Enantioselective hydrogenation over cinchona-modified Pt: The special role of carboxylic acids

Article abstract of DOI:10.1002/1521-3765(20020315)8:6<1430::AID-CHEM1430>3.0.CO;2-K

The influence of acetic acid (AcOH) and trifluoroacetic acid (TFA) on the hydrogenation of ethyl-4,4,4-trifluoroacetoacetate has been investigated by using Pt/ A1203 modified by cinchonidine and O-methylcinchonidine. We have shown that the sometimes dramatic changes in enantioselectivity and rate cannot simply be interpreted by protonation of the alkaloid modifier. We propose a new three-step reaction pathway, involving interaction of the carboxylic acid with the reactant and the chiral modifier. The mechanism is supported by IR spectroscopic identification of cyclic TFA-modifier ion pairs. This new approach can rationalise the poorly understood role of acids in the enantioselective hydrogenation of activated ketones over cinchona-modified platinum metals.

Full text of DOI:10.1002/1521-3765(20020315)8:6<1430::AID-CHEM1430>3.0.CO;2-K

FULL PAPER
Enantioselective Hydrogenation over Cinchona-Modified Pt:
The Special Role of Carboxylic Acids
Matthias von Arx, Thomas B¸rgi, Tamas Mallat, and Alfons Baiker*[a]
Abstract: The influence of acetic acid (AcOH) and trifluoroacetic acid (TFA) on the
hydrogenation of ethyl-4,4,4-trifluoroacetoacetate has been investigated by using Pt/
Al2O3 modified by cinchonidine and O-methylcinchonidine. We have shown that the
sometimes dramatic changes in enantioselectivity and rate cannot simply be
interpreted by protonation of the alkaloid modifier. We propose a new three-step
reaction pathway, involving interaction of the carboxylic acid with the reactant and
the chiral modifier. The mechanism is supported by IR spectroscopic identification of
cyclic TFA modifier ion pairs. This new approach can rationalise the poorly
understood role of acids in the enantioselective hydrogenation of activated ketones
over cinchona-modified platinum metals.
Keywords: carboxylic acids
spectroscopy ¥ enantioselectivity ¥
trifluoromethyl ketone
¥ IR
the acid remained unclear. Recently, it has been reported that
using trifluoroacetic acid (TFA) as an additive has an even
stronger effect than AcOH in the hydrogenation of ethyl
pyruvate,[20] but no feasible explanation for this behaviour has
been found.
Introduction
Cinchona alkaloid-modified Pt, suitable for the hydrogena-
tion of activated ketones, represents one of the most
successful heterogeneous enantioselective catalysts.[1
5]
In
the most studied reaction, the hydrogenation of ethyl
pyruvate, the influence of many reaction parameters has
been investigated in detail during the past decade. It has
already been recognised by Orito et al.[6] that pretreatment of
a catalyst with acetic acid (AcOH) is advantageous. More
recently, the best ee in a-ketoester hydrogenation was also
achieved in AcOH.[7, 8] In the hydrogenation of a-ketoes-
ters,[9, 10] a-ketolactones[11] and pyrrolidine triones,[12] increas-
ing solvent polarity diminished the enantioselectivity. Only
AcOH and HCOOH, which afforded good enantioselectiv-
ities, were exceptions to this negative correlation. This effect
was attributed mainly to protonation of the quinuclidine N
atom of cinchonidine (CD), and this information was used for
In the past decade the range of application of cinchona-
modified Pt has been remarkably broadened, and AcOH
proven to be the most suitable solvent in the hydrogenation of
a-ketoamides,[21] a-ketoacetals[22, 23] and trifluoromethyl ke-
tones,[24, 25] providing up to 97% ee. This development
prompted us to reinvestigate one of these reactions and to
clarify the role of acids in the reaction mechanism. Compared
to ethyl pyruvate, the positive effect of AcOH with respect to
apolar solvents such as toluene was much more pronounced in
the hydrogenation of ethyl-4,4,4-trifluoroaceto acetate.[24] We
therefore chose the hydrogenation of this compound
(Scheme 1) to investigate the role of AcOH and TFA.
16]
the development of a mechanistic model.[13
Analysis by NMR spectroscopy and theoretical calculations
revealed that protonation changes the distribution of the
alkaloid in various conformations in solution, favouring the
so-called ™open(3)∫ conformation (conformation in which the
quinuclidine nitrogen atom points away from the quinoline
19]
ring).[17 Despite this progress, some aspects of the role of
[a] Prof. Dr. A. Baiker, Dr. M. von Arx, Dr. T. B¸rgi, Dr. T. Mallat
Laboratory of Technical Chemistry
Swiss Federal Institute of Technology, ETH-Hˆnggerberg HCI
Wolfgang Pauli Strasse, 8093Z¸rich (Switzerland)
Fax : (‡41)1-632-11-63
Scheme 1. Hydrogenation of ethyl-4,4,4-trifluoroacetoacetate (1) over
chirally modified Pt/Al2O3.
E-mail: baiker@tech.chem.ethz.ch
1430
¹ WILEY-VCH Verlag GmbH, 69451 Weinheim, Germany, 2002
0947-6539/02/0806-1430 $ 17.50+.50/0
Chem. Eur. J. 2002, 8, No. 6

1430 1437
In a preliminary report we proposed, based on theoretical
calculations, the formation of a CD-TFA cyclic ion pair that
can interact with the 4-hydroxy-6-methyl-2-pyrone reactant
on the Pd surface in the enantio-differentiating step.[26] Here
the catalytic experiments are supported by IR spectroscopic
analysis and a general model is suggested to interpret the
special role of carboxylic acids in enantioselective hydro-
genation reactions over cinchona-modified metal catalysts.
