Direct aromatic C-N bond cleavage evidenced in the hydrodenitrogenation of 2,6-dimethylaniline over cobalt-promoted Mo/Al2O3 sulfide catalysts: A reactivity and FT-IR study

Article abstract of DOI:10.1006/jcat.2001.3271

The hydrodenitrogenation of 2,6-dimethylaniline (DMA) was studied over a series of sulfided Co(0-4.7%)-Mo(8.7%)/Al₂O₃ catalysts at 573 K under 4 MPa total pressure and 0-56 kPa H₂S partial pressure. Two NiMo samples were tested for comparison. The reaction network presents three parallel routes: dearomatization of DMA followed by either hydrogenation-elimination to dimethylcyclohexenes and dimethylcyclohexanes, or NH₃ elimination to mxylene, and disproportionation of DMA to 2-methylaniline and 2,4,6-trimethylaniline. We demonstrate that part of the xylene is formed by direct aromatic carbon-nitrogen bond cleavage through a nucleophilic substitution involving hydride species. On CoMo catalysts, in the presence of H₂S, the amount of extra xylene is independent of Co content, while the dearomatization is promoted. Without H₂S, this special substitution reaction is most important on the Mo catalyst, and strikingly Co acts as a poison. FT-IR spectroscopy of adsorbed carbon monoxide evidences a new type of sites on the sulfided catalysts after a mild hydrogen treatment. We propose that a site configuration located exclusively on unpromoted Mo atoms highly depleted in sulfur is responsible for the direct denitrogenation route. The NiMo couple behaves differently: xylene formation is independent of Ni content, which means that the specific Mo sites for direct C-N bond rupture are poisoned by nickel, even in the presence of H₂S. The location of Co and Ni on the MoS₂ slabs then appears different.

Full text of DOI:10.1006/jcat.2001.3271

Journal of Catalysis 202, 78–88 (2001)
doi:10.1006/jcat.2001.3271, available online at http://www.idealibrary.com on
Direct Aromatic C–N Bond Cleavage Evidenced in the
Hydrodenitrogenation of 2,6-Dimethylaniline over Cobalt-Promoted
Mo/Al2O3 Sulfide Catalysts: A Reactivity and FT-IR Study
J. van Gestel,1 C. Dujardin, F. Mauge´, and J. C. Duchet
Laboratoire Catalyse et Spectrochimie, UMR CNRS 6506, ISMRA-Universite´, 6, Boulevard Mare´chal Juin, 14050 Caen, France
Received December 12, 2000; revised March 23, 2001; accepted May 7, 2001
unsaturated sites (CUS) exposed on the surface. In this re-
spect, a nitrogen-containing model molecule is more appro-
priate to assessthe ultimate potentialofthe catalyst because
the released ammonia is believed to leave unchanged the
surface of the sulfide catalyst. Quinoline is the typical ni-
trogen model compound but, due to the complexity of the
reaction scheme, the reactivity for the individual reaction
steps is difficult to assess (3, 4). Therefore, we have chosen
an aniline compound, 2,6-dimethylaniline, which offers the
advantage of reacting through parallel routes (5). It is thus
possible to measure directlythe hydrogenation and the C–N
bond cleavage functions of the catalyst. Moreover, an ani-
line model molecule is representative of the o-propylaniline
intermediate in the quinoline reaction network.
Two processes are involved in the hydrodenitrogenation
of alkyl anilines: hydrogenation of the aromatic ring and
C–N bond cleavage. In the case of hydrogenation of aro-
matics, with or without an S or N heteroatom, H2S acts
as an inhibitor, but the reaction becomes zero order with
respect to H2S. This response to H2S has given valuable
information about the active species, hydrides and protons,
involved in the mechanism of toluene hydrogenation (6).
The method has been extended to aromatic sulfur com-
pounds(7, 8). With respect to C–N bond cleavage, the litera-
ture distinguishes saturated amines (Csp3–N) and aromatic
amines (Csp2–N). The former compounds are easily con-
verted into hydrocarbons. Since the reaction is enhanced
by H2S, SH groups are involved in the denitrogenation.
Nitrogen removal may proceed by Hofmann elimination
after quaternization of the nitrogen atom by a proton. The
protonated amine may alternatively undergo an SN2 nucle-
ophilic substitution by SH , followed by the hydrogenol-
ysis of the C–S bond (9–13). In contrast, the reaction of
aromatic amines is inhibited by H2S (5, 14), indicating an-
other mechanism. The direct Csp2–N bond rupture is still
questioned, because of the delocalization of the nitrogen
lone pair. An apparent direct heteroatom removal can nev-
ertheless occur through a dihydro intermediate, followed
by ammonia elimination, in a way similar to that proposed
for the apparent direct desulfurization of dibenzothiophene
The hydrodenitrogenation of 2,6-dimethylaniline (DMA) was
studied over a series of sulfided Co(0–4.7%)–Mo(8.7%)/Al2O3 cata-
lysts at 573 K under 4 MPa total pressure and 0–56 kPa H2S par-
tial pressure. Two NiMo samples were tested for comparison. The
reaction network presents three parallel routes: dearomatization of
DMA followed by either hydrogenation–elimination to dimethylcy-
clohexenes and dimethylcyclohexanes, or NH3 elimination to m-
xylene, and disproportionation of DMA to 2-methylaniline and
2,4,6-trimethylaniline. We demonstrate that part of the xylene is
formed bydirect aromaticcarbon–nitrogen bond cleavagethrough a
nucleophilic substitution involving hydride species. On CoMo cata-
lysts, in the presence of H2S, the amount of extra xylene is indepen-
dent of Co content, while the dearomatization is promoted. Without
H2S, this special substitution reaction is most important on the Mo
catalyst, and strikingly Co acts as a poison. FT-IR spectroscopy of
adsorbed carbon monoxide evidences a new type of sites on the sul-
fided catalysts after a mild hydrogen treatment. We propose that
a site configuration located exclusively on unpromoted Mo atoms
highly depleted in sulfur is responsible for the direct denitrogena-
tion route. The NiMo couple behaves differently: xylene formation
is independent of Ni content, which means that the specific Mo
sites for direct C–N bond rupture are poisoned by nickel, even in
the presence of H2S. The location of Co and Ni on the MoS2 slabs
c
then appears different.
