CATALYTIC CRACK ING OVER CRYSTALLINE A LUMINOSIL IC ATES I . Instantaneous Rate Measurements for Hexadecane Cracking D O N A L D
M . N A C E
Mobil Research and Development Corp., Paulsboro. N . J . 08066
A novel microreactor is described in which product distribution and instantaneous rates of cracking can be evaluated for hydrocarbon reactants having boiling points up to 600'F. Using this tool, the decline in activity of catalysts was measured a t on-stream times from less than a second to 2 minutes of n-hexadecane cracking. Although the rate of cracking decreases by a factor of 10 from 1 second to 1 minute on stream, the distribution of products does not change during this aging process. A crystalline aluminosilicate catalyst, having over 10 times the activity of a 46 AI amorphous silica-alumina catalyst for cracking n-hexadecane, declines in activity at the same rate as the amorphous catalyst. Product distributions from these two catalysts have significant differences which may be due to higher relative rates of hydrogen transfer reactions to carboncarbon scission reactions over the crystalline aluminosilicate catalyst.
PROBLEMS in defining relative activities and selectivities of cracking catalysts have concerned those engaged in catalysis research for many years. Since gas oil cracking reactions can be generalized by the use of a pure hydrocarbon reactant, such studies are numerous in the literature of catalysis. Because the activity of a cracking catalyst decays during reaction, microreactor techniques have been used in recent years to measure instantaneous reaction rates a t short intervals of catalyst residence time. This technique, involving gas chromatographic analysis of pulses of product, has been limited t o hydrocarbons having carbon numbers less than 10, as exemplified by the papers of Kokes et aE. (1955), Hall et a / . (1960), and Levy et a / . (1963). T h e present investigation uses a novel microcatalytic reactor to measure instantaneous rates of conversion of pure hydrocarbons having boiling points as high as 600" F. An application of this technique to show differences in activity and product selectivity in the cracking of n-hexadecane over the conventional amorphous silicaalumina catalyst and the new highly active crystalline aluminosilicate catalyst is demonstrated. Experimental
Microreactor. T h e apparatus used in this investigation consists of a glass reactor, a valve system for sampling the vapor-stream effluent from the reactor, and a modified commercial gas chromatograph (Figure 1). T h e glass reactor contains a fixed preheating section filled with Vycor chips and a removable catalyst bed holder which can be filled with 3.5 cc. of mixed Vycor and catalyst particles. A hypodermic pump with variable gear drive is capable of delivering liquid feed at a desired rate through a needle into the top of 24
I & E C PRODUCT RESEARCH A N D DEVELOPMENT
the preheating section. For pulse experiments, a slug of reactant can be injected instead from a hand hypodermic into a helium stream passing through, the preheater at a rate which produces the required contact time in the catalyst bed. The valve system used to sample the product stream from the pumped feed experiments consists of seven diaphragm packless valves. cam-operated and machined to contain very small dead space volumes. The stem, cam. and aluminum diaph,ragm are part o f a commercial valve (Hoke C-413); however. the valve body was specially made from stainless steel cylinders and contains I ,,inch inlet and outlet ports connected. as shown in Figure 1, by ',-inch tubing. Each of seven valve bodies is bolted to a 1,inch thick stainless steel plate, 5 x 6 inches, on the back of which are fastened two 150-watt strip heaters. The entire valve assembly is enclosed in an asbestos box. so that the temperature of the valves and connecting tubing can be held a t any constant temperature up to 650' F. \$-hen operated for slug injections. helium flows through the reactor. valves 2 , 1. 5, 6, and into valve 3, where it is split, a 60-cc. per minute fraction going through the chromatograph and the remainder being bypassed. An aliquot of the products from the entire injected hydrocarbon reactant, therefore, is carried into the chromatograph column by the 60-cc. per minute helium flow regardless of the total helium flow through the reactor. 1
