Manuscript for the special IECR issue CREVIII conference held in Barga, Italy on June 24-29, 2001

Catalytic Cracking of Cumene in a Riser Simulator

S.Al-Khattaf (2) and H. de Lasa (1.*) (1) Chemical Reactor Engineering Centre, Faculty of Engineering Science, University of Western Ontario, London, Ontario, Canada N6A 5B9 (2) Department of Chemical Engineering, King Fahd University of Petroleum & Minerals, Dhahran 31261, Saudi Arabia. (*) Corresponding author

A catalyst activity decay model

Abstract
The present study reports the catalytic cracking of cumene using a novel CREC Riser Simulator. Two sizes of zeolite crystals (0.4 and 0.9 m) are employed in the experiments. The goal is to develop and test, under relevant reaction conditions for FCC (temperature, contact time, reactant partial pressure, catalyst /oil ratio), a new catalytic activity decay model based on reactant conversion. Kinetic and decay model parameters are estimated using regression analysis. Activation energies, deactivation constants, and Arrenhius pre-exponential constants are calculated with their respective confidence intervals. The lack of transport constrains of cumene molecules, while evolving inside the zeolites, is confirmed evaluating a modified Thiele modulus and effectiveness factors.

Introduction
Catalyst deactivation in FCC can take place due to several factors. Catalyst pellets can lose its shape or mass due to attrition and high temperatures (sintering). Catalyst poisoning can also take place given the effect of impurities contained in the feedstock. Typical contaminants are hydrocarbons containing S, N, O, Ni and V [1-3]. However, under normal FCC conditions coke is the most important factor affecting catalyst activity. As catalytic reactions progress coke deposits on the catalyst surface, covering active sites and leading to catalyst decay. In this respect, important modeling efforts have been addressed to the modeling of catalyst decay due to coke. A first approach towards modeling catalyst deactivation is based on a time-on-stream decay model (TOS). This empirical model assumes that the coking rate is independent of reactant composition, extent of conversion, and hydrocarbon space velocity [4]. In this respect, classical decay models include an exponential decay law, =exp (-t) [5] and a power decay

law = t-n . More recent contributions consider modified exponential functions, =exp(-tn) with n=0.4 [6] and n=0.5 [6]. Regarding TOS models, Szepe and Levenspiel [8] showed that decay functions can be derived by assuming that the rate of activity decay is a function of the number of active sites involved in the deactivation event. In this respect, d /dt = - kd n and d /dt = - kd1 C2go were suggested[8]. Wojciechowski [9] considered the integration of d /dt = - kd n and this yielded = [1+ kdt(n-1)]-N. A second approach to catalyst decay involves a catalyst activity decay function related to the coke content. With this end, several relations were reported such as = exp (-Cc), =1/(1+Cc), =1/(1+Cc)2, =1-Cc, =(1-Cc)2 [10,11]. With the same objective, there are a number of two parameter models based on coke concentration as follows: a) =1/(1+Ccn) [12], b) =R + (1-R)exp(-Cc) [13], c) = (B+1)/(B+exp(Cc) [14] d) d/dCc = F + G (1) [15]. Corma et al.[16], studying catalytic cracking of paraffins, argued about interrelating activity decay, coke on catalyst and reactant conversion using : a) d/dt = -kmd m Ci, b) dCc/dt = r0c which is equivalent for m=2 and a first order rate of coke formation to = exp (-Cc). In conclusion, while the mechanism of coke formation is not well understood yet and there is still debate about suitable decay models, the present study proposes, a sound approach for catalyst deactivation with a decay function based on reactant conversion. This decay model is tested, in the present study, in the context of the catalytic cracking of cumene in a novel CREC Riser Simulator

Modeling Cumene Cracking in the CREC Riser Simulator

Cumene cracking is a first order reaction [17]. Its disappearance in the Riser Simulator can be represented by the following species balance equation;

V dCA = k in in C A Wc dt

(1)

where CA represents cumene concentration in the Riser Simulator (Kmole/m3), V the volume of the riser (45.10-6 m3 ), Wc the mass of catalyst (0.81.10-3 gcat), t the reaction time in seconds, the effectiveness factor , in the intrinsic deactivation function, kin the intrinsic rate constant in m3/Kgcat.s.

