~ APPLIED

CATALYSSI
A:GENERAL
ELSEVIER Applied Catalysis A: General 122 (1995) 61-76

High surface area a-aluminas

III. Oxidation of ethylene, ethylene oxide, and
acetaldehyde over silver dispersed on high surface
area a-alumina
C h i n g - F e n g Mao, M. A l b e r t V a n n i c e *
Department of Chemical Engineering, The Pennsylvania State University, University Park, PA 16802, USA
Received 20 July 1994; revised 22 September 1994; accepted 22 September 1994

Abstract

Silver nanocrystallites dispersed on high surface area ( H S A ) a-alumina were prepared, and the
activity and selectivity to ethylene oxide (EO) during ethylene oxidation over these catalysts were
determined in the absence of C1 promoters. Cesium was also added to some of the samples to neutralize
any surface acidity. Over all the A g / ( H S A ) a-A1203 catalysts, the selectivity to EO was negligible
regardless of the addition of cesium while the apparent activation energy for carbon dioxide formation
of I l0 kJ mol ~over the unpromoted samples was higher than that of 80 kJ mol ~over a low surface
area (LSA) a-alumina-supported silver catalyst. When this latter catalyst with selectivity to EO was
mixed with ( H S A ) a-A1203, only carbon dioxide and water were obtained as a final product. EO
isomerization and oxidation as well as acetaldehyde oxidation were also studied, and both silver and
the support were found to be active in these reactions; hence these ( H S A ) a-aluminas are poor
supports for ethylene epoxidation. In contrast to a promoting effect obtained at low concentrations in
commercial catalysts, the higher loadings of cesium on these A g / ( H S A ) a-A1203 catalysts increased
the turnover frequency of ethylene oxidation as well as the rates of EO and acetaldehyde oxidation,
while the apparent activation energy for ethylene combustion decreased to 80 kJ mol - ~in the presence
of cesium.

Kevwords: Acetaldehyde oxidation Alumina Ethylene oxidation; Ethylene oxide isomerization; Ethylene oxide
oxidation; Silver

1. Introduction

Commercial a-alumina is a unique support material for silver as it provides a

highly selective ethylene epoxidation catalyst with appropriate promoters. This is

* Corresponding author. Tel. ( + 1-814) 8634803, Fax. ( + 1-814) 8657846

0926-860X / 95 / $09.50 © 1995 Elsevier Science B.V. All rights reserved

SSDI O926-860 X ( 94 ) 0 0 2 1 4 - 2
62 C.-F. Mao, M.A. Vannice / Applied Catalysis A: General 122 (1995) 61-76

attributed to its inertness for isomerization of ethylene oxide (EO) to acetaldehyde

rather than its low surface area [ 1]. Relatively high silver loadings on low surface
area (LSA) a-aluminas ( < 1 m 2 g 1) are typically used in commercial ethylene
oxidation catalysts, in which silver is poorly dispersed. Kanoh et al. [2] have
demonstrated that silver supported on alumina with surface areas less than l0 m 2
g-1 ( a phase) exhibited selectivity to ethylene oxide while silver dispersed on
aluminas with surface areas greater than 10 m 2 g - 1 (yphase) showed no selectivity.
An interesting question is whether a high surface area form of a-alumina would
also be relatively inert to secondary reactions and also provide stable, highly dis-
persed silver. Recently, high surface area (HSA) a-aluminas (ca. 100 m 2 g i)
have been developed by Alcoa which can provide highly dispersed silver, i.e., more
efficient silver utilization, and the adsorption of O2, C2H4, CO2, and CO on these
catalysts has been investigated [ 3 ]. In the present study, ethylene oxidation activity
and selectivity to EO over these catalysts are evaluated, and EO oxidation and
isomerization as well as acetaldehyde oxidation are also examined.
The influence of the support on the selectivity to EO has been known for many
years, and Harriott [4] reported that silver supported on alumina derived from
boehmite was a non-selective catalyst. Kanoh et al. [2] attributed the loss in
selectivity to a secondary oxidation step involving EO isomerization to acetalde-
hyde on the alumina and the subsequent combustion of acetaldehyde. Recently,
silver catalysts involving a variety of support materials have been investigated, and
the selectivity to EO has been associated with the activity of EO isomerization and
oxidation over the supports; among the materials studied, only a-alumina showed
no activity [ 1,5 ] ; thus EO isomerization and oxidation can serve as a probe reaction
to test if the support is active in the secondary oxidation step.
This paper examines whether silver supported on (HSA) a-alumina is a selective
catalyst for ethylene oxidation. Since these (HSA) a-aluminas exhibit a certain
level of acidity as revealed by carbon monoxide adsorption [ 6] and by the poten-
tiometric titration method [ 7], cesium was added to one of the silver samples in
an effort to neutralize surface acidity. The effect of cesium on the silver samples
was elucidated by examining the change in oxygen chemisorption and hydrogen
titration on these catalysts. Ethylene oxidation over these catalysts as well as (LSA)
a-AlzO3-supported silver was conducted, and activity and selectivity to EO were
compared. In addition, EO isomerization and oxidation over chosen catalysts and
the pure supports were performed to evaluate the inertness of the supports, and
acetaldehyde oxidation was also examined to gain information about the secondary
oxidation step.

