Regio- and enantioselective allylic alkylation of an unsymmetrical substrate: A working model

Article abstract of DOI:10.1021/ja9828713

The evolution of a model for understanding asymmetric allylic alkylations catalyzed by palladium with the use of ligands derived from chiral diamines and 2-diphenylphosphinobenzoic acid provides a basis for attacking the problem of regio- and enantioselec

Full text of DOI:10.1021/ja9828713

VOLUME 121, NUMBER 19
MAY 19, 1999
© Copyright 1999 by the
American Chemical Society
Regio- and Enantioselective Allylic Alkylation of an Unsymmetrical
Substrate: A Working Model
B. M. Trost* and F. D. Toste
Contribution from the Department of Chemistry, Stanford UniVersity, Stanford, CA, 94305-5080
ReceiVed August 10, 1998
Abstract: The evolution of a model for understanding asymmetric allylic alkylations catalyzed by palladium
with the use of ligands derived from chiral diamines and 2-diphenylphosphinobenzoic acid provides a basis
for attacking the problem of regio- and enantioselective alkylations proceeding through the intermediacy of
1-monosubstituted allyl complexes. The model predicted that in the kinetic ionization of an achiral precursor
the major enantiomer of the product resulting from attack at the more substituted terminus would be the mirror
image of that obtained under Curtin-Hammett conditions. Experimentally, the ee was rationally varied from
66% of one enantiomer to 83% of the mirror image using the same ligand. Nonpolar solvents and the absence
of counterions that coordinate to palladium favor the kinetic product. More polar solvents and counterions that
coordinate well to palladium favor Curtin-Hammett conditions. For maximum regio- and enantioselectivity,
the chiral racemic 3-substituted-1-alkene is the preferred substrate.
Selectivity, especially enantioselectivity, continues to be a
major area of focus in organic chemistry. Transition metal
catalyzed enantioselective reactions are among the most power-
ful in this regard.1 In general, the enantioselectivity in metal
catalyzed reactions is derived from the differentiation of
enantiotopic faces of a π-system. This has been achieved with
the π-systems of olefins (hydrogenation, hydroboration, epoxi-
dation, cyclopropanation, cycloadditions), carbonyl compunds
(hydride reductions, organozinc additions), and enolates (aldol
reactions, alkylations, protonations). Unlike most transition metal
catalyzed processes, allylic alkylations do not rely solely on a
single mechanism as a source of asymmetry.2 Although dif-
ferentiation of olefin faces is one possible mechanism, the source
of the enantioselectivity is complicated by the possibility that
one or more other steps in the catalytic cycle may be the
enantiodiscriminating step(s). Furthermore, the choice of transi-
tion metal (Pd, Ir, W, Mo, Rh)3,4 may also have an effect on
the mechanism of enantioselectivity.
The mechanism by which the catalyst imposes its chirality
upon the product for metal catalyzed allylic alkylations remains
obscure in most cases. Most discussions focus on the nature of
the steric interactions in a static model derived from a knowledge
of the structure of the π-allylmetal intermediate. In devising
our family of chiral ligands, we envisioned the creation of chiral
space. In this model, the sum of the interactions associated with
the motions going from η2- to η3-metal complexes or the reverse
are envisioned to dictate the chirality of the product. Unfortu-
nately, all attempts to characterize directly the intermediates
either by X-ray crystallography or NMR spectroscopy have
(3) For lead references dealing with regioselectivity, see: Mo: Trost,
B. M.; Lautens, M. J. Am. Chem. Soc. 1987, 109, 1469. W: Trost, B. M.;
Tometzki, G. B.; Hung, M.-H. J. Am. Chem. Soc. 1987, 109, 2176. Ir:
Takeuchi, R.; Kashio, M. Angew. Chem., Int. Ed. Engl. 1997, 36, 263. Rh:
Evans, P. A.; Nelson, J. D. J. Am. Chem. Soc. 1998, 120, 5581.
(4) For enantioselective allylic alkylations, see: Mo: Trost, B. M.;
Hachiya, I. J. Am. Chem. Soc. 1998, 120, 1104. W: Lloyd-Jones, G. C.;
Pfaltz, A. Angew. Chem., Int. Ed. Engl. 1995, 34, 462. Ir: Janssen, J. P.;
Helmchen, G. Tetrahedron Lett. 1997, 38, 8025.
(1) (a) Noyori, I. Asymmetric Catalysis in Organic Synthesis; Wiley: New
York, NY, 1994. (b) Ojima, I. Catalytic Asymmetric Synthesis; VCH:
Weinheim, Germany, 1993.
(2) (a) Trost, B. M.; Van Vranken, D. L. Chem. ReV. 1996, 96, 395. (b)
Trost, B. M. Acc. Chem. Res. 1996, 29, 355.
10.1021/ja9828713 CCC: $18.00 © 1999 American Chemical Society
Published on Web 04/30/1999

4546 J. Am. Chem. Soc., Vol. 121, No. 19, 1999
Trost and Toste
Table 1. Solvent Effects with Achiral Substrate 2 and Phenol 1a
(X ) Och3)a
failed. Thus, we have been confronted with learning more about
the nature of the process by indirect, albeit circumstantial,
evidence.
entry solvent EtN(30) 3:4 4a(R) 4a(S) % ee t1/2 (min)
1,3-Symmetrically substituted substrates are among the most
studied and successful in allylic alkylations. The success of these
substrates lies in the fact that from both enantiomers of the chiral
racemic substrate the same symmetrical π-allyl intermediate is
accessible. This transforms the problem of enantioselectivity
to one of regioselectively attacking one of the enantiotopic
terminal carbons of the allyl. A more challenging case is that
of the unsymmetrically 1-monosubstituted π-allyl intermediate
(eq 1). In this case the allyl intermediate no longer possesses a
1
2
3
4
toluene
THF
CH2Cl2
0.099 60:40
0.207 48:52 12.7 39.3 51 (S)
0.309 42:58 32.5 25.5 12 (R)
6.8 33.2 66 (S)
45
240
5
CH3CN 0.460 23:77 50.1 26.9 30 (R)
a Reaction performed with ligand and palladium complex as outlined
in eq 2.
intermediates are diastereomeric and that the nucleophilic
addition step can be considered to be the microscopic reverse
of the ionization step. If ionization of A in the presence of a
chiral complex preferentially generates C, then path a would
be the expected “matched” route for nucleophilic attack from
C. To the extent that ionization of ent-A with this same chiral
complex to form D is unfavorable, nucleophilic attack, according
to path c, would be less favorable or “mismatched” compared
to path a. Furthermore, it may be argued that the complex that
preferentially ionizes A to form C will preferentially ionize the
regioisomeric substrate B to form D (vide infra). The fact that
path b represents the microscopic reverse of that ionization then
suggests that, for nucleophilic attack on D, path b should be
favored or matched compared to path c. This scenario suggests
that, to control regioselectivity of attack, it is necessary to control
which of the two complexes, C or D, is attacked, i.e., chiral
ligands can influence regioselectivity in ways not possible with
achiral ligands. In this paper, we report a catalytic system that
shows these expectations can be realized experimentally.
plane of symmetry, and thus the question of enantioselectivity
is much more complex. A second issue arises from the fact that
the two terminal carbons of the π-allyl intermediate are no
longer equivalent; one is a primary carbon while the other is
secondary. Several factors will determine the regioselectivity
of nucleophilic attack on a monosubstituted intermediate.5,6
Steric factors will direct attack to the less sterically demanding
primary site, while electronic factors tend to favor attack at the
more electropositive secondary carbon. Which of these domi-
nates depends on several factors, including the choice of metal,
the R group in the allyl substrate, and the nucleophile. This
paper describes factors inherent to chiral ligands, which are not
present in achiral ones, that allow control of the regioselectivity
of nucleophilic attack in palladium catalyzed allylic alkylations.
