The influence of the calcination conditions on the catalytic activity of Bi2MoO6 in the selective oxidation of propylene to acrolein

Article abstract of DOI:10.1016/j.molcata.2006.04.030

The activity of Bi₂MoO₆ catalysts with small amounts of excess bismuth in the selective oxidation of propylene is strongly dependent on the calcination conditions. These catalysts show reasonable activities and high selectivities after calcination for a short period of time or after calcination at a moderate temperature. However, these catalysts become low active and low selective after calcination for a long period of time or after high temperature calcination. This difference is ascribed to relative low Bi/Mo-ratios at (parts of) the surface after calcination for a short period of time. The active surface becomes enriched with bismuth during prolonged calcination or reaction, resulting in a deactivation of the catalyst. The deactivated catalysts can be reactivated by reduction, this supports that the deactivation is due to a surface enrichment with bismuth. The dependence of the catalytic activity on the calcination conditions is likely one of the reasons behind the conflicting results on the activity of Bi₂MoO₆ found in the literature.

Full text of DOI:10.1016/j.molcata.2006.04.030

Journal of Molecular Catalysis A: Chemical 256 (2006) 1–8
The influence of the calcination conditions on the catalytic activity of
Bi2MoO6 in the selective oxidation of propylene to acrolein
Willy J.M. van Wella,∗, Minh Thang Leb, Niels Christian Schiødtc, Serge Hosted, Per Stoltzea
a Aalborg University, Esbjerg, Niels Bohrs Vej 8, DK-6700 Esbjerg, Denmark
b Hanoi University of Technology, Department of Petrochemistry, Dai Co Street 1, Hanoi, Vietnam
c Haldor Topsøe A/S, Nymøllevej 55, DK-2800 Lyngby, Denmark
d Gent University, Department of Inorganic and Physical Chemistry, Krijgslaan 281-S3, B- 9000 Gent, Belgium
Received 29 March 2006; accepted 3 April 2006
Available online 24 May 2006
Abstract
The activity of Bi2MoO6 catalysts with small amounts of excess bismuth in the selective oxidation of propylene is strongly dependent on
the calcination conditions. These catalysts show reasonable activities and high selectivities after calcination for a short period of time or after
calcination at a moderate temperature. However, these catalysts become low active and low selective after calcination for a long period of time
or after high temperature calcination. This difference is ascribed to relative low Bi/Mo-ratios at (parts of) the surface after calcination for a short
period of time. The active surface becomes enriched with bismuth during prolonged calcination or reaction, resulting in a deactivation of the
catalyst. The deactivated catalysts can be reactivated by reduction, this supports that the deactivation is due to a surface enrichment with bismuth.
The dependence of the catalytic activity on the calcination conditions is likely one of the reasons behind the conflicting results on the activity of
Bi2MoO6 found in the literature.
© 2006 Elsevier B.V. All rights reserved.
Keywords: Bismuth molybdates; Selective oxidation; Catalytic activity; Deactivation; Acrolein
1. Introduction
drying and sol–gel methods. Within these methods, there are
again a large number of variations like pH, concentrations
and method of mixing of the bismuth and molybdenum solu-
tions. Moreover, theconditions(temperature, durationandatmo-
sphere) under which the bismuth molybdenum precursors are
calcined to obtain the microcrystalline bismuth molybdenum
oxide vary also greatly in different studies [11]. The influence
of these different synthesis conditions on the activity has only
rarely been mentioned in the literature [12].
The activity of Bi2MoO6 in the butene oxidation has
been shown by Batist et al. to be strongly dependent on the
conditions at which the catalyst has been prepared [13,14].
They explained this strong dependence partly by different
oxygen coordination of molybdenum in the aqueous phase
during the synthesis, and partly by the amount of bismuth at
the surface of the catalyst. The second argument is supported
by latter publications, which show that an excess of bismuth in
Bi2MoO6 leads to a pronounced enrichment of the surface with
bismuth. The result of this enrichment is a catalyst with a low
activity and low selectivity [15–18]. This enrichment causes
therefore the activity of the catalyst to be very sensitive to the
Bismuth molybdates form the base of commercial catalysts
for propylene oxidation to acrolein, for propylene ammoxidation
to acrylonitrile and for the oxidative dehydrogenation of butene
to buta-di-ene [1,2]. The reactivity of the three model catalysts,
Bi2MoO6 (␥-phase), Bi2Mo2O9 (␤-phase) and Bi2Mo3O12 (␣-
phase) in these reactions has been studied extensively in the
literature [3–10]. Despite the large number of studies, there
is still debate about the relative activity of these model cata-
lysts. Most researchers agree that the Bi2Mo2O9 phase is more
active than the Bi2Mo3O12 phase [3–8]. However, the activity of
Bi2MoO6 is disputed. Some researchers claim it to be the most
active phase [3], some claim it to be much less active than the
two other phases [4–6,10] and others find it to have an activity
similar to Bi2Mo3O12 [7–9].
Several methods are used in the literature to synthesize
the model catalysts: precipitation, solid-state reactions, spray-
∗
Corresponding author. Tel.: +45 79127633; fax: +45 75453643.
E-mail address: wvw@aaue.dk (W.J.M. van Well).
1381-1169/$ – see front matter © 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.molcata.2006.04.030

2
W.J.M. van Well et al. / Journal of Molecular Catalysis A: Chemical 256 (2006) 1–8
ratio of bismuth and molybdenum in the catalyst, and thereby,
also very sensitive to the preparation conditions.
