Theoretical and experimental studies of asymmetric organozinc additions to benzaldehyde catalyzed by flexible and constrained γ-amino alcohols

Article abstract of DOI:10.1021/jo0262210

The cis-decalin based γ-amino alcohols, 1-5, were synthesized, resolved, and employed as catalysts in the addition of organozincs to benzaldehyde. Despite large facial differentiation in the RZn adducts of the ligands, the enantioselectivities were found

Full text of DOI:10.1021/jo0262210

Th eor etica l a n d Exp er im en ta l Stu d ies of Asym m etr ic Or ga n ozin c
Ad d ition s to Ben za ld eh yd e Ca ta lyzed by F lexible a n d Con str a in ed
γ-Am in o Alcoh ols
Manoranjan Panda, Puay-Wah Phuan, and Marisa C. Kozlowski*
Department of Chemistry, Roy and Diana Vagelos Laboratories, University of Pennsylvania,
Philadelphia, Pennsylvania 19104
marisa@sas.upenn.edu
Received J uly 20, 2002
The cis-decalin based γ-amino alcohols, 1-5, were synthesized, resolved, and employed as catalysts
in the addition of organozincs to benzaldehyde. Despite large facial differentiation in the RZn adducts
of the ligands, the enantioselectivities were found to be moderate. Transition structure calculations
were done using several quantum chemical methods to examine the underlying causes of the
selectivity for three γ-amino alcohols, 1, 2, and 3. The tricyclic µ-oxo (6/4/4), rather than the bicyclic
six-membered (6/6), transition structures were found to explain the observed enantioselectivity for
2 and 3. MP2 calculations gave the best correlation to the experimental results compared to B3LYP
and HF calculations. The conformational flexibility of the chiral ligands was found to be an important
factor in the selectivity.
In tr od u ction
The identification of new chiral ligands for asymmetric
synthesis remains an important endeavor. Using a
database mining protocol to find novel ligand architec-
tures,1 we have identified the cis-decalin architecture as
potentially useful (Figure 1).
F IGURE 1. cis-Decalin.
cis-Decalins provide a flexible framework for the
construction of a wide variety of interesting chiral ligands
including diamines,2 diols, amino alcohols, phosphites,
and phosphines. Several examples of non-C2-symmetric
γ-amino alcohols based upon the cis-decalin skeleton can
be imagined (Figure 2). In this study, the utility of cis-
decalin derivatives 1 and 2 was investigated in the
asymmetric addition of organozincs to aldehydes on the
basis of PM3 transition structure calculations which
predicted high levels of enantioselectivity for 2. Further
constrained analogues of 2 (3-5) were also prepared and
examined in order to determine the role of conformational
flexibility on the selectivities of 1 and 2.
F IGURE 2. cis-Decalin γ-amino alcohols, constrained ana-
logues, and DAIB.
The discovery that chiral amino alcohols, such as
DAIB, are excellent catalysts for the enantioselective
addition of dialkylzincs to aldehydes3,4 has stimulated a
host of efforts directed at understanding the mechanism,
stereochemical induction, and scope of this reaction. Due
to its relative simplicity, this reaction is also an ideal case
for addressing the necessary architectures and elements
required for high stereochemical induction. â-Amino
alcohols have typically been employed in this reaction
although a number of γ-amino alcohols3b,5 have also been
used successfully.
Rationalizations of the observed stereoselectivity of
asymmetric dialkylzinc additions to benzaldehyde cata-
lyzed by amino alcohols have been reported by various
research groups utilizing transition structure analyses.6
However, there is sparse information regarding the
theoretical investigation of the reaction with γ-amino
alcohol catalysts. We wished to determine if the reported
(1) Kozlowski, M. C.; Panda, M. J . Mol. Graphics Modell. 2002,
20, 399.
(2) (a) Li, X.; Schenkel, L. B.; Kozlowski, M. C. Org. Lett. 2000, 2,
875. (b) Li, X.; Yang, J .; Kozlowski, M. C. Org. Lett. 2001, 3, 1137.
(3) For reviews, see: (a) Soai, K.; Niwa, S. Chem. Rev. 1992, 92,
833. (b) Noyori, R. In Asymmetric Catalysis In Organic Synthesis; J ohn
Wiley & Sons: New York, 1994; p 260.
(5) Some examples of γ-amino alcohols used in organozinc alkylation:
(a) Oppolzer, W.; Radinov, R. N. Tetrahedron Lett. 1988, 29, 5645. (b)
Muchow, G.; Vannoorenberghe, G.; Buono, G. Tetrahedron Lett. 1987,
28, 6163. (c) Cho, B. T.; Kim, N. Tetrahedron Lett. 1994, 35, 4115. (d)
Hulst, R.; Heres, H.; Fitzpatrick, K.; Peper, N.; Kellogg, R. M.
