Operando Attenuated Total Reflectance FTIR Spectroscopy: Studies on the Different Selectivity Observed in Benzyl Alcohol Oxidation

Article abstract of DOI:10.1002/cctc.201500432

Au, Pd, and AuPd supported on TiO₂ were prepared by a sol immobilization route. Operando attenuated total reflectance (ATR) IR spectroscopy and catalytic batch reactor experiments were performed in parallel to elucidate the different catalytic

Full text of DOI:10.1002/cctc.201500432

DOI: 10.1002/cctc.201500432
Full Papers
Operando Attenuated Total Reflectance FTIR
Spectroscopy: Studies on the Different Selectivity
Observed in Benzyl Alcohol Oxidation
[a]
[b]
[a]
[a]
[c]
Alberto Villa,* Davide Ferri, Sebastiano Campisi, Carine E. Chan-Thaw, Ye Lu,
[b, d]
[a]
Oliver Krçcher,
and Laura Prati
Au, Pd, and AuPd supported on TiO were prepared by a sol
py evidenced that the presence of Au facilitates the desorption
of byproducts, which thus reduces the extent of deactivation
of the active sites caused by the irreversible adsorption of
benzoate species. Although benzaldehyde was the main prod-
uct in both catalysts, the nature of the byproducts differs. Pd/
2
immobilization route. Operando attenuated total reflectance
(
ATR) IR spectroscopy and catalytic batch reactor experiments
were performed in parallel to elucidate the different catalytic
performance of the catalysts in the liquid-phase oxidation of
benzyl alcohol. Pd/TiO exhibited a higher activity than AuPd/
TiO favored the deoxygenation of benzyl alcohol to produce
2
2
TiO and Au/TiO , but the modification of Pd with Au demon-
toluene as the main byproduct. Conversely, AuPd/TiO promot-
2
2
2
strated a significant stability enhancement. ATR-IR spectrosco-
ed the transformation of benzaldehyde to benzoic acid.
Introduction
The oxidation of alcohols to the corresponding carbonyl com-
pounds catalyzed by supported noble metals has been a sub-
been attributed to the presence of isolated single Pd sites
[
8,12]
formed in alloyed nanoparticles.
[1–4]
ject of growing interest recently.
In the last decades, re-
The selective oxidation of benzyl alcohol to benzaldehyde is
a model reaction used widely and it is an important industrial
process for benzaldehyde production as a key factor for fine
search has focused on green methods that use “clean” oxi-
dants such as O in combination with supported metal nano-
2
[
1,2]
particles as the catalyst. Au and Pd have been studied exten-
sively as active catalytic metals for alcohol oxidation; Au
mainly for its good resistance to deactivation, and Pd for its
chemicals production.
The main byproduct of the oxidation
of benzyl alcohol to benzaldehyde is toluene, which is report-
[
13]
ed to be formed by the disproportionation of benzyl alcohol
or by the reaction of the intermediate metal hydride with the
[
1–4]
high activity and selectivity to the corresponding aldehyde.
[
14,15]
However, both suffer from limitations. In particular, in the case
of Au-catalyzed alcohol oxidation, a basic environment is
needed to accelerate the rate-determining step (i.e., hydride
abstraction), which thus enhances carboxylates as the main
alcohol instead of O2.
A bimetallic AuPd catalyst has been
[
10,16,17]
shown to limit the formation of toluene.
Vibrational spectroscopy is ideally suited to interrogate the
binding of molecular species on solid catalysts even if the solid
is surrounded by a liquid environment. The reaction mecha-
nism of the liquid-phase benzyl alcohol oxidation and the de-
tailed observation of the evolution of adsorbed species have
been studied extensively using attenuated total reflection
(ATR) IR spectroscopy in both flow-through and batch-reactor
[
5–7]
products.
phenomena.
However, Pd suffers from severe deactivation
2]
[
It has been shown that alloyed Au-Pd catalysts have im-
proved catalytic activity and selectivity to the desired product
[8–11]
and also resistance to deactivation.
This phenomenon has
[
15]
cells. A commercial Pd/Al O catalyst has been largely used
2
3
[
18–20]
for these measurements.
Alcohol dehydrogenation to
[
a] Dr. A. Villa, S. Campisi, Dr. C. E. Chan-Thaw, Prof. L. Prati
Department of Chemistry
benzaldehyde is accompanied by both decarbonylation and
oxidation of the benzaldehyde product to give benzoic acid. A
number of adsorbed species have been detected and assigned
to the different reaction paths at work during this reaction.
Università degli Studi di Milano
via Golgi 19, 20133 Milan (Italy)
E-mail: alberto.villa@unimi.it
[
b] Dr. D. Ferri, Prof. O. Krçcher
[21]
Nowicka et al. studied benzyl alcohol oxidation on bimetallic
Paul Scherrer Institut
Au-Pd nanoparticles dispersed on TiO , MgO, ZnO, and carbon
5
232 Villigen PSI (Switzerland)
c] Dr. Y. Lu
Laboratory for Solid State Chemistry and Catalysis
2
by using an ATR-IR probe in a batch reactor to monitor the
liquid phase by diffuse reflectance infrared Fourier transform
spectroscopy (DRIFTS) and inelastic neutron scattering (INS).
Only the oxidative dehydrogenation to benzaldehyde was ob-
served on MgO and ZnO, whereas toluene was detected on
the other supports, which confirmed the benzaldehyde/tolu-
ene selectivity issue.