The surface area of the samples was measured by using a
Sorptomatic 1990 from ThermoQuest. The samples were out-
gassed overnight at 180 ◦C. The surface area was calculated
by using the two-parameter BET model for adsorption points
between p/p0 = 0.05 and 0.33.
The surface composition was determined by XPS. The XPS
spectra were recorded using an ESCA spectrometer equipped
with an Al anode (hυ = 187.850 eV).
The bulk composition of the samples was determined by
ICP. For this purpose, 20 mg of Bi2MoO6 sample was dis-
solved in 50 ml of 7.5 M nitric acid and afterwards diluted to
500 ml. Two or three solutions were made for each sample.
The standard solutions were made from Bi(NO3)3·5H2O and
(NH4)6Mo7O24·4H2O.
We will focus on Bi2MoO6 in this article, since the activity
of this catalyst is the most disputed issue in this context. Dur-
ing our previous work with bismuth molybdate catalysts in the
propylene oxidation [19,20], we have observed that the activity
of catalysts with a Bi/Mo-ratio close to 2 can be very sensitive
to the calcination conditions. This dependence is, according to
our knowledge, not reported previously and it could possibly
explain some of the ambiguities on the activity of Bi2MoO6 cat-
alysts found in the literature. This dependence will be the main
issue of this paper, but we will start with a comparison of the
activity of different samples with Bi/Mo around 2 prepared by
co-precipitation and spray-drying. The results reported in this
paper could also contribute to the general understanding of the
catalytic activity of bismuth molybdate catalysts.
2.3. Activity measurements
The catalysts were tested in a vertical down-stream quartz
tube reactor with an internal diameter of 4 mm. The quartz tube
was heated by an iron tube placed in a Carbolite oven. The
grain size of the catalyst was between 200 and 500 ␮m and
0.05–0.1 g of catalyst was used. The reactant gasses, consist-
ing of 5% O2 in N2 and 5% C3H6 in N2, were added in a ratio
1:1. The total gas flow varied between 20 and 60 ␮mol/s. The
conversion was kept between 3 and 15%, to enable calculation
of reaction rates from the results. The composition of the outlet
gasses was determined by using an on-line Shimadzu 17A gas
chromatograph with a thermal conductivity detector. The mole
fraction of propylene, acetaldehyde and acrolein in the effluent
gas was determined every 13 min on Carbowax 20 M on Chro-
mosorb G and Chromosorb 101 packed columns in series. The
mole fraction of carbonmonoxide and carbondioxide was deter-
mined on Carboxen 1000 and Porapak N packed columns in
series. The latter analysis was not performed continuously.
The reaction rate for the formation of acrolein was calculated
by assuming differential operation of the reactor. The reaction
rate, rw, based on the catalyst mass was calculated as:
2. Experimental
2.1. Synthesis of catalysts
Co-precipitation and spray-drying were used to synthe-
size several batches of Bi2MoO6 catalyst. In the precipitation
method, three solutions consisting of 0.49 M Bi(NO3)3·5H2O in
1.5 M HNO3, 0.035 M (NH4)6Mo7O24·4H2O in water and 4 M
ammonia were mixed simultaneously using a peristaltic pump.
The solutions were mixed simultaneously in order to assure the
correct stochiometric ratio between Bi and Mo during the mix-
ing of the solutions. The pH of the mixture was kept between
pH 3 and pH 6 by controlling the amount of ammonia during
mixing. The final pH of the mixture was adjusted to 5 (P1, P4)
or 5.5 (P2, P3, P5) and the mixture was filtrated after 2 h (P5),
6 h (P2, P3) or after 16 h (P1, P4).
In the spray-drying method, Bi(NO3)3·5H2O and
(NH4)6Mo7O24·4H2O were dissolved in a nitric acid solution
with a pH of lower than 1. The clear solution was spray-dried
¨
in a Buchi 190 spray-dryer with a 0.5 mm nozzle and a
liquid feeding rate of 5 ml/min under a flow of heated air
(700 l/min (STP) and 225 ◦C). Two samples were prepared by
spray-drying, sample 1 had a Bi/Mo-ratio equal to 2.1 and
sample 2 had a Bi/Mo-ratio equal to 2.0.
The amount of catalyst prepared per batch by precipitation
and spray-drying was between 5 and 10 g. The samples were
calcined in static air at temperatures between 350 and 550 ◦C
for 10 h.
Y
C3H4O
rw = Ftot
×
(1)
m
in which Ftot is the total mol flow (which is assumed to be con-
stant during reaction), YC H O the mole fraction of acrolein in
the outlet stream and m is the catalyst mass. The reaction rate,
rs, based on the catalyst surface area was calculated as:
3
4
Y
C3H4O
rs = Ftot
×
(2)
m SA
in which SA is the surface area of the catalyst.
