Nonradical zinc-barbier reaction for diastereoselective synthesis of vicinal amino alcohols

Article abstract of DOI:10.1021/ja054706a

A new protocol for the synthesis of vicinal amino alcohols is described. The method employs a Barbier-type reaction between an imine and 3-benzoyloxyallyl bromide in the presence of zinc metal. The addition products are debenzoylated to afford amino alcohols in good yields and with diastereomeric ratios greater than 85:15 in favor of the anti isomer. A Hammett study has been performed which strongly indicates that the allylation does not follow a radical mechanism, but instead an organometallic reagent is formed which subsequently reacts with the imine. A computational study based on this mechanism reproduces the observed diastereoselectivity with high accuracy, but only when a sufficiently large portion of the substrate is included.

Full text of DOI:10.1021/ja054706a

Published on Web 10/25/2005
Nonradical Zinc-Barbier Reaction for Diastereoselective
Synthesis of Vicinal Amino Alcohols
Lise Keinicke, Peter Fristrup, Per-Ola Norrby,* and Robert Madsen*
Contribution from the Center for Sustainable and Green Chemistry, Department of Chemistry,
Technical UniVersity of Denmark, Building 201, KemitorVet, DK-2800 Lyngby, Denmark
Received July 14, 2005; E-mail: pon@kemi.dtu.dk; rm@kemi.dtu.dk
Abstract: A new protocol for the synthesis of vicinal amino alcohols is described. The method employs a
Barbier-type reaction between an imine and 3-benzoyloxyallyl bromide in the presence of zinc metal. The
addition products are debenzoylated to afford amino alcohols in good yields and with diastereomeric ratios
greater than 85:15 in favor of the anti isomer. A Hammett study has been performed which strongly indicates
that the allylation does not follow a radical mechanism, but instead an organometallic reagent is formed
which subsequently reacts with the imine. A computational study based on this mechanism reproduces
the observed diastereoselectivity with high accuracy, but only when a sufficiently large portion of the substrate
is included.
Scheme 1. Some Synthetic Strategies for Production of Vicinal
Amino Alcohols
Introduction
Vicinal amino alcohols are important structural elements in
organic chemistry.1 Many natural products, pharmacologically
active compounds, and chiral auxiliaries contain amino alcohols
as major recognition elements. Amino alcohols are also
significant components in ligands for asymmetric catalysis.
Current strategies for stereoselective synthesis of amino
alcohols are mainly based on olefin oxidation (Scheme 1), either
through direct aminohydroxylation2 or via other reactive func-
tionalities such as epoxides3 or aziridines.4 These can be
performed with good stereoselectivity but not always with
complete regiocontrol. Alternatively, carbon nucleophiles can
be added to protected R-hydroxyimines5 or R-aminoaldehydes.6
We have recently become interested in another strategy, where
the amino alcohol is built in a convergent fashion by formation
of the C-C bond. Recent applications of this approach include
addition of dipolar reagents7 or suitably functionalized orga-
nometallic reagents8 to imines. Catalytic asymmetric Mannich-
type reactions of R-hydroxyketones and imines using chiral
ligands have also been reported.9
homoallylic alcohols and amines, respectively. It has recently
been shown that acyloxyallyl bromides, derived from acyl
bromides and acrolein,10 can also be added to aldehydes to yield
diols (Scheme 2, left).11 Of particular interest to us was a recent
report where in one case acetyloxyallyl bromide was found to
add to a protected imine formed in situ from an R-amidosulfone,
yielding an amino alcohol (Scheme 2, right).8
Under Barbier-type conditions using metals such as Zn and
In, allyl bromides add to aldehydes and aldimines, yielding
We have previously performed Barbier-type allylations of
carbohydrate-based aldehydes, which combined with ring-
closing metathesis forms a general route to cyclic polyols.12
Amino alcohols could also be generated by allylation of
hydroxyimines using allyl bromide in the presence of Zn, In,
or Mg, which was used as a critical step in a recent synthesis
of the calystegine alkaloids.13 As an extension of this work, we
(1) Bergmeier, S. C. Tetrahedron 2000, 56, 2561-2576 and references therein.
