Solar light photocatalytic hydrogen production from water over Pt and Au/TiO2(anatase/rutile) photocatalysts: Influence of noble metal and porogen promotion

Article abstract of DOI:10.1016/j.jcat.2009.11.006

Hydrogen production from water under artificial solar light irradiation was performed over a series of Pt and Au/TiO₂(anatase/rutile) photocatalysts. Different TiO₂ supports with varying anatase/rutile contents were compared, based on either sol-gel synthesis or commercial TiO₂. The influence of template promotion on sol-gel TiO₂ synthesis has been studied using different porogens or templates. Among various factors influencing the hydrogen evolution efficiency, it was pointed out that the following parameters were crucial to enhance H₂ evolution: (i) the nature and content of the metallic co-catalyst, (ii) the surface, crystallographic, and porosity properties of the TiO₂ anatase/rutile support, (iii) the anatase/rutile ratio, (iv) the metal-support interactions, and (v) the relative amount of methanol added as a sacrificial reagent. The influence of these different factors was studied in detail. In optimized conditions, important H₂ production efficiency (120 μmol/min) was obtained over days without deactivation and with very low amounts of methanol.

Full text of DOI:10.1016/j.jcat.2009.11.006

Journal of Catalysis 269 (2010) 179–190
Contents lists available at ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Solar light photocatalytic hydrogen production from water over Pt and
Au/TiO (anatase/rutile) photocatalysts: Influence of noble metal and porogen
2
promotion
1
Olivier Rosseler, Muthukonda V. Shankar , Maithaa Karkmaz-Le Du, Loïc Schmidlin,
*
Nicolas Keller , Valérie Keller
Laboratoire des Matériaux, Surfaces et Procédés pour la Catalyse – European Laboratory for Catalysis and Surface Sciences, CNRS-University of Strasbourg,
25 Rue Becquerel, 67087 Strasbourg Cedex, France
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 30 July 2009
Revised 2 November 2009
Accepted 6 November 2009
Available online 9 December 2009
Hydrogen production from water under artificial solar light irradiation was performed over a series of Pt
and Au/TiO
were compared, based on either sol–gel synthesis or commercial TiO
tion on sol–gel TiO synthesis has been studied using different porogens or templates. Among various fac-
tors influencing the hydrogen evolution efficiency, it was pointed out that the following parameters were
crucial to enhance H evolution: (i) the nature and content of the metallic co-catalyst, (ii) the surface,
crystallographic, and porosity properties of the TiO anatase/rutile support, (iii) the anatase/rutile ratio,
iv) the metal–support interactions, and (v) the relative amount of methanol added as a sacrificial
reagent. The influence of these different factors was studied in detail. In optimized conditions, important
2
(anatase/rutile) photocatalysts. Different TiO
2
supports with varying anatase/rutile contents
2
. The influence of template promo-
2
2
Keywords:
H production
2
Photocatalysis
Water splitting
2
(
H
2
production efficiency (120
lmol/min) was obtained over days without deactivation and with very low
Pt/TiO
Au/TiO
2
amounts of methanol.
2
Template-assisted sol–gel synthesis
Methanol
Ó 2009 Elsevier Inc. All rights reserved.
Sacrificial reagent
Solar light
1
. Introduction
With increasing concern for the use of renewable resources and
hydrogen production, especially by direct water splitting is con-
sidered as a dream process because it enables pure hydrogen
production from renewable sources without the application of
an external potential. In the last years much effort has been
put in the development of photocatalysts that are able to har-
vest solar energy and to convert it into chemical energy. More
recently, several works have reported new photocatalysts with
improved efficiency toward water splitting under solar light exclu-
sively [2–6]. However, highly efficient hydrogen production by
photocatalytic water splitting is still far from practical application
and is closely linked to the development and improvement of
nanomaterials.
in particular for finding ways to use solar energy, solar light-acti-
vated photocatalysis is becoming an attractive tool. About 95% of
the present hydrogen needs are covered by steam reforming of
methane, a non-renewable source. Sustainable hydrogen produc-
tion systems are of great importance for the future. Photocatalysis,
with its potential to use sunlight for generating hydrogen, repre-
sents one of the promising technologies for clean and environmen-
tally friendly hydrogen production and provides a way to use
sunlight for generating hydrogen as a renewable green fuel.
Photocatalysis, as opposed to other renewable energy production
technologies, such as bio-fuel reforming, operates at ambient
conditions without complex processing requirements. Since the
discovery of photocatalytic water splitting on semi-conductor
catalysts by Fujishima and Honda [1], efficient photocatalytic
The photocatalytic activity of TiO
2
can be increased or de-
creased by the presence of a noble metal co-catalyst deposited
on its surface, depending on many factors such as the metal con-
tent and the nature of the reactant as well as the way of preparing
the catalyst. The enhancement of the photocatalytic activity has
been attributed to the formation of a Schottky barrier at the me-
tal/semi-conductor interface, which leads to electron trapping
and efficient charge separation [7]. Several metals have been tested
and reported in numerous publications [8–13], emphasizing the
*
Corresponding author.
E-mail address: nkeller@chimie.u-strasbg.fr (N. Keller).
Present address: Department of Materials Science and Nano-Technology, Yogi
1
Vemana University, Vemanapuram, Kadapa 516 003, India.
2
best properties of the Pt/TiO systems.
0
021-9517/$ - see front matter Ó 2009 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2009.11.006