2001 Academic Press
Key Words: hydrodenitrogenation; sulfide catalysts; dimethyl-
aniline; mechanism; active sites; CO; IR spectroscopy.
INTRODUCTION
Hydrotreating on sulfide catalysts is studied mostly in
the presence of H2S in order to simulate industrial HDS
conditions. The catalysts are then evaluated on the ba-
sis of the reactivity of sulfur model molecules, classically
thiophene and dibenzothiophene (1, 2). However, the full
potential of the bare surface of the catalyst cannot be ex-
plored in this way since conversion of the model molecule
produces H2S, which readily saturates the coordinatively
1 To whom correspondence should be addressed. Fax: 33 2 31 45 28 22.
E-mail: vangestel@ismra.fr.
78
0021-9517/01 $35.00
c
Copyright
2001 by Academic Press
All rights of reproduction in any form reserved.

MECHANISM OF THE HDN OF 2,6-DIMETHYLANILINE
79
(15–17). Nevertheless, Moreau et al. (18) showed that, in a
The partial pressure of 2,6-dimethylaniline (DMA) was
series of substituted benzenes, hydride species realize a di- kept constant at 13 kPa for all experiments. Reaction con-
rect aromatic carbon–N bond cleavage by nucleophilic sub- ditions at steady state were varied by changing the con-
stitution. This conclusion is based on Hammett correlations tact time (55–250 h g/mol) at a fixed H2S partial pressure.
and quantum chemical calculations. Jian and Prins (14, 19) Thereafter, the H2S partial pressure (0–56 kPa) was varied
suppose a direct hydrogenolysis of the Csp2–N bond. They at a fixed contact time. Conversions of DMA were kept
report a negative effect of nickel on this reaction and con- below 20% .
clude that the hydrogenolysis requires highly unsaturated
FT-IR Spectroscopy of Adsorbed CO
molybdenum atoms.
We report here a study of the reaction pathways in the
hydrodenitrogenation of 2,6-dimethylaniline on a series of
cobalt-promoted Mo/Al2O3 sulfide catalysts, and two NiMo
catalysts. Our aim is to elucidate the reaction mechanism,
especially with respect to the Csp2–N bond cleavage, and
to propose a configuration for the active sites. These sites
are characterized by IR spectroscopy of adsorbed carbon
monoxide, a proven and powerful probe to describe sulfide
catalysts (20–22).
The catalyst powder was pressed into a self-supported
wafer ( 10 mg, 2 cm2) and placed into the IR cell. Prior
to sulfidation, the catalyst was dried in situ at atmospheric
pressure at 423 K under N2, and cooled to 298 K. The cata-
lyst was then sulfided with 10% H2S/H2 (30 ml/min) at a
rate of 3 K/min up to 623 K. After 2 h of sulfiding, the cata-
lyst was flushed with N2 for 0.25 h and the temperature was
decreased to 298 K. The sulfided catalyst was rapidly evac-
uated at 573 K (10 K/min) and finally cooled to 100 K for
CO adsorption. The catalyst was contacted with 133 Pa CO
at equilibrium. IR spectra of adsorbed CO were recorded
with a Nicolet Magna 550 IR spectrometer equipped with
a MCT detector. For comparison, spectra were normalized
to a disc of 10 mg.
Alternatively, after the sulfidation ofthe catalyst at 623K,
the nitrogen flush was followed by a hydrogen treatment
(30 ml/min) at 573 K for 2 h. The catalyst was then cooled
under N2. CO adsorption was performed at 100 K as de-
scribed above. The two activation procedures are called
respectively “sulfided” and “H2-treated.”
EXPERIMENTAL
Catalyst Preparation
The Mo/Al2O3 catalyst (8.7 wt% Mo) was prepared by
pore-filling impregnation of a -alumina support (258 m2/g,
0.66 cm3/g) with an ammonium heptamolybdate solution.
The sample was dried at 393 K overnight and calcined at
773K for 3h. The seriesofCoMo catalystswere prepared by
impregnation of the molybdenum catalyst with a cobalt ni-
trate solution, drying, and finally calcining at 773 K for 3 h.
The catalysts contained 0.1, 1.7, 3.2, and 4.7 wt% cobalt as
determined by elemental analyses performed at the Service
Central d’Analyse (CNRS, Vernaison). Two NiMo/Al2O3
catalysts with 0.5 and 1.5 wt% nickel were prepared by im-
pregnation with a nickel nitrate solution, using the same
Mo/Al2O3 base sample.
RESULTS AND DISCUSSION
Reaction Network of 2,6-Dimethylaniline (DMA)
Figure 1 shows the typical product distribution obtained
on the Co(3.2% )Mo/Al2O3 catalyst in the presence of
56 kPa H2S at various contact times. Dimethylcyclohex-
enes (DMCHe) and dimethylcyclohexanes (DMCHa) are
2,6-Dimethylaniline Activity Measurement
The reactor (stainless steel reactor system from Sotelem, lumped and are the major denitrogenation products. The
L = 300 mm, i.d. = 9 mm) was loaded with 0.25–0.40 g of other denitrogenation product is m-xylene (Xyl), formed
catalyst grains (0.2–0.5 mm) and pressurized at 4 MPa. The in small quantity. We checked that under our conditions,
sample was first dried overnight in situ at 423 K under N2 m-xylene is not hydrogenated. Finally, large quantities of
(30 ml/min), and the temperature was decreased to 298 K. 2-methylaniline and 2,4,6-trimethylaniline are observed, in
Sulfidation was carried out in flowing 10% H2S/H2 approximately equal amounts. These products result from
(30 ml/min) at 623 K for 2 h with a temperature ramp of the disproportionation (DIS) of DMA. All three groups
3 K/min. Thereafter, the catalyst was cooled under H2S/H2 of products appear to be primary products. We deter-
to the reaction temperature of 573 K, and the H2S content mined the reaction orders for the formation of DMCHe +
was adjusted at 56 kPa (H2S/H2 = 1.4% ). The liquid feed DMCHa (k1), Xyl (k2), and DIS (k30 ) as zero, zero, and
(10 vol% 2,6-dimethylaniline, reactant; 80% heptane, sol- 0.4, respectively. Previously we reported first-order kinet-
vent; and 10% decane as internal standard) was introduced ics (5). The difference is due to the DMA partial pres-
byan HPLC pump and vaporized in the H2S/H2 stream. The sure, 13 kPa instead of 1.3 kPa. Activities are expressed
reaction products were condensed at the reactor exit and by the rate constants. For the sake of uniformity of units,
the liquid was analyzed on a gas chromatograph equipped the activity for DIS is given by k3 = k30 (P
)
mol h
0.4
1
DMA
1
with a CPSIL-5CB capillary column and a FID detector.
kg
.

