Efficient synthesis of 8-Oxa-3-aza-bicyclo[3.2.1]octane hydrochloride

Article abstract of DOI:10.1021/op9002642

A four-step process to generate 8-oxa-3-aza-bicyclo[3.2.1]octane hydrochloride starting with 5-hydroxymethyl-2-furfuraldehyde is described. Raney nickel-mediated reduction of 5-hydroxymethyl- 2-furfuraldehyde afforded 2,5-bis(hydroxymethyl)tetrahydrofuran with a predominantly cis configuration. Conversion of the diol to a ditosylate, followed by cyclization with benzylamine generated the 8-oxa-3-aza-bicyclo[3.2.1]octane core. The title compound was prepared via hydrogenolysis of the N-benzyl group with Pearlman's catalyst and isolated as the hydrochloride salt from a mixture of 2-propanol and heptane. The overall yield for the process was 43-64%. An impurity generated during scale up highlighted the need for a strict in-process specification for acetonitrile content prior to performing the final hydrogenolysis. Ethanol conditioning of all processing equipment that had acetonitrile contact prior to the introduction of the substrate stock solution is highly recommended.

Full text of DOI:10.1021/op9002642

Organic Process Research & Development 2010, 14, 459–465
Efficient Synthesis of 8-Oxa-3-aza-bicyclo[3.2.1]octane Hydrochloride†
Terrence J. Connolly,* John L. Considine, Zhixian Ding, Brian Forsatz,§ Mellard N. Jennings, Michael F. MacEwan,
Kevin M. McCoy, David W. Place, Archana Sharma, and Karen Sutherland
Chemical DeVelopment, Wyeth Research, 401 North Middletown Road, Pearl RiVer, New York 10965, U.S.A.
Abstract:
Reduction conditions were screened using a parallel, mul-
tireactor synthesizer using 10 mL cells with a fill volume of
1.75 mL operating at 600 rpm.4 The screening focused on
conditions that could be used in our facility so all reactions
were examined at 60 psig hydrogen and ambient temperature.
Multiple types of palladium on carbon and Raney nickel were
examined as catalysts in an array of solvents. A summary of
the screening is presented in Table 1.
In-process analysis using GC/MS identified an intermediate
in the reduction sequence having an exact mass of 130, assigned
as the half-reduced furan 7 (Scheme 2), and also revealed that
product 3 was formed as a mixture of two isomers.5 NMR of
isolated 3 indicted that the major isomer was consistent with a
meso structure, in line with the expected cis product (cis-3 in
Scheme 2). Evaluating the results of our screening, palladium
on carbon catalysts were not effective in completely reducing
2 to 3. A significant amount of intermediate 7 was formed and
extending the reaction time from 3 to 24 h did not improve
further conversion to 3.6
A four-step process to generate 8-oxa-3-aza-bicyclo[3.2.1]octane
hydrochloride starting with 5-hydroxymethyl-2-furfuraldehyde is
described. Raney nickel-mediated reduction of 5-hydroxymethyl-
2-furfuraldehyde afforded 2,5-bis(hydroxymethyl)tetrahydrofuran
with a predominantly cis configuration. Conversion of the diol to
a ditosylate, followed by cyclization with benzylamine generated
the 8-oxa-3-aza-bicyclo[3.2.1]octane core. The title compound was
prepared via hydrogenolysis of the N-benzyl group with Pearl-
man’s catalyst and isolated as the hydrochloride salt from a
mixture of 2-propanol and heptane. The overall yield for the
process was 43-64%. An impurity generated during scale up
highlighted the need for a strict in-process specification for
acetonitrile content prior to performing the final hydrogenolysis.
Ethanol conditioning of all processing equipment that had aceto-
nitrile contact prior to the introduction of the substrate stock
solution is highly recommended.
Contrasting with the palladium on carbon results, the reaction
profile was very clean when Raney nickel (entries 11-14) was
used as the catalyst. Best results were seen in MeOH, IPA, or
a mixture of EtOAc and MeOH, all of which showed complete
conversion to the desired product within 3 h at 60 psig hydrogen
and ambient temperature. The use of methanol produced the
cleanest cis-3 and was chosen for evaluation at a larger scale.
Initial scale-up runs were performed in a 500 mL Parr shaker
type hydrogenator. Scaling the reaction up to 10 g of starting
material required that the reaction temperature and hydrogen
pressure be increased to 60 °C and 70 psig respectively in order
to have complete conversion of 2 to 3. Using these conditions,
the reaction was complete within 16-18 h with a Raney nickel
charge of 0.3-0.4 mass equivalents.