1
80
O. Rosseler et al. / Journal of Catalysis 269 (2010) 179–190
Some works performed in the last decades have shown that it
ether (Brij 56), polyethylene glycol (PEG), polyvinyl alcohol
(PVA), or hexadecyltrimethylammonium bromide (CTAB) was then
added after the aging step and the reactant mixture was first dried
at room temperature and then further dried at 110 °C before the fi-
nal treatment in air at 400 °C for 2 h. Table 1 summarizes the dif-
ferent porosity promoters added during the sol synthesis and their
was possible to increase the hydrogen production reaction rate
for water splitting using water–alcohol mixtures, particularly
water–methanol systems [13–17]. In these cases, the rate of hydro-
gen production was found to be much higher, compared to that ob-
tained from photocatalytic cleavage of pure water. This increase
has been attributed to the effect of the alcohol acting as a sacrificial
agent (electron donor and holes and/or oxygen scavenger), by
main characteristics. The resulting samples are denoted TiO
Brij56, TiO –PEG, TiO –PVA, and TiO –CTAB, respectively. More
generally, TiO -porogen-promoted samples will be denoted TiO -P.
The platinum deposition was performed by wet impregnation
using an aqueous solution of H PtCl . After drying at 110 °C for
2
–
2
2
2
decreasing both electron–hole recombination and O
/H
2 2
back reac-
2
2
tion rates [18,19]. In most of these studies, the relative ratio of
methanol to water was relatively high. Although this process is
of interest in terms of environmental purposes and reduction in
the use of fossil fuels, it still relies on a fossil-derived fuel, metha-
nol or more generally on alcohols. A more sustainable route should
involve either pure water splitting or water splitting assisted by
lower amounts of methanol. Trying to lower and optimize the rel-
ative content of methanol while keeping high efficiency for water
splitting is one of the goals of our study.
2
6
1 h in air, and oxidation at 350 °C for 2 h, the material was finally
reduced in flowing hydrogen at 350 °C for 1 h. Amongst many re-
ported deposition methods [22], gold nanoparticle deposition
was performed here following the Direct Anionic Exchange (DAE)
method developed by Ivanova et al., which consists in hydroxyl
surface groups exchange with gold complexes [23]. The TiO
port was immerged into an aqueous chloroauric acid solution
(HAuCl ) and was further heated at 70 °C for 1 h. The solution
2
sup-
This work presents the results obtained for efficient photocata-
4
lytic hydrogen production on TiO
Au photocatalysts, including the search for optimized low ratio of
methanol. The influence of both metals, on the one hand, and of
2
anatase/rutile-supported Pt and
was subsequently washed with ammonia (35 vol.%) for 20 min, fil-
tered, and the resulting material was further dried at 120 °C and
calcined at 300 °C for 4 h in air (heating rate of 5 °C/min). In this
deposition process, the pH range strongly influences the gold com-
2
the TiO anatase/rutile-based support properties (anatase/rutile ra-
ꢀ
ꢀ
tio, specific surface area, porosity, etc.), on the other hand, will be
studied in detail and is correlated to complementary characteriza-
tion methods. High rate of hydrogen production could be sustained
over days without deactivation of the system on 0.3 wt.% Au/TiO
anatase–rutile photocatalyst.
plex nature from AuCl4 to AuðOHÞ and the ammonia washing
4
leads to gold complex grafting on the support surface by exchange
with hydroxyl groups. The final calcination reduces the gold com-
plexes to metallic gold nanoparticles.
2
2
. Experimental
.1. Material synthesis
The TiO photocatalysts were based on the TiO
2.2. Characterization techniques
2
Structural characterization was done by powder X-ray Diffrac-
tion (XRD) measurements, carried out on a D8-Advance diffrac-
tometer equipped with a Vantec detector, in a h/2h mode using
2
2
–P25 standard
from Degussa (now Evonik) with an anatase/rutile ratio of about
1
the Ka radiation of Cu at 1.5406 Å. The rutile and anatase contents
2
8
0/20 and a specific surface area of about 50 m /g, and sol–gel
nanosized anatase/rutile template-promoted materials.
Nanosized sol–gel TiO -template-promoted photocatalysts
were derived from the intensity of the most intense diffraction
peaks, the rutile (1 1 0) and anatase (1 0 1) peaks at 27.6° and
25.4°, respectively, according to the JCPDS files 21–1276 and 21–
1275.
2
were synthesized following a liquid phase low temperature disso-
lution–reprecipitation process using a sol–gel procedure in which
the titanium isopropoxide was first hydrolyzed by the addition of
a 2 M HCl aqueous solution, followed by a 48-h aging of the ob-
tained hydrosol [20]. Even at room temperature aging, the thermo-
dynamically stable rutile phase with fine particle size could be
obtained [21]. The porogen template, polyoxyethylene(10)cetyl
The surface area measurements were performed on an
2
ASAP2010 Micromeritics porosimeter using N as an adsorbent at
liquid N temperature. Prior to N adsorption, the material was
2
2
outgassed at 200 °C for 1 h in order to desorb impurities or mois-
ture from its surface. The surface areas were calculated by applying
the B.E.T. and t-plot methods.
Table 1
Different porogen promoters/templates used and added after the aging step during the sol–gel synthesis.
Name
Formula
Usual characteristics
PEG
Water soluble polymer used as thickener in cosmetics
O
Poly(ethylene glycol)
HO
H
n
PVA
Non-ionic surfactant lubricant for contact lens
OH
Poly(vinyl alcohol)
HO
OH
n
Brij56
Amphiphilic surfactant stable in acid and basic media
Cationic amphiphilic surfactant
O
Polyoxyethylene(10)cetyl ether
H C
OH
10
3
16
CTAB
H C
3
Br -
Hexaecyltrimethylammonium bromide
+
N
H C
CH3
3
CH3
12

O. Rosseler et al. / Journal of Catalysis 269 (2010) 179–190
181
Scanning Electron Microscopy (SEM) was carried out on a Jeol
JSM-6700F working at 3 kV voltage, equipped with a CCD camera.
The sample was previously coated with gold and then deposited on
a standard aluminum holder for observation.
tested, varying in the 0–25 vol.% range. For a better comparison,
all the photocatalytic tests were carried out with 1 g of catalyst
and 1 L of total liquid.
Transmission Electron Microscopy (TEM) was performed on a
Topcon 002B microscope working with a voltage of 200 kV and a
point-to-point resolution of 0.17 nm. The sample was sonically dis-
persed in an ethanol solution for 5 min and a drop of the solution
was deposited onto a copper grid covered by a holey carbon mem-
brane for observation.
3
. Results on Pt/sol–gel TiO
2
(anatase/rutile)-porogen modified,
Pt/TiO -P
2
2
3.1. Characterization results of the sol–gel TiO -P supports
The compared Thermal Gravimetry Analysis (TGA) performed
on TiO –Brij56, TiO –PEG, TiO –PVA, and TiO –CTAB samples after
The XPS measurements were performed on a MULTILAB 2000
2
2
2
2
(
TERMO) spectrometer equipped with Mg K
a
anode (h
m
= 1456 eV).
drying at 110 °C and calcination at 400 °C revealed that whatever
the templates or porosity promoters used, they were completely
eliminated after calcination (graphs not shown, see in [24] in the
Charging effects were corrected using the C1s peak at 284.6 eV as a
reference for the graphitic carbon. All the spectra were decom-
posed assuming several contributions, each of them having a
Doniach–Sunjic shape and a Shirley background subtraction.
Thermal Gravimetry Analysis (TGA) was performed using a
SETARAM thermo-analyser. Each sample was placed in a platinum
crucible and heated from room temperature to 900 °C with a heat-
ing rate of 10 °C/min, using a 20/80 (v/v) O
min flow stream.
The UV–Vis absorption spectra of the materials were recorded
on a Cary 100 Scan UV/Vis spectrophotometer (Varian).
2
case of the TiO –PEG material). The exothermic heat evolution ob-
tained on the dried samples indicated the combustion of the poro-
gen below 400 °C, whereas this exothermic peak did not appear
anymore on the samples calcined at 400 °C, confirming the re-
moval of the porogen during the calcination step.
The nitrogen adsorption/desorption curves provided informa-
tion related to the porous texture of the different porogen-modi-
fied samples, taking PEG, PVA, and Brij56 as examples (Fig. 2). All
3
2
/N
2
mixture at 50 cm /
2
the TiO -P materials revealed type-IV isotherms, characteristic of
mesoporous solids with a uniform pores distribution (even in the
case of CTAB, not represented here). Fig. 2 represents the nitrogen
2
.3. Experimental device and procedure
adsorption/desorption curves of non-modified sol–gel TiO
iting also type-IV behavior.
2
, exhib-
The liquid-phase hydrogen production was carried out in a Pyr-
It has already been shown, using this low temperature dissolu-
tion–reprecipitation sol–gel process, that the thermodynamically
stable rutile phase with a fine particle size appeared after
ex reactor, equipped with a plunging quartz tube containing a
50 W metal halide lamp (solar light simulation) located vertically,
1
under 500 mL/min nitrogen flow (Fig. 1). This lamp was simulating
the solar illumination, with a very low UV-A content and a large
part in the visible light range. The measured irradiance was
3
00 min of aging, even at room temperature [25]. Table 2 summa-
rizes BET and XRD characterizations of the different TiO -P and
unmodified samples, compared to the commercially available
TiO –P25 reference. One could note that for a same aging duration,
2
2
3
0 mW/cm . All the experiments were performed in the same irra-
2
diance conditions. This tube was surrounded by a water circulation
in a cylindrical Quartz jacket, in order to avoid heating of the aque-
ous media. The reaction was performed under a 500 rpm mechan-
ical stirring. Prior to any experiment, oxygen was removed by
nitrogen flushing. The reaction products were analyzed on-line
every 2 min by thermal conductivity detectors on a micro-gas
chromatography (Model Agilent P200 Series) allowing detection
and quantification of hydrogen, oxygen, CO, CO
5
catalytic tests were performed using pure water as a hydrogen
source, and methanol as a sacrificial additive agent for enhancing
hydrogen production. Different methanol/water ratios have been
the nature of the template had a huge influence on the anatase/ru-
tile ratio. More precisely, the addition of PEG porogen yielded the
formation of a large ratio of the rutile phase (95%), whereas this
proportion decreased in the order PVA (70%), CTAB (60%), and
Brij56 (50%). Without the addition of a template, pure rutile phase
was obtained, but with higher surface area than for usual rutile
phase obtained by heating anatase phase at high temperature (30
2
, and organics on
2
2
m /g against 8 m /g), pointing out the interest of these low tem-
perature sol–gel syntheses. The addition of template also delayed
the anatase–rutile phase transition. It is also worth mentioning
that crystallites obtained by this low temperature sol–gel synthesis
A molecular sieve, OV1, PlotQ, and Stabilwax columns. The photo-
exhibited a much lower size than in the case of the reference TiO
2
–
P25.
Adding a template during the low temperature dissolution–rep-
recipitation synthesis enhanced the specific surface areas, confirm-
ing that the templates acted as porogens or porosity promoters;
this positive effect is induced by the template decomposition dur-
ing the post-synthesis thermal treatment. The fingerprint, after
decomposition of the porogen or template will, of course be spe-
Water
circulation
N2
cific, depending on both size and chemical nature. The correspond-
2
ing measured specific surface areas ranged from 70 m /g (TiO
2
–
2
2
CTAB) to 190 m /g (TiO –PEG). The mean pore diameter values,
in the 4–10 nm range confirmed the presence of mesopores and
the influence of the nature of the porogen/template on the result-
ing pore sizes.
Body of the
reactor
Solar lamp
Quartz tube
Computer
2
UV–Vis absorption spectra recorded for these sol–gel TiO -P-
based samples were clearly shifted toward the visible light part
of the spectrum (also called red-shift), compared to the reference
TiO
absorption was extended up to 550 nm but is stronger in the
00–450 nm range. The graphical determination of the cut-off
wavelengths was performed for estimating the band gap energy
2
–P25 (Fig. 3). Whatever the template or porogen used, the
Magnetic stirrer
4
Fig. 1. Water-splitting photoreactor coupled with an on-line microGC.