I t is possible to activate both solenoid valves shown in Figure 1 and to open and close two of the diaphragm valves in order to switch to a separate 60-cc. per minute helium carrier stream which does not pass through the reactor. This can be done after the hydrocarbon slug reaches the chromatographic column in order t o exc!lude any slowly desorbed products from the catalyst. However, experiments have shown that allowing the helium stream to continue through the catalyst bed during chromatographic analysis does not produce results significantly different from those in which helium purges the catalyst
Figure 1 . Schematic of microreactor for only a minute. The :hydrocarbon contact time is determined by the volume o f catalyst in the holder and the helium rate: and a t fixed helium flow rate, the catalyst on-stream time is determined by the volume of hydrocarbon reactant injected. i5'hen the apparatus is used as a continuous flow reactor, the connections shown by broken lines in Figure 1 are attached. A helium flou- of 60 cc. per minute passing through valves 5, 6, and 3 is the chromatograph carrier gas. T h e hydrocarbon stream leaving the reactor passes without condensation through valves 2 . 1, 5, :3, 4, and 2 and enters a small glass trap. At the instant that an analysis is desired, valves 1 and 4 are closed and 2 is opened simultaneously, valves 3 and 5 are opened, and valve 6 is closed. K i t h this operation, the vapor sample in the coil between valves 5 and 3 ' s carried into the gas chromatograph column by the 60-cc. per minute helium carrier gas. The hydrocarbon feed to the reactor can be stopped and the catalyst purged a t any time after the sample is taken. Analysis. T o scan a wide boiling range of products and ensure a quantitative analysis of components, a specially modified gas chromatograph technique was adopted. I n place of the heated detector cell in a Model A-ii50 Aerograph dual-column temperatureprogrammed gas chromatclgraph, dual copper oxide combustion tubes were connected to the tubing leading from the dual columns. The effluent hydrocarbon peaks are oxidized a t 1300'F. in one of the copper oxide tubes, and the carbon dioxide is detected in a room temperature thermal conductivity cell after removal of' water in a Drierite tcbe. This general technique has been employed by Eggertsen a r d Groennings (1958,and Eggertsen et a i . (19601.
Standardization test's confirmed t h a t combustion is 99+'< complete with butane and higher molecular weight hydrocarbons, 98+', complete with propane. and 90 to 92' complete with ethane. Separate gas collection and analysis showed that conversion to methane, ethane, and ethylene accounts for less than 1' conversion of reactants
at conditions used in this study, and the error due to insufficient combustion is ignored in the rate calculations and product distributions for most experiments. The copper oxide was regenerated with oxygen after each analysis was completed. Temperature programming during analysis was started below room temperature (column precooled with a stream of vaporizing liquid nitrogen) and heated a t 7.2"F. per minute until all products and unconverted reactant had been detected. Using a &foot column of silicone gum rubber (SE-30) on firebrick. retention times of noncyclic compounds (paraffins and olefins) are a nearly straightline function of true boiling point, the n-hexadecane being eluted at 55 minutes. Cyclic compound boiling points deviate from this line by not over 10' F. a t a given retention time. The recorded chromatogram can, therefore, be resolved into various fractions which divide the products according to carbon number, since practically no overlap of isomers from one carbon number to another exists in the products of paraffin cracking. From quantitative blends of Cy to C . , hydrocarbons, average weight deviations of + 0 . 3 ' ~ have been measured. The unconverted n-hexadecane peak after 10 to 30'; conversion is very large. but quantitative calibration experiments have established that the hydrocarbon weight-peak area relationship is linear throughout the entire range encountered. A separate analysis required for the cracking experiments is that of coke on catalyst. Because of the small weight of catalyst used in each experiment (0.00; to 0.7 gram), the magnitudes of the weight of carbon amounted in most cases to between 0.00005 and 0.010 gram. The analysis scheme used involves burning the entire catalyst VOL. 8 NO. 1 M A R C H 1 9 6 9
