Transformation of a model FCC gasoline olefin over transition monometallic sulfide catalysts

Article abstract of DOI:10.1016/j.jcat.2007.03.009

The selective HDS of FCC gasoline is a sensible option for reducing sulfur content in commercial gasoline. For such application, a minimum activity of the catalyst toward olefin hydrogenation is required to preserve the high octane number of the feedstock. The conversion of a model FCC olefin (2,3-dimethylbut-2-ene:23DMB2N) under close HDS conditions was investigated over a series of unsupported transition monometallic sulfides (FeS, Ni₃S₂, PdS, Co₉S₈, Rh₂S₃, RuS₂, PtS, and MoS₂). The results reveal for the first time that a volcano curve relationship exists between the ab initio calculated sulfur-metal bond energy, E (MS), descriptor of bulk TMS, and their activities in olefin hydrogenation under the conditions of HDS of FCC gasoline. In particular, Rh₂S₃, with an intermediate metal sulfide bond energy of 119 kJ/mol, was the most active catalyst in hydrogenation of the model olefin. In a similar spirit as volcano curves obtained for the HDS of dibenzothiophene and hydrogenation of toluene and recently reported in the literature, a microkinetic model furnished a rational interpretation of this trend.

Full text of DOI:10.1016/j.jcat.2007.03.009

Journal of Catalysis 248 (2007) 111–119
www.elsevier.com/locate/jcat
Transformation of a model FCC gasoline olefin over transition monometallic
sulfide catalysts
A. Daudin a, S. Brunet a,∗, G. Perot a, P. Raybaud b, C. Bouchy c
a
UMR CNRS 6503, Catalyse en Chimie Organique, Université de Poitiers, Faculté des Sciences Fondamentales et Appliquées, 40 avenue du Recteur Pineau,
86022 Poitiers Cedex, France
b
IFP, Direction Chimie et Physico-Chimie Appliquées, 1–4 avenue de Bois Préau, 92858 Rueil Malmaison, France
c
IFP, Direction Catalyse et Séparation, BP 3, 69390 Vernaison, France
Received 15 November 2006; revised 7 March 2007; accepted 11 March 2007
Abstract
The selective HDS of FCC gasoline is a sensible option for reducing sulfur content in commercial gasoline. For such application, a minimum
activity of the catalyst toward olefin hydrogenation is required to preserve the high octane number of the feedstock. The conversion of a model
FCC olefin (2,3-dimethylbut-2-ene:23DMB2N) under close HDS conditions was investigated over a series of unsupported transition monometallic
sulfides (FeS, Ni S , PdS, Co S , Rh S , RuS , PtS, and MoS ). The results reveal for the first time that a volcano curve relationship exists
3
2
9
8
2
3
2
2
between the ab initio calculated sulfur–metal bond energy, E(MS), descriptor of bulk TMS, and their activities in olefin hydrogenation under the
conditions of HDS of FCC gasoline. In particular, Rh S , with an intermediate metal sulfide bond energy of 119 kJ/mol, was the most active
2
3
catalyst in hydrogenation of the model olefin. In a similar spirit as volcano curves obtained for the HDS of dibenzothiophene and hydrogenation
of toluene and recently reported in the literature, a microkinetic model furnished a rational interpretation of this trend.
© 2007 Elsevier Inc. All rights reserved.
Keywords: Hydrogenation; Isomerization; Gasoline; Transition monometallic sulfide catalysts; 2,3-Dimethylbut-2-ene
1. Introduction
theoretical foundation for the observed periodic trends in TMS
activities, such as TMS enthalpy of formation [2], sulfur coor-
dination number [6], and bond energy model [7].
Transition monometallic sulfide (TMS) catalysts are known
to be active in various hydrotreating reactions [1]. Pecoraro
and Chianelli [2] were the first to report a systematic study of
the intrinsic activity of bulk TMS in a model reaction, the hy-
drodesulfurization (HDS) of dibenzothiophene at 400 ◦C. They
established a relative activity scale for TMS as a function of the
periodic position of the transition metals. These results were
further confirmed by several experimental studies of both un-
supported [3] and supported sulfides [3–5] in HDS, hydroden-
itrogenation (HDN), and hydrogenation (HYD) of aromatics
that found that group VIII metal sulfides of the first row (e.g.,
Fe, Co, Ni) were poorly active, whereas those of the second
row (e.g., Ru, Rh) were much more active. These early results
triggered several attempts in the 1980s and 1990s to develop a
Recent progress brought by application of density functional
theory to realistic models of sulfide catalysts [8] have over-
come the limitations of previous approaches in properly in-
vestigating the structure and electronic properties of numerous
TMS catalysts [9,10]. In particular, a general approach called
the “yin-yang” ab initio descriptor rationalized experimental
trends observed in HDS and HYD activity of those catalysts
[10–12]. This descriptor, termed the bulk sulfur–metal bond en-
ergy, E(MS), of the TMS, shows how experimental activity pat-
terns can be correlated with a computed-bond-energy descriptor
in accordance with the Sabatier principle [13]. Volcano-type
relationships have been established between the ab initio calcu-
lated metal–sulfur bond energies, E(MS), and various catalytic
activities relevant for hydrotreatment processes, including HDS
of dibenzothiophene [10–12,14] and HYD of biphenyl [8,15]
and, more recently, toluene [16,17]. Indeed, Pecoraro et al. [2]
*
Corresponding author. Fax: +33 549453897.
E-mail address: sylvette.brunet@univ-poitiers.fr (S. Brunet).
