Niobia-modified aluminas prepared by impregnation with niobium peroxo complexes for dimethyl ether production

Article abstract of DOI:10.1016/j.cattod.2012.02.062

Use of a water-soluble niobium peroxo complex allowed the preparation of niobium-modified aluminas containing up to 90% of the theoretical niobia monolayer in one impregnation step. There was a maximum in the density of surface Lewis acid sites at 45% of the theoretical monolayer. FTIR of adsorbed pyridine and adsorbed CO₂ suggest the vertical growth of the Nb ₂O₅ layer for the largest niobium contents. The addition of 22.5% of the theoretical monolayer eliminated about 80% of the basic surface hydroxyls, inhibiting the adsorption of gas phase CO₂ by the samples. The niobia/alumina catalysts suffered less inhibition by CO₂ than the pure alumina in the methanol dehydration reaction, confirming that Nb ₂O₅ is mainly deposited on sites where the CO₂ adsorption is stronger, leaving free sites that are active in catalytic dehydration and less inhibited by CO₂, however none of the niobia/aluminas was more active than the pure alumina. Nevertheless, in the direct syngas to DME conversion using a mixed catalyst system comprised of a CuZnAl methanol synthesis catalyst and a methanol dehydration component, the activity was significantly larger with a niobia/alumina as a dehydrating component than with the pure alumina.

Full text of DOI:10.1016/j.cattod.2012.02.062

Catalysis Today 192 (2012) 104–111
Contents lists available at SciVerse ScienceDirect
Catalysis Today
journal homepage: www.elsevier.com/locate/cattod
Niobia-modified aluminas prepared by impregnation with niobium peroxo
complexes for dimethyl ether production
Angela S. Rochaa,b, Aline M. da S. Forrestera, Elizabeth R. Lachtera, Eduardo F. Sousa-Aguiarc,
Arnaldo C. Faro Jr.a,∗
a Instituto de Química, Universidade Federal do Rio de Janeiro, Ilha do Fundão, CT, Rio de Janeiro, RJ CEP 21949-900, Brazil
b NUCAT/PEQ/COPPE/Universidade Federal do Rio de Janeiro, Caixa Postal 68502, CEP 21941-972, Rio de Janeiro, RJ CEP 21949-900, Brazil
c CENPES/Petrobras, Ilha do Fundão, Cidade Universitária, Quadra 7, Rio de Janeiro 21949-900, Brazil
a r t i c l e i n f o
a b s t r a c t
Article history:
Use of a water-soluble niobium peroxo complex allowed the preparation of niobium-modified aluminas
containing up to 90% of the theoretical niobia monolayer in one impregnation step. There was a maximum
in the density of surface Lewis acid sites at 45% of the theoretical monolayer. FTIR of adsorbed pyridine
and adsorbed CO2 suggest the vertical growth of the Nb2O5 layer for the largest niobium contents. The
addition of 22.5% of the theoretical monolayer eliminated about 80% of the basic surface hydroxyls,
inhibiting the adsorption of gas phase CO2 by the samples. The niobia/alumina catalysts suffered less
inhibition by CO2 than the pure alumina in the methanol dehydration reaction, confirming that Nb2O5
is mainly deposited on sites where the CO2 adsorption is stronger, leaving free sites that are active in
catalytic dehydration and less inhibited by CO2, however none of the niobia/aluminas was more active
than the pure alumina. Nevertheless, in the direct syngas to DME conversion using a mixed catalyst
system comprised of a CuZnAl methanol synthesis catalyst and a methanol dehydration component, the
activity was significantly larger with a niobia/alumina as a dehydrating component than with the pure
alumina.
Received 31 October 2011
Received in revised form
29 December 2011
Accepted 9 February 2012
Available online 7 May 2012
Keywords:
Alcohol dehydration
Dimethylether
Syngas
Niobia
Alumina
© 2012 Elsevier B.V. All rights reserved.
1. Introduction
Methanol synthesis is performed on CuO–ZnO–Al2O3 based
catalysts with high activity and selectivity [4–6], while alcohol
Interest on dimethyl ether (DME) production has grown signifi-
cantly in recent years, due to its potential use as an alternative fuel
to LPG (liquefied petroleum gas) and diesel [1–3], or as an important
intermediate in the production of dimethyl sulfate, methyl acetate
and light olefins [3]. As a fuel, its use is particularly attractive due
to its high cetane number and storage facility. Diesel motors could
burn DME after some small modifications, reaching low particu-
lates (soot) and NOx emission, and no sulfur compounds.
Syngas (CO + H2) is the main source for DME production. This is
an additional attractive feature of DME utilization, as syngas can be
produced from both natural gas and biomass. DME may be obtained
from syngas in one or two steps. The usual process goes through
two steps, where methanol is first produced from syngas (CO + H2)
and then dehydrated to DME:
dehydration takes place on acid catalysts [7–9]. ␥-Alumina is the
traditional catalyst for dehydration of alcohols.
An alternative route is the coupled synthesis of methanol and
DME from synthesis gas using dual catalyst systems in one step,
the so-called STD (synthesis gas to DME) process. In this system,
the catalysts could be a physical mixture of the methanol synthesis
catalyst and an acid solid or a bi-functional catalyst with activity to
produce methanol, and also dehydrate methanol to DME and water
[3,10–17]. Besides reactions (1) and (2) above, reaction (3) below,
the water-gas shift (WGS) reaction, also occurs in this system, as the
normally used methanol synthesis catalyst (CuZnAl) is also active
for the WGS reaction:
CO + H2O → CO2 + H2
(3)
CO + 2H2 → CH3OH
(1)
(2)
␥-Alumina would be a natural choice as a dehydrating compo-
nent in such a dual catalyst system. However, previous work from
our group with alumina-supported niobia in isopropanol dehydra-
tion [18] highlighted the importance of basic sites for this reaction.
In fact, in the case of alumina-supported niobia, a linear correlation
was found between the specific isopropanol dehydration activity
and basic site density. The fact that carbon dioxide adsorbs strongly
2CH3OH → CH3OCH3 + H2O
∗
Corresponding author. Tel.: +55 21 25627821.
E-mail addresses: farojr@iq.ufrj.br, farojr@gmail.com (A.C. Faro Jr.).
