Sustainable production of acrolein: Gas-phase dehydration of glycerol over Nb2O5 catalyst

Article abstract of DOI:10.1016/j.jcat.2007.06.016

Gas-phase dehydration of glycerol to produce acrolein was investigated at 315 °C over Nb₂O₅ catalysts calcined in the temperature range of 350-700 °C. The catalysts were characterized by nitrogen physisorption, TG-DTA, XRD, and n-butylamine titration using Hammett indicators to gain insight into the effect of calcination temperature on catalyst texture, crystal structure, and acidity. Calcination at 350 and 400 °C produced amorphous Nb₂O₅ catalysts that exhibit significantly higher fractions of strong acid sites at - 8.2 ≤ H₀ ≤ - 3.0 (H₀ being the Hammett acidity function) than the crystallized Nb₂O₅ samples obtained by calcination at or above 500 °C. Glycerol conversion and acrolein selectivity of the Nb₂O₅ catalysts were dependent of the fraction of strong acid sites (- 8.2 ≤ H₀ ≤ - 3.0). The amorphous catalyst prepared by the calcination at 400 °C, having the highest fraction of acid sites at - 8.2 ≤ H₀ ≤ - 3.0, showed the highest mass specific activity and acrolein selectivity (51 mol%). The other samples, having a higher fraction of either stronger (H₀ ≤ - 8.2) or weaker acid sites (- 3.0 ≤ H₀ ≤ 6.8), were less effective for glycerol dehydration and formation of the desired acrolein.

Full text of DOI:10.1016/j.jcat.2007.06.016

Journal of Catalysis 250 (2007) 342–349
www.elsevier.com/locate/jcat
Sustainable production of acrolein: Gas-phase dehydration of glycerol over
Nb2O5 catalyst
Song-Hai Chai, Hao-Peng Wang, Yu Liang, Bo-Qing Xu ∗
Innovative Catalysis Program, Key Lab of Organic Optoelectronics & Molecular Engineering, Department of Chemistry, Tsinghua University,
Beijing 100084, PR China
Received 28 February 2007; revised 10 May 2007; accepted 13 June 2007
Available online 6 August 2007
Abstract
◦
Gas-phase dehydration of glycerol to produce acrolein was investigated at 315 C over Nb O catalysts calcined in the temperature range of
2
5
◦
350–700 C. The catalysts were characterized by nitrogen physisorption, TG-DTA, XRD, and n-butylamine titration using Hammett indicators to
gain insight into the effect of calcination temperature on catalyst texture, crystal structure, and acidity. Calcination at 350 and 400 C produced
◦
amorphous Nb O catalysts that exhibit significantly higher fractions of strong acid sites at −8.2 ꢀ H ꢀ −3.0 (H being the Hammett acidity
2
5
0
0
◦
function) than the crystallized Nb O samples obtained by calcination at or above 500 C. Glycerol conversion and acrolein selectivity of the
2
5
Nb O catalysts were dependent of the fraction of strong acid sites (−8.2 ꢀ H ꢀ −3.0). The amorphous catalyst prepared by the calcination
2
5
0
◦
at 400 C, having the highest fraction of acid sites at −8.2 ꢀ H ꢀ −3.0, showed the highest mass specific activity and acrolein selectivity
0
(51 mol%). The other samples, having a higher fraction of either stronger (H ꢀ −8.2) or weaker acid sites (−3.0 ꢀ H ꢀ 6.8), were less
0
0
effective for glycerol dehydration and formation of the desired acrolein.
© 2007 Elsevier Inc. All rights reserved.
Keywords: Niobium oxide (Nb O ); Solid acid; Glycerol (glycerin); Acrolein; Alcohol dehydration; Sustainable technology
2
5
1. Introduction
(SPA), have been tested for the dehydration of glycerol in either
gaseous or liquid phases [9–11]. The dehydration of glycerol
also has been investigated in sub-supercritical and supercritical
water (250–390 ◦C and 25–35 MPa) in the presence of low con-
centrations of liquid acids or salts [12–14]. Glycerol is usually
produced as a mixture with water. The direct use of glycerol in
water is advantageous over pure glycerol for the production of
acrolein, but a highly water-tolerant solid acid catalyst would
be developed for the purpose.
Niobium oxide (Nb2O5) has been used as a water-tolerant
solid acid catalyst for various water-involving reactions, such as
esterification, hydrolysis, dehydration, and hydration [15–18].
