Catalysis Today 338 (2019) 31–39

Contents lists available at ScienceDirect

Catalysis Today
journal homepage: www.elsevier.com/locate/cattod

From qualitative to quantitative understanding of support effects on the T

selectivity in silver catalyzed ethylene epoxidation
⁎
J.E. van den Reijena, W.C. Versluisa, S. Kanungob, M.F. d’Angelob, K.P. de Jonga, P.E. de Jongha,
a
Inorganic Chemistry and Catalysis, Debye Institute for Nanomaterials Science, Utrecht University, Utrecht, the Netherlands
b
Laboratory of Chemical Reactor Engineering, Department of Chemical Engineering and Chemistry, Eindhoven University of Technology, Eindhoven, the Netherlands

A R T I C LE I N FO A B S T R A C T

Keywords: In the epoxidation of ethylene, catalyzed by supported Ag particles, the support plays not only a beneficial role,
Ethylene oxide but can also negatively impact the selectivity to ethylene oxide. Especially high surface area supports, and
Silver supports containing acid groups, seem detrimental for the selectivity, which is attributed to secondary reactions
Heterogeneous catalysis on the support surface. We prepared Ag nanoparticles on supports with a wide range of specific surface areas and
Kinetic model
different chemical nature, but with similar Ag metal loading and particle size. Indeed, a strong dependence of
Support effects
Epoxidation
selectivity on both the specific surface area, and the chemical nature of the oxide support was found, with α-
Al2O3 giving much more selective catalysts than SiO2. Furthermore, the conversion was an important factor in
determining the selectivity, while on the other hand the support had no influence at all on the ethylene con-
version. We described our experimental selectivities with a reaction model that used the rate constant for the
secondary reaction of oxidation of ethylene oxide to carbon dioxide and water as the only variable parameter.
The model was able to adequately describe the experimental results. It gave insight into the dependence of
selectivity on conversion, how the secondary reaction depended on the chemical nature of the support, and how
its rate scaled linearly with the support’s specific surface area. Building on this insight we modified a SiO2
support to passivate its OH-groups, which indeed yielded a 94% decrease in the rate of the secondary reaction,
performing even better than α-Al2O3 with similar specific surface area would. We have shown that the selectivity
towards ethylene oxide is dependent on a metal specific, intrinsic selectivity, and the decrease in selectivity with
conversion to be support dependent. This decrease is caused by the isomerization of ethylene oxide, which is
found to be first order in ethylene oxide and the active sites on the catalyst support.

1. Introduction Silver has remained the key component in commercial catalysts, but
other components have been added [14]. The original patent of Lefort
Ethylene oxide is a key chemical feedstock, which is produced proposed powdered or colloid silver, but catalysts used in industrial
worldwide on a million-ton scale [1–5]. Every year 16% of the ethylene applications nowadays are supported on low surface area α-alumina.
produced worldwide is converted to ethylene oxide, yielding a com- Furthermore, they are promoted by a range of metal-based promotors
pound annual growth rate of 4.3% in the first half of this decade, re- including cesium, and chlorine is continuously co-fed in the gas stream
sulting in a production of 34.5 million ton in 2016 [2]. Ethylene oxide during the process [1,15–17]. These changes resulted in an increased
is primarily used to produce ethylene glycol, which is mainly used as selectivity to ethylene oxide from 68 to 70% for early unpromoted
antifreeze, but also used in the production of various polymers [1,6,7]. silver catalysts [3], to up to 90% for present industrial catalysts
Its synthesis from ethylene is challenging, as full combustion of ethy- [2,3,18].
lene leads to the formation of water and carbon dioxide and has to be As selectivity is key, it is interesting to investigate the reaction paths
avoided. Hence, achieving high selectivities is key for a viable process to ethylene epoxide and CO2 formation. In 1945 Twigg was the first to
[8,9]. propose isomerisation of ethylene oxide to a short lived intermediate,
The current industrial process is based upon ethylene oxidation which he proposed was acetaldehyde formed by isomerization of
using a silver catalyst discovered in 1931 by Theodor Lefort [10], and ethylene oxide [19]. Lee et al. showed for a wide range of metal oxide
industrialized by Union Carbon in 1936 and Shell in 1958 [11–13]. support materials that the support material catalysed the ring opening

⁎
Corresponding author.
E-mail address: p.e.dejongh@uu.nl (P.E. de Jongh).

https://doi.org/10.1016/j.cattod.2019.04.049
Received 30 October 2018; Received in revised form 20 March 2019; Accepted 13 April 2019
Available online 14 April 2019
0920-5861/ © 2019 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/BY-NC-ND/4.0/).
J.E. van den Reijen, et al. Catalysis Today 338 (2019) 31–39

