Activation and isomerization of hydrocarbons over WO3/ZrO2 catalysts. II. Influence of tungsten loading on catalytic activity: Mechanistic studies and correlation with surface reducibility and tungsten surface species

Article abstract of DOI:10.1016/j.jcat.2008.02.026

We studied the correlation among the catalytic behavior of WO₃/ZrO₂ samples toward unsaturated and saturated hydrocarbons transformation, tungsten surface species oxidation states, and the crystallographic structure of the zirconia support. Different tungsten-loaded catalysts were studied, from 9 wt% (near-monolayer coverage) to 30 wt%. The resulting WO₃/ZrO₂ materials were obtained by impregnation of a tungsten salt on either a commercially available monoclinic zirconia or an amorphous hydroxide, ZrO_x(OH)_4-2x, followed by a calcination step (according to the Hino and Arata procedure), leading to a tetragonal structure. In contrast to previous works, here we demonstrate that the crystallographic structure of zirconia has no influence on catalytic properties. Correlations with XPS analyses revealed two aspects of catalytic behavior that depend strongly on the catalyst reducibility and thus on the W surface species oxidation states. First, on hardly reducible (tungsten loadings <15 wt%) or slightly reduced (below 423 K) surfaces, a purely acidic monomolecular mechanism for both isomerization (largely predominant) and cracking reactions, associated with W⁶⁺ and W⁵⁺ surface species, was demonstrated. Second, on easily reducible (tungsten loadings >15 wt%) or deeply reduced (over 723 K) surfaces, a bifunctional mechanism associating dehydrogenating/hydrogenating properties occurring on metallic tungsten and acidic isomerization and cracking on W⁵⁺ and W⁶⁺ surface species was observed. However, in this last case, we could not exclude the participation of a purely metallic isomerization mechanism occurring through σ-alkyl adsorbed species on the β-W metallic phase. A more pronounced reduction then led to an increase in the extensive hydrogenolysis mechanism, causing catalyst deactivation.

Full text of DOI:10.1016/j.jcat.2008.02.026

Journal of Catalysis 256 (2008) 159–171
www.elsevier.com/locate/jcat
Activation and isomerization of hydrocarbons over WO3/ZrO2 catalysts
II. Influence of tungsten loading on catalytic activity: Mechanistic studies
and correlation with surface reducibility and tungsten surface species
François Di Gregorio, Nicolas Keller, Valérie Keller ∗
LMSPC, Laboratoire des Matériaux, Surfaces et Procédés pour la Catalyse and ELCASS, European Laboratory for Catalysis and Surfaces Sciences,
25, rue Becquerel 67087, Strasbourg Cedex 2, France
Received 30 October 2007; revised 27 December 2007; accepted 27 February 2008
Abstract
We studied the correlation among the catalytic behavior of WO /ZrO samples toward unsaturated and saturated hydrocarbons transformation,
3
2
tungsten surface species oxidation states, and the crystallographic structure of the zirconia support. Different tungsten-loaded catalysts were
studied, from 9 wt% (near-monolayer coverage) to 30 wt%. The resulting WO /ZrO materials were obtained by impregnation of a tungsten salt
3
2
on either a commercially available monoclinic zirconia or an amorphous hydroxide, ZrO (OH)
, followed by a calcination step (according
x
4−2x
to the Hino and Arata procedure), leading to a tetragonal structure. In contrast to previous works, here we demonstrate that the crystallographic
structure of zirconia has no influence on catalytic properties. Correlations with XPS analyses revealed two aspects of catalytic behavior that depend
strongly on the catalyst reducibility and thus on the W surface species oxidation states. First, on hardly reducible (tungsten loadings <15 wt%) or
slightly reduced (below 423 K) surfaces, a purely acidic monomolecular mechanism for both isomerization (largely predominant) and cracking
6+
5+
reactions, associated with W and W surface species, was demonstrated. Second, on easily reducible (tungsten loadings >15 wt%) or deeply
reduced (over 723 K) surfaces, a bifunctional mechanism associating dehydrogenating/hydrogenating properties occurring on metallic tungsten
5+
6+
and acidic isomerization and cracking on W and W surface species was observed. However, in this last case, we could not exclude the
participation of a purely metallic isomerization mechanism occurring through σ-alkyl adsorbed species on the β-W metallic phase. A more
pronounced reduction then led to an increase in the extensive hydrogenolysis mechanism, causing catalyst deactivation.
© 2008 Elsevier Inc. All rights reserved.
Keywords: Tungsten oxide-supported zirconia; Isomerization; Mechanistic studies; Acidic mechanism; Metallic mechanism; Difunctional mechanism; Correlation
with XPS; Surface reducibility
1. Introduction
alytic performance, but they generated many problems, includ-
ing corrosion and storage issues. The need for materials in
agreement with environmental concerns and regulations has led
to the preparation of solid acid catalysts with particular ca-
pacities, including (i) high conversion and selectivity toward
isomerization, (ii) stability under reaction, (iii) low sensitivity
to poisoning, and (iii) easy regeneration. Metal oxide catalysts
using oxoanions appear to be suitable candidates for the izomer-
ization of alkanes heavier than C4. Zirconia modified by sulfate
anions, first introduced by Hino and Arata [17], was widely
studied because of its high activity and very interesting isomer-
ization selectivity. But sulfate loss during thermal treatment, as
well as deactivation of the material during the catalytic reac-
tion, brought Hino and Arata [18] to take an interest in tungsten
Catalytic transformations of hydrocarbons, such as alkyla-
tion and cracking, are very important in the field of petro-
leum chemistry. An isomerization process involving the skele-
tal rearrangement of the starting hydrocarbon molecule to ob-
tain high-octane isoparaffins has been investigated [1–16]. This
process, which requires acidic properties, was first investigated
using liquid-phase systems such as HF, H2SO4 superacids, and
halogen-containing solids. Those catalysts exhibited good cat-
*
Corresponding author. Fax: +33 3 9 0 24 27 61.
E-mail address: vkeller@chimie.u-strasbg.fr (V. Keller).
0021-9517/$ – see front matter © 2008 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2008.02.026

