Applied Catalysis B: Environmental 90 (2009) 213–223

Contents lists available at ScienceDirect

Applied Catalysis B: Environmental

journal homepage: www.elsevier.com/locate/apcatb

Highly active CoMoS phase on titania nanotubes as new

hydrodesulfurization catalysts
J.A. Toledo-Antonio *, M.A. Cortés-Jácome, C. Angeles-Chávez, J. Escobar,
M.C. Barrera, E. López-Salinas
Molecular Engineering Program, Instituto Mexicano del Petróleo, Eje Central Lázaro Cárdenas # 152, San Bartolo Atepehuacan, G. A. Madero, 07730 México, D.F., Mexico

A R T I C L E I N F O A B S T R A C T

Article history: A series of CoMoS on nanotubular TiO2 (NT) catalysts were prepared by incipient impregnation at
Received 1 September 2008 Mo = 3–5 atoms/nm2 and Co/(Co + Mo) = 0.2–0.4. A nanotubular TiO2 (NT) of 5.5 nm inner diameter,
Received in revised form 4 March 2009 236 m2/g surface area and 0.5 cm3/g total pore volume was prepared and used as support. The CoMoS/NT
Accepted 9 March 2009
catalysts were characterized mainly by Raman and XPS spectroscopies, HRTEM and HAADF (high angle
Available online 27 March 2009
annular dark field) detector. The sulfided catalysts were tested in the hydrodesulphurization (HDS) of
dibenzothiophene at 320 8C and 56 kg/cm2. Inorganic fullerene-like MoS2 and/or CoMoS particles, made
Keywords:
up of 2–6 structural layers, depending on the Co/(Co + Mo) ratio, were observed, smeared along the
HDS catalysts
CoMoS phase
nanotubes. An estimation of CoMoS phase surface density (d), after taken into account all sulfided
Titania nanotubes species quantified by XPS spectra, indicated that highest HDS activity (43  104 m3/kgMo s) is attained
X-ray photoelectronic spectroscopy when Mo = 5 atoms/nm2 and Co/(Co + Mo) = 0.3. In comparison, at a lower (0.2) or higher (0.4) Co/
Raman Spectroscopy (Co + Mo) ratio, either segregated MoO3 or Co9S8 phases were detected, respectively, with Co
High resolution electron microscopy sulfidability decreasing 60–42%, respectively, and HDS activity being considerably lower.
ß 2009 Elsevier B.V. All rights reserved.

1. Introduction formation of highly dispersed CoMoS type 2 sites may result in

highly active CoMo sulfided catalysts. However, CoMoS phase
The more stringent environmental regulations concerning formation is not an easy task, and strongly depends on the
maximum sulfur content in liquid fuels, e.g., <15 ppm, have been interaction degree between Mo oxide (e.g., precursor to MoS2) and
the driving force for oil refineries to produce ultra-low sulfur fuels the support. In alumina-supported catalysts for example, the
at affordable cost, upgrading existing technologies and developing surface is often modified with additives such as phosphorous,
new processes and more active hydrodesulfurization (HDS) fluorine or boron [5,6] to adequately obtain improved Mo oxides
catalysts. The application of the later could enhance the dispersion, and in order to control morphology, size and stacking
productivity and improve the product quality without negative number of MoS2 particles.
impact on capital investment. Other strategies applied to generate highly active CoMoS type 2
Industrially, HDS reaction is carried out on CoMoS/Al2O3 or sites involve replacing the conventional alumina support by other
NiMoS/Al2O3 catalysts. The properties of the sulfided catalyst metal oxides showing less interaction with deposited Mo oxides
strongly depend on the interaction with the support and on the such as SiO2 [7], TiO2 [8,9], ZrO2 [10] or mixed oxides [11,12]. Thus,
promotion degree of MoS2 by the adjacent Co and Ni sulfides great interest is being devoted to study the effect of the support on
yielding the so called ‘‘CoMoS’’ or ‘‘NiMoS’’ phases, [1,2] which are the HDS catalytic activity of CoMoS and NiMoS sulfided catalysts.
considered to contain the active sites for HDS reactions. It has been One of the most interesting supports for HDS catalysts is titania,
reported that coordinatively unsaturated sites occur at edges and since a considerable increase in intrinsic HDS activity has been
corners of MoS2 crystallites. For alumina-supported catalysts the reported for MoS2 when supported on anatase TiO2, in comparison
sites have been classified as type 1, when present on monolayer with a similar composition on an alumina carrier [8,13]. However,
MoS2 slabs, and type 2, when in highly stacked MoS2 particles [3]. the TiO2 low specific surface area (SSA) (e.g., <60 m2/g) limits its
CoMoS type 2 sites are about twice as active as type 1 [4]. Then, the industrial application as HDS catalyst support. Recently, TiO2 with
relatively high SSA (140–175 m2/g) have been developed, with
good enough textural properties to efficiently disperse CoMoS
* Corresponding author. phase [14,15]. These improved formulations may have a beneficial
E-mail address: jtoledo@imp.mx (J.A. Toledo-Antonio). impact on the production of ultra-low sulfur diesel (ULSD).

0926-3373/$ – see front matter ß 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.apcatb.2009.03.024
214 J.A. Toledo-Antonio et al. / Applied Catalysis B: Environmental 90 (2009) 213–223

