Kinetic characterization of unsupported ReS2 as hydroprocessing catalyst

Article abstract of DOI:10.1016/j.jcat.2010.09.005

The hydrodesulfurization (HDS) and hydrodenitrogenation (HDN) activities of an unsupported rhenium sulfide and a sulfided CoMo/Al₂O ₃-SiO₂ catalyst are compared using 4,6- diethyldibenzothiophene (46DEDBT) and 3-ethylcarbazole (3ECBZ) as probe molecules. The active site densities and adsorption-reaction rate constants are determined from transient experiments. It is found that ReS₂ has an unusually high selectivity toward hydrogenation. As such, it has a far higher activity than CoMo/Al₂O₃-SiO₂ for desulfurizing 46DEDBT in the absence of 3ECBZ. The higher activity of ReS₂ arises from a higher turnover frequency as the active site density on ReS₂ is about one-third that on CoMo/Al₂O₃-SiO₂. Due to ReS₂'s higher hydrogenation power, the HDS of 46DEDBT on ReS ₂ is less resilient to 3ECBZ inhibition than that on CoMo/Al ₂O₃-SiO₂. ReS₂ shows about a sevenfold activity advantage over the CoMo/Al₂O₃-SiO ₂ catalyst in the hydrodenitrogenation of 3ECBZ. The results shed some light on the HDS-HDN interactions in real-feed deep HDS.

Full text of DOI:10.1016/j.jcat.2010.09.005

Journal of Catalysis 276 (2010) 114–122
Contents lists available at ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Kinetic characterization of unsupported ReS2 as hydroprocessing catalyst
1
⇑
T.C. Ho , Q. Shen , J.M. McConnachie, C.E. Kliewer
Corporate Strategic Research Labs, ExxonMobil Research and Engineering Co., Annandale, NJ 08801, United States
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 8 July 2010
Revised 2 September 2010
Accepted 4 September 2010
Available online 16 October 2010
The hydrodesulfurization (HDS) and hydrodenitrogenation (HDN) activities of an unsupported rhenium
sulfide and a sulfided CoMo/Al2O3–SiO2 catalyst are compared using 4,6-diethyldibenzothiophene
(46DEDBT) and 3-ethylcarbazole (3ECBZ) as probe molecules. The active site densities and adsorption-
reaction rate constants are determined from transient experiments. It is found that ReS2 has an unusually
high selectivity toward hydrogenation. As such, it has a far higher activity than CoMo/Al2O3–SiO2 for
desulfurizing 46DEDBT in the absence of 3ECBZ. The higher activity of ReS2 arises from a higher turnover
frequency as the active site density on ReS2 is about one-third that on CoMo/Al2O3–SiO2. Due to ReS2’s
higher hydrogenation power, the HDS of 46DEDBT on ReS2 is less resilient to 3ECBZ inhibition than that
on CoMo/Al2O3–SiO2. ReS2 shows about a sevenfold activity advantage over the CoMo/Al2O3–SiO2 catalyst
in the hydrodenitrogenation of 3ECBZ. The results shed some light on the HDS–HDN interactions in real-
feed deep HDS.
Keywords:
Hydrodesulfurization
Hydrodenitrogenation
Rhenium sulfide
Catalyst inhibition
Diethyldibenzothiophenes
Ethylcarbazole
Ó 2010 Elsevier Inc. All rights reserved.
Competitive adsorption
1. Introduction
This, coupled with the dwindling supply of high-quality crude oils
and the growing need to reduce organonitrogen and polynuclear
Catalytic hydrodesulfurization (HDS) and hydrodenitrogenation
(HDN) are used in the refining industry for removing indigenous
organosulfur and organonitrogen species from various petroleum
fractions such as middle distillates (200–370 °C boiling range; die-
sel, jet fuel, heating oil, etc.). The majority of the catalysts used are
Al2O3-supported MoS2 and/or WS2 into which Co and Ni are incor-
porated as promoters [1]. The mounting concerns over the environ-
ment and the growing need to process heavier/dirtier crude oils
have made it mandatory to develop new HDS and HDN catalysts
for producing high-quality, ultra-clean fuels. For the HDS of a die-
sel molecule such as dibenzothiophene (DBT), the reaction pro-
ceeds along two pathways [1]. The hydrogenolysis pathway gives
biphenyl-type products (BP) through direct sulfur extraction. The
hydrogenation pathway produces cyclohexylbenzene-type prod-
ucts (CHB) via hydrogenation of one of the phenyl rings followed
by C–S bond cleavage. More than ten years ago, diesel sulfur spec-
ifications were between 350 and 500 wppm in many countries.
Refiners could meet such specifications by desulfurizing DBT and
some of its alkyl derivatives. However, today’s 10–15 wppm spec-
ifications in a growing number of countries require desulfurization
of 4-alkyl- and 4,6-alkyl-DBTs. Desulfurization of such sterically
hindered DBTs relies heavily on the hydrogenation pathway [2].
aromatic species, points to the importance of catalyst’s hydrogena-
tion function. Much effort has been made to increase hydropro-
cessing activity via enhancement of catalyst’s hydrogenation
function.
Unsupported sulfides in general are known to be more selective
toward hydrogenation than commercial CoMo/Al2O3 sulfide
catalysts [3–5]. For instance, Hermann et al. [3] reported that
unsupported sulfides of Mo, Ru, Nb, Rh, and Pd are more hydroge-
nation selective than sulfided CoMo/Al2O3 catalysts in the HDS
of 4,6-dimethyldibenzothiophene. Unsupported RuS2 shows
a
remarkably high activity for deep HDS of a tough-to-desulfurize
petroleum fraction at low hydrogen pressures [6]. An unsupported
metal sulfide called Nebula was recently commercialized for deep
HDS and HDN applications [7]. For exploration and development of
new unsupported sulfide catalysts, it is essential to gain a quanti-
tative understanding of the differences between unsupported sul-
fides and conventional Al2O3-supported catalysts in terms of
active site density and adsorptivity. Being in a poorly crystalline
and highly disordered state, transition metal sulfides are notori-
ously difficult to characterize. The most direct way of estimating
active site density is to conduct combined modeling and experi-
mental studies under transient conditions.
MoS2 and WS2 are both trigonal prismatic and have a layered
structure. The active sites are commonly believed to be sulfur
vacancies associated with exposed Mo or W cations and SHꢀ/S2ꢀ
groups on the MoS2/WS2 edge [1]. Recently, nonvacancy sites have
also been identified; they are metallic brim sites with remarkable
⇑
Corresponding author.
E-mail address: teh.c.ho@exxonmobil.com (T.C. Ho).
Present address: Department of Chemical Engineering, University of Texas,
1
Austin, TX 78712, United States.
