I
C. J.
PLANK, D. J.
SIBBETTI, and R. B. SMITH
Research and Development Laboratory, Soeony Mobil Oil
Co.,Inc., Paulsboro, N. J.
Comparison of Catalysts in Cracking Pure Methylcyclohexane and n-Decane Useful information on comparative behavior of three important types of catalysts: silica-alumina, silica-magnesia, and fluoride-treated silica-alumina
TO
RECOGXIZE fundamental differences between various types of cracking catalysts, it is almost essential to study the cracking of specific pure hydrocarbons over the catalysts. Catalytic cracking of pure hydrocarbons was first studied by Egloff and colleagues (2,4,20) followed by Greensfelder and coworkers (7-73, 27), and more recently by Gladrow, Krebs, and Kimberlin (5, 6). The first two groups restricted their study largely to silica-alumina catalyst modifications, while the third considered several types. However, in all this work the detailed analyses were restricted to the Cs and lighter products and narrow conversion ranges were studied. In the work described here the cracking of two pure hydrocarbons-methylcyclohexane and n-decane-has been studied a t 495' C. in considerably more detail. Both gaseous and liquid products were analyzed as completely as possible by the mass spectrometer. A broad range of conversions was covered with the hope of gaining insight into the primary products and reactions. The emphasis is on catalyst comparison. Three types of catalysts widely different in their gas oil cracking properties were used: silica-alumina, silica-magnesia, and fluoride-treated silica-alumina. The first is a typical commercial silicaalumina bead cracking catalyst. The second is a catalyst known to give much higher gasoline yields than silica-alumina at the same conversion level, but with lower octane number. I t has been made and tested on a commercial scale ( 3 ) .
Present address, Davison Chemical Co., Baltimore, Md.
742
The third catalyst is an experimental material, described by Hyman (75). Although very unstable, it is of interest because it gives increased yields of gasoline a t a given conversion level with decreased light gas and much lower coke yields. The results obtained from the pure hydrocarbon cracking go far toward explaining these cataIytic differences in gas oil cracking. Experimental Catalysts. Table I contains the available analytical information on test catalysts. All were crushed to 10- to 25-mesh (U. S. standard) to minimize diffusion difficulties. Fresh portions of the fluorided catalyst were used in each run. The silica-alumina and silicamagnesia catalyst samples were regenerated and used repeatedly, with no change in activity noted a t the end of the series of runs. The sources of the three types of catalysts were as follows:
SILICA-ALUMIXA, a commercial bead Table I.
catalyst produced prior to 1947 by Socony Mobil Oil Co. (77). SILICA-MAGNESIA, a fluid grade 73% silica-277, magnesia catalyst manufactured by Davison Chemical Co.; pelleted and then crushed before use. FLUORIDE-TREATED SILICA-ALUMINA, prepared by impregnating a 10- to 25mesh batch of crushed silica-alumina bead catalyst of Type 1, with an equal volume of a 10% (by weight) aqueous solution of ammonium fluoride. After the mass had been stirred for 3 hours, the sample was dried overnight at 140' C. and calcined for 3 hours at 600' C. The total sample (10 liters) was consolidated, then separated into uniform portions by riffling. Hydrocarbons. Methylcyclohexane, Phillips Pure Grade analyzing 99.8Yo by mass spectrometer, was used without purification. n-Decane, from Humphrey-Wilkinson, Inc., was distilled in a 100-plate column, then passed over silica gel and stored under nitrogen. Mass spectrometer analysis showed the resulting material to be 99.2YG n-decane with 0.8% other paraffins.
Properties of Test Catalysts Composition
Silica-aluminaSilica-alumina, 90-10
Silica-magnesia, 73-27
fluoride, 1.8
88-10-
BET surface area, sq. meter/ g.
Bulk density, g./ml. Real density, g./mI. Particle density, g./ml. Pore volume, ml./g. Av. pore diameter, A.
INDUSTRIAL AND ENGINEERIN6 CHEMISTRY
437 0.73 2.35 1.18 0.421 39
449
139
0.72
0.69
2.69 1.04 0.592
2.27 1.21
53
0.384 110
of conditions was divided into two equal portions. One portion was hydrogenated at room temperature and 2atm. pressure, with a palladium-onalumina catalyst. Both portions were distilled separately through a high temperature Podbielniak column into cuts with the following boiling point ranges:
Boiling Point Range, Cut
AV. CONTACT TIME (SEC.) Figure 1. Conversion of methylcyclohexane vs. average contact time
Cracking Process. The hydrocarbon conversions were carried out in a borosilicate glass reactor at 495' C. and atmospheric pressure. The tube was 60 cm. long, with an inside diameter of 2.5 cm. I t contained a coaxial thermowell 0.6 cm. in outside diameter. A 140-cc. catalyst section was used with Vycor chips (10- to 14-mesh) placed above and below the catalyst for preheating and support. The temperature of the catalyst bed was held within f 5 ' C. by a 24-inch electric furnace, Model M-30244, made by the Hevi-Duty Electric Co. It contained three independent 220-volt units totaling 3.0 kw., which were independently and automatically controlled by a Celectray (Weston Electric Instrument Corp.). The temperatures noted in Tables I1 and I V are the average bed temperatures for each series of runs. The liquid hydrocarbon was pumped down the vertical catalyst bed by a direct displacement pump, which fixed the feed rate within f0.5%. The product from the catalyst bed was passed through a water-cooled condenser into an ice-cooled, tared flask which acted as a gas separator. The gas passed into a gas-collecting system, where it was measured by the displacement of saturated salt solution. The charge pumped was always 100 cc. (76.9 grams of methylcyclohexane or 73.0 grams of n-decane); conversion was changed by varying space velocity. The catalyst was treated with dry nitrogen a t 495' C. for 2 hours before each run. At the end of each cracking' cycle the catalyst was purged for 1 hour with dry nitrogen a t 495' C. Coke on the catalyst was then determined as carbon dioxide by oxidation of a representative sample with oxygen. The carbon monoxide produced in the combustion was converted to carbon dioxide by using a cupric oxide catalyst. Analytical Procedure. T h e composite liquid product obtained a t one set
1 2 3 4 5 6
7
174
O
C.
