The direct formation of H2O2 from H2 and O2 over colloidal palladium

Article abstract of DOI:10.1016/S0021-9517(02)00171-9

Additional evidence is presented for the catalytic role of aqueous colloidal palladium in the direct conversion of H₂ and O₂ to H₂O₂. Reactions typically were carried out at an O ₂/H₂ ratio of 2, and 25°C, by introducing the gases at 1 atm through a frit into the aqueous slurry. The source of palladium was either PdCl₂ or reduced palladium supported on silica gel. During the course of the reaction the palladium is distributed among PdCl ₄^2-, colloidal palladium, palladium deposited on the frit, and Pd/SiO₂ when SiO₂ is present. Although the amount of colloidal palladium differs, depending on its source and the time on stream, the rate of H₂O₂ formation is proportional to the amount of colloid present at a particular time. Maximum rates were observed for colloids prepared from Pd/SiO₂; however, the largest H ₂O₂ concentration of 0.7 wt% was attained when PdCl ₂ was the source of colloidal palladium. The ultimate steady-state concentration of H₂O₂ is, in part, affected by the rate of H₂O₂ decomposition, which is relatively large over the Pd/SiO₂ sample that was tested. The rate of decomposition also is determined by the H⁺ concentration. Results obtained with a mixture of ¹⁶O₂/¹⁸O₂ confirm that oxygen remains undissociated during the formation of H₂O₂, which may explain why palladium is uniquely suited as a catalyst for this reaction.

Full text of DOI:10.1016/S0021-9517(02)00171-9

Journal of Catalysis 214 (2003) 113–120
www.elsevier.com/locate/jcat
The direct formation of H2O2 from H2 and O2 over colloidal palladium
Dhammike P. Dissanayake 1 and Jack H. Lunsford ∗
Department of Chemistry, Texas A&M University, College Station, TX 77842-3 12, USA
Received 11 June 2002; revised 16 October 2002; accepted 22 October 2002
Abstract
Additional evidence is presented for the catalytic role of aqueous colloidal palladium in the direct conversion of H and O to H O .
2
2
2 2
◦
Reactions typically were carried out at an O /H ratio of 2, and 25 C, by introducing the gases at 1 atm through a frit into the aqueous
2
2
slurry. The source of palladium was either PdCl or reduced palladium supported on silica gel. During the course of the reaction the palladium
2
2−
is distributed among PdCl
, colloidal palladium, palladium deposited on the frit, and Pd/SiO when SiO is present. Although the amount
2 2
4
of colloidal palladium differs, depending on its source and the time on stream, the rate of H O formation is proportional to the amount
2
2
of colloid present at a particular time. Maximum rates were observed for colloids prepared from Pd/SiO ; however, the largest H O
2
2
2
2
concentration of 0.7 wt% was attained when PdCl was the source of colloidal palladium. The ultimate steady-state concentration of H O
2
2
is, in part, affected by the rate of H O decomposition, which is relatively large over the Pd/SiO sample that was tested. The rate of
2
2
2
+
16
18
decomposition also is determined by the H concentration. Results obtained with a mixture of O /
O confirm that oxygen remains
2
2
undissociated during the formation of H O , which may explain why palladium is uniquely suited as a catalyst for this reaction.
2
2
2003 Elsevier Science (USA). All rights reserved.
1. Introduction
al. [7], in which they describe a novel membrane reactor,
also is of interest because such a reactor would enable one
to avoid potentially explosive mixtures of H2 and O2 while
maintaining the most efficient ratio of these reagents in the
liquid phase.