0021-9517/$ - see front matter Ó 2010 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2010.09.005

T.C. Ho et al. / Journal of Catalysis 276 (2010) 114–122
115
Nomenclature
g
g
knNf/kHDN, defined in Eq. (11)
knNf =k0
qn
S
concentration of adsorbed nitrogen atom on catalyst
ꢀ
concentration of sulfur atom in the fluid phase (
lmole/
n
kHDS
kHDN
kn
surface HDS rate constant (1/s)
surface HDN rate constant (1/s)
adsorption rate constant for nitrogen (cc feed/s/
equivalent pseudo-first-order rate constant, defined in
Eq. (19).
adsorption rate constant for sulfur (cc feed/s/
desorption rate constant for nitrogen (1/s)
desorption rate constant for sulfur (1/s)
adsorption selectivity defined in Eq. (13)
cc feed or g/cc)
sulfur atom concentration in feed liquid, g S/cc or
lmole/cc feed
sulfur concentration in the fluid phase at Steady State I
time
Sf
l
mole)
b
ko
S
t
ks
kn0
ks0
p
l
mole)
v
z
superficial velocity
axial distance from the entrance of the reactor (cm)
Greek symbols
N
concentration of nitrogen atom in the fluid phase
mole/cc feed or g/cc)
feed nitrogen atom concentration, g N atom/cc or
mole/cc
catalyst maximum site capacity, lmole N adatom/g cat
e
qp
s
bed void fraction (0.3)
sulfided catalyst packing density in the bed (1.15 g/cc)
space time based on catalyst weight
(
l
Nf
l
hn
qn/qm, the fractional coverage of adsorbed nitrogen
qm
hydrogenation functionality [8]. Reduction of molybdenum in
MoS2 from Mo(IV) to Mo(III) transforms the molybdenum coordi-
nation from hexagonal to distorted octahedral, with concomitant
formation of localized Mo–Mo bonds. The fully reduced compound,
LiMoS2, has a structure similar to ReS2, which might be expected as
Re(IV) is isoelectronic with Mo(III) [9]. Rhenium in ReS2 is bound
by sulfur in an octahedral environment, and the rhenium atoms
connect through a diamond-chain motif of Re–Re bonds [10,11]
as shown in Fig. 1. Rhenium sulfide and molybdenum/tungsten
sulfides also share a common anisotropic characteristic, in that
only a small fraction of the total surface area is expected to be cat-
alytically active. By contrast, the isotropic structure of RuS2 crystal-
lites gives a high exposure of active surface metal atoms to
adsorbates.
Rhenium sulfide is active for desulfurizing thiophene and DBT
[12]. Jacobson et al.’s study using an indole-DBT-naphthalene mix-
ture has indicated that the HDN activities of five unsupported me-
tal sulfides decrease in the following order: ReS2 > MoS2 > RuS2 >
NbS2 ꢁ Co9S8 [13]. This activity ranking was rationalized in terms
of a bond energy model in which the metal–sulfur bond strengths
are measured by temperature-programmed reduction of H2S-pre-
treated catalysts. Jacobson et al. also reported the following HDS
activity ranking: RuS2 > ReS2 > MoS2 > NbS2 ꢁ Co9S8. Thus, ReS2 is
an active and selective HDN catalyst based on experiments with
the indole-DBT-naphthalene mixture. It is also more active than
MoS2 for both HDS and HDN. The performance difference between
ReS2 and MoS2 may be related to their reducibility [14]. It is un-
clear that the high activity of unsupported rhenium sulfide is due
to the number or the quality of active sites. Also, the performance
of rhenium sulfide under ultra-deep HDS conditions has not been
examined.
9
1
C2H5
4
5
2
8
6
3
7
2
7
S
N
H
1
8
C2H5
H5C2
Fig. 2. Model compounds used in the present study: 4,6-diethyldibenzothiophene
and 3-ethylcarbazole.
In practice, a viable ultra-deep HDS catalyst requires a high
hydrogenation power and a high resilience to inhibition by nitro-
gen compounds [4,15,16]. Of the multitude of organonitrogen
inhibitors, the most potent ones in ultra-deep HDS processes are
alkylcarbazoles, which are abundantly present in severely hydro-
treated oils [17–20]. Fig. 2 shows a pair of realistic model com-
pounds for probing deep HDS chemistry and process, as will be
discussed later. They are 4,6-diethyldibenzothiophene (46DEDBT)
and 3-ethylcarbazole (3ECBZ); both of them rely primarily on cat-
alyst’s hydrogenation function to remove their heteroatoms [5,21].
A trace amount (five wppm as nitrogen atom) of 3ECBZ can drasti-
cally reduce the activity of a commercial CoMo/Al2O3–SiO2 catalyst
for desulfurizing 46DEDBT [21]. It is of both fundamental and
practical importance to investigate 46DEDBT-3ECBZ competitive
adsorption/reaction on various unsupported metal sulfides.
The present work is a kinetic characterization of unsupported
ReS2 prepared from (NH4)4(Re4S22)ꢂ2H2O as deep HDS and HDN
catalysts. We aim to gain a quantitative understanding of compet-
itive adsorption dynamics using 46DEDBT and 3ECBZ as probe
molecules. Through analyses and modeling of transient response
data, the active site density and associated adsorption/reaction
rate constants are determined. The results are compared with
those for a commercial sulfided CoMo/Al2O3–SiO2 catalyst in terms
of intrinsic activity/selectivity and organonitrogen tolerance. The
results shed some light on the nature of HDS–HDN interactions
in ultra-deep HDS of petroleum fractions.
2. Experimental
2.1. Materials
The rhenium sulfide catalyst was prepared via a procedure dif-
ferent from those described in the review of Escalona et al. [12].
Specifically, it was prepared by the thermal decomposition of
Fig. 1. Structure of ReS2: gray ball, rhenium; yellow ball, sulfur. The metal–metal
bonding is highlighted in black.

116
T.C. Ho et al. / Journal of Catalysis 276 (2010) 114–122
(NH4)4(Re4S22)ꢂ2H2O {rhenate(4-), tetra-
l3-thioxohexakis[
l-(tri-
2.4. Steady-state and transient experiments
thio)]tetra-, tetraammonium, dihydrate}, which was synthesized
according to the procedure of Müller et al. [22]. The thus-synthe-
sized solid product, after being pressed into pellets at 20,000 psig,
was broken into granules and then heat treated in a tube furnace
with a ramp rate 25 °C/min to 400 °C and held for one hour under
a nitrogen flow at 200 cc/min. The resulting stoichiometry is
approximately ReS2.3. Upon exposure to the hydrogen atmosphere
in the reactor, it loses excess sulfur, resulting in a stoichiometry of
ReS2 and a weight loss of about 4%. The ReS2 catalyst, with a BET
surface area of 55 m2/g, was used in the form of 20–40 mesh gran-
ules in the transient response experiments. Both 46DEDBT and
3ECBZ were synthesized through known procedures.
