Alumina-, niobia-, and niobia/alumina-supported NiMoS catalysts: Surface properties and activities in the hydrodesulfurization of thiophene and hydrodenitrogenation of 2,6-dimethylaniline

Article abstract of DOI:10.1016/j.jcat.2007.09.012

The activity of NiMoS catalysts supported on niobia, alumina, and niobia/alumina was compared for the thiophene hydrodesulfurization (HDS) and 2,6-dimethylaniline (2,6-DMA) hydrodenitrogenation (HDN) reactions. To evaluate the acidity of the supports and identify the nature of the sulfide sites, adsorption of 2,6-dimethylpyridine, pyridine, and CO was performed and followed by IR spectroscopy. This study has shown that with niobia as a support, the activity of NiMoS catalysts in thiophene HDS and in HDN of 2,6-DMA was no longer promoted by the synergy between Ni and Mo. The absence of synergy between molybdenum and nickel on niobia can be explained by the strong interaction of each metal with niobia at the expense of interaction with each other. Moreover, it has been shown that on a niobia/alumina support, the formation of the NiMoS phase can be directly linked to the presence of alumina not covered by niobia. However, niobia is an interesting support for the HDN of 2,6-DMA, because it favors the formation of xylene through direct ammonia elimination involving low H₂ consumption. The activity for xylene formation on niobia is linked to the electron-deficient nature of the Mo sulfide site, as demonstrated by CO adsorption followed by IR.

Full text of DOI:10.1016/j.jcat.2007.09.012

Journal of Catalysis 252 (2007) 321–334
www.elsevier.com/locate/jcat
Alumina-, niobia-, and niobia/alumina-supported NiMoS catalysts:
Surface properties and activities in the hydrodesulfurization of thiophene
and hydrodenitrogenation of 2,6-dimethylaniline
Angela S. Rocha a, Arnaldo C. Faro Jr. a, Laetitia Oliviero b,∗, Jacob Van Gestel b, Françoise Maugé b
a
Departamento de Físico-Química, Instituto de Química, UFRJ, Ilha do Fundão, CT, Rio de Janeiro, RJ, CEP 21949-900, Brazil
Laboratoire Catalyse et Spectrochimie, ENSICAEN, Université de Caen, CNRS, 6, Bd Maréchal Juin, F-14050 Caen, France
b
Received 9 July 2007; revised 10 September 2007; accepted 11 September 2007
Available online 31 October 2007
Abstract
The activity of NiMoS catalysts supported on niobia, alumina, and niobia/alumina was compared for the thiophene hydrodesulfurization (HDS)
and 2,6-dimethylaniline (2,6-DMA) hydrodenitrogenation (HDN) reactions. To evaluate the acidity of the supports and identify the nature of the
sulfide sites, adsorption of 2,6-dimethylpyridine, pyridine, and CO was performed and followed by IR spectroscopy. This study has shown that
with niobia as a support, the activity of NiMoS catalysts in thiophene HDS and in HDN of 2,6-DMA was no longer promoted by the synergy
between Ni and Mo. The absence of synergy between molybdenum and nickel on niobia can be explained by the strong interaction of each metal
with niobia at the expense of interaction with each other. Moreover, it has been shown that on a niobia/alumina support, the formation of the
NiMoS phase can be directly linked to the presence of alumina not covered by niobia. However, niobia is an interesting support for the HDN of
2,6-DMA, because it favors the formation of xylene through direct ammonia elimination involving low H consumption. The activity for xylene
2
formation on niobia is linked to the electron-deficient nature of the Mo sulfide site, as demonstrated by CO adsorption followed by IR.
© 2007 Elsevier Inc. All rights reserved.
Keywords: Hydrotreatment; NiMo synergy; Niobium; HDS; HDN; IR spectroscopy
1. Introduction
promoted by sulfided cobalt or nickel and supported on alumina
[1–6]. The acid–base characteristics of the supports play a fun-
damental role in the activity of these catalysts; several groups
have studied modified aluminas as supports in an effort to im-
prove their activity [7].
Fluorine was used to increase the acidity of alumina and
amorphous silica alumina; however, the activity of NiMo phase
for the HDN reaction was not influenced by this variation in
acidity [8]. On the other hand, the addition of boron and phos-
phorous to NiMo/Al2O3 catalysts has shown positive effects on
hydrocracking and HDS activities [9,10].
Along this line, first niobium was tested as an additive to tra-
ditional hydrotreating catalyst systems. Niobium oxide has in-
teresting properties in reactions catalyzed by solid acid [11,12].
The addition of niobia (on surface or in the bulk) to alumina for
preparing hydrotreating NiMo/Al catalysts has been reported
by Weissman [13]. The enhanced hydrotreating activity with
the addition of niobia was attributed to the increase in support
Given the continuing worldwide effort to decrease pollu-
tant emissions, the studies of catalysts for hydrotreatment of
fuel fractions from petroleum remain relevant. Catalysts for hy-
drodenitrogenation (HDN) and hydrodesulfurization (HDS) are
fundamental for this purpose, because nitrogen and sulfur com-
pounds in fuels not only cause environmental problems, but
also are poisons for other catalysts on which these fractions are
treated. For these reasons, the allowable range of nitrogen and
sulfur content in fuels will become increasingly restrictive, re-
quiring the development of HDN and HDS catalysts with high
activity.
Conventional catalysts for hydrotreatment typically are com-
posed of one group VI transition metal sulfide (MoS2 or WS2)
*
Corresponding author. Fax: +33 231 452 822.
E-mail address: laetitia.oliviero@ensicaen.fr (L. Oliviero).
0021-9517/$ – see front matter © 2007 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2007.09.012

