Ind. Eng. Chem. Prod. Res. Dev. 1985, 2 4 , 544-549
544
The small isomerization activity for ethylbenzene conversion over A1203+ mordenite (containing no Pt) was attributed to iron impurities in the mordenite. Treatment of mordenite by dealuminating the zeolite resulted in a "postulated" widening of the side pores. This resulted in a change in the selectivity ratio of xylenes/ benzene (derived from ethylbenzene) from 1.2 to 0.14 as C6 and/or C2 migration through the side channels became possible. Acknowledgment We wish to express appreciation to William Carr and Jim Kosco of Engelhard Industries for preparation of many of the catalysts tested. Robert M. Yarrington and Antonio Eleazar, also of Engelhard, are to be thanked for obtaining and calculating the kinetic data. Registry No. EtPh, 100-41-4; o-MeC6H,Me, 95-47-6; p MeC6H4Me,106-42-3;Me2C6H4,1330-20-7;Pt, 7440-06-4;faujasite,
12173-28-3; mordenite, 12173-98-7. Literature Cited Barrer, R. M.; Peterson, D. L. Proc. R . SOC. London, A 1964, 280, 466. Beecher, R.; Voorhies, A. Ind. Eng. Chem. Prod.Res. Dev. 1969, 8 , 366. Carr, W.; Polinski, L.; Hindin. S. G.; Kosko, J. U.S. Patent 4 128591 (to Engelhard Minerals and Chemical Co.), Dec 5, 1978. Corma, A.; Cortes, A.; Nebot, I.; Tomas, F. J . Catal. 1979, 57,444. Edison, R. R.; Boyum, A. A. Oil Gas J. 1979, 77, 140. Hersh, C. K. "Molecular Sieves"; Reinhold: New York, 1961; p 44. Meier, N. M. Z . Krista//ogr. 1961, 115,439. Minachev, K.; Garanin, V.; Isokova, T.; Kharlamov, V.; Bogomolob, V. Adv. Chem. Ser. 1971, No. 102, 441. Polinski, L. Ind. Eng. Chem. Prod. Res. Dev. 1972, 7 1 , 107. Satterfield, C. N. "Mass Transfer in Heterogeneous Catalysis"; MIT Press: Cambridge, MA, 1970; p 42. Setzer, H. J.; Eggen, A. C. W. U.S.Patent 3450500 (to United Aircraft Corp.), June 17, 1969. Voorhies, A,; Hatcher, W. J., Jr. Ind. Eng. Chem. Prod. Res. Dev. 1969, 8 , 361. Weisz, P. B. "Advances in Catalysis and Related Subjects"; Academic Press: New York, 1962; Vol. 13, p 137.
Receiued for reuiew January 30, 1985 Accepted May 31, 1985
Coke Tolerance of Catalytic Reforming Catalysts In-Slk Nam, John W. Eldrldge," and J. R. Kittrellt Department of Chemical Engineering, University of Massachusetts, Amherst, Massachusetts 0 1003
Bimetallic naphtha reforming catalysts are noteworthy in their ability to tolerate up to 20% coke before reaching end-of-run activity. A multilayer coke model is presented that adequately fits accelerated coking rate data on reforming catalysts. The model is used to interpret the possible effects of hydrogen partial pressure and catalyst compositional variables on the coke tolerance of t h e catalyst.