conversion, based on empirical observations, without offering insight

into the origin of this relationship [32–35]. It was proposed that the
isomerization of ethylene oxide would be a first order reaction in
ethylene oxide and dependent on the support [21,23,36]. The iso-
merization rate of ethylene oxide in the absence of ethylene was de-
termined, but either only on bare support materials or with catalysts
prepared using a variety of preparation methods, with a wide range of
silver particle sizes and in the presence of (potential) promotors
[20–24,36–38].
To address this important question, we set out to show the effect of
the support properties and understand the empirical correlation be-
tween conversion and selectivity, by carefully excluding effects due to
differences in particle size, metal loading or the presence of promotors.
Fig. 1. Proposed reaction scheme for ethylene oxidation over supported silver To this aim we prepared series of catalysts on supports with different
catalysts comprising several reaction steps. In first instance the ethylene is ei- specific surface areas and surface groups, but having similar silver
ther selectively oxidized to ethylene oxide, or fully oxidized. The ethylene oxide
particle size and weight loading, and being prepared using the same
formed can be recovered as a product but can also undergo consecutive oxi-
metal precursor. The catalytic performance was described using a ki-
dation via a two step-reaction: first the isomerization of ethylene oxide to
acetaldehyde over the support and then the oxidation of acetaldehyde to water
netic model, enabling us to quantify the impact of the support material
and carbon dioxide over silver. on the selectivity, and understanding the correlation between conver-
sion and selectivity in the ethylene epoxidation reaction [32,36]. An
experiment in which silanol surface groups were blocked by organic
of ethylene oxide to form acetaldehyde [20]. However, since acet-
capping agents demonstrated the strength of this approach to limit the
aldehyde is rarely found under reaction conditions, it is thought that
secondary reactions and increased the overall selectivity of the reaction.
acetaldehyde is rapidly further oxidized to CO2 in the presence of silver
[18,21,22]. This led to a proposed scheme for the silver catalyzed
2. Methods
ethylene epoxidation as displayed in Fig. 1 [23].
Selective oxidation of ethylene to ethylene oxide (r1) competes with
2.1. Catalyst preparation
full oxidation to CO2 and H2O (r2). Full oxidation is thermodynamically
more favorable than selective oxidation (−1323 kJ/mole compared to
A selection of fumed silica supports was obtained from Safic-Alcan
−106.7 kJ/mole respectively) [1]. The ethylene oxide that is formed
Necarbo, specifically Aerosil OX-50, Aerosil 200, Aerosil 300, Aerosil
can also undergo oxidation (r3); it is believed that isomerization to
380, Aerosil R812 and Aerosil R972 (all produced by Evonik). These
acetaldehyde (r3,a) is followed by fast further oxidation (r3,b). While
materials primarily differ by their specific surface area, except the latter
reactions r1, r2 and r3,b are believed to be catalyzed by the metal par-
two which are silylated versions of Aerosil 300 and Aerosil 150 re-
ticle, the isomerization of acetaldehyde is believed to occur on the
spectively. A hydrophilic equivalent of Aerosil R812 was obtained after
catalyst support. A common strategy to limit side reactions on the
heat treatment of this material in a flow of oxygen (GHSV = 2000 h−1)
support is minimizing the support specific surface area [22]. Yong,
at 400 °C for 2 h to remove the organic capping agents. The hydrophilic
Kennedy and Cant showed a low selectivity towards ethylene oxide for
analogue is expected to be equivalent to Aerosil 300. Quartz was sup-
high surface area supports such as titania and γ-alumina and a high
plied by Sigma Aldrich (≥99.995% trace metals basis). 1 m2/g and
selectivity over catalysts supported on low surface area materials [24].
8 m2/g α-alumina were obtained from Sigma-Aldrich (1 m2/g;
They also reported an influence of the chemical nature of the support.
0.006 cm3/g; -100 mesh, Sigma Aldrich) and BASF (8 m2/g; pore vo-
Different rates of ethylene oxide isomerization for α-alumina and for
lume 0.02 cm3/g, AL4196E, BASF) respectively. All support materials
silica supported catalysts were found [24]. This is in agreement with
were pressed and sieved into a 38–90 μm sieve fraction prior to im-
the results from Lee et al. who reported increased rates of ethylene
pregnation.
oxide isomerization over acidic supports, with silica and α-alumina
20 m2/g α-alumina was prepared using a procedure described be-
yielding higher selectivities and the α-alumina catalyst outperforming
fore [39]. In a typical synthesis 1.5 g of PMMA colloidal crystal powder
the silica supported catalyst [20].
made by emulsion polymerization of methyl methacrylate (MMA, 99%,
Apart from limiting the support surface area, and choosing a fa-
≤30 ppm 4-methoxyphenol inhibitor, Sigma-Aldrich) was placed in a
vourable support chemistry, another strategy is to tune the surface
Buchner funnel and drop-wise wetted with approximately 3 mL solution
properties of a given support. Industrial catalysts contain caesium,
of 1 M aluminum nitrate nonahydrate in demineralized water/me-
which passivates acidic surface groups and limits the isomerization of
thanol (1:1 vol ratio) and soaked for 2 min. Thereafter the excess so-
ethylene oxide to acetaldehyde [17,22,25,23]. During preparation of
lution was removed by applying vacuum for 20 min resulting in a dry
the catalyst, caesium is co-impregnated with silver, and is therefore not
powder. The powder was wetted with ammonia/methanol (1:1 vol
only present on the support surface, but also decorates the silver surface
ratio) after which the excess solution is removed via vacuum suction
[1,17,26]. Özbek and van Santen showed that caesium also influences
and the sample was dried for 20 min. These steps were repeated three
the reactivity of the active metal surface, as higher caesium con-
times before drying the powder overnight at room temperature. The
centration leads to lower overall conversion [20,25,27–29]. Although
dried powder was calcined at 1150 °C under a nitrogen flow with a gas
caesium generally improves the selectivity, in high concentrations it
hourly weight velocity (GHWV of at least 20 Lgpowder−1 hr−1 for 6 h
can even lead to lower overall selectivies, probably because caesium
after a 5 °C/min heating ramp followed by a second calcination in a
introduces oxygen vacancies on the silver surface, which are known to
flow of oxygen with a GHWV of at least 20 L gpowder−1 hr−1 at 1150 °C
be active sites for full combustion [15,27]. It is clear that if one wants to
for 6 h after a 5 °C/min heating ramp.
study intrinsic support effects, promotors such as caesium should rather
All catalysts were prepared by incipient wetness impregnation.
be avoided.
Support materials were dried for 24 h at 120 °C in a static oven in air.
An open question that is seldomly addressed in academic open lit-
This was sufficient to remove adsorbed water from the macroporous
erature is the strong dependence of selectivity on conversion, the exact
pores of the alumina and from hydrophobic silica materials. For the
nature of this dependence and how it can be described [10,30,31].
hydrophilic silica some water was retained in the sample.
Industrial sources report a linear decrease of selectivity with increasing
Silver oxalate was prepared by adding a 20 mL, 12 M aqueous

32
J.E. van den Reijen, et al. Catalysis Today 338 (2019) 31–39

solution of silver nitrate (99+%, Sigma-Aldrich) to a 40 mL, 12 M 2.3. Catalytic testing

aqueous solution of oxalate dihydrate (99.5%, Merck) at 60 °C and fil-
