The origin of chemo- and enantioselectivity in the hydrogenation of diketones on platinum

Article abstract of DOI:10.1021/ja057586f

In the Pt-catalyzed hydrogenation of 1,1,1-trifluoro-2,4-diketones, addition of trace amounts of cinchonidine, O-methyl-cinchonidine, or (R,R)-pantoyl-naphthylethylamine induces up to 93% ee and enhances the chemoselectivity up to 100% in the hydrogenation of the activated carbonyl group to an OH function. A combined catalytic, NMR and FTIR spectroscopic, and theoretical study revealed that the two phenomena are coupled, offering the unique possibility for understanding the substrate-modifier-metal interactions. The high chemo- and enantioselectivities are attributed to the formation of an ion pair involving the protonated amine function of the chiral modifier and the enolate form of the substrate. DFT calculations including the simulation of the interaction of a protonated amine with the enolate adsorbed on a Pt 31 cluster revealed that only the C-O bond next to the CF₃ group of the substrate is in direct contact with Pt and can be hydrogenated. The present study illustrates the fundamental role played by the metal surface and indicates that also the enol form can be the reactive species in the hydrogenation of the activated ketone on chirally modified Pt.

Full text of DOI:10.1021/ja057586f

Published on Web 03/02/2006
The Origin of Chemo- and Enantioselectivity in the
Hydrogenation of Diketones on Platinum
Simon Diezi, Davide Ferri, Angelo Vargas, Tamas Mallat, and Alfons Baiker*
Contribution from the Department of Chemistry and Applied Biosciences, ETH Zurich,
H o¨ nggerberg, HCI, CH-8093 Zurich, Switzerland
Received November 7, 2005; E-mail: baiker@chem.ethz.ch
Abstract: In the Pt-catalyzed hydrogenation of 1,1,1-trifluoro-2,4-diketones, addition of trace amounts of
cinchonidine, O-methyl-cinchonidine, or (R,R)-pantoyl-naphthylethylamine induces up to 93% ee and
enhances the chemoselectivity up to 100% in the hydrogenation of the activated carbonyl group to an OH
function. A combined catalytic, NMR and FTIR spectroscopic, and theoretical study revealed that the two
phenomena are coupled, offering the unique possibility for understanding the substrate-modifier-metal
interactions. The high chemo- and enantioselectivities are attributed to the formation of an ion pair involving
the protonated amine function of the chiral modifier and the enolate form of the substrate. DFT calculations
including the simulation of the interaction of a protonated amine with the enolate adsorbed on a Pt 31
3
cluster revealed that only the C-O bond next to the CF group of the substrate is in direct contact with Pt
and can be hydrogenated. The present study illustrates the fundamental role played by the metal surface
and indicates that also the enol form can be the reactive species in the hydrogenation of the activated
ketone on chirally modified Pt.
nated alcohols,2
1-24
the compounds of which have attracted great
Introduction
25,26
attention in past years in agro- and pharmaceutical chemistry.
Although the ee’s are frequently lower than those achieved in
homogeneous catalysis,
Pt/Al2O3) and the cinchona alkaloids used as modifiers are cheap
and commercially available, and the technical handling is not
demanding. Thoroughly investigated substrate groups are aro-
matic trifluoromethyl ketones
sessing an additional (nonactivated) carbonyl group in the
At best, 92% ee was achieved in the hydrogena-
tion of 2,2,2-trifluoroacetophenone and 96% ee in the hydro-
Various strategies have been developed for the creation of
chiral surface structures that combine high catalytic activity and
stereochemical control.1-5 Among them, imparting chirality to
a catalytic active metal surface by adsorption of chiral organic
molecules (chiral modifiers) seems to be one of the most
27,28
the supported metal catalyst (usually
6-8
promising. This heterogeneous catalytic approach even allows
2
9,30
and aliphatic ketones pos-
continuous process operation.9
-11
In the hydrogenation of
CdO and CdC functions, chirally modified supported metal
31-33
â-position.
catalysts represent a successful approach with synthetic
30
1
2-20
potential.
3
4
genation of ethyl-4,4,4-trifluoroacetoacetate. The best chiral
modifiers for Pt are the naturally occurring cinchonidine (CD),
its O-methyl-derivative (MeOCD), and the synthetic modifier
An emerging field is the Pt-catalyzed enantioselective hy-
drogenation of R,R,R-trifluoromethyl ketones to chiral fluori-
(
R,R)-pantoyl-naphthylethylamine ((R,R)-PNEA). Recently, the
(
(
(
(
(
(
(
1) Baiker, A. Curr. Opin. Solid State Mater. Sci. 1998, 3, 86.
2) Bein, T. Curr. Opin. Solid State Mater. Sci. 1999, 4, 85.
3) Sherrington, D. C. Catal. Today 2000, 57, 87.
latter modifier afforded better than 99% chemoselectivity and
4) Davis, M. E. Top. Catal. 2003, 25, 3.
(21) Bal a´ zsik, K.; T o¨ r o¨ k, B.; Felf o¨ ldi, K.; Bart o´ k, M. Ultrason. Sonochem. 1999,
5, 149.
(22) von Arx, M.; Mallat, T.; Baiker, A. Angew. Chem., Int. Ed. 2001, 40, 2302.
(23) Varga, T.; Felf o¨ ldi, K.; Forg o´ , P.; Bart o´ k, M. J. Mol. Catal. A: Chem.
2004, 216, 181.
(24) Felf o¨ ldi, K.; Varga, T.; Forgo, P.; Bart o´ k, M. Catal. Lett. 2004, 97, 65.
(25) Hudlicky, M. Chemistry of Organic Fluorine Compounds; Ellis Horwood
Ltd.: Chichester, 1992.
(26) Filler, R.; Kobayashi, Y.; Yagupolskii, L. M. Organofluorine Compounds
in Medicinal Chemistry and Biomedical Applications; Elsevier: Amsterdam,
1993.
(27) Kuroki, Y.; Asada, D.; Iseki, K. Tetrahedron Lett. 2000, 41, 9853.
(28) Kuroki, Y.; Sakamaki, Y.; Iseki, K. Org. Lett. 2001, 3, 457.
(29) Mallat, T.; Bodmer, M.; Baiker, A. Catal. Lett. 1997, 44, 95.
(30) von Arx, M.; Mallat, T.; Baiker, A. Tetrahedron: Asymmetry 2001, 12,
3089.
(31) von Arx, M.; Mallat, T.; Baiker, A. J. Catal. 2000, 193, 161.
(32) Hess, R.; Diezi, S.; Mallat, T.; Baiker, A. Tetrahedron: Asymmetry 2004,
15, 251.
5) McMorn, P.; Hutchings, G. J. Chem. Soc. ReV. 2004, 33, 108.
6) B u¨ rgi, T.; Baiker, A. Acc. Chem. Res. 2004, 37, 909.
7) Murzin, D. Y.; M a¨ ki-Arvela, P.; Toukoniitty, E.; Salmi, T. Catal. ReV.-
Sci. Eng. 2005, 47, 175.
(
8) Baiker, A. Catal. Today 2005, 100, 159.
(
9) K u¨ nzle, N.; Hess, R.; Mallat, T.; Baiker, A. J. Catal. 1999, 186, 239.