Catalytic reforming is one of the major refinery processes for upgrading gasoline octane numbers and for producing aromatic hydrocarbons. Reactions that produce the desirable aromatic and isoparaffinic products with high octane numbers include dehydrogenation, dehydrocyclization, and isomerization. Reforming reaction kinetics and mechanisms have been discussed in detail elsewhere (Smith, 1959; Mayers et al., 1961; Selman and Voorhies, 1975; DePauw and Froment, 1975). Older reformers were typically high-pressure units operated with a platinum on alumina catalyst. The development of new, stable multimetallic catalysts has permitted the use of significantly lower hydrogen partial pressure (Cecil et al., 1972). While the addition of rhenium does not greatly affect initial catalyst activity, it significantly enhances activity maintenance and increases the tolerance of the catlyst to high coke levels, thus permitting reduced operating pressure. Since Chevron (Kluksdahl, 1968) introduced the platinum-rhenium catalyst, Johnson and LeRoy (1974) have suggested that rhenium removes coking precursors by hydrogenation. This view is supported by Ermakov et al. (1977), whose work indicates that rhenium increases the degree of dispersion of platinum. This was deduced both from the 2- or %fold increase in the H2:Pt ratio attained by H2 adsorption and from the degree of dispersion re'KSE, Inc., Amherst, MA 01004. 0196-4321/85/1224-0544$01.50/0
vealed by X-ray studies of Pt and Re-Pt/Si02 catalysts. They also suggest that rhenium modifies the alumina support so as to maximize the platinum surface area or to resist coke accumulation. The sequential tests of Bertolacini and Pellet (1980) involving comparisons of the cosupported platinum-rhenium catalyst with a mixture or a layered two-catalyst system also indicated that rhenium removes coke precursors during the dehydrogenation reaction. The role of rhenium in the catalyst can thus perhaps be explained by a coking precursor removal mechanism with hydrogen. It is well-known that catalyst coking represents one important mechanism of deactivation for reforming catalysts. The coke formation may occur by several mechanisms, producing the various forms of coke broadly classified as noncatalytic and catalytic coke (Trimm, 1982). Catalytic coke formation on metals is quite complex. It can cause encapsulation of the metal or the growth of carbon deposits forming filamentous or platelet coke a t a metallic site. Although it is clear that the stability of reforming catalysts largely depends on the carbon deposition rate, it is not certain whether coke is formed on metallic sites, acid sites, or both. Figoli et al. (1982) observed an initial rapid deposition of coke that mainly affected the metallic function of the catalyst. When the deposition rate was slower, the deposition of coke was observed to affect the acidic function. The influence of the total pressure and hydrogen to hydrocarbon mole ratio has also been widely investigated, since hydrogen partial pressure is a key variable in 0 1985 American Chemical Society
Ind. Eng. Chem. Prod. Res. Dev., Vol. 24, No. 4, 1985 545
reforming processes (Mayers et al., 1961; Selman and Voorhies, 1975). In order to establish the role of the catalyst deactivation by coking, it is useful to examine kinetic models of the deactivation rate. However, a noteworthy characteristic of the bimetallic catalysts is their tolerance for higher catalyst coke levels (up to 20%), as distinguished from their attractive low rates of deactivation. Therefore, in the present work, we describe an approach that can help elucidate the reasons for this high coke tolerance, without emphasis on the many complex interactions which can influence the deactivation rate. It is hoped that these developments can provide a basis for further experimental investigations of the coke tolerance of candidate reforming catalysts, leading ultimately to better insight into the mechanism of the multimetallic catalysts. Simplified Coke-Activity Model for Reforming To illustrate the impact of coke on activity for reforming, let us simplify the complex reactions to a simple dehydrogenation reaction. In fact, more than 200 chemical components are involved in the reforming process. Ramage et al. (1980) and Zhorov et al. (1965) discussed lumping schemes for reforming kinetics to simplify the reaction model. For our simplified dehydrogenation reaction
A
B
+ nH2
(1)
If S represents an active site on the catalyst, the following sequence of surface steps could be envisioned: the reactant A, such as methylcyclopentane, adsorbs on the catalyst and is converted to the adsorbed product B, such as benzene, with the associated production of adsorbed hydrogen. These products are then desorbed from the catalyst, regenerating the active sites. In fact, there is a dual site functionality in catalytic reforming, and there are several molecules of hydrogen formed, depending upon the specific reactant molecule, A. Although data to be discussed later may require consideration of additional mechanistic steps, the level of detail provided by eq 2-7 is sufficient for the present illustrative purposes. If it is assumed that the coking reaction occurs by parallel kinetics, the overall reaction can be written in terms of presumed individual reaction steps:
A+S*AS AS + 2 s e A*S + 2HS
(2)
A*S e BS
(4)
BSeB+S
(5)
-
2HS mA*S
+ 2s coke1 + nH,
F?