[
Empa, Swiss Federal Laboratories for Materials Science and Technology
Ueberlandstrasse 129, Dübendorf (Switzerland)
[
d] Prof. O. Krçcher
Institut des Sciences et IngØnierie Chimiques
Ecole Polytechnique FØdØrale de Lausanne (EPFL)
1
015 Lausanne (Switzerland)
ChemCatChem 2015, 7, 2534 – 2541
2534
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Ide et al. used ATR-IR spectroscopy to prove that the deacti-
vation of Pt in alcohol oxidation is caused by the presence of
species adsorbed strongly and not to metal reconstruction or
Table 1. Statistical median and standard deviation of the particle size
analysis for the Au, Pd, and AuPd catalysts.
[
22]
Catalyst
Au/TiO2
Statistical median [nm]
Standard deviation (s)
leaching.
To clarify the origin of the catalytic activity of Pd/TiO and
3.7
3.5
4.1
0.9
0.7
1.2
2
Pd/TiO
AuPd/TiO
2
AuPd/TiO2 in liquid-phase benzyl alcohol oxidation, in this
work we followed the reaction on the surface of the catalysts
using ATR-IR spectroscopy. We correlated the batch catalytic
results to the development of different surface species. Au, Pd,
and AuPd nanoparticles were prepared by a sol immobilization
procedure using polyvinyl alcohol (PVA) as a protective agent
2
adsorption to obtain a description of the native surfaces and
the influence of the liquid environment on the adsorption
properties of the catalysts. The DRIFT spectra recorded after
30 min adsorption under a flow of CO/He on the three cata-
lysts are shown in Figure 2.
and they were supported on TiO to ensure a substantially uni-
2
[23,24]
form particle size and distribution in the catalysts.
The corresponding spectra collected during desorption (He
Results and Discussion
À1
flow) are also shown. Signals that appear below n˜ =2150 cm
The Au, Pd, and AuPd catalysts were synthesized by sol immo-
can be attributed to CO adsorbed on the noble metal particles.
Typically, the adsorption of CO on metallic Pd results in signals
[23,24]
bilization using PVA as a protective agent.
In particular, the
bimetallic AuPd catalyst was prepared using a two-step proce-
dure that has been demonstrated to produce alloyed AuPd
NPs of a single composition and high dispersion if supported
associated with CO bound linearly (CO ) that absorbs above
L
À1
n˜ =2000 cm , bridged bound CO (or twofold bound, CO ),
B
and finally species adsorbed on hollow sites (threefold bond,
[
8]
[25–27]
on activated carbon.
CO ).
These two latter species give rise to a number of
B3
bands that can be divided qualitatively as follows: n˜ =2000–
À1
À1
[28–31]
1
880 cm for CO and n˜ =1880–1800 cm for COB3.
The
B
Catalyst characterization
assignment of each signal is complicated by the fact that CO
adsorption is site dependent, that is, the adsorption of CO
(CO , CO , or CO ) on facets will give rise to signals at different
The average diameter of TiO -supported Au, Pd, and AuPd
2
nanoparticles was determined by scanning transmission elec-
tron microscopy (STEM; Table 1). Au and Pd nanoparticles
showed a similar particle diameter (3.7 and 3.5 nm, respective-
ly).
L
B
B3
frequencies from CO adsorbed in the same geometry on
edges, steps, or defect sites. Coordination to facets produces
higher-energy signals than coordination to particles edges/
STEM data indicated a slight
increase of the average particle
size with the addition of Pd to
the Au nanoparticles (from 3.7
to 4.1 nm).
A
representative
overview image of AuPd/TiO is
2
given in Figure 1. Energy disper-
sive X-ray spectroscopy (EDX)
analysis of 10 nanoparticles se-
lected randomly indicated that
they all contained both Pd and
Au (Figure 1d).
The lattice spacing of a typical
AuPd nanoparticle with the
nominal
.4:2.6 wt%
composition
between the
7
Pd(111) plane (2.25 Š) and the
Au(111) plane (2.35 Š) was
ꢀ
2.29 Š (Figure 1c), which im-
[8]
plies the presence of an alloy.
The corresponding EDX spec-
trum (Figure 1d) indicates a Au/
Pd ratio consistent with the
nominal
.3:2.7 wt%).
Pd/TiO , AuPd/TiO , and Au/
value
(Au/Pd
7
Figure 1. a) Representative STEM image, b) PdAu loading [at%] in 10 random nanoparticles of AuPd/TiO
resolution TEM image and d) the corresponding EDX spectrum of a AuPd particle with a Au/Pd composition of
2
, c) high-
2
2
TiO2 were characterized by CO 7.4/2.6 wt%.
ChemCatChem 2015, 7, 2534 – 2541
www.chemcatchem.org
2535
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
À1
tained. The shoulder at n˜ =ꢀ2070 cm (edge sites) is now
more pronounced. The CO signal loses intensity at n˜ =
B
À1
1
1
952 cm to reveal more defined signals at n˜ =ꢀ1980 and
À1
900 cm . The removal of the gas-phase signals of CO shows
2+
À1
the signals of CO adsorbed on Pd ( n˜ =2145 and 2130 cm ).
It is recognized immediately that the amount of adsorbed
CO on AuPd/TiO is lower than that on Pd/TiO (Figure 2b).
