Suppression of the water splitting back reaction on GaN:ZnO photocatalysts loaded with core/shell cocatalysts, investigated using a μ-reactor

Article abstract of DOI:10.1016/j.jcat.2012.03.021

Using silicon-based μ-reactors, we have studied the photocatalytic water splitting reaction and the catalytic back reaction on the same catalysts. GaN:ZnO without cocatalyst and loaded with Rh, Pt, Cr₂O ₃/Rh, Cr₂O₃/Pt, and Rh-Cr mixed oxide has been tested for gas-phase photocatalytic water splitting. The results confirm the high activity observed in liquid-phase experiments with Cr₂O ₃/Rh and Rh-Cr mixed oxide as cocatalysts. To investigate the reason of this enhanced activity, the back reaction was studied by reacting stoichiometric H₂/O₂ and monitoring the water molecules produced. The comparison of the two experiments shows that the suppression of the back reaction with the core/shell cocatalysts and the Rh-Cr mixed oxide corresponds to an increase in the net photocatalytic water splitting activity. The fact that the back reaction is not completely suppressed with Cr ₂O₃/Pt compared to Cr₂O₃/Rh may be the cause of the higher net activity of the Cr₂O₃/Rh.

Full text of DOI:10.1016/j.jcat.2012.03.021

Journal of Catalysis 292 (2012) 26–31
Contents lists available at SciVerse ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Suppression of the water splitting back reaction on GaN:ZnO photocatalysts
loaded with core/shell cocatalysts, investigated using a l-reactor
Fabio Dionigi a, Peter C.K. Vesborg a, Thomas Pedersen b, Ole Hansen a,b, Søren Dahl a, Anke Xiong c,
Kazuhiko Maeda c,d, Kazunari Domen c, Ib Chorkendorff a,
⇑
a CINF, Department of Physics, Building 312, Fysikvej, Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark
b Department of Micro- and Nanotechnology, Nanotech, Building 345 East, Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark
c Department of Chemical System Engineering, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan
d Precursory Research for Embryonic Science and Technology (PRESTO), Japan Science and Technology Agency (JST), 4-1-8 Honcho Kawaguchi, Saitama 332-0012, Japan
a r t i c l e i n f o
a b s t r a c t
Article history:
Using silicon-based l-reactors, we have studied the photocatalytic water splitting reaction and the cat-
Received 31 January 2012
Revised 8 March 2012
Accepted 23 March 2012
Available online 3 May 2012
alytic back reaction on the same catalysts. GaN:ZnO without cocatalyst and loaded with Rh, Pt, Cr2O3/Rh,
Cr2O3/Pt, and Rh–Cr mixed oxide has been tested for gas-phase photocatalytic water splitting. The results
confirm the high activity observed in liquid-phase experiments with Cr2O3/Rh and Rh–Cr mixed oxide as
cocatalysts. To investigate the reason of this enhanced activity, the back reaction was studied by reacting
stoichiometric H2/O2 and monitoring the water molecules produced. The comparison of the two experi-
ments shows that the suppression of the back reaction with the core/shell cocatalysts and the Rh–Cr
mixed oxide corresponds to an increase in the net photocatalytic water splitting activity. The fact that
the back reaction is not completely suppressed with Cr2O3/Pt compared to Cr2O3/Rh may be the cause
of the higher net activity of the Cr2O3/Rh.
Keywords:
Photocatalysis
Hydrogen
Core/shell
Water splitting
GaN:ZnO
Ó 2012 Elsevier Inc. All rights reserved.
Hydrogen oxidation
1. Introduction
reaction) can be suppressed by cocatalyst modification. The photo-
catalyst that we have used for this purpose is GaN:ZnO loaded with
Photocatalytic water decomposition into hydrogen and oxygen
using solar light is a clean and renewable way to store energy from
the sun as chemical energy. Several approaches have been studied
in order to find a catalytic system able to perform this highly desir-
able reaction in an efficient way [1,2]. One of the promising sys-
tems consists of a powder catalyst made of a semiconductor
photocatalyst (i.e., a metal oxide or an oxynitride) with the surface
decorated by smaller cocatalyst nanoparticles. Photocatalytic
water splitting using this kind of catalyst is illustrated in Fig. 1a.
This system has several advantages originating basically from its
simplicity: there is only one photocatalyst material, there are no
wire connections, and a powder catalyst has a high surface-area-
to-mass ratio. However, it also presents some technical challenges.