The selectivity for the formation of acrolein was calculated
2.2. Characterization of catalysts
as:
The calcined samples were characterized by X-ray diffraction
using a Philips X’pert diffractometer equipped with Co-source
(wavelength of Co K␣ radiation: λ = 1.79 A). An obliquely cut
Y
C3H4O
SC H O
=
(3)
˚
3
4
Y
− Y
C3H6
C3H6,0
silicon disc was used as a sample support when the amount of
sample was small (less than 0.5 g). The error introduced in 2θ by
using a silicon disc was corrected by calculating the displace-
ment of the sample by comparing the measured diffractogram
with JCPDS-ICDD reference 84–0787 for Bi2MoO6 [21].
in which YC H and YC H are the mole fraction of propylene
3
6
3
6,0
in the outlet stream and in the inlet stream. The selectivity for
theformationofacetaldehyde, carbonmoxideandcarbondioxide
was calculated like the selectivity for acrolein, with the mole

W.J.M. van Well et al. / Journal of Molecular Catalysis A: Chemical 256 (2006) 1–8
3
fraction of carbonmonoxide and carbondioxide divided by 3 and
the mole fraction of acetaldehyde divided by 1.5.
The sum of the selectivities for acrolein, acetaldehyde, car-
bonmonoxide and carbondioxide formation was between 95 and
105%. Determination of the acrolein selectivity with Eq. (3)
becomes too inaccurate at low conversion, in these cases the
selectivity to acrolein was calculated by:
Y
C3H4O
SC H O
=
(4)
3
4
Y
+ (2/3) × Y
+ (1/3) × YCO + (1/3) × YCO
C3H4O
C2H4O
2
The selectivity was also calculated by Eq. (4) for experiments
that lasted for several days.
3. Results
3.1. Comparison of catalysts with Bi/Mo around 2
Fig. 1. XRD patterns of catalysts calcined at 550 ◦C.
We have prepared five samples by precipitation and two sam-
ples by spray-drying. The first sample made by spray-drying
(SD1) was prepared with Bi/Mo = 2.1, the second sample (SD2)
had a Bi/MO-ratio of 2. The spray-dried samples have been cal-
cined at 550 ◦C, while the calcination temperature of the samples
made by precipitation has been varied.
Tables 1 and 2 show the activity, at a reaction temperature
of 375 ◦C, as a function of calcination temperature for the dif-
ferent samples. Table 1 shows the rate of acrolein formation
based on the catalyst mass (rw), Table 2 shows the activity based
on the surface area (rs). The samples are ordered after increas-
ing activity. The results show that there is a large variation in
the activity for the samples calcined at 550 ◦C. The activity
of sample SD2 with Bi/Mo = 2 is about 10 times higher than
the activity of SD1 with Bi/Mo = 2.1. The activity of the sam-
ples made by precipitation (P1–P5) varies from 0.1 to less than
10−4 ␮mol/(m2 s).
calcination temperature. When the calcination temperature is
decreasedfrom550to450 ◦C, theactivityofsampleP1increases
from 5 × 10−3 to 2 × 10−2 ␮mol/(m2 s), while the activity for
sample P3 changes from no activity to 10−4 ␮mol/(m2 s). Cal-
cination of this latter sample at 350 ◦C leads to an additional
increase to 2 × 10−3 ␮mol/(m2 s). Such large increases in activ-
ity are not seen at all for samples P2 and P4.
Fig. 1 shows the XRD patterns for the samples after calci-
nation at 550 ◦C, the intensity is displayed as the square-root of
counts. All the diffraction peaks found for sample P2 and sample
SD2 correspond to Bi2MoO6 (reference 84-0787). All the other
samples have extra peaks around 2θ = 32.5 and 37.0◦. However,
sample P4 made by precipitation has a number of characteris-
tics that are absent for the other samples: it has extra peaks at
2θ = 15.4, 17.2, 20.9 and 21.5◦, the extra peak at 37.0◦ is sharper,
it has a small double peak around 54◦ and the peak at 64◦ is more
intense. All these characteristics are typical for Bi2Mo2O9 and
this means that sample P4 is polluted by Bi2Mo2O9. Because
of the absence of these characteristics, the other samples cannot
be polluted by Bi2Mo2O9, instead we ascribe the extra peaks to
␤-Bi2O3, which is the metastable tetragonal structure of Bi2O3.
The presence of ␤-Bi2O3 in Bi2MoO6 has also been reported
by Buker and Greaves [22]. We have estimated the amount of
␤-Bi2O3 or Bi2Mo2O9 in the samples from the ratio between the
peak areas at 32.5◦ (mean signal for ␤-Bi2O3 and Bi2Mo2O9)
and 32.9◦ (mean signal for the Bi2MoO6). The results are given
in Table 3.
The most remarkable feature in the results, however, is that
the activity of the low active samples depends strongly on the
Table 1
Reaction rate for acrolein formation at 375 ◦C based on catalyst mass, rw
(␮mol/(g s)), for different calcination temperatures
Temperature (◦C)
P3
SD1
P1
P2
SD2
P4
Uncalcined
350 ◦C
0.3
0.8
0.03
0.001
0
450 ◦C
0.12
0.007 0.005
0.20
0.07
0.4
0.11
550 ◦C
0.08
Table 3 shows the results of the estimated content of ␤-Bi2O3
and Bi2Mo2O9 and the Bi/Mo-ratio calculated from these per-
centages. The samples are ordered after decreasing Bi/Mo-ratio.
The table displays also the Bi and Mo content measured by ICP.
The results of the double measurements were within 5% of each
other. The bismuth content is below the stochiometric value for
Bi2MoO6 and is almost the same for all the samples measured.