(2) Li, G.; Chang, H.-T.; Sharpless, K. B. Angew. Chem., Int. Ed. Engl. 1996,
35, 451-454.
(3) (a) Larrow, J. F.; Schaus, S. E.; Jacobsen, E. N. J. Am. Chem. Soc. 1996,
118, 7420-7421. (b) Olofsson, B.; Somfai, P. J. Org. Chem. 2002, 67,
8574-8583.
(4) Hu, X. E. Tetrahedron 2004, 60, 2701-2743.
(5) Bloch, R. Chem. ReV. 1998, 98, 1407-1438.
(6) Reetz, M. T. Chem. ReV. 1999, 99, 1121-1162.
(10) (a) Neuenschwander, M.; Bigler, P.; Christen, K.; Iseli, R.; Kyburz, R.;
Mu¨hle, H. HelV. Chim. Acta 1978, 61, 2047-2058. (b) Bigler, P.;
Scho¨nholzer, S.; Neuenschwander, M. HelV. Chim. Acta 1978, 61, 2059-
2080.
(11) (a) Lombardo, M.; Girotti, R.; Morganti, S.; Trombini, C. Chem. Commun.
2001, 2310-2311. (b) Lombardo, M.; Girotti, R.; Morganti, S.; Trombini,
C. Org. Lett. 2001, 3, 2981-2983. (c) Lombardo, M.; Morganti, S.;
Trombini, C. J. Org. Chem. 2003, 68, 997-1006.
(12) Hyldtoft, L.; Madsen, R. J. Am. Chem. Soc. 2000, 122, 8444-8452.
(13) Skaanderup, P. R.; Madsen, R. J. Org. Chem. 2003, 68, 2115-2122.
(7) Torssell, S.; Kienle, M.; Somfai, P. Angew. Chem., Int. Ed. 2005, 44, 3096-
3099.
(8) Petrini, M.; Profeta, R.; Righi, P. J. Org. Chem. 2002, 67, 4530-4535.
(9) (a) List, B.; Pojarliev, P.; Biller, W. T.; Martin, H. J. J. Am. Chem. Soc.
2002, 124, 827-833. (b) Trost, B. M.; Terrell, L. R. J. Am. Chem. Soc.
2003, 125, 338-339. (c) Matsunaga, S.; Yoshida, T.; Morimoto, H.;
Kumagai, N.; Shibasaki, M. J. Am. Chem. Soc. 2004, 126, 8777-8785.
(d) Kobayashi, S.; Ueno, M.; Saito, S.; Mizuki, Y.; Ishitani, H.; Yamashita,
Y. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 5476-5481.
9
15756
J. AM. CHEM. SOC. 2005, 127, 15756-15761
10.1021/ja054706a CCC: $30.25 © 2005 American Chemical Society

Nonradical Zn
−
Barbier Synthesis of Amino Alcohols
A R T I C L E S
Scheme 3. Model System for the Allylation of Imines
Scheme 2. Barbier-type Synthesis of Amino Alcohols and Diols
Scheme 4. Allylation of Benzylideneanilines 3a and 3b
are currently investigating the scope of the C-C disconnection
approach depicted in Scheme 1. In the current work, we present
reactions of a model system and a mechanistic investigation of
the title reaction, whereas synthetic applications will be reported
separately.
As opposed to classical organometallic addition reactions such
as Grignard, where an organometallic reagent is prepared and
subsequently added to an electrophile, in allylation reactions
run under Barbier conditions the metal is added to a solution
already containing both the halide and the electrophile. Several
mechanistic pathways have been suggested. The initial reaction
could potentially be formation of a Grignard-type organometallic
reagent, but the reaction is unusually tolerant and can even be
performed in water, which would be expected to protonate the
organometallic reagent rapidly.14 Furthermore, formation of
dimerization products, both of the allylic reagent and the
electrophile, indicates a mechanism with single-electron transfer
(SET) from the metal as the first step.15 These observations make
a SET mechanism plausible, but cannot exclude the more
traditional mechanism via an initially formed, closed shell
organometallic species. We therefore decided to study the
mechanism of the title reaction in more detail.