1
82
O. Rosseler et al. / Journal of Catalysis 269 (2010) 179–190
TiO -PEG
2
1
1
1
40
20
00
anatase (5%)-rutile (95%)
TiO -PVA
2
anatase (30%)-rutile (70%)
TiO -Brij56
2
8
6
4
2
0
0
0
0
0
anatase (50%)-rutile (50%)
Sol gel TiO2
without template
0.0
0.2
0.4
0.6
0.8
1.0
Relative pressure, P/P0
2
Fig. 2. Adsorption/desorption isotherms of sol–gel TiO -P modified samples. Examples of PEG, PVA and Brij56 as templates.
Table 2
Specific surface area, mean crystallite size diameter, anatase/rutile ratio, mean pore diameter and total porous volume, depending on the starting template after 48 h aging.
2
TiO
2
Surface area (m /g)
Crystallite size (nm)
Anatase
Anatase:rutile ratio
Mean pore diameter (nm)/
porous volume (cm /g)
3
Rutile
TiO
TiO
TiO
TiO
2
2
2
2
–Brij56
–PEG
–PVA
100
190
150
70
30
50
7
–
9
11
–
10
11
10
10
10
52
64
50:50
5:95
10/0.35
4/0.64
4/0.28
6–7/0.18
/
/
/
30:70
40:60
0:100
80:20
0:100
–CTAB
Without template
P25
P25 calcined at 800 °C
32
–
8
2
2
1
1
0
0
.5
.0
.5
.0
.5
.0
values, both summarized in Table 3. The calculated band gap ener-
gies were all close to 2.98 ± 0.04 eV. The higher the rutile content,
the lower the apparent band gap energy (and the higher the cut-off
wavelength), as expected. Although the effect of carbon doping on
the visible light absorption is currently discussed in the literature,
it appeared in our case very unlikely, whatever the porogen pro-
moter used. Indeed, according to the TGA/TDA analyses, there
PEG
PVA
CTAB
Brij 56
P25
was no carbon elimination (as CO
one could assume that the shift in the visible light domain ob-
served on the TiO -P samples was certainly due to the formation
of the rutile phase, whose content was higher than that for TiO
P25. Furthermore, the addition of Pt (0.3 wt.% or 2 wt.%) to the
sol–gel TiO -P samples did not change the UV–Vis absorption char-
2
) up to 900 °C. Consequently,
2
2
–
2
acteristics (not shown in Fig. 3).
SEM images showed particle aggregations to form uniformly
sized 100–200 nm nanoclusters (Fig. 4, TiO
example), similar to that observed for the TiO
XRD measurements, particle size ranged from 7 to 11 nm, depend-
ing on the templating agent. It should also be reminded that the
porosity, measured previously in BET analyses, corresponds to
open voids between particles within these aggregates.
2
–Brij56 given as an
–P25. According to
2
00
300
400
500
600
700
800
2
Wavelength (nm)
2
Fig. 3. Light absorption properties of TiO -P materials.
Table 3
High crystallinity of the particles, for both anatase and rutile
phases, was confirmed by TEM analyses, showing nice and sharp
fringes (Fig. 5); the measured 0.35 and 0.45 nm interplane dis-
tances corresponded to the anatase (1 0 1) and rutile (1 1 0) crys-
tallographic planes, respectively [26].
Apparent band gap and cut-off wavelength of the different TiO
assisted materials.
2
sol–gel-template-
Template
Band
gap (eV)
Cut-off
wavelength
absorption (nm)
Anatase:
rutile ratio
To summarize, the above-described synthesis and characteriza-
TiO
TiO
TiO
TiO
2
2
2
2
–PEG
–PVA
–CTAB
–Brij56
2.94
2.97
2.99
3.02
422
419
415
411
5/95
2
tions of TiO using a low temperature sol–gel method modified
30/70
40/60
50/50
with different kinds of templates led to anatase–rutile composite
semi-conductors and mesoporous materials showing high crystal-
linity and varying ratios of anatase/rutile phases. Depending on the

O. Rosseler et al. / Journal of Catalysis 269 (2010) 179–190
183
35
30
25
20
15
10
5
0
2
0
,3wt%Pt/TiO -Brij56 (100 m /g)
2
2
0
,3wt%Pt/TiO -PEG (190 m /g)
2
2
0
,3wt%Pt/TiO -P25 (50 m /g)
2
2
0
,3wt%Pt/TiO -PVA (150 m /g)
2
2
0,3wt%Pt/TiO -CTAB (70 m /g)
2
2
2
wt%Pt/TiO -Brij56 (100 m /g)
2
2
0
,3wt%Pt/TiO -rutile obtained from P25 calcined at 900°C (8 m /g)
2
0
50
100
150
Time under solar illumination (min)
Fig. 6. On stream hydrogen production from methanol/water (250 ml/750 ml)
Fig. 4. SEM image of the sol–gel TiO
2
–Brij56 material.
mixture over 0.3 wt.% Pt/TiO
solar light illumination.
2 2
-P and 0.3 wt.% Pt/TiO –P25 catalysts under artificial
nature of the template added after the aging step during the sol–
gel synthesis, both surface area (from 70 to 190 m /g) and ana-
tase/rutile ratio (from 5/95 to 70/30) could be tuned. The last
parameter also had an influence on the band gap narrowing of
2
TiO . The role of the different organic polymers acting as pore reg-
ulating agents has already been reported in the literature [27],
pointing out the influence of their functional groups during the
hydrolysis step of the synthesis. A more pronounced polarity could
lead to stronger interactions with the surface of titania-based com-
plexes and with the gel in solution. Hydrogen bond formation
between templating agents and the growing titania particles in
formation into the sol–gel have also been detected using IR
spectroscopy. The different templating (amphiphilic, ionic, etc.)
molecules with varying sizes and functional groups yielded self-
assembly-like mesophases. Each template yields a specific meso-
phase, depending mostly on its size. During the drying step, these
2
illumination. Very negligible direct water splitting over bare TiO2–
P25 or TiO -P samples (<0.2 mol/min) was recorded over a few
minutes under illumination in pure water. Finally, it has been ver-
ified that methanol did not undergo photochemical decomposition,
even for wavelengths lower than 300 nm, as already mentioned in
the literature [29].
2
l
The different TiO
5 methanol/water ratio to the reference TiO
summarizes the results obtained on the 0.3 wt.% Pt/TiO
.3 wt.% Pt/TiO –P25 samples, tested under the same conditions,
keeping in mind that the only difference between the 0.3 wt.% Pt/
TiO -P samples is the template used for enhancing the porosity.
The results are nevertheless very different from one another.
Amongst these samples, the 0.3 wt.% Pt/TiO –Brij56 exhibits the
best efficiency (32 mol/min or 1920 mol/h), followed by the
.3 wt.% Pt/TiO –PEG catalyst (20 mol/min or 1200 mol/h),
both exhibiting higher activity than the 0.3 wt.% Pt/TiO –P25
reference sample (18 mol/min or 1080 mol/h). The 0.3 wt.% Pt/
TiO –PVA (16 mol/min or 960 mol/h) and 0.3 wt.% Pt/TiO –CTAB
mol/min or 480 mol/h) samples showed the worst activi-
ties for hydrogen production. It must also be mentioned that the
2
-P supports were first compared at a given 25/
–P25 material. Fig. 6
-P and
7
2
2
0
2
2
2
l
l
0
2
l
l
2
mesophases, trapped inside the TiO network, were used as im-
prints. Finally, their elimination or decomposition during the calci-
nation step led to the formation of potentially well-ordered
2
l
l
2
l
l
2
2
mesoporous TiO , as reported in the literature [28].
(
8
l
l
3.2. Photocatalytic water splitting results
0
2
.3 wt.% Pt/TiO rutile (100%) (obtained from calcined P25 at
8
00 °C) yielded very poor activity, certainly due to its low specific
Preliminary tests did not show any evidence of H
from water without photocatalysts, neither in the dark, nor under
2
production
surface area (Table 2). Increasing the Pt content to 2 wt.% on the
2
Fig. 5. TEM and high-resolution TEM (left) image of the sol–gel TiO –Brij56 sample. Example of an anatase particle (left) and of an anatase/rutile mixture (right).