As a result, the Bi/Mo-ratio calculated by using the bismuth
Table 2
Reaction rate for acrolein formation at 375 ◦C based on catalyst surface area, rs
(␮mol/(m2 s)), for different calcination temperatures
Temperature (◦C)
P3
SD1
P1
P2
SD2
P4
350
450
550
0.002
0.0001
0
0.02
0.004 0.005
0.04
0.03
0.1
0.1
0.05

4
W.J.M. van Well et al. / Journal of Molecular Catalysis A: Chemical 256 (2006) 1–8
Table 3
Comparison of Bi/Mo-ratio determined by XRD, ICP and XPS for samples calcined at 550 ◦C
Pollution
XRD
Bi/Mo
XRD
[Bi] (mmol/g)
ICP
[Mo] (mmol/g)
ICP
Bi/Mo
ICP
Bi/Mo from [Mo]
ICP
Bi/Mo
XPS
P3
SD1
P1
P2
SD2
P4
14% Bi2O3
9% Bi2O3
3% Bi2O3
–
2.3
2.2
2.1
2
2
1.8
2.9
n.d.
2.9
n.d.
2.9
3.0
1.5
n.d.
1.6
n.d.
1.8
2.0
1.9
n.d.
1.8
n.d.
1.6
1.5
2.2
n.d.
2
n.d.
1.8
1.6
2.6
2.7
2.4
2.4
2.0
1.9
–
11% Bi2Mo2O9
Stochiometric values for Bi2MoO6: 3.3 mmol Bi/g and 1.6 mmol Mo/g.
and molybdenum measurements is below 2 for all samples. The
molybdenum content in a bismuth molybdate sample is much
more sensitive to differences in the Bi/Mo-ratio than the bis-
muth content, this is reflected in a larger variation between the
results for the different samples. Therefore, and because of the
low solubility of bismuth, the Bi/Mo-ratio is also calculated from
the molybdenum content. The order in the Bi/Mo-ratios for the
different samples calculated in this way correspond to the XRD-
results; however, the values for the Bi/Mo-ratios are somewhat
lower. We conclude from the XRD and ICP measurements, that
P2 and SD2 are pure Bi2MoO6, P3, SD1 and P1 are polluted by
␤-Bi2O3 (P3 has the highest content, P1 the lowest) and sample
P4 is polluted by Bi2Mo2O9. It should be noted that the order in
the Bi/Mo-ratios of the different samples in Table 3 corresponds
to the order in activity in Table 2.
Table 3 displays also the Bi/Mo-ratio at the surface as mea-
sured by XPS. It shows that the Bi/Mo-ratio determined by XPS
is higher than the ratio determined by XRD for samples P3,
SD1, P1 and P2, while the two Bi/Mo-ratios are almost equal to
each other for SD2 and P4. The XPS results show therefore that
the samples that are polluted by ␤-Bi2O3 have a higher Bi/Mo-
ratio at the surface than in the bulk. However, samples with a
Bi/Mo equal or lower than 2, are not enriched with bismuth in
the surface. This is in agreement with previous reported results
[15–18]. Sample P2 does not show ␤-Bi2O3 in the XRD anal-
ysis; however, a Bi/Mo-ratio higher than 2 is measured for this
sample by XPS.
3.2. Influence of calcination temperature on the
crystallinity
Fig. 2 shows the XRD patterns for sample P3 after calcination
at 350, 450, 500 and 550 ◦C, together with the pattern for SD2
calcined at 550 ◦C (pure Bi2MoO6). After calcination at 350 and
450 ◦C, all and only the reflections of the Bi2MoO6 structure
are seen. The only difference with the pattern for the pure SD2
sample is the line broadening due to the smaller crystal size. The
figure shows that the extra reflections displaying the presence of
␤-Bi2O3 are only seen after calcination at 500 ◦C. This means
that it seems that pure Bi2MoO6 is obtained when sample P3 is
calcined at a temperature of 450 ◦C or lower.
The same phenomenon is happening in a smaller degree for
sample P4 with Bi/Mo lower than 2. Fig. 3 shows the XRD
patterns for P4 calcined at different temperatures together with
the pattern for pure Bi2MoO6. Also in this case, the pollution
with Bi2Mo2O9 is difficult to see when the sample is calcined
at a low temperature.
3.3. Activity of uncalcined samples: influence of calcination
temperature, duration and atmosphere on the activity
To investigate the influence of the calcination conditions fur-
ther, the activity of uncalcined samples is measured. This means
The XRD and ICP analysis show that our samples made
by spray-drying have approximately the intended Bi/Mo-ratio,
while this is not the case for most of the samples made by precip-
itation. The latter should be ascribed to the loss of metal ions,
especially molybdenum, during filtration. The loss of molyb-
denum was confirmed through the synthesis of P5. In this
synthesis, the mixture was already filtrated after 2 h and the
residue was analyzed for molybdenum and bismuth by ICP.
No bismuth, but 14% of the molybdenum was found in the
residue.
The results presented above confirm earlier reported results
that show that the activity of Bi2MoO6 is very sensitive to
the Bi/Mo-ratio [15–18]. A small excess of bismuth leads to
a large enrichment of the surface with bismuth and thereby to a
low activity. However, our results show that the activity of the
catalysts with an excess of bismuth is very dependent on the
calcination temperature. This dependence will be the issue of
the remainder of this article.
Fig. 2. Comparison of XRD patterns of sample P3 calcined at 350, 450, 500
and 550 ◦C with the XRD pattern of sample SD2 (pure Bi2MoO6) calcined at
550 ◦C.