0021-9517/$ – see front matter © 2007 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2007.03.009

112
A. Daudin et al. / Journal of Catalysis 248 (2007) 111–119
and Lacroix et al. [3] correlated higher HDS activity of DBT
with an optimal E(MS) of 120–140 kJ/mol, and greater HYD
activity of biphenyl was obtained for E(MS) of around 110–
120 kJ/mol. Consequently, it has become possible to rational-
ize and predict the performance of sulfide catalysts from such
volcano curve relationships. A recent review has described the
theoretical advances in this area in more detail [8].
tive of olefins found in FCC gasoline. Indeed, an olefin with a
branched skeleton and internal double bond that can give rise to
a rather limited number of isomers, such as 2,3-dimethyl-2-ene,
seems quite appropriate as a model molecule for determining
the various mechanisms and active sites involved. A kinetic
study using a model molecule should provide more information
than a study using a real feed for dissociating the various phe-
nomena involved [19–22]. The TMS catalysts were chosen to
cover a representative interval of metal–sulfur bond energies,
E(MS), 70–170 kJ/mol, as defined previously [11]. Unsup-
ported TMS catalysts were used to avoid any possible support
effect. The reactivity of the olefinic feed was measured under
conditions typical of FCC gasoline-selective HDS. Hydrogena-
tion activities with regard to the TMS catalysts evaluated are
discussed with the E(MS) descriptor proposed by Toulhoat and
Raybaud [11] to explore whether volcano curves may exist for
these new classes of reactions, as have been obtained recently
for toluene and biphenyl hydrogenation. Furthermore, in the
case of 23DMB2N hydrogenation, a microkinetic model to ex-
plain the E(MS)-activity relationship is proposed.
The European Union’s planned reduction of the sulfur level
in diesel and gasoline to 10 wt ppm in 2009 [18] will re-
quired the elaboration of new catalysts or optimization of con-
ventional catalysts. Emissions from motor vehicles (NOx and
SOx) contribute greatly to air pollution, and sulfur is a well-
known poison for catalytic converters. The gasoline fraction
produced from the FCC process represents 30–50% of the com-
mercialized motor fuel but contains up to 85–95% of the sul-
fur impurities. It is composed mainly of aromatics (30 vol%),
alkenes (30 vol%), and sulfur compounds such as alkylthio-
phenes (max. 5000 ppm) [19–21]. These olefins (composed
of C5–C10 and mainly C6 olefins) are iso-olefins and internal
double-bond olefins, which are more stable and less reactive
than terminal olefins for obtaining and maintaining the octane
number. A high hydrotreating selectivity, corresponding to a
high HDS activity and a low hydrogenation activity for olefins
(HYDO), is required. Indeed, significant alkene saturation can
occur during the HDS process, leading to a lower octane num-
ber in the final product. Consequently, hydrotreating catalysts
must meet selective criteria for HDS/HYDO—that is, achieve
deep HDS with minimum olefin saturation.
Recent work [22,23] has shown that the isomerization step
may be a key factor in the hydrogenation of alkenes. Indeed,
Mey et al. [23] showed that the first step in the transformation
of 23DMB2N is its isomerization in 23DMB1N, followed by
the hydrogenation step to produce 23DMB.
The selectivity for HDS/HYDO was increased by modify-
ing the acid–base properties of the support either by decreasing
the acidity of the support by alkaline elements (e.g., Li, K)
[23–25] or using a more basic support such as hydrotalcite [26],
or by poisoning selective hydrogenation sites by carbon depo-
sition [27,28] or adsorption of basic nitrogen compounds [29].
Indeed, Mey et al. [23] reported decreased hydrogenation activ-
ity on CoMo-supported catalyst modified by potassium. These
results have been explained by the inhibition of the olefin iso-
merization activity and consequent decreased activity of the
catalyst in olefin hydrogenation.
2. Experimental
2.1. Catalyst preparation
Co, Ni, Ru, Rh, Pd, and Pt sulfides were prepared using
the nonaqueous precipitation method reported by Pecoraro and
Chianelli [2]. The metal chloride precursor and lithium sulfide
were added in ethyl acetate, and the metal sulfide was obtained
after 4 h of stirring at 353 K. Then the solution was cooled to
room temperature and filtrated to recover the dark precipitate.
The corresponding sulfide solid was stabilized by a sulfidation
step in H2/H2S (10 mol% H2S) flow for 2 h at 673 K (heating
rate, 276.3 K/min). LiCl was removed by several washes with
acetic acid and vacuum filtration. Finally, the metallic sulfide
was obtained after a second sulfidation procedure under usual
conditions.
Preparation of molybdenum sulfide was carried out by
thermal decomposition of ammonium thiomolybdate [31,32].
First, the ammonium tetrathiomolybdate (NH4)2MoS4 (ATM)
was obtained by reaction between ammonium heptamolybdate
(NH4)6[Mo7O24]·4H2O (4 g in 20 cc of distilled water) with a
50 wt% of ammonium sulfide (NH4)2S in aqueous solution at
333 ◦C. The ATM precursor was precipitated as red crystals by
cooling the solution in iced water for 3 h, after which the pre-
cipitated red crystals were thoroughly washed with isopropanol
and dried.
To the best of our knowledge, the correlation between the ac-
tivities of TMS and their corresponding E(MS) descriptor has
not yet been investigated for the reaction of alkene hydrogena-
tion. Such a study is of practical and theoretical interest because
for MoS2 catalyst, Qu and Prins [30] suggested that hydrogena-
tion of aromatics occurs on sites different from those active in
the hydrogenation of cycloalkenes.