A key to the acidic and catalytic properties of Nb2O5 is its calci-
nation or pretreatment temperature [19]. In this paper we report
our investigation into the catalytic behavior of Nb2O5 catalysts
prepared by varying the calcination temperature of a hydrated
niobium oxide (Nb2O5·nH2O) for the gas-phase dehydration
of aqueous glycerol. Our data provide a basis for correlating
the catalyst acidity with the catalytic performance in glycerol
dehydration.
The use of renewable resources, such as biomass, as feed-
stock for the production of fuels and chemicals has become
an increasingly important focus in energy-related catalysis, be-
cause fossil resources will be exhausted in a few decades [1–4].
Glycerol is a main byproduct in natural triglyceride methanoly-
sis for biodiesel production. In recent years, the increasing use
and production of biodiesel has resulted in an increase of glyc-
erol production and a price decline, which makes glycerol a
particularly attractive molecule (building block) for the synthe-
sis of other valuable chemical products [5–8]. Catalytic conver-
sion of glycerol to acrolein by a double-dehydration reaction
(Fig. 1) could be an important route for using glycerol resources
and could offer a sustainable alternative to the present acrolein
technology based on propylene. Various solid acid catalysts, in-
cluding sulfates, phosphates, zeolites, and solid phosphoric acid
*
Corresponding author. Fax: +86 10 62792122.
E-mail address: bqxu@mail.tsinghua.edu.cn (B.-Q. Xu).
0021-9517/$ – see front matter © 2007 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2007.06.016

S.-H. Chai et al. / Journal of Catalysis 250 (2007) 342–349
343
Fig. 1. Schematic representation of the double dehydration of glycerol to acrolein.
2. Experimental
quartz reactor (9 mm i.d.) using 0.63 ml of catalyst. The cat-
alyst mass in the reactor varied with the catalyst calcination
temperature (T ) and were 0.56 g (T = 350◦C), 0.57 g (400 ◦C),
0.61 g (500 ◦C), 0.73 g (600 ◦C), and 0.84 g (700 ◦C). Be-
fore the reaction, the catalysts were pretreated at 315 ◦C for
1.5 h in flowing dry N2 (30 ml min−1). The reaction feed, an
aqueous solution containing 36.2 wt% glycerol (molar ratio
glycerol/water = 1/9), was fed into the reactor by a micro-pump
at a space velocity (GHSV) of glycerol of 80 h−1. The reaction
products were condensed in an ice–water trap and collected
hourly for analysis on a HP6890 gas chromatograph equipped
with a HiCap CBP20-S25-050 (Shimadzu) capillary column
(0.32 mm i.d., 25 m long) and a flame ionization detector. The
reaction was usually conducted for 10 h, and the condensed
products during the first hour of the reaction were abandoned
due to poor material balance. Conversion of glycerol (GL) and
product selectivity were obtained according to the following
calculations:
2.1. Catalyst preparation
Nb2O5 catalysts were prepared by calcination of a hydrous
niobium oxide (Nb2O5·nH2O), provided as a HY-340 product
from CBMM (Brazil). The calcination was done under flow-
ing air (80 ml min−1) in a horizontal tubular oven (50 mm i.d.)
using 3.0 g of Nb2O5·nH2O dispersed in a quartz boat (ca.
15 ml) placed in the middle of the oven. The heating rate was
10 ◦C/min, and the calcination continued for 4 h at each of the
selected temperatures (350–700 ◦C). The catalyst powders were
pressed, crushed, and sieved to 20–40 mesh before use.
2.2. Characterizations
BET surface areas, pore volumes, and average pore di-
ameters of Nb2O5 samples were derived from the nitrogen
adsorption–desorption isotherms at −196 ◦C measured on a
Micromeritics ASAP 2010C instrument. The samples were de-
hydrated under vacuum at 200 ◦C for 5 h before the measure-
ment. The average pore diameter data were calculated accord-
ing to the Barrett–Joyner–Halenda (BJH) method.
