Multiple cycle reaction mechanism in the enantioselective hydrogenation of α,α,α-trifluoromethyl ketones

Article abstract of DOI:10.1016/j.jcat.2011.03.009

The enantioselective hydrogenation of 2,2,2-trifluoroacetophenone (1) on cinchona-modified Pt, combined with the diastereoselective hydrogenation of cinchonidine and NMR analysis of the modifier-substrate-product interactions, revealed the key role of the

Full text of DOI:10.1016/j.jcat.2011.03.009

Journal of Catalysis 280 (2011) 104–115
Contents lists available at ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Multiple cycle reaction mechanism in the enantioselective hydrogenation
of a,a,a-trifluoromethyl ketones
1
⇑
Z. Cakl , S. Reimann, E. Schmidt, A. Moreno, T. Mallat, A. Baiker
Institute for Chemical and Bioengineering, Department of Chemistry and Applied Biosciences, ETH Zurich, Hönggerberg, HCI, CH-8093 Zurich, Switzerland
a r t i c l e i n f o
a b s t r a c t
Article history:
The enantioselective hydrogenation of 2,2,2-trifluoroacetophenone (1) on cinchona-modified Pt, com-
bined with the diastereoselective hydrogenation of cinchonidine and NMR analysis of the modifier–
substrate–product interactions, revealed the key role of the product (S)-1-phenyl-2,2,2-trifluoroethanol
(2) in enantioselection. We propose a multiple cycle mechanism including a racemic route (a) on the
unmodified sites and three enantioselective routes. In the enantioselective cycles, there is an N–H–O type
interaction between the quinuclidine N and the carbonyl O-atom of the substrate. At low conversion, the
alkaloid alone is the source of chiral information (route b). With increasing conversion, the weakly acidic
minor product (S)-2 forms an adduct with the alkaloid and this complex controls the enantioselection
(route c, lower ee). The frequently applied strong acid additive TFA replaces (S)-2 and the alkaloid–TFA
complex gives the highest ee (route d). The diastereoselective hydrogenation of cinchonidine disproves
a former mechanistic model proposed in the literature.
Received 2 February 2011
Revised 8 March 2011
Accepted 9 March 2011
Available online 13 April 2011
Keywords:
Enantioselective
Diastereoselective
Hydrogenation
Platinum
Cinchona alkaloid
NMR
Ó 2011 Elsevier Inc. All rights reserved.
Trifluoromethyl ketone
Trifluoromethyl alcohol
1. Introduction
From a mechanistic point of view, the differences fill a long list.
In the hydrogenation of a-ketoesters, blocking the basic quinucli-
Early after discovering the enantioselective hydrogenation of
dine N atom of the alkaloid by alkylation or arylation eliminates
the enantioselection, while O-methylation (to MeOCD) has only a
minor effect on the ee [11,12]. This difference has been commonly
interpreted as evidence for the involvement of the quinuclidine N
in the activated complex leading to enantioselection and for the
minor importance of the OH function [13–18]. N-methylation of
CD leads to a loss of ee also in the hydrogenation of 1 and ethyl-
4,4,4-trifluoroacetoacetate [19]. The influence of O-methylation is
more complicated. In the hydrogenation of five different aryl-
substituted 2,2,2-trifluoroacetophenones, the ee was almost com-
pletely lost or even the opposite enantiomer, the (S)-alcohol
formed in small excess (4–11%), when CD was replaced with MeO-
CD [5]. On the contrary, replacement of CD by MeOCD enhanced
the ee from 70% to 90% in the hydrogenation of ethyl 4,4,4-trifluo-
roacetoacetate [20]. In some other cases, the effect of O-methyla-
tion depended also on the substituents in the substrate and on
the solvent [8,19].
There are several more examples on the crucial role of the alco-
holic OH function in enantioselection. In the hydrogenation of tri-
fluoromethyl cyclohexyl ketone, replacement of toluene by EtOH
inverted the ee with both CD and CN [21]. In two other instances,
in the hydrogenation of adamantyl trifluoromethyl ketone and
tert-butyl-trifluoromethyl ketone, the major product was inverted
upon addition of 2-propanol [10]. Formation of the corresponding
a
-ketoesters by Orito’s group [1], it was generally considered that
the Pt–cinchona system is highly specific to the transformation of
the 1,2-dicarbonyl compounds -ketoesters, -ketoacids, and
-diketones [2,3]. The successful hydrogenation of 2,2,2-trifluoro-
a
a
a
acetophenone (1) to (R)-1-phenyl-2,2,2-trifluoroethanol (2,
Scheme 1) with cinchonidine (CD)-modified Pt/alumina was the
first evidence that the real structural requirement the substrate
has to fulfill is the presence of an activating function in
to the carbonyl group [4]. In the past years, the research in the
hydrogenation of -trifluoroketones has revealed unique char-
acteristics of this reaction class, compared with those of the mostly
investigated transformation, the hydrogenation of -ketoesters.
a–position
a,a,a
a
From a synthetic point of view, the most important deviation is
the unusual substrate specificity of the Pt–cinchona system:
Hydrogenation of 1 [5] and alkyl-4,4,4-trifluoroacetoacetates [6]
afforded up to 96% ee, while the reaction is poorly selective with
some aryl-substituted aromatic, benzylic, and particularly with ali-
phatic trifluoromethyl ketones [7–10].
⇑
Corresponding author. Fax: +41 44 632 1163.
E-mail address: baiker@chem.ethz.ch (A. Baiker).
Present address: On leave from Institute of Chemical Technology Prague,
1
Department of Organic Technology, Technicka 5, 16628 Prague 6, Czech Republic.
0021-9517/$ - see front matter Ó 2011 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2011.03.009

