ISSN 0023-1584, Kinetics and Catalysis, 2020, Vol. 61, No. 4, pp. 603–612. © Pleiades Publishing, Ltd., 2020.

Kinetic Models of Ethylene Oxide Production on Ag Catalysts:

A Review
A. Fazelia, *, A. Naseria, and F. Eslamjamala
aCaspian Faculty of Engineering, College of Engineering, University of Tehran, Tehran, Iran
*e-mail: [email protected]
Received May 25, 2019; revised January 13, 2020; accepted February 17, 2020

Abstract—Process modeling and design of the ethylene oxide production plant is mainly depended on the
kinetics of the ethylene oxide production reactions. In this article, the kinetics equations for partial oxidation
(epoxidation) of ethylene on silver catalyst were reviewed. There are three competitive reactions in this sys-
tem: ethylene partial and total oxidation and ethylene oxide total oxidation. The reaction rate equations for
these three reactions were compared and advantage and the disadvantage of kinetic models were discussed.
Dichloroethane (DCE) is used in these reactions to increase the ethylene oxide selectivity. Therefore, the
kinetics of the reactions considering the role of DCE is also reported. Most of the kinetics models have some
weaknesses. However, one of the reviewed models was the complete model because of including all three
reactions of partial and total oxidation of ethylene and total oxidation of ethylene oxide. Also, this model con-
sidered the concentration of selectivity promoter (DCE) and reverse reactions.

Keywords: ethylene, partial oxidation, kinetics, reaction rate, ethylene oxide

DOI: 10.1134/S0023158420040059

1. INTRODUCTION Zhu et al. studied the role of Re, Cs and Cu as pro-

Ethylene oxide (EO) is an intermediate compound moters on the selectivity of ethylene partial oxidation
that cannot be directly used generally. It can be con- on Ag catalyst using density functional theory (DFT)
verted mainly (about 70%) to ethylene glycols [1]. at different plans of Ag(111), Ag(110) and Ag(100).
Other products from this basic chemical intermediate Finally, it was determined that for Ag (110) catalyst,
are ethanolamines, polyethylene glycol, glycol ethers, selectivity follows the order: Re < no-promoter < Cs <
ethoxylates, and polyether polyols [2]. EO historically Cu, but in the case of Ag (111) and Ag (100), the
had been produced from the chlorohydrin route that is arrangement of catalysts selectivity is in the rank of Re
environmentally pollutant. After that, ethylene cata- < no promoter < Cu < Cs. It became clear that the per-
lytic direct oxidation technology was developed which formance of the Ag(111) catalyst plane is better than
can use air or oxygen. The oxygen-typed process is that of Ag(110) and Ag(100). These promoters have a
economical at higher production capacity. Con- high impact on adsorption of reaction components
trolling the oxidation reaction at medium economic that cause to change the selectivity of the catalyst. Zhu
conversion is an important point to reduce water and et al. have tried to provide a theoretical method that
CO2 production at the same time with suitable EO improves the selectivity of EO production and deter-
production [3]. Ethylene, oxygen and a diluent over mines the appropriate promoter [7].
Ag/α-Al2O3 catalyst at temperature about 250°C can
van den Reijen et al. determined that the Ag cata-
produce EO more selectively by using only 1 ppm of a
lyst alone has a very negative effect on the selectivity of
compound including chlorine [4]. Guseinov et al.
studied the ethylene oxidation by varying oxidation ethylene to EO and they said that Ag nanoparticles on
degree of silver on Ag/corundum catalyst [5]. Demi- catalytic support have a much better effect on the EO’s
dova et al. investigated the role of Ag particle size by in selectivity. Among the catalyst supports that are avail-
situ XPS and proved that small Ag particles (8 nm) able for the Ag catalyst in ethylene oxide production,
were inactive for ethylene oxidation. But Ag with a α-Al2O3 has a better performance than SiO2 [8]. Chen
particle size of around 40 nm had electrophilic and et al. investigated the effect of ethyl chloride on the
nucleophilic oxygen on the surface and it was highly performance of the Ag/α-Al2O3 catalyst. According to
active [6]. their study, chlorine in ethyl chloride increased EO
selectivity from 81.7 to 86.6% [9]. Main reactor types
Abbreviations: EO, ethylene oxide; DCE, 1,2-dichloroethane. that are used for EO production are multi-tubular

603
604 FAZELI et al.

Table 1. Langmuir–Hinshelwood kinetic parameters [13]

C2H4O formation CO2 formation
T, °C
k , mol g–1 h–1 K C2H4 , bar–1 K O2 , bar–1 k ' , mol g–1 h–1 K C' 2H4 , bar–1 K O' 2 , bar–1

180 0.11 ± 0.02 4.20 ± 1.06 1.40 ± 0.40 0.04 ± 0.008 10.80 ± 3.64 2.66 ± 0.79
185 0.36 ± 0.16 2.48 ± 0.94 0.86 ± 0.40 0.17 ± 0.06 6.02 ± 3.02 1.60 ± 0.82
210 0.70 ± 0.18 1.48 ± 0.30 0.77 ± 0.19 0.38 ± 0.12 2.42 ± 0.82 1.55 ± 0.56
180* 0.036 ± 0.007 6.03 ± 1.33 1.09 ± 0.27 0.08 ± 0.002 8.44 ± 2.57 1.42 ± 0.44
* Chlorinated catalyst.

