On the role of oxygen in the liquid-phase aerobic oxidation of alcohols on palladium

Article abstract of DOI:10.1016/S0021-9517(02)93723-1

The effect of oxygen on the liquid-phase aerobic oxidation of alcohol was studied using the conversion of cinnamyl alcohol over Pd-based catalysts as a sensitive test reaction. Dehydrogenation and oxidative dehydrogenation of cinnamyl alcohol followed the same reaction pathways independent of the presence or absence of oxygen and reaction conditions. Hydrogen was detected on the metal surface during reactions. Catalyst deactivation in argon was attributed to decarbonylation reactions and site blocking by CO. Introduction of molecular oxygen significantly enhanced alcohol conversion rate by a factor of ≤ 285 due to oxidative removal of CO. In situ attenuated total reflection-IR spectroscopy corroborated the strong adsorption of CO on Pd/Al₂O₃ and its rapid removal by oxygen. Oxygen played a key role in the continuous oxidative removal of CO and other degradation products from the active sites. This oxidative cleaning of the metal surface allowed a high rate of alcohol dehydrogenation even when the oxidation of the co-product hydrogen was slow and incomplete.

Full text of DOI:10.1016/S0021-9517(02)93723-1

Journal of Catalysis 211, 244–251 (2002)
doi:10.1006/jcat.2002.3723
On the Role of Oxygen in the Liquid-Phase Aerobic Oxidation
of Alcohols on Palladium
Csilla Keresszegi, Thomas B u¨ rgi, Tamas Mallat, and Alfons Baiker1
Laboratory of Technical Chemistry, Swiss Federal Institute of Technology (ETH), H o¨ nggerberg HCI, CH-8093 Zurich, Switzerland
Received April 9, 2002; revised June 7, 2002; accepted June 7, 2002
out by replacing oxygen with a hydrogen acceptor such as
The mechanism of alcohol oxidation was investigated using the quinone, nitrobenzene, or an olefin (8). This approach has
conversion of cinnamyl alcohol (1) over Pd-based catalysts as a been applied recently in the transfer dehydrogenation of
sensitive test reaction. Studies in a slurry reactor revealed that de-
hydrogenation and oxidative dehydrogenation of 1 follow the same
reaction pathways independent of the presence or absence of oxy-
gen and reaction conditions. Hydrogenation and hydrogenolysis
side reactions indicated the presence of hydrogen on the metal sur-
the secondary role of oxygen. M u¨ ller and Schwabe demon-
face during reactions. Catalyst deactivation in Ar is attributed to
decarbonylation reactions and site blocking by CO. Introduction
of molecular oxygen induced a dramatic enhancement of alcohol
conversion rate by a factor of up to 285 due to oxidative removal of
allylic and aromatic alcohols on Pd using a cheap olefin as
a hydrogen acceptor (9, 10).
In situ catalyst potential measurements provided fur-
ther support to the dehydrogenation mechanism and to
strated for the first time that during oxidation of ethanol or
2-propanolthepotentialofPtwasinthe“hydrogenregion”;
i.e., the metal was partially covered by hydrogen despite the
CO. Strong adsorption of CO on Pd/Al2O3 and its rapid removal by presence of gaseous oxygen (11, 12). Some more examples
oxygen were corroborated by in situ ATR-IR spectroscopy. All these of this situation can be found in the recent literature
observations conform to a model according to which oxidation of (13–15).
1
follows the classical dehydrogenation mechanism, and the key
However, an increasing number of reports contradict this
concept and suggest that oxidation of some alcohols pro-
ceeds on partially oxygen-covered metal surface and the
actual concentration of adsorbed hydrogen during reaction
seems to be negligible (16–20). These observations can-
not easily be interpreted by the classical dehydrogenation
mechanism. A feasible interpretation could be an oxidative
dehydrogenation mechanism in which the adsorbed oxygen
role of oxygen is the continuous oxidative removal of CO and other
degradation products from the active sites. This oxidative cleaning
of the metal surface allows a high rate of alcohol dehydrogenation
even when the oxidation of the co-product hydrogen is slow and
incomplete. It is likely that the observed effects are not limited to
the oxidation of 1 on Pd, and regeneration of the active sites by
oxygen generally plays an important role during aerobic oxidation
of alcohols on platinum metals.
ꢀc 2002 Elsevier Science (USA)
Key Words: oxidation; dehydrogenation; palladium; cinnamyl al- or OH species play a crucial role. The rate-determining el-
cohol; deactivation; decarbonylation; CO poisoning; ATR-IR spec- ementary step might involve direct interaction of the ad-
troscopy.
sorbed oxidizing species with the adsorbed reactant or its
partially dehydrogenated intermediate (3, 7).
Several kinetic studies indicated
a
Langmuir–
INTRODUCTION
Hinshelwood-typebehavior:asurfacereactionbetweenthe
adsorbed reactant and the dissociatively adsorbed oxygen
in the rate-determining step (see, e.g., (21, 22)). In recent
studies Markusse et al. suggested that oxygen-assisted
dehydrogenation would be the rate-determining step
20, 23). However, kinetic analysis requires working under
conditions free from mass-transport limitations, which
conditions lead to a rapid catalyst deactivation due to
Platinum-group metals (particularly Pt and Pd) and gold
are active catalysts in the liquid-phase oxidation of alco-
hols, diols, and carbohydrates with molecular oxygen un-
der close to ambient conditions (1–7). A dehydrogenation
mechanism of the reaction is mostly accepted, assuming
that the role of oxygen is to oxidize the co-product hydro-
gen formed in the rate-determining step and thus shift the
equilibrium toward the carbonyl compound. The secondary
importance of oxygen in this mechanism is supported by
Wieland’s early observation that the reaction can be carried
(
“
overoxidation” of the metal surface (24, 25)—an unpleas-
ant obstacle in determining the real initial rate. It was
also shown that multiple steady states may complicate the
kinetic analysis (26, 27).
