Kinetic study of the hydrodenitrogenation of carbazole over bulk molybdenum carbide

Article abstract of DOI:10.1016/S0021-9517(03)00072-1

Hydrodenitrogenation (HDN) of N-containing compounds (carbazole in this paper) in the presence of a small amount of sulfur (50 ppm) requires a strong catalytic hydrogenating function. Bulk molybdenum carbide, Mo₂C, behaves like a noble metal and has strong hydrogenating properties and can catalyse deep HDN if the S content is not too high. A kinetic investigation monitored both the hydrogenating function through carbazole consumption and the HDN of carbazole through two kinetically coupled cycles. Biphenyl was not observed. Kinetic coupling occurs by means of adsorbed orthocyclohexylaniline. The first route is the direct HDN reaction, leading to cyclohexylbenzene, while the second hydrogenation route leads to bicyclohexyl. At 633 K, a zero-order reaction is observed with respect to carbazole; total HDN is observed for 0.4 s, and the selectivity ratio of bicyclohexyl to cyclohexylbenzene is equal to 9.

Full text of DOI:10.1016/S0021-9517(03)00072-1

Journal of Catalysis 218 (2003) 24–31
www.elsevier.com/locate/jcat
Kinetic study of the hydrodenitrogenation of carbazole
over bulk molybdenum carbide
a,c
a
b
Agnieszka Szyma n´ ska, Marek Lewandowski, Céline Sayag,
b,∗
and Gérald Djéga-Mariadassou
a
Institute of Coal Chemistry of PAN, Sowi n´ skiego 5, 44-121 Gliwice, Poland
b
Laboratoire Réactivité de Surface, UMR CNRS 7609, casier 178, Université P. & M. Curie, 4 Place Jussieu, 75252 Paris Cedex 05, France
c
Silesian Technical University in Gliwice, Department of Chemistry, M. Strzody 9, 44-100 Gliwice, Poland
Received 10 July 2002; revised 7 November 2002; accepted 17 February 2003
Abstract
Hydrodenitrogenation (HDN) of N-containing compounds (carbazole in this paper) in the presence of a small amount of sulfur (50 ppm)
requires a strong catalytic hydrogenating function. Bulk molybdenum carbide, Mo C, behaves like a noble metal and has strong hydrogenat-
2
ing properties and can catalyse deep HDN if the S content is not too high. A kinetic investigation monitored both the hydrogenating function
through carbazole consumption and the HDN of carbazole through two kinetically coupled cycles. Biphenyl was not observed. Kinetic cou-
pling occurs by means of adsorbed orthocyclohexylaniline. The first route is the direct HDN reaction, leading to cyclohexylbenzene, while
the second hydrogenation route leads to bicyclohexyl. At 633 K, a zero-order reaction is observed with respect to carbazole; total HDN is
observed for 0.4 s, and the selectivity ratio of bicyclohexyl to cyclohexylbenzene is equal to 9.

2003 Elsevier Inc. All rights reserved.
Keywords: Molybdenum carbide; Hydrodenitrogenation; Carbazole
1
. Introduction
Transition-metal carbides and nitrides have been stud-
ied intensively for more than a decade; it has been shown
that they may be excellent potential substitutes for hy-
drotreating catalysts [2–6]. Among them, molybdenum ni-
trides and carbides stand out because they are more ac-
tive than commercial alumina-supported sulfided NiMo and
CoMo catalysts, both in HDN or HDS of simple model
molecules [3,7–9] and even in the removal of nitrogen from
coal-derived liquids [10]. Schlatter et al. [3] were the first
to report that unsupported molybdenum nitrides exhibit a
much higher selectivity towards aromatic products during
the HDN of quinoline than sulfided NiMo supported on alu-
mina and, therefore, lead to lower hydrogen consumption.
Lee et al. [8] reported that bulk γ -Mo2N was more active
for the C–N hydrogenolysis of 1,2,3,4-tetrahydroquinoline
to 2-propylbenzene during the HDN of quinoline under at-
mospheric pressure than a sulfided Ni–Mo/Al2O3 catalyst.
Choi et al. [7] studied the HDN of pyridine over γ -Mo2N
at higher pressure and reported that the molybdenum ni-
Hydrotreating processes, such as hydrodenitrogenation
(
HDN) and hydrodesulfurization(HDS), are the main proce-
sses leading to the removal of nitrogen- (N-) and sulfur-
S-) containing compounds present in petroleum feedstocks,
respectively. The demand for cleaner fuels is growing (S
50 ppm for 2005), and even if the activity of conventional
(
<
hydrotreating catalysts (alumina-supported sulfided NiMo–
CoMo) has improved considerably in recent years, a limit
seems to have been reached. The latter catalysts are hardly
active towards refractory molecules. A two-stage hydrotreat-
ing process could be considered using highly active hy-
drotreating catalysts that possess noble metal properties and
are tolerant to small amounts of S-containing compounds.