160
F. Di Gregorio et al. / Journal of Catalysis 256 (2008) 159–171
oxide-supported zirconia. They concluded that superacid sites
were created by the interaction of WO3 with ZrO2 when the
latter oxide was crystallized from an amorphous form into a
tetragonal form.
posited either directly on a commercial monoclinic ZrO2 zirco-
nium oxide (Johnson Matthey) or on an amorphous hydroxide,
ZrOx(OH)4−2x, prepared according to the procedure described
by Hino and Arata [18]. Whatever the preparation method, im-
pregnation with the ammonium salt was performed in large
excess of water, followed by a slow evaporation of the solvent
at 373 K and drying overnight at 393 K.
WO3/ZrO2 catalysts are usually prepared by wet impregna-
tion of zirconium hydroxide with an appropriate salt solution. It
was found that catalytic activity is strongly dependent on both
the precursor loading and the calcination temperature, the latter
allowing expulsion of the tungstate from the bulk of the aerogel.
Maximum catalytic activity was reported to occur for tungsten
loadings close to theoretical monolayer coverage of approxi-
mately 6.25 W-atoms/nm2 [19] calculated from the spreading
of a single tungsten oxide layer on the support; however, no
activity has been reported for catalysts prepared by direct im-
pregnation of crystallized ZrO2 as obtained when salts were
deposited directly on alumina by an equilibrium adsorption
technique.
The first part of this report [20] dealt mainly with prepa-
ration and characterization methods (in particular, XPS and
DRX measurements) of a series of WO3/ZrO2 catalysts in
which tungsten contents varied from 9 to 30% (-W wt%), ob-
tained by impregnation of a monoclinic zirconium oxide, ZrO2.
Spectroscopic studies revealed that modifications of the cata-
lyst surface occurring during reduction treatments were more
or less pronounced and were correlated with tungsten loading.
These surface modifications expressed changes in tungsten ox-
idation state as well as in atomic surface ratio. Consequently, it
seems obvious that the catalytic properties of WO3/ZrO2 sys-
tems strongly depend on the degree of reduction of tungsten
surface species. To determine the active sites, it seems neces-
sary to correlate the surface state and the structure obtained by
the various surface or volume characterization techniques with
the results of catalytic tests. Although the reactivity of alkanes
on WO3/ZrO2 systems has been extensively investigated, to the
best of our knowledge, a direct correlation with the various ox-
idation states of tungsten determined by X-ray photoelectron
spectroscopy (XPS) has not yet been reported.
2.1.1. Direct impregnation on commercial monoclinic ZrO2
The direct impregnation of (NH4)10W12O41·5H2O on a
commercial monoclinic ZrO2 was achieved by calcination of
the sample for 1 h in air at 623 K by applying a 5 ◦K/min ramp.
The impregnation of a theoretical monolayer of WO3 was done
supposing that one WO3 molecule occupied 23 Å2 [1]. Cata-
lysts with various initial WO3 contents were prepared by vary-
ing the concentration of the precursor salt to obtain initial tung-
sten weight percentages of 6 (equivalent of one monolayer),
9, 15, 20, and 30%. These catalysts samples were designated
WZAm, WZA9, WZA15, WZA20, and WZA30, respectively.
2.1.2. Impregnation on amorphous hydroxide, ZrOx(OH)4−2x
This synthesis proceeded in two successive steps: (i) pre-
cipitation of an amorphous phase of zirconium hydroxide,
ZrOx(OH)4−2x, obtained from the hydrolysis of ZrOCl2 using
the controlled addition of NH4OH (14,8 N) to maintain a con-
stant pH of 10, with the gel washed and then dried overnight
at 293 K, and (ii) wet impregnation of the dried amorphous
phase with a solution of (NH4)10W12O41·5H2O precursor. The
postsynthesis treatment involved drying the solid overnight at
293 K, followed by calcination for 1 h in air at 1073 K by ap-
plying a 10 ◦K/min ramp. Only one tungsten content sample
was prepared in this manner, yielding a 9 wt% W WO3/ZrO2
sample designated WZB9.
2.2. Characterization techniques
All of the samples were characterized by XPS, XRD, BET
specific surface area, and temperature-programmed reduction
(TPR) measurements, as described in detail elsewhere [20].
We also investigated the catalytic behavior of a series of
WO3/ZrO2 catalysts toward 2-methylpentane and its associated
olefin (4-methyl-1-pentene) reactivity. This reactivity was stud-
ied as a function of tungsten loading and of the preparation
method—direct impregnation of commercial monoclinic ZrO2,
or impregnation of an amorphous hydroxide ZrOx(OH)4−2x
(obtained from the hydrolysis of ZrOCl2)—depending on both
the prereduction and reaction temperatures. We also conducted
a detailed analysis of cracking and reaction product distribution
as a function of reduction duration and carbon chain length,
and investigated different reaction mechanisms and their possi-
ble correlations with tungsten surface oxidation states.
2.3. Apparatus and procedures
Experiments were performed in a glass flow reactor as
described previously [21], equipped with a purification trap
to remove oxygen traces. Reactions were carried out under
pressure slightly higher than the atmospheric pressure with a
30 cm3/min flow rate of hydrogen. Five microliters of reactant
were injected via a septum into a U-tube maintained at a con-
stant vapor pressure of hydrocarbon (around 6.5 Torr). Reaction
products were analyzed by online gas chromatography (Var-
ian 3300) using a dimethylsulfoxane column (50 cm long and
0.3 mm internal diameter) (Chrompack CP-SIL-5CB), with a
flame ionization detector and a flux of helium as the gas carrier.
Reaction products could be directed through a hydrogenator
(Pt Adams) in which molecules were hydrogenated, facilitat-
ing chromatographic analysis when working with olefins.
2. Experimental
2.1. Catalyst preparation
Supported catalysts samples were prepared by wet im-
pregnation with an ammonium metatungstate pentahydrate
(NH4)10W12O41·5H2O salt, at pH values close to 5, and de-

F. Di Gregorio et al. / Journal of Catalysis 256 (2008) 159–171
161
2.4. Probe molecules
and 3MP/nC6 ratio on the WZAm sample when starting from
4M1Pen reacting at 473 K. The catalyst was immediately active
for olefin conversion (57%) with high isomerization selectiv-
ity (85%). The isomerization ratio value of 3MP/nC6 was close
to 7. These parameters were quite stable with time on stream
under hydrogen at 473 K. A slight increase in isomerization se-
lectivity occurred, at the expense of cracking products, to reach
stable selectivity of around 93%.
The main hydrocarbons used for catalytic tests were 4-
methyl-1-pentene (4M1Pen) and 2-methylpentane (2MP). Com-
paring these two reactants provided information about the
acidic and metallic properties of bifunctional catalysts. We also
investigated the influence of carbon chain length by studying
the reactivity of the linear alkenes 1-hexene (nC6=), 1-heptene
(nC7=), 1-octene (nC8=), and 1-nonene (nC9=).
Fig. 1b shows the evolution of the same parameters shown in
Fig. 1a (αt , Sisom, Scrack, 3MP/nC6) plus products issued from
2.5. Calculations
carbon chain enlargement, S>C , as a function of prereduction
6
temperature on WZAm sample for 4M1Pen reaction occurring
at 473 K. It must be mentioned than the values reported for the
reduction temperatures of 473 and 623 K were obtained after
reduction of the catalyst for 24 h under hydrogen. In contrast,
the values for reduction at 723 and 823 K are associated with
1 h of catalyst prereduction. From Fig. 1b, it can be seen that
the olefin reactivity certainly revealed a stable catalyst surface
under hydrogen after reduction at 473 and 623 K, but reduction
at 723 K modified the catalytic properties of the WZAm system.
Indeed, a significant drop in conversion—from 63 to 38%—
occurred after reduction at 723 K. This decreased conversion
was accompanied by an increase in isomerization selectivity to
100%. A more severe reduction at 823 K led to quasi-inactivity
of the system toward conversion of 4M1Pen at 473 K.
The total conversion, αt (%), is defined as the percentage
of reactant transformed. The selectivity, S (%), of a product is
the percentage of this product among all the products formed.
The total reaction rate, r (mol/s g), is defined as the number
of moles of reactant transformed per second per gram of cat-
alyst. Starting from 2MP or from 4M1Pen, the 3MP/nC6 (3-
methylpentane/nC6) ratio can be representative of the acidity of
the catalyst, because under acidic reaction conditions, a skeletal
rearrangement leads preferentially to the most stable carboca-
tion yielding 3MP.
3. Catalytic results
3.1. Reactivity of 2MP and 4M1Pen
Like the WZAm catalyst, the WZA9, WZB9, WZA15,
and WZA30 catalysts were immediately active and stable for
4M1Pen isomerization at 473 K without previous reduction and
demonstrated high selectivity in isomers (>78%); thus, results
of these catalytic tests as a function of reduction duration at
473 K are not given here. In the same manner, results obtained
after reduction of the WZA9 catalyst at 623 K (Fig. 2a) are quite
similar to those obtained on the WZAm catalyst submitted to
the same treatment; however, unlike for the WZAm catalyst,
reduction at 723 K did not modify product selectivity to any
great degree. A drop in conversion from 58 to 22% was only
when reduction temperature was increased from 723 to 823 K.
While isomerization selectivity increased (>97%), selectivity
From a thermodynamic standpoint, isomerization reactions
are slightly exothermic and are favored at low temperature. The
catalysts used for this reaction also must work at the lowest
temperature. Based on these considerations, the main reaction
temperatures investigated were 473 and 623 K. Different pre-
reduction temperatures also were investigated: 473, 623, 723,
and 823 K. It is important to notice that whatever the prereduc-
tion duration at 473 and 623 K, WZAm, WZA9, and WZB9
catalysts were not active for 2MP transformation.
Fig. 1a shows the evolution of total conversion (αt ), iso-
merization (Sisom) and cracking (Scrack) product selectivities,
(a)
(b)
Fig. 1. (a) Evolution of total conversion (%) (F), of total isomerization (1) and total cracking (P) selectivities (%) and of 3MP/nC (×) ratio as a function of
6
reduction duration (min) at 473 K for 4M1Pen reaction at 473 K on WZAm catalyst. (b) Evolution of total conversion (%), of total isomerization and total cracking
selectivities (%), of 3MP/nC ratio and of enlargement reaction products selectivity (%) as a function of reduction temperature for 4M1Pen reaction at 473 K.
6

