A highly enantioselective phosphabicyclooctane catalyst for the kinetic resolution of benzylic alcohols

Article abstract of DOI:10.1021/ja021224f

A new class of chiral phosphines belonging to the P-aryl-2-phosphabicyclo[3.3.0]octane family (PBO) has been prepared by enantioselective synthesis starting from lactate esters and 2,2-dimethylcy-clopentanone enolate 5. A selective enolate alkylation method has been developed for preparation of 9 and 10 using a chelating ester substituent in the triflate alkylating agent 11. Subsequent conversion to the PBO catalysts 2 and 39 relies on a diastereoselective cyclization from the cyclic sulfate 17 and LiPHAr to afford the more hindered endo-aryl phosphines. These phosphines function as efficient catalysts for the kinetic resolutions of aryl alkyl carbinols by benzoylation (16, 21, 22) or iso-butyroylation in the case of the less hindered aryl alkyl carbinol substrates. With o-substituted aryl alkyl carbinols, the enantioselectivities exceed 100, and s = 380 ± 10 has been demonstrated in the case of methyl mesityl carbinol. The PBO- catalyzed acylations probably involve a P-acylphosphonium carboxylate intermediate and a tightly ion paired transition state.

Full text of DOI:10.1021/ja021224f

Published on Web 03/14/2003
A Highly Enantioselective Phosphabicyclooctane Catalyst for
the Kinetic Resolution of Benzylic Alcohols
Edwin Vedejs*,† and Olafs Daugulis‡
Contribution from the Chemistry Department, UniVersity of Wisconsin,
Madison, Wisconsin 53706, and Department of Chemistry,
UniVersity of Michigan, Ann Arbor, Michigan 48109
Received September 30, 2002; E-mail: edved@umich.edu
Abstract: A new class of chiral phosphines belonging to the P-aryl-2-phosphabicyclo[3.3.0]octane family
(PBO) has been prepared by enantioselective synthesis starting from lactate esters and 2,2-dimethylcy-
clopentanone enolate 5. A selective enolate alkylation method has been developed for preparation of 9
and 10 using a chelating ester substituent in the triflate alkylating agent 11. Subsequent conversion to the
PBO catalysts 2 and 39 relies on a diastereoselective cyclization from the cyclic sulfate 17 and LiPHAr to
afford the more hindered endo-aryl phosphines. These phosphines function as efficient catalysts for the
kinetic resolutions of aryl alkyl carbinols by benzoylation (16, 21, 22) or iso-butyroylation in the case of the
less hindered aryl alkyl carbinol substrates. With o-substituted aryl alkyl carbinols, the enantioselectivities
exceed 100, and s ) 380 ( 10 has been demonstrated in the case of methyl mesityl carbinol. The PBO-
catalyzed acylations probably involve a P-acylphosphonium carboxylate intermediate and a tightly ion paired
transition state.
In 1996, we reported the first examples of enantioselective
acyl transfer reactions catalyzed by a chiral phosphine. These
experiments demonstrated that 1-phenyl-trans-2,5-dimethylphos-
pholane activates m-chlorobenzoic anhydride for the chloroben-
zoylation of alcohols with significant enantioselectivity (s )
other hand, bicyclic derivatives based on the 2-phosphabicyclo-
[3.3.0]octane (PBO) skeleton proved to be remarkably more
reactive even though they are more hindered than their mono-
cyclic analogues.7-10 The phosphine-catalyzed acylations in-
volve P-acyl phosphonium carboxylate intermediates,11 and
the high reactivity of 1 appears to be due to a preference for
P-phenyl rotamers that allow easy access to the unshared
electron pair at phosphorus.
The gem-dimethyl catalyst 1 was exceptionally promising in
terms of reactivity and also gave encouraging enantioselectivity
results in acylations, but the racemic catalyst-borane complex
had to be laboriously separated into enantiomers by HPLC.10
Clearly, an enantioselective synthesis was needed before ex-
tensive effort to optimize variables would be worthwhile. Modi-
fications of the scheme developed for preparation of 1 from
1,1-dimethylcyclopentanone were therefore considered that
k
FAST/kSLOW ) 13-15).1 There had been many prior attempts
to develop chiral nucleophilic acylation catalysts,2 but this was
the first case where a nonenzymatic catalyst was shown to react
with s > 10.3 However, the reaction was very slow, so a search
for more reactive catalysts was initiated. While our study was
in progress, reports from other groups began to appear describing
chiral nitrogen nucleophiles that have reached impressive levels
of enantioselectivity.4,5
Between 1996 and 1999, work in our laboratory encountered
a variety of mono- and disubstituted phospholanes that were
not significantly better than the original lead structure.6 On the
(5) (a) Ruble, J. C.; Latham, H. A.; Fu, G. C. J. Am. Chem. Soc. 1997, 119,
1492. (b) Ruble, J. C.; Tweddell, J.; Fu, G. C. J. Org. Chem. 1998, 63,
2794. Bellemin-Laponnaz, S.; Tweddell, J.; Ruble, J. C.; Breitling, F. M.;
Fu, G. C. Chem. Commun. (Cambridge) 2000, 1009. (c) Kawabata, T.;
Nagato, M.; Takasu, K.; Fuji, K. J. Am. Chem. Soc. 1997, 119, 3169. (d)
Spivey, A. C.; Fekner, T.; Spey, S. E. J. Org. Chem. 2000, 65, 3154. Spivey,
A. C.; Maddaford, A.; Fekner, T.; Redgrave, A. J.; Frampton, C. S. J. Chem.
Soc., Perkin Trans. 1 2000, 3460. (e) Sano, T.; Miyata, H.; Oriyama, T.
Enantiomer 2000, 5, 119. Sano, T.; Imai, K.; Ohashi, K.; Oriyama, T. Chem.
Lett. 1999, 265. Oriyama, T.; Hori, Y.; Imai, K.; Sasaki, R. Tetrahedron
Lett. 1996, 37, 8543. (f) Jarvo, E. R.; Copeland, G. T.; Papaioannou, N.;
Bonitatebus, P. J., Jr.; Miller, S. J. J. Am. Chem. Soc. 1999, 121, 11638.