80
VAN GESTEL ET AL.
FIG. 2. Selectivity diagram for the reaction of 2,6-dimethylaniline in
the presence of 56 kPa H2S on the Co(3.2% )Mo/Al2O3 catalyst.
FIG. 1. Product distribution for the reaction of 2,6-dimethylaniline at
573 K in the presence of 56 kPa H2S on the Co(3.2% )Mo/Al2O3 catalyst.
over NiMo catalysts in the presence of H2S. More recently,
Rota and Prins(11) reported also a high selectivityin hydro-
genation products (95% ) in the conversion of o-toluidine
over NiMo catalysts. None of these studies mentioned the
disproportionation route in the product distribution.
With the exception ofthe disproportionation, allthe stud-
ies agree with the general HDN reaction scheme for aniline
type compounds with two parallel routes: hydrogenation
and apparent direct nitrogen removal. This looks very sim-
ilar to the generally accepted scheme for the hydrodesulfur-
ization of dibenzothiophene (16). Accordingly, these paral-
lel denitrogenation routes most probably proceed through
a preliminary dearomatization step yielding a dihydro in-
termediate. The proposed reaction network for DMA is
depicted in Scheme 1.
The selectivity diagram (Fig. 2) shows that the three
routes run strictly parallel up to 20% conversion of DMA.
The selectivity for xylene does not exceed 5% of total
conversion, including the disproportionation of DMA, and
12% if only the denitrogenated products are taken into ac-
count. These results compare well with those reported for
other Co(Ni)Mo/Al2O3 catalysts measured at much lower
DMA pressure (5). They contrast with those reported by
Moreau et al. (23) for the HDN of aniline in a batch reac-
tor at 613 K and 7 MPa, yielding 80% selectivity for ben-
zene over a CoMo catalyst in the absence of H2S. Over a
NiMo catalyst, under the same conditions, the hydrogena-
tion route was predominant (23). Pe´rot et al. (24) reported
a high selectivity for hydrogenation products in the study
of 2,6-diethylaniline in a flow reactor at 623 K and 7 MP a
SCHEME 1. Reaction network of 2,6-dimethylaniline.

MECHANISM OF THE HDN OF 2,6-DIMETHYLANILINE
81
In this scheme, dimethylcyclohexenes and dimethylcy-
clohexanes are formed through a complete hydrogenation
of the dihydro intermediate into dimethylcyclohexylamine,
followed by a fast elimination of ammonia. Alternatively,
the dihydro intermediate may undergo elimination of am-
monia, yielding xylene and thus restoring the aromaticity.
Eventually, asproposed byRota and Prins(11), substitution
of the amino group on the dihydro or hexahydro interme-
diates by SH may precede elimination of the heteroatom.
However, such thiol intermediates are not detected under
our conditions. The limiting step for both routes should
be dearomatization, because we did not observe any in-
termediate. The selectivity is governed by the relative rate
constants k1 and k2 and adsorption coefficients. Dispropor-
tionation does not contribute to nitrogen removal unless
the isomerized methylanilines follow a scheme similar to
that of DMA. Only traces of monomethyl and trimethyl
HDN products were detected.
FIG. 4. Influence of cobalt content on the rate constants for
DMCHe + DMCHa (k1), Xyl (k2), and DIS (k3) in the presence of
56 kPa H2S.
Influence of H2S
Figure 3 shows the influence of H2S partial pressure on
the three routes over the Co(3.2% )Mo/Al2O3 catalyst. The
rate constants k1 and k2 abruptly decrease with minute
amounts of H2S (up to 300 ppm), and exhibit zero order
with respect to H2S. It is worth mentioning that the effect
of H2S partial pressure on the reaction of DMA is totally re-
versible. The disproportionation route (k3) increases with
increasing H2S pressure and is not occurring on the bare
alumina carrier. It follows almost a half-order dependency
in H2S, implying that dissociated H2S is involved in the re-
action mechanism. H2S is indeed known to create Brønsted
acidity over sulfide catalysts, located on the sulfide phase
and on the alumina carrier (25–27). Because H2S acceler-
ates DIS, we are inclined to suggest an acidic mechanism
occurring on the sulfide phase and/or on alumina.
With respect to k1 and k2, the abrupt inhibition by H2S
followed by a stabilization looks very similar to the H2S sen-
sitivity reported for aromatic hydrocarbons hydrogenation
(6). It contrasts with the continuous decrease in activity for
hydrodesulfurization of model sulfur-containing molecules
(28). Thisdifferent behavior can be attributed to the relative
equilibrium adsorption constants, much larger for H2S than
for DMA, but comparable for H2S and organic sulfur com-
pounds. Although not the purpose of this paper, it is worth
mentioning that the zero order in H2S has drawn particular
attention in literature. Two main models interpreting the
change in reaction order with H2S pressure emerge from
kinetic studies. Kasztelan and Guillaume (6) proposed a
change in the rate-determining step in the sequence of el-
ementary reactions for toluene hydrogenation: first addi-
tion of H or H+ to the aromatic ring. Olguin-Orozco and
Vrinat (7) extended this model to dibenzothiophene. Alter-
natively, we proposed a gradual conversion of vacancy sites
into sulfur-saturated sites (2). However, under high hydro-
gen pressure, the kinetics cannot discriminate between the
two models, both of them satisfactorily explaining the zero
order in H2S.
Influence of Cobalt Content in the Presence of H2S
The reactivity of DMA over the series of Co(0–
4.7% )Mo/Al2O3 samples was studied in the range of 0–
56 kPa H2S partial pressure. Since nitrogen removal is zero
order in H2S, only data for 56 kPa H2S are reported here.
Figure 4 shows the influence of the cobalt content on the
FIG. 3. Influence of H2S on the activity of the Co(3.2% )Mo/Al2O3
catalyst for the three reaction routes: k1, formation of dimethylcyclohex-
enes and -anes; k2, formation of m-xylene; k3, disproportionation.