Tetrahedron: Asymmetry 1996, 7, 2755. (e) Cicchi, S.; Crea, S.; Goti,
A.; Brandi, A. Tetrahedron: Asymmetry 1997, 8, 293.
(4) Pu, L.; Yu, H.-B. Chem. Rev. 2001, 101, 757.
10.1021/jo0262210 CCC: $25.00 © 2003 American Chemical Society
Published on Web 12/24/2002
564
J . Org. Chem. 2003, 68, 564-571

Organozinc Additions to Benzaldehyde
SCHEME 1
SCHEME 2
analyses were reliable in predicting selectivity for this
slightly different ligand type. In this report, we compare
a theoretical evaluation of cis-decalin γ-amino alcohols
in this reaction with the experimental results.
SCHEME 3. X-r a y Str u ctu r es of 13 a n d 14 Used in
Assign in g th e Absolu te a n d Rela tive
Con figu r a tion s of 1 a n d 2
Resu lts a n d Discu ssion
Syn th esis of cis-Deca lin γ-Am in o Alcoh ols 1 a n d
2. Several syntheses of the basic structure found in 1 have
been reported.7 A modified approach was developed as
shown in Scheme 1 which was amenable to large scale
synthesis and provided 1 in high yield using simple
transformations. The application of Rh/Al2O3 as the
hydrogenation catalyst significantly improved the yield
and selectivity of the reduction of 6, and amino alcohol 7
could be recrystallized as a single isomer. For these
studies, the relatively nonpolar N-Boc derivative 8 was
generated and then resolved using chiral HPLC. Resolved
8 could easily be converted to N-methyl derivative 1 by
treatment with TFA followed by Eschweiler-Clarke
methylation.
The synthesis of 2 was accomplished starting with
ketal 10 which was generated following the method of
Momose et al.9 (Scheme 2). After the N-methyl amide of
ketal 10 was formed, treatment with 1 N HCl yielded
ketoamide 11. Selective reduction of ketoamide 11 was
accomplished using NaBH4 to provide the epimeric
alcohols 12r and 12â in ratio of 5:1. While 12r and 12â
were not separable by silica column chromatography, the
diastereomers and enantiomers of 12 could be separated
with chiral HPLC. The relative stereochemistry of the
desired R-epimer was secured by NOE differences unique
to 12r (H7 to H9). Reduction of 12r with LiAlH4 furnished
2 as an oil which could be used directly.
quaternary ammonium salts (Scheme 3). The salts 13
and 14 were obtained by treatment of 1 and 2, respec-
tively, with MeI.
Develop m en t a n d Syn th esis of Con str a in ed An a -
logu es of 2. Since cis-decalin 1 and 2 are conformation-
ally mobile, we sought to develop constrained analogues
to probe the role of conformational flexibility on reactivity
and selectivity. Examination of the three-dimensional
structure of 2 showed that addition of a methylene would
constrain the molecule (Figure 3). This constrained form
contains a 6-azabicyclo[3.2.1]octane framework. Hence,
amino alcohols (S,S,R)-3, (S,S,R,S)-4, and (R,R,S,S)-5
were synthesized following the procedure of Carroll et
al.10 These compounds had not been examined previously
as ligands in asymmetric synthesis.
The absolute and relative configurations of 1 and 2
were established via anomalous dispersion from the
single-crystal X-ray structures of the corresponding
(6) (a) Yamakawa, M.; Noyori, R. J . Am. Chem. Soc. 1995, 117, 6327.
(b) Yamakawa, M.; Noyori, R. Organometallics 1999, 18, 128. (c)
Goldfuss, B.; Houk, K. N. J . Org. Chem. 1998, 63, 8998. (d) Goldfuss,
B.; Steigelmann, M.; Khan, S. I.; Houk, K. N. J . Org. Chem. 2000, 65,
77. (e) Goldfuss, B.; Steigelmann, M.; Rominger, F. Eur. J . Org. Chem.
2000, 1785. (f) Goldfuss, B.; Steigelmann, M. J . Mol. Modell. 2000, 6,
166. (g) Vazquez, J .; Pericas, M. A.; Maseras, F.; Lledos, A. J . Org.
Chem. 2000, 65, 7303. (h) Brandt, P.; Hedberg, C.; Lawonn, K.; Pinho,
P.; Andersson, P. G. Chem. Eur. J . 1999, 5, 1692.