Copeland, G. T.; Jarvo, E. R.; Miller, S. J. J. Org. Chem. 1998, 63, 6784.
(6) Vedejs, E.; Daugulis, O. LatVian J. Chem. 1999, 1, 31.
† University of Michigan.
‡ University of Wisconsin.
(1) Vedejs, E.; Daugulis, O.; Diver, S. T. J. Org. Chem. 1996, 61, 430.
(2) Wegler, R. Liebigs Ann. Chem. 1932, 498, 62. Wegler, R.; Ru¨ber, A. Chem.
Ber. 1935, 68, 1055. Bird, C. W. Tetrahedron 1962, 18, 1. Weidmann, R.;
Horeau, A. Bull. Soc. Chim. Fr. 1967, 117. Stegman, W.; Uebelhart, P.;
Heimgartner, H.; Schmid, H. Tetrahedron Lett. 1978, 19, 3091. Mukaiyama,
T.; Tomioka, I.; Shimizu, M. Chem. Lett. 1984, 49. Ichikawa, J.; Asami,
M.; Mukaiyama, T. Chem. Lett. 1984, 949. Duhamel, L.; Herman, T.
Tetrahedron Lett. 1985, 26, 3099. Weidert, P. J.; Geyer, E.; Horner, L.
Liebigs Ann. Chem. 1989, 533. Potapov, V. M.; Dem’yanovich, V. M.;
Klebnikov, V. A.; Korovina, T. G. Zh. Org. Khim. 1986, 22, 1218 (1095
Engl). Evans, D. A.; Anderson, J. C.; Taylor, M. K. Tetrahedron Lett. 1993,
34, 5563. Ishihara, K.; Kubota, M.; Yamamoto, H. Synlett 1994, 611.
(3) Reviews of enzyme-catalyzed acyl transfer: Sih, C. J.; Wu, S.-H. Top.
Stereochem. 1989, 19, 63. Chen, C.-S.; Sih, C. J. Angew. Chem., Int. Ed.
Engl. 1989, 28, 695. Klibanov, A. M. Acc. Chem. Res. 1990, 23, 114.
Roberts, S. M. Chimia 1993, 47, 85.
(7) Preliminary communication: Vedejs, E.; Daugulis, O. J. Am. Chem. Soc.
1999, 121, 5813.
(8) Vedejs, E.; Daugulis, O.; MacKay, J. A.; Rozners, E. Synlett 2001, 1499.
(9) (a) Vedejs, E.; MacKay, J. A. Org. Lett. 2001, 3, 535. (b) Vedejs, E.;
Rozners, E. J. Am. Chem. Soc. 2001, 123, 2428.
(10) Vedejs, E.; Daugulis, O.; Harper, L. A.; MacKay, J. A.; Powell, D. R. J.
Org. Chem. 2003, 68, in press.
(11) Vedejs, E.; Diver, S. T. J. Am. Chem. Soc. 1993, 115, 3358.
(4) Reviews: Jarvo, E. R.; Miller, S. J. Tetrahedron 2002, 58, 2481. Spivey,
A. C.; Maddaford, A.; Redgrave, A. J. Org. Prep. Proced. Int. 2000, 32,
331.
9
4166
J. AM. CHEM. SOC. 2003, 125, 4166-4173
10.1021/ja021224f CCC: $25.00 © 2003 American Chemical Society

An Enantioselective Phosphabicyclooctane Catalyst
A R T I C L E S
Scheme 1
tion had occurred via enolization of 8 during the time that the
alkylation product was warmed to room temperature. However,
this complication could be avoided by taking care to control
the temperature prior to and during the reduction step, resulting
in good selectivity for 9. The stereochemistry of the major
product was assigned on the basis of X-ray crystallography of
intermediates later in the synthesis.
The next problem was to access the diastereomeric diol 10
by inverting enolate facial selectivity in the alkylation step. An
analysis of possible transition states for alkylation of enolate 5
by the ethyl lactate triflate 4 suggested that the major diaste-
reomer 8 may be formed by attack from the top face of enolate
5 as shown in the proposed transition state 6. This model
assumes that the triflate leaving group is displaced with inversion
of configuration and places the smallest substituent (hydrogen)
of the alkylating agent over the cyclopentene ring. Given these
constraints, structure 8 is the expected product if the ester group
is positioned to avoid the enolate oxygen (6). Presumably, this
arrangement minimizes unfavorable electronic interactions as
well as nonbonded steric interactions.
By the same logic, access to the other enolate face (attack
from below) would be expected if the apparent tendency of the
ethyl ester to avoid enolate oxygen could be reversed by
modifying the ester alkoxy group. The triflate 11 suggests a
possible solution, based on the notion that the methoxyethoxy-
ethyl ester group might have an affinity for lithium ion. To test
this possibility, 11 was prepared from methyl (S)-lactate by a
sequence involving transesterification with methoxyethoxyetha-
nol followed by triflic anhydride/pyridine. The triflate was
formed as a somewhat unstable liquid which had to be used on
the same day due to decomposition.
The alkylation of enolate 5 with 11 was explored in solvents
of varying lithium coordination ability. Generation of the enolate
in toluene using LDA followed by addition of 11 and immediate
reduction as in the ethyl ester case gave the desired, inverted
product ratio. The best result was obtained from -55 to -60
°C (ca. 10:1 10:9), although the yield was in the 30-40% range.
Better yields could be achieved at higher temperatures, but at
the expense of a decreased diastereomer ratio. When the
experiment was repeated in the more coordinating THF as a
solvent, the opposite ratio (1:10 10:9) was obtained in low yield
(10-15%), supporting the involvement of lithium coordination
in the toluene experiment. While the yield in toluene was
modest, the possibility of alkylating either face of the enolate
by using an ester group that is capable of lithium coordination
was successfully demonstrated. Subsequent experiments estab-
lished that the alkylation proceeds with clean inversion of
configuration at the triflate-bearing carbon, and without detect-
able racemization.
might allow the synthesis of analogues such as 2 (Ph-PBO) or
3 in single enantiomer form. In principle, the only change needed
would be to use a chiral lactate triflate 4 in place of the achiral
ethyl glycolate triflate alkylating agent in the reaction with the
lithium enolate (5) of 2,2-dimethylcyclopentanone.12a-c Depend-
ing on the degree of stereoselectivity in the enolate alkylation
step, this approach was expected to provide access to either or
both of the diastereomeric phosphines via ketoesters 7 or 8 and
diols 9 or 10 (Scheme 1).