2. Experimental
2.1. Catalyst preparation
Two different (HSA) a-A1203's were used as supports, denoted as a-AIzO3(U)
(Alcoa, 78 m2/g) and a-AI203(G) (Alcoa, 104 m2/g), which were prepared from
 C.-F. Mao, M.A. Vannice /Applied Catalysis A: General 122 (1995) 61-76 63

Table 1
Oxygen adsorption and hydrogen titration on silver catalysts at 443 K

Catalyst a Gas uptake (/xmol/g catalyst) Fraction of exposed Ag atoms

02 H2 titr. (O~d/Ag) (H2/Ag)

A-0.83% Ag/o~-A1203(U ) 16.5 33.0 0.43 0.43

A-0.83% Ag/o~-AI203(U) b 13.7 23.3 0.36 0.31
B-2.2% Cs/{0.83% Ag/ot-AI203(U) } 13.0 31.6 0.34 0.41
B-2.2% Cs/{0.83% Ag/a-AI203(U) }b 7.0 l 1.0 0.18 0.14
C-4.4% Cs/{0.83% Ag/ot-A1203(U) }b 6.4 13.7 0.17 0.18
D-0.68% Ag/{2.2% Cs/~-AI203(U) }b 4.9 7.7 0.[6 0.13
E-0.71% Ag/ot-AI203(G) 18.4 34.7 0.56 0.53
F- 12.7% Ag/ot-AIzO3(G) 61.3 149 0.10 0.12
G-16.5% Ag/c~-AI203(N) 1.1 2.0 0.0014 0.0013

Initial impregnation step within brackets.

b Pretreatment B.

unground and ground diaspore, respectively [6]. A (LSA) o~-A1203 (Norton SA-
5202, 0.92 mZ/g) was also used for comparison and is referred to as a-AI203(N).
All the supports were heated at 823 K in flowing air (Linde, dry grade) at 500 cm 3
STP/min for 2 h before impregnation with silver. The silver catalysts were prepared
by incipient wetness impregnation of these aluminas with AgNO3 (Aldrich,
99.9999%) dissolved in distilled deionized water followed by drying overnight at
383 K in an oven and storage in a desiccator. Catalysts with silver loadings below
1% and above 10% were studied, as shown in Table 1; the silver loadings were
determined by plasma emission spectroscopy at the Mineral Constitution Labora-
tory of the Pennsylvania State University. Since it has been found that these (HSA)
a-Al203's possess a certain level of surface acidity [6,7], some samples were also
impregnated with cesium to neutralize surface acid sites. The cesium-promoted
samples were also prepared by incipient wetness impregnation of cesium either
before or after the silver impregnation step using Cs20 (Aldrich, metallic impurity
< 1%) in distilled deionized water. The impregnation sequence of silver and cesium
in Table 1 is denoted by placing the first step within the brackets. The amount of
cesium was chosen by considering its capability to neutralize the surface acid sites.
On the (HSA) ot-AlzO3(U) sample, 2.2% and 4.4% Cs correspond to surface
coverages of 8.5% and 17%, respectively, based on a surface site density of
1.5 × 1 0 1 9 m 2 for a-alumina.

2.2. Catalyst characterization

Silver dispersion (in cesium-free catalysts) as well as the concentration of

surface silver atoms for ethylene oxidation were measured by both oxygen chem-
isorption and hydrogen titration at 443 K in a high-vacuum system which has been
described in previous papers [3,8]. Two different pretreatments were employed
prior to chemisorption measurements: The first pretreatment included calcination
64 C.-F. Mao, M.A. Vannice /Applied Catalysis A: General 122 (1995) 61-76

in 10% 02/90% He (40 cm 3 STP/min) at 773 K for 2 h followed by reduction in

hydrogen (20 cm 3 STP/min) at 673 K for 2.5 h, while the second pretreatment
consisted of reduction in hydrogen (20 c m 3 STP/min) at 573 K for 2.5 h followed
by an O2/H2 cycle, i.e., exposure to oxygen ( 10 c m 3 STP/min) for 10 min then
reduction in hydrogen (20 cm 3 STP/min) for 15 min at 443 K. The former is
referred to as Pretreatment A and the latter, milder procedure as Pretreatment B.

2.3. Catalytic measurements"

Reactions of ethylene oxidation, ethylene oxide isomerization and oxidation, and

acetaldehyde oxidation were conducted at atmospheric pressure in a Pyrex reactor,
with an internal flitted disc to support the catalyst powder, which was placed in a
heating mantle equipped with a temperature controller. After characterization by
chemisorption, the catalyst was passivated in oxygen prior to its exposure to air
and then loaded into the reactor (ca. 0.3 g). Pretreatments identical to those used
in the chemisorption study were performed before the reactants were introduced.
The gases used: oxygen (MG Ind., 99.99%) and helium (MG Ind., 99.999%) were
purified by passing through molecular sieve traps and helium was also passed
through an Oxytrap, ethylene (MG Ind., 99.5%) was used without further purifi-
cation while ethylene oxide (MG Ind., 99.7%) was passed through an Ascarite trap
to remove carbon dioxide. The flow-rates of 02, He and C 2 H 4 w e r e regulated by
mass flow controllers (Tylan, model FC-260) while that of EO was controlled by
a rotameter. Acetaldehyde (Aldrich, 99%) was stored in its liquid form and fed
into the reactor by passing a mixture of helium and oxygen through an acetaldehyde
saturator held at 260 K to obtain a saturation vapor pressure of 170 Torr. The
acetaldehyde in the saturator was replaced daily to avoid accumulation of acetals
and polymeric compounds [5]. The total feed flow-rate of 30 cm 3 STP/min was
composed of 9 cm 3 STP/min C2H4, 2.4 cm 3 STP/min 02, and 18.6 cm 3 STP/min
He for ethylene oxidation, while in the other two reactions ethylene was replaced
by 4 cm 3 STP/min EO or 6.9 cm 3 STP/min CH3CHO with the oxygen flow-rate
remaining the same and helium constituting the rest of the feed. This provided a
typical space velocity of 2000 h - J. In the Arrhenius runs an ascending temperature
sequence from 483 to 503 K was followed by a descending temperature sequence
and a second ascending temperature sequence to check for deactivation. The cal-
culation of activation energy was typically based on the descending temperature
sequence and the second ascending temperature sequence. The activity measure-
ments were made during a period of 2 h on stream after each change in temperature
for ethylene oxidation, while a shorter (40 min) period was used in the study of
EO and acetaldehyde oxidation. The conversions were routinely below 10% and
the system was treated as a differential reactor. The reactor effluent was analyzed
by a gas chromatograph (Perkin-Elmer, Sigma 3) equipped with TCD and Porapak
Q columns.
 C.-F. Mao, M.A. Vannice / Applied Catalysis A: General 122 (1995)61-76 65