In the last decade, anatase TiO2 has been converted into titania 380 8C, under a 10 vol.% H2S/H2 flow for 2 h. This step was carried
nanotubes or nanofibers through a relatively simple alkaline out in a Pyrex glass tubular down-flow reactor, the H2S/H2 mixture
hydrothermal method [16–19], representing a suitable route for being fed at a 24 L/h g constant flow rate. This routine could be
rapid manufacturing scale-up. The nanotubes exhibit large internal considered as standard in activating HDS catalysts [9,12].
and external surfaces, along with surface in the vertex and in the
interlayer regions that compose the nanotube walls [20,21]. In fact, 2.3. HDS catalytic test
the transformation of titania into nanotubes yields materials with
SSA as large as 400 m2/g [17,22], with high pore volume and inner Sulfided catalysts (0.2 g) were tested in HDS of dibenzothio-
diameters around 7 nm, opening possibilities to disperse a larger phene (DBT, 0.3 g), model molecule representing S-bearing
amount of transition metal oxides or sulfides. In a recent work, we compounds in middle distillates, in a SS316 batch reactor using n-
have shown that titania nanotubes can disperse efficiently 20 wt.% hexadecane (100 ml) as solvent (initial DBT concentration
of WO3 maintaining the nanotubular morphology and SSA being as 16.3 mmol/L, 674 ppm S). The reaction temperature was fixed
high as 210 m2/g after annealing at 400 8C [23]. at 320 8C and stirring rate of 1000 rpm. After operating
Present and future trends in commercial catalysts develop- temperature was attained, pure H2 (UHP) was fed to the reactor
ments for ULSD production, point out towards the increase of to attain 56 kg/cm2 (total pressure). Samples of liquid product
active phase loading from 3 Mo atoms/nm2, for conventional were analyzed in a Varian 3400 CX gas chromatograph, equipped
catalysts, to around 5–8 atoms/nm2 [24]. Recently, we have with a flame ionization detector and a dimethylpolysiloxane
optimized the impregnation procedure by adjusting the pH of (50 m  0.2 mm  0.5 mm) capillary column. HDS kinetic con-
the Co and Mo solutions at 5.6, for samples with Mo loading of stants were calculated assuming pseudo-first order kinetics
3 atoms/nm2 [25]. In this work, the effect of metal loading on referred to DBT concentration (x = conversion, t = time), taking
nanotubular titania support was studied by increasing Mo content into account that hydrogen was fed in such excess that its partial
from 3 to 5 atoms/nm2. As well, the (Co/Co + Mo) atomic ratio was pressure could be considered constant:
optimized in samples of higher metal loading that showed high
HDS activity. The influence of the nanotubular morphology of the lnð1  xÞ
k¼ (1)
support on the CoMoS particles was thoroughly examined by t
HRTEM. k values for various catalysts were normalized by considering
reaction volume and mass of catalyst used (k in m3/(kgcat s)). Even
2. Experimental though a Langmuir–Hinshelwood protocol could provide impor-
tant information about reaction kinetics, and that comparing HDS
2.1. Synthesis of nanotubular titania catalysts performance through a pseudo-first order kinetic model
could be considered an oversimplification, the later has been
Nanotubular titania, used as support, was synthesized by a recognized as a reliable method in discriminating among various
hydrothermal method, starting from a TiO2 anatase (Hombitec HDS catalytic formulations [12,15].
K03, provided by Sachtleben Chemie). TiO2 anatase powder was
suspended in an aqueous 10 M NaOH solution and the resulting 2.4. Catalysts characterization
suspension was placed in an autoclave. The hydrothermal reaction
was carried out at 100 8C, for 24 h under autogenous pressure and Chemical composition of the impregnated materials after
stirring. Thereafter, the white slurry was filtered and neutralized annealing at 380 8C was determined by atomic absorption
with a 1 M solution of nitric acid (HNO3) overnight. Then, the spectroscopy (AAS) in a PerkinElmer 2380 apparatus.
material was washed with deionized water and dried at 110 8C Textural properties were measured in an ASAP-2000 analyzer
yielding a hydrous titania powder with nanotubular morphology. from Micromeritics. SSA was calculated from N2 physisorption at
196 8C using the Brunauer–Emmet–Teller (BET) equation.
2.2. Catalysts preparation Average pore size was obtained by the Barrett–Joyner–Halenda
(BJH) method in the desorption stage. Materials calcined ex situ at
CoMo-based catalysts were prepared by incipient wetness 380 8C, were outgassed at 350 8C prior to measurements.
impregnation in two steps. In the first step, nanotubular titania Raman spectra were obtained using a Yvon Jobin Horiba
(NT) was put in contact with an aqueous ammonium heptamo- (T64000) spectrometer, equipped with a CCD camera detector. As a
lybdate (AHM) solution at pH 5.6 in order to have 3, 4 and 5 Mo source of excitation the 514 nm line of a Spectra Physics 2018
atoms/nm2, as indicated in Table 1. In the second step and after Argon/Krypton Ion Laser system were focused through an Olympus
drying at 110 8C, Co (through a cobalt acetate aqueous solution) BX41 microscope equipped with a 100 magnification objective.
was impregnated on the Mo-containing samples. After impreg- The laser power never exceeded 5 mW on the sample.
nation, further processing included drying at 120 8C and annealing Transmission electron microscopy (TEM) and scanning trans-
at 380 8C under air flow. Samples were labeled as NT-x-y, where NT mission electron microscopy (STEM) were performed both in a
stands for nanotubular titania support, ‘‘x’’ being the nominal JEM-2200FS microscope with accelerating voltage of 200 kV. The
amount of Mo atoms/nm2 and ‘‘y’’ being the Co/(Co + Mo) atomic microscope is equipped with a Schottky-type field emission gun
ratio. Catalysts sulfidation was carried out on oxidized samples at and an ultra high resolution configuration (Cs = 0.5 mm;

Table 1
Nominal and real chemical compositions of catalysts.

Catalysts Monom Co/(Co + Mo)nom Moreal (wt.%) Coreal (wt.%) Moreal Co/(Co + Mo)real
(atoms/nm2) (atomic ratio) (atoms/nm2) (atomic ratio)

NT-3-0.3 3 0.3 12.83 3.13 3.52 0.28

NT-4-0.3 4 0.3 13.50 4.73 3.98 0.36
NT-5-0.3 5 0.3 14.98 4.42 4.60 0.32
NT-5-0.2 5 0.2 15.14 3.23 5.03 0.26
NT-5-0.4 5 0.4 15.40 7.2 5.68 0.43
 J.A. Toledo-Antonio et al. / Applied Catalysis B: Environmental 90 (2009) 213–223 215