Absolute rate measurements for the title reaction are com-
plicated by the presence of a heterogeneous reagent, which
precludes many types of spectroscopic monitoring and which
may introduce mass transfer over phase boundaries as an
important but unknown factor in the kinetic equations. We have
elected to perform a Hammett study, which can be implemented
as a competition experiment between the substrates and therefore
is not sensitive to steps preceding the reaction with the
electrophilic substrate. We have previously applied the Hammett
methodology to a range of metal-assisted reactions.16 As our
current model system, we have chosen reactions with para-
substituted N-benzylidene anilines. Benzoate reagent 2 was
chosen over the corresponding acetate 1, due to ease of handling
and higher storage stability (Scheme 3).
performed for the unsubstituted and p-methoxy-substituted
substrates 3a and 3b, respectively. With the use of zinc in THF
the desired allylation of the imine took place, but during the
reaction the expected product 5a was partially converted to the
more stable amido alcohol 6a by migration of the benzoyl group
(Scheme 4). One major diastereomer of 6a was observed.
Heating the reaction mixture or changing the solvent to dioxane
changed the ratio of 5a/6a, but it was not possible to achieve
high selectivity for one of the products. The highest isolated
yields were 72% of 6a using the initial reaction conditions and
50% of 5a from the reaction in dioxane. However, the
diastereoselectivity was high for both products. Subsequent
debenzoylation gave 4a in high overall yield, and with one
diastereomer clearly dominating (90:10), showing that the major
isomer was the same for both 5a and 6a. The major diastereomer
could be assigned as anti by conversion of 4a into oxazolidinone
7a through treatment with 1,1-carbonyldiimidazole (CDI) fol-
lowed by hydrogenation of the double bond. The final reduction
1
was performed in order to avoid overlap of signals in the H
NMR spectrum. Comparison to literature data for the coupling
constant J4,5 in analogous compounds17 verified the stereochem-
ical assignment. Imine 3b was subjected to the same reaction
sequence as 3a (Scheme 4), and the coupling constant J4,5 in
oxazolidinone 7b indicated that the stereochemistry for this
compound was also anti. The anti/syn ratio of 4b was found to
be 95:5, even higher than that for 4a.
We have previously observed interesting changes in the
selectivity of the Barbier allylation when changing the metal
from Zn to In.12,13 In the current case, imine 3a was also
allylated with 2 and indium metal in anhydrous ethanol (Scheme
Results and Discussion
Synthesis of Amino Alcohols 4. Imines 3a-f were prepared
from condensation of benzaldehyde and the corresponding
anilines (see the Supporting Information for experimental
details). The initial reactions and optimization experiments were
(14) (a) Pe´trier, C.; Luche, J.-L. J. Org. Chem. 1985, 50, 910-912. (b) Wilson,
S. R.; Guazzaroni, M. E. J. Org. Chem. 1989, 54, 3087-3091.
(15) (a) Sjo¨holm, R.; Rairama, R.; Ahonen, M. J. Chem. Soc., Chem. Commun.
1994, 1217-1218. (b) Waldman, H. Synlett, 1990, 627-628.
(16) (a) Fristrup, P.; Le Quement, S.; Tanner, D.; Norrby, P.-O. Organometallics
2004, 23, 6160-6165. (b) Linde, C.; Koliai, N.; Norrby, P.-O.; A° kermark,
B. Chem. Eur. J. 2002, 8, 2568-2573. (c) Rasmussen, T.; Jensen, J. F.;
Østergaard, N.; Tanner, D.; Ziegler, T.; Norrby, P.-O. Chem. Eur. J. 2002,
8, 177-184.
