Direct synthesis of hydrogen peroxide on zirconia-supported catalysts under mild conditions

Article abstract of DOI:10.1016/j.jcat.2006.02.014

Palladium catalysts supported on multiscripts(SO, 4, mml:none(), mml:none(), =)-, Cl^--, F^--, and Br^--doped zirconia were tested for the direct synthesis of hydrogen peroxide under very mild (1 bar and 20 °C) and nonexplosive conditions. The catalysts were characterized by thermogravimetric/differential scanning calorimetry analysis, N₂ physisorption, and temperature-programmed reduction before and after catalytic tests to investigate the oxidation state of the metal. The catalytic tests were carried out in different solvents, and the effect of the Pd oxidation state was ascertained. The best catalytic results were observed in methanol, using H₂/O₂ mixtures containing a large excess of oxygen and using the sulfate-doped zirconia catalyst. Surface-oxidized Pd⁰ catalysts showed high catalytic activity and the highest selectivity.

Full text of DOI:10.1016/j.jcat.2006.02.014

Journal of Catalysis 239 (2006) 422–430
www.elsevier.com/locate/jcat
Direct synthesis of hydrogen peroxide on zirconia-supported
catalysts under mild conditions
Stefano Melada, Riccardo Rioda, Federica Menegazzo, Francesco Pinna, Giorgio Strukul ∗
Dipartimento di Chimica, Università di Venezia, and Consorzio INSTM, Dorsoduro 2137, 30123 Venezia, Italy
Received 20 February 2006; revised 21 February 2006; accepted 23 February 2006
Available online 23 March 2006
Abstract
=
−
−
−
Palladium catalysts supported on SO -, Cl -, F -, and Br -doped zirconia were tested for the direct synthesis of hydrogen peroxide under
4
◦
very mild (1 bar and 20 C) and nonexplosive conditions. The catalysts were characterized by thermogravimetric/differential scanning calorimetry
analysis, N physisorption, and temperature-programmed reduction before and after catalytic tests to investigate the oxidation state of the metal.
2
The catalytic tests were carried out in different solvents, and the effect of the Pd oxidation state was ascertained. The best catalytic results
were observed in methanol, using H /O mixtures containing a large excess of oxygen and using the sulfate-doped zirconia catalyst. Surface-
2
2
0
oxidized Pd catalysts showed high catalytic activity and the highest selectivity.
 2006 Elsevier Inc. All rights reserved.
Keywords: Hydrogen peroxide direct synthesis; Palladium; Zirconia; Sulfate
1. Introduction
Several patents concerning the direct synthesis of H2O2
[4–8] and a number of scientific papers [9–16] have appeared
over the last decades, with Pospelova and coworkers pioneering
the field in the 1960s [9,17]. Most of these patents and papers
deal with explosive hydrogen/oxygen mixtures and very high
pressures, which are necessary to improve H2O2 selectivity.
The first industrial attempt to prepare H2O2 via direct synthe-
sis was carried out in the 1980s by Du Pont [4,5,18,19]. For the
first time, the use of such additives as halogenides (especially
bromide [5]) and organophosphates [19] was suggested to im-
prove the selectivity of carbon-supported Pd [4] and Pd/Pt cat-
alysts [18]. The synthesis was carried out under high pressure
(up to 170 bar), but this process was never scaled-up, because
of serious explosion risks.
Direct synthesis using various technologies, including mod-
ified fuel cells [20–22], new water electrolysis processes able
to simultaneously produce hydrogen and hydrogen peroxide
[23,24], and nonequilibrium plasma systems [25], has been
proposed. These technologies, although fascinating, consume
very high amounts of energy and demonstrate—at least at the
moment—low overall efficiency. Some patents [26,27] and sci-
entific papers [28,29] have also addressed the use of catalytic
membranes to avoid direct contact between H2 and O2 in the
The direct synthesis of H2O2 has been known since the be-
ginning of the 20th century [1] but has not yet found industrial
application because of severe drawbacks related mainly to the
hazardous nature of H2/O2 gas mixtures and the low H2O2
yield that can be reached with known catalysts based on pal-
ladium. But the increasing demand for hydrogen peroxide [2]
from several industrial sectors, including the pulp and paper in-
dustry and wastewater treatment, as well as the general need
for greener oxidants has rekindled interest in finding an alterna-
tive to the anthraquinone route [3], which currently is almost the
only process used worldwide, with a >90% share. This process,
based on the anthraquinone/anthrahydroquinone redox pair, has
the advantage of avoiding direct mixtures of H2 and O2. Nev-
ertheless, it produces significant amounts of organic waste due
to the overreduction of anthraquinone, it requires several sepa-
ration and concentration steps, and it is economically feasible
only in large-scale plants.
*
Corresponding author. Fax: +39 041 234 8517.
E-mail address: strukul@unive.it (G. Strukul).
