Chemical looping epoxidation

Article abstract of DOI:10.1016/j.jcat.2017.12.030

Chemical looping epoxidation of ethylene was demonstrated, whereby the sole oxidant was a solid oxygen carrier, 15 wt% Ag supported on SrFeO₃. Ethylene reacted with a bed of carrier particles, without any O_2(g) in the feed, to produc

Full text of DOI:10.1016/j.jcat.2017.12.030

Journal of Catalysis 359 (2018) 1–7
Contents lists available at ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Priority Communication
Chemical looping epoxidation
Martin S.C. Chan a, , Ewa Marek b, Stuart A. Scott b, J.S. Dennis a
⇑
a Department of Chemical Engineering and Biotechnology, University of Cambridge, Philippa Fawcett Drive, CB3 0AS, United Kingdom
b Department of Engineering, University of Cambridge, Trumpington Street, CB2 1PZ, United Kingdom
a r t i c l e i n f o
a b s t r a c t
Article history:
Chemical looping epoxidation of ethylene was demonstrated, whereby the sole oxidant was a solid oxy-
gen carrier, 15 wt% Ag supported on SrFeO3. Ethylene reacted with a bed of carrier particles, without any
O2(g) in the feed, to produce ethylene oxide (EO) and CO2. Following the reduction by the C2H4 of the
SrFeO3, it was regenerated by passing air through the bed. The rate of reoxidation was slow, with full
regeneration being achieved only by prolonged oxidation at elevated temperatures. A striking synergy
between Ag and SrFeO3 was observed solely when they were in intimate contact, suggesting a basis
for a proposed reaction mechanism.
Received 10 October 2017
Revised 8 December 2017
Accepted 28 December 2017
Keywords:
Epoxidation
Chemical looping
Silver
Ó 2018 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license (http://
creativecommons.org/licenses/by/4.0/).
Perovskite
Redox
1. Introduction
temperatures. Lastly, the solid oxidant needs to be regenerable
without significant loss of oxygen-storage capacity and kinetic
Ethylene oxide (EO) is an industrially-significant chemical
intermediate, with a global production of 26 million tonnes in
2013 [1]. It is used in the synthesis of polymers, antifreeze, surfac-
tants, solvents, etc. The predominant technology for its production
is the epoxidation of ethylene with air or purified oxygen. Silver is
the only catalyst used because no other material matches its per-
formance [2].
A key intermediate species in the reaction mechanism is the
oxygen adatom (Oa), which is the selective oxygen species respon-
sible for the epoxidation and, depending on its electrophilicity,
combustion [3]. In the conventional process, where ethylene and
gaseous oxygen are co-fed to the reactor, Oa is generated from
the dissociative adsorption of O2(g). To date, there have been no
reports of Oa being directly supplied from a solid oxidant, probably
because of the challenging requirements imposed on such a mate-
rial. Working catalysts typically operate at partial pressures of O2
of up to 2 bar and temperatures of 200–300 °C [2], where evidence
suggest the existence of surface silver oxide phases [4–6]. Herein, it
is therefore posited that the solid oxidant needs to have a high
activity.
A recent, large-scale in silico screening of materials has identi-
fied SrFeO3 as a potential oxygen carrier for chemical looping com-
bustion [7]. It was found that SrFeO3 can release O2(g) reversibly
over many cycles of thermal decomposition followed by reoxida-
tion at low temperatures (ꢀ673 K, pO2 = 0.15 bar), demonstrating
its favourable lO2 and kinetics. Crucially, SrFeO3 does not react
with CO2 significantly [7] and thereby deactivate with the forma-
tion of carbonate, a problem afflicting other possible materials
with a similar lO2 such as SrO2/SrO and BaO2/BaO [8]. Resistance
against carbonation is necessary because significant levels of CO2
are generated from side reactions during epoxidation.
These favourable properties suggested the possibility of inter-
facing SrFeO3 with Ag; SrFeO3 could then supply oxygen to Ag as
bulk or subsurface oxygen (Oss), where the oxygen could then dif-
fuse to the silver surface to be presented as Oa to effect the epox-
idation in the absence of O2(g). The concept is summarised in
Fig. 1. The catalyst is then regenerated in air in a separate step to
complete the chemical looping cycle.
chemical potential of oxygen,
or at least to ensure a sufficiently high surface coverage of Oa. Such
a high O2 at these temperatures is rare among solid oxides. Suffi-
ciently fast ionic conduction is also required to replenish Oa during
the reaction, but oxide ion conduction is typically slow at these
l
O2, to stabilise these active phases,
It is advantageous to perform selective oxidations using a solid
oxidant, in the absence of O2(g), because this avoids flammable
mixtures and renders a safer process. The conversion of ethylene
is no longer limited by the supply of O2(g), decreasing the need to
recycle unreacted ethylene – industrial processes tend to operate
with limited amounts of O2 with ethylene in excess [2]. Further-
more, separation costs can be reduced by avoiding the need to sep-
arate N2 (from purifying either O2 or the products). Minimising
l
⇑
Corresponding author.
E-mail address: mscc3@cam.ac.uk (M.S.C. Chan).
https://doi.org/10.1016/j.jcat.2017.12.030
0021-9517/Ó 2018 The Authors. Published by Elsevier Inc.