Cc = 1.1 mm; point-to-point resolution = 0.19 nm) and in-column 3.1. Chemical structure of calcined CoMo/NT catalysts
omega-type energy filter. STEM is particularly useful in nanopar-
ticles studies by using a high angle annular dark field (HAADF) Raman spectra of support and calcined catalysts are shown in
detector, which collects electrons that undergo Rutherford Fig. 1. It is worth starting with the characteristics of the single
scattering; thus, the image can be acquired where the intensity support before the incorporation of Mo and Co where the typical
is approximately proportional to Z2 (Z being the atomic number of bands of anatase were evident on the support calcined at 400 8C,
the scattering atom). Therefore, elements with a high Z show particularly the band at 156 cm1 corresponding to the E1g
higher intensities and a white contrast in the image should occur. vibration. [28]. Other bands corresponding to anatase phase at 200,
This technique is useful to distinguish the presence of different 400, 505, and 640 cm1 were less intense and broad on the
chemical elements, when there is a large difference among their impregnated samples, merging with those of nanotubular struc-
atomic numbers, such as in supported catalysts. ture at 278, 450 and 700 cm1 [17,29]. Then, after calcination the
Local chemical analysis by energy dispersive X-ray spectro- structure of the nanotubular support is made up of a mixture of
scopy (EDXS) was performed in a NORAN energy dispersive X-ray anatase and orthorhombic layered phases, suggesting the forma-
spectroscope, which is attached to the microscope using the STEM- tion of anatase domains in some regions of the nanotubes walls, as
EDX combination. The samples were ground, suspended in reported elsewhere [22]. Apparently, the incorporation of CoMo on
isopropanol at room temperature, and dispersed with ultrasonic nanotubes stabilizes the orthorhombic structure of layered
agitation, then, an aliquot of the solution was dropped on a 3 mm titanates with nanotubular morphology against its transformation
in diameter lacey carbon copper grid. into anatase. In fact, the E1g vibrating mode of anatase decreased
X-ray photoelectronic spectroscopy (XPS) on sulfided catalysts in intensity as the amount of Mo and Co loaded increased. Beside
was determined in a THERMO-VG SCALAB 250 spectrometer the E1g band, two broad bands were observed around 660 and
equipped with AlKa X-ray source (1486.6 eV) and a hemispherical 935 cm1. The broad peak at 935 cm1, is likely to be the vibrating
analyzer. Experimental peaks were decomposed into components mode of terminal Mo = O in tetrahedral coordination [30]. The
using mixed Gaussian–Lorentzian functions and a non-linear intensity of the broad band around 660 cm1 increased consider-
squares fitting algorithm and a Shirley background subtraction ably compared with those at 278 and 450 cm1, then, the major
was applied. An area ratio of 2/3 and a splitting of 3.2 eV were used contribution to the band 660 cm1 comes from the oxidized CoMo
to fit the Mo 3d peaks. Binding energies were reproducible to phase, suggesting the presence of a well dispersed b-CoMoO4
within 0.2 eV and the C 1s peak at 284.6 eV was used as a reference phase [31,32] in all the samples, regardless of the amount of Mo
from adventitious carbon. Surface elemental composition was loaded (at constant Co/(Co + Mo) = 0.3 atomic ratio).
determined by fitting and integrating the Co 2p, Mo 3d, S 2p, O 1s The optimization of the Co/(Co + Mo) atomic ratio was
and Ti 2p bands and converting these values to atomic ratios using examined on a high Mo loaded catalyst (e.g., 5 atoms/nm2). The
theoretical sensitive factor provided by the manufacturer of the XPS Mo species loaded on the NT support strongly depend on the
apparatus [26]. All XPS curves are shown with the differential fitting amount of Co loaded, as shown in Fig. 2. At Co/(Co + Mo) = 0.2,
curve (lowest curve in each spectra) after adjusting to theoretical additional to the broad bands at 660 and 935 cm1, two narrow
curves. All manipulation of sulfided samples was done under inert bands were observed at 815 and 990 cm1, corresponding to the
conditions using a glovebox and a special box vessel, attached to the Mo–O–Mo vibration and to terminal Mo = O in MoO3 particles,
XPS equipment, to introduce samples into the ultra high vacuum respectively [33,34]. Other less intense bands at 125, 243, 288 and
chamber of the spectrometer. 335 cm1 confirm the presence of MoO3 phase. At higher Co
loading, (sample NT-5-0.4), the intensity of the band at 935 cm1
3. Results and discussion increased considerably, in comparison with that in NT-5-0.3,
suggesting a well defined CoMoO4 phase or an increase on the
Chemical compositions determined by AAS on the calcined crystallite dimension of CoMoO4 phase dispersed on this sample.
catalysts are shown in Table 1. In all cases real Mo and Co These results can be rationalized as follows: At low Co loading,
concentrations were quite close to the nominal ones. (e.g., NT-5-0.2), there are not enough Co atoms to react with all
Initially nanotubular support calcined at 400 8C showed SSA of loaded Mo, then, a fraction of Mo segregates as MoO3, while the
236 m2/g (Table 2). After impregnation and calcination, SSA and rest reacts with Co forming CoMoO4 phase. At high Co loading (e.g.,
pore volume of the materials decreased slightly in comparison
with the single support because of the impregnation of non-porous
CoMo precursors and also because of structural changes of the
carrier after calcinations [27]. As expected, the catalysts of higher
Co and Mo concentration showed lower pore volume and pore size.
From the SSA and real Mo loading, the real surface concentration
after calcination was between 3.5 and 5.68 atoms/nm2, as
indicated in Table 1.

Table 2
Textural properties and real Mo surface concentration of supports and
catalysts after calcinations.

Support/catalysts SSA (m2/g) TPV (cm3/g) APS (nm)

a
NT 236 0.52 6.8
NT-3-0.3 229 0.45 6.6
NT-4-0.3 213 0.38 5.8
NT-5-0.3 204 0.39 7.3
NT-5-0.2 189 0.35 6.1
NT-5-0.4 170 0.32 6.2
a
Annealed at 400 8C. SSA: specific surface area, TPV: total pore volume, Fig. 1. Raman spectra of support and calcined CoMo/NT catalysts, with different Mo
APS: Average pore size. loadings at constant Co/(Co + Mo) = 0.3 atomic ratio.
216 J.A. Toledo-Antonio et al. / Applied Catalysis B: Environmental 90 (2009) 213–223

3.2. MoS2 particles morphology

HRTEM technique has been used to study the orientation of

MoS2 crystallites on catalyst supports, mostly on g-Al2O3, widely
used as HDS support. As is shown in Fig. 3a, nanotubular features of
NT support were maintained after CoMo impregnation in NT-5-0.3
and after annealing at 380 8C. Small white dots dispersed
homogeneously on the walls of the nanotubes are observed in
the HAADF image of Fig. 3b. As the HAADF detector collect
electrons, that experiments Rutherford scattering, then, the white
dots on the walls of TiO2 nanotubes correspond to Co and Mo in the
CoMoO4 phase. The reason is that the scattered electron intensity
from the sample is directly proportional to square of atomic
number. So, the scattered electrons from the Co and Mo atoms are
intenser than those from Ti and O atoms. This phenomenon was
previously reported for tungsten nanoparticles deposited on a
similar NT-support [23]. The size of these nanoparticles was
Fig. 2. Raman spectra of calcined CoMo/NT catalysts, at constant Mo loading measured on several white dots by using the intensity profiles and
5 atoms/nm2, with different Co/(Co + Mo) atomic ratio.
it was below 3 nm.
In NT-5-0.2 (see Fig. 3c), with low Co concentration, two kinds
of Mo species were observed: One corresponds to Co and Mo atoms
NT-5-0.4), there are more Co atoms available to react with Mo, and in CoMoO4 nanoparticles homogeneously smeared on the surface
larger or well defined crystallites of CoMoO4 phase are produced. of the nanotubes, size below 3.0 nm, as determined in sample NT-
The optimum Co/(Co + Mo) atomic ratio was fixed at 0.3, where 5-0.3. The second one corresponds to large particles (see inset in
well-dispersed CoMoO4 phase was obtained. Then, using the Fig. 3c), which were mainly made up by molybdenum atoms, as
HRTEM technique the dispersion of the mixed oxide phase determined by EDXS analysis (not shown). Therefore, these large
CoMoO4, was studied in detail in the following electron micro- particles correspond to well crystallized and segregated MoO3 in
scopy analysis. agreement with our previous Raman spectroscopy results (see