3. Results

3.1. Oxygen chemisorption and hydrogen titration

Table 1 lists oxygen chemisorption and hydrogen titration uptakes at 443 K on

these silver catalysts. The H t i t r / O a titration ratio varied over a range of 1.6 to 2.4,
thus indicating that the stoichiometric reaction between adsorbed oxygen species
and hydrogen to form adsorbed water, which gives a titration ratio of two, was
frequently not completely descriptive of the chemistry that occurred. This variation
is due not only to the presence of cesium but also to the milder pretreatment used
in some of the samples. The fraction of exposed silver atoms (or apparent disper-
sion) may therefore serve as an indicator of the overall effect of different pretreat-
ments, the usage of cesium, the sequence of impregnation, and different supports
on available silver surface area. For the unpromoted 0.83% A g / ( H S A ) a-
A1203 (U) catalyst, the sample after Pretreatment B had a lower apparent dispersion
than that after Pretreatment A, which is attributed to the lower reduction temperature
[ 9] as well as the lack of a high-temperature pretreatment step during which the
silver nitrate can decompose (decomposition temperature 737 K). Similarly the
presence of cesium (Sample B) also led to a lower apparent dispersion compared
to the sample without cesium under the same pretreatment, which is probably due
to the blockage of adsorption sites by cesium. With both the presence of cesium
and the usage of Pretreatment B, apparent dispersions were even lower - - only
40% of that of the unpromoted sample under Pretreatment A; however, a higher
cesium concentration (Sample C) or a different impregnation sequence (Sample
D) gave similar apparent dispersions with the same pretreatment (B). A high
apparent dispersion was also obtained with the 0.71% A g / ( H S A ) ce-A1203(G)
catalyst (Sample E) whereas with a higher silver loading ( Sample F) the apparent
dispersion was significantly reduced. The apparent dispersion of the (LSA) a-
A1203-supported silver catalyst (Sample G) was the lowest among all the samples
and it was two orders of magnitude less than any of the (HSA) a-A1203-supported
silver catalysts. In summary, (HSA) ce-A1203's are better than (LSA) or-A1203 for
dispersing silver, as anticipated, and the fraction of exposed metallic silver surface
atoms decreases considerably if either cesium is added or a milder pretreatment is
used.

3.2. Ethylene oxidation

The Arrhenius plots for ethylene epoxidation to EO and complete combustion

to carbon dioxide over three samples are shown in Fig. 1. Some deactivation was
typically observed during the first ascending temperature sequence, thus the appar-
ent activation energy (Ea) was calculated based on the activity measured over the
descending temperature and the second ascending temperature sequence, during
which period the catalysts were more stable. Results for the ethylene oxidation
66 C.-F. Mao, M.A. Vannice / Applied Catalysis A: General 122 (1995) 61-76

4.Oxl o "7

1.0x10"'
"5
E

x
<

2.0x 10"8 I i
1.96 2.00 2.04 2.08
I O 0 0 / T (K -~)
Fig. 1. Arrhenius plots for ethylene oxidation: carbon dioxide formation: 16.5% A g / a - A 1 2 0 3 ( N ) ( © ) , 0,83%
A g / a - A I 2 0 3 ( U J (El), 2.2% Cs/{0.83% A g / a - A I 2 0 3 ( U ) } ( A ) ; EO formation: 16.5% A g / a - A I z O 3 ( N ) ( 0 ) .

reactions are summarized in Table 2, where selectivity (S) to EO is defined as the

percentage of reacted ethylene converted to EO, and the turnover frequency (TOF)
is based on the corresponding oxygen uptake to count surface silver atoms. Blank
experiments were conducted with all the supports and no activity was observed.
All the (HSA) a-A1203-supported silver catalysts showed no selectivity to EO
regardless of the presence of cesium, while the (LSA) a-Al203-supported silver
catalyst had a selectivity to EO of 60%, which is close to that reported in the
literature [ 10,11 ]. Again, no Cl-containing promoters were used. Thus, unlike
(LSA) a-Al203-supported silver, the (HSA) ~-A1203's are poor supports for the
selective ethylene epoxidation reaction, and only complete oxidation was obtained;
however, quite high oxidation activities could be attained compared to the high
Table 2
Catalytic behavior of silver catalysts for ethylene oxidation at 503 K (reaction conditions: Po2 = 61 Torr, Pc.~m= 228
Torr )

Catalyst C2H4 S~ (%) Activity TOF Ea

cony. ( m o l / g s × 1 0 v) (s -I ×103 ) (kJ/mol)
(%)
EO CO2 EO CO2 EO CO2

A-0.83% Ag/a-AlzO 3(U) 0.73 0 0 3.0 0 9.1 112

A-0.83% Ag/ot-AI203(U) ~ 0.58 0 0 2.3 0 8.6 116
E-0.71% Ag/ot-AI203(G) 4.7 0 0 17 0 47 103
F-12.7% Ag/ot-Al203(G) 0.77 0 0 29 0 24 107
B-2.2% Cs/{0.83% Ag/a-AI203(U) } 0.63 0 0 2.5 0 9.8 75
B-2.2% Cs/{0.83% Ag/a-AI203(U) }, 0.77 0 0 3.2 0 23 73
C-4.4% Cs/{0.83% Ag/a-AI203(U) }b 0.83 0 0 3.4 0 26 76
D-0.68% Ag/{2.2% Cs/ot-AI203(U) }~ 1.2 0 0 5.1 0 52 90
G-16.5% Ag/ct-AlzO3(N) 0.58 60 0.70 0.94 33 45 71 80
H-16.5% Ag/a-AI203(N) + o~-A1203(U) c 0.31 0 0 0.69 0 65 81

a S = mol EO formed/mol C2H4 reacted,

Pretreatment B.
0.5 g catalyst, physical mixture with 50% a-AI203(U).
 C.-F. Mao, M.A. Vannice /Applied Catalysis A: General 122 (1995) 61-76 67

