On the mechanism of the copper-catalyzed cyclopropanation reaction

Article abstract of DOI:10.1002/1521-3765(20020104)8:1<177::AID-CHEM177>3.0.CO;2-H

The selectivity-determining step in enantioselective copper-catalyzed cyclopropanation with diazo compounds has been studied by experimental and computational methods. The addition of the very reactive metallacarbene intermediate in an early transition st

Full text of DOI:10.1002/1521-3765(20020104)8:1<177::AID-CHEM177>3.0.CO;2-H

FULL PAPER
On the Mechanism of the Copper-Catalyzed Cyclopropanation Reaction
Torben Rasmussen,[b] Jakob F.Jensen, [a] Niels Èstergaard,[a, b] David Tanner,[a]
Tom Ziegler,[c] and Per-Ola Norrby*[a, b]
Abstract: The selectivity-determining
step in enantioselective copper-cata-
lyzed cyclopropanation with diazo com-
pounds has been studied by experimen-
tal and computational methods. The
addition of the very reactive metallacar-
bene intermediate in an early transition
state to the substrate alkene is concerted
but strongly asynchronous, with substan-
tial cationic character on one alkene
carbon in the neighborhood of the
transition state. Evidence from isotope
effects and Hammett studies supports
the nature of the transition state. For-
mation of a metallacyclobutane inter-
mediate by a [2‡2] addition is kineti-
cally disfavored. Ligand substrate in-
teractions influencing the enantio- and
diastereoselectivity have been identi-
fied, and the preferred orientation of
the alkene substrate during the addition
is suggested.
Keywords: asymmetric catalysis
¥
copper ¥ cyclopropanation ¥ density
functional calculations
mechanisms
¥ reaction
asymmetric cyclopropanation. However, as is often the case
for metal-mediated asymmetric catalysis, detailed mechanistic
understanding has lagged behind empirical catalyst develop-
ment. Since better mechanistic insight should aid in the
development of even more efficient catalysts, we now present
kinetic and computational studies of the copper-catalyzed
cyclopropanation of styrenes involving bis(oxazoline) ligands.
For the title reaction two points of mechanistic detail are
Introduction
In recent years, much effort has been devoted to the develop-
ment of methods for metal-catalyzed diastereo- and enantio-
selective cyclopropanation of olefins with diazo compounds.[1]
One of the most appealing catalytic systems relies on a copper
complex together with a C2-symmetric ligand, such as a
semicorrin[2] or a bis(oxazoline)[3] (Scheme 1). These ligands,
particularly the bis(oxazolines), which are easily available and
induce high levels of enantioselectivity for a range of olefinic
substrates, are currently the favorite ones for Cu-catalyzed
4]
widely accepted: that the actual catalyst is a CuI species,[1
even if CuII is used, and that a short-lived electrophilic
copper carbene intermediate is involved (Scheme 2).[1 6] We
CuI
Z
N2
R
N2
+
Z
R
ligand
ML*
R
Z
Scheme 1. Copper-catalyzed cyclopropanation.
N2
Z
ML*
R
Z
Scheme 2. Proposed catalytic cycle for the asymmetric cyclopropanation.
[a] Prof. P.-O. Norrby, J. F. Jensen, N. Èstergaard, Prof. D. Tanner
Department of Chemistry, Organic Chemistry
Technical University of Denmark
have used a combination of Hammett-type kinetic studies,
isotopic labeling, and high-level computational methods to
investigate two of the points of contention, namely the nature
of the step which controls the stereochemistry of the
reaction[2, 4, 7] and the participation of a metallacyclobutane
intermediate.[4]
Building 201, Kemitorvet, 2800 Kgs. Lyngby (Denmark)
Fax : (‡45)45933968
E-mail: pon@kemi.dtu.dk
[b] Prof. P.-O. Norrby, Dr. T. Rasmussen, N. Èstergaard
Department of Medicinal Chemistry
Royal Danish School of Pharmacy
Universitetsparken 2, 2100 Copenhagen (Denmark)
[c] Prof. T. Ziegler
Department of Chemistry, University of Calgary
2500 University Drive, N.W. Calgary
Alberta T2N1N4 (Canada)
Results and Discussion
Kinetic studies: The relative rates of cyclopropanation of the
substituted styrenes (3 8) used in the Hammett and isotope
Supporting information for this article is available on the WWW under
http://wiley-vch.de/home/chemistry/ or from the author.
Chem. Eur. J. 2002, 8, No. 1
¹ WILEY-VCH Verlag GmbH, 69451 Weinheim, Germany, 2002
0947-6539/02/0801-0177 $ 17.50+.50/0
177

P.-O. Norrby et al.
FULL PAPER
studies are shown in Table 1. All experimental procedures are
available as Supporting information. In calculations of rate
constants we assumed pure first-order kinetics in substrate,
Table 1. Rates of cyclopropanation of substituted styrenes relative to
styrene (2).
‡
Substrate
Substituent
s
s
krel
Figure 1. Competition of isotopically substituted styrenes.
