Key properties promoting high activity and stability of supported PdSb/TiO2 catalysts in the acetoxylation of toluene to benzyl acetate

Article abstract of DOI:10.1016/j.apcata.2011.03.025

The influence of thermal pretreatment in air, helium and 10% H ₂/He on the catalytic performance of a supported 9% Pd, 12% Sb/TiO₂ catalyst in the gas phase acetoxylation of toluene has been studied by integrated evaluation of catalytic tests and catalyst characterization using mainly XRD, TEM and XPS. A pretreatment temperature of 600 °C is beneficial for shortening the activation period, while the atmosphere during thermal pretreatment influences the long-term stability. Highest catalytic performance with a good long-term stability was reached with pretreatment in He, which creates Sb-containing Pd particles, on which a mixed Pd⁰/PdO surface is formed during time on stream. Sb stabilizes oxidized Pd species and leads to more stable catalysts in contrast to air-calcined samples. The catalyst pretreated in H₂/He was completely inactive, due to the formation of a stable Pd₈Sb₃ alloy.

Full text of DOI:10.1016/j.apcata.2011.03.025

Applied Catalysis A: General 398 (2011) 104–112
Contents lists available at ScienceDirect
Applied Catalysis A: General
journal homepage: www.elsevier.com/locate/apcata
Key properties promoting high activity and stability of supported PdSb/TiO2
catalysts in the acetoxylation of toluene to benzyl acetate
∗
∗
S. Gatla, N. Madaan, J. Radnik , V.N. Kalevaru, M.-M. Pohl, B. Lücke, A. Martin, A. Brückner
Leibniz-Institut für Katalyse e. V. an der Universität Rostock, Albert-Einstein-Str. 29a, D-18059 Rostock, Germany
a r t i c l e i n f o
a b s t r a c t
Article history:
The influence of thermal pretreatment in air, helium and 10% H /He on the catalytic performance of a
supported 9% Pd, 12% Sb/TiO2 catalyst in the gas phase acetoxylation of toluene has been studied by
2
Received 28 January 2011
Received in revised form 10 March 2011
Accepted 11 March 2011
integrated evaluation of catalytic tests and catalyst characterization using mainly XRD, TEM and XPS. A
◦
pretreatment temperature of 600 C is beneficial for shortening the activation period, while the atmo-
Available online 22 March 2011
sphere during thermal pretreatment influences the long-term stability. Highest catalytic performance
with a good long-term stability was reached with pretreatment in He, which creates Sb-containing Pd
Keywords:
Toluene
Benzyl acetate
Acetoxylation
TiO2 supported palladium
Deactivation
0
particles, on which a mixed Pd /PdO surface is formed during time on stream. Sb stabilizes oxidized Pd
species and leads to more stable catalysts in contrast to air-calcined samples. The catalyst pretreated in
H2/He was completely inactive, due to the formation of a stable Pd8Sb3 alloy.
© 2011 Elsevier B.V. All rights reserved.
Thermal pretreatment
1
. Introduction
Palladium particles supported on oxidic carriers are widely used
ditioning time [2,7]. For improving the industrial relevance of such
catalysts, the conditioning time must be shortened and the cat-
alyst stability must be increased while maintaining high activity
and selectivity.
Inspired by these previous findings, we have recently performed
a comprehensive study in which we explored the influence of dif-
ferent synthesis conditions and thermal pretreatment procedures
on the nature of the Pd particles [8]. As a main result it was found
that the conditioning time of a PdSb/TiO2 catalyst can be markedly
shortened, when the Pd particle size is purposefully increased prior
to the catalytic test by a dedicated thermal pretreatment in He
as catalysts for a variety of heterogeneous catalytic processes,
among them acetoxylation of hydrocarbons. Thus, acetoxyla-
tion of ethylene over supported Pd/SiO2 is the most important
technical process to produce vinyl acetate monomer [1]. In
the case of toluene, acetoxylation with molecular oxygen [2–5]
could be an attractive alternative for the production of benzyl
acetate and its hydrolysis product benzyl alcohol, which is still
based on the environmentally problematic chlorine chemistry [6]
(
Scheme 1).
or 10% H /He atmosphere. Unfortunately, the catalytic activity of
2
So far, a catalyst with a nominal content of 10 wt.% Pd and 8 wt.%
these catalysts was lower than that observed previously [2]. This
may have been due to the formation of surface sulfides during
thermal pretreatment in inert or reducing atmosphere, which was
negligible in the earlier studies, in which the catalysts were calcined
in air. These sulfides derived most probably from the reduction
of ammonium sulfate which, according to the previous prepara-
tion procedure [2], was used as an additive to remove chloride
Sb supported on TiO2 (anatase) has shown the best performance
with 86% selectivity of benzyl acetate (BA) at 68.5% toluene con-
version in the direct gas phase acetoxylation [2]. Remarkably, the
activity of this catalyst increased dramatically within the initial
1
0–12 h on stream and this went along with a marked increase
of the metal particle size. Unfortunately, this catalyst deactivated
rather quickly after reaching the maximum toluene conversion at
ca. 12 h. It has been shown that this deactivation can be effectively
suppressed by doping with bismuth or copper. However, this led
to a drop of the total activity and to a marked increase of the con-
(resulting from the SbCl starting compound) during thermal treat-
3
ment.
Therefore, we have developed a new synthesis procedure using
Sb O instead of SbCl3 as starting material, which made the addi-
2
3
tion of ammonium sulfate dispensable. In this work, we show that
thus prepared catalysts are much more active than those described
in [8] and, moreover, show a remarkably high long-term stabil-
ity, in contrast to previous catalysts [2] which deactivated rather
quickly.
∗ Corresponding author. Tel.: +49 0381 1281 244; fax: +49 0381 1281 51244.
E-mail addresses: joerg.radnik@catalysis.de (J. Radnik),
angelika.brueckner@catalysis.de (A. Brückner).
