Epoxidation of propylene with nitrous oxide on Rb2SO4-modified iron oxide on silica catalysts

Article abstract of DOI:10.1016/j.jcat.2007.01.001

The catalytic activity of alkaline and earth alkaline-modified silica-supported metal oxide was investigated for epoxidizing propylene with nitrous oxide. Iron oxide gave the best results, and surprisingly chromium oxide also produced propylene oxide (PO). Unmodified iron oxide catalyst showed low oxidation activity and produced propanal (57% selectivity) in concert with small amounts of acrolein, allyl alcohol, and acetone. After modification, the oxidation rate increased significantly, with PO the principal product. PO selectivities up to 85-90% and space-time yields of 0.25-0.53 mmol PO g^-1 h^-1 were obtained over supported iron oxide modified with Rb₂SO₄. A high throughput composition study revealed that other alkali and earth alkali salts were less effective modifiers. Isopropanol decomposition demonstrated that Rb₂SO₄ severely reduced the acidity of the catalyst. As a result of the neutralization, PO isomerization was drastically reduced. Accordingly, when feeding PO instead of propylene with N₂O over the catalyst, a similar reduction of consecutive PO reactions was observed on Rb₂SO₄ modification. Despite the excellent epoxidation results, a catalytic process remains infeasible due to the restricted service time of the catalyst. Thermogravimetric analyses of a spent catalyst showed carbonaceous residues, suggesting that cokes deactivate the catalyst. Feeding PO indicates that PO itself is a source of cokes. Catalyst regeneration is possible without significant loss of performance. UV-vis DRS and EPR were used to determine the local environment of Fe³⁺ in the (un)promoted iron oxide catalyst; the findings suggest well-dispersed distorted tetrahedral Fe³⁺ sites for epoxidation activity. Fe dispersion is ruled by the promoter salts, with both anions and cations being essential. Along with the structural influences, inspection of the catalytic data in concert with XPS and Raman analyses provides evidence of a direct (electronic) promoter effect on the catalytic activity.

Full text of DOI:10.1016/j.jcat.2007.01.001

Journal of Catalysis 247 (2007) 86–100
www.elsevier.com/locate/jcat
Epoxidation of propylene with nitrous oxide on Rb2SO4-modified iron oxide
on silica catalysts
Bart Moens a, Hendrik De Winne a, Steven Corthals a, Hilde Poelman b, Roger De Gryse b,
Vera Meynen c, Pegie Cool c, Bert F. Sels a,∗, Pierre A. Jacobs a,∗
a
Center for Surface Chemistry and Catalysis, Katholieke Universiteit Leuven, Kasteelpark Arenberg 23, B-3001 Heverlee, Belgium
b
Surface Physics and Thin Film Division, Ghent University, Krijgslaan 281-S1, B-9000 Ghent, Belgium
c
Laboratory of Adsorption and Catalysis, Department of Chemistry, University of Antwerp, Universiteitsplein 1, B-2610 Wilrijk, Belgium
Received 1 November 2006; revised 31 December 2006; accepted 2 January 2007
Available online 20 February 2007
Abstract
The catalytic activity of alkaline and earth alkaline-modified silica-supported metal oxide was investigated for epoxidizing propylene with
nitrous oxide. Iron oxide gave the best results, and surprisingly chromium oxide also produced propylene oxide (PO). Unmodified iron oxide
catalyst showed low oxidation activity and produced propanal (57% selectivity) in concert with small amounts of acrolein, allyl alcohol, and
acetone. After modification, the oxidation rate increased significantly, with PO the principal product. PO selectivities up to 85–90% and space–
−1 −1
were obtained over supported iron oxide modified with Rb SO . A high throughput composition
2 4
time yields of 0.25–0.53 mmol PO g
h
study revealed that other alkali and earth alkali salts were less effective modifiers. Isopropanol decomposition demonstrated that Rb SO severely
2
4
reduced the acidity of the catalyst. As a result of the neutralization, PO isomerization was drastically reduced. Accordingly, when feeding PO
instead of propylene with N O over the catalyst, a similar reduction of consecutive PO reactions was observed on Rb SO modification. Despite
2
2
4
the excellent epoxidation results, a catalytic process remains infeasible due to the restricted service time of the catalyst. Thermogravimetric
analyses of a spent catalyst showed carbonaceous residues, suggesting that cokes deactivate the catalyst. Feeding PO indicates that PO itself is a
source of cokes. Catalyst regeneration is possible without significant loss of performance. UV–vis DRS and EPR were used to determine the local
3+
3+
environment of Fe in the (un)promoted iron oxide catalyst; the findings suggest well-dispersed distorted tetrahedral Fe sites for epoxidation
activity. Fe dispersion is ruled by the promoter salts, with both anions and cations being essential. Along with the structural influences, inspection
of the catalytic data in concert with XPS and Raman analyses provides evidence of a direct (electronic) promoter effect on the catalytic activity.
© 2007 Elsevier Inc. All rights reserved.
Keywords: Partial oxidation; Nitrous oxide; Propylene epoxidation; Iron oxide; Promoter effects; Physicochemical characterization
1. Introduction
produced abundantly in industry [2], its use as an oxidant offers
a promising solution to the environmental problems associated
with this gas.
Catalytic oxidation preferably uses molecular oxygen as an
oxidant. Instead of partial oxidation, the high reactivity of the
surface activated oxygen species generated from O2 often leads
to consecutive oxidation and, consequently, to a loss in selec-
tivity. In the last decade, much attention has been given to the
use of mild oxidants, such as nitrous oxide (N2O). N2O is a gas
with a high greenhouse warming potential [1]. Because N2O is
Selective oxidation with N2O has been demonstrated for
benzene [3–19] and methane [19–23] hydroxylation, and for
alkane oxidative dehydrogenation over Fe-ZSM-5, and to a
lesser extent over ZSM-5 exchanged with other ions [24–29].
Although hydroxylation has first been associated with zeolite
framework acidity, N2O is activated at a metal site, leading to
N2 and atomically adsorbed oxygen, sometimes referred to as
α-oxygen [19]. The active O further oxidizes CH4 and benzene
to precursors of methanol and phenol, respectively. Phenol des-
orbs at reaction temperature [19] and thus is produced catalyti-
*
Corresponding authors. Fax: +32 16 32 1998.
E-mail addresses: bert.sels@biw.kuleuven.be (B.F. Sels),
pierre.jacobs@biw.kuleuven.be (P.A. Jacobs).
0021-9517/$ – see front matter © 2007 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2007.01.001

B. Moens et al. / Journal of Catalysis 247 (2007) 86–100
87
cally, whereas the methoxy intermediate decomposes into COx
2. Experimental
and water. Methanol is obtained only after hydrolysis at low
temperature [20,30]. Whereas the synthesis of phenol tested at
Solutia’s demonstration plant [31] points to an economically
feasible process, purification of industrial N2O and catalyst de-
activation seem to be an obstacle. In contrast to hydroxylation,
the oxidative dehydrogenation mechanism is unclear. Whereas
α-sites likely control hydroxylation, either alcohol dehydro-
genation or direct oxidative dehydrogenation is expected to oc-
cur [25].
Recently, epoxidation of propylene with N2O has been dis-
cussed [32–37]. Because current PO manufacturing suffers
from serious drawbacks [38], it is worthwhile to search for new
PO routes. The chlorohydrin route faces environmental pollu-
tion problems due to quantitative CaCl2 production, whereas
the hydroperoxide process is handicapped due to its dependence
on the market price of the alcohol byproduct. Propylene epox-
idation with N2O may provide an environmentally benign PO
production route as no byproducts are formed:
2.1. Catalyst preparation
Catalysts were prepared by incipient wetness impregnation
with a fresh aqueous solution of a metal chloride (or nitrate)
precursor (e.g., FeCl2 for iron oxide) together with the promoter
salt on commercial silica (Silica Kieselgel 60, Fluka; specific
surface area, 500 m2 g−1; pore volume, 0.75 ml g−1; average
pore diameter, 40–63 nm). Concentrations of the aqueous salt
solutions typically range between 9 and 180 mM in FeCl2 and
0.15 and 4.83 M in promoter salt. After impregnation, the cat-
alyst is dried for 3 h at 373 K and calcined in air at 973 K for
6 h.
2.2. Catalyst testing
Epoxidation was studied using 4 parallel fixed-bed flow
quartz reactors (3 mm i.d.) operating at atmospheric pressure
and 683 K. Each reactor tube contained 0.10 g catalyst pellets
ranging in diameter from 125 to 250 µm. N2O (purity 99.5%)
and O2 (purity 99.5%) were used as oxidants. After pretreat-
ment of the catalyst at 793 K in O2 atmosphere for 3 h, the
reaction was started by introducing a mixture of propylene and
N2O in He at reaction temperature. The gas rates were regu-
lated using electronic thermal mass flow controllers. A molar
ratio of C3/N2O/He (1/5/30) was used with a total flow rate
equivalent to 7.5 ml min−1 per reactor tube (GHSV [STP] =
4.5 L h−1 g−1). The outlet gases were analyzed via on-line GC
analysis using a Poraplot Q column connected to an internal
methanator and a FID. All connections and valves between the
reactor tube and the gas chromatograph were heated at about
423 K to prevent condensation. The results after 180 min are
presented and used for discussion, unless stated otherwise, de-
spite the imperfect steady-state conditions due to catalyst de-
activation (vide infra). Co-feeding of PO was carried out using
2.5 vol% PO in Ar.
Iron oxide loaded on silica has been proposed for the selective
epoxidation with N2O [32–37]. All authors agree on the ne-
cessity of alkali-promoted iron oxide. After modification with
sodium acetate, a PO yield and selectivity of 4.8 and 40% were
obtained. Almost no PO was formed over unpromoted silica-
supported iron oxide [36,37]. Cesium acetate was advanced as
the preferred additive, providing PO yields of 7.5% with 75%
PO selectivity at 650 K [32]. KCl was presented as a superior
modifier for iron oxide supported on structured mesoporous sil-
ica, such as SBA-15 and MCM-41, as well as on steam-treated
ZSM-5 [33–35]. A change in redox behavior of the iron catalyst
was observed as a result of the KCl promotion [34]. In addi-
tion, on modification with alkali metal salts, a dramatic shift
in oxidation mechanism, from allylic oxidation for unpromoted
catalysts to epoxidation for promoted catalysts, was proposed
[33,34]. A PO selectivity of 80% at 4% propylene conversion
was reported.
So far, only iron oxide and a few promoters have been stud-
ied. To gain a broader view of the potential of various metal
oxides, we have systematically investigated the catalytic per-
formance of a series of supported metal oxides modified with
various alkaline and earth alkaline additives. For practical rea-
sons, we have chosen to synthesize the catalyst via a single
impregnation step of the two constituents, that is, metal and pro-
moter salt.
2.3. Catalyst characterization
Isopropanol decomposition using temperature ranging from
523 to 573 K was studied with 0.10 g of sample and a flow of
7.5 ml min−1 consisting of 4.2 kPa isopropanol in He. Acetone
and propylene were analyzed with on-line gas chromatography
with a FID such as for the oxidation experiments.
Thermogravimetric tests were done in a TGA Q500 appa-
ratus (TA Instruments) on samples discharged from the reactor
after several hours on stream. The tests were made in a flow
of 90 ml min−1 O2, which passes tangentially to the pan of the
TGA apparatus containing approximately 25 mg of sample. The
heating rate was 5 K/min from 303 to 1073 K. The samples
turned from black to white after the TGA experiments, indicat-
ing complete C removal.
High-throughput screening provided us with a unique cat-
alyst composition consisting of iron oxide supported on silica
modified with Rb2SO4. Having determined an optimal formu-
lation, this paper attempts to correlate the catalytic features (ac-
tivity/selectivity) of the catalyst with its surface acid–base prop-
erties and structural/electronic properties. A reaction mecha-
nism of the iron-catalyzed epoxidation is proposed based on a
comparison of the products formed in a typical propylene epox-
idation process with the results of PO co-feeding experiments
in reaction conditions.
XRD measurements were carried out to study the eventual
formation of crystalline phases in the catalyst; a STOE STADI
P transmission diffractometer was used with CuKα radiation