fixed-bed. Catalysts are filled in tubes and shell-side (I)

of the reactor are cooled using water vapor or oil [10]. 0.5 Oxygen
Ethylene Ethylene oxide
Total and partial ethylene oxidation and total com-
bustion of EO may occur simultaneously in EO pro- (II) (III)
3 Oxygen 2.5 Oxygen
duction. These three reaction pathways are presented
in Scheme 1 and listed in Eqs. (1) to (3) [11]. Kolo-
2 Carbon dioxide + 2 Water
bashkin et al. modeled EO production considering the
transport phenomena inside the catalyst grain [12]. Scheme 1.
Despite the importance of kinetics of EO production
reactions, there is no review article to study the kinet- 2. KINETIC MODELS
ics of these reactions that lists the reaction rate equa-
tions and discusses the differences. Therefore, in this In this section, kinetic models for EO production
article, reviewing and discussing of the advantages and are reviewed in detail from the past to the present and
comprehensive conditions that the models are used
disadvantages of presented reaction rate equations for will be discussed briefly.
EO production are the main targets that can be useful
for mathematical modeling of EO production reactor
and process. 2.1. Ghazali Model
Ghazali et al. studied kinetic of 11.3 wt % Ag/1.0 wt
C2H4 + 1 O2  C2H4O, % Ba/SiO2 catalyst for ethylene epoxidation at 180–
2 (1) 210°C [13]. They used a U-tube that works as a fixed
ΔH = –1.23 × 10 kJ/kmol,
5
bed semi-differential reactor. They considered two
competitive reactions (I) and (II) and used the Lang-
C2H4 + 3O2  2CO2 + 2H2O, muir–Hinshelwood kinetic model (Eqs. (4) and (5))
(2) to fit their experimental data. The parameters of these
ΔH = –1.30 × 106 kJ/kmol, equations were reported at 180–185–210°C for the
general catalyst and at 180°C for chlorinated catalyst
(Table 1). Ghazali et al. [13] applied 39 experimental
C2H4O + 5 O2  2CO2 + 2H2O, data of Klugherz et al. [14] at 220°C and fitted the
2 (3)
parameters by non-linear regression. These parame-
ΔH = –1.17 × 10 kJ/kmol.
6
ters are listed in Table 2.
Ethylene oxide production reaction pathways [11]. kK C2H4 K O2 PC2H4 PO2
R (1) = , (4)
(1 + KC H PC H
2 4 2 4
+ K O2 PO2 )
2

Table 2. Fitted parameters of Eqs. (4) and (5) [13] accord-

ing to Klugherz et al. [14] (experiments at 220°C) k ' K C' 2H4 K O' 2 PC2H4 PO2
R ( 2) = . (5)
(1 + K ' )
2
Parameter Value
C2H4 PC2H4 + K O' 2 PO2

k , mol g–1 h–1 (31.6 ± 4.4) × 10–4 Ghazali et al. [13] developed a modified Lang-
muir–Hinshelwood kinetic model by adding a power
K C2H4 , bar–1 2.91 ± 0.29 expression to the denominator. Equation (6) is the
modified equation which the parameters were esti-
K O2 , bar–1 0.57 ± 0.11 mated using 211 experimental points of Klugherz et al.
[14] (Table 3).

KINETICS AND CATALYSIS Vol. 61 No. 4 2020

 KINETIC MODELS OF ETHYLENE OXIDE PRODUCTION 605

kK C2H4 K O2 PC2H4 PO2

R (1) = . (6)
( )
2
1 + K C2H4 PC2H4 + K O2PO2 + K C2H4OPC2H4 PO2 PC2H4O
0.19 0.35

A power correlation for the ratio of rates of reaction atmospheric pressure. No reaction was found at
(I) to (II) instead of the rate of reaction (II) was 210°C. The proposed kinetic model consists of Eqs.
reported in [13]. This correlation is shown in Eq. (7). (9) and (10) corresponding to reactions (I) and (II),
Another treatment of Eq. (4) by adding an ethylene respectively.
oxide inhibition term was offered as described by Eq.
(8). This modified model was fitted using Klugherz et k1 PO2 PC2H4 − k2PO2 PC2H4 PDCE
k7

al. [14] experiments and fit parameters are listed in Table R (1) = , (9)
1+ k5PO2 + k6PC2H4
4. The fit parameters of the Ghazali et al.’s models [13]
restricted to the kinetic factor and adsorption con- k3 PO2 PC2H4 − k4 PO2 PC2H4 PDCE
k8
stants that were not related to temperature using an R ( 2) = . (10)
expression. Therefore, no activation energy was 1+ k5PO2 + k6PC2H4
reported in [13]. The model did not include any term
to account for the relationship between the rate of Values of the pre-exponential factors and activa-
reaction and chlorine components generally used for tion energies were summarized in Table 5 according to
increasing the selectivity of the reactor. Eqs. (11) and (12). PO2 , PC2H4 and PDCE represents par-
tial pressure of oxygen, ethylene and dichloroethane
R (1) S (DCE) in atm. R(1) and R(2) denote reaction rates in
= = 2.43PC2H4 PO2 ,
0.16 0.08
(7)
R ( 2) 100 − S −1
mol h–1 gcat −1
. k1 and k3 in mol h–1 gcat atm–3 and k5, k6
k K C2H4 K O2 PC2H4 PO2 in atm–1 are pre-exponential factors.
R (1) =
(−RTE ) (for k − k ) ,
. (8)
(1 + KC H PC H
2 4 2 4
+ K O2 PO2 + K C2H4OPC2H4O )
2
k = k0 exp a
1 4 (11)

k = k0 exp  a  ( for k5 and k6 ) .