2.2. Characterization
The TMS catalysts were characterized before and after cat-
alytic activity measurements by TEM combined with EDX
(Philips CM 120 kV), X-ray diffraction (Bruker D5005), BET
surface area (Micromeritics ASAP 2010), and elemental analy-
sis (CE Instruments NA2100 Protein) at the University of
Poitiers (LACCO) to identify the exact nature of TMS and
This paper reports the activity of various unsupported TMS
(FeS, Ni3S2, PdS, Co9S8, Rh2S3, RuS2, PtS, and MoS2) cata-
lysts in the transformation of a model feed composed of 2,3-
dimethylbut-2-ene (23DMB2N) in n-heptane in the presence
of a partial pressure of H2S corresponding to 1000 ppm wt of
sulfur in the model feed. This alkene is considered representa-

A. Daudin et al. / Journal of Catalysis 248 (2007) 111–119
113
Table 1
Table 2
XPS binding energies of the various TMS
Partial pressures of the different components for the preliminary sulfidation step
and the transformation of 23DMB2N
TMS
Binding energy (eV)
Metal
Pressure (bar)
Sulfidation
Feed
Sulfur (2p
)
3/2
P
P
P
P
P
0
1.51
0.02
13.40
5.07
20
olefin
MoS
(MoO )
FeS
3d : 229.4
5/2
162.2
–
162
162.1
162.3
161.6
162.0
2
0.1
0.9
0
H S
2
3d : 232.5
x
5/2
H
2
2p : 706.6
3/2
nC
7
Ni S
2p : 852.4
3
2
3/2
1
total
Co S
2p : 778.3
3/2
9
8
PdS
3d : 336.3
5/2
RuS
3d : 280
5/2
2
Table 3
Products resulting from the transformation of 23DMB2N
verify that no modification was done after the activity mea-
surements. Comparison of the XRD patterns and the JCPDS
database allowed identification of the various crystalline TMS
phases.
Isomerization (Isom)
XPS spectra were recorded using a KRATOS AXIS Ul-
tra spectrometer equipped with a 300 W AlKα source (hν =
1486.6 eV). TMS samples were packed in Shlenk under argon
to avoid sulfate formation. They were identified with reference
samples drawn from the Handbook of X-Ray Photoelectron
Spectroscopy [33], NIST X-ray Photoelectron Spectroscopy
Database (NIST Standard Reference Database 20, Web Ver-
sion 3.4).
Calibration was done with the carbon peak of contamina-
tion identified at 284 eV. For each TMS, the metal and sulfur
peaks were identified with their binding energies (Table 1).
The elemental surface composition of the TMS and thus the
atomic ratio of sulfur/metal (S/Me) after reaction were deter-
mined (with a relative uncertainty of 20%) from the intensity of
the metal and sulfur peaks.
Hydrogenation (HYDO)
necessary under conditions in which linear relationships be-
tween conversions and contact time were obtained. Activity of
the TMS in isomerization is defined as the number of moles of
23DMB1N formed per second per square meter of catalyst, and
activity of the TMS in HYDO is defined as the number of moles
of 23DMB formed per second per square meter of catalyst.
2.4. Analysis
The reaction products were analyzed online using a Varian
3800 gas chromatograph equipped with an automatic sampling
valve, a 50-m PONA (HP) capillary column (i.d., 0.2 mm; film
thickness, 0.5 µm), a flame ionization detector, and a cryo-
genic system. Analytical conditions were taken from previous
work [23] to obtain accurate separation: a temperature pro-
gram from 273 K (15 min) to 293 K (274.3 K/min), from
293 to 323 K (275 K/min), and then from 323 to 423 K
(280 K/min). Product identification was performed by GC-MS
coupling (Finnigan INCOS 500). The products of transforma-
tion of 2,3-dimethylbut-2-ene are reported in Table 3. Prod-
ucts obtained in our experimental conditions for the various
TMS were in agreement with previous findings with commer-
cial CoMo/Al2O3 catalyst.
Transformation of the 2,3-dimethylbut-2-ene (23DMB2N)
led to the formation of isomerization products (mainly 2,3-
dimethylbut-1-ene [23DMB1N]) and hydrogenation products
(mainly 2,3-dimethylbutane [23DMB]). Thiols were not ob-
served under our experimental conditions. It should be em-
phasized that the limit of detection for such compounds was
estimated as ca. 10 ppm wt with the GC analysis. Consequently,
it was possible to successively measure isomerization activity
(with 23DMB1N formation <30% corresponding to the ther-
modynamic equilibrium between 23DMB2N and 23DMB1N
under the experimental conditions [22,34]) and hydrogenation
activity (with the formation of 23DMB).
2.3. Reaction conditions
Catalytic activity measurements were carried out in a fixed-
bed reactor at 523 K under a total pressure of 2 MPa. The
catalyst was presulfided at 673 K for 10 h with a mixture
of 10 mol% H2S in H2 under atmospheric pressure and then
cooled to the reaction temperature in the presence of the sul-
fiding mixture. The desired reaction conditions were adjusted,
and the model feed was injected into the reactor. The model
feed was composed of 20 wt% 23DMB2N in n-heptane. The
residence time varied from 0.24 (3.3 g 23DMB2N/g cata/s) to
164 s (4.9 × 10−3 g 23DMB2N/g cata/s). The H2/feed ratio
was 360 L/L.
All experiments were performed in the presence of H2S to
maintain the various catalysts in a sulfided state. The partial
pressure of H2S (1.9 kPa) chosen corresponded to a equivalent
of 1000 wt ppm S in the feed, a typical value for European FCC
gasoline. All partial pressures conditions of the various com-
ponents for sulfidation step and catalytic activity measurements
are reported in Table 2.
