Heterogeneous enantioselective hydrogenation of ethyl pyruvate catalyzed by cinchona-modified Pt catalysts: Effect of modifier structure

Article abstract of DOI:10.1021/ja003259q

The effect of the structure of chiral modifiers derived from natural cinchona alkaloids on the enantioselectivity and rate of the Pt/Al₂O₃-catalyzed hydrogenation of ethyl pyruvate was investigated. The influence of the following structural elements was studied: The cinchonidine versus the cinchonine backbone; effect of the nature and the size of substituents attached to C₉; effect of partial hydrogenation of the quinoline ring; effects of changes of the substituent at the quinuclidine moiety. The strongest effects on ee and somewhat less on rate were observed for changes in the O-C₉-C₈-N part of the cinchona alkaloid and for partial or total hydrogenation of the quinoline rings. The nature of the substituents in the quinuclidine part had a comparably minor influence. The solvent was found to have a significant effect on enantioselectivity and rate. In acetic acid, the best results were obtained with O-methyl-10,11-diydrocinchonidine (ee's up to 93%), whereas dihydrocinchonidine was the most effective modifier in toluene. In agreement with a basic model proposed by Pfaltz, it was concluded that the minimal-requirements for an efficient modifier for the hydrogenation of α-keto esters is the presence of a basic nitrogen center close to one or more stereogenic centers and connected to an aromatic system. The results are in qualitative agreement with mechanistic models based on hydrogen-bonding interactions between an adsorbed modifier molecule and adsorbed ethyl pyruvate or its half-hydrogenated intermediate.

Full text of DOI:10.1021/ja003259q

J. Am. Chem. Soc. 2000, 122, 12675-12682
12675
Heterogeneous Enantioselective Hydrogenation of Ethyl Pyruvate
Catalyzed by Cinchona-Modified Pt Catalysts: Effect of Modifier
Structure
H. U. Blaser,* H. P. Jalett, W. Lottenbach, and M. Studer
Contribution from the SolVias AG, P.O. Box, CH-4002 Basel, Switzerland
ReceiVed September 5, 2000
Abstract: The effect of the structure of chiral modifiers derived from natural cinchona alkaloids on the
enantioselectivity and rate of the Pt/Al2O3-catalyzed hydrogenation of ethyl pyruvate was investigated. The
influence of the following structural elements was studied: the cinchonidine versus the cinchonine backbone;
effect of the nature and the size of substituents attached to C9; effect of partial hydrogenation of the quinoline
ring; effects of changes of the substituent at the quinuclidine moiety. The strongest effects on ee and somewhat
less on rate were observed for changes in the O-C9-C8-N part of the cinchona alkaloid and for partial or
total hydrogenation of the quinoline rings. The nature of the substituents in the quinuclidine part had a
comparably minor influence. The solvent was found to have a significant effect on enantioselectivity and rate.
In acetic acid, the best results were obtained with O-methyl-10,11-diydrocinchonidine (ee’s up to 93%), whereas
dihydrocinchonidine was the most effective modifier in toluene. In agreement with a basic model proposed by
Pfaltz, it was concluded that the minimal requirements for an efficient modifier for the hydrogenation of R-keto
esters is the presence of a basic nitrogen center close to one or more stereogenic centers and connected to an
aromatic system. The results are in qualitative agreement with mechanistic models based on hydrogen-bonding
interactions between an adsorbed modifier molecule and adsorbed ethyl pyruvate or its half-hydrogenated
intermediate.
Introduction
Empirical structure-activity and selectivity correlations are
an important tool for developing enantioselective catalysts. This
is even more pronounced for heterogeneous systems because
of a serious lack of understanding of the mode of action of such
catalysts. Especially the nature of the chirality transfer from
the chiral modifier or ligand to the chiral product is not really
understood. In this contribution, we present results concerning
the effect of modifier structure for the enantioselective hydro-
genation of ethyl pyruvate (etpy) to ethyl lactate (etlac) using
Pt-alumina catalysts in the presence of cinchona-type modifiers
(Orito reaction; Figure 1).1 A preliminary report has appeared.2
This study should contribute to the understanding of the
enantioselective hydrogenation of functionalized carbonyl groups
with modified heterogeneous catalysts, a topic of interest from
both a preparative and a mechanistic point of view.3-6
Figure 1. Test reaction.
cinchonidine (Cd)/quinine (Qn) and cinchonine (Cn)/quinidine
(Qn) (see Figure 2) that served as starting points for preparing
derivatives thereof (see Figure 3). Roughly, one can distinguish
three parts where changes seem relatively easy: (i) the sub-
stituent at C9, (ii) the quinoline moiety, and (iii) the quinuclidine
moiety. It must be pointed out that some of these transformation
are rather difficult to carry out experimentally and very often
the resulting product mixtures had to be separated by HPLC
(see Experimental Section). Here we very briefly describe the
synthesis of the most important derivatives tested in this study
and discuss the evidence for our structural assignments.
Results
Two reaction types were employed for changing the sub-
stituent at C9: O-methylation or acylation and substitution of
OH. Methylation with MeI had to be carried out in the presence
of NaH in order to get good yields of MeO-HCd. Acylation
with acid chlorides or related derivatives were usually straight-
forward; O-substituted derivatives with large groups were
originally developed for the asymmetric dihydroxylation.7 We
assume that the substitution reactions to Cl-HCd, Br-HCd, and
probably also for F-HCd occur with inversion of configuration
as shown for 9-bromoquinine by Braje et al.8
Synthesis of Cinchona Derivatives: Structure and Purity.