We can conclude that the apparent general agreement
in the oxidative dehydrogenation mechanism of alcohol
oxidation includes contradictory opinions concerning the
1
To whom correspondence should be addressed. Fax: +41 1 632 11 63.
E-mail: baiker@tech.chem.ethz.ch.
244
0
ꢀ
021-9517/02 $35.00
c 2002 Elsevier Science (USA)
All rights reserved.

LIQUID-PHASE AEROBIC OXIDATION OF ALCOHOLS
245
real nature of elementary steps and the role of oxidizing in 5 ml toluene. Periodically withdrawn samples were an-
species. Some years ago we proposed the blocking of active alyzed by GC (Thermo Quest Trace 2000, equipped with
sites by strongly adsorbed alcohol degradation products as an HP-FFAP capillary column and an FID). Products were
a possible explanation for this contradiction (18, 19, 28, 29). identified by GC-MS and GC analysis of the corresponding
Oxidative removal of these impurities accelerates the target authentic samples.
reaction; thus, adsorbed oxygen apparently plays an active
ATR-IR Spectroscopy
(
direct) role in the reaction mechanism.
This question is addressed here involving catalytic and in
The transformation of cinnamyl alcohol (1) on Pd was
studied by in situ attenuated total reflection (ATR) infrared
spectroscopy (32). A catalyst layer on the internal reflection
element was used for this purpose. A slurry of 10 mg 5 wt%
Pd/Al2O3 in 5 ml toluene was stirred for 30 min. Then a
situ spectroscopic methods. The model reaction is the trans-
formation of an allylic alcohol, trans-cinnamyl alcohol. Aer-
obic oxidation of allylic alcohols on Pt- and Pd-based cata-
lysts under appropriate conditions affords sometimes
excellent yields to α,β-unsaturated aldehydes (29, 30). In
the absence of oxygen the selectivity is low due to side re-
actions in which the reactant alcohol acts as a hydrogen
acceptor (10). The high reactivity of reactant and product
offersthepossibilitytodemonstratetheimportanceofcom-
peting side reactions, and to clarify the role of oxygen in
these reaction pathways.
◦
ZnSeinternalreflectionelement(IRE, 45 , 50 × 20 × 2mm,
KOMLAS) was covered by dropping the slurry on its one
side. The solvent was allowed to evaporate and the pro-
cedure was repeated three times. After drying for several
hours in air the catalyst layer was ready for use. The cata-
2
lyst exposed to the reaction mixture, a film of 3.1-cm area,
amounted to 1–1.5 mg.
ATR spectra were recorded using a home-built stainless-
steel flowthrough cell with a total volume of 0.077 ml (33).
The cell was heated to 60 C by a thermostat during the mea-
EXPERIMENTAL
◦
surements. After assembling, the cell was mounted onto an
ATR attachment (OPTISPEC) within the Fourier trans-
form IR spectrometer (Bruker, IFS-66/S) equipped with
an MCT detector. At the entrance and exit of the IRE the
IR beam was collimated to minimize signal contributions
from the O-ring region. Liquid (toluene or a toluenic so-
lution of 1) was flowed over the catalyst sample at a rate
Materials
The following commercial catalysts were used for alcohol
oxidation: 5 wt% Pd/Al2O3 (Johnson Matthey 324; mean
Pd particle size, 3.4 nm determined by TEM; calculated
dispersion, 0.34 (31)), 5 wt% Pt/Al2O3 (Engelhard 4759;
mean Pt particle size, 2.9 nm determined by TEM; calcu-
lated dispersion: 0.40), and 4 wt% Pd–1 wt% Pt–5 wt%
Bi/C (Degussa, CEF 196 XRA/W). trans-Cinnamyl alco-
hol (Acros, 98%) was purified by recrystallization from
−1
of 0.5 ml min by a peristaltic pump (ISMATEC Reglo
100) located behind the cell. Liquid was provided from two
separate glass bubble tanks where the liquids could be sat-
urated with gases. The flow from the two tanks was con-
trolled by a pneumatically activated three-way Teflon valve
1
petroleum ether (99.3% by H NMR and GC). Ethylene
glycol diacetate (Aldrich, 99+%, internal standard) and
toluene (J. T. Baker, >99.5%) were used as received. Gases
were of 99.999% purity (PANGAS).
(
PARKER PV-1-2324). After the sample compartment of
the spectrometer was purged with dry air for about 1 h, neat
toluene was admitted to the sample for 20 min. Then the
catalyst was reduced in situ by admitting toluene saturated
with H2 for 10 min. After purging with Ar-saturated toluene
for 10 min the catalyst was oxidized by flowing toluene satu-
rated with air. Then the toluenic solution of cinnamyl alco-
hol was pumped over the pretreated catalyst and the spectra
were recorded.
Catalytic Studies
The reactions were performed in a 150-ml flat-bottomed
glass reactor equipped with gas inlet, reflux condenser, ther-
mometer, inlet for taking samples, and magnetic stirring.