Hydrogen and oxygen are produced in the same environment, and
they can back react to form water. This is an unfortunate event be-
cause it implies a loss of products and thus a loss of energy/effi-
ciency. In order to achieve high efficiency, this back reaction
must be avoided. In this work, we show through direct experi-
ments how the water formation reaction (water splitting back
five different cocatalysts: two noble metals, Pt and Rh, their corre-
sponding core/shell modification (the noble metal particles cov-
ered by a Cr2O3 shell), and the mixed oxide Rh2ꢀyCryO3. They
have previously been synthesized and tested in liquid phase by
Maeda et al., and the same authors have shown that the chromia
shell improves the activity for photocatalytic liquid-phase water
splitting [3,4]. So far, the GaN:ZnO with a band gap of 2.6–2.7 eV
is the photocatalyst with the highest reported activity in the visible
region for overall water splitting (5.1% at k = 410 nm when loaded
with Rh2ꢀyCryO3) [4]. For this reason, we focus the study of the
chromia core/shell cocatalyst on this photocatalyst. However, the
positive effect of the chromia shell is not limited to GaN:ZnO. For
example, it has been shown recently that overall water splitting
can be observed when GaN nanowires are loaded with a Cr2O3/
Rh cocatalyst, while no measurable activity has been observed
when only Rh was loaded [5]. Electrochemical experiments using
electrodes to model the cocatalyst nanoparticles showed that the
reason for the enhanced activity when a chromia shell is used to
protect the noble metals is the suppression of the water splitting
back reaction [6]. In this work, the water splitting back reaction
rate is analyzed directly measuring the catalytic water formation
from a stoichiometric 2:1 flow of H2 and O2 and on the same sys-
tems used for photocatalytic water splitting (cocatalyst deposited
⇑
Corresponding author. Fax: +45 4593 2399.
E-mail address: ibchork@fysik.dtu.dk (Ib Chorkendorff).
URL: http://dcwww.fys.dtu.dk/~ibchork/ (Ib Chorkendorff).
0021-9517/$ - see front matter Ó 2012 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.jcat.2012.03.021

F. Dionigi et al. / Journal of Catalysis 292 (2012) 26–31
27
on photocatalyst powder). This experiment is similar to hydrogen
oxidation experiments that have been extensively studied in heter-
ogeneous catalysts on model system, that is, Pt single crystals and
Pt nanoparticles on oxide support [7,8]. Thanks to those experi-
ments, a lot of information is available on surface coverage and
kinetics for this reaction. Differing from these experiments, our
purpose is not to study the fundamental aspects of the H2/O2 reac-
tion, but to perform the H2/O2 reaction on photocatalysts at condi-
tions similar to photocatalytic water splitting conditions (presence
of stoichiometric H2/O2 mixture, room temperature, 1 bar). Finally,
we compare those results with the photocatalytic activities ob-
tained for gas-phase water splitting using the same catalysts.
Nanoparticulate Rh–Cr mixed oxide (Rh2ꢀyCryO3), a cocatalyst
assisting H2 evolution, was loaded onto the as-prepared (Ga1ꢀxZnx)
(N1ꢀxOx) catalyst according to the method described previously
[14,15]. Briefly, 0.1 g of (Ga1ꢀxZnx)(N1ꢀxOx) powder and 3–4 ml of
distilled water containing an appropriate amount of Na3RhCl6ꢃnH2O
(Rh 17.8 wt.%) and Cr(NO3)3ꢃ9H2O were placed in an evaporating
dish over a water bath. The suspension was stirred using a glass
rod to complete the evaporation. The resulting powder was col-
lected and heated in air at 623 K for 1 h to convert Rh and Cr species
to Rh2ꢀyCryO3 [15]. Rh and Cr were loaded at rates of 1 and 1.5 wt.%
(metallic content), respectively.
2.3. Gas-phase photocatalytic water splitting reaction and back
reaction in l-reactors
2. Experimental
2.1. Preparation of the GaN:ZnO photocatalyst
Both the photocatalytic water splitting reactions and the dark
back reactions (water formation) were performed using silicon-
(Ga1ꢀxZnx)(N1ꢀxOx) solid solution (x ꢁ 0.12) was prepared by
heating a mixture of b-Ga2O3 (1.08 g) and ZnO (0.94 g) powders un-
der NH3 flow (200 ml minꢀ1) at 1098 K for 13.5 h according to a
similar method we have reported previously [9,10]. The as-synthe-
sized (Ga1ꢀxZnx)(N1ꢀxOx) powder was then subjected to post-calci-
nation in a static air atmosphere at 873 K for 5 h [11]. The effects
of preparation parameters on the physicochemical properties of
GaN:ZnO have been discussed in our previous papers [10,12] and
are reported briefly in the present study as supporting information.
XRD analysis confirms that the as-prepared sample exhibits a single
hexagonal wurtzite phase (Fig. S1). The sample has a featureless
morphology (Fig. S2) with a specific surface area of 7–8 m2 gꢀ1
determined by nitrogen adsorption at 77 K. As shown in Fig. S3,
the band gap of the material is estimated to be ca. 2.7 eV. In this
manuscript, the as-prepared (Ga1ꢀxZnx)(N1ꢀxOx) is referred to as
GaN:ZnO for simplicity.