0920-5861/$ – see front matter © 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.cattod.2012.02.062

A.S. Rocha et al. / Catalysis Today 192 (2012) 104–111
105
on basic sites (it is frequently used as a probe molecule for basic site
titration) is a potential problem for this use of alumina, since con-
siderable amounts of carbon dioxide are produced by reaction (3)
in the STD process.
used without modification as the methanol synthesis component
in the mixed catalyst system. A ␥-Al2O3 calcined at 773 K for 3 h or
11.1%NbAl were used as the dehydration component.
Previous work has shown that deposition of niobia on the alu-
mina surface leads to the elimination of basic sites [18,19]. On
one hand, this had a negative effect on the isopropanol dehy-
dration reaction [18] but, on the other, it also inhibited carbon
dioxide adsorption. A balance between these effects could lead to an
improved performance of niobia–alumina, as a dehydration com-
ponent in dual catalyst systems for DME production, as compared
to pure ␥-alumina.
2.2. Characterization
The surface areas and pore volumes of all catalysts were deter-
mined from nitrogen adsorption data at 77 K in an automated
volumetric apparatus, Micromeritics ASAP 2010 C. Before the anal-
yses all samples were treated in situ under vacuum at 673 K. The
catalysts were analyzed by X-ray diffraction in a Rigaku Miniflex
˚
diffractometer, with Cu K␣ radiation, ꢀ = 1.5418 A, 30 kV and 15 mA.
In this paper, we report on the effect of supported niobia on the
activity and selectivity of ␥-alumina in DME production, as a dehy-
dration component in a dual catalyst system for the STD process.
We also present results on the activity of these materials in the gas
phase methanol dehydration reaction and on the inhibitory effects
of carbon dioxide on this activity.
In the preparation of niobia/alumina catalysts, it is difficult to
introducehigh amountsof niobia in one impregnationstep, because
generally the niobia precursors have low solubility in water. It has
been recently reported that niobium peroxo complexes have a high
water solubility [20] and appear to be an attractive alternative for
the preparation of supported catalysts with high niobium load-
ings in a single impregnation step. So we also present results on
the characterization of niobia/aluminas prepared from ammonium
diperoxo-dioxaloniobate.
FTIR measurements were carried-out in a Perkin Elmer Spec-
trum One spectrophotometer, between 4000 and 400 cm−1
.
Self-supported wafers of the samples, containing 10 mg cm−2, were
evacuated at ca. 1.3 mPa (10−5 Torr) and 723 K for 4 h. After cooling
to room temperature, the spectrum was recorded. For the acid-
ity analysis, the sample at 323 K was then exposed to 0.133 kPa
(1 Torr) of pyridine vapors at equilibrium and a second spectrum
was recorded. The same wafer was submitted to evacuation for
10 min at 323 K and the spectrum was obtained. Subsequent evac-
uation was performed at 373, 423, 473, 573 and 673 K for 10 min
followed by the spectral acquisitions. The spectra presented were
obtained by subtracting the spectra recorded after from the one
recorded before pyridine introduction. For the basicity analysis,
essentially the same procedure was followed, replacing the pyri-
dine by CO2.
2.3. Methanol dehydration in a recirculation system
2. Experimental
Methanol dehydration experiments were performed at 101 kPa
total pressure, 523 K and 8.9 kPa methanol initial partial pressure
(balance helium) in a batch apparatus with external re-circulation
of the gas phase. The re-circulation loop is connected to a vacuum
line that can be evacuated to ca. 1.3 mPa (10−5 Torr). About 10 mg
of catalyst were used after in situ evacuation at 723 K. In a similar
series of experiments, carbon dioxide was added to the reactant in
a 1:1 proportion relative to methanol.
The composition of the reaction mixture was measured by
periodically withdrawing samples from the gas phase and inject-
ing them into a gas chromatograph (HP 6890) equipped with a
methyl siloxane capillary column (100.0 m × 250 ␮m × 0.50 ␮m)
and a mass-selective detector (MSD HP 5973). Under the applied
reaction conditions, the only observed product was DME. Activities
were calculated from the initial slope of conversion vs. time data.
2.1. Catalysts preparation
The ␥-alumina was obtained by calcination of a commercial
boehmite supplied by Condea Chemie (Pural SB RT04/111) at 823 K
for 3 h. Bulk niobium oxide was prepared by calcination at 773 K for
4 h of niobic acid (Nb2O5·nH2O), supplied by CBMM (Araxá, Brazil).
The niobium precursor used was the ammonium diperoxo-
dioxalo niobate complex (NH4)3[Nb(O2)2(C2O4)2], that was
synthesized in two steps using the method reported by Bayot
et al. [20]. First, the synthesis of ammonium tetraperoxoniobate,
(NH4)3[Nb(O2)4], was performed by treatment of 1 g of niobic acid
in 25 mL of distilled water with 25 mL of a 30 wt% solution of H2O2
and 6 mL of 25 wt% solution of ammonia. The mixture was agitated
until complete dissolution, when 50 mL of acetone were added,
yielding a white precipitate, which was filtered off, washed with
acetone and air-dried. This first complex was dissolved in an aque-
ous oxalic acid solution. The solution turns rapidly to a bright yellow
color. Addition of acetone yielded a yellow microcrystalline solid,
the niobium precursor, which was filtered off, washed with acetone
and air-dried.
Four Nb2O5/Al2O3 catalysts, containing between 22.5 and 90.1%
of the theoretical monolayer (6.3 niobium atoms per nm2 [21])
of niobia on alumina (between 5.9 and 20 wt%), were prepared
by incipient wetness impregnation of a ␥-alumina obtained from
boehmite by calcination at 823 K with a solution containing the
required amount of the niobium precursor complex dissolved in
10 wt% aqueous nitric acid. The catalysts were subsequently dried
at 393 K for 2 h and calcined at 723 K for 4 h. In preliminary tests (not
shown here for reasons of space) where the calcination tempera-
ture was varied between 623 and 773 K, a maximum in catalytic
activity for methanol dehydration was found for this temperature.
These samples were named X%NbAl, where X is the niobia content
in wt%.
2.4. Direct synthesis of DME in a batch reactor
The catalytic systems, comprised of a physical mixture of 2 g
of a commercial methanol synthesis catalyst (CuZnAl) and 4 g of
a methanol dehydration catalyst (␥-Al2O3 or Nb2O5/Al2O3), was
suspended in n-hexadecane and evaluated in the direct synthesis
of dimethyl ether from syngas (CO/H2 1:2 molar ratio) in a high
pressure batch reactor.