Furthermore, if ideal conditions can be created, reactions that
are both regio-and enantioselective may become possible. This
study was also envisioned to provide additional insight into the
enantiodiscriminating step.
Results
As a test case, a system which had been reported to provide
poor regioselectivity with achiral ligands was chosensthe allylic
alkylation of phenols, illustrated by the reaction of 4-methoxy-
phenol (1a) with methylhexenyl carbonate (2) (eq 2).7 The
In examining eq 1, if the metal fragment is achiral, the issues
affecting the regioselectivity for the two enantiomeric intermedi-
ates C and D will be identical. When the metal is palladium,
the normal bias in such a case leads to preferential formation
of the achiral product via path b. On the other hand, if the metal
fragment is chiral, then the intermediates C and D are diastereo-
meric, and the factors affecting regioselectivity for C may be
different than those for D. Using the model for the ligands
developed in these laboratories wherein the motions associated
with going from an η2- to η3-complex during nucleophilic attack
are responsible for the chiral recognition leads to the conclusion
that the differences in these two complexes may lead one
diastereomeric complex to favor attack via path a (or, alterna-
tively, path c) and the other via path b (vide infra).
π-allyl generated from this substrate would bear a propyl
substituent which, usually for steric reasons, generally disfavors
attack at the secondary site, although for electronic reasons
attack at the secondary carbon would be favored. Furthermore,
this system would allow the electronic and steric effects of the
nucleophile to be readily examined by simply changing the
substituents on the phenol.
A second aspect that must be considered derives from the
facts that, in the presence of chiral ligands, both the alkene-
metal complexes before ionization as well as the π-allyl
The first parameter we chose to examine was the solvent.
These experiments clearly show that a change in solvent polarity
resulted in a change in both the regio- and enantioselectivity of
the allylic alkylation reaction (Table 1). Furthermore, the change
in regioselectivity followed the same trend as the change in
enantioselectivity. That is, the more polar solvents favored the
formation of the more substituted product 4 and at the same
time showed an increase in the formation of 4a(R). As seen
(5) Trost, B. M.; Verhoeven, T. R. J. Am. Chem. Soc. 1980, 102, 4730.
(6) Recently, reversals in regioselectivity have been reported for 1-aryl-
prop-2-enyl acetates using non-C2-symmetric ligands. (a) Pretot, R.; Pfaltz,
A. Angew. Chem., Int. Ed. 1998, 37, 323. (b) Hayashi, T.; Kawatsura, M.;
Uozumi, Y. Chem. Commun. 1997, 561. Also see: (c) Hayashi, T.; Kishi,
K.; Yamamoro, A.; Ho, Y. Tetrahedron Lett. 1990, 31, 1743. (d) Hayashi,
T.; Kawatsura, M.; Uozumi, Y. J. Am. Chem. Soc. 1998, 120, 1681. (e)
Trost, B. M.; Krische, M. J.; Radinov, R.; Zanoni, G. J. Am. Chem. Soc.
1996, 118, 8, 6297.
(7) (a) Goux, C.; Lhoste, P.; Sinou, D. Synlett 1992, 725 (b) Goux, C.;
Massacret, M.; Lhoste, P.; Sinou, D. Organometallics 1995, 14, 4585.

Regio- and EnantioselectiVe Allylic Alkylation
J. Am. Chem. Soc., Vol. 121, No. 19, 1999 4547
Table 2. Phenol Substituent Effects on Regio- and
Enantioselectivity
entry
1 (X)
σ+
solvent
3:4
4 (R) 4 (S) % ee
1
2
3
4
5
6
7
8
OCH3 a
CH3 b
H c
-0.28 toluene 60:40
6.8 33.2 66 (S)
9.5 27.5 47 (S)
-0.14
63:37
0.00
0.24
62:38 13.1 24.9 31 (S)
55:45 20.3 24.7 10 (S)
Cl d
CO2CH3 e
OCH3 a
H c
0.44
36:64 39
35
22 (R)
-0.28 CH3CN 23:77 50.1 26.9 30 (R)
0.00
0.44
40:60 42
37:63 34.7 28.4 10 (R)
18
40 (R)
CO2CH3 e
Therefore, we examined the rate of the reaction in the
aforementioned solvents. The half-lives reported in Table 1
indicate no correlation between the half-life of the reaction and
the solvent polarity. Our results mirror those of Brown who
found turnover rates of 0.4, 8 and 1.2 min in CD2Cl2, THF and
toluene, respectively.11c The lack of correlation between the
regio- and enantioselectivity and rate does not exclude the
possibility that the rate of nucleophilic attack is decreasing with
solvent polarity. It does, however, suggest that the regio- and
enantiodetermining step (nucleophilic attack) is not the rate-
determining step in at least some of these solvents.
Although, in some (or all) solvents, phenol addition may not
be rate-determining, in all cases, the nucleophile is involved in
the regio- and enantiodetermining step. Therefore, changing the
substituent(s) on the phenol should have an effect on the regio-
and enantioselectivity. On the basis of a working hypothesis
(vide infra) that the major enantiomer in the case of toluene
arises from attack on the kinetically formed π-allyl intermediate
and, in acetonitrile, the opposite enantiomer arises from attack
on the equilibrated intermediate (e.g., corresponding to π-allyl
intermediates C and D respectively of eq 1), substituent effects
in the phenols, which would influence nucleophilicity, should
be observed.
If this hypothesis is correct, a decrease in the rate of
nucleophilic attack, while maintaining the rate of interconversion
of the two diastereomeric π-allyl complexes constant, will lead
to an increase in the products arising from trapping of the
nonkinetically formed intermediate. In the case of eq 2, the
prediction means that the amount of the S enantiomer compared
to that of the R enantiomer should decreasesin other words,
the ee should decrease! To probe the effect of decreasing
nucleophilicity, the substituents at the para position of the phenol
were varied (Table 2). In toluene as solvent, as the substituents
became less electron-donating (higher Hammett σ value), the
enantioselectivity deteriorated. It is reasonable to assume that
a decrease in the electron-donating ability of the para substit-
uents of phenols results in a decrease in the nucleophilicity.
Thus, the decrease in enantioselectivity may arise from an
increase in the rate of interconversion of the two diastereomeric
complexes C and D (eq 1) relative to that of nucleophilic attack.
Indeed, if the rate of nucleophilic attack is substantially
decreased (X ) CO2CH3) relative to that of isomerization, even
in toluene, the major enantiomer is the one derived from attack
on the equilibrated intermediate. The enantioselectivities cor-
relate very well with the Hammett σp values (see Figure 2).