E
2.2. Petrov Model (12)
 RT 
Petrov et al. [15] developed two kinetics rate
expressions for reactions (I) and (II) for a chlorine- k7 = 0.19 and k8 = 0.07 are powers of PDCE. These two
promoted silver catalyst. This catalyst has a 0.14 m2/g powers are dimensionless and constant at all tempera-
tures. Equations (9) and (10) are reported to be valid
pore surface area and 0.05 m2/g silver surface area. only at 0.2 to 28 ppm DCE concentrations. A kinetic
The experiments were performed in a fixed-bed reac- model for the Ag catalyst without using the DCE pro-
tor at a temperature of 210, 240, 264 and 292°C and motor has been presented in [16]. Inhibition of
adsorbed oxygen and ethylene was considered in this
Table 3. Fitted parameters of Eq. (6) [13] based on 211 model. The best advantage of this kinetic model is
experimental points (Klugherz et al. [14]) considering DCE concentration. Powers of PDCE at
reaction (I) and (II) are different and this difference
Parameter Value can predict the selectivity change well. DCE inhibits
k , mol g–1 h–1 (41 ± 4.1) × 10–4
the main reaction more than the side reaction. It is
believed that DCE inhibits dissociative adsorption of
K C2H4 , bar–1 3.40 ± 0.18 oxygen on silver active sites [15].
K O2 , bar–1 0.57 ± 0.06 Petrov et al. studied the kinetics of ethylene oxide
oxidation on the Ag/Al2O3 catalyst [17]. Their experi-
K C2H4O , bar–1 93.6 ± 12.2
ments were carried out in a fixed bed reactor under
atmospheric pressure at 100–300°C. In this kinetics,
the concentration of DCE and the rate of the back-
Table 4. Fitted parameters of Eq. (8) [13] based on Klugh- ward reaction are not considered. They have proven
erz et al. [14] experiments that at temperatures below 190°C, total oxidation of
Parameter Value ethylene oxide does not occur. In addition, at a tem-
perature higher than 300°C, decomposition of eth-
k, mol g–1 h–1 (29.5 ± 1.8) × 10–4 ylene oxide into ethylene and oxygen occurred. In this
kinetics, the conversion of ethylene oxide to formalde-
K C2H4 , bar–1 4.32 ± 0.27
hyde is a limiting step. The reaction rate of the total
K O2 , bar–1 0.88 ± 0.08 oxidation of ethylene oxide is given by Eq. (13). The
pre-exponential factor and activation energy related to
K C2H4O , bar–1 69.5 ± 10.2 Eq. (13) are listed in Table 6.

KINETICS AND CATALYSIS Vol. 61 No. 4 2020

606 FAZELI et al.

C1PO2 PC2H4O
R (3 ) = . (13)
(1 + C2PO2 +C3PO0.5
2
+C4 PC2H4O +C5PC2H4O PO−20.5 )2

2.3. Al-Saleh Model tions were performed at operating temperatures of 240

to 320°C and pressure of 21.7 bar. The Berty reactor
Al-Saleh et al. [18] studied ethylene oxidation reac- feed contained 3–6% of ethylene, 18–19.5% of oxy-
tions using the Berty reactor. They used a commercial gen and the balance of nitrogen. They developed two
Ag/α-alumina for their experiment. Their investiga- reaction expressions according to Eqs. (14) and (15)
for ethylene partial and total oxidation reaction,
respectively. These reaction rates are similar to the
Table 5. Values of the pre-exponential factors and activa-
Langmuir–Hinshelwood kinetic model and both are
tion energies of Eqs. (9) and (10) [15]
inhibited by CO2. The powers of oxygen partial pres-
Pre-exponential factor, k0 Activtion sure at these two equations are the same, but the power
Factor energy, of ethylene partial pressure in Eq. (14) is greater than
value units Ea, cal/mol that in Eq. (15). With increasing ethylene conversion,
k1 6.867 −1 9260 ethylene partial pressure remained at a low level;
mol h–1 gcat atm–2
therefore, the rate of reaction (I) damped more than
k2 1.073 × 102 −1 10 413 that of reaction (II) and hence selectivity decreased. It
mol h–1 gcat atm–3
−1
was found that temperature has only a minor effect on
k3 1.062 × 10 mol h–1 gcat atm–2 8792
the selectivity. Al-Saleh et al. [18] summarized the
k4 3.959 × 10 −1 9473 activation energies of these two reactions (Table 7).
mol h–1 gcat atm–3
They reported that their activations are higher than
k5 1.346 × 10–6 atm–1 9610
some values of this table. In Eqs. (14) and (15), reac-
k6 3.109 × 10–8 atm–1 15044 −1 –1
tion rates are expressed in mol gcat h , T is the tem-
perature in K, P is the partial pressure in Pa.
Table 6. Pre-exponential factor and activation energy total
oxidation reaction related to Eq. (13) [17]

T )
(2.259 × 10 ) exp ( − 9010
Pre-exponential factor, C0 Ea, 7 0.5 0.6
Factor (P ) ( P O2 C2H4 )
(cal/mol)/(kJ/mol) R (1) =
1 + (1.985 × 10 ) exp ( 2400 ) P
value units ,(14)
2
−1 CO2
C1 1.614 × 10 mol h–1 gcat atm–2 9565/39.87 T
C2 4.401 × 10–10 atm–1 19422/80.96
C3 9.913 × 10–7 atm–0.5 12913/53.83
C4 3.373 × 10–14 atm–1 32390/135.02
C5 7.031 × 10–11 atm–0.5 23422/97.64 (4.34 × 10 ) exp  −10800
7