1
84
O. Rosseler et al. / Journal of Catalysis 269 (2010) 179–190
Table 4
2
also be underlined on the TiO –P25 reference, confirming previous
2
Hydrogen evolution rate (l –Brij56 as a function of the
mol minꢀ1) on 0.3 wt.% Pt/TiO
observations that the optimum Pt loading lies in the 0.1–1 wt.%
range [8,12,30–32]. This optimum in Pt content is often explained
as the result of a competition between, (i) a decrease of electron/
ratio of methanol sacrificial reagent. Stabilized value over more than 42 h of reaction.
Stabilized hydrogen evolution rate
(lmol min )
Methanol volume
(mL)
Methanol
ratio
ꢀ
1
2
hole recombination at the TiO surface in intimate contact with
(
vol.%)
Pt (due to the electron trapping effect on Pt clusters) and (ii) elec-
tron/hole recombination on Pt. In addition, above a certain metal
loading, the particles begin to touch and overlap, thus decreasing
0
0
1
4
1
3
.2
.2
0.1
1
10
0.01
0.1
1
the metal–TiO
The steady-state regime of H
a function of the methanol/water ratio, ranging from 0.01 to
5 vol.%, over the most active 0.3 wt.% Pt/TiO –Brij56 catalyst. It
2
interface and the efficiency of charge transfers.
50
5
3
2
100
250
10
25
2
evolution is reported in Table 4 as
2
2
has already been illustrated in Fig. 6 that steady-state regime
was reached for 25 vol.% of methanol after about 40 min under
flow and illumination. No deactivation was observed over more
most efficient Pt/TiO
2
–Brij56 sample, led to a decrease in photocat-
alytic H production to very poor values. The same tendency could
2
2
Fig. 7. TEM images of 0.3 wt.% Au/P25 (a) and (b), 2 wt.% Au/P25 (c) and (d), 2 wt.% Au/TiO –Brij56 (e) and (f) samples.

O. Rosseler et al. / Journal of Catalysis 269 (2010) 179–190
185
than 43 h of continuous reaction. The stabilized H
was clearly enhanced by increasing the ratio of methanol sacrificial
reagent.
2
evolution rate
ments, and XRD and XPS analyses), for metal loadings varying be-
tween 0.3 and 2 wt.%.
Fig. 7 represents TEM images of 0.3 wt.% Au/TiO
Au/TiO –P25, and 2 wt.% Au/TiO –Brij56 photocatalysts at differ-
ent magnifications. First, increasing the gold amount from 0.3 to
2 wt.% on TiO –P25 led to an increase in the mean particle size
2
–P25, 2 wt.%
The increase in photocatalytic activity for H
.3 wt.% Pt/TiO –Brij56 photocatalyst could be explained putting
forward different possibilities: (i) a higher Pt dispersion due to
the higher porosity of TiO after the removal of the template,
2
evolution over the
2
2
0
2
2
2
from 3 to 8 nm (Fig. 7a–d), however maintaining a high dispersion
of the metallic nanoclusters. The same tendency is observed on
resulting in better electronic interactions between the Pt co-cata-
lyst and the support [33]; and (ii) the higher surface area, the high-
er the accessibility of the reactants to the actives sites; (iii)
coupling both anatase and rutile particles enhances charge separa-
tion and takes advantage of the activation of the rutile in the visi-
TiO
2 wt.% Au loading (Fig. 7e and f). It could be pointed out that
although the surface area of the TiO –Brij56 support was two times
larger, the metallic mean particle size was very similar at 2 wt.%
content on both TiO –Brij56 and TiO –P25 supports, in spite of
2
–Brij56, showing mean gold particle size of about 8 nm at
2
ble region of the spectrum. In addition, as the evolution of H
is expected to occur on different sites, on Pt/ anatase and rutile
particles, respectively, it should limit the thermodynamically fa-
vored reverse reaction between O and H , as already mentioned
by other authors [34]. This could contribute to a certain extend
to the higher activity of the TiO –Brij56 sample. Nevertheless, sur-
2
and
2
2
O
2
the higher dispersion expected on the former support.
UV–Vis light absorption spectra, after gold deposition, showed a
2 2
modification of the absorption properties of TiO –P25 and TiO –
2
2
Brij56 supports (Fig. 8). Unlike platinum, gold deposition mainly
yielded an additional absorption band around 550 nm. Qualita-
tively, directly after gold deposition the sample had a much darker
color (dark violet to almost black). Obviously, this colored material
necessarily absorbs visible light, confirmed by the presence of a
broad absorption peak centered around 550 nm. This absorption
has already been reported in the literature and is attributed to a
plasmon resonance phenomenon [41,42] due to collective oscilla-
tions of the conduction electrons located on the 6s orbital of gold
and induced by the incident electromagnetic wave. This phenome-
non is frequently observed for particle sizes between 5 and 50 nm,
the absorption band width theoretically depending on the particle
size. In addition to this plasmon absorption, deposition of gold also
shifted the absorption spectrum deeper into the visible light range,
in the 380–450 nm region. Consequently, the estimated band gap
value was decreased (Table 5). The band gap narrowing could be
2
face area and pore volume or distribution are obviously not the
only factors affecting the photocatalytic activity but the anatase/
rutile ratio also has to be considered. For example, TiO
TiO –PVA have the two largest surface areas (and TiO
highest pore volume) and also the highest rutile content with
5% and 70%, respectively. Consequently, a large rutile content,
2
–PEG and
2
2
–PEG the
9
even though it absorbs a larger part of visible light, is detrimental
to a high photocatalytic activity. As a result, anatase/rutile coupling
and intimate contact in an optimal ratio allows (i) the absorption of
a part of the visible light spectra via the rutile phase, by the so-
called ‘‘antenna effect” in photocatalysis [35,36], (ii) the spatially
separate evolution of H
ero-junction formation, of two semi-conductors, anatase and rutile
TiO , leading to a better separation of photogenerated charges
2 2
and O , and (iii) the coupling, via het-
2
through interparticle charge transfer [37–40]. The anatase–rutile
semi-conductor coupling under solar light irradiation (UV-A + vis-
ible radiations) underlines the importance of the relative position
of the valence and conduction bands of both semi-conductors. Ana-
tase particles can only be activated by the small UV-A fraction of
the solar irradiation, whereas rutile particles on the other hand
can absorb the UV-A illumination but, in addition, are also able
2 2
as large as 0.33 eV or even 0.67 eV on TiO –Brij56 and TiO –P25
supports, respectively. Thus, regarding only that visible light
absorption shift, without consideration of the plasmon resonance
phenomenon, the addition of 1.4 wt.% of gold seemed to be the
optimum ratio and led to the best absorption properties, whatever
the TiO
2
support be.
supports and any gold content (between 0.3 and
For both TiO
2
to absorb visible light up to 420 nm. When these two TiO
2
phases
2 wt.%), XRD characterizations did not show any diffraction peak
characteristic of gold, certainly due to the low gold content, below
detection limit. Furthermore, the results of BET and pore distribu-
tion analyses before and after gold deposition are not significantly
different, meaning that such low gold concentrations do not affect
are in contact and simultaneously illuminated with solar light,
both anatase and rutile semi-conductors can be activated at the
same time, resulting in a vectorial photogenerated charge transfer
and thus in a better charge separation. The more intimate the con-
tact between anatase and rutile nanoparticles, the more efficient
the electon/hole interparticle transfer phenomenon. But consider-
ing that anatase sites, or more precisely Pt/anatase sites, are the ac-
tive sites for hydrogen evolution, a too high rutile content is
therefore detrimental to hydrogen evolution. This assumption
may explain the existence of an optimum anatase/rutile ratio.
TiO -Brij 56
2
2
1
1
0
0
.0
.5
.0
.5
.0
P25
2
1
% Au/TiO -Brij 56
2
.4% Au/TiO -Brij 56
2
0.3% Au/TiO -Brij 56
2
4
2
. Results on Au/TiO -P samples
1
2
.4% Au/P25
% Au/P25
Two kinds of supports have been evaluated for gold deposition
and photocatalytic water splitting experiments, the above-men-
tioned TiO –Brij56, that showed the best results with platinum
promotion, and the reference TiO –P25.
2
2
4.1. Characterization results
In addition to the previous characterizations of the TiO
2
sup-
200
300
400
500
600
700
800
ports (Section 3.1) in terms of TGA, BET, UV–Vis absorption, XRD,
and SEM and TEM analyses, additional analyses have been further
performed after gold deposition (UV–Vis absorption, BET measure-
Wavelength (nm)
Fig. 8. Light absorption properties of Au/TiO
2
photocatalysts.