As the reaction was scaled up further, a Parr 2 L stirred
vessel was used. The reduction of 2 to 3 became more sluggish,
a demonstration of the significant impact of mass transfer on
the reduction. The reaction required 0.8 mass equivalents of
Raney nickel in order to reach completion and a new in-process
intermediate having a mass of 128 was identified. The new
intermediate was assumed to be furan derivative 6 since it did
convert to product 3. Comparing multiple runs conducted in
the stirred 2 L pressure reactor with the same lots of starting
material, catalyst, solvent ratio, and catalyst charge resulted in
Introduction
8-Oxa-3-aza-bicyclo[3.2.1]octane hydrochloride (1) was
required as a starting material for a program under development
at Wyeth. Although commercial sources of this compound were
available, the lead time proposed by external vendors was not
acceptable to Wyeth’s productivity model, details of which have
been published recently.1 To meet the expedited pre-IND
timeline, the compound was made in house. This communica-
tion describes the route chosen for scale up in order to meet
our delivery commitment.
Results and Discussion
The process developed to prepare 1 is shown in Scheme 1
and was based heavily on the work of Wiggins and Cope.2,3
These early reports on hydrogenation of 5-hydroxymethyl-2-
furfural (2) to produce 2,5-bis(hydroxymethyl)tetrahydrofuran
(3) used Raney nickel in diethyl ether at 130-160 °C and
1100-2000 psig hydrogen. Although the reported yield was
excellent, (90%) the conditions were not acceptable for scale
up in our facility.
† Part of this work was presented at the 2008 PhRMA API Workshop, San
Juan, Puerto Rico, April 28-May 1, 2008.
* Address correspondence to this author at the above mailing address, attention
B222/2125. E-mail: connolt@wyeth.com.
§ Wyeth Research, Analytical and Quality Science, Pearl River, NY. Wyeth
was acquired by Pfizer on 16-Oct-2009.
(4) Screening was performed using an Endeavor Catalyst Screening System
from Biotage/Argonaut.
(1) O’Brien, M. K.; Kolb, M.; Sutherland, K.; McCoy, K.; Pilcher, A.
Chimia 2006, 60, 518–522.
(5) 13C NMR supported the 95/5 ratio of isomers determined by GC. See
the Experimental Section for line listings of both the major and minor
isomer.
(2) Cope, A. C.; Baxter, W. N. J. Am. Chem. Soc. 1955, 77, 393–6.
(3) Haworth, W. N.; Jones, W. G. M.; Wiggins, L. F. J. Chem. Soc. 1945,
1–4.
(6) Higher reaction temperatures and additional catalyst charges were not
evaluated for the reactions that did not go to completion.
10.1021/op9002642  2010 American Chemical Society
Published on Web 02/16/2010
Vol. 14, No. 2, 2010 / Organic Process Research & Development
•
459

Scheme 1. Synthesis of 1
different reaction rates when the agitation speed of the Parr
reactor was varied. In fact, at lower agitation (<200 rpm), the
reaction did not just slow down but stopped completely.
To gain a more qualitative understanding of the reaction,
the hydrogenations were performed in an RC1e equipped with
a MP10 glass vessel and a 6-bar rupture disk. Use of this
transparent vessel confirmed suspicions that with slow to
moderate stirring (<400 rpm), a large portion of the Raney
nickel remained on the bottom of the reactor, which led to slow
and incomplete conversion of 2 to 3. At higher agitation
(600-700 rpm), the catalyst was lifted off the bottom of the
vessel, and complete conversion to 3 was straightforward.7
When a gas-diffusion stirrer was used (700 rpm), complete
conversion required as little as 4 h with 0.8 mass equivalents
of Raney nickel.
When the reduction was complete, the catalyst was filtered
from the reaction mixture, and the filtrate was concentrated,
diluted with acetonitrile, and concentrated again until the
methanol to acetonitrile ratio was e2:98 (GC area %), and the
water content was less than 1% (KF method). Levels of
methanol and water above these respective values resulted in
incomplete reaction in the next step.
Conversion of diol 3 to ditosylate 4 was performed using
slightly modified conditions from those reported by Cope.2 Our
first approach involved addition of solid tosyl chloride to a
mixture of 3 in acetonitrile and pyridine. Although complete
conversion of 3 to the monotosylate intermediate was achieved,
further conversion to 4 was slow. Even after 16 h at ambient
temperature, the reaction mixture still contained 20% of the
monotosylate intermediate. Increasing the amount of tosyl
chloride to 3 equiv and using catalytic dibutyltin oxide acceler-
ated the second step of the process, and complete reaction was
observed in less than one hour with 10 equiv of pyridine. The
heat output from this reaction was examined in an RC1 and
was found to be proportional to the dose of tosyl chloride
(Figure 1). The total heat output for the reaction was 152 J/g
of reaction mass, corresponding to a ∆Tad of 151 °C. However,
in light of the excellent dose-control over the exotherm, the
magnitude of the adiabatic temperature rise was not a concern.
Wyeth’s pilot-plant technical group provided feedback that
the tin-catalyzed process for the conversion of 3 to 4 had two
liabilities: Portionwise addition of solids to a reactor that
contained a flammable solvent was not the preferred mode of
addition, but more importantly, the use of tin in the reactors of
a multipurpose plant would require that specific cleaning and
testing procedures be developed to ensure that the processing
equipment was free of tin residues at the end of the process.