trating and washing the precipitate with miliQ water [31,40–42]. A In a typical catalysis experiment, 50–150 mg of catalyst grains
solution of 0.82 mmol/gram freshly prepared Ag2C2O4 in a mixture of (38–90 μm) was loaded in a quartz reactor between two layers of quartz
ethylene diamine (en) and water (0.73 g/g milliQ/(en), or 1:4:16 mol wool. To activate the catalyst, the sample was heated to 180 °C in a flow
ratio Ag2C2O4 : en : H2O) was used to impregnate the support materials. of 16.5 mL min−1 of 8.5% oxygen and 30% ethylene in helium. The gas
In the case of hydrophobic supports a mixture of ethylene diamine, hourly weight velocity (GHWV) per gram of silver was varied over the
water and ethanol was used in a volumetric ratio of 20 : 55 : 25 (en : 25–160 L gAg−1 hr−1 range in order to obtain ethylene conversions in
water : ethanol) to ensure wetting of the support. The volume of solu- the range of 0–10 % (after a stable conversion was obtained), and thus
tion added corresponded to the total pore volume as obtained from operate in differential conditions. The reaction products were char-
nitrogen physisorption. During impregnation, the solution was added acterized using online gas chromatography. The outlet gases of the
dropwise to the dry support material which was continuously mixed to reaction were analyzed every 15 min by an online Compact GC
ensure a homogenous distribution of the silver salt over the support. (Interscience) equipped with a Porabond Q column and a Molsieve 5 Å
The impregnated material was put in an oven at 60 °C in static air for column in two separate channels, both with a thermal conductivity
12 h to dry. After 5, 10 and 30 min the material was thoroughly mixed detector. Conversion data were calculated from data retrieved after
to limit macroscopic inhomogenities [43]. steady state was reached, which was typically after 8 h. A more detailed
The dry grains were heated in a U-shaped reactor to decompose the description of the calculations can be found in the Supporting in-
silver salt, forming a catalyst with 15 wt% silver and an average particle formation (Supporting information section E).
size of 60 nm. In typical heat treatment, 200 mg of sample was heated
with a 5 °C/min ramp to a temperature between 215 and 300 °C in an 2.4. Kinetic model describing selectivity trends
oxygen or nitrogen atmosphere with a gas hourly weight velocity of
120 L g−1 hr−1 in order to obtain the desired particle size [31]. All The overall reaction was modelled based on the reaction scheme
catalysts had a nominal weight loading of 15 wt% of silver, but due to presented in Fig. 1. A primary selectivity is determined by the ratio
the different support specific surface area the silver surface density (Ag/ between two parallel paths for the oxidation of ethylene: selective
m2support) differs greatly between the catalysts, an estimation of the oxidation forming the desired product ethylene oxide (r1) and full
silver surface density is given in the Supporting information section C oxidation to form carbon dioxide and water (r2). The ethylene oxide can
and Table S1. either be retrieved, or oxidized further in the reactor (r3) via the iso-
merization to acetaldehyde [19,36]. For the oxidation reactions r1, r2,
that are assumed to occur over the silver nanoparticles, quite some
2.2. Characterization information is available in literature.
Regarding r1, the partial oxidation of ethylene to ethylene oxide, the
Nitrogen physisorption isotherms for all support materials were reaction is taken to be first order in ethylene, which is a logical as-
measured at −196 °C on a Micromeritics TriStar 3000 apparatus. The sumption, but also experimentally reported by Campbell et al. for ex-
specific surface area of the support was calculated by fitting experi- perimental conditions similar to ours [45]. The reaction order in
mental data with the BET equation (0.05 < p/p° < 0.25). BJH pore size oxygen is taken as 0.5, being in the range of -0.5 to 1.5 as reported by
distributions were obtained from the corresponding adsorption iso- Stegelmann et al. [46]. Furthermore both reaction orders are in line
therm, with the total pore volume at a relative pressure of 0.995. For with the results of Linic and Barteau [47].
the high surface area α-alumina mercury porosimetry was performed For the full combustion of ethylene (r2) a reaction order in oxygen in
using Micromeritics AutoPore IV 9500 to probe the large pores of the the -1 to 1 range, and typically close to zero, is reported by several
material. Pore volume data was calculated over the range different authors [46,48,49]. This zeroth order in oxygen might appear
0.0007–227.5270 MPa. The contact angle between the support mate- at first sight surprising, given the consumption of oxygen in this reac-
rials and the mercury was assumed to be 130° and the mercury surface tion. However, it is the combination of both a negative and a positive
tension and density 0.485 N/m and 13.5335 g/mL, respectively. dependence on the oxygen partial pressure. The positive order stems
The size and spatial distribution of the obtained silver particles were from the consumption of oxygen in the reaction, but it is balanced by
measured using scanning electron microscopy (SEM). The images were the fact that a more oxidized silver surface favors selective oxidation
obtained using a FEI XL30 FEG SEM. The images were processed using instead of full combustion [27]. Stegelmann et al. observed the ethylene
iTEM Soft Imaging System software [44]. Diffuse-Reflectance UV/Vis reaction order in r2 lower than the ethylene order in r1 and is taken by
spectra of the supported silver catalysts were measured using a Varian us as 1/2 [46].
CARY 500 Scan UV/Vis-NIR spectrophotometer with an integrating Because both the r2 and r3 yield carbon dioxide and water at equal
sphere. In a typical analysis, 30 mg of analyte was diluted with 600 mg stoichiometry, it is not straightforward to derive the individual values
pristine α-Al2O3 (1 m2 g−1, -100 mesh, Sigma Aldrich) and ground to a from the overall catalytic data. Deconvolution of these rate constants
fine powder before taking a spectrum in the 1000 – 200 nm range. A was performed by extrapolating the obtained catalytic data of all
background measurement of pristine α-Al2O3 was taken and subtracted measured catalysts to a conversion of 0.04% ethylene. At such low
from the obtained spectrum. Crystal phase analysis was performed with conversion, the ethylene oxide partial pressures is close to zero, and
a Bruker D8 Phaser diffractometer (X-Ray Diffraction, Bruker D8 Phaser isomerization of this component, and hence r3, can be neglected. The
diffractometer equipped with a Co Kα source (λ = 0.1789 nm) using an resulting selectivities are therefore governed by the ratio between r1
angle range from 20° to 90° in 2θ and comparing XRD patterns with the and r2. Hence the rates for r1 and r2, and the corresponding rate con-
PDF-4 + 2016 database. stants k1 and k2, were calculated from the ratio r1/r2 and the overall
The weight loss due to burning of the organic capping agents from reactant conversion rate (r1+r2) at the extrapolated conversion of
the hydrophobic silicas was analyzed using a Perkin Elmer Pyris 1 TGA 0.04%.
under a 10 mL min−1 flow of 20% oxygen in nitrogen. For these ex- Using the obtained rate constants k1 and k2 and the experimental
periments the temperature was increased from 50 °C to 900 °C with a feed concentration as the initial conditions, the rates and partial pres-
ramp of 5 °C min−1. The decomposition products were analyzed using a sures of each component were calculated differentially over the catalyst
Pfeiffer Vacuum OmniStar mass spectrometer. bed. In each step the result of the previous calculation was used, with
initially values defined by the feed composition. The conversion and
selectivity were calculated for each step based on the calculated partial