Nature has provided us with the two diastereomeric families of
* Corresponding author: (e-mail) hans-ulrich.blaser@solvias.com; (tel)
++41 61 686 61 55; (fax) ++41 61 686 63 11.
(1) Orito, Y.; Imai, S.; Niwa, S. J. Chem. Soc. Jpn. 1982, 137 and
references cited therein.
(2) Blaser, H. U.; Jalett, H. P.; Monti, D. M.; Baiker, A.; Wehrli, J. T.
Stud. Surf. Sci. Catal. 1991, 76, 147.
(3) Blaser, H. U. Tetrahedron: Asymmetry 1991, 2, 843.
(4) For a recent comprehensive review of the enantioselective hydrogena-
tion of R-keto esters using cinchona-modified platinum catalysts and related
systems, see: Blaser, H. U.; Jalett, H. P.; Mu¨ller, M.; Studer M. Catal.
Today 1997, 37, 441.
(7) Berrisford, D. J.; Bolm, C.; Sharpless, K. B. Angew. Chem., Int. Ed.
Engl. 1995, 34, 1059.
(5) Pfaltz, A.; Heinz, T. Top. Catal. 1997, 4, 229.
(6) Baiker, A. J. Mol. Catal. A: Chem. 1997, 115, 473.
(8) Braje, W. M.; Wartchow, R.; Hoffmann, H. M. R. Angew. Chem.,
Int. Ed. Engl. 1999, 38, 2540.
10.1021/ja003259q CCC: $19.00 © 2000 American Chemical Society
Published on Web 12/02/2000

12676 J. Am. Chem. Soc., Vol. 122, No. 51, 2000
Blaser et al.
Figure 2. Structures, numbering, and abbreviations for the parent cinchona alkaloids.
Table 1. Reaction Conditions for the Various Test Series
prel series2 series 1 series 2 series 3 series 4
catalyst type
catalyst wt, mg
etpy vol, mL
solvent vol, mL 20
modifier wt, mg 10
4759
100
10
5R94
50
5
5R94
50
5
5R94
50
5
4759
50
5
20
20
25
10
1 and 10 1 and 10 1 and 10 1 and 10
temp, °C
pressure, bar
25-30
20
20
100
20
100
20-30
70-100
70-100 100
Figure 3. Derivatization of cinchonidine (for abbreviations, see Figures
4 and 5).
reason, the results (especially the rate data) of some series are
not directly comparable. However, the major effects discussed
below are significant and not affected by these variations.
Hydrogenation Results: Enantioselectivities and Rates
Using Different Modifiers. (1) Comparison of the Parent
Cinchona Modifiers. As already described by Orito,1 the results
shown in Table 2 confirmed that cinchonidine and quinine
always give preferentially ethyl (R)-lactate while the pseudo-
enantiomeric cinchonine and quinidine led to an excess of S
product. The same holds true for the partially hydrogenated
derivatives HCd and HCn. This means that the absolute
configuration at C8 and/or C9 controls the sense of induction.
Nevertheless, the relative position of the group R to the second
quinuclidine substituent and the nature of Z sometimes had a
strong effect on both the magnitude of the enantiomeric excess
and the reaction rate: The cinchonidine and quinine families
induced higher ee’s than the cinchonine and quinidine families,
respectively. The presence of a methoxy group in the quinoline
ring (Cd vs Qn, Cn vs Qd) led to a lower enantioselectivity.
The 10,11-dihydro derivatives in the Cd series gave slightly
higher ee’s, whereas the effect for the Cn series was ambiguous.
Bu¨rgi et al.10 reported that the introduction of a phenyl group
in the 2-position of the quinoline ring of 9-deoxycinchonidine
also led to a decrease in ee of ∼20%.
The quinoline moiety was partially hydrogenated using two
different catalysts: Raney nickel for the preferential hydrogena-
tion of the heteroaromatic ring to the two epimers of py-HCd
and platinum oxide for reducing the carbocyclic aromatic ring
to cyc-HCd as well as to the saturated DHCd isomers in the
presence of trifluoroacetic acid. In all cases, complex product
mixtures were obtained and purification was carried out using
chromatographic methods. For py-HCd, two epimers (called
A/B-py-HCd) are expected and were indeed observed in a ratio
of ∼2:3. While we were able to separate the two epimers, we
were not able to make a clear assignment of the absolute
configuration at C4′ on the basis of the available data. Cyc-
HCd exists only as one isomer; the eight isomers possible for
DHCd were not separated. The Vinyl group at C3 of the
quinuclidine moiety was selectively hydrogenated in high yields
in the presence of Pd/C to give HCd as an important intermediate
for many other derivatives. Ozonization followed by reductive
workup lead to norcinchol, one of the few water-soluble
cinchona derivatives. In all cases, structures were confirmed
by NMR.
Screening Strategy and Quality of Results. A preliminary
screening in ethanol indicated that several structural factors
influence the effect of cinchona modifiers for the Pt-catalyzed
hydrogenation of R-keto esters.2 Because the enantioselectivity
is also significantly affected both by the solvent type and by
the modifier concentration,4,9 a more elaborate procedure was
applied and the tests were carried out in toluene (1 and 10 mg
of modifier), in acetic acid (1 mg of modifier), and in ethanol
(10 mg of modifier), in some cases with additional experiments.
This choice was based on the finding that optimal concentrations
usually were very low in AcOH, medium in toluene, and high
in EtOH. Furthermore, because acceleration is an important
characteristic of the Pt cinchona catalysts,4 not only the ee’s
but also the initial rates were determined for several test series.