88
B. Moens et al. / Journal of Catalysis 247 (2007) 86–100
(ν = 1.54 Å) measuring between 5 and 60 degrees 2θ, with 2.5
degrees per step analyzing 500 seconds per step.
Table 1
Effect of alkali and earth alkali acetate modification of FeO on silica on the
epoxidation of propylene with N O
x
2
Surface analysis was performed by XPS using a Perkin
Elmer PHI ESCA 5500 electron energy analyzer system com-
bined with monochromatic 450 W AlKα X-ray radiation. Wide
scan surveys and high-resolution multiplexes were performed
at constant pass energies of 187.85 and 23.5 eV, respectively.
Samples were fixed in an indium foil to avoid contamination of
the carbon signals.
FT-Raman spectra were recorded on a Nicolet Nexus FT-
Raman spectrometer with a Ge detector and a 1064 nm Nd:YAG
laser. Samples were measured in a 180◦ reflective sampling
configuration. A total of 4000 scans were averaged for each
sample, with a resolution of 8 cm−1. The laser power was set
between 1.0 and 2.0 W. The samples that showed heating effects
(broad band around 3000 cm−1 Raman shift) were measured at
laser powers of 1 W maximum.
Entry Active element/
C conv. Selectivity (%)
3
promoter/support (%)
PO ACR PA AC AA CO CO Other
2
1
2
3
4
5
6
7
8
9
–/–/–
0
0
0
1
0
18
13
0
57
57
16
9
3
0
0
53
24
0
0
1
3
9
1
0
2
1
3
0
13
7
15
2
1
0
0
7
0
0
1
6
6
0
6
7
20
30
0
6
13
8
8
9
–/–/SiO
Fe/–/SiO
Fe/Li/SiO
Fe/Na/SiO
Fe/K/SiO
Fe/Rb/SiO
Fe/Cs/SiO
Fe/Mg/SiO
Fe/Ba/SiO
0.8
2.6
5.1
7.3
6.5
3.9
3.1
3.5
4.5
2
2
16 16
26 10
22
9
2
0
2
2
5
2
0
12
14
11 48
2
9 70 10
2
15 78
4
4
3
16
2
7
2
10
6
15
13 21
2
−1 −1
;
Notes. 1. Reaction parameters: C /N O/He = 1/5/30; GHSV = 4.5 L h
g
3
2
partial pressure of propylene and N O is 2.8 and 13.9 kPa, respectively; T =
2
683 K; TOS = 180 min.
2. Catalyst: Fe 0.1 wt%; promoter/Fe = 36 (mol/mol).
Diffuse reflectance UV–vis spectra were obtained on a Var-
ian Cary 5 spectrophotometer equipped with a diffuse–reflec-
tance accessory. Spectra were measured in the reflectance mode
3. PO = propylene oxide, ACR = acrolein, PA = propanal, AC = acetone,
AA = allylic alcohol, other = cracking products (CH , C H , C H + their
4
2
4
2 6
oxygenated species).
(R ) and converted according to the Kubelka–Munk function
∞
For a series of monovalent alkali metal acetates, activities
and PO selectivities/yields showed the following sequence (see
Supplementary material, Fig. S1): Na > K > Li > Rb > Cs.
This order emphasizes that Na is the modifier of choice among
the alkali metal acetates.
Among the Na acetate-promoted oxides of other potentially
active elements originating from the chloride and nitrate salts of
Zn, Mn, Co, Ru, Cu, Ni, Au, Rh, Cr, Sm, Eu, La, Nd, and Ag,
with a metal loading between 0.1 and 0.2 wt%, only chromium
oxide (from CrCl3 impregnation) was capable of producing PO.
A PO selectivity of approximately 10% at 0.5% conversion was
achieved. This low but distinct PO yield is worth mentioning,
although not understood. Indeed, to the best of our knowledge,
chromium oxide has never before been associated with selective
epoxidation processes.
F(R ). The spectra of the samples calcined in dry air at 793 K
∞
were collected between 200 and 2500 nm with BaSO4 as a ref-
erence.
EPR measurements were carried out with a cw X-band
Bruker ESP 300E instrument (microwave frequency ≈ 9.5 GHz)
equipped with a rectangular TE104 cavity. EPR spectra were
measured using a field modulation amplitude of 3.2 mT, a mod-
ulation frequency of 100 kHz, and microwave power of 0.1–
0.2 mW. EPR spectra of samples calcined in dry air at 793 K
were recorded at 120 K and at room temperature.
3. Results
3.1. Catalytic properties of various promoted silica-supported
iron oxide catalysts
Various support materials for Na acetate-promoted iron ox-
ide, including ZSM-5, Al2O3, TiO2, CaCO3, BaCO3, and amor-
phous SiO2/Al2O3 with varying Si/Al ratios, were used as well.
None showed activity for the selective (ep)oxidation of propy-
lene with N2O; only COx was formed. With the carbonate sup-
ports, the system failed to show any oxidation activity at all.
The Na acetate-promoted iron oxide on silica, showing su-
perior PO formation yields with N2O, also failed to show any
epoxidation selectivity with O2 as oxidant. Mainly CO2 was
formed, along with small amounts of acrolein and acetaldehyde.
Finally, the alkali metal acetate promoter salts were substi-
tuted for the corresponding nitrates (data not shown) and sul-
fates (Table 2). Surprisingly, on modification with Rb2SO4, the
catalyst showed a PO selectivity of 76% at 9.1% conversion,
corresponding to a PO formation rate of 0.37 mmol g−1 h−1
(Table 2, entry 3). This result is clearly better than that obtained
with Na acetate (0.10 mmol g−1 h−1 PO formation rate; Table 1,
entry 5). Modification of the iron catalyst with a series of sul-
fate salts (Table 2) yielded the following epoxidation activity
sequence: Rb ꢀ Cs > K > Tl > Na > VO. Rb and Cs sulfate
promoted the PO yield up to values of 5–6%. The different
The catalytic performances at 683 K of promoted FeOx/sil-
ica for propylene oxidation are summarized in Table 1. Under
the specified conditions, no reaction occurred over a reactor
tube filled with glass wool (entry 1). The silica support was
slightly active, although conversion was low (entry 2). The
modest activity is likely attributed to iron impurities (<100 ppm
wt%). When 0.1 wt% iron was introduced in the silica, ox-
idation was in agreement with earlier observations (entry 3)
[32–37], underlining the active role of iron. Despite its oxida-
tion ability, unpromoted FeOx/silica showed a PO selectivity
of <1% at a propylene conversion of 2.6% (entry 3), with
propanal (PA) the main product formed with 57% selectivity.
Lower amounts of acrolein (ACR), allylic alcohol (AA), and
acetone (AC) were also found.
Other data in Table 1 (entries 4–10) demonstrate that mod-
ifying the iron catalyst with acetate salts clearly produced PO
from propylene. Screening tests with various alkali (Li, Na, K,
Rb, and Cs) and earth alkali (Mg and Ba) acetate salts revealed
that univalent cations were generally better promoters than bi-
valent cations (entries 5 and 9).