0021-9517/$ – see front matter  2006 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2006.02.014

S. Melada et al. / Journal of Catalysis 239 (2006) 422–430
423
gas phase. The use of membranes is very promising, but far
from industrially applicable.
overnight. This material was impregnated by an incipient wet-
ness method with various anions, using (NH4)2SO4, NH4F,
NH4Br, and NH4Cl in amounts necessary to yield an 8 wt% an-
ion loading. Impregnated supports were then calcined in flow-
ing air (50 ml/min) at 500 ◦C (sulfate-doped zirconia at 650 ◦C)
for 3 h. Calcined supports were subsequently impregnated by
incipient wetness with a H2PdCl4 solution to give a nom-
inal 2.5 wt% Pd-loaded catalyst and finally calcined again
at 500 ◦C in flowing air for 3 h.
This work reports the preparation and use of Pd-supported
catalysts for the direct synthesis of hydrogen peroxide that can
be carried out under very mild conditions. Zirconia was cho-
sen as the support, because it has desirable properties, includ-
ing tuneable surface acidity/basicity (controlled by the addition
of different dopants), redox properties, and tuneable porosity
and surface area. In particular, sulfating zirconia [30] not only
causes modification of the acid properties, but also affects sur-
face features; sulfates retard crystallization, stabilize the tetrag-
onal phase, and improve surface area and pore size. Because
H2O2 is more stable under acidic conditions, a strongly acidic
support (doped with SO4=, Cl−, F−, or Br−) may be useful for
the direct synthesis of H2O2.
2.4. H2O2 synthesis
Catalytic tests were carried out at atmospheric pressure in
a 20 ◦C thermostatted glass reactor. Mixing was carried out
with a Teflon rotor operating at 1000 rpm. Air, hydrogen, and
nitrogen were bubbled by a gas diffuser directly into the liq-
uid phase, at a total flow of 50 ml/min. The final gas mixture
had the following composition: H2:O2:N2 = 10:10:80. (Cau-
tion: this is just outside the explosion range, but not outside the
flammable range.) Catalysts were introduced wet rather than in
the dry state. In addition, because a ring of Pd catalyst tends
to form above the reaction mixture during catalytic experi-
ments, the reactor slurry was swirled so as to remove the solid
from the reactor wall. For some experiments, a nonexplosive,
nonflammable H2:O2 4:96 mixture was used instead of the air-
hydrogen–nitrogen mixture. The reaction medium was 100 ml
of a 0.03 M H2SO4 aqueous, ethanolic, or methanolic solution
and was presaturated with the gas mixture before introduction
of the catalyst (135 mg). If not otherwise indicated, catalysts
were neither reduced nor pretreated before use.
2. Experimental
2.1. Materials
ZrOCl2 (Fluka), (NH4)2SO4 (Merck), NH4F (Fluka), NH4Br
(Fluka), and NH4Cl (Carlo Erba) were used as received for sam-
ple synthesis. All kinetic tests were performed in anhydrous
methanol (SeccoSolv, Merck, [H2O] < 0.005%) and ethanol
(SeccoSolv, Merck, [H2O] < 0.02%). Commercial standard so-
lutions of Na2S2O3 (Fixanal [0.01], Hydranal-solvent E, and
Hydranal-titrant 2E, all from Riedel–de Haen) were used for
iodometric and Karl–Fischer titrations.
2.2. Methods
During the catalytic tests, small aliquots of the liquid phase
were sampled through a septum and used for water and hydro-
gen peroxide determination. H2O2 concentration was measured
by iodometric titration, and water was determined by the vol-
umetric Karl–Fischer method. H2O2 selectivity at time t was
determined as follows:
Surface areas and pore size distributions were obtained
from N2 adsorption/desorption isotherms at −196◦C (using a
Micromeritics ASAP 2000 analyser). Calcined samples were
pretreated at 300 ◦C for 2 h under vacuum. The noncalcined
sample was pretreated at 150 ◦C for 4 h under vacuum. Sur-
face area was calculated from the N2 adsorption isotherm by
the BET equation, and pore size distribution was determined
by the BJH method also applied on the adsorption branch [31].
Total pore volume was taken at p/p0 = 0.99.
[H2O2]
SH O
=
.
2
2
[H2O2] + [H2O]
A simple model was used to fit the kinetic data and find the
apparent kinetic constants of the four reactions involved in the
catalytic process (see Scheme 1). We accounted only for H2O2
decomposition by H2 reduction and did not consider the H2O2
disproportionation reaction, because the latter is negligible un-
der our reaction conditions [47].
Thermogravimetric/differential scanning calorimetry (TG/
DSC) analyses were performed on a NETZSCH STA 409C
instrument in flowing air (40 ml/min) at a temperature rate
of 10 ◦C/min in the 25–1000 ◦C range. Temperature-pro-
grammed reduction (TPR) experiments were carried out in a
homemade apparatus. Samples (100 mg) were heated at a rate
of 10 ◦C/min from 25 to 450 ◦C in a 5% H2/Ar reducing mix-
ture (40 ml/min STP). Actual metal loading was determined
by atomic absorption spectroscopy after microwave disaggre-
gation of the samples (50 mg).
k
1
H + O −→ H O
2
2
2 2
k
2
1
2
H + O −→ H O
2
2
2
k
3
H O + H −→ 2H O
2
2
2
2
2.3. Catalyst preparation
k
4
1
2
H O −→ H O +
O
2
2
2
2
Zirconia support was prepared by precipitation from ZrOCl2
at constant pH (pH = 10), aged under reflux conditions [32,33],
washed free from chloride (AgNO3 test), and dried at 110 ◦C
Scheme 1.