A Parr commercial reactor, model 4560, equipped with a gas
burette, was modified for the catalytic measurements. The catalysts
were pre-treated in situ with 200 mL min−1 of hydrogen at 5.0 MPa,
533 K for 16 h, with heating rate of 10 K min−1. After reduction, the
reactor was depressurized, purged with the syngas, the tempera-
ture was raised to 553 K under agitation of 500 rpm and pressurized
to 5.0 MPa. The gas burette was previously adjusted to keep the
system at 5.0 MPa during the reaction and the syngas uptake was
obtained by the decrease in pressure in the gas burette.
After consumption of 1.0 MPa of syngas from the burette, a
gas phase sample was removed from the reactor and analyzed in
the same gas chromatograph described under Section 2.3. Under
A commercial methanol synthesis catalyst supplied by John-
son Matthey Catalysts (KATALCO 51-9), named here CuZnAl, was

106
A.S. Rocha et al. / Catalysis Today 192 (2012) 104–111
Table 1
Textural characterization of the alumina and niobia/aluminas.
Aa (m2 g−1
)
Acorr (m2 g−1
)
VP (cm3 g−1
c
)
Vcorr (cm3 g−1
)
dP (nm)
b
d
e
Catalyst
␥-Al2O3
199
197
188
186
184
199
209
211
221
230
0.58
0.54
0.50
0.47
0.43
0.58
0.57
0.56
0.56
0.54
11.6
11.0
10.6
10.1
9.3
5.9% NbAl
11.1% NbAl
15.9% NbAl
20% NbAl
a
Surface area per mass of catalyst.
Surface area per mass of support.
Pore volume per mass of catalyst.
Pore volume per mass of support.
Average pore diameter from 4VP/A.
b
c
d
e
the conditions used here, the detectable gas phase products were
methanol, DME, carbon dioxide and water. The n-hexadecane
phase was also analyzed for the identification of any heavy prod-
ucts that might be formed in the reaction, but no products could be
detected.
Activities were obtained from change in pressure vs. time
data, which were transformed into molar quantities using com-
pressibility factors. Selectivity measurements were performed at
iso-conversion by means of the chromatographic analysis. Negligi-
ble quantities of ethane were also observed.
Textural properties (surface areas, pore volumes and average
pore diameters) are shown in Table 1. Surface areas and pore vol-
umes, on a catalyst mass basis, decreased slightly with increasing
niobium loading. Even if pore blockage did not occur, some decrease
in these values is expected simply from the increase in density of
the materials due to niobium incorporation. To correct for this fac-
tor, surface areas per mass of support are also presented in Table 1.
Interestingly it is seen that corrected surface areas actually increase
significantly with increasing niobium content. At the same time,
corrected pore volumes and average pore diameters, calculated
from 4V/A, decrease in the same direction. If a porous amorphous
niobia phase was formed, separate from the alumina phase, one
would expect an increase, and not the observed decrease in the cor-
rected pore volume. Thus it seems that the supported niobia phase
is deposited inside the pores of the alumina, causing a decrease in
pore volume and in pore diameter, but increasing the surface area
per mass of support, presumably due to a roughening of the surface.
The FTIR spectra of the pure and niobia-modified aluminas
in the OH-stretching region, after evacuation at 723 K for 4 h,
are shown in Fig. 1. Peak maxima are observed with the pure
alumina at 3770, 3730, 3675 and 3587 cm−1. Maxima close to
these wavenumbers have been attributed to hydroxyl groups of
increasing basicity as the wavenumber increases [19,22,23]. The
assignment of these bands has been discussed before [22–25]. With
the niobia-modified aluminas it is clearly seen that the three max-
ima with the highest wavenumbers decrease pronouncedly as the
niobium content increases. The profiles obtained are very similar to
the ones reported by Burcham et al. [19], except that, here, the peak
at 3770 cm−1, corresponding to the most basic surface hydroxyls,
cannot be discerned already at a niobium content corresponding to
22.5% of the theoretical monolayer (5.9 wt% niobia).
3. Results and discussion
3.1. Identification of the ammonium diperoxo-dioxaloniobate
complex
The niobium peroxo-oxalate complex used in the synthesis of
all niobia/alumina was identified by FTIR using KBr wafer. The
spectrum is available as supporting information. The main bands
observed are assigned in accordance with the proposals of Bayot
et al. [20] for this type of complex. Thus, the asymmetric stretching
frequency of the carboxylate groups complexed to niobium appears
as a not well defined band at 1685 cm−1 and a symmetric mode
band appears around 1400 cm−1. The bands at 1707 and 1257 cm−1
are attributed to C O and C O stretching modes. The terminal
Nb O vibration could be identified by the band at 909 cm−1. Diper-
oxo species present bands in the region of 800–880 cm−1
.
The obtained complex was calcined at 873 K for 4 h and lost
29.9% of its initial weight, against an expected loss of 30.7% accord-
ing to the reaction expected from decomposition of ammonium
diperoxo-dioxalo niobate:
2(NH4)3[Nb(O2)2(C2O4)2] → 6NH3 + 8CO2 + 3H2O + Nb2O5
3675 3587
3730
3770
The solubility of the complex was found to be 0.226 g mL−1
in water (niobium oxide basis) and 0.385 g mL−1 in 10 wt% HNO3
solution (also niobium oxide basis). With these high values of sol-
ubility, it was possible to prepare all the niobia/alumina materials
using only one incipient wetness impregnation step. For compar-
ison, the water solubility of ammonium oxaloniobate, commonly
used as a precursor for aqueous impregnation of niobium, was ca.
0.15 g mL−1, niobium basis.
(e)
(d)
(c)
3.2. Characterization of niobia/aluminas with varying niobium
contents
(b)
(a)
For all alumina-supported niobia catalysts, X-ray diffraction pat-
terns only displayed the characteristic reflections of the ␥-alumina
support. Analysis by X-ray fluorescence provided results for the
niobium contents that differed by less than 7% from the theoreti-
cal ones. The niobia loadings used here are the ones obtained from
XRF, not the nominal ones.
4000
3800
3600
3400
3200
3000
Wavenumber / cm-1
Fig. 1. FTIR spectra in the OH stretching region. (a) ␥-Al2O3; (b) 5.9%NbAl; (c) 11.1%
NbAl; (d) 15.9%NbAl; (e) 20%NbAl.