The crystal structures of Nb2O5 samples were character-
ized by powder X-ray diffraction (XRD) using a Bruker D8
Advance X-ray diffractometer with a Ni-filtered CuKα (λ =
0.15406 nm) radiation source at 40 kV and 40 mA.
moles of GL reacted
GL conversion (%) =
× 100
moles of GL in the feed
and
product selectivity (mol%)
moles of carbon in a product defined
=
× 100.
moles of carbon in GL reacted
3. Results
Thermal analysis (TG-DTA) of Nb2O5·nH2O, as well as
temperature-programmed oxidation (TPO) measurements of
the used catalysts, were conducted on a Mettler-Toledo TG/
SDTA 851 thermal analyzer. The sample was placed in an α-
Al2O3 cumber and heated in flowing air (50 ml min−1) from
3.1. Nitrogen physisorption
Fig. 2 shows the nitrogen physisorption isotherms of the cal-
cined Nb2O5 samples. None of these isotherms falls into the six
types of physisorption isotherms defined by IUPAC; hysteresis
loops were observed, which are usually associated with capil-
lary condensation in mesopores [22]. For the samples calcined
at 350, 400, and 500 ◦C, the hysteresis loops at relative nitro-
gen pressure (P/P0) below 0.85 resembled type H2, but those
at higher pressures were similar to type H3. Nb2O5 samples
calcined at 600 and 700 ◦C also displayed type H3 hysteresis
loops at P/P0 > 0.80. The type H2 hysteresis loop, associated
with materials with complex interconnected networks of pores
of different sizes and shapes, is usually taken as the indication
for the presence of pores with narrow mouths (i.e., ink bottle
pores) or channel-like pores of relatively uniform size [22]. The
type H3 loop characterizes slit-shaped pores formed from ag-
gregates of plate-like particles [22].
room temperature to 800 ◦C at a rate of 20 ◦C min−1
.
The atomic H:C ratio in the carbon deposits was determined
by elemental analysis using an EAI CE-440 elemental analyzer.
The samples were dried overnight at 110 ◦C before the mea-
surements.
Measurements of catalyst acidity (acid strength and amount)
were based on the n-butylamine titration method using various
Hammett indicators, including anthraquinone (pKa = −8.2),
dicinnamalacetone (pKa = −3.0), and neutral red (pKa = 6.8),
as described previously [19–21]. Acid strength was expressed
by Hammett acidity function (H0) that was scaled by pKa val-
ues of the indicators. Before the measurement, the samples
were formulated to 100–180 mesh and pretreated at 315 ◦C for
4 h in flowing dry N2.
Textural properties of the Nb2O5 samples derived from the
nitrogen physisorption isotherms are given in Table 1. Increas-
ing the calcination temperature resulted in continuous decreases
in the sample surface area (from 115 to 7 m2 g−1) and pore
2.3. Catalytic reaction
The gas-phase dehydration of glycerol was conducted at
315 ◦C under atmospheric pressure in a vertical fixed-bed

344
S.-H. Chai et al. / Journal of Catalysis 250 (2007) 342–349
Fig. 2. Nitrogen physisorption isotherms of Nb O calcined at (!) 350, (2)
2
5
◦
400, (1) 500, (") 600, and (P) 700 C.
Table 1
Texture properties of calcined Nb O catalysts
2
5
a
Calcination BET surface Pore volume Average pore diameter (nm)
3
−1
temperature area
(cm g
)
BJH adsorption BJH desorption
◦
2
−1
)
( C)
(m g
350
400
500
600
700
115
99
42
12
7
0.14
0.13
0.11
0.05
0.03
5.1
5.6
10.3
21.1
20.7
4.7
5.0
8.0
16.6
22.0
Fig. 3. (A) TG-DTG and (B) DTA curves of the hydrated niobium oxide.
a
Measured at P/P = 0.995.
0
volume (from 0.14 to 0.03 cm3 g−1), which were especially re-
markable when the calcination was done at temperatures above
400 ◦C. The average pore diameter, ca. 5 nm for the samples
calcined at 350 and 400 ◦C, also increased remarkably after
calcination at 500–700 ◦C. The surface area data agree well
with those reported by Tanabe et al. [23], whose data were
126 m2 g−1 after the evacuation at 300 ◦C and 42 m2 g−1 af-
ter evacuation at 500 ◦C.
3.2. TG-DTA and XRD
Fig. 3 shows the (A) TG-DTG and (B) DTA curves of
the Nb2O5·nH2O precursor. The weight loss on the TG curve
closely matched the endothermic feature on the DTA curve,
which should suggest an endothermic dehydration process.
Most of the crystallization water in the hydrated precursor was
removed by heating up to 350 ◦C; the H2O/Nb2O5 ratio was ca.
3.3 (molar) according to the TG-DTG measurement (Fig. 3A).
The DTA curve also showed a sharp exothermic peak at ca.