Enolate Alkylations Studies Using Lactate Triflates
The alkylation experiments involving 4 and 5 were challeng-
ing due to the ease of product equilibration. After considerable
experimentation, the following reaction conditions were found
to be optimal. Thus, 5 was reacted with 4 from -55 to -60 °C
in THF followed by immediate reduction of the crude ketoesters
7 and 8 with LiAlH4 at -70 °C to give 9 and 10 in ratios as
high as 10-15:1. If the intermediate ketoester mixture (presum-
ably 10-15:1 8:7) was isolated and then subjected to reduction,
then a 1:1 mixture of 9:10 was obtained. Most likely, equilibra-
A proposed transition state for the alkylation using 11 is
illustrated by structure 12. The enolate lithium ion may be
coordinated to all three oxygens of the methoxyethoxyethyl
carboxylate. With hydrogen placed beneath the cyclopentane
ring to minimize steric interactions, the result is to position the
electrophile for alkylation of the desired face of the enolate.
Complexation involving the lithium ion is presumably broken
up in THF, resulting in the same sense of diastereoselectivity
as with the ethyl lactate triflate reagent via the geometry 6.
(12) â-Keto ester lithium enolates12a,b and acyloxazolidinone lithium enolates12c
have been alkylated with lactate triflates. We could find no case of the use
of lactate triflates for alkylation of simple ketone enolates. (a) Hoffman,
R. V.; Kim, H.-O. Tetrahedron Lett. 1993, 34, 2051. (b) Hoffman, R. V.;
Kim, H.-O. J. Org. Chem. 1995, 60, 5107. (c) Decicco, C. P.; Nelson, D.
J.; Corbett, R. L.; Dreabit, J. C. J. Org. Chem. 1995, 60, 4782.
We could find no prior examples where enolate facial selec-
tivity in analogous alkylations has been inverted by modifying
9
J. AM. CHEM. SOC. VOL. 125, NO. 14, 2003 4167

A R T I C L E S
Vedejs and Daugulis
Scheme 2
Phosphine 3 is the less stable diastereomer in terms of phos-
phorus configuration and proved to be somewhat sensitive to
epimerization at phosphorus.16 Thus, deprotection of 14 with
pyrrolidine at 50 °C always gave a trace of an isomeric phos-
phine, presumably having the exo-P-phenyl orientation. Reaction
times in the deprotection experiments had to be carefully
controlled to minimize this complication.
Preliminary evaluations of phosphine 3 as an acylation
catalyst indicated somewhat lower reactivity and enantioselec-
tivity compared to 1. Thus, a value of s ) 8.9 was found for
the m-chlorobenzoylation of 16 using 3 as the catalyst compared
to s ) 14 using 1 in the corresponding reaction. Further attention
was therefore focused on the diastereomer 2 (Scheme 2).
The synthesis of 2 (Ph-PBO) from the crude diol 10 used
the same treatment with SOCl2 followed by oxidation to afford
the cyclic sulfate 17 (21% yield overall from 2,2-dimethylcy-
clopentanone). Subsequent reaction of 17 with LiPHPh and
complexation using BH3-THF allowed the isolation of a borane
complex with a remarkable 36:1 diastereomer ratio of 18:19 in
favor of the desired endo-phenyl epimer. In the corresponding
cyclization step in the synthesis of 1, a 16:1 ratio of diastere-
omers had been obtained in favor of the endo phenyl isomer.10
The diastereomers 18 and 19 were separated by column
chromatography, and phosphine 2 was liberated by briefly
warming 18 with pyrrolidine. The absolute and relative con-
figuration of 2 was confirmed at the stage of 18 by X-ray
crystallography. In contrast to the borane complex of 1,10 the
crystal lattice of 18 contains only one conformer. The ee of 18
was 96-98% if nonrecrystallized cyclic sulfate 17 was used in
the synthesis, or 99.7% ee if 17 was recrystallized prior to use.
Alternatively, a single recrystallization of 18 with g96% ee
raised the ee to >99.9%.
the lithium coordination ability in the ester alkoxy substituent.
Possibly related examples of inverted facial selectivity have been
reported in alkylations of sulfoxide anions (MeI gave opposite
facial selectivity as compared to the alkylation with trimethyl
phosphate)13a and amide enolates13b,c (opposite facial selectivities
for alkylations with alkyl halides versus epoxides).13b Lithium
interactions with the different leaving groups have been invoked
in these examples.
With both diastereomers 10 and 9 available in enantiomeri-
cally enriched form, attention was turned to the synthesis of 2
and 3 (Scheme 2). The precedented14 synthesis of cyclic sulfate
13 from diol 9 was uneventful, and reaction of 13 with LiPHPh
followed by BH3-THF afforded the epimeric borane complexes
14 and 15 in a ratio of 7.7:1. The diastereomers were separated
by column chromatography, and the endo-phenyl phosphine 3
was liberated by brief warming with pyrrolidine.15 This isomer
has the less hindered exo orientation for the unshared electron
pair at phosphorus and is expected to be the most reactive
nucleophilic catalyst.10 The enantiomeric excess of 3 was
checked at the stage of 14 and was found to be >99.9% after
a single recrystallization, even though the starting lactate was
no better than 97% ee. The relative and absolute configuration
of 14 was then confirmed by X-ray crystallography, confirming
that the reaction of the 2,2-dimethylcyclopentanone enolate 5
with ethyl (S)-lactate triflate 4 had occurred with inversion.