(
10) Baiker, A.; K u¨ nzle, N.; Mallat, T. EP1063224, Hoffmann La Roche,
December 27, 2000.
(
11) Toukoniitty, E.; M a¨ ki-Arvela, P.; Kalantar Neyestanaki, A.; Salmi, T.;
Murzin, D. Y. Appl. Catal., A 2002, 235, 125.
(
(
(
(
(
12) Wells, P. B.; Wilkinson, A. G. Top. Catal. 1998, 5, 39.
13) Sugimura, T. Catal. SurV. Jpn. 1999, 3, 37.
14) Baiker, A. J. Mol. Catal. A: Chem. 2000, 163, 205.
15) Osawa, T.; Harada, T.; Takayashu, O. Top. Catal. 2000, 13, 155.
16) T o¨ r o¨ k, B.; Bal a´ zsik, K.; Felf o¨ ldi, K.; Bart o´ k, M. Ultrason. Sonochem. 2001,
8
, 191.
(
(
(
(
17) von Arx, M.; Mallat, T.; Baiker, A. Top. Catal. 2002, 19, 75.
18) Baddeley, C. J. Top. Catal. 2003, 25, 17.
19) Studer, M.; Blaser, H.-U.; Exner, C. AdV. Synth. Catal. 2003, 345, 45.
20) Sipos, E.; Tungler, A.; Fogassy, G. J. Mol. Catal. A: Chem. 2004, 216,
(33) Diezi, S.; Hess, M.; Orglmeister, E.; Mallat, T.; Baiker, A. J. Mol. Catal.
A: Chem. 2005, 239, 49.
(34) von Arx, M.; Mallat, T.; Baiker, A. Catal. Lett. 2002, 78, 267.
1
71.
4048
9
J. AM. CHEM. SOC. 2006, 128, 4048-4057
10.1021/ja057586f CCC: $33.50 © 2006 American Chemical Society

Hydrogenation of Diketones on Platinum
A R T I C L E S
Scheme 1. Hydrogenation of (Fluorinated) â-Diketones over
action. A major limitation of heterogeneous enantioselective
Pt/Al2O3 Modified by Me3N, Et3N, Quinuclidine, Quinoline,
45,46
catalysis is that truly in situ spectroscopic studies
supporting
(R,R)-PNEA, CD, and O-Methyl-CD
the mechanistic models are barely available due to technical
difficulties.
Hydrogenation of fluorinated â-diketones revealed that ad-
dition of the chiral amine or amino alcohol-type modifier not
only induced enantioselection but also enhanced the low
chemoselectivity to 99-100% in the hydrogenation of the
3
2,33,35
activated keto function.
Similar observations have been
reported for the hydrogenation of R,γ-diketoesters,47 and no
explanation for the excellent chemoselectivity has been found
yet. We assumed that chemo- and enantioselectivities in the
hydrogenation of fluorinated â-diketones are coupled phenom-
ena, and understanding how the modifier enhances the chemose-
lectivity can lead us to the mechanism of enantioselection. The
following mechanistic study based on catalytic experiments,
NMR and FTIR spectroscopic data, and high-level calculations,
indicates that the acid-base-type substrate-modifier interaction
is strongly affected by the adsorption on the metal surface.
9
3% ee in the hydrogenation of 1,1,1-trifluoro-acetylacetone
35
(1, Scheme 1) to the corresponding alcohol.
Experimental Section
Despite the impressive development in this field in the past
Materials. 1,1,1-Trifluoro-2,4-pentanedione 1 (Acros), acetylacetone
years, the mechanistic details at the base of the reaction are
still under debate. Several models have been developed, and
the conflicting opinions reflect the great scientific interest of
2
(Fluka), and 3-penten-2-one 4 (Alfa Aesar) were distilled before use.
Toluene (J. T. Baker) was dried and stored over activated molecular
sieves. Trimethylamine (Me N, Fluka), triethylamine (Et N, Fluka),
3
3
2
3,24,36,37
this topic.
There is some agreement on the formation
quinoline (Fluka), quinuclidine (Aldrich), trifluoroacetic acid (TFA,
Fluka), cinchonidine (CD, Fluka), and 1,1,1,5,5,5-hexafluoroacetyl-
acetone 3 (Acros) were used as received. O-Methyl-cinchonidine
of chiral sites on the metal surface upon adsorption of the
modifier, and also on 1:1 interactions between adsorbed modifier
48
3
8
(MeOCD) and (R,R)-pantoyl-naphthylethylamine ((R,R)-PNEA, Scheme
and substrate. However, contrasting hypotheses have been
formulated concerning the nature of substrate-modifier interac-
tions that lead to selectivity during hydrogenation of activated
ketones on Pt. Because the basic N atom of the modifier is in
most cases crucial for enantiodifferentiation (with only one
49
1
)
were synthesized by known procedures. The 5 wt % Pt/Al
E4759) catalyst was purchased from Engelhard.
Catalytic Hydrogenations. The hydrogenation reactions were
carried out in a mechanically stirred parallel pressure reactor system
Argonaut Technologies) or in a magnetically stirred stainless steel
autoclave controlled by computerized constant-volume constant-pressure
equipment (B u¨ chi BPC 9901). Optimally, the 5 wt % Pt/Al catalyst
was prereduced before use in a fixed-bed reactor by flushing with N
2 3
O
(
(
39
known exception ), interaction models have clustered around
the two main roles that can be played by this function: (i)
hydrogen-bond donor to the keto-carbonyl oxygen after proto-
2 3
O
2
nation,3
8,40
and (ii) nucleophile to the keto-carbonyl carbon as
at 400 °C for 30 min, followed by reductive treatment in H for 60
min at the same temperature. After being cooled to room temperature
in H (30 min), the catalyst was directly used for hydrogenation. The
2
metal dispersion was 0.32 and 0.20 before and after reductive heat
treatment, respectively, as calculated from the average particle size
determined by TEM.50 According to standard conditions, 42 mg of
catalyst, 1.84 mmol of substrate, 6.8 µmol of modifier, and 5 mL of
solvent were stirred (1000 rpm) at 10 bar and room temperature (23-
2
23,37,41
free amine.
Steric effects are generally considered crucial
for selectivity, after formation of one of the aforementioned
interactions. It has also been proposed that the ketone might
interact via hydrogen bonding with the adsorbed aromatic
4
2
anchoring group. In this proposal, the anchoring group would
then become a second docking site, but the relevance of such
interactions still has to be proven. On the other hand, catalytic
2
5 °C) for 2 h. Deviations from the standard conditions are specified
in the text.
Conversion and enantiomeric excess (ee) were determined by gas
and basic organic chemistry studies questioned recently the role
of quinuclidine as a nucleophile,4
3,44
whereas the 1:1 model
operating via hydrogen bond seems to become an increasingly
durable concept. With this background, any experimental
observation that can shed some light on the nature of the 1:1
interactions between the ketone substrate and the chiral amine
modifier is of great interest for rationalization of the catalyst
chromatography using a Chirasil-DEX CB column (Chrompack). The
products were identified via GC/MS (HP 5973 mass spectrometer) and
NMR. The enantiomers were verified by comparing the sign of their
51,52
optical rotation (Perkin-Elmer 241 polarimeter) with literature data.
