Gas phase epoxidation of propene by nitrous oxide over silica-supported iron oxide catalysts

Article abstract of DOI:10.1006/jcat.1999.2769

Catalysts consisting of sodium-promoted silica gel-supported iron oxide catalyzed the gas phase epoxidation of propylene by N₂O. Selectivities to propylene oxide of 40-60% at propylene conversion of 6-12% were observed. The effect of different

Full text of DOI:10.1006/jcat.1999.2769

Journal of Catalysis 191, 93–104 (2000)
doi:10.1006/jcat.1999.2769, available online at http://www.idealibrary.com on
Gas Phase Epoxidation of Propene by Nitrous Oxide
over Silica-Supported Iron Oxide Catalysts
Viorel Duma and Dieter Ho¨nicke
Department of Chemical Technology, Technical University of Chemnitz, 09107 Chemnitz, Germany
Received August 11, 1999; revised November 23, 1999; accepted November 23, 1999
oxygen species will be necessary. However, the molecular
oxygen produces strong electrophilic oxygen species over
noble metals and nucleophilic species over transition metal
oxides.
Generally, the reaction between oxygen species and a
propene molecule will occur, dependent on the electronic
properties of the oxygen species, following one of three
possible reaction paths (Fig. 1):
Catalysts consisting of sodium-promoted silica gel-supported
iron oxide were found to catalyze the gas phase epoxidation of
propene by nitrous oxide. Selectivities to propene oxide of 40–60%
at propene conversions in the range of 6–12% were observed. The
influence of different parameters on the yield to propene oxide is
discussed: the amount and the dispersity of iron oxide, the proce-
dure of alkali promotion, the morphology of the support, and the
c
reaction parameters.
2000 Academic Press
1. The reaction between mild electrophilic oxygen
species and propene will take place at the -bond of
propene. In this way will emerge propene oxide, propanal,
and acetone. High temperatures and acidic or basic condi-
tions will favor, directly or by the isomerization of propene
oxide, the formation of propanal and acetone (1a, 2, 5),
so that such circumstances have to be avoided in order to
improve the selectivity to propene oxide.
2. Strong electrophilic oxygen species will attack all
C–C bonds. Thus, the reaction products will be carbon ox-
ides and organic molecules having one or two carbon atoms.
3. Nucleophilic oxygen species will attack preferentially
H atoms in the allylic position. This reaction path will lead
to products like allyl alcohol, acrolein, and acrylic acid.
Key Words: propene oxide; epoxidation; gas phase; nitrous oxide;
iron oxide catalysts.
INTRODUCTION
Propene oxide is one of the most important organic
intermediates and is used for the production of many
useful chemicals. Today propene oxide is mainly manu-
factured by one of two commercial processes viz. the
organic-hydroperoxide process or the chlorohydrin pro-
cess (1a, 2), both of them performed in the liquid phase.
The compulsory formation of a side product in the former
and the chlorine used in the latter process are substantial
drawbacks of these industrial processes. Therefore, the
development of a process for the direct gas phase epoxi-
dation of propene which overcomes these disadvantages
would be of great economical importance.
Some alkenes like ethene and styrene, which have no al-
lylichydrogen atoms, can be directlyepoxidized with molec-
ular oxygen in the gas phase over Ag catalysts. The process
for the direct epoxidation of ethene over Ag catalysts was
developed in the past (1b) and has been performed world-
wide for six decades. However, over the same or similar
catalysts the propene oxidation leads to propene oxide se-
lectivities of less than 15% at a conversion degree of less
than 15% (2). Some recent patents claim that reaching high
selectivities to propene oxide over novel supported pre-
cious metal catalysts is possible; however, the yields are
generally under 1% (3, 4).
A particular case is silver, which generates a mild elec-
trophilic, peroxide-like oxygen species but a nucleophilic
one as well. Therefore, alkenes which have no allylic hydro-
gen atoms can be selectively oxidized to epoxides over Ag
catalysts. However, alkenes such as propene, which possess
allylic hydrogen atoms are oxidized in the both positions,
vinyl and allyl.
Our approach to finding an effective method for the
gas phase epoxidation of propene was to use an oxygen-
containing compound as an oxidant and a solid catalyst
which can adsorb and activate it in such a manner that only
mild electrophilic species will be generated.
Nitrous oxide has certain properties which make it suit-
able as a selective oxidant for the epoxidation of propene.
Energetically, the release of an oxygen atom from N2O is
more favorable than from the oxygen molecule:
The interaction between molecular oxygen and met-
als or metal oxides will produce electrophilic and nucleo-
philic oxygen species. For the transformation of the vinyl
group of an alkene in an oxirane ring, a mild electrophilic
N2O → N2 + O
1r H2098 = 167.6 kJ/mol;
1r G0298 = 128.0 kJ/mol
93
0021-9517/00 $35.00
2000 by Academic Press
All rights of reproduction in any form reserved.
c
Copyright