Z. Cakl et al. / Journal of Catalysis 280 (2011) 104–115
105
N
N
OH
H
HO
H
H
8
H
8
9
9
R
R
6´
6´
N
N
8 (S), 9 (R)
Cinchonidine
Quinine
R
H
R
H
8 (R), 9 (S)
Cinchonine
Quinidine
(CD)
(QN)
(CN)
(QD)
OMe
OMe
Scheme 1. Hydrogenation of 2,2,2-trifluoroacetophenone (1) on cinchona-modified Pt/Al2O3.
hemiketal with the solvent can diminish the ee at high conversion
but cannot cause inversion of the major enantiomer [22]. NMR
analysis also proved that no hemiketal was formed with the prod-
Our opinion is fundamentally different. On the basis of DFT cal-
culations [30,31] and in situ spectroscopic measurements [32,33],
we assume an N–H–O type interaction between the quinuclidine
N and the carbonyl O-atom even in a non-acidic medium. We attri-
bute the frequently unpredictable behavior of the Pt–cinchona sys-
uct
a,a,a,-trifluoromethyl alcohol, as expected in the presence of
the CF3 group.
A thoroughly investigated phenomenon is the unpredictable ef-
fect of acid additives and solvents in this reaction class. In the
hydrogenation of ethyl-4,4,4-trifluoroacetoacetate, replacement
of the solvent toluene with AcOH doubled the ee and addition of
TFA increased it further. In case of the 40-CF3 derivative of 1, how-
ever, carboxylic acids diminished the ee. For comparison, in the
hydrogenation of 1 and its aryl-substituted derivatives, replace-
ment of CD with CDꢀHCl increased the ee under all conditions
applied [5]. Obviously, the effect of carboxylic acids cannot simply
be attributed to protonation of the quinuclidine N of CD, but rather
H-bonding interactions have to be taken into account, as indicated
by IR measurements [23,24].
Recently, Bartók’s group reported numerous striking examples
on the inversion of the major enantiomer by the addition of the
strong acid TFA to the reaction mixture [10,25–28]. According to
their interpretation of the unexpected inversion, a ‘‘nucleophilic
intermediate complex’’ (N ? C@O type interaction) between the
alkaloid and ketone would be formed in the absence of TFA but
even in the presence of AcOH. In contrast, in the presence of TFA,
the protonated quinuclidine N of the alkaloid modifier would inter-
act with the carbonyl O-atom of the substrate via an N–H–O type
interaction. This assumption is rather astonishing, since an NMR
study proved the complete protonation of the quinuclidine N of
CD by 24 equivalents of AcOH [29]. The only additional effect of
TFA was the protonation of the quinoline N, the transformation
of which was negligible in AcOH. Bartók’s concept focusing on
the protonation of the quinuclidine N of the alkaloid by TFA also
cannot rationalize the unexpected inversions in alcohols [10,21],
whose solvents do not protonate CD.
tem in the hydrogenation of
a,a,a-trifluoromethyl ketones to
additional H-bonding interactions. This concept can rationalize
the special effect of carboxylic acids [6,24]. Here, we present our
novel observations on the role of the product in the hydrogenation
of 1 and the evolution of competing enantioselective cycles during
reaction.
2. Experimental section
2.1. Materials
2,2,2-Trifluoroacetophenone (1, 99%, Aldrich) was carefully dis-
tilled in vacuum before use. ( )-1-Phenyl-2,2,2-trifluoroethanol (2,
>98%, Fluka), (R)-2, (P99.0%, Fluka), (S)-2, (P99.0%, Fluka), cincho-
nidine (CD, 98% NT, Fluka), cinchonine (CN, P98% NT, Fluka), qui-
nine (QN, 99%, Fluka), quinidine (QD, >99%, Acros), toluene
(P99.7%, Fluka), trifluoroacetic acid (TFA, 99%, Acros), toluene D8
(99.94%, Cambridge Isotop Lab., INC.), and chloroform D (99.8%,
Armar Chemicals) were used as received. The 5 wt.% Pt/Al2O3
catalyst was purchased from Engelhard (Engelhard 4759).
2.2. Catalytic hydrogenations
The 5 wt.% Pt/Al2O3 catalyst was reduced at elevated tempera-
ture in a fixed-bed reactor prior to use. According to the standard
procedure, the catalyst was heated under flowing nitrogen up to
400 °C in 30 min, followed by a reduction in flowing hydrogen
for 60 min at the same temperature, and finally cooled down to

106
Z. Cakl et al. / Journal of Catalysis 280 (2011) 104–115
room temperature in flowing hydrogen in 30 min. At the end, the
freshly reduced catalyst was purged with nitrogen for 10 min
and then transferred immediately to the autoclave.
The enantioselective hydrogenations of 1 were carried out in a
25-ml stainless steel Parr autoclave equipped with a glass liner
with PTFE cover and a magnetic stirrer, and a valve for sample col-
lection or substrate injection. The pressure was controlled with a
constant pressure regulator valve (Buchi BPC 9901). Under stan-
dard conditions 21 1 mg of catalyst, 1.85 mmol of substrate and
For 19F,
D was set to 117.75 and 167.75 ms, respectively, with a
gradient recovery time of 100 ls. Sixteen scans were taken with
a recovery delay of 12–18 s (determined via a 19F T1 measurement),
and a total experimental time of ca. 2.5–4 h.
All the spectra were acquired using 32 k points and processed
with a line broadening of 1 Hz (1H) and 2 Hz (19F). The slopes of
the lines, m, were obtained by plotting their decrease in signal
intensity vs. G2 using a standard linear regression algorithm. Nor-
mally, 12–20 points have been used for regression analysis and
all of the data leading to the reported D values afforded lines whose
correlation coefficients were >0.9995. We estimate the experimen-
tal error in D values at 2%.
3.4
(1000 rpm) at 20 °C under a constant hydrogen pressure of 3 bars.
Hydrogenation of 2–4 mg (6.8–13.6 mol) of CD was performed
lmol of modifier in 5 ml of toluene were stirred magnetically
l
in a 50-ml stainless steel autoclave equipped with a glass liner
with PTFE cover and a magnetic stirrer. The pressure was con-
trolled with a constant pressure regulator valve (Buchi BPC
9901). Under standard conditions 42 2 mg of catalyst, 10 ml of
toluene were stirred magnetically (750 rpm) at 25 °C under a con-
stant hydrogen pressure of 3 bars. Hydrogenation of CD was fol-
The diffusion coefficients reported were determined using the
diffusion coefficient of HDO in D2O as
a
reference
slope of
(DHDO = 1.9 ꢁ 10ꢂ9 m2 sꢂ1),
which
afforded
a
1.976 ꢁ 10ꢂ4. The data obtained were used to calculate the D val-
ues of the samples according to Eqs. (1) and (2) [38].
ꢀ
ꢁ ꢀ
ꢁ
2
2
cx
cH
dx
ð
D
ꢂ d=3Þx
lowed in the presence of
1 under the same conditions as
f ¼
ð1Þ
ð2Þ
dHDO
ð
D
ꢂ d=3ÞHDO
described earlier for the hydrogenation of CD. 7.4 mmol of 1 was
added either before starting the reaction or after a time delay of
30 min in the transient experiments. The standard procedure was
applied also for the experiments in the presence of the hydrogena-
tion products of 1. Therein 3.7 mmol of either (R)-2, (S)-2 or 2 has
been added to the toluenic solution.
The conversion and enantioselectivity in the hydrogenation of 1
were determined by GC analysis, using an HP 6890 gas chromato-
graph equipped with a capillary column (CP-Chirasil-Dex CB,
DHDO
mHDO
Dx ¼ mx ꢃ
ꢃ f
From the D values, the hydrodynamic radii of the ions were ob-
tained via the Stokes–Einstein equation [39] and by introducing a
semi-empirical estimation of the c factor [40,41]. The solvent vis-
cosity (10ꢂ3 kg sꢂ1 mꢂ1) used for the calculation of the hydrody-
namic radii was
g (CHCl3) = 0.53.
25 m ꢁ 0.25 mm, i.d. 0.25
lm). Conditions: 80 °C for 2 min,
5 °C minꢂ1 to 120 °C, 120 °C for 5 min, 8 °C minꢂ1 to 180 °C, head
pressure 1.5 bar He. Retention times (min): 1 3.60, (S)-2 18.05,
(R)-2 18.32. Products were identified by comparison with authen-
tic samples and by GC–MS using an HP 6890 gas chromatograph
equipped with an HP 5-MS column (25 m ꢁ 0.20 mm, i.d.
2.3.2. 2D NOE measurements
1H,1H NOESY spectra were acquired on a Bruker Avance 700
spectrometer equipped with a multinuclear inverse probe. The
concentration of the sample was 10 mM. For the NOESY experi-
ment, the standard three-pulse (noesyph) [42] sequence was used
with phase cycling by the TPPI method [43]. A relaxation delay of
800 ms was applied and the mixing time was 600 ms. The number
of scans per increment was 32 (2 k data points), and 512 or 1 k
experiments were acquired in the second dimension. Total exper-
imental times were between 8–14 h. A QSINE weighting function
was used in each dimension prior to Fourier transformation into
a 2 k ꢁ 1 k data matrix.
0.33 lm) coupled to an HP 5973 mass spectrometer.
The identification of the hydrogenated products of CD and the
sample preparation is described elsewhere [34,35]. The estimated
error in the determination of the enantiomeric excess (ee) and
the diastereomeric excess (de) was about 0.5% (at above 10% con-
version) and that of the reaction rate (TOF) was in the range 10%.
2.3. NMR measurements
The 19F,1H HOESY spectra were acquired using the standard four-
pulse sequence (invhoesy) [44] on a Bruker Avance DRX-400 spec-
trometer equipped with a doubly tuned (1H, 19F) TXI probe. The con-
centration of the sample was 10 mM. A relaxation delay of 800 ms
was applied and the mixing time was 800 ms. Typically, 16 transients
were acquired into 2 k data points for each of the 512 or 1 k incre-
ments in t1. Total experimental times were between 8 and 12 h. A
QSINE(F1) and EM(F2) weighting function was used in each dimen-
sion prior to Fourier transformation into a 2 k ꢁ 1 k data matrix.
Nuclear magnetic resonance (NMR) data were recorded on Bru-
ker Avance 700 and DRX-400 spectrometers operating at the given
spectrometer frequency. The samples were measured as solutions
in CDCl3 at ambient temperature and in non-spinning mode. The
chemical shifts are expressed in parts per million (ppm) and refer-
enced to tetramethylsilane (TMS) for the 1H and 13C spectra, and to
CFCl3 for the 19F spectra [36]. Coupling constants J are given in Hz
as absolute values.
2.3.1. PGSE measurements
3. Results and discussion
All PGSE diffusion measurements were performed at a concen-
tration of 2 mM using the standard stimulated echo pulse sequence
[37] on a Bruker Avance DRX-400 spectrometer equipped with a
microprocessor-controlled gradient unit and an inverse multinu-
clear probe with an actively shielded Z-gradient coil. The shape
of the gradient pulse was rectangular, its duration d was 1.75 ms,
and its strength varied automatically in the course of the experi-
3.1. Enantioselective hydrogenation of 2,2,2-trifluoroacetophenone
(1): general aspects
We have mentioned in the introduction that in the hydrogena-
tion of trifluoromethyl ketones, strong interactions with the sol-
vent may blur the influence of other parameters. Because of the
extended H-bonding among the reaction components, it is impor-
tant to simplify the system with a weakly interacting solvent such
as toluene. Attractive features of this solvent are the weak polarity,
the weak H-bond donor and acceptor properties, and the reason-
ably good solubility of the cinchona alkaloids [45,46].
ments. In the 1H-PGSE experiments,
and 167.75 ms, respectively, with a gradient recovery time of
D [38] was set to 117.75 ms
100 ls. The number of scans per increment was 16 (in steps of
2–3% from 2–3% to 48–60%). A measurement of 1H T1 was carried
out before each diffusion experiment and the recovery delay set
to (at least) five times T1. Typical experimental times were 2–4 h.
In preliminary experiments, we reinvestigated the role of some
reaction parameters in toluene to provide a solid basis for the

