KINETICS OF THE CATALYTIC DEHYDRATION OF PRIMARY ALCOHOLS JOHN E. STAUFFERl AND W I L M E R
L. K R A N I C H
Worcester Polytechnic Institute, Worcester, Mass.
The reaction kinetics for dehydration of the homologous series of primary alcohols from ethyl to hexyl over ?'-aluminum oxide has been studied. The objective was to find a means of predicting unknown kinetic behavior of CI compound in a catalytic reaction from known data on a homologous compound with the same catalyst. Reaction rates for the entire series for dehydration to olefins could be predicted with a maximum error of 327; and an average error of 15% over the temperature range from 3 4 8 " to 4 2 8 ' C. using a single semiempirical equation. The activation energy for all members of the series was 30.8 kcal. per gram-mole.
many years invesiigators have been reporting the kinetics of particular surface reactions on solid catalysts. In more recent years, Eyring arid others have tried to explain kinetic behavior on the basis of sound basic theory. These efforts have been wholly successful in only the simplest cases. The present study investigates systematically the kinetics of a series of similar reactions on a single catalyst, in order to develop a semiempirical relationship which applies to the entire series. If this type of relationship were broadly available, the prediction of reaction rates for a number of chemical reactions in a homologous series would be possible-based on limited data on only one or two members of the series. If later investigations of other reactions and other catalysts shoiv similar relationships, broader generalizations may be possible and empiricism may yield at least in part to theory. The dehydration of primary alcohols to olefins and ethers over y-aluminum oxide catalyst was selected for this work because the kinetics of several members of the series had been studied and the reaction mechanism had been widely discussed. The side reactions are minor and the analyses can be accomplished by established chromatographic techniques. OR
Apparatus
To simplify interpretation of the data and improve accuracy, a shallow-bed (differential) type of reactor was used as recommended by Hougen and \Vatson (3) and Smith (72). While this type of reactor avoids the problems of changing temperature, pressure, and concentrations within the reactor, it is limited to small conversions per pass. Accordingly a recirculation system was provided to make the results less sensitive to 3rrors in anal)-sis, to increase the reactor throughput for a given fresh feed rate, and consequently to decrease the significance of mass transfer to the catalyst. I n the system as designed, the fresh feed was metered into the recirculating stream, preheated, and fed to the reactor. The exit stream from the reactor was split into two parts. hlost of the gas was recycled by a pump, to join with the feed
stream. The rest passed uncondensed through a sample valve and was vented. M'hen a n analysis was made, the sample valve introduced a small sample of gas into the chromatograph. A flowsheet of the apparatus is shown in Figure 1. A 3, 8-inch nominal pipe size nipple of stainless steel was used for the reactor. The diameter of the reactor was the minimum practicable to obtain the maximum flow rate per unit area Ivithout introducing a \\.all effect or channeling. Thus, the effect of diffusion upon the over-all reaction rate was reduced. Thermocouple wells were inserted into the inlet and exit ends of the reactor. 4 hot air oven maintained the reactor and a preheating tube coil at a desired temperature. T o extend the capabilities of the differential reactor, a recycle pump was incorporated into the design to return part of the exit stream from the reactor to the feed stream. The gas composition within the reactor could be varied by holding the recycle rate constant and adjusting the alcohol feed rate. The recycle design also facilitated the analysis of the exit stream by increasing the over-all conversion of feed with only a relatively small conversion per pass. TVith this apparatus the rate of reaction could be calculated from measurement of the alcohol feed rate and analysis of the exit gas stream. A recycle pump of original design was built (Figure 2). Two stainless steel belloivs of 2,006-inch O.D. (Robertshaw-
HELIUM VEHT
ROTAMETER
Present address, Stauffer Chemical Co., Richmond, Calif.
FEED
r_L/FEEI
PUMP
MUCTNITY
RED) VAL
ELL
A
RECYCLE
T I SFACTOR G A S VENT
1
ILCOiOL
S W L E VALVE
Figure 1.
Flowsheet of apparatus V O L . 1 NO. 2 M A Y 1 9 6 2
107
Figure 2.