162
F. Di Gregorio et al. / Journal of Catalysis 256 (2008) 159–171
(a)
(b)
Fig. 2. (a) Evolution of total conversion (%), of total isomerization and total cracking selectivities (%), of 3MP/nC ratio and of enlargement reaction product
6
selectivity (%) as a function of reduction temperature for 4M1Pen reaction at 473 K on WZA9. (b) Evolution of total conversion (%), of total isomerization and total
cracking selectivities (%), of and 3MP/nC ratio and of enlargement reaction product selectivity (%) as a function of reduction temperature for 4M1Pen reaction at
6
473 K on WZB9.
Fig. 3. Evolution of total conversion (%), of total isomerization and total crack-
Fig. 4. Evolution of total conversion (%), of total isomerization and total crack-
ing selectivities (%), of 3MP/nC ratio and of enlargement reaction product
6
selectivity (%) as a function of reduction temperature for 4M1Pen reaction at
473 K on WZA15.
ing selectivities (%), of 3MP/nC ratio and of enlargement reaction products
selectivity (%) as a function of reduction temperature for 4M1Pen reaction at
473 K on WZA20.
6
in cracking products and products emerging from carbon chain
enlargement reactions diminished to quasi-zero values. Fig. 2b
shows the conversions and selectivities obtained on WZB9
catalyst as a function of reduction temperature. The isomeriza-
tion selectivities obtained were lower than those found for the
WZA9 catalyst regardless of the reduction temperature. More-
over, the loss of activity after reducing treatments at 823 K was
much less pronounced on the catalyst prepared according to the
B way.
Table 1
Conversion α (%), total isomerization S
tivities (%), 3MP/nC ratio and enlargement reaction products selectivity S
(%) for 4M1Pen reaction at 473 K on WZA20 and WZA30 samples progres-
sively reduced at 723 K
and total cracking S
selec-
T
isom
crack
6
>C
6
Catalyst
Time
(min)
α
S
S
S
3MP/nC
6
T
isom
(%)
crack
>C
(%)
(%)
(%) 6
WZA20
60
59.0
32.5
22.5
66.5
95.0
98.5
22.0
2.5
11.5
2.5
1.9
16.4
44.8
345 (∼6 h)
1380 (∼24 h)
1.0
0.5
As for the WZA9 and WZB9 catalysts, a long reduction
treatment (24 h) at 473 K (Fig. 3) did not modify the cat-
alytic properties of the WZA15 system for reactions occurring
at 473 K. Furthermore, reductions at 623 K (24 h) and at 723 K
(1 h) led to the same evolution in properties as for the WZA9
catalyst. A more severe reduction at 823 K (1 h) yielded a
decrease in conversion, accompanied by an increase in isomer-
ization of up to 95%. Catalytic tests performed at 623 K when
starting from the saturated 2MP molecule revealed activity of
the WZA15 catalyst prereduced at 723 K. Conversion of 2MP
increased progressively from 12 to 23% in the course of reduc-
tion at 723 K, reaching a stable value after 23 h (detailed results
not shown). This reactivity of 2MP suggests the presence of cat-
alytic sites that are able to activate the saturated hydrocarbon.
WZA30
60
42.0
44.0
5.0
91.5
87.0
98.5
4.5
6.5
1.5
4.0
6.5
0
6.6
7.3
111.3
345 (∼6 h)
1020 (∼17 h)
Fig. 4 illustrates the evolution of conversion and product se-
lectivity of 4M1Pen reacting at 473 K as a function of reduction
temperature on the WZA20 catalyst. The figure does not show
the evolution of the same parameters observed on the WZA30
sample, because those parameters exhibited similar tendencies
in the two catalysts. Both catalysts (WZA20 and WZA30) were
active, stable, and selective toward isomerization of 4M1Pen
when prereduced at 473 K (24 h) and 623 K (24 h), even with
the diminished isomerization selectivity seen at high prereduc-

F. Di Gregorio et al. / Journal of Catalysis 256 (2008) 159–171
163
Fig. 5. Evolution of the catalytic properties obtained on WO /ZrO samples as a function of tungsten loadings and reducing treatments.
3
2
Table 2
Conversion α (%), total isomerization S
emerging from the two representative WZAm and WZA15
samples in detail. Due to the previously mentioned catalytic be-
havior, we studied the WZAm catalyst for low-temperature re-
duction treatments (at 473 and 623 K), and studied the WZA15
catalyst for reduction at temperatures of 723 K and above. We
investigated the catalytic properties of the WZAm system by
varying the carbon chain length from 1-hexene to 1-nonene. For
each molecule, a systematic analysis of the product distribution
and the isomerization and cracking activation energies was con-
ducted.
and total cracking S selec-
crack
T
isom
tivities (%), 3MP/nC ratio and enlargement reaction products selectivity S
6
>C
6
(%) as a function of reduction duration at 723 K for 2MP reaction at 723 K on
WZA20 and WZA30 samples
Catalyst
Time
(min)
α
S
S
S
3MP/nC
6
T
isom
crack
>C
(%)
(%)
(%)
(%) 6
WZA20
345 (∼6 h)
14.0
22.0
86.5
73.5
13.5
26.5
0
0
4.1
8.8
1380 (∼23 h)
WZA30
205
1020 (∼17 h)
16.5
10.0
87.0
42.5
13.0
57.5
0
0
6.9
–
3.2.1. Surfaces slightly reduced or hardly reducible:
WO3/ZrO2 catalysts with tungsten loadings <15 wt%
3.2.1.1. Conversion of 4M1Pen The distribution of reaction
products of 4M1Pen carried out at 473 K on the WZAm sample
prereduced at 473 K is shown in Table 3a. Several points can be
highlighted:
3MP was the most abundant isomer product obtained.
The 3MP/nC6 ratio was generally >1, indicating that mono-
branched isomer formation is favored over linear isomer forma-
tion. This behavior is characteristic of an acid catalysis. 3MP
and nC6 were not found in the thermodynamic ratio (0.96) at
473 K, suggesting that skeletal rearrangement reactions at the
surface were not equilibrated and that desorption was not the
rate-limiting step.
tion temperature. Table 1 presents the conversions and product
selectivities obtained on the WZA20 and WZA30 systems re-
duced at 473 K; the table clearly indicates that their activities
toward 4M1Pen reacting at 473 K were not stable and decreased
over the course of reduction. This decrease was more pro-
nounced on the WZA30 catalyst. Like the WZA15 catalyst, the
WZA20 and WZA30 catalysts were active for 2MP isomeriza-
tion at 623 K when activated at 723 K (Table 2). With increasing
duration of reduction at 723 K, the WZA20 catalyst exhibited
an increase in 2MP conversion from 14 to 22%, whereas, the
WZA30 catalyst showed a decrease in conversion accompanied
by a considerable decrease in isomerization selectivity.
Fig. 5 summarizes the evolution of catalytic properties of the
different WO3/ZrO2 samples as a function of W loading and of
reduction treatment. The WO3/ZrO2 systems can be classified
into two categories of catalysts: (i) those with tungsten loading
<15 wt% and (ii) those with tungsten loading >15 wt%. In the
first category, the catalysts are active for olefin conversion only
at 473 K, whereas the catalysts in the second category exhibit
conversion of the saturated 2MP hydrocarbon at 623 K after
activation under hydrogen at 723 K.
The main cracking products were isobutane, iC4, and isopen-
tane, iC5.
Molecules with higher molecular weight than the starting
molecule were formed in nonnegligible proportions. Among
these, methylhexanes, such as 2-methylhexane and 3-methyl-
hexane, were predominant. Among the dimethylpentanes,
23DMP was the most abundant. Molecules with eight carbon
atoms were formed at trace levels.
3.2. Product distributions
3.2.1.2. Conversion of linear olefins: Influence of chain length
Figs. 6a, 6b, and 6c show the results of catalytic tests per-
formed using pulses of 1-hexene (nC6=), 1-heptene (nC7=)
Based on the two categories of samples deduced from Fig. 5,
herein we discuss only the evolution of product distributions

