Acid-base properties and catalytic activity of nanophase ceria-zirconia catalysts for 4-methylpentan-2-ol dehydration

Article abstract of DOI:10.1039/a903104i

The room-temperature high energy ball-milling technique was used to prepare nanophase Ce(1-x)Zr(x)O₂ (x = 0; 0.2; 0.5; 0.8; 1) catalysts. The acid-base properties of these catalysts were investigated by means of adsorption microcalorimetry, using NH₃ and CO₂ as probe molecules. The catalytic activity for 4-methylpentan-2-ol dehydration was tested at atmospheric pressure in a fixed-bed flow microreactor. The inclusion of increasingly high contents of zirconium into the ceria lattice has a complex influence on the acidity and basicity of the pure parent oxide, in terms of both number and strength of the sites. A maximum in 1-alkene selectivity is observed for the ceria-rich catalyst and a minimum for the zirconia-rich sample. Catalytic results are correlated with the acid-base properties and can be interpreted in the light of the mechanism formerly proposed for zirconia, ceria and lanthania. Surface conditioning of the zirconia-rich catalyst occurs during the run, resulting in a remarkable variation of selectivity.

Full text of DOI:10.1039/a903104i

View Article Online / Journal Homepage / Table of Contents for this issue
Acid–base properties and catalytic activity of nanophase
ceria–zirconia catalysts for 4-methylpentan-2-ol dehydration
M. G. Cutrufello,a I. Ferino,*a V. Solinas,a A. Primavera,b A. Trovarelli,¤b A. Aurouxc and
C. Picciauc
a Dipartimento di Scienze Chimiche, Universita di Cagliari, via Ospedale 72, 09124 Cagliari, Italy.
E-mail: ferino=vaxca1.unica.it; Fax: ]39 070 675 8605
b Dipartimento di Scienze e T ecnologie Chimiche, Universita di Udine, via CotoniÐcio 108, 33100
Udine, Italy
c Institut de Recherches sur la Catalyse, CNRS, 2 avenue Albert Einstein, 69626 V illeurbanne
Cedex, France
Received 19th April 1999, Accepted 24th May 1999
The room-temperature high energy ball-milling technique was used to prepare nanophase Ce Zr O (x \ 0;
1~x
x
2
0.2; 0.5; 0.8; 1) catalysts. The acidÈbase properties of these catalysts were investigated by means of adsorption
microcalorimetry, using NH and CO as probe molecules. The catalytic activity for 4-methylpentan-2-ol
3
2
dehydration was tested at atmospheric pressure in a Ðxed-bed Ñow microreactor. The inclusion of increasingly
high contents of zirconium into the ceria lattice has a complex inÑuence on the acidity and basicity of the pure
parent oxide, in terms of both number and strength of the sites. A maximum in 1-alkene selectivity is observed
for the ceria-rich catalyst and a minimum for the zirconia-rich sample. Catalytic results are correlated with the
acidÈbase properties and can be interpreted in the light of the mechanism formerly proposed for zirconia, ceria
and lanthania. Surface conditioning of the zirconia-rich catalyst occurs during the run, resulting in a
remarkable variation of selectivity.
In an attempt to enhance the redox properties of ceria by
incorporating foreign cations in its lattice, ceriaÈzirconia solid
solutions have been prepared both by conventional1,2 and
innovative methods.3h5 The increased redox capability of
ceriaÈzirconia solid solutions has also been investigated with
computer simulation techniques.6 Little has been published
about the acidÈbase features of ceria, despite the key role of
this parameter in several catalytic applications. The surface
acidity and basicity of ceria has been investigated by
temperature-programmed desorption of ammonia and step-
wise thermal desorption of carbon dioxide and it is believed
that acidÈbase pairs are important in the activation of hydro-
carbons at high temperatures.7 Mechanistic studies with deu-
terated butan-2-ols indicated that acidÈbase sites are involved
in the dehydration reaction over ceria.8 The microcalorimetric
technique has been recently used to assess the acidÈbase
properties of ceria catalysts for the dehydration of 4-
methylpentan-2-ol.9 Temperature-programmed desorption of
monomer for the manufacture of thermoplastic polymers of
superior technological properties (TPX). Fine tuning of the
acidÈbase character of the catalyst is essential in order to
obtain high selectivity to the desired 1-alkene and avoid para-
sitic formation of oleÐns with internal double bond and/or
dehydrogenation to ketone;9,11 accordingly, 4-methylpentan-
2-ol conversion also represents a useful test reaction for char-
acterising the acidity and basicity of catalysts.12 Micro-
calorimetry has gained importance as one of the most reliable
methods for the study of surface acidity and basicity of solid
catalysts. Accordingly, direct assessment of the acidÈbase
properties of the ceriaÈzirconia solid solutions, in terms of
both site concentrations and strength distributions, has been
carried out by means of this technique, using ammonia and
carbon dioxide as probe molecules.