The excellent degree of correlation is somewhat surprising if
Figure 1. Effect of solvent polarity on the regio- and enantioselectivity
in the palladium-catalyzed allylic alkyaltion of 2.
from the graph (Figure 1), both of these trends correlate well
with the ETN(30) solvent polarity scale.8
The absolute configuration of the aryl ether 4a(R) was
determined by hydrogenation of the olefin and comparison of
the saturated product with an authentic sample (eq 3). The
saturated ether was prepared by a two-step protocol starting from
2-butanone. Enantioselective reduction of the ketone using
N-methylephedrine-modified lithium aluminum hydride9 fol-
lowed by Mitsunobu reaction afforded the desired ether 7 in
37% ee (by HPLC). This is slightly higher than originally
reported for the reduction of 3-hexanone to 3-hexanol, 18-
20% ee based on rotation. Hydrogenation of 4a (R) (74% ee,
see Table 4, entry 10) afforded the unsaturated ether in 75%
ee. The major isomer was identical to the authentic sample as
determined by HPLC.
Is the change in regio- and enantioselectivity vs solvent
polarity due to a change in the rate of nucleophilc attack? In
analogy to an SN2 reaction, one would expect the rate of
nucleophilic attack to increase with increasing solvent polarity.
However, in the case of allylic alkylation reactions, the analogy
may not hold since both the nucleophile and electrophile are
charged species. Polar solvents may, therefore, stabilize both
reacting partners better than nonpolar solvents. The stabilization
of the reactants by polar solvents may be greater than that felt
in the transition state in which the charge is destroyed. This
would lead to an overall decrease in the rate of nucleophilic
attack as solvent polarity increases. Bosnich proposed that, in
DMF, the step involving attack of the nucleophile on the
π-allylpalladium intermediate was rate-determining.10 If this is
true in all solvents, we should see a correlation between the
overall rates of the reaction and of the nucleophilic attack.
(11) For early transition state arguments, see; (a) Sprinz, J.; Kiefer, M.;
Helmchen, G. Tetrahedron: Asymmetry 1994, 5, 573. (b) Togni, A,;
Burckhardt, U.; Gramlich, V.; Pregosin, P. S.; Salzmann, R. J. Am. Chem.
Soc. 1996, 118, 1031. For late transition state arguements, see; (c) Brown,
J. M.; Hulmes, D. I.; Guiry, P. J. Tetrahedron 1994, 50, 4493. (d)
Steinhagen, H.; Reggelin, M., Helmchen, G. Angew. Chem., Int. Ed. Engl.
1997, 36, 2108. (e) Schaffner, S.; Muller, J. F. K.; Neuburger, M.; Zehnder,
M. HelV. Chim. Acta 1998, 81, 1223.
(8) Reichardt, C. SolVents and SolVent Effects in Organic Chemistry,
2nd ed.; VCH: Weinheim, Germany, 1988.
(9) Hiroi, K.; Kitayama, R.; Sato, S. Pharm. Chem. Bull. 1984, 32, 2678.
(10) (a) Auburn, P. R.; Mackenzie, P. B.; Bosnich, B. J. Am. Chem.
Soc. 1985, 107, 2033. (b) Mackenzie, P. B.; Whelan, J.; Bosnich, B. J. Am.
Chem. Soc. 1985, 107, 2046.

4548 J. Am. Chem. Soc., Vol. 121, No. 19, 1999
Trost and Toste
Table 3. Solvent Effect with Chiral Substrate
entry
solvent
3a:4a
4a (R)
4a (S)
% ee
1
2
3
4
Toluene
THF
CH2Cl2
CH3CN
33:67
33:67
33:67
20:80
40.9
40.9
45.6
61.2
26.1
26.1
21.4
18.8
22 (R)
22 (R)
36 (R)
53 (R)
with increasing solvent polarity (i.e., increasing the relative rate
of interconversion of C and D).
As can be seen in Table 3, the prediction that the major
enantiomer from this reaction should have the R absolute
configuration is validated. Furthermore, as anticipated, an
increase in solvent polarity resulted in both an increase in
enantio- and regioselectivity. The best results were obtained in
acetonitrile which provided the secondary product in an 80:20
ratio with 53% ee. These results reinforce the hypothesis that,
in polar solvents (acetonitrile), the reaction is approaching
Curtin-Hammett conditions.
Figure 2. Phenol substituent effects on enantioselectivity of allylic
alkylation of 8 in toluene.
one considers all of the steps, including deprotonation of the
pronucleophile phenol, involved in the allylic alkylation which
may be influenced by the substituents.
All of the above observations suggest that, to obtain high
enantio- and regioselectivity for the formation of the secondary
product 4, the reaction should be under Curtin-Hammett
conditions if possible; i.e., the rate of equilibration between
π-allyl intermediates must be greater than that of nucleophilic
attack. Therefore, to increase the enantio- and regioselectivity,
either the rate of interconversion must increase or the rate of
nucleophilic attack must decrease. While the proper choice of
solvent moves in the right direction, it is not sufficient. It has
been shown that anionic ligands (such as halides) can promote
the π-σ-π interconversion process via a σ-allyl intermediate
16.12 Åkermark reported that addition of 0.1 equiv of chloride
decreased the equilibration time of syn,anti-π-allyl complexes
to 0.5 h compared to 48 h when tetrafluoroborate was used as
the counterion.13 Recently, Togni has examined this anion effect
on the allylic alkylation using ferrocenyl ligands and found that
the addition of hard anions (such as fluoride) had a beneficial
effect on the enantioselectivity of allylic aminations.14 These
results prompted us to examine the effect of anions on the allylic
alkylation of phenols. Coordinating an anion to palladium to
promote isomerization of C and D leads to the prediction that
addition of a halide anion should result in a more regio- and
enantioselective reaction. Furthermore based upon palladium-
halide bond strengths, the order of effectiveness should be F-
> Cl- > Br- > I-.15
For eq 2, the major enantiomer produced when the allylic
alkylation is carried out in acetonitrile is 4(R). This enantiomer
is presumably produced by nucleophilic attack on the equili-
brated π-allyl intermediate which arises from π-σ-π isomer-
ization of the kinetic π-allyl intermediate. A decrease in the
rate of nucleophilic attack should allow for Curtin-Hammett
conditions to be approached. Thus, when acetonitrile is used as
solvent for the allylic alkylation, a decrease in nucleophilicity
should result in an increase in the enantiomer derived from the
equilibrated π-allyl. Indeed, this is the case when the electron-
donating ability of the para substituent is decreased from -OCH3
to -H (30 and 40% ee, respectively). However, when p-CO2CH3
was used as a substituent, the trend did not hold, and the
enantioselectivity decreased to 10%. Since, in this case, the
phenoxide would be a good leaving group, the decrease may
arise because, in acetonitrile, the reaction may become revers-
ible.