For accurate activity measurements, residence times were
chosen so as to also keep the overall conversion of 23DMB2N
into 23DMB1N (isomerization [Isom]) and into 23DMB
(HYDO) nearly constant around 15% in both cases. Neverthe-
less, for the more active TMS, two different experiments were

114
A. Daudin et al. / Journal of Catalysis 248 (2007) 111–119
2.5. Microkinetic modeling
the initial hydrogen transfer from the sulfhydryl group to the
adsorbed olefin and is assumed to be rate-determining (for the
optimal model). Step (E5) is the second hydrogenation from a
MH species. Finally, step (E6) corresponds to alkane desorp-
tion with simultaneous regeneration of M∗ sites. To equilibrate
the global reaction scheme (with full M∗ recovery and avoiding
H2S overconsumption), the following terminating step is pro-
posed:
For aromatic (e.g., biphenyl, toluene) HYD and dibenzoth-
iophene HDS over sulfide catalysts, an approach combining
sulfur–metal bond energy ab initio descriptor and microkinetic
modeling has proven successful [8,11,16,17]. These results sug-
gest a general interpretation of periodic trends in catalysis by
sulfides. To explore whether such a concept holds for olefin hy-
drogenation, we tested several possible models and identified
the one best able to reproduce the experimental catalytic re-
sults. In previous works [11,16,17] ab initio microkinetic mod-
els within the Langmuir–Hinshelwood formalism put forward
that E(MS), as defined previously [10–12], appears to be a rel-
evant chemical descriptor of the catalyst for recovering volcano
curve relationships for catalytic activity. To keep the formal-
ism as general as possible, the model proposed in the current
work is a one-site model and assumes that H2S and H2 ad-
sorb dissociatively at the surface, as suggested earlier [35]. The
sulfhydryl groups created by H2S dissociation take part in the
elementary steps of the reaction. In what follows, we describe
the proposed microkinetic model for olefin hydrogenation re-
action. This model is optimized by assuming a linear relation-
ship between adsorption constants or activation energies and
the sulfur–metal bond energies calculated by DFT in previous
work [10–12]; the approach is directly inspired by Brønsted–
Evans–Polanyi (BEP) formalism.
2MSH + H2 → 2M∗ + 2H2S.
(E7)
Consequently, H2S takes part in the reaction by producing the
MS and MSH species at the surface and is regenerated at the
end of the catalytic cycle, so it does not appear in the global
reaction. This model is the one leading to the best fit of the ex-
perimental results. Among other models tested but not reported
here (for the sake of clarity) and explored previously [38], we
have verified that assuming a different rate-determining step
[such as (E5)] and/or inverting the hydrogenation steps (E4) and
(E5) leads to less satisfactory results.
2.5.2. Equation rates and Brønsted–Evans–Polanyi
relationships
Within the proposed mechanism, steps (E0)–(E4) are as-
sumed to be equilibrated, whereas step (E5), corresponding to
the addition of the first hydrogen from the sulfhydryl group, is
rate-determining. The equation of site conservation is written as
θ∗ + θS + θSH + θH + θR = 1.
(1)
2.5.1. Elementary steps of the mechanism
The reaction rate is expressed as
Equations (E0)–(E6) are the relevant elementary steps of the
mechanism involved in the one-site (M∗) microkinetic model
of olefin hydrogenation:
r(T, pH S, pH ) = kSHθRθSH
2
2
αRαS1/2
α
H2S
1/2
K
S
= kSH
(2)
(3)
M∗ + H2S ←→ MS + H2,
(E0)
(E1)
(1 + αS + αR + α1/2
α
+ αH α1/2
α
1/2
H2S
−1/2
)
H2S
2
2
S
S
K
H
2
MS + M∗ + H2 ←→ MSH + MH,
with
αi = Ki
K
H
S
pi
p0
p
H2S
2
MS + M∗ + H2S ←→ 2MSH,
(E2)
(E3)
and αS =
.
KSpH
2
K
R
M∗ + R ←→ MR,
To improve the model fit, an additional term of thermal conver-
sion, rth, is included for plotting rate equation (2). This contri-
bution, which is negligible for most of the active systems and
does not depend on E(MS), accounts for the very low (but dif-
ferent from zero) TOF of catalysts with very low/high E(MS).
In the same spirit as proposed earlier [10,17], the adsorption
constants, Ki, and activation energies are expressed as a func-
tion of a relevant intrinsic parameter of the sulfide catalyst, the
E(MS), as defined in [10–12]. The adsorption constants, Ki,
and the kinetic constant kSH are thus written as
K
SH
MR + MSH −→ MRH + MS,
MRH + MH → MRH2 + M∗,
MRH2 → RH2 + M∗.
(E4)
(E5)
(E6)
Note that step (E0), corresponding to the creation of MS species
from the M∗ site, is directly related to the sulfidation equilib-
rium state of the catalytic surface, depending on the sulforeduc-
tive conditions as revealed by DFT calculations [36,37]. Indeed,
the possibility that the effect of HDS reaction conditions change
this surface state characterized by the distribution of MS and
M∗ species cannot be excluded. At the same time, this equation
assuming a link between MS and M∗ avoids the need to con-
sider a more complex model with two independent sites. We
also tested such a two-site model but found that it did not im-
prove the final fit.
Ki = eꢀS /R−ꢀH /RT = eꢀS /Re[ꢀE
(4)
i,0+βiE(MS)]/RT
i
i
i
and
=/RT
=+γ E(MS)]/RT
,
kBT
h
kBT
h
e−ꢀG
=
e−[ꢀG
(5)
SH
0
kSH
=
where kB and h are the Boltzmann and Planck constants, re-
spectively.
If a linear relationship holds between the adsorption energy
variation (respectively activation energy) and E(MS), resulting
Hydrogen and H2S are both activated through heterolytic
dissociation at steps (E1) and (E2). Step (E3) represents ad-
sorption of the 23DMB2N olefin, labelled as R. Step (E4) is

A. Daudin et al. / Journal of Catalysis 248 (2007) 111–119
115
from Brønsted–Evans–Polanyi (BEP) relationships [39–41],
then Ki and kSH depend on the catalyst via E(MS) and the
BEP parameters, βi and γSH, depending on the adsorbed mole-
cules i. Equation (2) is finally a parameterized expression of r
depending on T , pH , pH S, and E(MS). Achieving optimal fit
5 wt%), in accordance with its significantly higher surface area.