585 ◦C (Fig. 3B). This exothermic feature, which is similar to
the one at ca. 570 ◦C in the earlier report of Tanabe et al. [23],
characterizes a transformation of the sample from its amor-
phous state to crystalline Nb2O5.
Fig. 4. XRD patterns of Nb O samples calcined at (a) 350, (b) 400, (c) 500,
(d) 600, and (e) 700 C.
2
5
◦
X-ray diffraction detected with these two samples. Clear dif-
fractions at 2θ = 22.6◦, 28.4◦, 36.7◦, and 46.2◦ for pseudo-
hexagonal Nb2O5 crystals (TT phase) [17] appeared for the
samples calcined at 500 and 600 ◦C. The peaks at 2θ = 28.4◦
and 36.7◦ were found to split into two peaks when the cal-
cination temperature was increased to 700 ◦C. This splitting
indicates a transformation of the pseudohexagonal TT phase to
the orthorhombic phase (T phase) of Nb2O5 during the calci-
nation at 700 ◦C [17]. The pseudohexagonal TT phase has a
lower crystallinity and can be considered a modification of the
orthorhombic T phase [17].
Fig. 4 shows the effect of calcination temperature on the
XRD patterns of Nb2O5 catalysts. The samples calcined at 350
and 400 ◦C appeared as amorphous materials, with no distinct

S.-H. Chai et al. / Journal of Catalysis 250 (2007) 342–349
345
3.3. Catalyst acidity
only over the sample calcined at 350 ◦C. The fraction of strong
acid sites (−8.2 ꢀ H0 ꢀ −3.0) was as high as 0.6 (i.e., 60% of
the total acidity) over the catalyst calcined at 400 ◦C. The frac-
tion of medium-strong to weak acid sites (−3.0 ꢀ H0 ꢀ 6.8)
increased with increasing catalyst calcination temperature.
Based on their acid strength, acid sites at the catalyst sur-
face are divided into the following three groups: medium-strong
and weak (−3.0 ꢀ H0 ꢀ 6.8), strong (−8.2 ꢀ H0 ꢀ −3.0),
and very strong (H0 ꢀ −8.2). The total amount of acid sites
in the three groups is defined as the total acidity (H0 ꢀ 6.8).
Fig. 5 shows the effect of calcination temperature on the acidity
and acid site distribution of the Nb2O5 catalysts. The high-
est acid strength of Nb2O5 samples decreased with increas-
ing calcination temperature (T ): H0 ꢀ −8.2 for T = 350◦C,
−8.2 ꢀ H0 ꢀ −3.0 for T = 400–600◦C, and −3.0 ꢀ H0 ꢀ
6.8 for T = 700◦C. The increase in the calcination tempera-
ture also reduced the catalyst acidity (amounts) in the differ-
ent acid strength ranges. These acidity data (acid strength and
amount) are quite similar to those obtained by Tanabe et al. on
Nb2O5 samples pretreated/calcined at comparable temperatures
[23,24].
3.4. Catalytic reaction
The catalytic dehydration of glycerol over the present
Nb2O5 catalysts was characterized by a decrease in the con-
version of glycerol with the reaction time on stream (TOS),
along with an induction period to reach a stable selectivity
for acrolein production. Examples are shown in Fig. 7. With
the catalysts calcined at 400 and 700 ◦C, glycerol conversion
decreased with TOS but the selectivity for acrolein increased
during the first 3–6 h and stabilized at longer TOS. Thus, cat-
alytic reaction data at TOS of 1–2 and 9–10 h are presented in
Fig. 8 to show the effect of calcination temperature on the cat-
alytic performance of Nb2O5. The catalyst calcined at 400 ◦C
exhibited the highest selectivity (up to 51 mol% by normal-
ization to glycerol reacted (Fig. 8B)) for acrolein production;
despite this, the glycerol conversion based on the catalyst vol-
ume (0.63 ml) was slightly lower over this catalyst than over
the catalyst calcined at 500 ◦C (Fig. 8A).
Fig. 6 gives the fraction of acid sites with different acid
strengths. Very strong acid sites (H0 ꢀ −8.2) were observed
The effect of calcination temperature on the stabilized prod-
uct distribution (i.e., 9–10 h TOS) is shown in Table 2. In
addition to the desirable acrolein, the main product from the
Fig. 5. Effect of the calcination temperature on acidity of Nb O samples at
2
5
(!) H ꢀ −8.2, (P) −3.0 ꢀ H ꢀ 6.8, and (2) −8.2 ꢀ H ꢀ −3.0. The “E”
0
0
0
data give the total acidity at H ꢀ 6.8.