The reaction conditions used in this study are 1.83 MPa hydro-
gen pressure, 265 °C, eight liquid weight hourly space velocity
(WHSV), and 116 cc H2/cc liquid feed. The sequence of the experi-
ments is as follows. The reactor was started with feed
A
(1067 ppmw total sulfur as atoms) and pure hydrogen to obtain
the baseline data in the absence of 3ECBZ (Steady State I). Follow-
ing this, Feed A was replaced by Feed B (1067 ppmw sulfur plus
80 ppmw nitrogen as atoms) at time ‘‘zero” at the same conditions.
After the reactor reached a new steady state (Steady State II), Feed
A was put back on stream to strip the reversibly adsorbed nitrogen
species off the catalyst, in an attempt to recover the lost HDS activ-
ity. To speed up the slow desorption of chemisorbed nitrogen spe-
cies, the reactor temperature was raised to 300 °C for 5 h and then
lowered back to 265 °C. The sulfur and nitrogen concentrations in
the liquid effluents were intermittently measured throughout the
entire experiment.
2.2. Transmission electron microscope
Large agglomerates of ReS2 particles were crushed into fines
(ꢃ100 nm thick) using an agate mortar and pestle. The fines were
dusted onto a standard, holey-carbon-coated, 200 mesh, Cu grid
and transferred into a Philips CM200F TEM/(S)TEM. The micro-
scope was operated in the bright field TEM imaging mode at an
accelerating voltage of 200 kV. Randomly selected areas of the
material were imaged at an ‘‘on screen” magnification of
175,000ꢄ using a Gatan model 694 CCD camera system and
Gatan’s Digital Micrograph v.2.5 software. Stack heights were mea-
sured from over 300 ReS2 crystallites. Stack heights were plotted,
and mean and median stack height values were calculated using
KaleidaGraph v.3.0.
Since C4BP-to-C4CHB hydrogenation is very slow, the selectiv-
ity toward hydrogenation for an isothermal plug-flow reactor can
be approximately gauged by the following concentration ratio in
the liquid effluent
wt% C4CHB in product
wt% C4BP in product
cDEDBT
¼
ð1Þ
Note that cDEDBT in general is a decreasing function of temperature
[21] and that the C2CHB/C2BP ratio roughly tracks the C4CHB/C2BP
ratio.
3. Experimental results
2.3. Reactors, procedures, and analyses
3.1. TEM characterization
A cocurrent fixed-bed reactor, made of a 3/8-inch ID 316 stain-
less steel pipe, was operated isothermally in up-flow mode to
avoid incomplete catalyst wetting and bypassing. Glass beads were
charged in the fore and aft zones to achieve vapor–liquid equilib-
rium. To mitigate the axial dispersion effect, three grams catalyst
was uniformly mixed with an equal volume of glass beads in the
reactor central zone.
The liquid products were analyzed by GC/MS and GC using a
75% OV1/25% SW-10 fused silica capillary column. In addition,
the total nitrogen was quantified by combustion and chemilumi-
nescence using the Antek analyzer. Due to their low concentra-
tions, individual HDN products were not measured. The carrier
solvent dodecane (analytic pure grade) was essentially inert under
the reaction conditions studied. The product gases were vented
through a caustic scrubber followed by a wet test meter.
Table 1 lists the compositions of the feed mixtures (density =
0.72 g/cc), with the balance being dodecane.
Fig. 3 is an electron micrograph of ReS2. The bright field TEM
image presents numerous ReS2 crystallites. Each crystallite is com-
posed of parallel ‘‘dark” and ‘‘light” layers. In the bright field TEM
image, the ‘‘dark” layers correspond to rows of Re atoms, while
the ‘‘light” layers correspond to rows of sulfur atoms. It is worth-
while noting that at this magnification, the double layer of sulfur
atoms present between the layers of Re atoms is not resolved. It
is also clear that these layered features are not immediately evi-
dent in all regions in the image. This effect arises from slight thick-
ness changes across the field of view which affects focus. By
watching the sample while slightly moving though either side of
focus, it was possible to identify regions of the agglomerate where
Besides H2S, the main products obtained from the HDS
of 46DEDBT were identified to be diethylbiphenyls (C4BP),
diethylcyclohexylbenzenes (C4CHB), ethylcyclohexylbenzenes
(C2CHB), ethylbiphenyls (C2BP), ethylbenzenes, and ethylcyclo-
hexanes. These products accounted for more than 98% of 46DEDBT
converted. Diethylbicyclohexyl, ethylbicyclohexyl, biphenyl and
cyclohexylbenzene were present in trace amounts. Products from
partial hydrogenation of 46DEDBT, if present, are negligibly small
in concentration.
Table 1
Feed compositions (wt.%).
Probe molecules
Feed A
Feed B
10 nm
4,6-Diethyldibenzothiophene
3-Ethylcarbazole
0.8
0.0
0.8
0.112
Fig. 3. Electron micrograph of ReS2.

T.C. Ho et al. / Journal of Catalysis 276 (2010) 114–122
117
140
120
100
80
in the HDS level, the recovery of the HDS activity becomes increas-
ingly sluggish. The activity does not come close to its original level
even after 200 h.
Minimum
Maximum
Points
Mean
Median
1
7
As mentioned, a similar transient response experiment was
conducted with the sulfided CoMo/Al2O3–SiO2 catalyst at 2.4
WHSV and otherwise identical conditions [21]. The HDS level
dropped from 70% at Steady State I to less than 10% at Steady State
II. In terms of intrinsic gravimetric activity (pseudo-first-order rate
constant, cc liquid feed/g cat/s), ReS2 shows approximately a 9-fold
activity advantage over sulfided CoMo/Al2O3 at Steady State I. By
intrinsic, it is meant that the HDS activity is measured in the
absence of any inhibitor such as nitrogen or aromatic species.
317
2.9
3
60
40
20
3.3. Hydrogenation selectivity
0
0
2
4
6
8
10
Fig. 6 shows the behavior of cDEDBT during the breakthrough and
# of Stacks
stripping periods. Initially, cDEDBT is about 11, indicating an over-
whelming preference for the hydrogenation pathway. For perspec-
tive, cDEDBT ꢃ 5 for the CoMo/Al2O3–SiO2 catalyst at the same
conditions [21]. Referring to Fig. 6, cDEDBT drops rapidly from ꢃ11
to as low as ꢃ0.3 after the 3ECBZ assault. That is, the inhibiting
process is so selective toward the hydrogenation sites that the
hydrogenation pathway is almost completely shut off for HDS at
Steady State II. And the chemisorption of 3ECBZ is so strong that
the adsorbed nitrogen species would not come off the catalyst sur-
face easily. As a result, a complete recovery of the hydrogenation
function does not seem possible. This quasi-irreversible behavior
is not uncommon [23–25]. The long-lingering adsorbed nitrogen
species tend to polymerize and eventually form coke [26].