0
926-860X/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.apcata.2011.03.025

S. Gatla et al. / Applied Catalysis A: General 398 (2011) 104–112
105
cles (0.425–0.6 mm). The catalyst particles were diluted with the
fivefold amount of corundum particles to avoid hot spot forma-
tion during the reaction. The liquid reactants (toluene, >99.9%,
Roth, Germany, and acetic acid, >99.9%, Walter CMP, Germany) in
a molar ratio of 1:4) were dosed by a HPLC pump, while gases were
dosed by mass flow controllers, resulting in a total feed mixture
of toluene:acetic acid:oxygen:nitrogen = 1:4:3:16. Tests were per-
−
1
formed with a gas hourly space velocity of GHSV = 2688 h and a
residence time of ꢀ = 1.34 s. The product stream was collected in a
cold trap and analyzed off-line by gas chromatography (Shimadzu
GC-2010) using an WCOT fused silica capillary column. The column
outlet was connected to a methanizer (30% Ni/SiO2 catalyst) for
conversion of carbon containing products into methane, detected
by a Flame Ionization Detector (FID).
Scheme 1.
2
. Experimental
2.1. Catalyst preparation
Catalysts with a nominal content of 10 wt.% Pd and 16 wt.% Sb
on TiO2 (anatase, Kronos, containing about 2.3 wt.% S) were pre-
pared in two steps with an intermediate calcination after the first
step, in analogy to the procedure in Ref. [2]. While step 2 in both
2.3. Materials characterization
The elemental composition (Ti, Pd, Sb, Na, S) was determined
by ICP-OES using a Varian 715-ES ICP-emission spectrometer.
Prior to the analysis 10 mg of the sample was mixed with
5 ml of aqua regia and 3 ml hydrofluoric acid and treated in a
microwave-assisted sample preparation system “MULTI WAVE”
procedures is identical, Sb O3 was used instead of SbCl3 in the
2
present procedure and no ammonium sulfate was added in step
1
: (1) 1.9 g of Sb O powder (Sigma–Aldrich) was suspended in
2 3
2
0 ml distilled water and stirred for 20 min, thus, forming a homo-
◦
geneous slurry. Then 8.2 g of TiO2 were added to this suspension
and stirred for 1 h at room temperature. Upon addition of TiO2 the
pH of the suspension dropped to 1. This is most probably due to
(Anton Paar/Perkin–Elmer) at ≈200 C and ≈60 bar. The digested
solution was filled up to 100 ml and measured. The data analysis
was performed using the ICP Expert software. Carbon, hydrogen,
nitrogen and sulfur analysis was done with the CHNS micro-
analysator TruSpec (Leco).
the dissolution of traces of sulfate from the TiO support as sulfuric
2
acid. It should be mentioned that sulfur-free TiO was also tested as
2
support, however, led to inactive catalysts. Obviously, the surface
acidity of sulfated TiO2 is beneficial for the adsorption and conver-
sion of toluene. Thus, 25% ammonia solution was added dropwisely
The surface area (SBET) and pore size distribution of the cata-
lysts were determined on an ASAP – 2010 (Micromeritics) by N2
◦
physisorption at −196 C. Before the measurement, the catalyst was
◦
◦
to adjust pH = 7. The resulting suspension was heated at 70 C for
evacuated for 2 h at 150 C to remove physisorbed water.
1
h, followed by solvent removal in a rotary evaporator. The final
XRD powder patterns at ambient conditions were recorded in
◦
◦
˚
solid was dried at 120 C for 16 h and calcined in air at 400 C for
3
solved at 50 C in 12.1 ml H O (acidified by HCl to pH = 1). Despite
the fact that this introduces some chloride ions into the system,
PdCl2 was used as Pd source, since other Pd salts such as nitrate
or acetate led to poorly active catalysts [8]. As was shown in our
previous work, a small amount of chloride enables the intermedi-
ate formation of Pd chloro complexes which probably assist in the
segregation of Pd metal particles [8]. After cooling to room tem-
perature, pH = 4 was adjusted by dropwise addition of 1 N Na CO
transmission geometry with Cu K␣ radiation (ꢁ = 1.5406 A) in the
◦
◦
h. (2) In the second step, 1.7 g PdCl2 (>99.9%, Janapharm) was
2Â range of 10–55 (step width = 0.25 ) on a Stoe STADI P diffrac-
tometer, equipped with a linear Position Sensitive Detector (PSD).
Time per step was 25 s for the samples without Si as standard and
160 s with Si as standard). Processing and assignment of the pow-
der patterns was done using the software WinXpow (Stoe) and the
Powder Diffraction File (PDF) database of the International Centre
of Diffraction Data (ICDD).
◦
2
Transmission Electron Microscopy (TEM) investigations were
carried out at 200 kV using a CM-20 microscope (Philips) equipped
with an EDXS Noran Six (Thermo Scientific). Samples were pre-
pared by depositing the catalysts on a nickel grid (300 mesh) coated
with carbon film.
X-ray photoelectron spectra (XPS) were recorded on a VG
ESCALAB 220iXL instrument with Al K␣ radiation (E = 1486.6 eV).
The samples were fixed with a double adhesive carbon tape
on a stainless steel sample holder. The peaks were fitted by
Gaussian–Lorentzian curves after Shirley background subtraction.
The electron binding energy was referenced to the Ti 2p3/2 peak
of TiO2 at 458.8 eV. For quantitative analysis of the near-surface
region the peak areas were determined and divided by the element-
specific Scofield factor and the analysator-depending transmission
function.
2
3
.