1
86
O. Rosseler et al. / Journal of Catalysis 269 (2010) 179–190
Table 5
more precisely from an increase in effective positive charges
around Ti(IV). No contributions corresponding to Ti lower oxida-
tion states have been detected. It can also be supposed from
Fig. 9a that for both supports, the addition or grafting of gold by an-
ionic exchange did not modify the Ti(IV) binding energies, evidenc-
ing that there is no direct interaction between gold and titanium
atoms on the surface; indeed, no shift is detected for Ti 2p on
Band gap and cut-off wavelength absorption of the different Au/TiO
photocatalyst.
2
anatase–rutile
Photocatalyst
P25
Band gap (eV)
Maximum wavelength
absorption (nm)
3.23
3.02
2.69
2.76
2.91
2.56
2.73
385
411
461
450
427
485
455
2
TiO Brij56
1
2
0
1
2
.4 wt.% Au/P25
wt.% Au/P25
.3 wt.% Au/TiO
.4 wt.% Au/TiO
2
wt.% Au/TiO anatase–Brij56
1
.4 wt.% Au/TiO
2
–P25 and 2 wt.% Au/TiO
2
–Brij56. The very slight
–P25 is
binding energy shift (about 0.2–0.3 eV) on 2 wt.% Au/TiO
2
2
Brij56
–Brij56
2
not significant, compared to the experimental error to conclude
an interaction or bonding between Ti and Au atoms.
The O 1 s region of the XPS spectra (Fig. 9b and c) exhibited a
broad signal that has been deconvoluted into two contributions,
a dominant peak at lower binding energy, corresponding to oxygen
bonded to titanium atoms [43], and a weaker contribution, as-
signed to –OH surface hydroxyl groups at higher energy. The bind-
ing energy of the oxygen bonded to titanium contribution is
the specific surface area, nor the pore size or distribution by filling
or blocking them.
XPS characterizations have been performed on both TiO
and TiO –Brij56 supports before and after gold deposition. For all
these samples, Ti 2p (Fig. 9a), O 1 s (Fig. 9b and c), and Au 4f
Fig. 9d) spectra have been recorded, deconvoluted, and analyzed.
2
–P25
2
2 2
located at 529.1 and 529.6 eV on TiO P25 and TiO –Brij56, respec-
tively (Fig. 9b). This difference in binding energy between the two
supports is increased to 1 eV in the case of the –OH surface group
contribution, identified at 530 and 531 eV, respectively. Gold depo-
sition increased strongly this –OH surface contribution binding en-
(
The Ti 2p XPS spectra reported in Fig. 9a showed a signal that could
be fitted with only two components, related to the Ti 2p spin–orbit
components of Ti(IV) surface species. A first comparison between
the two kinds of supports revealed a consistent binding energy
ergy on TiO
Fig. 9c). In contrast, the addition of gold did not modify the posi-
tion of the –OH contribution for the TiO –Brij56 sample. Further-
more, gold deposition also decreased the relative contribution of
surface –OH groups, this effect being stronger on TiO –P25 than
on TiO –Brij56. The –OH contribution on TiO –P25 decreased from
8% on the bare support to 19% and 12% after deposition of 1.4 and
2
–P25 sample, showing a shift from 530 to 531.7 eV
(
shift of 0.7 eV. The Ti 2p3 –Ti 2p
/2
1/2
spin–orbit components ap-
peared at 458.0–463.7 and 458.7–464.2 eV for TiO –P25 and
TiO –Brij56, respectively. This shift to higher binding energies ob-
served on the sol–gel–Brij56-promoted support results from a dif-
2
2
2
2
2
2
2
ferent chemical environment of Ti(IV), compared to TiO –P25 and
3
A
C
P25
P25
TiO2 anatase(50)/rutile(50) - Brij56
19%
2
wt.% Au/TiO anatase(50)/rutile(50) - Brij56
1.4 wt.% Au/P25
2
12%
2
wt.% Au/P25
2
wt.% Au/P25
2
wt.% Au/TiO2
27%
anatase(50)/rutile(50) - Brij56
1.4 wt.% Au/P25
454
456
458
460
462
464
466
468
526
528
530
532
534
Binding energy, eV
Binding energy, eV
D
B
2
wt.% Au/P25 after
test with 10 mL methanol
30%
TiO2 anatase(50)/
rutile(50) - Brij56
2
wt.% Au/TiO anatase(50)/
2
rutile(50) - Brij56
38%
2
wt.% Au/P25
90
P25
526
528
530
532
534
80
82
84
86
88
Binding energy, eV
Binding energy, eV
Fig. 9. XPS spectrum of (A) the Ti 2p1/2–Ti 2p3/2 region, (B) the O 1 s region of Au-free TiO
, (C) the O 1 s region of Au/TiO
2
, and (D) the Au 4f7/2–Au 4f region.
5/2
2