Recycle pump
surface of the catalyst pellets-i.e., diffusion to the surface was not controlling. Because a small amount of coking. was observed on the catalyst pellets, fresh catalyst was charged for investigating each alcohol. The initial run for each alcohol was repeated at the end of the series of runs Tor the alcohol to check for changes in catalyst activity. At constant recycle rate, the ethanol feed rate was varied a t 388' and 348O C. to study the effect of alcohol concentration upon the reaction rate. In this manner the ethanol concentration was varied between 36.8 and 85.8%. Each alcohol was dehydrated a t close to 428', 388', and 348' C. With hexyl alcohol a fourth temperature, 319.5O C., was also examined. The pressure was essentially atmospheric far all experiments. At least five analyses were made for each set of conditions. The feed rate was the same for all alcohols at each temperature, hut it was increased when the temperature was raised to compensate for the increased rate of dehydration. I n this manner an attempt was made to hold the alcohol concentration well within the range over which it was investieated as an independent variable. Results and Discussion
Fulton Controls Go.) were used to displace the gas. Ball check valves were machined out of a single block of stainless steel to control the flow of gas to and from each bellows. The hellows were driven hy two eccentric cams 180" out of phase. Because all-stainless-steel construction was used far parts in contact with the hot gas, contamination of the gas by gasket material, packing, sealants, or lubricants was avoided. Since a positive displacement pump was used, the recycle rate could be directly determined. Far most of the experiment the pump was operated a t 31.6 r.p.m., a t which speed the capacity was 0,080 cu. foot per minute. One disadvantage of the fceed and recycle pump design was that the flaw ratewas pulsating. These pulsationsundouhtedly reduced the accuracy of the data. Considerable effort was spent in finding a column material for the chromatograph that would minimize the tailing of the water peak and a t the same time completely separate the components of the gas sample. The best column material tested was Renex 678 (a polyoxyethylene alkyl aryl ether supplied by Atlas Powder Co.) deposited on acid-washed Celite 545. Procedure
Five primary alcohols--ethanol, I-propanol, I-butanol, 1pentanal, and 1-hexanol-were dehydrated over the catalyst. The purity of each was checked by the chromatograph. The y-alumina catalyst (Harshaw Chemical Co.) was in the form of 1/8-inch tablets of 98% A1,Oa with a surface area of approximately 80 sq. meters per gram. T o check for thermal dehydration, a hlank run was made in which no catalyst was placed in the reactor. The temperature of the reactor was 429' C., the highest used in any run. Hexanol was selected as the feed for this test because it exhibited the highest rate of catalytic dehydration to olefin of all the alcohols. Not a trace of hexene or hexyl ether was detected in the blank run. The effect of the recycle rate was investigated by making one run with a recycle flaw rate of 0.155 cu. foot per minute compared with the usual rate of0.080. Since the results were the same for the fast and slow recycle flow rates within the accuracy ofthe analysis, itwas concluded that the recycle rateof0.080 cu. foot per minute was sufficiently great to avoid the effects upon the over-all reaction rate of diffusion between the gas and the 108
I&EC FUNDAMENTALS
Because of the pulsations of the feed and recycle pumps, the results of single analyses were subject to considerable scatter. It was determined statistically, however, that when groups of five successive analyses were made a t a given set of conditions their average reduced this scatter to within the limits of other experimental errors. Accordingly, all data points presented are actually the averag-es of five or more chromatographic analyses. I n the preliminary investigation made to determine the effect on the reaction rate of concentration OS ethanol in the feed, no significant variation was observed from 36.8 to 85.8%. I t may thus he concluded that the adsorption af reactant on the catalyst is not the rate-controlling step over the range of conditions studied and that the avai!able reactive sites on the catalyst can be essentially saturated with reactant over the range of concentrations investigated. This is in agreement with the conclusions of Maurer and Sliepcevich ( 7 ) , Miller (X), and Laihle ( 6 ) . While this variable was not independently studied for all alcohols, the results support the conclusion that for all 01 the alcohols studied the controlling step is (with minor exceptions) the rate ofsurface reaction. T h e dehydration of alcohols proceeds almost exclusively to two products, ethers and olefins. I n almost no cases did the chromatograph pick u p significant quantities of isomeric alcohols or other materials. The two important reactions are then:
-
RCHzCHzOH
RCH=CHa
+ H20
-- RCH2--CHzO-CH~.CHsR
2 RCH2CH90H
+ Hs0
(1)
(2)
The experimental results can he analyzed in terms of the over-all alcohol convenion or the rates of the separate reactions. Pease and Yung (9) have shown for ethanol that the rate of Reaction 2 is rapid compared to that of Reaction 1. Furthermore, although Reaction 1 is substantially irreversible a t the conditions studied, Reaction 2 is complicated by the fact that equilibrium concentrations are rapidly approached. The higher the alcohol, the lower is the equilibrium concentration of ether. I n view of the fact that the reverse of Reaction 2 is probably significant under the conditions of this study, it is more promising to direct principal attention to Reaction 1, where only the forward reaction is involved.