In the hydrogenation of 1, (R,R)-PNEA, CD, and MeOCD always
afforded the (S)-enantiomer in excess, and in the hydrogenation of 2
the (R)-enantiomer formed in slight excess.
(
35) Diezi, S.; Hess, M.; Orglmeister, E.; Mallat, T.; Baiker, A. Catal. Lett.
2
005, 102, 121.
(
(
(
(
(
36) Baiker, A. J. Mol. Catal. A: Chem. 1997, 115, 473.
(45) Bonalumi, N.; B u¨ rgi, T.; Baiker, A. J. Am. Chem. Soc. 2003, 125, 13342.
(46) Schneider, M. S.; Urakawa, A.; Grunwaldt, J.-D.; B u¨ rgi, T.; Baiker, A.
Chem. Commun. 2004, 744.
(47) Studer, M.; Burkhardt, S.; Indolese, A. F.; Blaser, H. H. Chem. Commun.
2000, 1327.
(48) Borszeky, K.; B u¨ rgi, T.; Zhaoui, Z.; Mallat, T.; Baiker, A. J. Catal. 1999,
187, 160.
(49) Orglmeister, E.; Mallat, T.; Baiker, A. AdV. Synth. Catal. 2005, 347, 78.
(50) Hess, R.; Krumeich, F.; Mallat, T.; Baiker, A. Catal. Lett. 2004, 92, 141.
(51) Forni, A.; Moretti, I.; Prati, F.; Torre, G. Tetrahedron 1994, 50, 11995.
(52) Besse, P.; Ciblat, S.; Canet, J. L.; Troin, Y.; Veschambre, H. Tetrahedron:
Asymmetry 2000, 11, 2045.
37) Vayner, G.; Houk, K. N.; Sun, Y. K. J. Am. Chem. Soc. 2004, 126, 199.
38) Schwalm, O.; Minder, B.; Weber, J.; Baiker, A. Catal. Lett. 1994, 23, 271.
39) Orglmeister, E.; Mallat, T.; Baiker, A. J. Catal. 2005, 233, 333.
40) Blaser, H. U.; Jalett, H. P.; Monti, D. M.; Baiker, A.; Wehrli, J. T. Stud.
Surf. Sci. Catal. 1991, 67, 147.
(
41) Augustine, R. L.; Tanielyan, S. K.; Doyle, L. K. Tetrahedron: Asymmetry
1
993, 4, 1803.
(
42) Lavoie, S.; Lalibert e´ , M. A.; McBreen, P. H. J. Am. Chem. Soc. 2003,
1
25, 15756.
(
43) Orglmeister, E.; Mallat, T.; Baiker, A. J. Catal. 2005, 234, 242.
44) Vargas, A.; Ferri, D.; Baiker, A. J. Catal. 2005, 236, 1.
(
J. AM. CHEM. SOC.
9
VOL. 128, NO. 12, 2006 4049

A R T I C L E S
Diezi et al.
Spectroscopy. Protonation of quinuclidine by TFA and the influence
of quinuclidine on the keto-enol ratio of 1 have been investigated by
H and C NMR. All NMR data were recorded on a Bruker Avance
Scheme 2. Reaction Scheme for the Hydrogenation of 1
1
13
5
00 with TMS as internal standard. The concentration of 1 mimics the
conditions in the general reaction procedure (0.37 mol/L).
Transmission infrared spectra were acquired with an IFS66 spec-
trometer (Bruker Optics) equipped with a DTGS detector. Spectra of
quinuclidine/1 solutions in CH Cl with increasing molar ratios were
2 2
recorded using a KBr cell with 1 mm path length. Similar experiments
were carried out with 2 and 3. For comparison, CD was also used.
Computational Methods. All calculations of theoretical infrared
spectra were performed with the Gaussian 98 set of programs,53 using
density functional theory (DFT) with a hybrid functional (B3LYP) and
a 6-31Gdp basis set.
of the functional was modeled using the Becke correction for the
Adsorption studies have been performed using the Amsterdam
66
67
exchange and the Perdew correction for the correlation. Energies
were compared by setting as reference the most stable structure, because
only relative energies were of interest for this work, and its values
have been shown to be more reliable (less variation with the method
5
4
Density Functional (ADF) program package, and the surface was
simulated using a Pt 31 cluster for which the spin state was optimized
(R- â ) 8). A frozen core approximation was used for the description
of the inner core of the atoms. The orbitals up to 1s were kept frozen
for the second row elements, while orbitals up to 4f were kept frozen
for platinum. Decreasing the Pt frozen core to 4d, which implies the
explicit calculation of 14 additional electrons per platinum atom, has
been shown to increase the adsorption energy by only about 5 kJ/mol
55
used) than absolute ones. All calculations were run unrestricted. To
partially account for the surface reaction to the adsorbates, the central
seven platinum atoms of the cluster were set free to optimize, while
the other bond distances for platinum were fixed to the experimental
68
69
value of 2.775 Å for bulk metal. Molden was used as graphical
interface.
5
5
for the adsorption of benzene. The importance of relativistic effects
5
6,57
has been shown for calculations involving platinum;
therefore, the
core was modeled using a relativistically corrected core potential created
with the DIRAC utility of the ADF program. The DIRAC calculations
imply the local density functional in its simple X-R approximation
without any gradient correction, but the fully relativistic Hamiltonian
is used, including spin-orbit coupling. Furthermore, the relativistic
scalar approximation (mass-velocity and Darwin corrections) was used
for the Hamiltonian, with the zero-order regular approximation (ZORA)
Results
Catalytic Experiments. In the hydrogenation of 1, the
chemoselectivity to 1b (Scheme 2) was only 25% in acetic acid
and 79% in toluene. Addition of small amounts of achiral or
chiral amines (modifiers) enhanced the chemoselectivity up to
100%; the results in the weakly polar solvent toluene are shown
5
8-62
formalism,
where spin-orbit coupling is included already in zero
in Table 1. As compared to the unmodified hydrogenation, both
chiral modifiers CD and (R,R)-PNEA (Scheme 1) enhanced the
reaction rate even at low concentration (modifier/1 ) 0.0037
mol/mol). Interestingly, CD almost doubled the yield to 1b,
although quinuclidine (the function of CD that interacts with
the substrate) barely affected the reaction and quinoline (the
anchoring moiety of CD) diminished the yield remarkably when
they were applied alone in the same concentration. The highest
rate of the formation of 1b was observed in the presence of
MeOCD, the derivative of CD which is a more effective
modifier of Pt in this reaction than the parent alkaloid (Figure
1).
Considering only the achiral amines, there is a positive
correlation between the pKa values of the amines and the rates
of the target reaction. At 10-fold higher concentration, all amines
enhanced the hydrogenation rate relative to the unmodified
reaction. Note that tertiary amines can induce rate acceleration
also in the Pt-catalyzed hydrogenation of ethyl pyruvate,7
although the reliability of those observations has been questioned
6
3
order. The first-order Pauli formalism was shown to have theoretical
deficiencies due to the behavior of the Pauli Hamiltonian at the nucleus,
which lead to variational collapse for increasing basis set size.64 It was
shown that the scalar relativistic correction could account for up to
7
0% of the total energy in the adsorption of carbon monoxide on
platinum, and that also the calculated adsorption site was influenced
5
6
by the use of relativistic corrections. The ZORA formalism requires
a special basis set, to include much steeper core-like functions, that
are implemented in the code. Within this basis set, the double-ú (DZ)
basis functions were used for platinum, and double-ú plus polarization
(DZP) basis functions for the second row elements. The local part of
the exchange and correlation functional was modeled using a Vosko,
6
5
Wilk, Nuisar parametrization of the electron gas. The nonlocal part
(
53) Frisch, M. J.; et al. Gaussian 98, A.7 ed.; Gaussian, Inc.: Pittsburgh, PA,
998.