(17) (a) Yamamoto, Y.; Schmid, M. J. Chem. Soc., Chem. Commun. 1989,
1310-1312. (b) Murakami, M.; Ito, H.; Ito, Y. J. Org. Chem. 1993, 58,
6766-6770. (c) Barrett, A. G. M.; Seefeld, M. A.; White, A. J. P.; Williams,
D. J. J. Org. Chem. 1996, 61, 2677-2685.
9
J. AM. CHEM. SOC. VOL. 127, NO. 45, 2005 15757

A R T I C L E S
Keinicke et al.
Table 1. One-Pot Procedure for the Synthesis of Amino Alcohols
4a-f (Scheme 3)a
isolated yield
entry
imine
X
amino alcohol
(%)
anti/synb
1
2
3
4
5
6
3a
3b
3c
3d
3e
3f
H
OMe
Me
4a
4b
4c
4d
4e
4f
76
70
82
85
73
86
90:10
95:5
85:15
95:5
95:5
95:5
Cl
COOEtc
CF3
a (i) 2, Zn, THF; (ii) MeONa, MeOH, 60 °C. b From 1H NMR. c EtONa
in EtOH was used for the deprotection, step (ii).
4). Interestingly enough, benzoyl migration could be suppressed,
and the amino ester 5a was isolated as the sole product in 80%
yield. However, the diastereoselectivity was only 1.2:1, in favor
of the syn product. Due to the low diastereoselectivity with this
substrate, allylations with indium were not pursued further.
To attain a single product, a one-pot procedure was developed
to give the deprotected amino alcohol 4 directly. Thus, 3a was
allylated in THF using zinc and 2, and the reaction mixture
was filtered. The crude reaction mixture was then treated with
a solution of sodium methoxide in methanol at 60 °C.
Subsequent workup and purification by column chromatography
yielded compound 4a in 76% with a diastereomeric ratio of
10:1 in favor of the anti isomer. Imines 3b-f were subjected
to the same procedure; the yields and anti/syn ratios are collected
in Table 1. The diastereomeric mixtures could not be separated
by flash chromatography. However, simple recrystallization in
all cases furnished the pure anti amino alcohols 4.
Figure 1. Kinetic data for competitive allylation of imines 3b-f.
Table 2. Rates and σ Values for Imines 3a
substrate
X
σ
σ+
σ•
krel
log(krel)
3b
3c
3d
3e
3f
OMe
Me
Cl
COOEt
CF3
-0.27
-0.17
0.23
0.45
0.54
-0.78
-0.31
0.11
0.48
0.61
0.24
0.11
0.12
0.35b
0.08
0.61
0.95
1.38
2.56
2.71
-0.214
-0.022
0.138
0.408
0.432
a From refs 18 and 19. b The σ• value for COOMe was used.
Hammett Study. In the competition studies, the reactions
can be followed by monitoring either appearance of products
or, if no significant side reactions are occurring, disappearance
of starting material. Since the reaction of each imine yields four
initial products (syn and anti isomers of 5 and 6, respectively),
and the isolated yields in the reactions uniformly are high (Table
1), we decided to monitor the concentrations of imines 3 using
GC with anthracene as internal standard. We postulate that the
reaction order in all reactants is the same for all imines.
Furthermore, we assume that the reaction is first order in imine.
Under these conditions, the imine concentrations will follow
eq 1, where subscript “0” denotes initial concentration, X is
the substituted imine 3b-f, H is the unsubstituted imine 3a,
and krel is the relative reaction rate of the two imines.
Figure 2. Hammett plot of log(krel) vs σ, σ+, and σ• values.
addition, the kinetic plots all yielded straight lines with
correlation coefficients of r2 > 0.99 (Figure 1). It is unlikely
that such straight lines could be obtained if significant amounts
of imines participated in side reactions. The relative rates and
different σ values for all substrates 3 are shown in Table 2.