(7) (a) Grob, C. A.; Kiefer, H. R. Helv. Chim. Acta 1965, 48, 799. (b)
Le Bel, N. A.; Caprathe, B. W. J . Org. Chem. 1985, 50, 3938. (c)
Comins, D. L.; Abdullah, A. H. Tetrahedron Lett. 1985, 26, 43.
(8) Caprathe, B. W.; J aen, J . C.; Wise, L. D.; Heffner, T. G.; Pugsley,
T. A.; Meltzer, L. T.; Parvez, M. J . Med. Chem. 1991, 34, 2736.
(9) Momose, T.; Uchida, S.; Kinoshita, M.; Imanishi, T. Chem.
Pharm. Bull. 1977, 25, 1797.
Asym m etr ic Or ga n ozin c Ad d ition s Ca ta lyzed by
th e γ-Am in o Alcoh ols. On the basis of the results of
PM3 transition structure calculations which predicted
(10) Pitner, J . B.; Abraham, P.; J oo, Y. J .; Triggle, D. J .; Carroll, F.
I. J . Chem. Soc., Perkin Trans. 1 1991, 1375.
J . Org. Chem, Vol. 68, No. 2, 2003 565

Panda et al.
F IGURE 3. Development of constrained analogues of 2.
TABLE 1. En a n tioselective Alk yla tion of Ben za ld eh yd e
Ca ta lyzed by 1-5a
entry
solvent
R
ligandb (L*) er (R):(S) conv (%)
1
2
3
4
5
6
7
8
PhCH3
Me
1
1
2
2
1
1
2
2
2
2
2
2
3
4
5
48:52
62:38
73:27
85:15
47:53
61:39
73:27
84:16
84:16
84:16
86:14
86:14
67:33
53:47
63:37
29
7
42% THF/PhCH3 Me
PhCH3 Me
42% THF/PhCH3 Me
PhCH3 Et
42% THF/PhCH3 Et
PhCH3 Et
10% THF/PhCH3 Et
25% THF/PhCH3 Et
42% THF/PhCH3 Et
50% THF/PhCH3 Et
90% THF/PhCH3 Et
51
33
94
18
87
50
30
32
18
16
36
59
31
F IGURE 4. Schematic of the complexation of 1 with Et2Zn
and benzaldehyde.
9
eomeric reaction pathways of 1 and 2. Alternatively the
conformational flexibility of these compounds may give
rise to multiple reactive species (Figure 4). From prior
work, compounds 1 and 2 were known to predominantly
occupy the N-in form in nonpolar media due to hydrogen
bonding.12 As such, we expected the addition of Et2Zn
would lead to predominantly 1-in -Zn and 2-in -Zn due
to strong metal coordination with the nitrogen and
oxygen heteroatoms. Attempts to observe the formation
of 1-in -Zn (Figure 4) by NMR spectroscopy upon addition
of 1 equiv of Et2Zn to 1 at 250 K produces complex spectra
which may reflect a dimer-monomer equilibrium as well
as competing zinc species (1-in -Zn and 1-ou t-Zn ).
Although 2 provides better enantioselection than 1, the
origin of the improved selectivity was not clear. We had
previously measured the conformational equilibria of 1
and 2 and found that, in general, the equilibrium ratio
of 2-in /2-ou t (12:1) is higher than 1-in /1-ou t (2.4:1) in
CDCl3 at 240 K.12 Thus, 2 may give rise to higher
enantioselection, because more of 2-in -Zn is present.
Alternatively, structural features of 2 may contribute to
a more selective process by stabilization/destabilization
of one diastereomeric pathway.
To assess the impact of conformational flexibility on
2, a constrained analogue (3) was designed and synthe-
sized. On this basis, (S,S,R)-3 would be expected to give
the (S) product but instead provided the (R) product
(entry 13, Table 1). In addition, 3 is a less selective ligand
than 2. Surprisingly, the incorporation of N-alkyl groups
with additional stereogenic centers (4, 5) in place of the
N-methyl group of 3 did not improve the selectivity
(entries 14 and 15, Table 1). Since we had expected the
conformationally less flexible 3 to be more selective than
2, we undertook further calculations of the transition
10
11
12
13
14
15
PhCH3
PhCH3
PhCH3
Et
Et
Et
a
Reaction conditions: After Et2Zn (1.4 mmol) and ligand (5 mol
%) were mixed 30 min at room temperature, PhCHO (1 mmol)
was added, and the reaction was halted after 8 h. The enantiomeric
b
ratio and conversion was measured by GC (Cylcodex â). (R,S,S)-
1 and (R,R,S)-2 were employed.
high levels of enantioselectivity for 2, compounds 1-2
and constrained analogues 3-5 were investigated in the
asymmetric addition of dimethyl- and diethylzinc to
benzaldehyde. The results are collected in Table 1.