2
2
This could mean that the fraction of available metal atoms for
adsorption is smaller. The spectrum measured during desorp-
2
+
tion also presents fewer signals. The absence of Pd species
À1
(
n˜ = >2100 cm ) suggests that all Pd present in AuPd/TiO is
2
less prone to reoxidation if exposed to air most probably be-
cause it is alloyed to Au. Importantly, we can argue that CO co-
ordination to particle corners is suppressed by the presence of
À1
Au because the CO signal at n˜ =2097 cm observed on Pd/
L
TiO is absent in this case. This allows a clear observation of
2
À1
the signal at n˜ =2079 cm flanked by a shoulder on the low-
energy side, which suggests adsorption on edge sites. The
À1
signal at n˜ =2079 cm is the evident shoulder of the major
À1
CO signal at n˜ =2097 cm of Pd/TiO . The CO contribution
L
2
L
À1
on AuPd/TiO is also larger than the CO signal ( n˜ =1979 cm ).
2
B
These observations indicate that small and less contiguous do-
mains of Pd are present on the bimetallic AuPd particles in
agreement with the deposition of Pd on the preformed Au
nanoparticles. Although there is no evidence of CO adsorption
on Au atoms, this cannot be ruled out completely given the
complexity of the spectrum (Figure 2). The large difference be-
tween the spectra shown in Figure 2a and b provides a qualita-
tive picture of the modification of Pd nanoparticles by the
presence of Au, which suggests the formation of a AuPd alloy.
Figure 2. DRIFT spectra of adsorbed CO on a) Pd/TiO
c) Au/TiO . d) Color map of the time-resolved spectra recorded during CO
adsorption on Au/TiO . Solid and dashed traces represent spectra recorded
2 2
, b) AuPd/TiO , and
2
2
under CO/He and He flow, respectively. Spectra are offset for clarity.
steps. This can be used on a qualitative base to assess the
morphology of the Pd nanoparticles using IR spectroscopy.
Moreover, as the absorption coefficient of bridged species is
larger than that of linear species, the CO /CO ratio represents
The adsorption of CO on Au/TiO is clearly different because
2
of the nature of the metal (Figure 2c). Adsorption first causes
À1
a signal at n˜ =2128 cm , which redshifts initially to n˜ =
L
B
[32]
À1
a qualitative estimation of the size of the Pd domains.
2106 cm . After 3 min on stream, this signal starts to trans-
À1
À1
Pd/TiO exhibits signals at n˜ =2097, 1981, and 1952 cm ,
form into a signal that grows at n˜ =2075 cm (Figure 2d). The
2
À1
which are assigned to various CO species adsorbed on Pd (Fig-
ure 2a). According to the discussion above, the signals ob-
isosbestic point at n˜ =2091 cm indicates that the species
that provides the signal initially at high frequency transforms
into the species that provides the new signal. A third signal be-
served during CO adsorption can be attributed to CO and
L
À1
CO , respectively. The high energy of the CO signal indicates
comes visible as a shoulder at n˜ ꢀ2020 cm as this transfor-
B
L
CO coordination to corner atoms, whereas the shoulder at n˜
mation occurs. All of the signals do not resist desorption, and
the spectrum measured in He flow does not exhibit any signals
À1
[25]
ꢀ
2080 cm is assigned to edge sites. The signals that over-
À1
lap with that of gas-phase CO are most likely because of CO
of adsorbed CO. The signal at n˜ =2106 cm is typically associ-
2+
[26]
adsorbed on Pd cations. This species might be present on
particles that are not capped perfectly or on uncovered parti-
ated with CO adsorbed in a linear geometry to deficient Au
[
34]
atoms in Au nanoparticles. These coordinatively unsaturated
atoms provide adsorption sites for CO. The intensity of the
cles that underwent surface oxidation. According to previous
À1
À1
reports, we assign the CO signal at n˜ =1981 cm to CO coor-
signal at n˜ =2128 cm increases and it redshifts, which is typi-
B
[26,33]
À1
[35]
dinated to Pd(100) facets.
The signal at n˜ =1952 cm is
cal for CO adsorption on Au. The change observed on Au/
thus attributed to coordination to edge/step sites. The signal is
unusually intense compared to that of the high-energy COB
species, which reveals that the Pd particles are relatively small.
The spectrum measured during adsorption confirms this obser-
TiO during CO adsorption is related to the reconstruction of
2
the Au nanoparticles under the high partial pressure used in
À1
the experiment. The signal at n˜ =2075 cm is assigned tenta-
0
[34]
tively to the formation of Au sites or to negatively charged
[
36]
vation because the CO signal is relatively intense compared to
Au carbonyl species upon species absorption. Similar to Au/
L
[
36]
dÀ
that of COB.
SiO2,
Au species may arise from the interaction of ad-
The spectrum obtained during the subsequent He flow pro-
vides information on the species adsorbed the most strongly.
The signal intensity is generally lower after desorption. Addi-
sorbed CO with functional groups present on the Au particles,
which are the OÀH and C=O groups of the PVA capping agent
in this case.
tionally, a more defined structure within the CO signal is ob-
L
ChemCatChem 2015, 7, 2534 – 2541
www.chemcatchem.org
2536
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Catalytic activity in a batch reactor
Table 2. Comparison of TiO
2
-supported catalysts activities in benzyl alcohol oxidation.