The reaction conditions were modified such that the aceto-
nitrile stock solution of 3 and 3 equiv of tosyl chloride were
combined in a reactor followed by temperature-controlled
addition of pyridine to the mixture. Increasing the pyridine
charge from 10 to 13 equiv and extending the reaction time to
16 h resulted in complete conversion of 3 to 4. Quenching the
reaction mixture with water resulted in precipitation of the
product, which was filtered directly and then rinsed with water
and ethanol. Isolation of 4 using these conditions effectively
purged the trans-ditosylate product that was generated from the
trans-3 carried into the tosylation stepsthere were no impurities
detected in the isolated product with a molecular weight of 440.
Evaluation of the modified process in the RC1 provided a
heat output of 116 J/g of reaction mass and ∆Tad of 55 °C for
the desired reaction (Figure 2). The heat flow profile is the result
of multiple exothermic eventsspresumed activation of the tosyl
chloride, the desired reaction, HCl neutralization, and a late
exotherm that was observed when stirring was increased to more
effectively agitate the thick slurry that had formed. However,
the majority of the exotherm occurred early in the pyridine
addition profile and was proportional to the dose of pyridine.
Having a dose-related exotherm alleviated any concerns regard-
ing control of the exotherm on scale.
Table 1. Screening Results for Hydrogenation of 2
SM intermediate trans-product cis-product
As noted above, tosylation of the diol produced a thick slurry
in the reactor that prevented quenching the reaction mixture
into water. Instead, the reaction had to be quenched by addition
of water to the reaction mixture. Evaluation of the quench in
the RC1 showed that for the entire water addition, the ∆Tad
was 45 °C, and Qmax was 90 W. However, as shown in Figure
3, the majority of the heat generated during the quench occurred
during the first charge of water. This first charge of water (8 g)
trial catalyst solvent (%)
(%)
(%)
(%)
1
2
3
4
5
6
7
8
9
5% Pd/Ca MeOH 54
14
50
44
53
16
61
59
38
33
33
0
3
5
8
0
5
5
0
0
10
9
25
42
39
12
34
32
32
0
57
55
95
93
39
94
5% Pd/Ca EtOAc
5% Pd/Ca THF
5% Pd/Ca PhMe
0
0
29
5% Pd/Cb MeOH 43
5% Pd/Cb EtOAc
5% Pd/Cb THF
5% Pd/C2 PhMe
5% Pd/Cb IPA
0
0
62
0
10 10% Pd/Cc MeOH
11 Raney Nid MeOH
12 Raney Nid EtOAce
13 Raney Nid THF
14 Raney Nid IPA
0
0
0
61
0
(7) At the suggestion of a reviewer, suspension calculations were
performed. The minimum stir speed required in the RC1 to just suspend
the Raney nickel (Njs) was calculated to be 1580 rpm. A Dynochem
model predicted that the Raney nickel would be fully suspended at
stir speeds above 1700 rpm. To have complete reactions in a regime
where there was less than optimal power for suspension presumably
indicates that kLa was quite high. Dynochem modeling also predicted
that at 180 rpm, the Raney nickel would be fully suspended in the
pilot-plant vessels. We are grateful to Dr. Albert Schwartz for
performing the DynoChem modeling.
5
7
0
6
0
0
0
a Water-wet (66 wt % H2O) catalyst (standard, reduced). b Water-wet (51 wt %
H2O) catalyst (eggshell, unreduced). c Commercial dry catalyst (uniform, reduced
metal). d Type 2800, 50 wt % water. e Methanol was added to facilitate dissolution
of the starting material in EtOAc.
460
•
Vol. 14, No. 2, 2010 / Organic Process Research & Development

Scheme 2. Proposed Intermediates Detected During Reduction of 2 to 3
corresponded to 1.8 mol equiv of water, and presumably, the
large heat output was due to hydrolysis of the excess tosyl
chloride. The second water charge, although 10 times larger in
quantity than the first water charge, resulted in a much smaller
heat flow. Although some of the heat generated during the
addition of water was due to heat of dilution, the magnitude of
the heat output, particularly during addition of the first 30 g, is
more indicative of a heat of reaction. The heat of dilution
appeared to be constant at approximately 20 W, visible during
the addition of the last 60 g of water in Figure 3.
The heat-flow data collected during this RC1 experiment
alerted us to the potential for a significant temperature rise
during the initial part of the quench when the reaction was scaled
up to the pilot plant. The operations group was advised to add
the initial 10% water charge slowly and gradually increase the
addition rate once control of the exotherm was observed. On
the pilot-plant scale, the water addition rate gradually increased
from 0.2 kg/min to 6 kg/min, once control of the batch
temperature was demonstrated.
of trans-3 formed was 5.4%. The catalyst was removed, and
the solvent was exchanged to acetonitrile by concentrating under
vacuum (20-30 Torr) with a maximum jacket temperature of
70 °C. Acetonitrile was charged and the vacuum distillation
repeated three times (100 kg and 2 × 65 kg charges of
acetonitrile). Residual water was 0.7% and 0.2% (in-process
target was <0.5%) after the second and third charges, respec-
tively. Residual methanol at the end of the last distillation was
0.01%. The rest of the processing proceeded as expected and
generated 28.3 kg of 4 (83% yield). Filtration and drying on a
0.25 m2 filter/dryer did require extended timess18.5 and 66.5 h,
respectively.