164
F. Di Gregorio et al. / Journal of Catalysis 256 (2008) 159–171
Table 3a
Distribution of reaction products for 4MlPen and nC = reactions at 473 K on WZAm catalyst progressively reduced at 473 K
6
Probe
t
α
S
S
S
Distribution of reaction products
3MP/nC
6
r
T
isom
crack
>C
6
molecule
(min)
(%)
(%)
(%)
(nmol/g/s)
(%)
C
(%)
iC nC iC nC
23DMB
(%)
3MP
(%)
nC
(%)
>C
6
(%)
3
4
4
5
5
6
(%)
(%)
(%)
(%)
4MlPen
1-Hexen
5
445
57.0
47.0
87.0
91.0
7.0
4.0
6.0
5.0
0.3
0.2
2.6
1.5
0.2
0.2
3.7
2.3
0.2
0.1
14
7.1
64.2
79.4
8.9
4.6
5.7
4.6
7.2
17.1
1.5
1.1
5
180
390
40.0
24.5
22.5
87.0
93.5
95.5
7.5
4.5
4.5
5
2
0
0.6
0.2
0.1
2.8
1.9
1.2
0.3
0.1
0
3.5
2.3
1.6
0.3
0.1
0
7.5
5.7
5.5
41.8
45.8
46.5
37.7
41.9
43.2
5
2
2.9
1.1
1.09
1.07
0.22
0.12
0.11
C : propane; iC : isobutane, nC : n-butane, iC : isopentane, nC : n-pentane, 23DMB: 2,3-dimethylbutane.
3
4
4
5
5
Table 3b
Distribution of reaction products for nC = reaction at 473 K on WZAm catalyst progressively reduced at 473 K
7
t
α
S
S
Distribution of reaction products (%)
r
T
isom
crack
(min)
(%)
(%)
(%)
(nmol/g/s)
C
2
C
3
C
4
C
5
C
6
22DMP 24DMP 223TMB 33DMP 2MH 23DMP 3MH 3EP
10
218
1500
74.0 89.0
71.5 93.0
57.0 98.5
11.0
7.0
1.5
0.3 1.2 3.9 2.8 2.8
1
7.9
8.1
7.3
0.3
0.3
0
0.1
0.2
0
29.9
30.2
31.9
12.7
13.3
12.4
34.5
37.3
42.5
2.7
2.8
3.6
0.42
0.39
0.26
0
0
0.3 2.2 1.9 1.6 1.1
0.4 0.9 0.1
0
1
C : ethane, C : total butanes, C : total pentanes, C : total hexanes, 22DMP: 2,2-dimethylpentane, 24DMP: 2,4-dimethylpentane, 223TMB: 2,2,3-trimethylbutane,
2
4
5
6
33DMP: 3,3-dimethylpentane, 2MH: 2-methylhexane, 23DMP: 2,3-dimethylpentane, 3MH: 3-methylhexane, 3EP: 3-ethylpentane.
Table 3c
Distribution of reaction products for nC = reaction at 473 K on WZAm catalyst progressively reduced at 473 K
9
t
α
S
S
Distribution of reaction products (%)
r
T
isom
crack
(min)
(%)
(%)
(%)
(nmol/g/s)
*
C
3
C
4
C
5
C
6
C
7
C
8
DM
4MC8
2MC8
3MC8
16
190
500
61.0
48.5
33.0
93.5
98.0
99.0
5.5
2.0
1.0
0.1
0
0
2
1
0.3
2
1
0.4
0.7
0
0
0.6
0.3
0.1
1.2
0
0.1
23.2
19
12.8
8.3
6.5
7.3
36.2
42.8
46.1
25.8
29.5
32.9
0.38
0.27
0.16
*
DM : dimethylpentane isomers, 4MC8: 4-methyloctane, 2MC8: 2-methyloctane, 3MC8: 3-methyloctane.
Table 4
were performed under atmospheric hydrogen pressure and at
progressively decreased temperature. Global activation energy
was determined by calculating the slope of log(r) = f (1/T )
function. The results for each hydrocarbon are gathered in
Table 4. It seems that the apparent activation energy depended
on both the branching and the carbon chain length. This ac-
tivation energy diminished with expanding molecule branch-
ing and increased as the carbon chain length went from C6 to
C8. The lowest activation energy was associated with 4M1Pen
(∼38 kJ/mol); the highest, with nC9= (109 kJ/mol). More-
over, it seems that for linear molecules with more than eight
carbon atoms, the activation energy was modified only slightly.
Whatever the molecule, the isomerization activation energy was
lower than the cracking energy. Partial order toward hydro-
carbons (not detailed here), deduced from the reaction rate at
different hydrocarbon concentrations considering the H2 car-
rier gas and reactant as constants, went from 0.76 to 0.97 for
nC6=, nC7=, nC8=, and nC9=.
Global, isomerization and cracking apparent activation energies of different hy-
drocarbons on the WZAm catalyst reduced at 473 K
Starting
molecule
Apparent activation energy (kJ/mol)
Global
Isomerization
Cracking
4M1Pen
38
82
35
74
57
144
102
190
219
C =
6
C =
91
82
7
C =
108
109
108
107
8
C =
9
and 1-nonene (nC9=), respectively, at 473 K. The results ob-
tained with 1-octene are similar to those obtained with nC9=
and thus are not shown here. Table 3 gives the product dis-
tributions emerging from nC6=, nC7=, and nC9=. It can be
seen that (i) the linear olefins reacted more slowly and had
lower conversions than 4M1Pen, (ii) isomers were predomi-
nant and increased progressively with reduction time at 473 K,
(iii) among the isomers, the monobranched ones were predom-
inant, and (iv) whatever the linear probe molecules tested, the
catalyst underwent deactivation, although this did not affect iso-
merization and cracking selectivity.
3.2.2. Surfaces deeply reduced or easily reducible: WO3/ZrO2
catalysts with tungsten loading ꢀ15 wt%
Table 5a shows the evolution of reaction distribution prod-
ucts for the 2 MP and 4M1Pen reactions at 623 K on the
WZA15 catalyst as a function of reduction time at 723 K. Im-
portant findings include that the olefin reacted more readily than
the corresponding alkane, and that methane and ethane (which
3.2.1.3. Kinetic data Apparent activation energies (global,
isomerization, and cracking) on the WZAm sample were de-
termined at a temperature range of 423–473 K. Catalytic tests