O. Rosseler et al. / Journal of Catalysis 269 (2010) 179–190
187
2
wt.% of gold, respectively, whereas it was only reduced from 30%
2
results obtained in Fig. 10 on Au/TiO –P25 with different gold con-
to 27% on the TiO –Brij56 support after the deposition of 2 wt.% of
2
tents (0.3, 1.4, and 2 wt.%), the optimal gold loading seems to be at
2 wt.%. It led both to the highest H evolution rate and to the low-
est deactivation, even if the H production efficiency still remained
very low, around 0.23 mol/min. We could also notice that hydro-
gen production was close to zero on the 1.4 wt.% Au/TiO –P25 sam-
ple. On the 0.3 wt.% Au/TiO –P25 photocatalyst, despite the less
important deactivation, the initial H production efficiency was
lower (Fig. 10). According to these preliminary tests, the 2 wt.%
Au/TiO –P25 photocatalyst has been chosen for further photocata-
gold. It must be underlined that gold deposition did not induce any
shift in the position of the first contribution (i.e. oxygen bonded to
titanium atoms). The strong intensity decrease and the large shift
2
2
l
in binding energy of the second peak, on TiO
2
–P25, suggest a direct
2
coordination of gold atoms with the –OH surface groups. Indeed,
2
during the ammonia washing step of DAE, gold complexes in solu-
2
ꢀ
tion are transformed into ½AuðOHÞ ꢁ and the complexed hydroxyls
4
are then exchanged with surface –OH groups. The same interaction
2
on TiO
2
–Brij56 may not occur, or not be so strong, explaining the
lytic tests, investigating the influence of methanol as a sacrificial
reagent.
much lower decrease of the –OH signal.
The presence of metallic gold was confirmed on both kinds of
Fig. 11 summarizes the effect of the relative ratio of methanol
7/2
5/2
support by the XPS spectra of Au 4f and Au 4f for 2 wt.% load-
ing (Fig. 9d). Nevertheless, a 0.4 eV shift was observed between the
two kinds of support, exhibiting binding energies of 82.9 and
sacrificial reagent on the 2 wt.% Au/TiO
pared to the 0.3 wt.% Pt/TiO –Brij56 sample, previously identified
as the most efficient Pt-containing photocatalyst for H produc-
2
–P25 photocatalyst, com-
2
2
8
3.3 eV, for TiO
2
–P25 and TiO
2
–Brij56, respectively. This shift
tion. Comparing these two series of Au- and Pt-containing photo-
catalysts, the results clearly indicate that the addition of methanol
clearly evidences different chemical environments and interac-
tions with surface gold atoms. Based on the previous observation
that, after gold deposition, mostly the –OH surface contribution
has a more beneficial effect for H
P25 catalyst. The addition of 1 vol.% (10 mL in 1 L) of methanol en-
hanced H evolution by several orders of magnitude (from
0.23 mol/min to 7200 mol/h). Taking the last photocatalyst,
for example, a long lasting experiment yielded, after 10 days of
continuous reaction on stream, 38.7 L (1.728 mol) of H produc-
2 2
production on 2 wt.% Au/TiO –
of the TiO
2
–P25 support was affected, it could be assumed that a
2
specific Au–OH interaction or bonding occurs with this support.
This is apparently not the case with the TiO–Brij56-promoted
support.
l
l
2
Fig. 9d also shows that the photocatalytic test does not affect
tion without deactivation. Of course, it is well known that under
anaerobic conditions, methanol can also undergo photocatalytic
reforming to produce hydrogen, according to Eq. (1) [12,45,46].
2
the oxidation state of gold on the 2 wt.% Au/TiO –P25 catalyst.
2
This means that the origin of the produced H is therefore uncer-
4
.2. Photocatalytic water splitting results
tain and may not be issued exclusively from photocatalytic water
splitting. After 10 days of photocatalytic reaction with 10 mL of
methanol (corresponding to 0.248 mol), the total detected amount
Firstly, without the addition of methanol as a sacrificial reagent
the pure water splitting efficiency of different Au/TiO photocata-
lysts, previously characterized, was compared (Fig. 10). In contrast
to Pt co-catalyst, only the TiO –P25 support led to H production.
However, on this TiO –P25 support, the hydrogen production rate
was very low, whatever the samples tested (less than 0.25 mol/
min), and dropped to zero after relatively short periods. Besides,
this decrease in H production is faster on Pt-containing catalyst,
2
of H
produced at best 0.744 mol of H
produced H over 10 days exceeded the total amount that metha-
nol reforming alone could provide (by a factor 2.3), evidencing
2
was 1.728 mol. Nevertheless, methanol reforming could have
2
. Thus, the total amount of the
2
2
2
2
l
2
that a large part of H was undeniably produced by photocatalytic
water splitting.
2
compared to the equivalent gold catalyst. This faster deactivation
is probably due to a more rapid poisoning of the active sites which
may result from oxygen species strongly adsorbed on the surface of
the metallic nanoparticles. Consequently, competitive side reac-
tions are becoming predominant with time on stream: oxygen
CH
3
OH þ H
2
O ! CO
2
þ 3H
2
ð1Þ
Consequently, the major part of methanol acts as an effective sacri-
ficial reagent by scavenging either O or holes, limiting the reverse
2
H + 1/2 O ? H O reaction, and increasing charge separation [47],
2
2
2
ꢀÅ
ꢀÅ
reduction by photogenerated electrons to form O2 and O2 oxida-
tion by photogenerated holes [44]. Nevertheless, comparing the
respectively. In the former case, one may thus consider the photo-
oxidation reaction of methanol, as expressed in Eq. (2).
CH
3
OH þ 3=2O
2
! CO
2
þ 2H
2
O
ð2Þ
0
0
0
0
0
0
0
.30
.25
.20
.15
.10
.05
.00
Assuming that methanol scavenges all the O
2
molecules produced
through Eq. (3), i.e. 0.864 mol, the required quantity of methanol,
according to Eq. (2) is 0.576 mol, which is about 2 times as much
as the volume of the added methanol (0.248 mol).
2
% Au/TiO -Brij56
2
0
1
.3% Au/P25
% Au/P25
H
2
O ! H
þ 1=2O
ð3Þ
2
2
2% Au/P25
As the initial amount of methanol was only 0.248 mol, two explana-
0
.3% Pt/P25
tions can be put forward: (i) H production observed is not exclu-
2
sively the result of water splitting but a part of it comes from
methanol reforming and/or (ii) only a part of methanol is used to
react with the O produced through water splitting, the other
2
part acting as hole scavengers. Table 6 summarizes the products
continuously detected under steady-state conditions. The rates of
O
2
and CO
presence of O
that methanol photocatalytic oxidation does not consume all the
produced can be confirmed. It also proves that H and O are effi-
ciently separated on Pt/anatase and rutile, respectively. It must be
reminded that O is produced by water oxidation by holes (Eq.
2
evolution are constant and non-negligible. From the
2
in the outlet gas, the second hypothesis assuming
0
200
400
600
Time under solar illumination (min)
O
2
2
2
Fig. 10. On stream H
2 2
production from pure water on Au, Pt/TiO –P25, and 2 wt.%
Au/TiO –Brij56 catalysts under artificial solar light illumination.
2
2