2
−
(
>99%, Alfa) solution, which results in the formation of PdCl4
Pd–Cl bond scission seems not to occur during this step, since no
color change was observed and Pd remains coordinated by 4 Cl lig-
ands in the PdCl42− anions as it is the case, too, in solid PdCl . The
2
calcined powder prepared in step 1 was added to the above solution
and stirred for 1 h at room temperature, followed by solvent evapo-
ration in a rotary evaporator. The resulting solid was dried at 120 C
for 16 h. For testing the influence of different thermal pretreatment
procedures, one part of the sample was calcined in situ in the cat-
alytic reactor prior to the catalytic test in a flow of 27.6 ml/min air
◦
◦
at 300 C for 2 h, according to the previously used pretreatment [2].
The remaining parts were pretreated for 4 h externally in an oven at
◦
6
00 C under (a) oxidizing (air, 50 ml/min), (b) inert (He, 50 ml/min)
or (c) reducing (10%H /He, 50 ml/min) atmosphere. Highly purified
2
(
>99.99) gases were used for thermal treatment and for catalytic
tests.
ICP-OES analysis of the final catalyst revealed a Pd and Sb con-
3
. Results
3.1. Catalytic tests
tent of 9.1% and 12.4% respectively, suggesting the loss of some
volatile Sb during calcination in the first step.
In Fig. 1 the catalytic performance of the 9Pd12Sb/TiO2 catalyst
obtained after the three different thermal pretreatment proce-
dures (see Section 2.1) is compared. The catalytic performance after
thermal pretreatment at 300 C in air is similar to that of the con-
2
.2. Catalytic tests
◦
Catalytic tests were performed in a fixed-bed Hastelloy C micro
◦
reactor at 210 C and 2 bar total pressure with 0.8 g catalyst parti-
ventionally prepared PdSb/TiO2 catalyst from Ref. [2], which was

1
06
S. Gatla et al. / Applied Catalysis A: General 398 (2011) 104–112
between 0 and 2%). Equilibration time and maximum catalytic
performance do not change when the atmosphere of the thermal
◦
pretreatment at 600 C changes from air to He (Fig. 1c). However,
a significant improvement of the catalyst stability is observed in
the latter case, reflected by only a 10% drop of the BA yield within
2
4 h on stream. The most surprising effect was seen with a catalyst
◦
after thermal pretreatment in 10% H /He at 600 C. This sample
2
revealed to be totally inactive with a toluene conversion less than
1
%. Therefore, these results are not plotted in Fig. 1.
From these results it is clearly evident that the thermal pre-
treatment has a significant impact on the catalytic performance.
While the pretreatment temperature is crucial for the duration
of catalyst equilibration, the surrounding atmosphere is more
important for the BA selectivity as long as it has no reducing
properties which lead to complete activity loss. To get deeper
insights into this unexpected behavior, detailed characterization of
catalysts was performed after thermal pretreatment, after reach-
ing maximum catalytic performance at 9 h and 6 h on stream for
samples calcined in air and He, respectively, and after 30 h on
stream. For this purpose, the reaction feed was stopped and the
system was cooled slowly to room temperature in inert atmo-
sphere.
3
3
.2. Catalyst characterization
.2.1. Elemental analysis and surface area
Inspection of Table 1 shows that the Pd content of calcined and
used catalysts is almost equal to the nominal value. In contrast, the
corresponding Sb value after calcination is slightly lower, point-
ing to some loss of Sb as volatile species during thermal treatment,
especially for the He treated samples. No further change in the Pd
and Sb contents is observed during use in the reaction. A mini-
mum percentage of 1.5% sulfur is still detected, which is however
much smaller than the 6–7% detected in the catalysts of Ref. [8],
in which (NH ) SO was used as additive. This small amount of
4
2
4
residual sulfur is part of the TiO2 support, yet it might not have a
detrimental impact on the catalytic performance, since sulfur-free
anatase, which has also been tested as support, does not lead to
improved catalysts.
An interesting trend is observed for the BET surface areas
(Table 1). A marked decrease of the surface area and pore volume
of the pure support is evident upon loading with the metal com-
ponents. This may be partly due to pore blocking of the support
with the metal components. In addition, the intermediate calcina-
tion between step 1 and 2 in the preparation may also have affected
the BET surface area. The decrease of the surface area and pore vol-
ume is partly reversible after calcination in air but not in helium
Fig. 1. Toluene conversion (X-Tol), selectivity (S-BA) and yield (Y-BA) of benzyl
◦
acetate during time on stream over the 9Pd12Sb/TiO2 catalyst pretreated (a) at
at 600 C. On the other hand, these values decrease again when the
◦ ◦ ◦
00 C in air for 2 h, (b) at 600 C in air for 4 h and (c) at 600 C in helium for 4 h.
3
air-calcined sample is used in the acetoxylation reaction while only
a slight effect is observed for the helium-pretreated catalyst. This
suggests that high-temperature treatment in oxidizing atmosphere
may have an impact on the support structure, most probably on the
TiO2 crystallite size which is smaller for the samples after air pre-
treatment (see supplementary data, Fig. S1 and XRD results below).
This may be the main reason for the difference in the BET surface
areas.
pretreated after the same procedure but prepared using SbCl3 and
NH ) SO . The maximum BA yields are continuously increasing
(
4
2
4
and reach a maximum after about 10–12 h, followed by rapid deac-
tivation. This confirms clearly that the new preparation method
used in this work leads to equally effective catalysts as the conven-
tional procedure used in Ref. [2].
A significant difference can be seen when precalcination in air
is done at 600 C. In this case, the conditioning period shortens
In former studies, coke deposition was found as main reason
for the deactivation of the PdSb/TiO2 catalysts [2]. Therefore, the
amount of C was also analyzed in differently pretreated cata-
lysts after different time on stream (Table 1). As expected, coke
deposition increases with time on stream for all samples. The high-
◦
to about 5–6 h during which the BA yield steeply increases and
then reaches a plateau. The maximum BA selectivity is higher than
◦
after calcination at 300 C (Fig. 1b).The main reason for the higher
◦
selectivity is most probably the less pronounced carbonization of
est amount of carbon is found on the sample calcined at 300 C,
◦
◦
toluene over the catalysts pretreated at 600 C leading to a sig-
whereas for both samples treated at 600 C nearly identical C con-
nificantly lower amount of carbon deposits in the used samples
tents were found, independent on the pretreatment atmosphere
(Table 1). Nevertheless, the catalyst calcined in He at 600 C is more
◦
(
Table 1). Other byproducts are CO2 and small traces of benzy-
laldehyde, benzylidene acetates and benzyl alcohol (selectivities
stable than the one calcined in air at the same temperature (Fig. 1).