Table 3
Catalytic behavior of silver catalysts for ethylene oxide isomerization and oxidation at 503 K (pretreatment A,
reaction conditions: Po., = 61 Torr, PEO = 99 Torr, 0.2 g catalyst)

Catalyst r(mol/gs×107) TOF(s IXl03)~ E.(kJ/mol)

CH3CHO CO2 CH3CHO CO2 CH3CHO CO,

a-AI203(U) 3.6 0.35 - - 71 68

A-0.83% Ag/a-AI203(U ) 4.2 0.66 1.8 0.94 82 89
B-2.2% Cs/{0.83% Ag/ot-AI203(U) } 1.5 1.3 - - 60 72

TOF for EO reaction on Ag, corrected for contribution from ,~1203.

loading silver catalysts. This behavior is similar to that found with silver dispersed
on transition alumina and can be attributed to the presence of active sites on the
support surface. In the absence of cesium, the (HSA) a-A1203-supported silver
sample had an apparent activation energy (Ea) around 110 + 6 kJ mol- 1for carbon
dioxide formation and the corresponding turnover frequencies (TOFs) for the a-
A1203(G)-supported silver samples were higher than those for the a-A1203(U)-
supported silver samples. When cesium was added to the (HSA) a-A1203(U)
catalysts, EO was still not observed although activity for the combustion reaction
tended to increase, the TOF clearly was enhanced, and the apparent activation
energy dropped noticeably to around 80 kJ m o l - l. An increase in cesium loading
from 2.2% to 4.4% (Samples B and C) had no significant effect on the reaction;
however, Pretreatment A gave a lower TOF more similar to that of Sample A (no
cesium added), and impregnating with cesium first (Sample D) appeared to
increase activity and the activation energy. With a physical mixture (H) containing
50% of the selective silver catalyst (Sample G) and 50% pure c~-A1203(U), the
production of EO was completely suppressed and the combustion reaction was
enhanced; consequently, this clearly demonstrates that (HSA) ot-A1203 is very
active for the combustion of EO.

3.3. Ethylene oxide and acetaldehyde oxidation

Pretreatment A was used prior to any rate measurements of EO isomerization

and oxidation as well as acetaldehyde oxidation. In the study of EO oxidation, pure
ce-A1203(U) along with two selected silver catalysts were examined to study both
isomerization of EO to acetaldehyde and complete oxidation to carbon dioxide. A
complete analysis of the reaction products which would include any polymerized
products such as those reported by Yong et al. [5 ] was not conducted since the
primary purpose was to determine rates of isomerization versus rates of combustion
to carbon dioxide; however, a small amount of an unknown brown liquid deposited
on the reactor wall during the reaction may have represented EO polymerization
products. The results for EO isomerization and oxidation are summarized in Table
3 and the Arrhenius plots for EO isomerization and combustion are shown in Figs.
68 C.-F. Mao, M.A. Vannice / Applied Catalysis A: General 122 (1995) 61-76

1.0xl 0 -6

O
E

o
"~ l.Oxl 0-7
,<

5.0xi0-8 I i
1.96 2.00 2.04 2.08
I 0 0 0 / T (K -I)
Fig. 2. Arrhenius plots for ethylene oxide isomerization: a-AI20~(U), (C)); 0.83% Ag/a-AI~O3(U), ([]); 2.2%
Cs/{0.83% Ag/ot-Al203(U) }, (A).

2 and 3, respectively. On pure ce-Al203(U) the rate of isomerization was ten times
higher than that of oxidation, thus indicating that isomerization is the dominant
reaction over the support. The presence of silver on this alumina (Sample A)
doubled the rate of oxidation, while the rate of isomerization increased by a smaller
amount. The TOF for the EO reaction on silver can be estimated based on the
difference between the rate of EO reaction on the silver catalyst and the support.
In contrast, the addition of cesium to this Ag/c~-A1203 sample (Sample B) signif-
icantly decreased the rate of isomerization while it enhanced the oxidation rate.
The former can be attributed to the neutralization of surface acid sites by cesium
atoms while the latter may be explained by the high activity of acetaldehyde
oxidation on the cesium-doped sample, which will be shown later.
During the oxidation of acetaldehyde, carbon dioxide, along with methanol as a
minor product, was monitored. A blank experiment with an empty reactor dem-
onstrated that significant activity due to the homogeneous gas-phase reaction
occurred in the temperature range studied (above 483 K). The Pyrex reactor had
5.0xl 0 -7

u 1.0x10"7 [] /x

0
£
X
<

1.0xl 0 "8 I I
1.96 2.00 2.04 2.08

1 0 0 0 / T (K "1)
Fig. 3. Arrhenius plots for ethylene oxide complete oxidation: ot-A1203( U ), (C) ) ; 0.83 % Ag / a-Al203 ( U ), ( [] ) ;
2.2% Cs/{0.83% Ag/ot-Al203(U) }, ( ~ ) .
 C.-F. Mao, M.A. Vannice /Applied Catalysis A." General 122 (1995) 61-76 69

Table 4
Catalytic behavior of silver catalysts for acetaldehyde oxidation at 503 K (pretreatment A, reaction conditions:
Po.~=61 Torr, Pc,~cHo = 170 Torr, 0.3 g catalyst)

Catalyst (mass) r (mol/cm 3 s X 108) r (mol/g s X 107) a Ea (kJ/mol)

CH3OH COz CO2 CO_,

Gas-phase reaction (none) 0.24 1.8 - 108

a-A1203(U ) (0.3 g) 0.13 2.5 1.2 -
A-0.83% Ag/a-AI203(U) (0.3 g) 0.48 5.6 6.3 105
B-2.2% Cs/{0.83% Ag/ot-AlzO3(U) } (0.01 g) 0.89 6.7 260
E-0.71% Ag/ot-Al203(G) (0.1 g) 0.44 6.0 26 89

a Corrected for gas-phase reaction.