Hz
(3)
(6)
(7)
In eq 3, the reactant A, or relevant compounds contained within the feed naphtha, are converted to a reactive intermediate adsorbed on the surface, A*S. As discussed by Mayers et al. (1961), this could be cyclopentadiene formed by dehydrogenation of cyclopentane. Bertolacini and Pellet (1980), for example, have shown that naphtha spiked with cyclopentane or naphthalene rapidly deactivated a platinum catalyst but that the cyclopentane-spiked feed did not increase the deactivation of the platinum-rhenium catalyst. Therefore, rhenium could be preventing coke formation caused by cyclopentane. In the above scheme, this reactive intermediate, A*S, not only produces a final product, B, by eq 4 but also forms coke by eq 7. If eq 4 were the controlling reaction for the overall conversion of naphtha and eq 7 were the rate-controlling step for deactivation, a pseudoequilibrium relationship
would govern the concentration of coke precursor on the surface from eq 3:
The surface hydrogen concentration, CHS, would be similarly governed by a pseudoequilibrium relationship through eq 6, leading to (9) Hence, as is well-known in platinum reforming, an increase in the hydrogen partial pressure will reduce the coke precursor concentration on the surface and thus the deactivation rate. Similarly, if platinum-rhenium were to increase the adsorption of hydrogen on the surface (Ermakov et al., 1977) represented by an increased KH2or were to reduce the magnitude of KA*,a reduction in CA*swould also occur. It will be shown below that not only do these effects reduce the deactivation rate, but they also increase the coke tolerahce of the catalyst. That is, they permit the catalyst activity to be maintained at higher coke levels on the catalyst. In the mechanism suggested above, it is seen that any such reduction of CA*Swill also decrease the rate of the controlling reaction, eq 4, for the production of B. However, since that reaction is of different order than the coke formation reaction, eq 7, its rate will be reduced much less than the deactivation rate. Note, however, that we propose these effects to be qualitative in nature a t this stage, for the model reactions of eq 1-7 are highly simplified. In particular, we have not included another possible mechanism for the removal of A*S from the surface, the hydrogenolysis of the coke precursor, which may be very important for the bimetallic catalysts (Bertolacini and Pellet, 1980). We also recognize the qualitative nature of the assumption of ideal thermodynamics reflected in eq 8 and 9 by the absence of activity coefficients. To examine the relationship between activity and catalyst coke content, the approach of Nam and Kittrell (1984) is to be used. This approach has been successful in describing coking in the dehydrogenation of l-butene to butadiene over a Cr2O3-A1,O3 catalyst, in the hydrogenation of isobutylene over a nickel on silica-alumina catalyst, and in ethylene reactions over a silica-alumina catalyst . Accordingly, the total number of active sites is distributed among vacant sites, those covered by adsorbed A, B, hydrogen, or reactive intermediate, and sites inactivated by coke. Thus =
cs + CAS + CBS + CHS + CA*S + CCS
(10) where Ccs is the concentration of active sites covered by monolayer coke on the surface. The catalyst activity is defined as the fraction of sites not irreversibly inactivated by monolayer coke (cS)O
Substituting the pseudoequilibrium assumptions for eq 2, 3, 5, and 6 into eq 10 with the activity definition, eq 11, we obtain
546
Ind. Eng. Chem. Prod.