With a reliable method to generate ditosylate 4 in hand,
attention shifted to conversion of 4 to 1. Although there are
reports for the direct conversion of 4 to 1 with ammonia in
alcohol (150 °C and 650 psig),8-10 we were not successful in
adapting that chemistry into a process that could be run in our
facilities. Instead, a two-step process involving cyclization of
the ditosylate with benzylamine followed by hydrogenolysis
was developed.11-13
Ditosylate 4 was cyclized to 5 in acetonitrile at reflux in the
presence of 4 equiv of benzylamine. Attempts to use less than
4 equiv led to incomplete conversion. When the cyclization was
complete (<5% ditosylate), 1 equiv of p-toluenesulfonic acid
was added to scavenge the extra equivalent of benzylamine.
The solvent was exchanged to MTBE, the benzylamine/p-
toluenesulfonic acid salt was removed via filtration, and the
solvent was exchanged to ethanol.
Scale up of the process to the pilot plant started with 10 kg
of 2. Hydrogen uptake continued for 14.5 h, after which time
in-process analysis indicated that there was no starting material
or intermediates present in the reaction mixture and the amount
The ethanol stream of 5 was combined with Pearlman’s
catalyst, and debenzylation was performed at 60-65 °C/60-65
psig hydrogen. Some lots of 5 had up to 3% residual benzyl-
amine present, but this did not have a negative impact on the
desired reaction as benzylamine was reduced to toluene and
ammonia under the reaction conditions. To gain a better
understanding of the time required for reaction completion and
to evaluate the impact of extended reaction time of product
stability, hydrogenolysis of 5 was examined using a MonARC
in situ FTIR system. The instrument was supplied by Mettler-
Toledo and was fitted to the bottom of a standard Parr stirred
tank hydrogenator. Starting material 5 had strong IR peaks at
1455, 1166, and 797 cm-1. Hydrogenation led to a decrease in
intensity of these peaks with growths at 1402 and 1294 cm-1
Figure 1. Heat Flow for 3 to 4: TsCl added to Reaction Mass.
(8) Miller, A. D. U.S. Patent 3,856,783, 1974.
(9) Miller, A. D. U.S. Patent 3,953,596, 1976.
(10) Newth, F. H.; Wiggins, L. F. J. Chem. Soc. 1948, 155–158.
(11) Feurer, A.; Flubacher, D.; Weigand, S.; Stasch, J.-P.; Stahl, E.;
Schenke, T.; Alonso-Alija, C.; Wunder, F.; Lang, D.; Dembowsky,
K.; Straub, A.; Perzborn, E. WO 03/004503, 2003.
(12) Feurer, A.; Luithle, J.; Wirtz, S.-N.; Konig, G.; Stasch, J.-P.; Wunder,
F.; Lang, D.; Stahl, E.; Schenke, T.; Schreiber, R. WO 2004/031186,
2004.
(13) Zask, A.; Kaplan, J. A.; Verheijen, J. C.; Curran, K. J.; Richard, D. J.;
Ayral-Kaloustian, S. U.S. 2009/0098086, 2009.
Figure 2. Heat flow and pyridine dosing profiles for conversion
of 3 to 4.
Vol. 14, No. 2, 2010 / Organic Process Research & Development
•
461

Figure 3. Heat flow during aqueous quench of 4.
Figure 4. Chromatograms showing reference spectra of SM and toluene vs the ConcIRT calculated product spectrum.
(Figure 4). The trend for the hydrogenation followed at 1402
and 797 cm-1 is shown in Figure 5. The Sentinal probe, located
on the bottom of the tank, was initially blinded by the catalyst.
After the vessel was purged of air, the atmosphere was
converted to hydrogen and stirring was initiated. An immediate
growth for the starting material was detected at 797 cm-1 as
the catalyst was slowly cleared off the probe. The concentration
of starting material at 797 cm-1 gradually decreased as the
product component grew at the same rate at 1402 cm-1 as would
be expected for a reaction without an intermediate. Extending
the reaction time for hours after the reduction was complete by
IR did not result in a significant decrease in the intensity of
product at 1402 cm-1, providing evidence that product stability
under the reduction conditions was excellent.
Interestingly, toluene was not identified as a separate product
of the reaction by the ConcIRT algorithm. However, as shown
in Figure 4, the IR spectrum of the product component did show
IR bands common to toluene. Since both the product and
toluene would appear at the same rate, it is likely that the
ConcIRT component spectrum for product is actually a com-
posite spectrum of toluene and the secondary amine product.
Once the hydrogenation was complete, the catalyst was
removed and the product was isolated as a hydrochloride salt.