Z. Cakl et al. / Journal of Catalysis 280 (2011) 104–115
107
present mechanistic study. Under standard conditions with
1.85 mmol of 1, variation of the amount of modifier in the range
1.7 to 13.4 mol gave an optimum of 44% ee to (R)-2 with 3.4 mol
A
l
l
CD (0.68 mmol/l). The reaction rate slightly decreased with
increasing amount of CD. These results are in line with earlier data
obtained under different conditions for 1 [4] and trifluoromethyl
cyclohexyl ketone [21].
A comparison of the enantioselective hydrogenations with the
racemic reaction (Fig. 1) revealed higher rate (‘‘ligand accelera-
tion’’) [27,47–51] by addition of CD and CN, but a small rate decel-
eration in the presence of the 60-methoxy derivatives QN and QD
(Scheme 1). Under different conditions but also in toluene, Bartók
and coworkers found rate acceleration with all four alkaloids in the
same order as in Fig. 1 (CD > CN > QN > QD) [25]. Clearly, the pres-
ence of the 60-methoxy function in QN and QD diminishes the reac-
tion rate, and also the enantioselectivity as will be shown later.
Next, the effect of substrate concentration was investigated in
the presence of CD. At conversions of 90–95% the ee increased from
30% to 52% by increasing the amount of 1 from 0.46 mmol to
5.5 mmol under otherwise standard conditions (not shown). In
addition, increasing substrate concentration had a negative effect
on the reaction rate, in agreement with former observations [4,21].
The behavior of all four modifiers is compared in Fig. 2 at two
different concentrations. The reaction rate in the hydrogenations
with CN, QN, and QD showed the same behavior as in the case of
CD: higher substrate concentrations were detrimental to the rate.
In the presence of CD and QN, a decrease in the substrate concen-
tration by a factor of four diminished the ee by ca. 10–15%, but the
major product remained (R)-2. In case of CN and QD, however, the
lower substrate concentration resulted in the inversion of the ma-
jor product from (S)-2 to (R)-2. (Note that the differences with QD
were small but well reproducible). This inversion has the astonish-
ing consequence that at low substrate concentration, all cinchona
alkaloids afford the same major product, (R)-2, independent of
the configurations at C8 and C9 (Scheme 1).
B
Bartók and coworkers used low substrate concentration in tolu-
ene and also found (R)-2 as the major product with all four alka-
loids [25]. Interestingly, this is the starting point of their
mechanistic model mentioned in the introduction [27], where they
assume that protonation of the quinuclidine N of the alkaloids by
TFA is necessary to arrive at the ‘‘usual’’ product distribution:
(R)-2 with CD and QN and (S)-2 with CN and QD. However, the data
Fig. 2. Effect of initial concentration of 1 on the enantioselectivity. A: 3.7 mmol of
1; B: 0.93 mmol of
(ꢃ after the abbreviations). Conditions: 21 mg catalyst,
3.4 mol CD, 5 ml toluene, 20 °C, and 3 bar, reaction time: 6 h.
1
l
in Fig. 2B demonstrate that inversion of the major enantiomer of 2
may simply result from a change in the substrate concentration;
addition of a strong acid is not a requirement. It is very unlikely
that a shift in the substrate concentration would lead to a different
reaction mechanism, i.e. a different type of substrate–modifier
interaction (N ? C@O or NAHAO@C type interaction [25]). This
‘‘unexpected inversion’’ upon changing the substrate concentration
may be related to the presence of dimers, trimers, and tetramers of
1 on the Pt surface (at high substrate concentration), as observed
recently using scanning tunneling microscopy under UHV condi-
tions [50] and confirmed by NMR in solution, in the absence of
Pt and modifier [25]. These observations fit to our concept that
the origin of the unusual behavior of the Pt–cinchona system in
the hydrogenation of
a,a,a-trifluoromethyl ketones is the exten-
sive H-bonding interaction among the reaction partners, including
also the product as shown in the next chapter.
3.2. The role of substrate–modifier–product interactions during
hydrogenation of 1
A specific feature of the hydrogenation of 1 is the enhancement
of enantioselectivity at low conversion. In this initial transient
period, the ee doubled in 1,2-dichlorobenzene [4]. The phenome-
non strongly depends on the reaction conditions and solvent, and
variation of the ee with conversion may also be minor [9]. We
Fig. 1. Effect of different modifiers on the conversion rate of 1 as compared with the
racemic hydrogenation (diamonds); CD (squares), CN (circles), QN (up triangles)
and QD (down triangles). Conditions: 21 mg catalyst, 3.4 lmol CD, 3.7 mmol 1, 5 ml
toluene, 20 °C, and 3 bar.