B. Moens et al. / Journal of Catalysis 247 (2007) 86–100
89
Table 2
Table 3
Propylene conversion and product selectivities using sulfate salts as promoter
Influence of promoter anion and anion concentration on the catalytic perfor-
for FeO on silica
mance of 0.1 wt% FeO on silica in the epoxidation of propylene with N O
x
x
2
Entry Active element/
C
conv. Selectivity (%)
Entry Active element/ Rb/Fe
promoter/support
C conv. Selectivity (%)
3
3
promoter/support (%)
(%)
PO ACR PA AC AA CO CO Other
PO ACR PA AC AA CO CO Other
2
2
1
2
3
4
5
6
7
Fe/Na SO /SiO 1.5
73
90
76
76
67
0
0
0
0
0
0
0
0
0
0
2
3
8
0
0
2
2
1
0
0
0
0
0
0
0
0
0
0
0
0
0
2
0
0
24
3
0
9
7
1
2
3
4
5
6
7
8
9
Fe/RbCl/SiO
2
Fe/RbOH/SiO
Fe/RbF/SiO
2
36
36
36
1.9
1.7
2.2
2.3
3.2
2.3
9.1
6.1
7.3
3 0
6 0
5 0
0
0
0
0
0
1
2
2
4
0
0
0
0
0
0
2
1
4
0
0
0
0
0
0
0
0
2
22 75
18 76
19 76
19 81
0
0
0
0
23
1
9
5
8
2
4
2
Fe/K SO /SiO
3.0
10
11
12
12 10
100
15
2
4
2
2
Fe/Rb SO /SiO 9.1
2
4
2
a
Fe/Rb SO /SiO 8.5
Fe/Rb CO /SiO 36
0 0
2
4
2
2
3
2
Fe/Cs SO /SiO
6.9
Fe/VO(SO )/SiO 0.3
Fe/Rb SO /SiO 2.3
62 0
91 0
76 0
82 0
66 2
7
0
8
7
2
4
2
2
4
2
0
10
0
7
Fe/Rb SO /SiO 4.5
4
2
2
4
2
Fe/Tl SO /SiO
2.8
68
Fe/Rb SO /SiO
9
0 11
0 10
2 12
2
4
2
2
4
2
Fe/Rb SO /SiO 18
2
4
2
−1 −1
;
Notes. 1. Reaction parameters: C /N O/He = 1/5/30; GHSV = 4.5 L h
g
3
2
Fe/Rb SO /SiO 36
2
4
2
partial pressure of propylene and N O is 2.8 and 13.9 kPa, respectively; T =
2
−1 −1
683 K; TOS = 180 min.
Notes. 1. Reaction parameters: C /N O/He = 1/5/30; GHSV = 4.5 L h
g
;
3
2
2. Catalyst: Fe 0.1 wt%; promoter cation/Fe = 9 (mol/mol).
partial pressure of propylene and N O is 2.8 and 13.9 kPa, respectively; T =
2
a
Second run after regeneration of the catalyst by burning the cokes in air.
683 K; TOS = 180 min.
2. Catalyst: Fe 0.1 wt%; promoter cation/Fe = 36 (mol/mol).
sequences for the alkali metal acetates and sulfates and com-
bination of Rb with other anions (Table 3, entries 1–4), suggest
that both cation and anion play roles as activity/selectivity-
enhancing promoters. To the best of our knowledge, the su-
perior properties of Rb2SO4-promoted iron oxide on silica as
catalyst in the epoxidation of propylene with N2O have not been
reported before.
The effect of an increased Fe loading of the catalyst at con-
stant atomic Si/Rb ratio of 50 is shown in Fig. 1. A sharp
maximum in the PO yield curve is seen for an iron content
of 0.15 wt%, corresponding to an average Fe surface den-
sity of 0.032 Fe atoms per nm2 and a PO formation rate of
0.49 mmol PO g−1 h−1. The act−iv1ity per iron site remained
around 18 mmol PO mmol−1 Fe h , much higher than the pub-
lished values (∼5 mmol PO mmol−1 Fe h−1 [33–35]). The de-
pendence of activity on iron surface density supports the con-
cept that the degree of dispersion of the iron species on the
surface of SiO2 contributes to the epoxidation properties of the
catalyst. Despite the slight revival of oxidation activity at iron
loadings >0.6 wt% Fe, PO yields decreased gradually with in-
creasing Fe content. Fe2O3 and Fe2(CO3)3 showed negligible
epoxidation activity. The overall parallel changes of propy-
lene conversion and PO yield curves, at least between 0.05 and
0.6 wt% Fe loading (i.e., between 0.011 and 0.13 Fe atoms per
nm2) indicate that PO selectivities remained almost unchanged
between 60 and 75%.
The influence of Rb/Fe molar ratio was investigated for
0.1 wt% iron catalyst (Table 3, entries 5–9). For Rb/Fe ratios
of 2.3–36, propylene conversion clearly showed an optimum of
9% for an Rb/Fe ratio of 9, resulting in a PO yield of 7%, corre-
sponding to a PO formation rate of 0.37 mmol g−1 h−1. A sim-
ilar exercise was carried out for the catalyst with 0.15 wt%
iron. A conversion of 13% for a Rb/Fe molar ratio of 12 was
found, resulting in an optimal PO yield of 8.1% and a PO for-
mation rate of 0.44 mmol g−1 h−1. In summary, these results
Fig. 1. Effect of iron loading on catalytic performance of Rb sulfate-modified FeO on silica in the epoxidation of propylene with N O. Reaction parameters:
x
2
−1 −1
C /N O/He = 1/5/30; GHSV = 4.5 L h
g
, T = 683 K, TOS = 180 min. Catalyst: Si/Rb = 50 (mol/mol). (2) C H conversion, (F) PO yield.
3
2
3 6

90
B. Moens et al. / Journal of Catalysis 247 (2007) 86–100
Table 4
shown in Fig. 2. The conversion of propylene gradually de-
creased from nearly 17% after 80 min to 8% after 360 min
of reaction under the conditions shown in the figure. The
highest PO yield measured in this study (9.9%) was ob-
tained after 80 min and corresponds to a PO formation rate of
0.53 mmol PO g−1 h−1. The PO selectivity gradually increased
from 59 to 66% in the same time span. Similar observations
have been reported for benzene hydroxylation and alkane ox-
idative dehydrogenation with N2O [3–19,24–29]. The rapid
decline of catalytic activity has been attributed to coke forma-
tion blocking the active sites.
A closer look at the spent catalyst showed a black coloration
of the initially white catalyst after 7 h on stream. Thermogravi-
metric analysis of the black residue (Fig. 3, insert) showed an
initial low-temperature weight loss associated with water re-
moval; a second change at 673–873 K, with a maximum in the
derivative curve at about 805 K, corresponds to a further weight
loss of 3.5%, indicating combustion of carbonaceous deposits.
The correlation curve between the actual PO yield and the
accumulated PO production (Fig. 3) yields information on cat-
alyst deactivation. The slope of the curve is a measure of deacti-
vation rate, whereas the x- and y-intercepts are indicative of the
theoretical PO production capacity of the catalyst and the theo-
retical maximum PO yield per unit amount of catalyst assuming
no deactivation, respectively. The 0.15 wt% iron oxide on sil-
ica promoted with Rb2SO4 (Rb/Fe = 12) was compared with
a literature catalyst with a 1 wt% iron oxide on SBA-15 mod-
ified with KCl (K/Si = 0.04) [33]. Both catalysts clearly show
deactivation, and thus a limited catalyst service time; moreover,
the Rb2SO4-modified catalyst shows a higher potential for PO
production.
Influence of reaction temperature on the catalytic performance of Rb SO -
modified FeO supported on silica in the epoxidation of propylene with N O
2
4
a
x
2
Entry Temp.
(K)
C conv. Selectivity (%)
3
(%)
PO ACR PA AC AA CO CO Other
2
1
2
3
4
5
598
623
653
683
693
0.9
2.3
7.2
13
15.7
88
85
71
62
56
0
0
0
0
0
0
1
3
4
4
0
1
2
7
2
0
0
1
2
2
2
3
6
9
10
9
9
0
1
8
8
8
11 10
15
−1 −1
Note. Reaction parameters: C /N O/He = 1/5/30; GHSV = 4.5 L h
g
;
3
2
partial pressure of propylene and N O is 2.8 and 13.9 kPa, respectively; TOS =
2
180 min.
a
Catalyst: Fe 0.15 wt%; Rb/Fe = 12 (mol/mol).
suggest that an optimum molar Rb/Fe ratio of approximately 10
is needed. Higher Rb concentrations have only a minor effect
on the PO yield, whereas lower Rb concentrations are clearly
detrimental.
The effect of reaction temperature was investigated in the
range of 598–693 K (Table 4). At increasing temperatures, the
selectivity for PA, AC, AA, CO, and other unidentified prod-
ucts gradually increased at the expense of PO, whereas CO2
selectivity did not change with temperature. CO2 seems to be
formed independently from PO due to reactive lattice oxygen.
An Arrhenius plot of the experimental data in Table 4 provides
the apparent activation energy (79.8 kJ mol−1) for the Rb2SO4-
promoted 0.15 wt% Fe on silica, indicating the absence of mass
transfer limitations.
3.2. Stability in time of catalytic performances during
epoxidation of propylene
After calcining the spent catalyst in O2 at 823 K, the activity
of the Rb2SO4-modified FeOx/silica was almost fully restored,
as can be derived from the data in Table 2 (entries 3 and 4).
The catalytic behavior of a 0.15 wt% Fe on silica catalyst
modified with Rb2SO4 (Rb/Fe = 12) with time on stream is
Fig. 2. Catalytic behavior of Rb SO -modified FeO supported on silica as function of time-on-stream at 683 K. Reaction parameters: C /N O/He = 1/5/30;
x
2
4
3
2
−1 −1
GHSV = 4.5 L h
g
. Catalyst: 0.15 wt% Fe; Rb/Fe = 12. (2) C H conversion, (P) PO selectivity, (F) PO yield.
3
6

