the diamine ligand would chelate the Mg atom and that the
phenyl and N-alkyl groups would be preferentially oriented
anti to each other on the resulting five-membered ring.
Approach of the ketone to this reagent from the less hindered
face could then take place with the CF group oriented axial
3
or equatorial in a six-membered ring transition state. Since
relatively large aryl groups would prefer to be equatorial,
similar to that shown in Scheme 2 leads to the conclusion
that ligands of (S)-configuration should give products of (R)-
configuration. In fact, reductions of 1-naphthyl trifluoro-
methyl ketone using COMAs derived from (S)-proline (e.g.,
4, Scheme 3) consistently gave rise to (R)-3b as the major
product.
3
the transition state with an axial CF group is energetically
more favorable and gives rise to the major product. On the
basis of this model, ligands of the (S)-configuration should
afford (S)-carbinol as the major product. This is what is
observed.
Scheme 3
The enhanced selectivity observed with N-benzyl ligand
1f compared to ligands with smaller or larger N-alkyl groups
can also be explained using this model if one invokes a
stabilizing effect on one of the possible transition states by
the benzyl group. Specifically, an Ar-H‚‚‚F-C interaction
3
may help to stabilize the transition state where the CF is
axial. Such interactions, although known to be very weak,
have been observed in the solid state.11 In our case, the
presence of such interactions helps to explain the differences
in selectivities observed for ligands 1f, 1g, and 1h. Thus,
using the model shown in Scheme 2, it would be expected
that replacement of the benzylic pro-(R) hydrogen in 1f (to
give 1g) should have little influence on selectivity but that
replacement of the pro-(S) hydrogen (to give 1h) should give
decreased selectivity since steric interactions would make it
difficult for the favorable conformation shown to be attained.
The experimental evidence, namely, that 1f and 1g give
essentially identical selectivities while 1h gives significantly
lower selectivity (Table 2, entries 6-8), is consistent with
these expectations.
While the arguments presented above explain the differ-
ences in enantioselectivities observed with different ligands
and trifluoroacetophenone, they do not explain why there
are differences between aryl groups. Perhaps with acyclic
ligands, the orientation of the N-alkyl group is always anti
and the transition states shown in Scheme 2 pertain. Thus,
larger aryl groups will show a stronger preference for
adopting an equatorial orientation and be more likely to show
high (S)-selectivity. With trifluoroacetophenone, the two
With trifluoroacetophenone, the major product formed
using ligands 1a-f (all of the (S)-configuration) was
sometimes R and sometimes S. In general, ligands with
3
groups (Ph and CF ) are relatively similar in size, and thus
the selectivity may be more sensitive to changes in ligand
structure. However, the changes in enantioselectivities
observed, particularly reversals, with different N-alkyl groups
on the ligand are difficult to accommodate with this model.
Thus, neither explanation is completely satisfactory.
2
smaller N-alkyl groups (e.g., Me, Et, PhCH ) were (R)-
selective, while ligands with larger groups (e.g., t-Bu, Ph)
were (S)-selective. This may be because ligands with smaller
N-alkyl groups can adopt conformations in which the N-alkyl
group and the phenyl group of the phenylglycine-derived
ligand are cis. Approach of the ketone to the reagent would
then be on the opposite face to that depicted in Scheme 2,
and the opposite (R)-enantiomer would predominate. Thus,
with trifluoroacetophenone, larger N-substituents were needed
to obtain good selectivities, and the best numbers were
observed with ligand 1e (R ) Ph, Table 2, entry 5).
Other aryl trifluoromethyl ketones were also reduced with
the i-Pr COMA derived from 1f (Table 3). In general,
selectivities were excellent, particularly with larger aryl
groups. Product 3e, derived from aryl ) 9-anthryl, is
especially interesting since it is a useful chiral auxiliary and
it is known that enrichment to high enantiomeric purity is
12
easily achieved by recrystallization. The selectivity ob-
The idea that the relative orientation of the N-alkyl group
and the phenyl (side-chain) group is important in determining
the absolute configuration of the product is supported by
results using derivatives of proline as ligands. With such
ligands, the “side-chain” and N-alkyl group are confined in
a five-membered ring and may be considered cis-oriented
in a COMA complex. An analysis of possible transition states
served with COMA i-Pr
with results reported for other established chiral reducing
agents: BINAL-H, 92-98% ee; DIP-Cl, 82% ee; CBS-
catecholborane, 94% ee.
When an alkynyl trifluoromethyl ketone was reduced with
2
i-Pr Mg/1f, only moderate selectivity was observed and the
major isomer was the (R)-enantiomer (Table 3, entry 6),
2
Mg/1f (96% ee) compares favorably
12
13
14
reflecting the relatively small size of an alkynyl group
(9) (a) Bradshaw, C. W.; Hummel, W.; Wong, C.-H.; J. Org. Chem.
1
5
992, 57, 1532-1536. (b) Kanth, J. V. B.; Brown, H. C. Tetrahedron 2002,
8, 1069-1074.
(12) (a) Chong, J. M.; Mar, E. K. J. Org. Chem. 1991, 56, 893-896.
(13) Ramachandran, P. V.; Teodorovic, A. V.; Brown, H. C. Tetrahedron
1993, 49, 1725-1738.
(10) Kuroki, Y.; Sakamaki, Y.; Iseki, K. Org. Lett. 2001, 3, 457-459.
11) (a) Shimoni, L.; Glusker, J. P. Struct. Chem. 1994, 5, 383-397.
(
(b) Parsch, J.; Engels, J. W. J. Am. Chem. Soc. 2002, 124, 5664-5672.
(14) Corey, E. J.; Bakshi, R. K. Tetrahedron Lett. 1990, 31, 611-614.
Org. Lett., Vol. 4, No. 23, 2002
4141