Fig. 4. Stack heights were measured from over 300 ReS2 crystallites. Mean and
median stack height values were calculated using KaleidaGraph v.3.0. The mean
and median stack height values were about 2.9 and 3.0 layers per ReS2 crystallite,
respectively.
entire ReS2 crystallites were visible. Their stack heights were mea-
sured from these regions, and the values are shown in Fig. 4. Crys-
tallite stack heights ranging from 1 to 7 layers were observed in
this analysis, and the mean and median stack height values were
about 2.9 and 3.0 layers per ReS2 crystallite, respectively.
3.2. HDS activity
4. Modeling of inhibition dynamics
Fig. 5 displays the transient response of DEDBT desulfurization
on ReS2 to a step jump in the feed nitrogen atom content from zero
to 80 ppmw. Initially, at Steady State I (Feed A), the HDS level is
about 96% at a WHSV of eight. During the breakthrough period,
the HDS activity drops precipitately. When the system reaches
Steady State II, sulfur removal decreases to about 17%. Evidently,
the presence of 3ECBZ has a catastrophic effect on ReS2’s activity
for desulfurizing 46DEDBT. As shown in Fig. 5, the catalyst was
subsequently stripped of chemisorbed organonitrogen with Feed
A at 265 °C to recover the HDS activity. After about 150 h of strip-
ping, the reactor temperature was raised to 300 °C for five hours
and then lowered back to 265 °C. As can be seen, after an initial rise
4.1. Simplifying assumptions
Before setting out the model equations, we make the following
simplifying assumptions. (1) The catalyst surface is Langmuirian
and energetically uniform; (2) Hydrogen is adsorbed on sites that
are different from those for the adsorption of sulfur and nitrogen
species; (3) For adsorption of organosulfur and organonitrogen
species, only the hydrogenation sites are considered due to
overwhelming dominance of the hydrogenation pathway; (4)
Because of the high hydrogen/46DEDBT volume ratio, the hydrogen
100
12
A
80
St St I
A
300oC
5 h
300oC 5 h
60
8
Stripping
A
γ
40
Stripping
B
4
A
20
265oC
200
265oC
St St II
B
0
150
0
200
250
300
350
400
150
250
300
350
400
time, hr
time, hr
Fig. 5. Percentage of HDS vs. time (h) obtained with ReS2; Feed A was used at
Steady State I, Feed B was used in Steady State II; 265 °C, 8 WHSV, 1.83 MPa, and
116 cc H2/cc liquid feed; Feed A was used for stripping; After the 370th hour, the
stripping temperature was raised to 300 °C for 5 h.
Fig. 6. cDEDBT vs. time (h) obtained with ReS2; Feed A was used at Steady State I,
Feed B was used in Steady State II; 265 °C, 8 WHSV, 1.83 MPa, and 116 cc H2/cc
liquid feed; Feed A was used for stripping; After the 370th hour, the stripping
temperature was raised to 300 °C for 5 h.

118
T.C. Ho et al. / Journal of Catalysis 276 (2010) 114–122
N = Nf . The initial conditions correspond to Steady State I. Thus,
at t = 0 (z P 0), we have N = qn = 0 and
pressure is constant and the surface hydrogen coverage is high; (5)
The effect of molar change due to reaction is negligible; (6) The
inhibiting effects of H2S, NH3, and heteroatom-free hydrocarbons
are relatively insignificant compared with that of organonitrogen
species; (7) Axial dispersion effect and the velocity changes due
to adsorption are negligible; (8) The concentrations of sulfur and
nitrogen compounds are constant throughout the catalyst particles;
(9) The reactor is isothermal and isobaric, with negligible mass
transfer effects; (10) Both HDS and HDN occur irreversibly on the
catalyst surface and hydrogen addition is not rate limiting; (11) Cat-
alyst deactivation by coking is negligible on the time scale of the
inhibition event; and (12) Liquid feed and hydrogen reach physical
equilibrium before entering the catalyst bed.
dS
dz
v
¼ ꢀð1 ꢀ
eÞ
qPksqmS
ð5Þ
where S = S(z) at t = 0. Eq. (5) can be readily integrated with the
reactor inlet condition S(0) = Sf to yield
ꢀ
ꢁ
ð1 ꢀ
e qpksq
v
Þ
SðzÞ ¼ Sf exp ꢀ
m z
ð6Þ
With the above boundary and initial conditions, Eqs. (2)–(4) were
solved numerically by using an implicit finite difference scheme.
The parameter estimation was carried out through optimization
with constraints derived from steady-state experiments. The details
can be found in an earlier paper [27].
Let S and N be the sulfur atom and nitrogen atom concentra-
tions in the flowing stream (micromole/cc feed), respectively. Here,
both S and N include sulfur- and nitrogen-containing heterocyclic
species formed from conversions of 46DEDBT and 3ECBZ. Also,
let qn and qs be the adsorbed nitrogen and sulfur atom concentra-
tions, respectively. Denoting qm as the catalyst’s total chemisorp-
tion capacity on the hydrogenation sites, then hn = qn/qm is the
In practice, what matter are the HDS and HDN behaviors at
Steady State II, which are dictated by the following equations:
dS
dz
ð1 ꢀ
e qPksqm
Þ
v
v
¼ ꢀ
S
ð7Þ
ð1 þ knN=kHDN
Þ
fractional coverage of adsorbed nitrogen. For sulfur species, ks, k0 ,
s
dN
dz
ð1 ꢀ qPknqm
eÞ
and kHDS are the adsorption rate constant (cc liquid/s/lmole),
¼ ꢀ
N
ð8Þ
ð1 þ knN=kHDN
Þ
desorption rate constant (sꢀ1), and surface HDS rate constant
(sꢀ1), respectively. The corresponding rate constants for nitrogen
Thus, HDS is inhibited by nitrogen species, whereas HDN is self-
inhibited. Recall that a basic tenet of the present treatment is that
the inhibiting effect of sulfur species is negligibly small compared
with that of nitrogen species. And the desorption of nitrogen species
species are kn, k0 , and kHDN
.
n
To construct a model that is as simple as possible necessitates
further pruning. To this end, the slow desorption of organonitrogen
species can be considered as frozen on the time scale of the inhibi-
tion transient. At the other extreme, the surface HDS is too fast to
influence the sluggish inhibition process (kHDS ꢅ kHDN because the
C–S bond is much weaker than the C–N bond). The desorptions of
nitrogen or sulfur compounds are not fast enough to attain a local
adsorption–desorption equilibrium as the nitrogen and sulfur
adsorption fronts travel down the bed during the breakthrough
period. The consequence is that the inhibition transient is mainly
governed by the interplay of sulfur adsorption, nitrogen adsorp-
tion, and surface HDN.
is too slow (k0 ꢆ kHDN) to be kinetically relevant. Hence, the extent
n
of organonitrogen inhibition is driven by kHDN, which governs sur-
face concentration of chemisorbed nitrogen species.
