Applied Catalysis, 67 (1991) 179-188 179

Elsevier Science Publishers B.V., Amsterdam

Influence of carbon dioxide addition upon decay of

activity of a potassium-promoted iron oxide catalyst for
dehydrogenation of ethylbenzene

Junshi Matsui*‘, Toshiaki Sodesawa and Fumio Nozaki

Graduate School of Science and Technology, Chiba University, Chiba 260 (Japan)

(Received 4 January 1990, revised manuscript received 8 June 1990)

Abstract

In the catalytic dehydrogenation of ethylbenzene to styrene, the detailed influence of the addition of
carbon dioxide upon the decay of activity of commercial potassium-promoted iron oxide catalyst has
been investigated in the presence of steam. In addition, the rates of potassium component losses from
the catalyst bed were experimentally determined by measuring potassium concentration in the layer of
condensed water from the effluent during the catalytic reaction. The close correlation between the rate
of potassium loss and decay of catalytic activity suggested that catalyst deactivation is mainly due to a
decrease in the amount of potassium ferrite (K,Fe,O,) which plays an important role, providing active
sites for the catalytic reaction of ethylbenzene dehydrogenation. Furthermore, the addition of small
amounts of carbon dioxide to reactant feed gas (in concentrations ranging from 0.1 to 0.5 mol-% of CO*/
total mol of feed) resulted in a slight decrease in the rate of styrene formation; however, it. was found to
be effective to a certain extent for depression of the decay of the catalyst’s activity.

Keywords: ethylbenzene dehydrogenation, iron-potassium, carbon dioxide addition, ageing, deactivation.

INTRODUCTION

In the industrial production of styrene, the process of ethylbenzene dehy-

drogenation is presently employed, using an adiabatic or isothermal reactor
and a potassium-promoted iron oxide catalyst. The dehydrogenation reaction
in the adiabatic reactor is usually practiced in the presence of excess steam
under the conditions in which steam/ethylbenzene ratio, temperature and
pressure are 7-12 mol/mol, 600-650°C and 0.5-1.0 atm respectively, at the
reactor inlet.
In the commercial potassium-promoted iron oxide catalyst the potassium
component is added in the form of carbonate or oxide and it greatly enhances
the activity of the iron oxide catalyst for ethylbenzene dehydrogenation in the

‘Present address: Technical and Engineering Department, Idemitsu Petrochemical Co., Ltd., l-l
Anesakikaigan, Ichihara, Chiba 299-01 Japan.

0166-9834/91/$03.50 0 1991 Elsevier Science Publishers B.V.

180

presence of steam. During the course of the catalytic reaction, however, the
potassium promoter migrates gradually from the catalyst surface towards the
center or it is scattered in the reactant fluid. The resulting shift of potassium
component leads to a decay of activity. The present authors [l] have already
reported that the shift of potassium component is the major cause of the decay
of activity. Published reports [l-4] have also pointed out that the potassium
component exists in the catalyst pellet as a variety of chemical species such as
potassium carbonate, oxide or ferrite; and the shift of potassium component
occurs with the migration of potassium oxide or hydroxide, in which the latter
(potassium hydroxide) is formed by the reaction of potassium oxide and steam.
It is known that the carbon dioxide, small amounts of which are formed as
a by-product in the ethylbenzene dehydrogenation, depresses the catalytic ac-
tivity [ 4,5]. In addition, it is known that the existence of carbon dioxide in the
reactant feed depresses the rate at which the potassium component migrates
due to the formation of potassium carbonate by the interaction of potassium
hydroxide with carbon dioxide [3]. Although carbon dioxide greatly affects
both the catalytic activity and the decay of activity, very few detailed investi-
gations on the addition of carbon dioxide have been published to date.
In the present study, the influence of the addition of carbon dioxide upon
the catalytic activity and decay of activity upon ethylbenzene dehydrogenation
in the presence of excess steam has been investigated in detail and the results
obtained are compared with those of our previous paper [ 11.

EXPERIMENTAL

Catalyst

The material used was commercial potassium-promoted iron oxide catalyst,

the potassium content of which is less than half that employed in the previous
work [ 11. The catalyst pellets were 3.2 mm in diameter, 4-6 mm in length and
were used without further crushing.
When the catalyst life was tested under the condition of carbon dioxide ad-
dition, the material used was same as the catalyst employed in the previous
report [ 11, in which case the catalyst contains about 25 wt.-% potassium
carbonate.