Res. Dev., Vol. 24,
No. 4, 1985
In order to examine the dominant role of hydrogen pressure in the reforming process, let us simplify eq 12 by assuming (KH2PH,)1/2 is the dominant term in the denominator of eq 12. Hence the active site balance becomes
coke content and activity, it is difficult to use the rate of coke formation. Since a relationship between catalyst activity and total coke content is desired, the division of eq 17 by 20 yields aa -_ a4
Note that the other terms in the denominator could also be retained, in order to describe hydrogen to hydrocarbon ratio, for example, but the emphasis here is on hydrogen partial pressure effects. Now, the coke precursor, such as cyclopentadiene, is viewed to be an adsorbed component that is converted to coke by an irreversible reaction, the rate of which is proportional to the surface concentration of the coke precursor:
-Pck&A*KAa k&A*K,(Cs)oa -I- ~ L K $ H ~ [ ~ ( K H ~ P H , ) ~ / ~ ] ( C-Sa))O ( ~ (21) Equation 21 can be integrated for high excess H2 in the feed (PH2 can be treated as a constant) to provide q = al(l - a) - a2 In a (22) where
(Cdo “2
a1=--
Pc
The parameter p e is the stoichiometric factor converting the weight of monolayer coke produced to the number of sites inactivated, Ccs = pcCc (Nam and Kittrell, 1984). Pseudoequilibrium assumptions for eq 2 and 3 give
and from eq 13
With the definition of activity by eq 11
Note that the above derivations are based upon a monolayer coke concept due to the Langmiur assumptions of eq 2-7. Indeed, the monolayer coke cannot explain the high observed levels of coke on spent catalysts. For actual reforming operations, the Pt catalyst can exhibit 5 wt 7’0 coke. In view of this, Nam and Kittrell(1984) successfully employed a multilayer coke concept to describe the high levels of coke on catalysts associated with dehydrogenation and cracking reactions. Following their development, multilayer coke is deposited at a rate
Note that, as discussed in detail by Nam and Kittrell (1984), the multilayer coke is formed from monolayer coke in combination with additional reactant, A. The total coke content of the catalyst becomes q = cc
+ CL
a2
(cs)o kLKH2pH2[1 + = P C
(KH2PH2)1’21
~DKA*KA
(24)
The resulting eq 22 does not directly reflect the rate of coke formation on the catalyst but instead shows the tolerance of catalyst activity to total coke content of the catalyst. Note that the parameters a1and a2depend upon both temperature and hydrogen partial pressure, with the simplifying assumption of eq 13. In general, lower hydrogen pressures will tend to reduce the size of cy2 relative to al.Hence, accelerated deactivation tests in the laboratory using low hydrogen pressure might be expected to provide a linear coke vs. activity relationship, which would not be applicable at higher hydrogen pressures used commercially. The temperature dependence of a1 and a2 will be a complex function of the activation energies of kL and kD and of the heat of adsorption reflected in KHpand KA. Nam and Kittrell (1984) observed that the activation energies more commonly dominate the cy2 temperature dependence. Application to Reforming Catalyst Deactivation Data Cooper and Trimm (1980) presented data on the rate of coking vs. catalyst coke content, which can be used to test the utility of eq 22. To do so, the catalyst activity must be related to the coking rate. Catalyst activity can be calculated on the basis of the coking rate observed by Cooper and Trimm for n-hexane reforming on 0.3 wt 5% Pt-A1203 catalyst. The initial coking rate, producing only monolayer coke, can be deduced from eq 20 with a = l:
(19)
and the rate of total coke formation is obtained from eq 11, 16, 18, and 19
Since a relationship between activity and the rate of total coke formation is desired, the division of eq 20 by 25 yields
-rq-
(20)
Note that the qualitative effect of PH2 on the rate of coke formation is as expected in reducing the rate of coke formation. However, the coke formation rate is also dependent upon the prior history of the catalyst through its activity, a. In the examination of the effect of hydrogen partial pressure and the interaction of bimetallic catalyst
-a+
kLKH2PH2[1+(KH2pH2)”21
(1 - a ) (26) (rc)o ~DKA*KA where rq is the total coking rate at any time and (rc)ois the initial coking rate. Note that if the coefficient of (1 - a) exceeds unity (e.g., due to a high kL and/or high PHJ, the total coking rate will be greater than the initial (monolayer) coking rate. Under these conditions, the rate of total coke formation will increase as the activity de-
Ind. Eng.