based
l
-reactor technology [16]. The
l-reactor used for this work
is a flow reactor and consists in a 350-
lm-thick silicon chip with
an area of 16 mm by 20 mm. Channels that allow the gas to flow
are etched in the silicon as well as a circular (diameter = 10 mm)
3-lm-deep reaction chamber (240 nl total volume). The catalyst
is dispersed, sonicated in water solution, dropped in the chamber,
and allowed to dry using a circular mask of 8 mm in diameter. The
amount of material deposited is ꢂ60
lg for each catalyst. After cat-
alyst deposition, the wafer is bonded to a Pyrex lid by the cold
bonding technique [17]. The Pyrex lid is basically transparent at
wavelengths k > 300 nm, and thus, it allows photocatalytic experi-
ments with solar light [18]. The chip presents two inlets and two
outlets. One of the outlets is connected to a quadrupole mass spec-
trometer (QMS) by a flow-limiting capillary (the magnitude of the
flow conductance is ꢂ5 ꢄ 1015 molecule/s at 1 bar reactor pres-
sure) and defines the flow through the reaction chamber. The
two inlet channels mix on the chip into a main channel. The major
part of the flow of the main channel goes through the second outlet
where a pressure controller sets the pressure (1 bar in the experi-
ments presented here). The water vapor used in the photocatalytic
experiments is obtained by bubbling helium carrier gas (AGA, sci-
entific quality 6.0) through pure, liquid water (Millipore,
2.2. Deposition of cocatalyst nanoparticles
The Rh- and Pt-loaded GaN:ZnO was first prepared by a photo-
deposition method from Na3RhCl6 and H2PtCl6 solution, respec-
tively, in which 1 wt.% metal was dissolved in each case. Using
the as-prepared metal-loaded samples, photodeposition of
K2CrO4 was done. Considering the self-limiting nature of Cr2O3
photodeposition, it is expected that ꢂ0.3 wt.% Cr is deposited [13].
18.2 M
X cm). This water-saturated helium flows in the first inlet
channel, while pure, dry helium flows in the second inlet channel.
Each flow is regulated by conventional flow controllers and the
humidity adjusted via control of the ratio of dry to water-saturated
helium. For the back reaction experiments, He is flowing in the first
input channel, while H2 (or O2) is flowing in channel 2. Then, O2 (or
H2) is introduced in channel 1. Both H2 and O2 are of scientific qual-
ity 6.0 (AGA). The temperature of the l-reactor is monitored by a
thermocouple and controlled by a heating band and a thermo-elec-
tric (Peltier) cooling element. Thermo-grease is used to assure good
heat conductivity between the back side of the reactor and the
cooling element. The calibration procedure to convert the raw
ion currents measured by the QMS into molecule/s is described
in a previous publication [19]. The light source used in the photo-
catalytic experiments is a high-power UV LED (Hamamatsu model
LC-L2) assembled with a focusing lens (Hamamatsu L10561-220),
able to produce an average irradiance on the sample area of
ꢂ460 mW/cm2, as measured using a photodiode (Thorlabs model
S120VC). The peak wavelength is k ꢂ 367 nm, and the FWHM is
ꢂ9 nm. This light source has been used instead of visible light in
order to have a better signal-to-noise ratio, even if all the cocata-
lyst-loaded samples were tested to be active with a 1-kW Xe-arc
source with a longpass 420-nm cutoff filter. The LED was used at
20% (92 mW/cm2) of its full power for all the photocatalytic water
splitting experiments.
Fig. 1. Photocatalytic water splitting and catalytic formation of water on
a
photocatalyst (yellow) loaded with a cocatalyst (green). (For the interpretation of
the references to color in this figure legend, the reader is referred to the web version
of this article.)

28
F. Dionigi et al. / Journal of Catalysis 292 (2012) 26–31
from which the catalyst is drop-casted on the lid and errors in
the estimation of the relative humidity.
3.2. Water splitting back reaction
Water formation from the catalytic reaction of H2 and O2 has
been measured for all the samples. This experiment consists in
measuring the amount of water evolved as detected by a QMS.
Since this reaction evolves high amount of energy, it can be dan-
gerous to perform in a macroscopic reactor. In our case, however,
the reactor chamber is only 240 nl in volume, essentially eliminat-
ing the safety issues. We have performed the experiments at room
temperature (25 °C, 298 K), diluting the H2 (2 ml/min)/O2 (1 ml/
min) flow mixture with 15 ml/min of He and at 1 bar of total pres-
sure. Fig. 3a and b presents typical examples of a back reaction
experiment. In Fig. 3a, the raw data of an experiment with the
reactor loaded with Pt/GaN:ZnO are shown. At the beginning of
the experiment, only O2 and He are flowing through the reactor.