The increasing utilization of hydrogen peroxide as a
bleach in the pulp and paper industry, as an oxidant in waste-
water treatment, and as a source of oxygen for epoxida-
tion reactions has resulted in renewed interest in the direct
formation of H2O2 from H2 and O2. The current com-
mercial method for the production of H2O2 is a circuitous
process that involves the use of alkylanthroquinone and hy-
droquinone intermediates [1]. The raw product mixture con-
tains only about 2 wt% H2O2. Although recent advances
in the direct formation of H2O2 have been reported in the
patent literature [2–5], the fundamental studies that have
been carried out on this reaction are limited in both scope
and number. Most notable is the work of Thompson and co-
workers [6], who investigated the reaction over a catalyst
containing hafnium phosphate and viologen phosphonate
with supported Pd. This group reported that 13 wt% H2O2
could be attained by carrying out the reaction at atmospheric
pressure in anhydrous methanol. A paper by Choudhary et
In a recent paper, we reported on the direct formation of
H2O2 at atmospheric pressure, with the unexpected conclu-
sion that the reaction was mainly catalyzed by colloidal pal-
ladium [8]. Palladium metal is unique in its ability to cat-
alyze this reaction to any significant extent, and it is usually
introduced in a supported form. Typical supports include sil-
ica, silicic acid, carbon, and alumina. Generally, the reaction
is carried out in an aqueous slurry that also contains an acid
such as HCl or HBr. In the presence of HCl, supported Pd0
2−
initially is oxidized, with the formation of PdCl4 ions in
solution. The palladium ions then are reduced back to Pd
metal by H2, and some of the metal is present as a col-
loid. Evidence for the predominant role of the colloid in the
catalytic reaction principally came from two observations:
(1) after a period of reaction, the rate of H2O2 formation
remained essentially the same after the remaining Pd/SiO2
had been removed from the slurry, and (2) even larger rates
of reaction were observed when the source of palladium
was PdCl2; i.e., no SiO2 support was present in the system.
*
Corresponding author.
E-mail address: lunsford@mail.chem.tamu.edu (J.H. Lunsford).
On leave from the Department of Chemistry, University of Colombo,
1
Colombo 3, Sri Lanka.
0021-9517/03/$ – see front matter
doi:10.1016/S0021-9517(02)00171-9
2003 Elsevier Science (USA). All rights reserved.

114
D.P. Dissanayake, J.H. Lunsford / Journal of Catalysis 214 (2003) 113–120
Moreover, electron microscopy confirmed the presence of a
colloid in the aqueous phase.
10–15 min through the solution that was usually acidified
with HCl. The Pd/SiO2 materials consisted of 5 wt% Pd
supported on either Davison grade 57 silica or grade 03
silica. The resulting materials are referred to as Pd/SiO2(57)
and Pd/SiO2(03), respectively. The surface areas and pore
diameters of grade 57 silica are ca. 300 m2 g−1 and 15 nm,
respectively; whereas, those of the grade 03 silica are 625
m2 g−1 and 3 nm, respectively. The preparation of the
materials, which included as a final step reduction in H2,
has been described previously [9]. After addition of the Pd
to the aqueous phase, the solution or slurry was rapidly
stirred with an overhead stirrer (∼ 500 rpm) that had a glass
impeller. For longer duration runs, parafilm was loosely
placed over the top of the reactor to inhibit evaporation.
After different times on stream, small aliquots of the liquid
phase were removed from the reactor, and the H2O2 formed
was analyzed by colorimetry after complexation with a
TiOSO4/H2SO4 reagent. The estimated error in the analysis
of H2O2 is 2 × 10−4 wt% ( 5 × 10−5 M).
To determine whether the predominant role of colloidal
palladium in the formation of H2O2 is a more general phe-
nomenon, the influence of other factors on colloid concen-
trations and reaction rates has been explored. These factors
include types of silica, acid concentration and reaction time.
In addition, experiments involving 16O2/18O2 mixtures pro-
vide insight into the reaction mechanism.
2. Experimental
Most of the reactions reported here were carried out at
◦
760 Torr and 25 C in a Pyrex reactor, shown in Fig. 1,
that was 22 mm id. The oxygen and hydrogen gases were
allowed to flow through a Kimax 20F fine glass frit that
was located 100 mm below the top of the open reactor. The
flow rates, unless noted otherwise, were maintained at 20 mL
min−1 for O2 and 10 mL min−1 for H2 using Scott mass flow
controllers. Note that this is a potentially explosive mixture
and appropriate precautions should be taken, some of which
are described below.
Contact of the 2:1 O2/H2 mixture with a dry Pd/SiO2
catalyst may result in an explosion. After several hours
on stream a small amount of Pd/SiO2 is deposited on the
reactor wall above the liquid phase. This deposited material
usually has a water film, but to ensure that it did not become
dry, the stirring occasionally was stopped, the reactor was
lowered, and the slurry was swirled so as to remove the
solid from the wall. More recently we have found that the
initial introduction of Pd/SiO2 into the solution can result
in an explosion because the dry catalyst contacts the exiting
O2/H2 gas mixture. To avoid this problem, ca. 15 mL of
the aqueous solution was removed and slurried with the
Pd/SiO2. This slurry was then introduced into the remaining
aqueous phase.