The resulting krel values are used in the Hammett plots shown
in Figure 2. It can be seen that the fit is best using the regular
σ values, with a correlation coefficient of r2 ) 0.94 and a slope
of F ) 0.79. Trying to employ the σ+ values results in an inferior
correlation with r2 ) 0.67 (line not shown). One of the main
goals of the current Hammett study is to differentiate between
closed shell or radical mechanisms, which can be done by
comparing the fits to σ and σ• values, respectively. Several
different sets of σ• values are available, each with its own
advantages. In our case, the conclusions are independent of
which set is used, so only the results using the Creary σ•
values19a are discussed here. It can be clearly seen that the fit
to the σ• values is much worse, excluding the possibility of a
X0
H0
ln
) k ln
(1)
( )
( )
rel
X
H
Thus, by plotting ln(X0/X) versus ln(H0/H), we obtain the
relative reaction rate krel ) kX/kH as the slope. We then plot
log(krel) versus σX in the standard Hammett plot to obtain the
reactivity parameter F as the slope and also to check the
correlation against different types of σ values.
The five competition experiments between imine 3a and
imines 3b-f were performed in THF at ambient temperature.
Initial studies indicated that the allylation, after an initial lag
phase, was too rapid to allow a representative sampling. The
allylic reagent 2 was therefore added in portions, ensuring that
samples could be withdrawn at evenly spaced conversions. We
were concerned that in the absence of reagent 2, the metal might
consume the imine in pinacol-type reactions, but a stability test
showed that the imine 3a was stable in the presence of zinc. In
(18) Hansch, C.; Leo, A.; Taft, R. W. Chem. ReV. 1991, 91, 165-195.
(19) (a) Creary, X.; Mehrsheikh-Mohammadi, M. E.; McDonald, S. J. Org.
Chem. 1987, 52, 3254-3263. (b) Dust, J. M.; Arnold, D. R. J. Am. Chem.
Soc. 1983, 105, 1221-1227. (c) Jiang, X.-K.; Ji, G.-Z. J. Org. Chem. 1992,
57, 6051-6056. (d) Hansch, C.; Gao, H. Chem. ReV. 1997, 97, 2995-
3059.
9
15758 J. AM. CHEM. SOC. VOL. 127, NO. 45, 2005

Nonradical Zn−Barbier Synthesis of Amino Alcohols
A R T I C L E S
single-electron transfer (SET) mechanism. In particular the
COOEt and CF3-groups provide an interesting comparison. The
two groups have similar σ and σ+ values, but differ strongly
for σ•; the ester can stabilize a radical through resonance,
whereas the CF3 group cannot. Thus, the two groups should
induce very different rates in a radical reaction, but in the current
study, the rates of the two substrates are similar. In addition,
none of the groups employed here has a negative σ• value, but
still we see that electron-donating substituents retard the reaction.
Going back to the Hammett plot using the regular σ values,
the weakly positive F is consistent with a nucleophilic attack
on the imine and indicates that the electrophilicity of the imine
has a higher influence on the rate than the ability of the imine
nitrogen to coordinate to the metal.
The Hammett study strongly indicates that the allylic bromide
initially reacts with zinc to form an organometallic reagent,
which subsequently reacts with the imine. This opens up the
possibility to study the reaction computationally with a fair
accuracy. The SET mechanism would in principle require high-
level methods able to calculate accurate structures and energies
for open shell species on a metal surface. Such calculations are
possible today, but well beyond our resources. However, with
the current data in hand, we estimate that the reaction path can
be studied starting from the isolated organometallic reagent. We
have previously been involved in several computational studies
on the addition of isolated zinc reagents to carbonyl com-
pounds.20 The goal of the current study is to understand the
observed diastereoselectivity in the reaction.
Computational Study. We have initiated the study using
small model complexes, first to enable us to select a suitable
computational methodology, then to screen all potential paths
for the reaction. Subsequently, the most promising reaction paths
have been studied using the full substrate 3a, to give final
predicted selectivities. All calculations have been run in Jaguar21
using the B3LYP functional22 in combination with the LACVP*
basis set.23 All gas-phase transition states were verified by
normal analysis obtained from a mass-weighted force-constant
matrix built up with analytical energy second derivatives. Each
transition state displayed exactly one mode with a negative
eigenvalue, as required. The corresponding eigenmodes were
inspected visually in Molekel24 and found to correspond to
Zimmerman-Traxler transition states, as indicated in Figure
3. As a further validation, a sample transition state structure of
the small model system was displaced forward and backward
along the negative eigenmode and subjected to energy mini-
mization, yielding the expected reactant and product.