A local minimum in the O2 occurs instantly after the H2 flow in
the He channel is started (ꢂ90 min after we have mounted the
reactor into the gas manifold). At the same time, a peak in the
He is measured. These features are due to the sudden increase in
the flow of channel 1 (passing from 15 ml/min of He to 17 ml/
min (15 of He and 2 of H2)) and therefore not related to any cata-
lytic activity (the same feature is observed in Fig. 3b). After the
introduction of hydrogen, the water signal increases. At the same
time, the oxygen is consumed, confirming the occurrence of the
back reaction:
Fig. 2. Hydrogen and oxygen evolution rates for overall photocatalytic water
splitting with GaN:ZnO loaded with different cocatalysts. Relative humidity 85%. UV
LED at k ꢂ 367 nm (I ꢂ 92 mW/cm2). The inset a is an enlargement of the results for
Pt and Rh. In the histogram, the values of the initial rate (the peak that can be seen
in the time evolution of the QMS currents in inset b appearing immediately after the
light is switched on) are shown with full colored columns, while the values of the
stable rate obtained after 5 min of illumination are shown with hatched colored
columns. (For interpretation of the references to color in this figure legend, the
reader is referred to the web version of this article.)
3. Results
3.1. Photocatalytic gas-phase water splitting
2H2 þ O2 ! 2H2O
ð1Þ
GaN:ZnO with no cocatalyst and GaN:ZnO with five different
cocatalysts were tested for gas-phase overall water splitting using
an UV LED (ꢂ367 nm) as light source. In a previous publication, we
have shown that if the relative humidity, and thus the adsorption
of water on the catalyst, is increased, the activity also increases
[20]. It has been proposed that the increase in adsorbed water
can facilitate surface proton conduction, a necessary reaction step
for the water splitting mechanism. As a consequence, relative
humidity is an important parameter for gas-phase water splitting
and its value has been kept constant at ꢂ85% (constant water par-
tial pressure ꢂ27.3 mbar and temperature of the reactor ꢂ25 °C,
298 K) in all the tests. No hydrogen and oxygen photocatalytic evo-
lution could be detected with GaN:ZnO powder without a cocata-
lyst. All the other samples were active for gas-phase water
splitting, and the comparison of their activities is shown in Fig. 2.
Both the initial peak value and the stable value for both hydrogen
and oxygen obtained after 5 min of irradiation are shown for every
sample. The ratio between the evolved H2 and O2 is ꢂ2.5 for the
samples loaded with the Cr2O3/Rh, Cr2O3/Pt, and Rh2ꢀyCryO3 for
both the initial and stable values. The noble metal–loaded samples
show a stronger departure from the stoichiometric ratio for the ini-
tial activity rate, while the ratio decreases during illumination. The
GaN:ZnO samples loaded with only noble metals have the lowest
activity, while the samples where Cr is present have a higher activ-
ity. In particular, the Cr2O3 shell improves the activity of both the
Pt sample and the Rh sample. Cr2O3/Rh has a higher activity than
Cr2O3/Pt and comparable to Rh2ꢀyCryO3. Apart from Rh2ꢀyCryO3/
GaN:ZnO, none of the materials have previously been reported in
the literature to work in gas phase for photocatalytic overall water
splitting. The activity of Rh2ꢀyCryO3/GaN:ZnO is similar to what has
been reported in our previous study for the same catalyst obtained
with slightly different preparation conditions [20]. The presented
experiments have been repeated several times, and the reproduc-
ibility of the results with the optimized catalysts is usually within
10–15%. Possible sources of error can be attributed to small
differences in the concentration of the catalysts in the solution
for the Pt-loaded sample (Fig. 1b). Notice that this spontaneous
reaction ( G < 0) occurs without the need of photoexcited carriers,
D
but it is facilitated by the catalytic properties of the surface of the Pt
nanoparticles. The dominant mechanism is a Langmuir–Hinshel-
wood type, involving the dissociative adsorption of hydrogen and
oxygen. In Fig. 3b, the experiment has been repeated for
Rh2ꢀyCryO3/GaN:ZnO. In this case, the introduction of hydrogen
does not cause any noticeable evolution of water or consumption
of oxygen, meaning that the back reaction is strongly suppressed
with the Rh2ꢀyCryO3 cocatalyst. Notice that this second experiment
shows also that if reaction 1 occurs on the hot filament in the QMS,
the rate is negligible and below our detection limit. Thus, there is no
need to repeat the experiment with an empty reactor.
This experiment was repeated for all the other samples includ-
ing the GaN:ZnO without cocatalyst. In this latter case, no back
reaction activity is observed. The highest back reaction activity is
observed for Pt cocatalyst. Both core/shell cocatalyst exhibited less
catalytic activity for the back reaction than their respective pure
noble metal. A similar back reaction experiment was also per-
formed for all the samples by changing the order of introduction
of H2 and O2. In this experiment, the initial flow is H2 and O2 is
introduced later. So in this experiment, the sample is in a dry H2
atmosphere when O2 is injected, while in the experiments dis-
cussed before it was in a dry O2 atmosphere when H2 was injected.
