T H E D E C O N I P O ~ I T I OOF ~ DIMETHYL ETHER ON T H E SURFACE OF PLATINUM BY E. W. R. STEACIE AND H. A . REEVE
Introduction I t is of considerable interest to compare the velocities of homogeneous and heterogeneous gas reactions. In this connection it has been shown by Hinshelwood that there is a general tendency for homogeneous bimolecular reactions to become unimolecular on the surface of a catalyst, the heat of activation of the reaction falling to about one-half of its former value. This change in the characteristics of the reaction is accompanied by a modification of the reaction path. The homogeneous decomposition of hydrogen iodide, for example, proceeds as indicated by the equation 2
HI
=
Hz $- 12.
The possible homogeneous unimolecular decomposition
HI
=
H
+ I,
followed by a subsequent recombination of the atoms, is ruled out since the formation of atomic hydrogen would be a highly endothermic process. The momentary concentration of a large amount of energy in the molecule would thus be necessary, and hence a very large heat of activation would be associated with the reaction. The presence of a metal surface, however, renders such a change possible since the atomic hydrogen and iodine formed can be held on the surface in an adsorbed condition, and later the atoms evaporate in pairs as molecular hydrogen and iodine. The heat of activation of the homogeneous bimolecular decomposition is 44,000 calories per gram molecule.1 In order that two molecules may decompose, they must therefore have a combined energy in excess of 44,000calories. For the heterogeneous unimolecular decomposition on the surface of gold the heat of activation is 25,000 calories.2 Hence one molecule in order to decompose must have an energy in excess of 2 5 , 0 0 0 calories. The lowering of the activation energy by the catalyst is therefore accompanied by a change in the mechanism of the reaction, in the one case only a single molecule being activated, while in the other case two molecules must be activated simultaneously. I n consequence the effect of the catalyst on the magnitude of the activation energy is left unanswered. 1 2
Bodenstein: Z. physik. Chem., 29, 295 (1899). Hinshelwood: J. Chem. SOC.,127, 1552 (1925).
DECOMPOSITION OF DIMETHYL ETHER ON PLATINUM
307.5
It is therefore of interest to compare the homogeneous and heterogeneous reactions in the case of a substance which decomposes homogeneously in a unimolecular manner. Previous investigations of this kind have been made with acetone; ethyl ether,' and propionaldehyde.6 The present paper deals with a similar investigation using methyl ether on the surface of platinum. Methods of investigating Heterogeneous Gas Reactions. There are two main methods of investigating heterogeneous gas reactions. (a) The catalyst is contained in a bulb in finely divided form, The reactant, or reactants, is admitted to the bulb, or passed through it continuously, and the reaction is followed by the pressure change which accompanies it or by analysis. This method is the simpler of the two, and is the more usual when the course of the reaction, yield, etc!, is of primary importance. On account of the lack of definite knowledge of the surface, etc. it is not a suitable method of investigating the molecular statistics of the reaction. (b) The catalyst consists of a fine filament stretched axially through the reaction vessel. The vessel containing the gas is kept a t or near room temperature, while the filament is heated electrically to the desired temperature. This method is by far the most suitable FIQ.I when an accurate knowledge of the Electrical Diagram molecular statistics of the reaction is desired, and is the one used in this case. I n this method the solid material can function in two ways, either by adsorbing the reactant and permitting a catalytic reaction, or else by merely acting as a source of energy. I n the latter case the question of energy transfer between gas molecules and the solid sudace will be of paramount importance. Apparatus. The apparatus was similar, with the exception of the electrical set-up, to that used in previous investigations. It consisted of a reaction bulb of about IOO cc capacity, through which a platinum wire (0.05 to 0.10mm diameter) was sealed axially. The bulb was connected by capillary tubing and stopcocks to a capillary manometer, a supply of methyl ether, and a pumping system. The connecting tubing was wound with nichrome wire and heated electrically to prevent condensation. The reaction bulb was immersed in an oil bath, which was maintained at 45'C. Temperature Measurement and Control. The temperature of the heated filament was obtained from its resistance in the following manner. The filament A (Fig. I ) was made one arm of a Wheatstone Bridge A B C D . Since Taylor: J. Phys. Chem., 33, 1793 (1929). Steacie and Campbell: Proc. Roy. Soc., 128 A, 451 (1930); Taylor and Schwartz: J. Phys. Chem., 35, 1044(1931). Steacie and Morton-Can. J. Research, 4, 582 (1931).
