Oxidation of n-Heptane in a Premixed Laminar Flame

Article abstract of DOI:10.1021/jp984191s

A premixed laminar n-heptane/air flame (1 atm and φ= 1. 0 ± 0.05) was investigated experimentally (using on-line GC/MS) and computationally. Ethene, propene, 1-butene, 1-pentene, 1-hexene, and 1-heptene showed broad peaks with maxima at a distance around 1000 μm from the burner. Methane, ethane, and propane showed more narrow peaks and maxima around 1075 μm. The following compounds with their maxima were found (mol %): ethene (0.99), methane (0.069), propene (0.127), ethyne (0.103), ethane (0.080), 1-butene (0.040), 1-pentene (0.021), and 1,3-butadiene (0.014). Concentrations of less than 0.01 mol % were detected for the components 1-heptene, propyne, propadiene, methanol, and acetaldehyde. The flame was modeled using the semiempirical mechanism by Held et al.¹ The model produced peaks of the organic intermediates at about 400-500 μm. Important measured species not present in the mechanism were propane, propyne, methanol, isobutene, 2-butene, and 1-heptene. The computed maxima of 1-butene, 1,3-butadiene, and 1-pentene were higher than the experimental maxima by a factor of about 2, 2, and 3, respectively. The model gives very similar results for two different pressures, 1 and 0.06 atm, except for a common scaling factor.

Full text of DOI:10.1021/jp984191s

8222
J. Phys. Chem. A 1999, 103, 8222-8230
Oxidation of n-Heptane in a Premixed Laminar Flame
A° sa T. Ingemarsson, J o1 rgen R. Pedersen, and Jim O. Olsson*
Department of Physical Chemistry, Chalmers UniVersity of Technology, SE-412 96 G o¨ teborg, Sweden
ReceiVed: October 26, 1998; In Final Form: August 16, 1999
A premixed laminar n-heptane/air flame (1 atm and æ ) 1. 0 ( 0.05) was investigated experimentally (using
on-line GC/MS) and computationally. Ethene, propene, 1-butene, 1-pentene, 1-hexene, and 1-heptene showed
broad peaks with maxima at a distance around 1000 µm from the burner. Methane, ethane, and propane
showed more narrow peaks and maxima around 1075 µm. The following compounds with their maxima
were found (mol %): ethene (0.99), methane (0.069), propene (0.127), ethyne (0.103), ethane (0.080), 1-butene
(0.040), 1-pentene (0.021), and 1,3-butadiene (0.014). Concentrations of less than 0.01 mol % were detected
for the components 1-heptene, propyne, propadiene, methanol, and acetaldehyde. The flame was modeled
1
using the semiempirical mechanism by Held et al. The model produced peaks of the organic intermediates
at about 400-500 µm. Important measured species not present in the mechanism were propane, propyne,
methanol, isobutene, 2-butene, and 1-heptene. The computed maxima of 1-butene, 1,3-butadiene, and 1-pentene
were higher than the experimental maxima by a factor of about 2, 2, and 3, respectively. The model gives
very similar results for two different pressures, 1 and 0.06 atm, except for a common scaling factor.
1. Introduction
building blocks. The mixing station was equipped with HI-TEC
F-201 mass flow meters/regulators connected to two four-
channel controllers (Bronkhorst, Ruurlo, The Netherlands). A
precision liquid chromatography pump, Rheos 4000 (Flux,
Karlskoga, Sweden), delivered the liquid fuel. The pump exit
was connected to a 3 m long polyetheretherketone (PEEK)
tubing (0.13 mm i.d., 1.6 mm o.d. Alltech, Deerfield, IL)
coupled to a heated line (Job-Tec, Stockholm, Sweden). The
heated line contains inner inert (Teflon) tubing (6 mm i.d.). The
temperature of the heated line was set to 180 °C. The pump
delivered 70.0 µL/min of (0.5% n-heptane, 99% pure (Kebo-
Lab, Stockholm, Sweden), and the mixing station delivered 588
Modern reformulated gasoline and high-quality diesel oil
contains 70% and 95% alkanes, respectively. n-Heptane is
suitable as a model compound in combustion investigations of
1
n-alkanes. It is also used as a primary reference fuel for octane
rating in engines and has a cetane number similar to that for
2
conventional diesel fuel. Kinetic mechanisms for n-heptane
have been developed for modeling of different combustion
conditions: flow reactors, premixed flames, diffusion flames,
1
-5
and shock tubes.