This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

2
M.S.C. Chan et al. / Journal of Catalysis 359 (2018) 1–7
The bed was packed above the sintered disc, with a bottom layer of
-Al2O3 (Boud Minerals, 425–710 mm) to position the active bed in
the isothermal region of the reactor, then a middle layer with the
catalyst, then finally another layer of 3 g of -Al2O3 to distribute
and preheat the feed. Both ends of the quartz tube were connected
to Swagelok Ultra-TorrÒ vacuum fittings and sealed with fluorocar-
bon FKM O-rings. A 7 mm filter (Swagelok, SS-2TF-7) was fitted at
the outlet of the reactor. Ex situ regeneration of the active material
was performed by dismounting the reactor tube, containing the
packed bed, from its fittings and placing it in a box furnace to be
calcined at 400 °C for 2 h in static air.
C2H4
EO
a
Oa
a
Oss
Ag
O2-
SrFeO3-δ
Fig. 1. Chemical looping epoxidation using a silver-modified oxygen carrier.
The feed gases to the reactor were supplied from cylinders
(BOC), and consisted of 5.16 vol% C2H4 in balance N2 (certified to
5% uncertainty, BOC), N2 (purity > 99.998%), and bottled purified
air. Gas flows were manipulated by calibrated rotameters and
checked using a bubble film flowmeter at the start of each experi-
ment. Gases were switched using digitally-controlled solenoid
valves. Continuous online analysis of the gaseous products was
achieved by a Fourier transform infrared (FTIR) analyser (MKS
Instruments, Multigas 2030) equipped with a liquid N2-cooled
mercury-cadmium-telluride detector. Measurements were col-
lected 2.5 h after filling the liquid N2 dewar, once the background
has settled. The 5.11 m gas cell was heated to 150 °C. Each mea-
surement consists of 8 scans of the band 800 – 4600 cmꢂ1, lasting
1.87 s, at a resolution of 0.5 cmꢂ1. The collected spectra were anal-
ysed for C2H4, CO2, ethylene oxide, CO and H2O, using software
(MKS, MG2000). The analysis regions for each quantified species
were adjusted to exclude interfering peaks from other species to
eliminate any cross-sensitivities. Negligible amounts of CO were
detected, <40 ppm. H2O was not quantified; its quantity was
inferred from the stoichiometry of combustion, C2H4 + 3/2O2 ?
CO2 + H2O, i.e. 1 mol of CO2 generated implied that 1 mol of H2O
was also generated, and that 3 mol of O was consumed. No
acetaldehyde was detected.
recycling, and combining unit operations of reaction and separa-
tion, are both examples of process intensification [9–11]. Lastly,
improved selectivities may be achieved by the removal of O2(g)
[12].
In this work, we show that the epoxidation may be performed
in the absence of O2(g) by using a solid oxygen carrier. The aim is
not to show competitive selectivities, so common promoters were
not used (e.g. Cs and Cl), but rather to demonstrate the concept. We
use the term ‘catalyst’ as an abbreviation for the ‘catalyst-oxygen
carrier’ materials prepared in this study, where Ag acts as the cat-
alyst and SrFeO3 acts as the oxygen carrier.
2. Experimental methods
SrFeO3 was prepared by a solid-state synthesis method. Stoi-
chiometric amounts of SrCO3 (0.72 mol) and Fe2O3 (0.36 mol) were
mixed in a ball mill for 3 h at 25 Hz. Ethanol (50 mL, 99.8%, Fisher
Scientific) was added as a binder to improve mixing. The mixture
was dried for 24 h at 50 °C, and sieved to 180–355 mm. The parti-
cles were then calcined in four stages, with each stage consisting
of calcination at 1000 °C for 3 h, followed by cooling to room
temperature.
In the chemical looping experiments, the reactor was packed
with 1.0 g
a-Al2O3 for the bottom layer, with 2.000 g catalyst in
Incipient wetness impregnation was used to dose Ag onto sup-
the middle layer. A constant flow rate of 200 mL/min (measured
at 293 K, 1 atm), was maintained at all times. The feed gases were
switched automatically, with the base case cycling times of (i)
reduction with 5.16 vol% C2H4 in balance N2 for tred = 1.5 min, (ii)
purge with N2 for 2 min, (iii) oxidation with compressed air for
tox = 15 min, and (iv) purge with N2 for 2 min.