Effect of TFA in toluene: Remarkable increases in ee were
achieved with both modifiers (15% with CD, 35% with
MeOCD) by adding only 0.1 vol% of TFA; this corresponds
to a TFA/modifier molar ratio of 9.6 (Figure 2). The differ-
ence between the two modifiers was larger above the
optimum amount of acid: with 2 vol% TFA (TFA/CD molar
ratio ˆ 192) the enantioselectivity was almost completely lost
in the presence of CD, while with MeOCD the ee was still
better than in pure toluene.
Results and Discussion
Effect of AcOH in toluene: The efficiency of Pt/Al2O3
modified by cinchonidine (CD) or O-methylcinchonidine
(MeOCD) in the hydrogenation of trifluoromethyl ketone (1)
in toluene/AcOH mixtures is illustrated in Figure 1. In both
Figure 2. Hydrogenation of 1 in toluene using CD and MeOCD as chiral
modifiers: changes in ee and reaction rate (r) upon addition of TFA. Open
symbols: reaction rate; closed symbols: ee.
Parallel to the changes in enantioselectivity, the reaction
rates dropped for both modifiers following addition of small
amounts of TFA, independent of the modifier used (Figure 2).
In Figure 3the relative rates, defined as the reaction rate in
the presence of TFA related to the rate in pure toluene, are
compared. Over the whole concentration range, the rate
deceleration induced by addition of TFA was more pro-
nounced with CD than in the reactions with MeOCD
Figure 1. Hydrogenation of 1 in AcOH/toluene mixtures using CD or
MeOCD as chiral modifiers. Open symbols: reaction rate (r); closed
symbols: ee.
series the changes in enantioselectivity can be described by
saturation type curves. In AcOH, MeOCD is more effective
than CD, in agreement with our recent report.[24] More
importantly, both modifiers provide the highest ee in pure
AcOH. Former NMR spectroscopic studies indicated that 2 or
24 molar equivalents of AcOH were necessary to completely
protonate the quinuclidine N atom of CD.[10, 27] The latter
value, corresponding here to 0.2 vol% in toluene, seems to be
more accurate as it is based on a comparison of the shift
induced by TFA, which affords full protonation of the
aliphatic N already at one molar equivalent. Independent of
this difference, the curves in Figure 1 indicate that changes in
the solvent properties are crucial; protonation of the quinu-
clidine N atom of the modifier cannot account for the whole
selectivity enhancement observed.
Concerning the reaction rates (r), there is a significant
difference between the two series. While the reaction rate
with MeOCD was independent of the solvent composition,
the rate decreased with increasing acid concentration when
CD was used. This difference is attributed to a modifier
solvent interaction in which the OH group of the alkaloid is
involved.
Figure 3. Hydrogenation of
comparison of the reaction rates related to the rates in pure toluene.
1 in the presence of CD or MeOCD:
Chem. Eur. J. 2002, 8, No. 6
¹ WILEY-VCH Verlag GmbH, 69451 Weinheim, Germany, 2002
0947-6539/02/0806-1431 $ 17.50+.50/0
1431

A. Baiker et al.
FULL PAPER
(deceleration by a factor of 17 and 6, respectively). This
behaviour is in qualitative agreement with the results in
toluene/AcOH mixtures shown in Figure 1, where a consid-
erable rate deceleration was observed only with CD. How-
ever, much greater amounts of AcOH are needed to obtain
similar changes in ee and reaction rates relative to that used of
TFA.
Effect of TFA in polar solvents: We have reported recently
that in the hydrogenation of 1, higher enantioselectivities can
be obtained in polar solvents than in toluene.[28] For example,
70% ee was achieved in THF with MeOCD as modifier
(Figure 4). Addition of relatively small amounts of TFA
increased the ee to 90% although the full beneficial effect
leading to 93% ee was obtained only at a TFA/MeOCD molar
ratio of 192 (2 vol% TFA). In this solvent the ee did not drop
at higher TFA concentration. In addition, the reaction rate did
not decrease with an increasing amount of TFA as was
observed in toluene (Figure 2), but remained almost constant
(within Æ10%).
Figure 5. n(OH) region of the IR spectra of solutions of CD and MeOCD
in CH2Cl2 containing different amounts of TFA (TFA/modifier molar ratio
between 0 and 6). Modifier concentrations are higher by a factor of 2.5
compared to the standard reaction procedure. Spectra are offset.
Figure 4. Hydrogenation of 1 using MeOCD as chiral modifier: changes in
ee and reaction rate (r) upon addition of TFA to THF or AcOH. Open
symbols: reaction rate; closed symbols: ee.
Addition of TFA induced a small enhancement of ee even
in AcOH (Figure 4). Interestingly, the same maximum of
93% ee was achieved in both AcOH and THF. The reaction
rate was practically constant in AcOH above the optimum
amount of TFA.