Palladium was introduced as PdCl2 or via a Pd/SiO2
material to 10 or 20 mL of an aqueous phase. Prior to adding
the catalyst, the O2/H2 mixture was allowed to pass for
The selectivity for H2O2 formation was obtained by
converting the single-pass reactor to a recycle reactor which
made it possible to follow the loss of H2 and O2 during
the reaction. The gas-phase ratio, which initially was 2:1 for
O2/H2, was followed by gas chromatography. The reaction
was carried out for 1 h. From the reactions
yH2 + yO2 → yH2O2,
(1)
(2)
2xH2 + xO2 → 2xH2O,
it follows, that the H2 selectivity to H2O2 is given by
ꢀ
ꢁ
SH O = y/(y + 2x) × 100.
(3)
2
2
The values of x and y were determined from the O2/H2
ratios before and after reaction, the volume of the system,
the initial and final total pressures, and the amount of H2O2
formed.
The amount of colloid that was formed in the aqueous
phase during the reaction was determined by first analyzing
for the amount of Pd2+, using the method of Onishi [10].
Then the colloid was oxidized by exposing the solution to
air for 12 h, and the solution was again analyzed for Pd2+
.
Fig. 1. Schematic of the reactor.

D.P. Dissanayake, J.H. Lunsford / Journal of Catalysis 214 (2003) 113–120
115
The difference between the amounts of Pd2+ is taken to be
the amount of colloidal Pd. The estimated error in the colloid
concentrations is 3 × 10−6 M.
The systems prepared from 10−3 M PdCl2, 44 mg of
Pd/SiO2(57), or 44 mg of Pd/SiO2(03) all contained 2.2
mg of Pd; whereas, the system prepared for 10−4 M PdCl2
contained 0.22 mg of Pd. The liquid volume was 20 mL. It
will be shown subsequently that Pd was distributed among
the colloid, the frit, and the SiO2 when it was present. At
shorter times on stream (e.g., 2 h) the activity follows the
sequence Pd/SiO2(57) > Pd/SiO2(03) > 10−3 M PdCl2 >
10−4 M PdCl2, but as the reaction proceeded, the system
derived from 10−3 M PdCl2 maintained the largest activity.
Selectivities for H2O2 formation, based on H2 consumption,
were 61 and 66% for the Pd/SiO2(57) and 10−3 M PdCl2
systems, respectively.
The variation in activity for the Pd/SiO2(57) and
Pd/SiO2(03) systems is shown in Fig. 3. When 22 mg
Pd/SiO2(03) was introduced into the 10 mL of 1 M HCl so-
lution, there was an induction period of about 5 min, during
which time only a small amount of the colloid was formed
and, correspondingly, the rate of H2O2 formation was low.
After about 15 min, the colloid concentration and the rate
of H2O2 formation reached a maximum and thereafter both
slowly declined. By contrast, with the Pd/SiO2(57) cata-
lyst, the formation of the colloid was rapid, and no induc-
tion period was observed. But, with this system as well, the
decrease in the colloid concentration at times greater than
15 min was accompanied by a decrease in activity.
Electron micrographs of the colloidal palladium were
obtained with a Zeiss 10C TEM instrument operating at
100 kV. After 1 h on stream an aliquot of liquid was removed
from the aqueous phase and deposited on either a carbon-
coated copper or gold grid. The liquid was rapidly dried in a
vacuum oven at ca. 80 ◦C.
The oxygen isotope distribution in the hydrogen perox-
ide that forms from the reaction of H2 with an 16O2/18O2
mixture was determined by Raman spectroscopy. A Holo-
probe Raman spectrometer (Kaiser Optical) system that was
equipped with a Nd: YAG laser was used for this purpose.
Hydrogen peroxide in concentrations as small as 0.05 wt%
could be detected. The isotopically labeled H2O2 was pre-
pared by reacting 18O2 (340 Torr) and 16O2 (400 Torr)
with H2 (400 Torr) in a catalyst derived from 5 mg of
Pd/SiO2(57) in 10 mL of 1 M HCl. The reaction was car-
ried out in a closed 100-mL Pyrex round-bottom flask. The
slurry was stirred for 8 h with a magnetic stirrer, and dur-
ing this period 0.4 wt% H2O2 was produced. The H2O2 was
concentrated by pumping on the system for 0.5 h with the
solution at 30 ◦C.