In anticipation of a lengthy optimization study for catalyst
2, a qualitative method was sought that would allow a rapid
estimate of enantioselectivity by NMR assay of a mixture
containing 13CH3-labeled 20 and the enantiomeric 12CH3 alcohol
(Scheme 3). Thus, (S)-13CH3-20 was prepared from the labeled
2-acetonaphthone with (-)-DIP chloride,17 while unlabeled (R)-
20 was made by reduction of unlabeled 2-acetonaphthone using
(+)-DIP chloride. The two quasi-enantiomers were easily
distinguished because of the large 13C-CH coupling, resulting
in a doublet of doublets for the 13CH3 group of (S)-13CH3-20,
spaced symmetrically on both sides of the simple doublet due
to the 12CH3 group of unlabeled (R)-12CH3-20.18a The NMR
sample of the quasi-racemate 20 was then partially benzoylated
using benzoic anhydride and catalyst 2, and the resulting mixture
1
of benzoate ester and alcohol was monitored over time by H
NMR integration. The methyl region was well resolved, and
separate signals due to the quasi-enantiomeric benzoate esters
and unreacted quasi-enantiomeric alcohols were observed,
allowing assay for conversion as well as % ee of starting
material and product from a single 1H NMR spectrum by
(16) Mislow, K. Trans. N. Y. Acad. Sci. 1973, 35, 227.
(17) Brown, H. C.; Park, W. S.; Cho, B. T. J. Org. Chem. 1986, 51, 1934.
(18) (a) A conceptually similar method using 13C NMR assay to measure ee
values and conversion was reported during preparation of this manuscript:
Evans, M. A.; Morken, J. P. J. Am. Chem. Soc. 2002, 124, 9020. (b) Guo,
J.; Wu, J.; Siuzdak, G.; Finn, M. G. Angew. Chem., Int. Ed. 1999, 38, 1755.
Reetz, M. T.; Becker, M. H.; Klein, H.-W.; Sto¨ckigt, D. Angew. Chem.,
Int. Ed. 1999, 38, 1758. Diaz, D. D.; Yao, S.; Finn, M. G. Tetrahedron
Lett. 2001, 42, 2617.
(13) (a) Chassaing, G.; Lett, R.; Marquet, A. Tetrahedron Lett. 1978, 19, 471.
(b) Myers, A. G.; McKinstry, L. J. Org. Chem. 1996, 61, 2428. (c) Askin,
D.; Volante, R. P.; Ryan, K. M. Tetrahedron Lett. 1988, 29, 4245.
(14) Burk, M. J.; Feaster, J. E.; Nugent, W. A.; Harlow, R. L. J. Am. Chem.
Soc. 1993, 115, 10125.
(15) Wolfe, B.; Livinghouse, T. J. Am. Chem. Soc. 1998, 120, 5116.
9
4168 J. AM. CHEM. SOC. VOL. 125, NO. 14, 2003

An Enantioselective Phosphabicyclooctane Catalyst
A R T I C L E S
Scheme 3
Table 2. Acylation of Alcohols R1R2CHOH with (RCO)2O Using
Catalyst 2a
1
2
entry alcohol anhydride R
R
R
relative rateb
s
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
16 Ph
16 Ph
16 Ph
Ph
Ph
Ph
t-Bu
t-Bu
t-Bu
t-Bu
t-Bu
t-Bu
t-Bu
1
24
1 (heptane) 21
0.1 (-40 °C) 67
0.1 (CH2Cl2) 14
0.02
0.6
16 m-ClC6H4 Ph
16 1-naphthylc Ph
25
16
4.9
23
22
3.7
15
13
7
13
5.8
16 Me
16 i-Pr
21 Ph
22 Ph
23 Ph
24 Ph
24 i-Pr
20 Ph
Ph
Ph
Ph
0.01
1-adamantyl 0.6
4-MeOC6H4 t-Bu
Ph-ethynyl t-Bu
Ph
Ph
0.8
8
12
0.1
24
2
10
NR
3
0.04
2
17
5
3
8
4
i-Pr
i-Pr
2-naphthyl Me
20 1-naphthylc 2-naphthyl Me
20 2-naphthylc 2-naphthyl Me
20 phthalic
20 i-Pr
20 t-Bu
25 i-Pr
26 Ph
26 i-Pr
27 i-Pr
28 i-Pr
29 i-Pr
30 i-Pr
31 i-Pr
2-naphthyl Me
2-naphthyl Me
2-naphthyl Me
1-naphthyl Me
Ph
Ph
Ph
Ph
C6F5
c-C3H5
i-Pr
12
6.9
21
9.8
11
3.0
5.9
3.9
1
n-Bu
n-Bu
CO2Me
CF3
Me
Me
6
0.1
Me
1
a All reactions used 0.13 M substrate in toluene with 2.5 equiv of
anhydride at room temperature unless noted. b The relative rate (as compared
to entry 1) was estimated by calculating rate constants from individual %
conversion and time data points, and assuming pseudo-first-order conditions
and linear behavior with respect to the catalyst. The (R) alcohol was the
faster-reacting enantiomer for entries 1-21. Assignments were not made
for the less selective reactions (entries 22-26). c The naphthoic anhydride
was used as a stirred suspension.
Table 1. Solvent Effects in Benzoylation of (S)-13CH3-20/
(R)-12CH3-20 by NMR Assay (Room Temperature)
solvent
enantioselectivity
C6D6
toluene-d8
CD3CN
CD2Cl2
EtOAc
s ) 8-9
s ) 8-9 (7)a
s ) 5
Acylations of representative aryl alkylcarbinols catalyzed by
2 were investigated in the optimal hydrocarbon solvents, as
summarized in Table 2. The hindered substrate 16 was used
for the initial optimization studies, and several anhydrides were
evaluated. The highest enantioselectivity at room temperature
was obtained using 1-naphthoic anhydride (entry 5), but the
reagent was inconvenient because of limited solubility, lack of
a commercial source, and low reactivity requiring many hours
at room temperature. Benzoic anhydride was far more reactive,
and nearly as enantioselective as 1-naphthoic anhydride, while
iso-butyric anhydride (entry 7) was the least reactive anhydride
tested with 16. Enantioselectivity was considerably lower with
iso-butyric anhydride (s ) 4.9) and intermediate with acetic
anhydride (s ) 16; entry 6). The benzoic anhydride procedure
was briefly evaluated with other hindered aryl t-alkyl carbinol
substrates at room temperature, and comparable results were
obtained with 2119a and 2219b as with 16. On the other hand, an
alkynyl tert-butyl alcohol 2319c reacted with low selectivity
(entry 10), suggesting that the aryl substituent plays a large role
in recognition of the enantiomers.