These results demonstrate the occurrence of carbon deposition
during the experiment.
3.2. Transformation of 23DMB2N over sulfides catalysts
2
2
of the BEP parameters involves minimizing the deviation of the
theoretical values given by Eq. (2) from the experimental HYD
activities.
Finally, from the rate law’s expression, it is possible to ex-
tract an estimated value of apparent activation energies, Eapp, as
Fig. 1 shows the formation of the main products versus
residence time over rhodium sulfide: 2,3-dimethylbut-1-ene
(23DMB1N: main isomerization product) and the formation
of 2,3-dimethylbutane (23DMB: main hydrogenation product).
Skeletal isomerization products and their hydrogenation prod-
ucts were also detected in very small amounts (about 1 mol%).
Similar results were obtained for all of the TMS catalysts eval-
uated.
As reported previously [42], the isomerization of 23DMB2N
into 2,3-dimethylbut-1-ene (23DMB1N) over supported CoMo
catalyst occurred readily under the conditions of the reac-
tion, and thermodynamic equilibrium between the two iso-
mers was reached rapidly (residence times <0.2 s). Conse-
quently, the equilibrium mixture of both isomers (23DMB2N
+ 23DMB1N) is considered the reactant when measuring the
hydrogenation activity of the catalysts. After 0.2 s of resi-
dence time, 23DMB became the main product. 23DMB2N also
led to various minor products (skeletal products). The reac-
ꢀ
ꢁ
/RT
r T,pH S, pH , E(MS) = Ae−E
,
(6)
app
2
2
where A is the apparent prefactor. For given (T, p) reaction
conditions, the model allows determination of the variation of
Eapp as a function of E(MS).
3. Results and discussion
3.1. Characterization of the TMS catalysts
The crystalline structure, stoichiometry, and surface areas of
the different monometallic TMS catalysts are given in Table 4.
Their respective compositions, according to chemical analy-
sis and XRD pattern, correspond to FeS, PdS, Co9S8, Rh2S3,
RuS2, PtS, and MoS2 [2,31,32]. A mixture of two phases,
Ni9S8 and NiS, was observed for the nickel sulfide. FeS, PdS,
Co9S8, Rh2S3, RuS2, and PtS have similar specific surface ar-
eas (around 30 m2/g); nickel sulfide and palladium sulfide have
the lowest specific surface areas (5 m2/g), and MoS2 has the
highest (80 m2/g).
Table 5
Comparison of the characterization of the various TMS before (by XRD and
elemental analysis) and after the transformation of 23DMB2N (by XRD and
XPS)
TMS before catalytic activity
measurement
TMS after catalytic activity
measurement
Traces of chlorine (<1 wt%) were detected in all of the cat-
alysts prepared by the nonaqueous method.
TMS
FeS
S/Me atom
TMS
C (wt%)
S/Me atom
0.7
0.9
0.9
0.8
1.7
1.9
1.1
2.2
FeS
Ni S
3 2
1.3
0.4
0.7
0.4
0.8
1.1
1.1
2.6
1.1
0.5
0.9
1
After catalytic activity evaluation by X-ray diffraction, el-
emental analysis, and XPS, all of the catalysts were character-
ized to show the stability of the TMS (Table 5). No modification
of structure during the activity measurement was noted except
for the reduction of nickel sulfides (Ni9S8, NiS) in Ni3S2 sul-
fide phase. This was also in accordance with the decrease of the
sulfur/metal ratio from 0.9 before the test to 0.5 after the test as
measured by XPS. For the other unsupported TMS, XPS results
also confirmed a similar composition of the catalyst surface and
the bulk. The carbon content was relatively low, 0.4–1 wt%,
for all TMS except MoS2, which had a higher content (about
Ni S , NiS
9
8
Co S
Co S
9
8
9
8
PdS
PdS
Rh S
Rh S
−
2
3
2
3
RuS
PtS
RuS
PtS
1.9
1
2
2
MoS
MoS
1.8
2
2
Table 4
Physico-chemical properties of the unsupported monometallic TMS determined
by XRD, elemental analysis and specific surface area
Metal (Me)
TMS
S/Me atom
Specific surface
2
area (m /g)
Fe
Ni
Co
Pd
Rh
Ru
Pt
FeS
0.7
0.9
0.9
0.8
1.7
1.9
1.1
2.2
20
3
22
5
26
29
29
80
Ni S , NiS
9
8
Co S
9
8
PdS
Rh S
2
3
Fig. 1. Transformation of 23DMB2N over rhodium sulfide (Rh S ). Forma-
tion of 2,3-dimethylbut-1-ene (23DMB1N), 2,3 dimethylbutane (23DMB) and
2
3
RuS
PtS
MoS
2
2
products resulting of skeletal isomerization (T = 523 K, P = 2 MPa, H /feed
2
Mo
= 360 L/L).

116
A. Daudin et al. / Journal of Catalysis 248 (2007) 111–119
Scheme 1. 23DMB2N transformation, isomerization and hydrogenation products. 23DMB2N, 2,3-dimethylbut-2-ene; 23DMB1N, 2,3-dimethylbut-1-ene;
33DMB1N, 3,3-dimethylbut-1-ene; 2MP1N, 2-methylpent-1-ene; 2MP2N, 2-methylpent-2-ene; 4MP1N, 4-methylpent-1-ene; 4MP2N, 4-methylpent-2-ene;
3MP1N, 3-methylpent-1-ene; 3MP2N, 3-methylpent-2-ene; 22DMB, 2,2-dimethylbutane; 2MP, 2-methylpentane; 3MP, 3-methylpentane [23].