HCl was prepared in situ by addition of chlorotrimethylsilane
462
•
Vol. 14, No. 2, 2010 / Organic Process Research & Development

Figure 5. Relative trends showing normalized IR signals from SM 5 (797 cm-1) and product 1 (1402 cm-1).
to the solution of 1 that contained some alcohol.14,15 The
solubility of the product hydrochloride salt was quite high in
ethanol (23.7 mg/mL) as determined by HPLC using a Charged
Aerosol Detector (CAD).16,17 Losses to the mother liquors would
not be acceptable if the product was isolated directly from
ethanol. Compound 1 was not soluble in heptane, and addition
of IPA up to a 1:1 ratio only increased the solubility to 1.8
mg/mL. On the basis of these data, the solvent was exchanged
to heptane after the reduction was complete, and IPA and
chlorotrimethylsilane were added to effect precipitation of the
product as the HCl salt. Lab yields for the conversion of 4 to
1 were between 60 and 75%.
The process was scaled up to the pilot plant and performed
on 27.5 kg of 4. The cyclization required 48 h on this scale to
reach completion compared to 24 h that was observed in the
lab. The batch was concentrated at atmospheric pressure, and
MTBE was added. At this point in the laboratory, the batch
was typically a mobile slurry. In the plant, the slurry was quite
thick, and some difficulties were encountered during the
transfersprecipitation around the reactor bottom valve com-
plicated the initial transfer of the reactor contents to the filter.
However, these problems were resolved, and normal processing
resumed. The MTBE solution was then passed through a pad
of silica gel, and the solvent was exchanged to ethanol at
atmospheric pressure. A sample of the concentrate was assayed
and was shown to contain 12.3 kg of 5, which represented a
97% yield from 4. The concentrate was diluted with ethanol,
combined with Pearlman’s catalyst, and reduced at 65-70 °C
and 70 psig hydrogen. In-process analysis (GC) after 8 h showed
that starting material was not present. However, in addition to
the expected product, a new peak was observed that accounted
for 17 area %. Data generated using GC/MS and NMR
supported that the impurity was the N-ethyl derivative of 1.
Precipitation of 1 as the HCl salt did not purge any of the
N-ethyl derivative, and a total of 8 kg of material was isolated.
Fortunately, 1 was an early intermediate en route to the API,
and the presence of the impurity did not interfere with the
downstream chemistry. The only adjustment was to ensure that
the charge of 1 in the next step was corrected for the wt %
purity of 1.
The presence of the impurity, especially at such a high level,
was surprising and an investigation into the root cause followed.
The sample of the concentrate from the batch that had been
tested for in-process assay yield determination had also been
examined for residual solvents. GC analysis showed there to
be 0.59% and 0.53% acetonitrile and MTBE present respec-
tivelyslevels that were within the specifications set for the pilot-
plant batch based on past laboratory experiments. There was
some speculation that the ethylation could have been related to
acetaldehyde in the ethanol, either as a contaminate or from
oxidation of the solvent.18 The concentrate from the plant was
reduced using the same batch of catalyst and solvent used during
the scale-up run. The pilot-plant cycle time, including times
for nitrogen purges and hydrogen charges, was mimicked to
ensure that prolonged contact of solvent, substrate, and catalysts
before the introduction of hydrogen was not the cause of the
side reaction. The reductions in the lab were cleansno N-ethyl
impurity was formed when the same amounts of substrate,
solvent, and catalyst were combined and reduced in the
laboratory.
The only time we were successful in generating the N-ethyl
impurity was when acetonitrile was added to the reaction
mixture prior to performing the hydrogenation. In fact, ethy-
lation of 1 was quite efficient as the relationship between the
level of N-ethyl impurity and the molar equivalents of aceto-
nitrile present was found to be linear (Figure 6).19 In light of
our inability to repeat the side reaction when pilot-plant materials
(18) Andersen, D.; Storz, T.; Liu, P.; Wang, X.; Li, L.; Fan, P.; Chen, X.;
Allgeier, A.; Burgos, A.; Tedrow, J.; Baum, J.; Chen, Y.; Crockett,
R.; Haung, L.; Syed, R.; Larsen, R. D.; Martinelli, M. J. Org. Chem.
2007, 72, 9648–9655.
(14) Connolly, T. J.; Constantinescu, A.; Lane, T. S.; Matchett, M.;
McGarry, P.; Paperna, M. Org. Process Res. DeV. 2005, 9, 837–842.
(15) Luo, F.-T.; Jeevanandam, A. Tetrahedron Lett. 1998, 39, 9455–9456.
(16) Dixon, R.; Peterson, D. Anal. Chem. 2002, 74, 2930–2937.
(17) Medved, A.; Dormon, F.; Kaufman, S.; Pocher, A. J. Aerosol Sci.
2000, 31, S616–S617.
(19) Reports to generate oxazolidines from ephedrine and nitriles under
comparable reductive conditions are known. See: Henin, F.; Letinois,
S.; Muzart, J. Tetrahedron Lett. 1997, 38, 7187–7190.