The results of the two kinds of back reaction experiments are sum-
marized in Fig. 4 together with the photocatalytic water splitting
results. For the latter, only the H2 evolution is shown not to over-
load the plot with information, while for the back reactions exper-
iment, the water formation rate measured after 30 min from the
introduction of the second reactant gas (H2 or O2) is shown. All
the raw data are available as Supplementary information. In the
case when O2 is introduced last, the water signal for the samples
that have a finite back reaction rate presents an initial peak (with
the maximum at ꢂ4 min after the introduction of O2). After that,
the signal reaches a lower steady state. We noticed that neither
the peak value nor the steady state is equal to the steady state of

F. Dionigi et al. / Journal of Catalysis 292 (2012) 26–31
29
Fig. 3. QMS current for hydrogen molecules (m/z = 2), oxygen molecules (m/z = 32), water molecules (m/z = 18), and helium (m/z = 4) as a function of time for a water splitting
back reaction experiment. The flow of H2 is 2 ml/min, of O2 1 ml/min, and of He 15 ml/min. Water is evolved in the case of Pt/GaN:ZnO. No water could be detected with
Rh2ꢀyCryO3/GaN:ZnO.
Rh2ꢀyCryO3, no back reaction rate can be detected even in the pres-
ence of light. A different situation occurs with Pt and Cr2O3/Pt
cocatalysts. With these samples, an increase in the water evolution
rate can be detected under illumination when the system is ex-
posed to the stoichiometric H2/O2 mixture. This effect is particu-
larly strong with Pt when we start flowing H2. In that case, the
water signal increases to a value approximately equal to the max-
imum of the initial peak and a new steady state is reached. To
check whether this effect is due to a cleaning of the surface of Pt,
from , that is, hydrocarbons contamination, we have repeated the
back reaction experiment with Pt immediately after the first exper-
iment, without dismounting the chip (Fig. S16). The sample
showed exactly the same behavior, which rules out the hypothesis
of photo-induced cleaning of the surface from contamination. We
have noticed that in both the experiments just described, mass
28 is increased when the light is on under stoichiometric H2/O2
mixture together with a small CO2 (m/z = 44) evolution. Since the
time evolution of mass 28 is different than the one of mass 44,
Fig. 4. Initial hydrogen evolution rate for overall photocatalytic water splitting
(blue squares) and water evolution rate for the dark back reaction experiments
we assigned this signal mainly to N2 (instead of CO) and thus to
photocorrosion of the surface of the catalysts. This attribution is
also supported by the ratio of the measured mass 14 and 28 (equal
to 0.15). Notice, however, that these signals are two orders of mag-
nitude smaller than the water signal. This small but detectable
photocorrosion is also present with the Pt sample when the back
reaction experiment starts with O2, but no evidence of photocorro-
sion has been seen during the photocatalytic water splitting exper-
iments or with other samples.
starting with H2 (green triangles) or O2 (black circles) diluted in He. Notice that the
3 set of data are taken from three different kinds of experiments, as indicated in the
labels. For the two back reaction experiments, H2 and O2 are mixed in the
stoichiometric ratio. In the x-axis is indicated the cocatalyst loaded on the GaN:ZnO.
The X indicates the bare GaN:ZnO powder without cocatalyst. (For the interpre-
tation of the references to color in this figure legend, the reader is referred to the
web version of this article.)
the other case, when O2 is flowing as first reactant gas and H2 is
introduced later.
As previously mentioned, the reactions are carried out in the
dark so the water formed in these experiments is not the result
of light excitation of the catalyst. However, to study the effect of
light, we have illuminated the sample with the UV LED light source
used in the photocatalytic water splitting experiment with the
maximum power (ꢂ460 mW/cm2) before the second reactant gas
is introduced and after at least 30 min of this introduction (Figs.
S4–S15). In the case of light exposure when only one of the reac-
tant gases was present, we have performed that experiment first
of all to clean eventually small amount of dirt accumulated on
the surface from the air and second to have a reference to compare
the light effect in the stoichiometric mixture. No effect of light can
be seen for the GaN:ZnO without cocatalysts. The same holds for all
the experiments with the Rh-based catalyst. If the light has an ef-
fect with these samples, it is negligible. It is noticeable that in the
case of pure GaN:ZnO and GaN:ZnO loaded with Cr2O3/Rh and
4. Discussion
The trend of photocatalytic activities in Fig. 2 for the gas-phase
water splitting reaction is in agreement with what has been re-
ported for liquid phase [3,21]. In liquid-phase experiments, stoichi-
ometric H2 and O2 production has been measured with GaN:ZnO
loaded with Rh–Cr mixed oxide and core/shell cocatalysts, while
in the reported gas-phase experiments, the ratio of H2 and O2
slightly exceeds the stoichiometric value. This behavior has been
also reported and discussed in a previous publication about gas-
phase water splitting with Rh2ꢀyCryO3/GaN:ZnO and could not be
associated with any substantial amount of hydrocarbon contami-
nation or H2O2 formation [20]. As shown in Fig. 2, overall photocat-
alytic water splitting was observed with all the samples loaded
with cocatalysts but not with the GaN:ZnO alone. This fact shows

30
F. Dionigi et al. / Journal of Catalysis 292 (2012) 26–31
the importance of the cocatalyst in the photocatalytic production
of H2 and O2 from pure water. The reason for this is that the cocat-
alyst improves catalytic kinetics at the surface and helps separat-
ing the photoelectrons from the photoholes, that otherwise
would recombine. Following these considerations, one of the aims
of this work is to obtain a better understanding of different cocat-
alysts and how their positive effect can be optimized. By compar-
ing the different activities for photocatalytic overall water
splitting, we can observe that there are marked differences be-
tween GaN:ZnO loaded with bare Pt and Rh and their modifica-
tions. Both Pt and Rh are known as good hydrogen evolution
reaction (HER) catalysts, while the surface of GaN:ZnO acts as the
oxygen evolution site (Fig. 1a). It is known by the experimental
work of Trasatti for electrochemical hydrogen evolution [22] and
by the theoretical study using DFT calculations of Nørskov et al.