Figure 3. Model systems employed in the computational study.
Calculations in solvent use the PB-SCRF method25 in Jaguar,
employing parameters suitable for THF (dielectric constant 7.43,
probe radius 2.52372 Å). PB-SCRF is a continuum solvation
model, where the molecule is put into a reaction field consisting
of surface charges on a solvent accessible surface constructed
using a hypothetical spherical solvent probe molecule with the
indicated radius.26 The wave function and the reaction field
charges are solved iteratively until self-consistency is reached.
In our experience, the major role of this electrostatic model is
to screen excessive charge separation. For other metal-assisted
reactions, we have shown that solvation models of this type
give agreement with experimental data in cases where gas-phase
calculations deviate strongly.27 In the current case, we found
that for the small model system, the neutral complex (including
the bromide counterion) gave very similar results in the gas
phase and in solvent. The six-membered rings of the Zimmer-
man-Traxler transition states were virtually superimposable.
Furthermore, the energy difference between the boat and chair
forms was found to be the same in the gas phase and in solvent
(16 kJ/mol in favor of the chair). The position of the bromide
did differ between the two calculations, and the loose coordina-
tion of the bromide was found to give severe convergence
problems in the solvent calculations. We also tested the cationic
model system (without the bromide counterion). Gratifyingly,
the structures obtained with the cationic model system in solvent
were very close to those obtained with the bromide ion present,
and the relative energy of the boat and chair forms only shifted
slightly, to 14 kJ/mol. Not unexpectedly, when the cationic
system was reconverged in the gas phase, the structures were
distorted, even though the relative energy of the boat and the
chair was still reasonable, 20 kJ/mol.
From previous projects,27 we have found that energies
calculated with the continuum solvent model gives a fair
correspondence with experimental ratios, but for a quantitative
agreement we frequently also need to account for the vibrational
component of the free energy. With our current resources, this
component must be determined from gas-phase calculations, and
the combination of this contribution with the solvated energies
is only valid if the geometries in the gas phase and solvent are
similar. Thus, in the current approach, we have determined all
transition states both in solvent and in the gas phase, verified
the similarity of the structures by superimposition, and per-
formed a vibrational analysis in the gas phase. The vibrational
analysis was used both to verify that the eigenmode correspond-
ing to the single negative eigenvalue indeed corresponds to the
We first studied the neutral and cationic forms of the small
model system (Figure 3), in the gas phase and in solvent.
(20) (a) Rasmussen, T.; Norrby, P.-O. J. Am. Chem. Soc. 2001, 123, 2464-
2465. (b) Rudolph, J.; Rasmussen, T.; Bolm, C.; Norrby, P.-O. Angew.
Chem., Int. Ed. 2003, 42, 3002-3005. (c) Rudolph, J.; Bolm, C.; Norrby,
P.-O. J. Am. Chem. Soc. 2005, 127, 1548-1552. (d) Cozzi, P. G.; Rudolph,
J.; Bolm, C.; Norrby, P.-O.; Tomasini, C. J. Org. Chem. 2005, 70, 5733-
5736.
(21) Jaguar 4.2; Schro¨dinger, Inc.: Portland, OR, 2000. For the most recent
version, see: http://www.schrodinger.com.
(22) (a) Becke, A. D. J. Chem. Phys. 1993, 98, 5648-5652. (b) Lee, C.; Yang,
W.; Parr, R. G. Phys. ReV. B 1988, 37, 785-789. (c) Stephens, P. J.; Devlin,
F. J.; Chabalowski, C. F.; Frisch, M. J. J. Phys. Chem. 1994, 98, 8, 11623-
11627.
(25) Marten, B.; Kim, K.; Cortis, C.; Friesner, R. A.; Murphy, R. B.; Ringnalda,
M. N.; Sitkoff, D.; Honig, B. J. Phys. Chem. 1996, 100, 11775-11788.
(26) For an illuminating discussion about implicit solvation models in general,
see: Cramer, C. Essentials of Computational Chemistry: Theories and
Models; Wiley: New York, 2002.