¨
DUMA AND HONICKE
94
to improve the adsorption capacity of silica the use of a
highly dispersed transition metal oxide as an active compo-
nent should be effective. In a comparison (13) between the
activity of oxides of several 3d transition metals in the N2O
decomposition, it is shown that Fe3+ and Fe2+ ions have
the lowest level of activity. In another investigation (14) of
the N2O decomposition activities of several catalysts, sup-
ported iron oxide shows again one of the lowest rates of
activity. Drago et al. (15) found as well a very low activity
level of silica-supported iron oxide in the decomposition of
N2O. Furthermore, alkali promotion and high temperature
calcination are suitable measures for reducing the acidity
of the catalyst. For these reasons, catalysts were prepared
in the present study by doping various silica gels with small
amounts of iron oxide, followed by sodium impregnation
and calcination (16, 17).
FIG. 1. Possible reaction paths of the reaction between propene and
oxygen species.
1 O2 → O
1r H2098 = 249.2 kJ/mol;
1r G0298 = 231.7 kJ/mol.
2
EXPERIMENTAL
Materials
The distribution of electrons in the N2O molecule can be
described by two mesomeric forms in resonance with each
other (1c):
A. Catalyst Preparation
Group a. The silica gel support for this group of cata-
lysts was prepared after (18) from tetraethylorthosilicate
(TEOS), ethanol, i-propanol, 1-dodecylamine, and water.
After calcination in air at 873 K, portions of this silica
gel support (further named D) were impregnated by in-
cipient wetness method with solutions of different con-
centrations of Fe(III)–acetylacetonate in toluene. The so-
prepared catalysts were dried and then calcined at 873 K in
air. To decrease the acidity of the catalysts (neutralization
procedure), they were impregnated for 48 h in an aqueous
solution of sodium acetate (Naac) and, after filtration and
washing with distilled water, calcined at a temperature of
973 K for 6 h in air.
+
== + ==
–
O
O ↔ N
.
N
N
N
This distribution makes possible the adsorption of the N2O
molecule at cationic sites (6) and, consequently, the oxygen
atom will obtain electrophilic properties (Fig. 2).
Several authors have already used N2O as a selective
oxidant but not for propene oxidation. Ohtani et al. (7)
partially oxidized various organic substrates viz. alcohols,
ethers, and amines in an aqueous suspension of platinum
particles. Other authors (8–12) used N2O for the partial oxi-
dation of benzene to phenol. They chiefly used metal-doped
zeolites as catalysts. Such catalysts cannot be used in the
epoxidation of propene for two reasons: the high acidity of
the catalysts and the strong activities in the decomposition
of N2O with the formation of very reactive oxygen species.
The catalyst for the epoxidation of propene must adsorb
and activate the N2O molecule without decomposingit. Fur-
thermore, the catalyst must not have strong acidic and basic
properties because under such conditions propene oxide
can easily isomerize to propanal and acetone (1a, 2, 5) as
already mentioned. Thus, silica gel was chosen as the cata-
lyst support, which offers a high surface area, low acidity,
and no intrinsic activity in N2O decomposition. In order
Group b. The support used for this group of catalysts
was a silica gel from Degussa (Aeroperl 380/50, hereafter
referred to as A), produced from SiCl4 by the flame hy-
drolysis process. The catalysts were prepared by the same
procedure as the group a catalysts.
Group c. For this group of catalysts a silica gel with an
iron content of 6 ppm from Merck (M60 reinst, hereafter
referred to as Mr) was used as support. The catalysts were
prepared by the same procedure described for the group a
catalysts.
The iron loadings of the catalysts were 100, 300, 1000, and
10,000 ppm. The concentrations of the aqueous solutions
of Naac used in the neutralization procedure were 0.01 M
Naac (abbreviated N1), 0.1 M Naac (N2), and 1 M Naac
(N3).
Nomenclature of the catalysts. The name of the cata-
lysts was formed from a symbol for the support (D for the
group a, A for group b, and Mr for group c), followed by the
FIG. 2. Reaction step between propene and nitrous oxide toward
propene oxide.