The results reported below were obtained over the course of
several years. The conditions, the ethyl pyruvate, and catalyst
lots as well as the autoclaves differed slightly for the four
different series of experiments (see Table 1). In addition, two
different commercially available 5% Pt/Al2O3 catalyst types
were used that in our hands always gave high ee’s. For this
The solvent had a significant effect on the magnitude of the
enantioselectivity and on the rate of the modified catalytic
system. Generally, acetic acid gave the best ee values but in
most cases lower rates than ethanol or toluene. The modifier
concentration affected rates more strongly than ee’s. As a trend,
in EtOH, 10 mg of modifier often gave better enantioselectivi-
ties; in AcOH and toluene, addition of 1 mg of the modifier
was usually superior.
The rates for the unmodified systems varied between 2 and
5 mmol/min‚g, depending on the solvent and the series. As
described before,4 all modifiers had a rate-accelerating effect,
but with obvious differences for the four families. Generally,
higher ee’s correlated weakly with higher rates, but more
interesting is the fact that the presence of a ring methoxy group
affected the rates more than the ee’s.
(2) Effect of the Substituents X and Y at C9 (Figure 4). In
this series of experiments, we can distinguish two cases. In the
first, the OH group at C9 of HCd was replaced with relatively
(10) Bu¨rgi, T.; Zhou, Z.; Ku¨nzle, N.; Mallat, T.; Pfaltz, A.; Baiker, A.
(9) Blaser, H. U.; Garland, M.; Jalett, H. P. J. Catal. 1993, 144, 569.
J. Catal. 1999, 183, 405.

EnantioselectiVe Hydrogenation of Ethyl PyruVate
J. Am. Chem. Soc., Vol. 122, No. 51, 2000 12677
Table 2. ee (%), Rate (mmol/min‚g), and Absolute Configuration of etlac for the Hydrogenation of Ethyl Pyruvate Catalyzed by Pt Catalysts
Modified with Cinchona Alkaloidsa
AcOH
EtOH
toluene
1 mg
10 mg
rate
1 mg
10 mg
1 mg
10 mg
rate
modif
ee
rate
ee
ee
rate
ee
rate
ee
rate
ee
ac
ser
Cd
HCd
Cn
HCn
Qn
Qd
88
90
81
63
81
56
110
120
40
70
7
89
90
78
83
160
160
60
71
67
52
43
220
170
90
82b
72b
42
40
45
70b
50b
80
90
19
85
86
70
67
73
51
170
220
90
120
12
81
83
67
69
73
46
90
120
30
60
10
5
(R)
(R)
(S)
(S)
(R)
(S)
2
2
2
2
3
3
90
110
3
27
14
5
a For conditions, see Experimental Section; for abbreviations, see Figure 2. b Series 1.
Figure 4. Structures and abbreviations of cinchona modifiers with different substituents at C9.
Table 3. HCd Derivatives with Small- and Medium-Sized
Substituents at C9: Effect on ee (%), Rate (mmol/min‚g), and
Absolute Configuration of etlaca
QD, respectively. Bartok et al. also described the effect of using
iso-Cn and iso-Qd, where a ring is formed between C9-O and
C10.12,13 Under our test conditions, a strong decrease in ee was
observed for all cases except MeO-HCd and AcO-HCd. An
additional stereogenic center in the substituent did not affect
the enantioselectivity: both (R)- and (S)-Lac-HCd give the same
ee’s. The corresponding benzyl ethers (R)- and (S)-Bnlac-HCd
also showed the same behavior but with much lower ee’s. As
a rule, oxygen-containing substituents gave better performances
and more bulky substituents led to significantly lower enantio-
selectivities. The effect on the rates was similar as discussed
above: without significant exceptions, modifiers with low ee’s
also had low rates.
Ligands developed by Sharpless for the asymmetric dihy-
droxylation of olefins carrying very large substituents at C9
(Table 4) showed distinctly different behavior. In many cases,
the major enantiomer was opposite to what was expected from
the absolute configuration of the cinchona backbone. This
reversal was more pronounced for HQd than for HQn derivatives
and, in contrast to all other modifiers, not only the magnitude
of the ee values but also the sense of induction showed a strong
solvent dependence. The largest differences were found for Phn-
HQn (40% ee in AcOH vs 4% in EtOH or toluene) and for
Meq-HQd (13% (S) in AcOH vs 18% (R) in toluene). A similar
observation was recently reported by Collier et al.14 for the
toluene
AcOH
1 mg
EtOH
10 mg
1 mg
10 mg
modifier
ee rate ee rate ee rate ee rate ac ser
67 30 54 90 63 40 81 90 (R)
22 (R) prel
44 19 43 50 45 30 62 38 (R)
46 23 44 50 43 34 65 25 (R)
AcO-HCd
PivO-HCd
(R)-Lac-HCd
(S)-Lac-HCd
(R)-Bnlac-HCd
(S)-Bnlac-HCd
ClbzO-HQn
ClbzO-HQd
Cl-HCd
2
4
4
4
4
3
3
2
10
12
15
0
9
5
7
20 14 15
24 12 21
4
6
3
(R)
(R)
(R)
24
13 10
1
13
2
4
10
22
0
30
10 (S)
44 30 37 30 23 10 16 30 (R)
Cl-HCd
F-HCd
Br-HCd
44 nab
(R) prel
(R)
(R) prel
10 10 13
5
30
5
24
1
2
20 na
MeO-HCd
NH2-HCd
H-HCd
93 190 70 120 70 70 67 60 (R)
10 10 10 21 (R)
2
2
8
5
4
1
44 na
43 na
(R) prel
(R) prel
keto-HCd
a For conditions, see Experimental Section; for abbreviations, see
Figure 4. b na, not available.
small substituents (Table 3); the second group of modifiers
consists of ligands developed by Sharpless for the dihydroxyl-
ation reaction7 with large substituents at C9 (Table 4).