322
A.S. Rocha et al. / Journal of Catalysis 252 (2007) 321–334
acidity. Gaborit et al. concluded that niobium as a dopant of an
industrial NiMo/Al2O3 catalyst increased its activity in HDS
and hydrogenation (HYD) on high-pressure sulfidation with
CS2 [14]. The authors verified the presence of NbS2 species,
with specific acidic properties. Second, niobia- and alumina-
supported Ni, Mo, and NiMo catalysts were tested in HDS and
cumene hydrocracking [15]. The higher activity in the HDS re-
action of alumina-supported NiMo catalysts compared with that
supported on niobia was attributed to the fact that synergy be-
tween Ni and Mo sulfides occurred on the alumina-supported
catalysts but not on the niobia-supported ones. Finally, Allali
et al. [16] prepared niobium sulfide supported on carbon or
alumina that had HDS activity and found that the genesis of
niobium sulfide phases was strongly dependent on the sulfida-
tion method.
Recently, we studied the activity of molybdenum sulfide cat-
alysts supported on niobia, alumina, and niobia/alumina in the
thiophene HDS and 2,6-dimethylaniline (2,6-DMA) HDN re-
actions. 2,6-DMA Mo catalysts supported on niobia containing
supports were more active for the XYL route of 2,6-DMA de-
composition that the one supported on alumina. We have shown
the correlation between this activity improvement and the modi-
fication of the electronic properties of the Mo sulfide phase with
the support nature [17]. In the present contribution, we aim to
study the effect of the support on the promoting effect of nickel
on supported Mo catalysts. Toward this purpose, we prepared
NiMo sulfide catalysts supported on niobia, pure alumina, and
alumina modified with niobia. Their activity in HDS and HDN
reactions were investigated. We also studied the effect of H2S
on the activity of the NiMo sulfide catalysts for 2,6-DMA HDN.
To evaluate the acidity of the supports and identify the nature of
the sulfide sites, adsorption of 2,6-dimethylpyridine, pyridine,
and CO was performed and followed by IR spectroscopy.
Table 1
Composition and textural characteristics of the supports and catalysts
a
b
c
c
Material
S
PV
Atomic ratio wt% Ni wt% Mo
BET
−1
−1
)
cat
2
3
(m g
)
(cm g
Ni/(Ni + Mo)
–
cat
Al O
2
200
175
171
185
189
179
179
130
0.57
0.44
0.44
0.47
0.48
0.49
0.43
0.32
0.33
0.35
0.37
0.38
0.12
0.13
0.12
0.10
0.11
0.13
–
–
3
Mo/Al
0
–
12.9
9.9
6.7
3.5
–
25% NiMo/Al
50% NiMo/Al
75% NiMo/Al
Ni/Al
0.25
0.50
0.75
1.00
–
2.0
4.1
6.3
8.7
–
Nb O /Al O
–
2
5
2 3
Mo/NbAl
0
–
11.7
9.0
6.1
3.1
–
25% NiMo/NbAl 137
50% NiMo/NbAl 147
75% NiMo/NbAl 156
0.25
0.50
0.75
1.00
–
1.8
3.8
5.8
7.9
–
Ni/NbAl
Nb O
159
26
27
26
25
27
31
–
2
5
Mo/Nb
0
–
2.0
1.5
1.0
0.5
–
25% NiMo/Nb
50% NiMo/Nb
75% NiMo/Nb
Ni/Nb
0.25
0.50
0.75
1.00
0.3
0.6
0.9
1.2
a
Surface area determined by BET method.
Pore volume.
Introduced by impregnation.
b
c
The pH was adjusted to 2 by adding HNO3, to avoid any precip-
itation of molybdenum compounds. The resulting solids were
dried at 393 K and calcined at 723 K. Then the nickel was in-
corporated by impregnation with an aqueous solution of nickel
nitrate, Ni(NO3)2·6H2O, dried at 393 K and calcined at 723 K.
The nomenclature and nominal composition of the catalysts are
presented in Table 1.
2.2. Catalytic activity measurements
The catalysts were tested by means of two reactions: hy-
drodesulfurization (HDS) of thiophene and hydrodenitrogena-
tion (HDN) of 2,6-dimethylaniline. The thiophene HDS activ-
ity test was carried out in a flow microreactor working at at-
mospheric pressure and 623 K. Appropriate amounts of catalyst
(15–30 mg) were used to keep the conversion <5%. Before the
reaction, the catalysts were sulfided in situ with 30 mL min−1
2. Experimental
2.1. Preparation of the catalysts
Three supports were used: pure oxides γ -Al2O3 and Nb2O5
and alumina-supported Nb2O5. The alumina was obtained by
calcination of a boehmite (PURAL SB RT04/111) at 823 K
for 3 h. The niobia was prepared by calcination of niobic acid
(provided by CBMM) at 773 K for 4 h. The Nb2O5/Al2O3 was
prepared by incipient wetness impregnation of the alumina with
an aqueous solution containing the appropriate amount of a nio-
bium ammonium oxalate complex, (NH4)3[NbO(C2O4)3], to
obtain a final support with 20 wt% of Nb2O5, followed by dry-
ing at 393 K for 1 h and calcination at 773 K for 4 h. The 20
wt% of Nb2O5 corresponds to around 90% of the monolayer on
Al2O3 according to an atomic density of 5 Nb nm−2 [18].
The NiMo catalysts were prepared using incipient wet-
ness sequential impregnation to obtain catalysts with a total
of around 5 metal atoms per nm2 of support surface area. In
this method, first the molybdenum impregnation onto the sup-
ports was carried out using an aqueous solution of ammonium
heptamolybdate, (NH4)6Mo7O24·4H2O, and 6% of hydrogen
peroxide, the latter to increase the heptamolybdate solubility.
of H2S/H2 (10 vol%) at 623 K (heating rate of 3 K min−1
)
for 1.5 h. Thiophene was introduced into the reactor by flow-
ing hydrogen (70 mL min−1) through a thiophene saturator
maintained at 291 K (Pthiophene = 6920 Pa) and mixed to 20
mL min−1 of H2S/H2 (10 vol%). Products were analyzed online
using a Varian gas chromatograph equipped with a flame ion-
ization detector (FID) and a CPSIL-5CB capillary column. The
specific rate at steady state was calculated as rHDS = (F/m)X,
where F/m is the molar flow rate of reactant per gram of cata-
lyst and X is the thiophene conversion.
The 2,6-dimethylaniline HDN activity test was carried out
in a stainless steel reactor (Sotelem) at 4 MPa and 573 K.
Around 0.35–0.40 g of catalyst was first sulfided in situ un-
der 60 mL min−1 of H2S/H2 (10 vol%) at 4 MPa and 623 K
(temperature ramp of 3 K min−1) for 2 h. Then the catalyst
was cooled under H2S/H2 to the reaction temperature, and the
H2S content was adjusted at 1.4% in hydrogen. The liquid feed

A.S. Rocha et al. / Journal of Catalysis 252 (2007) 321–334
323
(10 vol% 2,6-dimethylaniline, reactant; 80% n-heptane, sol-
vent; and 10% n-decane, internal standard) was introduced by
an HPLC pump and vaporized in the H2S/H2 stream. The reac-
tion products were condensed at the reactor exit, and the liquid
was analyzed with a Varian gas chromatograph equipped with
a CPSIL-5CB capillary column and a FID. No new products
besides those observed during 2,6-dimethylaniline decomposi-
tion on the alumina-supported catalysts were observed on the
niobia-containing catalysts.
The partial pressure of 2,6-dimethylaniline was kept con-
stant at 13 kPa for all experiments. Reaction conditions at
steady state were varied by changing the contact time (55–250
h g mol−1) at 56 or 0 kPa of H2S. We determined the reaction
orders in 2,6-dimethylaniline for the hydrogenation (HYD), xy-
lene (XYL), and disproportionation (DIS) routes to be 0, 0, and
0.4, respectively [19]. Activities were expressed by the initial
rate.
a second dose of H2S/H2 mixture was placed in contact with
the catalyst overnight. Finally, a third dose of H2S/H2 mixture
was contacted with the catalyst for 1 h, after which the catalyst
was evacuated at the same temperature for 45 min and cooled
to 100 K. The subsequent procedure for CO adsorption was the
same as for the supports.
Spectra were recorded in 256 scans using a Nicolet FTIR
spectrometer. The bands of adsorbed species were obtained by
subtracting to the spectra recorded after that recorded before
introduction of the probe. Band intensities were corrected for
slight differences in the weight of the wafer and normalized to
a disc of 5 mg cm−2
.
3. Results
3.1. Textural and structural characteristics
The preparation of niobia on alumina is not an easy task,
because the common niobium compounds are poorly soluble,
and sequential impregnations are required to obtain supports
well covered with niobium oxide. This procedure often gener-
ates poorly dispersed niobia.
The surface areas and pore volumes of the supports and cat-
alysts for the present study are presented in Table 1. The high
surface area of the alumina contrasts with the low surface area
of niobia. The niobia/alumina showed only a small decrease in
surface area compared with the pure alumina.
The NiMo catalysts on γ -Al2O3 and Nb2O5/Al2O3 exhib-
ited smaller surface areas than their respective supports; this
decrease can be explained mainly by the increased catalyst
density due to incorporation of the metals. Support pore block-
age by metal particles also can contribute to this decreas in
surface area. It appears that the pore blockage occurred pref-
erentially with Mo. With Ni, incorporation into the alumina
structure would explain a smaller effect [15]. The interpreta-
tion is somehow more difficult for the Nb2O5 support, because
for Mo or Ni catalysts, the pore volume is larger than that of
the pure niobia. Impregnation of the metal phase in acidic me-
dia could be a reason for redissolution and reprecipitation of the
support.
XRD indicated the presence of hexagonal Nb2O5 phase
(TT-Nb2O5) in niobia and of γ -alumina for Nb2O5/Al2O3 and
Al2O3 supports, as reported previously [17]. These results cor-
roborate the BET results and strongly indicate preservation of
the alumina structure after niobium deposition.
The Raman spectra of the supports are in agreement with the
XRD results (Fig. 1). Alumina was not active in Raman. The
spectrum of Nb2O5 showed a sharp, intense band at 695 cm−1
assigned to the symmetric stretching mode of niobia polyhe-
dra [21]. The higher wavenumber compared with amorphous
niobia is characteristic of the ordered structure of the hexag-
onal phase [21]. The Nb2O5 spectrum also presents bands at
306 and 230 cm−1 characteristic of the bending mode of Nb–
O–Nb bonds. The Nb2O5/Al2O3 support presented three broad
bands at 912, 679, and 234 cm−1. The most intense band at
912 cm−1 can be attributed to the symmetric stretching mode
νs([–O–Nb–O–]n) of polymerized polyhedra (monolayer-type
2.3. Characterization
The supports were analyzed by XRD in a Rigaku Miniflex
diffractometer, with CuKα radiation, at λ = 1.5418 Å, 30 kV,
and 15 mA. Raman characterization of the support was per-
formed under ambient conditions. The spectra were recorded on
a Labram 300 Raman spectrometer (Jobin Yvon) equipped with
a confocal microscope and a CCD camera. The laser was used
at λ = 532 nm with an energy of 30 mW without filtration. No
characterization of the oxide catalysts was performed, because
the sulfidation modifies the dispersion of the metal phase [15].
The surface areas and pore volumes of the supports and oxidic
precursors were determined using the BET method from nitro-
gen adsorption data at 77 K in a Micromeritics ASAP 2010 C
automated volumetric apparatus. Before the analyses, all sam-
ples were treated in situ under vacuum at 673 K.
Support acidity was characterized by FTIR spectroscopy of
adsorbed CO, pyridine (Py), and 2,6-dimethylpyridine (2,6-
DMP). The samples were ground and pressed into self-sup-
ported wafers (ca. 10 mg, for a disc of 2 cm2). The supports
were activated in situ in the IR cell under vacuum at 723 K
for 4 h. After activation, 133 Pa of Py or 2,6-DMP was intro-
duced into the cell at ambient temperature. After adsorption, the
sample was evacuated at increasing temperatures up to 673 K.
Details of the experiments were given previously [20]. Carbon
monoxide IR experiments were recorded after the same surface
pretreatment and subsequent cooling at 100 K. Small calibrated
doses of CO were introduced into the IR cell up to an equi-
librium pressure of 133.3 Pa. The system was subsequently
evacuated at 100 K down to a residual pressure of 10−3 Pa.
Sulfided catalysts were characterized by FTIR of adsorbed
CO. The oxidic catalysts were pressed into self-supported
wafers (ca. 10 mg, for discs of 2 cm2) and placed into a quartz
IR cell equipped with CaF2 windows and a cryogenic chamber.
Before the adsorption experiments, the samples were sulfided in
situ, according to the following procedure. The catalysts were
evacuated from room temperature up to 623 K, then contacted
at 623 K for 1 h with 13.3 kPa of a gas mixture containing
10 vol% H2S in H2. After a rapid evacuation of the gas phase,