3. Results
In a separate experiment, the amount of Pd present in
the colloid, on the silica, and, by difference, on the frit was
determined for a Pd/SiO2(57) system after 10 and 24 h of
reaction in 1 M HCl. Of the 2 mg Pd in the system, after
10 h there was 0.2 mg as the colloid, 0.8 mg as Pd/SiO2,
and 1 mg on the frit; whereas, after 24 h there was 0.08 mg
as the colloid, 0.3 mg as Pd/SiO2 and 1.6 mg on the frit.
The frit became dark as a result of the Pd deposit. In the
presence of the 2:1 O2 to H2 ratio, the amount of palladium
3.1. Kinetic results
The formation of H2O2 as a function of time on stream
for four different catalyst systems is depicted in Fig. 2.
Fig. 2. Comparison of H O production over palladium derived from
Fig. 3. Formation of H O and variation in colloid concentration during
2 2
2
2
−3
−4
M PdCl ; ꢂ, 5 wt%
different sources: ꢀ, 10
M PdCl ; ꢁ, 10
2
the reaction of H and O in aqueous slurries of (ꢂ, ꢀ) Pd/SiO (57) and
2
2
2
2
2
Pd/SiO (57); ꢃ, 5 wt% Pd/SiO (03).
(ꢃ, ꢄ) Pd/SiO (03).
2
2

116
D.P. Dissanayake, J.H. Lunsford / Journal of Catalysis 214 (2003) 113–120
as PdCl42− was small after an initial induction period. Most
of the Pd resides on the frit for times on stream greater than
about 10 h, but the catalytic activity of this form of Pd is
relatively small. After allowing the palladium to form on the
frit for 5 h during the reaction of H2 and O2 in a system
derived from 10−3 M PdCl2, the solution with the colloid
was replaced by a fresh 1 M HCl solution, and the rate was
determined. The rate of H2O2 formation was 1.4 × 10−5
M min−1, which may be compared to the original rate of
1.9 × 10−4 M min−1. Thus, the activity of the Pd on the
frit was more than an order of magnitude less than that of
the colloid in solution. Even this value for the Pd/frit is an
upper limit since a small amount of the Pd was reintroduced
into the solution as a colloid.
The rates of H2O2 formation for the systems derived from
the two supported catalysts and from PdCl2 have been de-
termined, and the results are shown in Fig. 4 as a func-
tion of the colloid concentration. Within experimental error,
there is a linear relationship between the rate and the colloid
concentration, which further supports the predominant cat-
alytic role of the colloidal −p1alladium. A specific activity of
80 mmol H2O2 g−1 Pd min was determined from the lin-
ear relationship of Fig. 4. The earlier catalytic results, which
were obtained with 2.2 mg Pd/SiO2(57), indicated that the
specific activity of the solid catalyst itself was only about 3%
of the specific activity of colloidal palladium [8].
activities at the lower acid concentrations were not as stable
with respect to time on stream. Even the neutral solution
was active after an induction period of several minutes.
In this case, as well as with the 0.01 M HCl solution,
there was a decrease in activity that resulted from H2O2
decompositionat the lower proton concentrations. Pospelova
and Kobozev [11] have shown that the decomposition of
H2O2 is strongly inhibited by protons in similar catalytic
systems. This interpretation for the decrease in net H2O2
formation as depicted in Fig. 5A is confirmed by the results
of Fig. 5B. After the steady state was reached, H2O2 was
added to a level of about 0.04 wt%, and it was observed
that decomposition of this added H2O2 rapidly occurred,
down to the level of the original steady-state concentration.
Subsequent acidification of the solution to 1 M HCl caused
a net formation of H2O2 at a rate comparable to that found
for the original 1 M HCl solution (Fig. 5A).
The question arises as to whether the decrease in activity
of the 0.1 M solution also resulted from the approach to
steady state (i.e., the more rapid decomposition of H2O2) or
whether colloidal Pd was less stable in the 0.1 M solution. In
a separate experiment, the reaction was carried out for 7 h,
during which time the rate had decreased significantly. Then
0.1 mg Pd was added as PdCl2 (the same amount of PdCl2 as
was originally present) and the rate increased to the original
value that is indicated in Fig. 5A. This result confirms that
in the 0.1 M solution, the decrease in net rate results mainly
from the loss of palladium rather than the decomposition of
H2O2.