(11) Bartok, M.; Felfo¨ldi, K.; Szo¨llo¨si, G.; Bartok, T. React. Kinet. Catal.
Lett. 1999, 68, 371.
(12) Bartok, M.; Felfo¨ldi, K.; Szo¨llo¨si, G.; Bartok, T. Catal. Lett. 1999,
61, 1.
(13) Bartok, M.; Felfo¨ldi, K.; To¨ro¨k, B.; Bartok, T. Chem. Commun.
1998, 2605.
(14) Collier, P. J.; Hall, T. J.; Iggo, J. A.; Johnston, P.; Slipszenko, J.
A.; Wells, P. B.; Whyman, R. Chem. Commun. 1998, 1451.
As shown in Table 3 and described in our preliminary report,2
replacing the X or Y groups of HCd by other small substituents
did not change the sense of induction but significantly affected
the ee values. A recent publication described similar observa-
tions: For 9-deoxy-Cd, Bu¨rgi et al.10 reported an ee of 57%,
Bartok et al.11 obtained ee’s of 42 and 22% for epi-Qn and epi-

12678 J. Am. Chem. Soc., Vol. 122, No. 51, 2000
Blaser et al.
Table 4. HCd Derivatives with Large Substituents at C9: Effect on ee (%), Rate (mmol/min‚g), and Absolute Configurationa
toluene
AcOH
1 mg
EtOH
10 mg
1 mg
10 mg
modifier
acb
ee
rate
ee
rate
ee
rate
ee
rate
ser
Phal-(HQn)2
Phal-(HQd)2
Pyr-(HQn)2
Pyr-(HQd)2
Phn-HQn
Phn-HQd
Meq-HQn
Meq-HQd
(R)
(S)
(R)
(S)
(R)
(S)
(R)
(S)
16 (R)
4 (S)
6 (S)
6 (R)
40 (R)
4 (S)
28 (R)
13 (S)
26
22
20
28
3
1
2
7 (R)
26
23
26
24
7
9
7
10 (R)
1 (R)
5 (R)
6 (R)
4 (R)
2 (R)
4 (R)
11 (R)
15
30
27
29
4
5
4
16 (R)
29
13
11
32
2
3
1
4
4
4
4
3
3
3
3
4 (R)b
2 (S)
3 (R)
4 (R)
9 (R)
4 (R)
2 (R)
2 (R)
18 (R)
7 (R)
5 (R)
18 (R)
5 (R)
11 (R)
2
7
4
1
a For conditions, see Experimental Section; for abbreviations, see Figure 4. b Absolute configuration expected on the basis of the parent cinchona
c
alkaloid. Boldface type indicates reversed absolute configuration.
Table 5. HCd Derivatives with a Modified or Lacking Quinoline
Moiety: Effect on ee (%), Rate (mmol/min‚g), and Absolute
Configuration of etlaca
toluene
AcOH
1 mg
EtOH
10 mg
1 mg
10 mg
modifier
HCd
A-py-HCd
B-py-HCd
cyc-HCd
DHCd
M11
quinuclidine A
quinuclidine B
ee rate ee rate ee rate ee rate ac ser
90 120 72b 50b 86 220 83 120 (R)
2
1
1
1
1
4
4
4
82 10 35 10 50
47 10 12 10 48
4
4
5
52
57
7
4
4
5
4
(R)
(R)
(R)
(R)
(R)
Figure 6. Structures of cinchona modifiers with different substituents
at the quinuclidine part.
20 10
5
2
20
10
1
5
6
25
5
3
3
nac
Table 6. Changes at the Quinuclidine: Effect on ee (%), Rate
(mmol/min‚g), and Absolute Configuration of etlaca
2
1
na
na
11 na (R)
na (R)
5
toluene
AcOH
1 mg
EtOH
10 mg
a For conditions, see Experimental Section; for abbreviations, see
1 mg
10 mg
Figure 5. b Series 1. c na, not available.
modifier
ee rate ee rate ee rate ee rate ac ser
di-Br-Cd
60 nab
40 na
72 na
(R) prel
(R) prel
(R) prel
10-AcO-Cd
10-ald-Cd
norcinchol
N-Bn-Cd-Cl
91 37 60 26 75 36 76 44 (R)
2
1
na (R) prel
a For conditions, see Experimental Section; for abbreviations, see
Figure 6. b na, not available.
substitution at the 2′-position of the quinoline ring has only a
minor effect on ee: 2′-phenyl-9-desoxy-HCd gave 48% ee as
compared to 57% for 9-deoxy-Cd.
Figure 5. Structures of cinchona modifiers with replaced quinoline
moiety.
(4) Effect of Changes at the Quinuclidine Moiety. The
limited number of examples (see Figure 6 and Table 6) indicate
that the nature of the R substituent at C3 does not have an
important effect on ee as long as arrangement of the substituents
of the quinuclidine remains the same (as opposed to the
quinidine/cinchonine series). When the quinuclidine nitrogen
was alkylated, the modifier was no longer effective.
(5) Detailed Investigations of the Partially Hydrogenated
HCd Derivative. Two aspects were investigated in more detail
because we became aware of their importance in the course of
our project: (i) the variation of the enantioselectivity with
reaction time (or conversion); (ii) the effect of the modifier
concentration on ee and reaction rate.