For each alcohol the reaction rate with respect to olefin formation, r, in units of gram-moles per hour per gram of catalyst was plotted logarithmically us. the reciprocal of the absolute temperature (Figure 3). I n this study an over-all reaction rate constant was employed. Since this constant was found to be insensitive to variables other than temperature over the range investigated, a more elaborate breakdown into adsorption, reaction, desorption, and diffusion steps was not considered justified. I'Jith the exception of one run for hexanol and one for pentanol, the points for each alcohol lay on a straight line. Thus the reaction obeys the Arrhenius exponential lair or the more fundamental Eyring theory for the case of gaseous reactions. . b o t h e r striking feature of Figure 3 is that the lines for all alcohols have the same slope, and therefore the energies of activation are the samc for each alcohol. This suggests that the alcohols were dehydrated by the same mechanism and that the same bond ruptures were involved. The slope of the lines was measured, and from this value the energy of activation was calculated to be 30.8 kcal. per gram-mole of alcohol. Schwab and Sch\+-ab-.\gallidis (70) reported that the energy of activation was 31 .O kcal. per gram-mole for the dehydration of ethanol over y-aluminum oxide Kabel and Johanson ( g ) found 30.5 kcal. per gram-mole as the activation energy for dehydration of ethanol to diethyl ether over Dolvex 50 ion exchange resin for temperatures from 80' to 120' C. This may .indicate that the breaking of the same bond may be the rate-controlling step over a wide temperature range and over \+-idely(different catalysts whether the product is olefin or ether. Further work would be required for confirmation. The points for the rates of pentanol and hexanol dehydration to olefin at 423.5' C. were below the position predicted by the Arrhenius exponential law. 4 n explanation was that at the faster reaction rates for these higher molecular weight alcohols at the higher temperatures, the influence of diffusion to the surface or of pore diffusion may have become significant as compared to the surface reaction rate. In accordance with the main objective of the study, the relationship between the reactivity of the alcohol and the position of the alcohol in the homologous series was investigated. I n Figure 4 the reaction rate of olefin formation in units of grammoles per hour per gram of catalyst was plotted logarithmically us. the number of carbon atoms in the alcohol. The temperature was arbitrarily chosen as 383' C. (Selection of another temperature would have changed only the height of the curve.) The reactivity of the alcohol increased with molecular weight. LVith the exception of Adkins and Perkins (I), all researchers have qualit,atively observed the same results ( 2 , 5, 6, 8,7 7 ) . The equation of the straight line which best fitted the data was determined by the rnethod of least squares and found to be 0.0019j
1
I
0-\
L
I \I
1 1
I
BUTANOL PLYTAUOL "EXIYOL
0
.
0
0
U I
.. w 0 0 IW
2
.01
z
0 U IY
a
I
,001
,0014
14 ,0016
.0011
I/T,
Figure 3.
1
.m,r
*K-'
Rate of olefin formation as a function of temperature
0
U
\ Y
0 0
: 01
4
a z
0 u w
eo.4iz
.v
where I is the rate of reaction and is the number of carbon atoms in the alcohol. The maximum error in r incurred by approximating the data with the straight line was 32% (for propanol), and the average error for all the alcohols was 15%. The effect of the number of carbon atoms in the alcohol and the effect of temperature upon the reaction rate were combined into the following equation: r = 3.65
x
107 e0.47Z
S
-
30.81RT NUMBER
where r equals the rate (of olefin formation N is the number of carbon atoms in the alcohol, R is the gas law constant, and T i s the absolute temperature.
