Enantioselective Hydrogenation of α,β-Unsaturated Carboxylic Acids over Cinchonidine Modified Palladium: Nature of Modifier-Reactant Interaction

Article abstract of DOI:10.1006/jcat.1999.2584

The mechanism of enantiodifferentiation in the hydrogenation of alkenoic acids over cinchona-modified Pd has been investigated using the tiglic acid → 2-methyl-butanoic acid transformation as test reaction. Application of simple derivatives of cinchonidin

Full text of DOI:10.1006/jcat.1999.2584

Journal of Catalysis 187, 160–166 (1999)
Article ID jcat.1999.2584, available online at http://www.idealibrary.com on
Enantioselective Hydrogenation of , -Unsaturated Carboxylic Acids
over Cinchonidine Modified Palladium: Nature
of Modifier–Reactant Interaction
K. Borszeky, T. Bu¨rgi, Z. Zhaohui, T. Mallat, and A. Baiker1
Laboratory of Technical Chemistry, Swiss Federal Institute of Technology, ETH-Zentrum, CH-8092 Zu¨rich, Switzerland
Received March 8, 1999; revised June 4, 1999; accepted June 4, 1999
lective hydrogenation of alkenoic acids (7). The best enan-
tioselectivities are around 50% for alkenoic acids (8) and
over 70% for an aryl substituted derivative (9). Pd/silica
modified by chiral silyl ethers (10) and Raney Ni modified
by tartaric acid (11) were shown to be far less effective cata-
lysts.
The mechanism of enantiodifferentiation in the hydrogenation of
alkenoic acids over cinchona-modified Pd has been investigated us-
ing the tiglic acid → 2-methyl-butanoic acid transformation as test
reaction. Application of simple derivatives of cinchonidine, modi-
fied at the (C-9)–OH and/or the quinuclidine nitrogen, proved that
both functional groups are involved in the enantiodiscriminating
step. Addition of a strong base (1,8-diazabicyclo[5.4.0]undec-7-ene,
DBU) to tiglic acid prior to hydrogenation revealed that one cin-
chonidine molecule interacts with a dimer of tiglic acid on the metal
surface. Ab initio calculations corroborate the existence of an ener-
getically favored acid dimer–cinchonidine intermediate stabilized
by hydrogen bonding, involving both the OH and the quinuclidine
The advances made so far have been based exclusively
on “trial and error” approaches. Understanding of the re-
action mechanism could put the development of new and
more efficient catalysts on a more rational basis. To date,
only a few papers have reported on the possible interac-
tions between reactant and chiral modifier, both adsorbed
on the metal surface. We proposed first (12) that, analo-
gous to the well studied triethylamine–acetic acid interac-
tion (Scheme 1 (13)), the alkenoic acid dimer protonates
the quinuclidine nitrogen of cinchonidine (CD) in apolar
medium. This proposal could rationalize the observed loss
of ee in polar medium, due to the shift of dimer–monomer
equilibrium to the monomer side. Independently, Wells and
coworkers (14) proposed an acid–base interaction between
an alkenoic acid and a CD molecule. This 1 : 1 type interac-
tion has been supported by simple calculations.
Recently, we provided spectroscopic evidence for the
presence of alkenoic acids as dimers in apolar solutions
(12), and we assumed that the dimer structure is preserved
on the Pd surface during the enantiodifferentiating step.
On the basis of this 2 : 1 reactant–modifier interaction we
proposed a generally applicable empirical rule to predict
the absolute configuration of the major enantiomer of the
product. Besides, XRD analysis of a CD– , -unsaturated
acid salt indicated a chain-like structure, stabilized by a hy-
drogen bonded network involving both amino and alcohol
functional groups of the modifier. However, the possible
role of the OH group in the enantiodiscriminating step re-
mained unresolved.
c
nitrogen of cinchonidine.
1999 Academic Press
Key Words: , -unsaturated carboxylic acid; dimer; cinchoni-
dine derivatives; hydrogenation; palladium-on-alumina; ab initio
calculation.
INTRODUCTION
Asymmetric catalysis has made enormous progress dur-
ing the past decades fostered by a growing demand on chiral
substances. An important target of research has been the
enantioselective hydrogenation of , -unsaturated acids,
affording useful building blocks for the synthesis of non-
steroidal anti-inflammatory agents (1, 2). Transition metal
complexes with chiral ligands are the most powerful cata-
lysts for the production of optically active carboxylic acids.
In particular ruthenium(II) complexes with chiral diphos-
phine BINAP ligands received much attention (3). The de-
velopment of new atropisomeric bis(triarylphosphine) lig-
ands (H8-BINAP) provided outstanding catalysts affording
enantiomeric excess (ee) over 97% (4, 5).