The overall carbon balance in a particular reduction stage was
calculated according to (assuming dilute gases and small conver-
sions so that the total molar flow rate in and out of the reactor is
constant)
ports, consisting of either SrFeO3 or
a-Al2O3. Prior to impregnation,
SrFeO3 was sieved to a size range of 212 – 300 mm, and
a-Al2O3
(Alfa Aesar, product code 43862) was crushed and sieved to a size
range of 212 – 300 mm. The volume of solution added was equal to
the pore volume of the support, which was determined empirically
beforehand by adding deionised water dropwise with stirring to a
mass of support. The endpoint was taken to be when the granular
matter just started to cohere, i.e. when the internal pores have
been filled and the mixture is at the onset of the pendular regime,
when the volume of water added was taken to be equal to the pore
volume. The pore volume of SrFeO3 was measured to be 0.24 mL/g,
Total carbon detected
Carbon balance ¼
Total carbon fed
ꢀ
ꢁ
and the pore volume of a-Al2O3 was measured to be 0.26 mL/g. In a
R
tend
tstart
1
yC þ yEO
þ
12 yCO þ 2 yCO dt
typical preparation of 15 wt% Ag/SrFeO3, 1.3895 g of AgNO3
(ꢁ99.0%, Sigma-Aldrich) was dissolved in 1.2 mL water, which
was then added dropwise to a batch of 5.0000 g SrFeO3 with agita-
tion by a spatula. Correspondingly, in a typical preparation of 15 wt
% Ag/Al2O3, 1.3895 g of AgNO3 (ꢁ99.0%, Sigma-Aldrich) was dis-
solved in 1.3 mL water, which was then added dropwise to a batch
of 5.0000 g Al2O3 with agitation by a spatula. The impregnated
solids were then dried at 120 °C for 12 h in static air, before calci-
nation at 500 °C for 5 h with a ramp rate of 5 °C/min in static air.
A packed bed reactor was used to perform the epoxidation
experiments, and was operated in either chemical looping mode
or co-feeding mode. The reactor consisted of an 8 mm i.d., 200
mm long quartz tube mounted vertically with a sintered disc fixed
75 mm from the bottom. The tube was wrapped with a high tem-
perature dual-element heating tape (Omega, DHT052020LD). The
output of the heating tape was controlled by a type K thermocou-
ple with a mineral-insulated Inconel sheath inserted into the bed.
2H4
2
¼
ð1Þ
yC
ꢃ tred
2H4;feed
where yi is the mole fraction of species i, t is time, and yC
is the
2H4;feed
mole fraction of C2H4 in the feed. The start and end time for integra-
tion was the first and final points in time where EO was detected
(the range of this time was not necessarily equal to tred because of
the response of the analyser). The carbon balance was always
within 95 5%, and usually within 98 5%, which verifies that the
rate of accumulation of coke in the reactor is small and may be
neglected in subsequent analyses.
The cumulative amount of oxygen released by time t, in a par-
ticular cycle starting at time tstart, was calculated by integrating
all the oxygenated gaseous products according to
Z
t
ꢀ
ꢁ
Mol O releasedðtÞ ¼ F
yEO þ 3yCO þ yCO dt
ð2Þ
2
tstart

M.S.C. Chan et al. / Journal of Catalysis 359 (2018) 1–7
3
d23;2
where F is the total molar flow rate, and the term 3 yCO accounts for
oxygen released from both CO2 and H2O by assuming these two
products are generated only from the combustion of C2H4.
qs
ꢄ
p
p
number of surface Ag atoms
total number of Ag atoms
6qs 1
2
D ¼
¼
¼
qb d3;2
d33;2
qb
ꢄ
6
6 ꢄ 1:14 ꢄ 1019
1
1:17
d3;2½mꢅ d3;2½nmꢅ
In the co-feeding experiments, 2.0 g
a-Al2O3 was used for the
¼
¼
ð6Þ
5:87 ꢄ 1028
bottom layer with 1.000 g catalyst. The flows of C2H4 and air were
mixed to a composition of 4 vol% C2H4, 4 vol% O2 in balance N2, at a
total flow rate of 200 mL/min. The instantaneous carbon balance
was calculated according to
where qs ¼ 1:14 ꢄ 1019 mꢂ2 is the surface atom density of silver
averaged over the (1 0 0), (1 1 0) and (1 1 1) planes, and
qb ¼ 5:87 ꢄ 1028 mꢂ3is the bulk atom density of silver (fcc lattice
parameter is 0.40853 nm). The moles of surface Ag atoms per unit
mass of catalyst, ns, can then be calculated from the loading of silver
using ns ¼ Dw=mAg where w ¼ 15 wt % is the loading of Ag on the
Instantaneous carbon balance; co ꢂ feedingðtÞ
ꢂ
ꢂ
ꢂ
ꢂ
ꢂ
1
1
y
þ yEO þ 2
y
þ 2 yCO
C2H4
CO2
¼
ð3Þ
y
C2H4;feed
t
catalyst and mAg ¼ 108 g molꢂ1 is its molecular weight.
The carbon balance during co-feeding was always within 97
3%. The definitions of instantaneous selectivity and conversion
were equivalent to those used for the chemical looping case (see
Supporting Information).
The turnover frequency (TOF) for the rate of production of EO
and conversion of C2H4 was calculated according to
Power X-ray diffraction was performed on the supported silver
samples using Cu-Ka radiation with a voltage of 40 kV and current
40 mA (Empyrean, PANalytical). The diffractogram was collected in
the range of 2h from 10° to 150° over 67 min. The mean particle
size of the silver crystallites was measured by Rietveld refinement
of the diffraction patterns. The refined parameters were particle
size, strain, unit cell and atomic displacement. The surface area-
weighted mean diameter was calculated using the method
reported by Popa and Balzar (2002) [15]. The lack of measurable
peak-broadening in the Al2O3-supported sample suggested that
the particles were at least 300 nm. Nevertheless, measurements
for both Al2O3- and SrFeO3-supported samples were consistent
with SEM imaging. The refined XRD pattern for Ag/SrFeO3 can be
found in the Supporting Information.
rate of production of EO
number of surface Ag atoms nsm
FyEO
TOF EO ¼
¼
ð4Þ
ꢃ
ꢄ
F yEO
þ
12 yCO2
nsm
rate of conversion of C2H4
number of surface Agatoms
TOF C2H4 ¼
¼
ð5Þ
where ns is the moles of surface Ag atoms per unit mass of catalyst
and m is the mass of catalyst. The approximation that the rate of
reaction = Fyi was used because the change in the total molar flow
of gas can be neglected for dilute feeds with small conversion.