54) Baerends, E. J.; et al. ADF - Amsterdam Density Functional, Release 2004-
1
(
0
1 ed.; Scientific Computing and Modelling NV - Vrije Universiteit;
Theoretical Chemistry: Amsterdam, 2004.
0-72
(
(
(
(
(
(
(
(
55) Saeys, M.; Reyniers, M.-F.; Marin, G. B.; Neurock, M. J. Phys. Chem. B
2
002, 106, 7489.
56) Philipsen, P. H. T.; van Lenthe, E.; Snijders, J. G.; Baerends, E. J. Phys.
ReV. B 1997, 56, 13556.
7
3
recently.
57) Pacchioni, G.; Chung, S. C.; Kruger, S.; Rosch, N. Surf. Sci. 1997, 392,
A more detailed analysis revealed that the effect of amines
was strongly concentration-dependent, as illustrated by some
examples in Figures 1 and 2. In all cases, the reaction rate
1
73.
58) van Lenthe, E.; Baerends, E. J.; Snijders, J. G. J. Chem. Phys. 1993, 99,
597.
59) van Lenthe, E.; Baerends, E. J.; Snijders, J. G. J. Chem. Phys. 1994, 101,
783.
60) van Lenthe, E.; Baerends, E. J.; Snijders, J. G. J. Chem. Phys. 1996, 105,
4
9
(66) Becke, A. D. Phys. ReV. A 1988, 38, 3098.
6
505.
(67) Perdew, J. P. Phys. ReV. B 1986, 33, 8822.
61) van Lenthe, E.; van Leeuwen, R.; Baerends, E. J.; Snijders, J. G. Int. J.
Quantum Chem. 1996, 57, 281.
(68) Somorjai, G. A. Introduction to Surface Chemistry and Catalysis; Wiley:
New York, 1994.
62) van Lenthe, E.; Baerends, E. J.; Snijders, J. G. J. Chem. Phys. 1999, 110,
(69) Schaftenaar, G.; Noordik, J. H. J. Comput.-Aided Mol. Des. 2000, 14, 123.
(70) Margitfalvi, J. L.; Heged u¨ s, M. J. Catal. 1995, 156, 175.
(71) Blaser, H. U.; Jalett, H. P.; Monti, D. M.; Reber, J. F.; Wehrli, J. T. Stud.
Surf. Sci. Catal. 1988, 41, 153.
8
943.
(
(
63) Snijders, J. G.; Baerends, E. J.; Ros, P. Mol. Phys. 1979, 38, 1909.
64) te Velde, G.; Bickelhaupt, F. M.; Baerends, E. J.; Guerra, C. F.; van
Gisbergen, S. J. A.; Snijders, J. G.; Ziegler, T. J. Comput. Chem. 2001,
(72) Bond, G.; Meheux, P. A.; Ibbotson, A.; Wells, P. B. Catal. Today 1991,
10, 371.
(73) Toukoniitty, E.; Murzin, D. Y. Catal. Lett. 2004, 93, 171.
2
2, 931.
(
65) Vosko, S. H.; Wilk, L.; Nusair, M. Can. J. Phys. 1980, 58, 1200.
4050 J. AM. CHEM. SOC.
9
VOL. 128, NO. 12, 2006

Hydrogenation of Diketones on Platinum
A R T I C L E S
Table 1. Hydrogenation of 1 in the Presence of Achiral and Chiral Amines (Standard Conditions, Toluene)
yields [%]
mod/1
modifier
pK
a
ratio [%]
conv. [%]
1a
1b
1c
1
0
2
15
11
13
15
4
23
28
66
2
Me3N
Et3N
quinuclidine
quinoline
9.80
10.72
11.29
4.94
0.37
0.37
0.37
0.37
0.37
0.37
0.37
11
13
18
4
23
28
66
n.d.a
n.d.
1
n.d.
n.d.
2
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
(R,R)-PNEA
CD
MeOCD
Me3N
Et3N
quinuclidine
quinoline
9.80
10.72
11.29
4.94
3.7
3.7
3.7
3.7
33
38
60
37
1
1
1
1
29
33
56
34
3
4
3
2
a
Not detectable (<0.2%).
Figure 1. The influence of the concentration of CD and MeOCD on the
yield and enantioselectivity in the hydrogenation of 1 to 1b on Pt/Al2O3;
standard conditions, in toluene.
Figure 2. The influence of the concentration of Et3N and quinoline on the
yield and chemoselectivity to 1b in the hydrogenation of 1 on Pt/Al2O3;
standard conditions, in toluene.
characterized by the yield to 1b went through a maximum and
at sufficiently high concentration the yield dropped below 15%
achieved in the reference (unmodified) reaction. This apparent
catalyst poisoning is attributed to a too high coverage of the
active Pt sites by the amines, which leads to a reduced number
of free sites available for the target reaction. As compared to
CD and MeOCD, the rate acceleration was smaller with Et3N
and quinoline and the maximum of the reaction rate was shifted
to lower amine/substrate ratios. The chemoselectivity reached
quinuclidine, and quinoline mimic the influence of CD, MeOCD,
and (R,R)-PNEA, although they are less efficient than the chiral
modifiers.
To clarify the role of the CF3 group in the reactivity of 1,
acetylacetone (2) was also hydrogenated in the presence and
absence of CD (Table 2). In contrast to the hydrogenation of 1,
the reaction rate decreased by the addition of CD irrespective
of the solvent. The small but well reproducible ee reveals that
the position of the quinuclidine N of CD relative to that of 2
adsorbed on the Pt surface is not symmetric. A possible
explanation may be that not the keto but the enol form of 2
interacts with CD in the enantio-differentiating step.
It was assumed earlier that in the enantioselective hydrogena-
tion of the â-ketoester ethyl-4,4,4-trifluoroacetoacetate the keto
carbonyl group rather than the CdC bond of the enol form was
1
00% at low chiral amine concentrations (CD, MeOCD, and
(R,R)-PNEA, Table 1), while it increased monotonically in the
presence of achiral amines up to relatively high amine/substrate
ratios (Figure 2). The modifier concentration influenced also
the enantioselectivity; the highest ee’s in these series were 57%
for CD and 76% for MeOCD.
In summary, amines can accelerate the hydrogenation of the
activated and suppress the transformation of the nonactivated
keto group of 1 (Scheme 2). The achiral amines Me3N, Et3N,
22
the reactive species on cinchona-modified Pt. This assumption
is in agreement with all mechanistic models developed for the
enantioselective hydrogenation of activated ketones on Pt,
J. AM. CHEM. SOC.
9
VOL. 128, NO. 12, 2006 4051

A R T I C L E S
Diezi et al.