A recent review by Prins [1] emphasises the mechanisms of
HDN and discusses the recently introduced transition-metal
carbide and nitride catalysts.
*
2 3
tride was more active than a sulfided CoMo/Al O cata-
Corresponding author.
E-mail address: djega@ccr.jussieu.fr (G. Djéga-Mariadassou).
lyst, with a higher selectivity to C–N bond hydrogenolysis.
0021-9517/03/$ – see front matter  2003 Elsevier Inc. All rights reserved.
doi:10.1016/S0021-9517(03)00072-1

A. Szyma n´ ska et al. / Journal of Catalysis 218 (2003) 24–31
25
Moreover, these metal nitrides were shown to be sulfur-
resistant [10,11].
Liquide, Custom grade C, purity > 99.995%), helium (He,
Air Liquide, Custom grade C, purity > 99.995%), dioxygen
These catalysts can also be supported on oxides such as
alumina [12–15] or on carbonaceous supports [16–18] and
still behave similarly in hydrotreating reactions to the bulk
catalysts [12,17]. However, the pore size distribution of the
supported catalysts must be carefully designed to enable
the large N- or S-containing molecules to reach the active
sites [19].
It is well known that the HDN reaction is more diffi-
cult than the HDS reaction, the C–N bond energy being
higher than that of the C–S bond energy. Furthermore, ni-
trogen removal requires ring hydrogenation prior to the C–N
bond cleavage [20]. Perot [21] assumed that, during HDN
processes over commercial catalysts (alumina-supported
sulfided NiMo and CoMo), three types of reactions oc-
cur, leading to the hydrogenolysis of nitrogen compounds:
(O , Air Liquide, Custom grade C, purity > 99.5%), and
2
methane (CH , Air Liquide, Custom grade N30, purity >
4
99.9%).
Carbazole (Fluka, 98%), o-xylene (Fluka, ꢀ 98%), and
dimethyldisulfide (Fluka, ꢀ 98%) were used as reactants for
the kinetic study.
Molybdenum carbide was prepared according to a modi-
fied standard procedure of Volpe and Boudart [28]. Molyb-
denum trioxide (3 g) was placed on a porous disc in a quartz
reactor and carburized with a mixture of 10 vol% CH4/H2,
−1
at a volume hourly space velocity (VHSV) of about 85 h
.
The temperature was increased linearly from 300 to 973 K at
−1
a rate of 53 K h and was kept at 973 K for 1 h. Finally, the
sample was cooled to room temperature (RT) in the reactant
carburizing carrier gas. When exposure to air was necessary,
the sample was passivated with a 1 vol% O2/He gas mixture
to form a protective oxygen layer on its surface.
(
i) hydrogenation of nitrogen heterocycles, (ii) hydrogena-
tion of benzenic cycles, and (iii) C–N bond cleavage.
Therefore, bifunctional catalysts with both hydrogenating
and hydrogenolysing active centres are necessary. Another
mechanism, such as cracking, can be observed over oxides
2
.2. Characterization
(
MoO3, NiO), as indicated by Ledoux et al. [22].
Much less attention has been paid to carbides and nitrides
Structural characterization of the molybdenum carbide
in the study of the HDN of heavier N-containing hydrocar-
bons such as carbazole, which is representative of refractory
N-containing components of heavy fractions derived from
petroleum. The only publications thus far are those by Na-
gai and co-workers [4,23]. They studied alumina-supported
molybdenum carbide and nitride for the HDN of carbazole
in the absence of sulfur. They observed higher activity of
the carbides than of the corresponding nitrides or sulfides.
N-containing compounds are usually very strong inhibitors
of HDS reactions over alumina-supportedsulfided Ni–Mo or
Co–Mo catalysts [24–26] as well as over molybdenum car-
bide [27]. Therefore, the kinetic parameters for both HDN
and HDS reactions should be established over molybdenum
carbides.
The present work is concerned with the activity and se-
lectivity of the HDN of carbazole over bulk molybdenum
carbide in the presence of 50 ppm of sulfur introduced as
dimethyldisulfide into the feed to simulate the operating con-
ditions of a second-stage hydrotreating process. Investiga-
tions were performed at a wide range of conversions (from
was carried out by X-ray diffraction (XRD) using a SIE-
MENS D-500 automatic diffractometer with Cu-Kα mono-
chromatized radiation. The degree of crystallinity of the
obtained carbides was estimated from the intensity of all
◦
◦
reflections in the 2θ range from 10 to 90 .
High-resolution transmission electron microscopy
(
HRTEM) was carried out on fresh and used carbides. The
HRTEM study was performed using a JEOL 200 CX elec-
tron microscope. A Quantachrome-Quantasorb Jr apparatus
was used for the surface area (Sg) measurement under dy-
namic conditions. The Sg of the catalyst was obtained from
the nitrogen adsorption/desorption at different partial pres-
sures of nitrogen. A standard Sg measurement using the
BET method was based on a three-point analysis consist-
ing of mixtures of 10, 20, and 30 vol% N2/He. The material
(
100 mg) was outgassed in a nitrogen stream at 623 K for
2
h before measurement.