324
A.S. Rocha et al. / Journal of Catalysis 252 (2007) 321–334
shoulder at 3660 cm−1. Because the sample has a small surface
area, the quantity of OH groups is smaller on the surface than
on alumina.
In the spectrum of the Nb2O5/Al2O3 support, we can ob-
serve that alumina is partially covered; the same bands as for
alumina alone are observed, but their intensity and the propor-
tions between them are different than those of alumina alone.
Basic hydroxyls, showing bands in the high wavenumber range,
are preferentially consumed. The surface coverage of alumina
can be estimated from the decrease of the OH region (not taking
into account the increase due to Nb–OH); around 70% of the
alumina surface is covered by niobia. Then we can conclude
that the niobium is well dispersed on the surface of alumina.
However, the presence of Nb–OH cannot be distinguished in
the spectrum of Nb2O5/Al2O3. Adsorption of basic probe mole-
cules was used to demonstrate the presence of such groups.
3.2. FTIR spectroscopy of basic probes adsorbed on the
supports
Fig. 1. Raman spectra of supports.
Acidic properties of the pure supports were characterized
by adsorption of basic probe molecules followed by IR spec-
troscopy. Figs. 3–5 show the spectra after adsorption on the
supports of pyridine (Py), 2,6-dimethylpyridine (2,6-DMP)
and CO, respectively. The main band characteristics of the
probes adsorbed on Lewis acid sites (LAS) and Brønsted acid
sites (BAS) of the supports are presented in Table 2. Pyridine
is a commonly used probe for surface acidity characteriza-
tion [23,24]. In the 1650–1400 cm−1 range, the position of two
characteristic bands of Py ring vibration (ν8a and ν19b) serve
as fingerprints of the presence of LAS and BAS on the surface.
LAS sites are evidenced by a ν8a band at around 1610 cm−1 and
a ν19b band at around 1450 cm−1. Adsorption of Py on BAS
leads to a ν8a band at around 1630 cm−1 and to a ν19b band
at around 1545 cm−1. Adsorption of 2,6-DMP leads to coordi-
nated, protonated, and H-bonded species, and the IR spectra of
the formed species allow differentiation among LAS, strongly
acidic hydroxyl groups, and weakly acidic hydroxyl groups,
respectively [25]. In particular, the ν8a vibration, located at
1594 cm−1 in the spectrum of liquid 2,6-DMP, is very sensi-
tive to the adsorption mode; a ν8a wavenumber >1625 cm−1
characterizes protonated species (2,6-DMPH+) whereas lower
wavenumbers correspond to coordinated or H-bonded species
[26,27]. The ν8b band position at 1630 cm−1 is also character-
istic of the formation of 2,6-DMPH+ species. At low temper-
ature, CO undergoes H-bond interactions with acidic hydroxyl
groups, giving rise to OH...CO adducts. These interactions
lead to a high-frequency shift of the ν(CO) stretching mode
compared with the gas phase (2143 cm−1) [28,29]. The stronger
the BAS, the higher the wavenumber of the ν(CO/OH) mode.
Interaction with LAS Mn+ without d electrons leads to a ν(CO)
mode at around 2200 cm−1; the higher the wavenumber, the
stronger the LAS site.
Fig. 2. FTIR spectra of supports in the OH region.
surface), whereas the band at around 680 cm−1 can be attributed
to νs([–Nb–O–Nb–]n) [18]. The band at 234 cm−1 is close to
the value of 230 cm−1 observed for Nb2O5 and thus can be at-
tributed to the bending mode of Nb–O–Nb bonds.
The characteristic band of the Nb=O stretching vibration ex-
pected at around 980 cm−1 was not observed on either Raman
or IR spectra of the pure niobia (spectrum not shown), likely
due to the low surface area of the oxide. For the Nb2O5/Al2O3
support, by difference between its IR spectra and that of alu-
mina, a fundamental narrow band at around 995 cm−1 could
be observed indicating the presence of mono oxo species at the
support surface [18]. Thus, the IR and Raman results indicate
the high dispersion of niobia as previously suggested by the
BET and XRD results.
The IR spectra in the OH region of supports activated un-
der vacuum are shown in Fig. 2. The complexity of the alumina
spectrum, which is assigned to OH groups bound to aluminum
atoms in different environments, can be seen [22]. In compari-
son, the spectrum of the niobium oxide is simple and presents
a large and weak band with a maximum at 3710 cm−1 and a
The infrared spectra of pyridine on alumina (Fig. 3) are in
agreement with the literature values and show only the presence
of LAS [25] also evidenced by 2,6-DMP and CO adsorption.
In contrast, with 2,6-DMP and CO (which are more sensitive