that Pt is a very good catalyst for HER [23]. The same holds for
Rh. Even if the activity of Rh is lower than that of Pt, it is still
one of the best catalysts for this reaction, as can be seen in the vol-
cano plots obtained in the aforementioned works. On the other
hand, Cr2O3 has not been reported to be a good HER catalyst. These
considerations are not enough to explain the trend in Fig. 2, since
the activity of both Pt and Rh is smaller than those of Cr2O3/Pt,
Cr2O3/Rh, and Rh2ꢀyCryO3. However, we will show how the exper-
imental results can be justified by the presence of O2, formed by
the splitting of water, close to the HER cocatalyst. This situation
is different from the case of an electrochemical cell, where the
HER and the oxygen evolution reaction (OER) occur in separated
compartments. Instead in the system studied, H2 and O2 evolve
in the same environment. If we consider the comparison between
the noble metal and their core/shell modification shown in Fig. 5,
we can see how the Cr2O3 shell has the simultaneous effect of sup-
pressing the back reaction and enhancing the detected amount of
H2 and O2. This is in agreement with what was found for the model
electrodes, and explained by the fact that Cr2O3 can block O2 from
reaching the surface of the noble metal nanoparticle cocatalysts
while H+ can still get through. The latter can penetrate the shell, re-
act with the photoelectrons on the noble metal surface, and pass
back through the shell again as H2, due to its smaller molecular ra-
dius compared to O2. In this way, the oxygen reduction reaction
(ORR):
Fig. 6. Oxygen blocking by the proton- and hydrogen-permeable Cr2O3 shell. In the
absence of Cr2O3, the O2 can react on the Rh cocatalyst and recombine with the
hydrogen to form water (water splitting back reaction) (a). In the case of Cr2O3/Rh
cocatalyst, the shell prevents the O2 to reach the Rh, while protons can penetrate
the shell and react with the photoelectrons on the Rh core forming hydrogen (b).
as well as the catalytic reaction between H2 and O2 (reaction
Fig. 1b) is strongly suppressed. So the Cr2O3 acts as an oxygen-
blocking proton/hydrogen membrane (Fig. 6a and b). The fact that
Pt and Rh show high activity for the back reaction is well known
in fuel cell literature [24], where ORR is studied, and it is also illus-
trated in the volcano plot for ORR calculated using DFT by Nørskov
et al. [25]. It can be seen from the comparison of the two volcano
plots that even if the two noble metals, Pt especially, are very good
for HER, they are also very good for ORR. So the trend shown in
Fig. 2 is a trade-off between these two phenomena. If oxygen is able
to reach the HER cocatalyst, this one needs to be a very bad ORR
cocatalyst in order not to reduce it. It is very difficult to find a single
material that has both these properties. It is probably a better ap-
proach to try and protect the HER catalyst with a membrane, like
it has been done in the case with the Cr2O3 shell that separates it
from the O2 produced at the OER active sites.
The proposed explanation that the observed back reaction rates
are the main cause of the deviation from the HER trend implies
that the rate-determining step is the charge transfer at the cocata-
lyst molecules and not the one at the photocatalyst–cocatalyst
(where possible different Schottky barriers can be thought to play
a role). Indeed, Szklarczyk and Bockris have reported that the pho-
tocatalytic evolution of H2 with a photocathode of p-InP coated
with islets of different metals (Pb, Cd, Co, Au, Ni, Pt) correlates very
well with the dark electrocatalytic evolution of H2 from the bulk
metal electrodes [26]. From this comparison, the authors conclude
that the charge transfer at the photocatalyst–cocatalyst interface is
not the step that determines the photoelectrochemical evolution of
H2 in the case of a semiconductor cover with metal islets. This con-
clusion is in agreement with the characteristic times for primary
processes in TiO2 photocatalysis reported by Fujishima et al. in a
recent review [27]. From the comparison between the characteris-
tic times, it is clear how in Pt-loaded TiO2 the electron transfer
from TiO2 to the Pt clusters is much faster than the interfacial
charge transfer to adsorbates.
4Hþ þ 4eꢀ þ O2 ! 2H2O
ð2Þ
The behavior of Rh2ꢀyCryO3 cocatalyst is more difficult to ex-
plain than the one of the core/shell cocatalysts. The experiments
show that the chromia mixed with Rh is still preventing O2 to be
reduced. According to our previous study using XPS and XAFS,
the valence state of Rh species in Rh2ꢀyCryO3 is trivalent, which re-
mains unchanged even after water splitting reaction [28]. How-
ever, it is still not clear how the Rh species in the cocatalyst
work during the working state of water splitting.