n-Decane To 43
Methylcyclohexane To 43
43-66 66-89 89-100 100-150 150-174
43-66 66-89 89-100
100
+ (bottoms)
+ (bottoms)
-
' 0
20
40
60
80
CONVERSION (WT. %)
Figure 2. Conversion of methylcyclohexane to cracked products (Cl-C,)
The gas collected during the run and cut 1 were analyzed directly by the mass spectrometer. The remainder of the cuts were separated into two parts. One part was acid treated at 0' C. with a solution of phosphorus pentoxide in concentrated sulfuric acid. The other remained without treatment. All samples were analyzed by the mass spectrometer. Comparison of the analyses of treated and untreated products showed that isomerization of saturates during hydrogenation and acid treatment was negligible. No significant saturation of aromatics took place during hydrogenation. The method was devised because mass spectrometer analysis does not determine olefins with molecular weights in excess of 70. As hydrogenation saturates olefins and the acid treatment removes both olefins and aromatics, information from the analytical scheme is used as follows : Analysis of acid-treated samples determines the individual saturated products which boil below 110' C. Differences between analyses of hydrogenated and nonhydrogenated, acidtreated samples determine olefins. Analysis of the untreated product determines olefins of five carbons or less and aromatics. It supplies a check on olefin analysis by difference and on component analysis for paraffins. This analytical scheme accounts for individual hydrocarbons which boil below 110' C. Type analysis was used to determine the small amounts of product which boil above 110' C. T h e bottom cuts, largely (up to 99%) charged hydrocarbon, were calculated by a hand solution of the mass spectrometer data. Total weight recovery was better than 960/, and usually about 98%. The product yields were put on a no-loss basis after distillation by distributing the loss uniformly throughout the liquid products. The fact that the totals shown in Tables I1 and I V are not equal to 100% is due to round-off errors in mass spec-
trometer analyses and subsequent calculations. Results and Discussions
The experimental results are discussed separately for the two hydrocarbons. In each case the data are considered in terms of (1) conversion and (2) product distribution and reaction specificity. The reactions under consideration are carbon chain splitting, isomerization, hydrogen transfer, aromatization, and coke formation. By inference olefin polymerization is also shown to play a role. Thermal cracking is not significant in this work. At the longest contact time used, only about 2% cracking occurred in the absence of a catalyst. Cracking of Methylcyclohexane (MCH). CONVERSION.Figure 1 shows the conversion (100 minus weight per cent charge recovered) of methylcyclohexane at 495' C. as a function of the average contact time on the catalyst. Average contact time is defined as: Average contact time = tr X p X volume of catalyst nRT where t, is total run time, and n is average number of moles of hydrocarbon passed over the catalyst. The catalyst volume is defined as the bulk volume divided by two. LHSV (Tables I1 and IV), liquid hourly space velocity, is equal to volume of charge per bulk volume of catalyst per hour. Activity differences of the three catalysts are very great as shown by the average contact time (seconds) required to achieve 45y0 conversion of methylcyclohexane. Silica-alumina-fluoride Silica-alumina Silica-magnesia
1.0 5.6 24.5
I n general, the relationships are similar over the whole conversion range studied
.
VOL. 49, NO. 4
APRIL 1957
743
Table II.
Production Distribution (Weight Per Cent) in Conversion of Methylcyclohexane at 495” i: 1
Catalyst
LHSV, vol./vol.-hr. Average contact time, see. Conversion, weight % Hydrogen Methane Ethene Ethane Propene Propane Isobutane Butenes n-Butane Isopentane Pentenes n-Pentane Cyclopentane Cyclopentene n-Hexane
Silica-Alumina 2.4 0.52 1.4 6.1 27.1 47.0
6.2 0.57 17.3
0.057 0.093 0.15 0.031 1.26 0.74 1.93 0.38 0.89 1.04 0.25 0.063 0.057 0.027 0.066
0.20 0.56 0.51 0.37 1.84 2.95 5.96 1.13 1.13 2.13 0.25 0.16 0.12 0.02 0.11
0.66 1.90 0.74 0.86 2.87 5.08 8.44 2.71 2.00 4.21 0.44 0.28 0.40 0.04 0.23
0.18 0.05 0.11 0.60 0.10 0.037 0.070 0.071 0.15 0,052 0.01 0.009 0.033 0.12
0.54 0.3 0.11 0.19 1.46 0.54 0.20
0.07
...
... 0.32
0.34 0.14 0.12 0.06 0.95 1.62 0.05 0.11 0.22 0.48 0.09 0.01 0.01 0.07 0.35
0.12
0.020
0.04
0.57 1.42 0.52 0.75 8.02 82.74 0.67 0.79
0.43 1.07 1.22 1.61 8.15 72.94 1.18 0.32
0.008 0.010 0.012 0.003 0.41 0.09 0.52 0.25 0.52 0.20 0.061 0.011 0.081 0.038 0.007
2-Methylpentane 3-Methylpentane 2,3-Dimethylbutane Hexenes Methylcyclopentane Methylcyclopentene Cyclohexane Cyclohexene Benzene n-Heptane 2-Methylhexane 3-Methylhexane 3-Ethylpentane 2,2- and 2,4-Dimethylpentane 2,3-Dimethylpentane 3,3-Dimethylpentane Heptenes
1,l-Dimethylcyclopentane 1,2-Dimethylcyclopentane 1,3-Dimethylcyclopentane Ethylcyclopentane CTcyclopentenes Methylcyclohexene Methylcyclohexane Toluene Iso-octanes Octenes CS cyclopentanes CS cyclopentenes CS aromatics CYaromatics CIOaromatics Coke (as carbon) Totals
... ... ...
0.060 . a .
0,016 0.050 0.058 0.37 0.29 0,030
... ... 0.24
0.20
... 0.44 ...
... 0.53
0.68 1.70 1.66 1.28 7.44 53.02 4.27 0.35
... 0.24 ...
0.31 0.10 0.20 0.09 0.11 0.11 0.52 0.63
1.11 0.58 0.34 0.10 0.11
2.00 1.05 0.15 0.70
3.09 4.26 2.73 1.54 3.26 31.85 11.07 0.37 0.06 0.78 0.35 4.20 0.15 0.05 2.45
100.29
99.37
102.72
... 2.16
0.098 0.095 0.095 0.099 100.03
...
Product Distribution and Reaction Specificity, The weight percentages of the products obtained are shown in Table 11. The selectivity to specific products is given in Table I11 and Figures 2 to 8. Product selectivity is defined as moles of the indicated product
Table 111.