A.S. Rocha et al. / Journal of Catalysis 252 (2007) 321–334
325
Fig. 4. FTIR spectra of 133 Pa of 2,6-dimethylpyridine adsorbed on the supports
after evacuation at 323 K.
at such wavenumber could be assigned to either (i) the ν8a
vibration mode of pyridine coordinated to very strong Lewis
acid sites or (ii) the ν8b vibration mode of protonated pyridine.
The elimination of the 1627 cm−1 band after thermodesorption
at 573 K while that at 1607 cm−1 (which characterized weak
LAS) remains present shows that this band is specific for pyri-
dine interaction with BAS. Note that the absence of strong LAS
on Nb2O5 explains the clear observation of this vibration mode.
CO adsorption confirms the presence of weak LAS on Nb2O5,
because the wavenumber associated with CO interacting with
LAS is higher for alumina than for niobia (Table 2). The three
probes also show the presence of acidic Nb–OH that are strong
BAS because even after evacuation at 523 K, adsorbed pyridine
on BAS can be observed (Fig. 3). The results obtained with CO
adsorption also show that BAS of niobia are stronger than those
of alumina, because the band characteristic of CO/OH is ob-
served at higher wavenumber on niobia [ν(CO) = 2163 cm−1
]
than on alumina [ν(CO) = 2153 cm−1] (Fig. 5B).
The shift of the ν(OH) band due to CO interaction is also a
tool for comparing the strength of BAS; the greater the shift,
the stronger the site. Fig. 5A shows spectra obtained after the
addition of small doses of CO on Nb2O5 in the ν(OH) and
ν(CO) ranges. The shift of the ν(OH) band is evidenced by
the decrease of the initial band and the increase of the band
attributed to perturbed hydroxyls. The addition of small doses
helps to differentiate hydroxyl groups with different Brønsted
acid strengths in the same sample. In the case of niobia, two de-
creasing bands can be seen at 3719 and 3670 cm−1. The band at
3719 cm−1 shows a positive–negative effect for the first doses
of CO, indicating an indirect perturbation, likely due to CO
coordinated on LAS located in the vicinity of OH groups. In
contrast, at higher doses, the clear perturbation of the band at
3719 cm−1 gives rises to a new ν(OH) band at 3526 cm−1. The
band at 3670 cm−1 is perturbed by addition of the first doses
of CO and gives rise to the band at lower wavenumber (around
3460 cm−1). The variation in the ν(OH) due to the perturbation
by CO [ꢀν(OH)] can be calculated based on these attributions.
Fig. 3. FTIR spectra of 133 Pa of pyridine adsorbed on the supports and fol-
lowed by evacuation at different temperatures.
probes to weak BAS), such sites are revealed on alumina, in
accordance with previous findings [25].
On Nb2O5, adsorption of Py, 2,6-DMP, and CO indicates the
presence of LAS. These sites are weaker than those of alumina
(Table 2). This is illustrated by the wavenumber and stability
of the coordinated pyridine and 2,6-DMP species. As an exam-
ple, after evacuation at high temperature (623 K), no pyridine
(band at 1607 cm−1 in Fig. 3) remained adsorbed on the nio-
bia surface, whereas even after evacuation at 673 K the band
characteristic of Py on LAS at 1622 cm−1 remained on the alu-
mina spectrum. The detection of a band at 1627 cm−1 after
Py adsorption on Nb2O5 must be mentioned. Indeed, a band

326
A.S. Rocha et al. / Journal of Catalysis 252 (2007) 321–334
Fig. 5. (A) FTIR spectra of increasing doses of CO adsorbed at 100 K on Nb O . (B) Comparison of FTIR spectra of 133 Pa of CO adsorbed at 100 K on the
2
5
supports.
The values of 210 and 193 cm−1 thus obtained demonstrate that
the hydroxyl groups giving rise to the band at 3670 cm−1 are
the support, confirming that BAS are stronger on niobia than
on alumina.
stronger BAS than those giving rise to the band at 3719 cm−1
.
The spectra obtained after adsorption of the three probes on
Nb2O5/Al2O3 present bands characteristic of interaction with
LAS as strong as on alumina. In the case of Py adsorption
on Nb2O5/Al2O3, the massif attributed to the ν8a stretching
mode has at least two components as observed for alumina,
indicating LAS of different strength. The shift toward lower
wavenumbers of the ν8a massif compared with alumina may
be explained by preferential coverage of strong LAS of Al2O3
by niobium.
A ꢀν(OH) value of 193 cm−1, associated with a ν(CO) of
2163 cm−1, is in accordance with the previously reported corre-
lation between the ꢀν(OH) and the ν(CO) band corresponding
to a OH...CO adduct [30]. The position of the ν(CO) band
for the second type of hydroxyl is somewhat hazardous, due
to close proximity of the band characteristic of weak LAS.
The same study can be conducted for the other supports
(Fig. 5B). The ꢀν(OH) increases with the niobia content in

A.S. Rocha et al. / Journal of Catalysis 252 (2007) 321–334
327
Table 2
Characteristic IR bands of basic probe molecules adsorbed on Lewis acid site (LAS) and Brønsted acid sites (BAS) for the three supports (value given for a probe
pressure of 133 Pa)
Support
Pyridine
LAS (cm
2,6-DMP
CO
−1
−1
−1
−1
−1
−1
)
)
BAS (cm
)
LAS (cm
, ν
)
BAS (cm
, ν
)
LAS (cm
ν(CO)
2185
)
BAS (cm
ν(CO)
2153
ν
ν
ν
ν
ν
ν
8a 8b
8a
19b
8a
19b
8a 8b
Al O
1622
1614
1449
1616, 1582
1653, 1628
2
3
Nb O
1607
1446
1448
1642
1648
1543
1545
1603, 1582
1613, 1582
1645, 1628
1648, 1627
2174
2181
2163
2158
2
5
Nb O /Al O
2 2 3
1619
1610
5
Fig. 6. Comparison of FTIR spectra of 133 Pa of CO adsorbed at 100 K on
sulfided NiMo/Al catalysts.
Fig. 7. Comparison of FTIR spectra of 133 Pa of CO adsorbed at 100 K on
sulfided NiMo/NbAl catalysts.
For CO and 2,6-DMP adsorption on Nb2O5/Al2O3, the
bands associated with BAS have wavenumbers in between
those of alumina and niobia, indicating that these BAS are of
intermediate strength compared with the pure supports. Indeed,
according to the band position for the three probes (Table 2),
BAS sites are weaker on Nb2O5/Al2O3 than on pure Nb2O5,
but nonetheless these BAS are strong enough to protonate Py,
in contrast to the BAS of Al2O3.
According to the Raman results, niobia supported on alu-
mina formed polymerized polyhedra, as observed for a mono-
layer. A study by Jehng and Wachs [31] on similar system
found that BAS could be due to slightly distorted surface nio-
bium oxide octahedral, as well as NbO7 and NbO8, structures.
LAS correspond to highly distorted surface NbO6 octahedra
sites.
3.3. FTIR spectroscopy of adsorbed CO on sulfided catalysts
The sulfided NiMo catalysts were characterized by FTIR
spectroscopy of adsorbed CO to investigate the nature and the
amount of surface sites. The CO adsorption was performed at
about 100 K, to increase the coverage of the probe molecule on
weak sites such as the OH groups of the support and the sites of
the metal phase. The sulfided catalysts deposited on pure niobia
are not sufficiently transparent to the IR beam to allow transmit-
tance analysis of self-supported wafers.
Figs. 6 and 7 compare the spectra of CO adsorbed on the
sulfided NiMo catalysts supported on pure alumina and on nio-
bia/alumina, respectively, for adsorption of 133 Pa of CO at
equilibrium. Fig. 8 shows the spectra obtained after adsorption
of increasing small doses of CO on the sulfided 25% and 50%
NiMo catalyst supported on alumina and niobia/alumina.
Figs. 6 and 7 show that the presence of niobium in the
catalysts induces an increase of the acid strength of the OH
Resuming the acid properties of the supports, alumina has
strong LAS and weak BAS, niobia has weak LAS and strong
BAS and niobia/alumina has intermediate LAS and BAS sites.