Fig. 5. Comparison between the two noble metals tested as cocatalysts and their
core/shell modification. The cocatalysts were deposited on GaN:ZnO. In the figures,
the initial hydrogen evolution rate for overall photocatalytic water splitting (blue)
and the water evolution rate for the dark back reaction experiments starting with
O
2 (black) are shown. The arrows further underline the opposite tendency between
Comparing the results of the two kinds of back reaction
experiments where the order of H2 and O2 introduction is changed,
one can notice that they basically show the same trend: Pt has the
the H2 detected in the water splitting experiments and the H2O evolved in the back
reaction. (For the interpretation of the references to color in this figure legend, the
reader is referred to the web version of this article.)

F. Dionigi et al. / Journal of Catalysis 292 (2012) 26–31
31
highest rate of back reaction, the back reaction rate from Cr2O3/Pt
Acknowledgments
is reduced with respect to Pt, and the one from Cr2O3/Rh and
Rh2ꢀyCryO3 is strongly suppressed with respect to the one of Rh.
However, even if they present the same trend, there are some dif-
ferences between the two kinds of back reaction experiments. In
particular for Pt and Rh, the steady-state reaction rate reached
when the sample is subject to an initial H2 flow is lower in inten-
sity than the one when O2 is flown first. This difference could arise
from different surface coverages of oxygen and hydrogen on the
cocatalyst nanoparticles. Indeed, it is known that at low tempera-
tures, as the one in our experiment, hydrogen has the effect of poi-
soning the surface due to the weak temperature dependence of its
sticking coefficient with respect to the exponential temperature
dependence of desorption. A hysteresis cycle with two different
steady-state reaction rates has been reported for Pt, in agreement
with our results [8,29]. The steady state reached after illumination
in the case of Pt and Cr2O3/Pt can also be due to a different surface
coverage, this time induced by the interaction of the adsorbates
with the photoexcited carriers. Another difference that can be no-
ticed comparing the two types of back reaction experiments is that
the suppression of the back reaction due to the chromia shell
seems to be stronger when the experiments start with O2 instead
of H2. A possible explanation for this difference in behavior may
be a partial reduction of the chromia under the initial hydrogen
atmosphere. However, this case is less interesting than the one
starting with oxygen since in typical water splitting experiments
the catalyst is transferred from air to liquid water, so it is not sub-
jected to a reducing hydrogen atmosphere. From Figs. 4 and 5, it is
clear that the water formation rate is higher for the Cr2O3/Pt than
for the Cr2O3/Rh, where it is basically negligible. This fact can likely
be the cause for the higher net forward photocatalytic activity ob-
served for the Cr2O3/Rh system compared to the Cr2O3/Pt system.
Furthermore, this observation suggests that if Pt can somehow be
protected more efficiently, then this system would have a higher
activity, potentially even higher than protected Rh. This consider-
ation agrees with the fact that the back reaction activity for the
bare Pt is much higher than the one for the Rh, as expected from
the ORR volcano plot, while their photocatalytic water splitting
activity is comparable.
Center for Individual Nanoparticle Functionality (CINF) is fi-
nanced by the Danish National Research Foundation. This was also
financed by the Research and Development in a New Interdisci-
plinary Field Based on Nanotechnology and Materials Science pro-
gram of the Ministry of Education, Culture, Sports, Science and
Technology (MEXT) of Japan, and The KAITEKI Institute, Inc. This
project was further funded by the ‘‘Catalysis for Sustainable En-
ergy’’ (CASE) research initiative, which is funded by the Danish
Ministry of Science, Technology and Innovation.
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.jcat.2012.03.021.
References
[1] A. Kudo, Y. Miseki, Chem. Soc. Rev. 38 (2009) 253.
[2] M.G. Walter, E.L. Warren, J.R. McKone, S.W. Boettcher, Q. Mi, E.A. Santori, N.S.
Lewis, Chem. Rev. 110 (2010) 6446.
[3] K. Maeda, K. Teramura, D. Lu, N. Saito, Y. Inoue, K. Domen, Angew. Chem. Int.
Ed. 45 (2006) 7806.
[4] K. Maeda, K. Domen, J. Phys. Chem. Lett. 1 (2010) 2655.
[5] D. Wang, A. Pierre, M.G. Kibria, K. Cui, X. Han, K.H. Bevan, H. Guo, S. Paradis, A.-
R. Hakima, Z. Mi, Nano Lett. 11 (2011) 2353.
[6] M. Yoshida, K. Takanabe, K. Maeda, A. Ishikawa, J. Kubota, Y. Sakata, Y. Ikezawa,
K. Domen, J. Phys. Chem. C 113 (2009) 10151.
[7] B. Hellsing, B. Kasemo, V.P. Zhdanov, J. Catal. 132 (1991) 210.