Moreover, the concentration of cumene, CA can be expressed by the following relationship;

y W C A = A hc MWA V ( 2)

where yA represents the cumene mass fraction, Whc the total mass of hydrocarbons inside the riser (0.16.10-3 Kg), MWA the cumene molecular weight (120 Kg/Kmole). Furthermore, kin can be expressed using an Arrhenius relation as kin=k0 exp(-ER/RT). Then, eq (1) can be written as follows;

E V dy A = k o exp( R ) in y A Wc dt RT

(3)

In addition, and in order to reduce parameter cross-correlation it is frequent to adopt the the following form of eq(3):

E 1 1 V dy A = k'o exp[ R ( )] in y A Wc dt R T To

(4)

where To represents the average temperature used in the reaction experiments. It can be observed that eqs (1) and (4) account for the possible diffusion transport limitations on the cracking reaction. In this respect, an effectiveness factor () has to be used to describe the effect of diffusion on the cracking reaction [17]. Moreover and to rigorously model this problem an effectiveness factor for an FCC catalyst can be evaluated as follows [18]:

tanh(h ' ) h'

(5)

where h is the modified Thiele modulus, which is defined as;

n 1

(n +1)kincin C A h' = aext 2Deff 1

(6)

where c represents the catalyst density in Kg/m3, n the reaction order set at one for cumene cracking, kin the intrinsic kinetic rate constant in m3/Kgcat.sec, in the intrinsic deactivation function, aext the catalyst external surface area in 1/m. However, and for the case of catalytic cracking of cumene some important clarifications have to be introduced in eq(6) [19]. One important consideration is given by the fact that for a first order reaction the Thiele modulus becomes independent of the reactant concentration. Another important consideration concerns the characteristic parameter to be considered to calculate aext. Given that essentially all the cracking reactions take place in the zeolite crystal with little influence of the matrix, the aext should be defined using the characteristic dimension of the crystal, L. Thus, approximating the crystal geometry with an equivalent sphere aext = L/6 and eq(6) becomes,

h' =

L k in in c 6 D eff

(7)

with, c = zeolite density (825 Kg/m3) Regarding the various kinetic parameters contained in eq(7), one very relevant parameter is kin. This so-called, intrinsic kinetic constant, could be hypothesized to change with temperature following an Arrhenius expression, kin =k0 exp (-ER/RT), where k0 and ER represent the frequency factors and the energy of activation of the intrinsic cumene cracking respectively. Diffusivity in zeolites can be represented by the Eyring Equation as follows:
E D ) RT

D eff = D o exp(

(8)

where ED is the diffusion activation energy. ED is estimated to be around 26 KJ/mole and D0 is approximately 3.2x10-10 m2/sec [20]. It has to be stressed that this estimation is based on diffusivity measurements in faujasites performed with aromatic species with a molecular size in the cumene range. For example, the value assigned to cumene falls in an intermediate range with respect to the values measured for xylene and 1,3,5 tri-iso-methylbenzene [21].

One important matter while accounting for catalyst deactivation is to consider in eq(7) a deactivation function, in. One classical approach, while describing catalyst decay is to consider catalyst decay as a function of time-on-stream:

= exp(t )
on-stream).

