Origins of enantioselectivity during allylic substitution reactions catalyzed by metallacyclic iridium complexes

Article abstract of DOI:10.1021/ja212217j

In depth mechanistic studies of iridium catalyzed regioselective and enantioselective allylic substitution reactions are presented. A series of cyclometalated allyliridium complexes that are kinetically and chemically competent to be intermediates in the allylic substitution reactions was prepared and characterized by 1D and 2D NMR spectroscopies and single-crystal X-ray difraction. The rates of epimerization of the less thermodynamically stable diastereomeric allyliridium complexes to the thermodynamically more stable allyliridium stereoisomers were measured. The rates of nucleophilic attack by aniline and by N-methylaniline on the isolated allyliridium complexes were also measured. Attack on the thermodynamically less stable allyliridium complex was found to be orders of magnitude faster than attack on the thermodynamically more stable complex, yet the major enantiomer of the catalytic reaction is formed from the more stable diastereomer. Comparison of the rates of nucleophilic attack to the rates of epimerization of the diastereomeric allyliridium complexes containing a weakly coordinating counterion showed that nucleophilic attack on the less stable allyliridium species is much faster than conversion of the less stable isomer to the more stable isomer. These observations imply that Curtin-Hammett conditions are not met during iridium catalyzed allylic substitution reactions by ≠³-≠¹-≠³ interconversion. Rather, these data imply that when these conditions exist for this reaction, they are created by reversible oxidative addition, and the high selectivity of this oxidative addition step to form the more stable diastereomeric allyl complex leads to the high enantioselectivity. The stereochemical outcome of the individual steps of allylic substitution was assessed by reactions of deuterium-labeled substrates. The allylic substitution was shown to occur by oxidative addition with inversion of configuration, followed by an outer sphere nucleophilic attack that leads to a second inversion of configuration. This result contrasts the changes in configuration that occur during reactions of molybdenum complexes studied with these substrates previously. In short, these studies show that the factors that control the enantioselectivity of iridium-catalyzed allylic substitution are distinct from those that control enantioselectivity during allylic substitution catalyzed by palladium or molybdenum complexes and lead to the unique combination of high regioselectivity, enantioselectivity, and scope of reactive nucleophile.

Full text of DOI:10.1021/ja212217j

Article
pubs.acs.org/JACS
Origins of Enantioselectivity during Allylic Substitution Reactions
Catalyzed by Metallacyclic Iridium Complexes
Sherzod T. Madrahimov† and John F. Hartwig*,‡
†Department of Chemistry, University of Illinois, 600 South Mathews Avenue, Urbana, Illinois 61801, United States
‡Department of Chemistry, University of California, Berkeley, Berkeley, California 94720-1460, United States
S
* Supporting Information
ABSTRACT: In depth mechanistic studies of iridium
catalyzed regioselective and enantioselective allylic substitution
reactions are presented. A series of cyclometalated allyliridium
complexes that are kinetically and chemically competent to be
intermediates in the allylic substitution reactions was prepared
and characterized by 1D and 2D NMR spectroscopies and
single-crystal X-ray difraction. The rates of epimerization of
the less thermodynamically stable diastereomeric allyliridium
complexes to the thermodynamically more stable allyliridium stereoisomers were measured. The rates of nucleophilic attack by
aniline and by N-methylaniline on the isolated allyliridium complexes were also measured. Attack on the thermodynamically less
stable allyliridium complex was found to be orders of magnitude faster than attack on the thermodynamically more stable
complex, yet the major enantiomer of the catalytic reaction is formed from the more stable diastereomer. Comparison of the
rates of nucleophilic attack to the rates of epimerization of the diastereomeric allyliridium complexes containing a weakly
coordinating counterion showed that nucleophilic attack on the less stable allyliridium species is much faster than conversion of
the less stable isomer to the more stable isomer. These observations imply that Curtin−Hammett conditions are not met during
iridium catalyzed allylic substitution reactions by η3-η1-η3 interconversion. Rather, these data imply that when these conditions
exist for this reaction, they are created by reversible oxidative addition, and the high selectivity of this oxidative addition step to
form the more stable diastereomeric allyl complex leads to the high enantioselectivity. The stereochemical outcome of the
individual steps of allylic substitution was assessed by reactions of deuterium-labeled substrates. The allylic substitution was
shown to occur by oxidative addition with inversion of configuration, followed by an outer sphere nucleophilic attack that leads
to a second inversion of configuration. This result contrasts the changes in configuration that occur during reactions of
molybdenum complexes studied with these substrates previously. In short, these studies show that the factors that control the
enantioselectivity of iridium-catalyzed allylic substitution are distinct from those that control enantioselectivity during allylic
substitution catalyzed by palladium or molybdenum complexes and lead to the unique combination of high regioselectivity,
enantioselectivity, and scope of reactive nucleophile.
the presence of base.16,17 This class of catalyst converts linear
allylic carbonates to substitution products with high regiose-
lectivity for the branched isomers and with high enantiomeric
excess. Recently, 1b was also shown to be exquisitely selective
for the conversion of one enantiomer of a racemic mixture of
branched allylic benzoates to the allylic substitution products
with overall retention of configuration. This reaction of racemic
benzoates leads to high enantiomeric excess of both the
substitution product and the unreacted ester.18
I. INTRODUCTION
Iridium-catalyzed allylic substitution has become a powerful
method for the enantioselective formation of both carbon−
heteroatom1−8 and carbon−carbon9−15 bonds. The catalysts
developed for this transformation (Figure 1) contain a diolefin
ligand bound to a metallacyclic core that is formed by
cyclometalation of an iridium−phosphoramidite complex in
During the early mechanistic studies of iridium-catalyzed
allylic substitution with catalysts derived from phosphoramidite
ligands, the identity of the active species was revealed,19 and a
catalytic cycle, along with information on relative rates and
equilibria of the sequential steps, was deduced.20 As part of
these studies, a method for isolation and characterization of an
allyliridium complex that is an intermediate in this reaction was
Figure 1. Catalysts and ligands commonly used for iridium catalyzed
allylic substitution reactions.
Received: December 31, 2011
Published: April 9, 2012
© 2012 American Chemical Society
8136
dx.doi.org/10.1021/ja212217j | J. Am. Chem. Soc. 2012, 134, 8136−8147

Journal of the American Chemical Society
Article
Scheme 1
Scheme 2
developed.16 Following this work, an alternative method for
preparing analogous allyliridium complexes directly from
[Ir(COD)Cl]2 (COD = 1,5-cyclooctadiene) and the phosphor-
amidite ligand was published.21
the allyl intermediates, the relative reactivity of the diastereo-
meric intermediates leading to different enantiomers of the
substitution product, and the determination of the config-
urations of both intermediates and products that reveal the
stereochemical outcome of individual steps of the catalytic cycle
have not been determined.
Early studies on the stereochemical course of iridium-
catalyzed allylic substitution showed that the overall reaction of
1-phenylallyl acetate or cyclic allylic carbonates occurs with
predominant net retention of configuration when conducted
with the combination of [Ir(COD)Cl]2 and P(OPh)3 as
catalyst precursors.9,22 This overall stereochemical outcome
could result either from two steps occurring by inversion of
configuration or from all steps occurring with retention of
configuration. Reactions of substrates that discourage additions
or eliminations with inversion of configuration with the catalyst
derived from [Ir(COD)Cl]2 and P(OPh)3 gave low yields of
substitution product.22 However, these substrates are more
sterically demanding than those commonly used in iridium-
catalyzed allylic substitutions, and the catalyst contains a ligand
that is much different from those now used most commonly for
the iridium-catalyzed asymmetric allylic substitutions. Most
important, little information on the origins of enantioselectivity
has been gained during any of the prior studies. For example,
the enantioselectivity-determining step of the catalytic cycle,
the relative rates for epimerization of diastereomeric
intermediates versus the reactions of the nucleophiles with
We report the results of mechanistic studies conducted with
catalysts containing cyclometalated phosphoramidite ligands
that are the active form of the most commonly used iridium
catalysts for allylic substitution. These studies provide
information on (1) the stereochemistry of the individual
oxidative addition and nucleophilic substitution steps; (2)
isolation of the major allyliridium diastereomer in the catalytic
cycle and direct observation of both major and minor
diastereomers generated from alkyl- and aryl-substituted allylic
electrophiles, and (3) the relative rates for epimerization of
allyliridium complexes versus nucleophilic attack on the major
and minor diastereomers. These studies provide deep-seated
conclusions about the origins of the high stereoselectivity of
iridium-catalyzed allylic substitution. Our data show that the
factors controlling enantioselectivity in these iridium-catalyzed
reactions are distinct from those controlling enantioselectivity
in both palladium-catalyzed and molybdenum-catalyzed asym-
metric allylic substitutions and that the relationships between
diastereomeric intermediates and enantiomeric products are
8137
dx.doi.org/10.1021/ja212217j | J. Am. Chem. Soc. 2012, 134, 8136−8147