B. Moens et al. / Journal of Catalysis 247 (2007) 86–100
91
Fig. 3. Correlation between accumulated PO production per weight catalyst and the actual PO yield for modified FeO on silica in the epoxidation of propylene
x
−1 −1
+
with N O. Curve A: 0.15 wt% Fe on silica, Rb/Fe = 12; C /N O/He = 1/5/30; GHSV = 4.5 L h
g
, T = 683 K, TOS = 180 min. Curve B: K –1 wt%
2
3
2
FeO /SBA-15 (data obtained from [33]). Insert: thermogravimetric weight and derivative weight loss curves obtained in air flow, for a catalyst discharged after test
x
of Fig. 2 (Rb SO -modified 0.15 wt% FeO on silica with Rb/Fe = 12).
x
2
4
3.3. Mechanistic features: The stability of propylene oxide in
reaction conditions
Table 5
Product selectivities from propylene and PO
Entry Active element/ Feed
promoter/support
Conv. Selectivity (%)
(%)
Because propylene does not form coke by interaction with
the catalyst in absence of N2O, the carbonaceous residues prob-
ably originate from consecutive transformations of the oxidized
products. The origin of PO instability may be associated with
isomerization over Lewis or Brönsted acid sites. To rational-
ize the relationship between catalyst acidity and PO stability, a
control experiment with PO feed was carried out (Table 5). In
the presence of N2O, PO showed only slight thermal instability
in a reactor loaded with quartz wool. However, in the presence
of bare silica, PO instability was dramatically increased (en-
try 1), resulting in the formation of mainly PA and AA. AA can
be further oxidized into ACR. Nearly the same PA selectivity
was obtained when 0.1 wt% iron was added to the silica sup-
port (entry 2). The AA abundantly formed on the support was
now converted into COx (mainly CO) and unidentified prod-
ucts. On the other hand, addition of Rb2SO4 to the iron catalyst
partially restored PO stability (entry 4); only 28.9% of the PO
was converted into PA, AC, COx, and traces of AA. Thus an
important effect of the promoter is the prevention of secondary
reactions of PO. Similar observations can be made when com-
paring the behavior of the promoted and unpromoted catalysts
in a propylene reaction (entries 3 and 5). With the unpromoted
catalyst, PA rather than PO was the preferred product. Further-
more, black coloration of the catalyst occurred in each case,
indicating that PO induced coke.
PO ACR PA AC AA CO Other
x
1
2
3
4
5
–/–/SiO
Fe/–/SiO
Fe/–/SiO
PO
PO
100
100
–
–
6
49 1 28
3
9
8
13
25
13
2
10 48
2
1
6
7
3
0
2
2
Propylene 2.6 1 13 57
a
Fe/Rb SO /SiO PO
28.9 –
0
0
14 30
2
18 35
10 5
2
4
2
a
Fe/Rb SO /SiO Propylene 6.1 82
1
2
4
2
a
Catalyst: 0.1 wt% Fe; Rb/Fe = 18 (mol/mol).
the final product distribution. In this way a Rb2SO4-promoted
catalyst showed excellent PO selectivity, whereas the analogue
unpromoted catalyst gave isomerization products.
3.4. Catalyst characterization
3.4.1. Dehydration of isopropanol
PO instability plays an important role in the propylene re-
action with N2O. Acidity characterization of the catalyst can
support observations concerning PO isomerization. The iso-
propanol (IPA) decomposition is a model reaction for charac-
terizing acid and/or basic properties of the catalyst [39]. Dehy-
dration of IPA to propylene and/or di-isopropyl ether requires
Brönsted acid sites, whereas dehydrogenation to acetone occurs
on basic (redox) centers. Results of this experiment are listed in
Table 6. Silica and 0.1 wt% iron loaded silica at 573 K showed
IPA conversions of 9.4 and 14%, respectively (entries 1 and 2).
The high propylene selectivities of 98 and 99%, respectively,
suggest the presence of strong acidity. In contrast to the results
of PO feeding presented in Section 3.3, here the presence of
iron clearly influenced IPA conversion; adding iron increased
the Brönsted acidity of the catalyst. Promotion with Rb2SO4 re-
The PO feeding experiments also provide some elucidation
regarding the reaction mechanism. The amount of PO fed in the
experiment can be assumed to be formed initially in the epox-
idation reaction. Table 5 clearly suggests that initially PO was
the primary product, and that PO stability mainly determined

92
B. Moens et al. / Journal of Catalysis 247 (2007) 86–100
Table 6
mass at ∼973 K (=the calcination temperature). We attributed
this mass loss to the decomposition of Rb2SO4 into basic Rb2O
and SO3. However, the decomposition was not complete; Ra-
man measurements (vide infra) showed the presence of small
amounts of sulfate, which is probably essential for an enhanced
activity.
Isopropanol decomposition at 573 K as characterization of acidity
Entry
Active element/
promoter/support
IPA conv.
(%)
Selectivity (%)
Propylene
Acetone
CO
2
1
–/–/SiO
9.4
14
0.6
98
99
75
2
1
0
0
0
2
a
2
Fe/–/SiO
2
a
3
Fe/Rb SO /SiO
2
25
2
4
−1 −1
3.4.3. X-ray diffraction
Note. Reaction parameters: GHSV = 4.5 L h
g
; TOS = 30 min.
a
Catalyst: 0.1 wt% Fe; Rb/Fe = 18 (mol/mol).
X-ray diffraction was carried out to search for crystallites in
the catalyst powder. Fig. 5 shows the results of promoted FeOx
on silica with different amounts of Rb2SO4. Three distinct sig-
nals could be distinguished: d = 3.51 (25.3◦), 3.12 (28.6◦),
and 3.00 (29.7◦). The diffraction pattern of the catalyst with
Rb/Fe = 4.5, 12, and 18 also shows a small signal at d = 3.30
(26.5◦). After comparing these signals with a database [40], the
signals at d = 3.12 and 3.00 are assigned to crystalline Rb2SO4;
the small signal at d = 3.3 to RbCl; and the signal at d = 3.51,
appearing only at the largest promoter concentrations, to Rb2O.
In conclusion, XRD measurement suggests that the active iron
species formed an amorphous phase or well-dispersed crystal-
lites too small to detect, whereas an excess of Rb2SO4 led to
hexagonal Rb2SO4 crystallites.
sulted in neutralization of almost all acid sites, as evidenced by
the low IPA conversion of 0.6% at 573 K (entry 3), in agree-
ment with the mechanistic proposal. Thus, unpromoted iron
silica with acid sites only resulted in PO isomerization products,
whereas Rb2SO4-modified iron silica led to seriously reduced
PO isomerization. The proposed mechanism, with PO as main
product and acid-catalyzed PO isomerization determining the
final product selectivity, is thus supported.
3.4.2. Thermogravimetric analysis
Thermogravimetric analysis (TGA) is a useful tool for study-
ing catalyst synthesis and deactivation. TGA analysis of a used
Rb2SO4 promoted catalyst is shown in Fig. 3; coke formation
was already reported in detail in Section 3.2.
3.4.4. XPS
XPS was performed to study the surface composition and
examine the deposited cokes. Both wide-scan surveys and high-
resolution multiplexes per element were performed on unpro-
moted and Rb2SO4-promoted iron oxide on silica (see Supple-
mentary material, Figs. S2–S3). However, the surface densities
of Fe and S were too low (atomic concentrations ∼0.2 at%) to
allow identification of the exact oxidation valence of both ele-
ments on the catalyst surface. Under strict reservation, the bind-
ing energy [referenced to Si2p (SiO2) at 103.4 eV] of 709.8 eV
of iron on the Rb2SO4-promoted catalyst could be best com-
To study catalyst synthesis, a batch of Rb2SO4-promoted
iron oxide on silica was prepared as stated in Section 2, but was
not calcined. Calcination was simulated with the TGA device,
and the mass loss was compared to that of pure silica (Fig. 4).
The two samples showed the same initial mass loss course:
a large weight loss at low temperature (348–373 K) attributed
to physisorbed water, followed by a slow weight loss due to loss
of chemisorbed water at much higher temperature (623–673 K).
However, whereas the mass of silica remained nearly constant
from 923 K, the Rb2SO4-promoted catalyst started to lose more
Fig. 4. Thermogravimetric weight and derivative weight loss (insert) curves obtained in air flow, measured for a fresh, uncalcined catalyst (A: Rb SO -modified
2
4
0.15 wt% FeO on silica with Rb/Fe = 12) in comparison with blanco silica support (B).
x