Vol. 14, No. 2, 2010 / Organic Process Research & Development
•
463

follows: TsCl (2.61 min), 1 (5.80 min), benzylamine (7.97 min),
toluene (10.87 min), 5 (12.58 min), 4 (16.18 min). Melting
points are uncorrected. Commercially available reagents and
solvents were used without purification.
(Tetrahydrofuran-2,5-diyl)dimethanol (3). To a 2 L Parr
stirred pressure reactor, 5-(hydroxymethyl)furan-2-carbaldehyde
(2) (50 g, 397 mmol), Raney nickel 2800 (38 g, 1.5 equiv),
and methanol (1500 mL, 30 parts v/w)20 were added. The
reactor was flushed with nitrogen three times, then pressurized
with hydrogen to 70 psig and adjusted to 60 °C. The reaction
mixture was stirred at 600 rpm and 60 °C as the hydrogen
pressure was maintained at 70 psig until the conversion to 3
was greater than 95% (AN GC). The reaction solution was
cooled to 25 °C and filtered through Celite (50 g, 1 part w/w).
The reactor and Celite cake were rinsed forward with acetonitrile
(400 mL, 313 g, 8 parts v/w). Caution! Spent catalyst is
pyrophoric! Filtration must be performed under a blanket
of nitrogen. Catalyst bed should be rinsed with water and
kept water-wet prior to exposure to air and disposal. The
combined filtrate was concentrated to a volume of approxi-
mately 200 mL (∼4 parts v/w) under vacuum. The concentrate
of 3 was methanol/acetonitrile, ratio >2/98. An aliquot of the
mixture was concentrated to dryness and provided a clear,
slightly yellow liquid: 1H NMR (400 MHz, d6-DMSO) δ 1.61
(m, 2H), 1.8 (m, 2H), 3.33 (t, J ) 8 Hz, 4H), 3.83 (m, 2H),
4.57 (t, J ) 8 Hz, 2H) ppm. 13C NMR (400 MHz, d6-DMSO)
(major isomer): δ 27.15, 64.10, 79.69 ppm; (minor isomer)
27.36, 63.64, 79.18 ppm.
meso-(Tetrahydrofuran-2,5-diyl)bis(methylene)bis(4-me-
thylbenzenesulfonate) (4). To the concentrate of 3 in aceto-
nitrile obtained above was added TsCl (227 g, 3 equiv) then
pyridine (408 g, 13 equiv). Caution: Exotherm! The resulting
solution was stirred at 20 °C until the conversion to 4 was
greater than 98%. To the reaction mixture was added water (2
kg, 20 parts w/w) Caution: A significant exotherm is
observed during the initial water charge. The resulting
suspension was filtered. The filtered cake was washed with
water (2 × 250 g, 2 × 5 parts w/w), then with EtOH (150 g,
3 parts w/w), and dried to afford 4 as a white crystalline solid
with 97.2% purity (AN HPLC) in 71% yield from 2. 1H NMR
(400 MHz, d6-DMSO) δ 1.50 (m, 2H), 1.84 (m, 2H), 2.42 (s,
6H), 3.84 (m, 2H), 3.96 (m, 4H), 7.47 (d, J ) 8.0 Hz, 4H),
7.77 (d, J ) 8.0 Hz, 4H) ppm. mp 127.0-128.0 °C (lit.10
127.5-128.0).
Figure 6. Relationship between N-ethyl impurity level and
acetonitrile content.
and processing times were used, we concluded that the impurity
was formed through the inadvertent introduction of deleterious
acetonitrile. A review of the batch record from the pilot plant
revealed that several pieces of equipment that had acetonitrile
contact (pumps, hoses, holding pots, etc.) were used repeatedly
during the processing. However, not one piece stood out as the
potential source of the acetonitrile contamination. The most
obvious point where this could have occurred was during the
transfer from the concentrator to the hydrogenation vessel;
however, that remains speculation on our part.
Although discovered after the fact, the performance in the
pilot plant highlighted the need to have strict in-process controls
for the acetonitrile content. It is now recommended to condition
all equipment that had acetonitrile contact with ethanol prior
to introducing the hydrogenolysis feed-stock solution of 5. We
also recommend that the solvent exchange from acetonitrile to
ethanol be continued until the acetonitrile content in the ethanol
stock solution is reduced to below the limit of detection using
conventional GC-FID.
Conclusion
An efficient synthesis of 8-oxa-3-aza-bicyclo[3.2.1]octane
hydrochloride has been developed and demonstrated in the lab.
Minor modifications to literature conditions resulted in a process
that was easily scaled up. The process described was scaled up
to a pilot plant and was executed on a 10 kg scale. The overall
strength-corrected yield of 1 from 2-hydroxymethylfurfural was
53%. Unfortunately, upon scale up, some N-ethylation of the
product was observed during the final hydrogenolysis that
depressed the overall yield. It is now highly recommended that
the acetonitrile content of the hydrogenolysis stock solution be
reduced to below the level of detection when performing the
solvent exchange, and all equipment that had acetonitrile contact
be conditioned with ethanol before allowing contact with the
solution of 4.