CO uptake is a classical technique used to titrate the
metallic sites. Therefore, it is also widely used to titrate the
active sites of transition-metal carbide or nitrides [3,28–32].
The CO uptake was performed in situ, i.e., in the synthe-
sis reactor without exposing the fresh carbide to air or to
the passivating mixture. Pulses of a known quantity of CO
1
0 to 100%) and temperatures, which enable us to describe
the kinetics of the HDN reaction of carbazole as well as the
activity and selectivity to bicyclohexyl of the bulk molybde-
num carbide versus contact time.
(
17 µmol) were injected at regular intervals into the sample
−1
at RT in a stream of He (40 mL min ) purified by an oxygen
trap (Oxysorb, Messer Griesheim). After each injection, the
quantity of probe molecules that were not chemisorbed was
measured using a conventional device equipped with a TCD.
Injections were continued until the sites were saturated with
CO. Data were processed and the number of micromoles of
CO chemisorbed per gram of sample was determined.
2
. Experimental
2
.1. Catalyst
MoO3 (Fluka, ꢀ 99.5%) was used as the precursor for the
catalyst synthesis. The gases used were dihydrogen (H2, Air

26
A. Szyma n´ ska et al. / Journal of Catalysis 218 (2003) 24–31
2
.3. Kinetic study of HDN of carbazole
the molybdenum carbide was stabilized in situ for 10 h with
a mixture of 70 ppm of DMDS/o-xylene in the hydrogen
carrier gas (60 mL min ) at the temperature of the process
and at a total pressure of 6 MPa. The catalyst did not exhibit
deactivation versus time of run throughout the whole kinetic
study.
−1
The HDN of carbazole was conducted with a down-flow
fixed-bed microreactor in a high-pressure flow system. All
catalytic runs were performed in the presence of a small
amount of sulfur (50 ppm S), produced by the decomposition
of 70 ppm of dimethyldisulfide (DMDS) added to the feed.
The catalyst (density = 1.5 g cm , weight = 0.6 g, grain
size 0.25–0.315 mm) was mixed with carborundum (SiC) at
a ratio of catalyst to SiC = 1/5. The liquid feed, consisting
of 0.4 wt% carbazole in o-xylene, was fed to the reactor by
means of a high-pressure piston pump (Gilson model 307).
The hydrogen flow (30 to 360 cm min ) and total pressure
were controlled by a mass flow controller (Brooks 5850TR)
and a back-pressure regulator (Brooks 5866), respectively.
The temperature of the oven was regulated using a tempera-
ture controller (Hamyoung MX9).
−3
3. Results and discussion
3.1. Characterization of the catalyst
3
−1
The XRD pattern of the synthesized passivated carbide
(Fig. 1a) corresponds to the diffraction of the hexagonal
compact phase (hcp) β-Mo C structure. As can be observed,
2
new phases were not detected after the catalytic run in the
presence of sulfur (Fig. 1b). The small peak at around 2θ =
◦
The kinetic study was carried out at a total pressure of
36 corresponds to the diffraction line of SiC, i.e., residue of
6
MPa at temperatures between 553 and 653 K, with a
the carborundum (SiC) mixed with the catalyst, as described
in section 2.6.
H2/feed volumic ratio of 600 and contact times (tc) ranging
from 0.07 to 0.8 s. The contact time (tc) was defined as fol-
HRTEM showed that no MoS slabs could be observed in
2
3
lows: tc (s) = catalyst volume (cm )/(H2 flow + feed flow)
the sample used under the conditions of the HDN catalytic
3
−1
(
cm s ). The liquid products of the reaction were collected
run, in agreement with the work of Da Costa et al. [11]. How-
ever, the presence of sulfide species in the composition of the
active sites cannot be ruled out [33,34].
every hour in the condenser maintained at 288 K. Finally,
this liquid was analyzed by gas chromatography (HP 4890)
using a capillary column (HP1, 30 m × 0.25 mm × 0.25 µm)
and a FID detector. The product identification was confirmed
by both GCMS analysis (Finnigat Mat Model 800, capillary
column DB5, 30 m × 0.25 mm × 0.25 µm and HP5890-
HP5971A, capillary column PONA, 50 m × 0.20 mm ×
The specific surface area of the molybdenum carbide
2
−1
was 33 m g and CO titration led to 142 µmol of CO
chemisorbed per gram of sample. Assuming [3,28–32,35]
that one CO molecule titrates only one molybdenum atom
with metallic properties and that the average density of
surface molybdenum atoms on an hcp β-Mo2C is 1 ×
0
.5 µm) and injection of the standards. At each temperature,
Fig. 1. X-ray diffraction pattern of bulk molybdenum carbide (β-Mo C): (a) before catalytic run; (b) after catalytic run.