B. Moens et al. / Journal of Catalysis 247 (2007) 86–100
93
Fig. 5. X-ray diffraction of promoted FeO on silica with different amounts of Rb SO . Catalyst: 0.15 wt% iron; (A) 0, (B) 2.3, (C) 4.5, (D) 12, (E) 18, (F) 36, and
x
2
4
(G) 72 Rb/Fe.
pared to that of FeO (709.3–710.2 eV) and Fe2O3 (∼710.7 eV).
Due to the higher atomic mass, Rb was easier to detect; Rb
signals were clearly visible according to an atomic Rb concen-
tration of 0.8 at% on the promoted catalyst. No chloride could
be detected with XPS.
To analyze the cokes, XPS measurements of fresh and used
catalyst were compared (see Supplementary material, Fig. S4).
The atomic C concentration clearly rose with the use of this
catalyst, while the surface concentrations of other elements re-
mained unchanged. In principle, incorporation of O as carbonyl,
alcohol, or ether functions should be visible due to increasing
binding energy of the C signals. However, the binding energy
of the C signal did not rise as expected (fresh: 284.70 eV, used:
283.78 eV), indicating the presence of oxygen-free C deposits.
3.4.6. UV–vis spectroscopy
UV–vis DRS was carried out to study the coordination and
dispersion of the loaded iron. Fig. 7 shows the results of pro-
moted FeOx on silica with different amounts of iron, ranging
from 200 to 600 nm. Absorption in this region is attributed to
CT bands from ligand to isolated Fe3+ (O → Fe LMCT band,
200–335 nm) and d–d transitions (335–590 nm) [43], with the
latter covered by a strong tail of the LMCT bands [44]. In the
LMCT range, spectra were characterized by four distinct sig-
nals: 215–220, 245–250, 275, and 375 nm. According to the
literature [43,45–52], the first two signals can be assigned to
LMCT transitions (t1 → t2 and t1 → e) from isolated iron in
tetrahedral coordination, the signal at 275 nm can be assigned
to LMCT transition of isolated iron in octahedral coordination,
and the signal at 375 nm is usually attributed to octahedral Fe3+
in small oligonuclear clusters. The presence of large Fe2O3 par-
ticles is unlikely, because no LMCT feature above 450 nm is
present.
3.4.5. Raman spectroscopy
Raman spectroscopy was carried out to search for sulfate on
the calcined catalysts and study the optimization of promoter
concentration. Catalysts loaded with 0.1 wt% iron and differ-
ent amounts of Rb2SO4 were measured (Fig. 6). Below Rb
concentrations of Rb/Fe = 18, no clear signals could be dis-
tinguished, except for a small one at 1035 cm−1 first appearing
at Rb/Fe = 9. This signal did not originate from Rb2SO4, but
showed a strong resemblance to the νs1 vibration of Fe2(SO4)3
appearing at 1035–1045 cm−1. Starting from a concentration
of Rb/Fe = 18, five extra signals were detected at 445, 622,
977, 1105, and 1135 cm−1. According to the literature [41,42],
these signals can be assigned to the different S–O vibrations of
hexagonal Rb2SO4: symmetrical stretching (νs1 = 977 cm−1),
asymmetrical stretching (νs3 = 1105 and 1135 cm−1), and bend-
ing (νs2 = 445 and νs4 = 622 cm−1) vibrations. In conclusion,
for Rb/Fe < 9, no Rb2SO4 could be detected; sulfate remaining
after calcination was probably coordinated to iron sites, ex-
plaining the signal at 1035 cm−1. Raising the Rb concentration
above Rb/Fe = 9 clearly led to crystalline Rb2SO4, confirming
the XRD results.
The absorption observed for the silica support (Fig. 7, A),
with a band at 220 nm and a shoulder at 245 nm, is due to iron
impurities in the silica, attributed to framework Fe3+ in tetra-
hedral coordination. Raising the iron loading to 0.2 wt% iron
clearly caused a red shift (to 230 and 260 nm) (Fig. 7, B–E).
These bands can also be assigned to tetrahedral Fe, probably
located at extra-framework positions with asymmetric or dis-
torted coordination. The appearance of a shoulder at around
275 nm (Fig. 7, E) points to the presence of small amounts of
octahedral Fe3+. Above 0.2 wt% iron, the signals at 275 and
375 nm became dominant (Fig. 7, F), and they were clearly vis-
ible at 1 and 3 wt% iron (Fig. 7, G and H). Thus, at high Fe
concentration, isolated octahedral Fe3+ and small FexOy clus-
ters were formed in expense of tetrahedral (distorted) Fe3+
.
Fig. 8 shows the DRS spectra of FeOx on silica promoted
with different amounts of Rb sulfate compared with an un-
promoted catalyst. The latter (Fig. 8, A) showed a maximum
absorption around 255 nm with shoulders at 280 and 320 nm,
indicative for (distorted) tetrahedral Fe3+, octahedral Fe3+, and
small FexOy clusters. Up to Rb/Fe = 18 (Fig. 8, B–E), pro-

94
B. Moens et al. / Journal of Catalysis 247 (2007) 86–100
Fig. 6. Raman spectroscopy of promoted FeO on silica with different amounts of Rb SO . Catalyst: 0.1 wt% iron; Rb/Fe = 0 (A), 2.3 (B), 4.5 (C), 9 (D), 18 (E),
x
2
4
−1
36 (F), and 72 (G). Insert: Raman shifts enlarged between 1150 and 1000 cm
.
Fig. 7. UV–vis DRS spectra of Rb SO -promoted FeO on silica with different iron loadings, promoter/iron = 10. (A) 0 wt%, (B) 0.05 wt%, (C) 0.1 wt%,
x
2
4
(D) 0.15 wt%, (E) 0.2 wt%, (F) 0.4 wt%, (G) 1 wt%, and (H) 3 wt% iron.
motion with Rb2SO4 clearly caused a blue shift to a relatively
sharp absorption maximum around 245–255 nm in combina-
tion with a less obvious band at 230 nm, indicating the pres-
ence of tetrahedral Fe3+. At higher Rb concentrations (Rb/Fe >
18; Fig. 8, F and G) the signal at 230 nm became domi-
nant.

B. Moens et al. / Journal of Catalysis 247 (2007) 86–100
95
Fig. 8. UV–vis DRS spectra of promoted FeO on silica with different Rb SO concentrations. Catalyst: 0.15 wt% iron; Rb/Fe = 0 (A), 2.3 (B), 4.5 (C), 12 (D),
x
2
4
18 (E), 36 (F), and 72 (G).
The use of different sulfate promoters influences the disper-
sion of iron (see Supplementary material, Fig. S5). Although
all curves showed a maximum at 230–245 nm, corresponding
to well-dispersed (distorted) tetrahedral Fe3+, the major differ-
ence occurred at 300–450 nm, diagnostic of small iron oxide
clusters. Absorption intensities were highest for Na and Li and
lowest for Rb and K; thus, promotion with Rb2SO4 and K2SO4
produced the best iron oxide dispersion.
Kubelka–Munk transformation of the absorption data allows
the calculation of absorption edge energies or band gap energies
(Eg) [52–55], which give indirect information on the dispersion
and coordination of Fe. The edge energies were extracted using
Table 7
Edge energies in eV derived from the UV–vis DRS absorption spectra for
Rb SO (Rb/Fe = 10) promoted FeO on silica with different iron loadings
x
2
4
Entry Fe concentration Fe concentration Edge energy Number of nearest
2
a
b
(wt%)
(atom/nm )
(eV)
neighbors, N
Fe
1
2
3
4
5
0
0
4.21
4.23
4.14
4.15
4.06
3.85
3.58
3.58
1.86
3.94
4.13
4.12
4.17
4.22
4.33
4.34
0
0
0.05
0.1
0.15
0.2
0.4
1
0.011
0.022
0.032
0.043
0.086
0.22
0.25
0.22
0.56
1.36
2.41
2.40
8.92
1.01
0.33
0.34
0.15
0
6
7
8
3
–
0.65
–
c
d
e
f
9
10
11
12
13
14
15
16
0.15
0.15
0.15
0.15
0.15
0.15
0.15
0.032
0.032
0.032
0.032
0.032
0.032
0.032
the equation following Davis and Mott [56], [F(R ) × hν] ∼
∞
(hν − Eg)n, with n = 2, 3, 1/2, and 3/2 according to the pos-
sible electron transitions. The best fit was obtained for n = 2
and n = 1/2, suggesting spin allowed, Laporte (or symmetry)
partially forbidden (d–d) and spin allowed, Laporte allowed
(LMCT) electron transitions from O2− to Fe3+, respectively.
g
h
i
0
0
j
a
Edge energies are extracted by finding the intercept of the linearized low-
The edge energy values, derived from the x-intercept of the
1/2
1/2
energy region of a plot of [F(R ) × hν]
against hν, with hν is the incident
∞
linearized low-energy region of [F(R ) × hν]
against hν,
∞
photon energy and F(R ) the Kubelka–Munk expression for DRS reflection.
∞
are collected in Table 7 (entries 1–8; Supplementary material,
Fig. S6). An increase in Fe led to lower edge energies; an edge
energy of 1.86 eV was found for pure Fe2O3 (Table 7, entry 9).
Weber [53] and Gao and Wachs [55] proposed similar corre-
lations between Eg and the average number of neighbors and
local coordination structures in Mo6+ and V5+, respectively.
Accordingly, the decreased Eg with rising iron concentration
indicates a transformation of isolated tetrahedral Fe3+ into iso-
lated octahedral Fe3+ species and (tetrahedral or octahedral)
species with increased nuclearity. When varying the concentra-
tion of Rb sulfate on the catalysts, the edge energy clearly rose
with increasing Rb concentration (Table 7, entries 10–16) sug-
gesting an evolution toward perfectly isolated Fe species.
b
The average number of nearest Fe-neighbors is calculated using following
empirical expression [49]: number of nearest Fe-neighbors = 16 − 3.8 × E .
g
c
Pure Fe O .
2
3
d
e
f
Rb/Fe = 0.
Rb/Fe = 2.25.
Rb/Fe = 4.5.
Rb/Fe = 12.
Rb/Fe = 18.
Rb/Fe = 36.
Rb/Fe = 72.
g
h
i
j
3.4.7. EPR spectroscopy
EPR spectroscopy at X-band frequency was carried out to
study the local environment and coordination geometry of the