33
J.E. van den Reijen, et al. Catalysis Today 338 (2019) 31–39

Fig. 2. Electron microscopy images of silver on an α-Al2O3 support with a support specific surface area of 1 m2/g (Al2O3_1, left), silver on a fumed silica support with
a support specific surface area of 50 m2/g (SiO2_50, middle) and histogram of the diameter of the silver particles on both catalysts. These histograms are obtained
from measuring at least 300 particles and are subsequently normalized to accurately compare the distributions.

pressures at increasing catalyst bed lengths. A comparison of the cal- therefore a low amount of surface hydroxyl species per gram of mate-
culated selectivities and the real experimental selectivities in the rial. The α-alumina supports have surface areas ranging from 1 to
ethylene conversion range of 0–10 % allowed to derive the rate r3 and 20 m2/g, with the highest specific surface area belonging to a non-
corresponding rate constant k3 for the isomerization of ethylene oxide commercial ordered macroporous α-alumina that was made in-house
to acetaldehyde, which was the sole fitting parameter. This value for k3 [39]. Quartz has a support specific surface area of 0.1 m2/g as specified
for each individual catalyst. rate was obtained by a sum of least squares by the supplier. Due to the low surface area this value could not be
fit. The secondary reaction as represented by r3 and k3 is the process reliably verified due to the measurement error of nitrogen physisorp-
that determines the dependence of overall selectivity on support tion. The fumed silica materials have surface areas ranging from 50 to
properties, and the correlation between conversion and selectivity, and 380 m2/g. Note the low specific surface area of the silylated supports
is hence the main topic of this study. compared to the hydrophilic counterparts (entries SiO2_300,
MeSiO2_260 and MeSiO2_260-ht). These are expected to have similar
specific surface areas since the hydrophobic silica (support used in
3. Results
sample MeSiO2_260) is obtained via silylation of the hydrophilic silica
(a 300 m2/g Aerosil) and MeSiO2_260-ht is prepared by a heat treat-
3.1. Catalyst preparation and characterization
ment of MeSiO2_260. The slight gain in specific surface area after
converting hydrophobic (MeSiO2_260) into hydrophilic (MeSiO2_260-
Catalysts in this paper are coded via the method of Support_surface
ht) silica suggests that for the hydrophobic silica the organic species are
area, i.e. Al2O3_1 is a sample of α-alumina of 1 m2 g−1 impregnated
blocking part of the pores.
with silver and SiO2_380 is a sample of fumed silica of 380 m2 g−1
The density of hydroxyl species is slightly higher for the alumina
impregnated with silver. In Fig. 2 typical Scanning Electron Microscopy
than for the silica supports in Table 1, while the differences between the
(SEM) images of a low surface area α-alumina and a low surface area
crystalline and amorphous silica materials are small [54,61]. It must be
fumed silica supported silver catalysts are shown. The support material
said that the determination of the exact number of surface species is
is visible as a grey material, dotted with bright white silver metal
difficult for materials with such low specific surface area and can differ
particles. The Al2O3_1 catalyst in the left image is supported on large
significantly based on the preparation procedure [54]. It is expected
jagged chunks of non-porous material. The silica support (middle
that the density of these species is lower on crystalline surface as the
frame) has a fine structure and appears to have smaller silver particles
more reactive species are condensed [54,61]. When the amorphous
than the α-alumina supported catalyst. The histogram (obtained from
silica materials are silylated in order to block the surface hydroxyl
measuring at least 300 particles) in the right frame shows that the
species (MeSiO2_110 and MeSiO2_260), a sharp decrease in the density
average particle size is only slightly different (surface area corrected
of hydroxyl species is observed (compared untreated to Aerosil mate-
average particle size of 70 nm for Al2O3_1 compared to 60 nm for
rial). This significant drop in hydroxyl density is expected because these
SiO2_50), while the size distribution (fitted as a log-normal distribution)
species have reacted with the silylating agent, resulting in inert siloxy
is significantly broader on the former, with some particles with sizes of
groups. The catalyst supported on quartz (SiO2_0.1) does differ in
100–150 nm being present. The apparent difference between the silver
chemical nature from the rest of the silica samples but acts as a low
particle sizes in the left and middle frame is due to these large particles
surface silica which can be compared with the low surface area α-alu-
being more noticeable.
mina catalysts.
The large crystallites, in the case of the α-alumina support, can be
The catalysts tabulated in Table 1 have a wide range of support
explained by the preparation method of the support materials, as they
specific surface area (0.1 – 380 nm2), while the metal weight loading
are produced at high temperatures to crystallize to the most thermally
and silver particle sizes are very similar. This is achieved by carefully
stable α-alumina phase (at least 1000 °C before α-alumina is formed),
adjusting the heat treatment conditions during catalyst preparation, as
causing a collapse of most of the material porosity and condensing
was more extensively discussed in a previous publication [31]. In pre-
surface groups [50–52]. The silica support is prepared by sintering
vious studies large variations in support properties were always ac-
small silica spheres (5–50 nm in diameter) into an agglomerate which
companied by different particle sizes. In our case we have been able to
has no defined shape and exhibits interparticle nanoporosity [53]. The
tune the particle size to a limited range for which we reported earlier
temperature at which this flame hydrolysis is performed determines the
that variations within this particle size range are not expected to sig-
support specific surface area, but does not reach the temperatures
nificantly affect the selectivity [31]. This allows to attribute observed
needed to crystallize the silica [53]. XRD (see Figure A1 in the Sup-
differences in catalytic results as discussed in the next section solely to
porting information) confirms the highly crystalline nature of the α-
the differences in chemical nature and specific surface area of the
alumina and the amorphous nature of the fumed silica.
support material.
An overview of the main characteristics of the silver catalysts and
the different support materials used, is given in Table 1. The α-alumina
and quartz support materials exhibit low specific surface areas and