W.J.M. van Well et al. / Journal of Molecular Catalysis A: Chemical 256 (2006) 1–8
5
Fig. 5. Rate for acrolein formation rw (␮mol/(g s)) of uncalcined sample P1 at
(ꢁ) 375, (ꢂ) 425 and (×) 450 ◦C vs. time.
sample P3: the activity increases during calcination, it reaches a
maximum and decreases afterwards. The maximum reached for
sample P1 of rw = 0.8 ␮mol/(g s) at 375 ◦C is somewhat higher
than for sample P3. The deactivation, however, is much slower:
after 25 h, the activity is 0.54 ␮mol/(g s), which is nearly 70%
of the maximum rate. At this stage, the activity of sample P3
was only 20% of the maximum rate. Such a large decrease is
not even reached after 100 h for sample P1, where the activity
is 35% of the maximum rate. The selectivity (as determined by
Eq. (4)) decreased approximately from 95 to 85% during this
100 h period. At 425 and 450 ◦C, the deactivation of the catalyst
is much faster. The activity is only 3% of the maximum activity
after 25 h at 450 ◦C and the selectivity (Eq. (4)) has decreased
from 90 to 65%.
Fig. 3. Comparison of XRD patterns of sample P4 calcined at 375, 425, 450
and 550 ◦C with the XRD pattern of sample SD2 (pure Bi2MoO6) calcined at
550 ◦C.
that the uncalcined catalyst is heated in the reaction mixture con-
sisting of 95% N2, 2.5% O2 and 2.5% C3H6. The left part (first
25 h) of Fig. 4 shows the activity, conversion and selectivity of
sample P3 as a function of time, when it is heated in the reac-
tion mixture to 375 ◦C. The heating rate was 5 ◦C/min and the
final temperature is reached after 70 min. The figure shows that
the activity increases during the calcination and reaches a max-
imum of rw = 0.3 ␮mol/(g s) after 2 h at 375 ◦C. This maximum
rate is 10 times higher than the rate measured after calcination
at 350 ◦C for 10 h. The activity decreases rapidly after the max-
imum and is down to 0.06 ␮mol/(g s) after 25 h. The decrease
in activity is caused by a decrease in both the conversion and
the selectivity. The selectivity reaches a maximum of 84% after
95 min, after 25 h the selectivity has decreased to about 50%.
The maximum activities for the different uncalcined samples
are given in Table 1.
Sample P1 has also been calcined in static air at 450 ◦C for
1, 4 and 25 h to see the influence of the calcination time on the
surface area and on the activity. Table 4 shows the surface area,
activity and selectivity of these samples. The results show that
the surface area is almost independent on the calcination time,
but the activity of the samples decreases strongly with increas-
ing calcination time. The activity after 4 and 25 h of calcination
is, respectively, 42 and 8% of the activity after 1 h of calcina-
tion. This rate of deactivation in static air is the same as under
reaction conditions at 450 ◦C. Here, the activity decreases to,
respectively, 0.32 and 0.07 ␮mol/(g s) within 3 and 24 h after an
activity of 0.7 ␮mol/(g s) has been reached.
Fig. 5 shows the activity at 375, 425 and 450 ◦C as a function
oftimeforuncalcinedP1samples. Thesamepatternisseenasfor
We have calcined sample P1 in the reactor in a mixture of
5% O2 in N2 at 450 ◦C for 24 h, to investigate the influence of
the atmosphere on the deactivation. The activity of this sample,
0.10 ␮mol/(g s), corresponds well with the activity of a sam-
ple calcined under reaction conditions after 25 h (see Fig. 5).
The influence of the presence of water vapor on the deacti-
vation has been tested by calcining a sample of P1 for 1 h in
dry air and afterwards for 3 h in a mixture of approximately
1% water in nitrogen at 450 ◦C. The maximum activity of this
sample is 0.20 ␮mol/(g s) which is of the same order of magni-
tude as the activity of 0.49 ␮mol/(g s) measured after 4 h under
reaction. Comparing these values with the maximum activity of
2.0 ␮mol/(g s) in Fig. 5, we could say that 90% of the activity
is lost after calcination in 1% of water, while 76% of activity is
lost during reaction.
Fig. 4. (ꢀ) Rate for acrolein formation rw (␮mol/(g s)), (ꢁ) conversion and (ꢂ)
selectivity of uncalcined sample P3 at 375 ◦C vs. time. The second part of the
figure shows the results after reduction for 1 h in 5% of propylene in nitrogen.

