High pressure, high temperature shock tube studies of ethane pyrolysis and oxidation

Article abstract of DOI:10.1039/b110702j

A unique high-pressure single pulse shock tube has been used to obtain the first experimental data for ethane oxidation and pyrolysis at very high pressures and temperatures. Experiments were performed at two nominal reaction pressures of 340 bar and 613 bar in the temperature range 1050 K to 1450 K. The major stable species were identified and their concentrations determined using gas chromatography. Several minor species, with up to four carbon atoms and including oxygenates, were also observed in the oxidation studies. Three models based on literature mechanisms for hydrocarbon oxidation were used to simulate the experimental data. All of the models simulate the pyrolysis data well although only one of the models was capable of accurately describing the oxidation data.

Full text of DOI:10.1039/b110702j

View Article Online / Journal Homepage / Table of Contents for this issue
High pressure, high temperature shock tube studies of ethane
pyrolysis and oxidation
Robert S. Tranter,a Raghu Sivaramakrishnan,a Kenneth Brezinsky*a and Mark D. Allendorfb
a
Department of Chemical Engineering, University of Illinois at Chicago, 810 South Clinton St.,
Chicago, IL 60607, USA. E-mail: kenbrez@uic.edu; Fax: 312 996 0808; Tel: 312 996 9430
Combustion Research Facility, Sandia National Laboratories, Livermore, CA 94551, USA.
E-mail: mdallen@sandia.gov; Fax: 925 294 2276; Tel: 925 294 2895
b
Received 1st November 2001, Accepted 31st January 2002
First published as an Advance Article on the web 12th March 2002
A unique high-pressure single pulse shock tube has been used to obtain the first experimental data for ethane
oxidation and pyrolysis at very high pressures and temperatures. Experiments were performed at two nominal
reaction pressures of 340 bar and 613 bar in the temperature range 1050 K to 1450 K. The major stable species
were identified and their concentrations determined using gas chromatography. Several minor species, with up
to four carbon atoms and including oxygenates, were also observed in the oxidation studies. Three models
based on literature mechanisms for hydrocarbon oxidation were used to simulate the experimental data. All of
the models simulate the pyrolysis data well although only one of the models was capable of accurately
describing the oxidation data.
Department of Chemical Engineering at UIC.13 This shock
Introduction
tube has been designed to operate at pressures up to 1000
bar and in the temperature range 1000–2000 K. By using dilute
reaction mixtures at high reaction pressures bimolecular reac-
tions occur on time scales that are compatible with a shock
tube experiment, but without destroying the isothermal reac-
tion environment that is one of this shock tube’s main advan-
tages in combustion studies. The work presented here
represents the first experimental data obtained for ethane pyr-
olysis and ethane oxidation at nominal reaction pressures, P5 ,
of 340 bar and 613 bar and temperatures from 1050 to 1450 K.
In addition to attaining high-pressure and high-temperature
experimental data one of the goals of the high-pressure shock
tube group is to construct reaction mechanisms that can be
used to accurately simulate combustion and other industrial
processes over the wide range of pressures that are found in
practical devices. Such systems include catalytic processes used
to produce ethene from ethane based on short-contact-time
reactors, in which gas-phase reactions coupled with surface
chemistry are thought to play a critical role.14 To initiate the
construction of high-pressure hydrocarbon oxidation models
the results of the experiments were subsequently simulated
using three independently developed mechanisms for hydro-
carbon oxidation. These mechanisms have been validated
and optimised for much lower pressures than the reaction pres-
sures utilised in the experimental work presented here thereby
providing a good test of how well the unadulterated model can
simulate the experimental data at pressures much greater than
the mechanisms design conditions.
Ethane is not only a major constituent of some natural gases
but it is also formed during the oxidation of many hydrocar-
bons and it is necessary to include a comprehensive ethane
submechanism in a description of the combustion chemistry
of these species. Thus, ethane oxidation and pyrolysis have
received considerable attention from both experimentalists
and modellers in order to understand the combustion chemis-
try of C2H6 , e.g.1–12 In spite of all the theoretical and experi-
mental data available for the combustion of simple
hydrocarbons it is clear that these processes are still not well
understood, as described by Hughes et al.12 in a recent publica-
tion.
A wide variety of experimental techniques have been used to
investigate ethane oxidation, ranging from the measurement of
species profiles to the determination of bulk properties such as
laminar flame velocity and ignition delay times. These experi-
ments cover the temperature range from approximately 600
K up to around 2000 K i.e. slow combustion to ignition, and
all of the experiments have been performed at reaction pres-
sures below 10 bar. The temperature range is certainly compar-
able to temperatures experienced in practical combustion
devices; however, the pressure range is considerably below
the operating pressures of devices such as gas turbines. Thus,
although the models have been tested against a wide range
of experimental results and generally do a good job of predict-
ing the data, it has not been possible to validate existing com-
bustion models for use at high reaction pressures. Pressure can
have a large effect on the rate of a reaction due to fall-off effects
and it can also affect the branching ratio in a multi-channel
reaction. Consequently, if the pressure dependencies are not
accurately described in a reaction model then using the model
to make predictions of the reaction behaviour at significantly
higher pressures than those for which the model has been opti-
mised may lead to erroneous predictions.