30i6
E. W. R . STEACIE AND E. A , REEVE
the resistance of the various filaments used was low (from I to 1 5 ohms), the standard resistance B was of the same magnitude and was immersed in a large, well-ct'irred oil-bath. The heating of this resistance was negligible with the current used. Two high resistances, C and D,were used in the other arm of t,he bridge. C was a standard 10,000 ohm resist,ance, and D was a variable 1/10 to IOO,OOO ohm, resistance. A sensit'ive galvanometer, G, was used. The sensitivity could be varied by the resistance F in seiies, or by the variable shunt E. The voltage applied to the bridge was regulated by the potentiometer H , which was connected directly to the I I O volt D.C. supply. Methyl ether6 was prepared from sulphuric acid and methyl alcohol. The gas was bubbled through sulphuric acid saturated with methyl ether, passed through phosphorus pentoxide tubes, and fractionally distilled. During the course of the experiments it was stored as a liquid in a bulb immersed in a solid carbon dioxide-acetone mixture. Experimental Procedure The resistance of the filament was measured at' various temperatures, as determined by a Leeds and Northrup optical pyrometer, over a range from 700' to IIOO'C,at 50' intervals. A resistance-temperature curve was constructed from these results. On extrapolation this curve gave good agreement, with the experimentally determined resistance at room temperaiure. Even if the absolute values of the temperature of the filament are slightly in error, this will be unimport,ant for the present purpose provided that the relative temperatures are in good agreement. During any one run the bridge set'ting was such that a balance gavc a filament resistance corresponding to the temperature desired. On account of the changing thermal conductivity of the gas mixtures, as the reaction progressed, it was necessary to vary the impressed volt~geso as to maintain a balance in the bridge circuit. The reaction was followed by admitting methyl ether to the reaction vessel, maintaining the filament a t the desired temperature, and observing t,he variation of pressure with time. The Course of the Reaction. Hjnshelwood and Askey7 found that the homogeneous decomposition of methyl ether was mainly as represented by the equation CH30CISx = [CB4 -4-IICHO] =: CHJ 139 co.
+
+
Thus in a typical analysis of the products they found 3 2 . 0 % carbon monoxide, 33. j% hydrogen, and 34.5% methane. 5 We are indebted to Mr. J. S. Tapp of this laboratory for supplying the methyl ether used. 7 Proc. Roy. Soc., 115 A, 2x5 (1927).
DECOMPOSITION OF DIMETHYL ETHER ON PLATINUM
3077
In the present investigation analysis showed that the reaction was, in the main, the same as the above. Thus a typical analysis a t 9 7 7 O C gave the following result: CO = 34.6%, CHa = 33.0%, Hz = 32.5%. In the homogeneous reaction, in agreement with the foregoing equation, pressure increases a t completion of about 200% were obtained a t all temperatures. In this investigation, however, pressure increases at completion of 186 2% were invariably obtained. This somewhat lower value was undoubtedly due to the condensation of a small amount of paraformaldehyde on the walls of the reaction vessel. Since the final pressure increase was the same a t all temperatures, however, it is justifiable to use the pressure increase as a criterion of the extent to which the reaction has progressed. The times for various fractional pressure increases have therefore been used as a measure of the reaction velocity.
*
Experimental Results During the course of the investigation a number of different filaments were used. In general these filaments showed considerable differences on account of differing diameter, surface conditions, etc. In the case of a particular filament, after a certain amount of preliminary aging, a steady condition was finally reached and reproducible reaction velocity results could then be obtained. This steady condition might be upset, however, by too drastic heating. The results given below were obtained on various filaments which had reached a steady condition. In any particular series the runs were made in random order to obviate any error due to aging. T h e Effect of Pressure. Fig. 2 shows typical pressure-time curves for various initial pressures at II~o'K, on filament No.I. The complete data for a typical run are given in Table I.
TABLE I Filament No. Time mins. 0
I. I I goOK
Per cent decomposed
K
I
3 8
6 .o
0,0194
17.4
0.0239
I2
26.0
o.ozj0
I7
35.8 43.9 53.4 60 .o 65.1 72.3 77.5
0.0261
22
29
35 40 50
60
o ,0262
o ,0262 0.0262 0.0264 0.0258
0.0250
E.
3078
W.R. STEACIE AND H. A. REEVE
The values of per cent decomposition are calculated on the assumption that an increase in pressure of 186% corresponds to complete decomposition. The constants given in the last column are those calculated for a unimolecular reaction. As in the homogeneous reaction the constants rise in the early stages of the reaction while formaldehyde is accumulating, and finally become constant within the experimental error. The effect of pressure on the rate of reaction is indicated by Fig. 3 and some typical data are given in Table 11. As in the homogeneous reaction the velocity constants fall off at pressures below 300-400 mm.