S. Gatla et al. / Applied Catalysis A: General 398 (2011) 104–112
107
Table 1
Elemental composition, BET surface area and pore volume of 9Pd12Sb/TiO2 catalysts after different thermal pretreatments given in brackets and corresponding spent samples
after different time on stream. Sulfur content and surface area of pure TiO2 is also shown.
2
3
Sample-time on stream
Pure TiO2
Pd
Sb
S
Na
C
BET-surface area (m /g)
Pore volume (cm /g)
–
–
2.3
1.5
1.3
1.6
1.4
1.3
1.4
1.5
1.6
1.6
–
–
–
–
4.25
–
0.5
2.2
–
0.65
2.4
315
0.27
0.12
0.14
0.10
0.38
0.15
0.12
0.10
0.13
0.11
Fresh
9.1
7.7
8.1
8.5
8.6
8.8
9.3
8.2
9.1
12.4
10.5
9.7
12.2
12.2
12.5
9.3
5.3
4.6
5.3
5.1
5.0
5.2
5.7
5.9
5.35
56.6
81.5
64.6
115.3
71.6
52.8
28.6
40.6
35.5
◦
◦
◦
(
(
(
(
(
(
(
(
300 C-air)
300 C-air) – 32 h
600 C-air)
◦
◦
◦
600 C-air) – 9 h
600 C-air) –s 30 h
600 C-He)
◦
◦
600 C-He) – 6 h
10.2
8.8
600 C-He) – 32 h
This suggests that coke formation is not the only reason for deacti-
vation.
ration of Sb into the Pd lattice is possible for Pd/Sb ratios higher
than 5, while for Pd/Sb ratios between 3 and 0.5 stable alloy phases
such as Pd Sb , Pd Sb , Pd Sb and Pd20Sb7 are formed [17]. In
1
2
1
1
8
3
our case, this happened when the sample was pretreated in reduc-
3.2.2. X-ray diffraction
ing atmosphere, namely 10% H /He. In Fig. 2a the reflections of a
2
The XRD patterns of all samples show of course the reflections
Pd Sb alloy phase are clearly seen after this treatment.
8
3
of the anatase support [PDF No. 21-1272] besides those of Si which
was used as an internal standard (Fig. 2a) for the exact determi-
nation of peak positions and intensity. Besides, some small traces
Inspection of the intensity of the Pd reflections of the He-
pretreated catalyst after 6 h and 32 h on stream (Fig. 2b, Table 2)
◦
◦
clearly shows that the two subsignals at 39.3 and 39.5 increase at
◦
of NaCl [PDF No. 5-628] and Na SO4 [PDF No 75-914] are seen in
the expense of the peak at 39.8 , suggesting that the incorporation
2
certain samples, which might have formed from Na CO and PdCl
of Sb (and possibly also C) into Pd increases with time on stream.
When the catalysts had been pretreated in air instead of He, the
downshift of the Pd reflections with increasing time on stream is
less pronounced, suggesting that a major part of the metal phase
might persist as pure Pd.
2
3
2
or sulfate contained in the TiO2 support, respectively. All XRD pat-
terns of the spent samples and also of the fresh catalysts pretreated
◦
in He at 600 C show clearly the presence of a metallic Pd phase
[
6
PDF No. 88-2335], while Pd in the fresh catalyst calcined in air at
◦
00 C is present as PdO [PDF No. 41-1107]. In contrast, when ther-
Further information about composition, morphology and size of
the metal particles has been derived by TEM described below.
◦
mal pretreatment is performed in 10% H /He at 600 C, no Pd but
2
a Pd Sb alloy phase [PDF No. 89-2059] is formed. The Full Width
8
3
at Half Maximum (FWHM) of the anatase peaks is broader for the
samples calcined in air, suggesting lower crystallites size. This is
evident from Fig. S1 (supplementary data) in which the TiO2 crys-
tallite size, calculated by the Scherrer equation, for samples after
thermal pretreatment in air and He is compared. This could be one
reason for the higher BET surface area observed in the air-calcined
3
.2.3. Transmission electron microscopy
TEM/EDX results of the catalyst after calcination in air at 600 C
◦
and after 30 h use in the acetoxylation reaction are depicted in
Fig. 3. After calcination in air, Pd-containing particles with a size of
5
–10 nm are well dispersed over the whole support surface. With
regard to the XRD pattern (Fig. 2a), it is most probable that these
particles consist of PdO. From EDX alone this conclusion cannot
be derived since, due to their small size, the electron beam spot
cannot be exclusively focused on a Pd-containing particle alone
but hits always a certain area of the support as well. Therefore,
radiation emitted from oxygen in PdO and TiO2 cannot be distin-
guished. Besides areas with a rather high Pd/Sb ratio, probably
arising from the joint detection of small PdO particles and the
underlying Sb-containing TiO2 support (Pd/Sb = 5.4, Fig. 3a), there
are Sb-containing support areas without any Pd (Fig. 3a) or with
only traces of Pd (Fig. 3b). This suggests that in contrast to Pd which
forms a crystalline PdO phase, Sb is highly dispersed on the sup-
sample (Table 1). In addition to TiO crystallites, the changes of the
2
BET surface described above can be explained also by the different
Pd crystallite size. Very sharp Pd reflections indicating large crys-
tallites were found for the sample after He pretreament with the
lowest BET surface. The half widths of the Pd reflections after the
different stages of use show no significant differences. This could
explain why the BET surface areas of the samples pretreated in air
◦
and helium at 600 C do not differ much after 30 h and 32 h on
stream respectively (Table 1).