25
bed, including Vycor diluent, in oxygen in a microcombustion furnace, converting CO to CO? over hot copper oxide, and collecting the COj in a silica gel trap a t liquid nitrogen temperature. After purging out the oxygen with helium, the silica gel is warmed to room temperature with no helium flow, and on restarting helium flow, the Iiberated CO, is detected in a thermal conductivity cell. A calibration curve was made using varying portions of a coked catalyst of known carbon content. Kinetic Treatment of Data
Data for pure hydrocarbon cracking have generally been found to obey kinetic equations which are first-order in pressure of reactant. A review of various conversion equations tested with experimental data is given by Voge (19581, where, for example, total conversion of n-hexadecane cracking in an integral reactor fits moderately well the equation 0
hJ
h = - [-4 In (1 - x) - 3x1
gp
This equation, the derivation of which is shown by Voge (19581, expresses the reaction rate constant, k , in units of moles (kilograms of catalyst) - i (hour)-' (atmosphere) when S is in units of moles (liters of catalyst) -' (hour) -', g is catalyst density in grams (cc.)-', P is total pressure in atmospheres, and x is fraction converted. I n the derivation of the equation, the number of moles of product per mole cracked is assumed to have a value of 4 and to be independent of conversion level. The first-order rate constant, k , derived from the above equation using instantaneous measurements of conversion, is used throughout this paper as a measure of the activity of the catalyst a t a given instant. While it is realized that h is a pseudo rate constant comprising both physical and complex chemical kinetic effects on the rate of conversion, changes in k should approximate basic activity changes of the catalyst during the cracking cycle. Data obtained from continuous flow experiments appear to fit the equation moderately well, as values of k are independent of space velocity and change only with temperature and catalyst residence time. Pulse data, discussed below, show a dependency of the calculated k on space velocity a t very short catalyst residence time. Materials
The hydrocarbon reactant used in the cracking experiments was Humphrey-U'ilkinson olefin-free grade n-hexadecane which had been percolated through a silica gel column prior to use. By chromatographic analysis, a typical batch contained 0.3% isopentadecane, 2.7% n-pentadecane, and 2.1% isohexadecane. These impurities remained a t the same concentration in the cracked product. Substitution of an A P I sample of pure n-hexadecane resulted in < 0 . 2 5 iso-Cli lii hydrocarbons in the product a t 18% conversion. Conversions over two different catalysts are compared. One catalyst is a conventional amorphous silica-alumina which has been steamed to a surface area of 408 sq. meters per gram and has a sodium content of 0.2%. The alumina content is 9.7% by weight. I t is referred to as a 46 AI silica-alumina on the basis of a standard gas oil cracking test (CAT-A activity index) (Alexander, 1947). The second catalyst, R E H X catalyst, is a rare earth and 26
I & E C PRODUCT RESEARCH A N D DEVELOPMENT