T 

 (P ) (P O2
0.5
C2H4 )
0.5

R ( 2) =
1 + (1.08 × 10 ) exp (1530) P
. (15)
2
Table 7. Summary of activation energy of partial and total CO2
T
oxidation reaction [18]
Activation energy,
Author, year Ea, kJ/mol
2.4. Borman Model
R (1) R (2)
Borman et al. [19] performed a series of kinetic
Twigg (1946) 67 67 experiments for developing kinetic rate expressions for
Murray (1950) 50 36 ethylene partial and total oxidation reactions on
Wan Shen-Wu (1953) 81 – industrial Ag/α-Al2O3 catalyst. An internal recycle
Kurilenko (1962) 64 83 reactor similar to the Berty reactor was used. The tests
were performed in a temperature range of 181–254°C
Alfani (1970) 36 31
and a pressure range of 2–10 bar. Three models for
Dettwiler (1979) 77 63 ethylene partial and total oxidation reactions have
Al-Saleh (1988) 95 102 been proposed (Eqs. (16)–(18)):

KINETICS AND CATALYSIS Vol. 61 No. 4 2020

 KINETIC MODELS OF ETHYLENE OXIDE PRODUCTION 607

Table 8. Fitted parameters of the Borman et al. in three models [19] kri = kri, 0 exp ( −Ei /RT ) and K ij = K ij,0exp(Eads
i
/RT )
Model I Model II Model III
Parameter
R(1) R(2) R(1) R(2) R(1) R(2)
kr ,0 2.9 × 103 1.42 × 10 6 3.43 × 1013 8.1 × 109 4.72 × 1012 1.08 × 108
Model I: (mol kg–1 s–1 bar–n)
Model II–III: (mol kg–1 s–1)
Eai /R , K 8070 11380 18070 14250 17070 14650
K C2H4 (bar–1 ) 30.0 49.4 12.9 14.9 12.9 26.0

K C2H4,0 , bar–1 – – 2.67 × 10–8 5.30 × 10–3 2.05 × 10–7 4.0 × 10–3
Eads,C2H4 /R , K – – 9520 3780 8550 4180
K CO2 , bar–1 87.4 114.4 83.9 107 83.9 98.5
K C2H4O , bar–1 90.0 48.8 91.8 40.0 91.8 41.3

K C2H4O,0 , bar–1 – – – 1.49 – 1.21

Eads,C2H4O /R , K – – – 1570 – 1680
K H2O , bar–1 53.0 54.6 47.9 43.4 47.9 48.9

K H2O,0 , bar–1 0.368 4.04 × 10–2 1.99 7.45 × 10–3 8.25 0.109
Eads,H2O /R, K 2370 3430 1510 4130 840 2910
K O2 , bar–1 – – 9.97 8.76 84.8 66.7
n 0.127 0.140 – – – –

kri PC2H4 POn2i K Ci 2H4

R (i ) = , (16)
1 + K Ci 2H4 PC2H4 + K CO
i
P + K Hi 2OPH2O + K Ci 2H4OPC2H4O
2 CO2

kri PC2H4 PO2 K Ci 2H4 K Oi 2

R (i ) = , (17)
(1 + K Ci 2H4 PC2H4 + K CO
i
P + K Hi 2OPH2O + K Ci 2H4OPC2H4O + K Oi 2 PO2 )2
2 CO2

kri PC2H4 (PO2 )0.5 K Ci 2H4 (K Oi 2 )0.5

R (i ) = . (18)
( )
0.5 2
(1 + K C2H4 PC2H4 + K CO2 PCO2 + K H2OPH2O + K C2H4OPC2H4O + K O2 PO2
i i i i i
)

The parameters of these three models are summa- Borman et al. [19] developed a kinetic model for
rized in Table 8. The goodness of fitting at the different total oxidation of ethylene oxide according to Eq. (19):
models is compared in Table 9. The results show that ( −10500T )
these three models have almost the same power in pre- R (3) = 1.7 × 105 e PC0.4 mol kg −1s −1  . (19)
2H4O  
dicting the reaction rate, but model III based on the
Langmuir–Hinshelwood mechanism has a little bet-
ter performance in fitting the experimental data and 2.5. Lafarga Model
describing the reaction kinetics. Oxygen is suggested Lafarga et al. developed a kinetic model for eth-
to adsorb dissociatively and to react with adsorbed eth- ylene oxide production based on an Ag–Cs/α-Al2O3
ylene oxide. The activation energy of ethylene partial catalyst using a differential and integral fixed-bed
oxidation is smaller than that of the total oxidation reactor at a temperature range of 210–270°C and pres-
reaction. Therefore, ethylene oxide selectivity sure of 1 bar [20]. The two reaction rates (ethylene
increases with decreasing temperature. The lower par- epoxidation and total oxidation) are in the Langmuir–
tial pressure of ethylene oxide and higher partial pres- Hinshelwood dual-site type. Lafarga et al. modified
sure of CO2 increases the selectivity. the previous rate expressions by adding ni as a power of