96
B. Moens et al. / Journal of Catalysis 247 (2007) 86–100
by combining EPR with other characterization techniques like
UV–vis DRS or Mössbauer [57], can unambiguous statements
about the coordination be made. According to earlier work [18,
48,50,57,59–61], signals at geff ∼ 4.3 and geff ꢀ 6 could be as-
signed to isolated Fe3+ in strong rhombic or axial distortion,
respectively, irrespective of their coordination. Signals around
geff = 2 can arise from isolated Fe3+ in high symmetry (sharp
signal) or from Fe3+ in FexOy clusters (broadened signal). The
Curie–Weiss law (intensity ∼ 1/T ) can be used to discriminate
between these species.
As shown in Fig. 10, the transition around geff ∼ 2 clearly re-
acted differently on cooling. For the unpromoted catalyst, cool-
ing to 120 K left the signal around geff ∼ 2 unchanged (a slight
rise in the global intensity), while the signal at geff = 4.39
clearly rose. For the promoted catalyst, the signal intensity
at geff = 4.39 also increased, whereas a sharp signal around
geff ∼ 2 became dominant. The broad signal at the same geff
value clearly decreased on cooling. This behavior suggests that
the region around geff ∼ 2 of the spectrum comprised overlap-
ping signals from different Fe3+ sites. The anti Curie–Weiss
behavior of the broad part at geff ∼ 2 suggests magnetic inter-
action between electron spins of neighboring Fe3+ ions caused
alignment to the external magnetic field, resulting in increased
intensity at higher temperatures. The temperature behavior indi-
cates that the sharp transition at geff = 2.01 can be attributed to
highly symmetric Fe3+, whereas iron oxide clusters gave rise to
the broad part of the signal at geff = 2.01. Comparison with the
EPR spectrum of pure Fe2O3 (Fig. 9, I) allows us to make a ten-
tative distinction between Fe3+ in large clusters (a new signal
at geff = 2.31) and Fe3+ in small FexOy clusters (geff = 2.01).
The support signal (Fig. 9, A; geff = 2.01) resulted from
small amounts of iron impurities in silica corresponding to per-
fectly symmetric framework Fe3+, probably in tetrahedral co-
ordination. On raising the amount of iron (Fig. 9, B–E) up to
0.2 wt%, the intensities of isolated iron signals (geff = 6.05,
4.39) increased, whereas the intensity of the FexOy signal
(geff = 2.01, broad) remained constant. The signal of highly
symmetric, isolated Fe3+ (geff = 2.01, sharp) slowly decreased
in intensity when the iron concentration was increased from
0.05 to 0.2 wt%. A further rise in the iron concentration above
0.2 wt% iron led to a steep decrease of the signals at geff = 6.05,
4.39, and 2.01 in favor of a signal attributed to FexOy clusters;
the sharp signal at geff = 2.01 completely disappeared at these
Fe contents. In bulk iron oxide, only a signal at geff = 2.31 was
found. In addition, the low intensities of the signals at 1 and
3 wt% iron were in agreement with spin–spin coupling in ox-
ides, resulting in extinction of the signal. Note that the presence
of EPR-invisible Fe2+ is less likely in our samples due to the
strongly oxidizing conditions during synthesis and the thermal
pretreatment under O2 before the EPR measurement.
Fig. 9. X-band EPR spectra at 120 K of Rb SO -modified FeO on silica with
x
2
4
different iron loadings at (A) 0 wt%, (B) 0.05 wt%, (C) 0.1 wt%, (D) 0.15 wt%,
(E) 0.2 wt%, (F) 0.4 wt%, (G) 1 wt%, (H) 3 wt% iron, and (I) pure Fe O .
2
3
Fig. 10. X-band EPR spectra of unmodified and Rb SO -promoted FeO on
x
2
4
silica at different temperatures. Catalyst: 0.05 wt% iron. (A) Unpromoted FeO
x
on silica at 298 K, (B) unpromoted FeO on silica at 120 K, (C) Rb SO
x
2
4
(Rb/Fe = 12) modified FeO on silica at 298 K, and (D) Rb SO (Rb/Fe = 12)
x
2
4
modified FeO on silica at 120 K.
x
paramagnetic Fe(III) ions (with spin S = 5/2) of the Rb2SO4-
promoted catalyst. The X-band spectra of the Fe catalyst with
varying amounts of iron at 120 K are shown in Fig. 9. Fig. 10
shows the effect of temperature on the EPR transitions for an
unpromoted and Rb2SO4-promoted 0.05 wt% FeOx on silica.
Three signals can be distinguished in the spectra: geff
=
4.39, accompanied by a shoulder at geff = 6.05 and a broad
signal at geff = 2.01. As described in the literature [48,57],
identification from the geff values only is not possible, because
not only the coordination number, but also the local distortion
of the Fe3+ ions influence the number and position of the EPR
transitions. Only a simulation of the EPR spectrum by solv-
ing the spin Hamiltonian [57,58] with the appropriate zero field
splitting parameters D and E will provide reliable informa-
tion about the degree of distortion. Additional difficulties arise
because the distortion can originate from either tetrahedral or
octahedral Fe3+ [57,58]. Only after studying additional aspects,
like the change of the EPR signals on dehydration [48,58], or
Variation of the Rb2SO4 concentration in the catalyst com-
position was also followed with EPR (see Supplementary ma-
terial, Fig. S8). At low Rb concentrations (A and B), the signals
of isolated Fe3+ were very weak, and the broad part of the sig-
nal at geff = 2.01 indicates the presence of small iron oxide
clusters. On raising the amount of Rb (C, D, and E), the signals
of distorted isolated iron became dominant, and the presence

B. Moens et al. / Journal of Catalysis 247 (2007) 86–100
97
of iron oxide clusters was minimized. Rubidium sulfate appar-
4.3. Emphasis on the exact role of Rb2SO4
ently helps keeping the sites isolated from each other.
The use of different promoting salts (e.g., chloride, bromide,
sulfate, and nitrate instead of acetate) influence the catalytic
performance, in contrast to a literature claim that the anion
in the promoter salt had no drastic effect [32–37]. We expe-
rienced a change in catalytic behavior using different anions
(Na > K > Rb > Cs for acetates and Rb ꢀ Cs > K > Na for
sulfates), and clearly showed here that both cation and anion
are crucial for an optimal promoting effect. We believe the ef-
fect of Rb2SO4 is dual; first, it has a structural effect on the
catalyst by neutralizing surface acidity and facilitating the iron
dispersion, with both Rb and sulfate playing roles. In addition,
there is also evidence that it directly influences the activity of
the catalyst, probably due to electronic effects (see further).
4. Discussion
4.1. Choice of metal and promoter
As stated before, there is literature agreement on the neces-
sity of iron oxide in propylene epoxidation with N2O [32–37].
However, there is no agreement about the best promoting ele-
ment and the exact role of the promoter. As such, we under-
took an intensive screening of metals and promoters. Among
the numerous metals tested, only iron was able to produce PO
with high selectivity. Surprisingly, chromium showed activity
toward propylene epoxidation. Because the epoxidation activ-
ity of chromium was much lower than that of iron, no further
attention was given to this system.
Using iron oxide on silica and different acetate salts, the
obtained catalytic results were in good agreement with some
literature data [36,37], although in conflict with others advanc-
ing the use of KCl and CsAc [32–37]. Differences in prepa-
ration method, silica source, iron loading, and promoter/iron
ratio likely influence the promoting effect of alkali metal salts.
Screening of different alkali and earth alkali metal salt promot-
ers showed that with our synthesis procedure, Rb2SO4 was the
best promoter for FeOx/silica.
4.3.1. Neutralization of surface acidity
As suggested in the literature [32,36,37], the cationic part
of the promoter salt neutralizes the surface acidity of the sup-
port. By combining the information of the IPA decomposition
and the PO feed test, PO destabilization, and thus decreased
PO selectivity, can be correlated with enhanced surface acid-
ity. Consequently, there is the need for a promoter to neutralize
the acidity. Indeed, fed PO is not stable in the presence of silica
or FeOx/silica in reaction conditions. On addition of Rb2SO4
this acidity almost fully disappears, leading to a much higher
stability of PO.
4.3.2. Enhanced iron dispersion
4.2. Optimization of the catalyst’s composition
Using a combination of UV–vis DRS and EPR spectroscopy,
it became clear that the promoter has a second structural
effect: influencing the Fe dispersion. Without Rb2SO4, iron
forms mainly large FexOy clusters already at low Fe con-
tent, which are responsible for deep oxidation [62,63]. With
Rb2SO4, probably due to competitive cation exchange between
cationic Fe3+-species and Rb+, the formation of FexOy clusters
diminishes in favor of (distorted) isolated iron oxide. The cat-
alytic results indeed show that highly dispersed iron is required
for selective epoxidation, and that the amount of cation is crit-
ical to achieving good iron dispersion. Besides the amount of
cation, the cation’s nature also affected the epoxidation results.
Using different alkali metal sulfates as promoter results in fol-
lowing order of degree of Fe dispersion: Rb > Cs > K > Na >
Li. Because this order matches that of the catalytic activity, we
conclude that the epoxidation activity is governed by the metal
dispersion, isolated iron oxide being the active site for epoxida-
tion.
Importantly, one cannot exclude contribution of the anion
to the improved dispersion; indeed, the presence of sulfate
ligands between the iron sites might prevent direct Fe–O–Fe
bridging. Raman experiments clearly show the presence of
S=O vibrations in the promoted catalysts, indicating that sul-
fate is present on the catalyst. This is true for high Rb sul-
fate loadings (Rb/Fe > 12), where SO24− appears as hexagonal
Rb2SO4, whereas at low Rb concentrations, a distinct signal at
1035 cm−1 is visible indicative of iron(III)-sulfate. This bears
a structural resemblance to iron phosphate structures reported
Optimization of the composition revealed that a catalyst with
0.15 wt% FeOx/silica, promoted with Rb2SO4 (with Rb/Fe ∼
10) gave the best epoxidation results. Up to 0.15 wt% iron ox-
ide, the activity per iron site remained constant (18 mmol PO
mmol−1 Fe h−1), whereas higher iron concentrations (>0.2
wt% iron oxide) led to lower oxidation activity per iron. Metal
clustering (FexOy), as evidenced by UV–vis DRS (shift to
lower energy bands and lower edge energies) and EPR [de-
crease of (distorted) isolated FeOx signals], resulted in deep
oxidation. Thus, highly dispersed isolated iron oxide seems to
be the active site that allows selective epoxidation.
The amount of promoter is crucial to achieving good PO se-
lectivity; when the amount of Rb2SO4 is too low (Rb/Fe < 10),
Fe dispersion is not optimal, resulting in low activity; more-
over, the Brönsted acidity of the catalyst was not fully neu-
tralized here, resulting in PO isomerization. Excess Rb2SO4
(Rb/Fe > 12) has no further beneficial effect on the activity and
selectivity of the epoxidation, but forms a crystalline Rb2SO4
phase on the catalyst, as evidenced by XRD and Raman. At
very high promoter loadings (Rb/Fe > 18), formation of basic
Rb2O from excess Rb2SO4 during catalyst pretreatment likely
explains the slight activity decrease. Thus, a catalyst with Rb
concentration of Rb/Fe = 10–12, having the highest concen-
tration of distorted isolated iron oxide (from EPR) in tetrahe-
dral coordination (from UV–vis), resulted in the highest PO
yield.