REACTION TEMPERATURE 550°C
Chem. Prod. Res. Dev., Vol.
24, No. 4, 1985
547
Table I. Parameter Estimates with 95% Confidence Intervals reaction temp, "C parameter" lower limit final value upper limit
A 500'C 450T
450
a1 ff2
(rdo 500
ffl a2
5.9500 0.1050 0.0918 2.3198 0.8093 0.3041 1.7867 0.6808 0.5757
0.2100
(rd0 550
1.6500 0.0800 0.0785 1.7800
ffl
a2
(rdo
0.1800 1.5200 0.4750 0.3500
11.2100 0.5980 0.1051 4.6000 1.8600 0.5100 2.8500 1.6500
0.7100
"Units aland a2are expressed in mg of coke/g of catalyst; (rc)o is expressed in mg of coke/g of catalyst, min.
-
0-
I
9 5 % CONFIDENCE INTERVAL OF PARAMETER ESTIMATION
-I I
1
I
I
I
01
0 2
03
04
05
CATALYST COKE CONTENT (Wt%)
Figure 1. Comparison of theoretical predictions by eq 29 and experimental observations of coking by Cooper and Trimm (1980).
creases and the total coke content increases with time. The occurrence of this situation is easy to visualize from a physical standpoint. However, all the data used in this paper, which came from Cooper and Trimm (1980), was obtained under conditions for which rq/(rc)o< 1, so the rate of coke formation decreased as coke accumulated. With the definitions of a1 and cy2 in eq 23 and 24
W
8z
2-
a W
>.
4
10.8-
5
0.6-
2 8
0.4-
U
W
t;
0.2-
H
d
0.1-
1
000 006
Rearranging eq 27, catalyst activity becomes
I
I
I
+
x 103
I
(OK)
Figure 2. Temperature dependence of parameter a2 for eq 29.
Note that the activity is not simply the ratio of the coking rates unless a2is very small (no multilayer coking rate, kL = 0). Substituting eq 28 into eq 22 provides
Equation 29 is an implicit relationship between coking rate and catalyst coke content based upon the concept of multilayer coke formation. The platinum reforming data by Cooper and Trimm (1980) for n-hexane as a feed stream were fitted by nonlinear least squares to evaluate the parameters, al,cy2, and (rc)oof eq 29. The estimates of these parameters with 95% confidence intervals are listed in Table I. Figure 1shows the results of this effort with the best values of al,a2,and (rc)ofor each of these data sets, which were collected a t three different reaction temperatures. The symbols represent the original experimental data, and the lines represent the fit of eq 29. Whereas Cooper and Trimm (1980) used several empirical equations to represent these data,
the single model of eq 29, based upon multilayer coking kinetics, provides quite a satisfactory representation of all these data. It is interesting that the data of Figure 1 are nonlinear a t high temperatures but linear at the low temperature. In order to explain this behavior, the temperature dependence of parameter a2in eq 29 is presented in Figure 2. Because of the difficulties of parameter estimation with limited experimental data, the confidence intervals on a2 are quite large. Plotting the 95% confidence intervals on a2 for different reaction temperatures provides a temperature dependence compatible with the traditional Arrhenius behavior, as shown in Figure 2. The contribution of cy2 to eq 29 is negligible at 450 "C. However, the a2 parameter becomes important at higher reaction temperatures. This appears to cause the varying degree of linearity of the coking rate in Figure 1,as well as the increasing coking rate as the temperature is increased. The results discussed herein also support the use of the multilayer coke model, eq 22, on which eq 29 is based, to evaluate the effects of total coke content on catalyst activity for the reforming process. Discussion Equation 22 can also be used to examine the influence of hydrogen pressure on the relationship between the amount of coke and the catalyst activity. Thus it can
548
Ind. Eng. Chem. Prod. Res. Dev., Vol. 24, No. 4, 1985
H, PARTIAL PRESSURE (atmi
A COOPER
2
a
3
TR~MM
4
CATALYST COKE CONTENT ( W t Yo)
Figure 3. Predicted effect of H2pressure on P t reforming catalyst activity for constant KH2 (670 atm-') and KA.at 500 O C . Data from Cooper and Trimm (1980).