Fig. 3. (a) Bright field TEM image of oxidized CoMo/NT: NT-5-0.3. HAADF image of oxidized CoMo/NT in (b) NT-5-0.3, (c) NT-5-0.2 and (d) NT-5-0.4.
 J.A. Toledo-Antonio et al. / Applied Catalysis B: Environmental 90 (2009) 213–223 217

Fig. 2). When the amount of Co increased (e.g., in NT-5-0.4), where starting morphology of the support. The nanotubes are composed
more crystalline CoMoO4 phase was detected by Raman spectro- by 3 or 4 structural layers, with an inner diameter of 5.5 nm and an
scopy (see Fig. 2), the nanotubular support appeared embedded outer one of about 10 nm. After impregnation of the CoMoS
into a polymeric matrix constituted of CoMoO4 phase. Notice that precursors, the nanotubular morphology of support remained
the whole scanning area in the HAADF image (Fig. 3d) was unchanged, with MoS2 slabs distributed along the nanotubes, as
highlighted in a white polymeric matrix in which the nanotubes of shown in Fig. 4b–f. At low Mo loading (e.g., NT-3-0.3) in Fig. 4b,
TiO2 are embedded, as determined by EDXS analysis (not shown). short MoS2 slabs composed by just two layers were observed. As
Structural and morphological features of the single NT support expected, the density of the MoS2 layered structure increased with
have been discussed in detail elsewhere [22,35]. But a represen- Mo loading in NT-4-0.3, as shown in Fig. 4c. However, no indication
tative HRTEM image is presented in Fig. 4a in order to set forth a of a higher stacking level than two was observed at this Mo loading.

Fig. 4. HRTEM images of (a) NT support calcined at 400 8C. TEM images of (b) NT-3-0.3, (c) NT-4-0.3, (d) NT-5-0.3, inset A higher magnification of MoS layers where the layers
number are displayed in intensity profile, (e) NT-5-0.2 and (f) NT-5-0.4.
218 J.A. Toledo-Antonio et al. / Applied Catalysis B: Environmental 90 (2009) 213–223

When Mo concentration increased to 5 atoms/nm2 in NT-5-0.3, the layers and high stacking in the catalyst of high Mo loading were
stacking level of MoS2 slab increased between 3 and 6 structural observed, as shown in Fig. 5a–c. Also in this case, markedly curved
layers with an interlayer spacing of 0.61 nm, as indicated in structures were detected indicating a low interaction of Mo species
intensity profile inset A, in Fig. 4d. The length of the MoS2 slabs also with the nanotubular support, as observed in commercial type 2
increased considerably in comparison with NT-3-0.3 and NT-4-0.3, CoMoS/or NiMoS/Al2O3 catalysts [24]. Sakashita et al. observed
as depicted in Fig. 4b and c, respectively. Similar MoS2 features edge-bonded MoS2 crystallites species on anatase, and attributed
were observed for NT-5-0.2 and NT-5-0.4 samples, Fig. 4e and f, the cause to an epitaxial relationship between MoS2 and the
although in some regions, the long MoS2 slabs were curved anatase surface [36]. These observations indicate that the
producing onion-like or inorganic fullerene-like nanostructures, orientation and morphology of the MoS2 catalysts strongly depend
see Fig. 5, as those obtained after sulfiding bubbles of MoO3 at on the crystal plane on the surface of the support [36,8]. However,
800 8C, as reported by Tenne [20,21]. A magnified HRTEM image of the study of Sakashita et al. was carried out on a 3.3 wt.% Mo
a curved structure of MoS2 slabs is shown in Fig. 5b. Large MoS2 catalyst [36]; a low amount in comparison with industrial catalysts

Fig. 5. HRTEM images. (a) NT-5-0.2, (b) NT-5-0.3, (c) NT-5-0.4, (d) high magnification image of NT-5-0.4, (e) high magnification of region A of (d) and its Fourier transform and
(f) HAADF image of NT-5-0.3 sample showing a tridimensional image of almond seed like particles.
 J.A. Toledo-Antonio et al. / Applied Catalysis B: Environmental 90 (2009) 213–223 219