328
A.S. Rocha et al. / Journal of Catalysis 252 (2007) 321–334
Fig. 8. FTIR spectra of increasing doses of CO adsorbed at 100 K on sulfided catalysts.
groups of the support. Indeed, on supported NiMo catalysts,
the CO/OH band is observed at higher wavenumbers on
NiMo/NbAl [ν(CO) = 2159 cm−1] than on NiMo/Al [ν(CO =
2154 cm−1]. This result is in agreement with previous findings
of studies performed on supports alone. On alumina, a shift
of the CO/OH band toward the high wavenumbers is observed
going from the pure Ni to the pure Mo catalysts. This shift indi-
cates an increasing acidity of hydroxyl groups going from Ni to
Mo catalysts. The origin of this strengthening is not straightfor-
ward; interaction between sulfide SH groups and support OH
can be proposed. There is also an effect of CO coverage, be-
cause the shift is not observed for small doses (Fig. 8).
The interactions of CO with the sulfide phases are charac-
terized by the presence of bands in the 2130–2080 cm−1 range
(Table 3). Regarding the spectrum of Mo/Al (Fig. 6), the band
at 2114 cm−1 is attributed to CO interacting on molybdenum
sulfide, in agreement with previous studies [17,32]. From the
spectra of nickel-rich catalysts, 75% NiMo/Al and Ni/Al, the
band at 2094 cm−1 can be attributed to CO interacting with
nickel sulfide sites [33]. As shown in Figs. 6 and 7, the interac-

A.S. Rocha et al. / Journal of Catalysis 252 (2007) 321–334
329
Table 3
−1
Characteristic IR bands (cm ) of CO adsorbed on the metallic phase sup-
ported on alumina and niobia/alumina supports
Support
Al O
Mo
NiMo
Ni
2114
2119
+5
2127–2086
2128–2086
0
2094
2100
+6
2
3
Nb O /Al O
2
5
2 3
a
ꢀν(CO)
a
Variation of the ν(CO) band position going from the Al O support to the
2
3
Nb O /Al O support.
2
5
2 3
tion of CO with NiMo phase gives rise to a complex massif that
contains different bands. Analysis of the spectra corresponding
to adsorption of increasing doses allows discrimination of the
various components of this massif (Fig. 8). Thus, in the spec-
trum of CO adsorbed on 25% NiMo/Al, a broad band with two
maxima around 2116 and 2128 cm−1 and a shoulder around
2086 cm−1 can be seen. Adsorption of CO on the 50% NiMo/Al
catalysts also leads to a wide band with at least three compo-
nents, whereas the shoulder at around 2086 cm−1 is more in-
tense than that on 25% NiMo/Al. Thus, the two bands at around
2128 and 2086 cm−1, which are not present on unicomponent
catalysts, are attributed to the NiMoS phase, in accordance with
previous findings [32–34]. On 50% NiMo/Al, the greater in-
tensity of the band at 2086 cm−1 suggests a higher amount of
mixed phase on this catalyst. Nevertheless, the detection of a
signal at about 2116 cm−1 shows that not all of the sulfide phase
is promoted on these two catalysts.
Fig. 7 shows that the pure Mo sulfide phase supported on
NbAl is characterized by a band at 2119 cm−1, whereas the
Ni sulfide phase on this support is characterized by a band
at 2100 cm−1 (as deduced from analysis of the spectra for
Ni/NbAl and 75% NiMo/NbAl). Comparing these wavenum-
bers with those reported for Mo/Al and Ni/Al shows that
on NbAl-supported catalysts, the frequency of CO interacting
with these two sulfide phases is shifted toward higher values
(∼ +5 cm−1) (Table 3).
Fig. 9. Thiophene HDS activity of NiMo catalysts and supports.
catalyst supported on NbAl (Fig. 8). Thus, formation of NbSx
species is not evidenced by the IR study.
3.4. Thiophene hydrodesulfurization activity
Supported NiMo catalysts and supports alone were evalu-
ated in thiophene HDS. Because the catalysts have considerably
different surface areas, Fig. 9 presents the rate per square me-
ter of catalyst. The activities of pure supports are represented
by open symbols. As reported previously [2,4,15], alumina-
supported NiMo mixed sulfides are more active than the pure
nickel and molybdenum sulfides. This activity pattern results
from the synergy between nickel and molybdenum in contact
with each other on an adequate support, after sulfidation. In
agreement with previous results, activity is maximum for the
25% and 50% NiMo/Al catalysts. This suggests that the NiMoS
phase is present in higher amounts in these latter catalysts.
Activity was lower for the niobia/alumina and niobia-
supported catalysts with the same NiMo content compared with
the catalysts deposited on alumina. Thus, it can be deduced that
niobia hinders the formation of the NiMoS phase, giving rise to
the synergetic effect.
Another interesting observation is that the addition of niobia
increases the activity of pure nickel catalysts. Thus, the activity
of Ni on pure niobia is even greater than that of the correspond-
ing Mo catalyst. However, on the niobium catalyst series, the
contribution of niobium to the catalyst activity is noticeable.
Indeed, previous studies showed that niobium is partially sul-
fided, and niobium sulfide has significant activity [15,17]. Thus,
to better estimate the activity of the NiMo sulfide phase for
HDS of thiophene, support activities were subtracted from cat-
alyst activities. Table 4 presents these results for all materials.
Column 1 presents the same results as for Fig. 9, that is HDS
activity of the catalysts per square meter; column 2 presents
the catalyst activity minus the support activity; and column 3
presents the results of column 2 per total metal content. We
As can be seen in Figs. 7 and 8, interaction of CO with
NiMo catalysts supported on NbAl gives rise to a main band at
∼2118 cm−1. This demonstrates that unpromoted Mo sites are
predominant on 25% and 50% NiMo/NbAl. However, analysis
of the spectrum corresponding to the first CO doses provides
evidence of a band at ∼2128 cm−1, demonstrating the forma-
tion of NiMoS mixed phase. Note that this band is present only
as a weak shoulder on the spectrum corresponding to a CO
adsorption of 133 Pa, indicating that only a weak fraction of
the sulfide phase sites corresponds to the mixed NiMoS phase.
In parallel, the shoulder observed at ∼2086 cm−1 on alumina-
supported catalysts is only slightly visible on the spectra of the
25% and 50% NiMo/NbAl; this confirms the previous observa-
tion. In addition, note that although the bands attributed to the
NiMoS phase are weak on the niobia–alumina support, clearly
no shift due to the support can be seen, in contrast with what
was observed for the single metal catalysts (Table 3).
Finally, note that the band at 2135 cm−1 seen in the case
of Mo/NbAl that we have attributed to NbSx [17] cannot be
distinguished on the spectra of sulfided 25% and 50% NiMo