provide information for (1) interpreting accelerated laboratory deactivation tests in terms of commercial reactor behavior, thereby improving deactivation modeling capability, (2) predicting the operating time vs. activity decay in laboratory and industrial reactors, (3) examining the possible mechanistic effects of bimetallic catalysts through the influence of coke tolerance of the catalyst, and (4) understanding coking mechanisms involving coking precursor formation during reaction. in Figure 3 is shown a typical impact of hydrogen pressure on catalyst activity and coke content for a platinum reforming catalyst. The symbols in Figure 3 represent the laboratory data on the platinum on alumina catalyst of Figure 1 a t 500 "C,taken a t 5 atm hydrogen pressure. The line through these data represents the prediction of eq 22 with the parameters as defined in eq 23 and 24 and selected from Table I. A value of KH1= 670 atm-' was selected to represent these experimental data on the basis of H2 adsorption measurements by Harris (1962). For platinum-type catalyst, commercial reformers are typically operated a t about 30 atm. The relationship a t higher press&es between catalyst activity and its tolerance to various levels of catalyst coke content is shown in Figure 3, as predicted by eq 22 with parameters selected from Table I. It is widely recognized that low-pressure laboratory deactivation tests can be used to accelerate the rate of catalyst deactivation. However, as shown in Figure 3, these accelerated tests also influence the tolerance of the catalyst activity to any given level of catalyst coke. At 5 atm, the catalyst activity decreased nearly to zero when the catalyst contained about 1% coke for the platinum catalyst. However, for commercial conditions on Pt catalysts at 30 atm pressure, the catalyst still retains 50% of its original activity with 1% coke on the catalyst. Typically, commercial platinum reforming catalysts contain up to about 5 wt % coke by the end of a run. On the basis of the 30 atm curve of Figure 3, this is not surprising. Note also that the well-known increase in coking rate with low H2 pressure is predicted through eq 20 by the term for monolayer coke. However, Figure 3 does not involve this higher coking rate, only the above-discussed tolerance of the catalyst for any given total coke level. One of the theories for the role of the bimetallic catalyst is that the rhenium increases the hydrogen adsorption capacity of the catalyst relative to platinum catalysts
1
2
3
4
CATALYST COKE CONTENT (Wt %)
Figure 4. Possible effect of KH2on bimetallic catalyst activity tolerance to coke content for constant H2pressure (10 atm) at 500 "C.