0
Fig. 6. Effect of the calcination temperature on the fraction of acid sites at (!)
ꢀ −8.2, (2) −8.2 ꢀ H ꢀ −3.0, and (P) −3.0 ꢀ H ꢀ 6.8 over Nb O
H
Fig. 7. Time courses of (2) glycerol conversion and (1) acrolein selectivity
0
0
0
2 5
◦
catalyst.
over Nb O catalysts calcined at (A) 400 and (B) 700 C.
2
5

346
S.-H. Chai et al. / Journal of Catalysis 250 (2007) 342–349
nation of the used catalyst in flowing air was always effective,
a simple oxidation treatment with 20 vol% O2 in N2 at the re-
action temperature (315 ◦C) was found to be sufficient for a full
regeneration of the reacted Nb2O5 catalysts, irrespective of the
calcination temperature used in catalyst preparation.
3.4.1. Characterization of carbon deposits
Considerable carbon deposits were formed on the Nb2O5
catalysts during the catalytic reaction. TPO characterization of
the used catalysts (at 10 h TOS) was conducted using flowing
air in the TG-DTA mode; the results are compared in Fig. 9.
The weight change below 250 ◦C on the TG curves (Fig. 9A)
was due to the elimination of adsorbed water, because little en-
dothermic features were observed on the corresponding DTA
curves (Fig. 9B). The weight loss on the TG curves at 300–
500 ◦C was accompanied by a broad, strong exothermic peak
on the DTA curves in the same temperature range, due to burn-
off of the carbon deposits by oxidation.
The DTA curves of the used catalysts prepared by calcina-
tion at 350 and 400 ◦C (Fig. 9B) also showed an additional
small exothermic peak at ca. 585 ◦C but with no response on
the corresponding TG curves (Fig. 9A). This peak, the position
of which closely matched the exothermic feature on the DTA
curve of the catalyst precursor (Fig. 3B), indicated a crystal-
lization of the amorphous Nb2O5 catalysts (Fig. 4).
The amounts of carbon deposits measured from the TG
curves are plotted in Fig. 10 as a function of the catalyst calci-
nation temperature. Whereas the general trend is that catalysts
calcined at lower temperatures coked more severely than cat-
alysts calcined at higher temperatures, carbon deposition was
significantly heavier on the amorphous catalysts.
We also measured the atomic H/C ratio in the carbon de-
posits with elemental analysis. The measured H/C ratio ap-
peared to be similar (H/C = 0.50–0.53) for the deposits formed
over the catalysts calcined at 400 and 500 ◦C. Separate XRD
characterization of the used catalysts detected no signal for
graphitic carbon, suggesting that the carbon deposits existed as
amorphous carbon species on the catalyst surface.
Fig. 8. Effect of the catalyst calcination temperature on (A) glycerol conversion
and (B) acrolein selectivity at (") TOS = 1–2 h and (Q) TOS = 9–10 h.
double dehydration of glycerol, 1-hydroxylacetone from mono
dehydration appeared as the main byproduct. The selectivity of
1-hydroxylacetone increased from 10 to 20 mol% with increas-
ing calcination temperature. Other byproducts detected were
acetaldehyde, propionaldehyde, acetone, and allyl alcohol, all
of which with selectivities usually below 5 mol%. Moreover,
quite a large amount of complex compounds (27–45 mol%
by normalization to glycerol reacted) remained as unidentified
products, possibly formed by secondary reactions among prod-
ucts or between product and the reaction feed. It is noteworthy
that the selectivity for unidentified products was the lowest
(27 mol%) over the most selective catalyst for acrolein pro-
duction. In other words, the sample calcined at 400 ◦C was the
best-performing Nb2O5 catalyst for the production of acrolein
from glycerol.
4. Discussion
Our data show that Nb2O5 catalysts prepared by calcination
of a hydrated niobium oxide (Nb2O5·nH2O) are effective for
the gas-phase dehydration of glycerol to produce acrolein. Cat-
alyst performance in the dehydration reaction is significantly
affected by the calcination temperature of the Nb2O5 catalyst.