Journal of the American Chemical Society
Article
Scheme 3
distinct from those of prior catalytic asymmetric systems
studied mechanistically.
methyl cinnamyl carbonate 4b did not form the allyliridium
complex in detectable quantities in the absence of AgBF4;20 1a-
(R,R,R) remained unreacted. Thus, a small quantity of
allyliridium complex formed in an unfavorable equilibrium
appears to react irreversibly with AgBF4 to form thermody-
namically stable allyliridium complex with a non-nucleophilic
counterion 2c-BF4, or the allylic carbonate is activated by
association with the silver cation. The potential that the
carbonate is activated by association with silver cation was
tested by combining ethyl cinnamyl carbonate with AgBF4 in
THF-d8 and comparing the 1H NMR spectrum of the resulting
solution with that of ethyl cinnamyl carbonate. No difference
was observed between the two spectra, showing that little if any
adduct is formed. A small amount of Lewis acid−base complex
formed in an unfavorable equilibrium could be the species that
reacts with the Ir(I) complex, but the absence of an observable
adduct is at least consistent with reaction of the silver salt with a
small amount of allyliridium carbonate complex.
Two synthetic routes provided access to samples of
allyliridium complexes containing observable amounts of the
less stable diastereomer. A mixture of diastereomers of
cinnamyl complex 2e was generated from reaction of 2-
bromocinnamyl chloride, followed by addition of a silver salt
containing a non-nucleophilic anion. Because ortho-substituted
cinnamyl carbonates undergo iridium-catalyzed allylic sub-
stitution with lower enantiomeric excess than meta- or para-
substituted cinnamyl carbonates (typically 70%−80% ee for 2-
methoxy cinnamylcarbonates),1,8,15,23 we considered that the
stoichiometric reaction of the Ir(I) species with an ortho-
substituted allylic electrophile might generate a mixture of
diastereomers. Indeed, reaction of the ethylene complex 1a-
(R,R,R) with 2-bromocinnamyl chloride, followed by addition
of AgBF4, gave a mixture of diastereomers of 2e. The ratios of
diastereomers obtained from this process depended on solvent;
preparation of 2e in benzene gave a 60:40 ratio of
diastereomers, preparation of 2e in THF gave an 80:20 ratio
of diastereomers, and preparation in CH2Cl2 gave a 90:10 ratio
of diastereomers (part b of Scheme 2).
II. RESULTS AND DISCUSSION
1. Synthesis and Characterization of Allyliridium
Complexes. Our previous studies showed that the stoichio-
metric reactions of allylic carbonates with the Ir(I) ethylene
complex in Scheme 1 do not generate observable amounts of
allyliridium complexes. Thus, we prepared allyliridium complex
2a and crotyliridium complex 2b previously by the route shown
in part a of Scheme 1 involving the reaction of iridium−
ethylene complex 1a-(R,R,R), allyl or crotyl chloride, and silver
salts containing non-coordinating anions.16 Subsequent work
by Helmchen showed that crotyl complex 3a and cinnamyliri-
dium complex 3b could be isolated from the route in part b of
Scheme 1 starting from [Ir(COD)Cl]2, the phosphoramidite,
allylic carbonates, and silver perchlorate.21 In addition, we
prepared allyliridium complex 3b, generated from ligand L2, by
a procedure similar to that reported by Helmchen and co-
workers.21 However, the mechanistic studies reported here
required a route to allyliridium complexes that occurred under
sufficiently mild conditions to observe directly both minor and
major diastereomers containing the metal bound to the two
different faces of the allyl unit.
After conducting further studies with allylic chlorides, we
identified an alternative route to the allyl intermediates
involving the reaction of iridium−ethylene complexes with
allylic trifluoroacetates or with the combination of allylic
carbonates and AgBF4 (Scheme 2a). Both of these reactions
allowed the isolation of pure allyliridium products in high yields
and, in specific cases discussed later, as the mixtures of
diastereomeric allyliridium complexes needed for the envi-
sioned mechanistic studies. The reaction of ethylene complex
1a-(R,R,R) with cinnamyl trifluoroacetate 4a in THF, followed
by the replacement of the trifluoroacetate anion by
tetrafluoroborate anion, formed cinnamyliridium complex 2c
in 75% isolated yield. Similarly, reaction of the mixture of 1a-
(R,R,R) and methyl cinnamyl carbonate 4b with AgBF4 in THF
led to the precipitation of silver methyl carbonate and
An alternative method to prepare a mixture of diastereomeric
allyliridium complexes involves the reaction of one enantiomer
of a branched allylic trifluoroacetate with the enantiomer of
ethylene complex 1a that generates the less stable diastereo-
meric allyl complex. As shown in Scheme 3, the reaction of the
branched phenethyl-substituted allylic trifluoroacetic acid ester
5b-(S)24,25 with ethylene-ligated iridium complex 1a-(S,S,S)
(Scheme 3) at −40 °C gave the less stable diastereomer 2f-
TFA-(S,S,S,R) of the corresponding allyliridium complex as the
initial product of this oxidative addition. Over the course of
−
formation of 2c as the BF4 salt in 80% isolated yield. The
substituted cinnamyliridium complex 2d containing 2,6-difluoro
substituents was prepared from the cinnamyl chloride (part b of
Scheme 2).16 The reaction of 1a-(R,R,R) with 2,6-difluoro-
cinnamyl chloride 4c, followed by exchange of the chloride
anion by the tetrafluoroborate anion, gave allyliridium complex
2d in 65% isolated yield.
As noted in the introduction to this section, the combination
of ethylene complex 1a-(R,R,R) and an excess amount of
8138
dx.doi.org/10.1021/ja212217j | J. Am. Chem. Soc. 2012, 134, 8136−8147