The dependence of ee on the conversion was investigated
for three partially hydrogenated modifiers. Several studies have
shown that the enantioselectivity is often not constant during
the course of the reaction.4 In line with these observations, we
noted significant ee increases at the beginning of the reaction
for all three modifiers. After the initial increase, the enantio-
selectivity remained rather constant for A- and B-py-HCd around
80-84 and 60-64%, respectively, whereas it decreased strongly
during the very slow reaction in the presence of cyc-HCd from
∼40% to below 20%. While the reason for the initial increase
hydrogenation of etpy using Cd-modified Pd catalysts in
different solvents. However, one has to keep in mind that rather
low ee’s are observed and that only small changes of activation
energies can lead to a reversal of induction. Interestingly, the
dimeric ligands gave significantly higher rates even though the
enantioselectivities were comparable to the monomeric modi-
fiers.
(3) Effect of Change of the Quinoline Moiety. Several
potential hydrogenation products of HCd were isolated and
tested as modifiers (see Table 5 and Figure 5). With the
exception of the A-py-HCd isomer in AcOH, all ee’s were
considerably lower than for HCd, as also observed by Sun.15
As already pointed out, at the moment we cannot tell which
absolute configuration at the 4′-position of py-HCd is more
favorable. The loss in enantioselectivity was more pronounced
when the homoaromatic ring was hydrogenated (cyc-HCd) or
when a planar aromatic group was lacking completely. It is
noteworthy, that both quinuclidines A and B led to an excess
of ethyl (R)-lactate, albeit in low ee. All these modifiers showed
a very weak rate acceleration. Bu¨rgi et al.10 showed that
(15) Sun,Y. K. Merck Research Laboratories, personal communication.

EnantioselectiVe Hydrogenation of Ethyl PyruVate
J. Am. Chem. Soc., Vol. 122, No. 51, 2000 12679
Table 7. Optimal Concentration, Maximum ee, and Acceleration
eters have been gained. In the following paragraphs, first we
will describe the mechanistic models proposed so far, then
summarize the effects observed for the various structural
variations, and finally try to draw conclusions from the effects
described above and integrate these into a slowly emerging
general picture on the mechanism of enantioselection by
cinchona-modified Pt catalysts.
Factor for Different Modifiersa
ee (%) at
opt concn
(conversion)
opt concn,
acceleration
factor
ee
modifier
µM
(screening)
HCd
0.08
8
9
92 (79)
74 (37)
65 (43)
47 (28)
18
6
10
4
90
82
47
20
A-py-HCd
B-py-HCd
cyc-HCd
Several groups have developed structural models to explain
the mode of action of the cinchona-modified Pt catalysts or
analogues thereof.4-6,23-25 With the exception of the template
model of Wells that has been withdrawn,23 all other models
postulate specific interactions between one cinchona modifier
and one molecule of etpy. Augustine25 explained the observed
enantiodiscrimination by a two-point interaction between the
adsorbed modifier and adsorbed etpy (N1 lone pair with keto
group and lone pair of the oxygen at C9 with ester group).
Margitfalvi’s “shielding effect” model24 is based on a two-point
interaction between modifier and etpy in solution (interaction
N1 with keto group and a π-π stacking interaction quinoline
with ester group). The shielded etpy is thought to interact with
the Pt surface and hydrogen would be added to etpy from the
unhindered side of this adduct. Baiker6 and Pfaltz5 as well as
Wells23 have rationalized the results observed for cinchona
alkaloids and also for some analogues thereof by assuming a
strong interaction between adsorbed etpy (or a half-hydrogenated
intermediate) and the adsorbed modifier on the Pt surface via
hydrogen bridging as depicted in Figure 7. Molecular modeling
studies have been carried out for several of these mechanistic
models in order to confirm the feasibility of the proposed
interactions. From modeling results and NMR experiments27 and
from the effect of restricting rotation around the C8-C9 axis
using isocinchona derivatives,12,13 it was inferred that an open
conformation of the adsorbed modifier must be the optimal
active species (see Figure 7a). A similar situation was found
for aryl ethylamine modifiers (Figure 7b). In contrast, Margit-
falvy et al.24 proposed the interaction between etpy and a closed
conformation (where the lone pair of N1 points to the quinoline
ring) to be decisive. The calculated results were claimed to be
in agreement with both the “shielding effect” model24 and the
hydrogen bridge model!26-28
34
a Conditions: AcOH, JMC94, room temperature, and 100 bar.
is still under debate,16 the decrease at high reaction times is
probably due to further hydrogenation to the saturated DH-
Cd.4,15,17
To obtain more information concerning modifier adsorption
behavior and ligand acceleration, the effect of the modifier
concentration on ee and reaction rate was investigated in some
detail for three partially hydrogenated modifiers in acetic acid
as best solvent.9,18 In all cases, a maximum for both rate and ee
was observed similar to what was observed for several other
modifiers; i.e., very high modifier concentrations are detrimen-
tal.4 In Table 7 we have tabulated the optimal modifier
concentration, the corresponding ee and acceleration factor
(maximum/unmodified rate), and the ee values found in the
screening for HCd, cyc-HCd, A-py-HCd, and B-py-HCd. Two
facts are noteworthy: First, the ee’s observed in these detailed
investigations agree well with those found in the screening
series. This means that the differences in ee for different
modifiers are real and not due to a wrong choice of modifier
concentration or of the conversions chosen for screening.