Apparatus and catalytic tests

The catalytic activity tests were carried out using the same apparatus as that
described in the previous report [ 11.
The fresh catalyst pellets (12 g) were packed into the apparatus and exper-
iments were carried out at temperatures of 600, 610, 620 and 640” C in the
presence of excess steam with an ethylbenzene feed rate of 130 g/h at a pres-
 181

sure of 0.75 atm (1 atmz101.325 kN/m*). Under these reaction conditions

the observed conversion of ethylbenzene was kept at such a low value that
differential reactor conditions may be assumed.
The experimental data in the case of carbon dioxide addition (in the con-
centration of 0.0-0.5 mol-% COJtotal feed) were obtained after the catalyst
activity had been allowed to attain a steady value (after a run continued for 40
h). In any of the experiments, regardless of the carbon dioxide addition, the
composition of the feedstock liquid was: ethylbenzene 100 mol-% and steam-
to-liquid ratio 11.2 mol/mol. The formation rates of styrene, toluene and ben-
zene were calculated from the composition of the oil layer in the condensed
effluent by the same procedure as described in the previous report [ 11.

Loss of potassium component

The rates of potassium component losses from the catalyst bed were mea-
sured under various reaction conditions. The loss rates (Ek in mg. h-l) were
obtained from the correlation E K=CK-Fw, where CK (mg-g-‘) is the concen-
tration of potassium in the effluent-condensed water layer and F, (geh-l ) is
the feed rate of water into reactor. In this case, the Ck values were determined
by atomic absorption spectrometry.

PLAUSIBLE MODEL AND RATE EQUATION

Hirano [ 4,5] investigated the characteristics of the deactivation of a potas-

sium-promoted iron oxide catalyst in the dehydrogenation of ethylbenzene to
styrene under added carbon dioxide and he pointed out that potassium ferrite,
which plays a role in the active sites, would be formed and extinguished by the
following scheme:
K,CO, *K,O+CO,

KzO + Fe, O3 +Kz Fe, 0, (1)

K,Fe, 0, + COP + Kz CO, + Fe, O3
In the presence of a large excess of steam, almost all the potassium oxides
would convert to potassium hydroxide by reaction with the steam. Since the
melting point of potassium hydroxide is 36O”C, it should be molten at the
reaction temperature. Both the potassium hydroxide and the potassium oxide
will form a molten and/or solid solution with the other compounds. The actual
amount of these potassium compounds present in the molten phase or in the
solid phase will probably be determined by a complicated equilibrium corre-
lation corresponding to the partial pressure of carbon dioxide. However, at a
low carbon dioxide concentration range in the reactant feed gas, the following
simple correlation may be assumed:
182

( [KIT - [KID)/ [KID =Kl ’ [co2l (2)

where KI represents an equilibrium constant and [K],, [KIT represent the
amount of potassium component corresponding to the extent of potassium
carbonate decomposition at any concentration carbon dioxide ( [CO,] ) in the
gas phase and [CO,] = 0, respectively.
As has been previously described, the loss of potassium component from the
catalyst occurs during the course of catalytic reaction. The loss is supported to
occur as a result of the vaporizing of potassium hydroxide from the mixed
molten phase, because potassium hydroxide is more volatile than the other
potassium compounds. Since the loss rate of total potassium component in the
catalyst will obey first-order kinetics, the loss rate of the potassium component
in potassium ferrite will also be first order, as eqn. (3) shows. This is because
both the fractional ratio of the amount of potassium ferrite and the other po-
tassium compounds to the total amount of potassium are kept at constant
values in the equilibrium state.
- [K]./dt=B. [K]r (3)
where t and B are Time On Stream (TOS) in h and the decreasing rate con-
stant of potassium component in h-l, respectively, and [K]r represents the
amount of potassium component in the potassium ferrite on catalyst surface.
On the other hand, a relationship between the amount of potassium ferrite
on catalyst surface and [K ] r is also presented to be expressed by eqn. (4), in
a similar manner as eqn. (2 ) .

[Sl*/( [Sl.o- [Sl,) =& - WIF (4)

where K, represents an equilibrium constant, and [S],, [S], represent the
amount of active sites on the catalyst surface corresponding to [K] r at TOS = t
and at TOS = 0, respectively.
From eqns. (3 ) and (4) in the previous report [ 11, the authors have shown
that the rate of deactivation with process time is represented by eqn. (5), as-
suming that the activity (reaction rate) is proportional to the amount of po-
tassium ferrite.
ln[a/(l-cr)] =B.t+C (5)
where a is the specific activity, being represented by the ratio of the styrene
formation rate (rst) at TOS= t to that at TOS ~0, and C is an integration
constant.
On the other hand, the coexistence of carbon dioxide in the reactant feed gas
will convert the molten potassium hydroxide to potassium carbonate. There-
fore, the decrease in the extent of potassium carbonate decomposition will de-
press the rate of potassium loss from the catalyst, although the numbers of
active sites are lowered.
 183