4.2. Governing parameters and steady-state behavior
To identify governing parameters, Eqs. (7) and (8) are made
dimensionless with the following scaled variables: s = S/Sf, n = N/
Nf, and n = z/L, Also, the space time
. Eqs. (7) and (8) become
s
is defined as
s
= (1 ꢀ
e)qpL/
v
In light of the above, we set knNf ꢅ k0 , kHDN ꢅ k0 , kn ꢅ ks,
n
n
ds
dn
ksqms
1 þ gn
ksaSf ꢆ kHDS, and qs ꢆ qn. These stipulations were justified previ-
ously [16,27,28]. It should be noted that here we essentially use
organonitrogen adsorbates to ‘‘sample” (or ‘‘titrate”) the active
sites in situ. Thus, qm is a measure of the maximum site density
¼ ꢀ
s
ð9Þ
dn
dn
knqms
1 þ gn
¼ ꢀ
n
ð10Þ
with the units of
Reactor specifications are bed void fraction,
ticle density, p = 1.15 g/cc; and bed length, L = 3.82 cm. The con-
centrations of sulfur and nitrogen atoms in the feed are
Sf = 58.8 mole/cc and Nf = 4.1 mole/cc, respectively. With being
l
mole nitrogen adatom/g catalyst.
e
= 0.3; catalyst par-
Here, s is slaved to nitrogen concentration; s cannot be determined
unless n is obtained from solving Eq. (10). The corresponding
boundary conditions are s = n = 1 at n = 0. The dimensionless param-
eter g in the denominator is defined as
q
l
l
v
the superficial fluid velocity based on empty reactor, the transient
response of the system after introducing 3ECBZ (time t > 0) is de-
scribed by the following plug-flow, one-dimensional model [27]
knNf
kHDN
g ꢇ
ð11Þ
Thus, g represents the ratio of the nitrogen adsorption rate to the
surface HDN rate. It is a measure of the inhibition intensity. It is rel-
evant to point out that if the desorption is fast enough to maintain a
(i) S mass balance in the fluid phase
local adsorption–desorption equilibrium (k0 ꢅ kHDN), then the inhi-
@S
@z
@S
@t
n
v
v
þ
e
þ ð1 ꢀ
e
ÞqPksSðqm ꢀ qnÞ ¼ 0
ð2Þ
ð3Þ
ð4Þ
0
ꢀ
bition intensity should be gauged by g ꢇ knNf =kn ¼ KNf where
K ꢇ kn=k0 is the adsorption equilibrium constant.
n
(ii) N mass balance in the fluid phase
Dividing Eq. (9) by Eq. (10), we obtain
@N
@z
@N
@t
þ
e
þ ð1 ꢀ ÞqPknNðqm ꢀ qnÞ ¼ 0
e
ds
dn
s
n
¼ p
ð12Þ
(iii) N mass balance in the solid phase
At a given set of temperature, pressure, and hydrogen treat gas rate,
@qn
@t
¼ knNðqm ꢀ qnÞ ꢀ kHDNqn
s and n are functions of the space time
eter p is defined by
s
. The dimensionless param-
The model has four parameters (kHDN, kn, ks, and qm). Only one of the
four parameters is adjustable due to steady-state constraints [27].
The required boundary conditions at z = 0 (t > 0) are S = Sf and
ks
p ꢇ
kn
ð13Þ

T.C. Ho et al. / Journal of Catalysis 276 (2010) 114–122
119
Hence, p is an adsorption ratio reflecting the catalyst’s adsorption
affinity for sulfur species relative to that for nitrogen species.
It transpires from the above development that the inhibition
dynamics is characterized by g and p, both of which are functions
of catalyst properties for a given feed. Eq. (12) upon integration
from n = 0 to n = 1 (reactor outlet) yields a relationship between s
and n at the reactor outlet at Steady State II
ing the 3ECBZ-containing feed. As can be seen, the dynamic process
leading up to Steady State II is well predicted by the model.
5.2. Spatiotemporal behavior
Having determined the model parameters, we now examine the
spatiotemporal behavior of the system. Fig. 8 shows the sulfur con-
centration in the fluid phase as a function of bed position at succes-
sive elapsed times. Initially (t = 0 h), the reactor sulfur profile in the
absence of 3ECBZ shows an exponential decline. Subsequently, the
fluid sulfur content increases as the concentration wave travels to-
ward the reactor outlet. The advancing sulfur front is not at a local
adsorption–desorption quasi-equilibrium with the active sites.
After 10 h of nitrogen assault, the majority of the active sites in
the bed are occupied by nitrogen species; the bulk of 46DEDBT
stays in the fluid phase as a ‘‘spectator.” As a result, the sulfur pro-
file remains nearly flat and becomes only slightly concave down-
ward near the reactor outlet where some active sites have not
yet been blocked at that moment. The sulfur concentration in reac-
tor effluent increases as time goes by. The speed with which the
sulfur travels through the bed is faster than that calculated for
the CoMo/Al2O3–SiO2 catalyst.
p
sðs
Þ ¼ nð
sÞ
ð14Þ
The simple model-compound model represented by Eqs. (9) and
(10) captures two characteristic features of real-feed ultra-deep
HDS: sulfur removal is predominantly controlled by nitrogen spe-
cies [15] and nitrogen removal is largely governed by self-inhibition
[20].
5. Results and discussions
5.1. Rate parameters and active site densities
The model parameters obtained from data fitting are listed in
Table 2, which also includes the results obtained previously for
the sulfided CoMo/Al2O3–SiO2 catalyst [27].
The solid curves in Fig. 7 are model predictions, which compare
well with the measured nitrogen exit concentration (wppm) and
the percentage of HDS as functions of elapsed hour after introduc-
Fig. 9 shows multiple snapshots of nitrogen concentration in the
fluid phase resulting from the combined action of adsorption and
surface HDN reaction. The moving front is accompanied by a
decreasing N due to HDN whose rate plays a pivotal role in the
inhibition process. The model must account for surface HDN reac-
tion; otherwise, it would not predict the nitrogen breakthrough
curve and the steady-state nitrogen concentration. During the
breakthrough period, the catalyst bed is divided into two zones.
The upstream zone is saturated with organonitrogen and perform-
ing HDN, while the downstream zone has not seen organonitrogen
yet and is carrying out HDS.
Fig. 10 shows the percent coverage of active sites (100qn/qm) by
nitrogen species as a function of bed length at different elapsed
times. Prior to the breakthrough, the catalyst bed near the reactor
outlet is virtually free of nitrogen species as the upstream bed ad-
sorbs and denitrogenates all incoming nitrogen species. After 10 h,
the majority of active sites are occupied by nitrogen species, thus
diminishing HDS almost completely.