IR spectroscopic analysis: The formation and structure of
complexes between triethylamine and AcOH (existing pre-
dominantly as dimers in apolar media) were described in
1954.[29] The existence of similar TFA modifier interactions
is proven by the IR spectra in Figures 5 and 6. For comparison,
the effect of increasing amount of TFA is presented for both
cinchonidine (CD) and O-methylcinchonidine (MeOCD).
The spectra contain many features similar to those reported
for CD AcOH complexes.[30] Interaction of AcOH with the
quinuclidine N and the OH group of CD has been interpreted
ˆ
Figure 6. n(C O) region of the IR spectra of solutions of CD and MeOCD
in CH2Cl2 containing different amounts of TFA (TFA/modifier molar ratio
between 0 and 6). Modifier concentrations are higher by a factor of 2.5
compared to the standard reaction procedure. Spectra are offset.
by the formation of at least three types of complexes (1:1
cyclic, 2:1 cyclic and 2:1 half-cyclic). The analogous complexes
for TFA and CD are depicted in Figure 7.
1432
¹ WILEY-VCH Verlag GmbH, 69451 Weinheim, Germany, 2002
0947-6539/02/0806-1432 $ 17.50+.50/0
Chem. Eur. J. 2002, 8, No. 6

Cinchona-Modified Pt
1430 1437
À1
ˆ
n(C O) region is shown. The band at 1804 cm arises from
the non-hydrogen-bonded carbonyl group of free acid mol-
ecules, while the lower frequency band at 1784 cmÀ1 can be
assigned to the hydrogen-bonded carbonyl group of the acid
(e.g., acid dimers). The signal at 1670 cmÀ1, which shifts to
lower frequency with increasing acid concentration, belongs
to the carbonyl group of the carboxylate ion bound to the
‡
quinuclidine NH group of the modifier. The shift of this band
to lower frequencies is attributed to the formation of TFA
modifier complexes with molar ratios higher than 1:1.[30] All of
these signals appear in the spectra of both modifiers implying
that the acid molecules form complexes with both MeOCD
and CD. However, the differences in the relative intensities
show that the concentration of the various complexes are
unequal for CD and MeOCD for the same reasons given
above. Furthermore, a comparison of the region around
1800 cmÀ1 reveals that in the TFA/MeOCD mixture, with a
molar ratio of 1.8, the signal for the free acid and acid dimers
already appears. This is an indication that the TFA MeOCD
complexes are less stable than the TFA CD complexes. In
the latter case, at a TFA/CD molar ratio of 1.8, all acid
molecules are involved in complex formation and no free acid
molecules are present in solution.
Figure 7. Structures of TFA-CD complexes (complex A in Figure 8)
formed in solution at low TFA concentration.
It should be noted that toluene was used in the catalytic
experiments (Figures 1 3), while the spectroscopic study was
carried out in another weakly polar solvent, dichloromethane.
This is because IR spectroscopy in toluene is difficult due to
the low solubility of the modifier and the strong absorption of
The IR spectrum of CD in dichloromethane shows a peak at
3600 cmÀ1 due to the n(OH) vibration of its non-hydrogen-
bonded OH group (Figure 5). Upon addition of TFA the
intensity of this band decreased and a new, broad peak at
3220 cmÀ1 appeared. The latter signal is associated with the
formation of a hydrogen-bond between TFA and the OH
group of CD. This observation proves that, in addition to the
quinuclidine N that is protonated by one equivalent of
TFA,[27] CD binds to the acid also with its OH group forming
ˆ
toluene itself. Particularly, spectroscopy of the n(C O) region
is hampered by strong solvent absorption. In control experi-
ments, titration of cinchonidine with TFA showed similar
À
behaviour of the intensity of O H vibrations in toluene as in
À
dichloromethane. The intensity of n(O H) of the non-hydro-
gen-bonded OH group of cinchonidine decreased with
increasing acid concentration, whereas the bands due to
complex formation increased. The solvent had an effect,
however, on the band positions. In dichloromethane, bands
associated with hydrogen-bonded OH groups are observed at
3220, 3445 and 3575 cmÀ1, whereas in toluene the correspond-
ing bands appear at 3230, 3365 and 3530 cmÀ1. The other
possibility for ™bridging∫ experiments is a study of the effect
of carboxylic acids in the hydrogenation of 1 in a chlorinated
solvent (dichloromethane or dichlorobenzene). Unfortunate-
ly, the Pt was strongly poisoned, most likely as a result of
(minor) dehalogenation of the solvent.
‡
a NH ¥¥¥ OCOÀ ¥¥¥ HO cyclic complex as shown in Figure 7.
The new signals at 3575 and 3445 cmÀ1, which appear upon
further addition of TFA, belong to the hydrogen-bonded
n(OH) vibration of TFA molecules in the different CD TFA
complexes.[30] The weak n(OH) band of the free acid[31] at
3504 cmÀ1 is obscured by these signals and can barely be
detected.
In the spectra of MeOCD the signals at 3600 cmÀ1 and
3220 cmÀ1 are missing (Figure 5) because the OH group is
methylated. Due to the lack of this interacting function, cyclic
TFA modifier complexes are not possible. Even so, the
‡
From all of these observations we can conclude that both
modifiers form stable complexes involving one or more TFA
molecules, similar to those shown in Figure 7. In other words,
the quinuclidine N atom of the modifier is not only
protonated by TFA, but stable ion pairs exist in apolar media.