108
Z. Cakl et al. / Journal of Catalysis 280 (2011) 104–115
Fig. 4. Hydrogenation of equimolar mixtures of 1 and rac-2 using CD (squares), CN
Fig. 3. Conversion (squares), ee (spheres) and differential ee (down triangles) in the
(circles), QN (up triangles) or QD (down triangles). Reference line: CD⁄ (filled
hydrogenation of 1. Conditions 21 mg catalyst, 3.4
toluene, 20 °C, and 3 bar.
lmol CD, 5.5 mmol 1, 5 ml
squares, no rac-2). Conditions 21 mg catalyst, 3.4
1.85 mmol rac-2, 5 ml toluene, 20 °C, and 3 bar.
lmol CD, 1.85 mmol 1 and
nant product is the (S)-CDH6-A and the de is independent of the
conversion. The chemoselectivity, i.e. the ratio of CDH6-A to
CDH6-B (A/B) is very high, which indicates that saturation of the
homoaromatic ring of CD is a minor reaction.
In the presence of 1, the chemoselectivity dropped significantly
(Table 1). The initial rate did not increase but slightly decreased, in
found relatively small changes in toluene with CD and QN, but the
initial enhancement and the following drop in ee was always
clearly distinguishable. A typical example with CD is shown in
Fig. 3. The incremental ee indicates that the actual enantioselectiv-
ity reached a maximum below 20% conversion and dropped by
about 15% till the end of the reaction. Note that in the presence
of CN and QD, there was no detectable initial change of direction
in the ee–conversion curve, as shown in Fig. 2.
contrast to the effect of
a-ketoesters [35]. The most important
change is the inversion of the major diastereomer of CDH6-A from
(S) to (R). A similar inversion was observed earlier by changing the
solvent from toluene to acetic acid in the absence of any ketone
A feasible explanation for the drop in ee with conversion is the
interaction of the product
alkaloid modifier. Due to the presence of the CF3 group in
tion, the acidity of the OH function increases by several orders of
magnitude [52,53]. The polarity and H-bond donor ability ( ) of
a
,a,a-trifluoromethyl alcohol with the
substrate [34], or by addition of an
a-ketoester [35]. In addition,
a-posi-
the de and also the reaction rate (TOF) increased with increasing
conversion of CDH2. The conversion-dependent changes indicate
the strong influence of 2. An important point here is that the initial
de (deinit) is very small, although already this value is measured in
the presence of a small amount of 2, since a few percent conversion
is necessary to obtain reliable values by GC analysis. It indicates
that the de should be around zero and there is no preferred adsorp-
tion mode of CD, when the amount of 2 is close to zero. The mech-
anistic consequences of this observation will be discussed later.
The critical role of the product on the adsorption and hydroge-
nation of CD is confirmed by the experiment carried out in the
presence of equimolar amount of rac-2. The final de and the che-
moselectivity are the same as what was achieved in the presence
of 1, and the rate is also enhanced. The experiments in the presence
of (S)-2 or (R)-2 independently revealed that rac-2, (S)-2 and the
reaction mixture formed in the hydrogenation of 1 have similar ef-
fects on the rate of the hydrogenation of the quinoline ring and the
final des are also very similar. The influence of (R)-2 is different:
both the rate and the de of CDH2 hydrogenation are remarkably
higher. It indicates that (S)-2 interacts stronger with the alkaloid
and controls its hydrogenation even at a considerable excess of
the opposite enantiomer during hydrogenation of 1 (>50% ee to
(R)-2).
a
fluorinated alcohols are much higher, while the H-bond acceptor
ability (b) is remarkably lower than those of the corresponding
non-fluorinated alcohols.
In order to verify our hypothesis, the enantioselective hydroge-
nation of 1 was carried out in the presence of rac-2 with all four
alkaloids (Fig. 4). The initial racemic alcohol concentration was
subtracted before calculating the ee. In all these reactions, the ee
decreased from the beginning of the reaction. A comparison for
CD with and without rac-2 under otherwise identical conditions
is also shown to illustrate the shift in the ee–conversion lines.
The most interesting observation is that using CN as modifier
inversion of the ee was observed with increasing conversion. This
example suggests that the effect of increasing amount of product
is not simply a decrease of ee but rather a shift to a different mech-
anism that provides the opposite enantiomer in excess.
3.3. In situ diastereoselective hydrogenation of CD
A deeper insight into the nature of substrate–modifier–product
interactions was obtained by studying the chemo- and diastereose-
lective hydrogenation of the quinoline ring of CD (Scheme 2). This
transformation can be investigated under real reaction conditions,
in the presence of 1 and 2, and the product distribution provides
direct information on the adsorption mode of CD on the Pt surface
[34,35,54], which is not available by any other method.
The first step of the transformation is the fast hydrogenation of
the vinyl group of CD (Scheme 2), which is not investigated here.
The major characteristics of the hydrogenation of CDH2 in the ab-
sence of substrate or product are presented in Table 1. The domi-
Some important details of the analysis of the hydrogenation of
CD during the enantioselective hydrogenation of 1 are presented in
Figs. 5–7. In the experiment presented in Table 1, the hydrogena-
tion of 1 shows the same behavior (Fig. 5A) as in previous
experiments. After an initial increase, the ee decreased continu-
ously with conversion. The differential ee dropped from 60% to less
than 40% with conversion. The concomitant hydrogenation of the
alkaloid is shown in Fig. 5B, where the rate acceleration with time