6
W.J.M. van Well et al. / Journal of Molecular Catalysis A: Chemical 256 (2006) 1–8
Table 4
Surface area, rate for acrolein formation based on catalyst mass rw (␮mol/(g s)) and selectivity at 375 and 450 ◦C for sample P1 calcined at 450 ◦C for different
periods of time
Calcination time (h)
Surface area (m2/g)
rw (at 375 ◦C) (␮mol/(g s))
rw (at 450 ◦C) (␮mol/(g s))
Selectivity (at 450 ◦C) (%)
1
4
25
6
6
5
0.27
0.12
0.02
0.69
0.29
0.07
70
70
55
Table 5
Reaction rate for acrolein formation at 375 ◦C based on catalyst mass, rw (␮mol/(g s)), after reduction and ratio between reaction rate after and before reduction R,
for different calcination temperatures
Temperature (◦C)
P3
SD1
P1
SD2
P4
rw
R
rw
R
rw
R
rw
R
rw
R
Uncalcined
450 ◦C
550 ◦C
0.77
12
0.40
0.10
3.3
20
0.19
∞
0.07
10
0.09
1.2
0.11
1.0
3.4. XPS-characterization at different stages of
deactivation
leads to a strong increase in activity for the low active samples,
while the activity of the active samples is hardly influenced by
the reduction. It can also be seen that the activity of all samples
calcined at 550 ◦C have the same order of magnitude after reduc-
tion. In accordance to the results of Mitchell et al. and of Alsdorf
et al., XRD and DSC measurements show that reduction of the
samples leads to the formation of metallic bismuth [18,23]. The
The previous results show that the loss of activity is partly
due to a decrease in selectivity. This decrease in selectivity
implies that the surface composition changes during deactiva-
tion, in this case that the Bi/Mo-ratio increases, since surfaces
with a Bi/Mo > 2 are known to be unselective [15–18]. We
have measured the change in surface composition by measur-
ing the Bi/Mo-ratio at the surface by XPS at different stages of
deactivation. A sample of uncalcined P3 was taken out of the
reactor when the maximum activity was reached at 375 ◦C (see
also Fig. 4). The Bi/Mo-ratio for this sample was 2.4 which is
somewhat lower than the value of 2.6 measured for a sample
calcined at 550 ◦C. For P1, the Bi/Mo-ratio was measured after
2 h of calcination at 450 ◦C, the measured value of 2.3 is slightly
lower than the value of 2.4 measured after calcination at 550 ◦C.
Although the differences measured by XPS are small, they do
agree with an increase in the Bi/Mo-ratio at the surface during
deactivation.
˚
bismuth particles are several hundreds Angstrom large, as esti-
mated from the line width of the XRD peak of metallic bismuth
at 2θ = 31.7◦. The DSC measurements show that typically 3–5%
of the bismuth is reduced to metallic bismuth.
The P1 catalysts that are calcined at 450 ◦C for 1, 4 and 25 h
(see Table 3) were reduced at 450 ◦C. Fig. 6 shows the activity
at 450 ◦C of these samples after reduction. The figure shows
that all these catalysts have approximately the same activity,
and that they also deactivate with approximately the same
rate after reduction. This means that, although the catalysts
had very different activities before the reduction, the catalysts
(surfaces) are very similar after reduction. The deactivation
of the catalyst happens according to a typical pattern: first a
relative slow deactivation at constant high selectivity, afterwards
3.5. Reduction experiments
It has been shown in the literature that the activity and selec-
tivity of catalysts with Bi/Mo slightly above 2 can be increased
by reduction, while the activity and selectivity of catalysts with
Bi/Mo below 2 are unaffected [17,18,23]. The activation of the
catalysts is ascribed to a lowering of the Bi/Mo-ratio at the sur-
face, due to the formation of metallic bismuth [18,23,24].
We have reduced our samples for 30–60 min in a flow of 5%
of propylene in nitrogen. The second part of Fig. 4 shows the
result for reduction of sample P3 at 375 ◦C. The activity after
reduction is 2.5 times higher than the maximum activity and 12
times higher than the activity immediately before reduction. The
reduction leads also to a much higher selectivity (about 90%)
and a more stable catalyst. The activity of the samples after
reduction and the increase obtained by reduction for the differ-
ent samples is shown in Table 5. It can be seen that the reduction
Fig. 6. Rate for acrolein formation rw (␮mol/(g s)) at 450 ◦C after reduction of
samples of P1 that were calcined at 450 ◦C for (ꢁ) 1, (ꢂ) 4 and (ꢀ) 25 h, and
rate for acrolein formation for (×) uncalcined sample P1 vs. time.

W.J.M. van Well et al. / Journal of Molecular Catalysis A: Chemical 256 (2006) 1–8
7
a faster deactivation together with a decrease in selectivity. The
deactivation of the uncalcined sample at 450 ◦C (see Fig. 5)
is compared with the deactivation of the reduced samples at
450 ◦C in Fig. 6. This comparison shows that the uncalcined
sample deactivates with approximately the same rate as the
reduced samples in the second deactivation stage.
means that the bismuth that is in excess, is initially located in the
bulk of the catalyst or that is concentrated in larger particles of
bismuth oxide on the surface. Extensive surface studies, which
are beyond the scope of this work, would be required to solve
this latter question. The excess bismuth merges during calcina-
tion or reaction with the active parts of the surface, leading to a
deactivation of the catalyst.
4. Discussion
The deactivated catalyst can be reactivated by reduction,
while the activity of the active samples is hardly influenced
by reduction. Catalysts that are deactivated in different degrees
before reduction get approximately the same activity and deac-
tivation behavior after reduction. Combined with the results of
van Oeffelen et al. and Mitchell et al., this supports our opinion
that the deactivation is due to an increase in the Bi/Mo-ratio at
the catalytically active surface [17,18]. The comparable activi-
ties after reduction show that the properties of the surface after
reductionarenotdeterminedbythepropertiesofthecatalytically
active surface before the reduction, but rather by the properties
of a much larger part of the catalyst that is reduced.