ammonium chloride exchanged crystalline aluminosilicate of the synthetic 1 3 x faujasite type described by Plank et al. (1964) and has a surface area of 368 sq. meters per gram and a sodium content of 0.2%. The calcined catalyst analyzes 29.6 weight % A120j and 25.9 weight 'C (RE)?OI. Both catalysts were crushed and screened to 20- to 40-mesh after calcining a t 1000"F. overnight. Steaming of the zeolite catalyst reduces crystallinity. X-ray scattering data indicate that the steamed R E H X catalyst discussed in this paper retained 55'; of the crystallinity of typical unsteamed R E exchanged zeolite. This crystallinity agrees closely to the ratio of the B E T surface areas of the steamed zeolite (368 sq. meters per gram) and the unsteamed catalyst (660 sq. meters per gram), the latter value being in close agreement to the highest B E T area reported for fresh NaY zeolite (675 sq. meters per gram) by Yates (1968). The relative inactivity of the noncrystalline material in the steamed catalyst compared to the crystalline material is indicated by the ratio of the activity of steamed R E H X catalyst to t h a t of amorphous silica-alumina steamed a t the same severity. This ratio of activities for n-hexane cracking a t 1000" F. [the alpha value defined by Miale, Chen, and Weisz (1966)] is about 35 t o 1. Hexadecane Pulse Cracking
Rate Constant Dependency. When a quantity of n-hexadecane is pulsed through the catalyst bed, the various product peaks emerging from the chromatograph column are very sharp and not broadened as they would be if desorption of products from the catalyst were slow. Reasonably quantitative material balances can be calculated from pulse runs since products and unconverted reactant are detected by total count of carbon atoms. For example, with no catalyst in the reactor, a 0.012gram injection of n-hexadecane underwent a 1.5% conversion a t 900' F. and the total area of peaks in the chromatogram amounted to 52,000 units of area per 0.01 gram of hydrocarbon injected. Similar injections over various amounts of catalyst giving conversions of 10 to 2 6 5 produced areas deviating no more than the i3s;O accuracy of measuring the volume of hexadecane injected, an indication that no significant losses are incurred. T h e calculation of a pseudo-reaction rate constant from a fractional second pulse experiment assumes t h a t steadystate conditions exist during the conversion of practically all of the pulse. If adsorption isotherm kinetics play an appreciable part in the over-all reaction rate constant, comparison of h's for different catalysts or for a series of temperatures may differ considerably from such comparisons made after true steady-state conditions are established. By varying the volume of the pulse, the volume of the catalyst, and the rate of carrier gas flow through the reactor, the hydrocarbon contact time (expressed in this paper in terms of liquid hourly space velocity, cubic centimeters of reactant per cubic centimeter of catalyst-hour) and the catalyst residence time (onstream time in this paper) could be changed. Independent variation of catalyst residence time and space velocity can result in changes in catalyst-hydrocarbon ratio (cat/ HC). I n obtaining pulse data, the catalyst was diluted 25 to 1, or higher, with Vycor. Dilution had a negligible effect except a t higher conversion levels, where undiluted catalyst gave a lower value of k , an effect which is
Table
I.
Pulse Data for n-Hexadecane Cracking over REHX a t 900'
F.
Wt
Cat Bed Ddutmn"
LHSl', Vol Vol -Hr
On-Stream Time, See
Cat HC, Vol Vol
cc
k , Moles Kg Cat -Hr -Atm
ConL
5000 5000 2500 2500 1250 1250
CONSTANT CATALYST-HYDROCARBON RATIOPULSES 1.5 15.3 6.100 0.45 1.5 15.6 6,240 0.46 1.5 25.1 6.000 0.94 1.5 26.0 6,230 0.98 1.5 40.7 6.400 2.1 1.5 36.9 5.430 2.0
35:l Xone
5000 5000' 2500 2500 1250 1250
0.45 0.47 0.52 0.58 0.47 0.58
CONSTANT T I h l E OS-STREAM PULSES 1.5 15.3 6,100 1.6 14.8 5,700 2.9 36.7 10,500 2.5 28.6 7.180 6.3 71.4 20,700 5.0 56.7 12,000
2531
5000
0.45 2.0 3.7 7.0 11.3'
CONSTANT SPACE \.ELOCIT'I. PULSES 1.5 15.3 6,100 0.35 13.8 5.200 0.20 15.1 5.800 0.10 8.0 2,700 0.07 8.6 3.000
2531 I':one
25:l None
2531 None
25:l 70:1 None
cc
c
on cata1jst
0.49 0.33 0.32 0.27 0.32 0.25
53.800 49,500 49,400 49,500 53,100 47.500
0.49 0.21 0.40 0.48
53,800 52,800 48.600 49,000 54,000 47.800
... 0.30
53,800
0.49 0.55 0.42 0.58 0.69
... 5L1.900 55.900 52,200