CARBON
ATOMS
Figure 4. Rate of olefin formation a t 383" C. as a function of number of carbon atoms in alcohol molecule VOL. 1 NO. 2 M A Y 1 9 6 2
109
The maximum error in applying the equation to an estimate of reaction rates over the range of temperature and molecular weight studied was 32%, and the average error was 15%. Because of the high precision of the correlation of log r us. 1/T, these errors are the same as those associated with the graph of log 7 us. N . The net forward reaction rate with respect to ether formation was investigated at the same conditions as for olefin formation. In Figure 5 the rate is plotted logarithmically us. 1 / T for each alcohol. There was a marked irregularity of the data. Nevertheless one trend was apparent: The net rate of ether formation decreased with the increasing molecular weight of the alcohols. Three phenomena which could explain the results were pore diffusion controlling, reverse reaction, and steric effects of the alcohols in the surface reaction. Not enough was known about any of these phenomena to draw a conclusion. Equilibrium is almost certain to be of significance, since ether formation, as contrasted with olefin formation, is markedly reversible. I n Figure 6 the total decomposition of the alcohols to both olefins and ethers was plotted logarithmically us. 1 / T . Because the rate of olefin formation increased and the rate of ether formation decreased with increasing molecular weight of the alcohol, the rates of total decomposition for each alcohol should be closer together than those for olefin formation. This effect was observed, but there was an appreciable spread of the data. The data can be represented within 1 5 0 % by the single line shown, or by the equation for this line: lT =
97,800
where rT is the total rate of decomposition of alcohol. While further work will be necessary to relate the variation of the Arrhenius "frequency factor" with number of carbon atoms to a sound theory, some speculation is justifiable at this point.
"
Topchieva and Yun-Pin (13) and Adkins and Perkins ( I ) claim that olefins are not formed by a series reaction, alcoholether-olefin, but that Reactions 1 and 2 are truly independent. The former suggest that a single surface complex .41- OCHgR may break down, after surface rearrangement, into either olefin or ether. This suggests that a given catallst site may serve for either Reaction 1 or Reaction 2. The relative probabilit? that a surface complex would break according to Reaction 1 would then be equal to the reaction rate constant for Reaction 1 divided by the sum of the rate constants for Reactions 1 and 2. The reverse reaction should have no influence on this split between possible forward reactions, since it uould simply regenerate the common surface complex. -4lthough the rate data on ether formation given in Figure 5 are almost certainly obscured by the presence of the reverse reaction. they indicate the comparative behavior of the alcohols. This is best indicated at the lower temperature, where less foru ard reaction occurs during a pass through the catalyst bed and therefore the reaction is probably farther from equilibrium. At the lowest temperature, the rate of formation of ethyl ether is evidently considerably greater than that of the higher ethers. This means that the relative probability that a given site will enter selectively into Reaction 1 is less for ethanol than for the higher alcohols. Thus, per gram of catalyst. the number of sites participating in Reaction 1 is considerably smaller for ethanol than for the other alcohols. One might reason that the true reaction rate constant for Reaction 1 should be based fundamentally on the number of catalyst sites participating in this reaction. Since this number apparently is smaller for ethanol than for the other alcohols per unit weight of catalyst. the Reaction 1 rate constant for ethanol, if converted to a basis of the number of participating sites, would be raised more than the corresponding constant for
7
. ,001 ,0011
,0015
I/T,
,0016
,0017
.K-'
Figure 5. Rate of ether formation as a function of temperature 110
I&EC FUNDAMENTALS
I
/
T, * K - I
Figure 6. Rate of alcohol decomposition to olefin and ether as a function of temperature
the other alcohols, thus tending to bring the parallel lines of Figure 3 closer together. The data are inadequate to permit quantitative corrections, but they suggest that the Arrhenius frequency factors based on the number of sites participating in Reaction 1 may be much closer for the various alcohols than the more usual (and more practical) weight basis.