0.21 12.7 68.2
... 2.44 ...
2.0 1.8 11.9 0.020 0.040 0.016 0.040 0.18 0.096 0.23 0.19 0.062 0.10 0.010 0.022
...
0.013 0.026
...
...
...
0.010 0.078
...
0.031 0.066 0.13 0.50 0,023 0.07 0.16 0.001 0.25 0.03
Silica-Mr ignesia 0.26 0.11 0.60 1 2 . 8 25.7 6.7 34.2 47.0 23.0 0.086 0.17 0.11 0.067 0.81 0.33 1.10 0.22 0.58 0.41 0.15 0.043 0.031 0.013 0.096 0.048
...
0.001 0.054 0.34 0.44 0.028 0.048 0.041
...
0.027 0.09 0.005
...
0.099
...
... 0.16
0.092
0.23
0.063 0.060 0.018 1.05 6.20 88.15 0.53 0.34
0.70 1.97 0.65 1.99 6.54 77.04 1.18 0.016
99.90
...
0.46 0.17 0.10 0.10 0.10 0.076
0.17 0.41 0.17 0.17 0.79 0.60 1.38 0.81 0.42 0.76 0.08 0.01 0.10 0.02 0.21 0.15 0.05
...
0.13 0.60 0.13 0.07 0.04 0.11 0.08 0.25
0.77 0.31 0.12 0.05 0.19 0.02 0.56
0.43 0.44 0.22 0.26
1.80 3.59 0.98 0.50 8.61 65.83 2.26 0.13 0.98 0.39 0.52 0.95 1.01 0.50 0.75
97.26
98.53
... ... . . O
obtained per 100 moles of charge converted. This method of plotting seems most readily interpreted, but it magnifies errors by converting all values to ratios. The values shown for the lowest silicamagnesia point are more open to question than the others, as the percentage
Selectivity to Cracked Products in Methylcyclohexane Conversion
0.85 2.64 0.40 1.26 1.16 2.30 1.78 0.73 0.64 0.95 0.10 0.09 0.17 0.05 0.08 0.13 0.05 0.002 0.23 0.89 0.65 0.13 0.06 0.40 0.13 0.21 0.55 0.35 0.07 0.14 0.05 0.56 0.90 2.45 3.13 1.10
...
2.61 53.03 10.14 0.21 0.12 0.11 0.33 2.55 0.76
...
3.76
99.00
...
...
...
... ...
... ... ...
1.33 2.52 1.00 1.80 7.14 76.45 0.49
1.66 3.56 1.34 2.56 5.98 70.94 0.60 0.35
3.22 7.28 2.67 1.69 2.29 50.63 0.49 0.23
...
0.77
2.09
0.48 0.71 0.55 0.02 100.26
1.67 1.98 1.17
... ... 1.69
0.31 0.39 0.23 0,019 100.25
...
...
...
...
0.077 99.48
17.3
Ci C2 c 3
CI Cj Cs
0.3 0.4 6.8 13.2 3.2 4.1
Silica-Alumina 27.1 47.0 6 8 . 2 2.2 2.4 17.8 21.1 7.7 5.8
7.9 6.9 25.0 32.1 8.5 10.1
17.5 9.3 27.2 33.7 11.0 6.8
Moles indicated products 100 moles MCH converted’ This line represents conversion, weight %.
744
INDUSTRIAL AND ENGINEERING CHEMISTRY
Silica-Alumina-Fluoride 2 3 . 6 29.1 49 4 8 5 . 3 b 2.9 0.3‘ 0.5 1.0 1.1 1.0 2.8 4.0 10.2 12.5 19.0 21.5 16.1 21.6 30.1 28.8 6.3 9.2 13.5 14.6 5.2 6.7 9.1 8.7
1.61 2.66 2.85 9.95 8.75 14.69 4.16 0.07 0.20 1.81 0.60 4.18 4.06 2.23 0.42 99.82
errors were greatest in this case. CARBON CHAINSPLITTING is sho\vn in Table 111, which gives the selectivity to the various C1 to Ce products by carbon number, including saturates and olefins. Figure 2 summarizes the total Cl-CS production-Le., the “cracked products.” I t is clearly indicated here by extrapolating to zero conversion that cracking is not a significant primary reaction in the conversion of methvlcvclohexane over any of the three catalysts. Rather, it is a secondary reaction of about equal importance over all the catalysts up to about 50% conversion. Beyond -that conversion level the fluorided silicaalumina shows less tendency to give I
Selectivitya Silica-Magnesia 1 1 . 9 23.0 3 4 . 2 4 7 . 0 7.7 35.3 2.1 4.4 1.6 2.6 3 . 5 12.1 5.4 11.4 9.8 17.1 7.1 14.0 13.3 11.7 1.8 4.5 4.2 4.1 3.7 5.6 5.2 6.8
C.