3
4
5
6
7
8
p-MeO
p-Me
p-CF3
p-NO2
a-[D]
b,b-[D2]
À 0.28
À 0.78
2.52
1.54
À 0.14
À 0.31
0.53
0.81
0.61
0.79
0.49
0.42[a]
1.02 Æ 0.02[b]
0.95 Æ 0.02[b, c]
[a] Measured by NMR. [b] The rates for deuterated styrenes were
measured in competition with p-methylstyrene (4) and recalculated to
give rates relative to styrene. The limits are the sums of the standard errors
for the two determinations, as determined by linear regression using
Equation (1). [c] The isotope effect per deuterium atom is 0.974.
that is, ln([A]0/[A]) ˆ kt where [A] is any styrene concen-
tration. To avoid influences from other steps in the catalytic
cycle (such as formation of the metal carbene), relative rate
constants were obtained from competition experiments
according to Equation (1). The slopes in Figure 1 for isotopi-
ln([A]0/[A]) ˆ krel ln([B]0/[B])
(1)
‡
Figure 2. Hammett study, showing correlation with s and s .
cally substituted styrenes in competition with p-methylstyrene
(4) give rates relative to that standard, whereas Table 1 has
been recalibrated to show rates relative to unsubstituted
styrene (2). The Hammett results are plotted as log(krel)
isotope effects, but the data are
consistent with the concerted
but asynchronous transition
state X, where the new bond to
the b-carbon is formed well
before the bond to the a-carbon.
‡
versus s or s in Figure 2.
It is clear from Figure 2 that the data fit better to s than to
‡
s values. In combination with the modest 1 ˆ À0.51, this
indicates an early transition state with substantial positive
charge being developed at the benzylic position and is in
À
qualitative agreement with studies of Si H insertion, showing
Initial computational studies: Our first goal was to character-
ize the species in the postulated catalytic cycle (Scheme 2) for
a small model system.[10] We selected the model reaction of
ethene with diazoacetic acid catalyzed by the CuI complex of
1,4-diazabutadiene. For the initial studies, we employed the
B3LYP functional[11] in Gaussian98[12] together with the
6-311 ‡ G all-electron basis set for Cu and 6-31G set for the
remaining elements. The B3LYP level generally gives a good
performance for transition metal complexes,[13] and has
previously been shown to yield good bond energies for
copper compounds.[14] Both reactions in the catalytic cycle are
that the postulated copper carbene is a very reactive electro-
phile.[8] In an earlier Hammett study of cyclopropanation
employing an anionic ligand a correlation with s and a more
negative 1 value were found.[9] This difference is under-
standable, since the neutral Cu complex should be less
electrophilic than the cationic complexes employed here.
The isotope effects indicate that the b-carbon has been
somewhat rehybridized in the transition state (TS), whereas
no significant effect can be seen at the a-position (Table 1).
There can be many reasons for the absence of observable
178
¹ WILEY-VCH Verlag GmbH, 69451 Weinheim, Germany, 2002
0947-6539/02/0801-0178 $ 17.50+.50/0
Chem. Eur. J. 2002, 8, No. 1

Copper-Catalyzed Cyclopropanation
177 184
exothermic (Scheme 3), which is one of the requirements for
an efficient catalytic cycle. The geometries and raw energies
are available as Supporting information. Other important
factors which can influence catalytic efficiency include side
paths and the nature of the rate-limiting step,[10] but in this
work we were mainly concerned with factors affecting the
selectivity of the reaction.
COOH
NH
H
10
Cu
N2
HN
ˆ À
Figure 3. Overlay of the Cu C C moieties of 10 and 12a 12c.
E -58 kJ mol-1
E -138 kJ mol-1
Cu
HN
‡
À
Cu and C, so it should more properly be drawn as Cu C ,
corresponding to the lone pair of a singlet carbene coordi-
nated to a CuI cation. Again, this corresponds well with the
observed electrophilic character of 10/12. The analysis also
points to a slight donation from the carbonyl oxygen to the
carbene carbon as the major source of the strong orthogonal
preference of the carboxylate moiety.
NH
N2
COOH
COOH
11
Scheme 3. Computational small model system, with reaction energies.
Experimentally relevant models (12 and 13) of the two
metal complexes were also investigated utilizing the BP
functional[15] in ADF.[16] Most of the system is described by a
The selectivity-determining step must be the addition of the
metallacarbene (10 or 12) to the substrate alkene, even
though formation of the metallacarbene has been shown to be
turnover-limiting.[5, 10] We have therefore focused our atten-
tion on reaction of the metallacarbene with alkene. Early in
our studies it became clear that investigation of this step
posed some severe problems. In all our model systems, the
reaction between the metal carbene and alkene displays a
monotonic downhill energy profile, with no transition state on
the potential-energy surface. This is apparently in violation of
the experimental evidence that there is a real transition state
for the addition, since a substrate-dependent selectivity was
observed in the competition studies. In a diffusion-controlled
reaction, substrate properties should have no influence on the
relative rates of reaction. There can be several reasons for the
apparent failure to find a TS in the calculations, each of which
can be addressed by computation.