[8] E.M. Larsson, C. Langhammer, I. Zoric, B. Kasemo, Science 326 (2009) 1091.
[9] K. Maeda, T. Takata, M. Hara, N. Saito, Y. Inoue, H. Kobayashi, K. Domen, J. Am.
Chem. Soc. 127 (2005) 8286.
[10] K. Maeda, K. Teramura, T. Takata, M. Hara, N. Saito, K. Toda, Y. Inoue, H.
Kobayashi, K. Domen, J. Phys. Chem. B 109 (2005) 20504.
[11] K. Maeda, K. Teramura, K. Domen, J. Catal. 254 (2008) 198.
[12] K. Maeda, K. Domen, Chem. Mater. 22 (2010) 612.
[13] K. Maeda, K. Teramura, D. Lu, N. Saito, Y. Inoue, K. Domen, J. Phys. Chem. C 111
(2007) 7554.
[14] K. Maeda, K. Teramura, D. Lu, T. Takata, N. Saito, Y. Inoue, K. Domen, Nature
440 (2006) 295.
[15] K. Maeda, K. Teramura, D. Lu, T. Takata, N. Saito, Y. Inoue, K. Domen, J. Phys.
Chem. B 110 (2006) 13753.
[16] T.R. Henriksen, J.L. Olsen, P.C.K. Vesborg, I. Chorkendorff, O. Hansen, Rev. Sci.
Instrum. 80 (2009) 124101.
[17] P.C.K. Vesborg, J.L. Olsen, T.R. Henriksen, I. Chorkendorff, O. Hansen, Rev. Sci.
Instrum. 81 (2010) 016111.
5. Conclusion
[18] P.C.K. Vesborg, J.L. Olsen, T.R. Henriksen, I. Chorkendorff, O. Hansen, Chem.
Eng. J. 160 (2010) 738.
[19] P.C.K. Vesborg, S.-I. In, J.L. Olsen, T.R. Henriksen, B.L. Abrams, Y. Hou, A.
Kleiman-Shwarsctein, O. Hansen, I. Chorkendorff, J. Phys. Chem. C 114 (2010)
11162.
[20] F. Dionigi, P.C.K. Vesborg, T. Pedersen, O. Hansen, S. Dahl, A. Xiong, K. Maeda, K.
Domen, I. Chorkendorff, Energy Environ. Sci. 4 (2011) 2937.
[21] K. Maeda, K. Teramura, N. Saito, Y. Inoue, H. Kobayashi, K. Domen, Pure Appl.
Chem. 78 (2006) 2267.
[22] S. Trasatti, J. Electroanal. Chem. 39 (1972) 163.
[23] J.K. Nørskov, T. Bligaard, A. Logadottir, J.R. Kitchin, J.G. Chen, S. Pandelov, U.
Stimming, J. Electrochem. Soc. 152 (2005) J23.
[24] F.H.B. Lima, J. Zhang, M.H. Shao, K. Sasaki, M.B. Vukmirovic, E.A. Ticianelli, R.R.
Adzic, J. Phys. Chem. C 111 (2007) 404.
[25] J.K. Nørskov, J. Rossmeisl, A. Logadottir, L. Lindqvist, J.R. Kitchin, T. Bligaard, H.
Jonsson, J. Phys. Chem. B 108 (2004) 17886.
[26] M. Szklarczyk, J.O’M. Bockris, J. Phys. Chem. C 88 (1984) 5241.
[27] A. Fujishima, X. Zhang, D.A. Tryk, Surf. Sci. Rep. 63 (2008) 515.
[28] K. Maeda, K. Teramura, H. Masuda, T. Takata, N. Saito, Y. Inoue, K. Domen, J.
Phys. Chem. B 110 (2006) 13107.
Gas-phase photocatalytic water splitting experiments show, in
agreement with liquid-phase analogs, that Rh2ꢀyCryO3/GaN:ZnO
and Cr2O3/Rh/GaN:ZnO have a superior activity for this reaction
compared to Rh/GaN:ZnO, Pt/GaN:ZnO and Cr2O3/Pt/GaN:ZnO.
The experiments also show that Cr2O3/Pt/GaN:ZnO has a higher
activity than Pt/GaN:ZnO and Rh/GaN:ZnO. The results of the pho-
tocatalytic experiments find a correlation with the ones from the
water splitting back reaction (stoichiometric H2 oxidation). The
sample with the Pt cocatalyst has the highest rate of back reaction
and the Cr2O3 shell deposited on the Pt reduces this rate. In a sim-
ilar way, the water formation rate for Cr2O3/Rh and Rh2ꢀyCryO3 is
strongly suppressed and negligible compared to the one for Rh.
The combination of the photocatalytic water splitting and its cata-
lytic back reaction clearly shows the positive effect of the Cr2O3
shell in suppressing the water formation from H2/O2 back reaction,
thereby enhancing the net production of H2 and O2 from water
splitting.
[29] V.P. Zhdanov, Surf. Sci. 296 (1993) 261.