[
a]
[b]
[c]
The catalytic activity in the liquid-phase oxidation of
benzyl alcohol was tested in a glass batch reactor
using cyclohexane as the solvent. Under these reac-
Catalyst
Activity
Selectivity [%]
Toluene Benzaldehyde Benzoic acid Benzyl benzoate
[d]
[d]
[d]
Au/TiO2
14
532
124
–
10
1
80
84
88
14
1
3
3
2
Pd/TiO
AuPd/TiO
2
tion conditions, Au/TiO was not very active (14 mol
2
À1 À1
2
8
molmetal h ; Table 2) and reached a conversion of
À1
only 12% after 8 h (Figure 3a). This behavior con-
[a] Reaction conditions: alcohol/metal=500:1 molmol , 608C, pO
[b] Mol of reactant converted per hour per mol of metal calculated after 15 min of re-
action. [c] Selectivity at 90% conversion. [d] Selectivity at 10% conversion.
2
=2 bar, 1250 rpm.
[
5–7]
firmed previous results.
Conversely, Pd/TiO exhib-
2
À1 À1
ited a good activity (532 molmolmetal h ) and at-
tained 89% conversion after 8 h (Figure 3a).
The AuPd system showed a lower initial activity (124 mol
duction period, AuPd/TiO (Figure 3a) was as active as Pd/TiO2
2
À1 À1
molmetal h ; Table 2) than the Pd system. However, after an in-
after 1.5 h.
If the removal of the PVA layer was involved in the origin of
the induction time of AuPd/TiO , the same effect should have
2
[
37]
also been present in the case of Pd/TiO2 under the assump-
tion that PVA interacts equally with the two metals. Therefore,
the effect could also be because of the different nature of the
active site in the bimetallic system with respect to that in the
monometallic system.
With regard to the selectivity, benzaldehyde was the main
product over all the catalysts (80–88% selectivity). However,
the distribution of byproducts was different as can be seen
from the selectivity data in Table 2 and from the evolution of
the products during the reaction (Figure 3b and c). Monome-
tallic Pd promoted the formation of toluene (10% selectivity),
the formation rate of which is linear with conversion (Fig-
[
10,16,17]
ure 3b). Conversely, in agreement with earlier results,
the AuPd system did not produce toluene and promoted the
consecutive transformation of benzaldehyde to benzoic acid
(
(
Table 2 and Figure 3c) similar to the monometallic Au catalyst
Table 2).
Recycling experiments performed on the most active cata-
lysts (Pd/TiO2 and AuPd/TiO ) by filtering the catalyst and
2
adding a fresh solution of benzyl alcohol revealed the better
recycling performance of AuPd/TiO2 than Pd/TiO2 (Figure 4),
the activity of which declined very rapidly after the second
run. The positive effect of bimetallic AuPd catalysts was also
[
10]
observed for catalysts supported on activated carbon. Induc-
tively coupled plasma (ICP) analysis of the collected solution
after six runs showed a loss of <1 wt% metal in both cases.
Moreover, TEM images did not show any clear changes in the
morphology of the Pd and AuPd catalysts, which had a similar
particle size to the fresh materials (Table 1). Therefore, to un-
derstand the different catalytic performances, the possible role
of the adsorbed species on the catalyst surface was investigat-
ed.
Operando ATR-IR spectroscopy
The evolution of surface species was monitored during the oxi-
dation of benzyl alcohol using ATR-IR spectroscopy on particu-
late films deposited on an internal reflection element (IRE)
mounted in a batch-reactor cell. The film was in contact with
a cyclohexane solution of benzyl alcohol. As a result of the cell
design, the ATR-IR spectroscopy measurements were per-
formed at reflux (time of reaction 1 h) with oxygen bubbling
Figure 3. Reaction profiles of a) TiO
hol oxidation, and the product distribution on b) Pd/TiO
2
-supported catalysts during benzyl alco-
and c) AuPd/TiO
2
2
.
ChemCatChem 2015, 7, 2534 – 2541
www.chemcatchem.org
2537
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
the catalytic activity is Pd>AuPd>Au. The signals
that develop at the early stages of the measurements
À1
À1
at n˜ =1497 cm and ꢀ1397 cm are characteristic
of benzyl alcohol, most likely dissolved in cyclohex-
=
ane. The signals observed at n˜ =1600 (nC C), 1531
À1
(
= =
nCOOas), 1483 (nC C), 1452 (nC C), and 1417 cm (nCOOs)
belong to benzoate species, the product of benzalde-
hyde oxidation, which appeared to be coordinated
[
15]
predominantly to TiO2. This assignment is support-
ed by the spectra obtained upon the contact of
a benzoic acid solution with TiO2 and the corre-
sponding metal-based catalysts (Figure 6).
The signals exhibit shoulders that likely suggest
different adsorption geometries. The least pro-
nounced signals are observed on Au/TiO2 and the
2 2
Figure 4. Stability tests performed using Pd/TiO and AuPd/TiO .
instead of pressurizing the catalyst suspension as in the batch
reactor study. This should be considered if the two sets of data
are compared directly.
The ATR-IR spectra (Figure 5) obtained for the three samples
are very similar and only intensity changes are observed. A
À1
sharp signal at n˜ =1712 cm represents the benzaldehyde
product dissolved in cyclohexane. The increase in the intensity
of the signal is thus related to the progress of the catalytic re-
action, which can be monitored by the detection of the ad-
sorbed species. From the intensity of this signal, the order of
Figure 6. ATR-IR spectra recorded for benzoic acid (0.005m solution in cyclo-
hexane) adsorption on TiO , Pd/TiO , AuPd/TiO , and Au/TiO .