(9)

where is a constant and t is the time the catalyst is exposed to a reactant atmosphere (time

By substituting eq(9) into eq (4), setting =1 and integrating the resulting equation, the following is obtained:

Wck'o exp( yA = exp[

ER 1 1 ( )) R T To (1exp( t))] V

(10 )

In addition, it is demonstrated in the present study, that a catalyst activity decay function can be conveniently expressed as a function of converted reactant. First, coke formation can be postulated to take place as proportion of hydrocarbon conversion. This approach is a sound one if one considers that coke is mainly the result of the primary cracking reaction steps. This means that eq(3) can be modified to express the rate of coke formation using a stoichiometric number, c and the corresponding molecular weight ratio as follows:
dCc dt = ( c ky A Wc MWc V MWA ) (11)

where k is the reaction rate constant of cumene cracking and Cc the coke mass fraction based on the catalyst weight. Eq(11) can be divided by eq (3) and it results:
dC c dy A = c Whc MWc Wc MWA (12)

Since the right side of eq(12) is assumed as constant, then eq (12) can be written as,

dC c =A dy A
where A lumps a group of constant parameters.

(13)

By integrating equation (13) between 0 and CC and between 1 and yA, the following equation is obtained:

C c = A(1 y A )
with
W MWc A = c hc Wc MWA

(14)

(15)

Once this first step of relating the conversion of reactant to the coke fraction is established, the following step in the analysis is to consider a catalyst activity decay function based on coke concentration:
= exp(Cc ) (16)

However, given the relationship between the coke concentration on catalyst, Cc and the
weight fraction of cumene, yA as given by eq(14) the following equation is obtained:

= exp((1yA))
where = A

(17)

This result can be substituted in eq(4). Then, setting =1, the following equation, describing the rate of reactant consumption is obtained;

E 1 1 V dyA = k'o exp( R [ ])exp([1yA])yA Wc dt R T To

(18)

Thus, the catalyst activity decay model, based on reactant conversion, given by eq (18) involves three parameters k0, EA, and . These three parameters can be defined using a number of experiments in the Riser Simulator as it will discussed in the upcoming experimental section.

Experimental Procedures
Cumene was employed, in the present study, as model reactant compound. Regarding the FCC catalysts, commercially available Y-zeolites, small and large crystals, manufactured by the Tosoh company were used. Properties of these zeolites are reported in Table 1. These Na-zeolites were ion exchanged with NH4NO3 to replace the sodium cation with NH4 . Following this, NH3 was removed and the H form of the zeolites was spray-dried using kaolin as the filler and silica sol as the binder. The resulting 60-m catalyst particles had the following composition: 30 wt% zeolite, 50 wt% kaoline, and 20 wt% silica sol. The process of sodium removal was repeated for the pelletized catalyst. Following this the catalyst was calcined during 2 hr at 600C. Finally, the fluidizable catalyst particles (60-m average size) were treated with 100% steam at 760C for 5 hr. Table 2 reports the catalyst main properties following catalyst pretreatment. The unit cell size was determined by X-ray diffraction following ASTM D-3942-80. Surface area was measured using the BET method. It can be observed that the initial zeolite unit cell size of 24.5 was reduced, following steaming, for both zeolites, to around 24.28. The catalytic activity of the Y-zeolite catalyst, prepared using the various techniques described above, was measured in a Riser Simulator using cumene as the hydrocarbon feedstocks. The Riser Simulator is a novel bench scale with internal recycle unit. This reactor, invented by de Lasa [22] overcomes the technical problems of the standard micro-activity test (MAT). The catalytic conversion of cumene is modeled in the present study. The Riser Simulator consists of two outer shells, lower section and upper section that permits to load or to unload the catalyst easily. This reactor was designed in such way that an annular space is created between the outer portion of the basket and the inner part of the reactor shell. A metallic gasket seals the two chambers, an impeller located in the upper section. A packing gland assembly and a cooling jacket surrounds the shaft supports the impeller. Upon rotation of the shaft, gas is forced outward from the center of the impeller towards the walls. This creates a lower pressure in the center region of the impeller thus inducing flow of gas upward through the catalyst chamber from the bottom of the reactor annular region where the pressure is slightly higher. The impeller provides a fluidized bed of catalyst particles as well as intense gas mixing inside the reactor.
+