Pilot-Plant Performance. The transformation of 2 to 4 was
performed in a 300 gallon vessel set and used the following
charges: 5-(hydroxymethyl)furan-2-carbaldehyde (2) (9.8 kg),
Raney nickel 2800 (7.5 kg), and methanol (158 kg). Following
complete reduction, the catalyst was removed and the batch
concentrated as the solvent was exchanged to acetonitrile
through multiple acetonitrile charges and chases (all unde
vacuum). The solution of 3 in acetonitrile was charged to a
300 gallon vessel that contained p-toluenesulfonyl chloride (44.5
kg), and pyridine (80 kg) was added. Water was added to
Experimental Section
General. HPLC analysis of the intermediates and reaction
monitoring was carried out on a Waters Alliance HPLC with
PDA (monitoring at 205 nm) and charged aerosol detectors
equipped with a Primesep 200, 4.6 mm × 150 mm, column (5
µm particle size) at 30 °C. Elution was performed with a flow
rate of 1.5 mL/min and a gradient method using water/
acetonitrile (97:3) w/ 0.025% trifluoroacetic acid (mobile phase
A) and acetonitrile (mobile phase B) (initial: 90%A, 2 min,
90%A, 25 min, 10%A). Observed retention times were as
(20) This reaction can be performed successfully with 8 volumes of
methanol to improve throughput. The reaction was demonstrated at
this scale to support dilution of the process to the point that both
impellers of the dual-flight agitator on the pilot-plant vessel would be
submerged when the process was run on a 10 kg scale.
464
•
Vol. 14, No. 2, 2010 / Organic Process Research & Development

quench the reaction and precipitate the product. Product 4 (28.3
kg, 83% yield) was filtered and dried on a 0.25 m2 Rosenmund
filter/dryer. Filtration and drying of 4 required 18.5 and 66.5 h,
respectively.
heptane (100 mL) and the product pulled dry under nitrogen.
The product was transferred to a vacuum oven and dried at
<50 °C until loss on drying was <2%. Product 1 was isolated
as an off-white solid (25 g, 68% yield). Purity 98.1% (GC-
1
FID), wt % HCl: 23.9% (IC, theory 24.2%) H NMR (400
3-Benzyl-8-oxa-3-azabicyclo[3.2.1]octane (5). To a 5 L
reactor was charged 4 (285 g, 0.6469 mol) and acetonitrile (2.28
L). To the resulting slurry was charged benzylamine (277 g,
2.59 mol, 4 equiv) at room temperature. The slurry was heated
to reflux (80-85 °C) until the reaction was deemed complete
(<2% 4 by HPLC), typically 20-24 h. The reaction mixture
was cooled to 40-45 °C, and p-toluenesulfonic acid (123 g,
0.4494 mol, 1 equiv) was added. The mixture was heated to
reflux, and approximately half of the acetonitrile was distilled
out of the mixture. Toluene (700 mL) was charged, and the
mixture was cooled to ambient temperature. The solids that were
present were removed via filtration, and the collected solids were
washed with toluene (2 × 200 mL). The batch was concentrated
to approximately one-quarter of its original volume (∼400 mL).
The mixture was filtered through a plug of silica gel (290 g)
and washed with 2 × 250 mL of toluene. The filtrate was
concentrated to ∼400 mL of residual volume, and MTBE (400
mL) was added. The mixture was allowed to stir at ambient
temperature for 15-20 min and was clarified and concentrated
to afford 5 as a thick oil, 122 g, 93.5% yield.21 1H NMR (400
MHz, CDCl3) δ 1.85 (m, 2H), 1.97 (m, 2H), 2.33 (dd, J ) 11
Hz, J ) 2 Hz, 2H), 2.53 (ddd, J ) 11 Hz, J ) 1 Hz, J ) 1 Hz,
2H), 3.45 (s, 2H), 4.27 (m, 2H), 7.29 (m, 2H). A sample of the
product was isolated as the hydrochloride salt and used as a
reference standard: Purity 99.8% (AN LC), wt % HCl 14.6%
(IC, theoretical 15.2%); 1H NMR (400 MHz, d6-DMSO) δ 1.85
(m, 2H), 2.28 (dd, J ) 8.0 Hz, 2H), 3.15 (m, 4H), 4.25 (d, J )
6.0 Hz, 2H), 4.42 (d, J ) 1.0 Hz, 2H), 7.40 (m, 3H), 7.75 (m,
2H) ppm, mp 187.1-191.2 °C (lit.9 185-185.5 °C).