Ligand 2 gave the highest enantioselectivity of 68% ee
(entry 8) when THF was used as a cosolvent. Numerous
reports indicate that the enantioselectivity of dialkylzinc
additions is often influenced by solvent. Typically, non-
coordinating solvents (i.e., toluene) provide the best
results while the addition of coordinating solvents such
as THF decrease the selectivity.11 The enhancement of
selectivities in THF for 1 and 2 is thus unusual and does
not appear to be a medium effect (i.e., dielectric) since
even small amounts of THF cause a dramatic change
(entries 8-12).
The low to moderate enantioselectivity could be at-
tributed to low energy differences between the diaster-
(11) For specific cases involving THF see ref 17 and Zhang, F.-Y.;
Yip, C.-W.; Cao, R.; Chan, A. S. C. Tetrahedron: Asymmetry 1997, 8,
585.
(12) Phuan, P.-W.; Kozlowski, M. C. J . Org. Chem. 2002, 67, 6339.
566 J . Org. Chem., Vol. 68, No. 2, 2003

Organozinc Additions to Benzaldehyde
SCHEME 4. Tr icyclic a n ti a n d syn µ-Oxo Tr icyclic
Tr a n sition Str u ctu r es
structures of 1-3 to understand these experimental
results.
Com p u ta tion a l Meth od s Used in th e Tr a n sition
Str u ctu r e Ca lcu la tion s. The presence of two Zn atoms
along with the relatively large molecular framework of
most ligands employed in these reactions makes the use
of high level ab initio methods costly. To our knowledge,
most of the theoretical analyses reported so far involve
semiempirical methods, combined quantum mechanics-
molecular mechanics (QM/MM) methods, or ab initio
calculations on model systems. While it is not always
possible to choose suitable model systems, semiempirical
and QM/MM calculations may not properly consider all
the electronic and steric factors relevant to the enanti-
oselectivity. Indeed, initial PM3 transition structure
calculations with 2 predicted high selectivity, which
stimulated us to examine these compounds. In this study,
the relatively small size of ligands 1, 2, and 3 enabled
us to carry out the calculations at higher levels of theory,
which avoids the limitations with semiempirical and QM/
MM methods.
SCHEME 5. Six-Mem ber ed Tr a n sition Str u ctu r es
All the transition structures for the asymmetric dim-
ethylzinc13 addition to benzaldehyde catalyzed by ligands
1, 2, and 3 were fully optimized using the Hartree-Fock
(HF)/LanL2DZ14 method. The transition structures were
located using the synchronous transit-guided quasi-
Newton (STQN) method. The guess transition structure
geometries were taken from PM3 calculations.15 The fully
optimized transition structures were characterized by one
imaginary frequency which corresponds to the migration
of the suitably oriented methyl group to the carbonyl
carbon. Single point energy calculations were accom-
plished using B3LYP and MP2 to account for electron
correlation. All the calculations were executed using
Gaussian 98.16 Both tricyclic and bicyclic transition
structures were traced. The relative energies obtained
from the MP2 calculations are invoked in the discussion
below unless otherwise noted.
alcohols as the ligands, the corresponding tricyclic transi-
tion structure is composed of a fused 6/4/4 ring system.
The four low-energy diastereomeric transition structures
are illustrated in Scheme 4: anti R (the two unreacting
Me groups on the Zn atoms are anti, gives (R) alcohol),
anti S (gives (S) alcohol), syn R (the two unreacting Me
groups are syn and give (R) alcohol), and syn S.
The finding by Noyori et al. that the reaction requires
two zinc species per aldehyde17 led to transition structure
models where two Zn atoms are present in different
coordination spheres. The µ-oxo tricyclic transition struc-
ture models described by Noyori et al.6a explain the
observed absolute configuration as well as the level of
stereoselection in many of the cases. With γ-amino
Recently, Norrby et al.18 characterized bicyclic six-
membered transition structures of model systems using
B3LYP in conjunction with a higher basis set. The six-
membered bicyclic transition structures have one of the
Zn atoms in a trigonal planar arrangement, and the
aldehydic oxygen is coordinated to only one Zn atom
(Scheme 5). For the γ-amino alcohols, we refer to these
transition structures as 6/6. In the sections below, the
transition structures for ligands 1-3 using HF, density-
functional theory (DFT), and MP2 are described and
compared with the observed enantioselectivity.