RESULTS AND DISCUSSION

Catalyst activity decay

As described in the previous section, it may plausibly be expected from the

model that the addition of carbon dioxide will tend to suppress the decay of
activity. In order to confirm this speculation, the activity test was carried out
in the presence of additional carbon dioxide (0.1 mol- % total feed) under the
same reaction conditions and the same catalyst as those described in the pre-
vious report [ 1 ] over a period of 800 h at a temperature of 650 oC. The result
is shown in Fig. 1, together with the data which have been described in the
previous report (the activity change with process time at a temperature of
650°C in the absence of carbon dioxide). Fig. 1 shows that the addition of
carbon dioxide markedly depresses the decay of activity, although it can not
completely inhibit the catalyst deactivation,
Eqn. (5) indicates that the relationship between TOS (t) and In [a/ (1- a) ]
will give a straight line. The plots of In [a/ ( 1 - a! ) ] vs. t are shown in Fig. 2. A

Fig. 1. Change in the rates of styrene formation with time on stream (TOS 1. Temperature: 65O”C,
CO, concentration in feed: 0.1 ( l ), 0.0 mol-% (0 ).

TOS /h

Fig. 2. Plots of In ( CY/( la) ] vs. TOS. Symbols are the same as in Fig. I
184

good linear correlation was also obtained in the case of carbon dioxide addition.
Thus, the slopes (B) of the straight lines were found to be 8.7~10-~ (with
carbon dioxide), and 9.7. 10V3 h- ’ (without carbon dioxide), respectively. The
value of B in the case of carbon dioxide addition was clearly less than that in
the absence of carbon dioxide.
Since the difference in the rate of potassium loss with and without carbon
dioxide will be closely related the amount of potassium component formed by
the decomposition of potassium carbonate ( [K],), the value of B should be
proportional to [K], and the amount of potassium component capable of
forming potassium ferrite ( [K] r) for small amount of additional carbon diox-
ide. From eqn. (4) it can also be presumed that the catalytic activity will be
proportional to [K]r. Therefore, these speculations suggest that ratio of the
slope (B) in Fig. 2 in the case of additional carbon dioxide to that without
carbon dioxide should be similar to the ratio in the catalytic activity (rSt in
mol*h-‘). Since these values are (rst=1.11~10-2 with carbon dioxide) and
rs,=1.26.10W2 (without carbon dioxide), the ratio of B values (8.7*10W3/
9.7.10W3=0.90) is nearly equal to that of the rst values (1.11.10W2/
1.26. 1O-2 = 0.88)) and thus the speculation mentioned above is revealed to be
almost valid.

Activity change in the case of carbon dioxide

In order to further clarify the way in which addition of carbon dioxide affects
the decay of activity, catalytic activity tests in the presence of additional car-
bon dioxide were carried separately out, using fresh catalyst. The changes in
the rates of styrene formation are shown in Fig. 3. At all reaction temperatures,
styrene formation rates were slightly reduced with an increase in carbon diox-
ide concentration and a slight lowering of toluene formation rate was also ob-

Fig. 3. Influence of CO, addition upon the rate of styrene formation. Temperature: 600 (0 ), 610
(@),620, (0),64O”C (m).
 185

served. However, the change in benzene formation was negligible small. These
observations were similar to those in the decay test of catalytic activity de-
scribed in our previous report [ 11, in which the experiments were carried out
without additional carbon dioxide.
During the course of the decrease in the amount of potassium component
with process time, the correlation between the amount of active sites ( [S],)
and the amount of potassium component in potassium ferrite ( [ K]r) is ex-
pressed by eqn. (4 ), as described in the previous section,

[Sl,/C[Sl,o - ISI,) =Kz. [KIF (4)

This correlation will be applicable to the decrease of [KIF by carbon dioxide
addition. In this case, [S], and [S] AOrepresent the amount of active sites at
[CO,] andat [CO,] =O, respectively. Then eqn. (6) may be derived from eqns.
(2) and (4).
[Sld[Sl~= [co,ll(K, +KI&WIT)+~ (6)
Being a convenient expression of the ratio ( aco2) of rSt in the case of carbon
dioxide addition to that in the case without carbon dioxide eqn. (6) indicates
that the relationship between l/a coZ and [ COJ will give a straight line, be-
cause K,, K2 and [KIT are constant. From the data shown in Fig. 3, the cor-
relation between l/cuco, and [CO,] are illustrated graphically in Fig. 4. Good
linear correlations are presented in Fig. 4, and it is therefore found that eqn.
(6) is capable of representing the effect of carbon dioxide addition upon the
catalyst activity. The slopes of the straight lines in Fig. 4 decrease with a low-
ering of reaction temperature, and the difference in the slopes between tem-
peratures lower than 610 0C and temperatures higher than 620 “C was signifi-
cantly large.
Arrhenius plots for the rates of styrene formation ( rSt ) are illustrated in Fig.
5 at varying carbon dioxide concentrations. The Arrhenius plots showed a sig-
nificant increase in the apparent activation energy (Ea) with an increase of
carbon dioxide concentration and with a lowering of reaction temperature. Such