Journal of the American Chemical Society
Article
several hours at −40 °C, 2f-TFA-(S,S,S,R) converted to the
more stable allyliridium diastereomer 2f-TFA-(S,S,S,S). When
the same reaction between 5b-(S) and 1a-(S,S,S) was
conducted at 25 °C, 2f-TFA-(S,S,S,R) was not observed by
NMR spectroscopy; the more stable allyliridium complex 2f-
TFA-(S,S,S,S) was formed quantitatively.
The new allyliridium complexes 2c, 2d, and 2e prepared by
the methods in Scheme 2, and the new complex 2f-BF4
prepared as shown in Scheme 1, were characterized by
multinuclear 1D and 2D NMR spectroscopy. Complexes 2c,
2d, and 2e were also characterized by single-crystal X-ray
diffraction analysis. The related complex 3b was characterized
by multinuclear 1D NMR spectroscopy. The 31P NMR
spectrum of cinnamyliridium complex 2c in THF-d8 consisted
of a single resonance at 118.5 ppm, consistent with the
Figure 3. ORTEP diagrams of major and minor diastereomers of 2e
(counterions and hydrogens are omitted for clarity; ellipsoids are
drawn to 35% probability) determined by X-ray diffraction of a crystal
containing an 80:20 ratio of the major and minor diastereomers.
1
presence of a single observable diastereomer. Each of the H
crystal was not obtained, but the same ratio of diastereomers
was observed by 31P NMR spectroscopy of the bulk sample
from which the crystal was selected for X-ray crystallography.
The identity of 2f-TFA-(S,S,S,R) containing a phenethyl-
substituted allyl group was deduced from several pieces of
information. First, the 31P NMR spectra of the allyliridium
complexes consist of singlets near 120−125 ppm, and a
resonance assigned to 2f-TFA-(S,S,S,R) was observed at 120.2
ppm. Second, the signal at 120.2 ppm decays as the intensity of
the signal at 123.8 ppm increases. (The assignment of the 31P
NMR resonance at 123.8 ppm to 2f-TFA-(S,S,S,S) was
corroborated by the reaction of 1a-(S,S,S) with linear
electrophile 5l to form the same species, as shown in Scheme
3). Finally, studies described later in this paper show that
oxidative addition of the allylic esters occurs with inversion of
configuration, and the identity of 2f-TFA-(S,S,S,R) as the
kinetic product of oxidative addition is consistent with this
inversion of configuration.
NMR signals of the allylic protons of 2c was well resolved. The
chemical shifts of the allyl group are 5.61 ppm (substituted
terminus), 4.85 ppm (proton on central allylic carbon), 2.83
ppm (syn proton on the terminal allylic carbon), and 2.05 ppm
(anti proton on the terminal allylic carbon) in THF-d8.
Similarly, 2,6-difluorocinnamyliridium complex 2d was charac-
1
terized by H NMR, 13C NMR, 31P NMR, and 2D gCOSY-
NMR spectroscopy, as well as X-ray crystallography (vide
infra). Again, a single 31P NMR resonance (115.6 ppm) was
1
observed for samples in THF solvent. The H NMR signals of
the allylic protons for 2d in THF-d8 were also well-resolved and
resonated at chemical shifts similar to those of the allyl group in
2c. The solid-state structures of complexes 2c and 2d are
shown in Figure 2. The 31P NMR spectrum of 2f-BF4 consisted
2. Structural Characterization of Cinnamyliridium 2c,
2,6-Difluorocinnamyliridium 2d, and 2-Bromocinnamy-
liridium 2e. The solid-state structural data of the complexes
2c, 2d (Figure 2), and 2e-(R,R,R,S) (Figure 3), as well as that
of previously reported allyliridium complex 2a, crotyliridium
complex 2b,16 and crotyliridium complex 3a (which contains a
phosphoramidite ligand with 2-methoxyphenyl groups L1b21)
are listed in Table 1. An overlay of the structures of 2b and 2c is
provided in Figure 4. The position of C47 (the carbon of the
substituted allylic terminus) varies significantly among this
series of compounds (Table 1 and Figure 4). The position of
C45 (the carbon of the unsubstituted allylic terminus) varies to
a lesser degree. More subtle differences are found in the bond
angles, and these changes are analyzed in a footnote.26 The Ir−
C47 bond lengths increase with an increase in the size of the
Figure 2. ORTEP diagrams of 2c and 2d (counterions and hydrogens
are omitted for clarity; ellipsoids are drawn to 35% probability).
of a singlet at 120.6 ppm, and the 31P NMR spectrum of 3c
consisted of a singlet at 121.9 ppm which matches the value
reported by Helmchen et al.21
The 2-bromocinnamyl complex 2e-BF4 was characterized
after synthesis in a solvent that favored one diastereomer,
followed by recrystallization. Synthesis of 2-bromocinnamyl
complex 2e-BF4 in THF or CH2Cl2 from 2-bromocinnamyl
chloride, ethylene complex 1a-(R,R,R), and AgBF4, followed by
recrystallization from benzene gave samples containing the two
diastereomers with ratios as high as 95:5. The 31P NMR spectra
of samples of 2e consisted of two signals at 115.4 and 114.5
ppm for the major and minor diastereomers, respectively. The
sample containing a 95:5 ratio of diastereomers was
characterized by 1D 1H NMR, 1D 13C NMR, and 2D
gCOSY-NMR spectroscopy. The ratios of diastereomers of
the 2-bromocinnamyl iridium complex 2e generated in
benzene, CH2Cl2, and THF were determined by 31P NMR
spectroscopy. Preparation of 2e in benzene and recrystallization
from the same solution gave crystals containing an 80:20 ratio
of diastereomers that were suitable for single-crystal X-ray
diffraction (Figure 3). A 31P NMR spectrum of the individual
́
substituent on the allyl group, from 2.27 Å for parent
́
́
allyliridium 2a to 2.38 Å for crotyliridium 2b and 2.32 Å for
́
crotyliridium 3a to 2.46 Å for cinnamyl, 2-bromocinnamyl, and
2,6-difluorocinnamyliridium complexes 2c−e. Although large,
this difference in metal−carbon bond lengths does not affect
the other bond lengths around iridium; the other distances are
within 0.05 Å of each other. Moreover, this difference in Ir−
C47 distance does not lead to differences in the C−C bond
lengths of the allyl unit that would reflect varying degrees of
contribution of ene-yl resonance forms. As the difference
between Ir−C47 and Ir−C45 bond lengths in the series of
́
complexes increases from the small value of 0.07 Å for 2a to the
́
́
larger value of 0.14 Å for 2b and 0.11 Å for 3a and the largest
value of 0.27 Å for 2c−e, the C45-C46 bond lengths remain
between 1.40 and 1.44 Å for all complexes, and the C46-C47
́
́
8139
dx.doi.org/10.1021/ja212217j | J. Am. Chem. Soc. 2012, 134, 8136−8147

Journal of the American Chemical Society
Article
Table 1. Bond Distances and Angles around Ir for Allyliridium Complexes 2a−d,e-(R,R,R,S) and 3a
2a
2b
2c
2d
2e-(R,R,R,S)
3a
distances, Å
Ir−P
Ir−C45
Ir−C47
Ir−C21
2.2685(6)
2.204(3)
2.274(3)
2.125(3)
2.099
2.280(3)
2.240(10)
2.377(11)
2.114(15)
2.097
2.2582(14)
2.190(5)
2.461(5)
2.120(5)
2.099
2.2611(18)
2.191(6)
2.460(7)
2.111(6)
2.104
2.2633(14)
2.209(6)
2.445(6)
2.120(5)
2.100
2.248(4)
2.21(2)
2.319(17)
2.124(12)
2.132
Ir−cent1
Ir−cent2
2.244
2.215
2.263
2.258
2.264
2.211
C45−C46
C46−C47
1.407(4)
1.405(4)
1.414(16)
1.39(2)
1.409(7)
1.387(7)
1.395(9)
1.368(9)
1.397(9)
1.396(8)
1.44(3)
1.40(2)
angles, deg
P−Ir−C45
P−Ir−cent1
cent1−Ir−C47
C47−Ir−C45
cent2−Ir−P
cent2−Ir−C45
cent2−Ir−cent1
cent2−Ir−C47
C21−Ir−P
C21−Ir−C45
C21−Ir−C47
C21−Ir−cent1
C45−C46−C47
86.96(8)
106.6
85.9(3)
107.6
85.28
85.2(2)
107.16
106.11
62.3(3)
103.82
92.65
85.83(17)
106.87
105.13
62.8(2)
102.92
94.34
90.2(6)
105.42
101.02
66.3(8)
104.65
88.74
106.82
105.23
63.2(3)
103.22
94.26
102.2
103
66.4(1)
103.8
66.0(5)
104.6
90.0
90.9
82.3
81.4
81.87
82.63
81.37
81.34
95.8
97.3
91.79
92.11
91.85
95.91
75.73(8)
102.9(1)
90.9(1)
85.6
75.0(4)
102.5(5)
89.9(5)
86.4
76.39(14)
99.81
75.94(18)
100.2(3)
95.2(3)
85.06
76.59(15)
99.8(3)
96.1(2)
85.20
73.7(4)
104.7(7)
91.6(6)
86.03
96.09
84.7
121.3(3)
127.5(3)
122.1(2)
121.2(2)
122.1(4)
121.8(17)
3. Kinetic Studies of the Reactions of the Allyliridium
Complexes. To elucidate the origin of enantioselectivity by
the iridium catalyst in these substitution reactions, we measured
the rate constants for epimerization of allyliridium complexes
and the rate constants for nucleophilic attack on allyliridium
complexes. Direct comparison of this set of rate constants
would allow us to determine the enantioselectivity-determining
step. Our synthetic route to a nonequilibrium ratio of
diastereomeric allyl complexes allowed us to determine
explicitly the values of these rate constants.
3a. Rates for Epimerization of Allyliridium Complexes. As
described in section 1, the less thermodynamically stable
allyliridium complex 2f-TFA-(S,S,S,R) was obtained by the
oxidative addition of 5b-(S) (6 equiv) to 1a-(S,S,S) (Scheme
3). The conversion of 5b-(S) and 1a-(S,S,S) to the
diastereomeric allyl complexes was monitored at −40 °C and
at −30 °C. The initial >25:1 ratio of diastereomers of 2f
favoring the less stable isomer 2f-TFA-(S,S,S,R) converts over
the course of 5.5 h to a 5:3 ratio of diastereomers favoring the
more stable isomer 2f-TFA-(S,S,S,S) and eventually leads to
exclusive formation of thermodynamically favored 2f-TFA-
(S,S,S,S). The rate constant for this isomerization process was
assessed by monitoring the decay of the signal at 151.0 ppm in
the 31P NMR spectrum of the initial reaction mixture, the
changes in the intensity of the signal at 120.2 ppm for the less
stable diastereomer, and the appearance of the signal at 123.8
ppm corresponding to the more stable diastereomer.
Figure 4. Overlay of solid state structures of 2b and 2c.
́
bond lengths remain between 1.39 and 1.41 Å, with one
́
exceptionthe C46-C47 distance is only 1.368(9) Å for the
2,6-difluorocinnamyliridium complex 2d. This shorter C46−
C47 bond length in 2e could be due to the presence of two
strongly electron withdrawing units on the phenyl group, which
would weaken the σ donation from the π system to the metal.
Considering the proposed connection between the ene-yl form
of an allyl unit and the regioselectivity for attack at the more
substituted vs unsubstituted termini of an allyl group, the
observation of high regioselectivity for formation of a branched
isomer, despite the large changes in the difference between the
two Ir−C bond lengths, is noteworthy. In addition, the
structure of crotyliridium complex 3a is similar to that of
crotyliridium complex 2b, which contains the phosphoramidite
ligand L1a having plain phenyl groups, despite the often
distinct reactivity and selectivity.
The plots showing the species present during the oxidative
addition of 5b-(S) to 1a-(S,S,S) and during conversion of 2f-
TFA-(S,S,S,R) to 2f-TFA-(S,S,S,S) at −40 °C are shown in
Figure 5, and the plot corresponding to the same reaction
8140
dx.doi.org/10.1021/ja212217j | J. Am. Chem. Soc. 2012, 134, 8136−8147