Experimental
The high-pressure shock tube has been described in detail in a
previous publication13 and therefore only a brief description of
the apparatus will be given here. The shock tube consists of a
driver section (200 i.d.) that is separated from the driven section
(100 i.d.) by a diaphragm section. The diaphragm section has
been constructed to allow both the single diaphragm and dou-
To address the lack of experimental data at high pressures
and high temperatures, a high-pressure single pulse shock tube
has recently been constructed, tested and calibrated in the
DOI: 10.1039/b110702j
Phys. Chem. Chem. Phys., 2002, 4, 2001–2010
This journal is # The Owner Societies 2002
2001

View Article Online
ble diaphragm techniques to be used, although only the single
diaphragm mode was used in these experiments. The high-
pressure shock tube is operated as a single pulse shock tube
and therefore a large dump tank is situated close to the dia-
phragm section on the driven section side of the shock tube
to quench the reflected shock wave and prevent multiple heat-
ing events in the reaction zone.
In this apparatus the type of diaphragm separating the dri-
ver and driven sections effectively controls the pressure that is
generated behind the reflected shock wave assuming that the
driver and driven gases are the same from experiment to
experiment e.g. helium and argon. To generate 340 bar in
the reaction zone a soft brass diaphragm that is 0.03200 thick
and has two mutually perpendicular scores 0.01000 deep cen-
tered on the diaphragm is used. The diaphragm type is abbre-
viated to ‘‘soft brass 32/10’’. To attain 613 bar in the reaction
zone a soft brass 50/16 diaphragm is used.
The shock tube is fired by increasing the pressure in the dri-
ver section until the diaphragm breaks. Soft brass 32/10 dia-
phragms burst at approximately 205 bar and the soft brass
50/16 diaphragms burst at around 340 bar. To ensure that suf-
ficient driver gas is available to burst the diaphragms 10 high-
pressure tanks have been constructed (ca. 4 l each) to store the
driver gas at pressures up to 890 bar. These storage tanks are
filled by a gas booster that has a regular gas cylinder attached
to its inlet.
Oxidation mixtures were prepared in a similar way to pyro-
lysis mixtures with the addition of sufficient oxygen (AGA,
99.999%) to the mixing vessel to make a stoichiometric mix-
ture.
After the mixtures had been prepared they were allowed to
stand overnight before use and the actual species concentra-
tions were determined by withdrawing a sample from the mix-
ing vessel and analysing it with GC techniques.
Sampling and analytical methods
The very high pressures that are generated in this shock tube
are generally incompatible with real-time diagnostic techni-
ques. Consequently, samples of the gases in the shock tube
are withdrawn through a port centred in the endwall of the dri-
ven section by expansion into evacuated sample vessels. These
samples were analysed off-line to determine the nature and
concentration of the stable species. The sampling process could
cause heavy species to condense out of the gas phase and accu-
mulate on the walls of the sample vessels altering the composi-
tion of the gas phase mixture. This deliterious affect could be
ameliorated by heating the sample vessels and working with
sufficently dilute reaction mixtures to prevent condensation
of the reagents and products. In the current work the mixtures
are certainly dilute enough to prevent condensation of the
reagents and products and a carbon balance has been per-
formed for each experiment, discussed later.
For every experiment, a sample of the reagent mixture was
withdrawn from the shock tube prior to firing the shock tube,
thereby accounting for small variations in concentration due to
absorption of the reagents on the walls of the mixing vessel,
transfer lines and shock tube. A second sample was taken from
the shock tube immediately after it was fired but before the dri-
ver and driven gases had time to mix, verified by measuring the
xenon concentration in the pre-shock and post-shock samples.
The pre-shock sample was collected in a 50-cc electropolished
stainless steel vessel and the post-shock sample was collected in
a similar vessel but of 150-cc capacity. Care was taken to
ensure that the post-shock sample only contained gases from
the reaction zone. The necessary sampling conditions had been
determined experimentally in prior work with similar reaction
conditions.15
Quantitative analysis of the samples was performed using a
two-column GC method. Sample loops attached to two 10-
port gas sampling valves mounted in the same heated valve
box (50 ꢀC) were filled simultaneously from the same sample
vessel. When the GC run was started both valves switched at
the same time to the inject position. The effluent from one
valve was fed to a PLOT Q column (Agilent 30 m ꢁ 0.32
mm ꢁ 20 mm) that was attached to a FID. The second valve
was attached to a MolSieve 5 column (Agilent, 30 m ꢁ 0.25
mm ꢁ 12 mm) and this column eluted to a thermo-conduction
detector (TCD). The TCD was used to monitor the concentra-
tions of xenon and CO, all other species were monitored using
the FID. Both detectors were calibrated in parallel and if
changes were made to any part of the analytical system then
the calibrations were checked and if necessary the detectors
were recalibrated. Species identities were confirmed by match-
ing retention times with known pure samples of the analytes
and by GC-MS.
Reaction temperature, pressure and time
At the reaction pressures attained in the high-pressure shock
tube the gases exhibit non-ideal behaviour. Consequently,
the standard shock tube equations cannot be used to obtain
T5 , the temperature behind the reflected shock wave. Chemical
thermometers have been used to measure T5 in the shock tube
for a wide range of experimental conditions. These measured
T5s were correlated with the incident shock wave velocity to
produce a temperature calibration chart that can be used to
determine T5 for subsequent experiments. Details of the proce-
dure used to calibrate the reaction temperature at P5 ¼ 613
bar (9000 psi) and P5 ¼ 340 bar (5000 psi) can be found in a
recent publication.15
The velocity of the shock wave is determined for each
experiment by using high-speed pressure transducers mounted
in the side wall of the shock tube near the end of the driven
section. The separation between successive transducers is accu-
rately known and by measuring the time taken for the shock
wave to pass between transducers the velocity of the shock
wave can be calculated. The calculated velocity is then used
to determine the temperature behind the reflected shock wave
from the calibration chart. One additional presure transducer
is mounted in the end of the driven section and is used to moni-
tor the reaction pressure at the endwall. From this pressure
trace the reaction time is determined by the method described
by Hidaka et al.16 The velocities of the shock waves and the
rarefaction waves in the shock tube vary from experiment to
experiment resulting in a range of reaction times. By determin-
ing the reaction time for each experiment the experimental
data can be simulated using the exact reaction time rather than
a nominal or average reaction time.