F. Di Gregorio et al. / Journal of Catalysis 256 (2008) 159–171
165
(a)
(b)
(c)
Fig. 6. Evolution of total conversion (F) (%), of total isomerization (1), and total cracking (P) selectivities (%) as a function of reduction duration (min) at 473 K
for (a) 1-hexene (nC =), (b) 1-heptene (nC =) and (c) 1-nonene (nC =) reactions on WZAm catalyst.
7
6
9
Table 5a
Distribution of reaction products for 4MlPen and 2MP reactions at 623 K on WZA15 catalyst progressively reduced at 723 K
Probe
t
α
S
isom
S
crack
S
>C
6
Distribution of reaction products (%)
r
3MP/nC
T
6
molecule
(min)
(%)
(%)
(%)
(nmol/g/s)
(%)
C
1
C
2
C
3
C
4
C
5
22DB 23DB 3MP nC
>C
6
6
4MlPen
60
480
1400
62.0 87.5
51.5 89.0
52.5 82.0
11.5
10.5
17.5
1.0
0.5
0.5
0.5
1
1
5.3 2.4 2.4 1.8
9.3
9
8
35.6
55.6
52.2
40.6 1.4
22.1 0.3
19.7 0.3
0.9
2.5
2.7
1.7
1.9
1.2
1.3 2.2 2.1 3.8 2.5
2.2 2.5 3.8 3.7 5.5 2.2
2MP
60
480
1400
12.5 90.5
23.0 71.5
22.0 73.5
7.5
28.5
26
2
0
0.5
0.8 0.8 2.3 1.3 2.5 2.5
4.5 3.9 5.1 5.4 9.5 1.6
4.4 3.8 5.3 4.9 7.6 1.9
13.1
7
8.4
45.9
55.7
49.4
28.8
7.2
13.6 0.6
2
0
1.6
7.7
3.6
0.16
0.43
0.24
Table 5b
Distribution of reaction products for 2MP and 4MlPen reactions at 623 K on WZA15, WZA20 and WZA30 catalysts reduced one hour at 823 K
Probe
molecule
Catalyst
α
S
S
crack
(%)
Distribution of reaction (%)
iC
r
3MP/nC
T
isom
6
(%)
(%)
(nmol/g/s)
C
1
C
2
C
3
nC
4
iC
5
nC
5
22DMB 23DMB 3MP nC
6
4
2MP
WZA15 10.0 93.0
WZA20 35.5 24.0
7.0
76.0
76.0
1
5.7
12.9 10
0.6
5
1.4 0.7 0.9
6.4 3.2
16.6 9.1 3.8
1.5
4.6
13.2 10.5
0.9
3.6 2.1
1
10.8
6.8
8.9
73
48.2 11.4
12.4 2.6
8.2
9
4.2
4.8
0.13
0.45
0.07
3
WZA30
6.0 24.0
0
4MlPen
WZA20 48.0 81.5
WZA30 16.5 53.5
18.5
46.5
2.4
6.4
2.5
3.7 1.9 1.9
3.7
9.1
2.3 2.2
6.2 1.9
6.8
1.9
57.6 15
46.7
3.9
2.9 16.2
0.92
0.29
6.1 10.5 5.3 3.1
were not present in the previous experiments with slightly re-
duced surfaces) were formed when the catalyst was prereduced
at 723 K, and their relative proportions increased progressively
with reduction time at 723 K.
Table 5b shows the product distribution evolution for the
2MP and 4M1Pen reactions at 623 K as a function of reduc-
tion at 823 K on the WZA15, WZA20 and WZA30 catalysts.
The table clearly indicates that cracking product proportions

166
F. Di Gregorio et al. / Journal of Catalysis 256 (2008) 159–171
(a)
(b)
Fig. 7. (a) Initial conversion, total isomerization selectivity(%), and 3MP/nC ratio as a function of tungsten loading for 4M1Pen reaction at 473 K. (b) Evolution of
6
total conversion as a function of reduction temperature for 4M1Pen reaction at 473 K.
increased considerably with tungsten loading when starting
from 2MP at the expense of isomerization product selectivity.
The activation energies determined on the WZA15 catalyst at
573–623 K were 31 kJ/mol when starting from 4M1Pen and
109 kJ/mol when starting from 2MP (not shown).
ation between isomerization ability and the support’s tetragonal
structure.
4.2. Influence of tungsten loading on initial catalytic activity
Fig. 7a represents the initial total conversion, isomeriza-
tion selectivity, and 3MP/nC6 ratio as functions of tungsten
loading for the 4M1Pen reaction at 473 K. It seems that the
increase in tungsten species density at the surface of mono-
clinic ZrO2 support did not change the initial activity of the
WO3/ZrO2 samples, or their acidity. Those results are in contra-
diction with those reported in the literature [25–27] for systems
obtained by impregnation of an amorphous zirconium oxide (al-
though in previous studies, the reactant was not 4M1Pen, but
heptane, pentane, and xylenes). Hua and Sommer [25], Santi-
esteban et al. [26], and Barton et al. [27] reported a maximum
activity for tungsten loading slightly higher than the monolayer
coverage; beyond this value, the activity diminished. This max-
imum has been explained by increases in the number and the
force of acid sites with increasing WOx species density. The in-
corporation of a hydrogen atom through polytungstate species
structures is linked to the latter’s ability to delocalize negative
charges. In our case, the XPS studies of calcined and reduced
surfaces at 473 K [20] revealed that all of the catalysts were es-
sentially constituted of surface W6+ (∼70%) and W5+ cations
(∼30%). Those proportions were modified only slightly during
reduction at 473 K. Our results indicate that the activity was es-
sentially associated with the oxidation state of surface tungsten
cations, rather than with tungstate species structure or domain
sizes. 4M1Pen reactivity was associated only to Brönsted acid
sites with weak to moderate acidity. Hydroxyl groups were as-
sociated with W5+ cations and, to a lesser extent, to monomeric
and polymeric W6+ cations tetraedrally coordinated, as de-
scribed in the part I of this study [20].
4. Discussion
4.1. Influence of the ZrO2 structure on catalytic activity
As noted in Section 1, few papers deal with the study
of WO3/ZrO2 catalysts obtained by impregnation of a pre-
formed monoclinic zirconium oxide. WO3/ZrO2 is qualified
as superacid by Arata and Hino [22] when the catalyst is ob-
tained by impregnation of an amorphous zirconium hydroxide,
Zr(OH)4−2x. These authors asserted that acid sites are formed
by a combination of tungsten oxide with zirconium oxide when
the system adopts a tetragonal structure during the calcination
step. Several authors have noted that the presence of tungsten
oxide species on monoclinic zirconium oxide support does not
make the catalyst active for isomerization reactions [23]; the
monoclinic WO3/ZrO2 system has exhibited only a weak acid-
ity [24]. Although these findings are often cited, they have been
experimentally verified only rarely. At the opposite end, the
WO3/ZrO2 catalysts of our study with a monoclinic zirconium
oxide support demonstrated significant isomerization activity
both toward olefin (4M1Pen) and saturated molecule (2MP)
when reduced sufficiently (for tungsten loading ꢀ15 wt%).
Moreover, the results obtained for the WZA9 and WZB9 solids
demonstrate a similar behavior of both catalysts regardless of
the support structure (monoclinic or tetragonal). The absence
of activity toward 2MP on the WZA9 and WZB9 catalysts sug-
gest that these systems cannot be considered strong solid acids
or superacids, as they have been described according to the
Hammet indicators method. But the WZA9 and WZB9 cata-
lysts were immediately active for 4M1Pen conversion. It must
be noted that the support structure of the WZA9 remained mon-
oclinic after reducing treatments at 823 K, suggesting no associ-
Contrary to the conclusions of Barton et al. [27], Iglesia
et al. [28], and Kuba et al. [29], we believe that W5+ cations
are not able to initiate alkane isomerization. Indeed, Barton
et al. [27] proposed that carbenium ions are formed by a re-
dox process, leading to formation of organic radicals and W5+