FIQ.2 Pressure-Time Curves
TABLE I1 Filament No. I Temperature = I I 5o'K Preeaure mm
mins.
mins.
97 98 144 206 314 3'5
23 . o 25.5 19.3 14.3 10.6 12.6
53.6 64.4 48.4 35.3 25.4 30.0
T 6 0
TI00
Pressure mm
3 16 318 430 432 433
T 6 0
TI00
mins.
mins.
.o .o
29 .o 28.3 23 . o 25.4 25.4
I2 I2
10.0
10.7 10.7
DECOMPOSITION OF DIMETHYL ETHER ON PLATINUM
3079
The Temperature Coeficient. The heat of activation was calculated from two entirely separate series of results on different filaments. In the first series, for experimental reasons, it was impossible to use initial pressures above 400 mm. The heat of activation was therefore calculated by extrapolating the results at each temperature to the high pressure rate. In the second series of results runs were made a t initial pressures of about 700 mm. At these pressures the falling off was negligible and no extrapolation was necessary. Some typical data for the second series are given in Table 111. Fig. 4 shows a plot of log T 2 6 ,log T60 and log TIOO against the reciprocal of the absolute tm temperature. w F up
s
KO
30
f e eo
E Fm qr 300
eo0
100
0
Rossurc
m~
qoo
mm.
FIQ.3 The Effect of Pressure on the Rate of Reaction
I
106
FIG.4 The Temperature Coefficient Curve A-Calculated from TIOO Curve B-Calculated from TaO Curve C-Calculated from TU
TABLE I11 Filament No. 2 Initial pressures ca. 700 mm. Temperature
"K
I IO0
TI5
mins.
37.5 3 7 .o
I150
8.75 9.25
I200
2.53
1250
T5"
mins.
74 74 17.8 18.5
2.58
5.20 5.33
0.92
I .60
0.97 0.73
I .87
mms. 216
236 43 47 12.2
12.5 3.84 4.33
The heats of activation calculated for various fractional times for both series of results are given in Table IV.
3080
E. W. R . STEACIE AND H. A. REEVE
Series
I 2
TABLE IV Heat of Activation from Tzs from TbO calories per gram mol. 68500 68300 69500
-. -
from TIOO 70100 72200
On account of the complication due to the intermediate formation of formaldehyde, the best values of the heat of activation will be obtained by the extrapolation of the above values to initial rates. This is done in Fig. 5 . The values obtained in this way for the two series are 66900 and 67100 calories. The mean Tu 5 0 T~ value of the heat of activation is therefore 67000, as compared with 58500 found FIG j Extrapolation of the Heut of Actira- by Hinshelwood and Askey for the tion to Initial Rates homogeneous reaction. Discussion A compariEon of the heats of activation of homogeneous unimolecular decompositions with those of the same reactions in contact with hot filaments shows that the results fall into two classes. The data are summarized in Table V. T ~ B LvE of Activation _____Heat _________--__ Substance
Acetone Ethyl Ether (Steacie & Campbell) (Taylor & Schwartz) Propionaldehyde Methyl Ether
Homogeneous
68,500 53,000
Filament
68,400 52,000
57,000
96,500 67,000
I t will be seen that in the first two cases the heats of activation agree within the experimental error, while in the last two the heat of activation of the filament reaction is h i g h e y than that of the homogeneous reaction. This point will be discussed later in detail. There are two possible explanations of the action of the filament in these reactions: (a) Molecules are activated by collisions with the filament, statistical redistribution of energy being produced by such collisions. (b) The filament serves as a source of energy, the surrounding gas layer being maintained at a high temperature. Activation takes place in the hot gas layer by ordinary molecular collisions.