A.S. Rocha et al. / Catalysis Today 192 (2012) 104–111
107
Table 2
Amount of pyridine adsorbed on Lewis acid sites as a function of temperature.
Temperature
Catalyst
␥-Al2O3
5.9%NbAl
11.1%NbAl
15.9%NbAl
20%NbAl
Pyridine adsorbed on Lewis sites
423a
473b
573b
673b
117
84
57
110
75
28
134
91
57
107
85
48
130
80
37
10
11
15
11
6.8
a
Absolute amount in ␮mol g−1
Amount relative to the one adsorbed at 423 K.
.
b
Table 3
Integrated FTIR peak areas for the bands associated with pyridine adsorbed on aluminum (1623 and 1617 cm−1) and niobium (1610 cm−1) Lewis acid sites vs. niobium loading
as percentage of the theoretical monolayer (evacuation at 423 K for 10 min).
Catalyst
Integrated peak areas (a.u.)
% T.M.d
a
b
c
A1623 + A1617
A1610
% A1610
␥-Al2O3
1.03
0.57
0.54
0.40
0.51
0.00
0.21
0.76
0.57
0.59
0.0
26.5
58.4
58.5
53.6
0.0%
22.5%
45.0%
67.5%
90.1%
5.9%Nb/Al
11.1%Nb/Al
15.9%Nb/Al
20%Nb/Al
a
Added integrated peak areas for the bands at 1623 and 1617 cm−1
.
Integrated peak area for the band at 1610 cm−1
.
b
c
Integrated peak area for the band at 1610 cm−1 relative to the added area of the three bands.
Niobia loading as percentage of the theoretical monolayer (6.3 Nb atoms per nm2).
d
The acidities of the materials were studied by FTIR of adsorbed
particles of niobia dispersed on alumina, in the catalyst with the
highest niobia content.
pyridine after evacuation at different temperatures. Spectra
obtained after evacuation at 423 K for 10 min are shown in Fig. 2.
Absorbances were obtained after subtraction of the spectrum from
the dried wafer before pyridine adsorption.
Quantification of the LAS was obtained from the spectra of
adsorbed pyridine using the method proposed by Emeis [27].
Table 2 summarizes the results of such quantification. The figures
for the 423 K evacuation temperature correspond to the amount
of pyridine adsorbed on the LAS. For the remaining temperatures,
the figures represent the fraction of pyridine remaining on the LAS
after evacuation at a given temperature. No systematic variation
in total Lewis acidity was found as a function of niobium loading,
all values lying near to 120 14 ␮mol g−1. However, a maximum
acidic strength appears to exist for the catalyst containing 11.1%
niobia (45% of the theoretical monolayer, assuming monolayer cov-
erage to be 6.3 Nb atoms per nm2 [21]), since this catalyst retained
the largest fraction of the LAS-adsorbed pyridine after evacuation
at 673 K. Several studies [28–30] use the concept of cation elec-
tronegativity in order to explain the increase in acid strength of
oxides with addition of a dopant: the acid strength of the oxide
is thought to be increased upon addition of a more electronega-
tive cation, due to inductive effects. Using the method proposed
by Tanaka and Osaki [28] for the calculation of electronegativities,
values of 17.6 and 10.5 are obtained for Nb5+ and Al3+, respectively.
The increase in acid site strength when niobia partially covers the
alumina surface may therefore be related to the increase in acid
strength of weaker acidic sites on the alumina surface, caused by
the inductive effect of neighboring niobium atoms. The subsequent
decrease may be due to coverage of these LAS by niobia as suggested
by the growth of the band at 1610 cm−1 with increasing niobium
content.
The usual bands attributed to pyridine adsorbed on Lewis
acid sites (LAS) are observed on all catalysts in the region of
1600–1625 cm−1 (ꢁ8a
C C
C), 1575 cm−1 (ꢁ8b C), 1490 cm−1 (ꢁ19a
C
N) and 1450 cm−1 (ꢁ19b
C
N). Analyzing the group of bands
attributed to ꢁ8a
C C, it is possible to observe differences among
the samples with varying niobia contents. While this band appears
as a doublet around 1617 and 1623 cm−1 in the alumina spectrum,
a band around 1610 cm−1 appears as a shoulder that increases with
the niobia content. This band has already been reported as pyridine
coordinated to co-ordinatively unsaturated Nb5+ sites in niobium
phosphate [26] and niobia/alumina [18].
A small band at 1540 cm−1, attributed to pyridine adsorbed on
Brønsted acid sites, is observed only in the spectrum of 20%NbAl.
This result indicates the incipient formation of three-dimensional
The group of bands between 1600 and 1625 cm−1 was decom-
posed, using Lorentzian functions, into three constituent bands, one
of them at ca. 1610 cm−1, assumed to arise from pyridine adsorbed
on niobium LAS, and two others at ca. 1617 and 1623 cm−1, respec-
tively, attributed to pyridine adsorbed on aluminum LAS [18]. Fig. 3
illustrates the results of these decompositions.
The existence of two 8a vibration bands with alumina may
be attributed to pyridine adsorbed on octahedrally (1617 cm−1
)
Fig. 2. FTIR spectra of adsorbed pyridine after evacuation at 423 K on the NbAl
catalysts.
and tetrahedrally (1623 cm−1) coordinated Al3+ ions [31]. The

108
A.S. Rocha et al. / Catalysis Today 192 (2012) 104–111
1441
1615
1650
(a) ; R2= 0.99563
1624
1595
1578
1578
1230
1618
1624
1610
1589
(b) ; R2= 0.99627
(c) R2= 0.9898
(d) R2= 0.99708
1533
1340
1617
1623
1612
γ-Al2O3
1595
1591
1575
1578
5.9NbAl
1610
1618
1623
11.1%NbAl
20%NbAl
1610
1617
1622
(e) R2= 0.99674
1800
1590
1576
1580
1600
1400
1200
Wavenumber / cm-1
1640
1620
1600
1560
Fig. 4. FTIR spectra of adsorbed carbon dioxide after evacuation at 323 K.
wavenumber / cm-1
Fig. 3. Deconvolution of the FTIR bands of adsorbed pyridine in the range
1550–1650 cm−1. (a) ␥-Al2O3; (b) 5.9%NbAl; (c) 11.1%NbAl; (d) 15.9%NbAl; (e)
20%NbAl.
hydroxyl groups of the alumina surface, as reported before in the
literature [18,19].