Silica-Alumula-Fluoride 9.2 6.2 2.45 0.52 0.39 0.56 1.4 6.2 23.6 29.1 49.4 85.3 0.0068 0.019 0.002 0.0032 0.074 0.40 0,010 0.021 0.31 0.76 0.051 0.072 0.08 0.25 0.027 0.018 2.21 2.65 0.68 0.96 1.87 5.42 0.43 0.66 5.41 9.95 1.35 2.20 1.83 1.89 0.37 0.98 1.40 2.60 0.64 0.57 3.79 6.86 0.85 1.44 0.36 0.94 0.13 0.25 0.27 0.60 0.068 0.07 0.36 0.67 0.087 0.15 0.07 0.48 0.025 0.07 0.27 0.50 0.072 0.09 0.58 1.34 0.13 0.21 0.33 0.08 0.07 0.11 0.13 0.37 0.037 0.03 0.05 0.47 0.024 0.09 1.59 2.50 0.44 0.47 0.52 0.35 0.12 0.45 0.16 0.45 0.096 0.07 0.05 0.011 0.15 0.20 0.095 0.17 0.46 0.03 0.08 0.08 0.1 0.22 0.1 0.15 0.40 0.41 0.1 0.07 0.04 0.04 0.023 0.02 0.11 0.22 0.03 0.021 0.06 0.02 0.77 0.77 0.22 0.79 0.32 0.27 0.31 0.37
,
cracked~roductsthantheothercatal~sts~ The distribution of the cracked products shows some interesting catalyst differences. For example, silica-magnesia tends to favor terminal splitting, particularly at conversion levels above
COMPARISON O F C A T A L Y S T S
!." .**
' 0 *
20 40 60 80 CONVERSION (WT. %)
Figure 3. Conversion of methylcyclohexane to C7 cyclopentanes and C7 cyclopentenes
I
1
+
I
/
#*
0
40 60 80 CONVERSION (W T. %)
20
Figure 4. Conversion of methylcyclohexane to aromatics about 30%. However, the production of methane and Cz's by this catalyst is not matched by an equivalent yield of Cs-Ce's. In fact, only coke and aromatics yields parallel those of methane and C2. The yields of (22's are less than methane over any catalyst at a given conversion, although the general picture is similar. Both silica-alumina catalysts favor central cracking to c3-c~'~. However, the expected 1 to 1 ratio of C3 to C4 products is not obtained. Both catalysts give a sizable excess of (24's. This difference is apparently balanced by the comparatively high yields of C5 and Ce compounds compared to those given by silica-magnesia. This is another indication that a sizable proportion of the cracked products may come from polymeric precursors. For example, we may postulate reactions of the type 2',C c14+ Ce 2 c4. ~SOMERIZATION has been considered from two points of view: production of C7 cyclopentanes and cyclopentenes (Figure 3), and production of isoparaffins in the Cd io C7 range. The isonormal ratios are plotted in Figure G for the total C4-C7 paraffins produced. Extrapolating the curves of Figure 3 to zero contact time shows that isomerization to C7 cyclo-CS'Sis a primary reaction over the alumina-containing catalysts but may not be over silica-magnesia. The uncertainty lies in the weight to be placed on the lowest silica-magnesia point. The silica-alumina-fluoride catalyst is particularly active in this isomerization. The formation of isoparaffins is not directly a measure of the paraffin isomerization activity of the catalysts, since the iso-normal ratios obtained are well above the equilibrium values. Actually the formation of these isoparaffins represents one measure of the cata-
I
Figure 5. Conversion of methylcyclohexane to coke of methylcyclohexane converted a t 45% conversion are: SiO1/A1qO9/F
lysts' hydrogen-transfer activity. Therefore, this phase of the product distribution is discussed in connection with the hydrogen exchange reaction. AROMATIC FORMATION. Both plots of Figure 4 are important. At a given conversion silica-magnesia gives the highest yield of aromatics, and the fluoride-treated catalyst gives the lowest. However, at a given contact time the selectivity to aromatics is much lower on silica-magnesia than on alumina-containing catalysts. Thus the actual rates of aromatic formation are much the fastest over silica-alumina-fluoride and slowest on silica-magnesia. However, the relative rates of competing cracking reactions increase much more rapidly on catalysts containing alumina. As a result, net aromatization selectivity is more prominent with silica-magnesia a t a given conversion level. COKEFORMATION. Coke is the hydrocarbon residue of high molecular weight present on the catalyst a t the end of the cracking cycle. Its formation is a complex reaction which obviously involves polymerization. I n addition, as catalytic coke has been shown to be polyaromatic in nature, it involves hydrogentransfer steps as well as polymerization and cracking. The coke-forming properties of the catalysts are illustrated by Figure 5. The polymerization reaction has been shown to be important to ihe product distribution given by the fluoridetreated silica-alumina. Nevertheless, coke formation is negligible in converting methylcyclohexane with silica-alumina-fluoride. O n the other hand, silica-magnesia gives high coke yields and on silica-alumina they are intermediate. For example, the coke yields in gram-atoms of carbon per 100 moles
1.8 14.3
58.5
Cracking of the polymeric materials is much greater in the case of catalysts containing alumina; the competing hydrogen transfer reactions tending to convert the polymers to coke are greatly reduced. Several correlations not shown graphically here were developed. With all three catalysts the molar yields of hydrogen, methane, and total aromatics are linearly related to the yields of coke. The close relationship among hydrogen, aromatics, and coke formation is not surprising, but the relation between methane and coke is less expected. Furthermore, this relationship is essentially the same for all three catalysts. The ratio of moles of methane per gramatom of carbon in coke is 0.5 for silicaalumina and silica-magnesia, and 0.6 for silica-alumina-fluoride-for every 2 gram-atoms of carbon converted to coke, approximately 1 mole of methane is produced in the cracking of methylcyclohexane. Thus, it seems probable that methane is a principal product of the cracking of cyclo-olefin polymers during their conversion to coke. HYDROGEN TRANSFER, one of the primary reactions of carbonium ions, consists of the abstraction of a hydride ion from another hydrocarbon molecule by a given carbonium ion. The reaction is manifested in several aspects of the product distribution: saturation of cracked products, high yield of isoparaffins, aromatic and coke formation, and the fact that hydrogen is not a primary product, while methylcyclohexene is. The degree of saturation of cracked products is best illustrated by the ratio of paraffins to olefins in the cracked products (CZto (20). The comparison on the VOL. 49, NO. 4
APRIL 1957
745