1
88
O. Rosseler et al. / Journal of Catalysis 269 (2010) 179–190
Fig. 11. Comparison of hydrogen evolution rate (mol minꢀ1) on 2 wt.% Au/TiO
2 2
–P25 and 0.3 wt.% Pt/TiO –Brij56 photocatalysts as a function of the relative ratio of methanol
sacrificial reagent. Stabilized value over more than 42 h of reaction.
Table 6
reported by Iwase and colleagues that on NaTaO
reaction between O and H was 1500 times less important with
gold nanoparticles as a co-catalyst compared to platinum [47],
(ii) the importance of the contact or interaction of the TiO semi-
conductor with the precious metal, this contact depending on the
nature of the metallic nanoparticles. Platinum is known to intro-
duce ohmic contact facilitating a quick transfer of electrons to
the electrolyte [48]. On the other hand, gold nanoparticles undergo
charging and exhibit ability to store and transport electrons at the
3
:La, the reverse
Reaction products continuously detected on 2 wt.% Au/TiO
the stabilization period under artificial solar light illumination.
2
–P25 photocatalyst during
2
2
Amount of methanol
added (mL)
H
2
(l
mol/min)
O
/
5.4
70
2
(l
mol/min)
CO
2
(
lmol/min)
2
0
1
1
.1
1.1
14
120
0.4
2.8
18
0
2
interfacial semi-conductor (TiO )–metal composite interface. This
charge distribution at the interface results in Fermi level equilibra-
tion of the composite. As already reported by Chen and colleagues
(
4)). Its presence in the exhaust gas also confirms that methanol
does not scavenge all the photogenerated holes.
H
2
O þ 2hþ ! 2Hþ þ O
2
ð4Þ
[
49], as gold nanoparticles possess the particular property to un-
dergo quantitized charging, they are unique candidate to achieve
Fermi level equilibration. The mediating role of Au nanoparticles
Methanol is consumed by two reactions: (i) methanol reforming,
which consumes holes and produces H and CO and (ii) O scav-
enging leading to H O and CO . Nevertheless, the detection of O
proves that under illumination, more holes and O molecules are
produced than methanol (at this concentration) can scavenge.
It must also be mentioned that in the case of methanol reform-
ing, metal sites poisoning by CO is also to be expected, since it is
known that CO strongly adsorbs on noble metals. Nevertheless,
2
2
2
for electron storing and shuttling at the TiO
ing a negative shift in the Fermi level of TiO , has already been evi-
denced by Kamat and colleagues [50]. Under illumination, when
TiO and Au nanoparticles are in contact, the photogenerated elec-
trons are distributed between both kinds of particles, leading to
electron transfer from the excited TiO to Au, until the two systems
–Au interface, induc-
2
2
2
2
2
2
2
2
attain equilibration [49,51]. The electron accumulation on the Au
particles increases its Fermi level to more negative potentials, also
leading to a shift in the Fermi level of the composite closer to the
conduction band of the composite. As the Fermi level shifts to more
negative potential, the charge separation and reductive power of
CO should also easily be oxidized and released as CO
nation in the presence of O
2
under illumi-
2
.
5
. Discussion
the Au–TiO
the photocatalytic reduction of H
2
composite increases, which is of course beneficial to
O into H . It should also not be
From these experiments it is possible to draw a few conclusions
regarding the role of the metallic co-catalyst and the peculiar
behavior of gold nanoparticles supported on TiO anatase–rutile
2
2
forgotten that gold and platinum electrodes are usually reported
to display different electrochemical properties in the generation
2
support, either commercially available or synthesized by poro-
gen-assisted sol–gel method.
of H
2
[52,53].
The influence of the gold nanoparticle size on the H
2
production
It has been clearly evidenced that the highest H
activity was obtained with Au as a co-catalyst deposited on
TiO –P25. Pt-based photocatalysts are less active, whatever the
2
production
efficiency has also been evidenced in our case. As the loading in-
creased from 0.3 to 2 wt.%, the mean particle size increased from
3 to 8 nm, accompanied by an enhancement in the H evolution
2
2
support be. There are two possible and non-exclusive explana-
tions: (i) Pt is very active (more than gold) toward the reverse reac-
rate. This tendency is contrary to what is generally observed about
the size dependence of catalytic properties of gold nanoparticles
deposited on titania [54]. Furthermore, since the energy levels in
gold nanoparticles are discrete, a greater shift in energy is expected
tion H
2
+ ½ O
2
? H
2
O even at room temperature, thus decreasing
production. Indeed, it has already been
the efficiency of H
2

O. Rosseler et al. / Journal of Catalysis 269 (2010) 179–190
189
for smaller size nanoparticles than for the larger ones. The smaller
gold particles thus should have induced a greater shift in the
apparent Fermi level of the Au–TiO composite [51]. This clearly
means that in our case, other factors should also be considered,
in addition to the intimate contact or interaction between the
metallic and the semi-conductor nanoparticles.
60
2
4
0
0
0
2
TiO -P25
2
2
The compared UV–Vis characterizations of TiO -supported gold
and platinum photocatalysts showed that gold deposition shifted
the absorption edge of the photocatalyst, whereas the deposition
of Pt did not affect significantly the absorption spectrum. However,
this better light absorption in the visible region of the spectrum
cannot explain alone the better results obtained with gold-based
0
1
2
3
4
5
6
7
8
9
10
TiO
2
-Brij56
-20
TiO
2
-PEG
catalysts; the nature of the TiO
tance, especially if we consider the poor results obtained on Au/
TiO –Brij56 compared to the results obtained on Au/TiO –P25. In-
deed, as from the results reported in part 3.2, Pt/TiO –Brij 56
2
support is also of great impor-
-40
pH
2
2
2
Fig. 12. Zeta potential of TiO
2
P25, TiO
2
–Brij56, and TiO
2
anatase–PEG supports.
seemed to be the best combination for Pt-based catalysts, one
would have expected the same for Au-based photocatalysts. De-
spite a lower specific surface area, leading theoretically to a poorer
distribution of gold and also to lower accessibility of gold particles
ꢀ
complexes in solution have a high chloride content (AuCl ), the
4
direct anionic exchange is not efficient and the deposited gold
particles are prone to sintering.
We can also suppose that the lower activity observed on the
2
TiO –Brij56 support mainly results from the deposition method
on the support, Au/TiO
based catalysts. This is certainly due to the presence and the nature
of surface –OH groups of the TiO support, as we know that the an-
2
–P25 showed the best activity, among gold-
2
ionic exchange used for gold deposition requires a lot of hydroxyl
groups for an efficient anchoring of gold complexes. XPS studies of
the bare TiO –P25 supports revealed that both the relative amount
2
of surface –OH groups and their chemical environment (or interac-
tion with the surrounding atoms) are different on both kinds of
support. Furthermore, the important decrease in the proportion
of gold complexes and not from the support itself. Changing the
deposition method or using a support with higher anatase/rutile
ratio without modifying the large surface area would thus probably
lead to higher activities.
We have also put forward the very important role of methanol.
Methanol acts as an effective sacrificial reagent by scavenging
2
of surface–OH groups after gold deposition on TiO –P25, but only
either O
2
or holes, limiting the reverse H
2
+ 1/2 O
2
2
? H O reaction,
on this support, evidenced the formation of strong gold com-
plex–surface interactions, required for an efficient gold deposition.
As on the TiO –Brij56 support, the proportion of surface hydroxyl
2
and increasing charge separation. Even if a part of H
by photocatalytic methanol reforming, it has been demonstrated
that most of it is issued from direct photocatalytic water splitting.
2
is produced
groups was less affected by gold deposition, a less efficient deposi-
tion and anchoring of gold complexes can be supposed.
6
. Conclusion
Other points derived from the suspension of particles in a liquid
medium should also be considered: the surface chemistry, the
agglomeration, the sedimentation, and the dispersion of the
nanomaterials in the methanol/water solution and, as a result, a
possible effect on optical properties. Indeed, as the photocatalytic
properties in the liquid phase are sensitive to the state of disper-
sion of TiO -based catalysts because of optical property consider-
2
ations [55], complementary in situ characterization should be
performed.
Efficient H
ting using optimal and very small (1 vol.%) amount of methanol has
been achieved on low content (0.3 wt.%) Au/TiO anatase–rutile
photocatalysts. The addition of methanol is required to reach high
evolution rates but keeping this ratio as low as 1 vol.% seems to be
a good balance between sustainable high efficiency and limiting
the use of fuel-derived and non-renewable resources. The effi-
ciency can be tuned, the main factors affecting H evolution rates
2
are the nature and concentration of the noble metal co-catalyst,
the anatase/rutile ratio of the support, its specific area and poros-
2
production (up to 120 lmol/min) from water split-
2
One of the key factors may also be related to the anatase/rutile
ratio. Since the anatase phase has a higher density of surface hy-
droxyl groups, the anatase/rutile ratio (80/20 and 50/50 for
2
ity. In optimized conditions, interesting H production efficiency
can be maintained over days without deactivation and without
the addition of extra methanol.
2 2
TiO –P25 and TiO –Brij56, respectively) may play an important
role. Furthermore, as already mentioned in the literature [56], it
should also be reminded that the photogenerated active oxygen
species are different for rutile and anatase phases in aqueous solu-
tions containing alcohols as scavenger of holes. But another factor
that we consider determinant is the isoelectric point (IEP), to-
Acknowledgments
The authors thank the French Agence Nationale de la Recherche
gether with the deposition conditions. Indeed, gold complexes in
(
ANR) for supporting this study through the ANR-05-JCJC/0198 re-
ꢀ
solution are negatively charged (½AuðOHÞ Cl1ꢀxꢁ , with 0 < x < 4).
x
search program. P. Bernhardt (LMSPC, Strasbourg) is acknowledged
for performing XPS analyses. T. Dintzer (LMSPC, Strasbourg) and C.
Uhlaq (IPCMS Strasbourg) are acknowledged for performing,
respectively, SEM and TEM analyses.
From an electronic interaction point of view, the support has thus
to be positively charged for the deposition to be efficient. It can be
observed from Fig. 12 that the isoelectric point of the tested
supports decreases in the order TiO
2 2
–P25 (6.7) > TiO –Brij56
(
2.4) > TiO –PEG (1.8). The IEP is closely linked to the density of
2
References
surface hydroxyl groups. Since anatase has a higher density of sur-
face hydroxyl groups than rutile, this decrease in the IEP value
with increasing rutile content is not a surprise. Consequently,
the low IEP of the sol–gel TiO –Brj56 and TiO –PEG supports re-
2 2
quires very low pHs, lower than 2.4 and 1.8, respectively, for an
efficient anchoring of gold particles. Furthermore, at pH < 2.4, gold
[
1] A. Fujishima, K. Honda, Nature 238 (1972) 37.
[2] Z. Zou, J. Ye, K. Sayama, Nature 414 (2001) 625.
[
[
[
3] R. Abe, T. Takata, H. Sugihara, K. Domen, Chem. Comm. (2005) 3829.
4] A. Kudo, Pure Appl. Chem. 79 (2007) 1917.
5] M. Matsuoaka, M. Kitano, M. Takeuchi, K. Tsujimaru, M. Anpo, J.M. Thomas,
Catal. Today 122 (2007) 51.