In this ion pair the quinuclidine N atom (as well as the OH
group of CD) is ™blocked∫ by the acid molecule(s). Ab initio
calculations indicate that the binding energy of the CD TFA
formation of a stable ion-pair complex (CD carboxylate) by
protonation of the quinuclidine N is apparent and leads to a
characteristic increase of the baseline in the 3100 1900 cmÀ1
region (not shown in Figure 5). The appearance of two signals
at 3580 cmÀ1 and 3440 cmÀ1 implies the formation of addi-
tional TFA MeOCD complexes (e.g., 2:1 complexes) anal-
ogous to those formed by CD. However, due to the lack of the
free OH group in MeOCD, the structural arrangement of the
acid molecule(s) and MeOCD must be different from that
with CD. This is reflected in the spectra, where the relative
intensities of the two signals (at 3580 cmÀ1 and 3440 cmÀ1) are
different for the two modifiers.
cyclic ion pair is relatively high, 23kcalmol À1 [26]
Even so, no
.
enantioselectivity would be possible if access to the modifier
were prevented. The observed dramatic increase in enantio-
selectivity in Figures 2 and 4 implies that such a TFA mod-
ifier complex is the real chiral modifier, affording enhanced
enantioselectivities but lower reaction rates. Therefore, in the
The formation of TFA modifier complexes is further
evidenced by the spectra illustrated in Figure 6, in which the
Chem. Eur. J. 2002, 8, No. 6
¹ WILEY-VCH Verlag GmbH, 69451 Weinheim, Germany, 2002
0947-6539/02/0806-1433 $ 17.50+.50/0
1433

A. Baiker et al.
FULL PAPER
enantio-discriminating reaction step an arrangement of re-
actant, chiral modifier and acid has to be considered.
Interpretation of the catalytic experiments by the three-step
model
Comparison of cinchonidine (CD) and O-methylcinchonidine
(MeOCD): In any solvent, addition of a carboxylic acid led to
rate deceleration in the presence of CD as chiral modifier; the
effect was less pronounced or even negligible with MeOCD
(Figures 1 3). The ee increased in all experiments with both
modifiers. Evidently, in the presence of AcOH or TFA
protection of the hydroxyl group of CD (leading to MeOCD)
is beneficial for enantioselection and for the overall reaction
rate.
Proposed reaction mechanism: The catalytic experiments and
IR spectroscopic analysis indicate that hydrogenation of 1 on
the Pt surface is governed by specific molecular arrangements
involving the modifier, the reactant and one or more
carboxylic acid molecules. We suggest a three-step reaction
mechanism to interpret the rate and enantioselectivity of the
hydrogenation reaction, as shown in Figure 8. In the first step
The main difference between the two modifiers is that
MeOCD cannot form cyclic complexes with the carboxylic
acids (Figure 7). The cyclic complexes, feasible only with CD,
are expected to be more stable and also more rigid. For this
reason coordination of a reactant molecule to the acid
modifier ion pair A, which is necessary for the formation of
the enantio-differentiating complex B, is less favoured with
CD; this results in lower reactivity of the acid CD ion pair A.
However, the lower reactivity of complex A is not detrimental
to the enantio-differentiation. On the contrary, the three-step
reaction pathway in the presence of a carboxylic acid provides
higher enantioselectivities with both modifiers (Figures 1 and
2) as long as a large excess of TFA is not added.
Figure 8. Three-step reaction mechanism proposed for the enantioselec-
tive hydrogenation of trifluoromethyl ketones in the presence of carboxylic
acids. The presently accepted two-step reaction mechanism (valid in the
absence of carboxylic acids) is indicated by dashed lines.
Role of acid strength: Enhanced selectivity could be achieved
with both carboxylic acids tested (Figures 1 and 2). Much
smaller amounts of TFA were required to obtain similar or
even more pronounced effects compared with the use of
AcOH. With the stronger acid, TFA, the complex A is more
stable and, with respect to AcOH, the modifier‡ acid , A
equilibrium is shifted in favour of A. A further advantage of
TFA is its decreased tendency to form hydrogen-bonded
dimers in solution.[37] At a certain concentration, the number
of free acid molecules interacting with the modifier is
therefore higher than when AcOH is used.
The correlation between ee and acid/modifier molar ratio in
Figures 1 and 2 demonstrates that protonation of the quinu-
clidine N atom alone is not sufficient to achieve the maximum
ee. This conclusion is supported by the results obtained with
CD ¥ HCl (Table 1). Although the quinuclidine N atom is
protonated in this modifier, it is no more effective in the
hydrogenation of 1 than CD.
the chiral modifier and at least one carboxylic acid molecule
form an acid modifier complex (complex A) of the type
shown in Figure 7. The existence of such complexes in apolar
media is proven by IR spectroscopic analysis. The equilibrium
of formation of complex A (Figure 8) depends on the strength
of the acid, therefore a significant difference between AcOH
(pKa ꢀ 4.75) and TFA (pKa ꢀ 0.3) is likely. Acid strength also
influences the stability and structural rigidity of the alkaloid
acid complexes.
Some destabilisation of complex A, coupled with increasing
structural flexibility, can be expected upon solvation in polar
and polar protic solvents. This solvent effect facilitates the
access of reactant to the possible interaction sites of A, which
is a prerequisite for the formation of the enantio-discriminat-
ing ketone modifier acid complex (complex B in Figure 8).
Upon hydrogen uptake and enantioselective transformation
of the reactant, the product is released leaving complex A,
which is involved in the next cycle, behind.