Journal of the American Chemical Society
Article
two half-lives).29 Thus, the conclusions about the relative rates
for nucleophilic attack on the allyl intermediate versus
epimerization of the minor allyl diastereomer to the major
diastereomer apply to the catalytic reactions. Moreover, data
presented later in this paper imply that these relative rates are
maintained throughout a catalytic reaction conducted under the
standard conditions with a small excess of the nucleophile. The
rate constants were determined by measuring the decay of the
31P NMR signal of the starting iridium complex. The reaction
progress curves are provided in the Supporting Information,
and the rate constants are provided in Table 2.
Table 2. Rate Constants of Nucleophilic Attack on
Allyliridium Complexes 2c,d,f and 3b
Figure 5. Plot of concentration vs time for the oxidative addition of
5b-(S) to 1a-(S,S,S) and isomerization of 2f-TFA-(S,S,S,R) to 2f-
TFA-(S,S,S,S) at −40 °C in THF.
monitored at −30 °C is presented in the Supporting
Information. This process follows a consecutive irreversible
reaction sequence in which the two steps occur at similar
rates.27,28 The decay of 1a-(S,S,S) versus time corresponds to
the oxidative addition of the branched allylic ester 5b-(S) to 1a-
(S,S,S). A large excess of allylic ester 5b-(S) ensures that the
oxidative addition occurs under pseudo-first-order conditions in
1a-(S,S,S). This process occurs with an exponential, first-order
decay with a rate constant of 2.2 × 10−4 s−1 at −40 °C and with
a rate constant of 5.8 × 10−4 s−1 at −30 °C. These pseudo-first-
order rate constants for oxidative addition (k1) and the curve of
[2f-(S,S,S,R)] versus time were used to determine the rate
constant for isomerization of the allyliridium complex (k2).
a
complex
(M)
NHR1R2 or
TBAOAc
entry
solvent
T, °C
k
obs, s−1
1
2
3
2c (0.030)
2c (0.043)
2c (0.030)
PhNH2
PhNH2
PrNH2
CH2Cl2
THF
−30 6.0 × 10−4
−40 3.4 × 10−3
THF
−60 too fast to
measure
4
5
6
7
2d (0.030)
2d (0.043)
3b (0.030)
2c (0.030)
PhNH2
CH2Cl2
CH2Cl2
THF
−30 2.8 × 10−4
−30 2.0 × 10−4
−40 9.9 × 10−4
PhNH(Me)
PhNH2
d[2f‐(S, S, S, R)]/dt
TBAOAc
THF
−60 too fast to
= k1[1a‐(S, S, S)] − k2[2f‐(S, S, S, R)]
(1)
measure
8
9
2f-TFA
(0.046)
PhNH2
PhNH2
THF
THF
−40 2.5 × 10−3
[2f‐(S, S, S, R)]max /[1a‐(S, S, S)]0
2f-BF4
(0.037)
−40 9.9 × 10−5
= (k2/k1)[(k /k )/(1−(k /k ))]
2
1
2
1
(2)
a
The reactions of complexes 2c, 2d, 2f-BF4, and 3b were conducted
with the R,R,R,R diastereomers shown in the graphic. The reaction of
2f-TFA was conducted with the S,S,S,S diastereomer.
The change in the concentration of 2f-TFA-(S,S,S,R) over
time is expressed in eq 1. Integration and approximation of eq
1, according to Swain’s treatment,27 gives eq 2. By iteratively
fitting the experimental data to this equation, we determined
that the rate constant for conversion of the minor to the major
diastereomer is 5.4 × 10−5 s−1 at −40 °C and 3.4 × 10−4 s−1 at
−30 °C. Further details of this treatment of the data are
provided in the Supporting Information.
3b. Rates of Nucleophilic Attack on the Allyliridium
Complexes. The rate constants for nucleophilic attack on
allyliridium complexes 2c, 2d, 2f-(S,S,S,S), and 3b were
determined by conducting reactions of the complexes with 25
equiv of a primary or a secondary amine nucleophile and 25
equiv of triethylamine as proton acceptor. The large excess of
nucleophile creates pseudo-first-order conditions for the
reaction of the nucleophile with the allyliridium complexes.
Because the rate of these reactions surely depends on the
concentration of nucleophile, it is important to consider the
relative concentration of nucleophile in these stoichiometric
reactions and a typical iridium-catalyzed allylic substitution
reaction. The concentration of nucleophile in these stoichio-
metric reactions (0.75−1.15 M) is similar to that in the middle
of a typical iridium-catalyzed allylic substitution (after one to
The data in Table 2 reveal several trends: (1) The rate of
nucleophilic attack is sensitive to the polarity of the solvent. A
slight increase in the polarity of the solvent led to a large
increase in the rate of nucleophilic attack (entry 1 vs entry 2).
(2) Alkylamines react much faster than arylamines (entries 2
and 3). (3) N-Methylaniline and aniline react with nearly
identical rate constants (entries 4 and 5). (4) Nucleophilic
attack on cinnamyliridium complex 3b, which bears the
phosphoramidite ligand containing an anisyl group, is slower
than nucleophilic attack on cinnamyliridium complex 2c
(entries 2 and 6), which bears the phosphoramidite ligand
containing a phenyl group. This slower rate for attack on
complex 3b contrasts the often faster rate for catalytic allylic
substitution with the anisyl-substituted system,8,30 suggesting
that the origin of the difference in rate of the catalytic reactions
as a function of the substituent on the phosphoramidite is more
subtle than an effect on the rate of the turnover-limiting C−N
bond-forming step within the catalytic cycle.
8141
dx.doi.org/10.1021/ja212217j | J. Am. Chem. Soc. 2012, 134, 8136−8147