Fig. 2. Transformation of 23DMB2N over Ni S , PdS, Co S , FeS. Activity of
Fig. 3. Transformation of 23DMB2N over Rh S , PtS, RuS , MoS . Activity
2 3 2 2
3
2
9
8
the formation of 23DMB1N (T = 523 K, P = 2 MPa, H /feed = 360 L/L).
of the formation of 23DMB1N (T = 623 K, P = 2 MPa, H /feed = 360 L/L).
2
2
tional scheme of 23DMB2N transformation established over a
commercial CoMo supported catalyst in a previous work [22]
(Scheme 1) was in agreement with the product distribution ob-
served over unsupported TMS.
the hydrogenation activity was determined from 23DMB for-
mation when stabilization was reached. For the isomerization
step (measured by the formation of 23DMB1N), the most ac-
tive sulfide phase was Ni3S2, followed by PdS, Rh2S3, PtS,
RuS2, Co9S8, and MoS2. Conversely, for the hydrogenation
step (measured by the formation of 23DMB), Rh2S3 was the
most active phase, followed by RuS2, PtS, MoS2, PdS, Co9S8,
Ni3S2, and FeS.
For all sulfides studied, we report the evolution of the ac-
tivities for the isomerization (formation of 23DMB1N, Figs. 2
and 3) and hydrogenation reactions (formation of 23DMB,
Figs. 4 and 5) versus time on stream. No deactivation was noted
during the isomerization step for any of the TMS catalysts. For
the hydrogenation step, all of the TMS catalysts but nickel sul-
fide exhibited a more or less longer deactivation time, followed
by stabilization after 5–20 h. Indeed, PdS had the longest de-
activation and stabilization period, more than 24 h. All of the
other TMS catalysts stabilized after roughly 10 h. In each case,
3.3. Relationship between hydrogenation catalytic activity
with the metal–sulfur bond energy descriptor
The evolution of the hydrogenation activity as a function
of the E(MS) descriptor is shown in Fig. 6. The activity fol-

A. Daudin et al. / Journal of Catalysis 248 (2007) 111–119
117
Fig. 4. Transformation of 23DMB2N over Ni S , PdS, Co S , FeS. Activity
3
2
9 8
Fig. 6. Transformation of 23DMB2N over various unsupported monometallic
of the formation of 23DMB (T = 523 K, P = 2 MPa, H /feed = 360 L/L).
2
−2 −1
) (measured by the formation
sulfides. Hydrogenation activity (mol m
s
of 23DMB) versus the monometallic bulk sulfur–metal bond energy E(MS)
(T = 523 K, P = 2 MPa, H /feed = 360 L/L).
2
Table 6
Transformation of 23DMB2N—hydrogenation and isomerisation activities of
the various TMS
TMS
Activity isom
Activity HYDO
Isom/HYDO
−7
−2 −1
−7
−2 −1
10 ×(molm
s
)
10 ×(molm
s )
FeS
170
9000
2160
5430
3790
2500
3120
1120
4
5
43
1800
432
388
5
69
87
43
Ni S
3
2
Co S
5
9
8
PdS
14
720
36
36
26
Rh S
2
3
RuS
PtS
2
Fig. 5. Transformation of 23DMB2N over Rh S , PtS, RuS , MoS . Activity
2
3
2
2
MoS
2
of the formation of 23DMB (T = 623 K, P = 2 MPa, H /feed = 360 L/L).
2
not sufficient to avoid the isomerization from internal olefins
to external olefins, because external olefins are more reactive
toward hydrogenation.
lows a volcano-type curve from iron sulfide [lowest E(MS)] to
molybdenum sulfide [highest E(MS)]. Indeed, rhodium sulfide
(119 kJ/mol) has the highest hydrogenation activity. More par-
ticularly, Rh2S3 exhibits the highest hydrogenation activity, be-
tween 25 times (RuS2, PtS, MoS2) and 180 times (FeS, Ni3S2,
Co9S8) higher than the other TMS catalysts. These new results
suggest that the volcano curve concept previously reported in
the literature for the hydrogenation of aromatics [3,8,16,17] can
be extended to olefin hydrogenation. In both cases, a volcano
curve relationship was put forward between hydrogenation ac-
tivity of the different TMS catalysts and the ab initio E(MS)
descriptor, with an optimum at 119 kJ/mol corresponding to
Rh2S3. Whatever the unsaturated compounds (23DMB2N or
biphenyl and toluene), the hydrogenation activity follows a
similar trend as a function of the E(MS) descriptor. Sulfides
with E(MS) < 119 kJ/mol (e.g., FeS, Co9S8, Ni3S2) ex-
hibit the lowest hydrogenation activity. Rhodium sulfide [with
E(MS) = 119 kJ/mol] has the highest activity. Finally, sulfides
with higher E(MS) also exhibit weak hydrogenation activity.
A microkinetic modelling-based interpretation of this trend is
proposed in Section 3.4.
3.4. Microkinetic modeling
The BEP linear relationships between adsorption energies
[steps (E0)–(E3)] or activation energies of step (E4) and E(MS)
used for the kinetic modeling are plotted in Fig. 7. As ex-
pected, all adsorption energies are exothermic. The higher the
E(MS), the stronger the interaction of the active free site M∗
with 23DMB2N, –S, –SH. The trend in adsorption energies is
rather close to that obtained for toluene hydrogenation [17].