+
BP/DZ*
COOMe
H
BP/TZ+
BP/DZ
+
R
R
N
R
R
Cu
Cu
N
N
N
O
O
O
O
a: R = H
b: R = iPr
c: R = tBu
12
13
double-z Slater-type basis (DZ). For the heavy atoms in the
carbene moiety, we added a polarization function (DZ*),
whereas the copper atom was described by a triple-z basis
with added diffuse functions, TZ ‡ (types II, III, and IV in
ADF, respectively). The geometries and raw energies are
available as Supporting Information.
The geometries of the catalyst complexes 11 and 13 are
unsurprising, being near-planar and with the coordination
geometry determined by the ligand. Thus, in the small model
system 11 the N-Cu-N bond angle is constrained to 838,
whereas the same angle in the more realistic model systems 13
1) The real free-energy transition state could exist without a
corresponding TS on the potential energy surface. A TS
may be located by direct calculation of the free-energy
surface.
2) The small model system 10 may not reflect the true
reaction closely enough. However, 12a 12c correspond
to experimentally tested systems.
3) It may be necessary to include the effect of solvation to
find a TS.[18]
4) The level of theory may be insufficient.
À
is 1108 Æ 28. Similarly, the Cu N bond is 2.033 ä in 11 and
decreases to 1.90 1.91 ä in 13, very close to the observed
value (1.88 ä) for an oxazoline CuI bond.[17]
The calculated metallacarbenes 10 and 12 display some less
intuitively obvious features (Figure 3). The carbenoid carbon
ˆ À
is almost planar, with the Cu C C plane orthogonal to the
We do not have sufficient computational resources to
address all of these points in one calculation. Instead, we
implemented a combination method that has been successful
in earlier cases.[18] Several points on the path were located by
scanning along the reaction coordinate, whereupon all con-
tributions to the final energy were calculated at each point
and added together to form a composite high-level free
energy. We expect this method to yield barriers of sufficient
accuracy to allow correlation with experimental data and to
serve as a starting point for future development of a Q2MM
predictive model.[19]
N-Cu-N plane. The plane of the carboxylate moiety is in turn
orthogonal to the plane through the carbenoid carbon,
minimizing conjugation. This is consistent with a strong
cationic character on the carbenoid carbon. The force
counteracting conjugation is quite strong; in 12c, in which
the tert-butyl group interacts closely with the carboxylate, the
crowding is minimized mainly by distortion of the bis(oxazo-
line) ligand, not of the carboxylate.
A population analysis reveals that the d functions of Cu do
not participate to any significant degree in the bond between
Chem. Eur. J. 2002, 8, No. 1
¹ WILEY-VCH Verlag GmbH, 69451 Weinheim, Germany, 2002
0947-6539/02/0801-0179 $ 17.50+.50/0
179

P.-O. Norrby et al.
FULL PAPER
To achieve the smooth energy surface, without sudden
jumps in unconstrained coordinates, which is necessary for the
combination procedure to be successful, the scanning coor-
dinate(s) must be chosen with care and should also be as few
as possible to avoid the added complication of a multidimen-
sional surface. When the distance from the carbene carbon to
the approaching alkene was chosen as first scanning coor-
dinate, a discontinuous surface was obtained at which the
bond to the unconstrained carbon was formed suddenly
between one scanning point and the next. This problem could
be overcome by taking the angle of the alkene to the approach
vector as a second constrained coordinate. However, the
alternative of scanning the distance from the carbene carbon
to the alkene centroid was found to be more efficient, yielding
a smooth surface without application of additional constraints.
This latter method was employed for the model systems
reported here. For the small model 14 (reaction of 10 with
seven to remove the influence of the residual gradient along
the reaction coordinate in addition to translations and
rotations.[25] The B3LYP wavefunction was tested for triplet
instability. At all scan points, the energy required for vertical
excitation to the triplet exceeded 150 kJmolÀ1.
Solvation corrections for 14 were evaluated by using
B3LYP/LACVP[26] together with the PB-SCRF solvation
model[27] in Jaguar.[28] A continuum solvation treatment
similar to the one employed here is both necessary and
sufficient for correctly describing an ionic organometallic
reaction coordinate.[18]
Corrections for higher levels of theory were evaluated by
using CCSD(T)/LANL2DZ. Correlated calculations require
fairly large basis sets,[29] certainly larger than LANL2DZ. Our
computational resources were insufficient to go beyond this
basis set at the CCSD(T) level, but the effect of using a larger
basis can be estimated at a lower correlated level. Thus, we
calculated a basis set correction (DEbasis) at the MP2 level as
the difference between MP2 with SDD on Cu and 6-311G**
on the remaining atoms, and MP2/LANL2DZ. Combining the
basis set correction with the CCSD(T) energy, we obtained a
high-level potential energy E for the model system [Eq. (2)],
our best estimate of what the potential energy would be if we
could perform the calculations at the CCSD(T) level with a
TZVP basis set. With E as a foundation, the total free energy
along the reaction path was calculated according to Equa-
tion (2). The geometries and all relevant single-point energies
are available as Supporting information.