Brunauer, Emmett and Teller (BET) theory [13] was used to
model nitrogen physisorption isotherms measured at 77 K (Tristar
3000, Micromeritics). 1–2 g of the supported silver samples were
used in each analysis.
Thermogravimetric analysis (TGA) was conducted using a TGA/
DSC1 (Mettler Toledo). Approximately 20 mg of 15 wt% Ag/SrFeO3
was loaded into a 70 mL alumina crucible. The sample was heated
to 270 °C in air. 5 min after reaching the setpoint temperature,
the gas was switched to N2 for 2 min to purge. Ethylene was then
fed to the sample for 30 min, followed by a 2 min purge with N2,
then air was fed for 60 min. Throughout the experiments, the
TGA chamber was purged with protective and purge gas flows of
Ar (50 mL/min each, measured at 293 K and 1 atm). The reactive
gas was supplied through a capillary tube located above the cru-
cible at 50 mL/min (measured at 293 K and 1 atm). The reactive
gases fed from cylinders were air, N2, 5.16% C2H4 in balance N2
(all BOC, >99.99%). The actual concentrations of the specific gas
components listed above were approximately three times lower
when in contact with the solid sample due to the dilution with
the protective and purge gas.
3. Results
The catalyst, composed of 15 wt% Ag supported on SrFeO3, was
evaluated in a packed bed reactor (n.b here, the nonstoichiometric
perovskite SrFeO3ꢂd is abbreviated as SrFeO3 for brevity, its charac-
terisation was reported by Lau et al., (2017) [7]). A typical experi-
ment is shown in Fig. 2, which shows how the mole fractions
varied over 10 cycles. Within each cycle, air is first fed to the reac-
tor (indicated by the background level of ꢀ440 ppm CO2 in the
cylinder of compressed air) to charge the catalyst with oxygen, fol-
lowed by a purge with N2, then a switch to a feed of ethylene to
produce EO and CO2, followed by another purge. The selectivity
and conversion achieved by the catalyst decayed gradually with
cycling, before approaching an asymptote. This decay is probably
caused by the slow rate of reoxidation of the perovskite. Thermo-
gravimetric analysis, presented in Fig. 3, shows that the rate of oxi-
dation in air was indeed much slower than the rate of reduction in
ethylene. This is supported by the results in Fig. 4, which show the
effect of varying the duration of reduction and oxidation. When the
duration of oxidation was increased from tox = 7.5 min to 30 min,
both the selectivity and conversion were maintained at higher val-
ues. A similar improvement was observed when the duration of
reduction was decreased from tred = 3 min to 1 min. The decrease
in conversion is probably caused by the depletion of oxygen spe-
cies, owing to the catalyst gradually becoming more reduced with
cycling. Cycling with 5 min reduction and 10 min oxidation (data
not shown) led to the mole fraction of EO decaying to zero within
four cycles. Accumulation of coke, indicated by slight elevations in
the mole fraction of CO2 (of the order of 10 ppm) during oxidation
with air, may also have caused the decaying activity. The decrease
in selectivity might be explained by the depletion of subsurface
oxygen species, Oss, when the catalyst was reduced. The decrease
in abundance of Oss would cause the degree of electrophilicity of
Oa to decrease [16,17], resulting in the observed decrease in
selectivity.
Scanning electron microscopy (SEM) images were taken using a
Leo GEMINI 1530 VP. The particulate samples of 15 wt% Ag/SrFeO3
were mounted on copper tape. Images were taken at an accelerat-
ing voltage of 2.4 kV and a working distance of 5.4 – 5.6 mm.
Distributions of particle size were measured from the SEM images
with ImageJ [14]. The analysis was carried out semi-automatically,
using the ‘Analyze Particles’ function in ImageJ. The results were
then inspected manually to correct for any false measurements,
e.g. aggregated particles. Each analysed particle i was described
by its projected area Ai, from which an equivalent diameter may
pffiffiffiffiffiffiffiffiffiffi
be calculated di ¼ 2 Ai=
eter was then calculated using d3;2
p
. The surface area-weighted mean diam-
P
d3
i , which was taken to be
P
¼
di2
the mean crystallite size. 458 particles were measured for Ag/
SrFeO3, and 83 particles for Ag/Al2O3. The dispersion of silver, D,
was calculated by assuming spherical particles, so
The regenerability of the 15 wt% Ag/SrFeO3 catalyst was exam-
ined further by calcining the cycled sample ex situ in static air at

4
M.S.C. Chan et al. / Journal of Catalysis 359 (2018) 1–7
Fig. 2. Mole fraction, selectivity and conversion profiles from a packed bed reactor operated in chemical looping mode, using 15 wt% Ag/SrFeO3 as the catalyst. The
background of CO2 during oxidation steps was from compressed atmospheric air. Conditions: 270 °C, 1 atm, 2.000 g catalyst, gas feed of 200 mL/min (as measured at 293 K, 1
atm). Cycling times: (i) tred = 1.5 min reduction with 5.16 vol% C2H4 in balance N2, (ii) 2 min purge with N2, (iii) tox = 15 min oxidation with air, (iv) 2 min purge with N2.