All of these findings indicate that a stable ion pair complex
(A) is responsible for the increased enantioselectivity. Due to
the strong acidity of TFA, formation of complex A is favoured
at low acid concentration, leading almost exclusively to
product formation through the three-step reaction pathway.
Evidently, this is not the exclusive reaction pathway. In the
absence of a sufficient amount of acid the commonly accepted
two-step reaction mechanism dominates. This mechanism,
shown by dotted arrows in Figure 8, comprises only the
modifier and reactant in the enantio-discriminating com-
plex[16] and affords considerably lower ee×s in the hydro-
genation of 1 (by up to 48%, see Figure 1).
Table 1. Influence of modifier structure on enantioselectivity in the hydro-
genation of 1 and ethyl pyruvate in weakly polar and acidic media at standard
conditions.
Note that all these steps occur on the Pt surface (not
indicated in Figure 8) on which the chiral modifier is strongly
Compound
Solvent CD ee CD ¥ HCl ee MeOCD ee NMeCD ee
[%]
[%]
[%]
[%]
adsorbed by its aromatic ring system (™anchoring moi-
36]
ety∫).[32
Unfortunately, there is currently no in situ
1
1
toluene 46
AcOH 70
ethyl pyruvate toluene 86
ethyl pyruvate AcOH 94
42
43
90
80
95
< 2
< 2
< 2
< 2
information available on the adsorption mode of modifier
reactant complex on Pt, either with or without a carboxylic
acid.
89
1434
¹ WILEY-VCH Verlag GmbH, 69451 Weinheim, Germany,
2002
0947-6539/02/0806-1434 $ 17.50+.50/0
Chem. Eur. J. 2002, 8, No. 6

Cinchona-Modified Pt
1430 1437
The selectivity enhancement observed in this pathway is
intrinsically coupled with a lower reaction rate due to
™blocking∫ of the interaction sites of the modifier by the acid
molecule(s) in A. The increase in ee by 35%, together with a
drop in reaction rate by a factor of 3.5 upon addition of
9.6 equivalents of TFA (Figure 2), further suggests adherence
to the three-step reaction pathway.
In contrast, a large excess of AcOH is required to
effectively shift the equilibrium towards the formation of
complex A and thus favour the three-step reaction pathway.
This provides an explanation as to why the ee increased slowly
with increasing AcOH concentration in toluene/AcOH mix-
tures (Figure 1). The lower stability of this complex then
allows a faster coordination of the reactant and a smaller
decrease in reactivity compared to reactions carried out in the
presence of TFA.
most studied reaction on chirally modified Pt, are shown for
comparison. It is evident that the OH group of the modifier is
not crucial in enantio-differentiation, but quaternation of the
quinuclidine N atom of CD with MeI leads to a complete loss
of ee with both reactants, independent of the solvent (toluene
or AcOH). This is a strong indication that the quinuclidine N
atom of the modifier interacts with the reactant in both
solvents in the enantio-discriminating step. An N-H-O type
interaction involving the keto-carbonyl O atom has been
proposed for the hydrogenation of ethyl pyruvate,[5, 15, 16, 32]
ketopantolactone[11] and trifluoroacetophenone.[38]
On the basis of these data, the open and cyclic complexes
depicted in Figure 9 represent some possible structures of
complex B. The structures I VI illustrate possible interac-
tions of reactant, modifier and acid without considering the
adsorption on the metal surface. The complexes I V in
Figure 9 are formed from the 1:1 or 2:1 acid CD complexes
A (shown in Figure 7) by insertion of a reactant molecule into
the cyclic structure. The complexes I and III are considered as
the most probable cyclic structures when CD is used as
modifier. They include an N-H-O type interaction with the
reactant and an additional interaction with a carboxylic acid
molecule. In complexes II, IV and V the OH group of CD
plays a crucial role in the reactant modifier interaction; this
contradicts the experimental observations as methylation of
the OH function does not hinder the enantio-differentiation
(Table 1). In II and III the acid reactant interaction is
Solvent effect: When the hydrogenation of 1 over MeOCD-
modified Pt was carried out in polar solvents such as THF or
even AcOH, higher amounts of TFAwere needed to reach the
maximum in ee and no rate deceleration was observed
(Figure 4). These changes are attributed to the effect of
solvation of complexes A and B. As described above, a high
stability of complex A is needed to favour product formation
following the three-step reaction pathway. Upon solvation in a
polar solvent, the structural flexibility of A is increased
significantly. For this reason the reactivity of A towards
coordination of a reactant molecule (leading to B) and
transformation of the reactant to the product is enhanced. As
the overall reaction rate is controlled by the rate-determining
(™slowest∫) step and formation of complex A is very fast, the
increased reactivities of A and B in polar solvents eliminate
the dramatic drop in overall reaction rate observed upon TFA
addition to toluene (compare Figures 1 and 4).
The slight increase in ee after addition of TFA to AcOH
(Figure 4) can be interpreted in terms of the effect of acid
strength as discussed above. The small changes induced by
TFA indicate that exchanging AcOH by TFA in complexes A
and B has only a minor influence on the enantio-discrim-
ination.
In summary, the best reaction conditions in the hydro-
genation of the trifluoromethyl ketone (1) are those that
favour product formation following the proposed three-step
reaction mechanism, leading to enhanced enantioselectivities
without a considerable rate deceleration. This is achieved by
working in a polar solvent doped with small amounts of TFA,
or in pure AcOH. Compared with CD, the use of MeOCD is
also beneficial.