Tr a n sition Str u ctu r es for Liga n d 1. In the anti
transition structures (anti R and anti S) the two unre-
acting Me groups attached to Zn1 and Zn3 are anti, thus
the six-membered ring formed by the ligand with Zn1 is
anti to the Zn3-C4-C5-O6 four-membered ring (Figure
5). The most stable TS for 1 is anti R. The next most
stable TS, anti S, is 1.5 kcal/mol higher in energy due to
a steric interaction between phenyl ring and one of the
unreacting Me groups (Scheme 4, Figure 5, Table 2). The
two four-membered rings (Zn1-O2-Zn3-O6 and Zn3-
C4-C5-O6) are distorted to nullify this repulsion. In anti
(13) Transition structure calculations used dimethylzinc, which
simplifies the computational effort by avoiding the introduction of alkyl
rotamers.
(14) LanL2DZ denotes Los Almos effective core potential and
double-ú basis set for zinc and Dunning-Huzinaga double-ú basis set
for other atoms.
(15) See Supporting Information for a full discussion.
(16) Frisch, M. J .; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.;
Robb, M. A.; Cheeseman, J . R.; Zakrzewski, V. G.; Montgomery, J . A.,
J r.; Stratmann, R. E.; Burant, J . C.; Dapprich, S.; Millam, J . M.;
Daniels, A. D.; Kudin, K. N.; Strain, M. C.; Farkas, O.; Tomasi, J .;
Barone, V.; Cossi, M.; Cammi, R.; Mennucci, B.; Pomelli, C.; Adamo,
C.; Clifford, S.; Ochterski, J .; Petersson, G. A.; Ayala, P. Y.; Cui, Q.;
Morokuma, K.; Malick, D. K.; Rabuck, A. D.; Raghavachari, K.;
Foresman, J . B.; Cioslowski, J .; Ortiz, J . V.; Stefanov, B. B.; Liu, G.;
Liashenko, A.; Piskorz, P.; Komaromi, I.; Gomperts, R.; Martin, R. L.;
Fox, D. J .; Keith, T.; Al-Laham, M. A.; Peng, C. Y.; Nanayakkara, A.;
Gonzalez, C.; Challacombe, M.; Gill, P. M. W.; J ohnson, B. G.; Chen,
W.; Wong, M. W.; Andres, J . L.; Head-Gordon, M.; Replogle, E. S.;
Pople, J . A. Gaussian 98, revision A.9; Gaussian, Inc.: Pittsburgh, PA,
1998.
(17) Kitamura, M.; Suga, S.; Kawai, K.; Noyori, R. J . Am. Chem.
Soc. 1986, 108, 6071.
(18) Rasmussen, T.; Norrby, P.-O. J . Am. Chem. Soc. 2001, 123,
2464.
J . Org. Chem, Vol. 68, No. 2, 2003 567

Panda et al.
Tr a n sition Str u ctu r es for Liga n d 2. As in 1, the
anti R transition structure for 2 is more stable than anti
S (Figure 6, Table 2). Interestingly, the energy difference
is also about the same (1.5 kcal/mol). Again, this is
attributed, in part, to the cis arrangement of the phenyl
and the Zn1-Me group. The Zn1-O6-C5 angle is wider
in anti S (141°) than in anti R (122°). Thus, the
Zn3-C4-C5-O6 four-membered ring is puckered, the
Zn3-O6 bond length (2.466 Å) is longer, and the C4-C5
bond is shorter (2.289 Å) compared to those in anti R
(2.309 and 2.343 Å, respectively).
The syn transition structures for 2 reveal similar
structural as well as energetic (4-5 kcal/mol higher in
energy than their anti analogues) trends as 1. The
Zn1-O6-C5 angles in the syn are wider (syn S, 146°;
syn R, 159°) than the anti (anti R, 122°; anti S, 141°). In
syn R, the phenyl is closer to the ligand N-Me group,
raising its energy by 0.9 kcal/mol compared to syn S.
On the basis of the similarity in the relative stabilities
of the anti R and anti S transition structures containing
1 and 2, we would anticipate that both the ligands should
exhibit similar enantioselectivity.
Tr a n sition Str u ctu r es for Liga n d 3. For 3, anti R
is once again the most stable transition structure fol-
lowed by anti S (Figure 7, Table 2). The geometrical
features of the transition structures for 3 are similar to
those of 1 and 2. However, the steric distortions are
smaller as we go from lower to higher energy transition
structures.