Fig. 4. Effects of cycling times on the selectivity and conversion. Conditions: 270 °C,
1 atm, 2.000 g of 15 wt% Ag/SrFeO3, feed of 200 mL/min (as measured at 293 K, 1
Fig. 3. Variation of catalyst mass over a single cycle for a sample of 15 wt% Ag/
SrFeO3 measured by TGA. The temperature was fixed at 270 °C. The spikes in the
mass are artefacts caused by the switching of gases.
atm). Cycling times (unless stated otherwise): (i) tred = 1.5 min reduction with 5.16
vol% C2H4 in balance N2, (ii) 2 min purge with N2, (iii) tox = 15 min oxidation with
air, (iv) 2 min purge with N2.
400 °C for 2 h, before subjecting it to the same ten cycles. This was
repeated three times, shown in Fig. 5. It can be seen that this treat-
ment fully regenerated both the selectivity and activity of the cat-
alyst, confirming that the oxidation step is probably the rate-
limiting step in the cycle. Additionally, coke should have been
effectively removed by this period of oxidation. Fig. 5 also shows
that the selectivity gradually improved over time, which might
be related to the sintering of the Ag particles. The selectivity of
Ag particles, for epoxidation, increases with their size [18,19].
Scanning electron microscopy (SEM) images of the fresh and cycled
catalysts, presented in Fig. 6, show that Ag particles as small as 50
nm present on the fresh catalyst were no longer present on the
cycled catalyst.
Generally, over the course of one cycle, the instantaneous selec-
tivity of the carrier decreased as it became more reduced. This is
shown in Fig. 7 for the sample which has been regenerated thrice
at 400 °C for 2 h (i.e. cycles 31–40 in Fig. 5). The trend is consistent
with the mechanism proposed earlier, that reduction of the cata-
lyst led to a gradual depletion of Oa and Oss, resulting in the
decreasing selectivity. Interestingly, the first cycle after each calci-
nation step always exhibited a maximum in selectivity and conver-
sion (n.b. the early maxima for subsequent cycles are probably
artefacts of the response time of the gas analyser, or noise). This
gradual improvement in the selectivity during the first cycle in
Fig. 7 might be due to carbon deposition which selectively poi-
soned unselective sites [20].

M.S.C. Chan et al. / Journal of Catalysis 359 (2018) 1–7
5
Fig. 5. Effect of extended ex-situ reoxidation at 400 °C for 2 h, performed after every
10th cycle. Conditions: 270 °C, 1 atm, 2.000 g of 15 wt% Ag/SrFeO3, feed of 200 mL/
min (as measured at 293 K, 1 atm). Cycling times: (i) tred = 1.5 min reduction with
5.16 vol% C2H4 in balance N2, (ii) 2 min purge with N2, (iii) tox = 15 min oxidation
with air, (iv) 2 min purge with N2.
4. Discussion
The performance of Ag/SrFeO3 was in stark contrast with those
of (i) 15 wt% Ag/Al2O3 and (ii) a heterogeneous mixture of particles
of 15 wt% Ag/Al2O3 with particles of SrFeO3, shown in Fig. 8. The
15 wt% Ag/Al2O3 catalyst yielded only barely detectable levels of
EO (ꢀ10 ppm) with instantaneous selectivities of up to ꢀ30% in
the first cycle, and no detectable EO in subsequent cycles. The pro-
duction of CO2 also quickly fell to insignificant levels (<30 ppm).
This is because neither Ag nor Al2O3 can carry enough oxygen over
from the preceding oxidation step in the cycle, so the subsequent
reaction with ethylene became starved of oxygen. With the addi-
tion of separate particles of SrFeO3 to the bed, only CO2 was pro-
duced. This is because the perovskite could not generate Oa by
itself nor did it supply any oxygen to Ag/Al2O3, which prevented
epoxidation from occurring and so only combustion occurred. This
also shows that some CO2 evolved in the reaction with Ag/SrFeO3
must have originated from reaction with the part of the SrFeO3 sur-
face not covered with Ag. Therefore, the selectivity could be
improved simply by increasing the coverage of SrFeO3 with Ag.