Experimental
ammonia and CO , spectroscopic investigation of the adsorp-
2
CeO (Aldrich, 99.999% pure) was used as received. ZrO was
tion of CO and pyridine, titration by benzoic acid and test
2
2
2
prepared from hydrous zirconia (MEL chemicals) by calcina-
tion at 973 K. The solid solutions were prepared by milling
appropriate amounts of the two component oxides in a high-
energy vibratory ball-mill (Spex 8000) working at an oscil-
lation frequency of 20 Hz and an amplitude of 20 cm. Both
reactions were used to study the surface acidity and basicity of
CeO ÈCaO mixed oxides obtained by coprecipitation.10 To
2
the present authorsÏ knowledge, however, no papers have been
published concerning the acidÈbase features of ceriaÈzirconia
solid solutions.
the vial (V \ 50 cm3) and the balls (/ \ 10 mm) were ZrO
The present paper deals with the acidÈbase properties and
the catalytic activity for 4-methylpentan-2-ol dehydration of
nanophase CeO ÈZrO solid solutions, prepared by room-
temperature high energy ball-milling, whose redox and oxygen
storage properties had been formerly investigated.5 The con-
version of 4-methylpentan-2-ol is important from the practical
viewpoint as an alternative route to 4-methylpent-1-ene, a
2
made, in order to avoid leaching contamination with iron and
tungsten from the milling apparatus. The ball-to-powder ratio
was 18 : 1 by weight (i.e. 6 balls and 1 g of powder) and the
milling time 9 h. The composition of the solid solutions pre-
pared through this way, as well as their structural and textural
features, are summarised in Table 1. Further details are
reported elsewhere,5
2
2
Tian-Calvet heat Ñow equipment (Setaram) was used for
microcalorimetric measurements. Each sample was pretreated
¤ Present address: Dipartimento di Ingegneria Chimica, Universita
di Roma-La Sapienza, via Eudossiana 18, 00184 Roma, Italy.
Phys. Chem. Chem. Phys., 1999, 1, 3369È3375
3369

View Article Online
Table 1 Characteristics of the catalysts
decreases continuously for all the samples, with the exception
of ZrO , where a step is observed at about 140 kJ mol~1. In
Surface area
/m2 g~1
2
the region of high ammonia uptakes Q
is very low for all
Sample
Phase
diff
the samples and changes very slightly, which suggests the
occurrence of physical adsorption.
CeO
2
Ce Zr
0.5 0.5
Ce Zr
0.2 0.8
3
21
12
8
Cubic
Cubic
Ce Zr
O
O
O
To individuate in Fig. 1 the point at which the adsorption
shifts from chemical (i.e. at the acid sites) to physical (which
must be disregarded in the determination of the acidÈsites
0.8 0.2
2
2
2
Cubic (traces tetragonal)
Tetragonal (traces cubic
and monoclinic)
Monoclinic
concentration), plots of integral heat, Q , vs. ammonia
ZrO
28
int
2
uptake (Fig. 2a) and the adsorption isotherms (Fig. 2b) have
been drawn. Very low or negligible evolution of heat associ-
ated with remarkably increasing ammonia uptake (i.e. attain-
ment of a plateau in Fig. 2a) and the simultaneous deviation
upwards of the adsorption isotherm from the attained plateau
(i.e. the change of concavity in Fig. 2b) are an indication of the
establishment of physical adsorption. Inspection of Figs. 2a
and b and comparison with Fig. 1 allows one to conclude that
di†erential heats below ca. 30 kJ mol~1 are related to physical
adsorption.
overnight at 673 K under vacuum (10~3 Pa) before the suc-
cessive introduction of small doses of the probe gas (ammonia
or carbon dioxide). The equilibrium pressure relative to each
adsorbed amount was measured by means of a di†erential
pressure gauge (Datametrics). The run was stopped at a Ðnal
equilibrium pressure of 133.3 Pa. The adsorption temperature
was maintained at 353 K, in order to limit physisorption.
Testing of the catalytic activity was performed at atmo-
spheric pressure in a quartz-made Ðxed-bed Ñow microreactor
(10 mm id). All the equipment devices, including the stainless-
steel pipes to and from the reactor, had been passivated by
The total concentration of acid sites, n , has hence been
A
calculated for all the samples as the amount of ammonia
adsorbed with Q P 30 kJ mol~1; it is plotted vs. the oxide
diff
hot HNO treatment before the assembling. The catalyst was
composition in Fig. 3. It can be seen that n , which is the
3
A
activated in situ (6 h at 773 K under CO -free air Ñow). 4-
lowest for pure ceria, markedly increases upon addition of 20
2
Methylpentan-2-ol was fed in with an N stream (partial pres-
mol% of zirconia, then decreases as the ZrO content is
2
2
sure, p
\ 15.5 kPa, time factor, W /F \ 2.5 g
h
further increased up to 80 mol% and Ðnally attains the
0, alcohol
cat
g~1 ). For each catalyst, several reactor temperatures were
highest value on pure zirconia. The upwards shifting of the
alcohol
checked during the same run, until about 50% conversion was
Q
vs. ammonia uptake curves observed in Fig. 1 when
diff
reached. A new run with a fresh portion of the same sample
was then started and carried out at the temperature giving
50% conversion, in order to exclude inÑuence of the thermal
history of the sample on its catalytic behaviour and allow a
better comparison of the catalysts. On-line capillary GC
analysis conditions were: Petrocol DH 50.2 column, oven
temperature between 313 and 473 K, heating rate 5 K min~1.