2

A. Szyma n´ ska et al. / Journal of Catalysis 218 (2003) 24–31
27
Table 1
Distribution of intermediates and products (mol%) during the HDN of car-
bazole over bulk β-Mo C at 613 K, 6 MPa, and tc = 0.27 s
2
mol%
1. Nitrogen-containing compounds (Fig. 2)
Tetrahydrocarbazole (THC)
Orthocyclohexylaniline (OCHA)
13
1
2
3
4
. Direct denitrogenation (DDN, route 1, Fig. 2)
Biphenyl (BPh)
0
6
. Direct denitrogenation (DDN, route 2, Fig. 2)
Cyclohexylbenzene (CHB)
. Hydrogenation (HYD, route 3, Fig. 2)
Cyclohexyl-cyclohexene (CHCHe)
Bicyclohexyl (BCH)
1
22
5
. Isomerization
ꢀ
isomers of CHB and BCH
29
72
58
Carbazole conversion (%)
HDN conversion (%)
1
015 Mo cm , it is calculated that 26% of the surface
−2
molybdenum atoms were exposed as metallic sites on the
synthesized carbide [31].
3
.2. Distribution of products and intermediates: reaction
network of the HDN of carbazole over bulk molybdenum
carbide
Table 1 illustrates all the products and intermediates ob-
served for a run conducted at 613 K, with a contact time of
Fig. 2. Reaction pathway of the HDN of carbazole over bulk β-Mo C:
2
0
.27 s, a HDN conversion of 58%, and a total carbazole con-
(a) global step of hydrogenation; (b) undetected products. Acronyms:
version of 72%. Higher conversions were obtained but did
not enable us to detect all the intermediates involved in the
HDN carbazole reaction pathway.
tetrahydrocarbazole (THC), orthocyclohexylaniline (OCHA), cyclohexyl-
benzene (CHB), cyclohexyl-cyclohexene (CHCHe), and bicyclohexyl
(BCH).
Products are divided into five groups: N-containing com-
pounds, biphenyl from the direct denitrogenation (route 1,
Fig. 2), cyclohexylbenzene from the direct denitrogenation
to nC6-benzene, cyclopentyl-phenyl methane, and methyl-
cyclopentylbenzene occurred on this catalyst (29 mol%).
In accordance with the products observed, a reaction
scheme of the HDN of carbazole over bulk molybdenum
carbide is proposed in Fig. 2. For the sake of clarity, side
reactions are not reported but three very reactive intermedi-
ates are indicated in the reaction pathway (Fig. 2, products
labeled (b)). These latter intermediates are probably formed
(
route 2, Fig. 2), cyclohexyl-cyclohexene and bicyclohexyl
according to the hydrogenation (route 3, Fig. 2), and isomer-
ization products.
The hydrogenated nitrogen product tetrahydrocarbazole
(
THC) was detected in a significant amount (13 mol%). Suc-
cessive hydrogenatedcarbazole compounds, such as hexahy-
drocarbazole, octahydrocarbazole, decahydrocarbazole, and
perhydrocarbazole, were not detected under our operating
conditions due to their high reactivity, in agreement with the
work of Miyao et al. [23]. Orthocyclohexylaniline (OCHA)
was randomly observed due to its low concentration, de-
pending on both contact time and temperature. The major
denitrogenation product was bicyclohexyl (BCH, 22 mol%,
route 3, Fig. 2), with a small amount (1 mol%) of the inter-
mediate cyclohexyl-cyclohexene (CHCHe) and 6 mol% of
cyclohexylbenzene (CHB) through route 2 (Fig. 2). Side re-
actions, such as the isomerization of bicyclohexyl (BCH) to
C6-cyclohexyl isomers and methylcyclopentylcyclohexane,
as well as the isomerization of cyclohexylbenzene (CHB)
3
2
as C(sp )–N bond breaking is more favored than C(sp )–N
bond breaking [21]. Carbazole is hydrogenated to dihydro-
carbazole (DiHC) and then to THC, which very likely under-
goes partial hydrogenation prior to the hydrogenolysis of the
3
C(sp )–N bond to form OCHA. Then, from OCHA, kinetic
coupling takes place and the subsequent reactions can occur
either through the denitrogenation pathway (DDN, route 2,
Fig. 2) to give CHB as a product or through the hydrogena-
tion pathway (HYD, route 3, Fig. 2), leading to CHCHe and
finally to BCH. Routes 2 and 3 are, thus, parallel reactions.
Hydrogenation of CHB to BCH is not favored as long as
nitrogen-containing compounds are still present in the feed.