F. Di Gregorio et al. / Journal of Catalysis 256 (2008) 159–171
167
Table 6
Evolution of tungsten oxidation states, determined by XPS (M: metallic state) as a function of reduction temperature
Calcined
surface
Reduction temperature (K)
473
Catalyst
623
723
823
WZAm
WZ9
WZA15
WZA20
WZA30
+6, +5
+6, +5
+6, +5
+6, +5
+6, +5
+6, +5
+6, +5
+6, +5
+6, +5
+6, +5
+6, +5
+6, +5
+6, +5
+6, +5, +4
+6, +5, +4
+6, +5, +4
+6, +5, +4
+6, +5, +4
+6, +5, +4, M
+6, +5, +4, M
+6, +5, +4, M
+6, +5, +4
+6, +5, +4, M
+6, +5, +4, M
+6, +5, +4, M
cations, whereas Kuba et al. [29] proposed that W5+ cations
exist through (H+)x[(W5+)x(W6+
cient strength to directly initiate isomerization reactions starting
from an alkane.
• A significant decrease in 4M1Pen conversion (Fig. 7b)
occurred after a long reduction (>1400 min) at 723 and
823 K. This decrease was especially significant in highly
tungsten-loaded systems, like WZA20 and WZA30. This
decreased conversion can be explained by the transforma-
tion of W6+ and W5+ cations, associated with acid sites,
into metallic tungsten, or by tungsten diffusion into zir-
conium oxide. The decrease in conversion was delayed in
systems with intermediate tungsten loadings (9 and 15%),
in which important W6+ and W5+ proportions were pre-
served. Indeed, whereas the metallic tungsten proportion
reached ca 35% on the WZA20 and WZA30 catalysts (after
a long reduction treatment at 723 K), XPS revealed a value
of only 8% on the WZA15 catalyst [20]. This drop in ac-
tivity occurred after reduction at 823 K, in accordance with
the increase in metallic tungsten proportion [20]. Known
for its excellent alkane hydrogenolysis properties, metallic
tungsten in α-W led to WZA20 and WZA30 deactivation
by coke formation. Consequently, the decreased 4M1Pen
conversion is associated with the formation of α-W through
the reduction of WO3 crystallites on highly charged cat-
alysts. Present in larger proportions on the WZA20 and
WZA30 catalysts, these large crystallites could be more
easily reduced (to α-W) compared with the small crystal-
lites on the WZA15 catalyst, which are more directly in
interaction with the support.
)
n−xO3n]x− centers of suffi-
4.3. Influence of reduction temperature on catalytic activity
As expected based on the XPS studies, the catalytic prop-
erties of the WO3/ZrO2 systems were modified as a function
of reduction temperature. As noted in part I of this report [20],
the extent of reduction depends essentially on temperature and
tungsten loading. According to reduction intensity, four tung-
sten oxidation states were evaluated by means of XPS studies,
W6+, W5+, W4+, and metallic tungsten. Table 6 summarizes
the surface oxidation states of tungsten on each catalyst ob-
tained at different reduction temperatures. The correlation be-
tween tungsten oxidation states and 4M1Pen and 2MP reactiv-
ity allows us to propose the following:
• Regardless of the tungsten loading, and especially for cat-
alysts with tungsten loadings >9 wt%, reduction at 623 K
yielded a slight increase in 4M1Pen conversion at 473 K.
As indicated in Table 6, reduction at 623 K led to the forma-
tion of W4+ cations (except for the catalyst with monolayer
coverage). Among the oxidation states of tungsten, the
W4+ cation has received much attention in the literature,
especially regarding its catalytic properties [3,5,30]. Gen-
erally considered an intermediate state between the hexava-
lent W6+ cation and metallic tungsten, the W4+ cation is
known for its hydro/dehydrogenating properties when sup-
ported on alumina. These properties result from the pres-
ence of bulk WO2, the slightly deformed rutile structure of
which confers metallic properties via π electron delocal-
ization throughout the shortest W–W bonds. Generally, the
presence of an oxide support significantly affects the length
of those bonds, hindering formation of structures with such
properties at the interface of weakly loaded catalysts (under
monolayer coverage). In our case, the zero activity toward
the saturated molecule suggests that the W4+ cations did
not exhibit hydro/dehydrogenating properties, and that they
were formed not through partial reduction of bulk WO3,
but rather by reduction of an octahedric amorphous phase
(as described in part I [20]). We believe that W4+ cation
formation increased catalytic activity not because of the
cations’ hydrogenating/dehydrogenating properties, as on
bulk WO3, but because of their acidic cracking properties.
The general decrease in conversion on increasing duration
of reduction and increasing temperature also can be explained
by the continually decreasing W/Zr ratio observed for all of the
catalysts [20]. A decreased W/Zr ratio indicates a decrease in
the number of accessible active sites either by diffusion or by
the coverage of tungsten species by the ZrO2 support, or, in
the presence of tungsten metallic particles, by coalescence. The
WZAm and WZA30 catalysts were most strongly affected by
these phenomena (Figs. 1b and 7b).
4.4. Isomerization and cracking mechanisms on hardly
reducible (tungsten loading <15 wt%) or slightly reduced
(below 423 K) surfaces
The absence of metallic properties on slightly reduced or
hardly reducible surfaces implies a purely acidic isomeriza-
tion and cracking mechanism. Even if the formation of prod-
ucts with higher carbon chain length than the starting molecule
cannot be neglected (<6%), we believe that the isomerization
mechanism was essentially monomolecular under our operating

168
F. Di Gregorio et al. / Journal of Catalysis 256 (2008) 159–171
larized to their ionic form, these alkoxy species can lead to the
formation of protonated cycloprotanic intermediates. Fig. 9 il-
lustrates the mechanisms of the different isomers via “alkoxy”
species formation by means of a surface Brönsted acid site.
Starting from 1-hexene and considering that isomerization
proceeds through B-type isomerization, we can see that the dis-
tribution of methylpentanes is close to that expected statistically
(cf. Table 3a). Theoretical distribution predicts a 2MP/3MP ra-
tio equivalent to 1, whereas it ranged between 1.07 and 1.1
when catalytic tests were performed at 473 K. This ratio con-
firms the intervention of a monomolecular mechanism. An anal-
ogous scheme to that for 4M1Pen isomerization can explain
the formation of methylpentanes. 2-Methyl-3-pentylium and 3-
methyl-2-pentylium are formed via the cyclization of n-hexyl
cation. The formation of 23DMB and 22DMB, starting from
1-hexene, proceeds through consecutive 2-methyl-4-pentylium
and 3-methyl-2-pentylium ion isomerization within the same
adsorbed phase.
1-Heptene is the first molecule for which no products with
a mass exceeding that of the starting molecule have been de-
tected. Weitkamp et al. [32] studied the isomerization of n-
heptane over zeolithes, distinguishing between monomolecular
and bimolecular mechanisms based on the absence or pres-
ence of peculiar reaction products. They proposed that 22DMP,
33DMP, and 3EP formation are suppressed on catalysts with bi-
molecular isomerization properties. All of those hydrocarbons
were present in our reaction products, in agreement with the hy-
pothesis of a monomolecular mechanism for 1-heptene. B-type
isomerization led to a 1/2 ratio of 2MH and 3MH. Experimen-
tally, this ratio ranged between 0.86 and 0.75, not far from the
thermodynamic equilibrium value of 0.82. Deviation to statisti-
cal predictions can be explained by the consecutive transforma-
tion of the products resulting from the first isomerization step
(B-type isomerization) via A-type isomerization until thermo-
dynamic equilibrium is reached.
As reported by Zhang et al. [9], dibranched molecules with
one of the two ramifications set on the second carbon are pre-
dominant. Weitkamp et al. [33] explained this finding by a
decrease in B-type isomerization. The decreased proportion of
dimethylpentanes with reaction time was not seen for 3EP, the
proportion of which increased among the isomer products with
decreasing temperature. According to Weitkamp et al. [32],
this finding is in agreement with an isomerization mechanism,
and also implies the presence of cyclopropanic intermediates
and protonated cyclobutanic intermediates, as was proposed by
Martens and Jacobs [34].
At first sight, the increase in global activation energy with
carbon chain length (Table 4) may appear surprising if a classi-
cal acidic isomerization mechanism is assumed. The adsorp-
tion of hydrocarbon generally increases with increasing car-
bon chain length. This increasing adsorption is in accordance
with the increase in cracking products typically observed dur-
ing acid catalysis with increasing carbon chain length. But, in
contradiction to these usual findings, it appears that the products
emerging from isomerization reactions were formed preferen-
tially (>95%) even when starting from 1-nonene. This finding
is in agreement with a single mode of adsorption (via a double
Fig. 8. Methylpentanes and dimethylpentanes formation via protonated cyclo-
propane.
conditions. Protonation of the linear alkene led only to the for-
mation of a secondary carbocation that did not undergo crack-
ing. Indeed, D-type cracking, leading to the primary cation, was
slower than the isomerization of the linear carbocation. The
presence of low quantities of linear alkanes among the crack-
ing products is in agreement with those affirmations.
Regarding branched molecules, only a low proportion of
propane was observed among the reaction products. This
demonstrates that cracking of monobranched alkenes through
a C-type mechanism (leading to propane) is not likely. It fol-
lows that isomerization of the secondary carbocation is the first
step in ion transformation. High isomer product selectivity, re-
gardless of the carbon chain length, suggests that the formation
and desorption of the alkene from isomerized carbenium ion
is a fast process. Thus, the contact time of carbenium ions at
the surface is short, minimizing the likelihood of C–C bond
breaking and the consecutive formation of cracking products,
especially with increasing carbon chain length of the starting
molecule.
Isomerization of 4M1Pen via A- and B-type mechanisms led
to 23DMB/22DMB/3MP/nC6 isomers with a 1/1/1/1 statisti-
cal distribution. In our experiments, isomers were formed in
the following proportions: 3MP ꢀ 23DMB > nC6 > 22DMB.
This deviation with respect to statistical distributions can be
explained by the thermodynamical stability of the different
cationic intermediates. Cyclopropanic intermediates allow us to
explain the formation of all isomers without passing through
primary carbenium ions, as shown in Fig. 8. The higher se-
lectivity to 3MP compared with dibranched molecules can
be explained by a greater stability of the starting carbenium
ion, 2-methyl-3-pentylium (1), compared with the 2-methyl-4-
pentylium ion (2). The ethyl and isopropyl groups of the former
(1) have a greater stabilizing effect than the methyl and isobutyl
groups of the latter (2).
A similar hypothesis can be advanced to explain the predom-
inance of 3MP over n-hexane and of 2,3-dimethylbutane com-
pared with 2,2-dimethylbutane. This scheme, which implies
protonated cyclopropanic intermediates, does not take into ac-
count the interaction between molecule and the catalyst surface.
Thus we have interpreted the following isomerization mecha-
nism using OH groups that lead to “alkoxy” intermediates, as
was first proposed by Cheng and Ponec [31]. When strongly po-