Table 2. Enantioselective Hydrogenation of Acetylacetone (2) on
Unmodified and CD-Modified Pt/Al2O3 (Standard Conditions)
2a
2b
modifier
solvent
conv %
yield %
ee %
yield %
toluene
toluene
dichloromethane
dichloromethane
29
8
21
7
17
7
16
6
12
1
5
CD
CD
8 (R)
4(R)
1
Figure 3. 1H NMR investigation of the protonation of quinuclidine by 1
and TFA in CDCl3. The arrows mark the chemical shifts for a 1:1 mixture
of 1 and quinuclidine. The concentration of 1 mimics the conditions in the
general reaction procedure (0.37 mol/L).
Table 3. Hydrogenation of 3-Penten-2-one (4) on Unmodified and
CD-Modified Pt/Al2O3 (CH2Cl2 Solvent and Standard Conditions)
Addition of an amine to 1 in solution leads to the formation
78-80 1
of a protonated amine-enolate salt.
on the protonation of quinuclidine by 1 and trifluoroacetic acid
TFA) is shown in Figure 3. At least 4-6 mol equiv of TFA
H NMR investigation
yields %
(
modifier
conv %
4a
4b
was necessary for complete protonation of quinuclidine in
CDCl3. As compared to the shifts in the TFA/quinuclidine
mixture, a strong acidity of the enolic proton of 1 can be
estimated in this weakly polar medium.
1
100
00
97
99
3
1
CD
Scheme 3. Keto-Enol Equilibrium for 1
1
Furthermore, the H NMR spectrum of 1 in CDCl3 shows a
broad signal at δ ) 14.15 ppm, which belongs to the enolic
proton. After the addition of 1 mol equiv of quinuclidine, the
signal of the enolic proton disappeared and a new signal
appeared at 9.76 ppm (protonated nitrogen).
These NMR results confirm the strong acid-base interaction
between 1 and quinuclidine. There was no signal in the NMR
spectra giving evidence for a nucleophilic interaction between
the quinuclidine N and the electrophilic carbonyl C atom and
no zwitterion formation could be detected, in contrast to former
reports on the interaction of R,R,R-trifluoro-ketones and qui-
although no experimental evidence could be found yet to support
this hypothesis. To estimate the reactivity of the CdC and Cd
6
O bonds in the enol forms of 2, the hydrogenation of an
analogous compound 3-penten-2-one (4) was investigated under
standard conditions (Table 3). The CdC bond was completely
saturated in 2 h, while the hydrogenation of the carbonyl group
barely took place. Obviously, the reactivity of the CdC bond
is much higher than that of the (deactivated) carbonyl group of
2
3,37
nuclidine.
The 13C NMR signals of the two carbonyl carbons of 1 were
found at 194.76 ppm (-COCH3) and at 176.05 ppm (-COCF3).
After addition of 1 mol equiv of quinuclidine, the signal of the
activated carbonyl group shifted upfield to 168.84 ppm, whereas
the signal of the nonactivated carbonyl group remained constant
at 194.82 ppm. Further increase of the amount of quinuclidine
to 3 equiv had no influence on the chemical shifts, indicating a
1:1 interaction.
4, and thus the reactivity of one of the enol forms of 1 (Scheme
3
) on Pt cannot be excluded.
NMR Study. â-Diketones can be present in a keto- and two
enol-forms (Scheme 3), and the keto-enol equilibrium depends
on the solvent.7
4,75
In weak to medium polar solvents, 1 exists
FTIR Spectroscopy. The infrared spectrum of 1 in CH2Cl2
(Figure 4) displays bands at ca. 1645, 1601, 1206, 1157, and
mainly in its enol form (98% in toluene and 93% in dichloro-
3
2
methane). The amount of enol-1 and enol-2 in solution can
be quantified in CDCl3 using the formula:
-1
-1
1
109 cm , together with a weak envelope at around 1790 cm .
This last feature is given by the ν(CdO) mode of the free
diketone species, which however account only for about 7% of
pMe ) (208 - δ )/33
(1)
Me
3
2
the species in solution. The spectrum of compound 1 is
predominantly a combination of the spectra of the two enol
forms, whose calculated vibrational spectra are reported in
Figure 4c and d scaled by 0.967. The spectra agree well with
where δMe represents the chemical shift of the nonactivated
carbonyl group in the 13C NMR and pMe indicates the amount
of enol form on the side of the methyl group (enol-2). Using
this equation, we found that enol-1 was the dominant species
59%) in CDCl3. The slight shift of the equilibrium toward
enol-1 is due to the presence of the electron-withdrawing CF3
7
6
81
those already available in the literature. The linear combination
of the calculated vibrational spectra of the two enol forms in
(
1
the ratio 41:59 in favor of enol-1, as suggested by the H NMR
data, matches the experimental spectrum of 1. According to the
7
7
group.
(
78) Koltsov, A. I.; Golubev, N. S.; Milevskaya, I. S. Zh. Obshch. Khim. 1980,
(
74) Emsley, J. Structure and Bonding, Complex Chemistry: The Composition,
Structure and Hydrogen Bonding of the â-Diketones; Springer-Verlag:
Berlin, 1984.
50, 2504.
(79) Pashkevich, K. I.; Filyakova, V. I. Bull. Acad. Sci. USSR 1984, 33, 575.
(80) Filyakova, V. I.; Ratner, V. G.; Karpenko, N. S.; Pashkevich, K. I. Russ.
Chem. Bull. 1996, 45, 2163.
(81) Minoura, Y.; Nagashima, N.; Kudoh, S.; Nakata, M. J. Phys. Chem. A
2004, 108, 2353.
(
(
(
75) Rogers, M. T.; Burdett, J. L. Can. J. Chem. 1965, 43, 1516.
76) Koltsov, A. I. J. Mol. Struct. 1998, 444, 1.
77) Niwa, J.; Yamazaki, M.; Takeuchi, T. Chem. Lett. 1975, 4, 707.
4052 J. AM. CHEM. SOC.
9
VOL. 128, NO. 12, 2006

Hydrogenation of Diketones on Platinum
A R T I C L E S
and at 3594 cm- (ν(O-H)) are characteristic of the alkaloid.82
The presence of the mode of the free OH group of CD indicates
that this function is not involved in the interaction with 1.
The spectral changes observed upon titration of 1 by
quinuclidine (and CD) are associated with the formation of the
ion pair composed of the quaternary ammonium ion of quinu-
1
-
1
clidine (2600 cm ) and the enolate anion of 1 (1640, 1530,
-
1 79,82,83
1
204, 1175, and 1128 cm ).
The two major bands at 1640
and 1530 cm belong to groups displaying CdO79 and CdC
character. Although deprotonation of 1 does not allow the
localization of the two bonds (formation of the enolate ion),
the presence of molecular asymmetry in 1 induced by the CF3
group suggests that the charge distribution between the O-atoms
is also asymmetric.