Journal of the American Chemical Society
Article
Most important for the current study is the conclusion one
can draw by comparing the rate constants for interconversion
of the diastereomeric allyliridium complexes with the rate
constants for nucleophilic attack on allyliridium complexes.
Nucleophilic attack of alkylamines on the allyliridium
complexes was too fast to measure at −60 °C (Table 2:
entry 3), whereas conversion of the minor to the major
diastereomer of the allyliridium complexes required hours at
−40 °C. The rate constants for nucleophilic attack by
arylamines in THF (Table 2, entries 2, 6, and 8) were found
to be more than an order of magnitude larger than the rate
constants for conversion of the minor to major diastereomers.
For example, the rate constant for nucleophilic attack on the
major diastereomer of 2f-TFA-(S,S,S,S) was about fifty times
faster than the 5.4 × 10−5 s−1 rate constant at −40 °C for
epimerization of the minor diastereomer of 2f-(S,S,S,S)
deduced from the data in Figure 5. The rate constants for
nucleophilic attack by arylamines on 2c and 2d in CH2Cl2
(Table 2, entries 1, 4, and 5) were found to be comparable to
the rate constant for isomerization of the minor diastereomer of
2f to the major diastereomer of 2f at −30 °C. The identity of
the anion of the allyliridium complex had a strong influence on
the rate of nucleophilic attack; more than an order of
magnitude difference was observed between the rate constant
for nucleophilic attack on 2f-TFA and 2f-BF4 (Table 2, entries
8 and 9). Nucleophilic attack by acetate was too fast to measure
at −60 °C (Table 2 entry 7). This final result agrees with our
conclusion published previously in Communication form on
the reactions of amines with allylic carbonates16,20 indicating
that oxidative addition of allylic esters is reversible, as discussed
further later in this Article.
These rate constants were measured at temperatures that
were lower than those of the catalytic reactions conducted at
room temperature. Thus, we determined the activation
parameters for the nucleophilic attack of aniline on the major
diastereomer of 2f-BF4 and extrapolated the rate constants to
the 25−50 °C temperatures of the catalytic reactions. Because
data from the appearance and decay curves (Figure 5) would
not give data sufficiently accurate for this analysis of the relative
rates for epimerization and nucleophilic attack, we assessed the
temperature dependence of this process by assuming that the
entropy of activation for this intramolecular reaction was small.
The details of the analysis can be found in the Supporting
Information. This analysis shows that the rate for attack of
aniline on the major diastereomer of the allyliridium complex
2f-TFA is similar to the rate for epimerization of the minor
diastereomer of 2f-TFA to the corresponding major diaster-
eomer.31 Because we showed that nucleophilic attack on the
minor diastereomer is much faster than attack on the major
diastereomer (vide infra), this analysis implies that nucleophilic
attack on the minor diastereomer is faster than epimerization of
the minor to major diastereomer. Moreover, we have shown
that the major diastereomer is much more stable than the
minor diastereomer, so the rate of conversion of the major
diastereomer to the minor diastereomer is roughly 2 orders of
magnitude or more slower than the rate of conversion of the
minor diastereomer to the major diastereomer. Thus, attack of
an arylamine on the major diastereomer should be faster than
epimerization of this stereoisomer to the minor diastereomer
under the catalytic conditions.
Excesses of the Products of Nucleophilic Attack on the
Allyliridium Complexes. The samples of 2-bromocinnamyliri-
dium complex 2e and phenethyl-substituted allyliridium
complex 2f containing mixtures of diastereomers allowed us
to gain direct information on the relative rates for reaction of
diastereomeric allyliridium complexes with an amine as
nucleophile. The enantiomeric excesses from the reactions of
mixtures of 2e-(R,R,R,S) and 2e-(R,R,R,R) with aniline and
the sodium salt of dimethyl malonate are summarized in
Scheme 4. The reaction of aniline with an 80:20 ratio of
Scheme 4. Stoichiometric Reactions of 2e
diastereomers of 2e-BF4 generated by exchanging the chloride
anion with a more weakly coordinating BF4− anion formed the
N-phenyl allylic amine product in 60% ee. The enantioselec-
tivity of this reaction corresponds precisely to the ratio of
diastereomers. The reaction of aniline with a sample of complex
2e-BF4 consisting of a higher 95:5 ratio of diastereomers
obtained after recrystallization from benzene formed the
substitution product in a higher 92% ee. This enantioselectivity
also corresponds well with the initial ratio of diastereomers in
the sample of 2e-BF4. When the same sample of 2e-BF4
consisting of a 95:5 ratio of diastereomers was allowed to
react with sodium dimethylmalonate, the product of
nucleophilic attack was formed in a slightly lower 83% ee,
but this ratio of products (91.5:8.5 er) remains similar to the
ratio of diasteromers. These results are consistent with the
conclusion from kinetic measurements that the diastereomeric
complexes 2e containing noncoordinating anions react with the
nucleophiles faster than they interconvert.
4b. Rate Constants for Reactions of Individual Diastereo-
meric Allyliridium Complexes with Amine Nucleophiles. The
ability to generate mixtures of diastereomers that do not
interconvert allowed us to measure directly the rate constants
for reactions of the major and minor diastereomers of 2-
bromocinnamyl complex 2e with amine nucleophiles. The rates
of these reactions were measured at −60 °C in CH2Cl2. The
data were collected with 10 equiv of PhNH2 as nucleophile and
triethylamine to quench the acid formed by the substitution
process. We were unable to collect full reaction progress curves
for nucleophilic attack on both diastereomers because the
signal-to-noise ratio for the resonance of the minor isomer was
low after the first half-life. However, we were able to gain
information on the relative reactivity of the two diastereomers
toward nucleophilic attack from the initial rates.
4. Reactions of Diastereomeric Allyliridium Com-
plexes. 4a. Relationship between the Diastereomeric
Excesses of the Allyliridium Complexes and the Enantiomeric
8142
dx.doi.org/10.1021/ja212217j | J. Am. Chem. Soc. 2012, 134, 8136−8147

Journal of the American Chemical Society
Article
Plots showing the initial rates for reaction of the major and
minor diastereomers are shown in Figure 6. The initial rate of
The similarity of these rate constants for nucleophilic attack
on the two diastereomers (with reaction of the diastereomer
that forms the minor enantiomer being slightly larger) suggests
that the high enantioselectivity of the catalytic reaction must
result from a highly selective oxidative addition of the allylic
ester. This oxidative addition could be irreversible under the
catalytic conditions, in which case the enantioselectivity would
result from the kinetic ratio of diastereomers formed by this
initial step, or it could be reversible under the catalytic
conditions. In the latter case, Curtin−Hammett conditions
would be established through reversible oxidative addition. We
have shown that carboxylates react with the allyliridium
intermediate to form free allylic esters,16 showing that the
oxidative addition can be reversible (Table 2, entry 7). The
reversibility of the oxidative addition step likely depends on the
nucleophile. This step is clearly reversible for the reactions of
amines, but it could be irreversible for reactions with anionic
nucleophiles or reactions conducted with a precatalyst
containing a weakly coordinating anion that would exchange
with the carbonate anion.21 In either case, the enantioselectivity
is based on a high stereoselectivity for oxidative addition, and
this mode of enantioselection is different from that proposed to
control enantioselectivity with palladium complexes that
undergo facile η3-η1-η3 interconversions.32,33
If the high enantioselectivity results from a high stereo-
selectivity for oxidative addition, then the catalytic reactions of
2-bromocinnamyl carbonates should occur with ee’s that are
modest (the stereoselectivity for oxidative addition is low, as
described earlier in this paper). Consistent with this assertion,
the reactions of 2-bromocinnamyl ethyl carbonate 6 with
aniline, sodium dimethylmalonate, and lithium phenoxide
catalyzed by 1a-(R,R,R) form the substitution product with
25−29% ee (Table 3).34
Table 3. Catalytic Reactions of 2-Bromocinnamyl Carbonate
6 with Aniline, Lithium Phenoxide, and Na-
dimethylmalonate
nucleophile (product)
yield, %
7/8
ee, % T, °C time, h
PhNH2 (7a)
26
90
53
99:1
99:1
99:1
29
25
26
50
12
12
Figure 6. Initial rates of reaction of aniline with the major and minor
diastereomers of 2e-BF4 at −60 °C.
(CH3CO2)2CHNa (7b)
PhOLi (7c)
RT
50
0.5
reaction of the major diastereomer was 6.5 × 10−6 M·s−1, and
the initial rate of reaction of the minor diastereomer was 1.1 ×
10−6 M·s−1. The pseudo-first-order rate constants (kobs) for the
nucleophilic attack on each of the diastereomers can be
calculated from these rates by dividing them by the initial
concentrations of the two diastereomers. The two rate
constants are kmajor = 1.2 × 10−4 s−1 and kminor = 2 × 10−4
s−1. These values calculated from initial rates show that the rate
constant for reaction of the minor diastereomer is about two
times larger than that for reaction of the major diastereomer
when the reaction is conducted under the same conditions for
both major and minor allyliridium complexes.
We also sought to determine the relative rates for
nucleophilic attack on the two diastereomeric allyl complexes
in cases in which the enantioselectivity for the corresponding
catalytic reaction was high. To do so, we measured the rates of
nucleophilic attack on major and minor isomers of
phenethylallyl complex 2f (Scheme 5). A 1:1 ratio of 2f-
TFA-(S,S,S,S) and 2f-TFA-(S,S,S,R) was generated by the
reaction of Ir(I) ethylene complex 1a-(S,S,S) and branched
allylic trifluoroacetate 5b-(S). Reactions of the two diastereo-
meric allyl complexes 2f with aniline as nucleophile and
triethylamine as proton acceptor revealed that the nucleophilic
attack of aniline on 2f-TFA-(S,S,S,R) is much faster than the
8143
dx.doi.org/10.1021/ja212217j | J. Am. Chem. Soc. 2012, 134, 8136−8147