s ) 6-7 (5.5)a
s ) 7.5
C6F6
s ) 8-9
s ) 6-7
s ) 7
s ) 7-9
s ) 5-6
acetone-d6
3-methyl-3-pentanol
THF-d8
DMF-d7
a Value of s determined by HPLC assay.
comparison of integral intensities. The results are summarized
in Table 1. Enantioselectivities were somewhat overestimated
using NMR, judging from standard HPLC methods applied to
the toluene and dichloromethane entries (s values in parenthe-
ses). The source of the discrepancy is a combination of errors
in integration, alcohol ee values, and relative proportions of the
quasi-enantiomers, but the trends were clear. The reactions were
faster and more enantioselective in the nonpolar, aromatic
solvents (toluene, benzene). Saturated hydrocarbon solvents
could not be monitored conveniently by NMR due to solubility
problems with the benzoic anhydride.
The effort needed to prepare the isotopically enriched, quasi-
racemic mixture for NMR assay was worthwhile as long as s
values were modest, and numerous data points were required
for the same substrate-anhydride combination. Mass spectros-
copy provides an alternative for the rapid assay of isotopically
labeled mixtures using an automated approach,18b but the NMR
technique was convenient for our purposes even though it is
imprecise. However, optimization studies soon revealed condi-
tions resulting in much higher s values that required the precision
of HPLC assay.
The best kinetic resolution result with 16 was obtained with
benzoic anhydride in toluene solution at -40 °C (s ) 67).
Although the optimized experiment involves a temperature drop
of 60 °C from the room temperature conditions, the benzoylation
rate decreased by only ca. 8-fold. Similar behavior had been
(19) (a) Hellmann, G.; Beckhaus, H.-D.; Ru¨ckhardt, C. Chem. Ber. 1979, 112,
1808. (b) Rogers, E. F.; Brown, H. D.; Rasmussen, I. M.; Heal, R. E. J.
Am. Chem. Soc. 1953, 75, 2991. (c) Borden, W. T. J. Am. Chem. Soc.
1970, 92, 4898. (d) Armesto, D.; Horspool, W. M.; Manchen˜o, M. J.; Ortiz,
M. J. J. Chem. Soc., Perkin Trans. 1 1992, 2325.
9
J. AM. CHEM. SOC. VOL. 125, NO. 14, 2003 4169

A R T I C L E S
Vedejs and Daugulis
observed for the simpler PBO catalyst 1,10 resulting from an
unusual combination of low activation enthalpy and large,
negative activation entropy. This pattern has proven to be
general for all of the PBO catalysts encountered in this study.
A number of other aryl alkyl carbinols were resolved with
good selectivity using catalyst 2 in toluene (Table 2, entries
11-21). However, the optimum match of the anhydride reactant
with catalyst 2 for these substrates proved to be different as
compared to the substrates containing a tertiary alkyl substituent.
In the representative example of 2-naphthyl(-1-ethanol), 20, the
enantioselectivity of kinetic resolution using iso-butyric anhy-
dride (entry 17; s ) 12) was higher than that with benzoic
anhydride (entry 13), and almost as high as in the reaction with
1-naphthoic anhydride (entry 14; s ) 13). Because the benzo-
ylations were considerably faster, the benzoic anhydride method
was used with a number of other substrates. Temperature
optimization was not explored because a better catalyst was
discovered by varying the P-aryl group, as described below.
However, the reactions using 2 at room temperature helped to
establish qualitative enantioselectivity and reactivity trends that
have proven to be general.
Scheme 4
Rates of acylation were estimated from % conversion versus
time data by assuming pseudo-first-order kinetics10 and are
tabulated in Table 2 relative to the benzoylation of 16 (entry
1). The relative rate numbers are approximate,20 but they are
close enough to provide a qualitative ordering of substrate and
anhydride reactivity. Thus, the reactions using benzoic anhydride
were at least an order of magnitude faster than the iso-
butyroylations, an effect that was also seen with the simpler
catalyst 1. Increasing electron withdrawal in the substrate (entries
22-24) increased rates, while steric hindrance in the alcohol
or in the anhydride reagent resulted in considerably slower
reactions. Good enantioselectivity was observed if an aromatic
ring was attached to the hydroxyl-bearing carbon, but simple
steric differentiation of the alcohol substituents was not sufficient
for enantioselectivity, as shown in entries 25 and 26.
complex was isolated in 85% yield. The key cyclization step
was incredibly selective and gave a 91:1 ratio in favor of the
desired diastereomer 37, ca. 20% overall from 2,2-dimethyl-
cyclopentanone. The ee of 37 could not be measured with high
precision due to HPLC overlap of trace impurities with the
minor enantiomer, but the ee of the recrystallized precursor
sulfate 17 could be assayed by conversion to 18 (99.7% ee for
the batch of 17 used for most of the work described below), so
the same ee value was assumed for 39. Phosphine 39 was then
tested in acylations, as summarized in Table 3. It became
apparent that the new catalyst is superior to all of the PBO
derivatives described earlier, except for the case of the highly
hindered alcohol 16 where the combination of catalyst 2 and
benzoic anhydride gave better results (Table 2, entry 3). Aryl
alkyl carbinols that contain n-alkyl or sec-alkyl substituents were
resolved more efficiently using 39, and iso-butyric anhydride
was the best match for this catalyst. In all cases where direct
comparisons were made (see Supporting Information), the sense
of enantioselection was the same. Thus, the catalyst configu-
ration shown for 39 results in the preferential acylation of the
(R)-alcohol for the entries of Table 3.