424
S. Melada et al. / Journal of Catalysis 239 (2006) 422–430
Fig. 1. TG (a) and DSC (b) profiles of doped zirconia supports.
dc
3. Results and discussion
H
O
2
2
a
H
b
O
c
H
d
O
e
H
= −k c
c
c
− k c
c
c
− k c
c
H O
2
1
2
3
2
2
2
2
2
2
2
2
2
2
dt
3.1. TG/DSC analyses
dc
a
H
b
O
c
H
d
O
= −k c
− k c
1
2
dt
Initial TG/DSC analyses were carried out on uncalcined
supports to observe the material behavior during calcination.
Fig. 1 reports the TGA and DSC profiles for SO4=-, Br−-,
and F−-doped uncalcined samples. As can be seen (Fig. 1a),
all samples showed significant weight loss from room tem-
perature to about 300 ◦C due to water desorption and surface
dehydroxylation processes. At about 480–500 ◦C, ZF and ZBr
showed further weight loss, which can be reasonably attributed
to doping salt decomposition. This process shifted to higher
temperatures (600–650 ◦C) for the ZS sample. Above 600 ◦C,
the ZF and ZBr profiles showed no significant weight loss,
whereas the ZS support showed weight loss up to 1000 ◦C. In
the DSC profiles (Fig. 1b), the endothermic nature of water des-
orption (around 100 ◦C) and dehydroxylation (around 250 ◦C)
processes can be observed. The DSC profile of ZF sample
showed a sharp exothermic peak at 480 ◦C due to ammonium
salt decomposition, which can be associated with the weight
loss already observed on the TG analysis. Similarly, the DSC
spectrum of ZBr support exhibited an analogous exothermic
peak at slightly higher temperature (520 ◦C). Zirconia-phase
transition and doping salt decomposition for ZF and ZBr sam-
ples occurred at approximately the same temperature and were
associated with the exothermic peaks at 480–520 ◦C. The ZS
profile showed two distinct peaks [34–36]: an exothermic one
occurring without weight loss, which can be associated with
zirconia-phase transition (from amorphous to crystalline), and
a second endothermic peak at higher temperatures, associated
with weight loss, related to sulfate decomposition.
dc
H O
2
2
a
H
b
O
e
H
= k c
c
− k c
c
H O
2
1
3
2
2
2
2
dt
dc
H O
2
c
d
e
H
= k c
c
+ 2k c
c
H O
2
2
3
H
O
2
2
2
2
dt
Scheme 2.
In the system of differential equations coming from mass
balance (see Scheme 2), the concentrations of H2 and O2
into the liquid phase can be considered constant, because
these reagents are fed continuously into the semibatch reac-
tor. Their concentrations can be incorporated intoꢀ the kinetic
constants, kn, to give apparent kinetic constants, kn, and the re-
sulting system of differential equations can be solved for the
H2O2 and H2O concentrations to give
k1ꢀ
k3ꢀ
ꢀ
ꢁ
ꢀ
cH O − c0
=
1 − e−k t
3
2
2
H2O2
and
k1ꢀ
k3ꢀ
ꢀ
ꢁ
ꢀ
ꢁ
ꢀ
cH O − c0 = k2ꢀ + 2k1ꢀ + 2k3ꢀ c0
t − 2
1 − e−k t
.
3
2
H2O
H2O2
Catalytic tests data fitting was carried out with a computer
program OriginLab OriginPro7.5, equipped with a special fit-
ting tool (based on the Levenberg–Marquardt algorithm) to per-
form nonlinear regression analysis.
To prove this point, an undoped zirconia sample (Z) was
also prepared, but this is not reported in Fig. 1 for the sake
of simplicity. In agreement with the literature [37], this sample
showed the same weight loss of doped samples up to 400 ◦C, as-

S. Melada et al. / Journal of Catalysis 239 (2006) 422–430
425
intermediate effect; surface area was 85 m2/g, and mean pore
size was around 15 nm.
3.3. TPR analysis of fresh catalysts
After Pd deposition and calcination, TPR analysis was per-
formed to investigate the metal oxidation state. The TPR testing
setup requires that the reducing mixture pass through the sam-
ples to be analysed before starting the temperature ramp, to
clean the apparatus. This procedure can cause PdO reduction
already at room temperature. In fact PdO obtained after calci-
nation can be reduced already at −15◦C, and if the Pd metal
particles are large enough, they can lead to the formation of
Pd β-hydride [38,39]. It is well known that Pd β-hydride de-
composes at about 70 ◦C. Therefore, the calcined samples that
contain reducible PdO should show a negative peak at about
70 ◦C due to Pd β-hydride decomposition. Fig. 3a shows the
TPR profiles for all freshly calcined samples. In agreement with
the above, ZF–PdO and ZS–PdO samples showed a negative Pd
β-hydride decomposition peak at about 70 ◦C. The other sam-
ples did not show room temperature-reducible Pd(II) species;
they all possessed only PdO reducible at about 90–100 ◦C. This
species is present also in ZF–PdO and ZS–PdO, along with an-
other Pd(II) fraction that reduced at about 280–300 ◦C. This
may be residual palladium(II) oxychloride, formed during cal-
cination and not decomposed during thermal treatment.