Possible structures of the enantio-discriminating complex B:
The catalytic experiments carried out in the presence of a
carboxylic acid indicated that complex B has sufficient
conformational rigidity and structural flexibility to allow the
observation of high ee and good reaction rates. An intriguing
question is the real nature of reactant modifier acid inter-
actions in complex B. Valuable information on the structure of
B is provided by the hydrogenation of 1 over Pt/Al2O3
modified by some simple CD derivatives (Table 1). The
results of the hydrogenation of ethyl pyruvate, which is the
Figure 9. Possible structures of the reactant modifier carboxylic acid
complex (complex B in Figure 8) responsible for enantio-discrimination
(note that the interaction with the metal surface is not included).
Chem. Eur. J. 2002, 8, No. 6
¹ WILEY-VCH Verlag GmbH, 69451 Weinheim, Germany, 2002
0947-6539/02/0806-1435 $ 17.50+.50/0
1435

A. Baiker et al.
FULL PAPER
directed from the OH group of the acid towards the carbonyl
O atom of the reactant. IR spectroscopy has recently revealed
that such an interaction between TFA and another trifluoro-
methyl ketone (2,2,2-trifluoroacetophenone)[39] is very stable
in apolar media.
ester group is critical in the enantio-differentiating step, is
formed in the absence of acid. These results support our
proposal for the two-step and three-step reaction pathways in
the absence or presence of a carboxylic acid, respectively
(Figure 8).
Complex VI represents a feasible interaction between the
protonated MeOCD and 1. This singly-bonded complex is
expected to be less stable than the cyclic structures, but,
according to the catalytic experiments, affords higher enan-
tioselectivity (Figures 1 and 2). A possible explanation for this
increased efficiency is that the energy (or reactivity) differ-
ence between the singly-bonded pro-(S) and pro-(R) com-
plexes is bigger than in the case of the cyclic complexes.
Complex VI may be formed with CD, but with this modifier
the more stable cyclic complexes are favoured. Finally, the
singly-bonded complex VI is not only less stable but also more
reactive than the cyclic structures. This interpretation is
supported by the higher reaction rates observed in the
presence of MeOCD as modifier, relative to those achieved
with CD-modified Pt. Note that the reactant modifier
interaction in complex VI is similar to that suggested for the
hydrogenation of ethyl pyruvate in an acidic medium.[16]
The above interpretation may serve as a working hypoth-
esis for future studies necessary to prove the feasibility of any
of the proposed models.
Mechanistic implications for the hydrogenation of other
activated ketones: As mentioned in the introduction, the
enantioselectivities in the hydrogenation of a-ketoacetals, a-
ketolactones, pyrrolidine triones and a-ketoesters were out-
standing in acidic solvents. In the most studied reaction, the
hydrogenation of ethyl pyruvate, 97% ee has been achieved in
AcOH.[8, 40] The question arises as to whether the suggested
three-step reaction mechanism is also valid for this reaction in
acidic media. It has been reported that upon addition of
AcOH to toluene, and TFA to toluene or AcOH, the ee
increased significantly.[20] However, no reaction rates were
reported and the enantioselectivities were far from optimum.
We have also observed a strong positive effect of acid (AcOH,
TFA) addition in the enantioselective hydrogenation of a
pyrrolidine trione, but, again, only under conditions that were
far from optimal.[41] This is an important deviation from the
observations surrounding the hydrogenation of 1.
A comparison of the hydrogenations of ethyl pyruvate and
1 in toluene containing increasing amounts of AcOH reveals
some analogies as well as differences. A saturation-type curve
was observed in both reactions (Figure 11). However, in the
Influence of the size of ester group in the reactant: Our
suggestion for a three-step mechanism in the hydrogenation
of 1 in acidic media is supported by the striking results
obtained by using the methyl, ethyl or isopropyl esters of 1. In
an acidic medium over MeOCD-modified Pt/Al2O3 the ee×s
were high with all reactants independent of the size of the
ester group (92 93% in a TFA/THF mixture and 89 90% in
AcOH, Figure 10). Upon replacing MeOCD by CD in AcOH,
Figure 11. Comparison of the enantioselectivities in the hydrogenation of
1 and ethyl pyruvate in AcOH/toluene mixtures, using cinchonidine as
chiral modifier.
Figure 10. Enantioselectivities obtained in the hydrogenation of methyl-,
ethyl- and isopropyl-4,4,4-trifluoroacetoacetate in the presence or absence
of acid.
hydrogenation of ethyl pyruvate about half of the positive
effect was achieved already with 24 molar equivalents of
AcOH, which are needed to completely protonate the
quinuclidine N atom of CD,[27] and the optimum in ee at more
than an order of magnitude higher acid concentration was
only 2 3% higher. In contrast, in the hydrogenation of 1,
24 equivalents of acid afforded only a small enhancement
in ee.
Hydrogenation of ethyl pyruvate is almost as efficient in
toluene as in AcOH, making a comparison more difficult.
Although the two curves in Figure 11 are similar, the sharp
initial increase of ee in the hydrogenation of ethyl pyruvate
the ee decreased to 66 75% with increasing bulk of the ester.
