1182
S. W. BENSON AND G. N. SPOKES
Very Low Pressure Pyrolysis. 111. t-Butyl Hydroperoxide in
Fused Silica and Stainless Steel Reactors' by S. W. Benaon and G. N. Spokes Department of Thermochemistry and Chemical Kinetics, Stanford Research Institute, Menlo Park, California 94086 (Received August 88, 1967)
The technique of very low pressure pyrolysis (VLPP) was applied to study the pyrolysis of t-BuOnH in stainless steel and fused silica reactors over the temperature range 300-950". Evidence was found for both unimolecular and chain-type homogeneous reactions; the chain reaction involves free .OH radicals. Homogeneous reactions proceed very much faster than heterogeneous reactions in a fused-silica reactor. The products of the unimolecular reaction are acetone and water, together with some other low-mass ( 5 3 2 amu) substances (candidates are CHaOH,CHgO, CO, and Hz). The chain reaction yields, in addition, isobutylene oxide with HzO. The acetone-productiondata lead to a rate constant for t-Bu02H+ t-BuO OH, in excellent agreement with lower temperature data and thermochemical estimates of a 43 kcal 0-0 bond. The total range covered is from 500 to 1000°K and 11 powers of 10 in the rate constant. Pyrolysis in a stainless steel reactor produces predominantly t-BuOH 02. Smaller amounts of isobutylene and some acetone are also formed. This represents a heterogeneous reaction which is much faster than the homogeneous decomposition in silica. While the wall mechanism is uncertain, the high rate points to the existence of a loosely bound, physisorbed precursor which can undergo bond cleavage with a high A factor and a relatively low activation energy (Le,, ~ 1 kcal). 0
+
+
Introduction Experimental studies of the homogeneous decompositions of hydroperoxides have proven difficult to interpret conclusively. Chain reactions seriously interfere with the kinetics and lead to unsatisfactory Arrhenius plots. Begson2 applied kinetic and thermochemical reasoning to the extant hydroperoxide decomposition data and concluded that the 0-0 bond dissociation energies of hydroperoxides should be in the neighborhood of 43 lrcal/mole. This value is in disagreement with the values of Bell, et al.,3 who concluded from measurements of the heat of combustion and the pyrolysis of t-butyl hydroperoxide that the bond strength should be about 39 kcal/mole. Further work has subsequently been carried out on the thermal decomposition of t-BuOOH by Hiatt and Irwin,4 who studied the thermolysis in dilute solution. They found that the rate constants obtained were as much as a factor of 8 lower than those of Bell, et al., with activation energies approaching 43 kcal. We have extended measurements of the decomposition of t-butyl hydroperoxide to much higher temperatures, using a very low-pressure pyrolysis (VLPP) technique. From the observed rate constants and application of an appropriate correction to them, we have obtained data which support the high value of 0-0bond strength in hydroperoxides. Wall effects have often been invoked to explain unusual products, and for this reason, we have explored the effects brought about by changing the character of the walls of the reactor. &Butyl hydroperoxide was The Journal of Physical Chemistry
thus decomposed in a stainless steel (VLPP) reactor. This resulted in a totally different product spectrum.
Experimental Procedure Experimental procedures followed those outlined in ref 5. High purity (better than 99% by iodometry) gaseous t-BuOzH was passed at the rate of about 5 X 10-9 mole/sec through a (0.5-mm id., 1 cm long) capillary to the reactor whose temperature was maintained a t various values, up to lOOO", by means of three electrical heaters. Gas pressures in the reactors were very low so that gas-wall collisions predominated over gasgas reactions. Gaseous productsa were pumped from the reactor into the mass spectrometer for analysis. The gas density in the 91-collision fused silica and 108-collision metal reactors was approximately 10l2 moleccles/cm3, although this changed slightly as decomposition proceeded. Gas density in the 9200collision reactor was about 100 times this value. (1) (a) Supported in part by funds from a multisponsored project on hydrocarbon oxidation; (b) Paper 11: G. N. Spokes and S. W. Benson, J . Amer. Chem. SOC.,89, 6030 (1967). (2) S. W. Benson, J . Chem. Phys., 40, 1007 (1964). (3) E. R. Bell, J. H. Raley, F. F. Rust, F. H. Seubold, and W. E. Vaughan, Discussion8 Faraday SOC.,10, 242 (1951). (4) R. Hiatt and K. C . Irwin, submitted for publication in J. Org.