(23) LACVP* uses the 6-31G* basis set for all light elements and the Hay-
Wadt ECP and basis set for Zn and Br; see: Hay, P. J.; Wadt, W. R. J.
Chem. Phys. 1985, 82, 270-283.
(24) Molekel, version 4.3, 2002, Stefan Portmann, CSCS/ETHZ; see http://
www.cscs.ch/molekel/.
(27) (a) Hagelin, H.; A° kermark, B.; Norrby, P.-O. Chem. Eur. J. 1999, 5, 902-
909. (b) Norrby, P.-O.; Mader, M.; Vitale, M.; Prestat, G.; Poli, G.;
Organometallics 2003, 22, 1849-1855. (c) Kieken, E.; Wiest, O.; Helquist,
P.; Cucciolito, M. E.; Flores, G.; Vitagliano, A.; Norrby, P.-O. Organo-
metallics 2005, 24, 3737-3745.
9
J. AM. CHEM. SOC. VOL. 127, NO. 45, 2005 15759

A R T I C L E S
Keinicke et al.
Table 3. Calculated Relative Transition State Energies for the
Intermediate Model System (Figure 3), in kJ/mol
Table 4. Calculated Relative TS Energies in Solvent for the Large
Model System (Figure 3), in kJ/mol
gas phase
solvent (THF)
TS
allyl
C
d
to Zna
O rel
solvent (THF)
TS
allyl
CdO rel
to Zna
conformation
configuration
∆E
q
product
conformation
configuration
∆
E
q
∆∆G
q b
∆E
q
product
chair
chair
chair
chair
boat
boat
boat
boat
Z
Z
E
E
Z
Z
E
E
prox
dist
prox
dist
prox
dist
prox
dist
14
0
18
10
3
33
35
25
anti
anti
syn
syn
syn
syn
anti
anti
chair
chair
chair
chair
boat
boat
boat
boat
Z
Z
E
E
Z
Z
E
E
prox
dist
prox
dist
prox
dist
prox
dist
21
2
18
0
5
33
0
16
0
anti
anti
syn
syn
syn
syn
anti
anti
1
0
0
6
1
1
0
1
9
27
27
14
36
27
13
a Of the two possible orientations of the carboxylate-allyl bond, in one
the CdO bond will be oriented toward Zn (prox), in the other it will be
oriented away (dist).
a Of the two possible orientations of the carboxylate-allyl bond, in one
the CdO bond will be oriented toward Zn (prox), in the other it will be
oriented away (dist). b The relative vibrational corrections (including ZPE)
at 298 K, add to ∆Eq to obtain relative free energies.
expected reaction coordinate and to calculate the free energy
contribution (at 298 K, including ZPE). The free energy
adjustment was then added to the corresponding energy
determined in solvent, to arrive at a composite free energy which
is our best estimate of the free energy in solvent. For these
calculations, it is clear from the initial results that the gas-phase
calculations must be performed for the neutral complex. For
the calculations run in solvent, the presence of the counterion
did not seem to have an influence on the quality of the results,
since both the structure and energy were insensitive to the
presence of the bromide. We therefore elected to perform the
solvated calculations using the cationic model system, since the
presence of the loosely coordinated bromide ion had a severe
detrimental effect on the convergence properties of the transition
state optimizations.
Moving on to the intermediate model system, we have
investigated the possible variations in the Zimmerman-Traxler
transition state shown in Figure 3. Thus, the addition can proceed
via a boat or a chair transition state, the position of the
carboxylate in the allylic reagent can be either E or Z, and
finally, the carboxylate moiety can rotate around the bond to
the allyl moiety, pointing the carbonyl either toward or away
from the Zn atom. Coordinate scanning indicated that only these
two orientations are feasible for the carboxylate moiety. The
energies for the eight possible transition states of the intermedi-
ate model are shown in Table 3.
Figure 4. Best transition state with the large model system, the chair-Z-
dist isomer, leading to the anti product.