Besides these main bicarbonate bands, shoulders can be
observed in the spectrum of the alumina support (cf. Fig. 5) at the
low wavenumber side of the ꢁ2 vibration mode and the high wave
number side of the ꢁ3 vibration mode. These shoulders may be
attributed to another type of surface bicarbonate. According to the
calculations of Baltrusaitis et al. [33], the difference in wavenum-
ber between the ꢁ2 and ꢁ3 vibration modes would be smaller in
a bidentate surface bicarbonate species than for a bridge-bonded
one. Clearly these shoulders become relatively more important as
the niobia loading increases. It is reasonable to expect that, as the
coverage of the alumina surface by niobia increases, the probability
of formation of a bridge-bonded bicarbonate species, that requires
two aluminum atoms, decreases with respect to the formation of a
bidentate one, that requires a single aluminum atom.
Besides these shoulders, small and broad bands can also be
observed at ca. 1530 and 1350 cm−1, which are especially evi-
dent in the spectrum of the ␥-alumina support. As discussed by
Lavalley [32], the free carbonate anion has a ꢁ3 vibration mode at
1415 cm−1 and loss of the D3h symmetry on adsorption leads to
splitting of these band, producing a doublet symmetrically located
around 1415 cm−1. The average between 1350 and 1530 is close
to 1415, so these bands are consistent with adsorbed carbonate,
however there is not a general agreement on the assignment of
this band to a specific adsorption mode [32,34]. These bands also
integrated intensities of these bands, per unit catalyst area, after
evacuation at 423 K, are shown in Table 3. The intensities of the
bands at ca. 1617 and 1623 cm−1 were added, in order to obtain
a number proportional to the surface concentration of pyridine
adsorbed on aluminum LAS. Assuming molar extinction coefficients
to be the same for the bands assigned to pyridine adsorbed on alu-
minum sites and the one for pyridine adsorbed on niobium sites,
the proportion of pyridine adsorbed on niobium sites may be esti-
mated. From the last column in Table 3, it is seen that the percentage
of pyridine adsorbed on niobium LAS follows quite closely the nio-
bium content in terms of percentage relative to the theoretical
monolayer, up to a niobia loading corresponding to nearly 60% of
the theoretical monolayer. This suggests that, up to this point, the
niobia is spread essentially as a monolayer on the surface of the
alumina.
Chemisorption of CO2 has been used as a probe for basic sites
on solid surfaces [32]. While ␥-Al2O3 is capable to chemisorb this
probe molecule, no chemisorption could be observed for pure nio-
bia, which indicates the absence of basic sites on the niobia surface.
In this paper, carbon dioxide chemisorption on alumina and on the
niobia/aluminas was followed by FTIR spectroscopy. After the ini-
tial evacuation at 673 K, all materials presented strong absorption
bands at 1476 and 1569 cm−1. The small separation between the
peaks and the fact that these bands resist evacuation up to 673 K
suggest that they arise from a bulk carbonate, rather than from
un-removed adsorbed carbon dioxide [32]. All spectra shown after
carbon dioxide chemisorption and subsequent evacuation are dif-
ference spectra relative to each material after evacuation at 673 K
and before carbon dioxide chemisorption.
100
80
60
40
20
0
Fig. 4 shows the spectra obtained for ␥-alumina and the nio-
bia/aluminas after carbon dioxide chemisorption and subsequent
evacuation at 323 K for 10 min. In the region of 1100–1900 cm−1
,
the most important bands are observed at 1230, 1441 and
1650 cm−1. Bands near these values have been assigned by Bal-
trusaitis et al. [33], respectively, to the ı4(COH), ꢁ3(OCO)s and
ꢁ2(OCO)a vibration modes of adsorbed bicarbonate. From isotopic
labeling experiments and quantum mechanical calculations, these
authors propose that carbon dioxide chemisorption on surface
hydroxyls leads to the formation of a bridge-bonded bicarbonate
species. The intensity of these bands decrease strongly, even for
a niobium loading corresponding to only 22.5% of the theoretical
monolayer, confirming that niobia titrates preferentially the basic
0
50
100
150
200
250
Time / min
Fig. 5. Kinetic curves for methanol dehydration (373 K, 8.9 kPa methanol initial
pressure, 10 mg catalyst). ᭹, ␥-Al2O3;
ꢂ, 20%NbAl.
, 5.9%NbAl; ꢀ, 11.1%NbAl; ꢁ, 15.9%NbAl;

A.S. Rocha et al. / Catalysis Today 192 (2012) 104–111
109
Table 4
Integrated peak areas for the 1230 cm−1 band (bicarbonate) of adsorbed CO2 normalized by catalyst mass.a
Temperature (K)
Catalyst
Alumina
5.9%NbAl
22.5
11.1%NbAl
45.0
20%NbAl
90.1
% Theoretical monolayer
0
(Ads CO2)b
323c
2.82
1.49
0.61
0.24
0.57(0.20)
0.24 (0.16)
0.07 (0.11)
0.02 (0.08)
0.41(0.15)
0.15 (0.1)
0.02 (0.04)
0.02 (0.07)
0.22(0.08)
0.06 (0.04)
0.01 (0.02)
373c
423c
0.003 (0.01)
a
Numbers between brackets represent the mass normalized peak area relative to that of the pure alumina under the same conditions.
In the presence of 0.133 kPa (1 Torr) CO2 at room temperature.
After evacuation for 10 min at the specified temperature.
b
c
Table 5
decrease in intensity as the niobium loading increases, in line with
the expectation that the surface basicity decreases as the coverage
of the alumina surface by niobia increases.
Initial reaction rates for methanol dehydration (373 K, 8.9 kPa methanol initial pres-
sure, 10 mg catalyst).