Chem. (6) 5. W. Benson and G. N. Spokes, J . Amer. Chem. SOC.,89, 2525 (1967). (6) No visible change occurred t o the reactor's wall condition; that
is, there were no carbonaceous deposits and hence no products other than gas.
1183
VERYLOW PRESSURE PYROLYSIS O F t-BUTYL HYDROPEROXIDE Table I: Decomposition of tBuOzH in a 9200-Collision Silica Reactor Reactor temp, O Effluenta t-BuOiH Acetone Isobutylene oxided
Coldb (-30 2= 10')
300 f 50
81 13 6
96 f 4
4f4 0.25
C
7
760 f 60
800 f
60
600 rt 60
0 96
0 100
0 100
4
0
0 100 0
400 2= 6OC
500 =t
26 f 6 55 f 4 19 f 2
60
0
'
The apparent presence of acetone a t these temperatures is due to errors in the analysis and Mole per cent of CSand Cp effluent. Per cent of Data a t 400' are averaged from four experiments. is representative of the magnitude of the errors to be expected. isobutylene oxide may be systematically higher than quoted amounts, as discussed in text.
The cylindrical stainless steel7reactor was 10 cm long and had a 2.5-cm i.d.; the exit aperture was 1 cm in diameter. It was essential to keep the inlet capillary of the metal reactor below 100" to avoid catalyzed decomposition of the gases there. This cooling was accomplished by an air blast on the inlet capillary. Since the inlet end of the reactor was kept down to loo", even though the rest of the reactor was as hot as 850°, there was a temperature gradient of about 500 deg/cm near the inlet end of the reactor.
Results The results for each of these three reactors have been reduced in a slightly different way so that data and results for each reactor are treated separately. It has been necessary to cool both the top and the bottom of each reactor with an air blast and this has led to a rather large uncertainty in temperature, owing to the effectiveness of fairly weak drafts in affecting temperatures. About 9 of the 12 chromel-alumel couples for temperature measurement indicated a value within 20" of the central temperature, however. Decomposition of t-BuOzH in a 9200-Collision FusedSilica Reactor. Gas flowing a t about 3 X 1015 molemole/sec) was pyrolyzed in our cules/sec (5 X 9200-collision fused-silica reactor. Analysis of the results obtained is given in Table I. The stirred flow reactor equations are used in deriving the apparent first-order reaction rate. The details of treatment of data have been partly described elsewherees We may note that the average residence time in the reactor is just the reciprocal of k, and is here typically 0.4 sec. Mass spectral data from this reactor have been analyzed by means of a computer program on the basis of five sets of peaks and intensities derived from mass spectra of (1) residual gases (e.g. pump oil), (2) water vapor, (3) cold parent compound, (4) acetone and associated products presumed to be the only species a t 750-80O0, and (5) an "unknown" substance produced a t 400" whose set was derived by subtraction from the observed spectrum of appropriate amounts of sets 1-4. This unknown set resembled strongly that of isobutylene oxide. The only deficiency in this set was that the 43-amu peak was too Rmall by comparison with
our standard isobutylene oxide set (not surprising, since only a small error in acetone estimation will lead to this effect). An appropriate correction was made to this set at 43 amu and the resultant set 5 was used in the analysis. We have probably underestimated the per cent isobutylene oxide, since (in the absence of a direct calibration) we have taken the sensitivity of isobutylene oxide at 43 amu to be equal to that of acetone a t 43 amu. This would probably lead to underestimation of isobutylene oxide by no more than a factor of 2. The values in Table I are, therefore, a lower limit. If we regard the acetone produced in the 9200-collision reactor as coming from t-BuO, produced by unimolecular split of t-BuOzH, we can use the ratio of acetone to t-BuOzH as an indicator of first-order reaction rate and can construct Table 11.