GAS PHASE EPOXIDATION OF PROPENE BY NITROUS OXIDE
95
chemical symbol for iron, and a number which indicates the
Temperature-programmed desorption (TPD). The
iron content of the catalyst in ppm, followed by the abbre- TPD was carried out on catalyst samples (ca. 0.3 g) which
viation which indicates the concentration of the aqueous were filled in a TPD cell made of quartz. The TPD cell was
sodium acetate solution used in the neutralization proce- placed into an electrically heated oven and was connected
dure. For example, AFe300N2 is the name for a catalyst to a main system which permitted either the outgassing of
with a support consisting of Aeroperl 380/50. The iron con- the cell or the gas supplying. The catalyst sample in the TPD
tent is 300 ppm and the concentration of the neutralizing cell was outgassed at a pressure of less than 10 6 mbar for
solution was 0.1 M Naac.
30 min at room temperature. The TPD cell was then heated
to 973 K and evacuated at this temperature for 60 min.
The temperature of the cell was then adjusted to the
adsorption temperature and the adsorbate gas was per-
mitted to enter the TPD cell. After 30 min, the TPD cell
was cooled down to room temperature and evacuated
B. Reactant Gases
The reactant gases used were propene of 99.5 vol% mini-
mum purity, N2O of 99.5 vol% minimum purity, and helium
of 99.996 vol% minimum purity, all of them from Messer
Griesheim.
2
up to 10 mbar. The TPD cell was then connected to
a quadrupole mass spectrometer (QTMD-System from
Fisons Instruments) and evacuated up to a pressure of less
Catalyst Characterization
6
than 10 mbar for 2 h. The TPD cell was heated up to
973 K with a heating rate of 5 K/min and then held at this
temperature for 20 min. Six different molar masses (M)
were continuously monitored in the course of desorption.
The catalysts were characterized by means of nitrogen
physisorption, powder X-ray diffraction (XRD), temp-
erature-programmed reduction and oxidation (TPR-TPO),
and temperature-programmed desorption (TPD).
Oxidation Experiments and Product Analysis
Determination of textural properties. The specific sur-
face areas and the porosities of the supports and cata-
lysts were established by nitrogen physisorption at liq-
uid nitrogen temperature (77 K) according to the BET
method. The adsorption–desorption isotherms were re-
corded with a Sorptomatic 1990 apparatus (Fisons Instru-
ments), equipped with a Milestone 200 software. A surface
area requirement of 0.162 nm2 for the physically adsorbed
N2 molecule was used for the calculation of the BET sur-
face area. The pore size distribution was calculated with
the Dollimore–Heal method. The samples were outgassed
at 130 C for 2 h and then at room temperature overnight
prior to the measurements; the residual pressure was about
10 6 mbar.
The propene oxidation experiments were carried out in
a fixed-bed quartz-tube reactor (10 mm i.d.) under an ab-
solute pressure of 1200 mbar using a continuous-flow sys-
tem. The reactor, placed vertically, was surrounded by an
electrical heater. The reactor temperature was measured
and controlled via a thermocouple placed within the cata-
lyst bed. The gaseous reactants were fed using mass flow
controllers from Brooks. The reaction products were ana-
lyzed on-line using an IR-photometer from Rosemount to
detect CO2 and a gas chromatograph 5890 II Plus from
Hewlett-Packard for the other products. The gas chro-
matograph was equiped with two capillary columns viz.
Molsieb 5A and FFAP connected with a TCD and a FID,
respectively, which enabled the analysis of O2, N2, CO,
and N2O by the former and organic products by the latter
detector.
Powder X-ray diffraction. Powder XRD was carried out
in a HZG4 diffractometer from Seifert-FPM using a gener-
ator at 40 kV and 35 mA, a Ni filter, and CuK radiation
The catalyst (usually 1 g) was placed into the reactor, near
the bottom, on a porous quartz frit. Before each propene
˚
( = 1.540598 A).
Temperature-programmed reduction and oxidation oxidation experiment, the catalyst was oxidized at 793 K
(TPR-TPO). The apparatus consisted of a gas supplying in an air flow in order to remove remaining organic com-
unit, a quartz-tube reactor (6 mm i.d.) and a thermal pounds from the previous experiment and to yield a high
conductivity detector. The reactor with a catalyst sample oxidation state of the active component. When no more
(ca. 0.3 g) was placed into an electrically heated oven. A gas CO2 was detected in the outgas, the reactor temperature
stream of 30 ml/min with 10% H2 in N2 (for TPR) or of 40 was adjusted to the desired oxidation temperature for the
ml/min with 12.5% O2 in He (for TPO) was passed through reaction experiment. Then, the air flow was replaced by a
the catalyst sample. The temperature of the sample was flowmixture ofN2O (the needed flowvalue for the reaction)
controlled via a thermocouple and was raised from room and He (the rest to balance the needed total flow). When
temperature at 30 K/min to 973 K and then held constant the N2O concentration at the outlet of the reactor was con-
at this temperature for 30 min. The water produced by stant, propene was added to the flow mixture and the flow of
the reduction was condensed in a CO2(s)/i-propanol cooled He was correspondingly diminished to keep the total flow
trap. The amount of H2/O2 consumption was detected via constant. This moment was considered as the initial time
the thermal conductivity cell.
point (time-on-stream = 0) for the oxidation experiment.

¨
96
DUMA AND HONICKE
TABLE 1
Textural Parameters of Selected Supports and Catalysts
Support catalyst:
Group a
DFe300N2
Group b
AFe300N2
Group c
MrFe300N2
Parameter
D
A
Mr
Specific surface area (m2/g)
1167
1.84
511
0.74
314
1.28
229
1.38
420
0.83
332
0.63
Specific pore volume (cm3/g)
Relative volume (% ) for pore ranges
100–10 nm
10–5 nm
5–2 nm
2–1.5 nm
1.5–1 nm
<1 nm
55.1
1.8
4.2
8.7
30.0
0.0
64.4
7.3
6.9
5.2
16.0
0.0
79.3
6.4
6.7
0.8
2.0
0.0
86.8
10.1
2.7
0.0
0.0
0.7
4.4
87.7
5.3
1.7
0.0
0.3
4.2
89.7
4.2
1.3
0.0
0.1
During the oxidation experiments the reaction parameters pore volume, but after the iron and sodium impregnation
(pressure, temperature, gas flow composition, and velocity) and calcination the same parameters of the corresponding
were kept constant and the reaction products were contin- catalyst (DFe300N2) had less than 50% compared to the
uously analyzed by the IR-photometer and periodically by initial values. Especially the narrow pores of the catalyst
the gas chromatograph. After a period of time the experi- suffered a severe shrinking. The catalysts prepared with
ment was stopped; the flows of propene, nitrous oxide, and the supports A and Mr show surface area losses of less
helium were interrupted and the catalyst was regenerated than 30% compared to the related supports. They preserved
by oxidation at 793 K in a flow of air.
their specific pore structure with larger pores in the case
of A and with a narrow distribution in the range of slim
mesopores in the case of Mr.
The XRD spectra for support A, some selected catalysts,
and a sample of iron oxide prepared from the same iron
precursor, using the same procedure as with the catalysts,
RESULTS
Catalyst Characterization
The determined textural parameters of selected supports are shown in Fig. 3. The XRD spectra of the support and
and catalysts are summarized in Table 1. The support D catalysts showed no crystalline features for silicium oxide
had a very high specific surface area and a large specific or for iron oxide. In contrast, the iron oxide sample showed
FIG. 3. XRD spectra of support A, selected catalysts, and an Fe2O3 sample.