-1
This last observation becomes particularly important for the
discussion on how the chemoselectivity is controlled upon
addition of the chiral amine during the enantioselective hydro-
genation of 1. Therefore, interpretation of the changes observed
in Figure 5a (and Figure 8b, see later) required the aid of
vibrational analysis to resolve the structure of the enolate ion
and clarify which C-O bond has a partial CdO character in
the ion pair, the information of which is missing from the
available literature. Theoretical calculations were carried out
enabling the relaxation of the proton located on the amine toward
one of the two O-atoms of the enolate. The structure is shown
in the inset of Figure 5 together with the calculated spectrum
of the acid-base pair 1:Me3N, which matches well the
experimental spectrum in Figure 5a. The frequency of the sharp
-
1
-1
+
signal at 1951 cm (calcd 2018 cm ) for the ν(N-H ) mode
is underestimated as compared to the other vibrations probably
due to the fact that the calculations localize the proton on the
N-atom. Interestingly, the vibrational analysis suggests that the
C-O on the side of CF3 rather than that on the side of CH3 has
Figure 4. Experimental and calculated vibrational spectra of 1. (a)
Transmission IR spectrum of a solution of 1 in CH2Cl2 (5 mM). Calculated
vibrational spectra (scaled by 0.967) of (b) a mixture of the enol forms
5
9:41, (c) enol-2, and (d) enol-1 forms. (b) Vibrational spectrum obtained
by combining the calculated vibrational spectra of the two enol forms in
the ratio 59:41. The dotted spectrum in (a) corresponds to neat 1.
-
1
a partial CdO character. The signals at 1640 and 1530 cm
are then assigned to ν(CCF -CO)+ν(C-CCH ) and ν(C-CCH )+
calculated vibrational spectra, the signals at 1645 and 1601 cm-1
correspond to the asymmetric and symmetric stretching motions
3
3
3
ν(CCF -CO) modes of the enolate of 1, respectively.
3
Figure 6 shows the FTIR spectra of solutions of the diketones
of the CdO and CdC groups, whereas the set of signals
1
, 2, and 3 before and after addition of 1 mol equiv of amine.
-
1
between 1250 and 1100 cm is assigned to the vibrational
modes associated with the CF3 and CH groups.
The least reactive substrate toward deprotonation, that is, the
least acidic, is 2, in agreement with the effect induced by
substitution of a CH3 group by a CF3 group. In the FTIR spectra
of the neat substrates, this substitution results in the splitting of
the signals assigned to the asymmetric and symmetric stretching
The infrared spectrum of 1 exhibited striking changes upon
titration of a solution of 1 with a solution of quinuclidine (Figure
5
a). In the high-frequency region (left panel), a broad band
-
1
centered at 2640 cm increased with increasing quinuclidine
concentration. The changes observed in the middle panel of
Figure 5a suggest the immediate disappearance of the free
diketone species (1790 cm ) and the replacement of the bands
at 1645 and 1601 cm by signals at 1640 and 1530 cm
8
4
of the OdC-CdC-O group and in the complete disappear-
8
5
ance of the free diketone species in the case of 3.
-
1
Addition of quinuclidine does not affect the spectrum of 2
but substantially alters the spectra of 1 and 3. As in the case of
-
1
-1
,
-
1
the ion pair 1:quinuclidine, the signals at 1666 and 1549 cm
in Figure 6f are attributed to the formation of the enolate of 3.
respectively, at 1 mol equiv of base. In the right panel of Figure
-
1
5
a, the signals at 1157 and 1109 cm disappeared almost
-
1
Additional signals are also found between 3000 and 2000 cm
completely upon addition of quinuclidine, whereas new signals
+
-1
-
1
-1
(broad, ν(N-H )) and at 1192, 1144, and 1136 cm . The
enolate ions of 2 and 3 exist only in a single form due to
molecular symmetry. The enolate of 2 in aqueous solution
were found at 1175 and 1128 cm . The signal at 1206 cm is
strongly attenuated. Two isosbestic points can be clearly
-
1
observed at 1167 and 1113 cm that indicate the transition
from the species related to the spectrum before addition of
quinuclidine (dotted) to new species formed after addition of
the base (bold spectrum).
-
1 86,87
absorbs at frequencies below 1600 cm ,
that is, more than
(
82) Ferri, D.; B u¨ rgi, T.; Baiker, A. J. Chem. Soc., Perkin Trans. 2 1999, 1305.
(83) Bellamy, L. J. The Infrared Spectra of Complex Molecules. AdVances in
Infrared Group Frequencies; Chapman and Hall: London, 1980; Vol. 2.
Figure 5b shows that the interaction between CD and 1 is
identical to that between quinuclidine and 1, with only minor
differences in the position of the new signals. The additional
(84) Ogoshi, H.; Nakamoto, K. J. Chem. Phys. 1966, 45, 3113.
(85) Wallen, S. L.; Yonker, C. R.; Phelps, C. L.; Wai, C. M. J. Chem. Soc.,
Faraday Trans. 1997, 93, 2391.
(
86) Ernstbrunner, E. E. J. Chem. Soc. A 1970, 1558.
-
1
signals at 1593, 1572, and 1510 cm (quinoline ring modes)
(87) Dickie, S. A.; McQuillan, A. J. Langmuir 2004, 20, 11630.
J. AM. CHEM. SOC.
9
VOL. 128, NO. 12, 2006 4053

A R T I C L E S
Diezi et al.
Figure 5. Transmission infrared spectra of a solution of (a) 1 with 0 (dotted), 0.1, 0.2, 0.4, 1, and 2 (bold) equiv of quinuclidine and (b) 1 with 1 equiv of
CD. Arrows indicate the direction of the changes upon base addition. The calculated vibrational spectrum of the complex 1:Me3N is shown in (c) (scaling
factor 0.967). Spectra b and c are offset. C1 ) CCD ) 5 mM, Camine ) 0.5, 1, 2, 5, and 10 mM, ambient temperature, in CH2Cl2.
introduction of the electron-withdrawing CF3 group, thus further
supporting the validity of the vibrational analysis of the ion pair.
Adsorption of 1 on Pt. Adsorption of 1 was studied using a
Pt 31 cluster to model the surface to better understand the
catalytic behavior. Figure 7a shows the adsorption mode of 1
in the diketo-form, chemisorbed η2 via the activated keto-
carbonyl group (i.e., the keto-carbonyl in R-position to the CF3
group). The other possibility, the adsorption of the diketo-form
chemisorbed η2 via the nonactivated keto-carbonyl group, is
presented in Figure 7b. Consistently with previous results,88 the
former was more strongly adsorbed by ca. 3 kcal/mol than the
latter. This result was rationalized by the energy lowering of
the frontier orbitals of the activated keto-carbonyl, which can
better interact with the filled states of the metal in back-bonding
88,89
interactions.
This chemisorption mode of CdO double bonds
is also called di-σ to evidence the double σ binding of the keto
carbonyl moiety to the metal atoms. Also, η1 adsorption modes
are theoretically possible, but it has been shown that their
adsorption energy is much smaller than the η2 modes, and, in
particular, in such adsorption modes the keto-carbonyl moiety
is distant from the surface and cannot reach surface hydrogen;
therefore, they are less interesting in our discussion.88 On the
other hand, the η2 modes show rehybridization of the keto-
carbonyl group and have the right geometry for hydrogen
uptake. Another adsorption mode of the diketo-form is shown
in Figure 7c, where the diketone is adsorbed with both carbonyl
groups close to the surface. Because both carbonyls cannot
simultaneously find an optimal interaction with the surface Pt
atoms, the most strongly interacting moiety dominates the
adsorption. This mode is energetically disfavored as compared
to the two previous structures; nevertheless, we observe that
the result of this interaction is the rehybridization of the activated
Figure 6. FTIR spectra of solutions of (a) 2, (c) 1, and (e) 3. Addition of
equiv of quinuclidine afforded spectra b, d, and f, respectively. C ) 5
mM, ambient temperature, in CH2Cl2.