34
J.E. van den Reijen, et al. Catalysis Today 338 (2019) 31–39

Table 1
Catalysts with support characteristics and average metal particle sizes.
Catalyst Support material Specific surface area support (m2/g) Density of hydroxyl Ag particle size (nm)
species (OH nm−2)
Brand name Type Provided by Nitrogen TEM/SEM ± sigma UV/Vis (SPR Peak
supplier physisorption position)‡

Al2O3_1 N/A α-alumina 1 <5 6 [54,55] 71 ± 25 70 (457)

Al2O3_8 AL4196E α-alumina 8 n.d. 6 [54,55] 77 ± 22 72 (463)
Al2O3_20 N/A α-alumina N/A 20* n.d. n.d. 80 (480)
SiO2_0.1 N/A Quartz 0.1 <1 < 3 [56,57,58] 110 ± 70 108 (503)
SiO2_50 Aerosil OX50 Fumed silica 50 46 3 [53] 60 ± 15 70 (455)
SiO2_200 Aerosil 200 Fumed silica 200 196 3 [53] 42 ± 9 41 (401)
SiO2_380 Aerosil 380 Fumed silica 380 315 3 [53] 34 ± 11 41 (400)
MeSiO2_110 Aerosil R972 Hydrophobic fumed silica 110 97 0.71 [59] 68 ± 33 71 (435)
MeSiO2_260 Aerosil R812 Hydrophobic fumed silica 260 225 0.44 [60] 79 ± 39 77 (485)
MeSiO2_260-ht N/A Heat treated Hydrophobic N/A 278 n.d. n.d. 77 (485)
fumed silica

‡
Peak maximum of surface plasmon resonance peak in diffuse reflectance UV/Vis spectroscopy.
* Support specific surface area as determined by mercury intrusion.

3.2. Catalysis all catalysts. The results validate our starting point that our preparation
method leads to Ag nanoparticles that are very similar in properties,
The temperature dependent catalytic performance of two catalysts despite being prepared on very different supports in terms of chemistry
(Al2O3_1 and SiO2_200) is shown in Fig. 3. We selected these two ex- and surface area. Furthermore, it validates the assumption that the
amples as the specific support surface and nature of these two catalysts oxidation reaction of ethylene occurs solely on the silver particles, and
differ greatly. The oxidation of ethylene starts around 170 °C with si- that the primary selectivity (which fraction of ethylene is converted to
milar activity over the whole temperature range for these two catalysts the desired product, and which fraction is combusted) is independent of
(left frame Fig. 3). In the middle frame, the selectivities are plotted the support used.
versus the reaction temperature. Selectivities toward ethylene oxide The overall selectivity in the middle frame shows a very different
range from 40 to 75% for the α-alumina supported catalyst. For the picture, with a much higher selectivities for the low surface area α-
silica supported catalyst the selectivies are clearly lower, ranging from alumina than for the high surface area silica supported catalyst. This
20 to 50%. In the right frame of Fig. 3, the Arrhenius plots show the confirms earlier reports of support-dependence of the selectivity and is
activation energy for the epoxidation reaction for both catalysts, based attributed to a secondary selectivity loss mechanism via subsequent
on the yield of ethylene oxide (the explanation for this method can be oxidation of the formed ethylene oxide. It is likely that the isomeriza-
found in the Supporting information section D). The apparent activa- tion of ethylene oxide (r3,a) is catalyzed by surface groups of the sup-
tion energies for ethylene oxide formation of 74 and 67 kJ/mol for port. The subsequent oxidation of acetaldehyde (r3,b) is assumed to
catalysts supported on silica and alumina respectively are well in line occur on the silver particles rather than on the support surface, and
with the activation energies reported in literature, which range from 50 hence is expected to occur with a similar rate for all catalysts. The
to 80 kJ/mol [35,46,62,63]. decreasing selectivity with increasing conversion, as shown in the
The remarkably similar conversions over the entire temperature middle frame, has been reported before [33,39], stating a linear or first
range in the left frame of Fig. 3 strongly suggests that neither the nature order correlation [21,33,36,64]. However, no chemical or physical
of the support nor the silver surface density (Ag/msupport2) influence the explanation for this relation has been given. In the following section we
activity of the catalyst. The very similar activation energies observed on try to describe and understand the dependence of selectivity on support
the two totally different supports as represented in the right frame of properties and conversion. For a direct comparison of the selectivity,
Fig. 3 imply that the same is true for the rate r1; and since the overall catalytic performance should be compared at similar conversion instead
ethylene conversions are nearly identical, this also applies to r2. First of of at different temperatures, to exclude the contribution of differences
all, this shows the accuracy and reproducibility of the measurements. in activation energies due to the different temperatures. Hence, all
The same metal precursors and solvents were used in the preparation of further measurements discussed in this paper are obtained at a constant

Fig. 3. Conversion (left), selectivity (mid) and Arrhenius plots (right) for the formation of ethylene oxide of Ag_Al2O3_1 (blue circles) and Ag_SiO2_200 (orange
squares) measured at a GHWV of 43 L gAg−1 hr—1 at atmospheric pressure. Lines added to guide the eye. (For interpretation of the references to colour in this figure
legend, the reader is referred to the web version of this article).

35
J.E. van den Reijen, et al. Catalysis Today 338 (2019) 31–39

of the selectivity with conversion and support specific surface area, the
mechanism of this secondary reaction is not extensively studied and a
linear decay with conversion is often presumed. In the following section
we will discuss a precise model to describe the catalytic data, and the
dependence of selectivity on conversion and support properties, al-
lowing more insight in the underlying mechanisms and providing a
quantitative description of the rate of the secondary reaction.

3.4. Quantitative description and understanding of the selectivity trends

Our model to describe the experimental results consists of two

parallel and one consecutive reaction (with rates r1, r2 and r3 respec-
tively) and is given in Equations 1-4. In these equations the partial
pressures of the relevant compounds ethylene, oxygen, ethylene oxide
and acetaldehyde are represented as pC2 H4 , pO2 , pEO and pAA respec-
tively. The rationale for the reaction orders in r1 and r2 is explained in
the experimental section. The primary selectivity is determined by the
balance between the two possibilities for the oxidation of ethylene:
selective oxidation forming the desired product ethylene oxide (r1) and
full oxidation to form carbon dioxide and water (r2). The ethylene oxide
Fig. 4. Selectivity of a selection of catalysts over a 0–10 % range of ethylene product can either be retrieved or oxidized further in the reactor (r3).
conversions showing a support specific surface area and conversion dependent This secondary reaction involves isomerisation of ethylene oxide to
selectivity. Markers correspond to averaged conversion and selectivity obtained
acetaldehyde (r3,a) and subsequent full oxidation of the isomer, yielding
after at least 2 h of steady state operation after setting the GHWV. All data is
carbon dioxide and water (r3,b). Isomerisation to acetaldehyde is con-
obtained at an operating temperature of 180 °C. Different conversion were
obtained over a range of 25–160 hr−1 GHWV. Drawn lines correspond to first
sidered the rate limiting step [18,21,22], hence the rate of r3,b is con-
order fit of catalytic data using the method described. sidered to be infinitely fast, and the overall reaction rate r3 can be taken
as equal to r3,a, and is modelled as a first order rate equation. The rates
of the three reactions r1, r2 and r3,a, were obtained from fitting the
temperature of 180 °C, and desired conversions were obtained by
experimental data.
changing the gas hourly weight velocity (GHWV) rather than the tem-
1
perature. r1 = k1 pC2 H4 pO22 (1)
1
3.3. Influence of the conversion on the selectivity r2 = k2 pC22 H4 (2)