The experiment with bare SrFeO3 also eliminates the possibility
of SrFeO3 supplying oxygen to Ag via gaseous transport at this tem-
perature. Hence, the performance of 15 wt% Ag/SrFeO3 must have
originated from the intimate physical contact between Ag and
SrFeO3, suggesting possible mechanisms: (i) solid state oxygen dif-
fused from SrFeO3 to Ag, as hypothesised, (ii) the oxygen storage
capacity of Ag became enhanced, or (iii) Ag facilitated adsorption
and diffusion of ethylene to the SrFeO3 boundary to react with lat-
tice oxygen from SrFeO3, i.e. triple phase boundaries. However,
because of the magnitude of the effect, the mechanism of (ii) seems
unlikely; if one assumes that all the oxygen in the products origi-
nated solely from Ag, then the bulk chemical state of the silver
Fig. 6. SEM images of 15 wt% Ag/SrFeO3 (a) initial state after first calcination at 500
°C for 5 h, (b) cycled and regenerated multiple times, and (c) bare SrFeO3. Ag
crystallites as small as 50 nm are present in (a), whereas only crystallites > 100 nm
appear in (b). Selected individual and clusters of crystallites have been circled in red
for clarity.
much activity, probably because oxygen desorbed from Ag during
the N2 purge, which would be significant at these temperatures
[23]. Triple phase boundaries producing CO2 through the mecha-
nism of (iii) cannot yet be excluded, but the formation of ethylene
oxide through this mechanism is unlikely because epoxidation pro-
ceeds from the formation of oxametallacycles [24]. In conclusion,
SrFeO3 most likely donated oxygen to Ag to be presented as ada-
toms on the Ag surface through the mechanism of (i). This might
well be exploitable in catalyst-oxygen carrier systems other than
Ag-SrFeO3 shown herein and applicable to selective oxidations
other than the epoxidation of ethylene.
The Ag/Al2O3 and Ag/SrFeO3 catalysts were also examined in a
co-feeding mode, using gas-phase oxygen, where the feed was a
premixed stream of O2 (4 vol%) and C2H4 (4 vol%), with 200 mL/
min total flow (as measured at 293 K and 1 atm) and 1.000 g cata-
lyst (half the catalyst loading compared to the chemical looping
experiments). After 5 h on stream, Ag/Al2O3 achieved a conversion
phase would be written as AgO0.08 (for the first cycle with tred
=
3 min), which is far higher than the limit of solubility (O:Ag molar
ratio ꢀ10ꢂ6) [21]. Moreover, bulk silver oxide is not thermodynam-
ically stable at these conditions [21]. Adsorbed oxygen also cannot
account for the capacity because the amount of oxygen released
exceeded the number of surface Ag sites. As shown in Table 1,
the molar ratio of O released to surface Ag = 7.7 exceeded an
adsorption stoichiometry of 1, which is generally assumed for
chemisorption analyses [22]. Furthermore, adsorbed oxygen
should also be accessible by Ag/Al2O3 and yet it did not exhibit

6
M.S.C. Chan et al. / Journal of Catalysis 359 (2018) 1–7
of 1.6% at a selectivity of 27%, whilst Ag/SrFeO3 achieved a conver-
sion of 6% at a selectivity of 5.7%. Turnover frequencies (TOF) from
these co-feeding experiments are shown in Table 1. TOFs were also
calculated for the Ag/SrFeO3 sample in chemical looping mode,
shown in Fig. 9. Whilst the TOF apparently increased up to
ꢀ0.03 wt% of oxygen released, this is probably an artefact caused
by the response time of the analyser. It is feasible for the TOF to
have linearly decayed right from the start, which is typical for cer-
tain types of gas-solid reactions, and that convolution with the
analyser masked this initial period. TOF then decayed from max-
ima of 1.1 ꢄ 10ꢂ2 C2H4 sꢂ1 and 0.24 ꢄ 10ꢂ2 EO sꢂ1, because the
amount of oxygen depleted. The observed TOFs in this study were
within the generally-observed range of 10ꢂ3–10ꢂ1 sꢂ1 for unpro-
moted supported polycrystalline Ag catalysts [22,25]. It is not sur-
prising that the chemical looping TOF was on the lower end of this
range, because the lattice oxygen from SrFeO3 has additional resis-
tances for transport to the Ag active sites compared to gaseous
oxygen.
Whilst the selectivities and TOFs shown in the chemical looping
experiments were worse than those of the conventional co-feeding
experiments, they were within an order of magnitude and we note
that there are many degrees of freedom available to optimise the
performance of both, or separately, the catalyst and the oxygen
carrier, in addition to the operating parameters. No such optimisa-
tion was undertaken in producing these results, but will be
explored in a future study. It is noteworthy that the instantaneous
Fig. 7. Variation of instantaneous selectivity with degree of reduction of catalyst,
expressed as the total amount of oxygen released. Each line corresponds to one
cycle. Conditions: 270 °C, 1 atm, 2.000 g of 15 wt% Ag/SrFeO3, feed of 200 mL/min
(as measured at 293 K, 1 atm). Cycling times: (i) tred = 1.5 min reduction with 5.16
vol% C2H4 in balance N2, (ii) 2 min purge with N2, (iii) tox = 15 min oxidation with
air, (iv) 2 min purge with N2.
Fig. 8. Effect of the support on the performance in chemical looping mode.
Conditions: 270 °C, 1 atm, 2.000 g of catalyst, feed of 200 mL/min (as measured at
293 K, 1 atm). Cycling times: (i) tred = 1.5 min reduction with 5.16 vol% C2H4 in
balance N2, (ii) 2 min purge with N2, (iii) tox = 15 min oxidation with air, (iv) 2 min
purge with N2.