To further examine the same effects discussed above, a study
utilizing the chiral racemic substrate, 1-hexenyl 3-carbonate (8R/
8S) was initiated (eq 4). In this discussion, the possibility of
Table 4 summarizes the series of reactions involving the effect
of halide ions. As predicted, when the reaction was carried out
in methylene chloride, addition of 3% fluoride or chloride (entry
2 and 3) increased the enantioselectivity to 50 and 57% ee from
36% ee when no halide was present. Furthermore, the increase
in enantioselectivity was accompanied by an increase in
regioselectivity. Increasing the amount of chloride present to
30% (1500 mol % relative to that of palladium) resulted in a
further increase in the enantioselectivity to 65% and the
regioselectivity to 80:20. Interestingly, in acetonitrile, the effect
was observed to a much lesser extent. Addition of 3% tetra-n-
ionization to the anti complex is ignored since it normally
represents a higher energy transition state. To consider the
expectations, eqs 1, 2, and 4 must be examined. Arbitrarily,
assume that diastereomeric complex C is the kinetically formed
complex between substrate 2 and chiral ligand 6 and, therefore,
the one that should lead to attack at the primary carbon (eq 1,
path b). Therfore, diastereomeric complex D is the one that
should preferentially lead to the product of secondary attack,
but only after equilibration of the two complexes. Starting from
the chiral substrate 8, 50% of the substrate directly will form
diastereomer D, which preferentially leads to the product of
secondary attack, and only the remaining 50% which forms C
requires equilibration. Since the results with the chiral ligand 6
and substrate 2 show that S-4a is preferentially formed initially
and R-4a after equilibration, the expectation becomes that the
chiral racemic substrate should preferentially form the R isomer
in all cases and that the ee favoring the R isomer should increase
(12) (a) Vrieze, K. Dynamic Nuclear Magnetic Resonance Spectroscopy;
In Jackman, L. M., Cotton, F. A., Eds.; Academic Press: New York, NY,
1975. (b) Crociani, B.; Di Bianca, F.; Giovenco, A.; Boschi, T. Inorg. Chim.
Acta 1987, 127, 169.
(13) Sjogren, M. P. T.; Hansson, S.; Åkermark, B.; Vitagliano, A.
Organometallics 1994, 13, 1963.
(14) Burckhardt, U.; Baumann, M.; Togni, A. Tetrahedron: Asymmetry
1997, 8, 155.
(15) Karapet’yants, M. Kh.; Karapet’yants, M. L. Thermodynamic
Constants of Inorganic and Organic Compounds; Humphrey Science
Publishers: Ann Arbor, MI, 1970; p 208. Barin, I. Thermochemical Data
of Pure Substances, 3rd ed; VCH: Weinheim, 1995; p 1312.

Regio- and EnantioselectiVe Allylic Alkylation
J. Am. Chem. Soc., Vol. 121, No. 19, 1999 4549
Table 4. Effect of Halide Ion on Enantioselectivity in the Reaction
with Chiral Substrate 8 (Eq 4)
entry solvent (concn ((M))
additive
4a:3a % ee (R)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
CH2Cl2 (0.5)
CH2Cl2 (0.5)
CH2Cl2 (0.5)
CH2Cl2 (0.5)
CH3CN (0.5)
CH3CN (0.5)
CH2Cl2 (0.1)
CH2Cl2 (0.1)
CH2Cl2 (0.1)
CH2Cl2 (0.1)
CH2Cl2 (0.1)
CH2Cl2 (0.1)
CH2Cl2 (0.1)
CH2Cl2 (0.1)
67:33
71:29
73:27
80:20
80:20
78:22
86:14
84:16
84:16
84:16
84:16
71:29
88:12
84:16
36
50
57
65
53
57
69
76
83
74
86
59
77
75
3% TBAT
3% Bu4N+ Cl-
30% Bu4N+ Cl-
3% Bu4N+ Cl-
30% Me4N+ Cl-
30% Et4N+ Cl-
30% Bu4N+ Cl-
15% Bu4N+ Cl-
30% Hex4N+ Cl-
30% Bu4N+ ClO4
30% Bu4N+ Br-
30% Bu4N+ I-
-
pair 10 can also undergo π-σ-π isomerization, or it can
collapse to give product. As previously discussed, it is the
relative rates of these two processes that determine the enantio-
and regioselectivity of the allylic alkylation. In the presence of
tetraalkylammonium salts, both ion pairs 9 and 10 can undergo
a metathesis reaction with the ammonium salt to eventually
produce ion pair 11 and the tetraalkylammonium salt of the
phenoxide 13. When X is chloride in 11, it has already been
demonstrated to increase the rate of π-σ-π isomerization.
What is the effect of the tetraalkylammonium counterion in 13?
To study the effect, if any, of the tetraalkylammonium cation,
a series of experiments in which the R-group in the tetraalkyl-
ammonium salt was varied were conducted (see Table 4). The
general trend emerged that enantioselectivity increased as the
size of the R-group increased (cf. Table 4, entries 7-9). We
have previously observed the same trend for allylic alkylations
of malonate in the presence of tetraalkylammonium bromides.17
We propose that the tetrabutylammonium chloride effect is the
result of two complementary effects. The rate of π-σ-π
isomerization is increased in the presence of chloride, and the
rate of nucleophilic attack is decreased by the formation of the
tetrabutylammonium phenoxide 13. This conclusion is in
agreement with our earlier observations that higher alkyl groups
slow the rate of nucleophilic attack of malonate compared to
that of internal isomerization of palladium.17
butylammonium chloride increased the enantioselectivity from
53 to 57% ee. This observation is in accord with the previous
hypothesis that, in acetonitrile, the relative rate of π-σ-π
isomerization is greater than that in methylene chloride. Since
the rate of isomerization is already relatively fast in acetonitrile,
halide has a lesser effect on the rate of isomerization.
Lowering the concentration of the reaction mixture from 0.5
to 0.1 M resulted in a significant increase in the enantioselec-
tivity from 65 to 83% ee. The interconversion of C and D is a
unimolecular process; therefore, its rate should be unaffected
by concentration. On the other hand, nucleophilic attack is
bimolecular, and its rate depends both on the concentration of
the nucleophile and of the palladium π-allyl intermediate.
Lowering the concentration of the reaction mixture should,
therefore, significantly lower the rate of nucleophilic attack
relative to that of interconversion of C and D.
To further probe the hypothesis that chloride was acting as
an additional ligand for palladium, thereby promoting the
π-σ-π interconversion, a noncoordinating anion, perchlorate,
was examined. The use of perchlorate in place of chloride
dropped the enantioselectivity from 83 to 59% ee and the
regioselectivity from 84:16 to 71:29. We also examined the use
of other halides; moreover, our results paralleled those of
Togni14 with the enantioselectivities following the trend Cl- >
Br- > I-, in accordance with decreasing bond strength to
palladium.
These experiments further suggest that by approaching
Curtin-Hammett conditions, good regio- and enantioselectivity
should become independent of the nucleophilicity of the
phenolsi.e., of the substituent on the phenol. To exemplify this
point, the reactions of several phenols with chiral racemic
substrate 8 were performed. As shown in eq 6, both the
Although the enantioselectivity in the presence of tetrabutyl-
ammonium perchlorate (59% ee) was lower than when chloride
(83% ee) was used, it was better than that obtained in its absence
(36% ee). While the possibility of perchlorate behaving as a
weak ligand for palladium (a promoter for the π-σ-π
isomerization) cannot be excluded, this observation may suggest
that the tetraalkylammonium ion was also playing a role in the
enhancement of the enantioselectivity. We have previously
reported an enhancement in enantioselectivity in the allylic
alkylation of malonates in the presence of tetrabutylammonium
bromide.16 For reasons already discussed, in allylic alkylations
where the nucleophilic attack is the enantiodetermining step,
the nucleophile’s countercation may have an effect on the
enantioselectivity. Let us consider the ion pairs which may be
involved in the allylic alkylation of phenols with allyl carbonates
(eq 5). On ionization of allyl carbonates, the first ion pair formed
is that of the palladium π-allyl and the carbonate anion 9. In
the absence of tetraalkylammonium additives, 9 can undergo
π-σ-π isomerization to the diastereotopic allyl, or the carbon-
ate can deprotonate phenol to generate the ion pair 10. The ion
regioselectivity [(81 ( 3):(20 ( 3)] and enantioselectivity (80
( 3) were reasonably constant, suggesting that equilibration is
indeed being seen.