GAS PHASE EPOXIDATION OF PROPENE BY NITROUS OXIDE
97
FIG. 4. TPR spectra of Na-promoted and unpromoted silica supported Fe2O3 catalysts.
the characteristic peaks of different Fe2O3 crystalline reduction temperatures toward higher values for the Na-
phases.
impregnated catalysts than for the corresponding catalysts
The TPR spectra for catalysts with 300 and 1000 ppm Fe which were not Na-impregnated.
content as well as with and without sodium are shown in
The desorption spectra of propene using catalyst
Fig. 4. Despite the weak intensity of the signals due to the AFe300N2 is depicted in Fig. 5. It shows three desorption
small amounts of iron oxide, one can see a shifting of the regions with signals at about 350, 730, and 950 K in which
FIG. 5. TPD spectra of propene desorbed from the catalyst AFe300N2.

¨
98
DUMA AND HONICKE
FIG. 6. TPD spectra of nitrous oxide by the desorption from the catalysts (a) AFe300N2, (b) AFe10000N2, (c) MrFe300N2, and (d) AFe300N2,
DFe300N2, and MrFe300N2 (only the M44 signal).
various decomposition fragments of propene appear. Six
The desorption spectra of nitrous oxide (Fig. 6) from
molar masses of 12, 15, 16, 27, 41, and 42 were monitored. the catalysts AFe300N2 (Fig. 6a), AFe10000N2 (Fig. 6b),
The first desorption region for propene can be assigned to and MrFe300N2 (Fig. 6c) show the six monitored molar
the desorption of physically adsorbed propene and the sec- masses of 12, 14, 16, 28, 32, and 44 and mirrors two impor-
ond and third region to that of chemisorbed species.
tant features. On one hand, the N2O signal (molar mass 44,

GAS PHASE EPOXIDATION OF PROPENE BY NITROUS OXIDE
99
FIG. 6—Continued
M44) was always accompanied by signals stemming from sition of N2O and the recombination of two oxygen atoms,
nitrogen (M14) and oxygen (M16), which can emerge from and which did not necessarly accompany the N2O signal,
N2O decomposition in the vacuum chamber of the mass was used as an indication for a dissociative adsorption of
spectrometer. As far as these signals always accompanied N2O in the respective case.
the N2O desorption signal, they were not used to interpret
Three distinct groups of desorption peaks appear, one of
the nature of the N2O adsorption. On the other hand, the them slightly over room temperature (ca. 350 K) and the
appearance of O2 (M32), which necessitates the decompo- other two at high temperatures of about 650 and 950 K.