Journal of the American Chemical Society
Article
Scheme 5. Relative Rates of Nucleophilic Attack on 2f-TFA-
(S,S,S,S) and 2f-TFA-(S,S,S,R)
nucleophilic attack on 2f-TFA-(S,S,S,S) and that nucleophilic
attack on both diastereomers is faster than interconversion of
the two diastereomers containing the trifluoroacetate counter-
ion (vide supra). The reaction of 2f-TFA-(S,S,S,R) occurred to
completion in less than 3 min at −60 °C, while the reaction of
2f-(S,S,S,S) occurred to less than 5% conversion over the same
time. The rate constant for nucleophilic attack by aniline on the
more stable allyliridium diastereomer 2f-(S,S,S,S) was
determined quantitatively with the material prepared from the
reaction of ethylene-ligated complex 1a-(S,S,S) and linear
allylic trifluoroacetate 5l (Scheme 3). The rate constant at −40
°C in the presence of triethylamine as base was found to be 2.5
× 10−3 s−1 (Table 2, entry 8). The difference in rate constants
for attack on these diastereomers is larger than that for attack
on the 2-bromocinnamyl diastereomer and, again, favors
formation of the minor enantiomer, but the stereoselectivity
of the oxidative addition step is sufficiently high to cause the
overall process to occur with high enantioselectivity.
5. Investigation of the Stereochemistry of the
Individual Steps of the Catalytic Reaction. Prior studies
of the stereochemistry of iridium-catalyzed allylic substitution
reactions have shown that the reaction occurs with overall
retention of configuration.9,22 To determine if this stereo-
chemical outcome results from two steps of the catalytic cycle
occurring by inversion of configuration or two steps occurring
by retention of configuration, we determined the stereo-
chemical changes that occur during the reaction of ethylene-
ligated complex 1a-(R,R,R) with an enantioenriched, deute-
rium-labeled linear allylic trifluoroacetate 4a-D. In addition, we
determined the stereochemical outcome of nucleophilic attack
on the isolated allyliridium complexes that result from this
oxidative addition. Finally, we conducted the reactions of
enantioenriched, monodeuterated allylic carbonate 4b-D25
catalyzed by the two enantiomers of the cyclometalated iridium
catalysts 1a and 1b to determine the relationship between the
data gained on the stoichiometric reactions of the allylic
trifluoroacetate and the data gained on the true catalytic system
with allylic carbonates.
chemistry of the individual steps of molybdenum-catalyzed
allylic substitution reactions.25,35
Reaction of the Ir(I) complex 1a-(R,R,R) with the
monodeuterated, enantioenriched cinnamyl trifluoroacetate
4a-D (the enantiomeric ratio of the precursor alcohol was
>95:5),24,35,36 followed by anion exchange with AgBF4, formed
an 83:17 ratio of allyliridium complexes containing the
deuterium label syn and anti to the metal center, respectively
(eq 3). The ratio of the complexes was determined by
1
integration of the H NMR signals corresponding to syn and
anti protons at the terminal carbon of the allyl unit. This result
is consistent with an oxidative addition process that occurs with
inversion of configuration. The lack of exclusive formation of
one stereoisomer is likely due to partial racemization of the
substrate during esterification of the alcohol. We cannot
determine the enantiomeric excess of the trifluoroacetate
directly, but this result shows that the reaction occurs, at
least predominantly, with inversion of configuration.
The stereochemistry of nucleophilic attack on the allyliridium
intermediate was determined by reaction of aniline with the
cinnamyliridium complex consisting of an 83:17 ratio of syn and
anti isomers. This stoichiometric reaction gave the (R)
enantiomer of the allylic substitution product 9 in 96% ee.
The ratio of products Z-9-(R)/E-9-(R) was 8:1, with a small
amount of product formed that lacked the deuterium label.
This result is consistent with nucleophilic attack occurring with
almost complete inversion of configuration. These two results
show that the mechanism of iridium-catalyzed allylic sub-
stitution involves two inversions of configurations, one during
oxidative addition of the allylic ester and one during reductive
elimination that occurs by nucleophilic attack of the
nucleophile on the allyl intermediate.
5b. Catalytic Reactions of Enantioenriched, Deuterium-
Labeled Linear and Branched Allylic Electrophiles. Several
differences between the stoichiometric reactions of section 5a
and the catalytic system could compromise the conclusions
drawn from the reactions of the trifluoroacetate. First, the
stoichiometric reactions were conducted with the highly
reactive cinnamyl trifluoroacetate 4a-D, whereas catalytic
reactions are typically conducted with the less reactive alkyl
carbonates or acetates. Second, the stoichiometric reactions of
the allylic electrophiles with amine and malonate nucleophiles
were conducted with the cinnamyliridium complex containing a
weakly coordinating BF4− counterion and with added amine as
the proton acceptor, whereas the nucleophilic attack that takes
place during the catalytic reactions occurs on an intermediate
5a. Stoichiometric Reactions of the Ir(I) Ethylene Complex
1a-(R,R,R) with an Enantioenriched, Deuterium-Labeled
Linear Allylic Trifluoroacetate and Reactions of the Addition
Product with an Amine Nucleophile. The changes in
configuration from the oxidative addition and reductive
elimination steps were determined directly by conducting
reactions of enantioenriched, monodeuterated, linear allylic
trifluoroacetate 4a-D-(S) (eq 3). The position of deuterium in
the products of nucleophilic addition to these allyliridium
complexes, coupled with the absolute configuration of the
products, reveals the stereochemistry of the reaction of the allyl
complex with nucleophiles. These studies are based on the
system developed by Lloyd-Jones for determining the stereo-
8144
dx.doi.org/10.1021/ja212217j | J. Am. Chem. Soc. 2012, 134, 8136−8147