present work leads the authors to support the arguments of Voge, Good, and 6 Greensfelder. The high isoparaffin yield depends on the following factors: Olefin or carbonium ion isomerization is 4 very rapid, although possibly not rapid enough to maintain equilibrium between the olefin isomers [Gladrow, Krebs, 0 2 U and Kimberlin (S)]; and tertiary carIbonium ions are much more stable than 6 primary or secondary. Therefore, the E charge separation is greater than for 4 other carbonium ions, they react much more rapidly in hydrogen-transfer reactions, and they are less likely to be destroyed by cracking reactions. The over-all effect is to give a high ratio of is0 to normal paraffins when Figure 8. Conversion of methylcycloactive hydrogen-transfer catalysts are 8 16 24 32 hexane to hydrogen used in cracking. This ratio is far higher AV. CONTACT TIME (SEC.) for the cracked products than expected 60 Figure 6. Iso-normal ratio and p a r a f from thermodynamic data for all the 0fin-olefin ratio in conversion of methylcatalysts above 20y0conversion. Significyclohexane cant differences between the catalysts 3 show up mainly at low contact times. -40 Figure 6 shows that below 1-second averZ 0 age contact time the silica-alumina-fluocn ride is much more selective to isoparaffins $20 than silica-alumina. Silica-magnesia > Z reaches an equivalent selectivity only 0 after about 6-second contact. Possibly 0 0 4 8 12 0 the minimum in the curve for the fluoAV. CONTACT TIME-SEC. rided catalyst is not real, as all the values for this catalyst are yery high. Figure 9. Conversion of n-decane vs The hydrogen-transfer effect is also average contact time important to the actual yields of the isoparaffins (Table 11). Here the clear I ) I l l superiority of the alumina-containing catalysts is obvious. The rate of saturation is much the greatest on silicaalumina-fluoride. AROMATICAND COKE FORMATION. U b o ' ,/=The importance of hydrogen transfer Lin obtaining these products seems clear and has been discussed frequently .n 1 (8, 70, 12, 74, 20). The formation of these materials represents a series of hydrogen-transfer Figure 7. Conversion of methylcyclosteps rather than a single step. As a hexane to methylcyclohexene I result, the expected effect is not observed -the most effective hydrogen-transfer CONVERSION (WT.%] basis of conversion level (no graph) shows catalysts do not give the highest selectivno difference between the ratios given by ity to aromatics and coke, although they Figure 'I 0. Iso-normal ratios for C4-c6 silica-alumina and silica-magnesia. The give greater rates of aromatic formation. paraffins from n-decane ratio for the fluorided catalysts is signifiThe results are probably due to the fact cantly higher than the others only below that the rates of cracking of the cyclo3oy0conversion. olefins are very high on the catalysts Methylcyclohexene is a primary prodIn Figure 6 (bottom) the paraffin-olecontaining alumina; less chance is given uct but hydrogen is not. Figure 7 fin ratios (total C2-C6)are plotted against for polymerization and subsequent exshows selectivity to methylcyclohexene ; contact time. The three catalysts differ change to form coke or aromatics. it is apparent that methylcyclohexene is considerably-for example, the average Consideration of Table II illustrates the major primary product in the concontact times (seconds) necessary to version of methylcyclohexane. On the these points. The relative yields of reach a paraffin-olefin ratio of 2.5 are: Cs-Clo aromatics are high compared to other hand, Figure 8 shows that selectivthe toluene yields on the fluorided catalyst ity to hydrogen is essentially nil at low Silica-alumina-fluoride 2 Silica-alumina 7 but not on the other catalysts. This conversions, T o the authors this proves Silica-magnesia 24 that the initiating step in the reaction is shows the importance of polymerization not thermal conversion of methylcycloas a precursor of aromatic formation High Yield of Isoparaffins. The role hexane to the olefin plus hydrogen. The over this catalyst, while the low coke of the hydrogen exchange reaction in most reasonable alternative seems to be formation shows that polymerization forming isoparaffins has been discussed the formation of the initial carbonium cannot proceed very far before the crackby Voge, Good, and Greensfelder (27) ion by transfer of a hydride ion to the and Gladrow, Krebs, and Kimberlin (6), ing and hydrogen-transfer reactions take catalyst surface-Le., a Lewis acid over. whose views are very different. The
-
c-:
.'
I
746
INDUSTRIAL AND ENGINEERING CHEMISTRY
COMPARISON OF CATALYSTS mechanism of the type proposed by Milliken, Mills, and Oblad (78). This still leaves the question of what happens to the hydrogen when the carbonium ion converts to methylcyclohexene. No other products capable of using this hydrogen are formed a t very low conversions. The authors suggest that a proton is also transferred to the catalyst. This adsorbed proton must be sufficiently isolated from the hydride ion originally abstracted by the catalyst to resist molecular hydrogen formation, leaving both free to react in secondary hydrogen-transfer reactions. The capacity of the three catalysts for this type of reaction differs greatly, as seen by comparing the hydrogen and methylcyclohexene selectivities a t higher conversion levels. Hydrogen balance calculations show that both silica-alumina and silicamagnesia are in hydrogen balance above 20% conversion. With silica-aluminafluoride some hydrogen disappearance is shown even a t 29% conversion, though the hydrogen balance with this catalyst is very good (&I%) above 49% conversion.
Table IV. Catalyst LHSV, vol./vol.-hr. Average contact time, sec. Conversion, wt. yo
*
adsorption of molecular hydrogen on these catalysts would be completely unimportant a t reaction conditions. Cracking of n-Decane. CONVERSION. The conversions of n-decane are shown as a function of the average contact time on the catalysts in Figure 9. At short residence times the rates of conversion are in the order: fluorided silicaalumina > silica-alumina > silicamagnesia. However, as the residence time is lengthened the activity of the silica-alumina-fluoride declines more rapidly than either of the others. At residence times in the neighborhood of 15 seconds a short extrapolation shows the order of conversion rates to be: silica-alumina > silica-magnesia > silicaalumina-fluoride. The decrease in conversion rates with time is considered to be due to inhibitors formed by the cracking reactions. The fluoride-treated catalyst is most strongly affected by these inhibitors. Catalyst differences and the relative rates of n-decane us. methylcyclohexane conversion can best be considered in connection with kinetics, as discussed later. PRODUCT DISTRIBUTION AND REACTION
Product Distribution (Weight Per Cent) in Conversion of n-Decane at 5 7 0.9 11.5
Silica-Alumina 1.3 0 70 3.6 5 9 41 5 30.0
- _ _ _ Silica-magnesia 0 35 5 7 1.3 0.70 11 0 1.0 3.8 6.4 52.7 21.8 31.4 4.1
0.27 14.6 49.4
495" =t1 O C.