The loss of ee was even more pronounced when changing to
toluene. Increasing the size of the ester group, the ee
decreased from 64 to 17%. This is a strong indication that
hydrogenation of these compounds in toluene and AcOH
proceeds by two different reaction mechanisms. In AcOH the
enantio-differentiation is not, or is only to a small extent,
sensitive to the size of the ester group in 1. In contrast, a
structurally different complex, in which the bulkiness of the
1436
¹ WILEY-VCH Verlag GmbH, 69451 Weinheim, Germany, 2002
0947-6539/02/0806-1436 $ 17.50+.50/0
Chem. Eur. J. 2002, 8, No. 6

Cinchona-Modified Pt
1430 1437
[1] A. Baiker, H. U. Blaser in Handbook of Heterogeneous Catalysis,
Vol. 5 (Eds.: G. Ertl, H. Knˆzinger, J. Weitkamp), VCH, Weinheim,
1997, pp. 2422 2436.
indicates that in this reaction the effect of protonation of the
modifier is more important than in the hydrogenation of 1.
This being so, a spectroscopic analysis of the reactant cin-
chonidine AcOH system would be necessary to evaluate the
feasibility of the three-step reaction pathway in the hydro-
genation of ethyl pyruvate and other activated ketones over
chirally modified Pt.
[2] G. J. Hutchings, Chem. Commun. 1999, 301.
[3] T. Mallat, A. Baiker in Fine Chemicals through Heterogeneous
Catalysis (Eds.: R. A. Sheldon, H. V. Bekkum), VCH, Weinheim,
2001, pp. 449 460.
[4] G. V. Smith, F. Notheisz in Heterogeneous Catalysis in Organic
Chemistry, Academic Press, San Diego, 1999.
[5] P. B. Wells, K. E. Simons, J. A. Slipszenko, S. P. Griffiths, D. F. Ewing,
J. Mol. Catal. A. 1999, 146, 159.
[6] S. Niwa, S. Imai, Y. Orito, J. Chem. Soc. Jpn. 1982, 13 7.
[7] H. U. Blaser, H. P. Jalett, J. Wiehl, J. Mol. Catal. 1991, 68, 215.
Conclusion
¬
¬
[8] B. Tˆrˆk, K. Balazsik, G. Szˆllˆsi, K. Felfˆldi, M. Bartok, Chirality
The long-standing question of why AcOH, used as solvent or
additive in the enantioselective hydrogenation of activated
ketones over chirally modified Pt, affords the highest
enantioselectivities has been adressed. A systematic inves-
tigation of the role of AcOH and TFA in the hydrogenation of
trifluoromethyl ketone (1) revealed dramatic effects on
enantioselectivity and reaction rates. The results indicate
that–at least in the hydrogenation of 1–the reason for the
outstanding performance of the catalytic system can be traced
to the propensity of carboxylic acids to form acid modifier
ion-pair complexes. The existence of such complexes has been
confirmed by IR spectroscopy. These acid modifier ion-pair
complexes are suggested to act as the real chiral modifiers
over Pt. A new three-step reaction pathway, which can
rationalize the behaviour of the Pt cinchona system in the
presence of carboxylic acids is proposed.
1999, 11, 470.
[9] H. U. Blaser, H. P. Jalett, J. Wiehl, J. Mol. Catal. 1991, 68, 215.
[10] B. Minder, T. Mallat, P. Skrabal, A. Baiker, Catal. Lett. 1994, 29, 115.
[11] M. Sch¸rch, O. Schwalm, T. Mallat, J. Weber, A. Baiker, J. Catal. 1997,
169, 275.
[12] A. Szabo, N. K¸nzle, T. Mallat, A. Baiker, Tetrahedron: Asymmetry
1999, 10, 61.
[13] O. Schwalm, J. Weber, J. Margitfalvi, A. Baiker, J. Mol. Struct. 1993,
297, 285.
[14] O. Schwalm, B. Minder, J. Weber, A. Baiker, Catal. Lett. 1994, 23, 271.
[15] A. Baiker, J. Mol. Catal. A 1997, 115, 473.
[16] A. Baiker, J. Mol. Catal. A. 2000, 163, 205.
[17] G. D. H. Dijkstra, R. M. Kellogg, H. Wynberg, J. S. Svendsen, I.
Marko, K. B. Sharpless, J. Am. Chem. Soc. 1989, 111, 8069.
[18] G. D. H. Dijkstra, R. M. Kellogg, H. Wynberg, J. Org. Chem. 1990, 55,
6121.
[19] T. B¸rgi, A. Baiker, J. Am. Chem. Soc. 1998, 120, 12920.
¬
¬
[20] B. Tˆrˆk, K. Balazsik, K. Felfˆldi, M. Bartok, Stud. Surf. Sci. Catal.
2000, 130, 3381.
[21] G. Z. Wang, T. Mallat, A. Baiker, Tetrahedron: Asymmetry 1997, 8,
2133.
¬
¬
[22] B. Tˆrˆk, K. Felfˆldi, K. Balazsik, M. Bartok, Chem. Commun. 1999,
1725.
Experimental Section
[23] M. Studer, S. Burkhardt, H. U. Blaser, Chem. Commun. 1999, 1727.
[24] M. von Arx, T. Mallat, A. Baiker, J. Catal. 2000, 193, 161.
[25] M. von Arx, T. Mallat, A. Baiker, Tetrahedron Asymmetry 2002, 205,
213.