Assuming such BEP relationships, a volcano curve is recov-
ered for the olefin hydrogenation on the tested sulfide catalysts,
as shown in Fig. 8. The model is also able to reproduce the non-
symmetrical volcano shape, revealing a less pronounced rate
decrease on the right side of the volcano compared with the
left side. As indicated previously, this trend and the overall
fit cannot be correctly rendered by any other model tested in
the present study, for instance, assuming step (E5) as the rate-
determining step and/or inverting steps (E4) and (E5). We con-
firm that Rh2S3 is the optimum catalyst for intermediate values
of E(MS). This result can be interpreted within the framework
of the Sabatier principle and the analysis of the surface distrib-
ution of the different species as a function of E(MS).
It should also be emphasized that for all of the unsupported
TMS catalysts evaluated except Rh2S3, the isomerization activ-
ity is at least 10 times higher than the hydrogenation activity
(Table 6). This finding likely implies that for supported cata-
lysts, neutralization of the acidic sites of the support alone is

118
A. Daudin et al. / Journal of Catalysis 248 (2007) 111–119
Fig. 7. Linear BEP relationships between adsorption energies (steps (E0)–(E4))
and the activation energy (step (E4)) and the sulfur–metal bond energies,
E(MS).
(a)
(b)
Fig. 8. Microkinetic model according to Eq. (2) best fitting the experimen-
tal olefin hydrogenation activities plotted against sulfur–metal bond energies,
E(MS), in HDS conditions.
Fig. 9. Surface coverages of the different species as a function of E(MS):
(a) most predominant species: 23DMB2N (θ ), MSH (θ ) and MH (θ );
(b) less predominant species: M (θ ) and MS (θ ).
R
H
SH
∗
∗
S
Fig. 9a shows that for high E(MS), the surface is saturated
by MSH species, and for low E(MS), the surface is fully cov-
ered by MH species. These two extreme regions correspond to
very low numbers of 23DMB2N molecules adsorbed on the
surface. As expected from Eq. (2), the rate decreases when
θR is low at both high and low E(MS). The fact that the
rate-determining step involves hydrogenation of the adsorbed
olefin by MSH species saturating the surface at high E(MS)
implies that the rate decrease is less pronounced on the right
side of the volcano. The 23DMB2N coverage reaches a max-
imum for intermediate E(MS). At the same time, sulfhydryl
species coverage remains high at the surface, which maximizes
the product θSHθR and explains why the maximum HYD rate
is obtained for intermediate E(MS) values such as found for
Rh2S3. Also note that in this intermediate region, MH species
also occupy a nonnegligible fraction of catalytic sites. The non-
predominant MS and M∗ species also exhibit a maximum for
intermediate E(MS) values (Fig. 9b). Hence, the intermediate
sulfur–metal bond strength region provides the widest diversity
of adsorbed chemical species, which is an optimal situation for
heterogeneous catalysis, as has been suggested by Kasztelan
[43].
Fig. 10. Apparent activation energies as a function of E(MS). Line: value given
by the model, dots: experiments.
cording to the model, the variation of Eapp exhibits a minimum
close to Rh2S3, with a more pronounced increase of Eapp at
low E(MS) than at high E(MS). This trend is consistent with
the volcano curve discussed earlier. The experimental values re-
veal an increased Eapp at low E(MS) but an apparently constant
Eapp at high E(MS). Because the accuracy of experimental ac-
tivation energies depends on well-controlled (T, p) conditions
for determining these energies, perhaps we should not expect
better agreement between theoretical and experimental values
of apparent kinetic parameters.
Finally, we have estimated from the rate law given by Eq. (2)
the apparent activation energy, Eapp, resulting from the kinetic
modeling. Fig. 10 depicts the variation of Eapp as a function
of E(MS). The values of Eapp were determined experimentally
for three relevant TMS catalysts: Ni3S2, Rh2S3, and MoS2. Ac-

A. Daudin et al. / Journal of Catalysis 248 (2007) 111–119
119
4. Conclusion
[11] H. Toulhoat, P. Raybaud, J. Catal. 216 (2003) 63.
[12] H. Toulhoat, P. Raybaud, S. Kasztelan, G. Kresse, J. Hafner, Catal. To-
day 50 (1999) 629.
Based on our findings, we can conclude that it is possible to
generalize the volcano curve relationship between the ab initio
calculated metal–sulfur bond energy [E(MS)] of various un-
supported TMS catalysts and their activity to the hydrogenation
of olefins, such as 23DMB2N, under conditions typical of the
selective HDS of FCC gasoline. To the best of our knowledge,
this is the first time that such a volcano curve has been obtained
for this type of important reaction. In particular, Rh2S3, with
an intermediate metal–sulfide bond energy of 119 kJ/mol, was
the most active catalyst for the hydrogenation. The microki-
netic modeling combining ab initio metal–sulfur bond energy,
E(MS), and BEP formalism allowed a rational interpretation
of the volcano curve pattern obtained for olefin HYD. Analysis
of adsorbed species coverages show that the maximum of the
volcano correspond to the region where the widest chemical di-
versity of the surface is provided.
[13] P. Sabatier, Ber. Dtsch. Chem. Ges. 44 (1911) 2001.
[14] R.R. Chianelli, T.A. Pecoraro, T.R. Halbert, W.H. Pan, E.I. Stiefel,
J. Catal. 86 (1984) 226.
[15] M. Lacroix, N. Boutarfa, C. Guillard, M. Vrinat, M. Breysse, J. Catal. 120
(1989) 473.
[16] N. Guernalec, T. Cseri, P. Raybaud, C. Geantet, M. Vrinat, Catal. Today 98
(2004) 61.
[17] N. Guernalec, C. Geantet, P. Raybaud, M. Aouine, T. Cseri, M. Vrinat, Oil
Gas Sci. Technol. Rev. IFP 61 (2006) 515.
[18] Off. J. Eur. Commun. L76 (2003) 10.