The Riser Simulator operates in conjunction with a series of sampling valves that allow, following a predetermined sequence, to inject hydrocarbons and withdraw products in short periods of time. Reaction products are measured by a Hewlett Packard 5890A GC with an FID detector and a capillary column HP-1, 25 m cross linked methyl silicone with an outer diameter of 0.22 mm and an internal diameter of 0.33 microns. Coke content is evaluated, with a +/- 4% accuracy, in separate runs. Following this, a sample of coked catalysts is exposed to an oxidizing atmosphere with the CO2 formed being measured quantitatively. A detailed description of various Riser Simulator components, sequence of injection and sampling can be found in Kraemer [23]. Cumene cracking has been used extensively as a test reaction to investigate the characteristics of any newly developed cracking catalyst [16]. The main reaction pathway of cumene cracking involves the cleavage of the isopropyl group, producing benzene and propene.

Results and Discussion

In the case of the present study two catalysts, CAT-LC (catalyst with large zeolite crystals) and CAT-SC (catalyst with small zeolite crystals), were studied under typical cracking conditions using cumene as a model hydrocarbon reactant molecule. It was observed that cumene conversion increased with both temperature and reaction time (Figs 1 and 2) . An important result was the observed mild temperature effect with a relatively more significant influence of reaction times. For instance, at 5 seconds reaction time, cumene conversion was 23%, 32%, 38.5% and 41% when temperatures were set at 400C, 450C, 500C, 550C. In this respect, the conversion increased 9% from 400C to 450C, 6.5% from 450C to 500C and 2.5% from 500C to 550C. Once the decay model based on the time-on-stream, in the context of the Riser Simulator was established (eq.10), it was tested under a number of operating conditions as follows; a) four different reaction times (3, 5, 7, and10 seconds), b) four different temperatures (400, 450, 500, and 550oC), c) two catalysts prepared with two sizes of zeolite crystals (small crystal CAT-SC and large crystal CAT-LC), d) a single catalyst/oil ratio (catalyst/oil = 5).

The three model parameters ko, ER, and of eq(10) with =1, were determined using non-linear regression (MATLAB package). Table 3 reports the parameters obtained and the limited spans for the 95% confidence interval. Tables 4 and 5 (TOS model) report the very low correlation between ko and ER, ER and and the moderate correlation between ko and for both CAT-LC and CAT-SC. Fig 1 (CAT-LC) and Fig 2 (CAT-SC) show the comparison of the experimental and model predicted conversions (eq 10). The adequacy of the model and of the selected parameters to fit to data is shown in Table 3 given the 0.94-0.97 regression coefficients. Regarding the kinetic parameters obtained, an interesting result is the consistency of the energy of activation and the activity decay constant, 21.35 and 27.63 KJ/mole and 0.217 and 0.239 (1/s) respectively, for CAT-LC and CAT-SC. It has to be noticed that these values of ER and are in the same range to those found by Kraemer [23] while cracking alkylaromatics in the Riser Simulator and using a time-on-stream model. Following this, a non-linear regression involving eq(18), the same experimental data used with the TOS model and three adjustable parameters was considered using MATLAB software. Table 6 reports parameters obtained, while Tables 7 and 8 give the give the low to moderate cross-correlation for the parameters corresponding to the to the CAT-LC and CATSC respectively. Fig 3 reports a comparison of model results and experimental points for CATLC while Fig 4 shows the same type of analysis for CAT-SC. In addition, the adequacy of eq (18) to represent the experimental data is proven given the 0.94-0.97 regression coefficients (Table 6). It can be observed that the activation energy for cumene cracking based on the RC model is in the range of 52.75-61.96 KJ/mole with these results being in excellent agreement with the 41.8-83.7 KJ/mole activation energies reported for cumene cracking by a large number of researchers as mentioned in an extensive literature review by Corma and Wojciechowski [16]. In addition, the activity decay constant remains, for both CAT-LC and CAT-SC, in a close range of 5.5-6.2 (Table 6). Hence based on this model, at 50% conversion the catalyst have lost already 60% of its activity, while at 70% conversion there is a 72% activity loss.