Table 2
Model parameters for unsupported ReS2 and CoMo/Al2O3–SiO2.
Parameters
Unsupported ReS2
CoMo/Al2O3–SiO2
ks, cc feed/s/
l
mole
154 ꢄ 10ꢀ6
78.9 ꢄ 10ꢀ5
121 ꢄ 10ꢀ6
96.4
6.7 ꢄ 10ꢀ6
2.7 ꢄ 10ꢀ5
6.3 ꢄ 10ꢀ6
271.4
kn, cc feed/s/
lmole
kHDN, sꢀ1
qm, lmole N adatom/g cat
p
0.2
0.25
g
27
17
(265 °C, 1.83 MPa hydrogen pressure)
100
80
60
40
20
0
predicted %HDS
predicted %HDS
predicted effluent Nitrogen(wppm)
predicted effluent Nitrogen(wppm)
x 10-3
1.2
experimental %HDS
experimental %HDS
experimental effluent N(wppm)
experimental effluent N(wppm)
1
0.8
0.6
0.4
t = 0 (hr)
t = 0.67 (hr)
t = 2.68 (hr)
t = 4.69 (hr)
t = 10.47 (hr)
0.2
150
170
190
210
230
0
Time, (Hr)
0
0.5
1
1.5
2
2.5
3
3.5
4
Reactor length (cm)
Fig. 7. Percentage of HDS and total nitrogen concentration (wppm) at reactor exit
as functions of elapsed time following introduction of 3-ethylcarbazole; ReS2;
265 °C, 8 WHSV, 1.83 MPa, and 116 cc H2/cc liquid feed. Solid curves are predicted
from the model.
Fig. 8. Predicted moving sulfur concentration front s = S/Sf for different elapsed
times. ReS2 catalyst.

120
T.C. Ho et al. / Journal of Catalysis 276 (2010) 114–122
x 10-5
nometallic complex [29]. The latter serves as
-adsorption through six carbon atoms to a metal site on a catalyst
surface. Specifically, it was found that the relative binding powers
of 3ECBZ and 46DEDBT in CpRu(
6-arene)+ complexes are 6.5 and
1.1, respectively. So the binding (adsorption) ratio is 1.1/6.5 =
p = 0.17, which is comparable to p = 0.2 for ReS2. Rhenium sulfide
and ruthenium sulfide have similar metal–sulfur bond energies
[3]. Based on these observations, it appears that the relative bind-
ing (adsorption) tendencies of 46DEDBT and 3ECBZ on Re and Ru
are comparable.
a model for
9
8
p
t = 0 (hr)
t = 0.67 (hr)
t = 2.68 (hr)
t = 4.69 (hr)
t = 10.47 (hr)
g
7
6
5
4
To compare the relative susceptibility of HDS to organonitrogen
inhibition between ReS2 and sulfided CoMo/Al2O3–SiO2, we com-
pare the g values for the two catalysts. The ReS2 catalyst gives
g = 27 vs. g = 17 for CoMo/Al2O3–SiO2. This indicates that although
ReS2 is intrinsically more active than CoMo/Al2O3–SiO2 for desulfu-
rizing 46DEDBT, it is more vulnerable to 3ECBZ inhibition. As a
result, ReS2 suffers a greater loss of HDS activity than CoMo/
Al2O3–SiO2 in the presence of 3ECBZ. This is consistent with ReS2’s
3
2
1
0
-1
high hydrogenation power (high cDEDBT). A quantitative analysis of
0
1
2
3
4
the relative nitrogen tolerance of hydrogenation and hydrogenoly-
sis sites has been made for the sulfided CoMo/Al2O3–SiO2 catalyst
[16].
Reactor length (cm)
Fig. 9. Predicted moving nitrogen concentration front n = N/Nf for different elapsed
times. ReS2 catalyst.
Fig. 11 shows s(s) and n(s) as functions of s for the ReS2 catalyst.
The dashed line represents the HDS of 46DEDBT in the absence of
3ECBZ. In the presence of 3ECBZ, sulfur removal (solid curve) is se-
verely inhibited by nitrogen species. The self-inhibiting effect of
nitrogen species is also evident. When both 46DEDBT and 3ECBZ
are present, the selectivity of ReS2 for HDN is far higher than that
for HDS.
100
t = 0 (hr)
80
t = 0.67 (hr)
t = 2.68 (hr)
5.4. Interactions between HDS and HDN
60
40
20
0
t = 4.69 (hr)
t = 10.47 (hr)
This section aims to connect the 46DEDBT-3ECBZ model system
with deep HDS in the real world. Let us consider a hypothetical
middle distillate containing 1 wt.% sulfur (taking a round number).
Meeting the 10 wppm product sulfur specification means an ultra-
deep HDS level higher than 99.9% (s = 0.001). Even for low-sulfur
(e.g., prehydrotreated) feedstocks,
a reduction of from, say,
500 wppm feed sulfur to 10 wppm product sulfur means a deep
HDS level as high as 98% (s = 0.02). Polynuclear aromatics, due to
their heaviness/size [30] and high concentration, are generally
the dominant inhibitor when the HDS level is not deep (say, prod-
uct sulfur level >500 wppm). As the desulfurization gets deeper,
-20
0
0.5
1
1.5
2
2.5
3
3.5
4
Reactor length (cm)
1
Fig. 10. Percent site coverage by adsorbed nitrogen, 100qn/qm, for different elapsed
times. ReS2 catalyst.
s
5.3. ReS2 vs. sulfided CoMo/Al2O3–SiO2
n
s
For perspective, here we begin by a recapitulation of the results
on the sulfided CoMo/Al2O3–SiO2 catalyst [27]. Referring to Table 2,
the catalyst adsorbs 3ECBZ about 17 times faster (g = 17) than it
denitrogenates. It is highly selective toward the adsorption of
3ECBZ. As a result, the surface is swarmed with adsorbed nitrogen
species. Table 2 also indicates that ks and kn for unsupported ReS2
are much higher than those for CoMo/Al2O3–SiO2. ReS2’s kHDN is
higher than that of CoMo/Al2O3–SiO2 by a factor of about 19. More-
over, the density of active site qm on ReS2 is about one-third that on
CoMo/Al2O3–SiO2. This implies that the HDN surface turnover fre-
quency on ReS2 is much higher than that on CoMo/Al2O3–SiO2.
Relative to sulfided CoMo/Al2O3–SiO2, ReS2 has a higher selec-
tivity for 3ECBZ adsorption as can be seen from the p values listed
in Table 2. As an interesting aside, this result bears some resem-
blance to the binding powers of 46DEDBT and 3ECBZ on a Ru orga-
0.1
s
n
0.01
0
100
200
τ
300
400
Fig. 11. Semi-logarithmic plots of calculated
unsupported ReS2. The dashed line s is for the HDS of 46DEDBT in the absence of
3ECBZ.
s
and
n
vs. space time s for

T.C. Ho et al. / Journal of Catalysis 276 (2010) 114–122
121
Table 3
1
Equivalent pseudo-first-order gravimetric rate constants for HDN of 3ECBZ.