The brownish color of all of the samples after calcination
indicated that Pd was oxidized. After TPR analysis, all samples
were completely black. Notably, the high temperature region of
ZS–PdO TPR profile showed massive hydrogen consumption,
which can reasonably be attributed to sulfate reduction to SO2
and H2S [33].
Fig. 2. N physisorption isotherms of calcined samples and (insert) BHJ pore
2
size distributions of calcined samples: (P) ZS, (1) ZBr, (2) ZCl, (+) ZF, (")Z.
Table 1
Surface features of calcined samples and Pd loading on final catalysts
Sample
SSA
(m /g)
Pore diameter
(nm)
Pore volume
(cm /g)
Pd loading
(wt%)
2
3
ZS
ZF
ZBr
ZCl
134
50
135
85
10.2
29.0
9.9
0.341
0.364
0.336
0.349
2.64
2.45
2.65
2.82
16.4
Z
61
21.3
0.327
2.91
sociated with water desorption and dehydroxylation. At higher
temperatures, no further weight loss was observed. Only a sharp
exothermic peak at about 460 ◦C was recorded, associated with
crystallization of the amorphous oxide.
3.4. Catalytic data
All catalytic tests were carried out under very mild condi-
tions (1 bar and 20 ◦C) and with dilute H2/O2 mixtures, well
away from the explosive regime. Preliminary tests of H2O2
decomposition were carried out on pure calcined supports un-
der the same experimental conditions, to verify that they did
not decompose H2O2 over the 2-h test. In fact, as reported in
Scheme 1, numerous reactions are involved in the direct syn-
thesis of hydrogen peroxide (all catalyzed by Pd), and the major
problem of this process is the low selectivity.
3.2. N2 physisorption analyses
The N2 physisorption isotherms for the calcined samples
are reported in Fig. 2. As can be seen, all samples showed
type IV isotherms with hysteresis loops typical of mesoporous
materials. A noncalcined, nondoped zirconia precursor (not
shown) was mesoporous (BJH mean pore size 5.5 nm) but
showed a high fraction of microporosity and a high surface area
(424 m2/g). This fragile amorphous structure almost collapsed
after calcination to produce a mesoporous to macroporous ma-
terial, with surface area of 50–80 m2/g. All N2 physisorption
data are reported in Table 1, along with Pd loadings deter-
mined on the final catalysts. Treatment with fluoride anions
produced samples very similar to the nondoped sample; fluo-
ride did not prevent structure collapse and pore wall thickening.
The resulting materials had low surface areas (50–60 m2/g)
and a macroporous structure (pore size >20 nm). Sulfate and
bromide doping yielded materials retaining high surface area
(130–140 m2/g) and a mesoporous structure (mean pore size
around 10 nm). These ions strongly bind the material, avoiding,
or at least retarding, structure collapse. Chloride ions had an
The effect of different solvents was studied. The samples
were first tested in aqueous solution; the results are reported
in Fig. 4. As reported previously [29], in this solvent, estimat-
ing the amount of produced water, and hence the selectivity, is
impossible. During these tests, all catalysts turned from dark
brown to black in less than 15 min, with only minor differences
in activity observed among the samples. H2O2 concentrations
of 3.2–4.4 mM were produced within 2 h. Catalyst produc-
tivities, expressed as mmolH O /(gPd h), ranged from 42 for
2
2
ZCl–PdO to 66 for ZF–PdO after 120 min time on stream. Re-
activity tests in water showed that our samples have relatively
good catalytic activity toward H2O2 direct synthesis. Neverthe-
less, the production of H2O2 in solvents other than water is

426
S. Melada et al. / Journal of Catalysis 239 (2006) 422–430
Fig. 3. TPR profiles of fresh calcined catalysts (a) and exhausted catalysts (b).
Fig. 4. Catalytic activity during kinetics tests in aqueous solutions: (P)
ZS–PdO, (1) ZBr–PdO, (2) ZCl–PdO, (+) ZF–PdO, (") Z–PdO.
desirable for several reasons: (i) most oxidation reactions in-
volving H2O2 are carried out in organic solvents, like alcohols;
(ii) hydrogen and oxygen solubilities in organic compounds are
usually higher than those in water; (iii) dry organic media allow
measurement of the water produced, thus giving an estimate of
catalyst selectivity. For these reasons, the next tests were carried
out in alcohols, such as ethanol and methanol, which several
patents [4,40,41] and scientific papers [10,13] have reported to
be excellent solvents for H2O2 synthesis. Recently, for exam-
ple, Lunsford et al. reported that ethanol is superior to water
as the medium for direct formation of H2O2 from H2 and O2
over a palladium catalyst [42]. Under our experimental condi-
tions, we found a similar trend, as shown in Fig. 5a for ethanol
solutions. It can be observed that productivity is, as expected,
higher than in water, ranging from 59 to 155 mmolH O /(gPd h)
Fig. 5. Catalytic activity (a) and selectivity (b) towards H O during kinetics
2
2
tests in ethanol: (P) ZS–PdO, (1) ZBr–PdO, (2) ZCl–PdO, (+) ZF–PdO, (")
Z–PdO.
