1
Engineering Approaches
[
High-Temperature Pyrolysis of n-Butane
I
SAMUEL SANDLER and YU-HO CHUNG Department
of Mechanical Engineering, University of Toronto, Toronto, Ontario, Canada
Be w a r y of activation energy values determined under static conditions. Reaction rates are a function of heat transfer. Data should b e taken under conditions similar to use, as shown in this example for cracking n-butane
MOST
*
of the work reported in the literature concerning the kinetics of the thermal decomposition of n-butane has been carried out in the low-temperature region between 550' and 650' C. A generally accepted value of 58,700 cal. per gram mole for the activation energy of this reaction was obtained by Steacie and Puddington (73) using a static method. I n flow-system studies in this temperature range, activationenergy values varying from 61,400 to 73,900 were reported (6, 8, IO) and figures as low as 43,000 were calculated from data obtained from similar systems operated a t higher temperatures and very short residence times (4, 72). After considering the heat-transfer aspects of externally heated tubular reactors used for hydrocarbon pyrolysis, Calderbank (3) suggested that measurements of reaction rates under such conditions are, in reality, measurements of rates of heat transfer. This is due to the existence across the reactors of large temperature differentials resulting from the endothermic nature of the reactions. It follows that there should be a difference in the values of the activation energy as determined by static and flow methods, especially at high temperatures. Unfortunately, inherent difficulties in the application of the static method to high-temperature, low-residence-time studies do not permit a ready comparison of these two methods in the conditions that are of most interest commercially. I n the present investigation, which is a continuation of previous work on the thermal cracking of n-butane in the temperature range of 700' to 900' C. (7), the best value for the rate constant of the reaction is given by the expression
The values of the activation energy and frequency factor would appear to be low when compared with those normally accepted as valid. However, it is necessary, in practice, to modify the latter in order to correlate actual conversion data in commercial installations with predicted conversions. Thus, for example, Andrews and Pollock ( 7 ) , using a digital computer for tube design in a light-hydrocarbon cracking furnace, found that butane conversions predicted from some laboratory data (73) did not correspond to conversions actually obtained, An adjustment of the frequency factor in the Arrhenius equation was necessary for satisfactory predictions. Therefore, pyrolysis-coil designers should be wary of the indiscriminate use of activation-energy values normally obtained from bulb-method studies. There is also considerable justification for re-examining the methods used in kinetics studies, especially in high-temperature investigations. In any event, if the flow method, as generally applied, does not give a real value of the activation energy in a true kinetics sense because of a dependence on the heattransfer characteristics of the system, it may be advisable in pyrolysis-coil design to use the apparent value obtained under comparable temperature and flow conditions.
Experimental A flow method was employed. Tubular and annular reactors of Vycor glass were placed in short furnaces in which large temperature gradients existed and also in a long furnace having a constanttemperature zone in its center section. I n addition, experiments were performed using a Vycor tube packed with Vycor
filaments, a silica tube, a chromiumplated steel tube, and an annular reactor of porcelain, to determine the effect of surface area and composition on pyrolysis product distributions and rates of reaction, The nonisothermal furnaces had been used in a previous investigation (7) to find the conditions necessary for increased production of useful olefins. The more-refined isothermal furnace was adopted to verify the low value of the activation energy obtained in the previous work. Temperatures were measured by using thermocouples of Chrome1 and Alumel. In the nonisothermal furnaces, thermocouples were tied along the outside walls of the reactors a t half-inch intervals. At times, a single thermocouple, arranged so that it could be readily shifted along the tube wall from the outside, was used to measure the longitudinal temperature gradient. I n the isothermal reactors, gas-temperature readings were obtained, in preliminary runs only, by placing directly into the gas streamt a very fine thermocouple shielded from wall radiation effects by using silver foil. This thermocouple was removed during an actual run when samples were taken for analysis to prevent disturbance of the flow pattern and complications arising from possible reaction catalysis by the thermocouple materials. I t is felt that temperature readings were accurate to within f2' C. of the actual wall temperatures measured. I n the isothermal reactor, which was equipped with a small preheating section, the measured gas temperatures and the wall temperatures were nearly identical under low-conversion conditions. However, this may be due to a compensation of errors, and the actual gas temperatures may have been up to 5" C. lowerVOL. 53, NO. 5
*
MAY 1961
391!
I.