2 2 2 2
order of intensity is Pd>AuPd>Au, which is in agreement
with the activity order observed in the glass reactor. This prob-
ably indicates that benzaldehyde is oxidized effectively to ben-
zoic acid: the more benzaldehyde is produced, the more it is
oxidized. Therefore, the most active Pd/TiO catalyst exhibits
2
the most intense benzoate signals. The temporal behavior of
selected signals representative of dissolved benzaldehyde ( n˜ =
À1
À1
1
712 cm ), adsorbed benzoates ( n˜ =1600 and 1531 cm ), and
À1
dissolved benzyl alcohol ( n˜ =1496 cm ) are shown in Figure 7,
which demonstrates that the rate of accumulation of adsorbed
benzoate species on the three catalysts is different. Although
signals grow almost linearly with time on Pd/TiO , they level
2
off after a rapid initial growth on AuPd/TiO and Au/TiO . A
2
2
closer look at the spectra (Figure 6) reveals that in addition to
the continuous growth of the benzoate signals on Pd/TiO2
compared to that on AuPd/TiO2 and Au/TiO , the ratio be-
2
À1
tween the intensities of the signals at n˜ =1600 and 1531 cm
is not equal in the three cases. This observation excludes that
the signals originate only from the adsorption of benzoate
species on the support. The ratio of the intensities of the
Figure 5. ATR-IR spectra recorded during benzyl alcohol oxidation on a) Pd/
TiO , b) AuPd/TiO , and c) Au/TiO . Spectra are offset for clarity. Vertical
2 2 2
dashed lines indicate selected signals the kinetics of which is reported in
Figure 7.
À1
bands at n˜ =1600 and 1531 cm (I1600/I1531) increases in the
ChemCatChem 2015, 7, 2534 – 2541
www.chemcatchem.org
2538
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Au/TiO agrees well with the preferential adsorption of these
2
species on Pd. Previous experiments performed on Pd/Al O3
2
evidenced that benzoic acid is formed in the early stage of the
reaction and adsorbed strongly on the catalysts so that it re-
[
15]
mains undetectable in the effluent. In the case of AuPd/TiO2
Figure 3c), the benzoic acid concentration in the liquid phase
(
increased clearly only after 2 h. This induction time followed
the trend of the signal of benzoates observed at the surface of
the catalyst. Benzoate species accumulate initially on the cata-
lyst surface until the saturation of some specific sites, whereas
benzoates formed successively desorbed into the solution. This
behavior agrees with recent studies that showed that the for-
mation of the AuPd bimetallic system alters the electronic
structure of the metals, which potentially decreases the ten-
dency for the formation of intermediates that are adsorbed
[
38]
strongly. The higher rate of desorption of the products, in
particular that of benzoate species, on AuPd/TiO compared
2
with Pd/TiO could be a possible explanation for the lower du-
2
rability of the Pd catalyst compared with the bimetallic AuPd
system. A similar trend was obtained for Pt-catalyzed alcohol
[
22]
oxidation.
Batch reactor experiments showed that the Pd catalyst pro-
motes the formation of toluene (10%). In particular, close ob-
servation of the product distribution (Figure 3b) reveals that
the rate of formation of toluene increases with increasing con-
version, whereas the rate of formation of benzaldehyde shows
the opposite trend. Pang et al. observed that toluene is
formed in higher amounts by direct hydrogen abstraction at
Figure 7. Time response of selected ATR-IR signals observed during benzyl
alcohol oxidation on Pd/TiO
2
, AuPd/TiO
2
, and Au/TiO
2
: a) n˜ =1496, b) 1531,
À1
c) 1600, and d) 1712 cm
.
À1
[39]
order Au<AuPd<Pd, whereas the signal at n˜ =1600 cm
a high coverage of benzyl alcohol species. The presence of
seems to be accompanied by the signal of the symmetric
bimetallic AuPd nanoparticles limited the formation of toluene
À1
[10,16,17]
stretch of the carboxylate group ( n˜ =1417 cm ). This observa-
drastically, as reported in previous studies.
Following the
[
39]
tion suggests that an additional fraction of benzoate species is
mechanism proposed by Pang et al., the limited production
of toluene on AuPd could correspond to reduced intermolecu-
lar interactions between adsorbed benzyl alcohol species on
AuPd compared to Pd, which thus reduces the coverage of
benzyl alcohol. Indeed, the time response of the ATR-IR spectra
shown in Figure 7d shows that there is a higher amount of
benzyl alcohol on the surface of Pd than on AuPd, in particular
formed over the Pd-based catalysts, especially Pd/TiO , during
2
reaction. It is plausible to attribute this species to benzoates
adsorbed specifically on Pd particles rather than on TiO . These
2
species cannot be observed if benzoates are formed from ben-
zoic acid adsorption in the control experiments shown in
Figure 6.