2
2
after 120 min of time on stream. The nonpromoted Z–PdO sam-
ple showed the lowest H2O2 production; ZS–PdO demonstrated
the maximum H2O2 concentration, about 12 mM. The other
samples behaved similarly but with lower productivity, yield-
ing about 9 mM within 2 h. The profiles of H2O formation
shown in Fig. 5a allow calculation of selectivity trends as well
(Fig. 5b). The best initial selectivity was shown by the bromide-
and fluoride-promoted samples, with the latter preserving 9%
selectivity after 2 h. The nondoped and Cl− doped samples
showed the worst selectivity trends. During tests in ethanol, all
samples changed color from dark brown to black within a few

S. Melada et al. / Journal of Catalysis 239 (2006) 422–430
427
Table 2
Kinetic constants of tests in methanol using the hydrogen/air/nitrogen gas mix-
ture
ꢀ
4
ꢀ
4
ꢀ
2
Sample
k × 10
k × 10
k × 10
1
2
3
−1
(mol/(lmin))
(mol/(lmin))
(min
)
ZS–PdO
ZF–PdO
ZBr–PdO
ZCl–PdO
Z–PdO
5.3
5.6
3.7
7.0
5.5
8.7
7.8
8.9
10.8
9.4
1.1
1.3
1.0
2.6
1.2
3.5. Kinetic analysis
Fitting the catalytic curves shown in Fig. 6 according to
the procedure specified in the Experimental section yielded the
parameters reported in Table 2. Note than ZCl–PdO had the
highest kꢀ values. This means that it was the most active sam-
ple, but its activity toward H2O producing reactions (k2ꢀ and k3ꢀ )
was higher than that toward H2O2 synthesis (k1ꢀ ). ZBr–PdO, as
expected, showed the lowest kꢀꢀ values. Finally, ZS–PdO, ZF–
PdO, and ZPdO had similar k values. The k3ꢀ decomposition
constants were about the same for the SO42−-, F−-, and Br−-
doped samples and the nondoped samples, meaning that all of
these samples behaved similarly toward H2O2 decomposition,
in agreement with the observation of Lunsford et al. [43] that
decomposition rate is affected mainly by acidity and to a lesser
extent by the doping anion. Instead, ZCl–PdO had a k3ꢀ more
than double that of the other samples, accounting for the high
decomposition rate and a high H2O formation observed with
this catalyst.
Bromide ion was claimed to block very active water-
producing sites [5,18,44,45], thereby improving the selectivity
toward H2O2. However, excess bromide was detrimental to
catalytic activity [46,47]. In the present case, the starting 8%
bromide deposited on zirconia was probably high, because dur-
ing activity tests some bromide ions might be leached into
solution and readsorbed on the Pd particle surface, extensively
poisoning it.
Fig. 6. Catalytic activity (a) and selectivity (b) towards H O during kinetics
2
2
tests in methanol: (P) ZS–PdO, (1) ZBr–PdO, (2) ZCl–PdO, (+) ZF–PdO,
(") Z–PdO.
minutes. It is interesting to note that ZS–PdO and, to a lesser ex-
tent, ZF–PdO, showed a decreasing H2O2 concentration curve
after 100 min. This means that beyond that point, H2O2 decom-
position begins to prevail over H2O2 formation.
All calcined samples were also tested in methanol; the re-
sults are reported in Fig. 6. The following observations can be
drawn from Fig. 6a: (i) ZS–PdO, ZF–PdO, and nondoped ZPdO
have about the same productivity and selectivity, showing the
best catalytic results; (ii) the maximum H2O2 concentration ob-
tained was about 36 mM, which corresponds to a productivity
of 520 mmolH O /(gPd h); (iii) ZCl–PdO was quite active at the
3.6. Used samples
TPR analysis of the used samples, shown in Fig. 3b, can pro-
vide insight into some of the catalytic results reported above.
First, it is evident that Pd in all samples is completely reduced
during catalytic tests. This is demonstrated by the presence
of the Pd β-hydride decomposition peak and the absence of
any reduction peak in the same region. Only the ZCl–PdO
TPR profile showed a significant amount of Pd(II) species, still
nonreducible under the experimental conditions used. Interest-
ingly, most samples showed a reduction peak above 350 ◦C,
which can be associated with sulfate decomposition. This can
explain the similar behavior of ZS–PdO and, quite unexpect-
edly, Z–PdO. Sulfate ions, coming from sulphuric acid added to
acidify the solution in catalytic tests, can adsorb on the unpro-
moted zirconia surface, changing it to an actual sulfated zirco-
nia sample. In fact, it has been reported previously [45,48,49]
that unpromoted zirconia samples have a great ability to de-
2
2
beginning, but after 1 h time on stream, its productivity dropped
significantly (with the H2O2 decomposition rate predominat-
ing); and (iv) ZBr–PdO showed the worst performance. Fig. 6b
reports the selectivities for these reactions. The SO42−- and F−-
promoted and nonpromoted samples behaved very similarly.