82
VAN GESTEL ET AL.
section, show clearly a negative effect of cobalt on the rate
constant k2 for xylene formation. We think that it is unlikely
that this hydrogenolysis would be promoted in the presence
of H2S, while inhibited in the absence of H2S. For the hydro-
genation route, such a change is not observed: promotion
is observed in both cases. Moreover, addition or removal
of H2S in the reaction medium is completely reversible and
the intermediate region (between 0 and 300 ppm of H2S)
shows a continuous change in selectivity.
Second, one can propose that the m-xylene is formed
uniquelyin parallelto the DMCHe + DMCHa products, af-
ter dearomatization. In that case, cobalt addition promotes
the dearomatization step, yielding the common dihydro in-
termediate. The heteroatom removal occurs by a Csp3–N
bond cleavage (elimination) from the dihydro intermedi-
ate, much easier than the cleavage of the aromatic Csp2–N
bond. However, the absolute selectivity k2/k1, quantified
by the slope of the straight line, should be constant, what-
ever the cobalt content, and the extrapolated line should
go through the origin.
We believe therefore that the linear relationship in Fig. 5
corresponds to a combination of the two mechanisms. Xy-
lene is partly formed after dearomatization, with a con-
stant selectivity hydrogenation/elimination. This route is
effectively promoted by cobalt. The Y intercept shows that
dearomatization is not the only way to produce m-xylene.
The dearomatization route is represented in Fig. 5 by the
dashed line, going through the origin, and parallel to the ex-
perimental line. The difference with the experimental line
FIG. 5. Rate constant for xylene formation (k2) as a function of
DMCHe + DMCHa formation (k1) for the reaction performed in the pres-
ence of 56 kPa H2S over the series of Co(0–4.7% )Mo/Al2O3 catalysts (᭿)
and two NiMo/Al2O3 catalysts (0.5 and 1.5% Ni, ).
rate constants measured in the presence of H2S. Clearly, all
three routes are promoted by cobalt, including dispropor-
tionation. The activity is leveling off for Co contents above
about 3 wt% . Thus, the maximum of promotion is obtained
for a Co/Co + Mo ratio of 0.33, as expected for mixed sul-
fide catalysts. The rate constant increase for disproportion-
ation is in line with the higher efficiency of the promoted
catalyst for dissociation of H2S (25). Thus, the concentra-
tion of acidic sites induced by H2S is increased by cobalt,
which leads to a higher activity for the disproportionation
on the promoted catalyst. For the two other routes, the ac-
tivity enhancement is more important for hydrogenation
(promotion factor of 2.6) than for elimination (factor 1.7).
Different promotion factors were also observed in the case
of dibenzothiophene, the direct desulfurization route being
much more sensitive to the promoter than the hydrogena-
tion route (17, 29). The authors explained the difference in
the promotion factors by a change in the rate-limiting step:
C–S bond cleavage for the Mo catalyst against dearomati-
zation for the CoMo catalyst. With DMA, if (Fig. 5) the rate
constant for xylene formation (k2) is plotted against the rate
constant for the formation of DMCHe + DMCHa (k1), we
obtain a straight line. Addition of cobalt thus increases both
k2 and k1 in a constant ratio 1k2/1k1 = 0.070. This linear
relationship can be interpreted in different ways.
corresponds to a constant formation of xylene in the pres-
1
ence of H2S, amounting to 0.03 mol h
kg 1, and inde-
pendent of the cobalt content of the catalysts. A portion
of the observed xylene then comes directly from the DMA
molecule, without any preliminary dearomatization. The
correspondingreaction network isrepresented in Scheme 2.
Direct aromatic carbon–nitrogen bond cleavage may occur
by hydrogenolysis or by hydride substitution as discussed
later. Under our conditions, in the presence of H2S, the
contribution of this direct denitrogenation (DDN) route
to the total xylene formation is rather important since it
amounts to about 65% on the Mo catalyst and 40% on
First, if xylene is only the result of a direct cleavage of
the aromatic C–N bond, independently of hydrogenation,
the linear relationship simply means that the promoted sites
have a better ability to catalyze both hydrogenolysis and hy-
drogenation than the unpromoted Mo catalyst. However,
our results obtained without H2S, described in the following
SCHEME 2. Improved reaction network for the HDN of 2,6-
dimethylaniline.

MECHANISM OF THE HDN OF 2,6-DIMETHYLANILINE
83
the Co(3.2% )Mo catalyst. This DDN activity is indepen-
dent of cobalt content, and we conclude that it proceeds
on Mo sites distinct from those involved in dearomatiza-
tion. Moreover, these typical Mo sites are not influenced
by Co promotion. This will be discussed in view of the
decoration model in the Mechanism and Catalytic Sites
section.
One may argue that the k2–k1 relationship reported in
Fig. 5 for the CoMo catalysts is not general for mixed sul-
fide catalysts. In particular, the NiMo couple is known to
present a hydrogenolysis–hydrogenation selectivity differ-
ent from that of CoMo, and thus will not follow the same
line as for the CoMo catalysts in Fig. 5. Although it is not
the aim of this paper, focusing on CoMo catalysts, we mea-
sured the activity of two NiMo samples with a nickel con-
tent below the full promotion level. The corresponding data
are plotted in Fig. 5. The rate constant for xylene forma-
tion, k2, is fairly constant and very close to that measured
for the unpromoted Mo catalyst; the rate constant k1 for
the hydrogenation route increases with increasing Ni con-
tent. This may be interpreted as in the case of the CoMo
catalysts. Nickel promotes both k1 and k2 via dearomati-
zation through a common intermediate. However, the se-
lectivity factor 1k2/1k1 might be lower for NiMo than for
CoMo, due to its known hydrogenation properties. In that
case, dearomatization should be represented by a dashed
line going through the origin, but with a lower slope than
for the CoMo catalysts. However, the difference between
the observed k2 and the xylene formed after dearomati-
zation, representing the DDN reaction, is not constant for
the NiMo catalysts; it diminishes continuously with increas-
ing Ni content. The constant xylene formation measured
is thus the result of the compensating effect between the
promoted dearomatization route and the inhibited DDN
pathway. This means that the specific Mo sites responsible
for the direct formation of xylene, found to be insensitive
to cobalt, are poisoned in the case of the nickel promoter.
The surface sites on the “sulfided” CoMo catalysts were
characterized by infrared spectroscopy of adsorbed CO.