98
B. Moens et al. / Journal of Catalysis 247 (2007) 86–100
Fig. 11. Proposed reaction network for propylene epoxidation with N O and FeO catalyst.
x
2
for the selective oxidation of methane to methanol with N2O
[64,65]. Likewise, sulfate anions might form local structures in
which the iron cations are captured, thus preventing iron clus-
tering. As such, sulfate is involved in the dispersion of iron.
mation of propanal (and a little acetone), whereas allylic oxida-
tion products (such as allylic alcohol and acrolein) are formed
only in minor quantities. Moreover, acrolein should not be seen
as a primary product, because it can be formed from allylic al-
cohol. In light of the above observations (surface acidity and PO
stability), we propose a reaction mechanism (Fig. 11) in which
initially PO is the major reaction product. The epoxidation is
catalyzed by isolated tetrahedral Fe3+ sites. The presence of
FexOy clusters results in deep oxidation and formation of COx.
Small amounts of CH4, C2H4, and C2H6 and their oxygenated
counterparts (methanol, ethanol, and acetaldehyde) could be ex-
plained by acid-cracking reactions of propylene, followed by
oxidation. Due to acidity, PO isomerizes to propanal, acetone,
and allyl alcohol, which can be further oxidized to acrolein. In
addition, acid polymerization of secondary oxidation products
might lead to polymer structures eventually lying at the origin
of carbonaceous deposits, which deactivate the catalyst. Self-
polymerization of PO to polyether compounds, as proposed by
Ananieva and Reitzmann [32], also can contribute to deactiva-
tion. Analysis of the coke deposits should be easily made by IR
spectroscopy, but unfortunately, the amount of deposits is too
low to distinguish between both possibilities. XPS confirmed
formation of carbon deposits on a spent catalyst; however, the
binding energy of C was surprisingly not of oxygenated na-
ture. Perhaps graphite-like coke is formed, independent from
but dominant over the formation of oxygenated deposits. Fur-
ther study is required to unravel the deactivation mechanism.
For future application, because the catalytic activity can eas-
ily be restored on burning the coke, one might think of inte-
grating state-of-the-art continuous (as in FCC) or discontinuous
(with two reactor units) regeneration technologies.
4.3.3. Electronic effect
High iron dispersions are also obtained using acetate pro-
moters and not only by sulfate salts (see Supplementary mate-
rial, Fig. S7). As Rb2SO4 is a better promoter than Rb acetate,
other parameters besides iron dispersion influence the catalytic
activity. In literature, sulfate is known to enhance the catalytic
properties of vanadium catalysts in different oxidation reactions
[66–72]. Some authors have proposed that sulfate in the vicin-
ity of vanadyl sites electronically influences its redox proper-
ties, leading to enhanced catalytic performance [66]. Likewise,
the presence of a sulfate ligand on an iron site, explaining the
Raman signal at 1035 cm−1, might influence the electronic con-
figuration of iron and as such may explain the high activity
of rubidium sulfate-promoted iron catalysts. Further research
is ongoing to confirm this hypothesis.
4.4. Mechanistic proposal
As described in detail in the results section, PO selectivity is
primarily governed by acid–base properties of the catalyst. In
the absence of promoter, the relatively high acidity of the cat-
alyst results in consecutive PO isomerization, yielding low PO
selectivity; addition of promoter results in neutralization and
high PO selectivity (>70%). The amount of iron has no ma-
jor influence on PO selectivity; only at high iron concentration
(>0.6 wt% iron or 0.13 Fe atoms per nm2) are iron oxide clus-
ters formed, which are responsible for deep oxidation.
5. Conclusion
Previous papers [33–35] addressed a selective allylic oxi-
dation of propylene with N2O over unpromoted FeOx/silica.
Using H2-TPR, they observed that on modification with KCl,
the reduction of Fe3+ to Fe2+ shifts to higher temperature [33].
The lower iron reducibility was explained by inhibition of the
reactivity of lattice oxygen by the K+ promoter. Their allylic
oxidation is not entirely in line with our observations; in agree-
ment with other papers [32,36,37], we observed the main for-
The gas-phase epoxidation of propylene using N2O as an
oxidant was studied to screen for an ideal catalyst composition.
Only catalysts containing iron (and to some extent chromium)
oxide on silica show selective PO production when modified
with alkali and earth alkali salts. Using various promoters, it
became evident that both anion and cation are crucial for the

