h until the reaction
was complete (0.09 HPLC area % of the aryl bromide relative
to 15). Water (122 kg) and 95% ethanol (104 kg) were charged
Supporting Information Available
Analytical methods are fully described. This material is
available free of charge via the Internet at http://pubs.acs.org.
Received for review March 21, 2009.
OP900065C
728
•
Vol. 13, No. 4, 2009 / Organic Process Research & Development