Fig. 9. Turnover frequencies, based on the rates of C2H4 conversion and EO
production, varying with the degree of reduction during the first cycle. Conditions:
270 °C, 1 atm, 2.000 g of 15 wt% Ag/SrFeO3, feed of 200 mL/min (as measured at
293 K, 1 atm). Reduction in 5.16 vol% C2H4 over a period of 3 min.
Table 1
Characterisation and turnover frequencies (TOF), based on C2H4 conversion or EO production, of the prepared catalysts.
Chemical loopinga
Co-feeding
TOF C2H4
O
released
Sample
BET
Mean
Mean
particle
size
Silver
dispersion
SEM
Ag
Moles of
Oxygen
released
TOF C2H4
TOF EO
TOF EO
Surface Ag
surface particle
area
surface surface Ag
area
size
XRD
SEM
nm
m2 gꢂ1 nm
–
m2 gꢂ1 mol Ag gꢂ1 mol O gꢂ1
–
ꢄ10ꢂ2 sꢂ1 ꢄ10ꢂ2 sꢂ1 ꢄ10ꢂ2 sꢂ1 ꢄ10ꢂ2 sꢂ1
15 wt% Ag/SrFeO3 0.92
15 wt% Ag/Al2O3 0.22
110 40 107
>300
3
1.09 ꢄ 10ꢂ2 0.80
1.5 ꢄ 10ꢂ5
1.2 ꢄ 10ꢂ4 7.7
1.1
0.24
2.2
15
0.13
4.3
2740 80 4.27 ꢄ 10ꢂ4 0.031
5.9 ꢄ 10ꢂ7
b
b
b
b
a
First cycle with reduction time tred = 3 min. Maximum instantaneous values for TOF.
No detected activity.
b

M.S.C. Chan et al. / Journal of Catalysis 359 (2018) 1–7
7
1}, Chem. Phys. Lett. 367 (2003) 344–350, https://doi.org/10.1016/S0009-2614
(02)01699-8.
performance near the start of chemical looping operation briefly
surpassed the co-feeding operation, and that this could be reliably
reproduced with a long regenerating treatment at elevated tem-
peratures in air – we speculate that tuning the material properties
for faster reoxidation to sustain this early transient will be partic-
ularly important.
[5] S. Piccinin, S. Zafeiratos, C. Stampfl, T.W. Hansen, M. Hävecker, D. Teschner, V.I.
Bukhtiyarov, F. Girgsdies, A. Knop-Gericke, R. Schlögl, M. Scheffler, Alloy
catalyst in a reactive environment: the example of Ag-Cu particles for ethylene
epoxidation, Phys. Rev. Lett. 104 (2010) 1–4, https://doi.org/10.1103/
PhysRevLett. 104.035503.
[6] R. Reichelt, S. Günther, M. Rößler, J. Wintterlin, B. Kubias, B. Jakobi, R. Schlögl,
High-pressure STM of the interaction of oxygen with Ag(1 1 1), Phys. Chem.
Chem. Phys. 9 (2007) 3590–3599, https://doi.org/10.1039/B700432J.
[7] C.Y. Lau, M.T. Dunstan, W. Hu, C.P. Grey, S.A. Scott, Large scale in silico
screening of materials for carbon capture through chemical looping, Energy
Environ. Sci. 10 (2017) 818–831, https://doi.org/10.1039/C6EE02763F.
[8] W.B. Jensen, The origin of the brin process for the manufacture of oxygen, J.
Chem. Educ. 86 (2009) 1266, https://doi.org/10.1021/ed086p1266.
[9] J.-C. Charpentier, In the frame of globalization and sustainability, process
intensification, a path to the future of chemical and process engineering
(molecules into money), Chem. Eng. J. 134 (2007) 84–92, https://doi.org/
10.1016/j.cej.2007.03.084.
5. Conclusions
The epoxidation of ethylene in the absence of gaseous oxygen
depended crucially upon the intimate contact between the cata-
lyst, Ag, and the oxygen carrier, SrFeO3. This intimate contact sug-
gests that solid state transport of oxygen from SrFeO3 to Ag was
responsible for the observed performance, but further study is
needed to confirm this mechanism. Effects of the operation condi-
tions were studied, which showed that the oxidation step in the
cycle is probably the bottleneck, being slower than the reduction
step, and that further attention on this part of the process is crucial.
Comparable, although currently inferior, performance with con-
ventional co-feeding justifies further development of this
technology.
[10] T. Van Gerven, A. Stankiewicz, Structure, energy, synergy, time
- the
fundamentals of process intensification, Ind. Eng. Chem. Res. 48 (2009)
2465–2474, https://doi.org/10.1021/ie801501y.
[11] Y. Kim, L.K. Park, S. Yiacoumi, C. Tsouris, Modular chemical process
intensification: a review, Annu. Rev. Chem. Biomol. Eng. 8 (2017) 359–380,
https://doi.org/10.1146/annurev-chembioeng-060816-101354.
[12] S. Yusuf, L.M. Neal, F. Li, Effect of promoters on manganese-containing mixed
metal oxides for oxidative dehydrogenation of ethane via
a
cyclic redox
scheme, ACS Catal.
acscatal.7b02004.
7
(2017) 5163–5173, https://doi.org/10.1021/
[13] S. Brunauer, P.H. Emmett, E. Teller, Adsorption of gases in multimolecular
layers, J. Am. Chem. Soc. 60 (1938) 309–319, https://doi.org/
10.1021/ja01269a023.