The striking difference in this case is the smaller
relative energy difference between the transition struc-
tures compared to those of 1 and 2. The anti S is only
0.4 kcal/mol higher in energy than anti R. The syn R
and syn S possess almost the same energy and are 3.7
kcal/mol higher in energy than the most stable anti R.
On this basis, we would expect 3 to be less selective than
1 or 2.
The three different methods (HF, B3LYP, and MP2)
employed in this study revealed identical trends (Table
2). For all the cases studied [(R,S,S)-1, (R,R,S)-2, and
(S,S,R)-3], the anti R transition structure which leads
to the (R) alcohol is found to be the most stable transition
structure at each level of theory. The different theoretical
methods predict the anti S transition structure to be the
next most stable.
6/6 Tr a n sition Str u ctu r es. Norrby et al. reported
that the PM3 bicyclic transition structures for certain
â-amino alcohols are lower in energy than the tricyclic
structures. We had also located the 6/6 bicyclic transition
structures at the PM3 level19 for γ-amino alcohols, 1-3.
These PM3 structures are 4-8 kcal/mol lower in energy
compared to their 6/4/4 analogues.
To determine if these transition structures are relevant
to this case, we carried out HF/LanL2DZ optimizations
of the 6/6 transition structures for 3. Two 6/6 transition
structures, twist boat R and boat S, were located (Figure
8). In both, the methyl group migrates with retention of
configuration in a manner similar to the bicyclic transi-
F IGURE 5. HF/LanL2DZ optimized geometries of the tran-
sition structures with the ligand 1. Relative MP2 energies
(kcal/mol) are given in the parentheses. For clarity, only the
hydrogens at stereogenic centers are shown.
R, the Zn3-C4 bond of the migrating Me group is parallel
to the CdO leading to an almost planar four-membered
ring. In contrast, the Zn3-C4 bond of the anti S migrat-
ing Me group and the CdO bond are twisted about 40°,
leading to a nonplanar four-membered ring (Figure 5,
anti S). The Zn3-O6 bond in anti S (2.540 Å) is longer
than that of anti R (2.332 Å), and the C4-C5 (migrating
Me group) is shorter in anti S (2.277 Å) than in anti R
(2.341 Å). These distortions account for the higher energy
of anti S.
In the syn S and syn R transition structures, the two
unreacting Me groups attached to the Zn atoms are syn.
Consequently, the six-membered ring formed by the
ligand when chelated to Zn1 is syn to Zn3-C4-C5-O6
four-membered ring. This forces the substrate closer to
the ligand. As a result the syn transition structures are
4-5 kcal/mol higher in energy than the anti structures.
To reduce the severe steric interaction between the ligand
and the benzaldehyde substrate in the syn transition
structures, the Zn1-O6-C5 bond angle becomes wider
compared to anti structures (anti R, 120°; anti S, 139°;
syn S, 149°; syn R, 166°). While the phenyl ring avoids
the Zn3 Me group in both syn transition structures, the
proximity of the syn R phenyl group to the ligand raises
the syn R energy 0.8 kcal/mol compared to syn S.
The energy difference (1.5 kcal/mol, MP2) between the
two diastereomeric transition structures anti R and anti
S would indicate that this ligand should induce some
degree of enantioselectivity in the reaction, albeit not as
high as DAIB (3.0 kcal/mol, B3LYP).6b
(19) We independently located the 6/6 transition structures for the
ligands 1-3, similar to those reported by Norrby et al. in ref 18, while
locating the different possible 6/4/4 transition structures at the PM3
level (see Supporting Information). The PM3 calculations were done
using SPARTAN version 5.0. Wavefunction, Inc., 18401 Von Karman
Avenue, Suite 370, Irvine, CA 92612.