Product identiÐcation was conÐrmed by GC-MS. No signiÐ-
cant di†erences in conversion were observed in runs where
both the Ñow rate and the catalyst amount were signiÐcantly
changed while keeping the same W /F value; accordingly, the
occurrence of external di†usion limitations was ruled out.
passing from pure ceria to the solid solutions and then to pure
zirconia indicates an increase of the acid strength; the step at
140 kJ mol~1 indicates that a set of homogeneous sites of high
strength is present on pure zirconia, whereas on the other
samples the sites are continuously heterogeneous. To further
compare the acid strength of the samples, the population of
the sites with Q P 80 kJ mol~1, n
, has been evaluated
diff
Aw80
Results and discussion
Microcalorimetric characterisation
Information about the acid properties of the pure oxides and
the solid solutions is summarised in Fig. 1, where the di†eren-
tial heat of adsorption, Q , is plotted vs. ammonia uptake.
diff
The initial heat of adsorption is relatively low (76 kJ mol~1)
for CeO , lies in the range 131È108 kJ mol~1 for the solid
2
solutions and is rather high (175 kJ mol~1) for ZrO . In the
2
region of intermediate ammonia uptake the di†erential heat
Fig. 1 Di†erential heat of adsorption, Q , vs. ammonia uptake.
Fig. 2 Integral heat of adsorption, Q , vs. ammonia uptake (a) and
diff
int
(…), CeO ; (L), Ce Zr O ; (>), Ce Zr O ; (K),Ce Zr O ;
isotherms for ammonia adsorption (b). (…), CeO ; (L), Ce Zr O ;
2
(=), ZrO .
2
0.8 0.2
2
0.5 0.5
2
0.2 0.8
2
2
0.8 0.2 2
(>), Ce Zr O ; (K),Ce Zr O ; (=), ZrO .
0.5 0.5
2
0.2 0.8
2
2
3370
Phys. Chem. Chem. Phys., 1999, 1, 3369È3375

View Article Online
Fig. 3 Total concentration of acid sites, n , (…), and fraction of sites
Fig. 5 Di†erential heat of adsorption, Q , vs. CO uptake. (…),
A
diff
2
with di†erential heat of adsorption P80 kJ mol~1, n
: n , (=), vs.
CeO ; (L), Ce Zr O ; (>), Ce Zr O ; (K), Ce Zr O ; (=),
Aw80
A
2
ZrO .
2
0.8 0.2
2
0.5 0.5
2
0.2 0.8 2
oxide composition.
step is visible for Ce Zr O , whereas a plateau is present
(reasons for choosing such a limit stem from previous Ðnd-
0.2 0.8
2
on pure zirconia at ca. 145 kJ mol~1. The total concentration
ings,9 as will be discussed later on). In Fig. 3 the fraction of
of basic sites, n , as well as the fraction of sites with di†eren-
sites of this strength over the total sites, n
(as %) vs. the oxide composition. An increasing trend is
: n , is plotted
B
Aw80
A
tial heat P 80 kJ mol~1, n
: n , is plotted (as %) vs. the
Bw80
B
oxide composition in Fig. 6. It shows that, in comparison with
observed along the series: CeO (where no sites with Q
P
2
diff
pure ceria, the CeO -rich solid solution has a remarkable
80 kJ mol~1 are present) @ Ce Zr O \ Ce Zr O \
2
0.8 0.2
2
0.5 0.5 2
basic character, both in terms of n and n
: n ; the base
Ce Zr O \ ZrO .
0.2 0.8
In the case of CO adsorption, the shape of both the Q vs.
B
Bw80
B
2
2
2
properties of the solid solutions are attenuated as the zirconia
content is increased up to 80 mol% and then growing again
for pure zirconia. Note also (Fig. 5) that, in contrast with pure
ceria, whose sites are continuously heterogeneous, most of the
sites of the ceria-rich CeO ÈZrO form an homogeneous set of
int
uptake curves (Fig. 4a) and the adsorption isotherms (Fig. 4b),
points out the occurrence of chemical adsorption only. The
Q
vs. uptake plots in Fig. 5 show that the initial heats of
diff
adsorption are quite similar for all the samples (145È155 kJ
2
2
high strength; the population of this set lowers when passing
mol~1). SigniÐcant di†erences are observed as the CO uptake
2
to CeO ÈZrO of intermediate composition and Ðnally disap-
increases: the di†erential heat continuously decreases for
2
2
pears on the zirconia-rich solid solution, despite the fact that
pure zirconia has a remarkable population of homogeneous
sites of high strength.