5.7 0.9 16 0
Silica-Alumina-Fluoride 1 2 0.70 0 30 6.0 11.8 3.6 31.7 38.9 45.9
0.027 0.26 0.67 0.36 3.84 3.01 3.32 1.68 4.57 3.12 0.78 2.37
0.061 0.25 0.96 0.53 4.94 5.28 6.02 2.67 4.49 4-38 0.96 2.32
0.12 1.13 1.26 0.96 5.26 7.61 8.52 3.54 4.44 6.08 1.11 1.98
0.004 0.031 0.084 0.082 0.338 0.150 0.131 0.150 0.605 0.170 0.100 0,543
0.029 0.18 0.72 0.35 2.22 1.37 1.61 1.19 3.46 1.92 0.68 2.83
0.065 0.35 0.56 0.49 3.40 2.26 2.83 1.78 4.74 2.99 0.95 3.49
0.17 1.13 0.90 1.10 4.63 4.49 5.68 2.79 5.72 5.78 1.06 3.38
0.002 0.041 0.20 0.06 2.18 1.20 1.86
0.15 0.01 0.49 0.60 0.43 0.136 0.88 0.04
0.12 ... 0.45
0.007
0.13 0.01 0.41 0.54 0.27 0.13 0.88 0.01
0.028 0.005 0.099 0.042 0.022 0.003 0.312 0.002
0.16 0.02 0.62 0.53 0.28 0.06 1.27 0.01 0.31
0.18 0.02 0.75 0.63 0.36 0.11 1.46 0.01
0.07 0.01 0.26 0.36 0.20 0.068 0.79 0.006
0.09
0.49
0.34 0.03 0.75 0.94 0.54 0.16 2.05 0.02 0.18 0.11 0.49 0.57 0.24 0.14
0.03
Hydrogen Methane Ethene Ethane Propene Propane Isobutane n-Butane Butenes Isopentane n-Pentane Pentenes
0.008
C yclopentane Cyclopentene n-Hexane 2-Methylpentane 3-Methylpentane Dimethylbutanes Hexenes Benzene n-Heptane 3-Ethylpentane Dimethylpentares Heptenes
0.06 0.007 0.17 0.19 0.13
Toluene Iso-octanes n-Octane Octenes Ca aromatics Isononanes Nonenes COaromatics Isodecanes n-Decane Decenes CIOaromatics Coke (as carbon)
0.021 0.02
Totals
At the lowest conversions studied the hydrogen disappearance, defined as gram-atoms of hydrogen unaccounted for per 100 moles of methyjcyclohexane converted, is 66 for silica-dagnesia and 65 for silica-alumina-fluoride. In the case of the former catalyst this amounts to the retention of 1.1 meq. of hydrogen per gram of catalyst, with even more on the fluorided catalyst, because the conversion is higher. One possible alternative to the initiation theory proposed here is that molecular hydrogen forms but reacts with oxygen adsorbed on the catalyst surface. This alternative is refuted by the quantity of adsorbed oxygen which would be required even in the silicamagnesia case referred to-i.e., 0.27 mmole per gram. This is equivalent to 6 cc. (standard temperature and pressure) of oxygen adsorbed per gram of catalyst. Judging from the data of Benton ( 7 ) on oxygen adsorption on silica gel, the above figure is 50 times as great as the adsorption expected at these reaction conditions. Similarly, data of Reyerson (79) on the adsorption of hydrogen on silica gel indicate that the
0.05
0.18 0.08 1.40 0.90 0.99 0.62 1.84 1.35 0.29 1.33
... 0.60
... ... 0.08 6
1607 0.08 0.0006 0.001 0.14 0.53 88.54 0.29 0.09 0.02 100.11
... ... 0.12
0.14 0.10
... 0.27
0.50
0.004 0.01 0.55 0.02 69.95 0.62 0.44 0.42 100.02
0.92 0.55 0.18 0.40 0.04 0.10
... ... 0.35
... 0.43
0.073
0.25
0.18
0.28 0.11
0.41
... 0.03
0.75 0.01 0.003 0.86 0.03 58.52 0.52 0.51 1.60 99.62
... ...
... ... 0.52
0.103
... ... 0.38
0.49
... 0.14
0.001
0.03
... 0.05 ... 0.028
0.09
0.03 1.12 0.03
0.044 0.004 0.006 0.013 0.002
0.04 0.18 0.08
0.05 0.30 0.015 0.33
95.88 0.134
0.27 0.04 78.15 0.39 0.09 0.19
99.21
99.71
4 . .
1.20 0.04 47.30 1.15 0.65 2.76 100.21
... ... 0.017
...
... 0.09
1.00
2.38 1.85 0.48 1.24
... ... 0.22 0.12
...
...
0.01
0.05
0.02
0.50
0.74 0.02 0.03 1.06 0.06 50.61 1.11 0.96 2.42
0.12 0.001 0.004 0.16 0.01 84.04 0.29 0.18 0.06
99.71
100.45
...
0.04
68.63 0.88 0.34
99.52
0.003 0.19 0.48 0.25 3.83 2.95 4.69 2.13 3.78 4.05 0.89 2.17
0.005 0.32 0.59 0.37 4.41 4.00 5.88 2.56 3.99 4.78 0.97 2.15
0.009 0.65 0.85 0.56 4.77 5.40 7.77 3.32 4.08 6.09 1.15 1.93
0.13 0.01 0.37
0.12 0.01 0.37 0.78 0.53 0.25 0.82 0.01
0.35
... ... 0.22
0.11 0.05 0.48 1.11 0.71 0.24 0.34 0.02 0.03 0.006 0.30 0.11
0.14
0.18
0.27
0.04 0.02 0.59 0.02
0.06 0.01 0.81 0.01
0.70 0.006 61.13 1.11 0.52 0.82
0.88 0.01 54.08 1.82 ,0.65 1.63
98.54
100.31
0.77 0.54 0.164 0.88 0.01
... ... 0.20 e . .
0.03 0.03 0.42 0.01 0.01 0.46 0.02 68.34 0.79 0.33 0.52 99.94
VOL. 49, NO. 4
0.27
...
...
...
...
APRIL 1957
747
_______
~~~
Table V.
~~~
~
~~
Selectivity to Cracked Products in n-Decane Conversion S'nlnctivitvu
Silica-Alumina 11.5 30.0 41.5 52.7
Hz Ci
5.1 3.9 11.6 66.9 75.7 53.4 16.2
C? C3 C? Cs Ca a
6.5
7.7 16.9 75.6 80.0 42.6 12.8
10.5 11.9 17.6 81.0 78.5 38.3 10.7
16.4 19.0 20.7 80.4 77.5 35.2 8.2
Silica-Magnesia 4.1 21.8 31.4 49.4 9.2 8.3 24.3 48.8 66.7 51.0 24.3
14.7 10.0 16.5 60.0 74.4 48.7 18.7
9.0 7.2 24.5 54.7 71.9 51.1 21.3
24.5 20.1 19.7 60.9 71.6 42.5 15.3
Silica-Alumina-Fluoride 16.0 31.7 38.9 45.gb 0.8 2.2 8.1 70.3 82.0 50.5 17.9
0.7 5.3 11.4 71.2 83.2 45.7 14.6
0.9 7.4 12.1 71.7 79.3 41.2 12.2
1.3 12.5 15.2 73.1 81.8 40.2 10.8
Moles indicated products 100 moles n-Clo converted' This line represents conversion, weight %.