1
4
0 cm- below the frequency observed for the enolate of 1.
1
-
1
The enolate of 3 exhibits the ν(C-O) at 1666 cm and shows
-
1
a ∆ν with respect to the enolate of 1 of only 26 cm . The
difference in shift suggests that the enolate of 1 resembles more
the enolate of 3 and displays some carbonyl character on the
C-O bond near the CF3 group, in agreement with the usually
observed blue-shift of the ν(CdO) band of ketones upon
(88) Vargas, A.; B u¨ rgi, T.; Baiker, A. J. Catal. 2004, 222, 439.
(89) Hammer, B.; Nørskov, J. K. AdV. Catal. 2000, 45, 71.
4054 J. AM. CHEM. SOC.
9
VOL. 128, NO. 12, 2006

Hydrogenation of Diketones on Platinum
A R T I C L E S
Figure 8. Simulation of the interaction between Me3NH+ and adsorbed
enolate. Structure (a) is the starting geometry, in which the acidic proton is
equidistant from the two oxygen atoms. Structure (b) is an intermediate
Figure 7. Calculated adsorption modes of 1 on a Pt cluster. Structures (a),
(b), and (c) show the adsorption of the diketo-form, (d) and (e) the adsorption
+
point of the optimization, where the Me3NH is closer to the nonactivated
of the enol forms, and (f) the adsorption of the enolate. The energy of the
most stable structure (d) has been set to zero and taken as a reference for
the others. The numbers in brackets are the energy differences (kcal/mol)
between (d) and the other (less stable) structures.
moiety and removes it from the surface. Structure (c) is the converged
+
geometry, where the Me3NH is coordinated to the di-σ adsorbed CO.
with respect to the molecules that are not interacting with the
metal. Binding to the metal increases the difference of stability
by a factor of 2, raising it to 20 and 30 kcal/mol in favor of the
enol forms.
keto-carbonyl only, although also the nonactivated carbonyl is
close to the metal.
The adsorption of the enol forms of 1 (Figure 7d and e) was
the most energetically favored, with a preference for the
adsorption of enol-2 (Figure 7d), where the optimal chemisorp-
tion of the activated carbonyl is possible. It should be noted
that the adsorption of enol-1 (Figure 7e) does not lead to the
expected η2 interaction of the nonactivated carbonyl with the
metal, but rather with the formation of a σ-bond between the
carbon neighboring the CF3 group and a Pt atom. Also, in this
case, the competition for surface sites is in favor of the activated
carbonyl group, and the result is that the nonactivated carbonyl
undergoes no evident structural alteration induced by the metal.
Also, to be noted is that both adsorbed enol forms are stabilized
by internal hydrogen bonds. However, while in enol-2 the
H-bond length is 1.35 Å, it elongates to 1.71 Å in enol-1,
showing the stronger hydrogen-acceptor character of the ad-
sorbed activated carbonyl group.
Figure 7f shows the adsorption mode of the enolate, the
deprotonated form of the enol, thus a charged species. Unlike
the enols, the enolate has a more symmetric adsorption mode
where both oxygen atoms are bound to a metal atom, and with
a third σ binding site due to C(3) that is also bound to a metal
atom and rehybridized. Note that the relative energy is not given,
because this structure contains one hydrogen less than the others
and its energy cannot be simply compared. The negative charge
of the enolate is delocalized on the metal upon adsorption.
It has been discussed previously that introducing an amine
in the reaction system leads to an almost complete chemo-
selectivity in the saturation of the activated keto-carbonyl group
of 1. It has also been shown that in solution the amine
deprotonates the enol and generates an ammonium ion and the
enolate. Therefore, the interaction between the trimethylammo-
nium ion and the enolate was modeled in the presence of the
Pt surface. Figure 8 shows three steps of this simulation: Figure
8a shows the starting geometry whereby the (trimethyl)-
ammonium ion is positioned close to the adsorbed enolate ion,
such that the distance between its acid hydrogen and the oxygen
atoms of the enolate is equal (2.5 Å). During minimization, the
ammonium ion moves toward the surface and comes closer to
The structures of the diketo and enol forms without surface
interaction were also calculated using the same level of theory.
The two conformations of the diketo-form (not shown) are about
10 and 15 kcal/mol, respectively, less stable than the lowest
energy enol-2 at this level of theory. For our purposes, it is
important to note that after chemisorption the difference in
energy between the enol and keto forms of 1 is greatly enhanced
J. AM. CHEM. SOC.
9
VOL. 128, NO. 12, 2006 4055

A R T I C L E S
Diezi et al.
the nonactivated CO, as shown in Figure 8b. In this intermediate
position, the hydrogen bond formed with the nonactivated
moiety is 1.8 Å, while that formed with the activated moiety is
nation of 1. The real origin of enhanced chemoselectivity
induced by addition of amines could be clarified only by
considering the adsorption on the metal surface.
2
.3 Å. As the optimization continues, this preferential coordina-
The theoretical study of the adsorption of the different
(equilibrated) species of 1 on a model Pt cluster has shown that
the enol forms are more stable also on the surface with respect
to the diketo-form and adsorption even increased the difference
in their stability (Figure 7). More interestingly, the simulation
of the formation of the ion pair on the metal surface has revealed
that coordination of the ammonium ion to the adsorbed enolate
results in the displacement of the nonactivated CO from the
metal surface and at the same time the promotion of an η2
adsorption mode of the activated CO. In the equilibrium
geometry shown in Figure 8c, the ammonium ion coordinates
to the activated CO moiety adsorbed η2. It has already been
proposed that η2 adsorption modes of ketones are more activated
toward hydrogenation, and the results shown in Figure 8 seem
tion leads to a displacement of the nonactivated keto-carbonyl
from the surface, and when convergence is reached the
ammonium ion coordinates the activated CO bond that changes
its adsorption mode to η2 (Figure 8c).
Discussion
A big effort has been made in past years to understand the
functioning of chirally modified Pt in the enantioselective
hydrogenation of ketones. As discussed in the Introduction,
several mechanistic models have been proposed, but only one
of these models is supported by in situ experimental observa-
tions. A major difficulty is the high reactivity of activated
ketones in the presence of amine-type modifiers that complicates
or even hinders the study of substrate-modifier interactions in
solution.90 The present study is the first example where the
nature of the amine-ketone interaction in solution could be
refined by spectroscopic methods, and further modeled by state-
of-the-art calculations.
The hydrogenation experiments have shown that both car-
bonyl groups are reactive on Pt that may lead to poor
chemoselectivity, but only the activated carbonyl group is
reduced to an alcoholic OH function in the presence of the chiral
amine and amino alcohol-type modifiers CD, MeOCD, and
4
5
8
8
to be in good agreement with this proposal.