28
A. Szyma n´ ska et al. / Journal of Catalysis 218 (2003) 24–31
Fig. 4. Distribution of reactant, intermediates, and products observed dur-
Fig. 3. Distribution of reactant, intermediates, and products observed dur-
ing the HDN of carbazole versus contact time over bulk β-Mo C at
2
ing the HDN of carbazole versus contact time over bulk β-Mo C at
2
633 K: (ꢁ) carbazole, (2) tetrahydrocarbazole (THC), (Q) cyclohexylben-
zene (CHB), and (") bicyclohexyl (BCH).
5
73 K: (ꢁ) carbazole, (2) tetrahydrocarbazole (THC), (Q) cyclohexylben-
zene (CHB), and (") bicyclohexyl (BCH).
mum and, finally, the product concentrations (BCH, CHB)
increase with contact time.
This is due to competitive adsorption, as observed in HDS
reactions [36–38].
In domain I (tc < 0.27 s), the carbazole concentration de-
creases linearly. This means that carbazole saturates most of
the free active sites of the catalyst. Consequently, the number
of free pairs of adjacent active sites is too small to proceed to
the dissociation of dihydrogen (following a Polanyi–Horiuti
mechanism) and the carbazole conversion is very low. In do-
main I, both reaction rates (consumption of carbazole and
formation of THC) are equal. This rate corresponds to the
hydrogenation of carbazole to THC (Fig. 2).
For tc > 0.27 s (domain II), the carbazole conversion
increases and consequently the carbazole coverage must
slightly decrease, the rate of consumption of carbazole still
being constant. Therefore, a sufficient number of pairs of
adjacent active sites becomes free, allowing H2 to dissociate
and to speed up the reaction. The conversion of carbazole
increases significantly, from 10 to 80%, within a narrow
range of contact times (0.27 < tc < 0.47 s). Moreover, since
the decrease in the carbazole concentration versus contact
time is linear up to 80% conversion, a true zero-order re-
action with respect to carbazole must be considered. As no
partially hydrogenated nitrogen compounds (such as hexa-,
octa-, deca-, perhydrocarbazole, bicyclohexylamine) were
detected, it means that they are probably very reactive and at
a very low concentration on the surface. Consequently, ad-
sorbed carbazole must be considered as the most abundant
reaction intermediate (mari) according to Boudart’s con-
cept [43] and the following balance on active sites can be
considered:
The product of the direct denitrogenation pathway (DDN,
route 1, Fig. 2), i.e., biphenyl (BPh), was not observed in our
study over bulk molybdenum carbide, whatever the operat-
ing conditions. The absence of this DDN pathway was also
reported by Nagai et al. [4,23,39–41] in the HDN of car-
bazole in the absence of H2S over similar catalysts, mainly
bulk and supported molybdenum nitride. This is an inter-
esting result because the DDS pathway corresponding to
this DDN pathway (route 1, Fig. 2) was observed over bulk
molybdenum carbide for the hydrodesulfurization of diben-
zothiophene, which is the sulfur-containing compound cor-
responding to carbazole [42].
3
.3. Kinetic results
To study the kinetics of the carbazole HDN reactions,
the molar percentages of the reactant, intermediates, and
products were plotted versus contact time for various tem-
peratures, as shown, for example, in Fig. 3 (573 K) and
Fig. 4 (633 K). Both these figures are representative of the
two types of plots observed during the kinetic study of the
HDN of carbazole. The profiles of these plots depend on
the temperature of the process (see for example Fig. 5 for
carbazole consumption). The following sections discuss the
effect of temperature (T < 613 K and T > 633 K) on the ki-
netics of the HDN of carbazole for each type of compound:
the reactant (carbazole), the main intermediate (THC), and
the denitrogenated products (BCH, CHB, CHCHe, and their
corresponding isomers).
∗
[
L] ∼ [ carbazole],
(1)
∗
where [L] is the density of sites and [ carbazole] the ad-
sorbed carbazole concentration. Therefore, the rate of dis-
appearance of carbazole (rd) is
3
.3.1. Low temperatures (T < 613 K)
Fig. 3 is representative of runs conducted below 613 K;
two distinct domains can be observed: domain I for tc <
.27 s and domain II for tc > 0.27 s. In the latter domain,
0
∗
ꢀ
rd = k[ carbazole] ∼ k[L] = constant = k ,
(2)
the curves behave as if the reaction pathway consists of
successive reactions. The reactant (carbazole) concentration
decreases, while the THC concentration reaches a maxi-
considering a one-way elementary step for carbazole hydro-
genation and k as the corresponding rate constant.

A. Szyma n´ ska et al. / Journal of Catalysis 218 (2003) 24–31
29
observed once again. At a contact time of 0.4 s, carbazole
conversion is already complete. At 653 K, the decrease in
carbazole is not linear (Fig. 5). Since the catalyst is more
active at the higher temperature, the carbazole consumption
rate increases and the surface is no longer saturated by the
reactant.