46DEDBT-3ECBZ
ReS2
2.85
CoMo/Al2O3
0.42
ko ꢄ 103
Units: cc liquid feed/g cat/s.
Feed II
0.1
s
Feed I
parable to that of the trajectory for Feedstock II at high HDS levels
(>80% or s < 0.2). This confirms that 46DEDBT and 3ECBZ are realis-
tic model compounds for probing HDS–HDN interactions in the
ultra-deep HDS of difficult-to-desulfurize middle distillates. For
perspective, Fig. 12 also shows the ln(s) vs. ln(n) plot for the DBT-
3ECBZ probe over the CoMo/Al2O3–SiO2 catalyst. The slope is steep
because 3ECBZ is much less inhibiting to the HDS of DBT [21].
Parity Line
DBT-3ECBZ
HDS-Selective
HDN-Selective
0.01
1E-3
0.01
0.1
1
5.5. HDN activities
n
Fig. 12. ln(s) vs. ln(n) data on two different catalysts for two different petroleum
fractions, Feedstocks I and II [5]. The two solid straight lines are calculated from
model-compound data obtained with 46DEDBT-3ECBZ and DBT-3ECBZ probes. The
dashed line represents equal selectivity for HDN and HDS.
In practice, it is the performance of the catalyst at Steady State II
that matters. Fig. 7 shows that ReS2 gives rise to an effluent liquid
containing about 7 wppm nitrogen at Steady State II. This corre-
sponds to a remarkably high HDN of 91% at a WHSV of eight, not-
withstanding a strong self-inhibition effect. By contrast, the CoMo/
Al2O3–SiO2 catalyst gives a 58% HDN at a low WHSV of 2.4 [21,27].
The corresponding 46DEDBT HDS levels for ReS2 and CoMo/Al2O3–
SiO2 are ꢃ17% and ꢃ10%, respectively. These results indicate that
ReS2 essentially behaves as a deep HDN catalyst. In what follows
we compare the HDN activities of unsupported ReS2 and sulfided
CoMo/Al2O3–SiO2 based on results shown in Table 2.
residual nitrogen heterocycles emerge as the dominant inhibitor
[15], as alluded to earlier. In many practical situations, the objec-
tive of achieving less than 10 wppm product sulfur cannot be
met unless the product nitrogen level is reduced to a sufficiently
low level. Thus, here one deals with a process that requires
super-high conversions for both HDS and HDN. This situation can
be succinctly described by the ln(s) vs. ln(n) plot over multiple dec-
ades, as illustrated in Fig. 12. This figure shows how the HDS trajec-
tory becomes increasingly influenced by nitrogen inhibition as the
reaction (or space) time increases. Here, the experimental data on
real feeds (Feedstocks I and II) are taken from Fig. 16 in Ref. [5]. If a
catalyst happens to be equally selective toward HDS and HDN, then
its heteroatom removal trajectory follows the dashed line with
p = 1 (parity line). Above the parity line is the HDN-selective re-
gion. Typically, the trajectory starts in the HDS-selective region
(shaded area), in which reactive sulfur species are desulfurized.
As the HDS level gets deeper, the trajectory becomes more and
more curved toward the HDN-selective region due to increasingly
strong organonitrogen inhibition to the HDS of hindered DBTs.
Here, partially hydrogenated nitrogen heterocycles can be more
inhibiting than their parent molecules [20]. For difficult-to-desul-
furize feedstocks (e.g., catalytic light cycle oil that contains high
levels of alkylcarbazoles), a sufficiently deep HDS may not be
achievable within the operating constraints of a commercial HDS
reactor. An example is Feedstock II shown in Fig. 12. This feedstock
has a lower API gravity than Feedstock I. Both feeds have approxi-
mately the same level of total nitrogen, 705 wppm for Feed I and
740 ppm for Feed II. However, the nitrogen heterocycles in Feed I
were found to be dominated by six-membered species such as
quinolines, whereas those in Feed II are predominantly five-mem-
bered species such as carbazoles [5].
Integrating Eq. (10) from n = 0 to n = 1 gives n at the reactor
outlet
g½1 ꢀ nð
s
Þꢈ ꢀ ln nð
s
Þ ¼ knqms
ð16Þ
Since g ꢅ 1 for both catalysts, the first term on the left-hand side
dominates the HDN behavior over a wide range of HDN levels. Eq.
(16) can then be approximately written as
ꢂ
ꢃ
knqm
g
nð
sÞ ꢁ 1 ꢀ
s
¼ 1 ꢀ ko
s
ð17Þ
Thus, ko is viewed as an approximate overall HDN rate constant de-
fined as
knqm kHDNqm
ko ꢇ
¼
ð18Þ
g
Nf
Here, ko can also be regarded as an equivalent pseudo-first-order
gravimetric rate constant. The concept of equivalent first-order rate
constant was discussed in Ref. [5]. Note that ko is inversely propor-
tional to the feed nitrogen concentration Nf, which has a direct bear-
ing on real-feed HDN [20].
Based on the parameter listed in Table 2, the ko values for
unsupported ReS2 and sulfided CoMo/Al2O3–SiO2 can be calculated
and are listed in Table 3.
Since we neglect the second term on the left-hand side of Eq.
(16), the activity advantage of the unsupported ReS2 catalyst
shown in Table 3 is somewhat conservative. What we can say is
that ReS2 gives about a sevenfold HDN activity advantage over
the sulfided CoMo/Al2O3–SiO2 catalyst.
Now return to the model-compound experiments using the
46DEDBT-3ECBZ probe on unsupported ReS2. Eq. (14) can be
rewritten as
ln½sðs
Þꢈ ¼ p ln½nð
s
Þꢈ
ð15Þ
Thus, for the model-compound system, s(
s
) and n(
s) are linearly re-
6. Concluding remarks
lated to each other on a double logarithmic plot of ln(s) vs. ln(n),
with a slope of p. As Fig. 12 shows, the ln(s) vs. ln(n) plot is a rather
flat straight line due to the small p value of 0.2. Although not shown,
the same plot for the CoMo/Al2O3–SiO2 catalyst is also rather flat
because of a small p of 0.25. With either catalyst, the slope is com-
Using 46DEDBT and 3ECBZ as model compounds provides
useful information on the HDS–HDN interactions in the deep
HDS of petroleum middle distillates. The unsupported ReS2 catalyst
prepared from (NH4)4(Re4S22)ꢂ2H2O has an unusually strong

122
T.C. Ho et al. / Journal of Catalysis 276 (2010) 114–122
hydrogenation function. As a result, it gives about a 9-fold activity
advantage over a commercial sulfided CoMo/Al2O3–SiO2 catalyst
for desulfurization of 46DEDBT in the absence of 3ECBZ. But it suf-
fers a disproportionately greater loss of HDS activity than CoMo/
Al2O3–SiO2 in the presence of 3ECBZ. In other words, ReS2 is less
resilient to 3ECBZ inhibition than CoMo/Al2O3–SiO2. The latter
Acknowledgment
Q. Shen from the Department of Chemical Engineering at the
University of Texas at Austin acknowledges a summer internship
supported by ExxonMobil Research and Engineering Company.
has a higher selectivity for hydrogenolysis (lower cDEDBT). A recent
References
study has shown that hydrogenolysis sites are more capable of
mitigating the impact of organonitrogen inhibition than hydroge-
nation sites [16].