The role of the HCl in the formation of H2O2 also was of
interest, and in order to avert the complications introduced
by the dissolution of palladium from Pd/SiO2, PdCl2 was
used to form the colloid. The results of Fig. 5A show that
at 0.1 and 0.01 M HCl concentrations, the initial activities
were slightly larger than that of the 1.0 M HCl solution, but
The potential influence of the decomposition reaction on
the net formation of H2O2 was determined for solutions that
−4
Fig. 5. (A). Formation of H O in aqueous colloids derived from 10
M
2
2
−1
Fig. 4. Variation in catalytic activity for H O formation as a function
PdCl at different HCl concentrations: ꢁ, 1.0 M HCl; ꢀ, 10 M HCl; ꢂ,
10 M HCl; ✥, no HCl added. (B) After 80 min H O was added to give
2
2
2
−3
−2
of colloid concentration with the colloid being derived from: ꢀ, 10
PdCl ; ꢂ, 5 wt% Pd/SiO (57); ꢃ, 5 wt% Pd/SiO (03).
M
2
2
0.04 wt%; after 100 min the solution was acidified to 1 M HCl.
2
2
2

D.P. Dissanayake, J.H. Lunsford / Journal of Catalysis 214 (2003) 113–120
117
Table 1
a
Hydrogen peroxide decomposition
a
System
Decomposition rate
−1
(M min
)
−3
200 mg 5 wt% Pd/SiO (57)
2 × 10
2
−4 b
−5
20 mg 5 wt% Pd/SiO (57)
3 × 10
2
Pd/Frit
4 × 10
−3
c
−5d
1 × 10 M PdCl
−8 × 10
2
a
Decomposition of 2 wt% H O in 1 M HCl in the presence of flowing
2
2
O
and H .
2
2
b
−4
−1
The H O formation rate over this catalyst was 1.6 ×10 M min
2
2
after 5 h.
c
2+
was not exten-
The solution remained yellow, indicating that Pd
sively reduced.
d
A net formation of H O occurred.
2
2
contained a catalyst and a significantly larger H2O2 concen-
tration of 2 wt%. The experiments were carried out in 10 mL
of 1 M HCl with palladium in one of the relevant forms, as
well as O2 and H2 in a 2:1 ratio. The results are given in
Table 1. The rate of decomposition was approximately pro-
portional to the amount of Pd/SiO2(57) catalyst. Over 20 mg
Pd/SiO2(57), the rate of H2O2 decomposition was compa-
rable to the net rate of formation that was observed when
the level of H2O2 was much less. The palladium on the frit
was about an order of magnitude less active, and the col-
loid derived from 10−3 M PdCl2 actually resulted in the net
formation of H2O2. These results, together with those that
demonstrate the loss of palladium to the frit, indicate that at
the lower concentrations of H2O2 attained in the previous
experiments (e.g., those described in Fig. 2) and in 1 M HCl
the H2O2 concentration is limited more by the stability of
the colloid than by the decomposition of H2O2, especially
for the systems that did not contain Pd/SiO2.
Fig. 6. Variation in H O formation rate as a function of (ꢁ) O or (ꢀ) H
2
2
2
2
partial pressure in a slurry derived from 2.2 mg of 5 wt% Pd/SiO (57) in
2
−1
10 mL of 1 M HCl. Rates are expressed in M min
and pressures are
in Torr. For the variation in O pressure, the smallest and largest ratios
2
were O :H :Ar = 5:10:45 and 50:10:0, respectively; for the variation in
2
2
H
pressures the smallest and largest ratios were O :H :Ar = 30:2.5:27.5
2
2
2
and 30:30:0, respectively.
corresponds to an apparent reaction order of 0.8, although
for the reasons stated above, no fundamental significance
concerning the mechanism can be attached to this value.