These mechanisms are often used in
computer modeling of combustors such as engines. Ignition,
knocking, and emissions can be predicted. Modeling of combus-
tion processes necessitates a very wide range of detailed
experimental studies for evaluation/development of mechanisms.
(
10 mL/min dry air with 6 ( 0.30 mL/min krypton added.
6
In a pioneering study, Hamins and Seshadri measured the
The krypton was used as an internal standard. In the heated
line the liquid fuel was mixed with the gases coming from the
mixing station. The gas flows were calibrated before each
experiment using a DryCal DC-1 flow calibrator (BIOS, Pomton
Plains, NJ). The liquid flow of n-heptane was calibrated by
collecting the liquid flow for 10 min in a flask and then weighing
the flask. The fuel/air ratio corresponded to an equivalence ratio
of 1.0 ( 0.05.
composition profiles of several species in a counterflow flame
using on-line gas chromatography. However, there exist only a
limited number of more detailed studies, with many organic
compounds measured, of the pyrolysis/oxidation of n-hep-
tane.1
,7,8
For example, only one detailed investigation of
n-heptane in a premixed laminar flame has been performed (at
8
low pressure). More detailed experiments are necessary in order
to evaluate and improve the combustion kinetic mechanisms
further.
2.2. Burner Setup. An atmospheric premixed burner, de-
signed in-house and connected directly to the heated line above,
was used. The burner consists of a stainless steel cylinder (30
mm height and 14 mm o.d.). To minimize the risk for flash-
back, two screens were placed inside the cylinder at different
heights. These screens also induce a laminar flow of the fuel/
air mixture. A thin (0.25 mm) disk (diameter 10 mm) made of
stainless steel (Millipore, Bedford, MA) is clamped on top of
the steel cylinder. The disk has a large number of holes (diameter
150 µm), evenly spaced over the surface, constituting about 40%
of the area. The disk is exchanged with a new disk at the
beginning of each experimental series. The burner produced
The goal of this study was to analyze the formation of stable
organic compounds during combustion of n-heptane. A pre-
mixed laminar flame at stoichiometric composition burning at
atmospheric pressure was investigated in detail using on-line
GC/MS. The experimental results were compared with model
predictions using a semiempirical mechanism developed by Held
et al.1
2
. Experimental Section
.1. Fuel/Air Preparation. The gases were measured and
controlled using a station constructed of Swagelock (Solon, OH)
2
*
To whom correspondence should be addressed. Fax: +46 31 772 38
5
8. E-mail: jim.olsson@phc.chalmers.se.
1
0.1021/jp984191s CCC: $18.00 © 1999 American Chemical Society
Published on Web 09/29/1999

n-Heptane in Flame
J. Phys. Chem. A, Vol. 103, No. 41, 1999 8223
TABLE 1: GC/MS Setup
premixed laminar (flat) flames. The small size of the burner
obviates the need for extra cooling.
instruments and
methods
The burner is fastened on a motorized precision lift Stepper-
Mike model 18503 with a precision of 1 µm (Oriel, Stratford,
CT). A stepping motor controller (Oriel Stratford, CT) controlled
the height of the lift. The burner/lift combination can be placed
on top of or very close to the GC system with a minimum length
of the sampling line.
configuration and values
instruments
HP 5890 series II GC
HP 5989 A MS engine
Pora PLOT Q fused-silica
column
2
5 m × 0.25 mm i.d.
0
.33 µm film
injection
manual gas injection
Splitflow, 25 mL/min
250 °C
MS (280 °C, SIM mode)
50 °C initial
9
A sampling microprobe was positioned at the center of the
flame. It consisted of a quartz capillary (13 cm length, 2 mm
i.d., 3 mm o.d.) with a 0.1 mm inlet constriction in the tip. The
tip of the capillary was shaped as a cone with a height of 3 mm
and a top angle of about 30°. The microprobe was equipped
with an air cooling assembly made of copper. The cooling
assembly surrounded the microprobe above the tip. Changing
the airflow through the cooling assembly varied the temperature
of the probe. The airflow was set to create a temperature of
injection port
detector
oven ramp
3
2
4
min hold
0 °C/min to 130 °C
min hold
24 °C/min to 250 °C
2 min hold
1 mL/min; H
2
flow rate
2
40 °C inside the tip of the probe.
fused-silica column 25 m × 0.25 mm (Chrompack, Middelburg,
The Netherlands) with a coating of Poraplot Q. The carrier gas
was hydrogen (99. 99999% pure) delivered by a hydrogen
generator model 75-34 (Whatman, Maidstone, England). A HP
Vectra 486/66N controlled the system. The software used was
HP G1034C MS ChemStation rev A 03.03. The mass spectral
library used for comparison/identification was Wiley 138 K
provided by HP. Standards, liquids and gases, were available
for some of the important compounds. The standards were
injected into the systems in the same way as the sample from
the flame. The compounds for which standards were not
available were identified by use of the MS Wiley library. First,
a GC method was developed by fixing the retention times for
the different compounds (Table 1). Second, a selective ion
monitoring (SIM) method was set up. The peaks have been
integrated using the HP software. Generally, the integration
accuracy is lower for minor components due to increased noise
signal and more disturbances of baseline in the chromatograms.
Furthermore, some of the components, for example, the C4
species, are eluting as very narrow peaks in the chromatogram.
However, this could be resolved by using the specific masses
characteristic of the individual species. Formaldehyde was
determined semiqualitatively using Supelclean Lp DNPH (dini-
2.3. Gas Sampling. The sampling microprobe was connected
by polyetheretherketone (PEEK) sampling tubing and compo-
nents (Alltech, Deerfield, IL) to a GC/MS system from HP
(Hewlett-Packard, Palo Alto, CA).
First, the gas sample coming from the microprobe was
1
connected to a /8 in. male/male connector from Swagelock
(
Solon, OH). Teflon ferrules were used to fasten the quartz
microprobe to the connector. An 80 mm long Silco steel tubing
1.0 mm i.d., 1.6 mm o.d.; Alltech, Deerfield, IL) protrudes 35
(
mm into the microprobe. The choice of this large diameter was
made so that the restriction should be mainly in the tip of the
probe. The Silco steel tubing was fastened to a Swagelock
connector using /8 to /16 in. Teflon ferrules. The top of the
Silco steel tubing was connected to a PEEK tubing, 120 cm
1
1
(1.0 mm i.d., 1.6 mm o.d.; Alltech, Deerfield, IL). The tubing
is connected via the GC-1 injection valve to the GC-1 sampling
system. The injection valve was a Valco WT high-temperature
(max temperature 325 °C) six-port two-position valve (Valco,
Houston, TX) rotated by standard air pressure motor. The valve
temperature was 250 °C. The sampling loop was made of
stainless steel with a volume of 1 mL (1 mm i.d.). The injection
mode was split/splitless. The size of the injection was deter-
mined by the time setting on the purge valve.
10
trohydrazine) S10 cartridges (Supelco, Bellefonte, PA).