Solid catalyzed enantioselective hydrogenation would
be, however, preferable due to its technical and economic
advantages (6). Presently, supported Pd catalysts in the
presence of cinchona alkaloids as the source of chiral in-
duction are the only useful solid catalysts for the enantiose-
Very recently, Nitta and Shibata (15) proposed that
both the OH functional group and the quinuclidine nitro-
gen atom of cinchonidine are involved in hydrogen bond-
ing with -phenylcinnamic acid via 1 : 1 interaction. Note
that in the strongly polar medium, used under optimized
1 To whom correspondence should be addressed.
160
0021-9517/99 $30.00
c
Copyright
1999 by Academic Press
All rights of reproduction in any form reserved.

ENANTIOSELECTIVE HYDROGENATION OF UNSATURATED ACIDS
161
N-allyl cinchonidine chloride (3) and N-benzyl cin-
chonidine bromide (5): General procedure. Allyl chloride
(0.60 g, 8 mmol) (benzyl bromide 0.68 g, 4 mmol) was
added dropwise to a solution of CD (1.18 g, 4 mmol) in
THF (50 ml) under continuous stirring. After refluxing the
mixture for 1 week (12 h for benzyl bromide) the solvent
was evaporated in vacuo and the crude solid was purified
by flash chromatography using CH2Cl2 : MeOH = 7 :1 as
eluent (CH2Cl2 : MeOH = 18 : 1 for 5) (3, 0.20 g, 13% ; 5,
1.4 g, 75% ). Anal. Calcd for C22H27ClN2O (3): C, 71.2; H,
7.3; N, 7.6. Found: C, 70.7; H, 7.6; N, 7.1. Anal. Calcd for
C26H29BrN2O (5): C, 67.1; H, 6.3; N, 6.0. Found: C, 66.7; H,
6.4; N, 5.9.
SCHEME 1. Structure of triethylamine–acetic acid dimer species in
apolar solvents (13).
conditions, -phenylcinnamic acid may dominantly be
present as monomer due to stabilization by substrate–
solvent interactions. However, it is unclear whether in the
enantiodiscriminating step CD interacts with a monomer
or a dimer of acid on the metal surface.
N-methyl cinchonidine chloride (6). CD (6 g, 20.4 mmol)
was dissolved in methanol (110 ml), and methyl iodide
(5 g, 35.6 mmol) in methanol (50 ml) was added dropwise
under an argon atmosphere, at room temperature with con-
tinuous stirring and the resulting mixture was stirred for
24 h. The solvent was evaporated and the residue was re-
crystallized twice from ethyl acetate : methanol = 5 : 8 to af-
ford N-methyl cinchonidine iodide as white colorless crystal
(6.75 g, 76% ). Ion exchange using 1.3 eq of AgCl (Fluka)
in MeOH : water = 1 : 1 afforded 6. The modifier contained
8.5% iodide on the basisofelementaryanalysis. Anal. Calcd
for C20H25CIN2O (6): C, 69.7; H, 7.3; N, 8.1. Found: C, 68.4;
H, 7.52; N, 8.0.
Here we report results from our continued effort to un-
derstand the mechanism of enantiodifferentiation in the
hydrogenation of alkenoic acids. Several CD derivatives,
modified at the quinuclidine nitrogen and/or at the (C-9)–
OH, have been synthesized and tested in the hydrogena-
tion of trans-2-methyl-2-butenoic acid (tiglic acid). Beside
the catalytic experiments, ab initio calculations were per-
formed to confirm the possible structure of the diastere-
omeric transition complex formed between CD and the re-
actant carboxylic acid.
EXPERIMENTAL
N,O-dibenzyl- (7), N,O-diallyl- (8), and N-benzyl O-allyl
cinchonidine bromide (9): General procedure. Modifier 5
(0.93 g, 2 mmol) (N-allyl cinchonidine bromide for 8, which
was synthesized by the same method as 3 using allyl bro-
mide) was dissolved in methylene chloride (20 ml). Benzyl
or allyl bromide (6 mmol) was then added to the solution
followed by 50% aqueous NaOH (3.2 g, 40 mmol). After 4 h
reaction time the organicphase wasseparated and the aque-
ous phase was washed with methylene chloride (3 10 ml).
The combined methylene chloride extract was dried over
MgSO4. The solvent was evaporated and the crude product
was suspended in ether (30 ml), stirred for an additional
4 h, and filtered. Purification by column chromatography
(CH2Cl2 : MeOH = 10 : 1) afforded the title compounds (7,
0.64 g, 58% ; 8, 0.65 g, 71% ; 9, 0.599 g, 61% ). Anal. Calcd for
C33H35BrN2O (7): C, 71.4; H, 6.4; N, 5.04. Found: C, 69.2;
H, 6.2; N, 5.0. Anal. Calcd for C25H31BrN2O (8): C, 65.9; H,
6.9; N, 6.2. Found: C, 66.2; H, 7.0; N, 6.4. Anal. Calcd for
C29H33BrN2O (9): C, 68.9; H, 6.6; N, 5.5. Found: C, 67.2; H,
6.8; N, 5.7.