Significant differences are evident in the shape of the Pd (1 1 1)
reflections, which have been deconvoluted into several subsignals
(
Fig. 2b). The position, intensity and FWHM of these signals are
listed in Table 2. A single symmetric peak appears in the fresh
sample after thermal pretreatment in He at 40.01 , which is con-
Table 2
◦
Peak position, peak area and FWHM derived from XRD patterns of 9Pd12Sb/TiO2
samples after different time on stream.
sistent with the presence of a crystalline metallic Pd phase with
cubic structure and a Fm-3m (2 2 5) space group [9]. After use in
the catalytic reaction, this peak splits into three subsignals, which
are all shifted to lower values of 2Â in comparison to the peak of
pure metallic Pd. This behavior points to a change of the Pd lat-
tice during time on stream, which is most probably caused by the
incorporation of other components such as Sb, C and/or H. Thus, the
dissolution of C [10,11] and H [12] into the Pd lattice led to slight
downshifts of the 2Â values, as likewise observed in Fig. 2b, while
this was not observed for oxygen. In the latter case, only surface
or bulk Pd oxides were formed, the XRD reflections of which do
not appear in the metallic Pd region [13–16]. TEM measurements
shown below confirm the incorporation of Sb into the Pd lattice of
the He-calcined sample. According to the phase diagram, incorpo-
◦
Sample-time on stream
Sub peak position 2Â/
Peak area
FWHM
(
(
600-He) fresh
600-He) – 6 h
40.01
39.3
111.6
40.2
63.5
221.2
88.1
102.4
106.3
92.67
69.68
24.94
158.2
46.6
0.16
0.32
0.23
0.31
0.29
0.30
0.39
0.34
0.45
0.40
0.40
0.30
0.30
3
3
9.5
9.8
(600-He) – 32 h
39.3
39.5
39.8
(
600-air) – 9 h
39.5
3
4
9.9
0.2
(600-air) – 30 h
39.5
3
4
9.9
0.2
29.17

1
08
S. Gatla et al. / Applied Catalysis A: General 398 (2011) 104–112
Fig. 2. XRD patterns of calcined 9Pd12Sb/TiO2 samples in air, He, 10%H2/He and respective spent samples after 6–32 h of reaction (left) and deconvolution of the metallic Pd
◦
(
1 1 1) peak into subpeaks (right). Thermal pretreatment was done at 600 C for 4 h in all cases. 20% Si powder was added to all samples as internal standard: (+ Na2SO4, ꢀ
PdO; # Pd, ଝ Pd8Sb3, ꢀ NaCl, ꢂ TiO2).
port. This agrees well with the fact that no crystalline Sb O3 has
been observed by XRD.
in contrast to the latter catalyst, in which the formed Pd particles
contain only traces of Sb or Sb particles with traces of Pd (Fig. 3c
and d), those in the He-pretreated catalyst are richer in Sb, show-
ing uniform Pd/Sb ratios around 5 (Fig. 4c and d), while the rest of
the Sb is spread over the TiO2 support without forming a separate
phase. However, the Sb content detected by EDX on several spots of
the support is smaller than observed for the air-calcined catalyst,
which might be a consequence of the higher Sb enrichment in the
Pd particles. Furthermore, on certain Pd particles which are well
separated from the support, traces of oxygen have been detected
(Fig. 4c), suggesting a partial oxidation of their surface during time
on stream. This has been further explored by XPS described below.
2
After 30 h use in the catalytic reaction, Pd particles of about
4
0–100 nm are formed, which are very well crystallized showing a
facetted structure (Fig. 3c and d). These particles do contain only
some traces of Sb. On the other hand, nearly Pd-free Sb particles
were also observed. This shows clearly that the Pd and Sb compo-
nents remain widely separated in the catalyst.
Very different TEM/EDX results have been obtained from the
◦
catalyst pretreated in He at 600 C. Large Pd particles of up to 2 ␮m
diameter have been formed during pretreatment in He, leaving a
considerable part of the support free of Pd (Fig. 4a and b) but con-
taining Sb as for the air calcined samples. After 32 h on stream in
the catalytic reaction, surprisingly, a marked decrease of the par-
ticle size down to 40–100 nm was observed, which is equal to the
size of the Pd particles formed during time on stream from the air-
calcined catalyst (compare Figs. 3c and d and 4c and d). However,
3.2.4. X-ray photoelectron spectroscopy
In Fig. 5 XP spectra of the Pd 3d region are shown for samples
◦
after thermal pretreatment at 600 C in air or He, after stopping the
catalytic test in the state of maximum activity (6–9 h on stream)
◦
Fig. 3. Transmission electron micrographs of 9Pd12Sb/TiO2 samples after 4 h calcination in air at 600 C (a and b) and after 30 h use in the catalytic reaction (c and d).

S. Gatla et al. / Applied Catalysis A: General 398 (2011) 104–112
109
◦
Fig. 4. Transmission electron micrographs of 9Pd12Sb/TiO2 samples after 4 h thermal treatment in He at 600 C (a and b) and after 32 h use in the catalytic reaction (c and d).
and after 30–32 h on stream. The Pd 3d5/2 peak of the catalyst after
calcination in air falls at 335.8 eV. This value is in the range of oxidic
Pd. This agrees properly with the XRD results in which a crystalline
PdO phase has been detected for this catalyst (Fig. 2a). However,
this value is considerably lower than the binding energy of 336.5 eV
measured for pure PdO (Fig. 5a) [18]. A strong interaction between
TiO2 and the tiny PdO particles can be assumed as reason for the
lower binding energy compared to unsupported PdO. Remarkably,
the spectrum of the catalysts after pretreatment in He is almost
identical to the one after air-calcination (Fig. 5b), although metallic
Pd was detected by XRD in this case. This might be due to the forma-
tion of surface oxide species on the large Pd particles obtained after
the pretreatment in He. It is not likely that these species have been
formed by contact with ambient atmosphere at room temperature
since similar effects are not observed in other samples discussed
below.