Z. Cakl et al. / Journal of Catalysis 280 (2011) 104–115
109
2
2
2
2
S
R
6
6
6
2
2
2
12
Scheme 2. Chemo- and diastereoselective hydrogenation of 10,11-dihydrocinchonidine (CDH2).
Table 1
in the whole conversion range. This difference to Fig. 5B proves
that the increase in rate and de are due to the interaction of CD
with the increasing amount of 2 formed by hydrogenation of 1.
Finally, the nature of CD–2 interaction was investigated by the
addition of 3 eq. of TFA related to CD (Fig. 7). After an initial in-
crease, the ee in the hydrogenation of 1 remained almost constant;
only the differential ee indicates some small decrease above 85%
conversion. Also the rate and chemo- and diastereoselectivities in
the hydrogenation of the alkaloid are almost constant in the whole
concentration range (Table 1). Obviously, addition of only 3 eq. of
the strong acid TFA prevents the negative effect of 2 formed in
large excess (2/CD = up to 1088 molar ratio). In addition, the ee
at full conversion improved from 52% to 80% by the TFA additive.
These changes indicate that the interaction of CD and 2 involves
base–weak acid type interactions, respectively, which are hindered
by the interaction with the strong acid TFA.
Rates and selectivities in the hydrogenation of CDH2.a
Reaction components TOFinit (hꢂ1
)
TOF (hꢂ1
)
deinit (%) de (%) A/B
CD
CD + 1
CD + rac. 2
CD + (S)-2
CD + (R)-2
CD + 1 + TFA
1.0
0.8
2.1
1.5
6.2
1.5
1.0
1.2
1.6
1.3
5.9
1.4
10 (S)
7 (R)
10 (S)
31
45 (R) 12
45 (R) 12
42 (R) 10
72 (R) 22
78 (R) 34
37 (R)
38 (R)
65 (R)
69 (R)
a
Conditions: 42 mg catalyst, 6.8–13.6 lmol CD, 3.7 mmol of 1, (R)-2, (S)-2, or rac.
2, 10 ml toluene, 25 °C, and 3 bar. CD: TFA = 1:3. TOFinit was determined at
conversions below 10%, TOF in the conversion range 40–80%, deinit and de were
calculated similarly.
(conversion) is clearly observable. The large enhancement of the de
is also well observable and the differential de indicates stabiliza-
tion at around 50% conversion of 1. All these changes in the rate
and diastereoselectivity mirror the effect of increasing product
concentration.
The negative effect of rac-2 on the enantioselectivity is shown
in Fig. 6. The hydrogenation of CD was started in the presence of
rac-2 and after 30 min four eq. of 1 was added. The first measured
ee value was 49% (Fig. 6A), almost 10% less than what was achieved
without the addition of 2 (Fig. 5A). In addition, the ee decreased in
the whole conversion range without any initial increase as seen in
Fig. 5A. Hydrogenation of CD is affected by the addition of racemic
2, too (Fig. 6B). In particular, the significant increase especially of
the rate and the de to (R)-CDH6-A with conversion was not ob-
served here; the de varied in a relatively narrow range of 40–45%
3.4. NMR investigations
In order to gain more insight into the modifier–product interac-
tion during hydrogenation of 1 on Pt, the catalytic study was com-
pleted with NMR experiments. In these experiments, CDCl3 was
used as a solvent due to the low solubility of CD in toluene. The
negligible effect of solvent was proved in separate experiments
using the more soluble QN. A comparison of the 1H NMR spectra
of QN measured in deuterated toluene and CDCl3 revealed no
significant differences, implying that the conformations of QN in
both solvents are comparable (Supporting information).
The interaction of CD and QN with 1 and 2 was investigated by
pulsed gradient spin-echo (PGSE) NMR diffusion experiments.
Since the catalytic experiments revealed that the observed

110
Z. Cakl et al. / Journal of Catalysis 280 (2011) 104–115
A
B
Fig. 6. Conversion (squares), ee (spheres) and differential ee (triangles) in the
hydrogenation of 1 injected after 30 min to hydrogenation of CD started with
0.25 eq. of racemic 2 (A). The conversion (squares), de (circles), and differential de
(triangles) in the hydrogenation of CDH2 is shown in part B. Conditions: 42 mg
Fig. 5. Conversion (squares), ee (spheres) and differential ee (triangles) in the
hydrogenation of 1 in the presence of CD (A). The conversion (squares), de (circles)
and differential de (triangles) of the hydrogenation of CDH2 is shown in part B.
Conditions: 42 mg catalyst, 6.8
3 bar.
lmol CD, 7.4 mmol 1, 10 ml toluene, 25 °C, and
catalyst, 6.8 lmol CD, 7.4 mmol 1, 1.85 mmol rac-2, 10 ml toluene, 25 °C, and 3 bar.
catalytic behavior is dominated by the presence of (S)-2 (Table 1),
we considered this enantiomer for the diffusion study. Determina-
tion of the diffusion coefficients D allow to calculate the hydrody-
namic radii rH via the Stokes–Einstein equation (Eq. (3)) [39],
kT
D ¼
ð3Þ
c
p
grH
where k is the Boltzmann constant, T the absolute temperature, c
the friction constant,
namic radius.
g the solution viscosity, and rH the hydrody-
The comparison of the calculated hydrodynamic radii of (S)-2
from the experiments carried out with and without CD revealed
a significant bigger radius in the former case (Table 2). A similar ef-
fect was observed when QN was used as the modifier. In contrast,
no effect of the cinchona modifier on the diffusion behavior of 1
was detectable.
The increase of the hydrodynamic radius demonstrates the
interactions between the modifier and (S)-2. The fact that only
small changes of D and rH were observed can be rationalized by
assuming a fast equilibrium between solvated and complexed
(S)-2, whereby the equilibrium lies on the side of the solvated
(S)-2. The absence of significant changes of D and rH in the exper-
iments with 1 indicates that the interaction between 1 and the
Fig. 7. Conversion (squares), ee (spheres), and differential ee (triangles) in the
hydrogenation of 1 (filled symbols) in the presence of 3 eq. of TFA related to CD. The
concomitant conversion of CDH2 is shown with empty squares. Conditions: 42 mg
catalyst, 6.8 lmol CD, 7.4 mmol 1, 10 ml toluene, 0 °C, and 10 bar.