4.0 I

CO2 concentration /mol-%

Fig. 4. Relationship between 1 Ic+~, and CO, concentration. Symbols are the same as in Fig. 3.
186

-6.01I. IO L ’
\

-1 -1
T x103 /K

Fig. 5. Arrhenius plots for rSt. CO, concentration in feed: 0.0 ( 0 ), 0.2 ( A ),0.5 mol-% ( q 1.

an increase in Ea was also observed in our previous work [ 11, comparing the
activation energy over deactivated catalyst with that over fresh catalyst. A
similar variation in Ea with the decrease of potassium content of potassium-
promoted iron catalyst has also been reported in the investigation of ammonia
synthesis [ 61.
A similar inflection as that seen in the Arrhenius plots of Fig. 5 may arise
through a change in the rate-determining step from pore diffusion to catalytic
surface reaction. However, the value of Ea in the absence of carbon dioxide
addition, i.e., 28 kcal/mol, is too large for a temperature dependency in the
rates of pore diffusion. Furthermore, Ea in the presence of additional carbon
dioxide, i.e., above 50 kcal/mol, is also too large for a temperature dependency
in the rates of catalytic surface reaction. These large temperature dependen-
cies may be due to a significant change in the numbers of active sites, corre-
sponding to the extent of potassium carbonate decomposition in the temper-
ature range of 600-640 oC.

Rate of potassium loss

In our previous report [ 11, we observed that the decay rate of activity (the
slopes of the straight lines, i.e. the decreasing rate constant of potassium com-
ponents B in Fig. 2 of the preceding paper [ 11, which are obtained from the
decay rate of the catalyst) increases with an elevation of reaction temperature.
Such a variation in the rate constants B suggests that the rate of the potassium
component loss from the catalyst bed will increase with the elevation of reac-
tion temperature. Thus, in order to experimentally elucidate this speculation,
the rate of the loss (EK) of potassium component from the catalyst bed was
measured at 620,640 and 660°C. The results are shown in Fig. 6, together with
the values of 23 reported in the preceding paper [ 11. From Fig. 6, it is found
that the temperature dependency of EK is quite similar to that of the decreasing
rate constant of potassium component (B ) . This similar temperature depend-
 187

-3.0
1.05 1.10
T-' X103 /K-l

Fig. 6. Arrhenius plots for the rate of potassium loss, EK and for the constant of eqn. (4), B.

2.0’ 8 b c n
40 50 60
&’ / mol-‘.g.h

Fig. 7. Relationship between l/rst and l/EK.

ency strongly suggests the following argument: the decreasing rate of active
sites is closely related to the rate of potassium compound loss and the kinetic
model for the decay of catalytic activity is plausible.
With varying amounts of carbon dioxide in the reactant feed gas, the change
of the amount of active sites ( [S],) with the extent of potassium carbonate
decomposition ( [K] n ) can be expressed by eqn. (7) derived from eqn. (4))
because [K] n will be proportional to [K] r.

where Ku is an equilibrium constant. Since the rate of potassium loss (Ek)

should be proportional to [K],, it can be expected from eqns. (7) that the
relationship between l/EK and l/rst will give a straight line over a range of
[ C02]. In the range of 0.0-0.5 mol-% CO2 at a temperature of 640’ C, the l/
EK vs. l/rst plot gives a good linear correlation as shown in Fig. 7.
188

CONCLUSION

The influence of the addition of carbon dioxide upon decay in catalytic ac-
tivity of commercial potassium-promoted iron oxide catalyst in the ethylben-
zene dehydrogenation reaction has been investigated. The main conclusion is
that carbon dioxide depresses both the rate of potassium promoter loss from
the catalyst pellet as well as the decay in catalytic activity. In addition, it has
been speculated that both the depression in the rate of potassium loss and the
lowering of catalytic activity by the addition of carbon dioxide may be attrib-
uted to the same mechanism; i.e. these phenomena are mainly due to the de-
crease in the amount of potassium ferrite forming active sites on the catalyst
surface. Furthermore, the results obtained in this study have been satisfactor-
ily explained on the basis of the deactivation kinetic model which was de-
scribed in our previous report.

REFERENCES

1 J. Matsui, T. Sodesawa and F. Nozaki, Appl. Catal., 51 (1989) 203.

2 B.D. Herzog and H.F. Rase, Ind. Eng. Chem. Res. Dev., 23 (1984) 187.
3 W.-D. Mross, Catal. Rev. -Sci. Eng., 25 (1983) 591.
4 T. Hirano, Appl. Catal., 26 (1986) 65.
5 T. Hirano, Appl. Catal., 26 (1986) 81.
6 H. Kobayashi and 0. Nishijima, Kogyo Kagaku Zasshi, 57 (1934) 189.