We performed CO adsorption at 100 K, in order to increase
the coverage of the probe molecule on weak sites such as
those present on the unpromoted Mo phase. Under these
conditions, the IR spectra show the interaction of the probe
with the alumina carrier and with the sulfide phase (21, 22).
Results obtained on the series of sulfided CoMo catalysts
are presented in Fig. 6. For all samples, CO adsorbs on the
coordinated unsaturated (CUS) Al3+ sites (band at around
2190 cm 1) and interacts with the hydroxyl groups of the
alumina carrier (band at 2154 cm 1). The sulfide phase is
FIG. 6. FT-IR spectra of CO adsorbed at 100 K on “sulfided” Mo
and CoMo catalysts: (a) Mo (solid line); (b) Co(0.1% )Mo (dotted);
(c) Co(1.7% )Mo (solid); (d) Co(3.2% )Mo (dashed); (e) Co(4.7% ) Mo/
Al2O3 (solid).
unpromoted Mo sites. However, a fraction of unpromoted
sites still remains, even for the Co(4.7% )Mo catalyst, be-
yond the composition for maximum global activity. The sig-
nal at 2070 cm 1 appears to increase between 3.2 and 4.7%
Co. This may be due to a contribution of adsorbed CO on
excess cobalt. However, no signal at 2055 cm 1 for adsorp-
tion on cobalt sulfide, not involved in promotion, is detected
in our series of catalysts. Cobalt addition also leads to an
intensity decrease of the bands characterizing the carrier,
in particular these specific to acidic OH groups. This is in-
dicative of a decrease of the support acidity by cobalt spinel
formation.
With the exception of the sample with excess cobalt, the
intensity of the bands specific to the unpromoted and pro-
moted sites of the sulfide phase follows the denitrogenation
(k1 + k2)activityofDMA, asshown bythe linear correlation
in Fig. 7. This activity includes the extra xylene formation
by DDN, but this will not affect the correlation, since the
DDN activity is minor and constant for the CoMo series
(Fig. 5). This correlation means that the dearomatization
activity is a linear combination of the intrinsic activities of
unpromoted and promoted sites (30).
With respect to the sites responsible for the DDN
pathway, the spectra of the sulfided catalysts do not show a
specific signal of the sulfide phase accounting for the extra
xylene formation. This is attributed to a very low concentra-
tion of sites, as indicated by the low DDN activity measured
in the presence of H2S.
1
characterized by a band at 2110 cm assigned to the CUS
Influence of Cobalt Content in the Absence of H2S
1
located on molybdenum and by a signal at 2070 cm for
the promoted sites (20). The variation of the band intensi-
In the absence of H2S, cobalt influences the three routes,
ties with cobalt content confirms such an assignment, since i.e., hydrogenation, xylene formation, and disproportiona-
the number of promoted sites increases at the expense of tion, differently. The rate constants are plotted as a function

84
VAN GESTEL ET AL.
very selective toward xylene (60% of the denitrogenated
products), while the selectivity remains low (18% ) on the
Co(3.2% )Mo catalyst. The selectivity inversion indicates
that the parallel between k1 and k2, which was established
in the presence of H2S, is not observed anymore. It is a
strong indication that there is an extra xylene formation,
more pronounced for low cobalt content catalysts. Most
probably, the DDN route, revealed in the presence of H2S,
is more important without H2S.
To quantify the evolution of the DDN route with cobalt
content, we still consider that a fraction of the total xylene
is formed after dearomatization by a route strictly parallel
to hydrogenation. However, without H2S, the selectivity
factor 1k2/1k1 is unknown. We assume a value of 0.070,
as found in the presence of 56 kPa of H2S (Fig. 5). From
k1, we can calculate k2dearo for each catalyst, represent-
ing the activity for xylene formation via dearomatization.
The difference with the measured rate constant k2 for xy-
lene formation gives the DDN activity k2DDN. The val-
ues in Table 1 show that the contribution of DDN to to-
FIG. 7. Relationship between dearomatization activity (k1 + k2) and
the total intensity of the signal of adsorbed CO on the sulfide phase for
the series of Co(0–4.7% )Mo/Al2O3 catalysts.
tal xylene formation is very important on the Mo catalyst
1
(0.39 mol h
kg 1) and decreases with increasing cobalt
1
content to a level of 0.12 mol h
kg 1. This clearly ev-
of Co content in Fig. 8. DIS (k3) is considerably lowered,
and slightly promoted by cobalt. The hydrogenation route,
k1, is highly promoted by cobalt, as was the case in the pres-
ence of H2S. The striking result concerns xylene formation,
k2, for which cobalt exhibits a marked negative effect. Such
a negative promoter effect was already reported by Jian
and Prins (19) in the case of o-propylaniline over Mo and
NiMo catalysts. As a result, the xylene selectivity depends
strongly on Co content. Without H2S, the Mo catalyst is
idences a reversed promoting effect, a so-called “demot-
ing effect,” of cobalt on direct aromatic carbon–nitrogen
bond cleavage. We recall that in the presence of H2S, this
1
DDN activity is low and constant (0.03 mol h
kg 1).
This means that specific “DDN” sites are created in
much larger amounts in the absence of H2S, especially
on the Mo catalyst. The negative effect of cobalt con-
firms that these DDN sites are located on molybdenum
only.
Such drastic changes in activity and selectivity observed
without H2S could alternatively result from a modification
of both the structure and the number of surface sites, upon
contact with large amounts of hydrogen at high tempera-
ture. To answer this question, we carried out CO adsorp-
tion after a mild H2 treatment (573 K) of the sulfided cata-
lyst in order to mimic the reaction atmosphere as closely
as possible. These activation conditions considerably mod-
ify the CO spectra. As an example, the spectrum of the
“H2-treated” unpromoted Mo catalyst in Fig. 9 shows a
TABLE 1
Calculated Activities for the DDN Route in the Absence of H2S
1
1
Rate constant (mol h
kg
)
Co content (wt% )
k1
k2
k2DDN
0
0.26
0.32
0.66
0.88
0.82
0.41
0.32
0.22
0.19
0.18
0.39
0.29
0.17
0.13
0.12
0.1
1.7
3.2
4.7
FIG. 8. Influence of cobalt content on the rate constants for
DMCHe + DMCHa (k1), Xyl (k2), and DIS (k3) for the reaction of DMA
performed in the absence of H2S.