In Fig. 4, the selectivity to ethylene epoxide as a function of ethy- r3, a = k3, a pEO (3a)
lene conversion is shown for a selection of catalysts (Similar plots for
the other samples can be seen in Figure I1 in the Supporting informa- r3, b = k3, b pOn2 pAA
m
(3b)
tion). Only results up to 10% conversion are displayed, as for higher
if r3, a ≫ r3, b
conversions mass transfer limitations start to play a major role (see
Supporting information section F for the calculation of the expected
r3 ≈ k3 pEO (4)
mass transfer effects). Striking is that, although the selectivities vary
widely, at the lowest possible conversions measured all selectivities We considered the experimentally obtained selectivities as a func-
converge to a similar value (around 76%). This implies a metal specific, tion of conversion for all catalysts presented in Table 1. The drawn lines
intrinsic primary selectivity, which is independent on the type and the in Fig. 4 represent the calculated selectivities versus conversion for a
specific surface area of the support, and hence determined solely by the selection of these catalysts. Our experimental results clearly showed
silver metal. that at very low ethylene conversions (and hence negligible ethylene
Another general observation is that the low surface area alumina- oxide concentrations) results for all our supported silver catalysts
supported catalysts exhibit significantly higher selectivities than the converge to the same selectivity of 76%. This is in line with earlier
catalyst with the highest support specific surface area, SiO2_380. Also reports that at constant conversion there is no intrinsic effect of particle
within the family of alumina-supported catalysts, there is a clear cor- size on selectivity [31]. From this value, which represents the metal
relation with specific surface area, with decreasing selectivity with specific selectivity, the ratio between r1 and r2, and hence k1 and k2 was
larger surface area. This trend of decreasing selectivity with increasing derived. The average value of the k1/k2 values derived from the ex-
surface area also holds for all the hydrophilic silica supported samples trapolation of all individual catalysts in the dataset was used as a fitting
(Figure I1 in Supporting information). constraint for the rate constants of the r1 and r2 reactions for all sam-
It has been reported before that the selectivity of this process gen- ples.
erally decreases with increasing conversion, but there are no earlier We took a fixed value for the orders of reactions r1 and r2, but in fact
reports on the exact dependence. Based on empirical correlations for the variation in oxygen partial pressure does influence the reaction
low-surface area alumina-supported catalysts, such as Al2O3_1 in Fig. 4, order of both oxygen and ethylene in the r1 and r2 reactions [45–47,65].
a linear dependence is assumed, without a theoretical framework to To justify our simplification, we performed a robustness test in which
explain this linearity. However, from our results it is clear there is no the same calculations were performed with oxygen reaction orders over
simple linear correlation, but that in fact the correlation is more com- the ranges mentioned in literature (-0.5 to 1.5 in oxygen and 1 in
plicated. Considering the overall reaction steps describe earlier in this ethylene for r1 and -1 to 1 in oxygen and 0 in ethylene for r2)
work, it can be expected that the selectivity is inversely dependent on [27,45–47]. The reaction orders in r1 and r2 did not significantly in-
the ethylene oxide partial pressure. As with increasing ethylene oxide fluence the shape of the fitted curve but did change the absolute value
partial pressure, the rate r3 increases. While this trend is well accepted of k3. This did not influence the capability of the model to fit the data,
and provides a satisfactory qualitative description of the dependencies as the relative values did not change. The higher reaction orders in

36
J.E. van den Reijen, et al. Catalysis Today 338 (2019) 31–39

oxygen (above 0 and 0.5 for r1 and r2 respectively) could not adequately
describe the experimental data. These observations support our hy-
pothesis that the orders in both oxygen and ethylene in r1 and r2 are not
determining factors in our assessment of the role of the support on the
r3-reaction. Moreover, at industrially relevant and non-mass transfer
limited conversions, the partial pressures of oxygen and ethylene do not
significantly change and therefore only play an indirect role.
It is clear from Fig. 4 that our model is capable of adequately de-
scribing the overall decreasing selectivity with increasing conversion.
Also, the absolute selectivities calculated using our model correlate well
with the measured values for all catalysts. As the rate orders in r1 and r2
did not have a large influence on these data, it can be concluded that
the good correlation between the measurement and the calculated data
is primarily determined by the r3 reaction, the isomerization of ethylene
oxide. From the decrease in selectivity with increasing conversion, for
each catalyst a value for k3, was derived, which depended on the nature
and surface area of the support and is discussed in detail in the fol-
lowing section. The strong dependence on the support led to the con-
clusion that the rate limiting step for the oxidation of ethylene oxide,
the isomerization of ethylene oxide to acetaldehyde, occurs over the
catalyst support, rather than on the metal surface as is an often pro-
posed theory [21,46,66]. As stated before at low ethylene conversions Fig. 5. k3/k1 obtained from fitting catalytic data of supported silver catalysts
the support has no influence on selectivity, hence selective (r1) and full grouped by support material (α-alumina in blue and silica in orange) versus
oxidation of ethylene (r2) can therefore be assumed to solely occur on support specific surface area. Open symbols corresponding to silylated silica
the surface of the metal particle. with the heat-treated sample (alkyl surface groups removed again) indicated by
As an illustration of how well the model is able to describe the a square. Trendlines are added to guide the eye, these lines are forced through
experimental selectivities, it is interesting to have a close look, at the the origin and the dashed lines are extrapolated to more clearly indicate the
sharp drop in selectivity for the Al2O3_20 sample in Fig. 4 at a con- differences in the slope of the different families. (For interpretation of the re-
ferences to colour in this figure legend, the reader is referred to the web version
version of circa 9%, which is correctly predicted with our model. As the
of this article).
unselective reactions r2 and r3 consume more oxygen than the selective
reaction (r1), the atmosphere is deprived of oxygen at low selectivities
and high conversions (Figure G1 in the Supporting information). In the this by a relatively large error in the exact value for the specific surface
case of Al2O3_20, the resulting oxygen partial pressure is very low area value (provided by supplier). The open symbols correspond to the
(< 0.0034 atm compared to 0.085 atm in the feed, which corresponds catalysts supported on surface-treated silica with intermediate to high
to an oxygen conversion of 96%). This drop is correctly described in our SSA, but low values for k3/k1. MeSiO2_260-ht is depicted as the open
model by the positive reaction order in oxygen for the selective r1 re- square, with k3/k1 in line with the untreated silica. For both the alu-
action while the reaction order in oxygen for full oxidation r2 reaction is mina and silica supported catalysts, the data were accurately described
zero. These orders reflect the fact that a more oxidized silver surface with a linear relation between the rate constant and specific surface
favors the selective oxidation, and hence at these very low oxygen area of the support. The slope depends strongly on the type of material,
contents the full oxidation reactions are more dominant [27,67]. As the with a larger slope for the silica than for the alumina-supported and
reaction follows the Mars-Van Krevelen type mechanism, these low hydrophilic silica catalysts.
oxygen contents are insufficient to reoxidize the surface. At the sites Fig. 5 shows that the rate constant of isomerization is proportional
where these “oxygen vacancies” exist, the underlying silver is exposed to the specific surface area, but only for a given support. Deviations
which acts as an active site for the unselective r2 reaction, the reduced from the linear regression are small and can be ascribed to the ex-
rate of reoxidation causes a sharp drop in the selectivity towards perimental error in the fit by the model. However, a non-zero rate of
ethylene oxide [68]. isomerization in the absence of support material could also be con-
sidered. Possibly due to some minor isomerization occurring over the
3.5. Influence of the nature and specific surface area of the support metal particle [21,46,66]. This is an interesting topic for future in-
vestigations, but not considered further in this work. Mind that the rate
The data in Fig. 4 clearly show that the specific surface area of the constant of the isomerization of the product ethylene oxide to acet-
support has a strong influence on the selectivity, but also the chemical aldehyde, k3, is not a true rate constant of an elementary reaction.
nature of the support seems to play a role. To unravel these two factors, Assuming that within a set of materials, the surface density of the active
the values for k3 obtained from all catalysts mentioned in Table 1 surface groups can be considered constant [53], this linear relation
(catalytic data in Figures I1) were plotted versus the specific surface of suggests that the reaction can be considered as a first order reaction in
the support materials (Fig. 5). To minimize the error originating from the concentration of active surface species. Fig. 5 also shows that the
slight differences in overall activity, the obtained k3 values were di- slope is higher for silica than α-alumina. However, this cannot be ex-
vided by the corresponding value for k1 obtained from the fit. A high plained just based on the surface concentration of hydroxyl species on
value for k3 corresponds to a low selectivity while a low value corre- each material (see Table 1) [54–60,69]. This suggests that the hydroxyl
sponds to a very selective catalyst. species on α-alumina are more reactive than the hydroxyl species on
The three α-alumina supported catalysts are plotted in blue. These silica supports. This provides an explanation for the observations of Lee
catalysts have a low support specific surface and low values for k3/k1 and Yong, who state that silica support materials are relatively inert
(hence high selectivities). The silica supported samples are depicted in [20,66], and Chongterdtoonskul and Goncharova stating that high
orange, these catalysts have a wide range of support specific surface surface area of their alumina samples yield overall low selectivities
areas and values for k3/k1. The silica-supported catalyst with the lowest [70,71].
support specific surface area, the quartz supported SiO2_0.1, is an As the strength and the amount of the active species play a role, the
outlier with a higher k3/k1 ratio than expected. We tentatively explain observations of Lee and Yong and Chongterdtoonskul and Goncharova