Table
Frequency Factors and Activation Energies Were Obtained with Various Reactors under Isothermal and NonIsothermal Conditions Furnace Reactor Surface
Vycor
Annular
Porcelain
Tubular
Silica
Annular, capillary ends
Vycor
Tubular, capillary ends
Vycor
Dimensions
Type
Inner tube O.D. = 17 mm. Outer tube I.D. = 19 mm. Length = 12 in. Inner tube O.D. = 24 mm. Outer tube I.D. = 29 mm. Length = 14 in. I.D. = 16 mm. Length = 14 in. Inner tube O.D. = 17 mm. Outer tube Outer tube I.D. = 1 9 m m . I.D. = 19 mm. Length = 4in. I.D. = 13 mm. Length = 4 in.
Length, inches
Flow Rate, Temperature Ml./Min. Range, C. (Ambient)
Energy,
650-790
300
10.33
46,700
Nonisothermal
7.5
681-769
500
2.20
42,700
Nonisothermal
7.5
620-692
350
2.52
44,300
Isothermal
15
638-696
150
9.55
45,600
Isothermal
15
628-701
150
2.65
44,600
Results
Table II.
INDUSTRIAL AND ENGINEERING CHEMISTRY
Cal./G. Mole
5
actors, a more refined method was developed. This will be described in a separate publication. However, the assumptions involved are: constancy of rate of supply, relatively constant difference between outer-wall temperature and gas temperature, and operation a t low conversions to avoid secondary reactions. The second assumption is valid at low over-all conversions and constant feed rate. These are the conditions where the temperature drop due to reaction is not large, and the heat transfer coefficient is unchanged. Table I1 shows that, at the same outerwall temperatures, much shorter residence times are always required in the annular reactor to give the same conversions as in the tubular reactor. This may be attributed to better heat transfer in the narrow annular reactor: where the gas temperature is much closer to the wall temperature. The best estimate of reaction temperature, residence time, and apparent activation energy was obtained from the results of the experiments in the Vycor annular
392
Activation
Nonisothermal
than the corresponding wall temperatures under these conditions. Technical grade, 95%, and pure grade, 9970, n-butane in pressurized cylinders were obtained from the Phillips Petroleum Co. and used without further purification. The former was found to have a purity in excess of 97%. The only other constituent in both grades was isobutane. The reaction products were separated into several fractions by distillation in a semirobot Podbielniak low-temperature fractional distillation column. The resulting fractions containing methane plus hydrogen, ethane plus ethylene, propane plus propylene, and the butenes plus butanes were collected and subsequently analyzed by chemical methods. Hydrogen and methane were determined by combustion. Ethylene, propylene, and the butenes, in their respective fractions, were absorbed in a 25yo solution of sulfuric acid saturated with mercuric sulfate. The analyses were checked from time to time by mass spectrometry to determine the quantity of higher boiling components which might have been formed.
The values of the activation energy and frequency factor which were obtained in the various reactors under isothermal and nonisothermal conditions are summarized in Table I. In addition to the normal method of estimating average residence times and temperatures for first-order velocity constant calculations (2) in nonisothermal re-
Frequency Factor, Sec.-l X 10-9
reactor equipped with capillary inlet and outlet tubes. The over-all rate of the reaction in this high-temperature range may therefore be calculated from Equation l , which correlates the above information. The noncatalytic effect of the Vycor surface was demonstrated by using a Vycor tube packed with finely drawn Vycor threads. Table I11 gives a comparison between the conversions in this reactor and those in an annular reactor at similar temperatures. Despite the larger surface-to-volume ratio of the packed tube, there is actually a slight decrease in the conversion when compared with that obtained in the annular reactor. The product distributions in both reactors were almost identical. The effect of temperature on the product distribution in the effluent of an annular reactor is shown in Figure 1. The curves indicate that the methane, ethylene, ethane, hydrogen, and butenes contents of the effluent increase with temperature at characteristic rates for each substance. O n the other hand, the
Because of Better Heat Transfer, the Annular Reactor Gives Higher Conversion Than the Tubular Reactor
No.
Reactors
Flow Rate, Ml./hIin. (25' C., 760 M m . H g )
967 995 969 997 970 982
Tube Annulus Tube Annulus Tube Annulus
150 150 150 150 150 150
Expt.
Wall Temp., O C.
Conversion,
%
Residence Time, Sec.