¨
DUMA AND HONICKE
100
The low-temperature group of desorption peaks can be as-
signed to a physically adsorbed N2O form, while the high-
temperature groups of desorption peaks were assigned to
a chemically adsorbed N2O form. In the TPD of N2O from
the supports A, D, and Mr (not shown) only the low temper-
ature desorption peaks at ca. 350 K appeared. The catalyst
with an iron loading of 1% (AFe10000N2) showed a strong
desorption peak of molecular oxygen (M32) at about 650 K
(Fig. 6b). The two depicted desorption spectra for catalysts
with 300 ppm Fe (AFe300N2 and MrFe300N2) showed no
or veryweak desorption peaksofmolecular oxygen (Figs. 6a
and 6c). The catalysts with 1000 ppm Fe (not shown) pre-
sented weak M32 desorption signal intensity. The compar-
ison between the desorption spectra for M44 from three
catalysts with an iron content of 300 ppm (Fig. 6d) showed
similar intensitiesat ca. 350K but different onesat ca. 650K.
That means that the physisorption ability of these catalysts
is similar but the chemisorption ability increases in the or-
der DFe300N2 < AFe300N2 < MrFe300N2.
FIG. 8. Reaction rate ofpropene and selectivityto propene oxide over
catalysts with different iron loadings. Reaction parameters: 1% propene,
15% N2O, 84% He; GHSV = 4 L h 1 g 1 (STP); TOS = 77 min.
centration of 15% N2O and then remained approximately
constant. After reaching maximum values, the reaction rate
had a slight decrease with the time on stream. The selec-
tivity toward propene oxide also increased with the N2O
concentration up to a concentration of 15% and increased
in the first 100 min of TOS for all N2O concentrations. For
N2O concentrations of 10% or more the selectivity reached
a maximum after 100–200 min TOS and then began to de-
crease. The higher the N2O concentration, the higher the
decrease. As a consequence of these evolutions, the yield
to propene oxide generally reached the maximum at a con-
centration of 15% N2O. Similar results were obtained for
the other catalysts as well.
The influence of the iron content of the catalysts on the
catalytic performance is examplified in Fig. 8 for a series of
catalysts. All of them have the same support (A, i.e., Aerop-
erl 380/50) with large pores which do not restrict the iron ox-
ide particles grown during the preparation procedure and
were Na-promoted after the same procedure. The results
were recorded after a time on stream of about 77 min. The
reaction rate increased in the entire investigated tempera-
ture range with the iron loading and the selectivity reached
a maximum at an iron loading of 300 ppm. Catalysts with
an iron loading of 1% (10000 ppm; not shown) had a higher
activity than the ones with 1000 ppm but the selectivity to
propene oxide was very low.
In addition, CO adsorption measurements were carried
out over prereduced (in a flow of H2) catalyst samples in
order to determine the exposed surface and dispersity of
iron, but because of the very small amounts of iron it was
not possible to carry out quantitative determinations.
Oxidation Experiments
In order to investigate the influence of the N2O concen-
tration on the catalytic results, a series of catalytic runs were
carried out at a constant propene concentration of 1% and
different N2O concentrations. The effect of the N2O con-
centration and the time on stream (TOS) on the reaction
rate and selectivity for the catalyst DFe300N2 can be seen in
Fig. 7. For the set propene concentration, the reaction rate
increased with increasing N2O concentration up to a con-
As already mentioned, in order to neutralize the acidity
of the catalysts, they were impregnated with a Na-solution
and after that they were calcinated. In Fig. 9 a compar-
ison is shown between the activities (Fig. 9a) and selec-
tivities to propene oxide (Fig. 9b) for a series of catalysts
with an iron loading of 300 ppm but which have under-
gone different Na-impregnation treatments. The oxidation
activity expressed as propene reaction rate generally in-
creased with the sodium loading (Fig. 9a). The selectivity to
propene oxide was very low over the non-sodium-impreg-
nated catalyst, increased with the Na loading, reached
a maximum for the impregnation in the intermediate
FIG. 7. Reaction rate of propene and selectivity to propene oxide
at different N2O concentrations. Reaction parameters: 1% propene, N2O
variable, balance He; catalyst DFe300N2; reaction temperature 648 K;
GHSV = 4 L h 1 g 1 (STP).

GAS PHASE EPOXIDATION OF PROPENE BY NITROUS OXIDE
101
TABLE 2
Conversions and Selectivities over MrFe300N2
Temperature (K)
573 598 623 648 673 698 723
0.6 2.5 6.1 10.1 11.9 20.3 29.0
62.3 70.0 60.1 48.1 39.7 23.9 12.2
Conversion propene (% )
Selectivities (% )
Propene oxide
Propanal
Acetone
Allyl alcohol
Acrolein
Carbon monoxide
Carbon dioxide
9.1
5.3
7.0
5.7
6.1
2.9
2.9
2.5
6.8
7.8
5.1
3.2
6.6
9.6 11.0
7.1
5.0
8.2
7.9
5.2
3.4
8.2
5.6
3.9
10.9
8.8
5.2
3.8
5.8
4.5
0.0
0.0
6.9 10.7 14.8 26.3 39.0
4.3 9.5 8.9 15.5 19.5
Note. Reaction parameters: 1% propene, 15% N2O, 84% He; GHSV =
4 L h 1 g 1 (STP); TOS = 128 min.
In order to find out the regeneration capability of the
catalyst, two reaction experiments were carried out with
the same catalyst under identical conditions (Fig. 12). Af-
ter the first oxidation run was carried out, seven reaction–
regeneration cycles were performed at different tempera-
tures. The ninth reaction experiment was then carried out
FIG. 9. Reaction rate of propene (a) and selectivity to propene oxide
(b) over catalysts with different Na-impregnation treatments. Reaction
parameters: 1% propene, 15% N2O, 84% He; GHSV = 4 L h 1 g 1 (STP);
TOS = 77 min.
concentration of 0.1 mol/L aqueous solution of sodium ac-
etate, and decreased with further increase of the Na loading
(Fig. 9b).
In Fig. 10 a comparison is made between three catalysts
with the same iron content and which have undergone the
same neutralization treatment but with different supports.
The catalyst with slim mesopores (MrFe300N2) shows a
higher reaction rate and selectivity than the other two cata-
lysts (Fig. 10a). Consequently, the yield to propene oxide
over MrFe300N2 was considerably higher than over the
other two catalysts and reached values of ca. 5% (Fig. 10b).
In addition, data of conversions and selectivities to several
identified products at different temperatures over the cata-
lyst MrFe300N2 are given in Table 2.
The variation of the reaction rate and selectivity to
propene oxide with time on stream at different gas hourly
space velocities (GHSV) is represented in Fig. 11. The in-
crease of the GHSV from 2 to 4 L h 1 g 1 brought a signifi-
cant increase in the reaction rate and selectivity to propene
oxide. A further increase of GHSV to 6 L h 1 g 1 brought
no alteration of the reaction rate; however, the selectivity
was somewhat higher at the beginning of the experiment
but decreased faster with increasing TOS.
FIG. 10. Reaction rate of propene and selectivity to propene oxide
(a) and formation rate of and yield to propene oxide (b) over catalysts
with different supports. Reaction parameters: 1% propene, 15% N2O,
84% He; GHSV = 4 L h 1 g 1 (STP); TOS = 128 min.