Above 633 K, the concentration of THC goes through a
maximum, which is, however, smaller (13 mol%) than that
at 573 K (Fig. 3) as well as at a shorter contact time (around
0
.07 s). The conversion of carbazole is higher and the con-
centration of the intermediate decreases as it is quickly trans-
formed to products. Finally, at tc = 0.4 s, no THC is ob-
served and the HDN conversion reaches 100%.
As described above, BCH is the major product, while the
concentration of CHB remains very low (8 mol% for tc =
0.4 s). The concentration of BCH increases linearly as car-
bazole decreases, and the rate of formation of BCH is similar
to the rate of the disappearance of carbazole. At tc = 0.4 s,
the selectivity ratio between routes 3 and 2 (Fig. 2) is equal
to 9. Route 3 (HYD pathway) is still favored over route
Fig. 5. Molar percent of carbazole versus contact time during the HDN
of carbazole over bulk β-Mo C. (a) Low-temperature range: (ꢁ) 553 K,
2
(
(
2) 573 K, (Q) 593 K, and (×) 613 K. (b) High-temperature range:
∗) 633 K and (") 653 K.
Under our operating conditions, HDN conversion reaches
1
00% for a contact time as low as 0.8 s.
2
(DDN pathway) at the higher temperature. Above 0.4 s,
Fig. 5 presents the disappearance of carbazole versus con-
tact time and temperature. It indicates that the two-domain
profile is indeed observed when the temperature is below
when nitrogen compounds are no longer present, a slight
hydrogenation of CHB to BCH is observed (CHB concen-
tration decreases).
6
13 K (see Fig. 5a at 553, 573, 593, and 613 K). In do-
main I, the saturation of the active sites by carbazole is too
high for the reaction to proceed (as explained previously). As
expected, when the temperature increases (Fig. 5b at 633 and
3
.3.3. Detailed kinetic study of HDN of carbazole
At 633 K (Fig. 4) and between 0.07 and 0.27 s, the rates
of the hydrogenation of carbazole and of the formation of
BCH are constant and are approximately equal to 220 and
2
1
6
53 K), domain I is no longer observed due to a higher car-
bazole conversion and to the presence of free pairs of active
sites, leading to the dissociation of H2 (as explained previ-
ously).
For tc < 0.27 s (domain I), the concentration of THC is
low, in agreement with carbazole conversion. For tc > 0.27 s
−1
33 s , respectively. In this range of contact times, less than
0 mol% THC is present in the product. Thus, the major
global reaction seems to be at 633 K:
Carbazole + 9H2 = BCH + NH3.
(3)
(
domain II), THC behaves as an intermediate as it passes
This is quite different from the situation over a commercial
sulfided CoMo or NiMo alumina-supported catalyst [44,45].
The general pathway in Fig. 2 can then be simplified to
the catalytic cycle shown in Fig. 6. For the sake of simplic-
through a maximum (45 mol% at 0.4 s). Up to tc = 0.4 s,
THC increases and also reacts to BCH. For tc > 0.4 s, THC
starts to decrease and leads to both BCH and CHB; it dis-
appears completely at tc = 0.8 s. The molar percentage of
HDN products (BCH + CHCHe + isomers from route 3
and CHB + isomers from route 2) is presented versus con-
tact time in Fig. 3. The major product of the reaction was
BCH. Its concentration increases with contact time, concur-
rently with the decrease in the carbazole and THC concen-
trations. At tc = 0.8 s, when no more nitrogen compounds
are present, the selectivity between routes 3 and 2 can be
calculated from the ratio of their respective concentration
products; it is equal to 9. The DDN pathway (route 2) is not
as favored as the HYD pathway (route 3).
∗
ity, a global hydrogenationstep is included between OCHA
and BCHA. This is possible because the hydrogen pressure
∗
is in excess and remained more or less constant throughout
the reaction.
As is often the case in catalysis, one of the intermediates
can desorb before reacting on the surface, depending on the
corresponding rate constants:
3
.3.2. High temperatures (T > 633)
Fig. 4 represents a run conducted at 633 K. It is represen-
tative of runs conducted at temperatures higher than 633 K;
only domain II is observed. Indeed, at such temperatures, the
conversion is substantial, even at very short contact times.
At 633 K, the decrease in carbazole is linear up to 60% con-
version. The zero-order reaction with respect to carbazole is

30
A. Szyma n´ ska et al. / Journal of Catalysis 218 (2003) 24–31
Fig. 7. Arrhenius plot of the carbazole hydrogenating step.