(Ermakov et al., 1977). An increase in K H 2by a factor of 2-3 would seem plausible, but this would appear to be too small an increase to produce such profound improvements in operating behavior. In Figure 4 is shown the tolerance of catalyst activity to catalyst coke at pressures typical of commercial operation of bimetallic catalysts (10 atm), as calculated via eq 22. Values of KH2are also shown, ranging from 670 atm-', a typical value for a platinum catalyst, up to 2100 atm-', which might represent a bimetallic catalyst. However, it appears that a 3-fold increase in hydrogen adsorption capacity is not sufficient to permit the typical coke tolerance of the bimetallic catalysts, which can reach 20 wt % by the end of a run. Therefore, another consideration is necessary for eq 22 to describe the nature of the bimetallic reforming catalysts. A bimetallic reforming catalyst may also have a different value of KA* in eq 8 from that of a platinum catalyst. In fact, Bertolacini and Pellet (1980) reported that platinum-rhenium catalyst may remove coke precursors more effectively than platinum catalyst. Therefore, it is quite possible that a bimetallic catalyst has a small KA* compared to a platinum catalyst. A decrease in KAt by a factor of 3-10 would appear to be quite reasonable because a 10-fold increase in catalyst activity from physicochemical modifications of catalysts is not unachievable. For example, the addition of copper ions to hydrogen mordenite increases catalyst activity more than 10 times, as shown by Nam (1983). Also, the effect need not be due to the value of KA* as defined in eq 8. It is more likely that this is achieved by a hydrogenolysis or hydrogenation reaction that removes A* (e.g., cyclopentadiene) from the surface. In this case, the relevant reactions of A* become AS + 2 s A*S + 2HS (3) HS + A*S e D + 2 s (30) A*S .= BS (4) mA*S e coke1 + nHz (7) The reaction of eq 30 for a bimetallic catalyst can substantially reduce the surface concentration of A* relative to that on a platinum catalyst. A decrease in KA* in eq 8 will produce the same mathematical effect in the model of eq 22. Although a pseudo steady-state hypothesis would be involved to derive a new relationship for the surface concentration of A* from the hydrogenolysis mechanism, it cannot be tested by the available data and thus serves no purpose here. Hence, only the examination of the effect
Ind. Eng. Chem. Prod. Res. Dev., Vol. 24, No. 4, 1985 549
L
,
CATALYST COKE CONTENT (Wt 70)
Figure 5. Effect o f KA. on bimetallic catalyst activity tolerance to coke content for constant H2pressure (10 atm) and KHZ(2100 atm-') at 500 O C . Curves were calculated from eq 22.
of KA* on the surface concentration of A* will be presented below; it should be clearly recognized that the results are intended to portray those of a class of mechanisms and not those due to equilibrium arguments alone. In Figure 5, based upon eq 22, is shown the tolerance of catalyst activity to catalyst coke content a t 10 atm pressure with K H 2= 2100 atm-', typical of operating conditions and characteristics for a commercial bimetallic reforming catalyst. Values of KA*are changed to 1/3, 1/6, and l/lo of the KA* for the platinum catalyst. It is apparent that a 10-fold decrease in KA* is sufficient to increase markedly the coke tolerance of the catalyst, as well as to reduce its coking rate. Indeed, note from Figure 3 that the platinum catalyst activity at 30 atm is reduced to about 8% of the fresh catalyst activity at a 5 wt % coke content. From Figure 5, the bimetallic catalyst activity is not reduced to 8% of the fresh catalyst activity until 15 wt % coke a t 10 atm hydrogen pressure and 1/3 of the platinum-based KA*. It still retains 30% of the fresh catalyst activity a t that level of coke for the platinum-based