À1
The assessment of conversions in these experiments using
the IR signals is complicated by the overlap of the signals of
the reactant and products (which include adsorbed species).
at high conversion ( n˜ =1496 cm ). Unfortunately, ATR-IR spec-
troscopy is not a suitable technique for the direct detection of
toluene because of the overlap of the vibrational modes of tol-
uene with those of benzyl alcohol, benzaldehyde, and ad-
sorbed benzoate species and because of the low toluene con-
centration. Therefore, we could not assess the presence of tol-
uene directly from the operando ATR-IR spectroscopy experi-
ments and correlate it to the evolution of the observed spe-
cies. Finally, no adsorbed CO derived from benzaldehyde
decarbonylation was observed on all the samples, which re-
veals that this reaction is not favored on the catalysts prepared
The intensity of the signal of dissolved benzyl alcohol at n˜ =
À1
1
496 cm increases immediately after injection, faster than
that of benzoate species, and reaches a constant intensity on
the three catalysts. On Pd/TiO , this signal still tends to in-
2
crease slightly. Its intensity is most likely more influenced by
À1
the still increasing neighboring signal at n˜ =1531 cm , which
stabilizes after 30 min on Au/TiO and AuPd/TiO .
2
2
Despite the experimental differences, the different rates of
benzoate formation on Au/TiO and AuPd/TiO compared with
[
15]
by sol immobilization in contrast to commercial Pd/Al O .
2
2
2
3
that on Pd/TiO and the observation of benzoate species ad-
2
sorbed on Pd correlates with the selectivity obtained in the
glass batch reactor study. The quantity of benzoic acid detect-
ed in solution increases in the order Pd<AuPd<Au (Table 2),
which is opposite to the order detected by ATR-IR spectrosco-
py (Au<AuPd<Pd). The negligible amount of benzoic acid
obtained with Pd/TiO compared to that with AuPd/TiO and
Conclusions
Monometallic Au and Pd and bimetallic AuPd nanoparticles
were synthesized using a sol immobilization technique and
were supported on TiO . Operando attenuated total reflectance
2
(ATR) IR spectroscopy in a batch-reactor cell allowed us to gain
2
2
ChemCatChem 2015, 7, 2534 – 2541
www.chemcatchem.org
2539
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
an insight into the catalytic performance of the catalysts in the
liquid-phase oxidation of benzyl alcohol. In particular, the for-
mation of AuPd bimetallic nanoparticles drastically limited the
irreversible adsorption of products observed for Pd, which thus
decreased the extent of catalyst deactivation. This technique
also showed that the presence of Au facilitates the desorption
of the products, especially benzoate species. All the catalysts
demonstrated a good selectivity to benzaldehyde (78–83%).
However, toluene was produced as the main byproduct (16%)
on the monometallic Pd catalyst. The modification of Pd with
Au suppressed the formation of toluene by reducing the
amount of benzyl alcohol on the surface. Conversely, it pro-
moted the transformation of benzaldehyde to benzoic acid.
Catalytic test in a batch reactor
The reactions were performed in a thermostatted glass reactor
(
30 mL) provided with an electronically controlled magnetic stirrer
connected to a large reservoir (5000 mL) that contained oxygen at
2 bar (SIAD, 99.99%). The oxygen uptake was followed by a mass
flow controller connected to a PC through an A/D board, which
plotted a flow–time diagram. The oxidation experiments were per-
formed in the presence of cyclohexane (Sigma–Aldrich, puriss. p. a.
ACS reagent, ꢁ99.5%, GC) as the solvent and 0.3m benzyl alcohol
(
Sigma–Aldrich, puriss., meets analytical specification of Ph. Eur.,
BP, NF, 99–100.5%, GC; substrate/metal=500 mol/mol, 608C, pO =
2
2
bar). The periodic removal of samples from the reactor was per-
formed. Mass recoveries were always (98Æ3)% with this proce-
dure. For the identification and analysis of the products, a GC HP
7
5
820 A gas chromatograph equipped with a capillary column (HP-
, 30 m”0.32 mm, 0.25 mm film, Agilent Technologies) and thermal
conductivity detector (TCD) was used. Quantification of the reac-
tion products was performed by the external calibration method.
Experimental Section
Catalyst preparation
Monometallic catalysts
Catalyst characterization
Au catalyst preparation: Solid NaAuCl ·2H O (Aldrich, 99.99%
4
2
purity; 0.043 mmol) and PVA solution (MWPVA =13000–23000, 87–
9% hydrolyzed, Aldrich; 1% w/w; Au/PVA 1:0,5, w/w) were added
to H O (100 mL). After 3 min, NaBH4 (Fluka, >96%; Au/NaBH4
The morphology of the catalysts was observed by using a Jeol
2200FS transmission electron microscope equipped with a 200 kV
field-emission gun and a high-angle annular dark field detector in
STEM mode. The local composition was determined by EDX. The
samples were prepared by evaporation of an alcohol suspension
on a copper grid coated with a holey carbon film. The noble metal
particle size distribution was obtained by measuring particles by
using the software ImageJ.
8
2
1
:4 mol/mol) solution was added to the solution with vigorous
0
magnetic stirring. A ruby-red Au sol was formed immediately. The
UV/Vis spectrum of the Au sol was recorded to check the complete
À
0
reduction of AuCl4 and the formation of the plasmon peak of Au
nanoparticles. Within a few minutes from its generation, the colloid
acidified at pH 2 by sulfuric acid) was immobilized by adding the
(
2
À1
DRIFT spectra were collected by using a Vertex 70 (Bruker Optics)
spectrometer equipped with a liquid-nitrogen-cooled mercury cad-
mium telluride (MCT) detector and a commercial mirror unit (Pray-
ing Mantis, Harrick). CO adsorption from the gas phase was fol-
support (TiO , Degussa P25, 49 m g , 80% anatase) under vigo-
2
rous stirring. The amount of support was calculated to obtain
a final metal loading of 1 wt%. The catalysts were filtered, washed
several times, dried at 808C for 4 h.