Ph
BP/DZ*
COOH
NH
BP/TZ+
AMBER
BP/DZ
H
COOMe
H
Cu
Cu
HN
N
N
O
14
15
O
ethene), we could employ high levels of theory and calculate
the most relevant contributions to the total free energy
profile. The more realistic model 15 (reaction of 12b with
styrene) was too large to allow direct evaluation of high-level
contributions, but is closely related to the system employed in
our kinetic studies. We believe that a combination of the
results from the two model systems will allow a meaningful
comparison with the experimental results. The close corre-
spondence between 10 and 12b (the overlay in Figure 3 gave
an r.m.s. deviation of only 0.008 ä) lends credence to the
expectation that the reaction paths will be similar.
As a basis for our computational investigation, we chose an
IMOMM-like[20] QM/MM approach[21] for the model system
15, utilizing the same functional and basis set combination as
for metallacarbene 12b, with one exception; the steric
interaction is better described at the MM level (which
includes dispersion forces[22]). The isopropyl groups and their
interactions with the remainder of the system were therefore
described by AMBER,[23] with the QM/MM[21] method
implemented in ADF.[16]
The effect of contributions beyond the BP gas-phase
calculations were explored in Gaussian98[12] using the small
model system 14. Geometry optimization was performed by
using B3LYP[11] together with the LANL2DZ[24] basis set (to
correspond to the high-level correlation which was limited to
this basis set; vide infra). The geometries were optimized with
the same Z-matrix setup as in the ADF calculations, scanning
the distance from carbene carbon to alkene centroid. All
further contributions were calculated as single-point correc-
tions at the constrained B3LYP/LANL2DZ geometries.
Vibrational contributions for nonstationary points were
evaluated at the same level by using the projection method
of Handy and co-workers, reducing the degrees of freedom by
G ˆ E ‡ DGtherm ‡ DGSolv
(2)
E ˆ CCSD(T)/LANL2DZ//B3LYP/LANL2DZ ‡ DEbasis
DGtherm: vib, rot, and trans at B3LYP/LANL2DZ, 298 K, 3N À 7 harmonic
vibrations
DGSolv ˆ DGPB-SCRF(B3LYP/LACVP)//B3LYP/LANL2DZ
DEbasis ˆ (MP2/[SDD,6-311G**] À MP2/LANL2DZ)//B3LYP/LANL2DZ
To obtain the appropriate correction factor, all constrained
geometries of 14 were also calculated at the BP level with the
same combination of basis sets that was employed for model
system 15, that is, TZ ‡ for Cu, DZ* for all atoms close to the
reaction center, and DZ for the bidentate imine ligand. A
correction term DG was calculated as the difference between
G [Eq. (2)] and BP.
For all energies, we are interested only in the relative
contributions along the reaction path. To best illustrate the
effects of the various contributions, they were adjusted by the
value at an arbitrary zero point, set to the longest scan
distance used for the small system 14, r0 ˆ 3.284 ä. Thus, the
relative correction DDG was calculated as shown in Equa-
tion (3).
DDGr ˆ DGr À DBPr
(3)
DXr ˆ X(r) À X(r0), X ˆ G, E, B3LYP, BP
A third-order polynomial expansion around r0 was fitted to
the total correction energies DDGr and used to interpolate to
points on the finer scan employed for the large system 15. The
regression yielded a very good fit (r2 ˆ 0.9999), with a 95%
confidence interval of only 0.6 kJmolÀ1 for interpolated
180
¹ WILEY-VCH Verlag GmbH, 69451 Weinheim, Germany, 2002
0947-6539/02/0801-0180 $ 17.50+.50/0
Chem. Eur. J. 2002, 8, No. 1

Copper-Catalyzed Cyclopropanation
177 184
values. In no case was the polynomial used for extrapolations
beyond the range of the scan for system 14.
along the reaction path. Finally, the difference between
CCSD(T) and BP maximizes around the TS. This is consistent
with the concept of the TS being an avoided crossing of
reactant and product states, requiring extensive correlation
for an accurate description. A T1 test gave a value of 0.033,[30]
indicating that the system is borderline even for CCSD(T) and
might be better described by a multideterminant wavefunc-
tion, but such a calculation was unfortunately beyond our
resources.
With a well-behaved correction term DDG(r) in hand, we
initiated an extensive investigation of the larger model system
15. First, since the ligand is C2-symmetric, only one ligand
orientation needs to be considered. Inverting the carbene
substituent is exactly equivalent to rotating the entire carbene
Analysis of the results in Figure 4, which shows the relative
potential energies at appropriate levels, together with the
correction energies and the corresponding polynomial
DDG(r), demonstrates that the two DFT functionals agree
well, but differ slightly from the high-level potential energy E.
In no case can a transition state be located on the potential
energy surface (PES), but inclusion of the free energy terms
indicates a TS at around rˆ 2.5 ä.
ˆ
complex by 1808 around the Cu C axis. We have chosen to
illustrate the reaction by always using an approach from the
left. Figure 6 shows attack from the two faces of the carbene.