330
A.S. Rocha et al. / Journal of Catalysis 252 (2007) 321–334
Table 4
Thiophene HDS activity of NiMo catalysts and supports
a
b
Catalyst
r
Difference of activity
Difference of activity
HDS
−1
−6
−1 −2
−6
−1 −2
−1
)
(10 mol h
m
)
(10 mol h
m
)
(mol mol
th
h
metal
Al O
Mo/Al
25% NiMo/Al
50% NiMo/Al
75% NiMo/Al
Ni/Al
0.0
27.0
124.6
119.5
43.9
9.5
–
2
3
27.0
124.6
119.5
43.9
9.5
3.51
15.52
15.82
5.77
1.15
Nb O /Al O
4.4
–
2
5
2 3
Mo/NbAl
28.3
70.8
70.1
50.0
28.3
42.3
50.0
57.7
68.0
66.7
58.1
23.9
66.4
65.6
45.6
23.9
–
2.55
7.31
7.52
5.42
2.82
25% NiMo/NbAl
50% NiMo/NbAl
75% NiMo/NbAl
Ni/NbAl
Nb O
2
5
Mo/Nb
7.7
1.00
1.93
3.11
3.20
2.39
25% NiMo/Nb
50% NiMo/Nb
75% NiMo/Nb
Ni/Nb
15.4
25.7
24.4
15.8
a
−6
−1 −2
Difference of activity: r
− r
in 10 mol h
m
.
(catalyst)
(support)
−1
b
−1
.
Difference of activity per mole of metal (Ni + Mo) in mol
mol
h
thiophene
metal
should consider that this calculation underestimates the contri-
bution of metal phases, because the supports in the catalysts are
partially covered by the metals. In contrast, the support activity
can be modified by the presence of metal. Indeed, in the case
of niobium, sulfidation is known to increase in the presence
of metal, and niobium sulfide is known to be active in HDS
[15,17].
Based on the catalyst activity (column 1), molybdenum on
niobia is twice as active as Mo/Al and Mo/NbAl, but the sig-
nificant contribution of niobia to catalyst activity is responsible
for this superior activity, and in fact the activity per Mo atom
decreases with increasing niobium content (column 3).
Regarding the Ni catalysts, niobium-containing supports
increase nickel sulfide activity per metal atom. Ni/Nb and
Ni/NbAl catalysts are approximately two times more active
than Ni/Al (column 3).
Comparing the results in the last column for the NiMo cata-
lysts shows that the synergetic effect between nickel and molyb-
denum is maximum on alumina (25% and 50% NiMo/Al).
However, the higher activities of 25% and 50% NiMo/NbAl
compared with those of the corresponding monometallic cata-
lysts indicate that this effect is present in these catalysts as well.
On niobia, the synergetic effect appears for the 50% and 75%
NiMo/Nb, but it remains very low.
elimination (DDN), (ii) saturated hydrocarbons denitrogenated
via hydrogenation followed by elimination (HYG + HYD);
and (iii) nitrogen-containing products formed by dispropor-
tionation (DIS) [19]. The first route is very interesting, be-
cause the denitrogenation occurs with low hydrogen consump-
tion (Scheme 1). From 2,6-DMA, it leads to xylene formation,
which also occurs to some extent via the hydrogenation route
(HYG). In what follow, XYL represents the sum DDN + HYG.
The three routes XYL, HYD, and DIS are strictly parallel [19].
Previously, we found that niobium increased the selec-
tivity of molybdenum sulfide to xylene production in 2,6-
dimethylaniline HDN [17]. For the three routes, sulfided niobia
presented significant activity, whereas pure alumina was inac-
tive.
In the present work, we compared the activity of pure
nickel (Ni/supports), pure molybdenum (Mo/supports), and
25% NiMo/supports sulfided catalysts in 2,6-dimethylaniline
HDN. We also evaluated the supports alone.
3.5.1. Activity in the presence of H
2S
Table 5 reports the rate for each reaction route: HYD, XYL
and DIS in presence of H2S. In the same manner as for thio-
phene HDS, to estimate the activity of the supported phase the
activity of the support alone is subtracted from the catalyst ac-
tivity (columns 5–7).
First, we can observe the effect of the support on monometal-
lic catalysts. As reported previously, molybdenum on niobia
and niobia/alumina are very active for xylene production com-
pared with molybdenum on alumina. The activity of pure
molybdenum catalyst for the DIS route also increases with the
niobium content in the support, whereas the activity for HYD
is modified only slightly by the nature of the support (activity
by metal atom shows a slight decrease with increasing niobium
content). Thus, the selectivity of Mo catalysts for the XYL is
In conclusion, the synergetic effect of nickel and molybde-
num for thiophene HDS decreases with increasing niobia con-
tent in the support.
3.5. 2,6-Dimethylaniline (2,6-DMA) hydrodenitrogenation
activity
In previous work, it has been proposed that HDN of alkylani-
lines over NiMo catalysts can generate three types of products:
(i) aromatic hydrocarbons denitrogenated via direct ammonia

A.S. Rocha et al. / Journal of Catalysis 252 (2007) 321–334
331
Scheme 1. Reaction network of 2,6-dimethylaniline decomposition.
Table 5
respective monometallic catalysts, demonstrating a synergetic
effect between the two metals. The hydrogenation activity of
the 25% NiMo catalyst on niobia/alumina is greater than that on
alumina, but the activity per metal content without the support
contribution is the same in the two catalysts. A synergetic effect
between Ni and Mo for the HYD route is not observed on pure
niobia catalysts. However, the HYD activity of 25% NiMo/Nb
is significant and can be related to the stronger activity of pure
nickel on this support compared with other supports.
For the xylene route, the activity of NiMo catalysts increases
with increasing niobium content. However, on pure niobia, the
activity for XYL is much lower for the 25% NiMo than for the
pure molybdenum. We previously proposed that this special ac-
tivity of the molybdenum and niobium sulfide catalysts would
be a result of a synergy between the Mo sulfide phase and the
Nb sulfide (or oxysulfide), responsible for the C–N bond cleav-
age [17]. Apparently, the addition of nickel inhibits this effect.
For the disproportionation route, the activity of 25% NiMo
catalysts increases with increasing niobium content in the sup-
port, as was seen for monometallic catalysts. No synergetic
effect between Mo and Ni was observed for this route, what-
ever the support.
Activity of NiMo catalysts and supports in decomposition of 2,6-dimethyl-
aniline in presence of H S
2
Catalyst
Activity
(10 mol h
Difference of activity
−1
−6
−1 −2
−1
m
)
(mol
mol
h
)
DMA
metal
r
r
r
r
r
r
HYD
XYL
HYD
XYL
DIS
DIS
Al O
2
0
0
0
–
–
–
3
Mo/Al
25% NiMo/Al
Ni/Al
5.03
5.89
0.88
0.17
5.47
6.89
2.00
2.16
5.38
5.30
5.67
0.85
0.36
0.05
0.05
7.23
1.15
0.25
0.64
1.14
1.05
0.13
0.14
2.22
1.60
0.43
1.58
3.64
3.04
2.82
0.65
0.73
0.11
–
0.11
0.04
0.01
–
0.15
0.13
0.02
–
Nb O /Al O
2
5
2 3
Mo/NbAl
25% NiMo/NbAl
Ni/NbAl
0.56
0.74
0.22
–
0.77
0.12
0.02
–
0.22
0.16
0.03
–
Nb O
2
5
Mo/Nb
25% NiMo/Nb
Ni/Nb
10.97
3.40
1.54
0.42
0.39
0.53
1.34
0.35
0.14
0.27
0.18
0.19
greatly improved on the niobium-containing support at the ex-
pense of the HYD route. For nickel catalysts, the activity for
the three routes increases with increasing niobium content on
the support, but the selectivity for HYD remains the greatest
for the three supports, in contrast with the case of Mo catalysts.
Comparison between the two metals shows that for the three
routes, Mo is more active than Ni. Only on pure niobia support
is the hydrogenation activity of pure nickel catalyst comparable
to that of pure Mo catalyst. The greater hydrogenation activity
of Ni/Nb compared with Ni/Al is in accordance with the greater
activity for HDS of thiophene observed in the present study
and reported previously [15]. This greater activity of Ni cata-
lyst when supported on niobia compared with that supported on
alumina was attributed to an increase in the nickel present on
niobia surface, as demonstrated by XPS data [15], due to either
an absence of Ni loss in the support or the formation of smaller
Ni2S3 particles compared with alumina.
3.5.2. Effect of the absence of H2S in the reaction mixture
Table 6 reports the HDN activities of all supports and cat-
alysts in the absence of H2S. As in the presence of H2S, alu-
mina is inactive. The two niobium-based supports exhibit only
a slight decrease in activity in all routes in the absence of H2S
compared with the experiments carried out in the presence of
H2S. The hydrogenation activity of pure molybdenum catalysts
is only slightly modified by the absence of H2S. For the dispro-
portionation route, a moderate decrease in activity can be seen.
In contrast, the xylene formation on pure molybdenum catalysts
is significantly greater without H2S on the three supports, with
this effect relatively more pronounced on alumina (Fig. 10).
The activities of pure nickel catalysts (per metal atom) on all
supports are similar with and without H2S for the three routes.
Regarding the bimetallic catalysts, we found that 25% and
25% NiMo/NbAl are more active for the HYD route than the