been cleaned by aqua regia to eliminate any contamination by metals. Gas-phase

acetaldehyde oxidation has been investigated in a wide temperature range in the
past [12-14]. It has been demonstrated that acetaldehyde is oxidized to give
peracids as the major product below 423 K [ 12], and at high temperatures low
molecular weight products, like carbon dioxide and methanol, are formed [ 13,14].
Due to the contribution from this gas-phase reaction, the catalytic activity for
acetaldehyde oxidation cannot be obtained easily and the total measured activity is
listed. For carbon dioxide formation, the contribution from the catalyst can be
estimated by subtracting the activity of the gas-phase reaction from the total activity
to get activity per unit catalyst weight. For methanol, the total activity was some-
times less than that of the gas-phase reaction in the empty reactor and a similar
calculation cannot be made. The reaction rates in terms of reactor volume as well
as the corrected reaction rates are listed in Table 4, and the Arrhenius plot given in
Fig. 4 is based on the corrected values. Over pure ce-A1203(U) the formation of
methanol was suppressed compared to the gas-phase reaction while complete oxi-
dation to carbon dioxide also took place on the alumina. Over the silver catalysts
(Samples A and E) carbon dioxide along with methanol formation significantly
5.OxtO -6

G"

~ 1 . 0 x 10 -s

<

1.0x10-7 I I
1.96 2.00 2.04 2.08
1 0 0 0 / T (K 1)
Fig. 4. Arrhenius plots for acetaldehyde complete oxidation: 0.83% Ag/a-AI203(U), ( O ) ; 0.71% A g / a -
AI203(G), (F]).
70 C.-F. Mao, M.A. Vannice / Applied Catalysis A: General 122 (1995) 61 76

increased. The addition of cesium to a silver catalyst (Sample B) increased the rate
of carbon dioxide formation more than 40-fold compared to that of Sample A, thus
indicating that cesium strongly enhanced the complete oxidation of acetaldehyde.
This behavior can explain the high activity for EO oxidation over the cesium-doped
sample.

4. Discussion

Oxygen chemisorption shows that high dispersions of silver can be achieved on

these (HSA) a-aluminas, a result similar to that with silver on transition aluminas
[8,15]. The addition of cesium produced a decrease in oxygen and hydrogen
uptakes and a variation in Htitr/Oad ratios, regardless of pretreatment, as shown in
Table 1. Since the oxidation state of cesium and the morphology of cesium particles
after a pretreatment is not known, the effect of cesium on adsorption behavior is
not fully understood; however, it is reasonable to assume that cesium blocks oxygen
adsorption sites on silver, especially when using a lower-temperature pretreatment
(Pretreatment B) which should increase the chance that cesium stays on the silver
surfaces without sintering or migrating to the support. With supported silver, cesium
has been reported to exist in the form of small crystallites in the fresh catalyst while
under reaction conditions it spreads over the surfac~ of the catalyst [ 16]. The
structure and oxidation state of cesium deposited on silver single crystals has been
studied under ultra high vacuum (UHV) conditions [ 17-19]. Campbell [ 17]
reported that Cs-O interactions stabilize cesium adsorbed on Ag( 111 ), and small
doses of oxygen give a surface oxide with a stoichiometry near CsO3 which coa-
lesces into islands under reaction conditions. Grant and Lambert [ 18 ] also reported
that in the presence of oxygen a redistribution of cesium occurs on the Ag( 111 )
surface when heated to 500 K, and the growth of cesium oxide crystallites was
proposed. They suggested that a surface alloy of cesium and silver is produced
during the thermal decomposition of this cesium oxide. Finally, under reaction
conditions the formation of a stable surface Cs2CO3 has been proposed to account
for its promoting effect [ 19]. Our chemisorption study verifies that the addition of
cesium results in a decrease in oxygen adsorption sites on silver although the cesium
undoubtedly is distributed on both the silver and the alumina. Unfortunately, dif-
ferences in the reactivity of oxygen associated with the cesium and that chemisorbed
on the silver do not appear to be clearly distinguishable in the hydrogen titration
results because no reproducible trend in the titration ratio was observed.
The ethylene epoxidation data show that these (HSA) a-alumina-supported
silver catalysts are non-selective catalysts while the (LSA) a-alumina-supported
silver catalyst provided selectivity to EO consistent with that in the literature.
However, the mixture of this latter catalyst and pure (HSA) a-alumina also showed
no selectivity to EO, thus clearly indicating that EO oxidizes readily over these
(HSA) a-aluminas. This was later confirmed in the investigation of EO isomeri-
 C. -F. Mao, M.A. Vannice /Applied Catalysis A: General 122 (1995) 61-76 71