Upon alkene approach, the carbene will bend as a result of
becoming more sp3-hybridized. For attack from the Si face,
this will bring the carboxylate into close proximity with the
ligand isopropyl group.
Ph
Ph
COOMe
COOMe
H
H
iPr
iPr
iPr
iPr
Cu
Cu
N
N
O
N
N
O
O
O
Figure 4. Energies and correction (DDG) from model system 14.
Carbene-Re
Carbene-Si
Figure 6. Two possible approaches to a carbene with a C2-symmetric
ligand.
All four correction terms are of similar magnitude, and
follow the expected trend (Figure 5). Going along the reaction
coordinate towards the product, six degrees of freedom are
converted from rotations/translations to vibrations, increasing
the total free energy. The surface area decreases, which
lessens the stabilization by solvent. The basis set superposition
error (BSSE) is worsened by increased overlap, so correcting
for this by extending the basis set will also increase the energy
The approach vectors depicted in Figure 7 differ qualita-
tively: the open approach corresponds to a concerted but very
asynchronous addition, whereas the cyclic path leads to initial
formation of a metallacyclobutane, in analogy with known
Ph
Ph
COOMe
COOMe
H
H
Cu
Cu
Cyclic
Open
Figure 7. Two addition mechanisms.
metathesis catalysts.[1] The two approaches are identical
beyond approximately 2.7 ä, but separate into two distinct
paths at shorter distances. To differentiate the paths, energy
scans were started at a point at which the two approaches are
clearly different (approximately 2.3 ä) and conducted in both
directions to complete the reaction profiles.
Finally, a monosubstituted alkene may approach with four
different orientations (Figure 8). One might envision other
orientations in which the alkene has been rotated around the
approach vector, and at long distances the alkene will orient
to achieve the optimum nonbonded distance between the
phenyl and the ligand. On closer approach following the open
mechanism, the alkene will orient to avoid eclipsing the
Figure 5. Contributions to the relative free energy correction, Equa-
tion (2). The last term, DDECC, is the difference between CCSD(T)/
LANL2DZ and BP.
Chem. Eur. J. 2002, 8, No. 1
¹ WILEY-VCH Verlag GmbH, 69451 Weinheim, Germany, 2002
0947-6539/02/0801-0181 $ 17.50+.50/0
181

P.-O. Norrby et al.
FULL PAPER
Ph
Ph
COOMe
COOMe
N
H
H
R
R
Cu
Cu
exo
N
N
N
Alkene-Si
180¡
rotation
Alkene-Re
COOMe
COOMe
H
H
Ph
Ph
R
R
Cu
Cu
endo
N
N
N
N
Figure 8. Alkene approach orientations.
carbene substituents, whereas the cyclic mechanism can only
be realized when the two reacting bonds are parallel.
The relationship between the orientation of approach and
the configuration of the final product is not immediately
evident. The configuration of the phenyl-substituted cyclo-
propyl carbon depends only on which alkene face is attacked,
with the Re alkene always leading to 2S configuration.
However, the final configuration of the carbene carbon
depends on both the carbene face selectivity and the alkene
approach; a switch from exo to endo attack changes the final
configuration. Thus, the carbene-Re/exo and the carbene-Si/
endo combinations both lead to 1R configuration in the
product.
Figure 9. Comparison of the open and cyclic exo paths (compare Figure 7).
To avoid extrapolation, a constant value of DDG(r) ˆ DDG(2.284) ˆ
36.26 kJmolÀ1 was used for short scan distances, r< 2.284 ä.
In general, the endo approach is disfavored. For the open
type of TS, the phenyl group can stabilize the developing
charge only in an exo orientation. For the cyclic TS, the steric
interaction with the ligand becomes severe. Most scan
attempts for endo approaches would be converted sponta-
neously to the corresponding exo orientation by a 1808
rotation of the alkene around the approach vector (Figure 8).
However, exceptions to this rule were observed (vide infra).
Free-energy surface for model system 15: All free energies
were obtained by adding the correction function DDG(r) to
the QM/MM potential energies. We first compared the direct
addition to the initial formation of a metallacyclobutane
intermediate using exo approaches (Figure 9).
Figure 10. Asynchronicity of the concerted addition, exemplified by the
open exo-alkene-Re/carbene-Si path. The improper torsions correspond to
2
3
À ˆ
ˆ
the R C X ¥¥¥ R dihedrals of the three R2C X moieties, sp ˆ 1808, sp ˆ
1208.