Reagents
Reagent mixtures have been prepared manometrically in high-
pressure mixing vessels. For the pyrolysis experiments mix-
tures were prepared containing approximately 200 ppm ethane
(Matheson, 99.999%), 200 ppm xenon (AGA, 99.99%) with the
balance argon (AGA 99.999%). The argon was passed over an
Oxisorb trap (Messer Griesheim) prior to admission to the
mixing vessel to remove any traces of oxygen. Xenon acts as
an internal standard when the samples are analysed.
Results
Ethane pyrolysis
Approximately twenty experiments spanning the temperature
range 1025–1400 K have been performed for nominal P5s of
340 bar and 613 bar. A typical pressure profile from the pyro-
lysis experiments is shown in Fig. 1.
2002
Phys. Chem. Chem. Phys., 2002, 4, 2001–2010

View Article Online
Fig. 4 Carbon balances for 613 bar and 340 bar ethane pyrolysis
experiments. Closed triangles, 340 bar pre-shock; cross, 340 bar
post-shock; closed square, 613 bar pre-shock; open diamond 613 bar
post-shock. 613 bar values divided by 2.
Fig. 1 Pressure profile at the driven section endwall from a 340 bar
ethane pyrolysis experiment.
pressures the carbon balance is good up to T5 ¼ 1300 K indi-
cating that all of the species have been accounted for. Above
T5 ¼ 1300 K there is a small deficiency in the post-shock car-
bon total which suggests that at the higher temperatures either
not all of the species are being recovered from the shock tube
or that they are not being detected in the GC analytical work.
The analytical results indicate that at higher temperatures
methane is being formed in small quantities. However, the
peak shape for methane on the chromatogram was quite poor
leading to inaccuracies in determining the methane concentra-
tion. Hence, methane has not been included in the calculation
of the carbon totals and may be the source of the carbon total
deficiency at reaction temperatures greater than 1300 K.
Ethane oxidation
Fig. 2 Ethane pyrolysis, 340 bar. Closed squares, C2H6 ; open
squares, C2H4 ; open triangles, C2H2 .
A pressure profile from the oxidation experiments is shown in
Fig. 5. A similar number of experiments compared to the pyr-
olysis work have been performed at 340 bar and 613 bar using
stoichiometric mixtures of ethane and oxygen, Figs. 6 and 7.
In addition to the same species that are observed in the pyr-
olysis experiments, CO was also detected as a major product
and a number of new minor species were formed at higher
reaction temperatures. Several of the minor species have been
identified as acetaldehyde, ethylene oxide, propane, propene
and 1,3-butadiene. CO2 was not observed in any of the experi-
ments. This is most likely due to retention of CO2 on the Mol-
Seive column conncected to the TCD, which was used to
determine concentrations of the permeant gases, and not
because CO2 was not present in the post-shock sample. In
future work the analytical methodology could be improved
by including a methaniser to convert CO and CO2 to CH4 after
they elute from the Plot-Q column that is used for separating
the hydrocarbons prior to detection with an FID.
Carbon balances for each of the experiments at 340 bar and
613 bar are shown in Fig. 8. For the 340 bar experiments it is
clear that as T5 is increased the amount of carbon recovered in
the post shock sample decreases until at 1450 K only around
70% of the total carbon has been accounted for in the analysis
of the post shock sample. A similar trend is observed for the
613 bar experiments, although here the trend is a little less
clear since two different reaction mixtures with slightly differ-
ent initial concentrations of ethane (203 ppm and 180 ppm)
were used. All of the 340 bar experiments were performed with
one reagent mixture.
Fig. 3 Ethane pyrolysis, 613 bar. Closed squares, C2H6 , open
squares, C2H4 , open triangles, C2H2 .
Figs. 2 and 3 show the loss of the parent molecule, ethane,
and the formation of the major product species, ethene and
ethyne, with each set of three species points representing an
individual experiment. Below approximately 1125 K no appre-
ciable reaction was observed. Successive experiments were per-
formed up to T5 ꢂ 1400 K where around 90% of the ethane
had been consumed.
A possible explanation for the poor carbon balance is that
heavy species are being formed during the reaction and that
these species are condensing inside the shock tube or the sam-
For each experiment the carbon totals for the pre-shock and
post-shock samples were calculated, Fig. 4. At both reaction
Phys. Chem. Chem. Phys., 2002, 4, 2001–2010
2003

View Article Online
formed apart from soot would be sufficient to cause condensa-
tion.
When the carbon balance begins to deteriorate, T5 > 1150 K,
the minor peaks corresponding to C3 and C4 species first
appear in the analytical results. These peaks increase in size
as the temperatures of the experiments increase, indicating that
at higher reaction temperatures more of the C3 and C4 species
are formed. Due to poor baseline separation between the peaks
corresponding to these species they have not been quantified.