(e.g., 12 wt.%) which require higher concentrations of Mo to

attain better HDS efficiency. Likely, higher Mo loadings may lead to
larger MoS2 slabs lying on their basal planes.
Additionally, some isolated regions of fullerene-like MoS2 were
observed in the catalysts of the highest Mo content. The density of
these nanostructures markedly depended on the Co/(Co + Mo)
atomic ratio, as shown in Fig. 5. At Co/(Co + Mo) = 0.2, some
fullerene-like structures with 3 or 4 stacking level were observed
in Fig. 5a, marked by black arrows. At this Co content, a large
amount of MoO3 was segregated as indicated by our Raman results
in Fig. 2. However, MoO3 particles remained well dispersed on the
nanotubular support. After sulfidation, MoO3 nanoparticles
transformed into fullerene-like nanostructures. At higher Co
loading (e.g., NT-5-0.3), these isolated MoS2 nanostructures were
less frequently observed, as shown in Fig. 5b with fullerene-like
structures made up of 3 or 4 layers; as shown in Fig. 2, in this
sample CoMoO4 phase was highly dispersed. Apparently, the outer
layer is dislocated and became amorphous, which could be an
indication of the Co inclusion on the edges of MoS2 slabs forming
then the CoMoS phase, black arrows in Fig. 5b. In NT-5-0.4, where
well defined CoMoO4 phase was observed by Raman (see Fig. 2), a
large amount of fullerene-like MoS2 particles composed by 4–6
slabs were detected (see Fig. 5c and d). It is well known that
sulfidation of CoMoO4 phase generates into separate Co9S8 and
MoS2 phases, and null or low promotion effect of Co on the MoS2
phase is attained to generate the CoMoS phase [37,38]. A similar
result was found on NT-5-0.4 where Co-rich Mo particles were
present in the fullerene-like MoS2 particles as illustrated in Fig. 5d,
inset A. HRTEM image was obtained from the region marked by
‘‘inset A’’ in Fig. 5d and it is displayed in Fig. 5e, where 0.601, 0.358
and 0.298 nm interplanar distances and derived Fourier transform
(FT) were measured. Crystallographic analysis from FT indicated
that these distances correspond to the (1 1 1), (2–2 0) and (3–1 1)
planes of the cubic Co9S8 phase according to the JCPDS card No. 86-
2273. It should be noted that in this case the Co/Mo molar ratio of
oxidized precursor phase (0.66) is not very far from the
stoichiometric for the CoMoO4 phase. Sulfidation of this phase
results in segregated Co9S8 crystallites and curved MoS2 structures
are mainly formed, as previously reported for unsupported
sulfided CoMoO4 phase [37]. In NT-5-0.3, the segregation of
Co9S8 phase was not observed. In this case, the MoS2 nanoparticles
had ellipsoidal morphology yielding almond seed-like MoS2
nanoparticles, with 4 or 6 curved slabs, and interlayer distance
of 0.614 nm as shown in Fig. 5b. The top layer of MoS2 structure
appeared interrupted or dislocated suggesting that Co atoms are
located at the edge sites of MoS2 particles. Accordingly, HAADF–
STEM have been used to study the change in morphology on
triangular clusters containing only a S–Mo–S layer with addition of
promoting atoms like Co or Ni [39,40], where a triangular
morphology of MoS2 clusters showed higher truncation degree
when Co atoms were added exposing extended high index [1120]
truncation. A high resolution HAADF image of a Co–Mo–S
nanoparticle in NT-5-0.3 is showed in Fig. 5f. As aforementioned,
the sulfided nanoparticle has an almond seed-like morphology,
composed by several MoS2 slabs.
Fig. 6. Representative XPS of: Mo 3d spectrum from CoMo/NT Sulfided samples.
3.3. Sulfidability of surface species in CoMoS/NT catalysts

XPS spectra from Mo 3d, Co 2p and S 2p signals for the different doublets arise from two satellite signals, as shown in Fig. 7, in
NT catalysts are plotted in Figs. 6–8. The Mo 3d signal was fitted by agreement with literature elsewhere [43–45].
considering the following species: (i) Mo4+ from completely The surface composition, binding energy (BE) and full width at
sulfided MoS2, (ii) Mo5+ from molybdenum oxy-sulfides and (iii) half maximum (FWHM) values of the different components are
Mo6+ from unreduced molybdenum oxide [41,42]. reported in Tables 3 and 4, respectively. FWHM values around
Co 2p spectra was fitted with four doublets: (i) The one 1.0 eV suggest that the XPS signal is made up of only one
occurring at 778.5–779.7 from sulfided Co–S, (ii) another one at contribution or oxidation state of each component. As can be
781.5–781.9 from oxidized Co–O and (iii) the other two broad noticed, total Mo and Co dispersion increased along Mo loading.
220 J.A. Toledo-Antonio et al. / Applied Catalysis B: Environmental 90 (2009) 213–223

Fig. 8. Representative XPS of: S 2p spectrum from CoMo/NT Sulfided samples.

When increasing Mo loading from 3 to 5 atoms/nm2 and keeping a

constant Co/(Co + Mo) ratio at 0.3, sulfided Mo4+ increased from
5.8 to 11.6 at.% and sulfided Co–S increased from 1.3 to 2.7 at.%,
respectively; both components increased by a factor of 2.3,
suggesting that, even in samples of high metal loading, Co and
Mo sulfided species are efficiently dispersed on the nanotubular
Fig. 7. Representative XPS of: Co 2p spectrum from CoMo/NT Sulfided samples. support. The effect of the Co/(Co + Mo) atomic ratio was studied in
the catalysts of the highest Mo loading where sulfided Mo4+ was
13.6 and 15.0 at.%, as indicated in Table 3. The Co–S concentration
strongly depended on the Co/(Co + Mo) ratio, showing a maximum
for the sample where that ratio was 0.3 (NT-5-0.3).
 J.A. Toledo-Antonio et al. / Applied Catalysis B: Environmental 90 (2009) 213–223 221

Table 3
Atomic chemical composition (as determined by XPS) of sulfided catalysts.

Catalyst Surface chemical composition (at.%)

S Ti Cotot Co–Sa Co–Ob Motot Mo4+ Mo5+ Mo6+

NT-3-0.3 17.8 27.9 2.83 1.34 1.49 8.7 5.79 1.38 1.53
NT-4-0.3 15.6 22.5 4.77 2.54 2.23 12.93 10.45 1.62 0.86
NT-5-0.3 16.8 19.3 4.06 2.76 1.30 14.24 11.64 1.23 1.41
NT-5-0.2 17.4 18.3 3.81 1.62 2.19 21.12 13.63 4.11 3.38
NT-5-0.4 20.3 21.6 5.61 2.38 3.23 19.51 15.03 2.09 2.39
a
Sulfided Co.
b
Oxidized Co.