MECHANISM OF THE HDN OF 2,6-DIMETHYLANILINE
85
2-fold increase in the amount of Mo CUS as compared to
the “sulfided” sample, indicating that the removal of sul-
fur by hydrogen has created vacancies, as expected (31,
32). An interesting feature in the spectrum is the appear-
ance of a shoulder (2098 cm 1) at a lower wavenumber
than usually observed for Mo vacancy sites (2110 cm 1).
1
This new band at 2098 cm is better evidenced by sub-
tracting the spectra of CO adsorbed on the “H2-treated”
catalysts and on the “sulfided”. We have verified that the
original signal of the “sulfided” catalyst is restored after
contacting the “H2-treated” catalyst with H2S(10% )/H2 at
573 K. Hence, the structure of the MoS2 slab has been
preserved and only labile sulfur has been removed. In a
recent work, Travert et al. (32) suggested, on the basis of
density functional theory calculations, a lower sulfur coor-
dination of the site corresponding to the band at around
1
2095 cm than on the site corresponding to the (CO) at
2110 cm 1. It is worth mentioning that the activation con-
ditions of the “H2-treated” catalysts are very favorable for
the observation of a maximum of CUS and in particular
for the detection of the (CO) band at 2098 cm 1. Indeed,
catalysts were treated successively under a flow of H2S/H2
and of H2. Both sulfidation and H2 treatment steps were
followed by flushing under inert gas at high temperature
to avoid adsorption of H2S or contaminants like COS (al-
ways present as impurity in H2S) or H2O. This might be
the reason why previous studies did not report the pres-
FIG. 10. Difference spectra “H2-treated” “sulfided” for the CoMo
series: (a) Mo (solid line); (b) Co(0.1% )Mo (dashed); (c) Co(1.7% )Mo
(solid); (d) Co(3.2% )Mo/Al2O3 (dashed).
of the CoMo samples, representative of the promoting ef-
fect, are shown in Fig. 10. The band intensities highlight a
marked increase of the number of unpromoted sites rather
than promoted sites. This is even accentuated by the larger
molar extinction coefficient of the CO band characteristic
of the promoted sites (21). The intensity of the band at
1
1
2070 cm increases moderately, only upon H2 treatment
ence of the band at 2098 cm (31). Activation conducted
on catalysts containing the highest cobalt content (1.7 and
3.2% ). It is attributed to an increased number of promoted
sitesrather than to the creation ofunpromoted sitesin a spe-
cific environment. The number of promoted sites created
by H2 posttreatment increases more strongly above 1.7%
cobalt content, and accounts for the increased dearomatiza-
tion activity without H2S. On all the catalysts of the series,
under static conditions gives a broad shoulder at 2100–
2095 cm 1 of much lower intensity, which is difficult to ob-
serve (32). The hydrogen treatment also leads to a strong
increase of the number of CUS of the sulfided phase for the
series of CoMo catalysts (spectra not presented). The differ-
ence spectra between “H2-treated” and “sulfided” catalysts
1
the new band at 2098 cm is created upon H2 treatment
and, as observed in Fig. 10, the ratio between the number
of sites characterized by the bands at 2110 and 2098 cm 1 is
constant whatever the catalyst. The intensity of these two
bands is by far the most important for the Mo catalyst; it di-
minishes rapidly with cobalt content, and almost stabilizes
at 1.7% cobalt. The similarity between the intensity of the
1
band at 2098 cm and the decrease of the DDN activity
with cobalt in the absence of H2S (Table 1) is striking. It
strongly indicates that these Mo sites are involved in the
aromatic C–N bond rupture.
Mechanism and Catalytic Sites
In this section, we propose a mechanism, essentially for
the heteroatom removal. On one hand, dearomatization
followed by further elimination and hydrogenation, and
on the other hand, direct aromatic carbon–nitrogen bond
cleavage. However, the disproportionation reaction also
deserves some attention. Indeed, the reaction is unusual on
FIG. 9. FT-IR spectra of CO adsorbed on “sulfided” (a) and “H2-
treated” (b) Mo/Al2O3 catalysts. (c) Difference (b) (a).

86
VAN GESTEL ET AL.
sulfides, and yields substantial amounts of products under that proposed by the authors of Ref. (17), because DMA
our conditions. dearomatization is not very sensitive to H2S. The active
Disproportionation reactions are typically acid-cata- species for dearomatization (hydrogenation) are generated
lyzed, and have been well documented, for example in the by hydrogen dissociation, yielding a hydride on an anionic
case of xylene production from toluene on acidic zeolites vacancy, and a proton associated with a neighboring S2
.
(33). On sulfide catalysts, the acidity is too low to provoke The rate-limiting step is addition either of a hydride ion (6,
such a reaction, and the formation of 2-methylaniline and 7) or of a proton (2, 5, 8). With respect to elimination, the
2,4,6-trimethylaniline from DMA must occur somewhat required basicsite involved in the E2 mechanism isprovided
differently. Necessarily, the transition state for this reaction by S2 on the edges of the MoS2 slabs (29). In the absence
should present the bonding of one DMA molecule through of H2S, the concentration of vacancies is increased, as is the
a methyl group to the para position of the other DMA activity for dearomatization.
(Scheme 3). Our resultsindicate that aciditybrought byH2S
The difference between the DMA and dibenzothiophene
dissociation is clearly involved. Therefore, we may suggest behavior is the evidence for direct heteroatom removal
a protonation of the amino group of the DMA molecule. (DDN) without preliminary dearomatization of the DMA
The two methyl substituents in 2 and 6 positions favor this molecule. This particular reaction is of importance over the
protonation. Consequently, the methyl groups become pos- unpromoted Mo catalyst, in the absence of H2S. The DDN
itively charged. The effect may be reinforced by adsorp- activity is readily poisoned by H2S and, surprisingly, it de-
tion of the C3 or C5 carbon atoms. The positively charged creases with the cobalt content of the catalyst. This means
methyl group may then bind to the para position of a DMA that the reaction requires a specific configuration of sites lo-
molecule in the gas phase or physisorbed on the catalyst. cated exclusively on Mo, with a high degree of unsaturation
This intermediate may then easily split into the observed (14).
disproportionation products.