641 638 667 663 701 696
8.5 8.0 14.0 14.5 28.0 27.1
1.54 0.809 1.45 0.76 1.297 0.686
n-BUTANE P Y R O L Y S I S
c
propylene content reaches a maximum a t 810° C. Figure 2 gives a comparison of the product distributions obtained a t temperatures of 700' and 750' C. in a tubular reactor under isothermal conditions. The results indicate that the proportions of products formed a t different temperatures are slightly different. Higher temperatures favor the formation of ethylene while lower temperatures favor the formation of propylene. The important products are methane, hydrogen, ethylene, ethane, propylene, and various butenes. I n Vycor reactors, very small quantities of higher boiling products consisting chiefly of C S saturated hydrocarbons were detected by mass spectrometry. The total quantity of these substances did not exceed 1.5 mole % in the effluent at high-temperature conversions rising to 90%. Propane was also detected, but constituted less than 0.5% of the effluent at any time. Coke formation in the Vycor reactors was never excessive. Ample evidence of this is provided by the close approach of the carbon-hydrogen ratio of the gaseous products to the theoretical value of 0.400. The effect of some surfaces, notably stainless steel, to increase the rate of formation of secondary reaction products and cause excessive carbon deposition even a t moderate temperatures has been observed and described in detail (7). Chromium-plated stainless steel tubes behaved like Vycor tubes to a great extent in this respect. T h e present investigation has confirmed this behaviour. Some investigators (4, 5, 72) have reported that the molar quantities of ethylene produced were generally higher than those of ethane. They did not believe that the secondary process involving ethane dehydrogenation could account for this result and therefore concluded that the reaction : CaHio
2CzH4
+ Hz
Table 111.
--
0.25(CaHio O.O6(C4Hio
0.07(C4Hio .-t
+ C3Hs) CzHe + CzHa) 2C2H4 + Hz) C ~ H+ I Hz) CHI
Flow Rate, Ml./Min. (25' C., 760 Mm. Hg)
Expt.
No.
Reactor
1011 996 1012 98 1 1013 982
Packed
Tube
Packed
Tube
Annulus Packed
Wall Temp., O
300 150 300 150 300 150
Annulus
Tube
Annulus
c.
652 653 678 681 704 696
Conversion, yo
Residenoe Time, See.
9.4 11.2 16.0 21.0 25.3 27.1
1.09 0.79 1.02 0.719 0.945 0.686
A = Surface area, V = Volume, ml. A/V for Packed Tube = 580/17.8 = 32.6 ern.-' A/V for Annular Reactor = 129/6.46 = 19.9 om.?
MAXIMUM OUTER
Figure 1 .
770
730
690
810
85C
WALL T E M P E R A T U R E
- OC.
The effect of temperature of the effluent product distribution
Vycor annular reactor in 5-inch furnace.
n-Butane supplied at 300 ml. per min.
(2)
was a primary one. In the present work, the same effect has been observed. The relative amounts of the products formed in the various reactors differ slightly a t the same conversion owing to differences in temperatures and residence times required. Nevertheless, by extrapolating to zero conversion, as in Figure 3, the proportions of the primary reaction products may, in general, be represented by the following series of reaction equations : 0.62(C4H10
The Vycor Surface Does Not Act as a Catalyst
9
(5) (6)
440
2 v)
W
10
-I 0
rn
cy;;
IO
20
30
40
10
Yo DECOMPOSITION
(3) (4)
I
40
Figure 2.
6
I , , , , I
IO
20
30
40
SO
60
% DE C 0M P 0 S I T IO N
The effect of conversion on product selectivities a t temperatures of
700' and 750' C. Vycor tube, 13 mm. I.D., with capillary ends, in isothermal portion of a 12-inch furnace. supplied a t rates of 50 to 600 ml. per min.
VOL. 53, NO. 5
n-Butane
MAY 1961
393
0
w
90
-
80
-
70
-
50
-
v)
0 0
r
Hinsheiwood and his coworkers have supported the view that the thermal decomposition of paraffin hydrocarbons occurs by simultaneous chain and nonchain mechanisms (9, 74). The latter can be considered as direct molecular rearrangements, the result of which is the transfer of a hydrogen atom from one carbon atom to an adjacent one with a rupture of the bond between the two. The molecular reaction has a lower activation energy but a rather less favorable entropy factor, since the movements of the hydrogen and carbon atoms have to be coordinated to some extent. With these considerations in mind, the decomposition of n-butane in the low-temperature range may be considered to involve mainly the reactions described by Equations 3, 4, and 6 , above, while in the higher temperature ranges, another primary reaction, given by Equation 5, becomes important.