Regarding the activity decay model, based on reactant conversion, one main assumption is given by eq (15). This assumption considers A as a constant parameter, not affected by cumene conversion;

A=

Cc (1 y A )

(19)

In order to check the validity of this assumption a number of independent runs were developed and in each of them, in addition to the measurement of cumene conversion, the coke formed was determined. On this basis the Cc/(1-yA) group was calculated for each run independently. It was observed that as reported in Fig 5, A (for CAT-LC) is equal to 0.60.2 and this for the different temperatures and reaction times used in the study. Given that the confidence limits for A are in the 15% range, mainly the uncertainty on Cc measurements, the A parameter can be assumed as a constant. This validates the process of numerical integration of eq(18) adopted to calculate the ER, k0, and parameters. Finally and considering that for CAT-LC the determined parameter was 5.5 and adopting A =0.6, then it results that =8.33. Thus, eq(16) can be also reported as

= exp(8.33C c ) .
It can be also mentioned that Cc for cumene cracking has typically a value in range of 0.10.2 mass fraction. Then at Cc =0.15, = 0.2866 or 71% of the catalyst activity is lost at 0.15 % coke concentration. Eq(1) includes an effectiveness factor . Evaluation of kinetic parameters assumes in both cases =1. This important assumption has to be validated. With this end, the modified Thiele modulus (h) for cumene cracking was calculated using eq (7). For instance, the constant, the k0 and the ER parameters, based on time-on-stream, were set at 0.217 (1/s), 0.0076 (m3/kgcat.s) and 23.02 KJ/mole respectively (CAT-LC in Table 3). Substituting these parameter values in eq(7) for a reaction temperature of 400oC and reaction time of 5 seconds, the modified Thiele modulus was calculated as 0.095 and the effectiveness factor as 0.996. Similarly, the calculated , ko and ER parameters of eq(18) (CAT-LC in Table 6) were used to

10

recalculate the modified Thiele modulus and the effectiveness factor. For instance for a 0.22 cumene conversion, essentially the same h and , with very close to unit, were obtained. It can thus, be concluded that cumene cracking in FCC catalysts is controlled by the chemical reaction rate with diffusional transport constrains paying a negligible role. This is, by the way, consistent with the 6.8 estimated cumene critical diameter smaller than the 7.4, faujasite zeolite openings [24]. In summary, the decay models considered in the present study and based on cumene conversion and time-on-stream show to be adequate for predicting cumene conversions in a rather large range of conversions, temperatures, and reaction times. However, the decay model based on reactant conversion should be preferred given its sounder mechanistic basis. It has also to be stressed that the decay model based on reactant conversion was considered under cracking conditions free of transport constrains and in this context the derived decay function can be considered an intrinsic activity decay function.

Conclusions
a. Catalytic cracking of cumene is a model compound that allows to test FCC catalyst under conditions free o intra-crystalline diffusional constrains with the effectiveness factor remaining very close to unity. b. Experiments were developed in a novel Riser Simualtor using two sizes of faujasite crystals (0.4 and 0.9 m). The study was conducted under relevant reaction conditions for FCC in terms of temperature, contact time, reactant partial pressure and catalyst /oil ratio, c. A new catalytic activity decay model based on reactant conversion was proposed and tested. This phenomenological based model is most valuable given it allows defining the changes of chemical species without the specific extra requirement of measuring the coke concentration, d. Kinetic and activity decay parameters were estimated for this model using regression analysis, with parameters were established with their respective confidence intervals. The consistency of the kinetic parameters for the two-zeolite crystals is an encouraging result to support the validity of the proposed model.

11

Acknowledgments
The authors would like to express our appreciation to the Saudi Arabian Government that provided a postgraduate scholarship to Mr. S. Al-Khattaf. We are very grateful for the financial support of the Natural Sciences and Engineering Research Council of Canada. We also would like to acknowledge Dr A. Humphries from Akzo Nobel and J. Macaoay from Engelhard companies, which helped with the catalyst preparation used in the present study.