In the catalysts removed from the reactor after reaching maxi-
mum activity (Fig. 5c and d), two Pd 3d5/2 peaks are observed with
binding energies of 335.1 eV arising from pure metallic Pd [19]
and 336.5 being characteristic for PdO [18]. In the He-pretreated
sample, the metallic Pd peak is broader, indicating a higher het-
erogeneity of the Pd environment and the PdO peak is more
pronounced than in the air-calcined sample, in which the majority
of Pd on the surface is purely metallic with only a minor contribu-
tion of PdO. This could be due to the incorporation of Sb into the
Pd phase of the He-pretreated catalyst (as evidenced by TEM and
XRD), which might increase the heterogeneity of the Pd environ-
ment and obviously stabilizes PdO surface species. Both effects are
much less prominent in the XP spectrum of the air-calcined sample
(
Fig. 5d), in which TEM and XRD suggest that Pd and Sb are spatially
◦
Fig. 5. XP spectra of the Pd 3d peak of 9Pd12Sb/TiO2 samples: (a) after 4 h at 600
C
widely separated.
◦
in He, (b) after 4 h at 600 C in air, (c) He-pretreated sample after 6 h on stream, (d)
air-pretreated sample after 9 h on stream, (e) He-pretreated sample after 32 h on
stream and (f) air-pretreated sample after 30 h on stream.
After extended time on stream, the resolved PdO peaks disap-
peared completely in both samples (Fig. 5e and f). However, while
an essentially pure Pd surface is evidenced for the air-calcined cat-

1
10
S. Gatla et al. / Applied Catalysis A: General 398 (2011) 104–112
alyst, the slight shift of the Pd 3d peak to higher binding energy in
the He-pretreated catalysts indicates that the metal surface of the
latter remains still partially oxidized. Since both samples have been
exposed to ambient atmosphere between removal from the reactor
and insertion into the analysis chamber of the XP spectrometer, it is
clear that intermediate contact to air does not lead to a reoxidation
of metallic Pd in the used samples. The XP spectra of the samples
pretreatment procedures (Fig. 1), it is obvious that the latter do not
only influence the conditioning time needed to reach maximum
activity, but have a major impact also on the deactivation behavior.
The most surprising result was the total inactivity of the catalyst
pretreated in a flow of 10% H /He, over which the toluene conver-
2
sion was below 1%. As shown by XRD, this pretreatment converts
the whole Pd amount into a well defined crystalline Pd Sb3 alloy,
8
pretreated in H /He flow show a Pd peak at a binding energy of
which is obviously not active. Since XPS analysis of this sample
clearly reveals only one type of Pd before and after the reaction
evidenced by a binding energy of 335.6.eV being similar to that
found in other PdSb alloys [8], it is very probable that there is no
2
3
35.6 eV after both the pretreatment and after 8 h on stream, which
is similar to the results published in [8] and explained by the for-
mation of a PdSb alloy. No evidence for PdO was found at these
samples. In all used catalysts, Sb is present in oxidized form, evi-
denced by a Sb 3d3/2 peak with a binding energy of 540 eV, being
other Pd phase besides the inactive Pd Sb3 phase.
8
From Fig. 1 it is evident that the catalyst pretreated in air for
◦
characteristic for Sb O3 [20].
2 h at 300 C requires the longest conditioning period. Obviously
2
For exploring changes in the near-surface region in more detail,
the metal (Pd, Sb)-to-Ti surface atomic ratios have been calculated
based on the areas of the respective Pd3d, Sb3d3/2, Ti2p XPS peaks
these conditions are too mild to create Pd-containing particles
of suitable size. The conditioning time can be markedly short-
ened when the pretreatment is done at 600 C for 4 h. Despite
◦
(
see supplementary data, Fig. S2). For the catalyst pretreated in air
the fact that the size and composition of the Pd-containing par-
ticles differs markedly, depending on whether the pretreatment
was done in air (PdO particles of 5–10 nm) or helium (Pd parti-
cles of up to 1 ␮m), the conditioning time and also the maximum
toluene conversion and BA selectivity reached after this time are
rather similar (Fig. 1b and c). As we have shown recently [8],
there is no doubt that the conditioning time can be shortened
by pre-forming Pd particles of suitable size using a tailored ther-
mal pretreatment. However, inspection of Fig. 1 suggests that
a rather small size of the starting particles might be sufficient
for this effect. Despite the large difference in the starting parti-
cle size after air and helium pretreatment, the latter approaches
each other during conditioning, which leads to Pd particles of
about 40–100 nm independent of the pretreatment atmosphere
(Figs. 3 and 4).
◦
at 600 C, the Pd/Ti ratio decreases with time on stream, proba-
bly due to Pd particle growth observed with TEM. On the other
hand, a slight enrichment of Sb in the surface region was observed.
Hence, these opposite trends in the changes of the surface con-
centration (Fig. S2), suggest a separation of Pd and Sb, which is also
supported by TEM results. For the sample pretreated in He, the Sb/Ti
ratio increases during the conditioning period while the Pd/Ti ratio
remains nearly constant. Upon further time on stream, both ratios
decrease in parallel. This suggests spreading of Sb on the surface of
the metal particles during conditioning, followed by intermixing of
Pd and Sb during further time on stream.