Journal of the American Chemical Society
Article
containing either a carbonate, an acetate, or an alkoxide
counterion without an external base.
configuration would give (Z)-9-D-(S). Similarly, two different
products would form from nucleophilic attack on the syn-D-
major and syn-D-minor allyliridium complexes. Nucleophilic
attack on syn-D-major with retention of configuration would
give (E)-9-D-(S) as the major product, and nucleophilic attack
on syn-D-major with inversion would give (E)-9-D-(R) as the
major product. Nucleophilic attack on syn-D-minor with
retention would give (E)-9-D-(R) as the major product, and
nucleophilic attack on syn-D-minor with inversion would give
(E)-9-D-(S) as the major product.
Thus, we studied catalytic reactions of methyl cinnamyl
carbonate 4b-D-(R), an enantioenriched, deuterium-labeled
version of the allylic carbonate used in the catalytic process.
This substrate was prepared from the enantioenriched, labeled
allylic alcohol we had prepared for synthesis of the allylic
trifluoroacetate. The potential products of the reactions of this
labeled linear allylic carbonate are shown in Scheme 6.
Results from the catalytic allylic substitution of enantioen-
riched monodeuterated linear substrate 4b-D-(R) are provided
in Table 4. The reaction of 4b-D-(R) with aniline catalyzed by
Scheme 6
Table 4. Catalytic Reactions of Enantioenriched,
Monodeuterated Allylic Carbonates with Aniline Catalyzed
by Ethylene-Ligated Iridium Catalysts 1a and 1b
entry
catalyst
substrate
(Z)-9/(E)-9
ee, %
1
2
3
1a-(R,R,R)
1a-(S,S,S)
1b-(S,S,S)
4b−D-(R)
4b−D-(R)
4b−D
1:18
18:1
19:1
96 (R)
96 (S)
98 (S)
1a-(R,R,R) gave (E)-9-D-(R) with 96% ee and a 18:1 ratio of
E/Z isomers (Table 4: entry 1). The observed (R) and (E)
configurations of the major product from the reaction catalyzed
by the complex with (R,R,R) configurations is consistent with
the sequence involving two inversions or two retentions
(Scheme 6). The sequence involving two retentions can be
ruled out, based on the outcome of the stoichiometric reactions
(vide supra).
A similar stereochemical scenario can be proposed for the
allylic substitution of the same monodeuterated, enantioen-
riched linear allylic electrophile 4b-D-(R) with aniline catalyzed
by the antipodal catalysts 1a or 1b possessing (S,S,S) absolute
configuration. In this case, the opposite set of stereoisomers to
those in Scheme 6 would be expected to form by each of the
eight mechanisms. In that event, the reaction of 4b-D-(R) with
aniline catalyzed by 1a-(S,S,S) gave the allylic substitution
product (Z)-9-D-(S) with 96% ee and a 18:1 Z/E ratio (Table
4: entry 2), and the same reaction catalyzed by 1b-(S,S,S)
formed the substitution product (Z)-9-D-(S) with 98% ee and
a 19:1 Z/E ratio (Table 4: entry 3). These results are also
consistent with an iridium-catalyzed allylic substitution by the
sequence involving two inversions. Thus, our data from the
stoichiometric reaction and from the catalytic reaction are
consistent with each other.
6. Comparison of the Origins of Enantioselectivity in
Allylic Substitutions Catalyzed by Iridium, Palladium,
and Molybdenum Complexes. The origins of enantiose-
lectivity and the changes in configuration that occur during the
iridium-catalyzed allylic substitution are distinct from those of
catalysts based on complexes of the other metals most
commonly used for allylic substitution. Most mechanistic
studies have been conducted with palladium and molybdenum
catalysts. The step that controls enantioselectivity is distinct
from that of palladium systems, and the changes in
configuration that occur during the molybdenum-catalyzed
Oxidative addition with retention of configuration would
generate the allyl intermediate with deuterium in the position
anti to the central hydrogen of the allyl unit (labeled as anti-D-
major), and oxidative addition with inversion of configuration
would generate the allyl intermediate with deuterium in the
position syn to the central hydrogen (labeled as syn-D-major).
In addition, the minor allyliridium diastereomers similar to
those of 2-bromocinnamyl and phenethyl-substituted allyl
complexes 2e and 2f, respectively, can be formed (labeled as
anti-D-minor and syn-D-minor) through η3-η1-η3 interconver-
sion. An interconversion from the η3 form of anti-D-major to
the η1 isomer and back to the η3 generates syn-D-minor, and an
interconversion from the η3 form of syn-D-major to the η1
isomer and back to the η3 form generates anti-D-minor.
Nucleophilic attack with retention of configuration on the anti-
D-major isomer shown would give (Z)-9-D-(S) as the organic
product. Nucleophilic attack on anti-D-major with inversion of
configuration would give (Z)-9-D-(R) as the major product.
The opposite outcome would be expected from nucleophilic
attack on anti-D-minor. Nucleophilic attack on anti-D-minor
with retention of configuration would give (Z)-9-D-(R), and
nucleophilic attack on anti-D-minor with inversion of
8145
dx.doi.org/10.1021/ja212217j | J. Am. Chem. Soc. 2012, 134, 8136−8147

Journal of the American Chemical Society
Article
reactions are different from those that occur during the iridium-
catalyzed reactions.
major diastereomer, but the major diastereomer forms the
major enantiomer of the product. This scenario can occur in a
system that leads to high enantioselectivity because the kinetic
and thermodynamic selectivity for formation of the two
diastereomers is even higher than the ratio of rate constants
for reaction of the diastereomeric intermediates that favors
formation of the minor enantiomer.
Many versions of palladium-catalyzed asymmetric allylic
substitution have been reported. In general, the configuration of
the stereocenter formed at the allyl unit is controlled by the
step involving nucleophilic attack on the coordinated ligand.
Many palladium-catalyzed reactions involve intermediates
containing symmetric allyl groups. In this case, the
enantioselectivity is controlled by the relative rates for attack
at one terminus of the allyl group over the other. In other cases,
palladium-catalyzed reactions of linear or branched, mono-
substituted allylic esters that form products from attack at the
more substituted position of the allyl group occur with higher
enantioselectivity when the rate of η3-η1-η3 interconversions is
faster than the rate of nucleophilic attack, presumably because
the difference in rates of reactions of the diastereomeric allyl
complexes is much larger than the equilibrium ratio of
diastereomers.32,37,38 The rate of η3-η1-η3 interconversions in
palladium systems is typically faster than the rate of
nucleophilic attack, most likely because the first step of this
interconversion involves association of the counterion with the
16-electron metal center. In contrast to the mode of
stereoselection proposed to apply to most palladium-catalyzed
asymmetric allylic substitution reactions, the stereoselectivity of
iridium-catalyzed allylic substitution originates from the
oxidative addition step. The kinetic selectivity for formation
of the more stable diastereomer and the relative ratios of the
more stable diastereomer over the less stable diastereomer are
both sufficiently high to compensate for the faster rate of attack
of the nucleophile on the minor diastereomer. In these iridium-
catalyzed allylic substitutions, nucleophilic attack on the minor
diastereomeric allyl complex leads to the minor enantiomer of
the organic product.
A series of asymmetric allylic substitutions also have been
published with molybdenum catalysts,39−41 and many reactions
with stabilized carbon nucleophiles occur with high enantiose-
lectivity. Like the iridium-catalyzed allylic substitutions, the
molybdenum-catalyzed reactions tend to form high ratios of
branched to linear substitution products. Thus, one might
expect that the origins of stereoselection by the two systems
might be similar to each other. However, the studies we report
on the changes in configuration during the iridium-catalyzed
process show clearly that the iridium and molybdenum systems
are distinct. The oxidative addition and reductive elimination
steps of the molybdenum-catalyzed reaction occur with
retention of configuration,25 whereas these steps of the
iridium-catalyzed process occur with inversion of configuration.
Finally, the relative rates for reactions of the diastereomeric
intermediates in the iridium-catalyzed allylic substitutions are
much different from those measured previously for asymmetric
hydrogenation. The study by Halpern on the asymmetric
hydrogenation of methyl acetamidocinnamate showed that the
minor diastereomer formed the major enantiomer because the
two diastereomers equilibrate and reaction of the minor
diastereomer was several orders of magnitude faster than
reaction of the major diastereomer.42 In contrast, Bergens has
shown that the major diastereomer gives rise to the major
enantiomer in ruthenium-catalyzed hydrogenations of ketones,
and Curtin−Hammett conditions are not achieved because the
formation of the hydrogenation product is faster than the
interconversion of the diastereomeric intermediates.43 For the
iridium system that catalyzes allylic substitution analyzed in the
current work, the minor diastereomer reacts faster than the
III. SUMMARY
Studies on the origins of stereoselectivity from iridium-
catalyzed allylic substitution reactions showed that the factors
controlling enantioselectivity are distinct from those of systems
studied previously. These studies were enabled by the synthesis
and characterization of a series of allyliridium complexes
generated both as pure diastereomers and as mixtures of
diastereomers. The following conclusions were drawn from our
data.
(1) The more stable diastereomer forms the major
enantiomer of the organic product, but reaction of the
more stable diastereomer is slower than reaction of the
minor diastereomer. Thus, the oxidative addition of the
allylic carbonate principally controls the enantioselectiv-
ity of the process.
(2) Epimerization of the allyl intermediate containing a
weakly coordinating anion occurs slowly, even at room
temperature. This rate for epimerization is much slower
than the rates for epimerization with palladium systems.
This difference in rates of epimerization, most likely,
results from a difference in coordination sphere. The
iridium complex is a pseudo-octahedral, 18-electron d6
system, a class of complex that typically undergoes ligand
substitution by dissociative processes, while the palla-
dium complexes are pseudo square-planar, 16-electron d8
systems, which typically undergo ligand substitution by
associative processes.
(3) The rate for epimerization of individual diastereomeric
allyliridium complexes containing weakly coordinating
anions is slower than the rate for reaction of the
respective diastereomers by attack of typical nucleophiles
on the allyl group.
(4) Equilibration of the diastereomeric allyliridium com-
plexes likely occurs by reversible oxidative addition.
Nucleophilic attack of carboxylate anions on the
allyliridium intermediate is fast and thermodynamically
downhill, making the oxidative addition process uphill
and reversible.42
(5) The iridium-catalyzed allylic substitution occurs by an
overall retention of configuration that results from two
inversions. Oxidative addition of allylic esters occurs by
inversion of configuration, and attack of the nucleophile
on the allyl intermediate occurs by inversion of
configuration. These changes in configuration during
the reaction contrast those of the molybdenum system
analyzed previously in a similar fashion.
(6) The structures of the allyl intermediates containing
different substituents on the allyl unit vary by the length
of the metal−carbon bond to the substituted terminus,
but this change in distance does not significantly affect
the C−C bond distances, and the regioselectivity of the
substitution process is high when the difference in M−C
bond lengths is only 0.07 Å, as well as when this
difference is over 0.27 Å.
8146
dx.doi.org/10.1021/ja212217j | J. Am. Chem. Soc. 2012, 134, 8136−8147