À1
lowed at RT on as-received catalysts by admitting 50 mLmin
Pd catalyst preparation: Solid Na PdCl4 (Aldrich, 99.99% purity;
5 vol% CO/He after dehydration for 1 h at 1208C in He and by ac-
2
Pd 0.043 mol) and PVA solution (1% w/w, Pd/PVA 1:0.5 w/w) were
cumulating spectra (first 100 scans, 14 s/spectrum; then 200 scans,
ca. 180 s/spectrum; 4 cm resolution) over 30 min. Adsorbed CO
À1
added to H O (100 mL). After 3 min, NaBH4 (Pd/NaBH =1:8 mol/
2
4
mol) solution was added to the yellow-brown solution with vigo-
rous magnetic stirring. The brown Pd sol was formed immediately.
was then replaced by He to follow desorption under otherwise
identical conditions. The powder samples (ꢀ70 mg) were used
without further dilution. Spectra were normalized against a back-
ground spectrum recorded in He flow before the admittance of
CO. All spectra are presented in absorbance units and were cor-
0
A UV/Vis spectrum of the Pd sol was recorded to check the com-
2À
plete reduction of PdCl4 . Within few minutes from its generation,
the colloid (acidified at pH 2 by sulfuric acid) was immobilized by
adding the support under vigorous stirring. The amount of support
was calculated to obtain a final metal loading of 1 wt%. The cata-
lysts were filtered, washed several times, and dried at 808C for 4 h.
rected for contribution of atmospheric CO and water if needed.
2
A homemade batch-reactor cell was used to monitor alcohol oxi-
dation both on the surface of the catalyst and in solution using
[18]
the ATR mode. The cell is a modification of that described in
Ref. [40]. Stirring was provided by using a conventional laboratory
magnetic stirrer integrated in the base plate of the cell. The top of
the cell was fitted with a stainless-steel cover with a circular aper-
ture that allowed the insertion of a glass condenser (conical ta-
pered ground joint, 19/26) for experiments under reflux. The ZnSe
IRE (308, 50 mm”20 mm”2 mm; Crystran Ltd.) was coated with
a powder film obtained by the evaporation of a catalyst suspen-
sion (15 mg/2 mL suspension) and was placed on the horizontal
heatable base of the cell. After mounting the cell body and adding
Bimetallic catalysts
Bimetallic catalysts have been prepared following a procedure re-
[8]
ported previously. After the preparation of 0.73 wt% Au/TiO ac-
2
cording to the procedure reported above, the material was dis-
persed in water (100 mL), and Na PdCl (Pd 0.025 mol) and PVA sol-
2
4
utions (1% w/w, Pd/PVA 1:0.5 w/w) were added. H was bubbled
2
À1
(
50 mLmin ) under atmospheric pressure at RT. After 2 h, the
slurry was filtered, and the catalyst was washed thoroughly with
distilled water.
2
6
0 mL of cyclohexane solvent, the temperature was increased to
08C, and the system was left to equilibrate for ꢀ2 h under bub-
ICP analysis was performed on the filtrate to verify the quantitative
metal loading on the support. The final total metal loading was
bling N at reflux and stirring. Before the injection of benzyl alcohol
2
(190 mL), a background IR spectrum of the catalyst was recorded at
608C in cyclohexane. Then a series of consecutive ATR-IR spectra
1
7
wt% in all cases. For the bimetallic catalyst, the Au/Pd ratio was
.3:2.7 wt% (6:4 mol/mol).
À1
(200 scans, ꢀ4 min/spectrum, 4 cm resolution) were collected to
ChemCatChem 2015, 7, 2534 – 2541
www.chemcatchem.org
2540
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
follow the reaction. After ꢀ5 min from the injection of benzyl alco-
[19] C. Mondelli, J. Grunwaldt, D. Ferri, A. Baiker, Phys. Chem. Chem. Phys.
2010, 12, 5307–5316.
hol, N was replaced by O , and the system was allowed to react
2
2
[
[
20] C. Keresszegi, D. Ferri, T. Mallat, A. Baiker, J. Catal. 2005, 234, 64–75.
21] E. Nowicka, J. P. Hofmann, S. F. Parker, M. Sankar, G. M. Lari, S. A. Kon-
drat, D. W. Knight, D. Bethell, B. M. Weckhuysen, G. J. Hutchings, Phys.
Chem. Chem. Phys. 2013, 15, 12147–12155.
for ꢀ1 h. The metal content of the catalyst and the metal leaching
during the experiments were checked by the ICP analysis of the fil-
trate by using a JobinYvon JY24.
[
[
[
[
[
22] M. S. Ide, D. D. Falcone, R. J. Davis, J. Catal. 2014, 311, 295–305.
23] L. Prati, G. Martra, Gold Bull. 1999, 32, 96–101.
24] L. Prati, A. Villa, Acc. Chem. Res. 2014, 47, 855–863.
25] A. R. Denton, N. W. Ashcroft, Phys. Rev. A 1991, 43, 3161–3164.
26] T. Lear, R. Marshall, J. A. Lopez-Sanchez, S. D. Jackson, T. M. Klapçtke, M.
Bäumer, G. Rupprechter, H. J. Freund, D. Lennon, J. Chem. Phys. 2005,
Keywords: IR spectroscopy · gold · nanoparticles · oxidation ·
palladium
[
[
[
1] M. Besson, P. Gallezot, Catal. Today 2000, 57, 127–141.