Z. Cakl et al. / Journal of Catalysis 280 (2011) 104–115
111
Table 2
tween the quinuclidine N and 2. In both cases, the shielding effect
of the N lone pair is reduced probably leading to the observed
deshielding of H8.
Diffusion coefficients D (10–10 m2 sꢂ1) and hydrodynamic radii rH derived
from 1H and 19F diffusion experiments for (S)-2, 1 + modifier in CDCl3 at
298 Ka (mean values).
The presence of either (R)-2 or (S)-2 lead as well to an increase
2
of the vicinal 3JH
coupling constant, which is sensitive to the
8H9
D
rH (Å)
rotation of the quinuclidine moiety around the C8AC9 bond (Table
3). In comparison, the interaction of (S)-2 seems to introduce a lar-
ger rotation around the C8AC9 bond than (R)-2 as evidenced by its
(S)-2
(S)-2 + CD
(S)-2 + QN
17.05
16.37
16.27
2.4
2.5
2.5
bigger 3JH . Consequently, the corresponding 1H, 1H NOESY
8H9
1
experiments on samples of both modifiers with (R)-2 and (S)-2
showed NOE contacts between H1 and H8, H9, H10, H14 and
H16, and between H5 and H8, H9 and H16, respectively (Support-
ing information Figs. 6-1 to 6-4), indicating an Open(3) conforma-
tion of the modifier [57]. In addition, the strong NOE contact
between H1 and H10 supports close proximity of the latter to
the quinoline ring.
1
17.83
17.88
2.3
2.3
1 + CD
a
g
(CHCl3) = 0.53 ꢁ 10ꢂ3 kg sꢂ1 mꢂ1
.
In order to locate possible interactions between the modifier
and 2, 1H, 19F HOESY experiments were carried out. Besides the
intramolecular interactions of the CF3 group with the aromatic
ortho protons of 2 and its aliphatic proton on the C atom bearing
the OH group, a strong intermolecular NOE contact with the pro-
tons of the methoxy group on C60 in QN was observed (Fig. 10)
for both enantiomers of 2. In the case of CD and 2, there is no clear
evidence of an analogous interaction of the CF3 group with H6 due
to the presence of potentially overlapping NOE cross peaks to the
aromatic ortho protons of 2. However, a comparison of the chem-
ical shift differences (in the absence and presence of 2) of the pro-
tons in CD and the protons in QN suggests that 2 interacts with
both modifiers in a similar way (Figs. 8 and 9). The interaction of
the CF3 group with the methoxy protons in QN and the similarities
in the interaction of 2 with both modifiers may result from hydro-
gen bonding between the OH group of 2 and the quinuclidine N
atom.
modifier is much weaker than that between 2 and the modifier,
and the complexation equilibrium remains essentially completely
on the side of the non-interacting 1.
These results were corroborated by 1D 19F NMR measurements,
where the addition of modifier to the solution of 1 did not induce
any shift of the signal of the CF3 group. In the case of 2, however, a
significant shift by 0.17 ppm of the signal was observed (QN + (S)-
2, Supporting information, Figs. 5-1 and 5-2). The phenomenon has
already been reported by Abid et al., who used cinchona alkaloids
to determine the ee of a-trifluoromethylated-hydroxyl compounds
by 19F NMR spectroscopy [55].
To elucidate the interactions between 2 and the modifier, 1D 1H
NMR spectra of 10 mM solutions of QN or CD + (R)-2 or (S)-2 (mo-
lar ratio 1:1) were recorded in CDCl3. The graphical comparison to
the corresponding spectra of pure QN or CD in CDCl3 is shown in
Figs. 8 and 9, respectively. Major shift differences in both modifiers
were observed for the protons H8–H11 (Scheme 3). The shielding
of H9 and H11 and deshielding of H10 indicates a counterclockwise
rotation of the quinuclidine moiety around the C8AC9 bond
(Scheme 1), which places H9 and H11 over the aromatic ring of
quinoline and H10 next to it [56]. The possible reasons for the
deshielding of H8 are manifold. It may be caused partly by the
rotation of the quinuclidine fragment or by a hydrogen bond be-
In summary, based on our NMR data, we suggest a significant
interaction of the hydrogenation product 2 with the modifier. In
the complexation product, the HO function of 2 seems to be resid-
ing close to the quinuclidine N atom, presumably due to hydrogen
bonding of the OH group with the latter. This H-bonding interac-
tion between the alcohol and the modifier induces some conforma-
Fig. 8. Chemical shift differences in 1H NMR spectra of QN in the presence of (R)-2 and (S)-2 in CDCl3 (molar ratio 1:1). Numbering of protons is shown in Scheme 3.

112
Z. Cakl et al. / Journal of Catalysis 280 (2011) 104–115
Fig. 9. Chemical shift differences in 1H NMR spectra of CD in the presence of (R)-2 and (S)-2 in CDCl3 (molar ratio 1:1). Numbering of protons is shown in Scheme 3.
Scheme 3. CD and 2 proton numbering.
Table 3
Vicinal 3JH
coupling constants for quinine (QN) and cinchonidine (CD) in the
Fig. 10. 19F, 1H HOESY spectrum of (S)-2 with QN in CDCl3; molar ratio 1:1.
8H9
presence of (R)-2 and (S)-2 in CDCl3.a
QN
CD
Modifier
Modifier + (R)-2
Modifier + (S)-2
2.4
2.7
3.6
3.0
3.9
4.1
3.5. Mechanistic considerations
Three fundamentally different mechanistic concepts have been
brought up to interpret the stereochemical outcome of the hydro-
a
The accuracy of measured coupling constants is 0.1 Hz.
genation of
a,a,a-trifluoromethyl ketones on cinchona-modified
Pt. The most recent model proposed by Bartók and coworkers
[25–27] and its contradiction to some fundamental experimental
observations have already been discussed in the previous chapters.
The essence of our model developed for the hydrogenation of
activated ketones is a single attractive interaction, an NAHAO type
bond, between the quinuclidine N of the modifier and the keto
carbonyl group of the substrate [17]. The model is based on an
in situ spectroscopic observation [32]. The adsorption of the ketone
on the Pt surface is assumed to be controlled by electronic
interactions between the alkaloid and the activating (electron
tional changes in the modifier, the most prominent being the rota-
tion around the C8AC9 bond. This interaction is expected to be rel-
atively strong due to the acidity and increased H-bond donor
ability of the
a,a,a-trifluoromethyl alcohol 2 [52,53]. There is an
indication to an additional interaction between an H atom of the
C60-methoxy group of QN and the CF3 group of 2, but an analogous
interaction with the C6’ H atom of CD could not be evidenced.

Z. Cakl et al. / Journal of Catalysis 280 (2011) 104–115
113
withdrawing) function of the ketone. A direct involvement of the
quinoline ring of the modifier was rejected based on missing steric
effects in the hydrogenation of bulky -ketoesters [58]. The unique
behavior of the hydrogenation of -fluorinated ketones is attributed
to additional H-bonding interactions; the clearest example among
them is the sometimes striking changes in enantioselectivity in-
duced by addition of TFA [6,24].
rac
R
2
a
a
2
S
R
R
McBreen and coworkers described various prochiral complexes
responsible for enantioselection [50,59–61]. A common point in
these models, based on STM studies under ex situ conditions, is a
double substrate–modifier interaction involving not only the
quinuclidine N atom but also one or two H atoms of the quinoline
ring. This concept is, however, in contrast to the in situ investiga-
tion of the hydrogenation of CD. The results in Table 1 show that
during hydrogenation in toluene in the absence of substrate or
other additive, the adsorption mode of CD on Pt is slightly
dominated by the pro(S) geometry, leading to the formation of
(S)-CDH6-A upon hydrogenation. This adsorption mode is illus-
trated in Scheme 4, left. Notice, however, the very small excess:
at 10% de the product mixture corresponds to a 55:45 ratio. Addi-
tion of the substrate (1) barely changes this ratio. The minor initial
de of 7% to (R)-CDH6-A is probably due to the presence of small
amounts of the stronger interacting (S)-2, as discussed in the pre-
vious chapter. This result is in good agreement with those of ex situ
NMR diffusion experiments indicating weak interaction of the
alkaloid with 1 and stronger interaction with 2. By the end of the
reaction or by adding only (S)-2, the de increases to 42–45% (R)-
CDH6-A, which corresponds to a dominant pro(R) adsorption mode
of CD (Scheme 4, right). The major difference between the two
adsorption modes is the inverted position of the quinoline ring.
McBreen’s model implicitly assumes the pro(R) adsorption mode
of CD, which would enable a second interaction with the substrate.
In contrast, at the early stage of the hydrogenation of 1 the data in
Table 1 do not show a significant preference to this adsorption
2
2
Scheme 5. Multiple cycle mechanism in the hydrogenation of 2,2,2-trifluoroace-
tophenone (1) to 1-phenyl-2,2,2-trifluoroethanol (2) on the Pt–CD system.
mode. Obviously, the diastereoselective hydrogenation of CD dis-
proves the importance of a second interaction between 1 and an
aromatic H of CD.
3.6. Multiple cycle mechanism in the enantioselective hydrogenation
of 1
We propose a multiple cycle mechanism for the interpretation
of the unique behavior of a,a,a-trifluoromethyl ketones, compared
with those of other activated ketones. The starting point is the two-
cycle mechanism commonly accepted for chirally modified metals
[47,62]. The basic assumption is that not all surface metal sites are
chirally modified and the unmodified sites provide racemic
product (route a in Scheme 5). The contribution of this route is
important at low modifier concentration that leads to incomplete
surface modification and lower (overall) ee.
On the chirally modified surface, the substrate–modifier inter-
action results in the preferential formation of one enantiomer, in
the present case (R)-2 (route b in Scheme 5). This route is impor-
N
N
H
H
OH
H
O
OH
N
H
H
F3C
N
H
H
H
H
H
H
H
H
N
N
HO
HO
N
N
H
H
S
R
Scheme 4. Hydrogenation of 10,11-dihydrocinchonidine (CDH2) on the Pt surface alone (left) and in the presence of (S)-2 (right) in toluene.