' Volume of inert Vycor uolume of catalyst. ' Vapor uelocitj through catalJ,st bed double that of other runs in table. Multiple pulses of about 2-second duration mch uith 2 to 3 minutes of helium purge betueen pulses. Anaiisis made ojfinal 2-second pulse onlj
apparently due to surface temperature changes resulting from the endothermic heat of cracking. If space velocity and catalyst residence time are both varied in such a way that c a t / H C ratio is constant, the calculated h is fairly independent of space velocity. D a t a given in Table I show that a variation from 5000 to 1250 LHSV a t a fixed 1.5 c a t / H C ratio yields a n average k of about 6200. Although the rate constant does not change when catalyst residence time increases from 0.5 t o 2.0 seconds, there is also no increase in coke concentration on the catalyst. If the c a t / H C ratio is allowed to vary, as it does for constant time on-stream or constant space velocity data in Table I, k is found to increase with increasing c a t / H C ratio. A t 1.5 c a t i H C k is of the order of 6000, a t 3 cat/HC. k has increased t o 10,000, and a t 6 c a t / H C a value of ti as high as 20,000 was found. Table I d a t a demonstrate that dilution has a proconversion level; therefore, only found effect a t a the diluted catalyst data are considered t o be at a 900" F. catalyst temperature. C a t / H C ratio is found to have little, if any, effect on the reaction rate constant during continuous flow experiments or for the long duration pulses. Illustrating the change in dependency of k on space velocity-viz., on c a t i H C ratio a t constant catalyst residence timeis t h e Figure 7 exponential aging plot of i? rs. catalyst residence time for both pulse and continuous flow data. Apparently the reaction rate does not follow first-order kinetics during the first few seconds of catalyst operation, but it does a t longer times of catalyst use. Pulses to investigate the initial activity of the catalyst are of very short duration and are more distorted from a square shape than larger pulses. T h e nonlinear end effects result in non-steady-state adsorption conditions a t the beginning and the end of a pulse passing through the catalyst bed. T h e non-steady-state period is only a small fraction of a large pulse, but a very short pulse presents transient concentrations over most of its duration. For this reason
it is believed that the lack of constancy in k in the 0.5-second data results from the shape of the hydrocarbon pulse. A 0.5-second pulse a t the lowest space velocity (highest cat;HC ratio) requires a short pulse to travel through a long bed of catalyst. Distortion of the pulse would probably play less of a role there than it would passing through a shorter bed (lower c a t i H C ratio). Coke Formation
T h e quantity of coke formed on the catalyst does not vary greatly as space velocity or c a t j H C ratio is changed, nor is it sensitive t o changes in conversion from 14 to 5 i ' c (Table I ) . Even as time on-stream is increased t o 11 seconds, using multiple pulses, only a slight increase in coke occurs. as shown in Table I. Most of the coke is formed a t a very early time in the cracking cycle, t h e initial contact of the hydrocarbon with fresh active sites of' the catalyst probably resulting in a n irreversible adsorption of hydrocarbon on the active sites. T h e percentage of reactant converted t o coke decreases rapidly with increasing time on-stream, indicating rapid saturation of the catalyst with coke. Temperature Effects
T h e changes which occur when temperature is varied were studied briefly using the REHX catalyst. Conversion data a t constant c a t i H C ratio are given in Table 11. An Arrhenius-type plot of the reaction rate constant has Table II. Effect of Temperature on n-Hexadecane Cracking over REHX a t Constant Catalyst-Hydrocarbon Ratio
Temp., F. L H S V
900 825 750
5040 2440 12:30
Cat HC 13 1.;7
1.5
On-Stream Time. Sec. Conc. ((
0.46 1.0 2.0
15.6 20.5 26.5
h, ,Moles Kg. Cat -Hr.
6240 4330 3150
Wt. ( c c on Catalyst 0.33 0.29 0.34
VOL. 8 NO. 1 M A R C H 1969
27
z J51 0
E
30
c
I
?\
‘.>.&
0 -
QOOOF,
5,000 L H S V
( 1 6 % CONV.)
e-
825-F
2,500 L H S V
( 2 0 % CONV.)
A-
750’F
1,250 L H S V
(27% CONV.)