37
J.E. van den Reijen, et al. Catalysis Today 338 (2019) 31–39

could both be explained by variations of these two parameters in the times lower specific surface area. High surface area supports allow to
material. In Figure J1 in the Supporting information, we have tested a increase the silver weight loading, potentially leading to even higher
set of catalysts on low surface area support with relatively acidic sur- selectivities as the metal is expected to block reactive hydroxyl species
face species, such as niobia and strontiumtitanate. These types of sup- on the surface and to the possibility to load more active catalyst surface
ports are reported to be very selective by Chongtertoonskul [71,72]. In per reactor volume. Alternatively, these materials could be used to
our case, these catalysts exhibited selectivities towards ethylene oxide produce more stable catalysts as the high interparticle distance in-
similar to a silica catalyst with a specific surface area of 380 m2/g for creases the stability against deactivation due to sintering.
the niobia supported catalyst and 70 m2/g for the strontiumtitanate
sample (Figure J1 in the Supporting information), while both have a Acknowledgements
specific surface area lower than 20 m2/g. This is in accordance to the
observations of both Lee and Yong, where acidic support materials This research was founded by NWO Vici project 16.130.344. The
exhibit lower selectivities towards ethylene oxide [20,24]. We attribute authors would like to thank Jan-Willem de Rijk, Carlo Buijs (SCR
this to a higher rate of ethylene oxide isomerization over acidic support Technical University Eindhoven) and Marlies Coolen-Kuppens (SCR
species. Technical University Eindhoven) for their help on the epoxidation setup
A very striking and interesting feature in Fig. 5 are the low k3/k1 and Marjan Versluijs-Helder for SEM and TGA measurements.
values for MeSiO2-110 and MeSiO2-260. While the support specific
surface area of these materials is comparable to the other silica sup- Appendix A. Supplementary data
ported samples, the selectivity towards ethylene oxide is much higher;
the decrease in the k3/k1 value is 96%. This means that the 260 m2/g Supplementary data associated with this article can be found, in the
silica supported catalyst has a similar catalytic performance to a 20 m2/ online version, at https://doi.org/10.1016/j.cattod.2019.04.049
g α-alumina catalyst. In this sample the original silanol groups are
blocked by organic species. When these alkyl groups were burned off by References
heat treatment (square in Fig. 5), the sample, MeSiO2-260-ht, had a k3/
k1 value similar to a catalyst supported on untreated silica with a si- [1] S. Rebsdat, D. Mayer, Ethylene oxide, Ullmann’s Encycl. Ind. Chem. Wiley-VCH
milar specific surface area. This clearly shows that the difference in Verlag GmbH & Co, KGaA, Weinheim, Germany, 2001, pp. 547–572, https://doi.
org/10.1002/14356007.a10_117.
selectivity can be solely attributed to the difference in the support [2] Research In China, Global and China Ethylene Oxide (EO) Industry Report, 2017-
material, in this case the chemical nature of the surface. 2021, (2017), pp. 1–137 https://www.reportbuyer.com/product/5013139/global-
and-china-ethylene-oxide-eo-industry-report-2017-2021.html.
[3] H. van Milligen, B. VanderWilp, G.J. Wells, Enhancements in Ethylene Oxide /
4. Conclusion Ethylene Glycol Manufacturing Technology, (2016).
[4] J.C. Zomerdijk, M.W. Hall, Technology for the manufacture of ethylene oxide,
Supported catalysts with silver particle sizes of 40–70 nm were Catal. Rev. 23 (1981) 163–185, https://doi.org/10.1080/03602458108068074.
[5] H. Kestenbaum, A. Lange de Oliveira, W. Schmidt, F. Schüth, W. Ehrfeld,
prepared on silica and α-alumina supports with varying specific surface K. Gebauer, H. Löwe, T. Richter, D. Lebiedz, I. Untiedt, H. Züchner, Silver-catalyzed
areas. The silver weight-based activity of the catalysts was similar for oxidation of ethylene to ethylene oxide in a microreaction system, Ind. Eng. Chem.
all catalysts. However, the selectivity in the ethylene epoxidation, the Res. 41 (2002) 710–719, https://doi.org/10.1021/ie010306u.
[6] E. Alper, O. Yuksel Orhan, CO 2 utilization: developments in conversion processes,
most critical factor, depended strongly on the support and on the
Petroleum 3 (2017) 109–126, https://doi.org/10.1016/j.petlm.2016.11.003.
conversion. Although it was known in the literature that the selectivity [7] H.L. Greenberg, M.G. Ott, R.E. Shore, Men assigned to ethylene oxide production or
decreased with increasing conversion, this effect had not been de- other ethylene oxide related chemical manufacturing: a mortality study, Br. J. Ind.
scribed in detail. Using a kinetic model, we could accurately describe Med. 47 (1990) 221–230, https://doi.org/10.1136/oem.47.4.221.
[8] P.P. Mcclellan, Manufacture and uses of ethylene oxide and ethylene glycol, Ind.
the catalytic selectivity of our range of supported silver catalysts and Eng. Chem. (1950) 2402–2407, https://doi.org/10.1021/ie50492a013.
conversions. Even seemingly surprising features, such as a sudden drop [9] R.A. Sheldon, The e factor: fifteen years on, Green Chem. 9 (2007) 1273, https://
in selectivity at relatively high conversions due to a low oxygen con- doi.org/10.1039/b713736m.
[10] T.E. LeFort, Process for the production of ethylene oxide, US 1998878 A, 1935.
tent, was faithfully modeled. Our data fully supported that the primary [11] G.A. Perkins, Dichlobomksii ethees, US2042862, 1936.
selectivity (at very low conversions) is governed by the two alternative [12] J.M. Kobe, W.E. Evans, R.L. June, M.F. Lemanski, Epoxidation, Encycl. Catal.
reaction paths for the ethylene reactant on the silver catalyst surface, (2003) 246.
[13] M.J.F.M. Verhaak, Process for the production of ethylene oxide, US8969602B2,
while a further loss of selectivity at higher conversions is caused by a 2015.
secondary reaction involving the support surface, most likely ethylene [14] R. van Santen, Catalysis in perspective: historic review, Catal. From Princ. to Appl.
oxide isomerization to acetaldehyde followed by complete combustion. (2012), pp. 3–19 doi:978-3-527-32349-4.
[15] M.O. Ozbek, I. Onal, R.A. Van Santen, Ethylene epoxidation catalyzed by chlorine-
Building on the hypothesis of a first order in ethylene oxide for this promoted silver oxide, J. Phys. Condens. Matter 23 (2011) 404202–404209,
secondary reaction, the rate constants were shown to be linearly de- https://doi.org/10.1088/0953-8984/23/40/404202.
pendent on the support specific surface area but differed per material [16] J.T. Jankowiak, M.A. Barteau, Ethylene epoxidation over silver and copper–silver
bimetallic catalysts: I. Kinetics and selectivity, J. Catal. 236 (2005) 366–378,
type. The presumed active site for the isomerization reaction are hy-
https://doi.org/10.1016/j.jcat.2005.10.018.
droxyl groups, and the proportionality suggests a first order in the [17] R.P. Nielsen, J.H. La Rochelle, Catalyst for production of ethylene oxide,
hydroxyl species. The difference in rate constant for either silica or α- US3962136, 1976.
alumina did not simply scale with the hydroxyl density on these sup- [18] M.O. Ozbek, I. Onal, R.A. van Santen, Why silver is the unique catalyst for ethylene
epoxidation, J. Catal. 284 (2011) 230–235, https://doi.org/10.1016/j.jcat.2011.08.
ports, so probably the rate of isomerization does also depend on the 004.
specific properties of the surface sites. [19] G.H. Twigg, The mechanism of the catalytic oxidation of ethylene. I. Experiments
We clearly showed that the loss of selectivity with increasing con- using a flow system, Proc. R. Soc. A Math. Phys. Eng. Sci. 188 (1946) 92–104,
https://doi.org/10.1098/rspa.1946.0099.
version depends directly on the density of active surface groups, and [20] J.K. Lee, X.E. Verykios, R. Pitchai, Support participation in chemistry of ethylene
hence specific support surface area. In this paper, we kept the weight oxidation on silver catalysts, Appl. Catal. 44 (1988) 223–237, https://doi.org/10.
loading similar to the low surface area supported catalysts, but in in- 1016/S0166-9834(00)80055-X.
[21] O.O. Bernardini, E.A. Cherniak, Heterogeneous isomerization of ethylene oxide to
dustry, one would ideally be able to use a high surface area support acetaldehyde, Can. J. Chem. 51 (1972) 1371–1377.
(and hence silver loading and volume activity) without losing too much [22] C.-F. Mao, M. Albert Vannice, High surface area α-aluminas III. Oxidation of
selectivity. We showed that passivation of the silanol groups of a high ethylene, ethylene oxide, and acetaldehyde over silver dispersed on high surface
area α-alumina, Appl. Catal. A Gen. 122 (1995) 61–76, https://doi.org/10.1016/
surface area silica support, led to a selective, high surface area catalyst. 0926-860X(94)00214-2.
The selectivity to ethylene oxide for this catalyst supported on modified [23] D.A. Bulushev, E.A. Paukshtis, Y.N. Nogin, S. Bal, Transient response and infrared
silica was similar to that of an α-alumina supported catalyst with ten studies of ethylene oxide reactions on silver catalysts and supports, Appl. Catal. A