3.2. Electron microscopy
The effect of O2 and H2 concentration on the reaction
rate was studied in a reactor similar to that described
above except that the top was closed and the gases exited
through tubing. The glass stirrer passed through a gas
tight seal. The total flow rate was kept constant by using
argon as a diluent. Each reaction rate was determined in
a separate experiment, during which the rate was constant
for approximately 3 h after an initial transition period. The
palladium was introduced to the system as Pd/SiO2(57). The
results of Fig. 6 show that there is a positive order with
respect to both O2 and H2 up to a certain partial pressure,
but at higher pressures the order approaches zero. Here, the
concept of a reaction order needs to be understood in view of
the fact that the concentration of the colloid is also affected
by the O2 and H2 partial pressures. At the higher O2:H2
ratios, e.g., [12], the palladium was present in solution
mainly as PdCl42−. The yellow color of PdCl42− was clearly
evident. The plateau at the larger O2 partial pressures may
The electron micrographs shown in Fig. 7 confirm the
presence of colloidal palladium in the aqueous phase. For the
colloid formed from PdCl4 in the 0.01 M HCl solutions
2−
(Fig. 7A), a broad distribution in particle size was apparent,
with an average diameter of ca. 3 nm. For the colloid formed
from Pd/SiO2(57) in 0.1 M HCl, the average diameter was
ca. 6 nm (Fig. 7B). In the latter case, the particles are
more clustered. These results may be compared with those
reported by Henglein [12], who observed the formation of
2−
Pd colloids following the reduction of an aqueous PdCl4
solution with H2. The colloid was stabilized by sodium
citrate, and a narrow size distribution centered at about 4 nm
was observed.
3.3. Oxygen isotope results
Experiments using a mixture of 16O2 and 18O2 were
carried out to determine whether a dissociated form of
oxygen is involved in the formation of H2O2. If, indeed,
a dissociated form of oxygen were involved or if oxygen
dissociated and rapidly recombined to produce a diatomic
species which was then hydrogenated, one would expect to
observe a scrambling of oxygen isotopes, and H216O18O
2−
reflect the steady state between colloidal Pd and PdCl4
;
namely, most of the palladium is in the form of PdCl42−. The
plateau at the larger H2 partial pressures is more difficult to
interpret, although it may result from a saturation coverage
of hydrogen on the surface. The linear region of both curves

118
D.P. Dissanayake, J.H. Lunsford / Journal of Catalysis 214 (2003) 113–120
16
18
Fig. 8. Raman spectra of H O after reacting H with (A) an
mixture and with (B) O .
2
O
O
2
2
2
2
2
16
4. Discussion
4.1. Additional evidence for the role of colloidal palladium
As noted in the introduction, previous results indicated
that colloidal palladium may be the predominant active
phase during the catalytic conversion of H2 and O2 to
H2O2 [8]. Changes in the rate of H2O2 formation, deduced
from the slopes of the H2O concentration curves in Fig. 3,
support this conclusion since they correlate well with the
temporal changes in the colloid concentration. This is
particularly evident for the Pd/SiO2(03) catalyst, for which
there is a distinct induction period. Moreover, as shown
in Fig. 4, there is a linear relationship between the rate
of H2O2 formation and the colloid concentration for three
different catalysts, including the one derived from PdCl2.
The fact that the results for the PdCl2-derived system are
approximately on a common curve with those from the two
Pd/SiO2 sources indicates that the colloid can be formed
from all three sources, though not with equal efficiency.
The electron micrographs of Fig. 7 independently vali-
date the presence of the colloids in the aqueous phase. The
average diameters and the particle size distributions may not,
however, reflect the actual state of the aqueous phase since
the colloids were not stabilized and agglomeration of par-
ticles may occur during preparation of the samples. Fortu-
itously, the average particle diameter of 6 nm for the colloid
derived from Pd/SiO2(57) is the same as that reported pre-
viously for the supported palladium [9]. The agreement is
unexpected because the Pd first enters the aqueous phase as
PdCl42− and then is reduced to the colloidal palladium.
The palladium is believed to be distributed between
its several states according to Scheme 1. Even in a 2:1
oxygen-to-hydrogen gas mixture, palladium on the frit is
Fig. 7. Electron micrographs of palladium colloids obtained (A) after
−3
2−
exposure of a 10
M PdCl
solution, 0.01 M in HCl, to O /H for
4
2
2
◦
1 h at 25 C or (B) from the aqueous phase over 22 mg Pd/SiO (57) in
10 mL 0.1 M HCl, after exposure to O /H for 1 h at 25 C.
2
◦
2
2
would be a major product. Raman spectroscopy was used
to determine the isotopic distribution in the product, and
the results of Fig. 8 show that only H216O2 with a peak at
879 cm−1 and H218O with a peak at 830 cm−1 were present.