However, even generous estimates of their concentrations indi-
cate that each species is present in amounts <3 ppm, which are
insufficient to explain the carbon deficit. A possible explana-
tion for the carbon deficit based on CO2 will be proposed later
in this article.
Modelling
Fig. 5 Pressure profile at the endwall from a 613 bar ethane oxida-
tion experiment.
To the best of the authors’ knowledge the experimental results
presented above represent the first very high-pressure and
high-temperature data obtained for ethane pyrolysis and
ethane oxidation. The pressures covered by these experiments
are far greater than those encountered in practical combustion
devices and extend the range of experimental data which may
be used to test how well existing models for hydrocarbon com-
bustion can predict the oxidation and pyrolysis of ethane at
elevated temperatures and pressures. To simulate the experi-
mental data three mechanisms, described below, have been
selected from the literature.
It should be noted that it is not the authors’ intent to criti-
cize the models but to use them to understand where further
information is needed to develop a model suitable for accu-
rately predicting small hydrocarbon combustion over extended
pressure ranges. It should also be noted that the models used in
this work may have been optimised for specific problems in
combustion science and were never intended by their authors
for use at the extreme conditions represented by this experi-
mental work. However, these existing models do incorporate
the current understanding of small hydrocarbon oxidation,
including pressure dependent reaction rates, and thus represent
a good starting point for developing high-pressure models.
Fig. 6 Ethane oxidation f ¼ 1, 340 bar. Closed squares, C2H6 ; open
squares, C2H4 ; open triangles, C2H2 ; cross, CO.
Modelling approach
To simulate the shock-tube experiment the SENKIN code
from the Chemkin-III suite of programs17 has been used. Sen-
sitivity analysis was performed using the method employed by
SENKIN. The results of the analysis were normalized using
the SENKIN postprocessor ‘‘psenk,’’ which is described in
the SENKIN operating manual.17 Input concentrations,
post-shock temperatures and pressures, and reaction times
were all obtained from the experiments.
Three gas-phase models (with their associated thermochem-
istry) that have been optimized for use at pressures below 10
bar and perform well at those conditions were tested against
the experimental data presented here: GRI-Mech 3.0;18
a
mechanism for modeling aromatic hydrocarbon formation
in hydrocarbon flames developed by Marinov and co-
workers19–21 (Marinov98); and
a modified version of
Fig. 7 Ethane oxidation f ¼ 1, 613 bar. Closed squares, C2H6 ; open
the ‘‘M1A’’ mechanism of Pope and Miller21 which will be
referred to here as Miller 2001. The M1A mechanism was
developed to predict the formation of benzene in low-pressure
premixed flames of aliphatic fuels. GRI-Mech 3.0 was primar-
ily developed to simulate lean natural-gas flames and extends
only to C3H8 . Its performance for fuel-rich mixtures has not
been tested extensively. Marinov98 is a large mechanism con-
taining over 600 reactions and includes aromatic species as
large as C18 . Ref. 21 describes comparisons between Mari-
nov98 and data obtained in a sooting, atmospheric-pressure,
premixed butane–oxygen flame operating at an equivalence
squares, C2H4 ; open triangles, C2H2 ; cross, CO.
ple vessels. To test this hypothesis the shock tube was dis-
mantled between experiments for a few runs and the internal
walls in the reaction zone close to the end of the driven section
were cleaned. During the cleaning no soot or deposits other
than some fine brass particles from the diaphragms were
found. Of course, the initial concentration of ethane is so
low that it is unlikely that the partial pressure of any species
2004
Phys. Chem. Chem. Phys., 2002, 4, 2001–2010

View Article Online
Fig. 8 Carbon balances for 613 bar and 340 bar ethane oxidation,
f ¼ 1, experiments. Closed triangles, 340 bar pre-shock; cross, 340
bar post shock; closed square, 613 bar pre-shock; open triangles, 613
bar post-shock. 613 bar carbon totals divided by 2.
ratio of 2.6. Finally, Miller 2001 contains modified rate con-
stants for the oxidation of C2H5 as described in ref. 23 and
shown in Table 1. Miller 2001 also uses a different heat of for-
mation for C4H than the earlier version22 and an altered rate
constant for the reaction C4H + H2 $ C4H2 + H, neither of
which significantly affects the results of this investigation. A
copy of all three models may be obtained from the authors
and details of the optimisation procedure used to generate
Gri-Mech 3.0 can be found on the Gri-Mech 3.0 website.18
A possible complication that may need to be accounted for
in the modelling work is the finite time taken to quench the
reaction mixture in the shock tube. When the hot reaction mix-
ture is rapidly quenched in the shock tube prior to sampling
the post shock gases it takes approximately 1–1.5 ms for the
temperature to be reduced to a few hundred kelvin. During
this period, reactions can still occur and the composition of
the mixture may change. However, the greatest change in the
temperature of the reaction mixture occurs during the time
that the reaction pressure drops to 80% of its maximum value.
This change happens very rapidly and its effect was found to be
minimal during the temperature calibration studies.15 It is pos-
sible to model the effect of the complete cooling curve on the
reaction mixture composition. However, on the basis of the
earlier work in this apparatus it does not appear necessary to
explicitly account for reactions that may occur during the
cooling period.
Fig. 9 Comparison of experimental and predicted [C2H6] concentra-
tions for ethane pyrolysis: (a) 340 bar; (b) 613 bar. Closed triangle
expt, open triangle Gri-Mech 3.0, open square Miller 2001, cross Mar-
inov98 (not shown for 613 bar because the data are indistinguishable
from Gri-Mech 3.0).