F. Di Gregorio et al. / Journal of Catalysis 256 (2008) 159–171
169
Fig. 9. 4M1Pen isomerization through alkoxy species.
bond in position 1) and reaction of the molecules at the catalyst
surface whatever the molecule size. The increase in activation
energy from C6 to C9 can be explained by a steric effect due
to alkyl groups in the β-position with respect to the carbon
with the double bond. This steric difficulty increases with the
size of the alkyl group, leading to a decrease in the adsorp-
tion heat value (ꢀHads), as well as a consecutive increase in
activation energy with increasing carbon number of the starting
molecule—that is, Ea = E0 + ꢀHads, with Ea apparent activa-
tion energy, E0, absolute energy, and ꢀHads < 0.
formation of the main cracking products (i.e., isobutane and
isopentane) resulted from the β-scission of carbenium ion in-
termediates of high molecular mass formed by enlargement of
surface hexyl cations. This supposition is in accordance with
the presence of C7 among the reaction products. Indeed, β-
scission of carbenium ions favors the formation of fragments
leading to branched alkanes, iC4 or iC5, but not the formation of
primary carbenium ions leading to C1 or C2. Unlike the isomer-
ization mechanism, which implies intramolecular (monomole-
cular) transformations, cracking of C6 molecules must proceed
through intermolecular (bimolecular) transformations. Analo-
gous conclusions can be drawn for the analysis of cracking
products emerging from 1-heptene to 1-nonene.
4.5. Cracking mechanisms
When starting from 4M1Pen or 1-hexene, the cracking prod-
ucts ranged from C3 to C5. In parallel, products with mole-
cular mass higher than the starting molecule were detected.
The absence of methane and ethane starting from 1-hexene
indicates that hydrogenolysis and β-scission (not favored ener-
getically due to the formation of primary ions) did not occur.
Consequently, we believe, like Santiesteban et al. [35], that
4.6. Isomerization and cracking mechanisms on easily
reducible (tungsten loading >15 wt%) or deeply reduced (over
723 K) surfaces
Comparing the reactivity of 4M1Pen and 2MP at 623 K in-
dicates that on WO3/ZrO2 surfaces prereduced at 723 K, the

170
F. Di Gregorio et al. / Journal of Catalysis 256 (2008) 159–171
Fig. 10. Isomerization of 2MP via σ-alkyl adsorbed intermediates.
olefin reacted faster than the corresponding alkane (cf. Ta-
bles 5a and 5b). This finding suggests that the rate-determining
step was the alkane C–H bond rupture. As shown previously,
2MP activation occurs with metallic tungsten formation. So is
the isomerization mechanism purely metallic or bifunctional
(dehydrogenation/hydrogenation on metallic sites and crack-
ing/isomerization on acidic sites), or do both mechanisms oc-
cur? Several findings argue in favor of a bifunctional mecha-
nism in which carbenium ions are implied: (i) the global ac-
tivation energy value, Ea, obtained for 2MP (109 kJ/mol) is
characteristic of a bifunctional catalysis; (ii) the isomerization
ratio, 3MP/nC6, was nearly the same for 4M1Pen and 2MP;
and (iii) the proportion of 2DMB was always lower than that
of 23DMB, in agreement with the thermodynamic stability of
carbenium ions.
In the same manner, several findings suggest a metallic
monofunctional mechanism intervention. Among these are the
increased proportions of methane and ethane with increasing
duration of reduction at 723 K (Table 5). Indeed, the weak
possibility of D-type β-scission on acidic sites leading to C2
suggests formation of demethylation and deethylation prod-
ucts on metallic sites. Another important finding in this regard
is the progressive decrease in iC4/nC4 and iC5/nC5 with in-
creasing intensity of reduction treatment. On acidic cracking
(β-scission), hydrogenolysis does not favor the formation of
branched molecules.
Apparently, both the bifunctional and the metallic mono-
functional mechanisms must be taken into account simultane-
ously on surfaces reduced at 723 K. Even if determining the
relative proportion of each mechanism is difficult, it seems
reasonable to suggest that the formation of cracking products
is associated with both mechanisms. But the progressive in-
crease in the proportions of C1 and C2 with decreasing tem-
perature and tungsten loading (especially for the WZA20 and
WZA30 catalysts) indicates an evolution toward a purely metal-
lic mechanism. Both the activation of 2MP and the increase in
hydrogenolysis products clearly indicate the existence of two
kind of metallic sites on the catalysts reduced at 723 K. This
can be explained by, as observed in previous studies on WO3-
based materials [36–38], the progressive transformation during
reduction of metallic tungsten initially present in a metastable
β-W phase into a α-W phase. Unlike the W3O phase (β-W),
the pure α-W phase obtained by reduction of the W3O phase
essentially has the metallic properties of hydrogenolysis. Note
that tungsten metal phases β-W and α-W can coexist and have
dehydrogenating and hydrogenating properties.
Previous studies [38,39] have demonstrated that reduction
of bulk WO3 leads to bifunctional catalysts on which the de-
hydrogenation/hydrogenation function is provided by metallic
tungsten. Transformation of the alkene (emerging from alkane
dehydrogenation) then follows the aforementioned mechanism
with protonated cyclopropanic intermediates on acidic sites.
As was previously observed on partially oxidized tungsten
carbides, we associate a nonpredominant metallic-type isomer-
ization to adsorbed σ-alkyl intermediates, like those proposed
by McKervey et al. [40]. This occurs on a β-W metallic phase.
The σ-alkyl species most energetically favored for isomer-
ization reactions over metallacyclobutane intermediates imply
transition species with two electrons and three centers, like
those shown in Fig. 10.
Regardless of the duration of reduction, the amounts of
methane and ethane obtained were always greater than pre-
dicted by the statistical rupture of a single C–C bond (C5 + C1
or C4 + C2). This finding clearly suggests consecutive rup-
ture of carbon–carbon bonds of the molecule before its des-
orption, due to multiple adsorption on the metal. That scheme
was first proposed for noble metal particles [41]. Adaptation
of this model to our catalysts suggests the presence of metal-
lic tungsten capable of adsorbing the reactant to a significant
degree. We assign those metallic sites to a α-W phase, formed
by the ultimate reduction of WO3 crystallites. It is important to
note that demethylation and deethylation products were quasi-
unformed on the WZA15 catalyst, in which the proportion of
metallic tungsten was <9% even after the reduction treatment
at 823 K. We believe that for the 15% tungsten-loaded cata-
lyst, the WO3 crystallites formed during the calcination step
were not large enough to be transformed into an α-W phase
during reduction treatments. Rather, their reduction led to a β-
W phase that was stabilized due to metal–support interactions.
These metal–support interactions were less predominant on the