For the open TS, the QM/MM potential energy decreases
monotonically, but the free energy reaches a maximum
around 2.5 ä. The curvature is very gentle, and at longer
distances there is almost free rotation around the approach
vector. At short distances, the alkene bond has to align with
ization of the b-carbon starts at approximately 2.8 ä, and this
carbon atom is already essentially completely sp3-hybridized
around 2.0 ä. Over most of the range, the substituents on the
a-carbon show a very slight bending towards the reaction
center. At rˆ 1.7 ä it becomes completely planar and then
ˆ
the Cu C bond to form the final product, and the decrease in
À
the rotational freedom provides one reason why the vibra-
tional component of the free energy increases along the
reaction path (Figure 5), but at the TS there is still a certain
amount of flexibility in the orientation.
rapidly forms the second C C bond. The pyramidality of the
À ˆ
carbene carbon can likewise be measured as the C C Cu ¥¥¥
À
H improper torsion, at least for as long as a significant Cu C
bond (r> 1.5 ä) exists. The bending away from planarity has
already started beyond rˆ 3 ä, and ends at complete sp3
character around rˆ 2 ä. The degree of asynchronicity is
also shown by the difference (DCC) between the lengths of
Some of the geometrical changes along the reaction path
are depicted in Figure 10. The hybridization of the alkene
carbons can be measured by the improper torsional angle
À ˆ
À
R C C ¥¥¥ H, where the last H is the second substituent on the
the C C bonds that are forming, which reaches a maximum of
first carbon. A completely planar system has an improper
torsion of 1808, whereas for an idealized sp3 carbon it is 1208,
and the cyclopropyl carbons lie around 1478. The rehybrid-
approximately 0.84 ä at rˆ 2.1 ä. The small model system 14
is much more synchronous (maximum DCC ˆ 0.39 ä). This is
probably due to the lack of stabilization of the developing
182
¹ WILEY-VCH Verlag GmbH, 69451 Weinheim, Germany, 2002
0947-6539/02/0801-0182 $ 17.50+.50/0
Chem. Eur. J. 2002, 8, No. 1

Copper-Catalyzed Cyclopropanation
177 184
cationic charge on the primary carbon in the small model
system.
There is a significant barrier to formation of a metal-
lacyclobutane, even on the potential-energy surface. In an
absolute sense the barrier is still fairly low, and in the absence
of alternatives this would have been a feasible path. However,
as the direct addition through the open TS is about 40 kJmolÀ1
lower in energy, the possibility of an exo-metallacyclobutane
intermediate can be excluded, in agreement with our ob-
served isotope effects.
We concentrated on the open exo approaches. The four
possible vectors (Figure 11) were compared at one common
scan point, reaction coordinate 2.284 ä, at the the free-energy
curve turns sharply downward after a long, flat region. The
H
Ph
H
H
H
H
Ph
MeOOC
COOMe
Alkene-Si
Alkene-Re
H
Cu
H
H
iPr
iPr
iPr
iPr
Cu
0.0
5.5
Carbene-Re
Carbene-Si
Ph
H
H
H
Ph
H
H
MeOOC
COOMe
Figure 12. Gauche addition with collapse to a metallacyclobutane inter-
mediate. The ball-and-stick figure is viewed from the metallacyclobutane
apex (Cu hidden), as indicated on the right.
H
Cu
H
Cu
H
iPr
iPr
1.7
iPr
iPr
1.9
Figure 11. The four possible open exo approaches. Energies [kJmolÀ1] are
relative to the most favorable form (alkene-Si, carbene-Re).
comparison point (rˆ 2.284 ä), this path is 26 kJmolÀ1 higher
in energy than the most favored mechanism shown in
Figure 11, so for this substrate it should not contribute
significantly to product formation. This is also indicated by
the fact that it would produce the observed minor enantiomer.
However, this asynchronous path to the metallacyclobutane,
with an endo substituent, is substantially lower in energy than
the direct [2‡2] addition found for the cyclic exo path. It also
has a steric profile that differs substantially from the other
open paths. It is quite clear that the tert-butyl substituent (12c)
would prohibit the metallacyclobutane formation completely,
but with less hindered ligands and other alkene substituent
patterns the gauche path cannot be excluded with absolute
certainty. The hybridization changes and degree of bond
formation along the reaction path closely parallels that for the
open mechanism, so the two mechanisms are not expected to
be distinguishable by the type of kinetic studies we employed
in this study. The only valid mechanistic indicator in this case
seems to be the product pattern to be expected from the
different paths.
geometries and raw energies at this distance are available as
Supporting Information. At this point the addition is very
asynchronous, the length of the bond forming to the b-carbon
being 1.92 1.99 ä and the carbene a-carbon distance still
2.73 2.78 ä. The b-carbon is partially rehybridized towards
the sp3 form, whereas the a-carbon is essentially planar (sp2)
(Figure 10).