The Raman shift of 49 cm−1 is in good agreement with the
calculated value based on isotopic masses. No significant
peak was detected at about 852 cm−1, which is the position
expected for H2 16O18O. Clearly, H2O2 is derived from a
diatomic form of oxygen that presumably is adsorbed on the
palladium.
the dominant final state, but in the presence of pure O2,
2−
almost all of the palladium reenters the solution as PdCl4
.
Variations in the state of palladium, according to Scheme 1,

D.P. Dissanayake, J.H. Lunsford / Journal of Catalysis 214 (2003) 113–120
119
of H2O2 would not have been significant. At longer times
with larger concentrations of H2O2, the selectivities would
be less.
Choudhary and co-workers [13] have recently shown that
the support strongly influences both the rate of formation and
the rate of decomposition of H2O2, with PdO/fluorinated-
Al2O3 being among the best of the catalysts that they
studied. Fluorination of the alumina significantly decreased
the rate of H2O2 decomposition.
Clearly, the “support” material can play an important
role in the steady state concentration of H2O2, which
makes PdCl2 an attractive source of palladium for colloid
formation. The data of Fig. 2 show that after 15 h the largest
concentration of H2O2 was attained when the palladium was
introduced as PdCl2, and the results of Table 1 confirm
that the system derived from PdCl2 was the only one for
which there was a net increase in the amount of H2O2 when
the original solution contained 2 wt% H2O2. Most supports
are considered to be inert with respect to H2O2 formation;
however, Park et al. [14] reported that an H-beta zeolite had a
small amount of activity, but poor selectivity, for generating
H2O2.
Scheme 1.
are mainly responsible for the induction period (Fig. 3)
and the different rates of reaction as the O2 and H2
partial pressures are changed (Fig. 6). The induction period
depicted in Fig. 3 is more pronounced for Pd/SiO2(03)
than for Pd/SiO2(57) because the pore size of the former
2−
is much smaller. Consequently, the generation of PdCl4
and its transport to the bulk aqueous phase are inhibited.
For the practical utilization of this process, one of the
remaining challenges is to stabilize the colloidal palladium
so that it is not transformed into Pd/frit. Attempts to increase
the lifetime and/or the catalytic activity of the Pd system
by adding reagents that are known to stabilize colloids
were unsuccessful. These reagents included sodium citrate,
poly(vinyl alcohol) and poly(ethylene glycol).
4.3. Effect of pressure
Reactions were not carried out in this study at total
pressures greater than 760 Torr; although, Izumi et al. [2]
have reported in a patent the weight percent of H2O2 and
the selectivity at the end of 20 h for pressure up to 15.2 ×
103 Torr. Their catalyst was Pd/hydrous silicic acid and
the solution was 0.03 N in HCl and 1 N in H2SO4. The
O2/H2 ratio was 2.5. At 760 Torr they observed 0.78 wt%
H2O2, which is comparable to the amount shown in Fig. 2,
but the selectivity was only 30%. The amount of H2O2
increases almost linearly with pressure up to 7.6 × 103 Torr,
while the selectivity increased to a value of 79%. At 15.2 ×
103 Torr the wt% H2O2 and selectivity were 12.45 and 88%,
respectively. These results demonstrate that a significant
advantage in both H2O2 formation and selectivity can be
gained by operating at higher pressures
Although it is not explicitly indicated in Scheme 1, HCl
plays an important role in the formation of PdCl4 from
2−
the reduced forms of Pd. The previous study demonstrated
that > 0.01 M HCl was needed to form the colloid from a
Pd/SiO2(57) catalyst. The importance of protons in inhibit-
ing the decomposition of H2O2 has long been known [11],
and the results of Fig. 5B confirm this observation. In this
case, PdCl4 was used to produce the active phase; there-
fore, HCl was not required to generate PdCl4 as an inter-
mediate in the formation of colloidal Pd.