¨
DUMA AND HONICKE
102
FIG. 11. Reaction rate of propene and selectivity to propene oxide
FIG. 13. Conversion of propene and yields to products vs TOS. Re-
action parameters: 1% propene, 15% N2O, 84% He; catalyst MrFe300N2;
reaction temperature 648 K; GHSV = 4 L h 1 g 1 (STP).
as a function of TOS at different gas hourly space velocities. Reaction pa-
rameters: 1% propene, 15% N2O, 84% He; catalyst MrFe300N2; reaction
temperature 648 K.
nitrous oxide over iron-oxide-containing catalysts were suc-
cessful (16) and have led to the current in-depth investiga-
tion ofthe factorswhich influence the catalytic performance
in this reaction (17).
at the conditions of the first run. The results are depicted in
Fig. 12 and show that the reaction rate and the selectivity
to propene oxide had similar values in both runs.
The yields and selectivities to several products vs TOS
are given in Fig. 13. The yields to propene oxide, to its iso-
mers propanal and acetone, and to the carbon oxides have
a similar descendent course. On the other hand, the yield to
allyl oxidation products like allyl alcohol and acrolein has a
different course and remains approximately constant. From
Fig. 13 the selectivity-conversion behavior was derived and
depicted in Fig. 14. In spite of the deactivation during the
oxidation experiment indicated by the conversion decrease,
the selectivity to propene oxide of ca. 48% was nearly con-
stant in the range of 5–10% conversion degree.
The preparation of the catalysts was focused on obtaining
catalysts with different iron loadings, neutralization proce-
dures, and porosities. The reaction parameters were var-
ied to study the effect of reactant composition, reaction
temperature, gas hourly space velocity, and time on stream
on activity and propene oxide selectivity. Furthermore, the
possibility of regenerating the catalysts after a reaction ex-
periment and the products distribution was also studied.
The experiments with catalysts having different iron
loadings illustrated that the catalytic results depend
strongly on the iron loading of the catalysts (Fig. 8). The
increase of the reaction rate is not directly proportional to
the iron loading. This means that it is not the iron oxide
amount that directly determines the activity of the cata-
lysts. The evolution of the reaction rate and selectivity to
DISCUSSION
First investigations into the possibility of producing pro-
pene oxide by the gas phase reaction between propene and
FIG. 12. Reaction rate of propene and conversion and selectivity to
propene oxide for two oxidation runs at identical conditions. Reaction pa-
rameters: 1% propene, 15% N2O, 84% He; catalyst MrFe300N2; reaction
temperature 648 K; GHSV = 4 L h 1 g 1 (STP).
FIG. 14. Selectivities to products vs conversion. Reaction parameters:
1% propene, 15% N2O, 84% He; catalyst MrFe300N2; reaction tempera-
ture 648 K; GHSV = 4 L h 1 g 1 (STP).