SPECIFICITY.The product distribution in terms of weight percentages is shown in Table IV. The selectivity to specific products is shown in Table V and Figures 10 to 12. Most of the conclusions as to catalyst similarities or dissimilarities agree with those drawn from the methylcyclohexane work, but in general, differences between the catalysts are less pronounced in converting n-decane. CARBOKCHAIN SPLITTING.Unlike the case of methylcyclohexane, this is the primary reaction involved in the conversion of n-decane. From 83 to 92 weight yo of all products are aliphatic hydrocarbons with one to six carbon atoms per molecule. Over the conversion range studied, the yield of cracked products (C, to C,) is 2.4 =t0.1 moles per mole converted, for all three catalysts. As shown in Table V, silica-magnesia has a fairly high selectivity for end-chain cracking to give methane and CZ'Sa t low conversions compared to the other catalysts. The difference between the catalysts is much less at conversions above 30y0. The CS to CS aliphatic hydrocarbons are the chief primary products over all catalysts. The selectivity to C3's and C i s is greater on catalysts containing alumina, while silica-magnesia produces more Cs to C Saliphatics. The actual yields of C7 to C9 products of this class are low, varying only from 1.5 to 4.5 moles per 100 moles converted for the alumina-containing catalysts and 4 to 7 moles per 100 moles converted for silica-magnesia.
ISOMERIZATION. None of the catalysts give significant isomerization of n-decane to isodecanes a t 495' C. In every experiment except one, less than 0.2 mole of isodecanes was formed per 100 moles of n-decane converted, but, high yields of lower molecular weight paraffin isomers were obtained. This is attributed to olefin isomerization and hydrogen transfer. Figure 10 shows the is0 to normal ratio for the principal paraffin products (Cd to C6). The relative inferiority of silica-magnesia in this respect is apparent. AROMATICFORMATION. The formation of aromatics from n-decane over cracking catalysts is a relatively unimportant reaction. With all three catalysts aromatization is much slower and less important than in the conversion of methylcyclohexane. The maximum selectivity obtained was about 8 moles per 100 moles of n-decane converted. Silica-alumina is the most selective and silica-magnesia the least for this reaction, but differences are not great. Over 60% of all the aromatics formed are CS and Cgcompounds with all three catalysts. COKE FORMATION. Figure 11 is a plot of the weight percentage (on charge) of coke formed against conversion. The results are in sharp contrast with those observed in the cracking of methylcyclohexane. In n-decane conversion the coke yields are nearly the same over all three catalysts a t a given conversion, within the range studied. This strongly suggests that the coke advantage shown
c
- ------
8L G
0
3 OO
20 40 60 CONVERSION (WT. % ) Figure 1 1. Conversion of n-decane to coke 748
20
40
60
CON V E RS I O N ( W T. % ) Figure 12. Paraffin-olefin ratio from n-decane conversion
INDUSTRIAL AND ENGINEERING CHEMISTRY
by the fluorided catalyst in gas oil cracking is due very largely to its effect in the cracking of naphthenes. The coke advantage of this catalyst also holds in the cracking of Decalin (unpublished data). Silica-magnesia is the only one of the three catalysts which forms less coke from n-decane than from methylcyclohexane. Once again this correlates with the postulate that increasing the cracking rate lessens the coke formation. HYDROGEN TRANSFER. Saturation of Cracked Products. The ratio of paraffins to olefins in the cracked products (C, to Cg) is shown in Figure 12. Once again the superiority of the catalysts containing alumina in hydrogen-transfer efficiency is illustrated. As all catalysts give approximately 2.5 molecules per molecule of decane converted, the products would have a paraffin-olefin ratio of about 0.6, if no hydrogen transfer took place. The curve for silica-magnesia extrapolates to very near that value a t zero conversion. The silicaalumina and silica-alumina-fluoride curves, on the other hand, extrapolate to a much higher value. Furthermore, these catalysts retain their superiority a t higher conversions. The variation in relative rates of saturation is illustrated by the fact that the average contact times required to reach a paraffin-olefin ratio of 1.0 are: silica-alumina-fluoride 1.5 seconds, silicaalumina 2.5 seconds, silica-magnesia 8.0 seconds. HIGH YIELD OF ISOPARAFFINS. As shown in Figure 10, both the silicaalumina catalysts are considerably more active in this aspect of the hydrogentransfer reaction. There are three points of difference from the comparable comparison with methylcyclohexane. The silica-magnesia remains inferior to the other catalysts throughout the conversion range ; the fluorided catalyst does not give a significantly higher isomer ratio than silica-alumina a t very low conversions of n-decane; and, the absolute value of the is0 to normal ratio is much lower with n-decane for all catalysts over the whole range of conversions studied. This is doubtless due to the fact that n-decane cannot form any is0 compounds by direct carbon-carbon splitting. COKEAND AROMATIC FORMATION. The significance of hydrogen transfer to these reactions has been discussed under methylcyclohexane conversion. The main point added by the n-decane data is that the fluorided catalyst shows almost no difference from the other catalyst in coke formation. This apparently means that the relative rate at which the polymeric coke precursors are cracked is similar to that for the other catalystsa situation vastly different from the one obtaining in methylcyclohexane conversion.