Discussion
The working model for the class of ligands utilized herein is
depicted in Figure 3. This model evolved from our initial
hypothesis regarding the structural features required for creating
chiral space and molecular modeling. Numerous efforts to obtain
direct experimental support have failed. For example, attempts
to grow crystals have been thwarted by the chemical instability
(16) Trost, B. M.; Bunt, R. C. J. Am. Chem. Soc. 1994, 116, 4089. Trost,
B. M.; Bunt, R. C.; Zambrano, J. J. Am. Chem. Soc. 1996, 118, 6520.
(17) Trost, B. M.; Bunt, R. C. J. Am. Chem. Soc. 1998, 120, 70.

4550 J. Am. Chem. Soc., Vol. 121, No. 19, 1999
Trost and Toste
Figure 3. Model of chiral pocket and cartoon representation of complex derived from R,R-ligand 6.
of the π-allylpalladium complexes due to the facility of loss of
the elements of the alkene and HX to give simple palladium(+2)
complexes which are catalytically inactive. In fact, we believe
this event is the turnover-limiting event of the catalytic cycle.
Studies by NMR spectroscopy have been thwarted by the
complexity of the spectra due, in part, to the presence of multiple
species which have precluded assignment of specific signals to
specific complexes both in these laboratories as well as
elsewhere.18 Thus, understanding the nature of the asymmetric
inducing event must currently rely on circumstantial evidence
of studies of the type reported herein.
The cartoon model (Figure 3) derives from the ground-state
structure of the ligand-palladium-π-allyl complex based on
molecular modeling structures. In this model, the walls represent
the chiral space created by the propeller-like array of the phenyl
rings; the raised flaps represent the phenyls which lie in a plane
approximately parallel to the allyl, while the lowered flaps
represent phenyls which are somewhat perpendicular to the allyl.
One factor that must be addressed in developing a working
model for allylic alkylation is the preferred geometry of the
leaving group in the ionization reaction and the preferred
trajectory of the nucleophile in the alkylation reaction. Let us
consider the transition states for these processes. Based on the
Hammond postulate, the structure of the transition state can be
inferred from the stable structure on either side of the transition
state that lies closest in energy to the transition state. The smaller
the energy differences between the transition state and the
intermediate, the greater the structural similarities. Since the
nucleophilic attack step wherein charge on carbon is neutralized
is presumably exothermic, the transition state resembles, to some
extent, the π-allyl intermediate.11 Furthermore, the transition
state for ionization also should resemble the π-allyl intermediate,
since this process, which creates charge on carbon, is presumably
endothermic. Since these processes are, to a first approximation,
the microscopic reverse of each other, the factors which
influence one will do so to the other. Support for the underlying
assumptions derives from the fact that, while carboxylates are
good nucleophiles toward π-allylpalladium cationic complexes,
any equilibrium lies heavily on the side of the allyl ester and
the palladium(0) complex.
Figure 4. Transition-state trajectories for reactions with achiral
catalysts.
It is important to note that the metal does not sit directly
above the allyl but is situated toward the anti substituents, and
the allyl cants 5-15° moving the anti substituents away from
the metal.19 This allows for better overlap with the metal dxy
orbital. In systems bearing achiral ligands, there are two possible
trajectories as depicted in Figure 4. These two pathways differ
in their orientation relative to the π-allyl. In one approach, the
nucleophile comes in from the side of the syn substituents of
the π-allyl (exo), and in the other approach from the side which
bears the anti substituents of the π-allyl (endo).20 In nucleophilic
displacement reactions (SN2), the preferred trajectory is one that
places the leaving group at an angle of 180° to the approaching
nucleophile. The same stereoelectronic arguments should there-
fore be true in the allylic alkylation reaction. If one considers
palladium(0) as the nucleophile in the ionization and pal-
ladium(+2) the leaving group in the alkylation reaction, then
both ionization and nucleophilic attack should occur at an angle
(19) For several X-ray structures, see: Dictionary of Organometallic
Coumpounds; Macintyre, J. E., Ed.; Chapman and Hall: New York, NY,
1995; Vols 1-5.
(20) The terms exo and endo have been used to describe the orientation
of the allyl unit in the pocket of non-C2 symmetric ligands. For example,
see: Helmchen, G.; Kudis, S.; Sennhenn, P.; Steinhagen, H. Pure Appl.
Chem. 1997, 68, 513 and references 11d and e.
(18) We enlisted the aid of Professor A. Togni, but his group was also
thwarted.

Regio- and EnantioselectiVe Allylic Alkylation
J. Am. Chem. Soc., Vol. 121, No. 19, 1999 4551
Figure 5. Transition-state trajectories for reactions with chiral catalysts.
that closely approaches 180° relative to palladium in the π-allyl.
For these reasons, it is reasonable to propose the preferred
trajectory for ionization and nucleophilic attack is via the exo
mode.
In the case of C2-symmetric chiral ligands, there exist two
transition states which satisfy this requirement as depicted in
the cartoon of Figure 5, the front left quadrant proximal to the
wall and the front right quadrant proximal to the flap. Attack
from either rear quadrant corresponds to an endo mode and
therefore can be ignored as stereoelectronically unfavorable. The
question that remains is whether one exo ionization/nucleophilic
attack is favored over the other?
states is that in exo-15 the nucleophile enters under a raised
flap, while, in the exo-16 mode, the nucleophile encounters a
lowered flap. Therefore, some of the difference in the transition-
state energies may come from the steric requirements for
approach of the nucleophile.
That is to say, that in the exo-15 transition state, the
nucleophile approaches the π-allyl unit with an exo trajectory
and through an open quadrant of the chiral space. It is likely
that these factors are coupled with the steric interactions between
the allyl unit and the cavity of the ligand. In rotating from a
square planar η3-allyl to a tetrahedral η2-alkene, the R groups
rotate away from the “walls” in the exo-15 transition state but
toward the “walls” in the exo-16 transition state. Thus, the model
“predicts” that the product derived from the exo-15 transition
state should be the major enantiomer as observed in the reaction
of eq 8.
To gain some insight into this question, the reaction of eq 7,
Extrapolating this picture to the unsymmetrical monosubsti-
tuted case of eq 1 leads to the pathway depicted in Figure 6.