CeO ; a plateau at ca. 155 kJ mol~1 is observed for
2
Ce Zr O and Ce Zr O , rather short on the latter; no
0.8 0.2
2
0.5 0.5 2
The inclusion of increasingly high contents of zirconium
into the ceria lattice has hence a complex inÑuence on both
the acidity and basicity of the pure parent oxide. No general
rules for predicting the base character of binary oxides have
been proposed up to now. On the contrary, qualitative models
have been developed concerning the generation of new acidic
features upon mixing di†erent oxides. Most frequently cited
are the Tanabe13 and Kung models.14 Both models presume
dilute solid solutions of one cation in a matrix oxide. The
Tanabe model is based on a local approach: mismatches in
cation valency and coordination number along a hetero-
linkage MÈOÈM@ originate an excess charge at the site of the
minor component cation, resulting in acidity generation. For
solid solutions of cubic ceria and monoclinic zirconia the
Fig. 4 Integral heat of adsorption, Q , vs. CO uptake (a) and iso-
int
therms for CO adsorption (b). (…), CeO ; (L), Ce Zr O ; (>),
0.8 0.2
Fig. 6 Total concentration of base sites, n , (…), and fraction of sites
2
B
with di†erential heat of adsorption (80 kJ mol~1, n
: n , (=), vs.
2
2
2
Bw80
B
Ce Zr O ; (K),Ce Zr O ; (=), ZrO .
0.5 0.5 0.2 0.8
oxide composition.
2
2
2
Phys. Chem. Chem. Phys., 1999, 1, 3369È3375
3371

View Article Online
excess charge is ]1/2 for ceria-rich compositions and [4/7
for zirconia-rich solutions. According to the model, acidity
should be generated in both cases, of Lewis type in the former
and of BrÔnsted type in the latter. This prediction does not
match with the outcome of the microcalorimetric experiments,
which, though not able to discriminate between the nature of
the sites, show that an acidity larger than the sum of the acid-
ities of the component oxides is generated on Ce Zr
0.8 0.2
(Fig. 3). The assumption that zir-
O
2
but not on Ce Zr
O
0.2 0.8
2
conia maintains its monoclinic structure upon mixing with
cubic ceria seems too crude in the light of structural analysis,
which has shown5 that the e†ect of milling CeO and ZrO is
2
2
the formation of a Ñuorite-structured, cubic solid solution as
long as the zirconia content does not exceed 50 mol%, a
partial tetragonalisation being observed for higher zirconium
loading. On this basis calculation gives zero excess charge for
both ceria-rich and zirconia-rich solid solutions and no new
acidity should hence be expected, which is in contrast with the
experimental Ðndings. In no way is this model able to predict
the acidity trend for ceriaÈzirconia solid solutions.
Contrary to TanabeÏs model, KungÏs approach14 also takes
into account long range e†ects, by relating acidity generation
to both the electrostatic potential at the substituting cation
site and changes in the matrix necessary to balance stoichiom-
etry. A general conclusion of the model is that for oxides
having the same metalÈoxygen stoichiometry, new acidity (of
Lewis type) can be generated only if the host oxide is more
ionic than the guest. Accordingly, due to the higher ionic
character of CeO in comparison with ZrO , generation of
2
2
acidity should be expected for ceria-rich solid solutions but
not for zirconia-rich compositions, which qualitatively repro-
duces the observed experimental trend.
Fig. 7 Di†erential heats of adsorption, Q , vs. ammonia (a) and
diff
CO (b) uptakes for the as-prepared, (=), and reduced Ce Zr O ,
(L), samples.
2
0.8 0.2 2
Additional calorimetric runs have been carried out on
Ce Zr
O
and Ce Zr
O after exposure to hydrogen
0.8 0.2
2
0.2 0.8
2
atmosphere at 673 K overnight, i.e. conditions under which
Ce4`/Ce3` reduction is expected to occur to some extent.5
Comparison of the results for the reduced and as-prepared
samples is presented in Figs. 7 and 8 for the ceria-rich and
zirconia-rich solutions respectively. The most interesting
e†ects of the reduction of Ce Zr
weakening of its acid sites and the partial loss of homogeneity
of the base sites. The former e†ect is indicated by the down-
O
(Fig. 7) are the
0.8 0.2
2
wards shift of the Q
vs. the NH uptake curve (Fig. 7a);
diff
3
note that n
for the reduced sample is less than 60% of the
Aw80
original value for the untreated Ce Zr O . Fig. 7b shows
0.8 0.2
2
that after reduction, the initial di†erential heat for CO is
higher and the plateau in the Q vs. the CO uptake curve is
shorter, which indicates a more heterogeneous character of
the base sites. Weakening of the acid sites upon reduction is
2
diff
2
observed also for Ce Zr O , Fig. 8a (the Q
0.2 0.8 diff
uptake curve is shifted downwards; n is about one half of
vs. the NH
2
3
Aw80
the original value). No signiÐcant change of Q
vs. the CO
diff
2
uptake curve results from reduction (Fig. 8b), i.e. the strength
distribution of the base sites remains unchanged.