According to the general procedure, 0.10 g catalyst was pre-
reduced in situ by H2 for 10 min in 25 ml toluene, purged
with N2 for 5 min, and reoxidized by air for 30 min. The aim
of this reduction–oxidation cycle was to activate the cata-
lyst stored in air but to avoid the presence of any hydrogen
when introducing the reactant. The catalyst prereduction
RESULTS
Reaction Network
◦
temperature was lower by 20–25 C than the temperature
Preliminary studies revealed a complex reaction net-
of the alcohol oxidation reaction. After catalyst pretreat- work typical for the transformation of cinnamyl alcohol
−
1
ment the air flow was set in the range 5–60 ml min , or (1) to cinnamaldehyde (2) (Scheme 1). Changing the cata-
the reactor was purged with argon and a balloon was filled lyst composition (Pd/Al2O3, Pt/Al2O3, and Pd–Pt–Bi/C),
when the reaction was carried out in an inert atmosphere. reaction atmosphere (Ar or air), solvent (toluene, cyclo-
The reaction was started by injection of 1.0 g (7.5 mmol) hexane, or water with detergent), and reaction temperature
◦
cinnamyl alcohol and 1.1 g (7.5 mmol) internal standard (45–80 C) had large influences on the relative contribution

2
46
KERESSZEGI ET AL.
+
2 H
8
7
-
CO
- CO
+
+
2 H
- 2 H
OH
+ 2 H
O
O
5
4
- H
2
O
1
2
6
+
2 H
+
2 H
2 H
OH
- H O
3
2
SCHEME 1. Reaction network for the transformation of trans-cinnamyl alcohol (1) over alumina-supported Pd; all products were identified by
GC and GC-MS.
of various reactions but the overall scheme was valid for all Catalytic Dehydrogenation and Oxidation
cases.
Besides the target reaction, dehydrogenation of 1 to 2, a
of Cinnamyl Alcohol
Dehydrogenation of cinnamyl alcohol in Ar over
Pd/Al2O3 is characterized by a rapid loss of catalyst ac-
tivity. An example is shown in Fig. 1. Before reaction the
catalyst was activated by reduction with hydrogen and re-
oxidized in air to avoid the presence of surface hydrogen
at the beginning of alcohol transformation. After addition
of cinnamyl alcohol the first step is the reduction of the
oxidized metal surface by the alcohol reactant. The initial
rate of alcohol conversion related to the number of sur-
major reaction is the production of the saturated alcohol 3.
Formation of 3 may simply be interpreted as a transfer hy-
drogenation reaction in which the reactant itself plays the
role of hydrogen acceptor. A similar reaction is the transfor-
mation of cinnamaldehyde (2) to dihydrocinnamaldehyde
(6). We have shown recently (10) that many aromatic and
allylic alcohols can act as hydrogen donors and acceptors
on the Pd surface.
The appearance of 1-propenylbenzene (5) among the
products is a clear indication of hydrogenolysis of the C–O
bond in 1. Further hydrogenation of the alkenyl side chain
leads to propylbenzene (4), though this product may also
form from 3 by C–O-bond hydrogenolysis.
−
1
face Pd atoms (TOF) was 450 h . After 1 h the TOF was
−
1
only 10 h though the change in reactant concentration
was minor. Apparently, conversion of cinnamyl alcohol is
strongly poisoned on the reduced Pd surface. We propose
that the loss of activity is due to aldehyde decarbonylation
and the strong adsorption of CO, as shown in the follow-
ing text.
Decarbonylation of cinnamaldehyde (2) affords styrene
(8) and ethylbenzene (7) after hydrogenation. Alterna-
tively, decarbonylation of dihydrocinnamaldehyde (6) also
provides 7. Note that Pt group metals, and particularly Pd,
are good catalysts for aldehyde decarbonylation (34–36).
There was no indication of further oxidation of cin-
namaldehyde to cinnamic acid in toluene. The likely ex-
planation is that direct oxidation of the carbonyl group
is slow and the faster dehydrogenation via hydration of
the carbonyl group (the dominant mechanism in aqueous
medium) is hampered by the stability of the aldehyde and
the neutral apolar organic medium (29). Similarly, no C == C
After 90 min, Ar was replaced with air, which resulted in
a dramatic rate acceleration (TOF 2850 h ) and conver-
sion of the rest of the cinnamyl alcohol within 10 min. The
−
1
double-bond isomerization occurred, a reaction that is typi-
cal for aliphatic allylic alcohols, affording saturated aldehy-
des via enol-keto tautomerization (37–39). In cinnamyl al-
cohol the olefin function is in conjugation with the aromatic
electron sextet; thus neither isomerization of cinnamyl al-
cohol into dihydrocinnamaldehyde nor isomerization of
1
-propenylbenzene into allylbenzene took place.
In summary, formation of 3–7 indicates the presence of
FIG. 1. Conversion of cinnamyl alcohol (1) and selectivity to
cinnamaldehyde (2) in argon for 90 min and then in air. Conditions: 5 wt%
surface hydrogen during dehydrogenation or oxidation of
cinnamyl alcohol, and the appearance of 7 and 8 verifies
decarbonylation-type side reactions. This information is
used in the next section to interpret the catalytic results.
◦
Pd/Al2O3 prereduced by hydrogen and reoxidized by air in situ at 40 C;
◦
1
.0 g cinnamyl alcohol; 30 ml toluene; reaction temperature, 65 C; mixing
−
1
−1
frequency, 1250 min ; air flow rate, 60 ml min .