Since coke deposition as a possible reason for catalyst deactiva-
tion occurs during time on stream, the C1s region has been analyzed
in detail (Fig. 6a). In all samples, a C1s signal at 285 eV is evident,
which is typical for C- and H-bound C atoms [21]. The He-pretreated
catalyst in the state of its maximum activity shows exclusively this
peak. After 32 h on stream, during which a ≈10% drop in activity was
observed (Fig. 1c), an additional C1s signal at 289.5 eV is formed,
which is characteristic for C in carboxyl groups [21] (Fig. 6a). In
the air-pretreated catalyst, this peak is already present after 9 h
on stream. A comparison of the relative surface percentage of the
types of carbon at spent samples having been pretreated in air or
He is shown in Fig. 6b. It is clearly seen that the total amount of sur-
face carbon deposits in the air-calcined catalysts reached its final
value already in its state of maximum activity (after 9 h on stream)
and then remained constant. In contrast, the total C amount on the
surface of the He-pretreated sample increased by a factor of two
after 32 h on stream. Interestingly, the total bulk carbon content
The surface area of the TiO2 support seems to be less important
for the catalytic behavior. It decreases to a very different extent
◦
during thermal pretreatment at 600 C in air and He, namely to
115.3 m /g and 28.6 m /g, respectively (Table 1). Although this may
2
2
be one reason why the metal particles after treatment in helium are
larger in comparison to calcination in air (compare Figs. 3 and 4),
this does obviously not influence the catalytic behavior during con-
ditioning which is very similar in both cases (Fig. 1). This suggests
that the performance of the catalysts is governed by the nature
of the Pd particles and probably also by the surface acidity of the
support but not so much by its surface area.
More important, especially for the stability of the catalysts, is
obviously the composition of the metal particles, which changes
as well during conditioning. XRD and TEM results indicate that the
very large pure Pd particles present after pretreatment in helium,
restructure during time on stream into smaller ones. Obviously this
is caused by migration of Sb species from the support surface into
the Pd phase, since the latter contains a considerable percentage
of Sb after a certain time on stream, yet without formation of an
alloy phase. Sb species exposed on the particle surface are essen-
tially trivalent, as shown by XPS. However, it is probable that in
the bulk of the Pd particles they exist in zerovalent form like in
PdSb alloys, even though no crystalline alloy is formed in this cat-
alyst.
◦
for the two samples pretreated at 600 C (Table 1) does not differ
much, amounting to 2.2 and 2.4%. Obviously the carbon is a little
more enriched on the surface of the catalyst pretreated in He. Nev-
ertheless, its degree of deactivation after 32 h on stream is lower
compared to the air-pretreated catalyst (Fig. 1b and c). Probably,
carboxylic C deposits as formed in the latter catalysts are stronger
deactivating in comparison to C- and H-bound C deposits in the
He-pretreated catalyst.
4
. Discussion
When the catalyst was pretreated in air, the starting particles
are much smaller and consist essentially of PdO. These particles
are reduced to metallic Pd and agglomerate during conditioning,
yet without remarkable incorporation of Sb. Taking the catalytic
test result into account (Fig. 1b and c), it appears that the intermix-
ing of Pd and Sb within the metal particles is essential for ensuring
high catalytic performance and long-term stability, as long as the
incorporated Sb content remains low enough to prevent alloy for-
mation, which is well known for Pd/Sb ratios of 3:1, 2:1, 1:1 and
1:2 [17].
In this section, an integrated discussion of results of catalytic
tests and catalyst characterization is presented to identify reasons
of activation and deactivation. First of all, it is evident that the new
preparation procedure in which Sb O3 is used as Sb source instead
2
of SbCl3 without the addition of (NH ) SO as Cl-removing agent,
4
2
4
suppresses poisoning of the catalyst surface with sulfide and leads
to much more active catalysts than obtained in Ref. [8]. By compar-
ing the catalytic performance of the samples after different thermal

S. Gatla et al. / Applied Catalysis A: General 398 (2011) 104–112
111
Fig. 6. (a) XP spectra of the C1s state, (b) relative amount of carboxylic (289.5 eV) and C- and/or H-bound carbon (285 eV) present at the surface of spent 9Pd12Sb/TiO2
samples pretreated in air or He.
Besides the bulk composition of the metal particles, major dif-
ferences have also been observed in their surface composition by
XPS (Fig. 5). In the state of maximum activity pure metallic Pd and
pretreated in helium shows a more stable toluene conversion and
BA selectivity. This is most probably due to the co-existence of Pd0
and PdO in the latter. This might be favoured by the intermixing of
Sb into the Pd particles, which is assumed to stabilize oxidized Pd
in its vicinity.
Another reason for the more pronounced deactivation of the air-
pretreated catalysts results from the nature of the carbon surface
deposits. These contain a higher percentage of C O species which
could point to carboxylate-like intermediates that are accumulated
on the Pd particle surface, on which PdO with its higher oxidation
potential is missing. In summary, it can be stated that the bulk prop-
erties of the metal particles formed during pretreatment play an
important role for long-time stability by governing the formation
of distinct Pd species on the surface during time on stream.
0
oxidized PdO species coexist on the surface. This is still the case for
the He-pretreated sample after 32 h on stream (Fig. 5e) while the
surface of the air-pretreated catalyst lost its oxidized PdOx after the
same time. This goes along with a loss of about 25% in the BA yield,
while the latter decreases only by about 10% over the He-pretreated
catalyst under the same conditions (Fig. 1b and c). This suggests that
the co-existence of both Pd and PdO species is beneficial for high
catalytic performance. This has also been proposed for supported
Pd catalysts in other oxidation reactions, e.g. the total oxidation
of methane [22–25]. It was suggested that the hydrocarbons are
adsorbed on the metallic Pd surface sites, while PdO is responsible
for their oxidation. Probably, a similar mechanism is valid for the
acetoxylation of toluene. Considering the composition of the metal
particles, it may be assumed that dissolution of Sb within the Pd
particle surface stabilizes oxidized Pd in its vicinity.
Acknowledgement
Financial support by Deutsche Forschungsgemeinschaft (grant
no. BR-1380/13-1) is gratefully acknowledged.
Finally we have to consider the impact of the carbon deposits.