Second, the best ee’s and rates are observed at much lower
concentrations for HCd than for the three partially hydrogenated
modifiers; for the two py-HCd’s it is ∼100 and for cyc-HCd
∼400 times higher, respectively, than for HCd. This indicates
a much weaker adsorption of the partially hydrogenated
cinchona derivatives.
Discussion
Since the publication of our preliminary results on the effect
of the modifier structure on the catalyst performance for the
enantioselective hydrogenation of etpy,2 several groups have
published on this topic.5,6,10-13,17,19-21 Of special significance
are a number of rather efficient chiral amines prepared by Pfaltz
and Baiker5,6,21 that can be considered to be simple cinchona
models. Recently, Bartok et al.12,13 tested cinchonine and
quinidine derivatives with conformational restrictions and
Baiker’s group10 described a 2-phenylcinchonidine derivative.
In addition, vinca-type alkaloids20 and Tro¨ger’s base22 were also
found to induce moderate enantioselectivity whereas other
alkaloids were hardly effective.19 Even though these efforts have
not led to more selective modifiers compared to the cinchona
alkaloids, important insights into the decisive structural param-
Tables 8 and 9 present a summary of all pertinent results
obtained for structural variations of cinchona-type modifiers (see
Figure 8) for the hydrogenation of etpy with Pt catalysts. We
have grouped the results according to the simplified model of
Pfaltz,5 who distinguished two structural elements: an (ex-
tended) aromatic system (Ar) that is proposed to serve as
anchoring group on the catalyst surface and a basic amino
function (N-unit) that is thought to interact with the keto group
of the substrate. Table 8 summarizes the effects of various
combinations of the Ar and N-unit and Table 9 of changing the
substituents X and Y at C9.
From the results summarized in Tables 8 and 9 and Figure
9, it is evident that an (extended) aromatic system and a basic
(16) For a controversial discussion, see: Blackmond, D. J. Catal. 1998,
176, 267. Mallat, T.; Baiker, A. J. Catal. 1998, 176, 271.
(17) LeBlond, C.; Wang, J.; Liu, J.; Andrews, A. T.; Sun, Y. K. J. Am.
Chem. Soc. 1999, 121, 4920.
(18) Blaser, H. U.; Imhof, D.; Studer, M. Stud. Surf. Sci. Catal. 1997,
108, 175.
(19) Griffiths, S. P.; Johnston, P.; Vermeer, W. A. H.; Wells, P. B. Chem.
Commun. 1994, 2431.
(20) Farkas, G.; Fodor, K.; Tungler, A.; Mathe, T.; Toth, G.; Sheldon,
R. A. J. Mol. Catal. A: Chem. 1999, 138, 123 and references therein.
(21) Schu¨rch, M.; Heinz, T.; Aeschimann, R.; Mallat, T.; Pfaltz, A.;
Baiker, A. J. Catal. 1998, 173, 187.
(23) Simons, K. E.; Meheux, P. A.; Griffiths, S. P.; Sutherland, I. M.;
Johnston, P.; Wells, P. B.; Carley, A. F.; Rajumon, M. K.; Roberts, M. W.;
Ibbotson, A. Recl. TraV. Chim. Pays-Bas 1994, 113, 465.
(24) Margitfalvy, J. L.; Tfirst E. J. Mol. Catal. A: Chem. 1999, 139, 81
and references cited therein.
(25) Augustine, R. L.; Tanielyan, S. K.; Doyle, L. K. Tetrahedron:
Asymmetry 1993, 4, 1803.
(26) Bu¨rgi, T.; Baiker A. J. Am. Chem. Soc. 1998, 120, 12920 and
references cited therein.
(27) Wells, P. B.; Simons, K. E.; Slipszenko, J. A.; Griffin, S. P.; Ewing,
D. F. J. Mol. Catal. A: Chem. 1999, 146, 159 and references cited therein.
(28) Zuo, X.; Liu, H.; Guo, D.; Yang, X. Tetrahedron 1999, 55, 7787.
(22) Minder, B.; Schu¨rch, M.; Mallat, T.; Baiker, A. Catal. Lett. 1995,
31, 143.

12680 J. Am. Chem. Soc., Vol. 122, No. 51, 2000
Blaser et al.
Figure 7. Artist’s view of the structure of proposed adsorbed adduct complexes between modifier and etpy or its reaction intermediates: (a) HCd
in the open conformation with half-hydrogenated etpy;23 (b) protonated modifier (aryl ethylamine type) and etpy.5,6
Table 8. Best ee’s for Various Combinations of Aromatic Groups
and Amino Functions
Figure 8. Basic structure of effective cinchona-type modifiers and
analogues.
(ii) The presence of an N-atom in the aromatic ring is no
prerequisite; homo- and heterocyclic derivatives are of similar
effectiveness.
(iii) Modifiers with a simple benzene or pyridine ring show
no chiral induction. The minimum requirement seems to be an
aminobenzene moiety (see, e.g., Tro¨ger’s base or the nitrophenyl
derivative in Table 8, which under the reaction conditions is
reduced to 1-(4-aminophenyl)ethylamine)).
(iv) Whereas the substitution of small aromatic systems is
favorable, the addition of a methoxy or a phenyl group to the
quinoline has a slightly negative effect.
(v) If the aromatic system are not flat or are absent, only
very low ee’s can be obtained.
