P. Lauriol-Garbay et al. / Journal of Catalysis 280 (2011) 68–76
75
present should correspond to the anionic species [NbO(C2
O4)(OH)2(H2O)]À or [NbO(OH)4(H2O)]À [38], which could efficiently
exchange with [OH]À groups of the zirconium hydroxide Zr(OH)4
precursor. Such an exchange is not possible, or remains very limited,
for catalysts prepared by the impregnation of ZrO2 supports, with a
prior heat treatment at 600 °C, as in the case of the NbOx/ZrO2
catalysts.
The dissolution of Nb is accompanied by a decrease in the lat-
tice parameters of the zirconium oxide phase. This may be related
to the fact that the cationic size of Nb5+ (0.070 nm) is smaller than
the cationic size of the host Zr4+ (0.084 nm) ion. These results are in
agreement with those given in the literature [39], which also re-
port that at thermodynamic equilibrium the solid solution of Nb
in ZrO2 reached a concentration of at least 5% (Zr/Nb = 9.5).
Although the starting solid composition of the catalysts could only
allow the formation of such a solid solution, the latter is not com-
pletely formed. Because the Nb and Zr homogeneity in the formed
precursor is not optimal, and because the heating temperature, i.e.,
600 °C, is not sufficiently high, a completely homogeneous solid
solution is not formed and part of the niobium remains at the sur-
face of the catalyst in the form of a polymeric species.
The significant formation of the solid solution certainly allows
the number of strong acid sites (Lewis type acid sites) present at
the surface of the ZrO2 support phase to be decreased, as well as
the number of basic sites as evidenced by XPS spectroscopy. These
later sites have been shown to be unselective for glycerol dehydra-
tion to acrolein [16,40]. They may also be responsible like the
strong acid sites, but to a lower extend, for the low stability of
the catalysts. With that respect several studies have shown that
the coke loadings were higher on most acidic catalysts
[20,21,41]. The deactivation of catalysts for glycerol dehydration,
through the deposition of carbon on their surface during the cata-
lytic reaction, may also be influenced by several other parameters
such as the reaction temperature, the reactant feed, and the poros-
ity of the support. For the ZrNbO catalysts, a pore size of approxi-
mately 7 nm and a reaction temperature of 300 °C appeared to be
efficient parameters for long-term stability. A more complete study
would be needed, to determine whether these are the optimum
values for long-term stability. A high diluting gas flow rate, which
reduces the residence time of coke precursors in the reactor, has
also been shown to decrease the rate of deactivation. For such a
case, a compromise between the slower deactivation and the lower
productivity needs to be found.
The bulk Nb2O5 phase and Nb2O5 supported on silica have both
been shown to be active and selective catalysts for the dehydration
of glycerol [18,41]. It is thus not surprising that niobium oxide sup-
ported on zirconia is an efficient catalyst for this reaction. Zirconia
certainly has an effect on the strength of the supported species, or
on their stabilization, which leads to better catalytic properties. It
is noteworthy that niobium-based oxides have mainly Brønsted acid
sites, but also Lewis acid sites at their surface [42]. Although most
papers agree that Brønsted acid sites are the most active and selec-
tive for the dehydration of glycerol, the precise influence of Lewis
acid sites remains unknown. Studies have been undertaken in order
to better characterize the acid–base properties of the new catalyst
described here and to ascertain the effect of a zirconia support for
the niobium species, as well as the role of Lewis acid sites.
Nb is an effective agent for the stabilization of t-ZrO2, when pre-
pared above 600 °C [24]. In this respect, the relative t-ZrO2 content
increased considerably, in accordance with the Nb content of the
catalysts. However, it is rather difficult to ascertain the influence
of the ZrO2 structure (monoclinic or tetragonal) on the catalytic
performance of the catalysts. In the case of WOx/ZrO2 catalysts,
various discrepancies can be found in the literature. Some authors
have reported that the presence of tungsten oxide species on
monoclinic zirconium oxide does not activate the catalyst for isom-
erization reactions, because they exhibit only a weak acidity
[43,44], whereas other authors did not note any difference [45].
Presumably, in the study reported here, since the strength of the
active acid sites is weak, the nature of the zirconia phase may
not be a key factor, unless the non-selective zirconia sites are
neutralized.
The compound Nb2Zr6O17, which is a solid solution composition
with a modulated structure of the type Nb2ZrxÀ2O2x+1, with
7.1 6 x 6 10.3 [46], is characterized by a relatively high selectivity
to acrolein. This may be related to the presence of mainly weak
acid sites at its surface, with even fewer strong acid sites than
the most selective ZrNbO-12 catalyst. However, conversion on this
catalyst is not stabilized as a function of time, when on stream.
This may be explained by the presence of a higher content of basic
sites, as detected by XPS characterization, which could be respon-
sible for deactivation. Experiments conducted on basic catalysts
have shown that these types of catalyst are poorly active and
unstable [41].
5. Conclusion
This study has demonstrated that ZrNbO mixed oxides are
selective catalysts for the dehydration of glycerol to acrolein at
300 °C. These new catalysts appear to be particularly efficient,
since they deactivated only very slowly by coking. The efficiency
of the catalysts is related to that of the catalytic acid sites contrib-
uted by the niobium species, supported on zirconia, but also to the
neutralization of the non-selective sites of the support. The method
used to prepare the catalysts appears to be determinant in gener-
ating both features, since it allows the formation of a solid solution
of Nb in ZrO2, evidenced both by X-ray diffraction and Raman spec-
troscopy, and the spreading of polymeric niobium oxide species at
the surface of this solid solution. The results show that the same
catalyst could not be reproduced simply by grafting niobium oxide
onto the surface of zirconia. The deactivation has been shown to be
related to the formation of coke on strong Lewis acid sites and
eventually to the adsorption of polyglycerol on basic sites at the
surface of the catalysts. Both types of non-selective sites present
at the surface of the uncovered zirconia should be annihilated by
the dissolution of Nb into zirconia, which explains the high stabil-
ity of the prepared NbZrO catalysts. Although deactivation is
strongly slowed on these catalysts, it still occurs. However, a sim-
ple treatment with flowing air at a higher temperature than the
reaction temperature was found to be sufficient to completely
regenerate the deactivated catalysts back to their original level of
activity. This is a positive point for a possibly industrialized pro-
cess. Finally, it is noteworthy that the catalysts’ composition was
not fully optimized, such that there is room for further improve-
ment. The reaction conditions could certainly also be optimized.
It is important to note that a previous study of a Nb2O5-based
catalyst for an acid catalyzed reaction reported that it was partic-
ularly resistant to coking [42]. It is thus credible that the high sta-
bility of the catalyst is related not only to the neutralization of the
Lewis acid and basic sites of the zirconia support, by incorporation
of niobium into zirconia, but also to an intrinsic property of the
niobium oxide species.
Acknowledgments
The formation of the solid solution of Nb in ZrO2 has implica-
tions in terms of the nature of the zirconia phase formed, since
We gratefully acknowledge ADISSEO for financial support. We
thank C. Lorentz and P. Delichere for 13C NMR and XPS analyses.