114
Z. Cakl et al. / Journal of Catalysis 280 (2011) 104–115
tant only at the early stage of the reaction, until the concentration
of the -fluorinated alcohol is very low.
4. Conclusions
a
To interpret our observations in the enantioselective hydroge-
nation of 1 on cinchona-modified Pt, we have to introduce two
more cycles. The enantioselective hydrogenation of 1 and the
in situ diastereoselective hydrogenation of CD in the presence of
1 and 2 revealed a strong interaction between the alkaloid and
(S)-2. Importantly, this interaction is remarkably stronger than
the alkaloid–1 interaction (route b). Parallel to the formation of
the minor enantiomer (S)-2, the alkaloid is transformed to an alka-
loid–(S)-2 complex (Scheme 4, right) and this complex becomes
the actual modifier (route c in Scheme 5). This switch in the mech-
anism is reflected by lower ees to (R)-2 with increasing conversion,
and in case of CN even the major enantiomer is inverted (Fig. 4). It
is reasonable to assume that this deviation from the general behav-
ior of activated ketones is due to the acidic and strong H-bond
There have been several important experimental observations
in the enantioselective hydrogenation of
a,a,a-trifluoromethyl
ketones on cinchona-modified Pt that could not be rationalized
by the existing mechanistic models. In the present study, we focus
on the special role of the product, which is yet unprecedented in
the hydrogenation of other activated ketones. We propose a new
mechanistic concept for the enantioselective hydrogenation of
2,2,2-trifluoroacetophenone (1). This multiple cycle mechanism in-
cludes four competing cycles, one racemic and three enantioselec-
tive routes to 1-phenyl-2,2,2-trifluoroethanol (2). The primary
origin of enantioselection is (i) an NAHAO bond between the
quinuclidine N of the alkaloid and the carbonyl O of the substrate
and (ii) the chiral environment provided by the alkaloid. The un-
ique behavior of
a-fluorinated ketones is attributed to additional
donor character of the
a
-trifluoromethyl alcohol product [52,53].
H-bonding interactions of the quinuclidine N with the acidic prod-
uct or the acid additive. The importance of these cycles in enantio-
selection is determined by the relative strength of interactions:
CD–1 < CD–(S)-2 < CD–TFA. Since the basis of these additional cy-
cles is the acidic character of the fluorinated alcohol product, the
mechanistic model is expected to be valid also for other represen-
The acidic OH function should interact strongly with the basic
quinuclidine N of the alkaloid, and this assumption is supported
by the (ex situ) NMR measurements.
A less clear point is the second, weaker interaction between
(S)-2 and the alkaloid, which is not indicated in Scheme 4. The
NMR study confirmed the interaction involving the F atoms of
(S)-2 and the H atoms of the 60-methoxy group of QN, but a similar
interaction involving the H atoms at C6’ could not be clarified
unambiguously. Due to steric and electronic effects, the strength
of these interactions should be remarkably different for CD and
QN and may be related to the different efficiency of these modifiers
(see Figs. 1 and 2 and Refs. [25–27]). An important point is that the
remarkably stronger interaction of the (S)-2 isomer with CD, com-
pared with the (R)-2–CD interaction, proves unambiguously that
the second interaction is important and sterically demanding. In
case of a single interaction of the acidic OH function of 2 with
the quinuclidine N atom, no difference between the two enantio-
mers would exist. In this respect, a clear limitation of the NMR
study is that we could not use the same solvent, which was applied
in catalytic studies.
tatives of a-fluoromethyl ketones.
Acknowledgment
Financial support of this work by the Swiss National Science
Foundation is kindly acknowledged.
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.jcat.2011.03.009.
References
[1] Y. Orito, S. Imai, S. Niwa, N.G. Hung, J. Synth. Org. Chem. 37 (1979) 173.
[2] H.U. Blaser, Tetrahedron: Asymmetry 2 (1991) 843–866.
[3] G. Webb, P.B. Wells, Catal. Today 12 (1992) 319–337.
[4] T. Mallat, M. Bodmer, A. Baiker, Catal. Lett. 44 (1997) 95–99.
[5] M. von Arx, T. Mallat, A. Baiker, Tetrahedron: Asymmetry 12 (2001)
3089–3094.
[6] M. von Arx, T. Mallat, A. Baiker, Catal. Lett. 78 (2002) 267–271.
[7] K. Balàzsik, B. Török, K. Felföldi, M. Bartók, Ultrason. Sonochem. 5 (1999)
149–155.
[8] R. Hess, S. Diezi, T. Mallat, A. Baiker, Tetrahedron: Asymmetry 15 (2004)
251–257.
[9] T. Varga, K. Felföldi, P. Forgó, M. Bartók, J. Mol. Catal. A: Chem. 216 (2004)
181–187.
[10] A. Vargas, F. Hoxha, N. Bonalumi, T. Mallat, A. Baiker, J. Catal. 240 (2006)
203–212.
[11] H.U. Blaser, H.P. Jalett, D.M. Monti, A. Baiker, J.T. Wehrli, Stud. Surf. Sci. Catal.
67 (1991) 147–155.
[12] E. Talas, J.L. Margitfalvi, Chirality 22 (2010) 3–15.
[13] M. Studer, H.U. Blaser, C. Exner, Adv. Synth. Catal. 345 (2003) 45–65.
[14] T. Bürgi, A. Baiker, Acc. Chem. Res. 37 (2004) 909–917.
[15] D.Y. Murzin, P. Maki-Arvela, E. Toukoniitty, T. Salmi, Catal. Rev. Sci. Eng. 47
(2005) 175–256.
Addition of the strong acid TFA (pKa = 0.2) is commonly used to
improve the ee in the hydrogenation of
a-fluorinated ketones
[6,24]. TFA replaces the weak acid (S)-2 (pKa = 11.9 [63]), and the
base–strong acid type alkaloid–TFA complex provides higher ee
to (R)-2 (Fig. 7). Due to the large difference in acidity, this cycle
(route d in Scheme 5) is dominant already in the presence of only
three equivalents of TFA related to CD, despite the large excess of
(S)-2 in the reaction mixture. Comparing all cases depicted in
Scheme 5, the pro(R) adsorption mode of CDH2 is the most domi-
nant in case of the TFA–CD complex, as indicated by the highest
de of 78% to (R)-CDH6-A (Table 1). The structures of TFA–CD com-
plexes have been described elsewhere [64]. Note that similar struc-
tures have to be considered in case of AcOH, but those complexes
are less stable due to the weaker acidity, compared with TFA [24].
We assume that the aforesaid considerations are valid also to
other
a,a,a-trifluoromethyl ketones, since the origin of their un-
[16] M. Bartók, Curr. Org. Chem. 10 (2006) 1533–1567.
[17] T. Mallat, E. Orglmeister, A. Baiker, Chem. Rev. 107 (2007) 4863–4890.
[18] F. Zaera, J. Phys. Chem. C 112 (2008) 16196–16203.
[19] M. von Arx, T. Mallat, A. Baiker, in: D.C. Sherrington, A.P. Kybett (Eds.),
Supported Catalysts and their Applications, The Royal Soc. Chem., Cambridge,
2001, pp. 247–254.
[20] M. von Arx, T. Mallat, A. Baiker, J. Catal. 193 (2000) 161–164.
[21] K. Felföldi, T. Varga, P. Forgó, M. Bartók, Catal. Lett. 97 (2004) 65–70.
[22] M. von Arx, T. Mallat, A. Baiker, J. Catal. 202 (2001) 169–176.
[23] D. Ferri, T. Bürgi, A. Baiker, J. Chem, Soc. Perkin Trans. 2 (1999) 1305–1311.
[24] M. von Arx, T. Bürgi, T. Mallat, A. Baiker, Chem. Eur. J. 8 (2002) 1430–1437.
[25] K. Szöri, K. Balazsik, S. Cserenyi, G. Szollosi, M. Bartok, Appl. Catal. A: Gen. 362
(2009) 178–184.
ique behavior is the acidic (H-bond donor) character of the product
alcohol and the H-bond acceptor ability of the CF3 group. Note that
in the hydrogenation of other activated ketones, including the
mostly studied
a-ketoesters, the product is not acidic and thus
its involvement in the mechanism is not expected. This possibility
was excluded already in the nineties in case of ethyl pyruvate.
Margitfalvi and Hegedüs [65] found that ‘‘the form of the optical
yield vs. conversion dependencies was not altered by addition of
(R)-ethyl lactate’’, in contrast to our present observations. Further-
more, the biggest effect on the ee caused by the addition of ethyl
lactate was only 1.1%, barely exceeding the probable experimental
error. Clearly, ethyl lactate had no significant influence on the
enantioselection.
[26] G. Szöllösi, S. Cserènyi, M. Bartók, Catal. Lett. 134 (2010) 264–269.
[27] G. Szöllösi, S. Cserényi, I. Bucsi, T. Bartók, F. Fülöp, M. Bartók, Appl. Catal. A:
Gen. 382 (2010) 263–271.
[28] M. Bartók, Chem. Rev. 110 (2010) 1663–1705.