0 0 W
n v)
W -I
0
1E 0
0a W
n v)
W -I
0 I
I
I
I
I
I
I
I
1
I
I
IO
20
30
40
50
60
70
80
90
Oo /
Figure 3.
DE C 0 M PO S I TI 0 N
The effect of conversion on product selectivities
Chromium-plated steel tube, 2 8 mm. I.D., in 7.5-inch furnace.
of activation approaching zero for the reaction,
Discussion
The decomposition of paraffin hydrocarbons has been considered to proceed as a chain reaction involving free radicals ( 7 7 ) , the existence of which has been definitely established by the use of the techniques of nitric oxide inhibition and sensitized decomposition. For butane, the mechanism involved has been postulated as follows : C4H10+2R.E1 = 84kcal.
R.
+ C4Hio
n-Butane supplied at 5 0 0 ml. per min.
-+
+ H*
(13)
would account for the lower apparent energy of activation for the over-all reaction, there does not appear to be any theoretical or experimental basis for such an assumption. The aiternative reactions, C4Hg.--tC*H,.
+ C l H 4 E= 26kcal.
(14)
+
RH C4H9. Ez = 5.5 or 9 kcal.
C4Hg.+ R .
(7)
C4HQ.+ 2C2H4
+X E ,
=
26 kcal.
(8) (9)
By applying the usual steady state treatment, the first-order expression becomes :
would give an even higher figure for the calculated energy of activation. I t is also unlikely that a catalytic influence by the material of the reactor wall is responsible for the lowering of the activation energy because the same low value is obtained in reactors made of Vycor glass, silica, chromium-plated stainless steel, and porcelain. In addition, the noncatalytic effect of the Vycor surface has been demonstrated by experiments with a packed reactor. However, if the bond dissociation energy of the reaction C4H10 + 2CzHs.
and the energy of activation is:
(12)
which is in agreement with that of Steacie and Puddington (73) but is higher than our observed value. Although the postulaton of an energy
then the over-all activation energy would be lower i n the above calculation.
+ Ez + E3 -
E4)
=
58 kcal.
394
INDUSTRIAL AND ENGINEERING CHEMISTRY
The experiments described were carried out in the Department of Mechanical Engineering, University of Toronto, lvith the cooperation of G. R. Lord, head of the department, R. R. McLaughlin, dean of the Faculty of Applied Science, and R. 0. King, director of the Combustion Research Section. Literature Cited
(1) Andrews, A. J., Pollock, L. W., IND. ENG.CHEM.51, 125 (1959). (2) Benton, A. F., J . Am. Chem. SOC.53, 2984 (1931). (3) Calderbank, P. H., Chem. Eng. B o g . 50, Symposium Ser. No. 9, 53 (1954). (4) Cambron, A., Bayley, C. H., Can. J . Research 9, 173 (1933). (5) Crawford, V. A., Steacie, E. W. R., Can. J . Chem. 31, 937 (1953). (6) Frey, F. E., Hepp, H. J., IND.END. CHEM.25, 441 (1933). (7) King, R. O., Sandler, S., Chung, Y . H., Trans. Eng. Znst. Can. 3, 1 (1959). (8) Paul, R. E., Marek, L. F., IND.ENG. CHEM.26, 454 (1934). (9) Peard, M. G., Stubbs, F. J., Hinshelwood, C. N., Proc. Roy. SOC. (London) A214, 471 (1952). (10) Pease, R. N., Durgan, E. S., J . Am. Chem. Soc. 52. 1262 119301. (11) Rice, F. O., J . Am. Chhern. SOG.53, 1959 119311.
(16)
is less than 84 kcal. per gram mole, a value which has been assumed to be the same as that measured for the dissociation
E = l/z(Ei
Acknowledgment
RECEIVED for review September 6, 1960 ACCEPTED February 6, 1961 This work was sponsored by the Defence Research Board of Canada in accordance with Defence Production Contract No. CD800102. Part 37 of Investigation of the Mechanism of the Oxidation, Decomposition, Ignition, and Detonation of Fuel Vapors and Gases.