Notation
A aext B CA Cc Cgo Deff Do ED ER F G kd kd1 kin kmd ko ko group of parameters (-).Refer to eq (15) external surface area (1/m). parameter of catalyst activity model. Refer to reference [14]. gas phase concentration of cumene (Kmole/m3). coke mass fraction (Kg of coke/Kg of catalyst). gas oil concentration (Kmole/m3) effective diffusivity (m2/s). pre-exponential constant involved in the definition of he diffusivity coefficient (m2/s) activation energy for cumene diffusion (KJ/mole). activation energy for cumene cracking KJ/mole). parameter catalyst activity decay model. Refer to reference [15]. parameter of the catalyst activity decay model. Refer to reference [15]. deactivation rate constant (1/s). Refer to reference [8]. deactivation rate constant (1/s). Refer to reference [8]. intrinsic kinetic constant for cumene cracking (m3/ Kgcat. s). Units based on a first order reaction. deactivation rate constant (1/s). Refer to reference [16]. pre-exponential factor in Arrenhius equation (m3/ Kgcat. s). Units based on a first order reaction. pre-exponential factor in Arrenhius equation defined at an average temperature (m3/ Kgcat. s). Units based on a first order reaction.

12

L h

zeolite crystal size (micron). modified Thiele modulus, refer to eqs(6) and (7) (-).

MWA molecular weight of cumene (Kg/Kmole). MWc molecular weight of coke (Kg/Kmole). This molecular weight is set at 800 Kg/Kmole. n N R rc rco r2 t T To V Wc Whc yA in R c c reaction order or number of sites involved in the deactivation event (-). parameter decay model . reference [9]. gas universal constant (KJ/Kmole K). rate of coke formation (Kmole coke/Kgcat.s). rate of coke formation excluding catalyst decay (Kmole coke/Kgcat.s). regression coefficient (-) reaction time (s). reaction temperature (K). average temperature for the experiments of 475 C or 748 K. Riser Simulator volume (45.10-6 m3). catalyst mass (Kgcat). mass of hydrocarbon injected into the Riser Simulator (Kg). mass fraction of cumene (Kg of cumene/Kg of hydrocarbons). deactivation constant for TOS model (1/s). constant related to coke formation , catalyst decay function based on reactant conversion (RC model). effectiveness factor for cumene cracking (-). apparent deactivation function. intrinsic deactivation function. residual activity. stoichiometric number for coke (Kmoles coke/Kmole of cumene converted) zeolite crystal density (Kg/m3).

Greek Symbols

Abbreviations

CAT-LC catalyst manufactured with large zeolite crystals (refer to Table 2) CAT-SC catalyst manufactured with small crystals (refer to Table 2) CFL C/O confidence limit catalyst over reactant ratio (Kg catalyst/Kg reactant fed)