The highest initial selectivity (32%) was observed for the ZF–
PdO sample; ZS–PdO and Z–PdO started from about the same
value (29%). The final selectivity for these samples was about
the same, ranging around 20–21%. ZCl–PdO, despite its high
initial productivity, showed low selectivity (from 21 to 13%)
over the test duration. ZBr–PdO selectivity ranged from an ini-
tial value of 21% to a final value of 18%.

428
S. Melada et al. / Journal of Catalysis 239 (2006) 422–430
Fig. 7. Pd particle reduction during catalytic tests: possible structural changes during catalytic tests (a), Pd particle changes during catalyst pre-treatment (b).
compose H2O2. Sulfate adsorption on the surface, as in our
case, could block those coordinatively unsaturated Zr4+ sites
able to decompose H2O2 molecules. Moreover, the ZCl–PdO
and ZBr–PdO samples adsorbed significant amounts of sulfate,
which could indicate that halogenides were possibly released
in solution while sulfate ions were adsorbed onto the zirco-
nia surface. Therefore, bromide ion could adsorb onto the Pd
particle surface, poisoning it. Because the amount of Br− car-
ried by the catalyst was quite high (8 wt%), the extent of active
site poisoning likely was beyond the beneficial amount. Chlo-
ride ions can dissolve in methanol and readsorb onto the Pd
surface, but in this case the effect is more complicated. Choud-
hary [45] reported, in agreement with our data, that chlorination
increased the H2O2 decomposition rate by about one order of
magnitude with respect to that of brominated Pd/zirconia cata-
lyst. However, in the same paper, Choudhary also reported that
fluorinated Pd/zirconia was completely inactive toward H2O2
synthesis (producing only water), contrary to our results.
To verify whether our Pd-based catalysts were stable and
reusable, we carried out some catalytic tests with the ZS–PdO
sample already used in a reaction, after filtration and drying.
We found that both productivity and selectivity remained un-
changed after the third reaction cycle. This result not only
demonstrates the excellent reusability of our samples, but also
confirms the absence of Pd leaching during kinetic tests.
alysts became less selective as the reaction proceeded due to a
change in the Pd oxidation state.
A catalytic test on reduced ZS–Pd catalyst showed slightly
higher productivity with respect to fresh, nonreduced ZS–PdO
catalyst, but with more rapidly decreasing selectivity (Fig. 8).
These differences could be explained by the β-hydride phase
formation observed in Fig. 3b. During Pd reduction, a structural
rearrangement may occur (Fig. 7a), leading to the formation
of surface defects (more active sites), which could easily ac-
count for the improved hydrogenation performance as well as
the lower selectivity, as suggested previously [29].
We also tried a preactivation process consisting of particle
surface oxidation (Fig. 7b). ZS–PdO catalyst was reduced in
situ by passing a pure hydrogen flow into the reaction medium
(after the usual deaeration step). This led to a reduced catalyst
(with TPR measurements confirming complete reduction of the
catalyst). Pure oxygen was fed. Excess oxygen was removed by
passing pure nitrogen, after which the nonexplosive H2/air/N2
gas mixture was fed into the reactor and H2O2 synthesis started.
Fig. 8 shows the results and compares them with those of two
other tests (with fresh and reduced catalysts). Interestingly, sur-
face oxidation induced very high catalytic activity. In addition,
the selectivity was the highest (from 35 to 28%). The maxi-
mum H2O2 produced was 59 mM (0.2 wt%) with a productivity
>800 mmolH O /(gPd h). Table 3 provides results of the kinetic
2
2
analysis ꢀof the three catalytic tests. The reduced particles had a
higher k1 (almost double that of fresh catalyst), but also the
highest k2ꢀ and k3ꢀ . The reduced and surface-oxidized catalysts
showed the highest k1ꢀ and the lowest k2ꢀ , at least comparable to
that of a fresh catalyst, as was k3ꢀ . ZF–PdO and ZBr–PdO were
similarly preoxidized; Table 3 gives the corresponding kinetic
constants. Analysis of Table 3 and Fig. 8 leads to the following
observations: (i) surface-oxidized ZS–Pd was the best sample,
giving, as stated above, a 0.2 wt% H2O2 solution; (ii) both
surface-oxidized ZF–Pd and ZBr–Pd showed a significant im-
provement in catalytic performance with respect to fresh sam-
ples; (iii) selectivity always dropped—with different slopes—to
low values, indicating that the surface oxide layer is not stable
under our reaction conditions; and (iv) ꢀthe fluorinated sample
had the highest k1ꢀ but also the highest k3.
3.7. Effect of the Pd oxidation state
The effect of Pd oxidation state on catalytic activity has been
the matter of some debate in the scientific literature. Choudhary
[49,50] claimed that PdO is the best catalyst for the direct syn-
thesis of H2O2, whereas other authors, including Lunsford [43],
Burch [10], and Thompson [13], reported higher conversions
and higher selectivity toward H2O2 when Pd was in its reduced
state.