Considering that ionization is the microscopic reverse of
nucleophilic attack, the exo-15 transition state for nucleophilic
attack of Figure 5 leads to the formation of diastereomeric
complex 17 as the kinetically preferred one in the ionization of
Figure 6. Nucleophilic attack on 17 should favor path a over
path b for the same reason. On the other hand, the diastereomeric
complex 18, available by simple equilibration via a π-σ-π
sequence, should involve nucleophilic attack preferentially by
path c. Examination of complexes 17 and 18 suggests that the
kinetically preferred complex 17 may be less stable than 18
because the R group of 17 encounters a wall but does not do so
in 18. Molecular modeling calculations indicate that the differ-
ence between these two π-allyl complexes, 17 and 18, is
approximately 3.1 kcal/mol.23 To the extent that the transition-
state energy for attack of the nucleophile on 18 is lower than
that for attack on 17, then the overall reaction pathway may be
diverted to path c over paths a and b. It is to be noted that this
model predicts (1) the kinetic diastereomeric π-allyl complex
17 should favor attack at the primary carbon and (2) to the extent
attack occurs on the kinetic complex at the secondary carbon,
the enantiomer formed is the mirror image of that obtained by
attack on the thermodynamically more stable and kinetically
more reactive complex 18. All of the results support this model.
In toluene, the reaction favors primary (path a) over secondary
(path b or c) attack and, to the extent the latter occurs, product
formation favors enantiomer 19 (path b) over 20 (path c).
Increasing polarity of the solvent stabilizes the charged inter-
mediates and thereby should slow charge neutralization that
occurs upon nucleophilic attack. By increasing the lifetime of
the π-allyl intermediate, Curtin-Hammett conditions begin to
wherein the nucleophilic attack is the enantioselectivity-
determining step, was examined. The model must be consistent
with the observed depicted absolute stereochemistry of the
product and the observation that the magnitude of the enantio-
selectivity of the product has some dependence on the structure
of the nucleophile. Furthermore, it is important to note that
Z-olefin product has never been observed. This fact suggests
that either the transition state for the formation of anti,syn-π-
allyl complexes is too high in energy or that the anti,syn-π-
allyl is rapidly interconverting with the syn,syn-complex (by a
π-σ-π mechanism) and the relative rate for attack on the
syn,syn-complex is much faster. It may also be possible that
the syn,anti-complex undergoes nucleophilic attack exclusively
at the allyl terminus leading to the E-alkene product. In any
case, the anticipated higher energy for these processes involving
the anti,syn-complex leads us to consider the syn,syn-intermedi-
ate as the one generating product.21,22
There are four possible transition states for the nucleophilic
attack as depicted in Figure 5 for the desymmetrization step of
eq 8. Because of the proposed preferred trajectory of nucleo-
philic attack, we can ignore the two endo type transition states.
The first noticeable difference between the two exo transition
(21) In other systems,11,20 NMR experiments have been used to determine
which complex, syn,syn or syn,anti, is kinetically formed. However, all
attempts to utilize this approach with our ligand have proven to be
unsuccessful.
(22) In work published after submission of this manuscript, the possibility
that the product derives from regioselective attack on a syn,anti complex
has been discussed, see: Ramdeehul, S.; Dierkes, P.; Aquado, R.; Kamer,
P. C.; van Leewen, P. W. N. M.; Osborn, J. A. Angew. Chem., Int. Ed.
1998, 37, 3118. This paper concludes, as we have for our ligands, that the
steric interactions associated with the synchronous events of nucleophilic
attack and rotation due to the change of hapticity are responsible for the
chiral recognition.
(23) Hagelin, H., unpublished results in these laboratories.

4552 J. Am. Chem. Soc., Vol. 121, No. 19, 1999
Trost and Toste
Figure 6. Predicted path for asymmetric alkylation of chiral precursor for reaction using R,R-ligand 6.
Figure 7. Predicted path for asymmetric alkylation of chiral racemic precursor for reaction using R,R-ligand 6.
be approached, and path c begins to dominate. The observed
linear dependencies of regioselectivity and enantioselectivity
onsolvent polarity are in complete accord with these predictions.
While the role of solvent may extend beyond influencing the
rate of nucleophilic attack, such as affecting the rate of π-σ-π
epimerization,24,26 insufficient data exists to discuss any such
effects.
The model also makes an interesting prediction regarding the
effect of the dependence of selectivities on the regioisomeric
starting materials. Figure 7 depicts the pathways from the chiral
racemic precursor. Again, following the precedent derived from
Figure 5, enantiomer 21 should ionize to form diastereomer 17
which, as already indicated, is predisposed to react via path a
(also see Figure 6) to give the achiral product. To the extent it
reacts via the disfavored or mismatched path b, it will form 19,
the product of net retention. To form the chiral product via a
favored or matched pathway, the kinetic diastereomer must
isomerize to diastereomer 18, which will then react via path c
to give 20. On the other hand, enantiomer 22 should ionize
kinetically to form diastereomer π-allyl complex 18 directly.
Thus, 50% of the starting material requires no equilibration from
a less stable to a more stable and reactive diastereomer to form
enantiomer 20 and only 50% of the starting material requires
equilibration to follow the same path. This fact leads to the
prediction that intrinsically higher regio- and enantioselectivity
should be observed, starting from the racemic chiral substrate
than the achiral one. Of course, if true Curtin-Hammett
conditions are achieved, then it makes no difference. However,
because of the inability to reach this ideal situation, the
prediction should be valid. Experimentally, the prediction is
verified as demonstrated by the data in Tables 1 and 3. Under
our best conditions, the achiral substrate 2 with p-methoxyphenol
at 0.1 M, in the presence of 30 mol % tetra-n-butylammonium
chloride, gave an 80:20 ratio of 4:3 wherein R-4 of 62% ee
was the major enantiomer. Under the same conditions starting
with the chiral substrate 8, an 84:16 ratio of 4:3 (86% ee for
R-4) was obtained. Thus, it is clear that true Curtin-Hammett
conditions have been approached but not yet reached even under
our best conditions. Nevertheless, the results are synthetically
useful as shown by eq 6.
(24) (a) Faller, J. W.; Thomsen, M. E.; Mattina, M. J. J. Am. Chem.
Soc. 1971, 93, 2642. (b) Faller, J. W.; Tully, M. T. J. Am. Chem. Soc.
1972, 94, 2676.
(25) (a) Sjogren, M.; Hansson, S.; Norrby, P.-O.; Åkermark, B.;
Cucciolito, M. E.; Vitagliano, A. Organometallics 1992, 11, 3954. (b)
Åkermark, B.; Vitagilano, A. Organometallics 1985, 4, 1275.
(26) Andersson, P. G.; Harden, A.; Tanner, D.; Norrby, P.-O. Chem.s
Eur. J. 1995, 1, 12.
