T.K. Phung et al. / Journal of Catalysis 311 (2014) 102–113
111
A predominant role of corner and edge sites as carriers of highly
coordinatively unsaturated cationic sites, well established for other
oxides such as magnesia [61], is obviously very likely also on alu-
mina [1]. In fact on these ‘‘defective’’ situations, the most coordin-
atively unsaturated cationic sites are expected to exist. The likely
role of edge sites on alumina is further supported by the recent
observation by high-resolution TEM studies that also the (110) ex-
above, the differences in the catalytic behavior in ethanol dehydra-
tion are slight, showing that this reaction is not very sensitive or
demanding. Indeed the higher activity of the Puralox materials
P200 (when considered in terms of catalyst weight) and P200 or
P90 (when considered in terms of catalyst surface area) is likely
due to the high purity of these materials, resulting in high surface
acidity. IR spectra show the formation of ethoxy groups which
decompose to ethylene and water at the same temperature at
which we reveal the high catalytic activity to ethylene in the flow
reactor. It seems interesting to remark that the catalytic activity is
observed in the presence of water (which is a reaction product) at
quite a low temperature (473–673 K), i.e. in conditions where the
studies of the adsorption of CO does not reveal the presence of
Lewis acidity. We have, however, already remarked that pyridine
reveals strong Lewis sites also on largely hydroxylated samples,
due to its higher basicity with respect to not only CO but also
water. Indeed, we found that ethanol can adsorb as ethoxy groups
not only by dissociative adsorption on Lewis acid-base sites, but
also substituting hydroxyl groups on hydroxylated surfaces, caus-
ing desorption of water. In fact, light alcohols have similar acido-
basicity as water [53], and also similar volatility. This means that
the adsorption/desorption of water and ethanol are competitive.
In other words, ethanol can displace water and hydroxyl groups
producing ethoxy groups without any previous dehydroxylation.
Thus, the appearance of Lewis acidity does not need dehydroxyla-
tion if the basic probes are able to displace water. The same for
reactants: reactants which have sufficient basicity to compete with
water can be activated by alumina also in the presence of water.
This is not the case of reactants having very low basicity such as
hydrocarbons, whose conversion on alumina is usually inhibited
by water.
posed faces, supposed by most authors to be predominant on c-
Al2O3 nanocrystals, are not atomically flat but undergo a significant
reconstruction, forming nanoscale (111) facets [62] thus with a
very high density of edges.
The same sites are evident on the other alumina samples by
pyridine adsorption but are less evident, if at all, in our experimen-
tal conditions using CO adsorption. We suggest that this is due to
the higher amount of such sites on P200, as a result of its higher
surface area with respect to P90 and D100, and also a higher defect
typical of the surface of c-Al2O3 [63], partially reduced in the case
of d-Al2O3, and virtually absent in the non-defective structure of h-
Al2O3. In the case of V200, the lower amount of these sites can be
associated with the higher amount of sodium impurities in this
sample, which can result in the poisoning of some of them by
the balancing oxide ions or to a decrease in their acid strength
due to increased basicity of the oxide ligands (inductive effect).
Above, the spectra of the surface hydroxyl groups of the four
samples have been discussed and compared. While the spectrum
observed for P200 is the typical one of c-Al2O3, a role of different
impurities is likely in causing a modification of the relative inten-
sities of the bands, in the case of V200 (sodium) and D100 (chlo-
rine). The spectrum of P90 seems to be more typical of h-Al2O3,
with the almost total disappearance of the band at ca.
3770 cmꢀ1, and the mode at ca. 3730 cmꢀ1 sharper than usual
for
Al2O3 from
c
-Al2O3. This is likely associated with the crystallization of h-
Our data and our interpretation does not disagree with the data
reported by Kwak et al. [21] that found ethylene evolution from
ethanol during TPD experiments at slightly different temperatures
depending on the previous alumina pretreatment. These authors
identified the TPD peak found at 498 K on samples pretreated at
higher temperature to the evolution of ethylene from ethanol ad-
sorbed on Lewis sites, and that found at 523–533 K on samples
pretreated at lower temperature to the evolution of ethylene from
ethanol adsorbed on hydroxyl groups (they roughly define as
‘‘Brønsted acid’’ sites). Indeed, the two peaks may reflect the differ-
ence resulting from the overall hydroxylation state of the surface
on the ethylene desorption step. It seems in any case likely that
the difference between two more or less dehydroxylated surface
should be progressively removed on stream, due to the effect of
readsorption of water produced by the reaction and the desorption
of the pre-existing water, depending on the reaction temperature.
Our data well show that previous catalyst dehydroxylation is not
needed to produce ethylene from ethanol.
Characterization studies show that both Lewis and Brønsted
acidity are present over silica-alumina, while Lewis acid sites
only are present at the surface of aluminas. On the other hand,
the strength of the Lewis acid sites on silica-alumina is similar
or even stronger than that of the strongest Lewis sites of alu-
mina. Nevertheless, silica-alumina is nearly as active as the most
active alumina in terms of catalyst weight, but is significantly
less active in terms of catalyst surface area. Thus, it is supposed
that similar active sites (Lewis sites) but a lower active site den-
sity exist at the surface of silica-alumina with respect to alu-
mina. On the other hand, we can remark that at low ethanol
conversion, silica-alumina is more selective toward ethylene
than aluminas, that at low conversion produce more diethyl
ether. We suggest that this is due to the presence, at the surface
of alumina, of slightly stronger basic sites than on silica-alumina.
To produce ethyl ether, in fact, one ethanol molecule must be
activated as ethoxy groups, thus producing a nucleophilic spe-
c-Al2O3, with the loss of the typical defectivity of the
latter. These data confirm that the analysis of the OH stretching
spectrum is interesting as a typical fingerprint of the alumina
phase (in particular h-Al2O3 and c-Al2O3). On the other hand, we
agree with Dayn et al. [63] that showed (also in agreement with
the previous work of Tsyganenko and Mardilovich [64]) that many
kinds of different hydroxyl groups can be hypothesized to occur on
partially dehydroxylated alumina surfaces, thus suggesting that an
assignment of a ‘‘family’’ structure of the bands is more reliable
than the tentative identification of the planes on which they
should be located. Another remark is that the P200 and D100 mate-
rials, assumed by Wischert et al. to be essentially similar [50], are
actually very different in terms of phases, morphology, surface
area, impurities, and IR spectra of the surface hydroxyl groups.
Also for at least part of the surface hydroxyl groups, as for Lewis
sites, their location on edges and corners, more than on plane faces,
is very likely in our opinion, for the same reasons. In fact, the most
stable OH groups are likely bonded to highly uncoordinated hydro-
xyl groups. In this case, the (H)O-Al bond is expected to be stronger
and more covalent, thus being decomposed with more difficulty.
We may also mention that an OH group can quite likely be bonded
to an aluminum ion that also carries coordinative unsaturations
(e.g., an OH bonded to a Al3+ ions with overall coordination four),
and this may very likely occur on edges. Thus, Lewis sites and
OHs may, in some way, contribute to the same active site.
The catalytic data summarized above confirm that transitional
aluminas are excellent catalysts for the conversion of ethanol to
ethylene, in substantial agreement with the literature [21–23].
Yield to ethylene approaches 100% for all samples at 623 K in the
conditions of our experiments. The comparison of the data ob-
tained with aluminas and with silica-alumina suggests that Lewis
acidity represents the key property for these catalysts when ap-
plied to this reaction. In spite of the significant differences between
the surface properties of the four alumina samples, evidenced