F. Di Gregorio et al. / Journal of Catalysis 256 (2008) 159–171
171
largest WO3 crystallites, such as those formed on the WZA20
and WZA30 catalysts. This phenomenon makes the WZA15
catalyst a stable system for hydrocarbon isomerization.
[2] V. Adeeva, G.D. Lei, W.M.H. Sachtler, Appl. Catal. A 118 (1994) L11.
[3] A. Katrib, F. Hemming, P. Wehrer, L. Hilaire, G. Maire, J. Electron. Spec-
trosc. Relat. Phenom. 76 (1995) 195.
[4] E. Blomsma, J.A. Martens, P.A. Jacobs, J. Catal. 159 (1996) 323.
[5] A. Katrib, V. Logie, N. Saurel, P. Wehrer, L. Hilaire, G. Maire, Surface
Sci. 377 (1997) 754.
5. Conclusion
[6] A. Chica, A. Corma, J. Catal. 187 (1999) 167.
[7] M. Höchtl, A. Jentys, H. Vinek, Microporous Mesoporous Mater. 31
(1999) 271.
WO3/ZrO2 catalysts with different tungsten loadings were
evaluated for isomerization of hydrocarbons. In contradiction
to previous studies, here we found that the crystallographic
structure of the zirconium oxide had no influence on catalytic
properties of WO3/ZrO2 systems. Correlations with XPS and
XRD measurements [20] allow us to point out two catalytic be-
haviors:
[8] A. Lugstein, A. Jentys, H. Vinek, Appl. Catal. A 176 (1999) 119.
[9] W. Zhang, P.G. Smirniotis, J. Catal. 182 (1999) 400.
[10] S. Siffert, J.L. Schmitt, J. Sommer, F. Garin, J. Catal. 184 (1999) 19.
[11] J.F. Denayer, G.V. Baron, G. Vanbutsele, P.A. Jacobs, J.A. Martens,
J. Catal. 190 (2000) 469.
[12] V. Keller, F. Barath, G. Maire, J. Catal. 189 (2000) 269.
[13] V. Calemma, S. Peratello, C. Perego, Appl. Catal. A 190 (2000) 207.
[14] W. Hua, A. Goeppert, J. Sommer, J. Catal. 197 (2001) 406.
[15] A. Katrib, D. Mey, G. Maire, Catal. Today 65 (2001) 179.
[16] Y. Rezgui, M. Guemini, A. Tiguezza, A. Bouchemma, Catal. Lett. 87
(2003) 11.
[17] M. Hino, S. Kobayashi, K. Arata, J. Am. Chem. Soc. 101 (1979) 6439.
[18] K. Arata, M. Hino, J. Chem. Soc. Chem. Commun. (1987) 1529.
[19] R.A. Boyse, E.I. Ko, J. Catal. 171 (1997) 191.
[20] F. Di Gregorio, V. Keller, F. Garin, J. Catal. 225 (2004) 44.
[21] F. Di Gregorio, V. Keller, T. Di Constanzo, J.L. Vignes, D. Michel,
G. Maire, Appl. Catal. A 218 (2001) 13.
[22] K. Arata, M. Hino, in: M.J. Phillips, M. Ternan (Eds.), Proc. 9th Int. Congr.
on Catalysis, Calgary Alberta, 1988, The Chemical Institute of Canada,
Ottawa, 1988, p. 1727.
[23] Y. Huang, B. Zhao, Y. Xie, Appl. Catal. A 172 (1998) 327.
[24] J.C. Yori, J.M. Parera, Catal. Lett. 65 (2000) 205.
[25] W. Hua, J. Sommer, Appl. Catal. A 232 (2002) 129.
[26] J.G. Santiesteban, J.C. Vartulli, R.D. Bastian, D. Chang, J. Catal. 168
(1997) 431.
[27] D.G. Barton, S.L. Soled, G.D. Meitzner, G.A. Fuentes, E. Iglesia,
J. Catal. 181 (1999) 57.
[28] E. Iglesia, D.G. Barton, S.L. Soled, S. Miseo, J. Baumgartner, W.E. Gates,
G.A. Fuentes, G.D. Meitzner, in: J.W. Hightower (Ed.), 11th Int. Congr.
Catalysis, Baltimore, 1996, Elsevier, Amsterdam, 1996, p. 553, Stud. Surf.
Sci. Catal. 101 (1996) 553.
(i) Hardly reducible (tungsten loading <15 wt%) or slightly
reduced (below 423 K) surfaces were immediately ac-
tive for 4M1Pen conversion but inactive for the saturated
molecule transformation. These surfaces were essentially
associated with high oxidation surface states of tungsten
and especially to W6+ and W5+. The weak to moderate
acidity, associated mainly with Brönsted acidity, led to a
monomolecular mechanism and high isomerization selec-
tivity at 473 K regardless of the carbon chain length. The
isomerization selectivity is explained by the formation of
“alkoxy” species leading to protonated cyclopropanic inter-
mediates. Cracking reactions (not predominant) proceeded
on the most strongly acidic sites.
(ii) Readily reducible (tungsten loading >15 wt%) or deeply
reduced (over 723 K) surfaces were active for 2MP trans-
formation at 623 K. This activity is associated with the
formation of metallic tungsten through reduction of WO3
crystallites. Indeed, metallic tungsten has dehydrogenat-
ing/hydrogenating properties and confers a mainly bifunc-
tional behavior to the WO3/ZrO2 catalyst. Nonetheless, the
possibility of a minor effect of a purely metallic isomer-
ization mechanism implying both β-W metallic and more
reduced α-W phases cannot be excluded.
[29] S. Kuba, P. Lukinskas, R.K. Grasselli, B.C. Gates, H. Knözinger,
J. Catal. 216 (2003) 353.
[30] W. Grunert, E.S. Shpiro, R. Feldhaus, K. Anders, G.V. Antoshin, K.M.
Minachev, J. Catal. 107 (1987) 522.
[31] Z.X. Cheng, V. Ponec, Catal. Lett. 25 (1994) 337.
[32] J. Weitkamp, P.A. Jacobs, J.A. Martens, Appl. Catal. A 8 (1983) 123.
[33] J. Weitkamp, Ind. Eng. Chem. Prod. Res. Dev. 21 (1982) 550.
[34] J.A. Martens, P.A. Jacobs, in: J.B. Moffat (Ed.), Theorical Aspects of
Heterogeneous Catalysis, Catalysis Series, Van Nostrand–Reinhold, New
York, 1990, Chap. 2.
After the reduction treatment and tungsten loading, the cat-
alytic properties of WO3/ZrO2 were seen to evolve from a
purely monofunctional acidic catalysis to a purely metallic one,
passing through a bifunctional catalysis. A tungsten loading of
15% seems to be the critical value to ensure high activity, selec-
tivity, and stability of the catalyst on stream and good activity
for saturated hydrocarbon isomerization.
[35] J.G. Santiesteban, D.C. Calabro, C.D. Chang, J.C. Vartulli, T.J. Fiebig,
R.D. Bastion, J. Catal. 202 (2001) 25.
[36] M. Molière, Ph.D. thesis, University of Strasbourg, 1978.
[37] F. Hemming, Ph.D. thesis, University of Strasbourg, 1995.
[38] V. Keller, Ph.D. thesis, University of Strasbourg, 1993.
[39] G. Hägg, N. Schonberg, Acta Crystallogr. 7 (1954) 351.
[40] M.A. McKervey, J.J. Rooney, N.G. Samman, J. Catal. 30 (1973) 330.
[41] F.G. Gault, Adv. Catal. 30 (1981) 1.
References
[1] S. Meijers, L.H. Gielgens, V. Ponec, J. Catal. 156 (1995) 147.