zation and oxidation over a pure (HSA) a-alumina. Thus it is apparent that these
(HSA) a-aluminas are poor support materials for silver epoxidation catalysts.
Mesopore volume is another factor which can affect the performance of supported
silver for ethylene oxidation. The (HSA) a-aluminas have mesopore volumes
around 0.01 and 0.07 c m 3 g - 1, respectively [ 6 ]. Kanoh et al. [ 2 ] have demonstrated
that silver supported on aluminas with significant mesopore volume gave no selec-
tivity to EO. This effect is also implied by the higher rate of carbon dioxide
formation with the 0.71% Ag/a-A1203(G), which has a larger mesopore volume
compared to a-Al203 (U).
Conner et al. [20] have stated that selectivity to EO can be obtained with Ag/
A1203 catalysts if ethylene does not adsorb on the alumina to any appreciable
extent. In a previous paper, it was shown that ethylene adsorbs only weakly on
these (HSA) a-aluminas and coverages were low [6]; consequently, our initial
expectation was that these (HSA) a-aluminas should allow selectivity to EO.
However, the absence of EO as a product was mainly due to the secondary reaction
of EO oxidation, and the primary oxidation of ethylene directly to carbon dioxide,
as suggested by Conner et al. [ 20], was not the only cause for poor selectivity. The
secondary reaction including isomerization of EO to acetaldehyde on the support
and its subsequent oxidation on silver has been proposed as the principal reason
for poor selectivity [ 1,2,5]. Lee et al. [ 1 ] and Yong et al. [5] tested a variety of
supports and only those with poor activity for EO isomerization resulted in good
selectivity to EO. In addition, the activity for EO isomerization and oxidation was
found to be proportional to the surface acidity of these supports. These (HSA) a-
aluminas do possess a certain amount of surface acidity [6,7], thus their perform-
ance is inconsistent with the observation of Lee et al. [ 1]. Although a (HSA)
a-alumina prepared by another group was recently shown to be inactive for l-
butene isomerization under conditions where a y-alumina was very active [21],
this difference is probably due to the requirement of stronger acid sites for l-butene
isomerization compared to that for EO isomerization.
The use of cesium in commercial silver catalysts to promote selectivity to EO is
well known [ 16,22 ] ; however, all the cesium-impregnated Ag / (HSA) a-alumina
catalysts examined here gave no selectivity to EO although the rate of carbon
dioxide formation increased. This discrepancy is attributed to the high levels of
cesium employed. In a commercial low surface area catalyst, silver constitutes a
large portion of the total surface and only a small amount of cesium is required. In
contrast, with the (HSA) a-alumina supported sample, silver composes only 1%
of the total surface and a large amount of cesium was used in an effort to neutralize
acid sites on the support. As mentioned, the effect of cesium on ethylene oxidation
over silver has been extensively examined on silver single crystals [ 17-19,23-25 ].
On Ag( 111 ), Grant and Lambert [ 18] have reported that both the rate of EO
formation and the rate of carbon dioxide formation exhibited a maximum at a
cesium coverage of 0.25; however, in the presence of cesium and carbon dioxide a
decline in both rates was observed, which was attributed to the formation of stable
72 C.-F. Mao, M.A. Vannice /Applied Catalysis A: General 122 (1995) 61-76

cesium carbonate species [ 19,23]. The increase in ethylene combustion at low

cesium coverages was attributed to cesium acting as an electron donor to favor the
nucleophilic attack of adsorbed oxygen on the hydrogen in the ethylene molecule
[23] while the increase in EO formation was associated with the suppression of
the secondary EO isomerization and oxidation steps by cesium [23-25]. The
increase in selectivity to EO then results from cesium increasing the rate of EO
formation more rapidly than that of carbon dioxide formation. In contrast, Campbell
[ 17] reported that cesium decreases the activity for both EO and carbon dioxide
formation over Ag( 111 ), and this difference in behavior has been attributed to
lower reactant pressures in the former case. The cesium coverages on the silver
surfaces in our cesium-promoted catalysts can be estimated based on the decrease
in oxygen uptake at 443 K. For example, Sample B had a decrease in oxygen uptake
of 21% and 49% compared to Sample A after pretreatment A and B, respectively,
which may correspond to the cesium coverage on these silver surfaces; similarly,
Sample C was estimated to have a cesium coverage of 0.53. For Sample B at a
cesium coverage of 0.21, the rate of carbon dioxide formation was slightly lower
than the unpromoted catalyst while the catalysts with cesium coverages of 0.49 and
0.53 had a rate of carbon dioxide formation higher than the unpromoted catalyst.
Furthermore, on Sample A the addition of cesium increased the rate of EO oxidation.
These results contrast with the silver single crystal studies, and the differences can
be attributed to the presence of c~-alumina, which is also involved in these reactions
and can be altered by cesium in a different manner.
Since silver constitutes only a small fraction of the total surface area in these
catalysts, the effect of cesium on the support deserves more attention. In the absence
of cesium, the (HSA) a-alumina is inactive for ethylene oxidation but is active for
ethylene oxide isomerization and oxidation. In a commercial catalyst, one role of
cesium added at ppm levels is to neutralize acid sites on the support which would
otherwise catalyze EO isomerization to acetaldehyde [ 16]. In this study, because
of the relatively high concentration of acid sites due to the high surface area of
these (HSA) a-aluminas, a large amount of cesium was added to eliminate surface
acidity; however, this did not improve selectivity to EO in ethylene oxidation and
the cesium-promoted Ag/A1203 catalysts were still active for EO isomerization
and oxidation. This result is similar to that reported by Lee et al. [1 ] for Na-
impregnated Ag/alumina (boehmite) catalysts and they have suggested that alkali
impregnation of supported silver catalysts is not a proper method to eliminate
surface acidity. It is also possible that the large amount of cesium may form a
cesium oxide phase which is detrimental to EO formation as it allows EO isomer-
ization and oxidation. In the past, it has been shown that non-acidic metal oxides
supports, e.g., MgO, can also produce non-selective catalysts [ 1 ].
Table 3 shows that the (HSA) c~-alumina was active for both EO isomerization
and oxidation but the rate of EO isomerization was ten times higher. Both rates
increased with silver on this alumina, and the further addition of cesium decreased
the rate of EO isomerization sharply but increased that of EO oxidation, which
 C.-F. Mao, M.A. Vannice /Applied Catalysis A: General 122 (1995) 61-76 73

Table 5
Comparison o f rates o f ethylene oxide isomerization/oxidation and acetaldehyde oxidation

Catalyst Reaction conditions Rate ( m o l / m 2 s × 109) a Ref.