KINETICS AND CATALYSIS Vol. 61 No. 4 2020

608 FAZELI et al.

Table 9. Comparison of fitting criteria for three Borman et al.’s models [19]
Model I Model II Model III
Statistical parameters
R(1) R(2) R(1) R(2) R(1) R(2)
Log likelihood function 1404.6 1437.0 1390.7 1422.8 1396.4 1434.1
Average error, % 19.0 15.0 20.8 18.0 19.6 14.7
Standard deviation, % 16.8 14.8 19.8 13.9 16.9 12.9

oxygen partial pressure. The final reaction rate expres- kinetics of Lafarga et al. In addition, adsorption con-
sion is presented by Eq. (20): stant, K C2H4 was not taken into account as a function of
temperature.
ki PC2H4 POn2i
R (i ) = , (20)
(1 + K Ci 2H4 PC2H4 )2 2.6. Aryana Model
where i denotes the reaction type: (I) stands for eth- Aryana et al. combined the two most commonly
ylene epoxidation and (II) for ethylene total oxidation. used kinetics for full and partial ethylene oxidation
ki is the Arrhenius type reaction rate constant, PC2H4 with the third kinetics of total ethylene oxide combus-
tion. These three reaction pathways are presented by
and PO2 are ethylene and oxygen partial pressures in reactions (I) to (III) [11]. The rate of these three reac-
bar. K Ci 2H4 is adsorption constants of ethylene. The fit- tions was reported in the form of Eq. (21):
ted parameters based on differential reactor are listed R (i ) = fi (T , p j ) × ai (t ,T , pk ) × CFi , (21)
in Table 10.
Lafarga et al. used the parameters of Table 10 for where R (i ) is the overall reaction rate, fi is reaction
modeling the integral reactor and compare the model rates that are presented in Eqs. (22) to (24). ai is deac-
and experimental results and found that the kinetic tivation factor and CFi is a correction factor that must
parameters must be fitted again to approach the inte- be provided by catalyst vendor. DCE increases the
gral reactor experimental results [20]. The improved selectivity of the reaction and has an inhibitory effect
parameters are summarized in Table 11. The order of that was considered in a DCE power term.
PO2 in ethylene epoxidation reaction is greater than
K1PC2H4 PO2 − K 2PC2H4 PO2 PDCE
k13
that in ethylene total oxidation. Therefore, the selec- f (1) = − 1 ρcat , (22)
tivity is reported to decrease with increasing oxygen 2 1 + K 5PO2 + K 6PC2H4
partial pressure.
K 3 PC2H4 PO2 − K 4 PO2 PDCE
k14
Ethylene oxide oxidation reaction, addition of f ( 2) =−3 ρcat, (23)
DCE and reverse reactions were not considered in the 1 + K 5 PO2 + K 6PC2H4

K 7PC2H4OPO2 − K 8PC2H4 PC2H4OPDCEk15

f (3) =− 5 ρ .
−0.5 2 cat
(24)
2(1 + K 9PO2 + K10PO0.5
2
+ K11PC H
2 4O + K12 PC H
2 4O PO 2
)

PC2H4 , PO2 , PDCE and PC2H4O denote partial pressure of nius equation (Eq. (25)) and equilibrium adsorption
ethylene, oxygen, dichloroethane and ethylene oxide constant that can be presented in the Arrhenius form
respectively. K1 to K12 are stated based on the Arrhe- with positive activation energy (Eq. (26)).

 − E a, i 
K i =1−4 , 7–8 = K O2 , i e  , (25)
Table 10. Fitted parameters of Lafarga et al.’s model based  RT 
on differential reactor experiments [20]
E 
Parameter R(1) R(2) K i = 5, 6, 9 −12 = K O2, i e  a, i  . (26)
 RT 
k0 , mmol g–1 s–1(bar1+n)–1 3.51 × 105 1.06 × 10 6 Aryana et al. [11] say that K1 to K6 and power k13 are
Ea, kJ/mol 56.7 63.0 presented in [15] and K7 to K12 and power of k15 are
gotten from [17]. In this kinetics, the catalyst is of an
K C2H4 , bar–1 17.8 15.4 Ag/Al2O3 type. Aryana et al. used these kinetic expres-
sions for modeling a multi-tubular fixed-bed reactor
n 1.17 0.97 with 7750 tubes at a temperature of 230 to 240°C and

KINETICS AND CATALYSIS Vol. 61 No. 4 2020

 KINETIC MODELS OF ETHYLENE OXIDE PRODUCTION 609

a pressure of 20–25 atm and DCE with a concentra- Table 11. Modified parameters of Lafarga et al. model
tion of 1–3 ppm. In the proposed model, the effect of based on integral reactor experiments [20]
operating parameters such as DCE level, reactor con- Parameter R(1) R(2)
centration, reactor pressure, cooling temperature and
feed temperature on the reactor performance was k0 , mmol g–1 s–1(bar1+ n)–1 1.33 × 105 1.8 × 106
studied. Ultimately, their model results have the most Ea, kJ/mol 60.7 73.2
consistent with empirical data [11].
K C2H4 , bar–1 6.50 4.33