Modelling results
The above mechanisms were used to simulate each of the pyr-
olysis and oxidation experiments that have been performed.
Ethane pyrolysis
Fig. 9 indicates that Gri-Mech 3.0, Marinov98 and Miller 2001
do a good job of predicting the consumption of ethane during
the pyrolysis experiments. Although all three models tend to
overpredict the consumption at the higher reaction tempera-
tures attained in this work.
Table 1 Rate constants for C2H5 oxidation in the form k ¼ ATbexp(ꢃEa/RT)
Ab
b
Ea
c
Reaction
C2H5 + O2(+M) $ C2H5O2(+M)
2.02E10
8.49 ꢁ 1029
0.98
ꢃ4.29
ꢃ63.6
Low-pressure rate
220
Falloff parametersa 0.103; 601; 1.0 ꢁ 10ꢃ10
Enhanced third-body collision efficiencies: H2/2/CO/2/CO2/3/H2O/5/
C2H5 + O2(+M) $ C2H4 + HO2(+M)
High-pressure rate
1.41 ꢁ 107
1.09
6.53
ꢃ1975
ꢃ834
6.85 ꢁ 10ꢃ12
Falloff parametersa 0.45; 1.00 ꢁ 10ꢃ10; 1.00 ꢁ 1010
Enhanced third-body collision efficiencies: H2/2/CO/2/CO2/3/H2O/5/
C2H5O2(+M) $ C2H4 + HO2(+M)
7.14 ꢁ 104
8.31 ꢁ 1021
2.32
ꢃ0.651
27 955
22 890
Low-pressure rate
Falloff parametersa 0.0; 106; 106
Enhanced third-body collision efficiencies: H2/2/CO/2/CO2/3/H2O/5/
C2H5 + O2 $ C2H4 + HO2
5.81 ꢁ 106
1.57
20 578
a
b
c
Troe format used by CHEMKIN-III (1). Units of cmꢃ3 molꢃ1 sꢃ1
.
cal molꢃ1
.
Phys. Chem. Chem. Phys., 2002, 4, 2001–2010
2005

View Article Online
At 613 bar, Miller 2001 simulates the C2H4 concentration,
Fig. 10, better than Gri-Mech 3.0 does in the intermediate tem-
perature region 1100–1250 K. However, at higher tempera-
tures Miller 2001 predicts a decreasing C2H4 concentration
whereas Gri-Mech 3.0 continues to predict an increase in
C2H4 and at around 1400 K predicts the same maximum
C2H4 concentration as is found experimentally. The under pre-
diction of C2H4 by Gri-Mech 3.0 in the 1100–1250 K region is
consistent with the under prediction of the consumption of the
parent species in this region. The Marinov98 mechanism pro-
duces very similar results to Gri-Mech 3.0.
concentrations with the experimental post-shock carbon total
indicates that the high temperature carbon totals could be
balanced if the methane concentration can be measured accu-
rately experimentally, Fig. 12.
Ethane oxidation
The Miller 2001 model makes a very good prediction of the
ethane consumption, Fig. 13, at 613 bar. At 340 bar the Miller
2001 model does not show significant ethane consumption
until around 1130 K. Experimentally, significant consumption
The final stable species observed experimentally was ethyne,
C2H2 (Fig. 11), which begins to appear at temperatures above
ca. 1225 K. The C2H2 concentrations calculated by Gri-Mech
3.0 and Marinov98 are in good agreement with the experimen-
tal data whereas Miller 2001 greatly overpredicts the ethyne
concentration when it is formed in experimentally significant
amounts.
With the 340 bar pyrolysis experiments the general predic-
tions made by all three models are similar to those made for
the 613 bar experiments. In general, Gri-Mech 3.0 and Mari-
nov98 are somewhat closer to the experimental data e.g. Fig.
10 and Miller 2001 once again greatly overpredicts the ethyne
concentration at the higher reaction temperatures.
Both Miller 2001 and Gri-Mech 3.0 predict that above 1200
K methane should be formed in fairly significant quantities in
both the 340 bar and 613 bar experiments. It was noted earlier
that the methane concentration was not calculated from the
experimental data due to a poor peak shape for methane; how-
ever, the analytical data indicate that methane starts to be
formed around 1200 K, in agreement with the model predic-
tions. Summing the carbon content of the predicted methane
Fig. 11 Comparison of experimental and predicted [C2H2] concen-
trations for ethane pyrolysis: (a) 340 bar; (b) 613 bar. Closed triangle,
expt; open triangle, Gri-Mech 3.0; open square, Miller 2001; cross,
Marinov98 (not shown for 613 bar because the data are indistinguish-
able from Gri-Mech 3.0).
Fig. 10 Comparison of experimental and predicted [C2H4] concen-
trations for ethane pyrolysis: (a) 340 bar; (b) 613 bar. Closed triangle,
expt; open triangle, Gri-Mech 3.0; open square, Miller 2001; cross,
Marinov98 (not shown for 613 bar because the data are indistinguish-
able from Gri-Mech 3.0).
Fig. 12 Carbon totals including [CH4], from simulations, for pyroly-
sis experiments. Closed triangles, 340 bar preshock; cross, 340 bar post
shock; closed square, 613 bar pre-shock; open diamond, 613 bar post-
shock. 613 bar carbon totals divided by 2.