Mo 3d5/2 BE for sulfided Mo4+ species remained close at a S 2p3/2 BE of 161.5  0.3 (See Table 4) corresponding to S2 ions in
228.8  0.2 eV for all the samples (see Table 4) in agreement with BE MoS2 in agreement with literature [49]. The spectrum of NT-3-0.3
value reported for MoS2 phase [46]. The Co 2p3/2 BE was around (not shown), was made up of two broad peaks, that were fitted with
779.3 eV for samples with Co/(Co + Mo) = 0.3. This value shifted to two doublets revealing the presence of at least two sulfur species,
778.9 and 778.6 eV when the Co/(Co + Mo) varied from 0.2 to 0.4, associated to S2 and S22 ligands [41,43].
respectively. Accordingly, the BE value reported for segregated Co9S8 The sulfidability of Co increased from 47.3 to 68.0 as nominal
phase is around 778.5 eV [47,48]. On the other hand, when Co species Mo loading rose from 3 to 5 atoms/nm2, at Co/(Co + Mo) = 0.3, as
chemically interact with the edge sites of MoS2 particles (to form the indicated in Table 5. At higher Mo loadings, a large amount of edge
‘‘CoMoS’’ phase), BE shifts to a value 0.9 eV higher than that for sites is available to be promoted by Co atoms upon sulfidation (see
Co9S8. Then, the results shown in Table 4 strongly suggest that Co column 8 Table 3). Then, Co sulfidability increased. When Co/
species impregnated at Co/(Co + Mo) = 0.3 (in the final catalyst) give (Co + Mo) was varied from 0.3 to 0.2 or to 0.4, Co sulfidability
rise to sulfided species in intimate interaction with MoS2 particles, in decreased from 68% to 42%, respectively. At low atomic ratio, large
comparison with those obtained in materials at Co/(Co + Mo) = 0.2 or crystallites of MoO3 segregate (see Raman and TEM results, Figs. 2
0.4. Based on these facts, it can be assumed that a higher amount of and 5, respectively), being difficult to sulfide. In fact, Mo
‘‘CoMoS’’ phase is formed when Co and Mo are deposited at Co/ sulfidability dropped. Since there are not enough edge sites to
(Co + Mo) = 0.3. Presumably, at this condition, good promotion of be promoted, Co atoms aggregates in oxidized species decreasing
MoS2 by Co could be achieved. also its sulfidability. In contrast, at high atomic ratio, sulfided Co
These results can be rationalized as follows: (i) at low Co atoms aggregates into large Co9S8 crystallites, as observed by
concentration, Co/(Co + Mo) = 0.2, the amount of Co available to HRTEM (Fig. 5e), and sulfidability decreased considerably as
react with Mo atoms is too low (Co/Mo = 0.25) to form appreciable determined by XPS.
amounts of CoMoO4 (the precursor of CoMoS phase) and hence, The S/(CoS + MoS) atomic ratio, as determined by XPS, was
some MoO3 segregates. Thus, after sulfiding just a little amount of around 1.2–1.5 (see column 4, Table 5). Only in the catalyst with
promoted phase was observed. (ii) At high Co concentration, Co/ the lowest metal loading, NT-3-0.3, the ratio was over 2. As
(Co + Mo) = 0.4, (Co/Mo = 0.66), closer to the CoMoO4 stoichio- aforementioned, two different sulfur species were detected
metric atomic ratio, more CoMoO4 was observed in Raman spectra (spectrum not shown) in this sample.
and during sulfidation; Co9S8 segregates and curved MoS2 particles Co and Mo surface coverage was estimated by the surface
were mainly formed. Then, Co/(Co + Mo) = 0.3, (Co/Mo = 0.5), is the atomic ratio (Co + Mo)/(Co + Mo + Ti), from the total amount of Co
optimum atomic ratio to generate a well promoted CoMoS phase and Mo atoms determined in the corresponding XPS spectra. As
on our nanotubular support. These facts should be reflected in indicated in Table 5 (column 5), overall Co and Mo dispersion
enhanced HDS activity. increased from 0.3 to around 0.5 for nominal Mo loadings of 3 and
Representative S 2p spectra of different catalysts are presented 5 atoms/nm2, respectively. However, not all the Co and Mo atoms
in Fig. 8. Except for the catalyst with the lowest metal loading (e.g., exposed on the surface are sulfided, then, the amount of sulfided
NT-3-0.3) all spectra were fitted with just one S 2p doublet, having species dispersed on the surface can be determined by the atomic
ratio [(CoS + MoS)/(Cotot + Motot + Ti)] also reported in Table 5 (see
Table 4 column 6). As expected, the amount of sulfided Co and Mo species
Binding Energy and FWHM values of atomic components of sulfided catalysts. dispersed on the surface increased as metal loading increased.
Nevertheless, it is necessary to point out that not all the sulfided Co
Samples Binding Energy (eV)/FWHM (eV)a
and Mo species are forming the active CoMoS phase. The CoS/MoS
S (2p) Ti (2p 3/2) Co (2p3/2)b Mo (3d5/2)c atomic ratio also presented in Table 5 represents the promotion
NT-3-0.3 161.2 458.9 779.3 228.6
factor of MoS2 by Co atoms to yield the CoMoS phase. The
1.7 1.4 1.4 0.95 promotion factor was 0.23  0.01 for samples synthesized with a Co/
(Co + Mo) = 0.3, whereas this (CoS/MoS) atomic ratio decreased
NT-4-0.3 161.7 459.3 779.3 229.0
0.92 1.2 1.2 0.87
considerably in samples prepared with Co/(Co + Mo) = 0.2 or 0.4.
Noteworthy, samples prepared at Co/(Co + Mo) = 0.3 showed
NT-5-0.3 161.7 459.3 779.3 228.9
the same Co–S BE (779.3 eV) and same promotional factor (CoS/
0.92 1.2 1.5 0.83
MoS = 0.24) as reported in Table 5, whereas in samples where the
NT-5-0.2 161.6 459.1 778.9 228.9 Co/(Co + Mo) was 0.2 and 0.4, that is, NT-5-0.2 and NT-5-0.4, the
0.83 1.2 1.6 0.82
promotional factor decreased to 0.12 and 0.16, respectively, and
NT-5-0.4 161.5 458.9 778.6 228.7 the BE position was similar to that reported for Co9S8 (778.5 eV).
0.88 1.2 1.4 0.80 Multiplying the dispersion of sulfided species (CoS + MoS)/
a
Full width at half maximum.
(Cotot + Motot + Ti) by the promotional factor (CoS/MoS) allowed
b
Sulfided Co. us to propose a procedure to calculate the density of CoMoS phase
c
Sulfided Mo. on the surface of nanotubular titania (d), regardless of the Co/
222 J.A. Toledo-Antonio et al. / Applied Catalysis B: Environmental 90 (2009) 213–223

Table 5
Surface chemical composition of sulfides Co and Mo atoms derived from XPS measurements.

Catalyst Sulfidation (%) S/(CoS + MoS) Co + Mo/[Co + Mo + Ti] CoS + MoS/[Co + Mo + Ti] CoS/MoS da
(atomic ratio) (atomic ratio) (atomic ratio)
Co Mo

NT-3-0.3 47.3 66.6 2.5 0.29 0.18 0.23 0.041

NT-4-0.3 53.2 80.8 1.2 0.44 0.32 0.24 0.076
NT-5-0.3 68.0 81.5 1.2 0.49 0.38 0.24 0.091
NT-5-0.2 42.5 64.5 1.2 0.58 0.35 0.12 0.042
NT-5-0.4 42.4 77.0 1.2 0.54 0.37 0.16 0.059
a
CoMoS phase density.