The Synthesis and Evaluation of Me2Ph-PBO (36)
The phenyl group in Ph-PBO (2) was replaced by 3,5-
dimethylphenyl to evaluate acylation selectivity. The synthesis
was uneventful and followed the earlier precedent (Scheme 4),
although the diastereoselectivity in the cyclization from 17 was
somewhat improved compared to the Ph-PBO series (dr 43:1
34:35 versus 36:1 18:19). Deprotection of 34 in the usual way
with pyrrolidine provided the new catalyst 36. Acylations
catalyzed by 36 were somewhat faster compared to those using
2, but there was no significant advantage in enantioselectivity
in several examples at room temperature: benzoylation of 16,
s ) 8.1, toluene; iso-butyroylations of 20, s ) 14, toluene; 24,
s ) 14, heptane; and 32 (1-phenylethanol), s ) 12, heptane.
The exploration of this series was terminated, and no optimiza-
tion was attempted because a superior catalyst was discovered
as described in the next section.
The Synthesis and Evaluation of t-Bu2Ph-PBO (39)
Enantioselectivities were improved in heptane compared to
toluene, so the toluene conditions were used only where
necessary due to alcohol solubility. In the case of 1-phenyl-1-
pentanol (26, entries 16, 17), alcohol solubility in heptane was
sufficient to vary the concentration over a reasonable range.
This example was used to confirm that s at -40 °C is not
The synthesis of 39 (t-Bu2Ph-PBO) was carried out by the
usual method from 17 (Scheme 4), and the corresponding borane
(20) Reference 10 shows that pseudo-first-order conditions are satisfied using
6 equiv of the anhydride, while 2.5 equiv was used in experiments for
enantioselectivity determination.
9
4170 J. AM. CHEM. SOC. VOL. 125, NO. 14, 2003

An Enantioselective Phosphabicyclooctane Catalyst
A R T I C L E S
Table 3. Acylation of Alcohols R1R2CHOH with (RCO)2O Using
Catalyst 39a
39. At -40 °C, these alcohols were resolved with s in the range
of 100 or above.
anhydride
Early experiments with 41 gave an empirical s ) 39 at room
temperature or s ) 142-145 at -40 °C, but the catalyst 39
had been made from a batch of cyclic sulfate 17 known to have
99.7% ee as described earlier. With such high selectivity, we
wondered how s would be affected by the enantiomeric
contaminant (ent-39). Ismagilov’s mathematical treatment for
correcting s to the value expected for the pure catalyst (100%
ee) became available as these experiments were conducted,21
and the calculation indicated a corrected s ) 180-185. To test
the calculation, the catalyst 39 was recrystallized, and the
experiment was repeated. In this case, an empirical (uncorrected)
value of s ) 188 was determined, in excellent agreement with
the calculation if the recrystallized catalyst has >99.9% ee.
The correction is very sensitive to the precision of ee assay
for the impure catalyst. For example, if the empirical s ) 145
had resulted using a catalyst sample having 99.6% ee, then the
calculated enantioselectivity corrected to 100% catalyst ee
increases to s ) 204 versus s ) 185 for the assumption that the
catalyst has 99.7% ee. There is less sensitivity to catalyst ee,
and a smaller calculated correction to s, as the enantioselectivity
decreases. Below s ) ca. 40, the difference between the
empirical and the corrected s values becomes insignificant for
catalyst having 99.7% ee.
The o,o-disubstituted substrate 2,4,6-trimethylphenyl methyl
carbinol (42) also proved to be interesting. The room-temper-
ature experiments were not exceptional, and good selectivity
was observed in toluene (s ) 15) as well as in heptane (s )
22). Because of solubility limitations, the more promising
heptane conditions could not be used at lower temperatures, so
the toluene conditions were explored. This resulted in the largest
temperature effect seen for any of the alcohol substrates, as well
as the highest enantioselectivity. Starting with catalyst believed
to be 99.7% enantiomerically pure, an empirical enantioselec-
tivity s ) 220 was measured. Assuming eecat ) 99.7%, the
corrected21 value for 100% ee catalyst would be s ) 328.
However, if catalyst ee is taken as a slightly lower number
(99.6% ee), then the corrected enantioselectivity increases to s
) 392. Because the effect of catalyst purity is so large, the
experiment was repeated using the same batch of recrystallized
catalyst demonstrated to have >99.9% ee in the acylation of
41. In two experiments, enantioselectivities were determined
as s ) 370 and s ) 390 at -40 °C in toluene (Table 3, entries
36, 37). The calculated numbers are highly sensitive to error in
the measured ee values, but good agreement in s was possible
because the enantiomers of 42 are well separated by HPLC on
a chiral support and can be assayed with high precision. The
enantioselectivity with 42 is the highest value known to date
for a nonenzymatic acyl transfer reaction.