K**. Conclusions The mechanistic model of coke formation through eq 2-7 results in the relationships between catalyst activity and coke content and between coking rate and coke content expressed in eq 22 and 29. Such models may be useful for interpreting accelerated deactivation tests and for predicting useful catalyst operating lives. These models are qualitatively descriptive of the deactivation behavior of platinum and bimetallic catalyst functionalities (Ermakov et al., 1977; Bertolacini and Pellet, 1980). Furthermore, as shown in Figures 3-5, the models are descriptive of the higher coke tolerance of the bimetallic catalysts. However, we do not yet have sufficient data on the bimetallic catalysts, analogous to the Pt catalyst data base for Figure 1, to permit reconciliation of the theory underlying Figure 5 with experimental observations for bimetallic reforming catalysts. It is quite possible that when these experimental data are available further refinements in the mechanistic model of eq 2-7 will be required, such
as the addition of a hydrogenolysis step for the removal of the coke precursor from the catalyst surface. Nomenclature A, B, C, D = reactant, product, coke, and product intermediate, respectively A* = reactant intermediate a = catalyst activity CAS, CBS, CAIS,CHS = concentration of active sites occupied by A, B, A* and hydrogen, respectively, mol/g of catalyst CC = monolayer coke content of catalyst, mg of coke/g of catalyst Ccs = concentration of sites deactivated by monolayer coke, mol/g of catalsyt CL = multilayer coke content of catalyst, mg of coke/g of catalyst Cs = concentration of unoccupied active sites at any time, mol/g of catalyst (CS)= ~ concentration of fresh catalyst active sites, mol/g of catalyst KA, KB, KH2 = equilibrium adsorption constants of A, B, and hydrogen, respectively, atm-I KA* = dimensionless equilibrium constant for eq 3 kD = rate constant of coking by monolayer coke, min-' kL = rate constant of coking by multilayer coke, min-' atm-' PA,PB,.PH~= partial pressure of A, B, and hydrogen, respectively, atm q = total coke content of catalyst, mg of coke/g of catalyst (rc)o = rate of monolayer coke formation, mg of coke/g of catalyst, min rq = rate of total coke formation, mg of coke/g of catalyst, min S = active site t = operation time, h Greek Letters a', a2 = deactivation constants, mg of coke/g of catalyst pc = stoichiometry factor relating mol of sites deactivated by monolayer coke to mol of monolayer coke per g of catalyst Literature Cited Bertolacini, R. J.; Pellet, R. J. "Proceedlngs of the International Symposium, Catalyst Deactivation"; Elsevier: Amsterdam, 1980;p 73. Cecil, R. R.; Kmak, W. S.; Sinfelt, J. H.; Chambers, L. W. Oil Gas J . 1972, 70(32),50. Cooper, B. J.; Trimm, D. L. "Proceedings of the International Symposium, Catalyst Deactivation"; Elsevier: Amsterdam, 1980;p 63. DePauw, R. P.; Froment, G. F. Chem. Eng. Sci. 1975,30,789. Ermakov, Y. I.; Kuznetsov, B. N. Kinet. Katal. 1977, 18, 955. Figoil, N. S.; Beltramini, J. N.; Barra, A. F.; Martineili, E. E.; Sad, M. R.; Parera, J. M. ACS Symp. ser. 1982,No. 202,239. Harris, G.W. "Kinetic Study of the Hydrogenation of Benzene on a PiatlnumAlumina Catalyst", Chevron Report, 1962. Johnson, F. L.; LeRoy, V. M. J . Catal. 1974,35,434. Kluksdahi, H. E. U.S. Patent 3415737,1968. Mayers, C. G.; Lang, W. H.; Weisz, P. B. Ind. Eng. Chem. 1961,53, 4. Nam, In-Sik "Experimental Study and Theoretical Modeling of Catalyst Deactivation", Ph.D. Thesis, University of Massachusetts, Amherst, MA,
1983. Nam, In-Sik; Kittrell, J. R. Ind. Eng. Chem. Process Des. Dev. 1984,23,
242. Ramage, M. P.; Grazlani, K. R.; Krambeck, F. J. Chem. Eng. Sci. 1980,35,
41. Selman, D. M.; Voorhles, A., Jr. Ind. Eng. Chem. Prod. Res. Dev. 1975,
14, 118. Smith, R. B. Chem. Eng. Pfog. 1959,55(6),76. Trimm, D. L. "Progress in Catalyst Deactivation";Martinus Nijhoff: The Hague, 1980;"NATO AS1 Serles E: Applied Sciences", p 3. Zhorov, Y. M.; Panchenkov, G. M.; Zel'tser, S. P.; Tirak'yan, Y. A. Kin. Katal. 1965,6 , 986.
Received for review March 12, 1985 Accepted July 1,1985