38
J.E. van den Reijen, et al. Catalysis Today 338 (2019) 31–39

Gen. 123 (1995) 301–322. experiments, J. Catal. 214 (2003) 200–212, https://doi.org/10.1016/S0021-
[24] Y. Yong, E.M. Kennedy, N.W. Cant, Oxide catalysed reactions of ethylene oxide 9517(02)00156-2.
under conditions relevant to ethylene epoxidation over supported silver, Appl. [48] P. Harriott, The oxidation of ethylene using silver on different supports, J. Catal. 21
Catal. 76 (1999) 31–48. (1971) 56–65, https://doi.org/10.1016/0021-9517(71)90120-5.
[25] M.C.N.A. Carvalho, C.A. Perez, R.A. Simão, F.B. Passos, M. Schmal, The promoting [49] P.L. Metcalf, P. Harriott, Kinetics of silver-catalyzed ethylene oxidation, Ind. Eng.
effect of cesium on structure and morphology of silver catalysts, An. Acad. Bras. Chem. Process Des. Dev. 11 (1972) 478–484, https://doi.org/10.1021/
Cienc. 76 (2004) 19–27, https://doi.org/10.1590/S0001-37652004000100003. i260044a004.
[26] A.M. Lauritzen, Ethylene Oxide Catayst and Process for Preparing the Catalyst, [50] D.L. Trimm, Thermal stability of catalyst supports, Catal. Deactiv. (1991) 29–51.
US4761394, 1988. [51] M.M. Martín-Ruiz, L.A. Pérez-Maqueda, T. Cordero, V. Balek, J. Subrt, N. Murafa,
[27] M.O. Özbek, R.A. van Santen, The mechanism of ethylene epoxidation catalysis, J. Pascual-Cosp, High surface area α-alumina preparation by using urban waste,
Catal. Lett. 143 (2013) 131–141, https://doi.org/10.1007/s10562-012-0957-3. Ceram. Int. 35 (2009) 2111–2117, https://doi.org/10.1016/j.ceramint.2008.11.
[28] J. Wang, P.D. Ellis, Cesium-induced structural change of adsorbed ethylene on ce- 011.
sium-promoted silver catalyst studied by13C solid-state nuclear magnetic re- [52] L.L. Murrell, D.C. Grenoble, J.P. DeLuca, Process for preparing ultra-stable, high
sonance, J. Am. Chem. Soc. 113 (1991) 9675–9676, https://doi.org/10.1021/ surface area alpha-alumina, US4169883, 1976.
ja00025a044. [53] Evonik, Aerosil – Fumed Silica: Technical Overview, (2015).
[29] S. Linic, M.A. Barteau, On the mechanism of Cs promotion in ethylene epoxidation [54] M. Nguefack, A.F. Popa, S. Rossignol, C. Kappenstein, Preparation of alumina
on Ag, J. Am. Chem. Soc. 126 (2004) 8086–8087, https://doi.org/10.1021/ through a sol–gel process. Synthesis, characterization, thermal evolution and model
ja048462q. of intermediate boehmite, Phys. Chem. Chem. Phys. 5 (2003) 4279–4289, https://
[30] R.M. Carter, Process for the preparation of ethylene oxide, US2294383, 1946. doi.org/10.1039/B306170A.
[31] J.E. van den Reijen, S. Kanungo, T.A.J. Welling, M. Versluijs-Helder, T.A. Nijhuis, [55] W. Smit, C.L.M. Holten, Zeta-potential and radiotracer adsorption measurements on
K.P. de Jong, P.E. de Jongh, Preparation and particle size effects of Ag/a-Al2O3 EFG α-Al2O3 single crystals in NaBr solutions, J. Colloid Interface Sci. 78 (1980)
catalysts for ethylene epoxidation, J. Catal. 356 (2017) 65–74, https://doi.org/10. 1–14, https://doi.org/10.1016/0021-9797(80)90489-0.
1016/j.jcat.2017.10.001. [56] M.E. Grillo, D.S. Coll, J. Rodríguez, Effect of the environment on the hydroxyl
[32] H.R. Dettwiler, A. Baiker, W. Richarz, Kinetics of ethylene oxidation on a supported density of α-quartz (111), Chem. Phys. Lett. 522 (2012) 46–50, https://doi.org/10.
silver catalyst, Helv. Chim. Acta 62 (1979) 1689–1700, https://doi.org/10.1002/ 1016/j.cplett.2011.11.084.
hlca.19790620602. [57] S.M. Pimblott, N.J.B. Green, Recent advances in the kinetics of radiolytic processes,
[33] Y.C. Kim, N.C. Park, J.S. Shin, S.R. Lee, Y.J. Lee, D.J. Moon, Partial oxidation of Res. Chem. Kinet. (1993) 117–119.
ethylene to ethylene oxide over nanosized Ag/α-Al2O3 catalysts, Catal. Today 87 [58] F.A. Rodrigues, P.J.M. Monteiro, G. Sposito, Alkali-silica reaction the surface charge
(2003) 153–162, https://doi.org/10.1016/j.cattod.2003.09.012. density of silica and its effect on expansive pressure, Cem. Concr. Res. 29 (1999)
[34] M.A. Al-Juaied, D. Lafarga, A. Varma, Ethylene epoxidation in a catalytic packed- 527–530, https://doi.org/10.1016/S0008-8846(98)00220-8.
bed membrane reactor: experiments and model, Chem. Eng. Sci. 56 (2001) [59] E. Valencia, A. Maldonado, Adsorption of isobutene on partially hydrophobized
395–402, https://doi.org/10.1016/S0009-2509(00)00235-9. aerosil, J. Chem. Soc. Faraday Trans. 86 (1990) 539–543.
[35] D. Lafarga, M.A. Al-juaied, C.M. Bondy, A. Varma, Ethylene epoxidation on Ag-Cs / [60] J. Warren, S. Offenberger, H. Toghiani, C.U. Pittman, T.E. Lacy, S. Kundu, Effect of
α-Al2O3 catalyst : experimental results and strategy for kinetic parameter de- temperature on the shear-thickening behavior of fumed silica suspensions, ACS
termination, Ind. Eng. Chem. Res. 39 (2000) 2148–2156, https://doi.org/10.1021/ Appl. Mater. Interfaces 7 (2015) 18650–18661, https://doi.org/10.1021/acsami.
ie990939x. 5b05094.
[36] R.E. Kenson, M. Lapkin, Kinetics and mechanism of ethylene oxidation. Reactions of [61] G.E. Berendsen, L. De Galan, A geometrical model for chemically bonded tms and
ethylene and ethylene oxide on a silver catalyst, J. Phys. Chem. 74 (1970) pds phases, J. Liq. Chromatogr. 1 (1978) 403–426, https://doi.org/10.1080/
1493–1502, https://doi.org/10.1021/j100702a017. 01483917808060008.
[37] J.K. Lee, X.E. Verykios, R. Pitchai, Support and crystallite size effects in ethylene [62] W.M.H. Sachtler, C. Backx, R.A. Van Santen, On the mechanism of ethylene epox-
oxidation catalysis, Appl. Catal. 50 (1989) 171–188, https://doi.org/10.1016/ idation, Catal. Rev. 23 (1981) 127–149, https://doi.org/10.1080/
S0166-9834(00)80834-9. 03602458108068072.
[38] Y. Hui-Xing, C. Hua, H. De-Gang, The rate constant determination for the iso- [63] P. Borman, K. Westerterp, An experimental study of the kinetics of the selective
merization of ethylene oxide to acetaldehyde by chemical shock tube, Acta Chim. oxidation of ethene over a silver on alpha-alumina catalyst, Ind. Eng. Chem. Res. 34
Sin. (1989) 941–946. (1995) 49–58, https://doi.org/10.1021/ie00040a002.
[39] J.E. van den Reijen, P.H. Keijzer, P.E. de Jongh, Pore structure stabilization during [64] M.J. Palys, S.Y. Ivanov, A.K. Ray, Conceptual approach in multi-objective optimi-
the preparation of single phase ordered macroporous α-alumina, Materialia 4 zation of packed bed membrane reactor for ethylene epoxidation using real-coded
(2018) 423–430, https://doi.org/10.1016/j.mtla.2018.10.016. non-dominating sorting genetic algorithm NSGA-II, Int. J. Chem. React. Eng. 15
[40] S.M. Pourmortazavi, S.S. Hajimirsadeghi, I. Kohsari, R. Fareghi Alamdari, (2017), https://doi.org/10.1515/ijcre-2016-0041 aop.
M. Rahimi-Nasrabadi, Determination of the optimal conditions for synthesis of [65] R.J. Baxter, P. Hu, Insight into why the Langmuir-Hinshelwood mechanism is
silver oxalate nanorods, Chem. Eng. Technol. 31 (2008) 1532–1535, https://doi. generally preferred, J. Chem. Phys. 116 (2002) 4379–4381, https://doi.org/10.
org/10.1002/ceat.200800090. 1063/1.1458938.
[41] A.R. Fiorucci, L.M. Saran, É.T.G. Cavalheiro, E.A. Neves, Thermal stability and [66] Y.S. Yong, E.M. Kennedy, N.W. Cant, Oxide catalysed reactions of ethylene oxide
bonding in the silver complexes of ethylenediaminetetraacetic acid, Thermochim. under conditions relevant to ethylene epoxidation over supported silver, Appl.
Acta 356 (2000) 71–78, https://doi.org/10.1016/S0040-6031(00)00478-0. Catal. 76 (1991) 31–48, https://doi.org/10.1016/0166-9834(91)80003-F.
[42] V.V. Boldyrev, Thermal decomposition of silver oxalate, Thermochim. Acta 388 [67] R.A. Van Santen, H.P.C.E. Kuipers, The mechanism of ethylene epoxidation, J. Adv.
(2002) 63–90, https://doi.org/10.1016/S0040-6031(02)00044-8. Catal. Sci. Technol. (1987) 265–321, https://doi.org/10.1016/S0360-0564(08)
[43] E. Plessers, J.E. van den Reijen, P.E. de Jongh, K.P. de Jong, M.B.J. Roeffaers, 60095-4.
Origin and abatement of heterogeneity at the support granule scale of silver on [68] M.O. Özbek, I. Önal, R.A. van Santen, Ethylene epoxidation catalyzed by silver
silica catalysts, ChemCatChem. 9 (Dec(24)) (2017) 4562–4569, https://doi.org/10. oxide, ChemCatChem 3 (2011) 150–153, https://doi.org/10.1002/cctc.201000249.
1002/cctc.201700753. [69] S. Rojluechai, S. Chavadej, J.W. Schwank, V. Meeyoo, Catalytic activity of ethylene
[44] J. Hoekstra, A.M. Beale, F. Soulimani, M. Versluijs-Helder, J.W. Geus, oxidation over Au, Ag and Au-Ag catalysts: support effect, Catal. Commun. 8 (2007)
L.W. Jenneskens, Shell decoration of hydrothermally obtained colloidal carbon 57–64, https://doi.org/10.1016/j.catcom.2006.05.029.
spheres with base metal nanoparticles, New J. Chem. 39 (2015) 6593–6601, [70] S.N. Goncharova, E.A. Paukshtis, B.S. Bal’zhinimaev, Size effects in ethylene oxi-
https://doi.org/10.1039/C5NJ00804B. dation on silver catalysts. Influence of support and Cs promoter, Appl. Catal. A Gen.
[45] C.T. Campbell, M.T. Paffett, Model studies of ethylene epoxidation catalyzed by the 126 (1995) 67–84, https://doi.org/10.1016/0926-860X(95)00036-4.
Ag(110) surface, Surf. Sci. 139 (1984) 396–416, https://doi.org/10.1016/0039- [71] A. Chongterdtoonskul, J.W. Schwank, S. Chavadej, Effects of oxide supports on
6028(84)90059-1. ethylene epoxidation activity over Ag-based catalysts, J. Mol. Catal. A Chem. 358
[46] C. Stegelmann, N.C. Schiødt, C.T. Campbell, P. Stoltze, Microkinetic modeling of (2012) 58–66, https://doi.org/10.1016/j.molcata.2012.02.011.
ethylene oxidation over silver, J. Catal. 221 (2004) 630–649, https://doi.org/10. [72] A. Chongterdtoonskul, T. Suttikul, M. Santikunaporn, J.W. Schwank, S. Chavadej,
1016/j.jcat.2003.10.004. Effect of diluent gas on ethylene epoxidation activity over various Ag-based cata-
[47] S. Linic, M.A. Barteau, Construction of a reaction coordinate and a microkinetic lysts on selective oxide supports, J. Mol. Catal. A Chem. 386 (2014) 5–13, https://
model for ethylene epoxidation on silver from DFT calculations and surface science doi.org/10.1016/j.molcata.2014.02.003.

39