These trends as well as the results summarized in Table 7
concerning the optimal modifier concentration are in overall
agreement with the notion that strong adsorption on the Pt
surface is critical and that the aromatic rings adsorb via its
π-system and not via the quinoline or quinuclidine nitrogen
atoms. This is in agreement with deuteration experiments as
well as a recent NEXAFS study.29 Nevertheless, the results
summarized in Table 7 clearly show that the strength of
adsorption can by no means explain the differences in chiral
induction; otherwise, low ee’s could be improved simply by
increasing the modifier concentration and we and others have
shown that ee’s do not increase above a certain modifier specific
limit.4
Table 9. Effect on ee: Variation of the Substituents at C9
Therefore, the precise structure of the chiral unit and the
positioning of the basic nitrogen atom relative to the aromatic
moiety determines the intrinsic ee and acceleration factor. The
necessity for a basic nitrogen atom is inferred from the fact
that alkylation of N1 in cinchonidine or replacement of the amino
group by a hydroxy group in the amino alcohol modifiers5 leads
to racemic ethyl lactate. The three most effective modifier
classes as depicted in Figure 8 demonstrate that the quinuclidine
subunit obviously is very well suited as are the closely related
amino alcohols and the more simple aryl ethylamines. However,
there are remarkable differences: In the case of the cinchona
derivatives and the aryl ethylamines, the absolute configuration
of the major ethyl lactate enantiomer is determined by the
nitrogen center close to a stereogenic center is a necessary but
not sufficient prerequisite for an effective modifier. For the
aromatic moiety, the following trends can be identified:
(i) Larger aromatic systems generally give better ee’s than
smaller ones of the same type.
(29) Evans, T.; Woodhead, A. P.; Gutierrez-Sosa, A.; Thornton, G.; Hall,
T. J.; Davis, A. A.; Young, N. A.; Wells, P. B.; Oldman, R. J.; Plashkevych,
O.; Vahtras, O.; Agren, H.; Carraretta, V. Surf. Sci. 1999, 436, L691.

EnantioselectiVe Hydrogenation of Ethyl PyruVate
J. Am. Chem. Soc., Vol. 122, No. 51, 2000 12681
Figure 9. Structures of moderately selective modifiers.
absolute configuration at C8 and C1, respectively, i.e. the
stereogenic center R to the N atom. For the amino alcohols, the
sense of induction is controlled by the absolute configuration
of the stereogenic center R to the O atom, in contrast to the
cinchona derivatives where (except for very large groups) the
substituents at C9 affect only the magnitude but not the sense
of induction. Nevertheless, it is possible to rationalize the
effectiveness of these three classes with the hydrogen-bridging
model.
The effects of changing X and Y at C9 are summarized in
Table 9. Obviously, the presence of large substituents at the
oxygen atom leads to a strong decrease in enantioselectivity
and in some cases even to a reversal of the absolute configu-
ration of the major product enantiomer. If one accepts the open
conformation depicted in Figure 7a to be important, the effect
of replacing OH by a medium-sized group like acetate or lactate
can be explained by hindered adsorption and/or by deviation
from the optimal conformations by rotation around the C4′ -C9
bond due to steric effects. However, this simple picture does
not explain the significant loss in enantioselectivity for some
epi or deoxy derivatives where X and Y are about of the same
size. For the case of the Sharpless modifiers, where we
sometimes observe a reversal of the absolute configuration of
the product, one can speculate that the large aromatic substit-
uents at C9 compete with the quinoline ring for adsorption on
the Pt surface leading to a different adsorption complex. This
might explain the significant solvent effects on the absolute
configuration of the major enantiomer. However, one should
be very careful not to overinterpret effects that are the result of
a difference in activation energy of the order of only a few
hundred calories per mole.
As pointed out above, from a structural point of view, our
results do not allow us to discriminate rigorously between these
models. We think that the mechanistic picture of Pfaltz, Baiker,
and Wells, assuming adsorbed adducts as depicted in Figure 7,
can explain more known facts of this complicated reaction than
any of the other proposals. The finding of Pfaltz and Baiker
that the O-C9-C8-N fragment present in cinchona modifiers
is no prerequisite for obtaining good ee values is not in accord
with Augustine’s two-point interaction model. The following
points speak against the proposal of Margitfalvi: (i) the negative
effect of a methoxy substituent at the quinoline ring (electron-
donating groups should give a stronger π-π stacking interac-
tion) and (ii) the fact that very low modifier concentrations are
needed to get the highest ee’s and rates (for a 1:1 interaction in
solution, the best results would be expected at much higher
modifier concentrations). Nevertheless, it must be stressed that
none of the preferred models is able to explain all of the many
features and especially that it is not possible with any of the
models to single out specific interactions that are directly
responsible for enantiodiscrimination. This means that at the
moment it is not possible to predict in any meaningful way the
effect of structural variations of either the substrate or the
modifier on enantioselectivity. In other words, finding improved
catalytic systems is still a matter of educated guesses and much
trial and error!
Conclusions
From our detailed study of a variety of different cinchona
derivatives, we find that the best results are obtained with
cinchonidine or slightly changed Cd derivatives such as HCd,
MeO-HCd, or norcinchol. With few exceptions, modifiers
derived from cinchonidine lead to an excess of (R)-ethyl lactate
whereas cinchonine derivatives preferentially lead to the S
enantiomer. Three structural elements in the cinchona molecule
can be identified where variations affect rate and ee of the
enantioselective hydrogenation of ethyl pyruvate (or more
generally R-keto acid derivatives): (i) an aromatic moiety Q,
(ii) the substitution pattern of the quinuclidine, and (iii) the
substituents X and Y at C9.
Aromatic Moiety. For cinchona-based modifiers, a 4′-
substituted quinoline nucleus is optimal; for other chiral
modifiers, a larger aromatic ring system is of advantage. With
one exception (py-HCd), modifiers with partially hydrogenated
systems have much lower ee’s between 0 and 50% and low
acceleration ratios as well. Substituting the quinoline in the 6′-
or 4′-position leads to somewhat lower enantioselectivities.