8-Oxa-3-azabicyclo[3.2.1]octane hydrochloride (1). Com-
pound 5 (50 g) and ethanol (0.225 kg) were combined with
Pearlman’s catalyst (6.3 g, palladium hydroxide, 50% water-
wet) in a 2 L Parr stirred pressure reactor. The system was
purged of air with nitrogen three times, and hydrogen was
introduced into the system. The hydrogen pressure was adjusted
to 60-65 psig, and the reaction mixture was adjusted to 60-65
°C. The reaction mixture was maintained at 60-65 °C/60-65
psig hydrogen until the reaction was complete (<1% 5 AN GC).
The pressure vessel contents were cooled to 22 °C, flushed with
nitrogen three times and the catalyst was removed via filtration.
Caution: Spent catalyst is pyrophoric! Filtration must be
performed under a blanket of nitrogen. Catalyst bed should
be rinsed with water prior to exposure to air and disposal.
The filtrate was concentrated by atmospheric pressure distilla-
tion. Ethanol was exchanged for heptane until the ethanol level
is less than 2% (AN GC). After cooling to ambient temperature,
2-propanol (51 g) was added and the batch adjusted to 5-10
°C. Chlorotrimethylsilane (40 g) was added as the temperature
was maintained below 10 °C. Product 1 precipitated, and the
slurry was stirred at 5-10 °C for 1-2 h and was collected via
filtration. The flask and filter cake were rinsed forward with
MHz, d6-DMSO) δ 1.90 ppm (m, 2H), 2.10 (m, 2H), 2.81 (d,
J ) 8.0 Hz, 2H), 3.16 (dd, J ) 8.0 Hz, J ) 1.0 Hz, 2H), 4.39
(d, J ) 1.0 Hz, 2H), 9.50 (br, 2H) ppm; 13C NMR (100 MHz,
d6-DMSO) δ 26.48, 46.69, 71.41 ppm; mp 202.6-204.7 °C
(lit.9 202-204 °C).
Pilot-Plant Performance. The transformation of 4 to 1 was
performed in a 300 gallon vessel set and used the following
charges: 4 (27.5 kg), benzylamine (26.7 kg) and acetonitrile
(173 kg). The reactor contents were adjusted to 80-90 °C and
maintained at that temperature until conversion or 4 to 5 was
complete (4% 4 remained after 48 h at ∼84 °C). When the
reaction was complete, a solution of p-toluenesulfonic acid (11.8
kg) in acetonitrile (36 kg) was added to the reaction mixture,
and the resulting mixture was concentrated at atmospheric
pressure from approximately 340 to 115 L. To the concentrate
was added MTBE (81.5 kg), the resulting slurry was filtered,
and the reactor and cake were rinsed forward with MTBE (122
kg). The filtrates were concentrated at atmospheric pressure from
an original volume of approximately 400 L to a minimum
stirrable volume (∼ 20 L). The batch was cooled to 20-26
°C, and MTBE (20 kg) was added. The mixture was clarified,
and ethanol (65 kg) was added to the filtrate. The solution was
concentrated at atmospheric pressure from an initial volume of
∼115 L to ∼40 L. The reactor contents were adjusted to
ambient temperature, and ethanol (20 kg) was added. The
mixture was combined with Pearlman’s catalyst (1.5 kg, 50%
water-wet) in an inerted hydrogenation vessel, adjusted to 65
°C and 60-65 psig hydrogen, and maintained under these
conditions until the reaction was complete. The vessel was
cooled to ambient temperature, the atmosphere was purged of
hydrogen, and the catalyst was removed via filtration. The vessel
and filter were rinsed forward with ethanol (12.5 kg). The
combined filtrates were transferred to a reactor set up for
atmospheric distillation and the reactor contents concentrated
from an initial volume of approximately 140 L to approximately
114 L. Heptane (143 kg) was charged, and the concentration
was repeated as necessary until GC analysis showed there to
be NMT 2.0% ethanol present. The mixture was then adjusted
to ambient temperature and 2-propanol (12.5 kg) was added.
The mixture was adjusted to 4-10 °C, and chlorotrimethylsilane
(9.8 kg) was added. The resulting slurry was held for ap-
proximately 30 min, filtered on a 0.25 m2 Rosenmund filter/
dryer, rinsed with heptane (2 × 24 kg), and dried under vacuum
at 40-50 °C until in-process LOD was NMT 3%. After 20 h,
the LOD was 0.2%, and the filter was discharged to afford 1
(8.0 kg, purity 75% (wt %), 64% yield (corrected for wt %
purity of product)).
Acknowledgment
We are grateful to the technical staff from the Rouses Point
Pilot Plant for their contributions to this project. We are also
grateful to Mettler Toledo for providing a MonARC system
for demonstration purposes and to Dr. Wes Walker for technical
assistance with the MonARC system.
(21) As noted in the text, we highly recommend that this solvent
exchange to ethanol be continued until acetonitrile can no longer
be detected by GC before telescoping the ethanol stream of 5
forward. Additionally, all processing equipment that had acetonitrile
contact should be conditioned with ethanol before contact with this
ethanol stream of 5.
Received for review October 15, 2009.
OP9002642
Vol. 14, No. 2, 2010 / Organic Process Research & Development
•
465