Conclusion
The importance of the metal catalyzed allylic alkylation has
stimulated a great deal of effort to exercise control, especially
regio- and enantiocontrol, in a rational way. Achieving this goal
requires a detailed understanding of the factors that influence

Regio- and EnantioselectiVe Allylic Alkylation
J. Am. Chem. Soc., Vol. 121, No. 19, 1999 4553
both regio- and enantioselectivity.27 Regioselectivity is influ-
enced by normally opposing steric and electronic effects.28 Steric
effects lead to attack at the less substituted terminus, whereas
electronic effects lead to attack at the less electron-rich terminus
which is normally the more substituted terminus. The current
studies reveal a new strategy to influence regioselectivityschiral
ligands. Such ligands can exercise regiocontrol in addition to
enantiocontrol in ways not possible with achiral ligands. One
factor involves the effect of the chiral ligand on the preferred
trajectory of approach of the nucleophile on the π-allyl
intermediate by imposing a strong stereoelectronic requirement
for a linear SN2 transition state on the other regiochemistry-
determining factors. A second factor influencing regioselectivity
when chiral ligands are used, that does not exist with achiral
ligands, is the existence of diastereomeric π-allyl complexes
and their relative stability and reactivity. The intrinsic difference
in reactivity of these diastereomeric complexes will also affect
the regioselectivity of the process. Considering all of these
factors also leads to the initially surprising conclusion that, in
some cases, for maximum regio- and enantioselectivity, the
chiral substrate may be preferred over the achiral one (provided
that there exists a bias for forming a syn- vs anti-π-allyl
complex).
tories.29 The applicability of the concepts outlined herein to other
enantioselective palladium catalyzed reactions that have been
reported to have biases for attack at the more substituted
terminus will be interesting to explore to see if the selectivity
can be further enhanced.6 The model may find general applica-
tions in other asymmetric reactions which rely on similar chiral
environments.
The ease of cleavage of p-methoxyphenyl ethers to the
corresponding alcohols makes this reaction a deracemization
of allyl alcohols (eq 9, path a).30 By hydroboration-oxidation
or related hydration methods followed by cyclization, chro-
manes, chromanols, and ultimately chromanones are available
in nonracemic form (eq 9, path b).31 Oxidative cleavage-
cyclization should also provide access to the dihydrobenzofurans
of high enantiopurity (eq 9, path c).
The source of the enantiocontrol by the chiral ligands remains
a topic of much debate.1,22 For our ligands, we envision the
motions associated with the change in hapticity combined with
the sum total of the resultant steric interactions to account for
their generality and high selectivity in both ionization and
nucleophilic addition steps. The current results provide support
for this concept. This notion is gaining more prominence22 and
should be more carefully evaluated for related reactions.
Experimental Section
All reactions that involved palladium complexes were carried out
under an argon atmosphere in a degassed flask. All solvents were freshly
distilled from the appropriate drying agent and were further degassed
by bubbling a stream of argon through them for 15 min before use.
The phenols were purchased from Aldrich and recrystallized prior to
use. Allyl carbonate 2 and 8 were prepared as previously described.7b
General Procedure for O-Allylation. A degassed flask containing
4-methoxyphenol 1a (25 mg, 0.20 mmol), allyl carbonate 8 (32 mg,
0.20), Pd2dba3‚CHCl3 (2 mg, 1.9 µmol), the chiral ligand (4 mg, 5.8
µmol), and tetrabutylammonium chloride (17 mg, 0.06 mmol) is charged
with CH2Cl2 (2 mL). The purple reaction mixture is stirred at 0° for 8
h. The resulting yellow reaction mixture is directly applied to a silica
gel column. Flash chromatography eluting with 6:1 ether:petroleum
ether affords an 84:16 mixture of aryl ethers 4:3 (34 mg, 85% yield,
83% ee). Compounds were characterized as 4-5:1 mixtures of 4:3.
3-(4-Methoxyphenoxy)-1-hexene (4a):7b IR(film) 2980, 2933, 1507.
1H NMR (500 MHz, CD2Cl2) δ 6.88 (m, 4H), 5.86 (ddd, J ) 17.3,
10.5, 6.5 Hz, 1H), 5.25 (dt, J ) 17.3, 1.1 Hz, 1H), 5.20 (dt, J ) 10.5,
1.1 Hz, 1H), 4.45 (dt, J ) 6.5, 5.6 Hz, 1H), 3.78 (s, 3H), 1.30-1.80
(m, 4H), 0.98 (t, J ) 7.5 Hz, 3H). 13C NMR (125 MHz, CD2Cl2) δ
153.7, 152.4, 138.5, 116.2, 115.6, 114.4, 79.8, 55.6, 37.7, 18.5, 13.9.
Anal. Calcd for C13H18O2 C,75.69; H, 8.80. Found C, 75.72; H, 8.59.
Enantiomers were separated by HPLC using the Chiralcel OD column
eluting with 99.9:0.1 heptane:2-propanol (1 mL/min). The retention
times were major (R) 9.31 min, minor (S) 10.44 min, and regioisomer
18.02 min.
For maximum regio- and enantioselectivity, Curtin-Hammett
conditions must prevail. However, the employment of chloride
ion as a promoter for facial interconversion of the palladium
may be compromised because it may serve as a general base
catalyst. In this role, it would speed up the rate of nucleophilic
attack by the phenol and thereby counter its action in increasing
the rate of facial interconversion of the palladium π-allyl
intermediate. A promoter that would not function as a general
base catalyst would be ideal.
The choice of nucleophile will undoubtedly also play a major
role. For this study, we focused on oxygen to diminish steric
demands of the nucleophile influencing the regioselectivity. As
the nucleophile becomes bulkier, a point will undoubtedly be
reached where its steric requirements outweigh the restrictions
imposed by the ligand.
For a given nucleophile, choice of substrate, solvent, con-
centration, and promoter (chloride ion) can affect these relative
rates to the point of making the reaction useful. To illustrate,
for the synthesis of 3-(4′-methoxyphenoxy)-1-hexene, the initial
reaction employing “standard” conditions based on earlier
work16 produced a 42:58 ratio of achiral (3) to chiral product
(4), the latter in 12% ee. This reaction evolved into a new set
of conditions that produced a 16:84 ratio where the latter had
an ee of 83%. In fact, by rational modification of reaction
conditions, the ee varied from 66% favoring 4(S) to 83%
favoring 4(R) with the same chiral ligand! Thus, the model as
outlined herein provides a useful working framework to
understand and enhance the utility of asymmetric allylic
alkylations with the chiral ligands developed in these labora-
3-(4-Methylphenoxy)-1-hexene (4b):7b IR(film) 2960, 2931, 1509.
1H NMR (500 MHz, CD2Cl2) δ 7.08 (dd, J ) 9.3, 0.5 Hz, 2H), 6.84
(dd, J ) 9.3, 0.5 Hz, 2H) 5.87 (ddd, J ) 17.3, 10.3, 6.1 Hz, 1H), 5.27
(dt, J ) 17.3, 1.1 Hz, 1H), 5.21 (dt, J ) 10.3, 1.1 Hz, 1H), 4.57 (dt,
J ) 6.5, 5.6 Hz, 1H), 2.31 (s, 3H), 1.30-1.80 (m, 4H), 0.98 (t, J )
7.4 Hz, 3H). 13C NMR (125 MHz, CD2Cl2) δ 156.2, 138.3, 129.7, 115.9,
114.5, 78.9, 37.7, 20.4, 18.5, 13.9. Enantiomers were separated by
HPLC using the Chiralcel OD column eluting with 99.9:0.1 heptane:
(29) The results obtained herein provide impetus to reexamine the
conclusions for the intermolecular reactions reported in ref 6e and will be
reported in due course.
(27) Compare Pergosin, P. S.; Salzman, R. Coord. Chem. ReV. 1996,
155, 35.
(28) Trost, B. M.; Hung, M.-H. J. Am. Chem. Soc. 1984, 106, 6837.