[19] S. Hatanaka, M. Yamada, O. Sadakane, Ind. Eng. Chem. Res. 36 (1997)
1519.
[20] S. Hatanaka, M. Yamada, O. Sadakane, Ind. Eng. Chem. Res. 36 (1997)
5110.
[21] S. Hatanaka, M. Yamada, O. Sadakane, Ind. Eng. Chem. Res. 37 (1998)
1748.
[22] S. Brunet, D. Mey, G. Pérot, C. Bouchy, F. Diehl, Appl. Catal. A Gen. 278
(2005) 143.
Such a rational trend offers new ways of exploring the selec-
tivity of sulfides catalysts. We are currently further investigating
the activity of these TMS catalysts for the transformation of a
model feed of FCC gasoline to assess whether the selectivity
of TMS also can be rationalized using microkinetic modeling
combining E(MS) and BEP formalism.
[23] D. Mey, S. Brunet, C. Canaff, F. Maugé, C. Bouchy, F. Diehl, J. Catal. 227
(2004) 436.
[24] J.T. Miller, W.J. Reagan, J.A. Kaduck, J. Kropf, J. Catal. 193 (2000) 123.
[25] C. Sudhakar, M.R. Cesar, R.A. Heinrich, U.S. patent 5 525 211 (1996) to
Texaco.
[26] P.S.E. Dai, D.E. Sherwood Jr., R.H. Petty, U.S. patent 5 340 466 (1994) to
Texaco.
[27] O. Sadakane, EP 0 745 660 A1 (1996) to Mitsubishi Oil Corporation.
[28] S. Hatanaka, Catal. Surv. Asia 9 (2005) 87.
[29] S. Hatanaka, O. Sadakane, U.S. patent 6 120 679 (2000) to Mitsubishi Oil
Corporation.
[30] L. Qu, R. Prins, Appl. Catal. A Gen. 250 (2003) 105.
[31] G. Berhault, A. Mehta, A.C. Pavel, J. Yang, L. Redon, M.J. Yacaman, L.C.
Araiza, A.D. Moller, R.R. Chianelli, J. Catal. 198 (2001) 9.
[32] Y. Iwata, K. Sato, T. Yoneda, Y. Miki, Y. Sugimoto, A. Nishijma, H. Shi-
mada, Catal. Today 45 (1998) 353.
Acknowledgments
A. Daudin thanks the IFP and CNRS for a doctoral grant.
The authors also thank P. Lecour and C. Legens for the XPS
analysis (IFP Lyon) and fruitful discussions.
References
[33] C.D. Wagner, W.M. Riggs, L.E. Davis, J.F. Moulder, in: G.E. Muilenberg
(Ed.), Handbook of X-Ray Photoelectron Spectroscopy, Perkin–Elmer,
Norwalk, CT, 1979.
[34] M. Guisnet, P. Canesson, R. Maurel, Bull. Soc. Chim. Fr. 10 (1970) 3566.
[35] S. Kasztelan, Hydrotreating Technology for Pollution Control, Dekker,
New York, 1996, p. 29.
[36] P. Raybaud, J. Hafner, G. Kresse, S. Kasztelan, H. Toulhoat, J. Catal. 189
(2000) 129.
[37] P. Raybaud, J. Hafner, G. Kresse, S. Kasztelan, H. Toulhoat, J. Catal. 190
(2000) 128.
[38] (a) S. Kasztelan, D. Guillaume, Ind. Eng. Chem. Res. 33 (1994) 203;
(b) S. Kasztelan, D. Guillaume, Ind. Eng. Chem. Res. 34 (1995) 1500.
[39] J.N. Brønsted, K.J. Pedersen, Z. Phys. Chem. 108 (1924) 185.
[40] M.G. Evans, M. Polanyi, Trans. Faraday Soc. 32 (1936) 1333.
[41] M.G. Evans, M. Polanyi, Trans. Faraday Soc. 11 (1938) 34.
[42] D. Mey, S. Brunet, G. Pérot, F. Diehl, S. Kasztelan, Am. Chem. Soc. Prepr.
Div. Pet. Chem. 47 (1) (2002) 69.
[1] H. Topsøe, B.S. Clausen, F.E. Massoth, in: J.R. Anderson, M. Boudard
(Eds.), Hydrotreating Catalysis, Science and Technology, Springer-Verlag,
Berlin, 1996, p. 1.
[2] T.A. Pecoraro, R.R. Chianelli, J. Catal. 67 (1981) 430.
[3] M. Lacroix, H. Marrakchi, C. Calais, M. Breysse, C. Forquy, Stud. Surf.
Sci. Catal. 59 (1991) 277.
[4] J.P.R. Vissers, C.K. Groot, E.M. van Oers, V.H.J. de Beer, R. Prins, Bull.
Soc. Chim. Belg. 93 (1984) 813.
[5] M.J. Ledoux, O. Michaux, J. Agostini, P. Panissod, J. Catal. 102 (1986)
275.
[6] E.J.M. Hensen, H.J.A. Brans, G.M.H.J. Lardinois, V.H.J. de Beer, J.A.R.
van Veen, R.A. van Santen, J. Catal. 192 (2000) 98.
[7] J.K. Nørskov, B.S. Clausen, H. Topsøe, Catal. Lett. 13 (1992) 1.
[8] P. Raybaud, Appl. Catal. A Gen. 322 (2007) 76–91.
[9] R.R. Chianelli, G. Berhault, P. Raybaud, S. Kasztelan, J. Hafner, H. Toul-
hoat, Appl. Catal. A Gen. 227 (2002) 83.
[10] P. Raybaud, J. Hafner, G. Kresse, H. Toulhoat, J. Phys. Condens. Matter 9
(1997) 11107.
[43] S. Kasztelan, Catal. Lett. 2 (1989) 165.