Reaction rates (mmol g−1 min−1
)
Ratioc
The intensity of the ı4(COH) band at 1230 cm−1 may be taken as
a measure of the concentration of basic hydroxyl groups at oxide
surfaces [32]. In Table 4 are reported the peak areas per unit sur-
face area for several catalysts both in the presence of 0.133 kPa
(1 Torr) CO2 and after evacuation at different temperatures. For
the niobia/aluminas, the values between brackets represent the
normalized peak areas as a fraction of the one obtained with
the pure alumina under the same conditions. In the presence of
0.133 kPa CO2, the amount adsorbed on the catalysts decrease with
increasing niobium loading and, on the catalyst containing a nio-
bia loading corresponding to 22.5% of the theoretical monolayer
(5.9%NbAl), the amount of adsorbed CO2 is already only 20% of
the amount adsorbed on the pure alumina, confirming that niobia
titrates preferentially the hydroxyl groups responsible for carbon
dioxide adsorption, i.e., the basic hydroxyl groups. With increasing
evacuation temperature, this effect is intensified, so that after evac-
uation at 423 K, the carbon dioxide amount remaining on 5.9%NbAl
is only 8% of the one remaining on the pure alumina under the same
conditions. A similar effect is observed with the higher niobium
loadings. This shows that the basic hydroxyls remaining on the
uncovered portion of the niobia/alumina surface are not only fewer
but also weaker than the ones originally present on the alumina.
Catalyst
a
b
Without CO2
With CO2
␥-Al2O3
14.1
5.2
8.1
4.3
3.1
5.2
3.1
4.1
3.4
1.6
0.37
0.60
0.51
0.71
0.52
5.9%NbAl
11.1%NbAl
15.9%NbAl
20%NbAl
a
Without added CO2.
b
CO2 added to the reactant in a 1:1 ratio to methanol.
Ratio of the two reaction rates.
c
and Merrill [35] have proposed that methanol dehydration on
alumina may occur through two paths, one involving reaction
between two adsorbed methoxides and the other involving the
reaction between an adsorbed methoxide and a methanol molecule
hydrogen-bonded to the surface, with the latter predominating at
temperatures below 553 K. This mechanism may be represented by
Scheme 1 below, where Al+ represents a Lewis acid site and OH a
basic surface hydroxyl group:
In this mechanism, alkoxide formation requires methanol dis-
sociative chemisorption on an acid–base pair. According to the
results in Table 2, deposition of 5.9% niobia does not strongly affect
concentration or strength of the Lewis acid sites. However, the
results in Table 4 demonstrate that basic hydroxyls are strongly
affected by niobia deposition. The strong decrease in activity par-
allels the decrease in basic hydroxyl concentration. On the other
hand, the results in Table 4 show that, after the 5× decrease upon
the addition of 5.9% niobia, the basicity decreases more slowly with
increasing niobia loading. The fact that the activity decreased by a
factor of about 3, while the concentration of basic hydroxyl groups
decreased by a factor of about 5 (cf. 4th row in Table 4) upon addi-
tion of 5.9 wt% niobia to the alumina, may be explained if basic
surface sites other than hydroxyls, such as oxide anions, may also
participate in the reaction. It should be noticed that bands that can
3.3. Gas phase methanol dehydration
The only reaction products observed in methanol dehydration
on alumina and niobia/aluminas were DME and water up to 573 K
reaction temperature. With the most active catalyst, which was the
pure alumina, at 523 K equilibrium conversion was reached after ca.
50 min reaction time and no secondary products could be observed
after 4 h reaction. In order to illustrate the type of data obtained
in this work, Fig. 5 shows kinetic curves for methanol dehydration
obtained at 523 K.
The methanol dehydration activities determined from the initial
rates of reaction for all catalysts, both in presence and in the absence
of added carbon dioxide in a 1:1 ratio to methanol, are presented
in Table 5. Without carbon dioxide addition, there is a nearly 60%
decrease in activity with respect to the pure alumina upon adding
5.9% niobia. However, in the series of niobia/aluminas a maximum
in activity occurs with 11.1%NbAl.
CH3OH + Al+-OH → CH3O-Al-OH2
+
CH3O-Al-OH2+ → CH3O-Al+ + H2O
The strong decrease in activity upon supporting niobia on the
alumina surface may be ascribed to titration of the strongest alu-
mina basic sites by niobia, as revealed by the adsorbed CO2 FTIR
spectra. Previously we demonstrated that activity for isopropanol
dehydration correlates linearly with the basic site concentra-
tion of pure and niobia-modified alumina, as measured by CO2
chemisorption. Both olefin and ether formation are influenced in
the same direction, but olefin formation is more affected. Schiffino
CH3O-Al+ + CH3OH---OH-Al+ → (CH3)2O---Al+-OH + Al+-OH
(CH3)2O---Al+-OH → (CH3)2O + Al+-OH
Scheme 1.

110
A.S. Rocha et al. / Catalysis Today 192 (2012) 104–111
Table 6
Rate of carbon monoxide uptake during syngas conversion on single and mixed
catalyst systems (553 K, 5.0 MPa, 2 g methanol synthesis catalyst, 4 g methanol dehy-
dration catalyst).
0.22
0.21
0.20
0.19
Dehydration catalyst
Consumption ratea
(×103/mol min−1 g−1
)
–
0.34
1.02
3.01
Al2O3
11.1% NbAl
a
Rate obtained considering only the weight of the methanol synthesis catalyst.
from the decrease in pressure in the gas burette, assuming that
the 2:1 H2/CO ratio remained constant in the burette throughout
the experiment.
It is clear from Fig. 6 that, without the dehydration com-
ponent, the rate of syngas consumption initially decreases, but
subsequently attains a nearly steady behavior in terms of syngas
consumption vs. time. When the pure ␥-alumina was used as the
dehydration component, an initial induction time is observed dur-
ing which the rate of reaction is very slow, but subsequently it
becomes much larger than the one obtained with the methanol
synthesis catalyst alone. With 11.1%NbAl as the dehydration com-
ponent, the induction period is significantly less pronounced and
the rate of syngas consumption is larger than obtained with ␥-
alumina as the dehydration component. Furthermore, from the
curvature of the remaining CO vs. time curves, there appears to
be a smaller decrease in the rate of carbon monoxide consumption
as the reaction proceeds.
From the slopes of the linear portion of the remaining CO vs.
time curves, catalyst activities were established and are shown
in Table 6. All of the activity values are referred to the mass of
the methanol synthesis catalyst used in the experiments. It is
found that the overall activity obtained with the niobia/alumina
as the methanol dehydration component is ca. three times the one
obtained with the pure ␥-alumina.