Table I1 : First-Order Decomposition Rate Constant for t-BuOzH a t Very Low Pressure and a t Various Temperatures (T)in a 9200-Collision Fused-Silica Reactor
a
Temp,
+(MezCO)/
OK
+(l-BuOzH)
aec-1
570 f 50 670 f 50 770 f 50
0.16 2.2 >10
2.1 2.25 2.4
k e ( = 0.0874/T),a
h i ,
sec-1
0.34 4.95 >24
k , is the reactor escape rate constant for t-BuOzH (sec-1).
' kuni is the sought rate constant and is equal to ke[+(MezCO)/ +(t-BuOzH)] where +(x) is the rate of flow of effluent x.
Decomposition in a 91 -Collision Fused-Silica Reactor. Gas flows were about 5 X 1015molecules/sec. Decomposition was quite clean to give acetone and associated products.s The per cent decomposition was followed mass spectrometrically by monitoring two mass peak (7) The stainless steel type 304 used contains 19-20% chromium, 8-12% nickel, less than 2% Mn, and less than 0.08% P, 0.03% S, and 1.0% Si. (8) Acetone is formed by decomposition of the t-butoxy radical. The .CHS (from t-BuO.) and .OH from the initial step will react to yield low-mass substances which do not interfere with measurement of acetone flux.
Volume 78, Number 4 April 1968
1184
S. W. BENSONAND G. N. SPOKES
pairs a t 41 and 43 and a t 57 and 58 amu. From the known sensitivities of the spectrometer to the products and parent substances, we have determined the firstorder reaction rate constant of t-BuOzH. The data are presented in Table 111.
1-&OH
is0
C4H8
MepCO
t-BuO2H
Table I11 : First-Order Decomposition Rate Constant" for t-BuOzH a t Various Temperatures ( T ) in a 91-Collision Fused-Silica Reactor Temp, OK
(43-41) kuni (57-58) kuni (av)
kuni
300
0 0 0
673
2.7 1.44 2.07
773
873
933
958
973
72 96 84
550 582 566
1890 1750 1820
1840 2120 1930
2180 2980 2580
' kuni (43-41) and kuni (57-58) are rate constants calculated from the intensities of the 41-43 and the 57-58 amu peaks, respectively. kuni (av) is the simple average of these two. For this reactor geometry the escape rate constant k , for t-BuOzH was 8.831/? sec-1. The acetone to t-BuOzH ratio can be derived from k u n i by dividing by 8.83@. Decomposition in a 108-Collision Stainless Steel Reactor. The results obtained with the metal reactor are summarized graphically in Figure 1. The experiments were carried out with an air-cooled inlet capillary under conditions described above. The analysis of the mass spectral data from the metal reactor was subject to some error because the stability of the mass spectrometer was insufficiently great to permit a simple analytical procedure to be adopted. Results were analyzed by use of a regression program (which minimizes the squares of the errors) using standard mass spectral sets generated by running cold t-BuOzH, R!Ie2C0, HzO, t-BuOH, and isobutylene gas through the mass spectometer. The results reflect the general pattern of the decomposition, although it is most unlikely that any significant amount of parent was left unchanged at temperatures in excess of 700". We believe that apart from the parent concentrations a t higher temperatures the stated fractions of products are probably correct within a factor of about 2. I n addition to the acetone, t-butyl alcohol, and isobutylene, there were water (detected), oxygen (not detected, owing to air backgroundg), and low mass number species such as CzHG (not analyzed for). (Methyl ethyl ketone and isobutylene oxide both have an appreciable 72-amu peak; the complete absence of this peak shows that they were produced in proportions less than 1%.)
Discussion Pyrolysis in the 91- and 9200-Collision Quartz Reactors. Products of pyrolysis in the 91-collision reactor appear to be only acetone and associated lower mass number substances. The apparent first-order reaction rate constant as determined by measurements of the The Journal of Physical Chemistry
300
400
500 600 TEMPERATURE-%
700
80
Figure 1. VLPP of t-butyl hydroperoxide in a 108-collision stainless steel reactor.