We first note that the vibrational contributions, although large
in absolute value (see the Supporting Information), contribute
very little to the relatiVe energy of the transition states. The
only significant effect is for the transition states with Z-
configuration where the CdO bond is oriented toward Zn,
limiting the vibrational freedom and therefore incurring a slight
energy cost, ca. 5 kJ/mol. In the boat, the carbonyl oxygen
actually coordinates to Zn (distance Zn-O ) 2.16-2.17 Å).
In the gas phase, the prox-Z-boat is therefore one of the best
conformations and the only boat with any significance. However,
it is clear that this coordination is less favorable in solvent, where
only dist-chair conformations have a low enough energy to
contribute a significant amount of product. These two are
isoenergetic, whereas the experimentally observed ratios (Table
2) correspond to an activation free energy difference of 6-8
kJ/mol.
of the large model system, which includes the two phenyl groups
of substrate 3a, are shown in Table 4.
The data in Table 4 indicate that only three transition states
are of interest. The lowest energy TS is the chair-Z-dist isomer
(Figure 4), leading to the anti diastereomer, which is also
observed as the major product. Two low energy transition states
lead to the syn products, boat-Z-prox and chair-E-dist (Figure
5). We know from the vibrational analysis using the intermediate
model system that the vibrational contribution is significant for
the boat-Z-prox TS. Thus, for the three lowest energy structures,
we also located the transition states for the large model system
using the neutral gas-phase structures (including the bromide
counterion). After verifying the structural similarity with the
corresponding solvated structures, we performed a vibrational
analysis to arrive at the final composite free energies described
earlier. In this final analysis, the chair-Z-dist isomer is still the
global optimum, but now the boat-Z-prox is 8 kJ/mol higher,
and the isomer is shifted to 12 kJ/mol. These numbers are in
very good agreement with the experimental syn/anti difference
The above results were somewhat encouraging, but not
sufficiently accurate to be useful, so we decided to increase the
size of the model system. The solvent data for the 8 TS isomers
9
15760 J. AM. CHEM. SOC. VOL. 127, NO. 45, 2005

Nonradical Zn−Barbier Synthesis of Amino Alcohols
A R T I C L E S
has not been investigated computationally, but it has been
observed experimentally that the diastereomeric ratio of the
product is insensitive to the E/Z ratio in the allylic reagent.11c
Since none of the paths from the E-reagent to the anti product
have a low energy (Table 4), we regard the Curtin-Hammett
assumption as validated.
Conclusions
A new zinc-mediated addition reaction has been developed
for the synthesis of amino alcohols from imines and benzoy-
loxyallyl bromide. A combined Hammett and computational
study indicates that the addition takes place in homogeneous
phase between an initially formed organometallic allyl reagent
and the imine in a Zimmerman-Traxler transition state. The
main reactivity-determining factor in the addition is the elec-
trophilicity of the imine carbon, not the ability of the imine
nitrogen to coordinate to the metal. There is very good
agreement between the computed and observed selectivity in
the reaction.
Figure 5. Two best transition states leading to the minor syn products, the
boat-Z-prox and chair-E-dist isomers.
of ca. 8 kJ/mol and indicates that the minor syn-isomer
predominately arises from boat-Z-prox TS.
Acknowledgment. The Lundbeck Foundation is gratefully
acknowledged for financial support.
Finally, we want to note that a comparison of all potential
transition states is equivalent to implicitely postulating that the
system is under Curtin-Hammett conditions,28 that is, that
equilibration of all possible forms of the organometallic allyl
reagent is fast compared to reaction with the imine. This point
Supporting Information Available: Experimental procedures
and characterization of compounds; data from kinetic runs;
Cartesian coordinates and energies for all calculated structures.
This material is available free of charge via the Internet at
http://pubs.acs.org.
(28) Maskill, H. The Physical Basis of Organic Compounds; Oxford University
Press: Oxford, 1985; pp 293-295.
JA054706A
9
J. AM. CHEM. SOC. VOL. 127, NO. 45, 2005 15761