DECOMPOSITION O F DIMETHTL ETHER ON PLATINTJM
3081
The second explanation seems by far the more likely, since it is difficult to see why the velocity constants should fall off a t low pressures if collision with the filament were a necessary preliminary to reaction. That this is the correct explanation may be proved conclusively by an examination of the statistics of the reaction. Some typical data for the decomposition of methyl ether in contact with platinum follow: Filament temperature, I ISO'K; pressure, 3 16 mm; volume of reaction vessel, IOO cc; filament length, I O om; filament diameter, 0.06 mm; bath temperature, 318'K; rate of decomposition, o.033y0 per sec. Whence we obtain: ( a ) T h e number of molecules reacting which is 3.19 X 10~' molecules per sec. and (b) T h e number of molecules hitting the filament. According to Knudsen's equation we have
where m is the mass of gas striking the filament per sq. cm. per second, M is the molecular weight, T is the absolute temperature, and p is the pressure in bars. Evaluating the constants, and putting the pressure in atmospheres, we have m = 44.24-p. The value of T to be used is somewhat uncertain. Since we are only interested in the order of magnitude of the results, it is not of much importance, and the simplest assumption to make is that the temperature of the colliding molecules is that of the filament. I n any case an error of even 200' in T will not introduce more than a 10% error into the result. Whence we have m = 3.68 g per sec. per sq. cm., hence the number of molecules striking the filament is 4.85 X 1oZ2molecules per sq. cm. per sec. The surface area of the filament is 0.188 sq. cm., hence we have number of molecules striking the filament = 9.12 X 1oZ1per second. (c) T h e fraction of the molecules at I I gooK possessing the enerpy of actimtion. Using Hinshelwood's form of theory, this will be given by
Assuming that the heat of activation found by Hinshelwood and Askey for the homogeneous reaction is the true one, we have E = 58500, and n = ca. 12. Whence the fraction of the molecules a t I I 5o0K which possess the energy of activation = 1.36 X IO-*. Hence the total number of activated molecules, produced by collisions with the filament only is 9.12 X 1oZ1X 1.36 X IO-* = 1.24 x lol* molecules per sec. Hence about 2 6 0 0 molecules react for every one which obtains the necessary energy from the filament by direct collision. It may therefore be concluded that the reaction is occurring in a layer of hot gas surrounding the filament. The effective dimensions of this layer may be calculated as follows:
3082
E. W. R. STEACIE AND H. A. REEVE
The total number of molecules entering into collisions in the gas per cc per sec. is given by 2 dTr s2 un2, where s is the molecular diameter, u the root mean square velocity, and n the number of molecules per cc. The total number of activated molecules produced in the gas per cc per sec. is therefore given by
4 l Ts2un2e
E
+ (:n
6-m--
- I ) R T E f (in RT
i$n-
I)RT
)
'"-I
I
In order to evaluate this expression it is necessary to assign a value to s, the molecular diameter. Kinetic theory considerations indicate a value of about 5 X IO+ cm. I t is known, however, that molecules with a high energy content can transfer energy by resonance, and in general in dealing with energy transfers effective diameters much larger than kinetic theory diameters must be used. We may therefore assume S = 2 0 X IO-^ cm. Whence the number of activated molecules produced by collisions in the pas at I I jd°K is 3.54 X IO?? molecules per cc per sec. The number of molecules reacting per second is 3.19 X 1 0 Hence ~ ~ ~ the effective volume of the space in which reaction occurs is 3.19 X 3.54 x
1017
=
9.00
x
10-Gcc
IOZ2
Whence the effective thickness of the region surrounding the filament in which reaction occurs is 4.73 x 10-4 mm. * I t is well known that practically the entire temperature drop in the gas surrounding a filament of this type takes place in a "skin" about 0.5 mm thick by conduction. The last few degrees drop is much more gradual, and the heat transfer takes place mainly by convection. The total temperature drop here is 832'. If we assume that 7 j o o of this occurs in such a layer, then the reaction is occurring in a zone surrounding the wire which includes a temperature gradient of about IO. The volume, however, is a minimum value, as explained before. We may therefore conclude that the zone in which reaction occurs does not include a great enough temperature gradient to introduce any appreciable error into the calculated value of the temperature coefficient of the reaction, provided that the $lament i s in thermal eguilzbrium wzth the gas. Calculations for ethyl ether, acetone, and propionaldehyde yield similar results. In every case it is necessary to assume that the reaction occurs in a hot gas layer surrounding the filament. We may therefore consider that the dynamics of the decomposition of ethyl ether and of acetone are completely explained. In the case of propional-
* Since this volume is obtained by equating the number of molecules activated to the number reacting, it will be a minimum value. I t will, however, be approximately correct since the calculations have been made for a pressure at which the velocity constants have fallen off t,o an appreciable extent.