1
2
entry alcohol
R
R
R
solv/tempb
tol/RT
s
1
2
3
4
5
6
7
8
9
16
16
24
24
24
24
20
20
20
Ph
i-Pr
Ph
i-Pr
i-Pr
i-Pr
Ph
Ph
Ph
Ph
Ph
Ph
t-Bu
t-Bu
i-Pr
i-Pr
i-Pr
i-Pr
Me
Me
Me
Me
Me
Me
Me
10
9.8
6.9
18
tol/RT
tol/RT
tol/RT
hept/RT
20
hept/-40 °C 99 (117)c
n-Pr 2-naphthyl
i-Pr 2-naphthyl
t-Bu 2-naphthyl
i-Pr
i-Pr
i-Pr
Me
tol/RT
tol/RT
tol/RT
tol/RT
hept/RT
9.2
17
12
18
22
10 33
11 33
12 33d
13 33e
14 43
15 43
16 26
17 26
18 32
19 32f
20 28
21 28f
22 25
23 25
24 25
25 25
26 40
27 40
28 41
29 41
30 41g
31 41h
32 42
33 42
34 42
35 42
36 42h
37 42h
38 42e,h
Ph
Ph
Ph
Ph
Ph
Ph
Ph
Ph
Ph
Ph
Ph
Ph
hept/-20 °C 42 (45)c
tol/-40 °C 11
i-Pr
i-Pr
i-Pr
i-Pr
i-Pr
i-Pr
i-Pr
i-Pr
Ph
i-Pr
i-Pr
i-Pr
i-Pr
i-Pr
i-Pr
i-Pr
i-Pr
i-Pr
i-Pr
i-Pr
i-Pr
i-Pr
i-Pr
i-Pr
Me
CH2Cl hept/RT
CH2Cl hept/-20 °C 19
n-Bu hept/RT 18
n-Bu hept/-40 °C 55 (60)c
12
i-Bu
i-Bu
CF3
CF3
Me
Me
Me
Me
Me
Me
Me
Me
Me
Me
hept/RT
hept/-40 °C 31
hept/RT 7.5
14
hept/-20 °C 12
1-naphthyl
1-naphthyl
1-naphthyl
1-naphthyl
2-MeOC6H4
2-MeOC6H4
2-MeC6H4
2-MeC6H4
2-MeC6H4
2-MeC6H4
2,4,6-Me3C6H2 Me
2,4,6-Me3C6H2 Me
2,4,6-Me3C6H2 Me
2,4,6-Me3C6H2 Me
2,4,6-Me3C6H2 Me
2,4,6-Me3C6H2 Me
2,4,6-Me3C6H2 Me
tol/RT
tol/RT
hept/RT
6.0
35
42 (44)c
tol/-40 °C 99 (116)c
hept/RT
38
tol/-40 °C 82 (94)c
hept/RT
39 (41)c
hept/-40 °C 145 (185)c
hept/-40 °C 142 (180)c
hept/-40 °C 188h
hept/RT
tol/RT
22
15
tol/-10 °C 87 (100)c
tol/-40 °C 220 (328)c
tol/-40 °C 390h
tol/-40 °C 370h
tol/-40 °C 112h
a All reactions used 0.13 M substrate, 2.5 equiv of anhydride, and 38
with 99.7% ee unless noted. The (R) alcohol was the more reactive
enantiomer in all of the secondary alkyl aryl carbinols, and recoveries of
alcohol and ester were routinely >90%. b Solvents tol ) toluene; hept )
heptane; RT ) room temperature. c Selectivity corrected21 to 99.7% ee of
catalyst (the correction becomes significant at s > 40). d Concentration 0.033
M due to solubility limitation. e 0.6 equiv of anhydride was used. f Con-
centration 0.05 M. g Experiment using 41 on 7.5 mmol scale. h Recrystal-
lized catalyst 39, >99.9% ee.
affected by changes in concentration (over the range from 0.033
to 0.2 M) or % conversion (from 36 to 51%), and consistent
values of s ) 55 ( 2 were obtained in several experiments.
These findings support the kinetic assumptions mentioned
earlier.
As before, benzoylations were at least 10-fold faster than the
iso-butyroylations, but the latter were considerably more enan-
tioselective. Overall, the reactivity trends for the examples of
Table 3 were similar to those of Table 2. Thus, the effect of
temperature on the reaction rate was relatively small, and a 60
°C drop in temperature was tolerated with a rate decrease of
severalfold in typical examples. For example, the iso-butyro-
ylation of 41 in heptane at -40 °C was 5 times slower compared
to the room-temperature experiment (entries 28, 29).
In view of the high enantioselectivity in the iso-butyroylation
of 42, a similar experiment was carried out using acetic
anhydride. As shown in Table 3, entry 38, reasonably high
enantioselectivity was observed (s ) 112), in contrast to less
hindered alkyl aryl carbinols. For example, the acetylation of
1-phenyl-1-ethanol (33, entry 13) under similar conditions was
far less selective (s ) 11).
In the most dramatic advantage for catalyst 39, the enantio-
selectivity increased substantially as the temperature was
lowered. Alcohols such as 25, 40, 41, containing an o-substituted
phenyl group, showed the largest temperature dependence on
enantioselectivity and were among the best substrates for catalyst
The other entries in Table 3 show that typical aryl alkyl
carbinols are resolved with good to excellent enantioselectivity
(21) Ismagilov, R. F. J. Org. Chem. 1998, 63, 3772.
9
J. AM. CHEM. SOC. VOL. 125, NO. 14, 2003 4171

A R T I C L E S
Vedejs and Daugulis
using t-Bu2Ph-PBO (39). Almost any alkyl group is tolerated
(entries 4-19), with the best selectivity for R ) i-Pr. As
mentioned earlier, the R ) t-Bu example (entries 1, 2) is unusual
because better results are obtained with the P-phenyl catalyst
2. For primary or secondary alkyl groups, 39 is the more
effective catalyst. However, cyclic aryl alkyl carbinols such as
1-indanol (44) or 1-tetrahydronaphthol (45) are not resolved (s
< 1.5). Furthermore, the homoallylic alcohol 46 is not a good
substrate for acylation (s ) 1), nor is the tertiary alcohol 47 (s
) 1.3).
One other substrate category does react with useful enanti-
oselectivity. In a preliminary experiment, the allylic alcohol 48
reacted with s ) 52 at -40 °C. This topic has been discussed
in another publication from our laboratory, and kinetic resolu-
tions of a number of other allylic alcohols have been
demonstrated.9a These findings underscore the importance of a
double bond attached to the hydroxyl-bearing carbon for
enantioselection in the acyl transfer process.
We cannot comment in detail on the origins of enantiose-
lectivity without knowing other important details of transition
state geometry. In particular, a knowledge of carboxylate
placement and orientation would be important. Without evidence
bearing on this issue, specific proposals regarding the geometry
for acyl transfer would not be justified. We do know that cyclic
benzylic alcohols such as 44 and 45 are constrained to
geometries that do not fit the preferences of the PBO catalyst
39, judging from the negligible enantioselectivity in these
examples. High enantioselectivity requires a flexible, uncon-
strained π-system adjacent to the hydroxyl group. This could
be taken to imply that a π-stacking interaction is important
between the aryl rings of the substrate and the catalyst,25 but
several arrangements are possible where such effects could be
invoked. These geometries cannot be distinguished from the
evidence available at this time.
The lactate-based enantioselective synthesis of the new PBO
catalysts introduces an additional methyl substituent on the
phospholane ring compared to the original lead compound 1.