COMPARISON O F CATALYSTS
c
HYDROGEN FORMATION. Again this is a negative aspect of hydrogen transferthe better the hydrogen transfer, the less the hydrogen production. The selectivity to hydrogen us. conversion is shown in Table V. The picture is qualitatively very similar to the analogous comparison for methylcyclohexane. With silicaalumina-fluoride the hydrogen yields are very low and with silica-magnesia fairly high, silica-alumina being intermediate. Decenes are not major primary products over any of the catalysts. Thus, the data do not resolve the problem of the reaction-initiating mechanism for n-decane cracking-thermal olefin formation may be required (in very low concentration), or hydride ion transfer to the catalyst may be the initiator. To the authors, the high activity of the fluorided catalyst in spite of its low surface area argues that the latter mechanism is a t least an important contributor. Principal Catalyst Differences
Kinetics of Conversion. To make catalyst comparisons it was thought worth while to derive a kinetic expression for the conversion data obtained with both hydrocarbons. A simple firstorder equation fitted well the data for both hydrocarbons on all three catalysts. I t was derived according to the method outlined by Laidler (76),based on the assumption that the reactant is weakly adsorbed on the catalyst, certain products of conversion are strongly adsorbed and retard the reaction, and formation of the inhibiting materiala is directly proportional to hydrocarbon conversion. The resulting equation is dx/dt =
k(1 - x ) 1 K‘x
~
+
from which
where x is the fraction of the charge converted in t seconds, k is the conversion rate constant (sec.-l), and K’ is the poisoning constant. In all cases the plots of ( x ) / t against l / t log[l/(l - x ) ] were straight lines. Table VI shows the values of k and K‘ obtained from these plots. The approximate values arise from the fact that very small changes in slopes near 1.O give rise to large changes in k and K’. The differences among catalysts and hydrocarbons in conversion rates are more obvious from this comparison than from the graphs of conversion LIS. contact time. Of particular interest is the great superiority of the fluorided catalyst in the rate constant for both hydrocarbons, in spite of its relatively low surface area.
Table VI. Approximate Conversion Rate and Poisoning Constants Rate Poisoning Constant, Constant, K . Set.-' K‘ MCH SiOdMgO Si02/AlzOa SiOz/AIzOa/F
0.33
55
2.5+ 6.8
loo+
0.07
3.5
45
n-Decane SiOdMgO Si02/A120a SiOz/AIzOa/F
0.19 1.4+
7.0
loo+
The poisoning constants are high for all three catalysts in the cracking of methylcyclohexane, but only for the fluorided catalyst in n-decane cracking. One possible conclusion is that only Lewis acid sites are affected by the inhibitors. I n this case a corollary conclusion is that silica-alumina-fluoride cracks n-decane by a Lewis acid mechanism, but the other catalysts do not. It is further indicated that the inhibitors involved cannot be simply the coke formed, because the fluorided catalyst is strongly inhibited in converting both hydrocarbons. Coke formation, however, is almost nil in cracking methylcyclohexane. Silica-Alumina US. Silica-AluminaFluoride. These two catalysts are very similar in their selectivity to most products from both n-decane and methylcyclohexane. Their biggest selectivity differences are in hydrogen production from both hydrocarbons and coke from methylcyclohexane. Their principal differences are in reaction rates. The most important . reactions are carbon chain splitting, hydrogen transfer, and isomerization. The fluorided catalyst gives much higher rates for all three with methylcyclohexane. With n-decane the over-all conversion rates are fairly similar, although rate constants (Table VI) are different, because of the much greater poisoning effect on silica-alumina-fluoride. The authors attribute these rate differences to a considerably greater concentration of active Lewis acid sites on the fluoridetreated catalyst. Silica-Magnesia US. Silica-Alumina Catalysts. Here rate differences are further exaggerated. In converting methylcyclohexane the difference between silica-magnesia and silica-alumina is greater than between the latter and fluoride-treated silica-alumina. Silicamagnesia gives the lowest catalytic rates for all three of the important reactions listed above. O n the basis of the reaction mechanism postulated for methylcyclohexane this means that silica-magnesia has a very low concentration of Lewis acid sites of the necessary acid strength.
The differences in product selectivity between silica-magnesia and the other two catalysts are very great; and they stem from the reaction rate differences, particularly the cracking and hydrogen exchange rates. Compared with the others, silica-magnesia has low degree of product saturation (paraffin-olefin ratio), low degree of isomerization, and high hydrogen and methane production. Acknowledgment
The authors wish to express their appreciation to members of the Analytical Section of Socony Mobil Laboratories for help in obtaining the data quoted in this paper. In particular, the help of Dorothy Weber in calculating the mass spectrometer analyses is gratefully acknowledged. Literature Cited
45, (1) Benton, A. F., J . Am. Chem. SOC. 887 (1923). (2) Bloch, H. S., Thomas, C. L., Ibid., 66.1589 (1944). (3) Conn. A. L:. Meehan. W. F.. Shankland, R. V., Chem. ’Eng. Phgr. 46, 176 (1950). (4) Egloff, G., Morrell, J. C., Thomas, C. L., Bloch, H. S., J. Am. Chem. Sac. 61, 3571 (1939). (5) Gladrow, E. M., Krebs, R. W., Kimberlin. C. N.. IND. END. CHEM.45,’142 (1953). (6) Gladrow, E. M., Krebs, R. TY., Kimberlin, C. N.,. Preprints, General Papers, Division of Petroleum Chemistry, pp. 43-55, 124th Meeting, ACS, Chicago, Ill., September 1953. (7) G. M.. Voce. H. H.. Greens. , Good. felder, B. S , , IN&.’ENG.CHEM.39, 1032 (1947). (8) Greensfelder, B. S., chapter in “Chemistry of Petroleum Hydrocarbons,” vol. 11, Reinhold, h’ew York, 1955. (9) Greensfelder, B. S., Voge, H. H.,, IND.ENG.CHEM.37.514 (1945). (10) Ibid.. D. 983. (11) 2bid.l b. 1038. (12) Greensfelder, B. S., Good, G. M., 1 (1945). (13) Ibid.,41,2573 (1949). 114) Hansford. R. C.. chaDter in “Phvsical ’ Chemistry of ’Hyd>ocarbons,’; vol. 11, Academic Press, New York, 1952. (15) Hyman, J., U. S. Patent 2,331,479 (1943). (16) Laidler, K. J., chapter in “Catalysis,” vol. I, ed. by P. H. Emmett, Reinhold, New York, 1954. (17) Marisic, M. M., U. S. Patent 2,385,217 (1945). (18) Milliken, T. H., Jr., Mills, G. A., Oblad, A. G., Discussions Faraday SOG.No. 8,279 (1950). (19) Reyerson, L. H., J. Am. Chem. SOC. 5 5 , 3105 (1933). (20) Thomas, C. L., Ibid., 66, 1586 (1944). (21) Voge, H. H., Good, G. M., Greensfelder, B. S., IND.ENG.CHEM.38, 1033 (1946). ,
>
,
RECEIVED for review March 6, 1956 ACCEPTED October 6, 1956 VOL. 49, NO. 4
e
APRIL 1957
749