The 5 wt% Pt/Al2O3 catalyst (Engelhard 4759, metal dispersion after heat
treatment: 0.27, as determined by TEM) was pre-reduced in flowing
hydrogen for 90 minutes at 4008C. After cooling to room temperature in
hydrogen, the catalyst was transferred to the reactor without exposure to
air. Ethyl-4,4,4-trifluoroacetoacetate (1) was distilled, and THF was dried
over potassium before use. All other chemicals were used as received. O-
Methylcinchonidine (MeOCD)[42] and N-methylcinchonidinium chloride
(NMeCD)[43] were synthesised as described previously.
[26] W. R. Huck, T. B¸rgi, T. Mallat, A. Baiker, J. Catal., in press.
[27] W.-R. Huck, T. Mallat, A. Baiker, J. Catal. 2001, 200, 171.
[28] M. von Arx, T. Mallat, A. Baiker, Catal. Lett., in press.
[29] G. M. Barrow, E. A. Yerger, J. Am. Chem. Soc. 1954, 76, 5211.
[30] D. Ferri, T. B¸rgi, A. Baiker, J. Chem. Soc. Perkin Trans. 2 1999, 1305.
[31] J. D. S. Goulden, Spectrochim. Acta 1954, 6, 129.
[32] K. E. Simons, P. A. Meheux, S. P. Griffiths, I. M. Sutherland, P.
Johnston, P. B. Wells, A. F. Carley, M. K. Rajumon, M. W. Roberts,
A. Ibbotson, Rec. Trav. Chim. Pays-Bas 1994, 113, 465.
[33] B. Minder, T. Mallat, A. Baiker, G. Wang, T. Heinz, A. Pfaltz, J. Catal.
1995, 154, 371.
Hydrogenations were carried out at room temperature (approximately
208C) in an autoclave equipped with a 50 mL glass liner and a PTFE cover
in order to keep the system inert. Efficient magnetic stirring (1000 rpm)
was applied to avoid hydrogen transport limitation in the slurry reactor.
Total pressure (10 bar under standard conditions) and hydrogen uptake
were controlled by computerised constant-volume constant-pressure
equipment (B¸chi BPC9901). According to the general reaction procedure
the catalyst (42 Æ 2 mg) was added to a mixture of the modifier (6.8 mmol)
and the reactant (0.34 g, 1.85 mmol) in 5 mL solvent (reactant/modifier
molar ratio: 270). Conversion and enantioselectivities were determined by
direct gas chromatographic analysis of the reaction mixture using a chirasil-
DEX CB (Chrompack) capillary column in a HP6890 gas chromatograph.
For the hydrogenation reactions of 1, diglyme and decane were used as
internal standards in THF and toluene, respectively. Enantioselectivity is
¬
¬
[34] M. Bartok, T. Bartok, G. Szˆllˆsi, K. Felfˆldi, Catal. Lett. 1999, 61, 57.
[35] T. Evans, A. P. Woodhead, A. Gutierrez-Sosa, G. Thornton, T. J. Hall,
A. A. Davis, N. A. Young, P. B. Wells, R. J. Oldman, O. Plashkevych,
O. Vahtras, H. Agren, V. Carravetta, Surf. Sci. 1999, 436, L691.
[36] D. Ferri, T. B¸rgi, A. Baiker, Chem. Commun. 2001, 13, 1172.
[37] R. A. Nyquist, T. D. Clark, Vib. Spectrosc. 1996, 10, 203.
[38] M. Bodmer, T. Mallat, A. Baiker in Catalysis of Organic Reactions
(Ed.: F. E. Herkes), Marcel Dekker, New York, 1998, pp. 75 87.
[39] M. von Arx, T. B¸rgi, T. Mallat, A. Baiker, manuscript in preparation.
[40] X. Zuo, H. Liu, C. Guo, X. Yang, Tetrahedron 1999, 55, 7787.
[41] A. Szabo, N. K¸nzle, T. Mallat, A. Baiker, unpublished results.
[42] K. Borszeky, T. B¸rgi, Z. Zhaohui, T. Mallat, A. Baiker, J. Catal. 1999,
187, 160.
expressed as ee (%) ˆ 100 Âj (R À S) j (R ‡ S). Average reaction rates (r)
/
were calculated based on the time needed to obtain full conversion or, in
slow reactions, based on the conversion achieved in two hours of reaction
time. The estimated standard deviations are less than Æ0.5% for ee and
Æ5 20% for TOF (the higher values are related to the fast reactions).
[43] M. von Arx, T. Mallat, A. Baiker in Supported Reagents and their
Applications (Eds.: D. C. Sherrington, A. P. Kybett), Royal Society of
Chemistry, Special Publication, Cambridge, 2001, pp. 247 254.
Dichloromethane (Fluka) was used as solvent for the IR spectroscopy
experiments and was dried over molecular sieves (5 ä). A Bruker IFS-66
spectrometer was used at a resolution of 1 cmÀ1 and recording 200 scans.
The spectra were measured in a cell equipped with CaF2 windows with a
0.5 mm path length. The spectrum of pure dichloromethane was used as a
reference.
Received: September 3, 2001 [F3529]
Chem. Eur. J. 2002, 8, No. 6
¹ WILEY-VCH Verlag GmbH, 69451 Weinheim, Germany, 2002
0947-6539/02/0806-1437 $ 17.50+.50/0
1437