Tiglic acid (98% , Aldrich), DBU (1,8-diazabicyclo[5.4.0]
undec-7-ene, >99% , Fluka), and toluene (99.7% , Riedel-
deHae¨n) were used as received. Tetrahydrofurane
(>99.5% , Scharlau) and n-pentane (>99.5% , Fluka) were
purified over sodium wire and distilled prior to use.
Cinchonidine Derivatives
Cinchonidine hydrochloride (1, Sigma) and N-benzylcin-
chonidine chloride (4, Fluka) were used as received. The
other chiral modifiers were readily prepared from CD.
These compounds were identified by elemental analysis,
NMR, and mass spectroscopy. The results of the spec-
1
troscopic methods are not shown here. H NMR spectra
were recorded on a Bruker Advance DPX 300 operating at
300 and 75 MHz. Mass spectra were recorded on Hitachi-
Perkin-Elmer RMU-6M for EI; VG-ZAB2-SEQ for FAB
in 3-nitrobenzyl alcohol matrix (3-NOBA).
Cinchonidine hydrobromide (2). CD (1.18 g, 4 mmol)
was dissolved in methylene chloride (50 ml) and cooled
in an ice bath. Hydrobromic acid (48% , 0.67 g, 4 mmol)
9-Deoxy cinchonidine (10). 9-Deoxy cinchonidine was
was added dropwise with continuous stirring and the re- prepared according to a previously reported method (16).
sulting mixture was stirred for 1 h. The solvent was evapo- CD (3.00 g, 0.01 mol) was dissolved in a 2-propanol–2 M
rated and the residue was purified by flash chromatography aqueous HCl mixture (700 ml, 23% v/v for HCl) and ir-
(CH2Cl2 : MeOH = 10 : 1) to afford 2 as white colorless crys- radiated with a medium pressure UV lamp equipped with
tal (1.36 g, 91% ). Anal. Calcd for C19H23BrN2O (2): C, 60.8; a Pyrex filter and cooled by tap water. The solution was
H, 6.2; N, 7.5. Found: C, 60,7; H, 6.2; N, 7.4.
flushed with nitrogen for 20–30 min prior to and during

162
BORSZEKY ET AL.
irradiation. After 24 h stirring the solvent was evapo- Theoretical Calculations
rated under reduced pressure. The residue was treated with
The intermolecular interaction between CD and acrylic
6 M aqueous NaOH and then extracted with chloroform
and purified by flash chromatography (n-hexane : acetone :
diethylamine = 30 : 18 : 1) to afford 11 as an oily product
(320 mg, 11.5% ). Anal. Calcd for C19H22N2 (10): C, 81.9; H,
7.9; N, 10.0. Found: C, 80.0; H, 7.8; N, 10.2.
acid (the simplest model of an , -unsaturated carboxylic
acid) has been investigated by ab initio calculations at
Hartree Fock level of theory using a 4-31G standard ba-
sis set. Complete geometry optimizations were performed
by relaxing all intra- and intermolecular degrees of free-
dom. To mimic flat adsorption of the acid as well as the
quinoline moiety of CD on the Pd surface, some restricted
calculations were performed where the quinoline moiety
and the acid were kept coplanar relaxing all other degrees
of freedom. For the calculations only CD protonated at
the quinuclidine nitrogen was considered. The calculations
were performed using GAUSSIAN94 (17) on a HP/Convex
Exemplar SPP2000/X-32 and a HP Model 735 workstation.
O-methyl cinchonidine (11). Under nitrogen atmo-
sphere, 0.53 g KH (0.013 mol, 20% suspended in oil) was
washed three times with n-pentane, suspended in dry THF
(40 ml), and cooled to 0 C. CD (3.24 g, 0.011 mol) was
added to the slurry in small portions. The reaction mix-
ture was stirred for 30 min at 0 C and then 30 min at
50 C. To the cooled reaction mixture methyl iodide (2.41 g,
0.017 mol) was added gradually and stirred for 1 h at
room temperature. The mixture was then cooled to 0 C
and carefully quenched with 50 ml water. The aqueous
phase was extracted with ethyl acetate (3 15 ml). The or-
ganic layer was washed with brine and dried over Na2SO4.
After evaporating the solvent under reduced pressure,
the residue was purified by flash chromatography using a
n-hexane : acetone : diethylamine = 40 : 18 : 1 solvent mix-
ture. Recrystallization from n-hexane gave 11 as colorless
crystal (2.0 g, 65% ). Anal. Calcd for C20H24N2O (11): C,
77.9; H, 7.8; N, 9.0. Found: C, 77.7; H, 8.1; N, 9.3.
RESULTS
Hydrogenation of Tiglic Acid over Pd/Alumina, Modified
by CD Derivatives
The efficiency of CD and its derivatives as chiral mod-
ifiers has been tested in the enantioselective hydrogena-
tion of tiglic acid to 2-methyl-2-butanoic acid over a 5 wt%
Pd/alumina catalyst. Conversions and enantioselectivities
achieved in 1.5 h are shown in Table 1. In this time period
full conversion could not be reached when the modifier
contained a halide anion (1–9). Modifiers with bromide an-
ion had a stronger inhibitory effect on the catalyst than the
Hydrogenation of Tiglic Acid
The reactions were performed in a 100-ml stainless steel
autoclave equipped with magnetic mixing. A 50-ml glass in-
let with a Teflon cap and stirrer was used to keep the system
inert. Hydrogen consumption was followed by a constant
pressure–constant volume apparatus (Bu¨chi Pressflow Gas
Controller). A 5 wt% Pd/Al2O3 catalyst (Engelhard 40692,
D = 0.21 as determined by TEM, BET = 200 m2 g 1) was
used for hydrogenations. In reactions using CD derivatives
the conditionswere asfollows:500mgtiglicacid, 20mgcata-
lyst (without pretreatment), 2 mg modifier, 15 ml toluene,
15 C, and 40 bar hydrogen pressure.
TABLE 1
Conversion and Enantioselectivity in the Hydrogenation of Tiglic
Acid over 5 wt% Pd/Alumina Modified with CD Derivatives
The standard conditions for experiments with base ad-
ditive were as follows: 100 mg tiglic acid, 10 mg catalyst,
3 mg CD, 14 ml toluene, 15 C, and 60 bar. The catalyst was
reduced in the presence of CD in 12 ml toluene at 30 bar
hydrogen pressure in the autoclave, resulting in the forma-
tion of 10,11-dihydrocinchonidine. Then tiglic acid and the
appropriate amount of base (1,8-diazabicyclo[5.4.0]undec-
7-ene, DBU) were added in 2 ml toluene. After reaction
the catalyst was filtered off, the filtrate acidified with 5 M
aqueous HCl and the organic layer analyzed.
The reaction mixtures were analyzed by gas chromatog-
raphy using a chiral column coated with CP-cyclodextrin-
2,3,6-M-19 (Chrompack). Enantiomeric excess is expressed
as ee (% ) = 100 |(R S)|/(R + S), with a reproducibility
of 0.5% .
Modifier
R1
R2
X
Conversion (% )
ee (% ) (S)
CD
1
2
3
4
5
6
7
8
OH
OH
OH
OH
OH
OH
OH
O-benzyl
O-allyl
O-allyl
—
—
—
Cl
Br
Cl
Cl
Br
Cl
Br
Br
Br
—
—
100
47
21
29
40
23
35
27
25
7
38
26
29
17
18
24
1
1.5
2
2
H
H
Allyl
Benzyl
Benzyl
CH3
Benzyl
Allyl
Benzyl
—
9
10
11
100
100
1
2
OCH3
—

ENANTIOSELECTIVE HYDROGENATION OF UNSATURATED ACIDS
163
chloride salts. Incomplete reactions in the presence of dif-
ferent CD halide salts (1–9) can be explained by the general
observation that strongly adsorbing halogen ions (adsorp-
tion strength: I > Br > Cl > F ) can result in deactiva-
tion of platinum metal catalysts (18).
Beside the poisoning effect of these modifiers, also the
enantioselectivities were lower compared to that induced
by CD. When applying CD-hydrohalide salts (1 and 2) the
ee decreased by approximately 10% . Modifiers (3–5), in
which the quinuclidine nitrogen was quaternerized by an
allyl or a benzyl group, still induced a moderate ee. How-
ever, the N-methylated derivative (6) was inefficient as a
chiral modifier for Pd. Similar results were obtained when
either a benzyl or an allyl group was attached to the quinu-
clidine nitrogen as well as to the alcoholic OH of CD (7–9).
Derivatives of CD, in which the quinuclidine nitrogen was
not modified but the OH group was either removed (10)
or methylated (11), did not inhibit the hydrogenation reac-
tion, but provided essentially a racemic mixture of 2-methyl
butanoic acid. Apparently, protection by an allyl or a ben-
zyl group, or elimination of the (C-9)–OH group of CD,
destroys the enantiodifferentiating ability of modifiers.
The purpose of quaternarization of the quinuclidine ni-
trogen was to prevent the protonation of this nitrogen
atom by the reactant alkenoic acid. Moderate enantios-
electivities were still achieved with modifiers 3–5, which
would suggest that the quinuclidine nitrogen–tiglic acid
interaction via hydrogen bonding is not of vital impor-
tance in the enantiodiscriminating step. In a control ex-
periment the hydrogenation of tiglic acid in the presence
of N-benzyl cinchonidine chloride (4) was interrupted in
the early stage of the reaction (after 7 min) and the fil-
trate was analyzed by NMR spectroscopy. It was found
that already 60% of 4 underwent hydrogenolytic deben-
zylation (Scheme 2). Note that Pd is the most active cata-
lyst for debenzylation (19), and the ease of hydrogenolysis
of amines at room temperature and atmospheric pres-
sure follows the following order: quaternary ammonium
salts > tertiary > secondary > primary amines (20). At
higher pressure and elevated temperature this sequence
is reversed (21). It is assumed that hydrogenolysis of the
SCHEME 3. Protonation of DBU.
N-allyl derivative (3) occurs similarly. We can conclude
that when using modifiers 3–5 the actual effective modi-
fier which affords enantiodifferentiation is (dihydro)-CD
hydrochloride, which explains the unexpectedly high ee.
On the contrary, hydrogenolysis of the N-methyl group was
not detectable by NMR, which explains the very low final
ee (1% ). The O-allyl and O-aryl derivatives 7–9 are more
resistant to hydrogenolysis under the conditions applied,
as indicated by the very low ees. These results suggest that
both the quinuclidine nitrogen and the OH group of CD are
involved in the interaction with the reactant in the enantio-
differentiating step.
Hydrogenation of Tiglic Acid in the Presence
of Strong Base Additive
In order to gain more information on the interaction
between the alkenoic acid and CD we studied the effect
of a strong base on the enantioselective hydrogenation
of tiglic acid. The bulky, strong N-base, 1,8-diazabicyclo
[5.4.0]undec-7-ene (DBU, pKa = 23.9 (22)), was chosen
for this purpose. Note that the pKa value of the quinucli-
dine nitrogen of CD is only 10 (23). DBU is a sterically
hindered bicyclic amidin base which reacts with carboxylic
acidsto form a relativelylarge delocalized cation (Scheme 3)
(24). Besides, DBU was resistant against hydrogenation
under the reaction conditions applied. The base was added
to the reaction mixture prior to hydrogenation in an apolar
medium. The influence of DBU/tiglic acid molar ratio on
the ee is depicted in Fig. 1. Till about 0.5 eq of DBU the
SCHEME 2. Partial hydrogenation of N-benzyl- and allyl CD deriva-
tives (3–5) over Pd/alumina during tiglic acid hydrogenation.
FIG. 1. Effect of DBU on ee in the hydrogenation of tiglic acid in
toluene over Pd/alumina modified with CD.

164
BORSZEKY ET AL.
ee barely changed, while an abrupt decrease was observed
when this amount was exceeded. The base had only a
minor effect on the rate of hydrogenation: the initial rate
4
1
1
varied between 1.8 and 2.5 10 mol H2 g (cat.) s
independent of the amount of base additive.
,
In the concentration range where the variation of ee is
small (Fig. 1), there are two major species present in equilib-
rium: the acid dimer and DBU–(acid)2 (dimer) salt. Above
0.5 eq of DBU the dominant species in the liquid phase
are expected to be the DBU–(acid)2 (1 : 2) and DBU–acid
(1 : 1) salts. Apparently, the ee diminishes with increasing
amounts of 1 : 1 salt, but is barely influenced by the presence
of 1 : 2 salt.
FIG. 3. Coplanar arrangement of acrylic acid dimer–CD species.
to the quinoline moiety of CD, as shown in Fig. 3. The ge-
ometrical arrangement of the (C-9)–OH and the N–H of
protonated CD is ideally suited to form a hydrogen bonded
network (27).
Another adduct which could be located on the potential
energy surface is depicted in Fig. 4. Here, both the quinu-
clidine nitrogen and the OH group of CD form hydrogen
bonds to the O atoms of the acrylic acid monomer. Sev-
eral unsuccessful attempts were made to locate a structure
with both N–H O and O–H O hydrogen bonds by fix-
Theoretical Calculations
Ab initio calculations were performed to gain some in-
sight into the possible structure of the CD–alkenoic acid
activated complex. At first the structure of a CD–(acid)2
species was optimized. For this purpose the simplest unsat-
urated carboxylic acid, acrylic acid was used. As starting
geometry conformation Open(3) was chosen for CD, since
it has been shown by NMR experiments that CD adopts
this conformation in the presence of acids (25–27). A pos-
sible geometry of a CD–(acrylic acid)2 intermediate has
been calculated in which both the {CD}NH O{acid} and
{CD}OH O{acid} hydrogen bonds are involved, as shown
in Fig. 2. This adduct exhibits strong stabilization with re-
spect to species where only one acid molecule is interact-
ing with CD. The analogous activated intermediate was ob-
tained when the two acid molecules were fixed coplanar
==
ing the quinoline part of CD and the C C double bond of
acrylic acid in one plane. It seemed that while fixed in one
plane the hydrogen bonded network is too stressed, lead-
ing to rupture at the O–H O hydrogen bond which is the
weakest connection. The N–H O bond is much stronger
than the O–H O bond since the most important contribu-
tion to the bonding is ion–ion interaction in the former and
ion–dipole interaction in the latter. The binding energies of
the optimized species and acrylic acid dimer are shown in
Table 2. Note the high stability of the 2 : 1 species, and the
significant binding energy of the acid dimer.
FIG. 2. Hydrogen bonding in the energetically most stable acrylic acid
dimer–CD species.
FIG. 4. Acrylic acid monomer–CD species.

ENANTIOSELECTIVE HYDROGENATION OF UNSATURATED ACIDS
165
TABLE 2
vation clearly contradicts their final conclusion concerning
the crucial role of the basic quinuclidine nitrogen in the
reactant–modifier interaction, no explanation was given for
this discrepancy. We propose that under the conditions ap-
plied (prehydrogenation of catalyst and modifier for 20 min
prior to hydrogenation of (E)- -phenylcinnamic acid) a
facile debenzylation of 4 should occur, which results in CD
hydrochloride as the actual modifier.
Binding Energies of Acrylic Acid Dimer
and CD-Acrylic Acid Species
Species
Acid dimer
1E (kcal/mol)
20.8
19.2
40.7
27.1
CD–acrylic acid
CD–(acrylic acid)2
CD–(acrylic acid)2 planar
None of the CD derivatives protected at both the quinu-
clidine nitrogen and (C-9)–O atoms (7–9) were efficient as
chiral modifiers (Table 1). Two other modifiers, 9-methoxy-
CD (11) and deoxy-CD (10), were also tested in tiglic acid
hydrogenation. The very low ee achieved with these com-
pounds confirm that also the alcoholic OH of CD plays a
crucial role in enantiodifferentiation. Moreover, the inef-
ficiency of 11 indicates that the chiral modifier should be
able to act as a donor in a hydrogen bonding interaction.
On the basis of these experimental observations, com-
pleted with theoretical calculations at the ab initio level, we
propose that in the transition complex CD interacts with
an acid dimer via two hydrogen bonds. The quinuclidine
nitrogen is protonated by the acid dimer and the OH of CD
acts as a hydrogen donor. Possible structures of this adduct
are shown in Figs. 2 and 3. In the latter case the adduct
was optimized assuming a parallel adsorption on an ideal
flat Pd surface. However, as Fig. 2 demonstrates, a nonflat
surface would not hinder the appropriate interactions be-
tween Pd, CD and alkenoic acid. If we assume that the ad-
dition of hydrogen occurs from the catalyst side (“bottom-
DISCUSSION
Developing a mechanistic model for the enantiodiffer-
entiation in alkenoic acid hydrogenation over the Pd–CD
system requires clarification (i) whether in the enantiodis-
criminating step the reactant and the chiral modifier are an-
chored via single or double “docking” (H-bondingand ionic
interaction), and (ii) of the stoichiometry of the transition
complex. These questions have been addressed in some re-
cent papers (12, 14, 15) but the conclusions of various re-
search groups are contradictory. The experimental results
and theoretical calculations presented above suggest that
none of the former models (including our proposal (12))
are really correct.
Application of N- and O-derivatized CD as chiral modi-
fiers (3–11) demonstrated that both the quinuclidine nitro-
gen and the (C-9)–OH functional group of CD are involved
in the interaction with the alkenoic acid in the enantiodif-
ferentiating step over the Pd surface. N-methylation of CD
and alkylation or removal of the OH group resulted in an
almost complete loss of enantioselectivity. HCl and HBr
salts of CD (1 and 2) were less efficient than CD, but here
the drop in rate and ee is likely due to the poisoning effect
of halide ions, rather than protonation of the N-atom.
Application of N-allyl and N-benzyl CD (3–5) afforded
misleading results: hydrogenolysis of the C–N bonds of the
protecting groups was found to be faster than hydrogena-
tion of the alkenoic acid, and the unexpectedly “high” ee is
attributed to the enantiodifferentiation induced by 10,11-
dihydrocinchonidine hydrochloride formed in the rapid,
undesired reaction. (Note that saturation of the vinyl group
at 10,11 position of CD is always fast even at atmospheric
pressure.) In case of N-methyl CD (6) hydrogenolysis of
the alkyl group could be excluded by NMR analysis. As
the quaternarized quinuclidine nitrogen of 6 could not take
part in H-bonding with tiglic acid, the ee approached zero
indicating that an N–H O interaction between tiglic acid
and CD is essential for enantiodifferentiation.
==
side” syn addition (28)) then one of the C C double
bonds in the acid dimer should point toward the quinoline
ring of CD. As we pointed out earlier (12), only this ge-
ometry affords the (S) product as the major enantiomer,
assuming that the reactant acid molecules in the dimer
are present in the thermodynamically more stable trans
position.
In order to get further confirmation concerning the na-
ture of interaction between CD and alkenoic acid, we de-
signed a series of hydrogenation experiments in the pres-
ence of a bulky N-base (Fig. 1). This base (DBU) is orders of
magnitude stronger than the quinuclidine nitrogen of CD,
a feature which together with the steric hindrance imposed
by its bulkiness, prevents any interaction between the quin-
uclidine nitrogen of CD and the tiglic acid salt. However, up
to a DBU : tiglic acid = 1 : 2 molar ratio neither the reaction
rate nor the ee decreased. In this concentration range tiglic
acid ispresent asfree dimersand DBU : tiglicacid = 1 : 2salt
in varying ratio as dominant species. The almost constant
initial rate and ee, observed with increasing amounts of the
1 : 2 salt, indicate that this species does not disturb the enan-
tiodifferentiation. A feasible explanation is that CD can
replace the much stronger base DBU due to the stronger
interaction between CD and the dimer, as compared to the
interaction between DBU and tiglic acid dimer (Scheme 4).
Interestingly, N-benzyl cinchonidine (4) was already
tested in the enantioselective hydrogenation of (E)- -
phenylcinnamic acid over Pd/TiO2 (15). The authors re-
ported 57% ee, as compared to 61% achieved with CD
under otherwise identical conditions. Though this obser-

166
BORSZEKY ET AL.
unfavorable at high acid concentration and its adsorption
on the palladium surface is sterically hindered. Formation
of a CD : acid 1 : 2 transition complex is more feasible.
ACKNOWLEDGMENTS
SCHEME 4. Favored and unfavorable exchange between CD and
DBU in DBU–(tiglic acid)2 and DBU–tiglic acid salts, respectively.
We are thankful to M. von Arx for the synthesis of N-methyl cinchoni-
dine chloride. Financial support from the Swiss National Science Founda-
tion is gratefully acknowledged.
This behavior corroborates that not only a base–acid inter-
action exists between CD and the acid dimer (8, 12, 14), but
the H-bonding between the OH group of CD and the acid
dimer also contributes to the total interaction energy in the
adduct (15).
REFERENCES
1. Rieu, J. P., Boucherle, A., Cousse, H., and Mouzin, G., Tetrahedron 42,
4095 (1986).
2. Shen, T. Y., Angew. Chem. Int. Ed. Engl. 11, 460 (1972).
3. Noyori, R., and Takaya, H., Acc. Chem. Res. 23, 345 (1990).
4. Zhang, X., Mashima, K., Koyano, K., Sayo, N., and Kumobayashi, H.,
Tetrahedron Lett. 32, 7283 (1991).
5. Uemura, T., Zhang, X., Matsumura, K., Sayo, N., Kumobayashi, H.,
Ohta, T., Nozaki, K., and Takaya, H., J. Org. Chem. 61, 5510 (1996).
6. Baiker, A., Curr. Opin. Solid State Mater. Sci. 3, 86 (1998).
7. Wells, P. B., and Wilkinson, A. G., Top. Catal. 5, 39 (1998).
8. Borszeky, K., Mallat, T., and Baiker, A., Catal. Lett. 41, 199 (1996).
9. Nitta, Y., and Kobiro, K., Chem. Lett. 897 (1996).
10. Smith, G. V., Cheng, J., and Song, R., Catal. Lett. 45, 73 (1997).
11. Barto´k, M., Wittmann, G., Go¨ndo¨s, G., and Smith, G. V., J. Org. Chem.
52, 1139 (1987).
12. Borszeky, K., Mallat, T., and Baiker, A., Tetrahedron: Asymmetry 8,
3745 (1997).
13. Barrow, G. M., and Yerger, E. A., J. Am. Chem. Soc. 76, 5211 (1954).
14. Hall, T. J., Johnston, P., Vermeer, W. A. H., Watson, S. R., and Wells,
P. B. , Stud. Surf. Sci. Catal. 101, 221 (1996).
15. Nitta, Y., and Shibata, A., Chem. Lett. 161 (1998).
16. Stenberg and Travecedo, J. Org. Chem. 35, 4131 (1970).
17. Frisch, M. J., Trucks, G. W., Schlegel, H. B., Gill, P. M. W., Johnson,
B. G., Robb, M. A., Cheeseman, J. R., Keith, T., Petersson, G. A.,
Montgomery, J. A., Raghavachari, K., Al-Laham, M. A., Zakrzewski,
V. G., Ortiz, J. V., Foresman, J. B., Cioslowski, J., Stefanov, B. B.,
Nanayakkara, A., Challacombe, M., Peng, C. Y., Ayala, P. Y., Chen,
W., Wong, M. W., Andres, J. L., Replogle, E. S., Gomperts, R., Martin,
R. L., Fox, D. J., Binkley, J. S., Defrees, D. J., Baker, J., Stewart, J. P.,
Head-Gordon, M., Gonzalez, C., and Pople, J. A., Gaussian Inc.,
Pittsburgh PA 1995.
Above a DBU : tiglic acid = 1 : 2 molar ratio the ee drop-
ped linearly with increasing amount of base. The most fea-
sible explanation for the loss of ee is the presence of an
increasing amount of DBU : tiglic acid = 1 : 1 salt in this con-
centration range. The interaction with CD would require
then the breaking of hydrogen bonds in two DBU : tiglic
acid 1 : 1 salt in order to form a CD–(tiglic acid)2 interme-
diate. This is thermodynamically unfavorable compared to
the substitution of DBU in the DBU–(tiglic acid)2 species
(Scheme 4). Thus the hydrogenation in the presence of 1 eq
DBU results in drastic loss of ee (10% ee). Note that the
same low ee was obtained recently in the hydrogenation of
sodium (E)- -phenylcinnamate (15).
On the basis of a wealth of literature data (29) and our
IR study, we proposed earlier (12, 27) that the alkenoic
acid reactant is dominantly present as dimers in an apolar
medium, and this structure is preserved during interaction
with CD on the Pd surface. As no direct (spectroscopic)
confirmation is available yet for this assumption, the re-
sults obtained in the presence of DBU in toluene provide
important indirect evidence for the proposed mechanistic
model.
18. Rylander, P. N., “Catalytic Hydrogenation over Platinum Metals.”
Academic Press, New York and London, 1967.
CONCLUSIONS
19. Prelog, V., Helv. Chim. Acta 36, 308 (1953).
Various CD derivatives modified at the quinuclidine ni-
trogen and/or at the (C-9)–OH group were synthesized and
tested in the enantioselective hydrogenation of tiglic acid.
The experiments confirmed a former proposal (15) that the
alcoholic OH group as well as the quinuclidine nitrogen of
cinchonidine are hydrogen bonded to the reactant alkenoic
acid in the enantiodifferentiation step. Hydrogenation ex-
periments in the presence of a strong and bulky base addi-
tive provided indirect evidence for our former assumption
that cinchonidine interacts with an alkenoic acid dimer on
the Pd surface. A feasible structure of the acid dimer–CD
intermediate which can easily adsorb on the catalyst surface
has been rationalized by ab initio calculation. The calcula-
tions indicated that formation of a CD : acid 1 : 1 adduct is
20. O’Murchu, C., Tetrahedron Lett. 1969, 3231 (1969).
21. Mitsui, S., Kasahara, A., and Endo, N., J. Chem. Soc. Jpn. 75, 234
(1954).
22. Galezowski, W., Jarczewski, A., Stanczyk, M., Brzezinski, B., Bartl, F.,
and Zundel, G., J. Chem. Soc. Faraday Trans. 93, 2515 (1997).
23. Budavari, S., “The Merck Index,” 11th ed. Merck, Rahway, NJ, 1989.
24. Ono, N., Yamada, T., Saito, T., Tanaka, K., and Kaji, A., Bull. Chem.
Soc. Jpn. 51, 2401 (1978).
25. Bu¨rgi, T., and Baiker, A., J. Am. Chem. Soc. 120, 12920 (1998).
26. Dijkstra, G. D. H., Kellogg, R. M., and Wynberg, H., Recueil des
Travaux Chimiques des Pays-Bas 108, 195 (1989).
27. Ferri, D., Bu¨rgi, T., and Baiker, A., J. Chem. Soc. Perkin Trans. 2. 1305
(1999).
28. Siegel, S., Adv. Catal. 16, 123 (1966).
29. Patai, S., “The Chemistry of Carboxylic Acids and Esters.” Inter-
science, London, 1969.