1
90
O. Rosseler et al. / Journal of Catalysis 269 (2010) 179–190
[
6] T. Kida, G. Guan, A. Yushida, Chem. Phys. Lett. 371 (2003) 563.
7] A.L. Linsebigler, G. Lu, J.T. Yates Jr., Chem Rev. 95 (1995) 735.
8] P. Pichat, New J. Chem. 11 (1987) 135.
9] G.R. Bamwenda, S. Tsubota, T. Nakamura, M. Hurata, J. Photochem. Photobiol.
A: Chem. 89 (1995) 177.
[31] C.H. Lin, C. Lee, J.H. Chao, C. Kuo, Y. Cheng, W. Huang, H.W. Chang, Y. Huang,
M.K. Shih, Catal. Lett. 98 (2004) 61.
[32] T. Sreethawong, S. Yoshikawa, Int. J. Hydrogen Energy 31 (2006) 786.
[33] A. Abe, K. Sayama, K. Domen, H. Arakawa, Chem. Phys. Lett. 344 (2001) 339.
[34] L. Shi, D. Weng, J. Envir. Sci. 20 (2008) 1263.
[
[
[
[
[
10] M.C. Blount, J.A. Buchholtz, J.L. Falconer, J. Catal. 197 (2001) 303.
11] A. Patsoura, D.I. Kondarides, X.E. Verykios, Appl. Catal. B: Environ. 64 (2006)
[35] C.-Y. Wang, R. Pagel, J.K. Dohrmann, D.W. Bahnemann, Compt. Rend. Chim. 9
(5–6) (2006) 761.
1
71.
[36] T. Ohno, K. Sarukawa, K. Tokieda, M. Matsumura, J. Catal. 203 (2001) 82.
[37] T. Ohno, K. Tokieda, S. Higashida, M. Matsumura, Appl. Catal. A 244 (2003)
383.
[
[
[
[
12] L.S. Al-Mazroai, M. Bowker, P. Davies, A. Dickinson, J. Greaves, D. James, L.
Miljard, Catal. Today 122 (1–2) (2007) 46.
13] J.G. Highfield, M.H. Chen, P.T. Nguyen, Z. Chen, Energy Environ. Sci. 2 (2009)
[38] K. Komaguchi, H. Nakano, A. Araki, Y. Harima, Chem. Phys. Lett. 428 (2003)
338.
[39] D.C. Hurum, A.G. Agrios, K.A. Gray, T. Rajh, M.C. Thurnauer, J. Phys. Chem. B
107 (2003) 4545.
[40] S. Link, M.B. Mohamed, M.A. El-Sayed, J. Phys. Chem. B 203 (1999) 3073.
[41] P.V. Kamat, J. Phys. Chem. B 206 (2002) 7729.
9
91.
14] A. Sclafani, M.-N. Mozzanega, P. Pichat, J. Photochem. Photobiol. A: Chem. 59
1991) 181.
15] N. Strataki, V. Bekiari, D.I. Kondarides, P. Lianos, Appl. Catal. B: Environ. 77 (1–
) (2007) 184.
(
2
[
[
[
[
[
[
16] M. Ashokkumar, Int. J. Hydrogen Energy 23 (1998) 427.
17] H.-J. Choi, M. Kang, Int. J. Hydrogen Energy 32 (16) (2007) 3841.
18] J.R. Bolton, Solar Energy 57 (1996) 37.
19] T. Sakata, J. Photochem. 29 (1985) 205.
20] F. Bosc, A. Ayral, N. Keller, V. Keller, Appl. Catal. B: Environ. 69 (2007) 133.
21] F. Bosc, D. Edwards, N. Keller, V. Keller, A. Ayral, Thin Solid Films 495 (2006)
[42] G. Marci, M. Addamo, V. Augugliaro, S. Coluccia, E. Garcia-Lopez, V. Loddo, G.
Marta, L. Palmisano, M. Schiavello, J. Photochem. Photobiol. A 160 (2003) 105.
[43] A. Mills, G. Porter, J. Chem. Soc. Faraday 78 (1998) 3659.
[44] T. Kawai, T. Sakata, J. Chem. Soc. Chem. Comm. (1980) 694.
[45] P. Pichat, J.-M. Herrman, J. Disdier, H. Courbon, M.-N. Mozzagena, Nouv. J.
Chem. 5 (1981) 627.
2
72.
[46] M. Zalas, M. Laniecki, Sol. Energy Mater. Sol. Cells 89 (2005) 287.
[47] A. Iwase, H. Kato, A. Kudo, Catal. Lett. 108 (1–2) (2006) 7.
[48] D.E. Aspnes, A. Heller, J. Phys. Chem. 87 (1983) 4919.
[49] S. Chen, R.W. Murray, J. Phys. Chem.B. 103 (1999) 9996.
[50] V. Subramanian, E.E. Wolf, V. Kamat, T. J. Am. Chem. Soc. 126 (15) (2004) 4943.
[51] S. Chen, R.S. Ingram, M.J. Hostetler, J.J. Pietron, R.W. Murray, T.G. Schaaf, J.T.
Khoury, M.M. Alvarez, R.L. Whetten, Science 280 (1998) 2098.
[52] M.W. Breiter, in: W. Vielstich, H.A. Gasteiger, A. Lamm (Eds.), Handbook of
Fuel Cells – Fundamentals, Technology and Applications, vol. 2, John Wiley and
Sons, 2003, p. 4.
[
[
22] S. Ivanova, V. Pitchon, C. Petit, J. Mol. Catal. A: Chem. 256 (1–2) (2006) 278.
23] N. Keller, E. Barraud, F. Bosc, D. Edwards, V. Keller, Appl. Catal. B: Environ. 70
(
2007) 423.
[
[
24] F. Bosc, A. Ayral, P.-A. Albouy, C. Guizard, Chem. Mater. 15 (2003) 2463.
25] Y. Kotani, A. Matsuda, M. Tatsumisago, T. Minami, J. Sol–Gel Sci. Technol. 19
(
2000) 133.
[
26] D. Fattakhova-Rohling, M. Wark, T. Brezesinski, B.M. Smarsky, J. Tathousky,
Adv. Funct. Mater. 17 (2007) 123.
27] T. Klimnova, E. Carmona, J. Ramirez, J. Mater. Sci. 33 (1998) 1981.
28] H. Liu, J. Yuan, W.F. Shangguan, Energy Fuel 20 (2006) 2289.
29] V. Keller, P. Bernhardt, F. Garin, J. Catal. 215 (2003) 129.
30] O. Ishitani, C. Inoue, Y. Suzuki, T. Ibusuki, J. Photochem. Photobiol. A 72 (1993)
[
[
[
[
[53] O. Petrii, G. Tsirlina, Electrochimico Acta 39 (11) (1994) 1739.
[54] M. Haruta, Catal. Today 36 (1997) 153.
[55] T.A. Egerton, I.R. Tooley, J. Phys. Chem. B. 108 (2004) 5066.
[56] H. Goto, Y. Hanada, T. Ohno, M. Matsumura, J. Catal. 225 (2004) 223.
2
69.