The structures are not properly converged transition states,
so the energy differences should be interpreted only qual-
itatively. For this scan point, the results are in good agreement
with the model proposed by Pfaltz and co-workers.[2] The path
leading to the observed major diastereomer (1R,2R) is indeed
lowest in energy. The enantioselectivity at the carboxylate
substituent arises from the hindered distortion of the Si
carbene. This penalty is more severe in the tert-butyl-
substituted ligand 12c (Figure 3), in perfect agreement with
the higher experimental enantioselectivities observed with
this ligand. The alkene substituent does not interact signifi-
cantly with the ligand in any exo path; the cis trans selectivity
is solely an effect of the relatively weak steric repulsion
between carboxylate and alkene substituent in the transition
state. This is not true, however, for a 1,2-disubstituted alkene,
in which one substituent should be strongly affected by direct
interaction with the ligand.[7]
Conclusion
The selectivity-determining step in the copper-catalyzed
cyclopropanation proceeds by a concerted but very asynchro-
nous addition of a metallacarbene to the alkene. For styrene
substrates, the bond to the b-carbon is formed early in the
reaction, with final ring closure to the cyclopropane product
occurring late but still in a concerted manner. In the
neighborhood of the free-energy TS, the a-carbon develops
substantial cationic character. For the monosubstituted styr-
One additional low-energy path was found in the inves-
tigation of an endo approach. This path has all the character-
istics of the open mechanism, in that the bond to the styrene b-
carbon is formed first, but the alkene enters in a gauche-like
orientation which collapses to a metallacyclobutane inter-
mediate (Figure 12). Formation of the cyclopropane from this
intermediate is facile and exothermic. At the common
ˆ
ˆ
enes studied here, a parallel orientation of the C C and Cu C
Chem. Eur. J. 2002, 8, No. 1
¹ WILEY-VCH Verlag GmbH, 69451 Weinheim, Germany, 2002
0947-6539/02/0801-0183 $ 17.50+.50/0
183

P.-O. Norrby et al.
FULL PAPER
F. Lˆffler, B. Meynhardt, S. Kierse, M. M¸ller, C. N‰ther, U. L¸ning,
Eur. J. Org. Chem. 2001, 2151 2160.
[11] a) A. D. Becke, J. Chem. Phys. 1993, 98, 5648 5652; b) C. Lee, W.
Yang, R. G. Parr, Phys. Rev. B 1988, 37, 785 789.
bonds is preferred, with the phenyl group in an exo
orientation (not interacting with the ligand). The enantiose-
lectivity is determined by a direct repulsive interaction
between the carbene substituent and the ligand, whereas the
diastereoselectivity for monosubstituted alkenes is deter-
mined only by a relatively weak interaction between the
carbene and alkene substituents.[2] However, an alkene with
substituents on both carbon atoms could certainly interact
directly with the ligand.[7]
Metallacyclobutane formation through a direct [2‡2]
addition is kinetically disfavored, but if the approaching
alkene adds in a gauche-like orientation, the second bond may
be formed to Cu, leading to a metastable metallacyclobutane.
The gauche orientation is disfavored for the styrenes studied
here, but cannot be excluded for more highly substituted
alkenes. The question of whether a metallacyclobutane is an
intermediate in the reaction is of no importance to selectivity
predictions, since the selectivity is fully determined in the
initial addition, which is irreversible. However, a gauche
addition would yield a stereoselectivity quite different from
the regular open exo addition.
[12] M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb,
J. R. Cheeseman, V. G. Zakrzewski, J. A. Montgomery, R. E. Strat-
mann, Jr., J. C. Burant, S. Dapprich, J. M. Millam, A. D. Daniels, K. N.
Kudin, M. C. Strain, O. Farkas, J. Tomasi, V. Barone, M. Cossi, R.
Cammi, B. Mennucci, C. Pomelli, C. Adamo, S. Clifford, J. Ochterski,
G. A. Petersson, P. Y. Ayala, Q. Cui, K. Morokuma, D. K. Malick,
A. D. Rabuck, K. Raghavachari, J. B. Foresman, J. Cioslowski, J. V.
Ortiz, B. B. Stefanov, G. Liu, A. Liashenko, P. Piskorz, I. Komaromi,
R. Gomperts, R. L. Martin, D. J. Fox, T. Keith, M. A. Al-Laham, C. Y.
Peng, A. Nanayakkara, C. Gonzalez, M. Challacombe, P. M. W. Gill,
B. Johnson, W. Chen, M. W. Wong, J. L. Andres, C. Gonzalez, M.
Head-Gordon, E. S. Replogle, J. A. Pople, Gaussian98, Revision A.3,
Gaussian, Inc., Pittsburgh, PA, 1998.
[13] M. Diedenhofen, T. Wagener, G. Frenking, in Computational Organo-
metallic Chemistry (Ed.: T. Cundari), Marcel Dekker, New York,
2001, pp. 69 121.
[14] M. Bˆhme, G. Frenking, Chem. Phys. Lett. 1994, 224, 195 199.
[15] a) S. H. Vosko, L. Wilk, M. Nusair, Can. J. Phys. 1980, 58, 1200 1211;
b) A. D. Becke, Phys. Rev. A. 1988, 38, 3098 3100; c) J. P. Perdew,
Phys. Rev. B. 1986, 33, 8822 8824.
[16] ADF 2.3.0, Theoretical Chemistry, Vrije Universiteit, Amsterdam;
a) E. J. Baerends, D. E. Ellis, P. Ros, Chem.Phys. 1973, 2, 41;
b) G. te Velde, E. J. Baerends, J. Comp. Phys. 1992, 99, 84; c) C. G.
Fonseca, O. Visser, J. G. Snijders, G. te Velde, E. J. Baerends, in
Methods and Techniques in Computational Chemistry, METECC-95
(Eds.: E. Clementi, G. Corongiu), STEF, Calgari, 1995, p. 305.