GAS PHASE EPOXIDATION OF PROPENE BY NITROUS OXIDE
103
propene oxide with the iron loading suggests that the iron
oxide particle dimension plays an important role in the cata-
lytic activity. In other words, very small iron oxide particles
in the catalysts result in a low activity and selectivity toward
propene oxide, somewhat larger iron oxide particles (cor-
responding to an iron loading of ca. 300 ppm) have a much
higher activity and selectivity, and with a further increase of
particle dimension the activity continues to increase slowly,
but the selectivity to propene oxide goes down. Accord-
ing to the XRD investigations, the iron oxide particles in
all used catalysts have diameters less than 2 nm (13b). At-
tempts to determine more exactly the particle dimension
and the dispersity of the iron oxide by adsorption mea-
surements failed because of the experimental difficulties
bounded with the very small amounts of iron oxide that are
present in the catalysts.
The Na impregnation of the catalysts generally had a pos-
itive impact on activity. Less severe or no Na impregna-
tion led to a fast deactivation of the catalysts, so the activ-
ity was considerably lower after the same time on stream
(Fig. 9a). A more severe Na impregnation (as for the cata-
lyst AFe300N3) led to a lower activity at high temperatures.
This means that an excess of Na can also lead to a faster de-
activation of the catalysts. The evolution of the selectivity
with the Na-impregnation treatment showed that there is
an optimal impregnation treatment leading to a maximum
in selectivity.
The three silica gel supports used in the present investi-
gation had distinct textural properties which make them
suitable for studying the effect of support porosity on
the morphologic and catalytic properties of this kind of
catalysts (Table 1). The support D had an important pro-
portion of narrow pores. This was considerably diminished
after the catalyst preparation, where the first step was the
impregnation of the support with the iron precursor. That
leads to the assumption for the group a catalysts that an
important part of the iron oxide particles is located inside
the narrow pores. As a consequence, the localization of the
active centers in narrow pores can improve the adsorption
capacity of the catalyst but can also lead to the appearance
of diffusion problems. The catalysts from group b have large
pores, which permit a sufficient diffusion of the reactants
and products; however, this kind of pore cannot contribute
a great deal to the adsorption process. The catalysts from
group c combine the presumptive advantages of the cata-
lysts from groups a and b, creating a better contribution
of their slim mesopores to the adsorption process than the
larger pores of the catalysts of group b and less severe dif-
fusion problems than the catalysts of group a. A schematic
representation of the catalyst pore structures is shown in
Fig. 15. The described properties can be discussed also with
respect to the porosity of the catalysts which strongly influ-
enced the performance of the catalysts, as seen in Fig. 10.
The catalyst with a considerable proportion of narrow pores
(DFe300N2) showed the lowest activity and selectivity. This
FIG. 15. Schematic representation of the catalyst pore structure with
iron oxide particles smaller than 2 nm.
means that the narrow pores have a negative influence on
the oxidation of propene to propene oxide, probably due to
the appearance of desorption and/or diffusion difficulties.
The catalyst with large pores (AFe300N2) showed better
results than DFe300N2, but the catalyst with almost exclu-
sively slim mesopores (MrFe300N2) had the best perfor-
mance. Such pores can contribute to the adsorption process
of the reactants without considerably hindering the desorp-
tion and diffusion of products.
For a set propene concentration, the reaction rate of
propene increased with the concentration of the oxidant
(N2O). The selectivity to propene oxide initially increased
with the N2O concentration but beyond a certain N2O con-
centration, the selectivity to propene oxide decreased be-
cause of a stronger oxidation of propene to deeper oxy-
genated products.
Generally, the reaction rate increased and the selectiv-
ity to propene oxide decreased with increasing tempera-
ture (Figs. 8–10). The maximal yield to propene oxide was
reached at approx. 650–700 K (Fig. 10b).
The experiments carried out at different space velocities
1
led us to the conclusion that at values lower than 4 L h
g
1 there are some external diffusion hindrances that lim-
ited the reaction rate of propene and selectivity to propene
oxide (Fig. 11).
The results from the experiments depicted in Fig. 12
showed the good regeneration capability of the catalysts.
After a few reaction–regeneration cycles, the activity be-
comes somewhat lower but the selectivity becomes a little
higher, so the yield to propene oxide remainsapproximately
constant for more than 50 reaction–regeneration cycles.
The similar course for the yields (i.e., the rates of forma-
tion) to vinyl oxidation products (propene oxide, propanal,
acetone) and carbon oxides (Fig. 13) means that these prod-
ucts are formed over the same active centers and accord-
ing to similar formation rate equations. The decrease of the
rates of formation for these products can be associated with
a deactivation of the active centers, which are responsible
for the adsorption of N2O. In contrast, the allyl oxidation
products allyl alcohol and acrolein which are most prob-
ably produced by a Mars–van Krevelen mechanism over