DECOMPOSITION OF DIMETHYL ETHER ON PLATINUM
3083
dehyde and of methyl ether, however, it is still necessary to explain the fact that the temperature coefficients are higher than those of the homogeneous reactions. The heats of activation calculated for the filament reactions cannot be the true ones. Thus for propionaldehyde calculation shows that if the true value of E were 96,500 calories, then even if the whole reaction vessel were at the temperature of the filament the rate would still be 167 times slower than the observed value. The only possible explanation of the high temperature coefficients of these reactions would seem to involve the transfer of energy between the filament and incident gas molecules. Energy Transfer between Gas Molecules and the Filament. By means of heat conductivity measurements a t low pressures Soddy and Berry* came to the conclusion the accommodation coefficient for gases of high molecular weight was always in the neighbourhood of unity. Langmuir’s theory of adsorption indicates that all, or nearly all, the molecules hitting a solid condense and reevaporate. The accommodation coefficients are therefore virtually unity, and almost all the molecules leave in thermal equilibrium with the filament. This conclusion has been supported by a number of investigations. Recently, however, it has been shown that in certain cases molecular or atomic beams may be reflected specularly from solid surfaces without any transfer of energy.@ It therefore seems plausible to assume that energy transfer between complex gas molecules and solid$ may be highly specific, and that in certain cases the accommodation coefficients may be quite low. This is especially so in the case of molecules which decompose unimolecularly, since such molecules are notoriously specific in action insofar as the transfer of energy is concerned. The mechanism of the foregoing reactions may therefore be explained on the following basis. With acetone and ethyl ether the molecules are mostly adsorbed on collision. They evaporate after a very short mean life on the surface, and therefore leave in thermal equilibrium with the surface. The inner fiide of the gas layer next to the filament is therefore at a temperature which does not differ appreciably from the temperature of the filament. No appreciable error is therefore introduced into the calculation of the heat of activation of the reaction. With propionaldehyde and methyl ether the accommodation coefficients may be assumed to be low. (This assumption will be discussed later). Hence comparatively few of the molecules are adsorbed and reach thermal equilibrium with the filament. The majority of the molecules are reflected more or less specularly and very little energy is transferred to them. There is therefore a very abrupt drop in temperature a t the surface of the wire, followed by the usual more gradual skin effect. If this assumption is correct, the mean temperature of the layer in which reaction occurs will be considerably below the temperature of the filament itself. The high temperature coefficient of the reaction may therefore be explained in two ways: (a) We may interpret the high temperature coefficient as being due solely to the cause mentioned above. The temperature scale used in Proc. Roy. Sac., 84 A, 576 (1911). Bradley: Chem. Rev., 9, 47 (1931).
3084
E. W'. R. STEACIE AND H. A . REEVE
calculating the heat of activation of the reaction should therefore be shifted to somewhat lower temperatures. The heat of activation of the reaction is inversely proportional to (I/TI - I / T ~ ) .If we lower both temperatures by a constant amount, the calculated value of the heat' of activat,ion will therefore decrease. To explain the difference in the heat's of activation on this ground alone would require a very large temperature drop at the surface of the filament (for propionaldehyde 275°C). I t seems much more likely that the following explanation is the t'rue one. (b) I t has been observed that accommodation coefficients increase with increasing temperature, We may therefore explain the high temperature coefficient of the reaction in a much more reasonable way by assuming that the accommodation coefficient,s are low, but not excessively low, and vary with temperature. The temperature coefficient of the reaction is therefore a composite one, and includes the temperature coefficient of the accommodat'ion coefficient. It is thus unnecessary to assume such a pronounced drop in temperature at t'he surface of the filament. The specific action of the surface remains to be explained. If the foregoing assumptions are valid, the accommodat'ion coefficients for ethyl et'her and acetone are near unity while those for propionaldehyde and methyl ether are quite low. Accommodation coefficients usually rise in the presence of a layer of adsorbed gas. It, is therefore by no means improbable t'liat the specific action depends on the relative adsorption of the various products formed in the decompositions. I t has been noticed that a small amount of carbon is deposited in the decomposition of propionaldehyde and of methyl ether, but not in t,he other cases. This would seem to be the most likely explanation of the specificity. A number of the questions raised above might be settled by a determination of the heat loss from filaments in the presence of the various gases. Such experiments are in progress. SUWIary
The kinetics of the thermal decomposition of gaseous dimethyl ether in contact with heated platinum filaments have been investigated. The decomposition is unimolecular and occurs in a hot gas layer surrounding the filament. The heat of activation is found to be 67,000 calories as compared with 58,500 for the homogeneous react'ion. The high temperature coefficient may be explained on the assumption that the accommodation coefficient is low, and hence thermal equilibrium with the filament is not attained by colliding gas molecules. Physical Chemistry Laboratory, McGilZ University, Montreal, Canada.