2.7. Carucci Model n 0.58 0.30

Carucci et al. used a microchannel reactor for
kinetics study of ethylene oxide production on an Table 12. Kinetic parameters of Eq. (28) [21]
Ag/α-alumina catalyst at 220–300°C and atmo-
spheric pressure. Two kinetic models for ethylene par- Parameter* Value
tial oxidation have been suggested that are summa-
K C2H4 132 ± 40.1
rized in Eqs. (27) and (28) [21]:
K O2 77.4 ± 13.4
k ' CC2H4 CO2
R (1) = , (27) k' (9.75 ± 4.42) × 10–3
(1 + KC H CC H + K O2CO2 )
2
2 4 2 4 * The unites were not well determined in the reference.
k ' CC2H4CO2
R (1) = , (28)
(1 + KC H CC H
2 4 2 4
+ K O2CO2 )
2 oxygen is active component in the epoxidation reac-
tion, whereas dissociatively adsorbed oxygen is an
where R (1) is ethylene to ethylene oxide reaction rate, active site in the total oxidation reaction. Also, the eth-
Ci is the concentration of component i, Ki is adsorption ylene oxide selectivity is proportional to CO1/2 . The rate
constant of component i, ḱ is the apparent rate constant. 2
of reaction (I) based on the Langmuir–Hinshelwood
Kinetic parameters of Eq. (28) are listed in Table 12. mechanism is expressed by Eq. (29). In addition,
After non-linear curve fitting, the correlation coef- Eley–Rideal mechanism is used to present the rate of
ficients (R2) of Eqs. (27) and (28) are 93 and 96.7%, reaction (I) by Eq. (30). Salmi et al. [22] reported that
respectively. In Eq. (27), dissociative oxygen adsorp- the Eley–Rideal model of Eq. (30) has a higher coef-
tion is a rate determining step and in Eq. (28), ethylene ficient of determination (R2) than Eq. (29). The fitted
and molecular oxygen compete for adsorption and the parameters of these kinetic models are compared in
surface reaction is a rate determining step. In Eqs. (27) and Table 13.
(28), CO2 and CO2 are used, respectively and Eq. (28) is
k1'CC2H4CO2
better fitted to experimental data than Eq. (27). R (1) = , (29)
In these kinetic models, chlorine selectivity pro- (1 + KC H CC H )2
2 4 2 4
moters such as DCE are neglected, while these pro-
moters are used in the industrial production of eth- k1'CC2H4CO2
ylene oxide. In addition, in the proposed kinetic mod- R (1) = , (30)
els, no attention was paid to the reverse reaction term 1 + K C2H4CC2H4
and only the main reaction has been considered and the
k2'CC2H4 CO2
secondary reaction (total oxidation of ethylene) has been R ( 2) = . (31)
discarded, therefore, selectivity cannot be studied. (1 + KC H CC H )2
2 4 2 4

2.8. Salmi Model 3. COMPARISON OF MODELS

Salmi et al. [22] developed a kinetic model using a Summary of kinetic models and detailed condi-
microchannel reactor at a constant temperature of tions, advantages and disadvantages of models are
250°C. The feed contains 2–25% O2, 5–30% ethylene listed in Table 14. In the following subsections, the
and balance of He. Catalysts are pure silver micro main similarity and difference of the kinetic models
thickness plates (9.5 × 9.5 mm). Each plate is grooved are discussed.
by 9 parallel microchannels with a width of 460 μm
and a height of 80 μm. The pressure of 1–5 bar is 3.1. Catalyst
adjusted by a backpressure regulator. The experimen-
tal results at a given operating condition show that the The catalyst type of each model is presented in the
rate of reactions (I) and (II) is proportional to CO2 and first column of Table 14. This column shows that all
the catalysts use silver as a catalytic active agent. The
CO1/2
2
, respectively. This shows that adsorbed molecular active metal percent in two references is 20 and 11.3%

KINETICS AND CATALYSIS Vol. 61 No. 4 2020

610 FAZELI et al.

Table 13. Kinetic parameters of Eqs. (29) and (30) [22] This is a combination of the two works by Petrov et al.
Value [15, 17].
Parameter
eq. (29) eq. (30)
3.5. Considering DCE
K C2H4 , L/mol 0.228 × 10–3 0.362 × 10–3
The role of DCE in their models was considered in
k' * 32.7 178 [11] and [15], but the rest of the articles ignored the
2
R ,% 92.4 95.9 role of DCE.
* The unit was not well determined in the reference.
3.6. Considering Reverse Reaction
and is unclear in other catalysts. Also, α-Al2O3 is the Reverse reactions were inserted into the rate equa-
support of all catalysts except for the Ghazali model tions only in [11, 15], while the rest of the articles did
that uses SiO2 as catalyst support. It seems that all cat- not take them into account.
alysts are synthesized at a laboratory scale expect the
Al-Saleh model that applied a commercial catalyst. Ca 4. CONCLUSIONS
was used as a promoter in catalyst applied in Carucci
and Petrov models. However, Lafarga kinetic model is Ethylene oxide is mainly produced from ethylene
based on Cs promoter and the model of Ghazali is partial oxidation. The partial oxidation reaction com-
derived for a catalyst that used Ba as the promoter. petes with the total oxidation reaction. Both reactions
Petrov model also used a chlorined promoter for inhi- are highly exothermic and therefore temperature con-
bition of total oxidation side reaction and increasing trolling is very critical in selectivity and conversion
the selectivity. trade-off. Also, a high temperature of reaction results
temperature run away and low temperature causes to
reach a low conversion. Therefore, there is an opti-
3.2. Reactor mum temperature that is far from run-away tempera-
ture and the optimum process has a higher economic
Most of the available articles report using a fixed- efficiency that holds the selectivity at a high level and
bed reactor. Al-Saleh et al. [18] applied the Berty reac- at the same time keeps the conversion at economical
tor (CSTR), and in the Carucci model [21], a micro- value. To reach the optimum values, a reactor mathe-
channel reactor was used. Of course, articles that use matical modeling is necessary and the major problem
the fixed-bed reactors are also different; for example, of reactor modeling is achieving reactions rates. Some
Lafarga et al. [20] investigated both the integral and kinetic rates for reaction (I) to (III) are available in the
differential type of fixed-bed reactors, however literature, but they differ depending on the type of cat-
Ghazali et al. [13] studied only the semi-differential alyst, reactor and operating conditions. These param-
type. Aryana et al. [11] modeled a multi-tubular fixed- eters are necessary for fitting to experimental data.
bed reactor using the combination of works by Petrov Also, the kinetics models have a discrepancy in mech-
et al. [15–17]. anism assumptions that led to the diversity of formula-
rization. Therefore, there is a necessity to collect the
kinetic models from the past to the present and review
3.3. Operating Condition the strength and weakness of each model.
Typically, the used temperatures ranged from 210 Ghazali et al. [13] presented a modified Lang-
to 320°C, except [17] in which the temperature was in muir–Hinshelwood kinetic model by adding a power
the range of 100–300°C. It should be noted that at expression to the denominator. Lafarga et al. [20]
temperatures lower than 190°C, ethylene oxide total developed a kinetic model that has a power for PO2 and
oxidation reaction does not start significantly. Ghazali the powers for total and partial oxidation of ethylene
et al. [13] performed the reactions at 180–210°C. are different and this difference led to better prediction
About the pressure, all researches listed in Table 14 of selectivity. A kinetic model was developed in [20]
were carried out at atmospheric pressure except for that has the same power for oxygen partial pressure at
those by Aryana et al. [11] (P = 20–25 atm) and by Al- two competitive reactions of partial and total oxidation
Saleh et al. [18] (P = 21.7 bar). of ethylene. However, power of ethylene partial pres-
sure in the kinetics rate of ethylene partial oxidation is
greater than that in the total oxidation and this model
3.4. Reaction Type can predict the decrease of selectivity by increasing the
Most of the articles of Table 14 examined only ethylene conversion. Carucci et al. [21] developed two
reactions (I) and (II), except [17] and [11]. Only reac- kinetic models for partial oxidation of ethylene using a
tion (III) was considered by Petrov et al. [17]. Aryana microchannel reactor. In the two model equations,
et al. [11] considered all the reactions, i.e. (I) to (III). CO2 or CO2 were used. It was reported that the model