13

Literature Cited
[1] Larocca, M., Fast Catalytic Cracking with Nickel and Vanadium Contaminants. Ph.D. Thesis, The University of Western Ontario, London, Ontario, Canada (1988). [2] Larocca, M., Ng, S., and de Lasa, H., Cracking Catalyst Deactivation by Nickel and Vanadium Contaminants, Ind. Eng. Chem. Res. Vol, 29, pp2181, (1990). [3] Farag, H. Catalytic Cracking of Hydrocarbons With Novel Metal Traps. PhD Thesis Dissertation, the University of Western Ontario, London-Ontario, Canada, (1993). [4] Yates, S.J., Fundamentals of Fluidized Bed Chemical Processes, Butlewood Monographs in Chemical Engineering, London, England, 121 (1983). [5] Weekman, V.W., A Model of Catalytic Cracking Conversion in Fixed, Moving and Fluidbed Reactors, Ind. Eng. Chem. Proc. Des. Dev., 7, pp90-95 (1968). [6] Babitz, S.M., Kuehne, M.A., Kung, H.H., and Miller, J.T., Role of Lewis Acidity in the Deactivation of USY Zeolites During 2-Methylpentane Cracking, Ind. Eng. Chem. Res, 36, pp3027-3031(1997). [7] Hopkins, P.D., Miller, J.T., Meyers, B.L., Ray, G.J., Roginski, R.T., Kuehne, M.A., and Kung, H.H., Acidity and Cracking Activity Changes During Coke Deactivation of USY Zeolite, App Catal,136, pp29-48(1996). [8] Szepe, S. and Levenspiel, O., Catalyst Deactivation, Chemical Reaction Engineering, Proceeding of the Fourth European Symp., Brussels, Pergamon Press, Oxford, 265 (1971). [9] Wojciechowski, B.W., "The kinetic foundation and practical application of the time on stream theory of catalyst decay", Cat.Rev.Sci.Eng., 9(1), 79-113, (1974). [10] Froment, G. F., and Bischoff, K. B. Kinetics Data and Product Distribution from Fixed bed Catalytic Reactors Subject to Catalyst Fouling. Chem. Eng. Sci., 17, 105 (1962). [11] Froment, G. F., and Bischoff, K. B. Chemical Reactor Analysis and Design, Second Edition, J.Wiley, (1979). [12] Gross, B., Nace, D.M., and Voltz, S.E., Application of a Kinetic Model for Comparision of Catalytic Cracking in a Fixed Bed Microreactor and a Fluidized Dense Bed. Ind. Eng. Chem. Proc. Des. Dev., 13, pp199-203 (1974). [13] Acharya, D.R., Huges, R., and Li, K., Deactivation of Silica-Alumina Catalyst During the Cumene Cracking Reaction, Applied Catalysis A, 52, pp 115-129, (1989).

14

[14] Forissier, M. and Bernard, J.R., Deactivation of Cracking Catalyst with Vacuum Gas Oil , In Catalyst Deactivation, Editted by Bartholomew, C.H., and Butt, J.B., pp359-366 (1991). [15] Corma, A., and Wojciechowski, B.,W., The Catalytic Cracking of Cumene, Cat. Rev. Sci. Eng., 24 (1), pp. 1-65, (1982). [16] Corma, A., Miguel, P.J., and Orchilles, A.V. Kinetics of the Catalytic Cracking of Paraffins at Very Short Times on Stream. Journal of Catalysis, 145,pp.58-64 (1994). [17] Smith. J.M., Chemical Engineering Kinetics. McGraw-Hill, N.Y., (1981). [18] Aris, R. Elementary Chemical Reactor Analysis, First edition, Prentice Hall, 1969. [19] Al-Khattaf, S., and H.I. de Lasa, Activity and selectivity of fluidized catalytic cracking in a riser simulator: the role of Y-zeolite crystal size Ind. Eng. Chem. Res.38,1350 (1999). [20] Karger,J., and Ruthven, D.M., Diffusion in Zeolites and Other Microporous Solids.John Wiley & Sons, Inc, 1992. [21] Ruthven, D.M., and Kaul, B.K.,Adsorption of Aromatic Hydrocarbons in NaX Zeolite. 2. Kinetics. Ind. Eng. Chem. Res., 32, pp538-541 (1993). [22] de Lasa, H.I. "Novel Riser Reactor Simulator", USA Patent 5,102,628 (1992). [23] Kraemer, D. W., (1991). Modeling Catalytic Cracking in a Novel Riser Simulator, PhD Thesis Dissertation, University of Western Ontario, London, Canada. [24] Bhatai, S., Zeolite Catalysts, Principles and Applications, CRC Press, Boca Raton, 7-18, 239-250, (1990).