Ceria reduction by H treatment can be envisaged to occur
2
as follows:15 (1), dissociation of chemisorbed hydrogen to
form hydroxy groups; (2), formation of anionic vacancies and
reduction of the neighbouring cations; (3), desorption of water
by combination of hydrogen and hydroxy groups; (4), di†u-
sion of surface anionic vacancies into the bulk material. It is
also possible that Ce4`/Ce3` reduction occurs following H
2
dissociation without formation of oxygen vacancies in the Ðrst
step. However, such a possibility seems unlikely for the
present samples, as it would originate a hydroxylated surface,
i.e. an increased concentration of the acid sites which is not
experimentally observed (Figs. 7 and 8). The assumption that
the cerium cations act as Lewis sites is supported by the
observed acid-strength decrease after H treating, ascribable
to the lower charge/radius ratio of Ce3` (radius 1.14 Ó) in
comparison with Ce4` (radius 0.97 Ó).
Fig. 8 Di†erential heats of adsorption, Q , vs. ammonia (a) and
2
diff
CO (b) uptakes for the as-prepared, (=), and reduced Ce Zr O ,
2
0.2 0.8 2
(L), samples.
3372
Phys. Chem. Chem. Phys., 1999, 1, 3369È3375

View Article Online
According to the reduction mechanism outlined above, a
chemical pumping e†ect causes oxygen to di†use up to the
surface, where the vacancies are Ðrst originated upon water
elimination (step 3). Oxygen mobility should be favoured in
ceriaÈzirconia solid solutions because the substitution of Ce4`
with the smaller Zr4` cation (radius 0.84 Ó) causes shrinking
of the CeO Ñuorite-type lattice and formation of structural
2
defects. The energetics of oxygen ion transport in the
CeO ÈZrO solid solutions has been investigated over the
2
2
whole composition range by computer simulation methods;6
the activation energy for oxygen migration was found to
decrease with increasing zirconia content, a minimum being
attained for Ce Zr O , which was hence considered the
0.1 0.9
2
most easily reducible system. Decrease of the cell volume with
increasing zirconia content was experimentally observed for
the present samples and was considered responsible for the
higher reduction extent of Ce Zr
O in comparison with
0.2 0.8
2
Ce Zr O .5 The surface oxygen anions act as base sites in
0.8 0.2
2
the calorimetric experiments. The results in Figs. 7 and 8 can
be tentatively explained as follows. For the hydrogen-treated
Ce Zr O , anion vacancies not yet replaced by oxygen dif-
0.8 0.2
2
fusing from the bulk as well as oxygen anions are probably
present on the surface; a higher degree of heterogeneity
should hence be expected in comparison with the same sample
not treated with H . For the H -treated Ce Zr O , a
2
2
0.2 0.8 2
higher extent of bulk reduction occurs, which implies that the
anion vacancies formed at the beginning of the reduction
process (step 2) have already been replaced by oxygen anions
coming from the bulk; accordingly, no signiÐcant di†erences
in the strength distribution of the base sites should be
expected in comparison with the same sample not exposed to
Fig. 9 Temperature for initial conversion \ 50%, T , (a), and initial
50
selectivity to 4-methylpent-1-ene, S , (b), vs. oxide composition.
1A
H . In addition, in this sample the lower concentration of
2
cerium limits the total number of oxygen vacancies that can
be formed upon reduction. Therefore, the similarity of the
strength distribution of base sites after reduction should be
the result of (i) the increased di†usion of O2~ and (ii) the
decrease of the amount of reducible elements.
for Ce Zr O ). It is worthy of note that there is a close
0.2 0.8
2
resemblance between the 1-alkene selectivity trend on one
hand (Fig. 9b) and the trends of the acid-sites concentration
(Fig. 3), base-sites concentration (Fig. 6) and fraction of base
sites stronger than 80 kJ mol~1 (Fig. 6) on the other.
Previous Ðndings for the same reaction on zirconia,11 ceria
and lanthania9 allowed us to conclude that when the acid and
base functions of the catalyst are well balanced in terms of site
concentrations, a two-point adsorption of the reactant alcohol
occurs, in which the most acidic hydrogen (i.e. that of the ter-
minal methyl group) interacts with a base site while the acid
centre interacts with the OH group of the alcohol. The fate of
this adsorbed species depends on the relative strength of the
acid and base sites. If the acid sites are weak and the base sites
are strong enough, rupture of the CÈH bond with carbanion
formation occurs Ðrst and an E1cB mechanism sets in, leading
to the preferential formation of the 1-alkene (Hofmann
orientation). A basic strength of 80 kJ mol~1 seems to rep-
resent the threshold value required for establishing the above
sequence of steps. If the strength of the acid and base sites is
comparable, no intermediate carbanion is formed, the trans-
formation of the adsorbed species into the oleÐn being con-
certed (E2 mechanism); 2-alkene is then the preferred product
(Saytze† orientation).
Catalytic behaviour
4-Methylpent-1-ene (the desired product) and 4-methylpent-2-
ene were the main products formed upon reacting 4-
methylpentan-2-ol over the CeÈZr oxides. Other C -alkene
isomers with internal double bond were also formed, to a very
limited extent in all cases, with the exception of Ce Zr O .
6
0.2 0.8
2
Light oleÐns such as propene were also detected (their amount
being, however, negligible in all cases, with the exception of
CeO ), as well as trace amounts of pentene oligomers. Side
2
dehydrogenation of the reactant alcohol to 4-methylpentan-2-
one was practically negligible for all the catalysts. The conver-
sion of the reactant alcohol, as well as the selectivity to each
of the above products, was monitored as a function of time-
on-stream for each oxide in isothermal runs lasting 30 up to
80 h. The reaction temperature was such that the initial (1 h
on-stream) conversion was ca. 50% (see Experimental section).