E O (Ton') CH3CHO 02 ( T o n ' ) CO2 Temp. lsomerization to Oxidation

(Torr) (Ton.) (K) CH3CHO toCO2

o~-A1203 (0.5 m 2 / g ) 200 - 18 000 - 513 0.0 0.0 [ I]

AI203 ( b o e h m i t e ) 200 - 18 000 - 513 2.1 2.7 [I]
SiO2 200 - 18 000 - 513 0.10 0.11 [ I]
SiO2 18 - 144 30 512 1.5 0.71 [5]
10.8%Ag/SiO2 18 - 114 30 512 <0.01 1.5 [51
SiO 2 - 15 100 - 512 0.44 [51
cr-Al203(U) 99 - 61 - 503 4.6 0.45 t,
0.83% Ag/oL- 99 - 61 - 503 5.4 0.85
AIzO3(U)
ot-AIO3(U) - 170 61 - 503 1.5 t,
0.83% A g / oz- - 170 61 - 503 8.1 t,
AI203(U)

Based on the surface area o f supports.

b This work.

resulted in a 50% decrease in overall EO conversion. This result indicates that EO

isomerization and oxidation can occur on both the silver and the support, and cesium
has opposite effects on EO isomerization and combustion. Van Santen and Kuipers
[26] have reviewed studies showing that EO can isomerize to acetaldehyde which
may be further oxidized to give combustion products over silver surfaces, and
studies of EO on Ag( 111 ) confirm that silver itself is active for both EO isomeri-
zation and oxidation [ 24,25 ] ; however, dosing with cesium inhibited both reactions
[ 24 ]. Consequently, the observed enhancement for complete oxidation is attributed
to the (HSA) c~-alumina support and the presence of cesium on its surface, because
oxidation of any acetaldehyde formed is markedly increased, as shown in Table 4.
Similar behavior was observed by Lee et al. [ l ] in that the addition of sodium to
alumina strongly suppressed the rate of EO isomerization but increased the rate of
EO combustion, and they also found that calcium enhanced the rate of EO oxidation.
Yong et al. [ 5 ] found that the addition of cesium to silica reduced the rate of EO
isomerization but the rate of EO oxidation did not change. The reason for the
enhancement in EO oxidation by cesium in our catalysts is ascribed to the extremely
high activity for acetaldehyde oxidation produced by the addition of cesium. Thus,
it is concluded that cesium has a positive effect on the consecutive oxidation steps
of ethylene oxidation over these (HSA) a-alumina-supported silver catalysts;
however, its effect on the direct oxidation of ethylene cannot be addressed in this
study.
EO isomerization and oxidation has been studied over several oxide supports
[ 1,5 ], but few kinetic studies for supported silver have been reported. Table 5 lists
some of the literature data for comparison. Both alumina (boehmite) and silica can
catalyze EO isomerization and oxidation, and the rates of EO isomerization are
74 C.-F. Mao, M.A. Vannice /Applied Catalysis A: General 122 (1995) 61-76

close to the rates of EO oxidation, which is probably due to the relatively high
partial pressure of oxygen employed. A (LSA) c~-alumina had no detectable activity
for either isomerization or oxidation. The only study of supported silver shows an
almost zero rate of EO isomerization, or at least a very low selectivity to acetal-
dehyde compared to carbon dioxide, which contrasts with our result. This difference
may be due to the high silver loading and the different reaction conditions. Regard-
ing the apparent activation energy, the value of 82 kJ mol- ~ for EO isomerization
over silver on (HSA) a-alumina (Sample A) is higher than that of 29 kJ mol
on Ag ( 111 ) [ 25 ] and 41 kJ mol ~on a (LSA) c~-alumina-supported silver catalyst
in a static system [27], but similar to values of 75-84 kJ mol- 1 for other (LSA)
tr-alumina-supported silver catalysts [2]. The lower values may represent EO
isomerization over silver surfaces while the higher values reflect a contribution
from the support, particularly for Sample A. The value of 89 kJ mol-J for EO
oxidation over Sample A is somewhat higher than those previously reported for
(LSA) c~-alumina-supported silver catalysts (40-73 kJ mol 1) [ 2]. The difference
may well be due to the use of a pulse reactor and the high conversion on these
(LSA) o~-alumina-supported silver catalysts.
It is shown in Table 4 that acetaldehyde oxidation can proceed by a homogeneous
gas-phase reaction at these temperatures, a situation not previously discussed as a
route for EO oxidation. For the gas-phase reaction, it has been reported that at low
temperatures the overall activation energy for acetaldehyde oxidation is about 63
kJ tool ~ (peracid formation) [ 12], and that at high temperatures it is estimated
to be 94 kJ mol J based on the maximum rate of pressure increase [ 13 ]. Our value
of 108 kJ mol ~ for carbon dioxide formation is near the activation energy for
overall conversion at high temperatures. Table 4 also shows that both the support
and silver are active for acetaldehyde oxidation with the activity appearing to be
greater on the silver than on the support. In contrast, (LSA) o~-alumina is inactive
for acetaldehyde oxidation [ 5]. When cesium was added, the rate of acetaldehyde
oxidation was strongly enhanced, in contrast to the Cs/SiO2 system where cesium
inhibited acetaldehyde oxidation [5]. The promoting effect of cesium on acetal-
dehyde oxidation over our silver catalyst can explain the enhancement in the rate
of EO oxidation after the addition of cesium. Only one previous study has reported
data for acetaldehyde oxidation, and as shown in Table 5, pure silica has a specific
activity for acetaldehyde oxidation lower than that of (HSA) o~-alumina. The
apparent activation energy for acetaldehyde oxidation over silver supported on a
variety of aluminas has varied over a wide range of 61 to 136 kJ mol- 1 while that
on unsupported silver was 74 kJ mol t [ 2], thus our values of 89 and 105 kJ mol l
are consistent with these values.
The reaction route for complete ethylene oxidation over these silver catalysts
can be elucidated by comparing the kinetic data for ethylene oxidation, EO isom-
erization/oxidation and acetaldehyde oxidation. In general, the combustion of
ethylene to carbon dioxide on silver catalysts can happen via two routes - - direct
oxidation of ethylene (primary oxidation) and ethylene partial oxidation to EO
 C.-F. Mao, M.A. Vannice / Applied Catalysis A: General 122 (1995) 61-76 75

followed by EO oxidation (secondary oxidation). In this study the ratio of the

oxidation rates, rzo/rc2H4, are 0.22 and 0.52 for Samples A and B, respectively,
with similar pure feed concentrations. With the partial pressure of EO during the
EO oxidation reaction much higher than that produced during the ethylene oxidation
reaction, the rates of EO oxidation are still lower than those of ethylene oxidation
under similar reaction conditions; consequently, ethylene oxidation may result not
only from the oxidation of EO but also from the direct oxidation of ethylene. Thus
both the primary and the secondary oxidation routes take part in ethylene oxidation
over these silver catalysts. On the unpromoted sample (Sample A), the primary
oxidation route may be dominant, as suggested by the low rEo/rc2m ratio, while
on the cesium-promoted sample (Sample B) the ratio increases and the secondary
oxidation becomes more significant. With the mixture of Sample G and pure (HSA)
a-alumina, the production of EO as a final product is completely suppressed and
the rate of carbon dioxide formation increases compared to that over Sample G,
thus indicating that the oxidation of EO plays an important role in ethylene oxi-
dation.