The reaction can proceed either by hydrogenolysis or by
The routes for nitrogen removal are very much like the nucleophilic substitution by hydride. Although the former
hydrodesulfurization of dibenzothiophene (17). The ad- mechanism cannot be ruled out, the latter one is strongly
sorption site for dearomatization (hydrogenation) is a pro- supported bythe recent work ofMoreau etal. (18). Studying
moted or unpromoted Mo atom with one or several vacan- the carbon–heteroatom bond cleavage of a series of sub-
cies. Its degree of unsaturation is not necessarily as high as stituted benzenes over unsupported transition metal sul-
fides, they conclude from Hammett correlations and quan-
tum chemical modeling that a nucleophilic substitution by
hydrides takes place. Hydrides formed by dissociation of
hydrogen can react in two ways: nucleophilic substitution
leading to the DDN route, or following Kasztelan and
Guillaume (6), addition to the aromatic ring as the first step
for dearomatization. Since we observe a predominance of
DDN on the Mo catalyst, especially in the absence of H2S,
dearomatization is not favored. Probably, there is a lack
of protons in the near vicinity of the adsorbed molecule
to complete the formation of the dihydro intermediate. In-
deed, protons are attached to a sulfur atom and can migrate
only if a host sulfur is available in proximity. This situation
does not occur in the configuration depicted in Scheme 4 for
the unpromoted Mo catalyst in the absence of H2S. More-
over, in order to favor the nucleophilic substitution in the
DDN process, the aromaticity of the molecule should be
destabilized. This can be achieved by simultaneous adsorp-
tion of the nitrogen atom and the aromatic ring on two
adjacent vacancies. Additionally, hydrogen dissociation re-
quires a bridging sulfur atom and a vacancy, which are con-
verted into a SH group and a hydride adsorbed on the Mo
atom. The hydride attack on the Csp2 carbon cannot be fol-
lowed by proton addition due to the distance between the
SH group and the Mo with a high degree of unsaturation,
thus excluding hydrogenation. Finally, NH3 is evolved by
hydrogen, which regenerates the original adsorption site.
SCHEME 3. Mechanism for the disproportionation of 2,6-dimethyl-
aniline.

MECHANISM OF THE HDN OF 2,6-DIMETHYLANILINE
87
SCHEME 4. Proposed active sites and reaction mechanism for the direct denitrogenation (DDN) in the absence of H2S.
Therefore, all these features indicate that the crucial
The observed negative effect of cobalt on the DDN ac-
requirement for the DDN pathway is the presence of at tivity without H2S may be interpreted by a strong limitation
least three adjacent Mo sites with one Mo atom highly of the probability to create the arrangement of several ad-
depleted in sulfur. This configuration may well account for jacent Mo vacancies. If so, vacancies associated with cobalt
the IR signal detected on the unpromoted phase after H2 in the CoMo catalyst are not equivalent to vacancies on
1
posttreatment. Indeed, the bands at 2110 and 2098 cm
pure Mo. Indeed, CO adsorption provides evidence for a
reveal the presence of two Mo sites in a close but somewhat modification of the electronic properties of the site when
different environment, likely Mo sites situated on the same promoted. We can propose that these electronic proper-
1
edge of the MoS2 slab, the band at 2098 cm character- ties are unfavorable for the multipoint adsorption of the
izing Mo sites with a lower sulfur coordination. Taking aniline necessary for the DDN route. Even at 3.2 wt% Co,
into account ab initio calculations (32), it was proposed corresponding to a full promotion with respect to the global
that these sites are situated on the Mo edge of the MoS2 activity, the DDN activity without H2S is not completely in-
1
1
crystallites. The configuration depicted in Scheme 4 may hibited by Co. It remains at a level of 0.12 mol h
kg ,
account for the specificity of the sites required for DDN. which is 4 times higher than the activity in the presence
Nevertheless, we cannot exclude that some active sites are of H2S. This means that a part of the slab cannot be pro-
also situated on the sulfur edge. Indeed, on the Mo catalyst, moted, which is in agreement with the observations by IR
H2 treatment leads also to the appearance of a weak shoul- spectroscopy of adsorbed CO. We can propose that pro-
1
der at 2070 cm which, taking into account the previous motion by Co occurs only on one of the exposed edges of
¯
¯
work of Travert et al. (32), characterizes highly unsaturated the MoS2 slab, either the (1010)S-edge or the (1010)Mo-
Mo on the sulfur edge. We should also keep in mind that edge. Based on this reactivity study, we cannot distinguish
the H2 treatment is performed under atmospheric pressure between these edges, but a recent scanning tunneling mi-
in the IR cell, while the pressure reaches 4 MPa under the croscopy study indicated a preferential decoration of the
reaction conditions. The specific activity of sites presenting S-edge of the MoS2 slab (34).
a higher degree of unsaturation has already been proposed
(14).
The DDN reaction needs a highly unsaturated configu-
ration located on Mo atoms as depicted in Scheme 4. These

88
VAN GESTEL ET AL.
sites are expected to be very sensitive to H2S. Indeed, on genation, and acidity, make this aniline very valuable as
the unpromoted Mo catalyst, the high DDN activity in the a model molecule for sulfide catalysts. For example, one
absence of H2S diminishes readily after addition of H2S in can envisage its use for understanding the hydrogenation
the reactor, the highly unsaturated edge sites being easily of aromatics and hydrodesulfurization of sulfur compounds
poisoned. However, some activity remains on specific sites in gasoil.
that are insensitive to H2S, for which corner sites can be
proposed. Running the promoted CoMo catalysts in the
presence of H2S yields a low DDN activity on these specific
REFERENCES
sites, which can never be promoted by cobalt.
1. Topsøe, H., Clausen, B. S., and Massoth, F. E., in “Catalysis, Science
and Technology” (J. R. Anderson and M. Boudart, Eds.), Vol. 11.
Springer-Verlag, Berlin, 1996.
2. Leglise, J., Finot, L., van Gestel, J. N. M., and Duchet, J. C., Stud. Surf.
Sci. Catal. 127, 51 (1999).
On nickel-promoted catalysts, the situation is different.
Indeed, already in the presence of H2S, the DDN activity
decreases with Ni content. This points out that nickel inter-
acts differently than cobalt with the MoS2 slabs, and might
promote all the edge and corner sites of MoS2.