The initial catalytic results for all the samples are sum-
marised in Fig. 9, where both the temperature for 50% con-
version, T , and the initial selectivity to 4-methylpent-1-ene,
Such a co-operative action of the sites is disfavoured when
either the acid or the base sites are predominant. In the
former case, the reaction is initiated by the attack of the acid
sites to the hydroxy group of the alcohol: an intermediate car-
bocation is formed, which transforms into alkenes with inter-
nal double bonds (E1 mechanism, Saytze† orientation). When
the base sites on the catalyst surface are predominant and
strong, adsorption occurs by means of a hydrogen bond
involving the OH group of the alcohol and a base site;
50
S
, are plotted vs. composition. Increasingly high zirconia
1A
contents in the solid solution enhance its overall activity, as
indicated by the decreasing trend of T (Fig. 9a); though the
50
low activity of CeO might stem from its low surface area
2
(Table 1), there is no manifest correlation between the latter
and the overall activity of the samples (note for instance that
Ce Zr O is rather active despite its low surface area). Fig.
0.2 0.8
2
9b shows that the inclusion of 20 mol% of zirconia into the
ceria lattice results in a remarkable improvement of the selec-
abstraction of H by interaction with a positively polarised H
a
tivity to the desired 1-alkene (54% for CeO , 71% for
atom on the surface (originating from the previously split OH
2
Ce Zr O ); further addition of zirconia is not convenient
group of the alcohol) then occurs and a ketone is formed
instead of an alkene.
0.8 0.2
2
(60% selectivity for Ce Zr O ) or even detrimental (32%
0.5 0.5
2
Phys. Chem. Chem. Phys., 1999, 1, 3369È3375
3373

View Article Online
In light of this, it is not surprising that the values for n , n
remarkable carbanionic character) or via the formation of the
intermediate carbanionic species (E1cB).
A
B
and n
: n appear so relevant to the initial 1-alkene selec-
Bw80
B
tivity in the present case, as indicated by a comparison of
Figs. 3 and 6 with 9b. To check in detail how the above
picture Ðts the present results, the initial selectivity to 4-
methylpent-1-ene has been plotted vs. n : n in Fig. 10, where
It has already been noted that Ce Zr O does not lie on
the curve, its selectivity to 4-methylpent-1-ene being rather
low in comparison with that for the other samples of similar
0.2 0.8
2
n : n . For Ce Zr
0.2 0.8
sites is quite similar (n
Bw80
O the strength of the acid and base
B
A
B
A
2
previous data for ceria, lanthania and zirconia-based
catalysts9,11 have also been reported. (Both the present and
former data refer to 50% conversion). With the exception of
Ce Zr O , the two sets of data are Ðtted by the same
: n \ 55%; n
: n \ 61%,
B
Aw80
A
from Figs. 6 and 3). The fraction of alcohol that adsorbs on
these sites by a two-point mechanism does not undergo car-
banion formation, but rather one-step alkene formation
through a concerted E2 pathway. Both the latter and the
simultaneously present E1 mechanism lead to 2-alkene, which
explains the observed low selectivity to 4-methylpent-1-ene.
The conversion of the reactant alcohol, as well as the selec-
tivities to the various products, were not a†ected in the same
way for all the catalysts by time-on-stream. Some loss of the
0.2 0.8
2
volcano-shaped curve. In agreement with the picture outlined,
the maximum in 1-alkene selectivity corresponds to those
catalysts (Na-doped zirconia, ZrO /27;11 ceria prepared from
2
CeIII hydroxide by calcination in air, CeO /49 9 and
2
Ce Zr O ) where the acid and base functions are well-
balanced as to the number of sites (n : n is close to 1), but
unbalanced as to the strength (n
: n \ 0% for ZrO /27 from data in ref. 11; 70 and 31%
for CeO /49 from data in ref. 9, and 86 and 31% for
0.8 0.2
2
B
A
Bw80
: n \ 80% and
overall activity occurred over CeO , where conversion
decreased from 50 to 35% after 24 h on-stream and then did
not change further; the selectivities to 4-methylpent-1-ene,
B
2
n
Aw80
A
2
2
Ce Zr
O
from Figs. 6 and 3). These catalysts represent
S
, 4-methylpent-2-ene, S , C -alkene isomers, S , 4-
0.8 0.2
2
1A
2A
6
C6
somewhat of a limiting case in which dehydration through the
E1cB mechanism predominates. Other limiting cases are rep-
resented by the two zirconia samples ZrO /98 and ZrO /46,
prepared from di†erent precursors:11 on the latter, n : n is
so high that dehydrogenation is strongly favoured over dehy-
dration, on the former, n : n is so low that only dehydration
methylpentan-2-one, S , were nearly constant over the entire
K
run (54, 35, 3 and 1%, respectively). The catalyst appeared
grey-coloured after reaction; after treating at 773 K for 6 h
under air Ñow, the initial activity was restored. Probably,
small amounts of alkene oligomers that accumulated on the
surface during the early hours on-stream are responsible for
the observed activity loss (such an e†ect would be ampliÐed
by the low surface area of the catalyst). This is also suggested
by the presence in the reactor effluent of unusual amounts of
light alkenes (selectivity, S \ 5%), which were detected only
in traces for the other samples; these light alkenes are likely
originated by fragmentation (favoured by the high reaction
temperature) of the oligomers, whose presence was also
2
2
A
B
B
A
practically occurs, almost exclusively by the E1 mechanism.