2] T. Mallat, A. Baiker, Chem. Rev. 2004, 104, 3037–3058.
3] N. Dimitratos, J. A. Lopez-Sanchez, G. J. Hutchings, Chem. Sci. 2012, 3,
1
23, 174706–174713.
27] D. Tessier, A. Rakai, F. Bozon-Verduraz, J. Chem. Soc. Faraday Trans. 1992,
8, 741–749.
[
2
0–44.
8
[
[
4] S. E. Davis, M. S. Ide, R. J. Davis, Green Chem. 2013, 15, 17–45.
5] S. Carrettin, P. McMorn, P. Johnston, K. Griffin, C. J. Kiely, G. J. Hutchings,
Phys. Chem. Chem. Phys. 2003, 5, 1329–1336.
[
[
28] N. Sheppard, C. De La Cruz, Catal. Today 2001, 70, 3–13.
29] N. Sheppard, T. T. Nguyen in Advanced Infrared Raman Spectroscopy,
Vol. 5 (Eds.: R. J. H. Clark, R. E. Hester), Heyden, London, 1978, pp. 67–
[
[
6] F. Porta, L. Prati, J. Catal. 2004, 224, 397–403.
7] S. Demirel-Gülen, M. Lucas, P. Claus, Catal. Today 2005, 102–103, 166–
1
48.
[
[
30] C. De La Cruz, N. Sheppard, J. Mol. Struct. 1990, 224, 141–161.
31] E. Ozensoy, D. W. Goodman, Phys. Chem. Chem. Phys. 2004, 6, 3765–
1
72.
8] D. Wang, A. Villa, F. Porta, D. Su, L. Prati, Chem. Commun. 2006, 1956–
958.
[
[
3778.
1
[
[
32] M. A. Vannice, S. Y. Wang, J. Phys. Chem. 1981, 85, 2543–2546.
33] F. Di Gregorio, L. Bisson, T. Armaroli, C. Verdon, L. Lemaitre, C. Thoma-
zeau, Appl. Catal. A 2009, 352, 50–60.
9] D. I. Enache, J. K. Edwards, P. Landon, B. Solsona-Espriu, A. F. Carley, A. A.
Herzing, M. Watanabe, C. J. Kiely, D. W. Knight, G. J. Hutchings, Science
2
006, 311, 362–365.
10] A. Villa, D. Wang, N. Dimitratos, D. Su, V. Trevisan, L. Prati, Catal. Today
010, 150, 8–15.
[
[
[
[
[
[
[
34] E. Roze, P. Gravejat, E. Quinet, J. L. Rousset, D. Bianchi, J. Phys. Chem. C
[
2009, 113, 1037–1045.
2
35] F. Boccuzzi, A. Chiorino, M. Manzoli, P. Lu, T. Akita, S. Ichikawa, M.
Haruta, J. Catal. 2001, 202, 256–267.
36] K. Chakarova, M. Mihaylov, S. Ivanova, M. A. Centeno, K. Hadjiivanov, J.
Phys. Chem. C 2011, 115, 21273–21282.
[
[
11] W. Hou, N. A. Dehm, R. W. J. Scott, J. Catal. 2008, 253, 22–27.
12] M. Chen, D. Kumar, C. W. Yi, D. W. Goodman, Science 2005, 310, 291–
2
93.
[
13] M. Sankar, E. Nowicka, P. J. Miedziak, G. L. Brett, R. L. Jenkins, N. Dimitra-
tos, S. H. Taylor, D. W. Knight, D. Bethell, G. J. Hutchings, Faraday Dis-
cuss. 2010, 145, 341–356.
37] A. Villa, D. Wang, G. M. Veith, F. Vindigni, L. Prati, Catal. Sci. Technol.
2013, 3, 3036–3041.
38] M. B. Griffin, A. A. Rodriguez, M. M. Montemore, J. R. Monnier, C. T. Wil-
liams, J. W. Medlin, J. Catal. 2013, 307, 111 –120.
39] S. H. Pang, A. M. Roman, J. W. Medlin, J. Phys. Chem. C 2012, 116,
[
[
[
[
14] A. Savara, C. E. Chan-Thaw, I. Rossetti, A. Villa, L. Prati, ChemCatChem
2
014, 6, 3464–3473.
15] C. Keresszegi, D. Ferri, T. Mallat, A. Baiker, J. Phys. Chem. B 2005, 109,
58–967.
16] A. Villa, D. Wang, P. Spontoni, R. Arrigo, D. Su, L. Prati, Catal. Today
010, 157, 89–93.
1
3654–13660.
40] B. Panella, A. Vargas, D. Ferri, A. Baiker, Chem. Mater. 2009, 21, 4316–
322.
9
4
2
17] Q. He, P. J. Miedziak, L. Kesavan, N. Dimitratos, M. Sankar, J. A. Lopez-
Sanchez, M. M. Forde, J. K. Edwards, D. W. Knight, S. H. Taylor, C. J. Kiely,
G. J. Hutchings, Faraday Discuss. 2013, 162, 365–378.
Received: April 20, 2015
[
18] D. Ferri, A. Baiker, Top. Catal. 2009, 52, 1323–1333.
Published online on July 16, 2015
ChemCatChem 2015, 7, 2534 – 2541
www.chemcatchem.org
2541
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim