1432 J . Phys. Chem.

1990, 94, 1432-1439

Detailed Kinetic Modeling of Autocatalysis in Methane Pyrolysis

A. M. Dean
Corporate Research, Exxon Research and Engineering Company, Annandale, New Jersey 08801
(Received: May 10, 1989; In Final Form: August 3, 1989)

A detailed kinetic model is developed to explain the observed acceleration in rates of product formation during the pyrolysis
of methane at 0.58 atm and 1038 K. This model, obtained by a systematic simplification of a much larger one, consists
of 44 reactions involving 25 species. The model includes a number of reactions involving chemically activated complexes
formed by radical addition and recombination reactions, with rates for unimolecular reaction and bimolecular stabilization
of these energized complexes evaluated with a QRRK formalism. A combination of reaction rate analysis as well as sensitivity
analysis is used to show that these chemically activated reactions lead to surprisingly rapid production of cyclopentadiene
at very low extents of conversion ( ~ 0 . 1 % and
) that dissociation of cyclopentadiene accounts for the acceleration in rate.
Cyclopentadiene is produced by the addition of allyl radicals to acetylene; the chemically activated linear adduct undergoes
cyclization prior to collisional stabilization. An optimization procedure was used to obtain an accurate description of the
observed kinetics without compromising the theoretical plausibility of the rate constants.

Introduction progress recently in characterization of this rapid rate of soot

Although methane pyrolysis has been extensively studied,I some formation, attempts to accurately model this growth process have
aspects of the kinetics remain puzzling. In particular, Back and been only partially successful. For example, the extensive study
Back' have discussed a sharp increase in the rate of formation by Frenklach et al.3 required rate constants for radical addition
of ethane at very low levels of conversion (~0.1%). They per- reactions that are several orders of magnitude larger than usually
formed an exhaustive search for possible experimental artifacts observed for these types of reactions.
that might produce this acceleration and concluded that it was There appear to be three possible pathways to rapid molecular
probably a real effect, not attributable to reactions on surfaces, weight growth: (1) cycloaddition reactions between olefins and
etc. They speculated that it might be due to homogeneous re- dienes (Diels-Alder); (2) ion-molecule reactions with their very
actions of the reaction products but concluded that the mechanism large rate constant^;^ and (3) radical addition to unsaturates.
of this acceleration in rate was in considerable doubt. A major difficulty with (1) is that the equilibrium constants
Roscoe and Thompson2 addressed this autocatalysis in methane for this type of reaction suggest that, at the high temperatures
pyrolysis by proposing a detailed model that appeared to satis- where growth is observed to occur, the cyclic species would tend
factorily describe the acceleration in rate. This mechanism was to dissociate to the linear (smaller) fragments. A major difficulty
consistent with the spirit of Back and Back's discussion and ap- with an ionic mechanism is that it appears unlikely that the
peared to present a nice solution to the autocatalysis issue. concentration of ions in a pyrolytic environment (where chem-
However, a closer inspection of this mechanism, as described ionization reactions involving O2could not occur) could be suf-
below, indicated that it appeared to have been constructed in such ficiently high to have any impact. By default, our attention (like
a way that forward and reverse rate constants for a specific ele- that of Frenklach et a].) has focused upon radical addition re-
mentary reaction were not necessarily related by the appropriate actions. Previously, we demonstratedS that new reaction channels
equilibrium constant. When the reverse rate constants were could open up at higher temperatures in chemically activated
obtained explicitly from the forward values with microscopic reactions (such as radical addition to unsaturates). Methane
reversibility, it was observed that the model results of Roscoe and pyrolysis appeared to offer an opportunity to see if these new
Thompson were dramatically affected and virtually all of the channels could account for the observed autocatalysis.
autocatalysis disappeared.
This reaction system is particularly important since it has the Results and Discussion
prospect of providing an explanation for the process of rapid Roscoe-Thompson Mechanism. The mechanism presented by
molecular weight growth. It is well-known that methane pyrolysis Roscoe and Thompson2 was used to calculate concentration-time
can produce copious quantities of tar and coke. In fact, such profiles for the conditions (neat CH4, T = 1038 K, P = 0.58 atm)
production has been the primary reason for not using methane reported by Back and Back.l These calculations utilized the
as a feedstock in thermal cracking processes.' This propensity CHEMKIN package6 with the LSODE' stiff equation solver. Two
to form very high molecular weight species is directly attributable sets of calculations were done. In the first, following Roscoe and
to the stability of methane. At temperatures sufficiently high that Thompson, the system was treated as a set of irreversible reactions,
it is thermodynamically possible to produce significant yields of with use of their rate constants. Most of the reverse reactions
hydrocarbons such as acetylene and benzene from methane, the were also entered as irreversible reactions with rate constants
most favored thermodynamic products are graphite and hydrogen. explicitly assigned. In the second set of calculations, the system
It has proved exceedingly difficult to kinetically trap significant was treated reversibly, with the forward rate constants so desig-
quantities of such products without concomitant formation of nated by Roscoe and Thompson; here, all reverse reactions were
unwanted tar and coke. Thus, the data of Back and Back that automatically considered, with rate constants obtained via com-
illustrate this tendency toward an accelerating rate of production
of heavier hydrocarbons under conditions where a careful product
analysis has been done offer a nice opportunity to explore details (3) Frenklach, M.; Clary, D.W.; Gardiner, W. C., Jr.; Stein, S. Twenty-
of the kinetics of the growth process. first Symposium (International) on Combustion; The Combustion Institute:
The unusually rapid rate of growth at high temperatures is also Pittsburgh, PA, 1986; pp 1067-1076.
exhibited in flames as formation of soot. Although there has been (4) Calcote, H. F. Combust. Flame 1981, 42, 215-242.
(5) Dean, A. M. J . Phys. Chem. 1985,89, 4600-4608.
(6) Kee, R. J.; Miller, J. A.; Jefferson, T. H. CHEMKIN: A General-
( I ) Back, M. H.; Back, R. A. In Pyrolysis: Theory and Industrial Purpose, Problem-Independent, Transportable, Fortran Chemical Kinetics
Practice; Albright, L. F., Crynes, B. L., Corcoran, W.H., Eds.; Academic Code Package. Report SAND80-8003; Sandia National Laboratory: Liv-
Press: New York, 1983; pp 1-24 and references cited therein. ermore, CA; 1980.-
(2) Roscoe, J. M.;Thompson, M. J. Inr. J . Chem. Kinet. 1985, 17, (7) Hindmarsh, A. C. In Scientific Computing, Stepleman, R. S., et al.,
967-990. Eds.; North-Holland: Amsterdam, 1983; pp 55-64.

0022-3654/90/2094- 1432$02.50/0 0 1990 American Chemical Society

Kinetic Modeling of Autocatalysis in Methane Pyrolysis The Journal of Physical Chemistry, Vol. 94, No. 4, 1990 1433
n
larger than that observed for methyl.) Another result of chemical
(22 activation at high temperatures is the possibility of providing the
energy required for larger radicals to cyclize:I0

Detailed analysis5 indicated that the temperature and pressure

conditions under which rapid growth has been observed to occur
were such that these types of chemical activation were indeed
possible and that they could be responsible for rapid growth.
To apply this approach to the case of methane pyrolysis, a large
mechanism was assembled to attempt to describe the chemistry
in this system. This mechanism included our estimates for the
Lp I I I , I
rate constants for both radical addition and recombination rate
5000 10000 15000 20000 25000 30000 constants that were obtained with use of the methods described
TIME (SEC) in ref 5. The approach used was to estimate the unimolecular
Figure 1. Comparison of predicted and observed production of ethane reaction rates of the energized adduct formed in the reaction with
from methane pyrolysis at 1038 K and 0.58 atm. Key: (A)observed, a quantum version of RRK theory. Previous studies5*" have
from ref 8 (0.57 atm); (+) observed, from ref 1; (-D-) predicted using demonstrated the validity and utility of this method. One im-
Roscoe-Thompson irreversible mechanism; (-0-) predicted using Ros- provement in the estimation method was that the preexponential
coe-Thompson reversible mechanism. factors and barriers for reactions of the energized complexes were
obtained via a specific calculation of the appropriate equilibrium
puted equilibrium constants. The results of these calculations are constant as opposed to the generic estimates used earlier. Rate
compared to two sets of observations for ethane production in constants for radical isomerizations were assigned on the basis
Figure 1. One set of the observations is taken from a figure in of thermochemical kinetic arguments; A factors accounted for the
ref 1 , and the other is taken from a table in an earlier paper by loss of rotors in the transition state, and barriers were estimated
Chen et ai.* Although the irreversible calculation does a good on the basis of ring strain, enthalpy change, and intrinsic activation
job of replicating the observed increase in ethane production near energy for the hydrogen-transfer process. For example, the 1,3
1500 s, it is immediately obvious that virtually all of this accel- H shift
eration is removed when the system is treated reversibly. A
comparison of rate constants reveals that the most likely cause CH,CH=C'H * C'H2CHXH2
for the discrepancy is that the value used by Roscoe and Thompson
for the reaction was described as having A = 5 X 10l2s-' and E, = 36 kcal/mol.
The high-pressure limiting rate constants for the recombination
H + C2H4 -+ C2H5 and addition rate constants were taken from literature sources
is over a factor of 100 larger than that obtained via microscopic such as Tsang9 and the C R C tables.I2 Apparent rate constants
reversibility with their value for ethyl dissociation. (Much of the for the various reaction channels were computed as functions of
problem here seems to lie with the choice of rate constant for both pressure and temperature. These values obtained a t a
dissociation of ethyl; the value used in the mechanism is a factor pressure of 0.58 atm were then fitted to an expression in modified
of 20 lower than that recommended by TsangS9) Use of the Arrhenius form over the widest temperature range within the
mechanism employing microscopic reversibility forces the rate interval 300-2500 K such that this expression yielded rate con-
constant for addition of hydrogen atoms to ethylene to be un- stants within 10% of the QRRK-computed values over the range.
realistically low. Since the Roscoe-Thompson sensitivity analysis Since these modified Arrhenius parameters are fits to apparent
shows this addition to be important for ethane production at later rate constants and reflect the competition between various uni-
times and that it has a positive sensitivity coefficient, i.e., an molecular dissociation channels and collisional deactivation, no
increase in rate constant will increase the ethane yield, it is not physical or chemical significance should be attributed to the value
surprising that this lower value led to a substantially decreased of any specific parameter; these are best interpreted as simply a
rate of ethane production, thus minimizing the acceleration. convenient way to describe the complex temperature dependence.
Attempts to improve the fit by adjusting some of the forward Moreover, it is important to note that these QRRK estimates are
rate constants, consistent with theoretical guidelines, were not applicable only at 0.58 atm; the extent of collisional deactivation
successful. It appears that the mechanism proposed by Roscoe scales with pressure and thus the apparent rate constants must
and Thompson is not adequate to explain the observed acceleration be considered to be functions of pressure as well as temperature.
in rate for ethane production when proper account is taken of These estimates were combined with our best estimates for
microscopic reversibility. hydrogen-transfer and disproportionation rate constants, relying
Generation of Energized-Complex Mechanism. Another heavily on the compilations of T ~ a n g .Abstraction
~ reactions that
possible mechanism to explain the autocatalysis is the onset of formed resonantly stabilized radicals were assigned rate constants
new reaction channels in radical addition reacti0ns.j For example, based upon the CRC tabulations.12 Preexponential factors were
high temperatures significantly increase the probability that the scaled to reflect the number of abstractable hydrogens, and ac-
adduct formed when C H 3 adds to C2H4 will have enough energy tivation energies were adjusted to reflect abstraction of primary,
to break a C-H bond: secondary, or tertiary hydrogens. Similarly, activation energies
were lowered for abstraction reactions involving even weaker C-H
CH3 + C2H4 * C3H7' C3H6 +H bonds (such as in cyclopentadiene) with use of Evans-Polanyi
relationships. Other critical inputs to the model are the updated
Such a reaction opens up two possibilities for more rapid growth. estimates of the enthalpies and entropies of alkyl radicals; these
The first is prpduction of a stable molecule of higher molecular were based on Tsang.13 (These values were used to compute the
weight. The second, perhaps more important, aspect is that these equilibrium constants needed for the analysis of the Roscoe-
reactions convert relatively unreactive methyl radicals into much
more reactive hydrogen atoms. (For example, both abstraction
and addition rate constants for H atoms are orders of magnitude (IO) Westmoreland, P. R.; Dean, A. M.; Howard, J. B.; Longwell, J. P.
J . Phys. Chem., in press.
( 1 1 ) Westmoreland, P. R.; Howard, J. B.; Longwell, J. P.; Dean, A. M.
(8) Chen, C. J.; Back, M. H.; Back, R. A. Can. J . Chem. 1976, 54, AIChE J . 1986, 32, 1971-1979.
3 175-3 184. (1 2) Handbook of Bimolecular and Termolecular Gas Reactions; Kerr,
(9) Tsang, W. J . Phys. Chem. Ref. Data 1986, 15, 1087-1279; Ibid. 1987, J. A,, Ed.; CRC Press: Boca Raton, FL, 1981; Vol. I and 11.
16, 471-508. (13) Tsang, W. J . Am. Chem. Soc. 1985, 107, 2872-2880.
1434 The Journal of Physical Chemistry, Vol. 94, No. 4, 1990 Dean

I I I I
5000 10000 15000 20000 25000 30000 500.0 1000.0 1500.0 2000.0 2500.0 30000
TIME (SEC) TIME (SEC)
Figure 2. Comparison of predicted and observed production of ethane Figure 3. Comparison of predicted and observed production of ethylene
from methane pyrolysis at 1038 K and 0.58 atm. Key: (0) observed, from methane pyrolysis at 1038 K and 0.58 atm. Key: (0) observed,
from ref 8 (0.57 atm); (A)observed, from ref 1; (-m)predicted using from ref 8 (0.57 atm); (A)observed, from ref 1; (-D-)predicted using
Table I mechanism. Table I mechanism.

Thompson mechanism as well.) The final mechanism consisted

of 438 reactions involving 122 species. This mechanism was
systematically condensed via successive applications of sensitivity
ana1y~is.I~For the conditions of the Chen et al. experiments,
it was possible to obtain essentially the same results with a much
smaller mechanism involving 44 reactions among 25 species.
Optimization and Comparison to Observations. Although this
smaller mechanism gave definite indications of autocatalysis, the
I
magnitude was somewhat less than observed in the experiments. I
Since the sensitivity analysis indicated that a few reactions were
particularly important for all of the observed species and since
it is clear that there remains significant uncertainty as to the I

accurate assignment of rate constants, a constrained optimization

procedureI5 was employed to try to improve the model description 5000 10000 15000 20000 25000 30000
while requiring the rate constants to remain theoretically plausible. TIME (SEC)
It was observed that small changes in only three rate constants
markedly improved the fit. The first of these changes involved Figure 4. Comparison of predicted and observed production of acetylene
the initiation reaction: from methane pyrolysis at 1038 K and 0.58 atm. Key: (0)observed,
from ref 8 (0.57 atm); (A)observed, from ref 1; (--O-) predicted using
CH, C H 3 + H, k = 2.2 X IO7 s-I Table I mechanism.

This value is only 40% higher than that recommended by TsangI'

(suggested uncertainty of 50% for k,, presumably more when in
falloff region). The value used for
C2H6 + CH3 C2HS+ CH,, k = 5.7 X lo9 cm3 mol-] s-l

is approximately a factor of 2 lower than that recommended by

Tsang. However, there are few data points in this particular
temperature range, and the value used here is still consistent with
upward curvature on Arrhenius plots frequently observed with
/
i
hydrogen abstraction reactions; simple extrapolation of lower
temperature datal2 would suggest a value near 2 X lo9. The last
rate constant adjusted was for methyl abstraction from cyclo-
pentadiene:
5000 10000 15000 20000 25000 3( 10.0

+ CH3 = 0+ CH4, k = 2.2 x 10'' cm3 mol -' s-' TIME (SEC)
Fipre 5. Comparison of predicted and observed production of propylene
This rate constant is only 25% higher than one would estimate from methane pyrolysis at 1038 K and 0.58 atm. Key: (0) observed,
from ref 8 (0.57 atm); (A)observed, from ref 1; (-0-) predicted using
on the basis of C R C data on abstraction of species that form Table I mechanism.
resonantly stabilized radicals when one accounts for the unusually
weak bond in cyclopentadiene. Thus, it appears that the ad- This mechanism is listed in Table I, and the results for ethane
justments made in the mechanism are completely consistent with production are compared to the observations in Figure 2. Note
known uncertainties and do not represent excessive parametrization that the mechanism accurately predicts the autocatdlysis. Figures
in an attempt to achieve a better fit. 3-5 illustrate the comparisons between predictions and observa-
tions for C2H4,C2H2,and C3H6. The fit to ethylene is excellent,
(14) Lutz, A. E.: Kee, R. J.; Miller, J. A. SENKIN: A Fortran Program but the model overpredicts acetylene and slightly underpredicts
for Predicting Homogeneous Gas Phase Chemical Kinetics with Sensitivity propylene. Since both acetylene and propylene are present at much
Analysis; Report SAND87-8248; Sandia National Laboratory: Livermore, lower concentrations, it is possible that some of this discrepancy
CA, 1988. might be related to analytical difficulties. (Note the scatter in
( I 5) Byrne, G.D.; Dean, A. M.; McCroskey, P. S. In Proceedings ofthe
Second International Symposium on Computational Chemistry on Cray the C2H2data in Figure 4.) However, it is certainly possible that
Supercomputers; Cray Research, Inc.: 1988; 209-21 9. there remain some mechanistic problems. Attempts to improve
Kinetic Modeling of Autocatalysis in Methane Pyrolysis The Journal of Physical Chemistry, Vol. 94, No. 4, 1990 1435

n
0 = OVERALL REACTION
A = CY13PD=CY13PDS+H
+ = CH3+CY13PD=CH4+CY13PD5
X = H+CYl3PD=HZ+CY13PD5
0 0 = CYPENE4.=CY13PD+H
V = CCC +CY13PD5.=OCC+CY13PD
tX = C’CC +CZH2=CY13PD+H
m
-r , I I I 520 0 1040 0 1560 0 2000 0 2600 0
5000 10000 15000 80000 25000 30000 TIME (SEC)
TIME (SEC)
Figure 7. Reactions of cyclopentadiene with the largest net rates of
Figure 6. Comparison of predicted and observed product ion of ethane production and destruction are plotted as a function of time, based on
from methane pyrolysis at 1038 K and 0.58 atm. Key: (A)observed, the mechanism in Table I at 0.58 atm and 1038 K. Note that the
from ref 8 (0.57 atm); (+) observed, from ref 1; (-0-) predicted using reaction CH3 + CY 13PD = CH4 + CY 13PD5’ is running backwards to
Table I mechanism; (-O-) predicted using Table I mechanism without produce cyclopentadiene.
reactions involving species with more than four carbon atoms.
the fit to these species were not successful without resorting to
extensive manipulation of rate constants. In the author’s opinion, -1
such manipulation was unjustified and was not pursued seriously.
Nevertheless, the overall success of this mechanism in describing
most of the observations suggests that this mechanism indeed
captures the essence of the observed autocatalysis.
The results of a sensitivity analysis for this mechanism are
shown in Table I1 as an ordered list of the ten most important
reactions for the major products observed by Chen et al. (In the
case of methane, only seven reactions are shown because all others
had normalized sensitivity coefficients of less that 1 X 10” at all
times.) An interesting feature of the sensitivity analysis is that
the same group of reactions are seen to be important for all of
the major products. Note that only 14 reactions are listed in Table 520 0 1040 0 1560 0 2060 0 2600 0
11. If one extends the optimization procedure to include more TIME (SEC)
of these reactions, the fit to the data is not observed to improve Figure 8. Reactions of hydrogen atoms with the largest net rates of
significantly. More importantly, adding additional reactions production and destruction as a function of time, based on the mechanism
generally had little effect upon the ”optimized” values of the three in Table I at 0.58 atm and 1058 K. Note that two reactions, ethyl
rate constants described above. This provides additional evidence dissociation and cyclopentadiene dissociation, are the main sources of
that the values used to obtain the fits in Figures 2-5 have physical hydrogen.
significance. Of course, as in any modeling exercise, the rate
constants obtained should not be considered to be as reliable as also produces hydrogen atoms that we expect to be particularly
those directly measured in simpler systems designed to yield such reactive. Note that these rates become important just as the
rate data. acceleration in rate is observed for ethane. These two reactions ’

Reasons for Autocatalysis. We have seen above that the serve as an efficient route for radical production:
predicted profiles were sensitive to reactions of cyclopentadiene.
This was quite surprising since the overall extent of reaction in
this system is very low (0.09%conversion at 1600 s) and one might
not have expected the reactions to produce such a large molecule,
much less to have it play a dominant role in the system. This
Q=Q+H
sensitivity is illustrated in Table 11, where a number of reactions
involving cyclic C5 species are listed. This conclusion is reinforced
by removing the reactions involving the larger species (any reaction
Net: CH4 = CH3 + H

that involved a species with more than four carbons) from the Figure 8 shows the most important reactions for production of
mechanism; the autocatalysis is completely suppressed. These hydrogen atoms; note that dissociation of cyclopentadiene is as
results are shown in Figure 6. It is noteworthy that this behavior important as B scission of ethyl radicals. Thus, the essential role
is quite similar to that observed with the Roscoe-Thompson of cyclopentadiene in acceleration of methane pyrolysis becomes
mechanism when that mechanism incorporated microscopic re- clear: this molecule has the sufficiently weak C-H bond required
versibility. It illustrates that the main problem with that mech- to allow onset of a new chain-branching reaction.
anism is simply that it did not consider enough of the molecular Role of Energized Complexes in the Acceleration. Given the
weight growth chemistry. importance of cyclopentadiene, attention was focused on the routes
The most important reactions involved in the production and for its production. A potential energy diagram is shown in Figure
decay of cyclopentadiene are shown in Figure 7. Note that the 9. We want to ascertain the relative importance of “equilibrated”
major path for production is the abstraction of a hydrogen from vs “energized-complex” mechanisms in this reaction sequence. In
methane by cyclopentadienyl (reaction 25 running backwards). the conventional approach (here termed equilibrated to denote
Although this reverse reaction is very endothermic with a cor- the dominance of deactivating collisions), the cyclopentadiene
respondingly small rate constant, the large concentration of production would be considered to be the result of three distinct
methane shifts this reaction toward this direction. Once formed, reactions: (1) formation of a linear adduct from the addition of
the cyclopentadiene undergoes unimolecular dissociation. This allyl to acetylene in which the excess energy has been removed
dissociation is really the critical step for the acceleration in rate; in deactivating collisions; (2) isomerization of this adduct to the
not only does this regenerate the cyclopentadienyl radical, but it cyclic species, where the energy required for the isomerization
1436 The Journal of Physical Chemistry, Vol. 94, No. 4, 1990 Dean

TABLE I: Methane Pyrolysis Mechanism"

no. reactionb Ah n E, refs kA1038 K)h kA1038 K)h
1 CH4 = CH, + H (1038 K, 0.58 atm) 2.21E-07 0 0 see text 2.21 E-07 2.5 1E+13
2 C2H6 = 2CH3 (1038 K, 0.58 atm) 5.71E-03 0 0 9 5.71E-03 8.768+12
3 CHg + CH3 = C2H5 + H 1.80E+I 2 0 IO 400 9 I.l6E+10 4.32E+13
4 C=CC = C=CC' + H (900-2100 K) 5.888+47 -9.483 107 303 QRRK 3.74E-04 6.26E+13
5 C=CC = C2H3 + CH3 (900-2100 K) 1.478+56 -1 1.426 117905 QRRK 7.54E-04 1.77E+13
6 CECC = CZCC' + H (600-2100 K) 8.79E+29 -4.41 1 98 213 QRRK 9.10E-05 6.83E+13
7 CY I3PD = CY 13PD5' + H (600-2100 K) 8.13E+24 -2.981 78 682 QRRK 2.24E-01 5.96E+13
8 CH3 + CY13PD5' = CHD (900-1200 K) 1.01E+67 -1 5.663 35 064 QRRK 2.39E+l2 1.79E+02
9 CH, + CY 13PD5' = CYC6H7 + H (900-1200 K) 2.44E+41 -7.989 39 259 QRRK 1.06E+09 6.13E+ 13
IO CHD = CYC6H7+ H (600-2100 K) 4.35E+22 -2.230 79 806 QRRK 1.28E-01 9.94E+ 13
I1 2CY13PD5' = NAPH + H2 (900-2100 K) 4.30E+36 -6.268 45671 QRRK 1.29E+08 4.35E-05
I2 C2HS= C2H4 + H (1038 K, 0.58 atm) 8.94E+04 0 0 9 8.94E+04 5.39E+12
13 C2H3 = C2H2 + H (600-2500 K) 1.93E+28 -4.783 51 123 QRRK 1.24E+03 6.82E+Il
14 CYPENE4' = C=CCC=C' (900-2100 K) 1.52E+58 -1 3.062 60616 QRRK 1.O5E+06 2.99E+08
15 CYPENE4' = CY13PD + H (900-2100 K) 1.028+58 -1 3.066 60 155 QRRK 8.56E+05 6.24E+ 12
16 CYC6H7 = C6H6 + H (900-2100 K) 2.70E+57 -13.102 48 985 QRRK 3.97E+07 3.51E+ 12
17 CH4 + H = CH, + H2 2.20E+04 3.0 8 750 9 3.54E+Il 1.70E+ 10
18 C2H6 + H = C2H5 + H2 5.40E+02 3.5 5 200 9 1.568+12 3.788+09
19 CY13PD + H = H2 + CYl3PD5' 4.00E13 0 3 000 estC 9.34E+ 12 1.05E+06
20 C2H6 + CH, = C2H5 + CH4 2.70E-01 4.0 8 280 see text 5.66E+09 2.84E+08
21 C2H4 + CH, = CIH, + CH4 4.20E+11 0 1 1 100 12 1.93E+09 1.228+09
22 C=CC + CH, = C=CC' + CH4 3.80E+11 0 9 000 12 4.848+09 7.1 5E+06
23 C z C C + CH, = CcCC' + CH4 3.80E+11 0 9 000 estd 4.848+09 3.20E+07
24 C=C=C + CHj = C=CC' + CH4 3.80E+I 1 0 9 000 estd 4.848+09 8.86E+06
25 CY 13PD + CH, = CH4 + CY 13PD5' 3.11E+ll 0 5 500 see text 2.15E+10 5.04E+04
26 CHD + CH3 = CYC6H7 + CH4 5.00E+II 0 6 300 estC 2.368+30 1.61E+05
27 H + C=CC = CCC' (600-2500 K) 1.37E+42 -8.823 15 979 QRRK 1.45E+ 12 2.53E+05
28 H + C=C=C = C=CC' (600-2 100 K) 7.89E+23 -3.365 8 099 QRRK 1,10E+12 2.72E+00
29 H + C E C C = CC=C' (600-2500 K) 4.22E+35 -7.023 12441 QRRK 6.65E+11 4.92E+03
30 H + C=CC = C=CC' (600-2100 K) 2.41 E+29 -4.780 15505 QRRK 5.01E+ll 4.478+00
31 CH, + C2H4= CCC' (900-2100 K) 3.658+48 -10.896 27 860 QRRK 6.78E+09 3.628+06
32 CH3 + C2H4 = C=CC + H (600-2500 K) 1.98E+22 -2.879 25 442 QRRK 1.80E+08 5.51E+I 1
33 CH, + C2H2 = CC=C' (600-2500 K) 7.28D+34 -6.921 18 338 QRRK 1.34E+l0 1.448+05
34 CH, + C2H2 = C=CC + H (600-2100 K) 6.83D+16 -1.442 21 184 QRRK 1.06E+08 1.54E+ll
35 CH, + C2H2 = C=CC' (900-2100 K) 1,10D+45 -9.449 35412 QRRK 1.22E+09 1.58E+01
36 CH3 + C=CC = C=CCC + H (600-2500 K) 3.37E+24 -3.554 29 636 QRRK 3.70E+07 1. I2E+12
37 CH, + C=C=C = C2H3 + C2H4 (600-2500 K) 3.46E+33 -6.058 33915 QRRK 1.348+08 1.04E+07
38 C=CC' + C2H2 = C=CCC=C' (300-2500 K) 8.388+30 -6.242 12824 QRRK 2.49E+09 1.03E+07
39 C=CC' + C2H2= CYPENE4' (600-2500 K) 3.42E+52 -12.194 27 978 QRRK 7.30E+09 1.06E+05
40 C=CC' + C2H2 = CY 13PD + H (600-2500 K) 2.95E+32 -5.829 25 733 QRRK 2.95E+09 3.13E+ll
41 CH, + C2H5 = CH4 + C2H4 (900-1700 K) 5.50E+11 0 0 9 5.50E+11 2.92E-01
42 CH, + C=CC' = CHI + C=C=C 1.OOE+ 12 0 0 est/ 1.OOE+ 12 3.57E+03
43 C=CC' + C=CC' = c=cc + c=c=c 1.OOE+ 12 0 0 est/ 1.OOE+ 12 2.41E+06
44 C=CC' + CY 13PD5' = C=C=C + CY 13PD 1.OOE+ I2 0 0 estf 1.00E+12 1.53E+09
" k = A T " exp(-EJRT), units = s-' (first order) or cm3 mol-' s-' (second order). bAbbreviations: C=CC' = allyl; CY13PD = 1,3-cyclo-

on CH, + C=CC - -
pentadiene; CY 13PD5' = cyclopentadienyl;CHD = 1,3-cyclohexadiene; CYPENE4' = 4-cyclopentenyl; CYC,H7 = 1,3-hexadienen-5-y1;NAPH =
naphthalene. CBasedon H + C=CCC H2 + C=CC'C (ref 12) with lower E, to account for weaker C-H bond (Evans-Polanyi plot). dBased
CH4 + C=CC'. CBasedon CH, + C=CCC (ref 12) with A X 2 (twice the number of abstractable H) and lower E, to
account for weaker C-H bond (Evans-Polanyi plot). /Assumed faster than CH, + C2H5since involving resonantly stabilized radicals (Allara, D. L.;
Shaw, R. J . Phys. Chem. Ref Data 1980, 9, 523-559). fQRRK signifies that the rate constant listed is a modified Arrhenius fit to the calculated
chemically activated apparent rate constants at 0.58 a t m over the specified temperature range. hThese values follow the pattern 2.21E-07 = 2.21 X
I 0-7.

is obtained via collisional activation; and (3) collisional activation Equilibrated

of the cyclic species so that it could undergo p scission to cyclo-
pentadiene and hydrogen atoms. The energized-complex approach
does not assume that the initially formed linear adduct will always
be collisionally quenched; it allows the unimolecular cyclization
and/or p scission to compete with collisional stabilization. Note
that the two approaches will converge in the limit of high pressure
where all adducts must be collisionally deactivated prior to any Energized Complex Followed by Equilibrated
unimolecular reaction. So the question is really whether, under
the conditions of these experiments, the pressure is low enough (ECE1)
to make the time required for the bimolecular energy-transfer
processes implicit in the conventional approach to be longer than
that required for unimolecular reactions of the energized adducts.
0'--Q +H (ECE2)

Two separate analyses indicate that the energized-complex route Energized Complex
plays an important role under the conditions of the Chen et al.
experiments. The first of these is based upon the computed rate
constants for production of the various C, species by the addition The rate of production of cyclopentadiene was computed for the
of allyl to acetylene. The Q R R K treatment incorporated in the first two pathways by using a steady-state analysis in conjunction
mechanism shown in Table I has three ways to make cyclo- with the appropriate rate constants in Table I. These rates were
pentadiene: compared to that of the direct path ECl. It was found that the
Kinetic Modeling of Autocatalysis in Methane Pyrolysis The Journal of Physical Chemistry, Vol. 94, No. 4, 1990 1437

*
5;N
C=CC' + C,H, C=CCC=C"
0. Q + + H d l

100
3
L

-s
8
:
90

80
Lf
C=CCC=C'
7

P
LU 70

9
5000 10000 15000 20000 25000 3000
300000
60
TIME (SEC)
Figure 10. Comparison of the production of cyclopentadiene via a con-
50
ventional vs an energized-complex treatment. Key: (-O-) predicted
using Table I mechanism (includes energized-complex routes); (-0-)
Figure 9. Potential energy diagram for the addition of allyl radical to predicted using Table I mechanism where direct routes to cyclic products
acetylene, illustrating the possible pathways for reactions of the inter- from allyl addition to acetylene were omitted.
mediates.
equilibrated route E1-E3 only accounted for approximately 10%
of the total rate, while the "mixed" route ECE1-ECE2 accounted
for approximately 60% of the total rate. An inspection of Figure P I
9 offers an explanation of these results. Note that the isomeri-
zation barrier is below the entrance barrier and that there is a
deep well for the cyclic intermediate. This will result in a rapid
unimolecular reaction for this channel, sufficiently rapid to com-
pete favorably with deactivation. However, there remains a sizable
barrier for p scission of the cyclic radical. This, coupled with the
relatively low A factor for the 6-scission reaction (2.4 X 1O1j s-I),
results in significant deactivation to the cyclic species. This is
converted to the final products in a subsequent step. Thus, the
direct formation of cyclopentadiene as well as its cyclic precursor T

via unimolecular reaction prior to stabilization was found to be 500.0 1000.0 1500.0 2000.0 2500.0 3000.0
important in this system. TIME (SEC)
Another way to test the importance of these routes is to modify Figure 11. Comparison of the production of ethane via a conventional
the mechanism in Table I to reflect only the conventional routes vs an energized-complex treatment. Key: (-O-) predicted using Table
to cyclopentadiene. Thus, reactions 39 and 40 were removed, and I mechanism (includes energized-complex routes); (-e) predicted using
the rate constant for reaction 38 was increased (to 8 X lo9 cm3 Table I mechanism where direct routes to cyclic products from allyl
mol-' s-I) to reflect its rate if no other channels were available addition to acetylene were omitted.
for the initially formed linear adduct other than stabilization or
redissociation to the reactants. Figure 10 compares the production
of cyclopentadiene with this modification to the original one. Note
C=Cc' + C = C C=CCCC*' $ -
0" + H

that substantially more cyclopentadiene is formed when one allows

the chemically activated adducts to react. This change in pro-
duction of cyclopentadiene is also manifested in increased ethane
yield; Figure 1 1 clearly illustrates the role of the energized adducts
in increasing the magnitude of the observed acceleration.
ioi
In general, one finds that temperature is a much more dominant
variable than pressure in determining the importance of ener-
gized-adduct chemistry. The unimolecular rate constants scale
approximately exponentially in temperature while the stabilization ? 130 -
C.CC.C'

A
+ c c
I +? c.cc.ccsc~' p
I 0"-0+ H
g .
scales linearly with pressure. Thus, the observation that the
energized-complex reactions play a role in methane pyrolysis at
1038 K suggests that they are to be expected to play an even more
i
Y 110 -
I ! i
important role at higher temperatures, even if the pressure is raised.
It appears that the autocatalysis is a manifestation of the ex-
tremely rapid production of higher molecular weight species and
that this production is augmented via reactions of chemically
activated adducts. As we found in our earlier studies, these adducts
90 -

70 -
9iC=CC=CC.C'

have enough internal energy at these temperatures to react

unimolecularly before collisional stabilization can occur; this 50 -
significantly increases the overall rate of production of heavier Figure 12. Comparison of the potential energy diagrams for (a) addition
species because no time is required for collisional deactivation and of allyl radical to ethylene and (b) addition of butadienyl radical to
subsequent activation to get to these same products. acetylene. The differences in these diagrams account for the markedly
An interesting feature of the reaction mechanism is the fact different behavior of these two addition reactions.
that, although addition of allyl to acetylene is seen to be critical,
the analogous reaction, addition of allyl to ethylene, is not needed, cyclizations is the energetics. The potential energy diagram for
even though ethylene is present in much larger concentrations than allyl addition to ethylene is shown in Figure 12a, and can be
acetylene. Analysis of these two systems illustrates some important compared to that for addition to acetylene in Figure 9. Note the
ideas about cyclization of radicals. The major difference in these first difference is the depth of the initial well; it is deeper for
1438 The Journal of Physical Chemistry, Vol. 94, No. 4, 1990 Dean

TABLE 11. Sensitivity Information'

reaction times. s
no. reactionb 60 I 1204 1839 2400
CH,
I CH4 = CH3 + H -0.002 -0.008
26 H + C=CC = CCC' -0.002
24 CH, + CY13PD = CH4 + CY13PD5' -0.002
2 C2H6 = 2CH3 0.002
13 C2H3 = C2H2 + H -0.002
7 CY 13PD = CY 13PDS' + H -0.001
38 C=CC' + C2H2 = CYPENE4' -0.001
C2H6
1 CHI = CH, + H 0.701 0.797 1.49 1.70
19 C2H6 + CH3 = C2H5 + CH4 -0.592 -0.677 -0.605 -0.592
24 CH3 + CY13PD = CH, + CY13PD5' 0.025 0.245 0.632
26 H + C=CC = CCC' 0.006 0.069 0.321 0.500
13 C2H3 = C2H2 + H 0.001 0.040 0.249 0.390
7 CY13PD = CY13PD5' + H 0.032 0.210 0.357
38 C=CC' + C2H2 CYPENE4' 0.026 0.177 0.276
31 CHp + C2H4 = C=CC + H 0.003 0.030 0.140 0.2 18
11 2CY13PD5' NAPH + H2 -0.035 -0.212
39 C=CC' + C2H2 CY13PD + H 0.014 0.097 0.151
C2H4
1 CHI = CH3 + H 1.17 1.08 1S O 2.27
24 CH, + CY13PD = CH4 + CY13PD5' 0.009 0.134 0.560
19 C2H6 + CH3 = C2H5 + CH4 0.439 0.253 0.252 0.300
26 H + C=CC = CCC' -0.045 -0.065 0.07 1 0.406
2 C2H6 = 2CH3 -0.078 -0.028 -0.131 -0.384
7 CY 13PD = CY 13PDS' + H 0.012 0.129 0.381
13 C2H3 = C2H2 + H -0.029 -0.049 0.060 0.330
38 C=CC' + C2H2 = CYPENE4' 0.0 10 0.108 0.308
31 CH3 + C2H4 = C=CC + H -0.020 -0.028 0.031 0.177
39 C=CC' + C2H2 = CY 13PD + H 0.005 0.059 0.169
C2H2
1 CH4 = CH3 + H 1.69 1.52 1.61 1.86
13 C2H3 = C2H2 + H 0.678 0.561 0.559 0.682
26 H + C 4 C = CCC' 0.115 0.183 0.286 0.439
19 C2H6 + CH3 C2H5 + CH4 0.420 0.177 0.129 0.173
24 CH3 + CY13PD = CH4 + CYl3PD5' 0.009 0.102 0.407
2 C2H6 = 2CH3 -0.324 -0.227 -0.268 -0.398
7 CY 13PD = CY 13PD5' + H 0.01 1 0.092 0.259
31 CHI + C2H4 = C 4 C + H 0.050 0.080 0.125 0.191
28 H + C=CC = CC=C' -0.037 -0.074 -0.114 -0.138
34 CHg + C2H2 = C=CC' 0.086 0.107 0.100 0.106
c=cc
1 CHI = CHg + H 1.63 1.48 I .75 2.23
26 H + C=CC = CCC' 0.607 0.572 0.659 0.886
24 CHI + CY 13PD = CHa + CY 13PD5' 0.012 0.137 0.540
2 C2!& = ~ C H , -0.289 -0.202 -0.295 -0.498
19 C2H6 + CH, = C2H5 + CH4 0.410 0.164 0.140 0.210
31 CHI + C,Hd = C=CC + H 0.264 0.249 0.287 0.386
13 C2H3= d2H2+ H 0.006 0.024 0.137 0.340
7 CY13PD = CY13PD5'+ €I 0.013 0.115 0.317
38 C=CC' + C2H2 = CYPENE4' 0.006 0.08 1 0.229
5 C=CC = CzH3 + CHj -0.087 -0.140 -0.137 -0.095
H2
1 CH4 = CH, + H 1.1 1 1.18 1.69 2.52
26 H + C=CC = CCC' 0.021 0.076 0.272 0.639
24 CH3 + CY 13PD CH4 + CY l3PD5' 0.01 1 0.141 0.598
2 C2Hn = 2CH3 -0.05 1 -0.068 -0.216 -0.521
13 C;H; = C2H; + H 0.010 0.043 0.198 0.491
7 CY13PD = CY13PD5' + H 0.015 0.138 0.398
38 C=CC' + C2H2 = CYPENE4' 0.013 0.118 0.328
31 CH3 + C2H4 = C=CC + H 0.009 0.033 0.119 0.278
19 C2H6 + CHj = C2H5 + CH4 0.172 0.1 17 0.124 0.188
39 C=CC' + C2H2 = CY13PD + H 0.007 0.065 0.180
'Each entry is the normalized sensitivity coefficient, A i / X i (&Yi/&4,),. (Blank entries mean that the absolute value was less than 0.0005.) bSee
Table I for a list of abbreviations.