0
20
40
60
80
100
120
time / min
Fig. 6. Carbon monoxide uptake during syngas conversion (553 K, 5.0 MPa, 2 g
methanol synthesis catalyst, 4 g methanol dehydration catalyst). ꢀ, CuZnAl; ꢂ, CuZ-
nAl + ␥-Al2O3; ᭹, CuZnAl + 11.1%NbAl.
be assigned to surface carbonate, i.e., derived from surface oxide
rather than from hydroxide anions, were observed in our FTIR spec-
tra of adsorbed CO2.
The activity maximum observed with 11.1%NbAl appears to be
genuine, since it was also observed in a separate series of exper-
iments carried-out with carbon dioxide addition, as shown in the
third column in Table 5. This maximum may be related to the maxi-
mum in the concentration of Lewis acid sites able to retain adsorbed
pyridine up to 673 K, as shown in Table 2.
The initial rate results shown in the last column of Table 5 con-
firm the importance of the basic sites in methanol dehydration on
␥-alumina and the niobia/aluminas, as addition of carbon dioxide
in a 1:1 proportion to methanol strongly inhibited the reaction,
especially on alumina, where the rate in the presence of carbon
dioxide was approximately 1/3 of the rate in its absence. Although
carbon dioxide also inhibited the reaction on the niobia/aluminas,
the extent of inhibition was considerably smaller than in the case
of the pure alumina. This may be explained by the fact that nio-
bia titrates preferentially the strongest basic sites on the alumina
surface, that adsorb carbon dioxide more strongly.
The induction period observed with the mixed catalyst systems
may be related to the mechanism of methanol synthesis. Carbon
dioxide has been found to be the main reactant in carbon monoxide
hydrogenation [12,36,37]. Conversion of carbon monoxide actually
occurs mainly through the water gas shift (WGS) reaction, which
allows the maintenance of a steady-state concentration of carbon
dioxide in the system, as shown in Scheme 2 below:
This is expected to have an important impact in the use of these
catalysts in hybrid catalyst systems for the direct syngas to DME
conversion, as the main overall reaction occurring in this system is
reaction (4) below:
In the absence of added CO2, the shift reaction cannot be the path
for the initial CO2 formation. In this case, some CO could be con-
sumed in a Boudouard reaction (2CO → CO2 + C) to produce CO2,
as proposed previously [38]. This CO2 might be enough to initi-
ate methanol synthesis. Once the synthesis reaction is initiated,
the water produced in carbon dioxide hydrogenation may help to
retard carbon deposition by the Boudouard reaction via a carbon
reforming reaction (reaction (5) below):
3CO + 3H2 → CH3
O
CH3 + CO2
(4)
The large amount of carbon dioxide produced may lead to
strong competition with methanol for the basic sites involved in
methanol dehydration. It should be noticed, however, that, under
the conditions used here, the smaller inhibition by carbon dioxide
of methanol dehydration on the niobia/aluminas was not capable
of offsetting the decrease in activity derived from niobia deposition
on the alumina surface.
H2O + C → CO + H2
(5)
Thus, the induction period observed here with the mixed cata-
lyst systems may be explained by adsorption of water and carbon
dioxide by the dehydration component of the system. The smaller
induction period observed with the niobia/alumina catalyst would
then result from the inhibition of water and carbon dioxide adsorp-
tion by the surface niobia.
3.4. Direct synthesis of DME
Some preliminary results on the use of a mixed catalyst sys-
tem comprised of a commercial methanol synthesis catalyst and
11.1%NbAl, the most active NbAl methanol dehydration catalyst,
are shown in Fig. 6. The curves represent the amount of CO remain-
ing in a gas burette filled with syngas with a 2:1 H2/CO molar ratio
connected, via a pressure regulator, to the reactor containing the
catalyst system suspended in n-hexadecane, as described in detail
under Section 2.4. The amount of remaining CO was calculated
CO2 + 3H2 → CH3OH + H2O
CO + H2O → CO2 + H2
Scheme 2.

A.S. Rocha et al. / Catalysis Today 192 (2012) 104–111
111
In the direct synthesis of DME from syngas, the water produced
Acknowledgments
in the methanol dehydration reaction is responsible to drive the
shift reaction and to maintain a sufficient carbon dioxide concen-
tration level in the system. This explains the much larger overall
activity with the mixed catalyst system [12]. Furthermore, removal
of water by the WGS reaction helps to increase catalyst life [39].
It is interesting that the activity for DME production of the mixed
system CuZnAl plus 11.1NbAl was significantly higher than that
of the one with alumina, even though niobium addition causes a
reduction of the alumina activity for methanol dehydration. This is
likely due to the smaller inhibition of methanol dehydration by
carbon dioxide adsorption on the niobia/alumina catalysts than
on pure ␥-alumina, as reported in Table 5. Although in the gas
phase methanol dehydration tests reported here none of the nio-
bia/alumina catalysts was more active than the pure alumina, it
is conceivable that in the liquid phase and at the high pressures
used in the syngas to methanol experiments the inhibition factor
becomes more important than the activity factor.
Selectivity to carbon products (DME, methane and CO2) are very
similar for both mixed systems. Carbon dioxide and DME were pro-
duced in approximately 1:1 ratio, showing that the predominant
overall reaction is the aforementioned reaction (4). Methanol con-
stitutes 3–4% of the converted carbon monoxide. Therefore, niobia
addition to the alumina accelerates the whole reaction cycle, i.e.,
increased rate of methanol dehydration increases the production
of water, which in its turn increases the WGS reaction and conse-
quently the methanol synthesis reaction.
To PETROBRAS for financial support to this project. To Agência
Nacional do Petróleo, ANP, for a M.Sc. scholarship for A.M.F.
Appendix A. Supplementary data
Supplementary data associated with this article can be
found, in the online version, at http://dx.doi.org/10.1016/
j.cattod.2012.02.062.
References
[1] E.F. Sousa-Aguiar, L.G. Appel, C. Mota, Catalysis Today 101 (2005) 3–7.
[2] T.A. Semelsberger, R.L. Borup, H.L. Greene, Journal of Power Sources 156 (2006)
497–511.
[3] Q. Ge, Y. Huang, F. Qiu, S. Li, Applied Catalysis A 167 (1998) 23–30.
[4] R. Yang, X. Yu, Y. Zhang, W. Li, N. Tsubaki, Fuel 87 (2008) 443–450.