Attempts were made to regenerate the reacted Nb2O5 cat-
alysts by oxidation at elevated temperatures. Whereas recalci-
Table 2
Effect of calcination temperature on product distribution over Nb O catalysts (TOS = 9–10 h)
2
5
Calcination tem-
perature ( C)
Conversion
(%)
Product selectivity (mol%)
◦
a
Acrolein
Acetaldehyde
Propionaldehyde
Acetone
Allyl alcohol
1-Hydroxylacetone
Others
350
400
500
600
700
75
88
92
76
32
47
51
35
33
28
4.7
4.1
5.3
2.7
1.8
2.9
2.4
3.2
1.1
0.4
3.1
2.4
3.2
1.1
0.5
1.3
1.4
2.3
2.9
5.4
10
12
14
21
19
31.0
26.7
37.0
38.2
44.9
a
Selectivity for others (mol%) = 100 − total selectivity for all products identified.

S.-H. Chai et al. / Journal of Catalysis 250 (2007) 342–349
347
Fig. 11. Areal catalytic rates of Nb O catalysts for (P) glycerol consumption
2
5
and (2) acrolein formation at TOS = 9–10 h.
tallized catalysts (Figs. 5 and 6). These results clearly indicate
that the stronger acid sites are reduced more severely than the
weaker ones on Nb2O5 crystallization during high-temperature
calcination. In other words, the amorphous structure seems to
be crucial for maintaining a strong acidity of the Nb2O5 cata-
lyst [23].
The reaction data given in Table 2 and Figs. 7 and 8 were
based on the catalyst volume (i.e., 0.63 ml) used in our experi-
mental study of the reaction. The areal catalytic rates based on
glycerol consumption and acrolein formation at TOS of 9–10 h
are shown in Fig. 11. These rates, with distinct maxima on the
catalyst calcined at 600 ◦C, appear to have no clear correlation
with catalyst acidity, probably due to a disturbance of the se-
vere catalyst coking. We then made an attempt to normalize the
catalytic rates on the catalyst mass. The mass-specific catalytic
rates for glycerol consumption and acrolein formation, given in
Fig. 12, are better correlated with the fraction of acidity in the
range of −8.2 ꢀ H0 ꢀ −3.0 (Fig. 6). Thus, the amorphous cat-
alyst prepared by calcination at 400 ◦C exhibited the highest ef-
ficiency for acrolein production in the dehydration of glycerol,
which means that the strong acid sites at −8.2 ꢀ H0 ꢀ −3.0
more effectively catalyzed the selective dehydration of glyc-
erol for acrolein production compared with the stronger (H0 ꢀ
−8.2) or weaker acid sites (−3.0 ꢀ H0 ꢀ 6.8). The strong acid
sites of −8.2 ꢀ H0 ꢀ −3.0 on supported H3PO4 catalysts have
been claimed to be effective acid sites for the same reaction in
a U.S. patent [10].
Fig. 9. (A) TG and (B) DTA curves of the used Nb O catalysts calcined at
(a) 350, (b) 400, (c) 500, (d) 600, and (e) 700 C.
2
5
◦
Fig. 10. Effect of the catalyst calcination temperature on the amount of carbon
deposits over the used Nb O catalysts (TOS = 9–10 h).
2
5
The foregoing discussion does not contain any informa-
tion on the dependence of acrolein production on the nature
of acid sites, because the amine titration acidity using Ham-
mett indicators can not distinguish Brønsted and Lewis acid
sites. IR spectra of pyridine adsorbed on Nb2O5 samples pre-
treated/evacuated at 100–500 ◦C were measured by Tanabe
et al. [23]. Brønsted as well as Lewis acid sites were detected
over the samples evacuated at temperatures below 500 ◦C. The
Through proper selection of the calcination temperature for cat-
alyst preparation, the desirable acrolein product can be obtained
at 315 ◦C with a stable selectivity >50 mol%.
The catalysts prepared by calcination at 350 and 400 ◦C were
amorphous, as indicated by the TG-DTA and XRD data (Figs. 3
and 4), and also demonstrated the highest surface areas and pore
volumes. The catalysts prepared by calcination at higher tem-
peratures (500–700 ◦C) were crystallized and exhibited quite
low surface areas. The amorphous Nb2O5 catalysts also showed
significantly higher acidity and fractions of strong acid sites
in the range of −8.2 ꢀ H0 ꢀ −3.0 compared with the crys-
absorption associated with Brønsted acid sites (ca. 1540 cm−1
)
appeared to decrease remarkably with increasing evacuation
temperature and became invisible when the evacuation was
done at 500 ◦C. The absorption associated with Lewis acid sites

348
S.-H. Chai et al. / Journal of Catalysis 250 (2007) 342–349
This conclusion is further supported by our investigation of the
dependence of acrolein production on catalyst acid–base prop-
erties over a wide variety of solid acids and bases [27].