Z. Cakl et al. / Journal of Catalysis 280 (2011) 104–115
115
[29] W.-R. Huck, T. Bürgi, T. Mallat, A. Baiker, J. Catal. 200 (2001) 171–180.
[30] A. Vargas, T.B. Bürgi, A. Baiker, J. Catal. 222 (2004) 439–449.
[31] A. Vargas, T. Bürgi, A. Baiker, J. Catal. 226 (2004) 69–82.
[32] N. Bonalumi, T. Bürgi, A. Baiker, J. Am. Chem. Soc. 125 (2003) 13342–13343.
[33] I.C. Lee, R.I. Masel, J. Phys. Chem. B 106 (2002) 368–373.
[34] E. Schmidt, W. Kleist, F. Krumeich, T. Mallat, A. Baiker, Chem. Eur. J. 16 (2010)
2181–2192.
[35] E. Schmidt, T. Mallat, A. Baiker, J. Catal. 272 (2010) 140–150.
[36] R.K. Harris, E.D. Becker, S.M.C. De Menezes, R. Goodfellow, P. Granger, Pure
Appl. Chem. 73 (2001) 1795–1818.
[50] M.A. Laliberte, S. Lavoie, B. Hammer, G. Mahieu, P.H. McBreen, J. Am. Chem.
Soc. 130 (2008) 5386–5387.
[51] H.U. Blaser, M. Studer, Accounts Chem. Res. 40 (2007) 1348–1356.
[52] J.P. Begue, D. Bonnet-Delpon, B. Crousse, Synlett (2004) 18–29.
[53] R.E. Ramirez, C. Garcia-Martinez, F. Mendez, J. Phys. Chem. A 113 (2009)
10753–10758.
[54] E. Schmidt, F. Hoxha, T. Mallat, A. Baiker, J. Catal. 274 (2010) 117–120.
[55] M. Abid, B. Török, Tetrahedron: Asymmetry 16 (2005) 1547–1555.
[56] R.S. Macomber, A Complete Introduction to Modern NMR Spectrsocopy, Wiley,
New York, 1998.
[37] E.O. Stejskal, J.E. Tanner, J. Chem. Phys. 42 (1965) 288–292.
[38] P.S. Pregosin, Prog. Nucl. Magn. Reson. Spectrosc. 49 (2006) 261–288.
[39] P.S. Pregosin, P.G.A. Kumar, I. Fernandez, Chem. Rev. 105 (2005) 2977–2998.
[40] A. Gierer, K. Wirtz, Z. Naturfors, Sect. A: J. Phys. Sci. 8 (1953) 532–538.
[41] H.C. Chen, S.H. Chen, J. Phys. Chem. 88 (1984) 5118–5121.
[42] S. Macura, R.R. Ernst, Mol. Phys. 41 (1980) 95–117.
[43] S.J. Archer, D.M. Baldisseri, D.A. Torchia, J. Magn. Reson. 97 (1992) 602–606.
[44] P.L. Rinaldi, J. Am. Chem. Soc. 105 (1983) 5167–5168.
[45] Z. Ma, F. Zaera, J. Phys. Chem. B 109 (2005) 406–414.
[57] T. Bürgi, A. Baiker, J. Am. Chem. Soc. 120 (1998) 12920–12926.
[58] S. Diezi, S. Reimann, N. Bonalumi, T. Mallat, A. Baiker, J. Catal. 239 (2006)
255–262.
[59] S. Lavoie, G. Mahieu, P.H. McBreen, Angew. Chem. Int. Ed. 45 (2006)
7404–7407.
[60] S. Lavoie, M.A. Laliberte, I. Temprano, P.H. McBreen, J. Am. Chem. Soc. 128
(2006) 7588–7593.
[61] V. Demers-Carpentier, M.A. Laliberte, S. Lavoie, G. Mahieu, P.H. McBreen, J.
Phys. Chem. C 114 (2010) 7291–7298.
[46] L. Mink, Z. Ma, R.A. Olsen, J.N. James, D.S. Sholl, L.J. Mueller, F. Zaera, Top. Catal.
48 (2008) 120–127.
[47] M. Garland, H.U. Blaser, J. Am. Chem. Soc. 112 (1990) 7048.
[48] E. Toukonitty, D.Y. Murzin, J. Catal. 241 (2006) 96–102.
[49] M. Studer, S. Burkhardt, H.U. Blaser, Chem. Commun. (1999) 1727–1728.
[62] J.T. Wehrli, Thesis, ETH, Zürich, 1989.
[63] R. Stewart, R. van der Linden, Can. J. Chem.: Rev. Can. Chim. 38 (1960)
399–406.
[64] W.-R. Huck, T. Bürgi, T. Mallat, A. Baiker, J. Catal. 205 (2002) 213–216.
[65] J.L. Margitfalvi, M. Hegedüs, J. Mol, Catal. A: Chem. 107 (1996) 281–289.