(Co + Mo) ratio and on the Mo loading, which linearly correlated directly proportional to d, the density of CoMoS phase generated on
with the pseudo first order kinetic constant in the DBT HDS as the surface of NT.
shown in the following section. To construct the linear plot in Fig. 9, XPS and catalytic activity of
By varying Co loading in samples of high Mo concentration no two samples prepared at different acid and basic pH NTA-3-0.3 and
significant morphological changes in highly stacked MoS2 particles NTB-3-0.3, respectively, were included [25]. These catalysts have
(Fig. 5a–c) were evident. On the other hand, CoMoS phase density the same Mo loading and Co/(Co + Mo) atomic ratio, but were
clearly decreased (see column 8 in Table 5) when the Co/(Co + Mo) obtained through different methods (see [25] for details in
ratio did not correspond to the optimal value (0.3). Thus, d and materials preparation). Then, it can be concluded that, regardless
MoS2 stacking are likely to be independent parameters, even of the different preparation conditions, HDS of DBT strongly
though both increased with Mo and Co loading (samples at (Co/ depends on the surface density of CoMoS phase, e.g., d.
Co + Mo) = 3, see column 8 in Table 5 and Fig. 4b–d). In fact, a direct correlation can be established between the CoS
and sulfided Mo4+ determined by XPS and the HDS activity,
3.4. HDS on CoMoS/NT catalysts suggesting that higher dispersion in CoS and MoS2 correspond to
higher HDS activity. Moreover, as aforementioned, the BE of Co–S
HDS catalytic activity of CoMoS/NT catalysts on DBT is reported in NTA-3-0.3 occurred at 779.0 eV suggesting that most of the Co
in Table 6. HDS activity increased with Mo loading, at constant Co/ atoms are not in close interaction with MoS2 [25]. Thus, in this case
(Co + Mo) = 0.3, suggesting that all the metal loading was the formation of the ‘‘CoMoS’’ was limited and its corresponding
uniformly dispersed on the nanotubular support. However, when HDS activity was low. Meanwhile, in the other samples (NTB-3-0.3
Co/(Co + Mo) varied, to 0.2 and to 0.4, the HDS activity decreased and NT-3-0.3) the Co–S BE shifted to 779.3 eV suggesting a more
considerably. Clearly these differences in catalytic activity efficiently promoted MoS2 phase, this fact being well-correlated
correlate with the density (d) of CoMoS phase determined by with their enhanced HDS activity [25].
XPS. In fact, from Fig. 9, it can be observed that HDS activity is According to Fig. 9, the main factor determining HDS activity of
various materials is their concentration of CoMoS phase. The linear
Table 6 trend found when plotting these parameters indicate that the
Pseudo-first order kinetic constant, DBT HDS over various sulfided CoMo nature of active sites present in different materials is similar; the
catalysts supported on nanostructured titania.
amount of CoMoS phase dictating the catalytic behavior. The
Catalyst k  103 (m3/kgcat s) k0  105 (m3/kgMo s) available information precludes discrimination between sites type
NT-3-0.3 2.00 15.6
1 and 2 on our materials, although the high stacking degree
NT-4-0.3 3.58 26.5 observed for samples of high Mo loading (Fig. 4d–f) strongly
NT-5-0.3 6.44 43.0 suggests molybdenum sulfide particles in low interaction with the
NT-5-0.2 1.11 7.3 nanotubular support (e.g., type 2 sites, [50]). Formation of
NT-5-0.4 2.71 17.6
fullerene-like or onion-like MoS2 particles also point out in that
Batch reactor, 320 8C, 56 kg/cm2, 1000 rpm mixing speed, n-hexadecane direction (see Fig. 5). Thus, in our materials, type 2 sites are likely to
as solvent. be present even in the case of catalysts of moderate Mo loading
(3 atmos/nm2 nominal surface concentration).

4. Conclusions

Fullerene-like CoMoS and MoS2 nanoparticles were found

dispersed along nanotubular TiO2. Differently from other CoMoS
particles supported on g-Al2O3 or anatase TiO2, where hexagonal
CoMoS and/or MoS2 particles have been clearly identified, in
nanotubular TiO2, curved onion-like particles were observed
smeared along the nanotubes. The number of MoS2 slabs varied
from about 2 to 6, depending on the Mo loading. Dislocations in the
outer most zones of MoS2 particles strongly indicate the formation
of CoMoS sites, being the corner stone of highly active HDS
catalysts. Since not all MoS2 particles are promoted by Co, a
quantification of the CoMoS surface density (d) was achieved by
XPS spectroscopy and then it was correlated with HDS activity. The
independent variable Co/(Co + Mo), used in this work to optimize
CoMoS formation on nanotubular TiO2, was varied from 0.2 to 0.4,
Fig. 9. Reaction kinetic constant of HDS of DBT as function of CoMoS phase density and Mo loading from 3 to 5 atoms/nm2, with a view to approach
(d) determined by XPS. higher Mo contents as required for highly active industrial HDS
 J.A. Toledo-Antonio et al. / Applied Catalysis B: Environmental 90 (2009) 213–223 223