568 J . Org. Chem., Vol. 68, No. 2, 2003

Organozinc Additions to Benzaldehyde
TABLE 2. Tota l En er gies (Ha r tr ees) a n d Rela tive En er gies (in P a r en th eses; k ca l/m ol) of th e Tr icyclic Dia ster eom er ic
Tr a n sition Str u ctu r es for th e Liga n d s 1-3 a n d Bicyclic On es for 3 fr om a b in itio a n d DF T Ca lcu la tion s (La n L2DZ Ba sis
Set Used Th r ou gh ou t)
HF//HF
B3LYP//HF
MP2//HF
1
anti R
anti S
syn R
syn S
-1107.51724 (0.00)
-1107.51277 (2.80)
-1107.50814 (5.71)
-1107.51166 (3.45)
-1118.26217 (0.00)
-1118.25879 (2.12)
-1118.25477 (4.64)
-1118.25716 (3.14)
-1109.99443 (0.00)
-1109.99205 (1.49)
-1109.98574 (5.45)
-1109.98681 (4.78)
2
3
anti R
anti S
syn R
syn S
-1107.52172 (0.00)
-1107.51735 (2.74)
-1107.51149 (6.42)
-1107.51634 (3.37)
-1118.26609 (0.00)
-1118.26305 (1.91)
-1118.25759 (6.31)
-1118.26119 (3.06)
-1109.99640 (0.00)
-1109.99402 (1.49)
-1109.98635 (5.33)
-1109.98940 (4.44)
anti R
anti S
syn R
syn S
-1029.47038 (0.00)
-1029.46786 (1.58)
-1029.46438 (3.76)
-1029.46658 (2.38)
-1039.64436 (0.00)
-1039.64251 (1.16)
-1039.63893 (3.41)
-1039.64035 (2.52)
-1031.76558 (0.00)
-1031.76448 (0.44)
-1031.75969 (3.70)
-1031.75962 (3.74)
3 (6/6)
twist boat R
boat S
-1029.45914 (7.05)
-1029.46588 (2.82)
-1039.63584 (5.35)
-1039.64110 (2.04)
-1031.75774 (4.92)
-1031.76419 (0.87)
F IGURE 7. HF/LanL2DZ optimized geometries of the tran-
sition structures with the ligand 3. Relative MP2 energies
(kcal/mol) are given in the parentheses. For clarity, only the
hydrogens at stereogenic centers are shown.
F IGURE 6. HF/LanL2DZ optimized geometries of the tran-
sition structures with the ligand 2. Relative MP2 energies
(kcal/mol) are given in the parentheses. For clarity, only the
hydrogens at stereogenic centers are shown.
ligand. The most significant geometrical change between
the 6/4/4 and 6/6 transition structures occurs at the Zn3
atom which becomes trigonal planar and no longer
coordinates the carbonyl oxygen.
tion structures identified by Norrby. The chair transition
structure described by Norrby was not a saddlepoint for
these systems due to steric interactions with the chiral
In the HF/LanL2DZ calculations, the 6/6 transition
structures were found to be higher in energy than their
6/4/4 analogues (Table 2). In addition, the boat S transi-
J . Org. Chem, Vol. 68, No. 2, 2003 569

Panda et al.
Formation of six-membered chelates between the γ-ami-
no alcohols and zinc places the benzaldehyde substrate
further from the ligand compared to the â-amino alcohols,
which form five-membered chelates. Thus, the steric
effects offered by the ligand substituents in 1-3 are
reduced, resulting in lower enantioselectivity. These
ligands also differ from many of the successful â-amino
alcohols, in that they lack bulky substituents on the
carbons adjacent to the nitrogen and oxygen. As a result,
there are smaller repulsive interactions between the
chiral catalysts and the substrates.
On the basis of prior results and the results from
conformationally restricted 3 which match the calcula-
tions very well (Table 3), it appears reasonable that
transition structure calculations can be a predictive tool
for estimating reaction enantioselectivity. The discrep-
ancy between the calculated results and the experimental
data for 1 and 2 most likely arises from the intervention
of other ligand-metal species due to the conformationally
flexible nature of 1 and 2 (see Figure 4). These results
indicate that caution must be exercised in the design of
new systems, since the presence of other species which
may be or may not be catalytically relevant can alter the
basic reaction paradigm.
F IGURE 8. HF/LanL2DZ optimized geometries of the 6/6
transition structures leading to (R) and (S) products using the
ligand 3. The relative energies (kcal/mol) with respect to the
most stable 6/4/4 transition structure at the MP2 level are
given in parentheses.
TABLE 3. Com p a r ison of th e Ca lcu la ted Resu lts w ith
th e Exp er im en ta l Da ta a
calculated (MP2//HF)
Con clu d in g Rem a r k s
∆E (kcal/mol)
ee (%)
exp ee (%)
We have synthesized, resolved, and examined γ-amino
alcohols 1-5 in the asymmetric organozinc addition to
benzaldehyde. While preliminary PM3 calculations had
predicted that 2 would be highly selective, we observed
only moderate selectivity. Thus, the origin of the enan-
tioselectivity for γ-amino alcohols 1-3 was explored using
higher level quantum calculations. The calculations
predict that the (R) alcohol should be the predominant
product for all the three ligands. Results from 2 and 3
agree with this trend, whereas those from 1 do not. The
12% ee decrease from 2 to 3 is also reflected in the
calculations. Thus, the MP2//HF method can be useful
in estimating the relative selectivity of these γ-amino
alcohols.