Acknowledgements
[14] C.A. Schneider, W.S. Rasband, K.W. Eliceiri, NIH Image to ImageJ: 25 years of
image analysis, Nat. Methods. 9 (2012) 671–675, https://doi.org/10.1038/
nmeth.2089.
[15] N.C. Popa, D. Balzar, An analytical approximation for a size-broadened profile
given by the lognormal and gamma distributions, J. Appl. Crystallogr. 35
(2002) 338–346, https://doi.org/10.1107/S0021889802004156.
[16] M.O. Özbek, R.A. van Santen, The mechanism of ethylene epoxidation catalysis,
Catal. Letters. 143 (2013) 131–141, https://doi.org/10.1007/s10562-012-0957-
3.
MSCC acknowledges funding from an ESPRC Doctoral Training
Grant, and from an IChemE Andrew Fellowship. The research was
carried out with the financial support from EPSRC grant no. EP/
K030132/1. Maxime Duvieusart and Brian Graves are acknowl-
edged for help with the SEM images. Dr Robert B Grant is acknowl-
edged for help with the FTIR and for helpful discussions on the
epoxidation mechanism. Dr Michael W Gaultois is acknowledged
for the XRD Rietveld analysis. Zlatko Saracevic is acknowledged
for help with the nitrogen adsorption analysis.
[17] M.C.N. Amorim de Carvalho, F.B. Passos, M. Schmal, Study of the active phase
of silver catalysts for ethylene epoxidation, J. Catal. 248 (2007) 124–129,
https://doi.org/10.1016/j.jcat.2006.10.030.
[18] V.I. Bukhtiyarov, A. Knop-Gericke, Ethylene Epoxidation Over Silver Catalysts,
in: C. Hess, R. Schlögl (Eds.), Nanostructured Catal. Sel. Oxid., Royal Society of
Chemistry, Cambridge, 2011: pp. 214–247. doi:10.1039/9781847559876.
[19] P. Christopher, S. Linic, Shape- and size-specific chemistry of ag nanostructures
in catalytic ethylene epoxidation, ChemCatChem. 2 (2010) 78–83, https://doi.
org/10.1002/cctc.200900231.
[20] C.H. Collett, J. McGregor, Things go better with coke: the beneficial role of
carbonaceous deposits in heterogeneous catalysis, Catal. Sci. Technol. 6 (2016)
363–378, https://doi.org/10.1039/C5CY01236H.
[21] J. Assal, B. Hallstedt, L.J. Gauckler, Thermodynamic assessment of the silver-
oxygen system, J. Am. Ceram. Soc. 80 (1997) 3054–3060, https://doi.org/
10.1111/j.1151-2916.1997.tb03232.x.
[22] J. Lu, J.J. Bravo-Suárez, A. Takahashi, M. Haruta, S.T. Oyama, In situ UV-vis
studies of the effect of particle size on the epoxidation of ethylene and
propylene on supported silver catalysts with molecular oxygen, J. Catal. 232
(2005) 85–95, https://doi.org/10.1016/j.jcat.2005.02.013.
Appendix A. Supplementary material
XRD pattern of 15 wt% Ag/SrFeO3. Deconvolution of mole frac-
tions from analyser response. Definitions of selectivity and conver-
sion. Open data access available from https://www.repository.cam.
ac.uk. Supplementary data associated with this article can be
found, in the online version, at https://doi.org/10.1016/j.jcat.
2017.12.030.
References
[23] G.W. Busser, O. Hinrichsen, M. Muhler, The temperature-programmed
desorption of oxygen from an alumina-supported silver catalyst, Catal.
Letters. 79 (2002) 49–54, https://doi.org/10.1023/A:1015300520227.
[24] S. Linic, H. Piao, K. Adib, M.A. Barteau, Ethylene epoxidation on Ag:
identification of the crucial surface intermediate by experimental and
theoretical investigation of its electronic structure, Angew. Chemie Int. Ed.
43 (2004) 2918–2921, https://doi.org/10.1002/anie.200353584.
[25] S.R. Seyedmonir, J.K. Plischke, M.A. Vannice, H.W. Young, Ethylene oxidation
over small silver crystallites, J. Catal. 123 (1990) 534–549, https://doi.org/
10.1016/0021-9517(90)90149-E.
[1] C.H. McAteer, R. Murugan, Y. V. Subba Rao, Heterogeneously Catalyzed
Synthesis of Heterocyclic Compounds, in: Adv. Heterocycl. Chem., Elsevier
Ltd, 2017: pp. 173–205. doi:10.1016/bs.aihch.2016.03.003.
[2] S. Rebsdat, D. Mayer, Ethylene Oxide, Ullmann’s Encycl. Ind. Chem. (2012)
547–572. doi:10.1002/14356007.a10_117.
[3] R.M. Lambert, F.J. Williams, R.L. Cropley, A. Palermo, Heterogeneous alkene
epoxidation: past, present and future, J. Mol. Catal. A Chem. 228 (2005) 27–33,
https://doi.org/10.1016/j.molcata.2004.09.077.
[4] A. Michaelides, M.L. Bocquet, P. Sautet, A. Alavi, D.A. King, Structures and
thermodynamic phase transitions for oxygen and silver oxide phases on Ag{1 1