Journal of the American Chemical Society
Article
(26) The bond angles in the structures of allyliridium complexes 2a−
e involving C47 and C45 also change with substitution at the allyl
terminus. The bond angle between cent1-Ir-C47 (centroid of the
alkene unit cis to the allyl-Ir-branched allylic terminus) is 101−103 for
2a, 2b, and 3a and 105−106 for 2c−e. The C47-Ir-C45 angle which
corresponds to the bite angle of the allyl unit is ∼66 for 2a,b and 3a
and 62−63 for 2c−e. The angle between the Cent2 (centroid of the
alkene unit in the axial position of the complex) and C45 (the
unsubstituted terminus of the allyl unit) increases with increasing
substitution on the allyl: the Cent2-Ir-C45 bond angle is 89−91 for
2a,b and 3a and 93−94 for 2c−e. As a consequence, the angle
between C21 (cyclometallated carbon) and C45 decreases in the
series: C21−Ir−C45 bond angle is ∼103 for 2a,b, 105 for 3a, and
∼100 for 2c−e. The opposite can be observed for the branched allylic
terminus, where the bond angle between Cent2 (axial centroid) and
C47 (branched allylic terminus) decreases with the size of substitution
on the allyl: the Cent2−Ir−C47 bond angle is 96−97 for 2a,b, 3a and
∼92 for 2c−e. As a consequence, the angle between C21
(cyclometallated carbon) and C47 (branched terminus) increases in
the series: the C21−Ir−C47 bond angle is 90−91 for 2a,b, 92 for 3a,
and 95−96 for 2c−e. These differences between the structures can
also be seen in Figure 3. And finally, the angle between the allylic
carbons C45−C46−C47 is 121−122 for 2a,c−e, 3a and 127.5(3) for
2b.
(27) Swain, C. G. J. Am. Chem. Soc. 1944, 66, 1696.
(28) Neukom, J. D.; Perch, N. S.; Wolfe, J. P. J. Am. Chem. Soc. 2010,
132, 6276.
(29) The initial concentration of amine in a typical iridium-catalyzed
allylic substitution is 2−2.4 M.
(30) Polet, D.; Alexakis, A. Org. Lett. 2005, 7, 1621.
(31) See ref 29.
(32) Mackenzie, P. B.; Whelan, J.; Bosnich, B. J. Am. Chem. Soc. 1985,
107, 2046.
(33) Trost, B. M. Chem. Pharm. Bull. 2002, 50, 1.
Studies that further exploit these features of the catalytic
system, studies that alter the relative rates by variations in the
coordination sphere of the metal, and studies that reveal the
factors that control the site-selectivity of these reactions are in
progress.
ASSOCIATED CONTENT
* Supporting Information
■
S
Experimental procedures for the synthesis of 2c, 2d, 2e, and 2f-
BF4, kinetic studies, and studies on the stereochemical outcome
and catalytic reactions of 2-bromocinnamyl carbonate 5; along
with cif data. This material is available free of charge via the
Internet at http://pubs.acs.org.
AUTHOR INFORMATION
Corresponding Author
jhartwig@berkeley.edu
■
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
We thank the NIH (GM-55382) for support of this work and
Johnson-Matthey for [Ir(COD)Cl]2. We also thank Levi
Stanley for helpful discussions.
REFERENCES
■
(1) Ohmura, T.; Hartwig, J. F. J. Am. Chem. Soc. 2002, 124, 15164.
(2) Shu, C.; Leitner, A.; Hartwig, J. F. Angew. Chem., Int. Ed. 2004,
43, 4797.
(3) Pouy, M. J.; Leitner, A.; Weix, D. J.; Ueno, S.; Hartwig, J. F. Org.
Lett. 2007, 9, 3949.
(4) Yamashita, Y.; Gopalarathnam, A.; Hartwig, J. F. J. Am. Chem. Soc.
(34) As a side note, the yields of these reactions depended on the
strength of the nucleophile. The reaction of aniline with 6 gave a 26%
yield of product 7a, whereas the reaction of lithium phenoxide gave
the substitution product 7c in 53% yield, and that of sodium
dimethylmalonate gave the analogous product 7b in 90% yield.
(35) Malkov, A. V.; Gouriou, L.; Lloyd-Jones, G. C.; Stary, I.; Langer,
V.; Spoor, P.; Vinader, V.; Kocovski, P. Chem.Eur. J. 2006, 12, 6910.
(36) This trifluoroacetate was prepared from the known labeled
alcohol.
2007, 129, 7508.
(5) Ueno, S.; Hartwig, J. F. Angew. Chem., Int. Ed. 2008, 47, 1928.
(6) Pouy, M. J.; Stanley, L. M.; Hartwig, J. F. J. Am. Chem. Soc. 2009,
131, 11312.
(7) Stanley, L. M.; Hartwig, J. F. Angew. Chem. 2009, 121, 7981.
(8) Stanley, L. M.; Hartwig, J. F. J. Am. Chem. Soc. 2009, 131, 8971.
(9) Takeuchi, R.; Kashio, M. Angew. Chem., Int. Ed. 1997, 36, 263.
(10) Takeuchi, R.; Kashio, M. J. Am. Chem. Soc. 1998, 120, 8647.
(11) Takeuchi, R. Polyhedron 2000, 19, 557.
(37) Hayashi, T.; Kawatsura, M.; Uozumi, Y. Chem. Commun. 1997,
561.
(12) Tissot-Croset, K.; Polet, D.; Alexakis, A. Angew. Chem., Int. Ed.
2004, 43, 2426.
(38) Trost, B. M.; Toste, F. D. J. Am. Chem. Soc. 1999, 121, 3543.
(39) Trost, B. M.; Miller, J. R.; Hoffman, C. M. J. Am. Chem. Soc.
2011, 133, 8165.
(40) Belda, O.; Moberg, C. Acc. Chem. Res. 2004, 37, 159.
(41) Trost, B. M.; Dogra, K.; Franzini, M. J. Am. Chem. Soc. 2004,
126, 1944.
(13) Alexakis, A.; Polet, D. Org. Lett. 2004, 6, 3529.
(14) Graening, T.; Hartwig, J. F. J. Am. Chem. Soc. 2005, 127, 17192.
(15) Weix, D.; Hartwig, J. F. J. Am. Chem. Soc. 2007, 129, 7720.
(16) Madrahimov, S. T.; Markovic, D.; Hartwig, J. F. J. Am. Chem.
Soc. 2009, 131, 7228.
(42) Halpern, J. Science 1982, 217, 401.
(43) Daley, C. J. A.; Bergens, S. H. J. Am. Chem. Soc. 2002, 124, 3680.
(17) Kiener, C. A.; Shu, C.; Incarvito, C.; Hartwig, J. F. J. Am. Chem.
Soc. 2003, 125, 14272.
(18) Stanley, L. M.; Bai, C.; Ueda, M.; Hartwig, J. F. J. Am. Chem. Soc.
2010, 132, 8918.
(19) Leitner, A.; Shekhar, S.; Pouy, M. J.; Hartwig, J. F. J. Am. Chem.
Soc. 2005, 127, 15506.
(20) Markovic, D.; Hartwig, J. F. J. Am. Chem. Soc. 2007, 129, 11680.
(21) Spiess, S.; Raskatov, J.; Gnamm, C.; Brodner, K.; Helmchen, G.
̈
Chem.Eur. J. 2009, 15, 11087.
(22) Bartels, B.; Garcia-Yebra, C.; Rominger, F.; Helmchen, G. Eur. J.
Inorg. Chem. 2002, 2002, 2569.
(23) Lopez, F.; Ohmura, T.; Hartwig, J. F. J. Am. Chem. Soc. 2003,
125, 3426.
(24) Raminelli, C.; Comasseto, J. V.; Andrade, L. H.; Porto, A. L. M.
Tetrahedron: Asymmetry 2004, 15, 3117.
(25) Lloyd-Jones, G. C.; Krska, S. W.; Hughes, D. L.; Gouriou, L.;
Bonnet, V. D.; Jack, K.; Sun, Y.; Reamer, R. A. J. Am. Chem. Soc. 2003,
126, 702.
8147
dx.doi.org/10.1021/ja212217j | J. Am. Chem. Soc. 2012, 134, 8136−8147