Our samples were reddish-brown after calcination; these cat-
alysts are labelled ZX–PdO. During catalytic tests, the color
turned from reddish-brown to dark gray or black. TPR experi-
ments confirmed that the Pd reduction occurred during catalysis
(see above). These observations are summarized in Fig. 7a.
The reduction of the metal particles was clearly linked to the
observed change in catalytic behavior shown in Fig. 6; the cat-
We attempted to stabilize this surface-oxidized Pd particle
state by feeding an undiluted hydrogen/oxygen mixture with

S. Melada et al. / Journal of Catalysis 239 (2006) 422–430
429
Table 3
Kinetic constants for ZS–PdO pretreated in different conditions and tested under different gas feed
ꢀ
4
ꢀ
4
ꢀ
2
Sample
Gas feed
k × 10
k × 10
k × 10
1
2
3
−1
(mol/(l min))
(mol/(l min))
(min
)
ZS–PdO fresh
ZS–Pd reduced
ZS–PdO surface oxidized
ZS–Pd
ZF–PdO fresh
ZF–PdO surface oxidized
ZBr–PdO fresh
ZBr–PdO surface oxidized
H /air/N
5.3
9.0
9.9
2.7
5.6
14.2
3.7
9.2
8.7
10.4
7.9
3.5
7.8
8.3
8.9
9.9
1.1
2.4
1.5
0.051
1.3
3.5
1.0
1.7
2
2
2
2
H /air/N
2
H /air/N
2
4%H /96%O
2
2
H /air/N
2
2
2
2
2
H /air/N
2
H /air/N
2
H /air/N
2
cess of oxygen, undiluted H2/O2 mixture), catalyst selectivity
was extremely stable for several hours, meaning that the Pd
particle state did not change. It is also noteworthy that under
the reaction conditions of Fig. 8, all catalysts showed H2 con-
versions (calculated from H2O2 + H2O obtained vs. H2 fed) in
the range of 70–100%.
4. Conclusion
In this work, palladium-based catalysts supported on SO4=-,
Cl−-, F−-, and Br−-doped zirconia were successfully tested for
the direct synthesis of hydrogen peroxide under very mild con-
ditions and outside the explosion range. The 2.5% Pd loaded
on anion-doped zirconia samples showed good catalytic activ-
ity, selectivity, mechanical stability, and reusability. Investiga-
tion of the effects of water, ethanol, or methanol as a reaction
medium confirmed that best catalytic results could be achieved
in methanol, with the attainable H2O2 concentrations improved
from 3.2–4.4 mM in aqueous solution up to 36 mM under our
best experimental conditions.
The importance of surface-oxidized palladium for the reac-
tion can be deduced by TPR analysis and catalytic tests with dif-
ferent gas feeds. High H2O2 productivity (550 mmolH O /(gPd h))
2
2
and the highest selectivity (40%, still stable after 5 h) at 72%
conversion can be reached under suitable conditions. These re-
sults compare with the best findings obtained by Hutchings and
coworkers [16] using 5% Pd/TiO2 but working under pressure
(37 bar).
Fig. 8. Activity (a) and selectivity (b) in H /air/N of ZS–Pd in its different
2
2
Acknowledgments
states Pd (Q), PdO (∗), Pd–PdO surface ("), compared to catalytic results of
the same sample in 4% H and 96% O gas mixture (1).
2
2
The authors acknowledge the financial support of MIUR
(Rome). S. Melada and F. Menegazzo thank INSTM (Florence)
for postdoctoral fellowships.
oxygen in large excess (H2:O2 = 4:96). This is not only out-
side the explosion range, but also outside the flammable range,
and hence is intrinsically safer than the previous experiments.
We tested ZS–Pd, our best catalyst, under these new reaction
conditions; the results are given in Fig. 8 and Table 3. The cata-
lyst’s different behavior in the latter test is noteworthy. The pro-
ductivity was comparably lower at the beginning, but showed
very good values on long reaction times (550 mmolH O /(gPd h)
after 5 h). The most interesting result, however, is the very
low water production rate (50 times lower than in air), which
corresponds to a significantly higher selectivity toward H2O2
(around 40%). Moreover, under these test conditions (large ex-
References
[1] H. Henkel, W. Weber, US Patent 110875 (1914).
[2] Chemical Week, 2003, July 2/9, 44.
[3] H.-J. Riedl, G. Pfleiderer, US Patent 2215883 (1940).
[4] L.W. Gosser, US Patent 4681751 (1987).
[5] L.W. Gosser, J.T. Schwatrz, US Patent 4772485 (1988).
[6] J. Van Weynbergh, J.-P. Schoebrechts, US Patent 5447706 (1995).
[7] B. Bertsch-Frank, I. Hemme, S. Katusic, J. Rollmann, US Patent 6387364
(2002).
2
2

430
S. Melada et al. / Journal of Catalysis 239 (2006) 422–430
[8] G. Paparatto, R. D’Aloisio, G. De Alberti, R. Buzzoni, US Patent 6630118
[30] F. Moseley, P.N. Dyer, US Patent 4,336,240 (1982).
[31] S.J. Gregg, K.S.W. Sing, Adsorption, Surface Area and Porosity, 2nd ed.,
Academic Press, 1982, p. 111.