2−
2−
4.2. Decomposition of H2O2
At the largest acid concentrations used in this study (1 M),
the decomposition of H2O2 probably does not become
significant at H2O2 concentrations less than about 0.5 wt%,
but at larger concentrations the decomposition reaction, as
well as the loss of colloid, may cause the nonlinear behavior
shown in Fig. 2. At considerably larger concentrations, the
decomposition reaction, which is first order [11], becomes a
limiting factor, especially with Pd/SiO2(57) as the catalyst
(Table 1). Thus, the ultimate H2O2 concentration that can
be achieved depends on several factors including the source
of Pd, the concentration and stability of the colloid, and
the HCl concentration. Selectivities > 60% were obtained
after 1 h for the conditions of Fig. 2; hence, the H2O2
concentration was about 0.1 wt% and the decomposition
4.4. Insight into the reaction mechanism
Although important mechanistic questions remain unre-
solved, the Raman results shown in Fig. 8 provide strong ev-
idence that oxygen remains in a diatomic form during hydro-
genation. This is consistent with the mechanism suggested
by Pospelova and Kobozev [11,15], who proposed that O2
is adsorbed in the molecular form on Pd and reacts with ad-
sorbed H to yield HO2 as a surface intermediate. By con-
trast, on Pt, which is a catalyst for H2O production, O2 is
believed to adsorb in a dissociative form. The behavior of
O2 on Pd dispersed in an aqueous phase is different from
O2 on a Pd (111) surface under UHV conditions for which

120
D.P. Dissanayake, J.H. Lunsford / Journal of Catalysis 214 (2003) 113–120
Yates and co-workers [16] observed dissociative adsorption
at about 200 K. As expected, dissociative adsorption was
accompanied by extensive isotopic mixing when 16O2 and
18O2 were employed, but no 16O 18O cross-products were
detected when only molecular adsorption occurred. The ab-
sence of isotopic mixing is therefore a valid test for exclu-
sively molecular adsorption.
croscopy was carried out at the Electron Microscopy Center,
Texas A&M University.
References
[1] W.T. Hess, in: J.I. Kroschwitz, M. Howe-Grant (Eds.), Kirk-Othmer
Encyclopedia of Chemical Technology, Vol. 13, 4th. ed., Wiley, New
York, 1995, 961.
5. Conclusions
[2] Y. Izumi, H. Miyazaki, S. Kawahara (Tokuyama Soda Kabushiki
Kaisha), US Patent 4,009,252, 1977.
1. Under the conditions of these experiments, colloidal
Pd is the principal catalytic component for the direct
conversion of H2 and O2 to H2O2.
2. In the presence of H2, O2, and HCl, the palladium is
present in the colloidal form, as PdCl42−, as the metal
supported on the frit, and as Pd/SiO2 if silica is present.
The amount in each form varies with the time on stream
and the O2/H2 ratio.
3. Hydrochloric acid, together with O2, promotes the
dissolution of palladium metal, and protons inhibit the
decomposition of H2O2.
4. Oxygen remains in the diatomic form during the pro-
duction of H2O2 on palladium.
[3] W.F. Brill (Halcon), US Patent 4,661,339, 1987.
[4] L.W. Gosser, J.-A.T. Schwartz (DuPont), US Patent 4,832,938, 1989.
[5] L.W. Gosser (DuPont), US Patent 5,135,731, 1992.
[6] V.V. Krishnan, A.G. Dokoutchaev, M.E. Thompson, J. Catal. 196
(2000) 366.
[7] V.R. Choudhary, A. Gaikwad, S.D. Sansare, Angew. Chem. Int. Ed. 40
(2001) 1776.
[8] D. Dissanayake, J.H. Lunsford, J. Catal. 206 (2002) 173.
[9] F. Fajula, R.G. Anthony, J.H. Lunsford, J. Catal. 73 (1982) 237.
[10] H. Onishi, Photometric Determination of Traces of Metals, Wiley,
New York, 1989.
[11] T.A. Pospelova, N.I. Kobozev, Russs. J. Phys. Chem. (Trans.) 35
(1961) 584.
[12] A. Henglein, J. Phys. Chem. 104 (2000) 6683.
[13] A.G. Gaikwad, S.D. Sansare, V.R. Choudhary, J. Mol. Catal. A Chem.
181 (2002) 143.
[14] S.-E. Park, H. Lin, C.W. Lee, J.-S. Chang, Catal. Today 61 (2002) 117.
[15] T.A. Pospelova, N.I. Kobozev, Russ. J. Phys. Chem. (Trans.) 35 (1961)
262.
Acknowledgments
This research was supported by the Office of Basic En-
ergy Sciences, US Department of Energy. Electron mi-
[16] X. Guo, A. Hoffman, J.T. Yates, J. Chem. Phys. 90 (1989) 5787.