B. Moens et al. / Journal of Catalysis 247 (2007) 86–100
99
epoxidation activity and PO selectivity, and that Rb sulfate is
the modifier of choice. Variation of the amount of modifier and
Fe led to an optimal composition of 0.15 wt% Fe and Rb2SO4
(Rb/Fe = 12). This particular catalyst results in a propylene
conversion of 16.8% and a PO selectivity of 59% after 80 min
of reaction. As further raising the iron concentration leads to
FexOy formation and deep oxidation of propylene, only highly
dispersed iron oxide is capable of producing PO. This is con-
firmed by UV–vis DRS and EPR spectroscopy, suggesting the
active iron site has a distorted tetrahedral coordination. Raman
spectroscopy suggests the presence of sulfate in the vicinity of
iron, enhancing its redox properties. The promoter also plays a
structural role. Besides better iron dispersion, the promoter also
neutralizes acid sites, preventing consecutive reactions with the
formed PO. A mechanism for propylene epoxidation with N2O
is proposed based on acidity and PO stability experiments, and
suggests PO as the major primary product.
Because Rb2SO4-promoted FeOx/silica is the most promis-
ing catalyst combination, we have undertaken an extensive
physicochemical characterization study using additional tech-
niques (e.g., EXAFS, pulsed EPR, probe FT-IR, Mössbauer,
H2-TPR, magnetic circular dichroism and magnetization mea-
surements [73]) in an attempt to gain insight into the active iron
site and better understand the fundamentals of the catalyst de-
activation.
[13] L.M. Kustov, A.L. Tarasov, V.I. Bogdan, A.A. Tyrlov, J.W. Fulmer, Catal.
Today 61 (2000) 123.
[14] A. Ribera, I.W.C.E. Arends, S. de Vries, J. Perez-Ramirez, R.A. Sheldon,
J. Catal. 195 (2000) 287.
[15] V.I. Sobolev, K.A. Dubkov, E.A. Paukshtis, L.V. Pirutko, M.A. Rodkin,
A.S. Kharitonov, G.I. Panov, Appl. Catal. A 141 (1996) 185.
[16] V.I. Sobolev, A.S. Kharitonov, Ye.A. Paukshtis, G.I. Panov, J. Mol. Catal.
84 (1993) 117.
[17] A. Reitzmann, E. Klemm, G. Emig, Chem. Eng. J. 90 (2002) 149.
[18] P. Kubanek, B. Wichterlova, Z. Sobalik, J. Catal. 211 (2002) 109.
[19] V.N. Parmon, G.I. Panov, A. Uriarte, A.S. Noskov, Catal. Today 100
(2005) 115.
[20] B.R. Wood, J.A. Reimer, A.T. Bell, M.T. Janicke, K.C. Ott, J. Catal. 225
(2004) 300.
[21] K.A. Dubkov, V.I. Sobolev, G.I. Panov, Kinet. Catal. 39 (1998) 72.
[22] N.S. Ovanyesan, A.A. Shteinman, K.A. Dubkov, V.I. Sobolev, G.I. Panov,
Kinet. Catal. 39 (1998) 792.
[23] G.I. Panov, V.I. Sobolev, K.A. Dubkov, A.E. Parmon, N.S. Ovanyesan,
A.E. Shilov, A.A. Shteinman, React. Kinet. Catal. Lett. 61 (1997) 251.
[24] E.V. Kondratenko, J. Pérez-Ramirez, Appl. Catal. A 267 (2004) 181.
[25] R. Bulanek, B. Wichterlova, K. Novoveska, V. Kreibich, Appl. Catal.
A 264 (2004) 13.
[26] E.V. Kondratenko, M. Baerns, Appl. Catal. A 222 (2001) 133.
[27] S.S. Hong, J.B. Moffat, Appl. Catal. A 109 (1994) 117.
[28] M.B. Ward, M.J. Lin, J.H. Lunsford, J. Catal. 50 (1977) 306.
[29] E.V. Kondratenko, M. Cherian, M. Baerns, Catal. Today 112 (2006) 60.
[30] M.H. Groothaert, P.J. Smeets, B.F. Sels, P.A. Jacobs, R.A. Schoonheydt,
J. Am. Chem. Soc. 127 (2005) 1394.
[31] G. Ondrey, Chem. Eng. (New York) 111 (9) (2004) 17.
[32] E. Ananieva, A. Reitzmann, Chem. Eng. Sci. 59 (2004) 5509.
[33] X. Wang, Q. Zhang, Q. Guo, Y. Lou, L. Yang, Y. Wang, Chem. Com-
mun. 12 (2004) 1396.
[34] X. Wang, Q. Zhang, S. Yang, Y. Wang, J. Phys. Chem. B 109 (2005)
23500.
Acknowledgments
This work was supported by CECAT (Centre of Excellence,
Leuven University), IDECAT (European Centre of Excellence
Network), IAP, and GOA. The authors thank Kevin Huyghe for
assistance with the practical work. V. Meynen is grateful to the
FW-Flanders for financial support. Inside Pores (European Cen-
tre of Excellence Network) is also acknowledged.
[35] Q. Zhang, Q. Guo, X. Wang, T. Shishido, Y. Wang, J. Catal. 239 (2006)
105.
[36] D. Hönicke, V. Duma, W. Krysmann (Creavis Gesellschaft fuer Technolo-
gie und Innovation mbH), US Patent 6392065 (2002).
[37] Duma, D. Hönicke, J. Catal. 191 (2000) 93.
[38] D.L. Trent (The Dow Chemical Company), Kirk Othmer Encyclopedia
Chem. Technol. 20 (1996) 271.
[39] A. Gervasini, A. Auroux, J. Catal. 131 (1991) 190.
[40] Powder Diffraction File, Alphabetical Indexes, Inorganic Phases Sets 1–
45, ICDD, 1995.
Supplementary material
[41] D. Liu, H.M. Lu, J.R. Hardy, Phys. Rev. B 44 (1991) 7387–7393.
[42] Y.I. Yuzyuk, V.I. Torgashev, R. Farhi, I. Gregora, J. Petzelt, P. Simon,
D. De Sousa Meneses, L.M. Rabkin, J. Phys. Condens. Matter 10 (1998)
1157–1173.
[43] S. Bordiga, R. Buzzoni, F. Geobaldo, C. Lamberti, E. Giamello, A. Zec-
china, G. Leofanti, G. Petrini, G. Tozzola, G. Vlaic, J. Catal. 158 (1996)
486.
The online version of this article contains additional supple-
mentary material.
Please visit DOI:10.1016/j.jcat.2007.01.001.
References
[44] A.B.P. Lever, Inorganic Electronic Spectroscopy, Studies in Physical and
Theoretical Chemistry, vol. 33, second ed., Elsevier, 1984, pp. 452–453
and 161–173.
[1] T. Levander, Atmos. Environ. Part A Gen. Top. 24 (1990) 2707.
[2] W.C. Togler, Coord. Chem. Rev. 187 (1999) 303.
[3] J. Jia, K.S. Pillai, W.M.H. Sachtler, J. Catal. 221 (2004) 119.
[4] A. Waclaw, K. Nowinska, W. Schwieger, Appl. Catal. A 270 (2004) 151.
[5] Q. Zhu, R.M. van Teeffelen, R.A. van Santen, E.J.M. Hensen, J. Catal. 221
(2004) 575.
[45] G. Centi, S. Perathoner, R. Arrigo, G. Giordano, A. Katovic, V. Pedula,
Appl. Catal. A 307 (2006) 30.
[46] G. Centi, S. Perathoner, F. Pino, R. Arrigo, G. Giordano, A. Katovic,
V. Pedula, Catal. Today 110 (2005) 211.
[6] L. Capek, P. Kubanek, B. Wichterlova, Z. Sobalik, Collect. Czech. Chem.
Commun. 68 (2003) 1805.
[47] F. Arena, G. Gatti, G. Martra, S. Coluccia, L. Stievano, L. Spadaro, P. Fa-
mulari, A. Parmaliana, J. Catal. 231 (2005) 365.
[7] E.J.M. Hensen, Q. Zhu, R.A. Van Santen, J. Catal. 220 (2003) 260.
[8] D.P. Ivanov, V.I. Sobolev, G.I. Panov, Appl. Catal. A 241 (2003) 113.
[9] A.A. Ivanov, V.S. Chernyavsky, M.J. Gross, A.S. Kharitonov, A.K. Uri-
arte, G.I. Panov, Appl. Catal. A 249 (2003) 327.
[48] M.S. Kumar, M. Schwidder, W. Grünert, A. Brückner, J. Catal. 227 (2004)
384.
[49] C. Nozaki, C.G. Lugmair, A.T. Bell, T. Don Tilley, J. Am. Chem. Soc. 124
(2002) 13194.
[10] D. Meloni, R. Monaci, V. Solinas, G. Berlier, S. Bordiga, I. Rossetti,
C. Oliva, L. Forni, J. Catal. 214 (2003) 169.
[50] V. Umamaheswari, W. Böhlmann, A. Pöppl, A. Vinu, M. Hartmann, Mi-
cropor. Mesopor. Mater. 89 (2006) 47.
[11] K. Yoshizawa, Y. Shiota, T. Kamachi, J. Phys. Chem. B 107 (2003) 11404.
[12] B. Louis, P. Reuse, L. Kiwi-Minsker, A. Renken, Appl. Catal. A 210
(2001) 103.
[51] N. Maxim, A. Overweg, P.J. Kooyman, J.H.M.C. van Wolput, R.W.J.M.
Hanssen, R.A. van Santen, H.C.L. Abbenhuis, J. Phys. Chem. B 106
(2002) 2203.

100
B. Moens et al. / Journal of Catalysis 247 (2007) 86–100
[52] Q. Zhang, Q. Guo, X. Wang, T. Shishido, Y. Wang, J. Catal. 239 (2006)
105.
[62] F. Arena, G. Gatti, G. Martra, S. Coluccia, L. Stievano, L. Spadaro,
P. Famulari, A. Parmaliana, J. Catal. 231 (2005) 365–380.
[63] A. Parmaliana, F. Arena, J. Catal. 167 (1997) 57–65.
[64] Y. Wang, K. Otsuka, J. Mol. Catal. A Chem. 111 (1996) 341.
[65] A.V. Annapragada, E. Gulari, J. Catal. 123 (1990) 130.
[66] E. Garcia-Bordejé, J.L. Pinilla, M.J. Lazaro, R. Moliner, J.L.G. Fierro,
J. Catal. 233 (2005) 166–175.
[67] A. Martin, U. Bentrup, G.U. Wolf, Appl. Catal. A Gen. 227 (2002)
131–142.
[68] D. Klissurski, N. Abadzhijeva, React. Kinet. Catal. Lett. 2 (1975)
431–438.
[53] R.S. Weber, J. Catal. 151 (1995) 470.
[54] X. Gao, S.R. Bare, B.M. Weckhuysen, I.E. Wachs, J. Phys. Chem. B 102
(1998) 10842.
[55] X. Gao, I.E. Wachs, J. Phys. Chem. B 104 (2000) 1261.
[56] E.A. Davis, N.F. Mott, Philos. Mag. 22 (1970) 903.
[57] B.M. Weckhuysen, In-Situ Spectroscopy of Catalysts, American Scien-
tific Publishers, Stevenson Ranch, CA, 2004, pp. 220–222 and 240–
243.
[58] D. Goldfarb, M. Bernardo, K.G. Strohmaier, D.E.W. Vaughan, H. Tho-
mann, J. Am. Chem. Soc. 116 (1994) 6344.
[69] U. Bentrup, A. Martin, G.U. Wolf, Catal. Today 78 (2003) 229–236.
[70] D. Monti, A. Reller, A. Baiker, J. Catal. 93 (1985) 360–367.
[71] D.W.B. Westerman, N.R. Foster, M.S. Wainwright, Appl. Catal. 3 (1982)
151–160.
[59] A.M. Ferretti, C. Oliva, L. Forni, G. Berlier, A. Zecchina, C. Lamberti,
J. Catal. 208 (2002) 83.
[60] H. Guo, X. Zhang, M. Cui, R. Sharma, N. Yang, D.L. Akins, Mater. Res.
Bull. 40 (2005) 1713.
[72] N.R. Foster, M.S. Wainwright, D.W.B. Westerman, Aust. J. Chem. 34
(1981) 1325–1330.
[61] P. Selvam, S.E. Dapurkar, S.K. Badamali, M. Murugasan, H. Kuwano,
Catal. Today 68 (2001) 69.
[73] G.D. Pirngruber, P.K. Roy, R. Prins, Phys. Chem. Chem. Phys. 8 (2006)
3939–3950.