2006
Phys. Chem. Chem. Phys., 2002, 4, 2001–2010

View Article Online
Fig. 14 Comparison of experimetal and predicted [C2H4] concentra-
tions for ethane oxidation: (a) 340 bar; (b) 613 bar. Closed triangle,
expt; open triangle, Gri-Mech 3.0; open square, Miller 2001; cross,
Marinov98.
Fig. 13 Comparison of experimetal and predicted [C2H6] concentra-
tions for ethane oxidation: (a) 340 bar; (b) 613 bar. Closed triangle,
expt; open triangle, Gri-Mech 3.0; open square, Miller 2001; cross,
Marinov98.
of ethane occurs at around 1100 K . However the deviation
between the model and the experimental data is never more
than about 20%.
Mech 3.0, consistent with the C2H6 profile. Miller 2001 over-
predicts the maximum C2H2 concentration by about a factor
of 3.5, although the temperature at which the peak concentra-
tion appears matches the experimental data and for T5 < 1250
K Miller 2001 and the experimental data are in good agree-
ment. The C2H2 predictions made by Marinov98 (Fig. 15)
resemble a combination of the low temperature portion of
the Gri-Mech 3.0 results and the high temperature Miller
2001 results.
For CO, Miller 2001 predicts the experimental data well
(Fig. 16) and Gri-Mech predicts that CO is formed in a very
narrow temperature range from 1200–1250 K consistent with
C2H4 . Marinov98 consistently underpredicts CO concentra-
tions by 20–80%, except at T5 ¼ 1450 K where all three models
are in agreement with the experimental data.
As mentioned previously, the experimentally observed car-
bon balance for both sets of ethane oxidation experiments
while good at low T5’s is rather poor at high temperatures.
The relatively large carbon deficiency in the post-shock sam-
ples indicates that at least one species is not being observed
during the analysis of the samples. A possible candidate for
a missing species is CO2 that was not detected but would be
expected to be formed in an oxidation experiment. Both Miller
2001 and Gri-Mech 3.0 predict that significant amounts of
CO2 are formed at the higher temperatures attained in this
work. The Miller 2001 model generally predicts the concentra-
tions of the major species that are formed in ethane oxidation
better than Gri-Mech 3. Consequently, to test the hypothesis
that CO2 could improve the carbon balance, the predicted
CO2 concentrations from Miller 2001 have been added to the
experimental post-shock carbon total for the 613 bar experi-
ments (Fig. 17). Clearly, the predicted CO2 concentrations
greatly improve the carbon balance for the oxidation reactions
In contrast to the Miller 2001 data Gri-Mech 3.0 and Mar-
inov98, Fig. 13, predict that ethane will only begin to be con-
sumed at approximately 1170 K in both the 613 bar and 340
bar experiments. In fact, the experiments show that at 1170
K over half of the initial ethane present in the reaction mixture
has been consumed. Both of the models, in good agreement
with Miller 2001 and the experimental data, predict that all
of the ethane will have been consumed by 1275 K.
Miller 2001 simulates the ethene formation well, Fig. 14,
particularly at 613 bar although at this pressure the model
overpredicts the peak ethene concentration by around 25%.
The location of the ethene maximum is at around 1225 K,
which is in good agreement with the experimental data. For
the 340 bar experiments the maximum ethene concentration
is again overpredicted by about 25% by Miller 2001 and the
location of the maximum is about 25 K higher than the experi-
mentally observed maximum.
At both 613 and 340 bar Gri-Mech 3.0 predicts a very slow
build up of ethene followed by a rapid rise and decay (Fig. 14),
thus indicating that all of the ethene is being formed and con-
sumed within a very narrow temperature range, consistent with
Gri-Mech 3.0’s prediction of the ethane consumption. A
slower consumption of C2H4 than that predicted by Gri-Mech
3.0 is predicted by Marinov98 for T5 > 1250 K. However, for
T5 < 1250 K Marinov98 and Gri-Mech 3.0 predict similar
C2H4 concentrations.
Gri-Mech 3.0 does a very good job of predicting the C2H2
concentrations found in the oxidation experiments for both
of the pressure ranges (Fig. 15) for T5 > 1250 K. However, at
lower temperatures C2H2 is greatly underpredicted by Gri-
Phys. Chem. Chem. Phys., 2002, 4, 2001–2010
2007

View Article Online
Fig. 15 Comparison of experimetal and predicted [C2H2] concentra-
tions for ethane oxidation: (a) 340 bar; (b) 613 bar. Closed triangle,
expt; open triangle, Gri-Mech 3.0; open square, Miller 2001; cross,
Marinov98.
Fig. 16 Comparison of experimetal and predicted [CO] concentra-
tions for ethane oxidation: (a) 340 bar; (b) 613 bar. Closed triangle,
expt; open triangle, Gri-Mech 3.0; open square, Miller 2001; cross,
Marinov98.
indicating that if the analytical technique is improved to mea-
sure CO2 , possibly with a methaniser, then a good carbon bal-
ance can be obtained.
the worst prediction of [C2H2] of the three models and matches
the low temperature predictions of GRI-Mech 3.0 as well as
the high-temperature predictions of Miller2001.
To provide qualitative insights into the key rate-limiting
pathways for the production of ethyne under oxidizing condi-
tions we performed sensitivity analyses for all three models and
examined rates of production for C2H2 . In each case, the ana-
lysis was conducted for an experiment that corresponds to the
maximum C2H2 concentration observed in the 613 bar oxida-
tion experiments (T5 ¼ 1325 K).
Isothermal reaction conditions
In addition to the species concentrations the modelling soft-
ware has been used to calculate the reaction temperature at
the time corresponding to the mixture being quenched in the
shock tube. The calculated temperatures were compared with
the measured reaction temperatures and only at the highest
temperatures used in this study does the difference in tempera-
ture rise above a few kelvins. In the worst case the temperature
increase is only 10 K indicating that the reactions are carried
out at nearly isothermal conditions.
The results indicate that two reactions common to all three
models exhibit some of the largest sensitivities: first, the pres-
sure dependent reaction
H þ C2H2þðMÞ ¼ C2H3þðMÞ
ð1Þ
Fig. 18 shows the results from simulating [C2H6] for the 613
bar ethane oxidation experiments with a systematic increase of
10 K in the reaction temperature. Clearly, there is very little
difference between the modeling results with and without the
10 K increase in T5 .
and second, the formation of C2H3 by reaction of OH with
ethene:
OH þ C2H4¼ H2O þ C2H3
ð2Þ
Reaction-path analysis indicates that reaction (1) is the pri-
mary route for C2H2 production under stoichiometric oxida-
tion conditions in all three models, while reaction (2) is a
major pathway for the production of C2H3 . For both of these
reactions there appears to be quite some uncertainty in the lit-
erature rate coefficients, suggesting that closer examination of
the rate parameters may be necessary.
The ethane concentration is also quite sensitive to the con-
sumption of C2H3 radicals by reaction with O2 , although the
reaction with the greatest sensitivity differs among the three
models. GRI-Mech 3.0 and Marinov98 each show the greatest
sensitivity to C2H3 + O2 $ O + CH2CHO, while Miller 2001 is
most sensitive to C2H3 + O2 $ HCO + CH2O. Gri-Mech 3.0
also shows a reasonably large sensitivity to the pressure-depen-
Sensitivity analysis
It is the ethyne concentrations that show the largest discrepan-
cies between the models and the experimental data. For the
pyrolysis experiments [C2H2] is overpredicted by all three mod-
els, although Gri-Mech 3.0 and Marinov98 are closer to the
experimental data than Miller 2001 and they are in agreement
with each other. However, for the oxidation experiments the
situation is rather different. Miller 2001 matches the experi-
ments well for T5 < 1250 K, while GRI-Mech 3.0 predicts
[C2H2] very well for T5 > 1250 K. For T5 > 1250 K, Miller
2001 overpredicts [C2H2] by a factor of 3.5. Marinov98 gives
2008
Phys. Chem. Chem. Phys., 2002, 4, 2001–2010

View Article Online
ethane oxidation experiments, although the deficit was much
larger than that found in the pyrolysis experiments.
The experimental data were simulated using three models
for hydrocarbon combustion from the literature without
adjusting the rate constants to improve the fit to experimental
data. The unadjusted models do a surprisingly good job of pre-
dicting the high-pressure experimental data, given that none of
the models were designed to simulate experiments at the pres-
sures used in this work.
All of the models did a good job of predicting the stable spe-
cies concentrations observed in the ethane pyrolysis experi-
ments, although Gri-Mech 3.0 and Marinov98, which give
very similar results, do appear to do a slightly better job than
Miller 2001, especially for C2H2 .
For the oxidation experiments Miller 2001 is superior to
both Gri-Mech 3.0 and Marinov98, and accurately simulates
the 613 bar data with the exception of C2H2 , which it overpre-
dicts by about a factor of 3.5. Gri-Mech 3.0 and Marinov98
predict a significantly different consumption pattern for ethane
than that observed experimentally and this is reflected in the
prediction of other species. However, Gri-Mech 3.0 does pre-
dict the observed C2H2 profiles very well at temperatures
above 1250 K whereas Marinov 98 imitates the worst predic-
tions of C2H2 from both Gri-Mech 3.0 and Miller 2001.
The simulations of the pyrolysis experiments tend to suggest
that all three models use similar reaction schemes to describe
the pyrolysis of ethane, although some differences must exist
between Miller2001 and the other two models. Further differ-
ences must also exist between Miller2001 and Gri-Mech 3.0
and Marinov98 in the C2H2 chemistry, as shown by the greater
overprediction of C2H2 by Miller 2001. A similar pattern is
observed in ethane oxidation. However, here Miller 2001
makes the most accurate predictions for all species apart from
ethyne. Gri-Mech 3.0 and Marinov98 makes fairly similar pre-
dictions for [C2H6] but the profiles for [C2H2] are very different
for the two mechanisms. The different performances of Mari-
nov98 and Miller 2001 are somewhat surprising given that they
share many of the same reactions and rate coefficients.
A sensitivity analysis for [C2H2] predictions in the oxidation
experiments has highlighted several elementary steps that may
require further study to improve the prediction of [C2H2].
This work forms the first part of an ongoing series of inves-
tigations into ethane oxidation and pyrolysis over a wide range
of temperature, pressure and mixture stoichiometries.
Fig. 17 Carbon totals including [CO2], from simulations, for oxida-
tion experiments. Closed square, 613 bar pre-shock; open triangle,
613 bar post-shock.
Fig. 18 The effect of a 10 K increases in T5 on the predicted ethane
concentrations from simulations with Gri-Mech 3.0 for stochiometric
mixtures and P5 nominally 613 bar. Closed squares, actual T5 . Open
trangles T5 + 10 K.
dent recombination of CH3 radicals to form C2H6 . Neither
Miller 2001 nor Marinov98 show any appreciable sensitivity
to this reaction.
This preliminary sensitivity analysis suggests avenues for
improving the performance of all three models with respect
to the C2H2 concentration. However, we emphasize that there
are significant differences among the three models, as revealed
by the sensitivity analysis, and a careful examination of all
three will be necessary to determine where the deficiencies lie.
Acknowledgement
The authors would like to thank Dr James E. Miller of Sandia
National Laboratories and Dr N. M. Marinov (currently of
Corning, Inc.) for providing electronic versions of their
mechanisms. Additionally, we would like to thank Mr H.
Ramamoorthy and Mr A.. Raman for their assistance with
the experiments.
The authors are grateful for the financial support of this
work provided by the US Department of Energy Office of
Industrial Technologies Chemical Industry of the Future
Team; the US Department of Energy, Basic Energy Sciences,
Chemical Sciences Division, Grant # DE-FG0298ER14897
and the National Science Foundation, Combustion and
Plasma Systems Division, Grant # CTS-9974308.
Conclusions
Ethane oxidation and ethane pyrolysis have been investigated
experimentally for the first time at high-pressures (340 bar and
613 bar) and high temperatures (1100 K–1450 K) using a high-
pressure, single pulse shock tube. The major stable species
detected during the reaction are C2H2 , C2H4 , C2H6 and CO,
and their nature and concentrations were determined by gas
chromatographic techniques. For the pyrolysis experiments
the carbon balance indicates a small deficit at high tempera-
tures. Similarly, a carbon deficit was also observed in the
References
1
A. Burcat, R. W. Crossley and K. Scheller, Combust. Flame, 1972,
18, 115.
2
3
D. F. Cooke and A. Williams, Combust. Flame, 1975, 24, 245.
D. Noetzold and J. Algermissen, Combust. Flame, 1981, 40, 315.
Phys. Chem. Chem. Phys., 2002, 4, 2001–2010
2009

View Article Online
4
R. R. Baldwin and R. W. Walker, 18th Symposium (International)
on Combustion, The Combustion Institute, Pittsburgh, PA, 1981,
p. 819.
K. Takahashi, T. Inomata, T. Moriwaki and S. Okazaki, Bull.
Chem. Soc. Jpn., 1989, 62, 2138.
R. Renato, F. Bonini, A. Servida, M. Morbidelli and S. Carra,
Ind. Eng. Chem. Res., 1994, 33, 2540.
G. Barbieri, F. P. Maio and P. G. Lignola, Combust. Sci. Tech-
nol., 1994, 98, 95.
Y. Tan, P. Dagaut, M. Cathonnet and J.-C. Boettner, Combust.
Sci. Technol., 1994, 103, 133.
T. B. Hunter, T. A. Litzinger, H. Wang and M. Frenklach, Com-
bust. Flame, 1996, 104, 505.
15 R. S. Tranter, R. Sivaramakrishnan, N. Srinivasan and K. Bre-
zinsky, Int. J. Chem. Kinet., 2001, 33, 722.
16 Y. Hidaka, S. Shiba, H. Takuma and M. Suga, Int. J. Chem.
Kinet., 1985, 17, 441.
5
6
7
8
9
17 R. J. Kee, F. M. Rupley, J. A. Miller, M. E. Coltrin, J. F. Grcar,
E. Meeks, H. K. Moffat, A. E. Lutz, G. Dixon-Lewis, M. D.
Smooke, J. Warnatz, G. H. Evans, R. S. Larson, R. E. Mitchell,
L. R. Petzold, W. C. Reynolds, M. Caracotsios, W. E. Stewart, P.
Glarborg, C. Wang and O. Adignun, Chemkin Collection;
Release 3.6 ed.; Reaction Design, Inc., San Diego, CA, 2000.
18 Available at http://www.me.berkeley.edu/gri_mech/.
19 N. M. Marinov, W. J. Pitz, C. K. Westbrook, M. J. Castaldi and
S. M. Senkan, Combust. Sci. Technol., 1996, 116–117, 211.
20 N. M. Marinov, M. J. Castaldi, C. F. Melius and W. Tsang, Com-
bust. Sci. Technol., 1997, 128, 295.
10 P. Dagaut and M. Cathonnet, Combust. Flame, 1999, 118, 191.
11 Y. Hidaka, K. Sato, H. Hoshikawa, T. Nishimori, R. Takahashi,
H. Tanaka, K. Inami and N. Ito, Combust. Flame, 1999, 120, 245.
12 K. J. Hughes, T. Turanyi, A. R. Clague and M. J. Pilling, Int. J.
Chem. Kinet., 2001, 33, 513.
21 N. M. Marinov, W. J. Pitz, C. K. Westbrook, A. M. Vincitore, M.
J. Castaldi, S. M. Senkan and C. F. Melius, Combust. Flame,
1998, 114, 192.
13 R. S. Tranter, D. Fulle and K. Brezinsky, Rev. Sci. Instrum., 2001,
72, 3046.
14 D. K. Zerkle, M. D. Allendorf, M. Wolf and O. Deutschmann, J.
Catal., 2000, 196, 18.
22 C. J. Pope and J. A. Miller, Proc. Combust. Inst., 2000, 28, 1519.
23 J. A. Miller and S. J. Klippenstein, Proc. Combust. Inst., 2000, 28,
1479.
2010
Phys. Chem. Chem. Phys., 2002, 4, 2001–2010