The apparent mismatch between the spectroscopic results
obtained in solution and the adsorption study correctly interpret-
ing the observed changes in chemoselectivity may be explained
as follows. As mentioned, the ion pair formed with the
ammonium ion and the enolate in solution is characterized by
a closer interaction between the acidic proton and the nonac-
tivated CO. This interaction is also present when the ion pair is
interacting on the metal surface, but only in the first phase of
the interaction (Figure 8b). In contrast, the activated CO, which
is di-σ adsorbed, dominates the interaction with the acidic proton
at equilibrium, consistent with the electron transfer to the metal,
making this group a better hydrogen-bond acceptor. Comparison
of these results to the catalytic experiments strongly suggests
that the ammonium ion promotes the hydrogenation of the
activated keto-carbonyl by selective coordination, with a
fundamental contribution from the metal surface enhancing its
proton affinity that cannot be neglected. On the other hand, the
calculated geometry of the enolate explains why hydrogenation
of the nonactivated carbonyl lying well above the metal surface
is hindered in the presence of an amine. This observation
highlights the influence of the metal surface on the substrate-
modifier interaction.
(R,R)-PNEA (Table 1, Figure 1). The remarkable enhancement
in chemoselectivity is due to acceleration of the hydrogenation
of the activated and deceleration of the hydrogenation of the
nonactivated carbonyl functions by the basic amine. It has also
been proved by catalytic and spectroscopic methods that simple
achiral amines (e.g., Me3N, Et3N, quinuclidine) could improve
the chemoselectivity, although to a smaller extent (Figure 2),
and the OH function of CD was not involved in the substrate-
modifier interaction (Figures 4 and 5). The better efficiency of
the chiral modifiers as compared to simple amines at identical
(low) solution concentrations is attributed to the stronger
adsorption of the modifiers on the Pt surface. Note that strong
adsorption leading to high surface concentration is one of the
key characteristics of effective chiral modifiers.
The unambiguous relation between chemoselectivity and
metal-substrate-modifier interaction is valuable to assess the
debated model of the interaction between cinchona alkaloids
and activated ketones during enantioselective hydrogenation on
cinchona-modified platinum. In fact, a similar N-H-O-type
interaction is expected between the protonated quinuclidine
moiety of the alkaloid and an adsorbed activated ketone
substrate. This can occur also in a nonacidic medium (such as
toluene). It has been shown recently that the alkaloid can be
protonated by uptake of activated hydrogen from the Pt
NMR and FTIR spectroscopy has revealed that, in contrast
91,92
to former reports,
1 is about as strong of a proton donor as
TFA (Figure 3) and formation of the enolate-ammonium ion
pair, where the quaternary ammonium ion coordinates to the
enolate, is responsible for enantioselection. The structure
assigned to this ion pair partially localized the proton between
the N-atom of the amine and the O-atom of the nonactivated
keto-carbonyl of 1. For comparison, the small but significant
ee observed in the hydrogenation of 2 (Table 2) indicated that
the position of the N atom relative to the O atoms in the
modifier-substrate complex is nonsymmetric even in the
absence of the electronic effect of fluorine atoms.
4
4,45
surface.
The hydrogen-bonding interaction has already been inter-
93,94
preted as the trigger for rate acceleration,
which is observed
A striking result of this study is that the substrate-modifier
interaction identified in solution by spectroscopic methods
cannot interpret the chemoselectivity observed in the hydroge-
also in the experiments discussed here. The enantioselectivity
in the presence of the adsorbed modifier is to be interpreted as
the result of different interaction energies that the prochiral faces
of the ketone substrate show within a chiral environment
(
(
(
90) Ferri, D.; B u¨ rgi, T.; Borszeky, K.; Mallat, T.; Baiker, A. J. Catal. 2000,
93, 139.
91) Pashkevich, K. I.; Saloutin, V. I.; Postovskii, T. Y. Russ. Chem. ReV. 1981,
0, 180.
92) Shukla, J. P.; Sharma, R. S. Electrochim. Acta 1986, 31, 1449.
1
(93) Vargas, A.; B u¨ rgi, T.; Baiker, A. New J. Chem. 2002, 26, 807.
(94) Vargas, A.; B u¨ rgi, T.; von Arx, M.; Hess, R.; Baiker, A. J. Catal. 2002,
209, 489.
5
4056 J. AM. CHEM. SOC.
9
VOL. 128, NO. 12, 2006

Hydrogenation of Diketones on Platinum
A R T I C L E S
generated by the alkaloid on the surface. As already pro-
between the modifier and 1 are different in solution and on the
Pt surface. The outstanding chemoselectivity is attributed to the
selective coordination of the protonated amine (modifier) to the
adsorbed activated keto-carbonyl group and prevention of the
interaction of the nonactivated carbonyl group in this position
with the metal surface. The fundamental differences in the
structures of the substrate-modifier ion pair in solution and on
the Pt surface underline that the reaction mechanism cannot be
understood without considering the adsorption on the metal
surface.
9
3,94
posed,
enantiodiscrimination should arise from two com-
bined effects, both related to the strength of the hydrogen-bond
interaction: on one hand, the best fitting prochiral face has the
lowest energy and therefore a higher fractional coverage; on
the other hand, the stronger hydrogen-bond interaction should
show the largest rate enhancement, generating kinetic resolution.
A quantification of the two effects is only possible via modeling
of the interaction between the alkaloid and the substrate both
adsorbed on the metal, which is the subject of ongoing
investigation in our laboratory. The better performances obtained
using cinchona alkaloids as carriers for the amino-group show
the unique feature of these molecules that provide the reactive
site (basic nitrogen) in proximity of the metal, still allowing
the presence of free metal sites. The structural features of the
cinchona alkaloids allow reshaping of the surface upon adsorp-
tion and generation of new surface sites without annihilation
of the availability of the metal for the contact with the substrate,
unlike quinoline (the “anchoring group” of the alkaloid), which
is a catalyst poison when used alone (Table 1, Figure 2).
Therefore, the cinchona alkaloids are properly called surface
chiral modifiers, whereas a simple alkylamine is a reaction
additive.
The presented model shows important similarities to those
suggested for the enantioselective hydrogenation of other
activated ketones on Pt modified by chiral amines.1
4,49,95
In all
cases, an N-H-O-type hydrogen bond between the amine
modifier and the keto-carbonyl group of the adsorbed substrate
is critical for rationalizing the outcome of the reaction, concern-
ing both the rate enhancement and the enantioselectivity.
Acknowledgment. We kindly acknowledge the Swiss
National Science Foundation for financial support and the Swiss
Center for Scientific Computing (SCSC) in Manno and ETHZ
for computing time.
Supporting Information Available: Detailed assignment of
the vibrational modes of the two enol forms of 1 and of the
Conclusions
1:Me3N, Cartesian coordinates of the structures shown in Figures
Catalytic experiments combined with spectroscopic and
computational structural analysis have been applied to under-
stand the coupled phenomena: the enantioselection and the
enhancement of chemoselectivity in the hydrogenation of the
activated keto-carbonyl group of 1. The study revealed that the
structures of ammonium ion-enolate-type ion pairs formed
7
and 8, and full refs 53 and 54. This material is available free
of charge via the Internet at http://pubs.acs.org.
JA057586F
(95) Bodmer, M.; Mallat, T.; Baiker, A. In Catalysis of Organic Reactions;
Herkes, F. E., Ed.; M. Dekker: New York, 1998; p 75.
J. AM. CHEM. SOC.
9
VOL. 128, NO. 12, 2006 4057