¨
DUMA AND HONICKE
104
the redox centers of the catalyst, are not affected by this
deactivation.
SUMMARY
The present study describes a novel method of forming
propene oxide by the heterogeneously catalyzed gas phase
oxidation of propene with nitrous oxide. The catalysts con-
sisted ofsilica-supported, sodium-promoted iron oxide. The
optimal iron loading of the catalysts was in the range of 100–
1000 ppm and the iron oxide particle dimension was smaller
than 2 nm. The reaction occurred by the nondissociative
adsorption of N2O at the active centers of the catalysts
followed by the reaction between N2O and propene. The
impregnation of the catalysts with alkali (sodium) was of
great importance in order to minimize the side reactionsand
thereby to improve the selectivity toward propene oxide.
The reaction rate of propene always decreased slightly
with the TOS but the selectivity to propene oxide was ap-
proximately constant. After the regeneration of the catalyst
by oxidation in air, the catalyst regained its initial activity.
In the course of the air regeneration of the catalyst, car-
bon oxides were detected in the exhaust from the reactor.
The carbon oxides can originate from carbonaceous de-
posits and/or strongly adsorbed hydrocarbon species. Two
factors could contribute to the reversible decrease of the
reaction rate with TOS. On one hand, the physical blocking
of some adsorption centers by these deposits and strong
adsorbed species can gradually lead to a decrease of the
reaction rate with time. On the other hand, the chemical
reduction of the active centers under the reaction condi-
tions could also contribute to a decreasing number of active
centers with time. The impregnation with sodium reduced
the acidity and increased the reduction temperature of the
catalysts (Fig. 4). As a result, the reaction rate was higher
(Fig. 9a). An excess of basicity can have a reverse effect, as
evidenced bythe catalyst AFe300N2at higher temperatures
(Fig. 9a).
The selectivity to propene oxide increased with TOS at
the beginning of an experiment. In this period of time (ap-
prox. 100–200 min) the most active centers, which lead to
a deeper oxidation, are probably deactivated and this has
positive effects on the selectivity (Figs. 7, 11, and 12). Af-
ter this period, the selectivity to propene oxide went slowly
down. The decrease of the propene oxide selectivity is more
pronounced at higher temperatures, N2O concentrations,
and space velocities. The decrease of the propene oxide se-
lectivity can be explained by a further deactivation of the
adsorption centers, in contrast to the redox centers which
produce allyl oxidation products and which are less affected
by a deactivation process if at all (Fig. 13).
ACKNOWLEDGMENT
The authors express their gratitude to the Fonds der Chemischen
Industrie for financial support of this work.
REFERENCES
1. “Ullmann’s Encyclopedia of Industrial Chemistry,” 5th ed. VCH,
Weinheim. (a) Kahlich, D., Wiechern, U., and Lindner, J., Vol. A22,
p. 239 (1993); (b) Rebsdat, S., and Mayer, D., Vol. A10, p. 117 (1987);
(c) Thiemann, M., Scheibler, E., and Wiegand, K. W., Vol. A17, p. 332
(1991).
2. Trent, D. L., in “Kirk–Othmer Encyclopedia of Chemical Technology,”
4th ed., Vol. 20, p. 271. Wiley, New York, 1996.
3. (a) Clark, H. J., Maj, J. J., Bowman, R. G., Bare, S. R., and Hartwell,
G. E., WO Patent 00415, 1998; (b) Bowman, R. G., Maj, J. J., Clark,
H. W., Hartwell, G. E., Womack, J. L., and Bare, S. R., WO Patent
00414, 1998; (c) Bowman, R. G., Womack, J. L., Maj, J. J., Clark,
H. W., and Hartwell, G. E., WO Patent 00413, 1998.
4. Gaffney, A., Kahn, A., and Pitchai, R., WO Patent 30552, 1998.
5. Coxon, J. M., Maclagan, R. G. A. R., Rauk, A., Thorpe, A. J., and
Whalen, D., J. Am. Chem. Soc. 119, 4712 (1997).
6. Ramis, G., Busca, G., and Bregani, F., Gazz. Chim. Ital. 122, 79
(1992).
7. Ohtani, B., Takamiya, S., Hirai, Y., Sudoh, M., Nishimoto, S., and
Kagiya, T., J. Chem. Soc. Perkin Trans. II 2, 175 (1992).
8. Iwamoto, M., Hirata, J., Matzukami, K., and Kagawa, S., J. Phys. Chem.
87, 903 (1983).
9. Suzuki, E., Nakashiro, K., and Ono, Y., Chem. Lett. 6, 953 (1988).
10. Panov, G. I., Uriarte, A. K., Rodkin, M. A., and Sobolev, V. I., Catal.
Today 41, 365 (1998).
11. Burch, R., and Howitt, C., Appl. Catal. A 106, 167 (1993).
12. Ha¨fele, M., Reitzmann, A., Roppelt, D., and Emig, G., Appl. Catal. A
150, 153 (1997).
TPD experiments for propene and nitrous oxide were
carried out in order to investigate the nature ofthe adsorbed
species and thereby to contribute to the elucidation of the
reaction mechanism. The silica gel supports showed only a
desorption of the physisorbed N2O form, while the iron-
oxide-containing catalysts showed a desorption of both
N2O forms, physisorbed and chemisorbed. The catalysts
with an iron content up to 1000 ppm (0,1% ) showed lit-
tle or no O2 desorption. The catalysts with 10000 ppm (1% ) 13. “Handbook of Heterogeneous Catalysis” (G. Ertl, H. Kno¨zinger, and
J. Weitkamp, Eds.). VCH, Weinheim. (a) Cimino, A., and Stone, F. S.,
Vol. 2, p. 845 (1997); (b) Bergeret, G., and Gallezot, P., Vol. 2, p. 439
(1997).
Fe showed a relatively strong O2 desorption signal. This
indicates that the iron-free silica gels have no chemisorp-
tion capacity for N2O. Furthermore, the presence of small
amounts of iron oxide on the surface of the catalysts lead to
14. Yamashita, T., and Vannice, A., J. Catal. 161, 254 (1996).
15. Drago, R. S., Jurczik, K., and Kob, N., Appl. Catal. B 13, 69 (1997).
the nondissociative adsorption of N2O, and the presence of 16. Ho¨nicke, D., Duma, V., and Krysmann, W., DE Patent Appl. 198 54
615.7-43, 11/26/1998.
17. Duma, V., Ph.D. thesis, in preparation, Technical University Chemnitz,
larger amounts of iron oxide lead, in complement with the
latter, to a dissociative adsorption of N2O. In correlation
with the catalytic experimental results, one can say that the
Germany.
18. Gontier, S., and Tuel, A., in “Studies in Surface Science and Cata-
nondissociative adsorption of N2O is a necessary condition
for the epoxidation of propene in this system.
lysis” (L. Bonneviot and S. Kaliaguine, Eds.), Vol. 97, p. 157. Elsevier,
Amsterdam, 1995.