15

Table 1 Properties of the small and large zeolites of this study Small zeolite 4.1 5.6 24.49 0.4 Large zeolite 0.25 5.7 24.51 0.9

Na2O (wt%) SiO2/Al2O3 (mol/mol) Unit cell size () Crystal size (m)

Table 2. Properties of CAT-SC and CAT-LC catalysts used in this study CAT-SC (Catalyst with small crystals) 24.28 148 Negligible 0.4 CAT-LC (Catalyst with large crystals) 24.28 155 Negligible 0.9

Unit cell size () BET Surface Area (m2/g) Na2O (wt%) Crystal Size (m)

Table 3 Kinetic constants for both CAT-LC and CAT-SC based on time on stream (TOS) Catalyst k0.103 (m3/Kgcat.s) 95% CFL ER (KJ/mole) 95% CFL (1/sec) CAT-LC CAT-SC 7.61 7.24 1.3 1.16 21.35 27.63 0.96 0.89 0.217 0.239 0.066 0.94 0.064 0.96 95% CFL r2

Table 4 Correlation matrix for CAT-LC (TOS-model) ko ko ER 1 0.095 0.96 ER -0.095 1 -0.043 0.96 -0.043 1

16

Table 5 Correlation matrix for CAT-SC (TOS-model) ko ko ER 1 -0.125 0.96 ER -0.125 1 -0.046 0.96 -0.046 1

Table 6 Kinetic constants for both CAT-SC and CAT-LC based on reactant conversion (RC) Catalyst CAT-LC CAT-SC k0.103 (m3/Kgcat.s) 13.0 14.12 95% (CFL) 6.3 6.62 ER (KJ/mole) 52.75 61.96 95% CFL 2.88 3.2 5.5 6.2 95% CFL 1.9 1.9 0.94 0.97 r2

Table 7 Correlation matrix for CAT-LC (RC-model) ko ko ER 1 0.73 -0.98 ER 0.73 1 -0.75 0.98 -0.75 1

Table 8 Correlation matrix for CAT-SC (RC-model) ko ko ER 1 0.82 0.98 ER 0.82 1 -0.84 0.98 -0.84 1

17

Figure Captions

Fig 1. Comparison of experimental and of modeling cumene conversions for CAT-LC based on time-on-stream (eq(10)). C/O=5. Reported experimental points are average value of at least 3 measurements. Typical errors are 2%. ( ) 400C, () 450C,. ( ) 500C, ( ) 550 C.

Fig 2. Comparison of experimental and of modeling cumene conversions for CAT-SC based on time-on-stream (eq(10)). C/O=5. Reported experimental points are average value of at least 3 measurements. Typical errors are 2%. ( ) 400C, () 450C,. ( ) 500C, ( ) 550 C.

Fig 3. Comparison of experimental and of modeling cumene conversions for CAT-LC based on reactant converted (eq(18)). C/O=5. Reported experimental points are average value of at least 3 measurements. Typical errors are 2%. ( ) 400C, () 450C,. ( ) 500C, ( ) 550 C.

Fig 4. Comparison of experimental and of modeling cumene conversions for CAT-SC based on reactant converted (eq(18)). C/O=5. Reported experimental points are average value of at least 3 measurements. Typical errors are 2%. ( ) 400C, () 450C,. ( ) 500C, ( ) 550 C.

Fig. 5. Coke selectivity for cumene conversion as a function of reaction time. C/O=5, ( ) 400C, ( ) 500C, ( ) 550C.

18

0.6 0.5 Cumene conversion 0.4 0.3 0.2 0.1 0 0 5 10 Reaction time (sec) 15 20

Fig. 1

19

0.6

0.5 Cumene conversion

0.4

0.3

0.2

0.1

0 0 5 10 Reaction time (sec) 15 20

Fig. 2

20

0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 5 10 Reaction time (sec)
Fig.3

Cumene conversion

15

20

21

0.7 0.6 Cumene conversion 0.5 0.4 0.3 0.2 0.1 0 0 5 10 Reaction time (sec)
Fig.4

15

20

22

0.8

Cc/(1-yA)

0.6

0.4

0.2

0 0 2 4 6 8 10 12
Reaction time (sec)

Fig. 5

23