ratios of the mass peak pairs 43-41 and 57-58 agree satisfactorily, probably within experimental error. These data, together with those obtained from the 9200-collision reactor, are presented in Table IV. Also included in Table IV are high-pressure activation energies derived from these data assuming log A = 15.75 and = 15.0. The derived high-pressure activation energies take into account the fact that our experiments were conducted in the region of pressure falloff. The correction for falloff involves determining Eofrom our observed rate constants and is done by using eq 1 and 2 which have been derived from the Rice, Ramsperger, and Kassel formulation of unimolecular reactions, and whose applicability has been demonstrated elsewhere.4*'0 kuni =
___ kw
2d%
[
(S
eEm ]'-'exp(-E,/RT) - l)RT
(1)
and
where kuni is the observed apparent first-order rate constant; e = 2.7183; E, is the mean energy required for molecular decomposition in a time N k W - l , where k , is the frequency of wall collisions; and s is the Kassel s, the number of active oscillators in a molecule. Finally, the high-pressure Arrhenius formulation IC, = A exp( -Eo/RT) serves to define A and Eo. It turns out that while eq 1 has no analytic solution, it is a simple matter to solve for Em (given the other unknowns) by use of a slide rule with a log scale. E , (9) The air leak did not affect the chemistry of the processes being described, since it was from the atmosphere direct into the mass spectrometer and not through the reactor. (IO) S. W. Benson and G. N. Spokes, Mth International Symposium on Combustion, The Combustion Institute, Pittsburgh, Pa., 1967, p
. 96
1185
VERYLow PRESSURE PYROLYSIS OF BUTYL HYDROPEROXIDE Table IV : Derivation of Effective High-pressure Activation Energies from Experimental VLPP Data on t-Bu02H4 Temp, OK kuni, sec-l
S
E,, kcal/mole Eo,kcal/mole LogA = 15.75 LogA = 15.0
570
670
673
773
873
933
968
973
0.34 22 54
4.95 23 60.5
2.07 24 62.5
84 26 64.5
566 29 69.6
1820 31.5 75.4
1930 32.6 78.8
2580 33 76.2
38.0 36.5
41.3 39.7
42.7 41.0
42.0 40.3
42.3 40.6
43.4 41.3
44.3 42.5
42.7 41.2
kuni is the experimentally a Runs a t 570 and 670'K were made in the 9200-collision reactor, the others in the 91-collision reactor. observed first-order decomposition rate constant. s is the effective number of molecular oscillators (s is equal to '/z(C, 8); C, was determined from group values which we have generated). E , is the critical energy in the t-Bu0~Hmolecule which leads to decomposition in a time roughly equal to the time between wall collisions. EOis the activation energy deduced from our experimental data using eq 1 and 2, the only assumptions being (i) the molecules were fully equilibrated to the stated temperature and (ii) the Arrhenius A factor was either 1016.76 or 1016,0.
-
thermochemistry as deduced by use of group values for heats and entropies of the relevant species involved.12 It is, of course, presumed that the reaction proceeds by the simple mechanism
6
+ *OH (CHJ3CO. +(CH3)zCO + *CH3
5
(CH3)3COOH --f (CHa)aCO*
4
3 -I
-8 2
(CH&CO +escapes t o mass spectrometer
I
1
0 -I
-2
1
THESE DATA\-
'OH and .CHI +escape to the mass spectrometer and lead to low mass number products, e.g., C2Ha,H20, CH30H
-3 -4 -5
1.0
1.2
1.4
1.6
1.8
2.0
2.2
Figure 2. Arrhegus plot of VLPP data for t-butyl hydroperoxide: m-1,experimental data points with 91-collision reactor; +A+, experimental data points with 9200-collision reactor; 0, data of 91-collision reactor corrected for falloff assuming log A = 15.75; 0, data of 91-collision reactor corrected for falloff assuming , data of 9200-collision reactor corrected log A = 15.0; . for falloff assuming log A = 15.75; A, data of 9200-collision reactor corrected for falloff assuming log A = 15.0. The data of Kirk and Knox and of Hiatt and Irwin are included for comparison.