literature Cited
(1) Adkins, Homer, Perkins, P. P., J . Am. Chem. SOC. 47, 1163-7 (1925). (2) Brown, A. B., Reid, E. E., J . Phys. Chem. 28, 1077-81 (1924). (3) ,Hougen, 0. A., Watson, K. M., "Chemical Process Princiules." Vol. 111. Wilev. New York. 1947. (4) Kabel, R. L.,' Johinson, L. N.; Preprint 48, 54th Annual Meeting A. I. Ch. E., New York, 1961. (5) Komarewsky, V. I., Stringer, J. T., J . Am. Chem. SOC.63, 921-2 (1 941 \ \ - - '-/'
Acknowledgment
The authors express appreciation for very generous financial support from the Esso :Educational Foundation. Nomenclature
S = number of carbon atoms reaction rate of individual reaction, gram-moles per hour per gram of cat,alyst r T = reaction rate for gross decomposition of alcohol, same units as r R = gas constant = 1.98 X lop3 kcal./(gram-mole) (" K.) T = absolute temperature, " K. t
=
(6) Laible, J. R., Dissertation Abstr. 20, 974 (1959). (7) Maurer, J. F., Sliepcevich, C. M., Chem. Eng. Progr. Symposium Ser. 48, No. 4, Reaction Kinetics and Transfer Processes, 31-7 (1952). (8) Miller, D. N., Dissertation Abstr. 16, 926 (1958). (9) Pease, R. N., Yung, C. C., J . Am. Chem. SOC.46, 390-403 (1924). (10) Schwab, G. M., Schwab-Agallidis, E., Zbid., 71, 1806-16
I
11) Senderens, J. B., Ann. chim.phys. 25, 449-529 (1912). 12) (1949). Smith, J. M., "Chemical Engineering Kinetics," McGrawHill. New York. 1956. (13) Topchieva, K. V., Yun-Pin, K., Vestnik Moskou. Univ., 7, No. 12, Ser. Fit.-Mat. i Estestven. Nauauk No. 8, 39-48 (1952); C. A. 47, 9125 (1953). RECEIVED for review June 19, 1961 ACCEPTEDFebruary 7, 1962
T H E REACTION OF AMMONIA WITH CARBON A T ELEVATED TEMPERATURES T H O M A S K . S H E R W O O D A N D
ROBERT 0 . M A A K l
.tfasrachusetls I n d l u l e of Technology, Cambridge, Mass.
A simple heated-filament reactor was employed in studies of the reaction of ammonia with carbon to form HCN, and in the pyrolysis of both liquid and gaseous hydrocarbons. Reaction occurred in a thin layer of fluid in contact with the heated surface; rapid quenching of the reaction products resulted from the flow of cold flluid reactant past the wire. Eleven volume per cent acetylene was obtained b y pyrolyzing methane. Ammonia reacted with both graphite and pyrolitic carbon to form HCN, the reaction being rate is controlled b y the surfirst order in clmmonia. At low temperatures-below about 1600" K.-the face reaction; at higher temperatures-to 2300" K.-mass transfer to the surface controls the rate. Pyrolysis of liquid heptane gave large yields of ethylene, conversion passing through a maximum of 2.4 moles of ethylene per mole of heptane at a wire temperature of 1650" K.
hydrogen cyanide, and other endothermic compounds of cominercial interest can be formed at high temperatures, but recovery requires that the products be quenched rapidly to prevent decomposition. Both the high temperature and the rapid quench can be attained in electric arcs: shock tubes, or by devices equipped to provide adiabatic compression and expansion of gases. S o n e of these are simple to construct and use in experimental investigations of an exploratory character. The present study describes results obtained Lvith a simple reactor employing a heated filament, so arranged as to obtain the necessary rapid cooling of the reaction products. hfost of the data reported are for the reaction of ammonia with carbon; a few results for the pyrolysis of methane and of heptane are included.
A
CETYLEXE,
Present address, Esso Research Laboratories, Baton Rouge, La.
Experimental
The reactor, shown in Figure 1, was designed for batch operation. It consisted of a vertical borosilicate glass cylinder having a volume of 680 cu. cm., fitted with a ground glass flange. Gas circulation within the reactor was provided by rotating an annular borosilicate glass cup (rotor) consisting of cylinders 2.5 and 10 cm. in diameter. This was driven at 0 and 1025 r.p.m. by a vertical shaft passing out through a ground-glass joint holding a Teflon bearing made gas-tight by a rubber 0ring. A second Teflon bearing supported the rotor at the bottom. The fixed heated filament, 45 mm. long at a radius of 25 mm., was supported vertically and under slight tension by 4.8-mm. steel rods passing out through seals in the cover. These rods were wound with fine copper wire to reduce their electrical resistance and so hold them below reaction temperatures. Currents up to 60 amp. were used, supplied from a 110-volt source and controlled by a step-down transformer and Variac. ~
VOL.
1 NO. 2 M A Y 1 9 6 2
111