\,I
3 ‘
2‘
4 ‘
‘5
6 ‘
‘7
‘8
‘9
‘10
‘12
‘11
Figure 2 . Effect of temperature on product distribution from n-hexadecane cracking over REHX
a slope equivalent to a n 8.5 kcal. per mole activation energy, which agrees with the 10.0 kcal. per mole value reported by Blanding (1953) a t 3 seconds on-stream time for gas oil cracking. While this activation energy is much lower than the commonly cited range of 20 to 30 kcal. per mole found in integral reactor experiments involving much longer on-stream times, Blanding suggests that the higher values may reflect a combination of a true chemical rate change and the effect of temperature on the rate of coke formation. T h e product distribution is altered by temperature (Figure 2). As temperature decreases, the average molecular weight of cracked products increases. The decrease in the number of moles of product per mole cracked when temperature is lowered is believed to be caused by a decrease in the rate of carbon-carbon scission relative to the rate of hydrogen transfer of intermediate carbonium ions-that is, the chain process of beta-scission reactions of the hexadecane carbonium ion is stopped a t an earlier stage a t low temperatures because of proton exchange and desorption of intermediate molecular weight carbonium ions. Catalyst Effects
T h e R E H X catalyst is much more active a catalyst than 46 AI silica-alumina. With the latter catalyst a t
the lowest space velocity attainable in the microreactor for pulse runs, 100 LHSV, 40.45 conversion of n-hexadecane was obtained, compared t o 36.9‘; conversion using R E H X catalyst at 1250 LHSV. T h e calculated h’s for R E H X and for 46 AI silica-alumina for these two runs are 3430 and 500. respectively. Catalyst composition also has a marked effect on product distribution. I n Table 111, the product distribution. from the silica-alumina is shown to agree fairly well with the data of Greensfelder (1949). Less CI-C? make and a lower Cr:C, ratio are the main disagreements, part of which may be due to the 32” temperature difference. T h e R E H X catalyst alters this distribution markedly by reducing C X , yields and increasing C&, yields. T h e lower resulting moles of product per mole cracked is characteristic of crystalline aluminosilicate catalysts compared to amorphous silica-alumina. Conversion level has only a small effect on product distribution. T h e distribution differences are compared in Figure 3. T h e decrease in the number of moles of product per mole cracked when temperature is lowered and when the crystalline aluminosilicate is substituted for the amorphous silica-alumina is believed to be caused by changes in the relative rates of carbon-carbon scission and hydrogen transfer of intermediate carbonium ions. I t is not necessary t o assume that when the number of moles of product per mole cracked is greater than 2, primary cracked products desorbed from the catalyst are cracked again a t another catalyst site. T h e carbonium ion mechanism allows the initially formed carbonium ion t o continue t o split off short-chain olefinic molecules by beta-scission reaction until the residual carbonium ion is desorbed by a hydrogen exchange reaction. If hydrogen exchange to desorb the carbonium ion should occur after only two olefin molecules have been split from the original hexadecyl carbonium ion, the number of moles of product per mole cracked would be lower than if four propylene molecules, for example, had time to split off before hydrogen exchange removed the residual fragment. Reducing temperature or replacement of the amorphous with the crystalline catalyst results in higher molecular weight products because of
Table Ill. Product Distribution from n-Hexadecane Cracking
Literature Data‘
COP Si-Al-Zr CatalyAt, 98P” F .
Time on stream, see. LHSV c; conversion Product distribution, mole i
3600 10 24
C: C? C, Ci Ci
C, C-
C, C9
C,,, C,I C,? C,, C.4
Moles product mole CI, cracked Wt. ”
28
c;
C on catalyst
1.4 3.4 27.0 28.4 17.8 13.9 2.2 2.2 0.9 0.8 0.6 0.6 0.5 0.3 3.59
...
3600 2 54 3.2 4.8 30.0 30.8 16.0 r
-
1.1
2.4 1.3 1.1 0.8 0.8 0.5 0.3 0.2 3.76 1.1
Greensfelder 11949~.‘ Continuous flou run.
I & E C PRODUCT RESEARCH A N D DEVELOPMENT
16 ‘4 I SiO? Ai.0 C‘nta/)ht,900’F. 2 100 40.4
loo 100 30.1
...
...
1.1 23.0 36.8 22.2 11.1 2.7 1.o 1.0 0.7 0.3 0.2
1.3 25.9 31.9 21.3 13.2 2.5 0.8 1.0 0.9 0.7 0.2 0.1
...
... 3.55 0.03
...
3.54 0.10
REHX Catalyst, 9UO’F. 2 1250 40.7 .
I
15’ 1300 15.5
...
...
13.7 29.4 23.1 16.3 8.5 4.1 2.5 1.4 0.6 0.3
12.1 27.7 23.6 17.5 9.7 4.3 2.8 1.3 0.6 0.4
... ...
3.11 0.32
... ...