[1] H. Topsøe, B.S. Clausen, F.E. Massoth, Hydrotreating Catalysis, Springer-Verlag,
1996.
[2] B.C. Gates, H. Topsøe, Polyhedron 16 (1997) 3213.
[3] N. Hermann, M. Brorson, H. Topsøe, Catal. Lett. 65 (2000) 169.
[4] T.C. Ho, Catal. Today 98 (2004) 3.
The unsupported ReS2 catalyst shows about a sevenfold HDN
activity advantage over the sulfided CoMo/Al2O3–SiO2 catalyst. It
is also much more selective for HDN in simultaneous HDS and
HDN experiments. This can be attributed to a high adsorption
affinity for organonitrogen species coupled with a fast C–N bond
cleavage rate. The high HDN activity results from a greater intrinsic
activity per active site, rather than a higher site density. The nature
of the active sites on ReS2 and on CoMo/Al2O3–SiO2 should be very
different, even though ReS2 and MoS2 are structurally similar in
certain respects. It is of great interest to know whether metallic
brim sites exist on ReS2.
A highly hydrogenative ReS2 should be regarded as a catalyst
primarily for deep HDN. It is relevant to point out that the nitrogen
content of refinery hydroprocessing feedstocks has been increasing
over the years. The objectives of two key refinery conversion pro-
cesses, fluid catalytic cracking and hydrocracking, cannot be
achieved unless the feed nitrogen level is reduced to a sufficiently
low level. Developments of new or improved HDN catalysts and
processes have been and will continue to be important in the years
ahead [31].
As for ultra-deep HDS applications, ReS2 should be used with a
hydrogenolysis-selective catalyst in a properly configured stacked
bed to give an activity synergism. This catalyst-stacking strategy
has been demonstrated with a highly hydrogenative Ni0.5Mn0.5Mo
sulfide catalyst [4]. Also, the low nitrogen tolerance of the ReS2 cat-
alyst may be improved through incorporation of promoter metals
such as cobalt and/or nickel. Co- or Ni-promoted Re sulfide cata-
lysts, in unsupported or supported form, are more HDS-selective
than their unpromoted counterparts, suggesting that they may
be more hydrogenolysis selective [32,33].
[5] T.C. Ho, Catal. Today 130 (2008) 206.
[6] T.C. Ho, Catal. Lett. 89 (2003) 21.
[7] S. Eijsbouts, S.W. Mayo, K. Fujita, Appl. Catal. A: Gen. 322 (2007) 58.
[8] J.V. Lauritsen, M. Nyberg, J.K. Nørskov, B.S. Clausen, H. Topsøe, E. Lægsgaard, F.
Besenbacher, J. Catal. 224 (2004) 94.
[9] V. Petkov, J.L. Billinge, P. Larson, S.D. Mahanti, T. Vogt, K.K. Rangan, M.G.
Kanatzidis, Phys. Rev. B 65 (2002) 092105.
[10] H.H. Murray, S.P. Kelty, R.R. Chianelli, C.S. Day, Inorg. Chem. 33 (1994) 4418.
[11] H.J. Lamfers, A. Meetsma, G.A. Wiegers, J.L. de Boer, J. Alloys Compd. 241
(1996) 34.
[12] N. Escalona, M. Vrinat, D. Laurenti, F.J. Gil Llambias, Appl. Catal. A: Gen. 322
(2007) 113.
[13] C.J.H. Jacobsen, E. TÖrnqvist, H. Topsøe, Catal. Lett. 63 (1999) 179.
[14] M. Breysse, E. Furimsky, S. Kasztelan, M. Lacroix, G. Perot, Catal. Rev. 44 (2002)
651.
[15] T.C. Ho, G.E. Markley, Appl. Catal. A: Gen. 267 (2004) 245.
[16] T.C. Ho, Q. Liang, J. Catal. 269 (2010) 291.
[17] I. Ignatiadis, M. Kuroki, P.J. Arpino, J. Chromatogr. 366 (1986) 251.
[18] S.D. Sumbogo Murti, H. Yang, K.H. Choi, Y. Korai, I. Mochida, Appl. Catal. A:
Gen. 252 (2003) 331.
[19] W. Kanda, I. Siu, J. Adjaye, A.E. Nelson, M.R. Gray, Energy Fuels 18 (2004) 539.
[20] T.C. Ho, Appl. Catal. A: Gen. 378 (2010) 52.
[21] T.C. Ho, J. Catal. 219 (2003) 442.
[22] A. Müller, E. Krickemeyer, H. Bögge, Z. Anorg. Allg. Chem. 554 (1987) 61.
[23] E. Furimsky, F.E. Massoth, Catal. Today 52 (1999) 381.
[24] T.C. Ho, J. Sobel, Catal. Lett. 99 (2005) 109.
[25] P. Steiner, E.A. Blekkan, Fuel Proc. Technol. 79 (2002) 1.
[26] E. Furimsky, F.E. Massoth, Catal. Rev. 47 (2005) 297.
[27] T.C. Ho, D. Nguyen, J. Catal. 222 (2004) 450.
[28] T.C. Ho, D. Nguyen, Chem. Eng. Commun. 193 (2006) 460.
[29] M.G. Choi, T.C. Ho, R.J. Angelici, Organometallics 27 (2008) 1098.
[30] S. Korre, M. Neurock, M.T. Klein, R.J. Quann, Chem. Eng. Sci. 49 (1994) 4191.
[31] E. Furimsky, F.E. Massoth, Catal. Rev. Sci. Eng. 47 (2005) 297–489.
[32] N. Escalona, J. Ojeda, P. Baeza, R. Garcia, J.M. Palacios, J.L.G. Fierro, A. Lopez
Agudo, F.J. Gil-Liambias, Appl. Catal. A: Gen. 287 (2005) 47.
[33] J. Ojeda, N. Escalona, J.M. Palacios, M. Yates, J.L.G. Fierro, A. Lopez Agudo, F.J.
Gil-Liambias, Appl. Catal. A: Gen. 350 (2008) 6.