Substitution Pattern of the Quinuclidine. The absolute
configuration at C8 of the quinuclidine controls the absolute
configuration of the major product enantiomer (exceptions are
large substituents at C9). The nature of the substituent at C3
and the relative configuration at C3 and C8 (cinchonidine relative
to cinchonine series) have a moderate but significant effect on
ee. If N1 of the quinuclidine moiety is alkylated, optical
induction is lost completely.
Substituents X and Y at C9. The presence of an OH or OMe
group at C9 is optimal. Larger substituents on the O-atom or
replacing OH with other substituents leads to lower enantio-
selectivities.
The choice of the solvent has a significant effect on
enantioselectivity and rate. In acetic acid, best results were
obtained with O-methyl-10,11-diydrocinchonidine (ee’s up to
93%), whereas dihydrocinchonidine was the most effective
modifier in toluene. In EtOH, ee values were generally lower
but rates often higher than in AcOH.
Our results are in qualitative agreement with the closely
related mechanistic models proposed by Pfaltz, Baiker, and
Wells, namely that a successful modifier needs two structural
elements: first, an (extended) aromatic system, which is able
to form a strong adsorption complex with the Pt surface; second,
a suitable chiral amino function able to interact with the keto
group of the adsorbed etpy or its half-hydrogenated intermediate
via a hydrogen bridge. This activated complex is thought to be
responsible both for enantiocontrol and for rate acceleration.

12682 J. Am. Chem. Soc., Vol. 122, No. 51, 2000
Blaser et al.
Hydrogenation Procedure. The test reactions were carried out in
a 50-mL three-phase-slurry reactor with a magnetic stirring bar and a
small baffle. Etpy was added to the solvent followed by the pretreated
catalyst and the modifier. The autoclave was closed, and the air was
displaced with argon (2 times). Then the autoclave was pressurized
with H2, and the reaction was started by turning on the magnetic stirrer
(∼1000 rpm). After hydrogen uptake had stopped, the pressure was
released and the hydrogen was displaced by argon. Initial turnover
frequencies (tof; s-1) were calculated in terms of accessible platinum
atoms obtained from CO adsorption data. For series 1-3, conversion
was determined by GLC (column: OV 101, 2 m, 50 °C), ee’s by gas
chromatography on a chiral capillary column (Chirasil-(L)-Val, 50 m,
150 °C) after derivatization of the ethyl lactate with isopropylisocyanate.
For series 4, both conversion and ee of were determined without
derivatization by GLC (â-cylodextrin).
Experimental Section
Materials. Catalysts: 5R94 (Johnson Matthey) 5% Pt/Al2O3, Pt
dispersion 0.22 (measured by CO adsorption), particle size 10-30 µm,
SBET 131 m2/g; 4759 (Engelhard) 5% Pt/ Al2O3, Pt dispersion 0.24
(measured by CO adsorption), particle size 50-120 µm, SBET 168 m2/
g.2 Before use, the catalyst was reduced for 2 h in flowing hydrogen at
400 °C. Ethyl pyruvate (Fluka purum) was distilled and stored not
longer than 2 weeks at 5 °C. Ethanol, toluene, and acetic acid (all Fluka
puriss p.a.) were used as received.
Cinchona Derivatives. The following derivatives were obtained
from commercial sources and used as received: Cd, Cn, Qn, Qd, N-Bn-
Cd+Cl- (Fluka purum). The following AD ligand derivatives were
obtained either from commercial sources or from the laboratory of Prof.
K. B. Sharpless: ClBzO-HQn, ClBzO-HQd, Meq-HQn, Meq-HQd,
Phal-(HQn)2, Phal-(HQd)2, Phn-(HQn)2, Phn-(HQd)2, Pyr-(HQn)2, and
Pyr-(HQd)2. M1130 was obtained from Prof. F. Diederich (ETH Zu¨rich);
quinuclidines A and B were from Buchler GmbH (D-38110 Braun-
schweig).
Acknowledgment. We thank Prof. B. Sharpless (Scripps)
for various samples of AD ligands, Prof. F. Diederich (ETH
Zu¨rich) for a sample of the spirofluorene derivative M11, Prof.
A. Pfaltz (University of Basel) for a critical reading of the
manuscript, and S. Burkhardt, D. Imhof, J. Salathin, and A.
Schwendemann for careful experimentation.
The following cinchona derivatives were synthesized according to
the method summarized in Figure 3, and their spectroscopic data are
in good agreement with the proposed structure (for synthesis and spectra
see Supporting Information): HCd,31 MeO-HCd, AcO-HCd, PivO-HCd,
(R)-Bnlac-HCd, (S)-Bnlac-HCd, (R)-Lac-HCd, (S)-Lac-HCd, F-HCd,
cyc-HCd, A-py-HCd, B-py-HCd, DHCd, di-Br-Cd,32 and norcinchol.
Supporting Information Available: Synthetic procedures
and spectral data of new cinchona derivatives. This material is
available free of charge via the Internet at http://pubs.acs.org.
(30) B. Winter-Werner, Dissertation ETH-Z, Zu¨rich, 1996.
(31) Emde, H.; HelV. Chim. Acta 1932, 15, 557.
(32) Braje, W. M.; Frackenpohl, J.; Schrake, O.; Wartchow, R.; Beil
W.; Hoffmann, H. M. R. HelV. Chim. Acta 2000, 83, 777.
JA003259Q