Fig. 6. Major catalytic cycle at 633 K for the global reaction: carbazole +
4
. Conclusion
The present study shows that bulk molybdenum carbide
9
H = BCH + NH according to the data presented in Fig. 4. Carb, car-
2
3
bazole; DiHC, dihydrocarbazole; THC, tetrahydrocarbazole; HHC, hexahy-
drocarbazole; OCHA, orthocyclohexylaniline; BCHA, bicyclohexylaniline;
CHCHe, cyclohexylcyclohexene; BCH, bicyclohexyl.
is active in the HDN of carbazole in the presence of a small
amount of sulfur (50 ppm). This sulfur concentration cor-
responds to the 2005 European regulations. Under such an
atmosphere, no MoS2 slabs were detected by HRTEM. The
direct denitrogenation pathway of carbazole transformation
(
rate constants are defined according to Fig. 6). If kdesorption >
k4 reaction, then a fraction of THC can desorb before being
∗
hydrogenated to HHC. In the opposite case, if kdesorption <
k4 reaction, there is no chance for the intermediate to desorb
as it is quickly transformed. This explains why DiHC, HHC,
and BCHA were not detected.
In the catalytic cycle (Fig. 6) all the elementary steps turn
over at the same rate. If an i-elementary step occurs σi times
in the cycle to obtain the global reaction, represented by
Eq. (3), the turnover rate is equal to ri,net/σi , σi being the
stoichiometric number [43]. In the present case, σi is always
equal to unity, and the rate of reaction can be written as
(
route 1, Fig. 2), leading to biphenyl, was not observed on
the bulk molybdenum carbide, in contrast to commercial
hydrotreating catalysts. The main product of HDN of car-
bazole was bicyclohexyl (route 3, Fig. 3). The kinetic study
showed that a minimum contact time (0.27 s) or temperature
(
633 K) is required to proceed to a substantial conversion
and to the complete disappearance of carbazole and tetrahy-
drocarbazole. For tc < 0.27 s and T < 633 K, a first domain
is observed (domain I) in which the HDN conversion of
carbazole is low, due to a high carbazole coverage and, con-
sequently, to a low hydrogen coverage, preventing the HDN
reaction to proceed. When the contact time is increased, the
HDN reaction of carbazole starts; the hydrogen coverage is
higher and a second domain (II) is observed. Furthermore, if
the temperature is increased (T > 633 K), only domain II re-
mains for the whole range of contact times, and a substantial
reaction rate is reached. At 633 K, a zero-order reaction is
observed with respect to carbazole, and the selectivity ratio
between routes 3 and 2 is equal to 9. A detailed kinetic analy-
sis is given. The rate-determining step is the first carbazole
hydrogenationstep, in agreement with the corresponding ex-
perimental energy of activation.
∗
∗
r = k2[ carbazole] = k5[ CHCHe] = k2[L]
kexp(zero order with respect to carbazole).
=
The cycle (Fig. 6) is kinetically equivalent to the following
two-step catalytic reactions [43]:
K1
Carb + ∗
∗
Carb,
(4)
(5)
k2
∗
Carb + H2
···
∗
with [ carbazole] = [L]. Therefore, the rate-determining
step (Eq. (5)) leads to
∗
r = k2[ carbazole] = k2[L] = kexp.
The Arrhenius plot, ln kexp versus 1/T , applied in domain II,
permits calculation of the activation energy corresponding to
the rate constant k2.
Acknowledgments
The slopes of the carbazole curves were determined in
domain II between 553 and 633 K (Figs. 5a and 5b). The
energy of activation of the reaction, defined above, can be
calculated from the Arrhenius plot (Fig. 7); it is equal to
We greatly acknowledge the French National Center for
Scientific Research (CNRS), the Polish Academy of Science
(PAN) and Science Committee (KBN), the French Foreign
Office (Warsaw French Ambassy), and TotalFinaElf for their
financial support in the Framework of the PICS 508 and
Jumelage “Materials for environment.”
−1
8
6.6 kJ mol . This value is in agreement with the activa-
tion energy of the hydrogenationof such nitrogen-containing
aromatic compounds [46]. This result is consistent with the
rate-determining step (Eq. (5)).

A. Szyma n´ ska et al. / Journal of Catalysis 218 (2003) 24–31
31
References
[24] C.N. Satterfield, M. Modell, J.F. Mayer, AIChE 21 (1975) 1100.
[
[
[
25] M. Nagai, T. Kabe, J. Catal. 81 (1983) 440.
26] V. LaVopa, C.N. Satterfield, J. Catal. 110 (1988) 375.
27] P. Del Gallo, C. Pham-Huu, A.P.E. York, M.J. Ledoux, Ind. Eng.
Chem. Res. 35 (1996) 3302.
[
[
[
1] R. Prins, Adv. Catal. 46 (2001) 399.
2] J.S. Lee, M. Boudart, Appl. Catal. 19 (1985) 207.
3] J.C. Schlatter, S.T. Oyama, J.E. Metcalfe III, J.M. Lambert Jr., Ind.
Eng. Chem. Res. 27 (1988) 1648.
[28] L. Volpe, M. Boudart, J. Solid State Chem. 59 (1985) 348.
[29] D.J. Sajkowski, S.T. Oyama, Appl. Catal. A 134 (1996) 339.
[30] C. Sayag, G. Bugli, P. Havil, G. Djéga-Mariadassou, J. Catal. 167
[
[
4] M. Nagai, T. Miyao, Catal. Lett. 15 (1992) 105.