catalysts. A linear correlation between HDS activity of DBT and d, [22] J.A. Toledo-Antonio, S. Capula, M.A. Cortés-Jácome, C. Angeles-Chavez, E. López-
Salinas, G. Ferrat, J. Navarrete, J. Escobar, J. Phys. Chem. C 111 (2007) 10799–
indicates that the highest CoMoS formation (@ Co/(Co + Mo) = 0.3 10805.
and Mo = 5 atoms/nm2) in fact results in the most active catalyst of [23] M. Cortés-Jácome, M. Morales, C. Angeles-Chávez, L.F. Ramı́rez-Verduzco, E.
a series, d being a trustworthy parameter to screen potential HDS López-Salinas, J.A. Toledo Antonio, Chem. Mater. 19 (2007) 6605–6614.
[24] S. Eijsbouts, L.C.A. van den Oetellaar, R.R. van Puijenbroek, J. Catal. 229 (2005)
catalysts. 352–364.
[25] M.A. Cortés-Jácome, J. Escobar, C. Angeles Chávez, E. López Salinas, E. Romero, G.
Ferrat, J.A. Toledo-Antonio, Catal. Today 130 (2008) 56–62.
Acknowledgment [26] XPS and Auger Handbook, Doc. Number 600001, ThermoVG Scientific: West
Sussex U.K., 2 (2003).
The authors would like to acknowledge the IMP for the financial [27] J. Ramirez, L. Ruiz-Ramirez, L. Cedeño, V. Harle, M. Vrinat, M. Breysse, Appl. Catal.
A 93 (1993) 163–180.
support through the projects: IMP-D.00446 and D.00447. [28] A. Kukovecz, M. Hodos, Z. Konya, I. Kiricsi, Chem. Phys. Lett. 411 (2005) 445–449.
[29] Y.V. Kolen’ko, K.A. Kovnir, A.I. Gavrilov, A.V. Garshev, J. Frantti, O.I. Lebedev, B.R.
Churagulov, G.V. Tendeloo, M. Yoshimura, J. Phys. Chem. B 110 (2006) 4030–4038.
References [30] R.A. Nyquist, C.L. Putziq, V. Leugers, Handbook of Infrared and Raman Spectra of
Inorganic Compounds and Inorganic Salts, 1, Academic Press, San Diego, USA,
[1] H. Topsøe, B.S. Clausen, F.E. Massoth, J.R. Anderson, M. Boudart (Eds.), Catalysis 1997, p. 18.
Science and Technology, Springer, Berlin, 1996, p. 11. [31] E. Payen, M.C. Dhamelincourt, P. Dhamelincourt, J. Grimblot, J.P. Bonnelle, Appl.
[2] H. Topsøe, B.S. Clausen, Catal. Rev. Sci. Eng. 26 (1984) 395–420. Spectrosc. 36 (1) (1982) 30.
[3] S. Eijsbouts, Appl. Catal. A 158 (1997) 53–92. [32] Z.B. Wei, W. Yan, H. Zhang, T. Ren, Q. Xin, L. Li Zhongcheng, Appl. Catal. A 167
[4] S.M.A.M. Bouwens, F.B.M. van Zon, M.P. van Dijk, A.M. van der Kraan, V.H.J. de (1998) 39–48.
Beer, J.A.R. van Veen, D.C. Koningsberger, J. Catal. 146 (1994) 375–393. [33] P.A. Spevack, N.S. McIntyre, J. Phys. Chem. 96 (1992) 9029–9035.
[5] M. Breysse, C. Geantet, P. Afanasiev, J. Blanchard, M. Vrinat, Catal. Today 130 [34] G. Mestl, P. Ruiz, B.H. Delmon Knozinger, J. Phys. Chem. 98 (1994) 11269–11275.
(2008) 3–13. [35] M.A. Cortés-Jacome, G. Ferrat-Torres, L.F. Flores Ortiz, C. Angeles-Chavez, E.
[6] M. Sun, D. Nicosia, R. Prins, Catal. Today 86 (2003) 173–189. Lopez-Salinas, J. Escobar, M.L. Mosqueira, J.A. Toledo-Antonio, Catal. Today 126
[7] R. Cattaneo, Th. Weber, T. Shido, R. Prins, J. Catal. 191 (2000) 225–236. (2007) 248–255.
[8] H. Shimada, Catal. Today 86 (2003) 17–29. [36] Y. Sakashita, Y. Araki, K. Honna, H. Shimada, Appl. Catal. A 197 (2000) 247–253.
[9] J. Ramirez, L. Cedeño, G. Busca, J. Catal. 184 (1999) 59–67. [37] J.R. Günter, O. Marks, T.I. Korányi, Z. Paál, Appl. Catal. 39 (1988) 285–294.
[10] M. Vrinat, D. Hamon, M. Breysse, D. Durand, T. Courieres, Catal. Today 20 (1994) [38] A.I. Dugulan, E.J.M. Hensen, J.A.R. van Veen, Catal. Today 130 (2008) 126–134.
273–282. [39] M. Brorson, A. Carlsson, H. Topsøe, Catal. Today 123 (2007) 31–36.
[11] S. Damyanova, L. Petrov, P. Grange, Appl. Catal. A 239 (2003) 241–252. [40] A. Carlsson, M. Brorson, H. Topsøe, J. Catal. 227 (2004) 530–536.
[12] Y. Saih, M. Nagata, T. Funamoto, Y. Masuyama, K. Segawa, Appl. Catal. A 295 [41] J.C. Muijsers, Th. Weber, V. van Hardeveld, H.W. Zandbergen, J.W. Niemantsver-
(2005) 11–22. driet, J. Catal. 157 (1995) 698–705.
[13] J. Ramı́rez, S. Fuentes, G. Dı́az, M. Vrinat, M. Breysse, M. Lacroix, Appl. Catal. 52 [42] Th. Weber, J.C. Muijsers, J.H.M.C. van Wolput, C.P.J. Verhagen, J.W. Niemantsver-
(1989) 211–224. driet, J. Phys. Chem. 100 (1996) 14144–14150.
[14] S. Inoue, A. Muto, H. Kudou, T. Ono, Appl. Catal. A 269 (2004) 7–12. [43] A.F.H. Sanders, A.M. de Jong, V.H.J. de Beer, J.A.R. van Veen, J.W. Niemantsverdriet,
[15] S. Dzwigaj, C. Louis, M. Breysse, M. Cattenot, V. Belliere, C. Geantet, M. Vrinat, P. Appl. Surf. Sci. 144–145 (1999) 380–384.
Blanchard, E. Payen, S. Inoue, H. Kudo, Y. Yoshimura, Appl. Catal. B 41 (2003) 181– [44] J.E. Herrera, D.E. Resasco, J. Catal. 221 (2004) 354–364.
191. [45] A.D. Gandubert, C. Legens, D. Guillaume, S. Rebours, E. Payen, Oil Gas Sci. Technol.
[16] T. Kasuga, M. Hiramatsu, A. Hoson, T. Sekino, K. Niihara, Langmuir 14 (1998) 62 (2007) 79–89.
3160–3163. [46] R. Candia, O. Sorensen, J. Villadsen, N. Topsøe, B.S. Clausen, H. Topsøe, Bull. Soc.
[17] D.V. Bavykin, J.M. Friedrich, A.A. Lapkin, F.C. Walsh, Chem. Mater. 18 (2006) 1124– Chim. Belg. 93 (1984) 763–773.
1129. [47] Y. Okamoto, T. Kubota, Catal. Today 86 (2003) 31–43.
[18] G.H. Du, Q. Chen, R.C. Che, Z.Y. Yuan, L.M. Peng, Appl. Phys. Lett. 79 (2001) 3702. [48] H. Topsøe, B.S. Clausen, N. Topsøe, P. Zeuthen, Stud. Surf. Sci. Catal. 53 (1989) 77–
[19] J. Escobar, J.A. Toledo, M.A. Cortés, M.L. Mosqueira, V. Pérez, G. Ferrat, E. López- 102.
Salinas, E. Torres-Garcı́a, Catal. Today 106 (2005) 222–226. [49] A.M. De Jong, H.J. Borg, L.J. Van IJzendoorn, V.G.F.M. Soudant, V.H.J. De Beer, J.A.R.
[20] R. Tenne, Nature 431 (2004) 640–641. Van Veen, J.W. Niemantsverdriet, J. Phys. Chem. 97 (1993) 6477–6483.
[21] R. Tenne, Angew. Chem. Int. Ed. 42 (2003) 5124–5132. [50] H. Topsøe, Appl. Catal. A 322 (2007) 3–8.