[17] D. A. Evans, K. A. Woerpel, M. J. Scott, Angew. Chem. 1992, 104,
439 441; Angew. Chem. Int. Ed. Engl. 1992, 31, 430 432.
[18] H. Hagelin, B. äkermark, P.-O. Norrby, Chem. Eur. J. 1999, 5, 902
909.
Acknowledgement
Support from the Danish Technical Research Council is gratefully
acknowledged.
[19] P.-O. Norrby, J. Mol. Struct. (THEOCHEM) 2000, 506, 9 16.
[20] F. Maseras, Top. Organomet. Chem. 1999, 4, 165 191.
[21] T. K. Woo, L. Cavallo, T. Ziegler, Theor. Chem. Acc. 1998, 100, 307
313.
[22] A correct description of London dispersion in ab initio studies
requires a very high level of theory: G. Chalasinski, M. M. Szczesniak,
Chem. Rev. 2000, 100, 4227 4252. The DFT methods employed here
do not account for dispersion, whereas in molecular mechanics the
effect is included as an empirical term.
[23] S. J. Weiner, P. A. Kollman, D. T. Nguyen, D. A. Case, J. Comput.
Chem. 1986, 7, 230 252.
[24] P. J. Hay, W. R. Wadt , J. Chem. Phys. 1985, 82, 299 310.
[25] W. H. Miller, N. C. Handy, J. E. Adams, J. Chem. Phys. 1980, 72, 99
112.
[26] LACVP employs the same Hay Wadt ECP as LANL2DZ for Cu,[24]
but uses instead the Pople 6-31G basis set for lighter elements.
[27] B. Marten, K. Kim, C. Cortis, R. A. Friesner, R. B. Murphy, M. N.
Ringnalda, D. Sitkoff, B. Honig, J. Phys. Chem. 1996, 100, 11775
11788.
[28] Jaguar v.3.5, Schrˆdinger, 1998, www.schrodinger.com.
[29] F. Jensen, Introduction to Computational Chemistry, Wiley, Chi-
chester, 1999.
[1] a) M. P. Doyle, M. A. McKervey, T. Ye, Modern Catalytic Methods for
Organic Synthesis with Diazo Compounds, Wiley, New York, 1998;
b) A. Pfaltz, in Transition Metals for Organic Synthesis (Eds.: M.
Beller, C. Bolm), Wiley-VCH, Weinheim, 1998, pp. 100 113; c) A.
Pfaltz, in Comprehensive Asymmetric Catalysis, Vol. II, Springer,
Berlin, 1999, pp. 513 538.
[2] H. Fritschi, U. Leutenegger, A. Pfaltz, Helv. Chim. Acta 1988, 71,
1553 1565.
[3] a) R. E. Lowenthal, A. Abiko, S. Masamune, Tetrahedron Lett. 1990,
31, 6005 6008; b) D. A. Evans, K. A. Woerpel, M. M. Hinman, M. M.
Faul, J. Am. Chem. Soc. 1991, 113, 726 728; c) D. M¸ller, G.
Umbricht, B. Weber, A. Pfaltz, Helv. Chim. Acta 1991, 74, 232 240.
[4] T. Aratani, Pure Appl. Chem. 1985, 57, 1839 1844.
[5] M. M. DÌaz-Requejo, T. R. Belderrain, M. C. Nicasio, F. Prieto, P. J.
¬
Perez, Organometallics 1999, 18, 2601 2609.
[6] B. F. Straub, P. Hofmann, Angew. Chem. 2001, 113, 1328 1330;
Angew. Chem. Int. Ed. 2001, 40, 1288 1290.
[7] K. Ito, T. Katsuki, Synlett 1993, 638 640.
[8] a) L. A. Dakin, S. E. Schaus, E. N. Jacobsen, J. S. Panek, Tetrahedron
Lett. 1998, 39, 8947 8950; b) L. A. Dakin, P. C. Ong, J. S. Panek, R. J.
Staples, P. Stavropoulos, Organometallics 2000, 19, 2896 2908.
¬
[9] M. M. DÌaz-Requejo, P. J. Perez, M. Brookhart, J. L. Templeton,
[30] As a rule of thumb, T1 values greater than 0.02 indicate a multi-
reference situation: T. J. Lee, P. R. Taylor, Int. J. Quantum Chem.
Symp. 1989, 23, 199 207.
Organometallics 1997, 16, 4399 4402.
[10] While we were finalizing this manuscript, it came to our attention that
the entire reaction surface has been investigated at the B3LYP level:
a) J. M. Fraile, J. I. GarcÌa, J. A. Mayoral, L. Salvatella, J. Am. Chem.
Soc. 2001, 123, 7616 7625. Very recently, a study at the BP86 level
appeared, with slightly different conclusions: b) M. B¸hl, F. Terstegen,
Received: July 6, 2001 [F3398]
184
¹ WILEY-VCH Verlag GmbH, 69451 Weinheim, Germany, 2002
0947-6539/02/0801-0184 $ 17.50+.50/0
Chem. Eur. J. 2002, 8, No. 1