is then deduced from relation 2. Figure 2 presents experimental rate constants along with those obtained by using the Arrhenius formulation with the derived values for Eo. Data of Kirk and Knoxll and Hiatt and Irwin4 are also given in Figure 2 for comparison. We see that our corrected data fall on a line corresponding to a high-pressure rate constant equal to 1015.6*0.6 10-(42.2*2)'RT sec-'. These Arrhenius parameters are consistent with data obtained by the different techniques used in work reported in ref 4 and 11. I n addition these parameters are consistent with the
Chain reactions are ruled out in the 91-collision reactor because molecules make on the average only 1 gas-gas collision per 100 wall collisions. Pyrolysis studies in the 9200-collision quartz reactor were complicated by the appearance of isobutylene oxide as well as acetone. These known products of the chain decomposition can be explained by secondary reactions of HO. and Me. .OH (or .CHJ
+ (CH&CO2H .CHzC(CH3)202H
0
O2H
I
*CHz-C(CHs)2
+ HzO (or CHI)
/\
--t
H2C--C(CH3)2
--t
H2C=C(CHa)z
+ *OH + HO23
The isobutylene oxide is identified primarily by its contribution to mass peak 72 amu. (All mass peaks (11) A. D. Kirk and J. H. Knox, Trans. Faraday Soc., 56, 1296 (1960). (12) Values used differ slightly from those of Benson (J.Chem. Phys., 40, 1007 (1904)), inasmuch as we have used +9.3 kcal/mole for A H P 8 of *OHand -22 kcal/mole for AHrzesoft-BuO. The data of Kirk and Knox may have been affected by reason of the rather short residence time at their highest temperatures. This would lead to an anomalously low reaction rate at these temperatures, owing to the slowness of heat transfer between walls and gas. Volume 72, Number 4
April 1968
1186 corresponding to isobutylene oxide were included in the computer analysis of the data.) For the chain decomposition reaction to proceed, .OH free radicals must survive several collisions with the quartz reactor walls. The depletion of parent compound by the competing unimolecular reaction path leads to a continual drop in the rate of chain reaction, in accord with our observations that the amount of the secondary reaction falls to a low value a t higher temperatures. Pyrolysis in the 108-Collision Stainless Steel Reactor. The change in product spectrum and the increased rate of reaction in the stainless steel reactor is quite dramatic. The mechanism of the decomposition in the 1OS-collision metal reactor is not certain. There are (as in the case of the 91-collision silica reactor) very few gas-gas collisions and, therefore, we can ignore contributions by gas-phase chain reactions. An adequate mechanism must account for (1) the completely different product spectrum, (2) the production of isobutylene, (3) the production of acetone, and (4) the complex variation of product distribution with temperature. Furthermore, the high probability of reaction per gaswall collision imposes special constraints on the reaction mechanism. Thus at 450" the decomposition in the stainless steel reactor (Figure 1)leads to about 50% decomposition after 10s collisionsfor a collision efficiency of approximately 1in 46. If we interpret such a collision efficiency from the point of view of a bimolecular reaction with a specific wall site and allow a mean roughness factor of 2, then it implies either a steric factor of 1 and 6 kcal of activation energy or else zero activation Either of these energy and a steric factor of
T h e Journal of Physical Chemistry
S. W. BENSONAND G. N. SPOKES steric factors is unreasonably large for so large a molecule as t-Bu02H. The alternative to a direct bimolecular event is a precursor, physisorbed state, which can competitively dissociate or reevaporate 1
t-BuOzH(g)
02-(S)
JT t-BuOzH(S) -1
+ t-Bu02H(S)--%t-BuOz-(s) + HO-(S)
(or t-BuO-
+ H02-)
If reaction 2 is irreversible, it can account for the high collision efficiency if its rate is 10-2k-1. Since I C - ~ and k2 are essentially unimolecular acts involving simple bond breaking, they can readily have similar A factors. Hence the activation energy of step 2 need only be 6 kcal larger than the heat of reevaporation to account for the data. A reasonable value of AH1 is probably around 15 kcal, so that a value of E, = 21 kcal would be implied and is not unreasonable. By way of contrast, we note that the low-temperature metal oxide catalyzed dehydration of t-BuOH appears not to occur to any significant extent over the temperature range of the reaction. This was verified by direct introduction of t-BuOH. This is quite reasonable in terms of our consecutive unimolecular acts since the reaction t-BuOH(g) i-butene(g) HzO(g) is endothermic by 14 kcal, which would represent a minimum activation energy for the system. By contrast, the reaction t-BuOzH(g) 4 t-BuOH(g) 1/202(g)is 19 kcal exothermic. Such analyses can be extended to explain the relative rarity of wall-catalyzed reactions in our VLPP systems at high temperatures.
+ +