(30) Georg, G. I.; Mashava, P. M.; Akgun, E.; Milstead, M. W.
Tetrahedron Lett. 1991, 32, 3151.
(31) Trost, B. M.; Toste, F. D. J. Am. Chem. Soc. 1998, 120, 9074.

4554 J. Am. Chem. Soc., Vol. 121, No. 19, 1999
Trost and Toste
2-propanol (0.9 mL/min). The retention times were major (R) 5.51 min,
minor (S) 6.24 min, and regioisomer 11.71 min.
tion. To a solution of 4-methoxyphenol (307 mg, 2.45 mmol), (R)-3-
hexanol9 (210 mg, 2.06 mmol) and triphenylphosphine (640 mg, 1.45
mmol) in THF (20 mL) at 0° is slowly added diethyl azodicarboxylate
(385 µL, 2.45 mmol), and the resulting orange-tinged solution stirred
at room temperature for 6 h. The solution is diluted with ether (25
mL), washed with 2 N NaOH (2 × 20 mL) and water (25 mL), dried
(MgSO4), and concentrated in vacuo. The resulting solid is suspended
in hexanes, filtered, and concentrated in vacuo. Flash chromatography
eluting with ether affords the aryl ether (120 mg, 28%) as a colorless
liquid. IR(film) 2960, 2935, 1506. 1H NMR (500 MHz, CD2Cl2) δ 6.85
(m, 4H), 4.07 (quint, J ) 5.7 Hz, 1H), 3.79 (s, 3H), 1.35-1.70 (m,
6H), 0.98 (t, J ) 7.5 Hz, 3H) 0.95 (t, J ) 7.4 Hz, 3H). 13C NMR (125
MHz, CD2Cl2): δ 153.7, 152.8, 117.4, 114.6, 80.0, 55.7, 35.7, 26.5,
18.7, 14.2, 9.5. Enantiomers were separated by HPLC using the
Chiralcel OD column eluting with 99.9:0.1 heptane:2-propanol (1 mL/
min.). The retention times were major (S) 9.47 min (68.6%) and minor
(R) 11.5 min (31.4%).
By Hydrogenation of 4a(R). A suspension of 10% Pd/C (10 mg)
and 4a(R) (100 mg, 0.425 mmol) in ethyl acetate (1 mL) is stirred
under 1 atm of hydrogen at room temperature for 10 h. The resulting
reaction mixture is filtered through a plug of Celite to afford the
unsaturated aryl ether (97 mg, 97%). Enantiomers were separated by
HPLC using the Chiralcel OD column eluting with 99.9:0.1 heptane:
2-propanol (1 mL/min). The retention times were major (S) 10.06 min
(86.5%) and minor (R) 11.73 min (13.5%).
1
3-Phenoxy-1-hexene (4c):7b IR(film) 2960, 2933, 1494. H NMR
(500 MHz, CD2Cl2) δ 7.27 (m, 2H), 6.95 (m, 3H), 5.88 (ddd, J )
17.4, 10.6, 6.1 Hz, 1H), 5.29 (dt, J ) 17.4, 1.3 Hz, 1H), 5.22 (dt, J )
10.6, 1.3 Hz, 1H), 4.64 (q, J ) 6.1, 1H), 1.30-1.80 (m, 4H), 0.99 (t,
J ) 7.5 Hz, 3H). 13C NMR (125 MHz, CD2Cl2) δ 158.4, 138.2, 129.2,
116.0, 114.7, 78.6, 37.7, 18.5, 13.9. Enantiomers were separated by
HPLC using the Chiralcel OD column eluting with 99.9:0.1 heptane:
2-propanol (1 mL/min). The retention times were major (R) 5.77 min,
minor (S) 7.03 min, and regioisomer 19.67 min.
3-(4′-Chlorophenoxy)-1-hexene (4e):7b IR(film) 2960, 2932, 1490.
1H NMR (500 MHz, CD2Cl2) δ 7.22 (d, J ) 9.0, Hz, 2H), 6.85 (d, J
) 9.0 Hz, 2H) 5.84 (ddd, J ) 17.4, 10.4, 6.2 Hz, 1H), 5.27 (dt, J )
17.4, 1.1 Hz, 1H), 5.22 (dt, J ) 10.4, 1.1 Hz, 1H), 4.57 (q, J ) 6.2,
1H), 1.40-1.82 (m, 4H), 0.98 (t, J ) 7.3 Hz, 3H). 13C NMR (125
MHz, CD2Cl2) δ 166.9, 162.3, 137.4, 131.4, 124.2, 116.7, 115.3, 78.7,
51.8, 37.6, 18.5, 13.9. Enantiomers were separated by HPLC using the
Chiralcel OD column eluting with 99.9:0.1 heptane:2-propanol (0.8 mL/
min). The retention times were major (R) 5.30 min, minor (S) 5.97
min, and regioisomer 7.26 min.
3-(4-(Methoxycarbonyl)phenoxy)-1-hexene (4e): IR(film) 2960,
1
2930, 1790, 1470. H NMR (500 MHz, CD2Cl2) δ 7.97 (d, J ) 8.8,
Hz, 2H), 6.93 (d, J ) 8.8 Hz, 2H) 5.84 (ddd, J ) 17.4, 10.4, 6.4 Hz,
1H), 5.28 (dt, J ) 17.4, 1.1 Hz, 1H), 5.23 (dt, J ) 10.4, 1.1 Hz, 1H),
4.71 (q, J ) 6.4, 1H), 3.90 (s, 3H), 1.40-1.82 (m, 4H), 0.97 (t, J )
7.3 Hz, 3H). 13C NMR (125 MHz, CD2Cl2) δ 157.0, 137.8, 129.1, 125.5,
117.3, 116.5, 79.2, 37.6, 18.5, 13.9. Enantiomers were separated by
HPLC using the Chiralpak AD column eluting with 99:1 heptane:2-
propanol (1 mL/min). The retention times were minor (S) 6.47 min,
major (R) 7.27 min, and regioisomer 8.29 min.
Acknowledgment. We thank the National Institutes of
Health, General Medicinal Sciences Institute, and the National
Science Foundation for their generous support of our programs.
F.D.T. is a Stanford Graduate Fellow. Mass spectra were
provided by the Mass Spectrometry Facility, University of San
Francisco, supported by the NIH Division of Research Re-
sources.
1H NMR Experiments. A degassed screw-capped NMR tube
containing 4-methoxyphenol (12 mg, 0.096 mmol), Pd2dba3‚CHCl3 (0.2
mg, 0.19 µmol), and the chiral ligand (0.4 mg, 0.58 µmol) is charged
1
with degassed CD2Cl2 (1 mL). A background H NMR (500 MHz) is
measured, and then allyl carbonate 2 (15 µL, 0.095 mmol) is added.
Single scan spectra are acquired at 30-s intervals by arraying the
preacquisition delay (pad). The conversion at each interval is determined
by integrating the signal at 4.59 for the starting allyl carbonate.
3-(4-Methoxyphenoxy)-1-hexane (7a). By the Mitsunobu Reac-
Supporting Information Available: Experimental details
for Tables 1-4 (PDF). This material is available free of charge
via the Internet at http://pubs.acs.org.
JA9828713