ZrO , CeO and Ce Zr
O
as well as CeO /48 (a pure
2
2
0.5 0.5
2
2
ceria sample prepared from CeIII nitrate by calcination in air9)
lie on the left-hand branch of the curve in Fig. 10, showing a
tendency of the selectivity to 4-methylpent-1-ene to increase
with n : n . Both Saytze†- and Hofmann-oriented pathways
LA
B
A
are simultaneously present, as indicated by the values of 1-
alkene selectivity (47È60%). As n : n increases, the two-point
adsorption of the reactant alcohol is enhanced, which favours
E2 and E1cB over E1. However, the strength of the acid and
detected in the reactor effluent (selectivity, S \ 2%).
B
A
HA
Inclusion of zirconia in the ceria lattice originates a stable
catalyst for zirconia contents up to 50 mol%; the case of
base sites is not balanced over these catalysts (n
: n \
Ce Zr O , i.e. the more selective catalyst for the desired
Bw80
B
0.8 0.2 2
85% and n
: n \ 55% for CeO /48, from data in ref. 9;
1-alkene, is shown in Fig. 11 as an example. For Ce Zr
O
Aw80
A
2
0.2 0.8 2
n
: n \ 54% and n
: n \ 0% for CeO , 100 and
a continuous increase of conversion with time-on-stream from
ca. 15% up to more than 60% was observed (Fig. 12a), associ-
ated with a remarkable variation of the selectivities (Fig. 12b):
Bw80
B
Aw80
A
2
60% for ZrO , and 81 and 39% for Ce Zr O , from Figs.
2
0.5 0.5 2
6 and 3), which makes 2-alkene formation through the E2
mechanism unlikely. Hofmann orientation could result either
from a concerted pathway (E2 with a transition state having a
growth of 4-methylpent-2-ene, drop of C -alkene isomers,
6
decrease of the desired 1-alkene; only trace amounts of ketone
were formed at any time-on-stream. As already observed for
the acidÈbase properties and the initial selectivity to 1-alkene,
the on-stream features of Ce Zr
O also seem unique in
0.2 0.8
2
comparison with the other CeÈZr samples. It is worthy of note
that this behaviour is also completely di†erent from that of
the parent pure zirconia. For the latter, neither the conversion
nor the selectivities signiÐcantly changed with time-on-stream:
4-methylpent-1-ene was always slightly favoured over 4-
methylpent-2-ene (S \ 55%;
S
\ 43%); C -alkene
1A
2A
6
isomers were almost negligible (S \ 2%); only traces of 4-
C6
methylpentan-2-one were detected.
The observed changes in the catalytic behaviour of
Ce Zr
O with time-on-stream indicate that some kind of
0.2 0.8
2
surface conditioning occurs during reaction. The possibility
should be considered that this stems from a modiÐcation of
the acidÈbase parameters relevant to the reaction selectivity,
originated by a Ce4`/Ce3` shift occurring on the surface as a
consequence of the reducing atmosphere experienced by the
catalyst in contact with the reactant alcohol. However, even
assuming that the reaction conditions make the sample
undergo reduction to an extent similar to that occurring after
overnight exposure to hydrogen atmosphere at 673 K, the
Fig. 10 Initial selectivity to 4-methylpent-1-ene, S , vs. n : n for
1A
B
A
various catalysts. Full symbols refer to previously investigated cata-
lysts: two zirconia samples, ZrO /98 and ZrO /46, prepared from dif-
changes in the values of n : n and n
the observed selectivity changes with time-on-stream. For the
: n do not justify
B
A
Bw80
B
2
2
ferent precursors;11 Na-doped zirconia, ZrO /27;11 two ceria
2
H -exposed sample n : n and n
respectively (from data in Fig. 8); accordingly, a progressive
shift towards a more E1cB character should be expected upon
: n are 1.0 and 63%
samples, CeO /49 and CeO /48, prepared from di†erent precursors9
2
B
A
Bw80
B
2
2
and lanthania, La O /50.9 Open symbols represent the samples inves-
2
3
tigated in this work.
3374 Phys. Chem. Chem. Phys., 1999, 1, 3369È3375

View Article Online
sure and reaction temperature) are probably too mild for
allowing signiÐcant reduction. Actually, the colour of the
sample after reaction had only a green shade over the original
yellow, which seems to indicate that surface reduction has
occurred to a negligible extent.
Surface conditioning might stem from strong interactions
between the active sites and the reaction products. In no way,
however, would the interaction with the product alkenes
result in the increase of conversion with time-on-stream.