Our and previous reported results show that bismuth molyb-
datecatalystswithBi/Molargerthan2areenrichedwithbismuth
at the surface [15,17,18]. This results in samples which have
normally low activities and low selectivities. However, after
calcination at a low temperature or for a short period of time,
these samples show relative high activities and high selectivities.
These samples deactivate during calcination and this means that
the activity is very dependent on the calcination conditions. This
dependence of the activity on the calcination conditions is very
likely one of the reasons for the conflicting results found in the
literature about the activity of Bi2MoO6. The fact that signifi-
cant amounts of pollution (either ␤-Bi2O3 or Bi2Mo2O9) cannot
be seen by XRD after calcination at temperatures up to 500 ◦C,
makes this effect even more important.
The strong deactivation of samples with Bi/Mo > 2 is accom-
panied by a loss of selectivity. Together with the XPS results,
this shows that the Bi/Mo-ratio at the catalytic active surface
increases during the calcination. This change in Bi/Mo-ratio at
the surface is shown to be an activated and relative slow process.
The question is, what is happening during this deactivation pro-
cess?
One of the reasons for the increase in the Bi/Mo-ratio at the
surface could be the loss of molybdenum from the surface. The
loss of molybdenum due to the sublimation of molybdenum
compounds is a well-known cause for the deactivation of the
Fe2(MoO4)3 catalyst in the methanol to formaldehyde synthe-
sis. This sublimation is known to be strongly enhanced by the
presence of water due to the formation of the relative volatile
MoO2(OH)2 compound [25,26]. The loss of molybdenum from
Bi2Mo3O12 has been reported recently by Yanina and Smith
[26]. However, the loss of molybdenum cannot explain the fact
that catalysts with Bi/Mo significantly higher than 2 in the bulk,
do show high initial activities and selectivities. Furthermore, we
have shown that the rate of deactivation is approximately the
same when the catalyst is under reaction conditions (final water
content maximal 0.4%), under a dry O2 in N2 flow, or when it
is calcined in static air. Calcination in a 1% water–air mixture
leads to a somewhat higher deactivation (90% of the maximum
activity is lost) than under reaction conditions (76% of the activ-
ity is lost) after 4 h at 450 ◦C. According to the results of Yanina
et al. and to previous results on the Fe2(MoO4)3 catalyst, the
presence of water should have a very strong influence on the
deactivation [25,26].
CatalystshavingdifferentBi/Mo-ratiosinthebulkobtainalso
approximately the same activity after reduction. This shows that
reduction creates surfaces with a composition that is more or
less independent on the bulk composition, or surfaces where
the activity is more or less independent on the surface com-
position (Bi/Mo < 2). The slow initial deactivation at a high and
stable selectivity after reduction, is an indication that the Bi/Mo-
ratio at the catalytically active surface is below 2 after reduction.
The reduction leads to the formation of metallic bismuth [18]
and, due to the low melting point of 271 ◦C of bismuth, to the
formation of bismuth particles [23,24]. The formation of these
particles will lead to a decrease in the Bi/Mo-ratio in the remain-
ing surface, making this part of the surface active and selective.
After reduction, the bismuth particles are re-oxidized and merge
with the active part of the surface, leading to a deactivation [23].
The comparable rate of deactivation of the reduced samples in
the second stage, and of the deactivation of the uncalcined sam-
ples indicates that the processes by which excess bismuth is
polluting the active parts of the surface are similar in both cases.
We have shown that relative large amounts of excess bis-
muth and molybdenum cannot be seen by XRD as ␤-Bi2O3 and
Bi2Mo2O9 when the samples are calcined at temperatures below
500 ◦C. A reason for this could be that the areas of crystalline ␤-
Bi2O3 and Bi2Mo2O9 formed at low temperatures are not large
enough to be seen by XRD. Comparing the temperatures where
␤-Bi2O3 is observed by XRD with the much lower temperatures
where deactivation occurs, it is clear that XRD visible ␤-Bi2O3
is not a prerequisite for the deactivation. Spectroscopic methods
like Infrared and Raman, may be more suitable for detecting
these impurities [27].
When we compare the activity of pure Bi2MoO6 at
375 ◦C (0.08 ␮mol/(g s) or 0.05 ␮mol/(m2/s) for sample SD2)
with the activities of pure ␣-Bi2Mo3O12 (0.09 ␮mol/(g s) or
0.06 ␮mol/(m2/s)) and pure ␤-Bi2Mo2O9 (0.08 ␮mol/(g s) or
0.07 ␮mol/(m2/s)) synthesized by spray-drying, we can con-
clude that all three model catalysts have comparable activities.
This conclusion was also found in our previous work [20]. The
comparable activities makes that the ordering after activity is
The relatively high initial activities and selectivities can
instead be explained by relatively low initial Bi/Mo-ratios at the
catalytically active surface. This means that after a short period
of calcination (parts of) the surface is not or hardly enriched
with bismuth, resulting in an active and selective catalyst. This

8
W.J.M. van Well et al. / Journal of Molecular Catalysis A: Chemical 256 (2006) 1–8
difficult, and can therefore easily be influenced by factors like
small amounts of impurities in the catalyst and uncertainty in
the determination of the surface area.
References
[1] J.L. Callahan, R.K. Grasselli, E.C. Milberger, H.A. Strecker, Ind. Eng.
Chem. Prod. Res. Dev. 9 (1970) 134.
[2] R.K. Grasselli, J.D. Burrington, in: D.D. Eley (Ed.), Advances in Catal-
ysis, vol. 30, Academic Press, New York, 1981, p. 133.