LIQUID-PHASE AEROBIC OXIDATION OF ALCOHOLS
247
TABLE 1
Product Distribution in the Transformation of trans-Cinnamyl Alcohol (1) over Pd/Al2O3a
Yield (%)b
Selectivity (%)
Entry
Atmosphere
Time (min)
Conv. (%)
2
3
4
5
6
7
8
2
3–7
1
a
Argon
Air
Air
90
+25
55
10.7
99.8
100
5.7
62.8
63.7
21.9
4.1
34.9
34.0
1.5
0
0.2
0.2
<0.1
0.5
0.5
0.7
0.2
0.1
0.6
0.4
0.2
<0.1
0.8
0.9
0.2
<0.1
0.1
53
63
64
89
45
37
36
9
1
b
c
2
3
Air
85
24.7
0.2
0.7
a
◦
Conditions: 0.10 g 5 wt% Pd/Al2O3 (prereduced by hydrogen and reoxidized by air in situ at 40 C); 1.0 g cinnamyl alcohol; 30 ml
◦
−1
−1
toluene; reaction temperature, 65 C; air flow rate, 20 ml min ; mixing frequency, 750 min . For identification of the products see
Scheme 1.
b
No other products were detected by GC.
The catalyst was prereduced before reaction with the reactant alcohol under N2 atmosphere for about 3 min.
c
major products were cinnamaldehyde (2) and 3-phenyl-1- 1 was even slower (entry 3). Note that the rate-enhancing
propanol (3). Interestingly, changes in selectivity with time effect of catalyst prereduction before the aerobic oxidation
and gas atmosphere were small (Fig. 1). The selectivity to of alcohols on platinum-group metals is well known (40).
2
was higher by 5–7% on the oxidized Pd surface at the More important is that the cumulative selectivity to hydro-
beginning of the reaction and after introduction of air. genation and hydrogenolysis products 3–7 was 9%, demon-
The reaction was repeated at a lower rate of oxygen trans- strating the presence of surface hydrogen on the partially
port by reducing the mixing frequency and the air flow rate oxidized Pd surface during aerobic oxidation.
(
Table 1, entry 1). The average TOF of alcohol conversion
These experiments indicate that the reaction rate and
−1
in air decreased from 2850 to 1150 h but the product dis- product distribution strongly depend on the amount of oxy-
tribution barely changed. Note that in the second part of gen present on the Pd surface, but in all cases alcohol dehy-
both reactions, in the aerobic oxidation, the rate-limiting drogenation (leading to partial hydrogen coverage) seems
step is the oxygen transport from the gas phase to the Pd to be the dominant mechanism. On the basis of the prod-
surface, which is necessary to avoid too high oxygen cov- uct distribution shown in Table 1 we assume that the rapid
erage leading to catalyst “overoxidation” and deactivation deactivation of Pd/Al2O3 in Ar and the sharp rate enhance-
(
24, 25).
The product distribution before the introduction of air the alcohol degradation product CO and to oxidative re-
and at the end of the aerobic oxidation is shown in Table 1 moval of these species, respectively. In the following we
entries 1a and 1b). As discussed earlier, products 3–7, show more evidence for this concept.
formed by hydrogenation and hydrogenolysis reactions, ev- The striking effects observed with Pd/Al2O3 could not
ment by introduction of air are due to catalyst poisoning by
(
idence the presence of hydrogen on the Pd surface. Intro- be corroborated with Pt/Al2O3 due to its very low activity.
duction of air after a 90-min reaction time disfavored the For example, repeating the experiment shown in Fig. 1 with
more demanding reaction, the hydrogenolysis of the C–O Pt/Al2O3 the TOF values were lower by up to three orders
bond (to produce 4 and 5) but barely hampered the satu- of magnitude. Thus, the differences due to variations in the
ration of the C == C bond (to produce 3, 6, and 7). Similarly, catalyst pretreatment procedure could not reliably be mea-
oxygen only moderately retarded the decarbonylation of sured. It has already been reported that Pt/Al2O3 is a poor
aldehydes 2 and 6 to the hydrocarbons 7 and 8. A feasible catalyst for cinnamyl alcohol oxidation in aqueous alkaline
explanation is that the amount of oxygen was sufficient to medium (29).
(
partially) oxidize the impurities on the Pd surface but the
Dehydrogenation and oxidation of cinnamyl alcohol on
surface oxygen concentration remained low and removal a Bi-promoted catalyst 4 wt% Pd–1 wt% Pt–5 wt% Bi/C
of hydrogen formed in the rapid dehydrogenation of 1 was provided further support for the presence of hydrogen on
incomplete.
the metal surface during reaction (Table 2). The amount of
The reaction in Table 1, entry 1, was repeated with only products 3–7 formed by hydrogenation or hydrogenolysis-
a 3-min period in Ar (Table 1, entry 2). The time required type reactions is considerable both in Ar and in air. The
for complete conversion of 1 in air increased but the prod- higher selectivity to cinnamaldehyde (2) in air (Table 2 and
uct distribution was very similar. The lower reaction rate is Fig. 2) is mainly attributed to suppression of transfer hydro-
attributed to the incomplete reduction of the Pd surface by genation of cinnamyl alcohol. Decarbonylation affording
the reactant alcohol.
7 and 8 was barely detectable over Pd–Pt–Bi/C. The much
When the aerobic oxidation was started without any pre- lower decarbonylation activity of this three-component
reduction with the reactant alcohol in Ar, the conversion of catalyst than of Pd/Al2O3 (last lines in Tables 1 and 2) is

2
48
KERESSZEGI ET AL.