3.07 0.99
t40
* - 46 25 U I-
20
-
A I SILICA-ALUMINA (40 % CONV.) g O O ' F , 2 SEC. ON-STREAM
,A- R E H X (41 % CONV.) 9 0 0 ° F , 2 SEC. ON-STREAM
15
+20
f
o
W
0
-20
4
3
8
2a
I
r
-
U
-40
lx J
-60 1
0.40 CC BED DEPTH (325 LHSV) - 0
I-
,"
1.3
-80
0
Figure 3. Product distribution from n-hexadecane cracking over different catalysts
the increased ratio of hydrogen transfer rate to scission rate. Instantaneous Measutements .during Continuous Flow Runs.
Bed Temperature Influence. The microreactor, when operated with continuous flowing hydrocarbon reactant as described above, could be used to measure instantaneous conversion rates a t much higher times on stream. The high dilution of catalyst with Vycor in the pulse experiments (25 to 1) was not possible with the larger amounts of catalyst needed for continuous flow runs. Initial studies were made without any dilution of the 20- to 40-mesh catalyst particles in t3he catalyst holder; however, data obtained suggested that temperature gradients of considerable magnitude might exist in these catalysts beds. The catalyst surface temperatures were determined for a number of these undilutc$ catalyst runs by running a fine wire (KO.30 gage) thermocouple with a spot-welded junction into the catalyst bed. Several inches of the thermocouple wire next t o the junction were coiled into a ring small enough to be located in the lower end of the catalyst bed, with the junction just above it in the center of the bed. Lead wires ran through the Vycor chips above the catalyst and along the catalyst bed wall, so that vapor channeling would. not occur near the couple junction. With this arrangement of an interior catalyst bed thermocouple, temperatures could be measured which were not in error from heat leak along the thermocouple wire. Temperatures were recorded a t several junction positions in the catalyst and a t several times on stream. Plots of' temperature in a 0 . 0 5 , 0.1-, 0.4-, and 1.3cc. catalyst bed as a function of time on stream, as shown in Figure 4. indicate an initial exothermic heat effect followed by a rapid decline of temperature with a minimum a t 10 seconds or less, after which temperature rises to a steady-state level held for most of the 2 minutes on stream. The initial temperature rise is probably associated with an exothermic coke-forming reaction.
20
cc
BED DEPTH (110 LHSV)
40 60 80 100 120 TIME ON STREAM (SEC)
Figure 4. Temperature profiles in catalyst bed for n-hexadecone cracking over REHX at 900" F.
experiments over the less coked catalyst. T h e apparent activation energy, which is the observed relationship of temperature of the reaction rate, includes the effect of temperature on the kinetic rate constant (true activation energy) as well as any effect of temperature on the rate of decline of activity of the catalyst. If the catalyst loses activity faster a t a low temperature than a t a higher temperature, the apparent activation energy would be increased when longer on-stream times are used. Catalyst Comparisons
The striking differences in cracking activity and in rate of coke formation which have been demonstrated in pulse experiments for the crystalline and the amorphous catalyst still exist a t longer times on stream. The instantaneous rate constant obtained from continuous flow runs sampled from 7 seconds to 2 minutes on stream is plotted in Figure 5. Although runs were made a t several space velocities and cat:HC ratios, the rate constant is primarily a function of time on stream. The c a t l H C ratio a t these longer on-stream times does not have the dominant influence on the rate constant that it had in pulse experiments-for example, a t 30-second on-stream time, increasing c a t . < H C ratio from 0.09 to 0.37 (decreasing LHSV from I300 to 325) affected the calculated rate constant less than 20';. Simultaneously with the large drop in catalyst activity the concentration of coke on catalyst increases (Figure 5 ) . Data from the two techniques, pulse and continuous flow, plotted in Figure 6, show the activity decline over
0 0
1300 L H S V
A
325 L H S V
v
cn
Rate Constant Measurement
/
T o compare values of the reaction rate constant calculated from instantaneous samples, correction had to be made for the temperature variation in the catalyst bed. T o make this correction, the apparent activation energy was determined from runs a t 825", 865", and 905" F. sampled a t :'-minute on-stream time. T h e apparent activation energy of 22.8 kcal. pel' mole which was derived is nearly three times the value obtained with