[3] J.R. Monnier, G.W. Keulks, Preprints Am. Chem. Soc. Div. Petr. Chem.
24 (1979) 19.
[4] Z. Bing, S. Pei, S. Shishan, G. Xiexan, J. Chem. Soc. Faraday Trans.
86 (1990) 3145.
[5] D. Carson, G. Coudurier, M. Forissier, J.C. Vedrine, A. Laarif, F.
Theobald, J. Chem. Soc. Faraday Trans. I 79 (1983) 1921.
[6] I.K. Kolchin, E.L. Gal’perin, S.S. Bobkov, L.Ya. Margolis, Kinet. Catal.
6 (1965) 794.
[7] K. German, B. Grzybowska, J. Haber, Br. Acad. Pol. Sci. Chim. 11
(1973) 319.
[8] P.A. Batist, B.C. Lippens, G.C.A. Schuit, J. Catal. 5 (1966) 55.
[9] L.D. Krenze, G.W. Keulks, J. Catal. 64 (1980) 295.
[10] J.D. Burrington, R.K. Grasselli, J. Catal. 59 (1979) 79.
[11] B.R. Sant, S.B. Rao, J.R. Rao, R.S. Thakur, K.M. Parida, J. Sci. Ind.
Res. 43 (1984) 542.
5. Conclusions
The activity and selectivity of Bi2MoO6 catalysts, in the
selective oxidation of propylene to acrolein, are very sensitive
to the presence of small amounts of excess bismuth. This excess
leads to a surface enrichment with bismuth, and thereby to a low
active and low selective catalyst. However, such catalysts with a
Bi/Mo slightly above 2, do show reasonable activities and selec-
tivities after calcination at moderate temperatures and/or for a
short period of time. These catalysts loose their activity after cal-
cination at a higher temperature or for a longer period of time.
This makes that the activity and selectivity of these catalysts
are strongly dependent on the calcination time and temperature.
The above mentioned dependences are most likely to be some
of the main reasons behind the conflicting results reported in the
literature on the activity of Bi2MoO6.
The reasonable initial activities of catalysts with a Bi/Mo
slightly above 2, are ascribed to relative low Bi/Mo-ratios at
parts of the surface after calcination for a short period of time.
The excess bismuth merges with the active parts of the surface
during calcination, leading to the deactivation of the catalyst.
The sublimation of molybdenum from the catalyst may play a
minor role in the deactivation process.
[12] G.I. Aleshina, C. Joshi, D.V. Tarasova, G.N. Kustova, T.A. Nikoro,
React. Kinet. Catal. Lett. 26 (1984) 203.
[13] P.A. Batist, J.F.H. Bouwens, G.C.A. Schuit, J. Catal. 25 (1972) 1.
[14] P.A. Batist, J. Chem. Technol. Biotechnol. 29 (1979) 451.
[15] I. Matsuura, R. Schut, K. Hirakawa, J. Catal. 63 (1980) 152.
[16] P.M. Burban, G.C.A. Schuit, T.A. Koch, K.B. Bischoff, J. Catal. 126
(1990) 326.
[17] D.A.G. van Oeffelen, J.H.C. van Hooff, G.C.A. Schuit, J. Catal. 95
(1985) 84.
[18] A.G. Mitchell, P.M. Lyne, K.F. Scott, C.S.G. Phillips, J. Chem. Soc.
Faraday Trans. I 77 (1981) 2417.
[19] M.T. Le, J. Van Craenenbroeck, I. Van Driessche, S. Hoste, Appl. Catal.
A Gen. 249 (2003) 355.
[20] M.T. Le, W.J.M. Van Well, P. Stoltze, I. Van Driessche, S. Hoste, Appl.
Catal. A Gen. 282 (2005) 189.
[21] F. Theobald, A. Laarif, A.W. Hewat, Ferroelectrics 56 (1984) 219.
[22] R.A. Buker, C. Greaves, J. Catal. 108 (1987) 247.
[23] E. Alsdorf, I. Grohmann, K. Jancke, K.H. Schnabel, M. Selenina, D.
Vollgraf, Z. Phys. Chem. Leipzig 271 (1990) 35.
[24] M. El Jamal, M. Forissier, G. Coudurier, J.C. Vedrine, in: M.J. Phillips
(Ed.), Proceedings of Ninth International Congress on Catalysis, 1988,
p. 1617.
In accordance with literature, the deactivated samples can be
reactivated by reduction, while the activity of the active samples
is hardly affected by reduction. This reactivation is a result of
the formation of bismuth particles leading to a decrease in the
Bi/Mo-ratio at other parts of the surface. The deactivation during
calcination and after reduction occur at similar rates indicating
that these deactivation processes are similar.
Acknowledgements
[25] A.P.V. Soares, M.F. Portela, A. Kliennemann, L. Hilaire, Chem. Eng.
Sci. 58 (2003) 1315.
[26] S.V. Yanina, R.L. Smith, J. Catal. 213 (2003) 151.
[27] R. Olier, G. Coudurier, M. El Jamal, M. Forissier, J.C. Vedrine, J. Chem.
Soc. Faraday Trans. I 85 (1989) 2615.
Sune Ebbesen is acknowledged for performance of initial
experiments. Søren Aagaard Pedersen is acknowledged for per-
formance of the reduction experiments.