5. Summary

High dispersion of silver can be obtained using (HSA) a-aluminas as the support
material compared to silver supported on a (LSA) a-alumina. The addition of
cesium to some of the samples to neutralize the surface acidity of the support
significantly reduced oxygen chemisorption, which is attributed to the blockage of
the silver surface by cesium species. A wide variation in the H2/O2 titration ratio
was observed which is dependent on the presence of cesium, the sequence of
impregnation, the pretreatment, and the supports; consequently, the portion of the
oxygen associated only with the silver could not be clearly ascertained.
All the A g / ( H S A ) a-alumina catalysts showed no selectivity to EO in ethylene
oxidation; hence, these (HSA) a-aluminas are poor support materials for ethylene
epoxidation. The main reason for the lack of selectivity to EO is attributed to the
high activity of these (HSA) a-aluminas for EO isomerization to acetaldehyde,
which is then quickly oxidized to carbon dioxide. A physical mixture of a selective
silver catalyst and a (HSA) a-alumina gave only complete combustion products
thus providing further proof that (HSA) a-aluminas are detrimental to EO selec-
tivity because of their high reactivity with EO. Impregnation of cesium did not
enhance selectivity but increased the tumover frequency of carbon dioxide for-
mation and decreased the corresponding activation energy. This unexpected cesium
effect is attributed to the high loadings used and its promoting effect on the rate of
acetaldehyde oxidation.
Both silver and the alumina support were active for EO isomerization to acetal-
dehyde as well as EO oxidation. In the presence of cesium, the rate of EO isomer-
ization decreased while the rate of carbon dioxide formation increased compared
76 C.-F. Mao, M.A. Vannice /Applied Catalv,sis A: General 122 (1995)61-76

to the unpromoted sample. In the study of acetaldehyde oxidation, it was shown

that oxidation occurred on both silver and the support while silver was more active
than the support; in addition, a gas-phase reaction occurred at these temperatures.
On the cesium-impregnated sample, the rate of acetaldehyde oxidation markedly
increased, which can explain the enhancement in the rate of EO oxidation after the
addition of cesium to these catalysts.

Acknowledgements

This study was partially sponsored by the Aluminum Company of America. We

would like to thank Dr. A.J. Perrotta for providing the (HSA) a-A1203 samples.

References

[ 1 ] J.L. Lee, X.E. Verykios and R. Pitchai, Appl. Catal., 44 (1988) 223.
[2] H. Kanoh, T. Nishimura and A. Ayame, J. Catal., 57 (1979) 372.
[3] C.-F. Mao and M.A. Vannice, Appl. Coral. A, 122 (1995) 41.
[4] P. Harriott, J. Catal., 21 ( 1971 ) 56.
[5] Y.S. Yong, E.M. Kennedy and N.W. Cant, Appl. Catal., 76 ( 1991 ) 31.
[6] C.-F. Mao and M.A. Vannice, Appl. Catal. A, 111 (1994) 151.
[7] C. Contescu and J.A. Schwarz, Appl. Catal., 118 (1994) L5.
[8] S.R. Seyedmonir, D.E. Strohmayer, G.L. Geoffroy and M.A. Vannice, Adsorption Sci. Technol., 1 (1984)
253.
[9] J.K. Plischke and M.A. Vannice, Appl. Catal., 42 (1988) 255.
[ 10] R.D. Srivastava, Heterogeneous Catalytic Science, CRC Press, Boca Raton, 1988.
[ 11 ] S.R. Seyedmonir, J.K. Plischke, M.A. Vannice and H.W. Young, J. Catal., 123 (1990) 534.
[ 12] J.F. Griffiths and G. Skirrow, Oxid. Combus. Rev., 3 11968) 47.
[ 13] R.R. Baldwin, M.J. Matchan and R.W. Walker, Combust. Flame, 15 (1970) 109.
[ 14] E.W. Kaiser, C.K. Westbrook and W.J. Pitz, Int. J. Chem. Kinet., 18 (1986) 655.
[ 15] K.L. Anderson, J.K. Plischke and M.A. Vannice. J. Catal., 128 ( 1991 ) 148.
[ 16] W.-D. Mross, Catal. Rev. Sci. Eng., 25 (1983) 591.
[ 17] C.T. Campbell, J. Phys. Chem., 89 (1985) 5789.
[ 18] R.B. Grant and R.M. Lambert, Langmuir, 1 (1985) 29.
i 19] S.A. Tan, R.B. Grant and R.M. Lambert, Appl. Catal., 31 (1987) 159.
[20] P.T. Conner, S. Kovenklioglu and D.C. Shelly, Appl. Catal., 71 ( 1991 ) 247.
[ 21 ] T. Tsuchida, Appl. Catal., 105 (1993) L141.
[ 22] R.P. Nielsen and J.H. LaRochelle, US Patent 3 962 136 (1977).
[ 23] R.B. Grant, C.A.J. Harbach, R.M. Lambert and S.A. Tan, J. Chem. Soc., Faraday Trans. 1, 83 (1987) 2035.
]24] S.A. Tan, R.B. Grant and R.M. Lambert, J. Catal., 106 11987) 54.
[ 25] R.B. Grant and R.M. Lambert, J. Catal., 93 (1985) 92.
[ 26] R.A. van Santen and H.P.C.E. Kuipers, Adv. Catal., 35 11987) 265.
[ 27] R.E. Kenson and M. Lapkin, J. Phys. Chem., 74 (1970) 1493.