Hydrogen adsorption, reported to occur dissociatively on
ZrO to form ZrOH, ZrH and ZrHZr surface species,3 can
2
also be ruled out in the present case, due to the negligible
extent of the dehydrogenation reaction. On the contrary,
water is rather an abundant side product. Assuming its disso-
ciative adsorption on site pairs formed by metal and oxygen
ions, an increasingly high extent of surface hydroxylation
would be expected during the run. The resulting increase of
the concentration of the acid sites on the surface would
explain the growth of conversion and the overall enhancement
of the Saytze† products. The progressively favoured formation
of 4-methylpent-2-ene over C isomers with time-on-stream,
6
i.e. the progressive limitation of the consecutive isomerisation
reactions, would result from a fast inactivation of the stronger
acid sites, possibly due to strong interaction with the reaction
products. The possibility for water to be dissociatively
adsorbed on metal oxides is well documented and has also
been reported for ZrO .16 It can take place on site pairs
2
formed by coordinatively unsaturated Zr and oxygen ions and
is preceded by the undissociative adsorption of H O on Zr
cations. The tentative hypothesis could be proposed that on
Fig. 11 Conversion (a) and selectivities (b) vs. time-on-stream for
2
Ce Zr O . Symbols in (b): (=), 4-methylpent-1-ene; (…), 4-
0.8 0.2
2
Ce Zr
O the extent of surface hydroxylation is particu-
methylpent-2-ene; (K), C alkene isomers; (L), 4-methylpentan-2-one.
0.2 0.8
2
6
larly high because the distance between the acid (cations) and
base (oxygen ions) sites is such as to assist the dissociation
mechanism of the water molecules previously adsorbed undis-
sociatively on the cation site.
reduction, which is in contrast with the selectivity trends in
Fig. 12b. Moreover, the increasing trend of conversion (Fig.
12a) suggests an increase of the concentration of the active
sites on the surface during the reaction, which cannot be
related to a possible reduction of the surface. It should also be
added that the reaction conditions (i.e. alcohol partial pres-
Acknowledgements
One of the authors (C. P.) gratefully acknowledges a scholar-
ship from the European Union (PO Alta Formazione FSE).
References
1
P. Fornasiero, R. Di Monte, G. Ranga Rao, J. Kaspar, S.
Meriani, A. Trovarelli and M. Graziani, J. Catal., 1995, 151, 168.
Y. Sun and P. A. Sermon, J. Mater. Chem., 1996, 6, 1025.
F. Zamar, A. Trovarelli, C. de Leitenburg and G. Dolcetti, Stud.
Surf. Sci. Catal., 1996, 101, 1283.
2
3
4
5
6
C. de Leitenburg, A. Trovarelli, F. Zamar, S. Maschio, G. Dol-
cetti and J. Llorca, J. Chem. Soc., Chem. Commun., 1995, 2181.
A. Trovarelli, F. Zamar, J. Llorca, C. de Leitenburg, G. Dolcetti
and J. T. Kiss, J. Catal., 1997, 169, 490.
G. Balducci, J. Kaspar, P. Fornasiero, M. Graziani, M. S. Islam
and J. D. Gale, J. Phys. Chem. B, 1997, 101, 1750.
7
8
V. R. Choudary and V. H. Rane, J. Catal., 1991, 130, 411.
K. Tomke, in Proceedings of Sixth International Congress on
Catalysis, ed. G. C. Bond, P. B. Wells and F. C. Thomkins, The
Chemical Society, London, 1977, vol. 1, p. 105.
9
A. Auroux, P. Artizzu, I. Ferino, R. Monaci, E. Rombi, V. Solinas
and G. Petrini, J. Chem. Soc., Faraday T rans., 1996, 92, 2619.
10 P. Kaner and M. Baerns, Appl. Catal., 1996, 139, 107.
11 A. Auroux, P. Artizzu, I. Ferino, V. Solinas, G. Leofanti, M.
Padovan, G. Messina and R. Mansani, J. Chem. Soc., Faraday
T rans., 1995, 91, 3263.
12 A. Auroux, P. Artizzu, I. Ferino, R. Monaci, E. Rombi and V.
Solinas, Microporous Mater., 1997, 11, 117.
13 K. Tanabe, T. Sumiyoshi, K. Shibata, T. Kiyoura and J. Kita-
gawa, Bull. Chem. Soc. Jpn., 1974, 47, 1064.
14 H. H. Kung, J. Solid State Chem., 1984, 52, 191.
15 A. Trovarelli, Catal. Rev.-Sci. Eng., 1996, 38, 439.
16 V. Bolis, C. Morterra, M. Volante, L. Orio and B. Fubini, L ang-
muir, 1990, 6, 695.
Fig. 12 Conversion (a) and selectivities (b) vs. time-on-stream for
Ce Zr O . Symbols in (b): (=), 4-methylpent-1-ene; (…), 4-
0.2 0.8
2
methylpent-2-ene; (K), C alkene isomers; (L), 4-methylpentan-2-one.
Paper 9/03104I
6
Phys. Chem. Chem. Phys., 1999, 1, 3369È3375
3375