[32] S. Melada, M. Signoretto, F. Somma, F. Pinna, G. Cerrato, G. Meligrana,
C. Morterra, Catal. Lett. 94 (2004) 193.
[33] M. Signoretto, S. Melada, F. Pinna, S. Polizzi, G. Cerrato, C. Morterra,
Microporous Mesoporous Mater. 81 (2005) 19.
[34] B.H. Davis, R.A. Keogh, R. Srinivasan, Catal. Today 20 (1994) 219.
[35] R. Srinivasan, R.A. Keogh, D.R. Milburnand, B.H. Davis, J. Catal. 153
(1995) 123.
(2003).
[9] T.A. Pospelova, N.I. Kobozev, E.N. Eremin, Russ. J. Phys. Chem. 35
(1961) 143.
[10] R. Burch, P.R. Ellis, Appl. Catal. B 42 (2003) 203.
[11] D.P. Dissanayake, J.H. Lunsford, J. Catal. 214 (2003) 113.
[12] P. Landon, P.J. Collier, A.J. Papworth, C.J. Kiely, G.J. Hutchings, Chem.
Commun. (2002) 2058.
[13] V.V. Krishnan, A.G. Dokoutchaev, M.E. Thompson, J. Catal. 196 (2000)
366.
[14] A.G. Gaikwad, S.D. Sansare, V.R. Choudhary, J. Mol. Catal. A 181 (2002)
143.
[15] J.K. Edwards, B. Solsona, P. Landon, A.F. Carley, A. Herzing, M. Watan-
abe, C. Kiely, G. Hutchings, J. Mater. Chem. 15 (2005) 4595.
[16] J.K. Edwards, B. Solsona, P. Landon, A.F. Carley, A. Herzing, C. Kiely,
G. Hutchings, J. Catal. 236 (2005) 69.
[36] S. Ardizzone, C. Bianchi, E. Grassi, Colloids Surf. 135 (1998) 41.
[37] P.D.L. Mercera, PhD thesis, University of Twente, The Netherlands,
(1991) p. 19.
[38] W. Palczewska, in: Z. Paal, P.G. Menon (Eds.), Hydrogen Effects in Catal-
ysis, Dekker, New York, 1988, p. 373.
[39] Z. Karpinski, Adv. Catal. 37 (1990) 45.
[17] T.A. Pospelova, N.I. Kobozev, Russ. J. Phys. Chem. 35 (1961) 262.
[18] L.W. Gosser, J.-A.T. Schwartz, US Patent 4,832,938 (1989).
[19] L.W. Gosser, M.A. Paoli, US Patent 5,135,731 (1992).
[20] K. Otsuka, I. Yamanaka, Electrochim. Acta 35 (1990) 319.
[21] I. Yamanaka, T. Onizawa, S. Takenaka, K. Otsuka, Angew. Chem. Int.
Ed. 42 (2003) 3655.
[22] I. Yamanaka, T. Hashimoto, K. Otsuka, Chem. Lett. (2002) 852.
[23] Y. Ando, T. Tanaka, Int. J. Hydrogen Energy 29 (2004) 1349.
[24] M.J. Kirkpatrick, B.R. Locke, Ind. Eng. Chem. Res. 44 (2005) 4243.
[25] J. Zhou, H. Guo, X. Wang, M. Guo, J. Zhao, L. Chen, W. Gong, Chem.
Commun. (2005) 1631.
[40] G. Paparatto, R. D’Aloisio, G. De Alberti, R. Buzzoni, EP 1160196 A1
(2001).
[41] T. Haas, G. Stochniol, J. Rollmann, US Patent 6,764,671 B1 (2004).
[42] Y.-F. Han, J.H. Lunsford, Catal. Lett. 99 (2005) 13.
[43] S. Chinta, J.H. Lunsford, J. Catal. 225 (2004) 249.
[44] V.R. Choudhary, C. Samantha, A.G. Gaikwad, Chem. Commun. (2004)
2054.
[45] P. Paredes Olivera, E.M. Patrito, H. Sellers, Surf. Sci. 313 (1994)
25.
[46] S. Melada, F. Pinna, G. Strukul, S. Perathoner, G. Centi, J. Catal. 237
(2006) 213.
[26] J.A. McIntyre, US Patent 5512263 (1996).
[27] V.R. Choudhary, S.D. Sansare, A.G. Gaikwad, US Patent 6,448,199
(2002).
[47] V.R. Choudhary, C. Samanta, J. Catal. 238 (2006) 28.
[48] V.R. Choudhary, A.G. Gaikwad, S.D. Sansare, Catal. Lett. 83 (2002)
235.
[28] V.R. Choudhary, A.G. Gaikwad, S.D. Sansare, Angew. Chem. Int. Ed. 40
(2001) 1776.
[49] A.G. Gaiwad, S.D. Sansare, V.R. Choudhary, J. Mol. Catal. A 181 (2002)
143.
[29] S. Abate, G. Centi, S. Melada, S. Perathoner, F. Pinna, G. Strukul, Catal.
Today 104 (2005) 323.
[50] V.R. Choudhary, A.G. Gaikwad, S.D. Sansare, US Patent 6,534,440
(2003).