Although the total amount of surface carbon is lower over the
air-pretreated catalyst, its deactivation is more pronounced. XPS
revealed that the nature of the carbon deposits after the pre-
Appendix A. Supplementary data
◦
treatments in air and He at 600 C is different (Fig. 6). In the
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.apcata.2011.03.025.
air-pretreated catalyst, the relative percentage of carboxylic car-
bon is higher than after pretreatment in helium. Possibly, product
0
desorption is less easy on the pure Pd particle surface of the former
References
catalyst, leading to a more pronounced enrichment of carboxylate
species, compared to the He-pretreated catalyst.
[1] T. Wang, J.A. Broussard, US Pat., 5,700,753 (1996).
[
2] J. Radnik, A. Benhmid, K.V. Narayana, M.-M. Pohl, A. Martin, B. Lucke, U.
Dingerdissen, Angew. Chem. Int. Ed. 44 (2005) 6771–6774.
5
. Conclusions
[
[
[
[
3] K. Ebitani, K.-M. Choi, T. Mizugaki, K. Kaneda, Langmuir 18 (2002) 1849–1855.
4] T. Miyake, T. Asakawa, Appl. Catal. A 280 (2005) 47–53.
5] T. Komatsu, A. Onda, Catal. Surv. Asia 12 (2008) 6–15.
6] T. Okada, M. Hanaya, A. Hattori, H. Hamaji, T. Miyake, S. Tokumaru, N. Nagira,
S. Ikumi, T. Hori, N. Norimasa, EP. Pat., 0 778 257 B1, 1996.
The conditioning time of PdSb/TiO2 catalysts can be markedly
shortened by a thermal pretreatment which is severe enough
◦
(
600 C, 4 h) to initiate the pre-formation of Pd-containing particles
[7] J. Radnik, M.-M. Pohl, V.N. Kalevaru, A. Martin, J. Phys. Chem. C 111 (2007)
0166–10169.
1
from the PdCl2 precursor on the TiO2 support. However, the per-
formance and particularly the stability of the catalysts depend on
the atmosphere of the thermal pretreatment. Reducing atmosphere
[
8] S. Gatla, J. Radnik, V.N. Kalevaru, M.M. Pohl, M. Schneider, B. Lucke, A. Martin,
N. Madaan, A. Brückner, Phys. Chem. Chem. Phys. 12 (2010) 4833–4842.
[9] J. Haglund, A.F. Guillermet, G. Grimvall, M. Körling, Phys. Rev. B 48 (1993)
1685–11691.
1
(
10% H /He) leads to the formation of completely inactive PdSb
2
[
[
10] N. Krishnankutty, M.A. Vannice, J. Catal. 155 (1995) 312–326.
11] N. Krishnankutty, J. Li, M.A. Vannice, Appl. Catal. A 173 (1998) 137–144.
alloy particles. Calcination in air creates PdO particles of 5–10 nm,
0
which are reduced to almost pure Pd with only traces of incor-
[12] B.M. Khanuja, R. Mehta, P. Agar, P.K. Kulriya, D.K. Avasthi, J. Appl. Phys. 106
(2009) 093515–093518.
porated Sb during time on stream and which grow to 40–100 nm.
In contrast, a helium pretreatment atmosphere forms very large
[
[
13] V.A. Bondzie, P. Kleban, D.J. Dwyer, Surf. Sci. 347 (1996) 319–328.
14] E.H. Voogt, A.J.M. Mens, O.L.J. Gijzeman, J.W. Geus, Surf. Sci. 373 (1997)
210–220.
0
initial Pd particles (up to 1 ␮m) with a partially oxidized surface,
that are restructured to smaller ones of 40–100 nm by incorpora-
tion of Sb up to a ratio of Pd/Sb ≈ 5, yet without formation of stable
alloy phases. The surface of these particles contains both, metallic
[15] F.P. Leisenberger, G. Koller, M. Sock, S. Surnev, M.G. Ramsey, F.P. Netzer, B.
Klotzer, K. Hayek, Surf. Sci. 445 (2000) 380–393.
[
16] B. Klotzer, K. Hayek, C. Konvicka, E. Lundgren, P. Varga, Surf. Sci. 482–485 (2001)
37–242.
[17] H. Okamoto, J. Phase. Equilib. 13 (1992) 578–579.
2
0
Pd and PdO species while the surface of the Pd particles of the air-
[
[
18] C.J. Powell, Appl. Surf. Sci. 89 (1995) 141–149.
pretreatedcatalyst, thoughcontainingsome oxidized Pd inthe state
of highest activity, becomes essentially free of PdO after extended
time on stream. In contrast to the air-pretreated catalyst, the one
19] H. Gabasch, W. Unterberger, K. Hayek, B. Klotzer, E. Kleimenov, D. Teschner, S.
Zafeiratos, M. Havecker, A.K. -Gericke, R. Schlogl, J. Han, F.H. Ribeiro, B.A. -Kiss,
T. Curtin, D. Zemlyanov, Surf. Sci. 600 (2006) 2980–2989.

1
12
S. Gatla et al. / Applied Catalysis A: General 398 (2011) 104–112
[
20] R. Delobel, H. Baussart, J. Leroy, J. Grimblot, L. Gengembre, J. Chem. Soc. Faraday
[23] P. Salomonsson, S. Johansson, B. Kasemo, Catal. Lett. 33 (1995) 1–13.
[24] J.G. McCarty, Catal. Today 26 (1995) 283–293.
[25] K. Okumura, T. Kobayashi, H. Tanaka, M. Niwa, Appl. Catal. B 44 (2003)
325–331.
Trans. I 79 (1983) 879–891.
[21] NIST Standard Reference Database 20, version 3. 5 http://srdata.nist.gov/xps.
22] S.H. Oh, P.J. Mitchell, R.M. Siewert, J. Catal. 132 (1991) 287–301.
[