TABLE 2
Product Distribution in the Transformation of trans-Cinnamyl Alcohol (1) over Pd–Pt–Bi/Ca
Yield (%)b
Selectivity (%)
Entry Atmosphere Conv. (%)
2
3
4
5
6
7
8
2
3–7
1
2
Argon
Air
7.1
34.6
4.9
31.2 2.0
1.1
0.1
<0.1
0.8
1.0
0.1
0.2 <0.1
0
<0.1
0.1
69
90
30
9.5
a
◦
Conditions: 0.10 g catalysts (prereduced by hydrogen and reoxidized by air in situ at 25 C); 1.0 g
◦
−1
cinnamyl alcohol; 30 ml toluene; reaction temperature, 45 C; mixing frequency, 750 min ; air flow rate,
−
1
2
0 ml min ; reaction time, 5 h. For identification of the products see Scheme 1.
b
No other products were detected by GC.
likely due to the promoting effect of Bi. It has been shown lyst seems to be negligible, based on the poor activity of
that Bi promotion can suppress CO formation during elec- Pt/Al2O3.
trocatalytic oxidation of ethyleneglycol (41).
The reaction in air was considerably faster than that in ATR-IR Spectroscopy
Ar, as expected, but the sudden catalyst deactivation after a
The importance of decarbonylation reactions during de-
few minutes of dehydrogenation in Ar, as observed over Pd/
hydrogenation of cinnamyl alcohol has been investigated
Al2O3, was absent in the three-component catalyst (Fig. 2).
by ATR-IR of the solid–liquid interface. Figure 3 shows
We attribute this difference to suppression of side reactions
the in situ ATR spectra, which were recorded during flow-
that result in strongly adsorbing by-products or to weaker
ing a solution of cinnamyl alcohol in toluene over the
adsorption of these by-products on the Bi-promoted Pd
Pd/Al2O3 catalyst. The solution was first saturated with Ar
surface. ThecontributionofPtinthethree-componentcata-
(
Fig. 3a), then with air (Fig. 3b), and subsequently with
CO (0.5% in Ar, Fig. 3c). The bands in Fig. 3a at 2000,
−
1
1
900, and 1840 cm are attributed to adsorbed CO that
forms during feeding the alcohol over the catalyst. The
FIG. 3. (a) In situ ATR spectra recorded while flowing a solution of
0.056 mol/L cinnamyl alcohol in Ar-saturated toluene over Pd/Al2O3. The
time lag between first (bottom) and last spectrum (top) is 17 min. (b) Spec-
tra recorded while subsequently flowing an identical solution but saturated
with air over the same sample. The time lag between first (top) and the
last spectrum (bottom) is 2 min. (c) Spectra recorded while subsequently
flowing dissolved CO (0.5% in Ar) in the same solution over the same
catalyst. Time between first (bottom) and last (top) spectrum is 10 min.
The background for the spectra shown in (a) and (b) was recorded before
admitting the alcohol. For the spectra in (c) the background was recorded
before admitting CO.
FIG. 2. (a) Conversion of cinnamyl alcohol (1) and (b) selectivity to
cinnamaldehyde (2) in argon and in air over a Pd–Pt–Bi/C catalyst. Con-
ditions: catalyst prereduced by hydrogen and reoxidized by air in situ at
◦ ◦
5 C; 1.0 g cinnamyl alcohol; 30 ml toluene; reaction temperature, 45 C;
2
−
1
−1
mixing frequency, 750 min ; air flow rate, 20 ml min .

LIQUID-PHASE AEROBIC OXIDATION OF ALCOHOLS
249
latter two bands are ascribed to bridged CO and the former significantly hampering the side reactions involving surface
one to linearly bound CO (42). The sharp bands at 1640 hydrogen. These observations question the validity of the
−1
and 1690 cm are associated with cinnamyl alcohol and general assumption that the role of oxygen would be oxi-
species adsorbed on the Al2O3 support. It should be em- dation of the co-product hydrogen, and thus regenerating
phasized that the spectra were recorded truly in situ, that is, free active sites and accelerating the overall reaction.
while the catalyst was working. GC analysis of the solution,
The present study corroborates our former proposal
which was passed through the cell, showed the production (18, 19, 28, 29) that in many liquid-phase alcohol oxidation
of cinnamaldehyde and 3-phenyl-1-propanol, the two ma- reactions the major role of oxygen is the oxidative removal
jor products over Pd, with yields over 0.5% for each. The of degradation products. The dramatic rate acceleration
spectra in Fig. 3a confirm that CO is built up on the alumina- achieved by introducing oxygen (Fig. 1) is not an indica-
supported Pd catalyst. The time difference between the first tion of a direct involvement of adsorbed oxidizing species
(
bottom) and the last recorded spectrum (top) was 17 min. inthealcoholdehydrogenationmechanismbutisratherdue
Figure 3b reveals that CO formed during dehydrogena- to removal of strongly adsorbed species and an increase in
0
tion of cinnamyl alcohol is rapidly removed when air is the number of active Pd sites available for alcohol dehy-
introduced into the system. The time difference between drogenation. Decarbonylation of cinnamaldehyde (2) and
thefirst(top)andthelastmeasuredspectrumisabout2min. dihydrocinnamaldehyde (6) affords CO and hydrocarbons
For comparison, Fig. 3c shows the CO spectra obtained on (styrene (8) and ethylbenzene (7), respectively, Scheme 1).
the same catalyst when the reactant solution was saturated Introduction of molecular oxygen suppressed but did not
with CO. The relative intensity of the various CO bands and eliminate this side reaction (Tables 1 and 2). In situ ATR-IR
their behavior with coverage is qualitatively the same in spectroscopy revealed a considerable CO coverage on the
Figs. 3a and 3c. For example, the low-wavenumber bridged Pd surface during dehydrogenation of cinnamyl alcohol in
CO band grows first, whereas the high-wavenumber CO Ar and proved the subsequent oxidative removal of CO by
band grows faster at higher coverage.
introducing air (Fig. 3).
Pd is one of the best decarbonylation catalysts (34–
3
6, 43). That is, the transformations 2 → 8 and 6 → 7 in
DISCUSSION
Scheme 1 are not only fast but also selective. The produced
hydrocarbons and CO2 (after oxidation of CO) can eas-
ily desorb from the metal surface and thus regenerate the
Oxidative Removal of Degradation Products
According to the classical dehydrogenation mechanism active sites. However, in other reactions and on other met-
of alcohol oxidation over platinum-group metals, the rate- als, oxidative removal of CO may not be sufficient to re-
limiting step is dehydrogenation on the metal surface and generate the alcohol dehydrogenation activity. In general,
the role of oxygen is to oxidize the co-product hydrogen to a decarbonylation reaction produces CO, hydrogen, and
water, thereby accelerating the overall reaction. Dehydro- Cn−1 hydrocarbon species, as shown by Barteau and co-
genation and oxidative dehydrogenation of primary unsat- workers under UHV conditions (35, 36). In addition to CO,
urated alcohols, such as cinnamyl alcohol, are a sensitive the hydrocarbon fragments formed in the decarbonylation
test reaction for studying the reaction mechanism and the reaction can cause catalyst poisoning. We have reported re-
role of oxygen in the aerobic oxidation of alcohols. The cat- centlythatalcoholdehydrogenationmaybeacceleratedbut
alytic studies, presented in Tables 1 and 2 and Figs. 1 and 2, also retarded by addition of an olefin as hydrogen acceptor
revealed that transformation of cinnamyl alcohol runs over (10). For example, styrene reduced the rate of dehydro-
the same reaction pathways independent of the presence or genation of 1-phenylethanol to acetophenone on Pd/Al2O3
absence of molecular oxygen, reaction conditions, and cata- by a factor of more than 30, compared to the reaction in
lyst composition. Formation of by-products 3–7 (Scheme 1) Ar without olefin. Note that styrene is the decarbonylation
demonstrated that there was always a sufficient amount of product from cinnamaldehyde (8 in Scheme 1).
hydrogen on the metal surface to allow hydrogenation and
Electrocatalyticstudiesoftheadsorptionandreactionsof
hydrogenolysis side reactions. Hence, these results prove allyl alcohol on Pt and Pd indicated that removal of strongly
that the classical dehydrogenation mechanism is valid for adsorbed hydrocarbon fragments (decarbonylation prod-
the oxidation of cinnamyl alcohol on Pd.
ucts) as easily desorbing hydrocarbons requires strongly
Formation of 3–7 via hydrogenation and hydrogenolysis reducing conditions (44, 45). It was also reported that the
of reactant and products even in the presence of molecular best regeneration process for a catalyst deactivated in al-
oxygen indicates that oxidation of the co-product hydrogen cohol oxidation is a high-temperature treatment in hydro-
by oxygen on the Pd surface is unexpectedly slow. The ex- gen (46). Hence, we assume that the observed low activity
periments in Fig. 1 and Table 1, entry 1, demonstrate that of Pt/Al2O3 in cinnamyl alcohol dehydrogenation may be
dehydrogenation of 1 can be accelerated by more than two due to catalyst poisoning by strongly adsorbed hydrocarbon
orders of magnitude by introducing oxygen and without species and by the possible oligomerization products. This

2
50
KERESSZEGI ET AL.
assumption explains the minor positive effect of oxygen on
Catalyst deactivation during aerobic oxidation of alco-
the dehydrogenation rate. Apparently, in this reaction Pd hols is commonly interpreted as overoxidation of the metal
is less sensitive to poisoning by the hydrocarbon fragments. catalyst. On the basis of Fig. 4 the distinction between cata-
A possible alternative explanation to the positive effect lyst poisoning by surface impurities and deactivation due
of oxygen introduction during dehydrogenation of 1 would to too-high oxygen coverage (overoxidation) is simple. An
be that in air the Pd surface is partially covered by oxy- unambiguous test is to follow the reaction rate as a func-
gen. Partial coverage of the active sites by oxygen can, for tion of the oxidation state (potential) of the catalyst (15).
example, reduce the size of active site ensembles and thus In situ catalyst potential measurements revealed that dur-
suppress side reactions, and shift the catalyst potential lead- ing oxidation of -sorbose the reaction rate was negligi-
L
ing to different adsorption strength and reaction rates. To ble on a reduced, partially H-covered Pt surface (19). By
estimate the contribution of oxygen-covered metal surface increasing the rate of oxygen supply (i.e., moving from
to the overall performance of the catalyst we have to keep left to right on the bell-shaped curve in Fig. 4), the rate
in mind that the reduced metallic sites are more active than of sorbose conversion reached a maximum corresponding
the oxidized sites by at least an order of magnitude (40). to a relatively clean surface and moderate oxygen cover-
During our experiments the selectivity to products formed age, then decreased again due to too-high oxygen coverage
by hydrogenation and hydrogenolysis was in the range 9– of Pt.
3
7% even in the presence of air (Tables 1 and 2). This is an
Another important question is the scope of reactants
indication that the major part of the Pd surface was in a re- prone to decarbonylate on Pt or Pd. We propose that cata-
duced state and partially covered by hydrogen. Hence, the lyst poisoning by alcohol degradation should be expected in
key role of oxygen is assumed to be the oxidative removal allreactionswheretheproductisanaliphatic(35, 36)oraro-
of degradation products.
matic aldehyde (47), or an activated ketone (43) (for exam-
ple, an α-ketoacid, a typical oxidation product in carbohy-
drate oxidation). There is substantial literature on the facile
decarbonylation, or more generally on the degradation, of
carbonyl compounds on platinum-group metals affording
CO and a hydrocarbon residue (34–36, 43, 47, 48). Unex-
pectedly, even the dehydrogenation of simple secondary
aliphatic alcohols, such as 2-propanol, on Pt (49) and Pd
Generalization of the Observations
The role of oxygen in the aerobic oxidation of alcohols
can be illustrated by the bell-shaped curve in Fig. 4. It was
early discovered that the oxidized metal (region a) is poorly
active in alcohol oxidation (40). Prereduction of the sup-
ported metal catalyst by hydrogen, or by the reactant itself
in an inert atmosphere, can accelerate the reaction dramat-
ically (moving from right to left on the curve in Fig. 4). The
presentresults, andalsoformerobservations(18, 19, 28, 29),
indicate that below a certain oxygen coverage of the metal
(
50) may be accompanied by degradation to CO and hy-
drocarbon fragments. For example, exposure of Pd(110) to
-propanol at 120 K leads to the formation of CO, CH4,
2
hydrogen, and surface carbon (50). Hence, poisoning of
platinum-group metals by CO and hydrocarbon fragments
seems to be a general phenomenon during dehydrogena-
tion and oxidative dehydrogenation of alcohols.
Finally, we have to emphasize an interesting consequence
of reactant and product degradation and the positive effect
of oxygen on the reaction rate. A positive order of oxygen
concentration on the alcohol oxidation rate is frequently
used to support a reaction mechanism in which adsorbed
oxygen is directly involved. However, this conclusion is ac-
ceptable only when retardation by degradation products
and their oxidative removal can be excluded.
(
region b) the rate decreases again due to catalyst poison-
ing (region c). An important role of oxygen is to remove
the poisoning species and thus increase the number of ac-
0
tive M sites available for alcohol dehydrogenation. During
development of an alcohol oxidation process, the aim is to
operate the reactor at around region b by fine-tuning the
rate of oxygen supply to the actual rate of alcohol oxidation.
CONCLUSIONS
Dehydrogenation and oxidative dehydrogenation of cin-
namyl alcohol over Pd-based catalysts and in situ ATR-IR
spectroscopic analysis of adsorbed species on Pd/Al2O3 un-
der reaction conditions indicate the following mechanistic
model:
1. The reaction obeys the classical dehydrogenation
mechanism in which adsorbed oxidizing species do not play
a role.
FIG. 4. Schematic representation of the rate of alcohol dehydrogena-
tion as a function of the oxidation state of Pd.

LIQUID-PHASE AEROBIC OXIDATION OF ALCOHOLS
251
2
. Contrary to the general assumption, oxidation of the 12. M u¨ ller, E., and Schwabe, K., Kolloid-Z. 52, 163 (1930).
1
1
3. DiCosimo, R., and Whitesides, G. M., J. Phys. Chem. 93, 768 (1989).
co-product hydrogen by oxygen is not a necessary require-
ment for achieving high reaction rates.
4. Mallat, T., Bodnar, Z., and Baiker, A., in “Catalytic Selective Oxida-
tion” (S. T. Oyama and J. W. Hightower, Eds.), Vol. 523, p. 308. Am.
Chem. Soc., Washington, DC, 1993.
3
. The major role of oxygen is the oxidation of strongly
adsorbed CO produced by decarbonylation-type side reac-
15. Mallat, T., and Baiker, A., Catal. Today 24, 143 (1995).
tions. Continuous oxidative removal of CO ensures a high 16. van Dam, H. E., and van Bekkum, H., React. Kinet. Catal. Lett. 40, 13
0
(
1989).
number of free Pd sites available for the dehydrogenation
reaction.
1
1
7. Schuurman, Y., Kuster, B. F. M., van der Wiele, K., and Marin, G. B.,
Appl. Catal. A 89, 47 (1992).
8. Mallat, T., Bodnar, Z., Maciejewski, M., and Baiker, A., Stud. Surf.
Sci. Catal. 82, 561 (1994).
4
. The Cn−1 hydrocarbon fragments (co-products of the
decarbonylation reactions), or the dimers and oligomers
formed from them on the metal surface, can also lead to 19. Br o¨ nnimann, C., Bodnar, Z., Hug, P., Mallat, T., and Baiker, A.,
J. Catal. 150, 199 (1994).
catalyst deactivation. Removal of these species by oxygen
is inefficient.
2
2
2
0. Markusse, A. P., Kuster, B. F. M., and Schouten, J. C., J. Mol. Catal. A
58, 215 (2000).
1
1. Nondek, L., Zdarova, D., Malek, J., and Chvalovsky, V., Collect. Czech.
Chem. Commun. 47, 1121 (1982).
2. Schuurman, Y., Kuster, B. F. M., van der Wiele, K., and Marin, G. B.,
Appl. Catal. A 89, 31 (1992).
We propose that this model can be applied to many other
alcohol oxidation reactions over platinum-group metal
catalysts. The large differences observed in the reactivity of
various metal catalysts may partly be due to the different 23. Markusse, A. P., Kuster, B. F. M., and Schouten, J. C., Stud. Surf. Sci.
Catal. 126, 273 (1999).
extent of catalyst poisoning by degradation products and
2
2
4. van Dam, H. E., Kieboom, A. P. G., and van Bekkum, H., Appl. Catal.
3, 361 (1987).
to the efficiency of oxygen to regenerate the active sites.
The contradictory observations concerning the role of oxy-
gen in the reaction mechanism may be explained by the
3
5. Dijkgraaf, P. J. M., Rijk, M. J. M., Meuldijk, J., and van der Wiele, K.,
J. Catal. 112, 329 (1988).
dramatic effect of oxidative removal of strongly adsorbed 26. Jelemensky, L., Kuster, B. F. M., and Marin, G. B., Catal. Lett. 30, 269
(
1995).
7. Jelemensky, L., Kuster, B. F. M., and Marin, G. B., Chem. Eng. Sci. 51,
767 (1996).
8. Mallat, T., Br o¨ nnimann, C., and Baiker, A., Appl. Catal. A 149, 103
1997).
degradation products on the reaction rate. A positive or-
der to oxygen in the kinetic analysis does not necessarily
indicate a Langmuir–Hinshelwood mechanism with direct
involvement of oxygen. An unambiguous interpretation of
2
2
1
(
the role of oxidizing species requires in situ investigation of 29. Mallat, T., Bodnar, Z., Hug, P., and Baiker, A., J. Catal. 153, 131 (1995).
the surface species and the role of the oxidation state of the 30. Ebitani, K., Fujie, Y., and Kaneda, K., Langmuir 15, 3557 (1999).
3
3
1. Borodzinski, A., and Bonarowska, M., Langmuir 13, 5613 (1997).
2. Harrick, N. J., “Internal Reflection Spectroscopy.” Interscience,
New York, 1967.
catalytic metal surface.
ACKNOWLEDGMENT
33. Ferri, D., B u¨ rgi, T., and Baiker, A., J. Phys. Chem. B 105, 3187 (2001).
4. Tsuji, J., and Ohno, K., J. Am. Chem. Soc. 90, 94 (1968).
3
The authors thank Dr. Frank Krumeich for TEM measurements and
ETH-Zurich for financial support.
35. Davis, J. L., and Barteau, M. A., Surf. Sci. 187, 387 (1987).
36. Shekhar, R., Barteau, M. A., Plank, R. V., and Vohs, J. M., J. Phys.
Chem. B 101, 7939 (1997).
37. Eadon, G., and Sheikh, M. Y., J. Am. Chem. Soc. 96, 2288 (1974).
REFERENCES
38. Martinek, T., Molnar, A., Katona, T., Bartok, M., and Lovas, A., J.
Mol. Catal. A 112, 85 (1996).
1
. van Bekkum, H., in “Carbohydrates as Organic Raw Materials” 39. Kraus, M., Collect. Czech. Chem. Commun. 37, 460 (1972).
(
F. W. Lichtenthaler, Ed.), p. 289. VCH, Weinheim, 1991.
40. Dirkx, J. M. H., van der Baan, H. S., and van der Broek, M. A. J. J.,
Carbohydr. Res. 59, 63 (1977).
41. Jiang, X., Chang, S.-C., and Weaver, M. J., J. Chem. Soc., Faraday Trans.
89, 223 (1993).
42. Gelin, P., Siedle, A. R., and Yates, J. T., J. Phys. Chem. 88, 2978 (1984).
43. Leung, L.-W. H., and Weaver, M. J., Langmuir 6, 323 (1990).
44. Arevalo, M. C., Rodriguez, J. L., and Pastor, E., J. Electroanal. Chem.
505, 62 (2001).
45. Pastor, E. S. W., Iwasita, T., Arevalo, M. C., Gonzalez, S., and Arvia,
A. J., J. Electroanal. Chem. 353, 81 (1993).
2
. Vinke, P., de Wit, D., de Goede, A. T. J. W., and van Bekkum, H., Stud.
Surf. Sci. Catal. 72, 1 (1992).
. Mallat, T., and Baiker, A., Catal. Today 19, 247 (1994).
. Gallezot, P., Catal. Today 37, 405 (1997).
. Biella, S., Castiglioni, G. L., Fumagalli, C., Prati, L., and Rossi, M.,
Catal. Today 72, 43 (2002).
. Besson, M., and Gallezot, P., Catal. Today 57, 127 (2000).
. Kluytmans, J. H. J., Markusse, A. P., Kuster, B. F. M., Marin, G. B., and
Schouten, J. C., Catal. Today 57, 143 (2000).
3
4
5
6
7
8
9
. Wieland, H., Ber. Dtsch. Chem. Ges. 46, 3327 (1913).
. Hayashi, M., Yamada, K., and Nakayama, S., J. Chem. Soc., Perkin
Trans. 1 1501 (2000).
46. Nicoletti, J. W., and Whitesides, G. M., J. Phys. Chem. 93, 759 (1989).
47. Jung, K. J., and Gaset, A., Biomass 16, 63 (1988).
48. Hoffman, N. E., Kanakkanatt, A. T., and Schneider, R. F., J. Org.
Chem. 27, 2687 (1962).
1
1
0. Keresszegi, C., Mallat, T., andBaiker, A., NewJ. Chem. 25, 1163(2001).
1. M u¨ ller, E., and Schwabe, K., Z. Elektrochem. Angew. Phys. Chem. 34, 49. Wood, B. J., Niki, H., and Wise, H., J. Catal. 26, 465 (1972).
70 (1928).
50. Shekhar, R., and Barteau, M. A., Catal. Lett. 31, 221 (1995).
1