KINETICS AND CATALYSIS Vol. 61 No. 4 2020

 Table 14. Comparison of different kinetics models
Rate Eq. Reaction Reverse
Catalyst Reactor T/P, οC/bar or atm DCE Remark Ref., Year
(no.) type reaction

Rate of reaction in Eqs. (25) and (26) is pro-

(29) (1)
Pure silver Microchannel portional to CO2 and CO2 that shows O2
250/1–5 bar (30) (1) No No [22], 2013
microplates reactor and O as an active component in ethylene
(31) (2)
epoxidation and total oxidation, respectively.

Microchannel (27) (1) In Eqs. (23) and (24), CO2 and CO2 were

KINETICS AND CATALYSIS

Ag–Ca/α-Al2O3 220–300/1 atm No No used, respectively and Eq. (24) was better fit- [21], 2010
reactor (28) (1)
ted to experimental data than Eq. (23)
(22) (1) Kinetics [15] and [17] was used together for

Vol. 61
Multi-tubular Yes
Ag/α-Al2O3 230–240/20–25 atm (23) (2) Yes modeling of an industrial multi-tubular [11], 2009
fixed-bed reactor (1–3 ppm)
(24) (3) fixed-bed reactor

No. 4
Adding a power to PO2 Langmuir–Hinsel-
wood dual-site type.
Differential and

2020
(20) (1) Adsorption constant, K C2H4 , is not taking
Ag–Cs/α-Al2O3 integral fixed-bed 210–270/1 bar No No [20], 2000
(20) (2) into account as a function of temperature.
reactor
Different parameters were reported for dif-
ferential and integral reactors
Internal recycle (16) (1)
They proposed three models for ethylene
Ag/α-Al2O3 reactor similar to 181–254/2–10 bar (17) (2) No No [19], 1995
partial and total oxidation reactions
Berty reactor (18) (3)
Feed contains 3–6% ethylene, 18–19.5%
Commercial Berty reactor (14) (1) oxygen and the balance of nitrogen.
240–320/21.7 bar No No [18], 1988
Ag/α-Al2O3 (CSTR) (15) (2) Langmuir–Hinshelwood kinetic model.
Both reactions are inhibited by CO2
KINETIC MODELS OF ETHYLENE OXIDE PRODUCTION

20%Ag/Ca/α-Al2O3 Fixed-bed reactor 100–300/1 atm (13) (3) No No Total oxidation of ethylene oxide [17], 1988
Powers of PDCE at reaction (1) and (2) are
Chlorine-promoted (9) (1)
Fixed-bed reactor 210, 240, 264, 292/1 atm Yes Yes different and this difference can predict the [15], 1986
20%Ag/Ca/α-Al2O3 (10) (2)
selectivity changing well
A modified Langmuir–Hinshelwood
U-tube fixed-bed (4) (1)
kinetic model was developed by adding a
11.3%Ag–1%Ba/SiO2 semi-differential 180–210/1 atm (5) (2) No No [13], 1983
power expression to the nominator that
reactor (6) (1)
results Eq. (6)
611
612 FAZELI et al.