The amorphous carbon deposits formed on Nb2O5 cat-
alysts could result from consecutive reactions of the prod-
ucts or side reactions between the reactant with such product
molecules as acrolein, acetaldehyde, propionaldehyde, and 1-
hydroxylacetone, and unidentified byproducts as well [27]. The
quite low atomic H/C ratios (H/C = 0.50–0.53) from our ele-
mental analysis of the coked Nb2O5 catalysts suggest that the
carbon deposits are polyaromatic-like surface species [26]. This
seems quite unusual because in the transformation of hydro-
carbons over solid acid catalysts, polyaromatic carbon deposits
with similar H/C ratios are usually formed in pseudographitic
phase at much higher temperatures (above 420 ◦C) [26]. Nev-
ertheless, it is noteworthy that a simple switch of the coked
Nb2O5 catalysts to a flow of 20 vol% O2/N2 at the reaction
temperature (315 ◦C) was sufficient for a full regeneration of
the deactivated Nb2O5 catalysts. This property could be impor-
tant from the standpoint of potential applications.
Fig. 12. Mass specific catalytic rates of Nb O catalysts for (P) glycerol con-
sumption and (2) acrolein formation at TOS = 9–10 h.
2
5
(1446–1448 cm−1) was still distinct on the sample pretreated
at 500 ◦C, however. The ratio of Brønsted to Lewis acid sites
(B/L), defined as the intensity ratio of the IR absorption band at
ca. 1540 cm−1 (Brønsted) to that at 1446–1448 cm−1 (Lewis),
decreased with increasing evacuation temperature [23]. The in-
formation is applicable to our study because the surface area,
XRD, and n-butylamine titration data of our samples (Table 1,
Figs. 3 and 5) were similar to those of Tanabe et al. on the sam-
ples subjected to similar pretreatments [19,23,24].
5. Conclusion
Our findings demonstrate that Nb2O5 catalysts prepared by
calcination of hydrated niobium oxide are effective for the gas-
phase dehydration of glycerol to produce acrolein. The catalyst
performance for the dehydration reaction is significantly af-
fected by the catalyst calcination temperature that induces the
changes in surface acidity and crystallization of Nb2O5. The
calcinations at 350 and 400 ◦C produce amorphous Nb2O5 cat-
alysts that exhibit significantly higher acidity and fraction of
strong acid sites at −8.2 ꢀ H0 ꢀ −3.0 compared with the crys-
tallized Nb2O5 samples obtained by calcination at 500–700 ◦C.
The amorphous catalyst calcined at 400 ◦C shows the highest
fraction of acid sites at −8.2 ꢀ H0 ꢀ −3.0 and gives the highest
catalytic efficiency for the formation of acrolein. Deactivation
was observed for all of the Nb2O5 catalysts, but a simple treat-
ment with flowing air at the reaction temperature was found to
be sufficient to regenerate the deactivated catalysts to their orig-
inal activity.
Thus, the percentage of Brønsted acid sites at the surface
of the amorphous Nb2O5 samples obtained by calcination at
350 and 400 ◦C in the present study was higher than that at
the surface of the crystallized samples prepared at higher calci-
nation temperatures. The better catalytic performance of these
amorphous Nb2O5 catalysts could imply that Brønsted acid
sites can be advantageous over Lewis acid sites for acrolein
production from glycerol dehydration. However, this point re-
mains to be confirmed with in situ discrimination of the active
acid sites in the dehydration reaction, because the assessment
of Brønsted and Lewis acid sites was based on well-dehydrated
Nb2O5 samples [23]. It is not impossible that under the reac-
tion conditions, some percentage of the Lewis acid sites on
the calcined/dehydrated Nb2O5 catalysts could be converted
to Brønsted ones by reaction with H2O molecules [25], be-
cause the dehydration reaction of glycerol in the present study
was conducted in the presence of a large excess of H2O mole-
cules (a molar water/glycerol ratio of 9!). According to Tanabe
et al. [23], Brønsted acid sites of niobic acid or hydrated nio-
bium oxide are much more active than Lewis acid sites for the
isomerization of 1-butene, but those Brønsted acid sites elimi-
nated by evacuation at 300 ◦C can be quantitatively regenerated
with suitable rehydration. However, the regeneration of Brøn-
sted acid sites becomes impossible once the evacuation temper-
ature is increased to 500 ◦C. Tanabe et al. [23] explained the
irreversible loss of Brønsted acid sites by the irreversible for-
mation of the TT phase during the high-temperature (500 ◦C)
treatment. Nevertheless, our present data show that Brønsted
acid sites can be superior to Lewis acid sites for the formation
of acrolein in the dehydration of glycerol over Nb2O5 catalysts.