[5] J. Lee, K. Lee, S. Lee, Journal of Catalysis 144 (1993) 414–424.
[6] I. Wender, Fuel Processing Technology 48 (1996) 189–297.
[7] M. Xu, J.H. Lunsford, D.W. Goodman, A. Bhattacharyya, Applied Catalysis A 149
(1997) 289–301.
[8] Q. Sun, Y. Fu, H. Yang, A. Auroux, J. Shen, Journal of Molecular Catalysis A 275
(2007) 183–193.
[9] J. Khom-in, P. Praserthdam, J. Panpranot, O. Mekasuwandumrong, Catalysis
Communications 9 (2008) 1955–1958.
[10] J. Fei, Z. Hou, B. Zhu, H. Lou, X. Zheng, Applied Catalysis
49–54.
A 304 (2006)
[11] Y. Ohno, M. Yoshida, T. Shikada, O. Inokoshi, JFE Technical Report 8 (2006)
34–40.
[12] S. Lee, A. Sardesai, Topics in Catalysis 32 (2005) 197–207.
[13] J.-L. Li, X.-G. Zhang, T. Inui, Applied Catalysis A 147 (1996) 23–33.
[14] D. Mao, W. Yang, J. Xia, B. Zhang, Q. Song, Q. Chen, Journal of Catalysis 230
(2005) 140–149.
[15] D. Mao, W. Yang, J. Xia, B. Zhang, G. Lu, Journal of Molecular Catalysis A 250
(2006) 138–144.
[16] F.S. Ramos, A.M.D. de Farias, L.E.P. Borges, J.L. Monteiro, M.A. Fraga, E.F. Sousa-
Aguiar, L.G. Appel, Catalysis Today 101 (2005) 39–44.
[17] Q. Xu, T. Li, Y. Yan, Journal of Fuel Chemistry and Technology 36 (2008)
176–180.
[18] M. Abdel Rehim, A.C.B. dos Santos, V.L. Camorim, A.C. Faro Jr., Applied Catalysis
A 305 (2006) 211–218.
[19] L.J. Burcham, J. Datka, I.E. Wachs, Journal of Physical Chemistry B 103 (1999)
6015–6024.
[20] D. Bayot, B. Tinant, M. Devillers, Catalysis Today 78 (2003) 439–447.
[21] K. Asakura, Y. Iwasawa, Chemistry and Technology of Fuels and Oils 15 (1986)
859–862.
[22] J.B. Peri, Journal of Physical Chemistry 809 (1965) 220–230.
[23] H. Knözinger, P. Ratnasamy, Catalysis Review 17 (1978) 31–70.
[24] G. Busca, V. Lorenzelli, V.S. Escribano, R. Guidetti, Journal of Catalysis 131 (1991)
167–177.
[25] C. Morterra, G. Magnacca, Catalysis Today 27 (1996) 497–532.
[26] A.S. Rocha, A.M.S. Forrester, M.H.C. de la Cruz, C.T. da Silva, E.R. Lachter, Catalysis
Communications 9 (2008) 1959–1965.
[27] C. Emeis, Journal of Catalysis 141 (1993) 347–354.
[28] K. Tanaka, A. Osaki, Journal of Catalysis 8 (1967) 1–7.
[29] G. Connell, J.A. Dumesic, Journal of Catalysis 105 (1987) 285–298.
[30] C.L.T. da Silva, V.L.L. Camorim, J.L. Zotin, M.L.R. Duarte Pereira, A.D.C. Faro,
Catalysis Today 57 (2000) 209–217.
[31] M.I. Zaki, M.A. Hasan, F.A. Al-Sagheer, Colloids and Surfaces 190 (2001)
261–274.
[32] J.C. Lavalley, Catalysis Today 27 (1996) 377–401.
[33] J. Baltrusaitis, J.H. Jensen, V.H. Grassian, Journal of Physical Chemistry B 110
(2006) 12005–12016.
[34] A.M. Turek, I.E. Wachs, E. DeCanio, Journal of Physical Chemistry 96 (1992)
5000–5007.
[35] R.S. Schiffino, R.P. Merrill, Journal of Physical Chemistry 97 (1993)
6425–6435.
[36] S. Fujita, M. Usui, H. Ito, N. Takezawa, Journal of Catalysis 157 (1995)
403–413.
[37] Q. Sun, C.-W. Liu, W. Pan, Q.-M. Zhu, J.-F. Deng, Applied Catalysis A 171 (1998)
301–308.
4. Conclusions
The use of a peroxo-oxalo niobium complex allowed the prepa-
ration of alumina-supported niobium oxide containing up to 90% of
a theoretical monolayer in a single incipient wetness impregnation
step. The niobia on these materials was well dispersed, probably
as a monolayer at least up to a niobia loading of nearly 60% of the
theoretical monolayer.
No systematic trend in total Lewis acidity could be observed
as a function of niobium loading, but a maximum in Lewis acid
strength was found at a niobia loading around 45% of the theoretical
monolayer. Deposition of niobia resulted in a strong decrease in the
surface concentration of basic hydroxyl groups and the most basic
ones were preferentially eliminated.
Deposition of niobia also resulted in a strong decrease in the
activity for methanol dehydration, paralleling the effect on basicity,
but an activity maximum was found within the series of nio-
bia/aluminas, which may be related to the Lewis acidity results.
This shows that relatively strong acid–base pairs are required
for methanol dehydration. On the other hand, the elimination of
strongly basic hydroxyls by niobia deposition markedly decreased
the inhibiting effect of carbon dioxide on the methanol dehydration
activity.
In the direct production of DME from a mixture of carbon
monoxide and hydrogen using a mixed system comprised of a
methanol synthesis CuZnAl component and a dehydration com-
ponent, the expected effect of the addition of the dehydration
component on activity and selectivity was observed, but the overall
activity was significantly larger with a niobia/alumina as the dehy-
dration component than with the pure alumina. This is probably the
result of the smaller inhibition of methanol dehydration by carbon
dioxide, rendering more efficient the reaction cycle comprised of
methanol synthesis, methanol dehydration and the WGS reaction.
[38] K. Klier, V. Chatikavanij, R.G. Herman, G.W. Simmons, Journal of Catalysis 74
(1982) 343–360.
[39] T. Ogawa, N. Inoue, T. Shikada, Y. Ohno, Journal of Natural Gas Chemistry 12
(2003) 219–227.