with CO2 is better fitted to experimental data than the CO2 Concentration of O2 component;
model including CO2 and this proves that adsorption CC2H4 Concentration of C2H4 component;
of oxygen is molecular against dissociative adsorption
of oxygen. Petrov et al. modified their model of [16] by k13 , k14 , k15 Power of PDCE;
adding the effect of DCE [15] and then modified their K1 to K12 Reaction rate constant or adsorption constant.
model by considering the total oxidation of ethylene
oxide [17]. Aryana et al. combine models of Petrov
et al. [15, 16] and use the combination to simulate an REFERENCES
industrial multi-tubular fixed-bed reactor [11]. Their 1. Ren, D., Xu, H., Li, J., Li, J., and Cheng, D., Mol.
models are the complete model because of including Catal., 2017, vol. 441, p. 92.
all three reactions of partial and total oxidation of eth-
ylene and total oxidation of ethylene oxide. Also, their 2. Devanney, M.T., Ethylene Oxide, CEH Marketing Re-
port, Chemical Economics Handbook, SRI Consulting,
model considered the concentration of selectivity pro- 2019, p. 15.
moter (DCE) and reverse reactions. 3. Zomerdijk, J.C. and Hall, M.W., Catal. Rev., 1981,
vol. 23, p. 163.
NOMENCLATURE 4. Kilty, P.A. and Sachtler, W.M.H., Catal. Rev., 1974,
vol. 10, p. 1.
PO2 Oxygen partial pressure; 5. Guseinov, S.L., Frolkina, I.T., Vasilevich, L.A., Avet-
isov, A.K., and Gel’bstein, A.I., React. Kinet. Catal.
PCO2 Carbon dioxide partial pressure; Lett., 1977, vol. 6, p. 409.
PC2H4 Ethylene partial pressure; 6. Demidov, D.V., Prosvirin, I.P., Sorokin, A.M., Rocha, T.,
Knop-Gericke, A., and Bukhtiyarov, V.I., Kinet. Catal.,
PDCE Promoter partial pressure; 2011, vol. 52, p. 855.
PC2H4O Ethylene oxide partial pressure; 7. Zhu, L., Xu, H., Nan, Y., Xie, Y., Zhu, J., and Cheng, D.,
Appl. Surf. Sci., 2019, vol. 476, p. 115.
K C2H4 Adsorption constant for Langmuir–Hinshel- 8. van den Reijen, J.E., Versluis, W.C., Kanungo, S.,
wood; d’Angelo, M.F., de Jong, K.P., and de Jongh, P.E.,
Adsorption constant for Langmuir–Hinshel- Catal. Today, 2019, vol. 338, p. 31.
K O2
wood; 9. Chen, C.-J., Harris, J.W., and Bhan, A., Chem. Eur. J.,
2018, vol. 24, p. 12405.
K C' 2H4 Adsorption constant for Langmuir–Hinshel-
10. Partopour, B. and Dixon, A.G., AIChE J., 2017,
wood; vol. 63, p. 87.
' Adsorption constant for Langmuir–Hinshel- 11. Aryana, S., Ahmadi, M., Gomes, Vincent G., Romag-
K O2
wood; noli, Jose A., and Ngian, K., Chem. Prod. Process Mod-
Adsorption constant for Langmuir–Hinshel- el., 2009, vol. 4, no. 1, article 14.
K C2H4O
wood; 12. Kolobashkin, V.S., Avetisov, A.K., Shub, F.S., and
Slinko, M.G., Khim. Prom-st., 1989, no. 12, p. 888.
K C' 2H4O Adsorption constant for Langmuir–Hinshel- 13. Ghazali, S., Park, D.W., and Gau, G., Appl. Catal.,
wood; 1983, vol. 6, p. 195.
k, k ' , ki Reaction rate constant; 14. Klugherz, P.D. and Harriott, P., AIChE J., 1971,
vol. 17, p. 856.
R (1) Reaction rate of ethylene partial oxidation;
15. Petrov, L., Eliyas, A., and Shopov, D., Appl. Catal.,
R (2) Reaction rate of ethylene total oxidation; 1986, vol. 24, p. 145.
R (3 ) Reaction rate of ethylene oxide total oxidation; 16. Petrov, L., Eliyas, A., and Shopov, D., Appl. Catal.,
1985, vol. 18, p. 87.
T Temperature; 17. Petrov, L., Eliyas, A., Maximov, C., and Shopov, D.,
Ea Activation energy; Appl. Catal., 1988, vol. 41, p. 23.
Reaction rate correction factor for reaction i; 18. Al-Saleh, M.A., Al-Ahmadi, M.S., and Shalabi, M.A.,
CFi Chem. Eng. J., 1988, vol. 37, p. 35.
ni Oxygen order in reaction i; 19. Borman, P.C. and Westerterp, K.R., Ind. Eng. Chem.
Adsorption constant for component j in reac- Res., 1995, vol. 34, p. 49.
K ij 20. Lafarga, D., Al-Juaied, M.A., Bondy, C.M., and Var-
tion i;
ma, A., Ind. Eng. Chem. Res., 2000, vol. 39, p. 2148.
DCE 1,2-Dichloroethane that is equal to ethylene;
21. Carucci, J.R.H., Halonen, V., Eränen, K., Wärnå, J.,
Dichloride (EDC) Ojala, S., Huuhtanen, M., Keiski, R., and Salmi, T.,
Ind. Eng. Chem. Res., 2010, vol. 49, p. 10897.
22. Salmi, T., Carucci, J.H., Roche, M., Eränen, K.,
Wärnå, J., and Murzin, D., Chem. Eng. Sci., 2013,
vol. 87, p. 306.

KINETICS AND CATALYSIS Vol. 61 No. 4 2020