Acknowledgments
We thank CBMM (Brazil) for kindly providing us with the
Nb2O5·nH2O (HY 340) sample. This work was partly sup-
ported by NSF China (Grant 20590362).
References
[1] D.L. Klass, Biomass for Renewable Energy, Fuels and Chemicals, Acad-
emic Press, San Diego, 1998, p. 1.
[2] G. Centi, S. Perathoner, Catal. Today 77 (2003) 287.
[3] H. Van Bekkum, P. Gallezot, Top. Catal. 27 (2004) 1.
[4] G.W. Huber, J.A. Dumesic, Catal. Today 111 (2006) 119.
[5] I. Diaz, C. Marquez-Alvarez, F. Mohino, J. Perez-Pariente, E. Sastre,
J. Catal. 193 (2000) 295.
[6] T. Miyazawa, Y. Kusunoki, K. Kunimori, K. Tomishige, J. Catal. 240
(2006) 213.
[7] J.M. Clacens, Y. Pouilloux, J. Barrault, Appl. Catal. A 227 (2002) 181.

S.-H. Chai et al. / Journal of Catalysis 250 (2007) 342–349
349
[8] T. Hass, A. Neher, D. Arntz, U.S. patent 5 426 249, 1995, to Degussa Ak-
tiengesellschaft.
[9] H. Adkins, W.H. Hartung, Org. Synth. Coll. 1 (1941) 15.
[10] A. Neher, T. Haas, D. Arntz, H. Klenk, W. Girke, U.S. patent 5 387 720,
1995, to Degussa Aktiengesellschaft.
[18] K. Tanabe, Catal. Today 78 (2003) 65.
[19] K. Tanabe, M. Misono, Y. Ono, H. Hattori, New Solid Acids and Bases:
Their Catalytic Properties, Elsevier, Amsterdam, 1989, pp. 5 and 61.
[20] H.A. Benesi, J. Am. Chem. Soc. 78 (1956) 5490.
[21] H.A. Benesi, J. Phys. Chem. 61 (1957) 970.
[11] E. Schwenk, M. Gehrke, F. Aichner, U.S. patent 1 916 743, 1933, to
Scheering-Kahlbaum.
[22] K.S.W. Sing, D.H. Everett, R.A.W. Haul, L. Moscou, R.A. Pierotti, J. Rou-
querol, T. Siemieniewska, Pure Appl. Chem. 57 (1985) 603.
[12] S. Ramayya, A. Brittain, C. DeAlmeida, W. Mok, M.J. Antal, Fuel 66
(1987) 1364.
[13] M.J. Antal Jr., W.S.L. Mok, G.N. Richards, Carbohydr. Res. 199 (1990)
111.
[14] L. Ott, M. Bicker, H. Vogel, Green Chem. 8 (2006) 214.
[15] K. Tanabe, CHEMTECH 21 (1991) 628.
[16] K. Tanabe, S. Okazaki, Appl. Catal. A 133 (1995) 191.
[17] I. Nowak, M. Ziolek, Chem. Rev. 99 (1999) 3603.
[23] T. Iizuka, K. Ogasawara, K. Tanabe, Bull. Chem. Soc. Jpn. 56 (1983) 2927.
[24] Z.-H. Chen, T. Iizuka, K. Tanabe, Chem. Lett. 13 (1984) 1085.
[25] M. Craus, in: G. Ertl, H. Knözinger, J. Weitkamp (Eds.), Handbook of
Heterogeneous Catalysis, vol. 5, VCH, Weinheim, 1997, p. 2370.
[26] M. Guisnet, in: G. Ertl, H. Knözinger, J. Weitkamp (Eds.), Handbook of
Heterogeneous Catalysis, vol. 2, VCH, Weinheim, 1997, p. 626.
[27] S.-H. Chai, H.-P. Wang, Y. Liang, B.-Q. Xu, Green Chem. (2007), doi:10.
1039/b702200j, in press.