3. Ho, T. C., Catal. Rev.-Sci. Eng. 30, 117 (1988).
4. Girgis, M. J., and Gates, B. C., Ind. Eng. Chem. Res. 30, 3021 (1991).
5. Van Gestel, J., Leglise, J., and Duchet, J. C., Appl. Catal. A 92, 143
(1992).
6. Kasztelan, S., and Guillaume, D., Ind. Eng. Chem. Res. 33, 203 (1994).
7. Olguin-Orozco, E., and Vrinat, M., Appl. Catal. A 170, 195 (1998).
8. Van Gestel, J., Finot, L., Leglise, J., and Duchet, J. C., Bull. Soc. Chim.
Belg. 104, 189 (1994).
9. Nelson, N., and Levy, R. B., J. Catal. 58, 485 (1979).
10. Cerny´, M., Collect. Czech. Chem. Commun. 49, 2387 (1984).
11. Rota, F., and Prins, R., Topics Catal. 11/12, 327 (2000).
12. Pe´rot, G., Catal. Today 10, 447 (1991).
13. Portefaix, J. L., Cattenot, M., Guerriche, M., Thivolle-Cazat, J., and
Breysse, M., Catal. Today 10, 473 (1991).
CONCLUSION
HDN of 2,6-dimethylaniline shows the parallel forma-
tion of three groups of products: (i) dimethylcyclohex-
enes and dimethylcyclohexanes, (ii) m-xylene, and (iii) 2-
methylaniline and 2,4,6-trimethylaniline. The isomerized
anilinesresult from disproportionation, promoted bycobalt
and by acidity provided by the dissociative adsorption of
H2S.
DMA is dearomatized, and the formed dihydro interme- 14. Jian, M., and Prins, R., Catal. Lett. 50, 9 (1998).
15. Singhal, G. H., Espino, R. L., Sobel, J. E., and Huff, G. A., J. Catal. 67,
diate reacts in two ways: further hydrogenation to dimethyl-
cyclohexylamine, undergoing a rapid NH3 elimination, re-
sulting in dimethylcyclohexenes and -anes; or elimination
of NH3 to form m-xylene. The common dearomatization
step is promoted by cobalt, and inhibited by H2S.
We highlight a second route for xylene formation, direct
denitrogenation (DDN) of the aromatic aniline, occurring
by a nucleophilic substitution by a hydride species. The sites
responsible for this reaction are located uniquely on highly
unsaturated Mo atoms. These sites are detected by FT-IR
spectroscopy of adsorbed carbon monoxide after a mild
hydrogen treatment of the sulfided catalyst, giving rise to
a new band at 2098 cm 1, not observed on the “sulfided”
catalyst. The intensity variation of this band correlates with
457 (1981).
16. Meille, V., Schulz, E., Lemaire, M., and Vrinat, M., J. Catal. 170, 29
(1977).
17. Bataille, F., Lemberton, J. L., Michaud, P., Pe´rot, G., Vrinat, M.,
Lemaire, M., Schulz, E., Breysse, M., and Kasztelan, S., J. Catal. 191,
409 (2000).
18. Moreau, C., Joffre, J., Saenz, C., Afonso, J. C., and Portefaix, J. L.,
J. Mol. Catal. A 161, 141 (2000).
19. Jian, M., and Prins, R., Catal. Today 30, 127 (1996).
20. Bachelier, J., Tilliette, M. J., Cornac, M., Duchet, J. C., Lavalley, J. C.,
and Cornet, D., Bull. Soc. Chim. Belg. 93, 743 (1984).
21. Mauge´, F., and Lavalley, J. C., J. Catal. 137, 69 (1992).
22. Delgado, E., Fuentes, G. A., Hermann, C., Kunzmann, G., and
Kno¨zinger, H., Bull. Soc. Chim. Belg. 93, 735 (1984).
23. Moreau, C., Joffre, J., Saenz, C., and Geneste, P., J. Catal. 81, 168
(1983).
the activity of the catalyst in the absence of H2S for the 24. Pe´rot, G., Brunet, S., Canaff, C., and Toulhoat, H., Bull. Soc. Chim.
direct Csp2–N bond cleavage.
The DDN pathway, important in the absence of H2S,
is readily inhibited by H2S. The reaction without H2S is
decreased by addition of cobalt. In the presence of H2S,
Belg. 96, 865 (1987).
25. Petit, C., Mauge´, F., and Lavalley, J. C., Stud. Surf. Sci. Catal. 106, 157
(1997).
26. Topsøe, N. Y., Topsøe, H., and Massoth, F. E., J. Catal. 119, 252 (1989).
27. Topsøe, N. Y., and Topsøe, H., J. Catal. 139, 641 (1993).
the contribution of this DDN route for xylene formation is 28. Leglise, J., Finot, L., van Gestel, J. N. M., Duchet, J. C., and Dubois,
J. L., Catal. Today 45, 347 (1998).
29. Mijoin, J., The´venin, V., Garcia Aguirre, N., Yuze, H., Wang, J., Li,
W. Z., Pe´rot, G., and Lemberton, J. L., Appl. Catal. A 180, 95 (1999).
30. Kasztelan, S., Toulhoat, H., Grimblot, J., and Bonnelle, J. P., Appl.
low, but independent of Co content.
In the case of NiMo catalysts, inhibition by Ni is observed,
even in the presence of H2S. Our results indicate a prefer-
ential location of Co on only one type of MoS2 edges (either
Catal. 13, 127 (1984).
S or Mo edge), resulting in an incomplete decoration of the 31. Mu¨ller, B., van Langeveld, A. D., Moulijn, J. A., and Kno¨zinger, H.,
J. Phys. Chem. 97, 9028 (1993).
32. Travert, A., Dujardin, C., Mauge´, F., Cristol, S., Paul, J. F., Payen, E.,
and Bougeard, D., Catal. Today, in press.
33. Fraenkel, D., Ind. Eng. Chem. Res. 29(9), 1814 (1990).
slab. On the other hand, there is no preference for the Ni
location. The reaction of DMA is very valuable since it sug-
gests different interactions of Co and Ni as promoters of the
MoS2 slabs. Furthermore, the different catalytic functions
measured with DMA, i.e., hydrogenation, direct denitro-
34. Lauritsen, J. V., Helveg, S., Lægsgaard, E., Stensgaard, I., Clausen,
B. S., Topsøe, H., and Besenbacher, F., J. Catal. 197, 1 (2001).