1
2
3
1.49
1.49
0.44
85 (R)
85 (R)
35 (R)
4 (S)
46 (R)
34 (R)
a
∆E is the energy difference between the two lower energy
transition structures giving (R) and (S) alcohols. Calculated
enantiomeric excess is obtained from a Boltzmann distribution of
the transition structures at 298 K.
tion structure is lower in energy compared to twist boat
R, which predicts the (S) alcohol as the dominant product
for 3 in contrast to the experimental finding. These
results imply that the 6/6 transition structures are not
playing a major role in the reactions of these γ-amino
alcohol ligands.
The low enantioselectivities for these ligands may be
attributed to two factors. First, the energy difference
between the two lowest energy transition structures to
the enantiomeric products is smaller compared to other
highly selective ligands. For instance, the energy differ-
ence between the two lowest energy transition structures
incorporating DAIB6b is about twice that of 1, 2, and 3.
Second, the conformational flexibility of 1 and 2 appears
to give rise to other catalytic ligand-zinc species with
transition structures (Figure 4) different than the µ-oxo
tricyclic structures used in the calculations.
While 6/6 transition structures may be low-energy
pathways in some systems, they were found to be higher
in energy than the 6/4/4 transition structures for 1-3
when HF, DFT, and MP2 methods were employed. The
6/6 transition structure of 3 leading to (S) product is also
lower in energy than that corresponding to the (R)
product (in contrast to experiment), which further con-
traindicates the intervention of the 6/6 transition struc-
tures for these ligands.
Com p a r ison of th e Ca lcu la ted a n d Exp er im en ta l
Da ta for Liga n d s 1-3. For the DAIB ligand which gives
very high enantioselectivity (95% ee), a 3.5 kcal/mol
energy difference between the two lowest energy transi-
tion structures leading to the (R) and (S) products was
reported by Yamakawa and Noyori6a from B3LYP calcu-
lations using a basis set similar to the one employed here.
For the ligands 1, 2, and 3, we found smaller energy
differences of 2.1, 1.9, and 1.2 kcal/mol, respectively,
using B3LYP calculations. The MP2//HF results reveals
even smaller energy differences (1.5 kcal/mol for 1 and
2, and 0.4 kcal/mol for 3). Thus, these calculations predict
that our ligands would be less selective than DAIB. In
toluene, the diethylzinc addition to benzaldehyde cata-
lyzed by 1, 2, and 3 give enantiomeric excesses of 4% (S),
46% (R), and 34% (R), respectively. Qualitatively, the
calculated selectivities match the observed sense of
stereoselection, except for the case of 1, which provides
a minor selectivity (4% ee) for the (S) alcohol (Table 3).
From 2 to 3, the energy difference between anti R and
anti S transition structures decreases by 1.1 kcal/mol.
This general trend is qualitatively consistent with the
experimental 12% ee decrease from 2 to 3.
Overall, we have found that PM3 calculations are not
reliable for anticipating the selectivity in asymmetric
alkylation of amino alcohol catalysts which diverge from
570 J . Org. Chem., Vol. 68, No. 2, 2003

Organozinc Additions to Benzaldehyde
the â-amino alcohol framework. Among the quantum
chemical calculations examined (HF, DFT, MP2), MP2
proved most useful in estimating the selectivity for the
γ-amino alcohols. Even so, caution must be employed in
the calculations of conformationally flexible amino alco-
hols, as other catalyst forms may intervene.
was provided by the NCSA in the form of a supercom-
puting time allotment. The invaluable assistance of Dr.
George Furst in collecting the NMR spectroscopic data
and of Dr. Patrick Carroll in obtaining the X-ray
structures is gratefully acknowledged.
Su p p or tin g In for m a tion Ava ila ble: Full experimental
details and characterization, X-ray structure methods and data
tables for 13 and 14, results from PM3 calculations along with
the coordinates for the HF transition structures, and geom-
etries of the transition structures at HF/LanL2DZ. This
material is available free of charge via the Internet at
http://pubs.acs.org.
Ack n ow led gm en t. Financial support was provided
by the National Institutes of Health (GM-59945) and
Merck Research Laboratories. Acknowledgment is made
to the donors of the Petroleum Research Fund, admin-
istered by the American Chemical Society, for partial
support of this research. We thank Dr. Bishwajit Gan-
guly for performing the initial PM3 6/4/4 transition
structure calculations with 1-2. Computational support
J O0262210
J . Org. Chem, Vol. 68, No. 2, 2003 571