5] C.C. Yu, S. Ramanathan, F. Sherif, S.T. Oyama, J. Phys. Chem. 98
(1994) 13038.
(1997) 372.
[6] V. Schwartz, V.L.T. Da Silva, J.G. Chen, S.T. Oyama, Stud. Surf. Sci.
Catal. A 130 (2000) 467.
[
[
[
31] J.S. Choi, G. Bugli, G. Djéga-Mariadassou, J. Catal. 193 (2000) 238.
32] E.I. Ko, R.J. Madix, Surf. Sci. 109 (1981) 221.
33] P.A. Aegerter, W.W.C. Quigley, G.J. Simpson, D.D. Ziegler, J.W.
Logan, K.R. McCrea, S. Glazier, M.E. Bussell, J. Catal. 164 (1996)
[7] J.G. Choi, J.R. Brenner, C.W. Colling, B.G. Demczyk, J.L. Dunning,
L.T. Thompson, Catal. Today 15 (1992) 201.
[
[
8] K.S. Lee, H. Abe, J.A. Reimer, A.T. Bell, J. Catal. 139 (1993) 34.
9] J.G. Choi, J.R. Brenner, L.T. Thompson, J. Catal. 154 (1995) 33.
109.
[34] R.R. Chianelli, G. Berhault, Catal. Today 53 (1999) 357.
[
[
10] S. Ramanathan, S.T. Oyama, J. Phys. Chem. 99 (1995) 16365.
11] P. Da Costa, J.L. Lemberton, C. Potvin, J.M. Manoli, G. Perot, M.
Breysse, G. Djéga-Mariadassou, Catal. Today 65 (2001) 195.
12] J.S. Lee, M.H. Yeom, K.Y. Park, I.S. Nam, J. Catal. 128 (1991) 126.
13] M. Nagai, T. Miyao, T. Tuboi, Catal. Lett. 18 (1993) 9.
[35] G.S. Ranhotra, G.W. Haddix, A.T. Bell, J.A. Reimer, J. Catal. 108
(1987) 24.
[
[
[
36] L.D. Rollman, J. Catal. 46 (1977) 243.
[
[
[
[
[
37] P. Michaud, J.L. Lemberton, G. Pérot, Appl. Catal. A 169 (1998) 343.
38] P. Da Costa, C. Potvin, J.M. Manoli, J.L. Lemberton, G. Pérot, G.
Djéga-Mariadassou, J. Mol. Catal. A: Chem. 184 (2002) 323.
39] M. Nagai, Y. Goto, A. Irisawa, S. Omi, J. Catal. 191 (2000) 128.
40] M. Nagai, Y. Goto, O. Uchino, S. Omi, Catal. Today 43 (1998) 249.
41] M. Nagai, Y. Goto, A. Miyata, M. Kiyoshi, K. Hada, K. Oshikawa, S.
Omi, J. Catal. 182 (1999) 292.
14] C.W. Colling, L.T. Thompson, J. Catal. 146 (1994) 193.
15] B. Dhandapani, S.T. Oyama, Catal. Lett. 35 (1995) 353.
16] D. Mordenti, D. Brodzki, G. Djéga-Mariadassou, J. Solid State Chem.
[
[
[
141 (1998) 114.
[17] B. Dhandapani, S. Ramanathan, C.C. Yu, B. Frühberger, J.G. Chen,
S.T. Oyama, J. Catal. 176 (1998) 61.
[
18] J. Trawczynski, Catal. Today 65 (2001) 343.
[42] H.K. Park, D.S. Kim, K.L. Kim, Korean J. Chem. Eng. 15 (1998) 625.
[43] M. Boudart, G. Djéga-Mariadassou, Kinetics of Heterogeneous Cata-
lytic Reactions, Princeton University Press, Princeton, NJ, 1984.
[44] H. Schulz, M. Shon, N.M. Rahman, Stud. Surf. Sci. Catal. 27 (1986)
201.
[45] Z. Sarbak, Acta Chim. Hung. 127 (1990) 359.
[46] J. Shabtai, G.J.C. Yeh, C. Russel, A.G. Oblad, Ind. Eng. Chem. Res. 28
(1989) 139.
[19] C. Sayag, S. Suppan, J. Trawczynski, G. Djéga-Mariadassou, Fuel Pro-
cess. Technol. 77–78 (2002) 261.
[
[
[
[
20] J.R. Katzer, R. Sivasubramaniam, Catal. Rev. Sci. Eng. 20 (1979) 155.
21] G. Perot, Catal. Today 10 (1991) 447.
22] M.J. Ledoux, A. Bouassida, R. Benazouz, Appl. Catal. 9 (1984) 41.
23] T. Miyao, K. Oshikawa, S. Omi, M. Nagai, Stud. Surf. Sci. Catal. 106
(1997) 255.