addition to acetylene by about 8 kcal/mol. Furthermore, the These differences in energetics continue with respect to formation
difference in energy between the linear and cyclic intermediates of final products; addition to acetylene is overall exothermic while
is greater in the acetylene case by about 9 kcal/mol. Most of this addition to ethylene is slightly endothermic. Each of these dif-
difference can be attributed to the fact that C=CCC=C' is a ferences contribute to enhancement in rate for formation of cyclic
relatively high-energy vinyllic radical whereas C==CCCC' is not. products for addition to acetylene. The shallower well for ethylene
Kinetic Modeling of Autocatalysis in Methane Pyrolysis The Journal of Physical Chemistry, Vol. 94, No. 4, 1990 1439

addition will lead to more stabilization of the linear adduct since at extremely small extents of reaction, is in fact extremely com-
the barrier to cyclic species is comparable to the entrance channel; plicated. There appear to be three reasons for the success of the
for acetylene it is 11 kcal lower. Similarly, the difference in energy present mechanism:
between linear and cyclic intermediates means that the equilibrium (1) It was possible to use a quick but accurate method to
constant for the cyclization involving C=CCC=C' is much more estimate apparent rate constants for addition and recombinations
favorable, by a factor of 60 at 1038 K, than for the cyclization reactions, thus making assembly of the large initial mechanism
of C=CCCC'. Thus, we expect cyclization, whether it occurs feasible.
via stabilized or energized adducts, to be much more favored for (2) By starting with a very large mechanism and then rigorously
the case of C=CCC=C'; under the present conditions, this pruning via sensitivity analysis, it was possible to identify important
difference is enough to make contributions from allyl addition to reactions of large species such as cyclopentadiene that were not
ethylene unimportant. In general, the shallower wells in the considered by earlier workers, who simply never considered that
ethylene system, coupled with the greater entropy of the reactants the mechanism could be so complicated.
with respect to the intermediates, favor redissociation of the ad- (3) The mechanism included the possibility of unimolecular
ducts back to reactants, thus making this an inefficient channel reactions of energized adducts arising from radical addition re-
for molecular weight growth. actions prior to their collisional stabilization; these reactions
These observations can be generalized to other addition reac- substantially increased the rate of production of cyclopentadiene,
tions. The following factors will increase the probability that direct which then serves as a new chain-branching center.
production of cyclic species can occur via an energized-complex It appears that chemically activated reactions like those iden-
mechanism: (1) a deep well for the linear adduct-this will tend tified in this analysis might be expected to play a major role in
to make the barrier to cyclization lower than the entrance barrier, a variety of systems. Given the importance of these reactions in
thus increasing its unimolecular rate; (2) an increase in the ex- this system at 1038 K, one expects an even greater role at higher
othermicity of the cyclization reaction-this is needed to offset temperatures where the unimolecular rates of the initially formed
the entropy loss upon cyclization; and (3) a final cyclic product adducts will be even faster. One way to estimate the potential
with high stability-this will result in low-energy exit channels impact is to observe the shift in falloff curves for radical recom-
relative to the entrance and will lead to faster unimolecular rates. bination reactions with temperature. This behavior directly tracks
One such system where these factors are especially significant is the competition between unimolecular reaction and bimolecular
the formation of benzene via the following sequence: stabilization, and we see dramatic effects here. For example,
increasing the temperature from 900 to 1350 K for methyl radical
C=CC=C* + C2H2eC=CC=CC=C*' Z recombination shifts the falloff curve by a factor of 50 toward

0" 0 === 0 + ti*

higher pressures-an increase from 1350 to 2200 K shifts it
another factor of 60.16 Thus, it is to be expected that one will
need to account for the chemically activated reaction channels
in some circumstances even at pressures of many atmospheres.
The potential energy diagram is shown in Figure 12b. Here, the
linear adduct is seen to have a much deeper well than that resulting It is also important to note that these effects are not limited
from the allylic addition shown in Figure 12a. There are two to small molecules. Although one associates a slower unimolecular
reasons for this: (1) addition to acetylene is generally more rate with an increasing number of vibrational modes since there
exothermic than to ethylene; (2) the adding radical here is vinyllic, is less likelihood of localizing the critical energy in the reaction
as contrasted to the resonantly stabilized allyl; thus there is no coordinate, this is only to be expected at a fixed energy. In
loss of resonance upon addition as is the case with allyllic species. chemically activated systems, the energy in the complex is com-
The cyclization is also much more favored in the benzene system posed of both the energy released when the new bond is formed
since the combination of an unstable vinyllic linear radical and and the internal energy of the reactants. As the reactants become
a very stable cyclic radical results in an exothermicity of 43 larger, their internal energy necessarily increases, and thus, the
kcal/mol for the cyclization. Furthermore, note that the final energy in the complex will increase. As a result, one finds that
exit channel for benzene production, due to its unusual stability the possibility of unimolecular chemistry is not particularly sen-
as an aromatic molecule, is much lower than the entrance. As sitive to molecular complexity. All of these factors suggest that
a result, this path to benzene can be very important. In fact, at it is important to explicitly consider the possible role of these types
1 atm and 1200 K, the rate of production of benzene via the direct of reactions in virtually any gas-phase system at high temperatures.
reaction of the energized complex accounts for over 90% of the Acknowledgment. I have benefited from many stimulating
total reaction of the initially formed linear adduct. Here, the discussions on various aspects of molecular growth with J. W.
overall exothermicity is sufficient to compensate for the loss in Bozzelli, P. R. Westmoreland, and R. L. Woodin.
entropy upon cyclization. Such compensation is particularly
important at high temperatures where the TPS term plays a larger Registry No. CH,, 74-82-8; C2Hz,74-84-0 cyclopentadiene, 542-92-7.
role.
Concluding Remarks. This description of methane pyrolysis (16) Warnatz, J. In Combustion Chemistry; Gardiner, W. C., Jr., Ed.;
illustrates that even this seemingly simple system, where one works Springer-Verlag: Berlin, 1984; pp 197-360.