Modeling studies suggest that two distinct conformations of 1
are likely (designated as conformers A and C in ref 10) and
that they have similar energies, but rather different environments
near phosphorus.10 The corresponding conformers of 2 or 39
are shown in Scheme 4 as the structures 2-A or 2-C, and 39-A
or 39-C. Conformers in the C-series (2-C or 39-C) accept the
pseudoequatorial methyl substituent in the phospholane ring with
no significant change in geometry along the bicyclic (PBO)
framework. This geometry corresponds closely to the conforma-
tion of 18 determined by X-ray crystallography and is regarded
as the most likely ground-state geometry for 2 and 39. A similar
PBO geometry is plausible in the transition state for catalytic
anhydride activation because structures such as 2-C or 39-C
maintain an easily accessible unshared electron pair at phos-
phorus.
Conformers similar to 2-A or 39-A are less likely to serve as
the reactive catalysts because the phospholane methyl group
would be closer to the unshared electron pair at phosphorus.
Furthermore, the P-aryl substituent is closer to the plane of
maximum unshared electron pair density in the A-series of
conformers. However, a small (ca. 10-15°) change in the P-aryl
dihedral angles would be enough to obscure this difference
between the A and C conformers, while a modification of
phospholane twist envelope conformations would accommodate
the methyl substituent in a more nearly pseudoequatorial
orientation. Such geometries may be accessible and cannot be
ruled out as participants in the acyl transfer process. More rigid
structures will need to be investigated to further clarify transition
state preferences.
A substantial effort was invested to probe the relationship
between catalyst purity and enantioselectivity.21,26 The magni-
tude of the correction for the most selective reactions (Table 3;
entries 29-31, 34-37) initially surprised us, but the explanation
is simple enough. Samples of 39 having 99.7% ee contain 0.15%
of ent-39, a contaminant that is the best match for catalytic
conversion of the (S)-carbinols in Table 3 (the slower reacting
substrates for 39). The competing pathway resulting from
catalysis by the contaminant lowers the enantiomeric purity of
Discussion
A new class of chiral phosphines belonging to the P-aryl-2-
phosphabicyclo[3.3.0]octane family (PBO) has been prepared
by enantioselective synthesis starting from methyl lactate and
2,2-dimethylcyclopentanone. In the course of the synthesis, an
enolate alkylation method has been developed that allows
reversal of the alkylation facial selectivity by introducing a
chelating ester substituent in the electrophile. Subsequent
conversion to the PBO catalysts 2, 3, and 39 relies on a
diastereoselective cyclization from the cyclic sulfates and
LiPHAr reagents to afford the more hindered endo-aryl phos-
phines. These phosphines have proven to be efficient catalysts
for the kinetic resolutions of aryl alkyl carbinols and allylic
alcohols9a by benzoylation (16, 21, 22) or iso-butyroylation in
the case of the less hindered aryl alkyl carbinol substrates. With
o-substituted aryl alkyl carbinols, the enantioselectivities exceed
100, and s ) 370-390 has been measured in the case of methyl
mesityl carbinol. For the best substrates, the enantioselectivities
using PBO catalysts have reached the levels usually associated
with enzymatic acyl transfer reactions.3 Several other nonen-
zymatic acyl transfer kinetic resolution catalysts have recently
been reported that function with potentially useful enantiose-
lectivity for the best substrates.4,5
As discussed elsewhere,10 the PBO-catalyzed acylations are
believed to involve a tight ion pair transition state, derived from
a P-acylphosphonium carboxylate intermediate. Additional
evidence consistent with this proposal is provided by a number
of examples tabulated using the new catalysts 2, 3, and 39 where
the fastest reactions are observed in the least interactive solvent
(heptane). This finding suggests that tight ion pairing in the
acylation transition states allows the most effective participation
by carboxylate anion in the eventual proton transfer that must
occur as the alcohol is recognized, deprotonated, and acylated.
However, we cannot specify the geometric details of the
transition state, nor is it clear whether the reaction involves a
cyclic transition state, a conventional tetrahedral intermediate,22-24
or a concerted displacement of the phosphonium leaving group
by the alcohol oxygen acting as the nucleophile.
(25) Hunter, C. A.; Sanders, J. K. M. J. Am. Chem. Soc. 1990, 112, 5525.
(26) Luukas, T. O.; Girard, C.; Fenwick, D. R.; Kagan, H. B. J. Am. Chem.
Soc. 1999, 121, 9299. Johnson, D. W.; Singleton, D. A. J. Am. Chem. Soc.
1999, 121, 9307. Blackmond, D. G. J. Am. Chem. Soc. 2001, 123, 545.
(22) Williams, A. Acc. Chem. Res. 1989, 22, 387.
(23) Hengge, A. C.; Hess, A. C. J. Am. Chem. Soc. 1994, 116, 11256.
(24) Marlier, J. F. Acc. Chem. Res. 2001, 34, 283.
9
4172 J. AM. CHEM. SOC. VOL. 125, NO. 14, 2003

An Enantioselective Phosphabicyclooctane Catalyst
A R T I C L E S
the ester product as well as the unreacted alcohol relative to
the values expected for pure catalyst. The result is to lower the
uncorrected value calculated for s. The difference between
empirical and corrected values of s becomes especially striking
as the inherent enantioselectivity becomes exceptionally high.
When reactions are so selective that even the impure catalyst
affords products with an apparent value of s ) 220 (entry 35),
the use of purified catalyst is not so important in the practical
context because the enantioselectivity is already good enough
for most purposes. We emphasize the corrected s values
primarily because this allows a more meaningful comparison
with the analogous enzymatic acyl transfer reactions. Presum-
ably, natural acyl transfer catalysts have .99.9% ee. Optimized,
highly selective synthetic catalysts would be at a disadvantage
in the comparisons unless they are used as enantiomerically pure
substances.
Acknowledgment. This work was supported by the National
Science Foundation. The authors also thank Mr. J. A. MacKay
for the data of Table 2, entry 6, and Dr. R. Ismagilov for his
interest in developing the mathematical treatment of kinetic
resolutions using impure catalyst.
Supporting Information Available: Experimental procedures
and tabulated data for kinetic resolutions (PDF). This material
is available free of charge via the Internet at http://pubs.acs.org.
JA021224F
9
J. AM. CHEM. SOC. VOL. 125, NO. 14, 2003 4173