332
A.S. Rocha et al. / Journal of Catalysis 252 (2007) 321–334
Fig. 10. 2,6-Dimethylaniline HDN: activity in xylene formation route per metal content of molybdenum based catalysts, in presence and absence of H S.
2
Table 6
active catalysts. Strong interaction of pure metal with niobia
is also evidenced by the CO results; compared with alumina-
supported catalysts, on niobia–alumina, the ν(CO) band of CO
in interaction with the metal is blue-shifted. Thus, the absence
of shift of the ν(CO) band of CO in interaction with the NiMoS
phase when the support is niobia–alumina instead of alumina
suggests that the NiMoS phase is supported on the part of alu-
mina not covered by niobia.
Activity of NiMo catalysts and supports in decomposition of 2,6-dimethyl-
aniline in absence of H S
2
Catalyst
Activity
(10 mol h
Difference of activity
(mol(DMA) mol(metal)
−6
−1 −2
−1 −1
m
)
h
)
r
r
r
DIS
r
r
XYL
r
DIS
HYD
XYL
HYD
Al O
2
0
5.63
11.87
0.40
0.11
0
0
–
–
–
3
Mo/Al
25% NiMo/Al
Ni/Al
7.99 0.49 0.73
1.05 0.18 1.48
0.01 0.04 0.05
1.04
0.13
0.00
–
0.06
0.02
0.01
–
0.11
0.06
0.01
–
The higher activities of 25% NiMo/Al and 25% NiMo/NbAl
catalysts for hydrogenation of 2,6-DMA compared with mono-
metallic catalysts are in agreement with the HDS results and
reveal the synergetic effect between the two metals. This com-
parable behavior suggests that the active sites for hydrogenation
of nitrogenated compound are also the active sites for HDS,
that is, the NiMo sulfide. For this HYD route, the effect of the
presence of H2S is moderate. However, it can be seen that the
synergy effect between Ni and Mo is stronger in absence of
H2S. This can be explained by a change in the reaction mech-
anism as proposed by van Gestel et al. [35]. In the absence of
H2S in the flow, hydrogenation occurs after heterolytic dissoci-
ation of H2. In the presence of H2S in the flow, H2S can act as
a poison for the site of dissociation of H2 or can be dissociated
to SH− and H+ species and thus lead to hydrogenation.
Disproportionation reactions are typically catalyzed by
BAS. The high activities of niobium-based catalysts for this
route are in agreement with our IR characterization results
obtained by pyridine, 2,6-DMP, and CO. Indeed, niobia and
niobia/alumina present strong BAS, whereas alumina has only
weak BAS. Thus, we propose that the BAS related to nio-
bia are responsible for the high disproportionation activities of
niobium-based catalysts.
Nb O /Al O
0.05 0.08
–
2
5
2 3
Mo/NbAl
5.32 26.49 1.08 0.56
2.82
1.54
0.02
–
25% NiMo/NbAl 10.83 14.03 0.63 1.18
Ni/NbAl
Nb O
2.93
1.60
0.22 0.16 0.33
0.68 1.21
–
2
5
Mo/Nb
25% NiMo/Nb
Ni/Nb
4.94 44.96 2.73 0.43
5.78 37.60 2.61 0.52
4.59
5.74
4.63
0.10
0.20
0.18
0.15
1.31 2.18 0.45
As observed for pure molybdenum catalysts, the activity of
25% NiMo catalysts in disproportionation route decreases with-
out H2S, whereas the xylene formation markedly increases. For
HYD, activity increases in absence of H2S, and the synergetic
effect between Ni and Mo can be seen on the three supports in
this case.
4. Discussion
Our findings clearly demonstrate the effect of the support na-
ture on the HDS activity of NiMo catalysts. The synergy effect
of Ni and Mo is much lower on niobia-containing supports than
on alumina, in accordance with the findings of Faro and dos
Santos [15]. In all likelihood, nickel and molybdenum interact
preferentially with niobium compared with each other, inhibit-
ing the formation of the NiMoS phase. The CO FTIR results are
in agreement with this hypothesis; the band at 2128 cm−1 and
the shoulder at 2086 cm−1 which may be attributed to CO ad-
sorbed on the NiMo sulfide sites, are more intense for the most
The disproportionation trend for monometallic catalysts in-
dicates that molybdenum sulfide is more active than nickel sul-
fide. This result is in accordance with previous results of van
Gestel et al. verifying that H2S dissociation is involved in the
generation of surface acidity on Mo(Co)/Al2O3 catalysts [19].
Their study showed that such H2S dissociation could be re-

A.S. Rocha et al. / Journal of Catalysis 252 (2007) 321–334
333
sponsible for protonation of 2,6-DMA molecule, leading to an
intermediate species for the disproportionation reaction. In con-
trast, Ni has been shown to inhibit the activity of Mo/Al2O3
even in the presence of H2S, and thus H2S dissociation is hin-
dered on Ni sites [35]. Thus, the higher activity of Mo catalysts
compared with Ni catalysts can be explained by the former’s
ability to dissociate H2S. This ability can be linked with the
strength of metal LAS demonstrated by the IR results. The fre-
quency of CO adsorbed on sulfided molybdenum is higher than
that of CO adsorbed on sulfided nickel, indicating that MoSx is
a stronger LAS than NiSx. Strong LAS could be responsible for
H2S dissociation on Mo catalysts.
Alternatively, we can propose a strengthening of the OH
BAS with Mo. Indeed, on Al2O3, a slight blue shift of the band
characteristic of CO/(OH) going from pure Ni to pure Mo can
be seen in Fig. 6. However, the involvement of H2S dissociation
in the DIS mechanism is supported by the fact that the presence
of H2S in the reaction medium favors the DIS route. The effect
is more pronounced for catalysts supported on alumina, which
has on its own the weaker BAS.
As mentioned previously, the DDN route leading to xylene
formation is a particularly interesting route for denitrogena-
tion, because it involves less hydrogen consumption than other
routes. The proposed mechanism for DDN implies a C–N bond
rupture on unpromoted molybdenum sulfide sites. Accordingly,
pure molybdenum catalysts are the most active for this route.
The Mo sites involved in the DDN route are described as
highly unsaturated edge sites that can be easily poisoned by
H2S. This hypothesis is in concordance with our results, be-
cause activities for DDN increase significantly without H2S
for all Mo-containing catalysts. However, it must be pointed
out that when supported on niobia-containing supports, pure
nickel catalyst appears slightly active for XYL formation. Be-
cause pure niobia is itself active for the XYL route, we attribute
the apparent activity of Ni catalysts supported on niobia or nio-
bia/alumina to enhanced niobia activity rather than to nickel
activity.
the lower amount of molybdenum and the promoting effect of
Mo by Ni, which inhibits Mo activity for DDN [19]. The first
factor should decrease the activity of the 25% NiMo/Al to 75%
of the activity of the Mo/Al [expected activity, 0.78 mol(2,6-
DMA) mol(metal)−1 h−1]. The promotion of Mo by Ni explains
the lower observed activity, which corresponds to only around
17% of the expected activity. On niobia-supported catalysts, the
decrease in the XYL route is much lower when going from Mo
to 25% NiMo; the activity decreases by a factor of 1.25. This ac-
tivity decrease is the same order of magnitude as the decrease
in Mo content. In this case, intrinsic activity of molybdenum
is not affected by nickel. Accordingly, we have demonstrated
the low promotion of Mo by Ni on this support for the HYD
route of 2,6-DMA and for thiophene HDS. For Nb2O5–Al2O3,
as for Al2O3, we can suppose that the decrease in activity for
the XYL route is due to the lower amount of molybdenum and
to the promoting effect of Mo by Ni on alumina not covered by
niobia. The first cause should decrease the activity of the 25%
NiMo/NbAl to 75% of the activity of the Mo/NbAl [expected
activity, 2.11 mol(2,6-DMA) mol(metal)−1 h−1]. The interest-
ing point is that we can estimate the repartition of molybdenum
on alumina not covered by niobia and on niobia. Indeed, on
alumina, we can speculate that the promotion of Mo by Ni
induces a reduction of activity of 83%, whereas no further de-
crease of activity is expected on niobia, because no promotion
is expected. By this method, we obtain that 67% of Mo is on
niobia and 33% is on alumina, and thus that the repartition of
Mo is homogeneous on all the surface of the support, because
70% alumina coverage by niobia has been estimated by IR. Ac-
cordingly, for the monometallic Mo catalysts, the activity of the
Mo/NbAl is near the sum of 67% of the activity of Mo/Nb and
33% of the activity of Mo/Al. The coherence of these calcu-
lations validates the previously mentioned hypothesis that the
promoted phase NiMo is formed only on alumina not covered
by niobia.
5. Conclusion
In contrast, in any case, the activity of support alone for
XYL formation is much lower than that of the molybdenum-
containing catalyst. However, the nature of the support has a
significant effect on the catalyst activity for XYL formation.
Xylene formation is greatest on niobia-supported catalysts in
both the absence and presence of H2S. Because the electronic
properties of pure metal sulfides are modified on niobia, as re-
vealed by the shift of the IR band associated with CO in interac-
tion with supported metal LAS, we propose that the strong ac-
tivity for XYL formation on niobia-supported catalyst is linked
to this phenomenon. Comparing the favorable effect of niobia
support for the XYL route with the decreased synergy effect on
niobia-supported catalyst for HDS and HYD routes, we can in-
fer that both metals interact preferentially with niobium than
with each other.
The present study has shown that using niobia as support,
the activity of NiMo catalysts in thiophene HDS and in HDN of
2,6-DMA is no longer promoted by the synergy between Ni and
Mo. The lack of synergy between molybdenum and nickel on
niobia is explained by the strong interaction of each metal with
niobia at the expense of interaction with each other. Neverthe-
less, niobia is an interesting support for the HDN of 2,6-DMA
because it favors the XYL route, a way to remove nitrogen with
low H2 consumption. The activity for the XYL route on niobia
is linked to the electron-deficient character of the Mo sulfide
site, as revealed by CO adsorption followed by IR.
Therefore, using niobia/alumina as a support is a good com-
promise. Indeed, this study shows that on this support, the Ni-
MoS phase is formed on the fraction of alumina not covered
by niobia, whereas the molybdenum sulfide phase is located on
the niobia. Consequently, compared with conventional NiMo
supported on alumina, catalysts supported on niobia/alumina
demonstrate acceptable HDS and HYD properties, along with
high activity for direct denitrogenation.
Another interesting result for the XYL route is the effect of
promotion of Mo by Ni versus the support nature. Indeed, on
alumina, the XYL activity is divided by 8 when going from
Mo to 25% NiMo. In this support, we can suppose that the de-
crease in activity for the XYL route is due to two phenomena:

334
A.S. Rocha et al. / Journal of Catalysis 252 (2007) 321–334
Acknowledgments
[13] J.G. Weissman, Catal. Today 28 (1996) 159.
[14] V. Gaborit, N. Allali, C. Geantet, M. Breysse, M. Vrinat, M. Danot, Catal.
Today 57 (2000) 267.
[15] A.C. Faro Jr., A.C.B. dos Santos, Catal. Today 118 (2006) 402.
[16] N. Allali, A.M. Marie, M. Danot, C. Geantet, M. Breysse, J. Catal. 156
(1995) 279.
[17] A.S. Rocha, A.C. Faro Jr., L. Oliviero, M.A. Lélias, A. Travert, J. van
Gestel, F. Maugé, Catal. Lett. 111 (2006) 27.
[18] L.J. Burcham, J. Datka, E. Wachs, J. Phys. Chem. B 103 (1999) 6015.
[19] J. van Gestel, C. Dujardin, F. Maugé, J.C. Duchet, J. Catal. 202 (2001) 78.
[20] G. Crépeau, V. Montouillout, A. Vimont, L. Mariey, T. Cseri, F. Maugé,
J. Phys. Chem. B 110 (2006) 15172.
[21] J.M. Jehng, I.E. Wachs, J. Mol. Catal. 67 (1991) 369.
[22] H. Knozinger, C. Ratnasamy, Catal. Rev. Sci. Eng. 17 (1978) 31.
[23] E.P. Parry, J. Catal. 2 (1963) 371.
[24] S. Khabtou, T. Chevreau, J.C. Lavalley, Micropor. Mater. 3 (1994) 133.
[25] L. Oliviero, A. Vimont, F. Romero, M. Gaillard, F. Maugé, J.C. Lavalley,
Phys. Chem. Chem. Phys. 7 (2005) 1861.
[26] P.A. Jacobs, C.F. Heylen, J. Catal. 34 (1974) 267.
[27] C. Morterra, G. Cerrato, G. Meligrana, Langmuir 17 (2001) 7053.
[28] M.I. Zaki, H. Knözinger, Mater. Chem. Phys. 17 (1987) 201.
[29] K. Hadjiivanov, G. Vayssilov, Adv. Catal. 47 (2002) 307.
[30] O. Cairon, T. Chevreau, J.C. Lavalley, J. Chem. Soc. Faraday Trans. 94
(1998) 3039.
The CNRS supported this work through the International
Program for Scientific Cooperation with South America (proj-
ect “Hydrotraitement et Développement Durable”); PETRO-
BRAS also provided partial financial support. The authors
thank Guillaume Clet for the Raman characterization, Valérie
Ruaux and Wenbin Chen for their technical support, and the
NUCAT/UFRJ for the XRD measurements.
References
[1] L. Zhang, G. Karakas, U.S. Ozkan, J. Catal. 178 (1998) 457.
[2] B.M. Vogelaar, P. Steiner, A.D. van Langeveld, S. Eijsbouts, J.A. Moulijn,
Appl. Catal. A 251 (2003) 85.
[3] S.H. Yang, C.N. Satterfield, J. Catal. 81 (2003) 168.
[4] P. Baeza, M.S. Ureta-Zañartu, N. Escalona, J. Ojeda, F.J. Gil-Llambías, B.
Delmon, Appl. Catal. A 274 (2004) 303.
[5] D. Zuo, M. Vrinat, H. Nie, F. Maugé, Y. Shi, M. Lacroix, D. Li, Catal.
Today 93–95 (2004) 751.
[6] C. Dujardin, M.A. Lélias, J. van Gestel, A. Travert, J.C. Duchet, F. Maugé,
Appl. Catal. A 322 (2007) 46.
[7] M. Breysse, P. Afanasiev, C. Geantet, M. Vrinat, Catal. Today 86 (2003)
5.
[8] L. Qu, M. Flechsenhar, R. Prins, J. Catal. 217 (2003) 284.
[9] D. Ferdous, A.K. Dalai, J. Adjaye, Fuel 85 (2006) 1286.
[10] F. Dumeignil, K. Sato, M. Imamura, N. Matsubayashi, E. Payen, H. Shi-
mada, Appl. Catal. A 287 (2006) 135.
[31] J.M. Jehng, I.E. Wachs, Catal. Today 8 (1990) 37.
[32] F. Maugé, A. Vallet, J. Bachelier, J.C. Duchet, J.C. Lavalley, J. Catal. 162
(1996) 88.
[33] M. Angulo, F. Maugé, J.C. Duchet, J.C. Lavalley, Bull. Soc. Chim.
Belg. 96 (1987) 925.
[34] A. Travert, C. Dujardin, F. Maugé, E. Veilly, S. Cristol, J.F. Paul, E. Payen,
J. Phys. Chem. B 110 (2006) 1261.
[35] J. van Gestel, J. Leglise, J.C. Duchet, Appl. Catal. A 92 (1992) 143.
[11] M. Ziolek, Catal. Today 78 (2003) 47.
[12] M.A. Abdel-Rehim, A.C.B. dos Santos, V.L.L. Camorim, A.C. Faro Jr.,
Appl. Catal. A 305 (2006) 211.