Pumping at the exit of the GC injection valve with various
pumps resulted in peristaltic effects with increased sampling
errors. The problem was circumvented by using a large vacuum
vessel (25 L) kept at a constant pressure (30 mbar) by a
SOGEVAC SV25 rotating pump (Leybold, K o¨ ln, Germany)
connected to the vessel with a HI-TEC pressure controller
system (Bronkhorst, Ruurlo, The Netherlands). The exit from
the GC sampling valve was connected first to a restrictor
consisting of a 75 cm length PEEK (0.5 mm i.d., 1.6 mm o.d.)
tubing and then to another a fixed 3.5 m length PEEK (0.75
mm i.d., 1.6 mm o.d.) tubing. The longer tubing was connected
directly to the vacuum vessel, and the length of the shorter
tubing was chosen to yield a measured pumping rate of 50 (
The identification of the compounds was made either by using
standards and/or by comparing mass spectra with a mass spectra
library (Wiley 138K). Table 2 shows the method used for
identification.
2.5. Quantification of Ion Peaks. To increase the sensitivity
of the experiments, the SIM (selected ion monitoring) mode
was used. The selected ions measured constituted only the
largest mass fragments of the mass spectrum. The efficiency in
collecting ions for a species using a specific SIM method will
be decreased compared to the case that all the ions were
collected. For a molecule m, the measured ion current, IS,m, is
a fraction, ST,m, of the total ion current, IT,m, obtainable. ST,m is
defined as
5
mL/min. This rate resulted in a maximum signal strength.
IS,m
The 0.5 mm i.d. PEEK tubing was kept at a constant temperature
of 25 °C in a water bath. It is reasonable that the signal versus
pumping rate should go through a maximum. When the pumping
rate is zero, no sample reached the GC, and at high pumping
rates the pressure in the sample loop becomes very low, resulting
in a very small injection.
ST,m
)
(1)
IT,m
To compute ST,m factors for a molecule m, the ratio of measured
ions, according to the SIM method used, is divided by the total
amounts of ions produced. The ratio factors, ST,m, were
computed using the averages of the spectra for the molecule,
m, given in the MS library used (Wiley). Table 2 lists the ST
factor for the largest components found in the n-heptane flame.
2
.4. GC/MS System. In the GC/MS analysis a Hewlett-
Packard (HP) 5890 series II GC connected to a HP 5989A MS
engine (Palo Alto, CA) was used. The GC column was a PLOT

8
224 J. Phys. Chem. A, Vol. 103, No. 41, 1999
Ingemarsson et al.
can be computed at corresponding heights from
TABLE 2: GC Retention Times and MS Factors Used for
Quantificationa
time
F
a
) ST,Kr
Q
Kr
X
Kr/IS,Kr
(5)
(min)
name
methane
krypton
ethene
Q
S
T
SIM
1
1
2
2
2
5
6
6
6
7
7
8
8
8
9
1.39
1.79
2.56
4.75
5.43
6.09
6.10
.35
.39
.28
.33
.79
.84
.07
.26
.39
.16
.59
.56
.69
.93
.12
4.432
3.50
5.862
4.402
7.322
8.752
10.212
7.292
7.292
5.532
0.471
0.511
0.417
0.764
0.538
0.560
0.611
0.685
0.730
0.862
0.713
0.704
0.521
0.721
0.699
0.128
0.587
0.629
0.455
0.535
0.367
0.378
16
84
26
26
taking ST,Kr and QKr from Table 2. By inserting in eq 3 the
computed value of Fa at a specific height z, the mole fraction
Xm for a specific molecule m can be computed at the same
height.
The use of an internal standard corrects for general variations
in the GC/MS instrument such as variations in GC injection
and MS detection (decreased about 7%/h). Krypton elutes
essentially undisturbed by other gases entering the MS ion
source later than the nitrogen and oxygen. However, it was
necessary to add krypton, since the concentration of krypton in
air is only 1.14 ppm. This addition of krypton also allows the
detection of flame disturbances, air leaks, etc. Such disturbances
will increase the argon-to-krypton ratio.
.7. Total Electron Ionization Cross Sections. It is necessary
to use an additivity rule for computing Qm the total electron
ionization cross sections for the measured molecules. Fitch and
Sauter compared 179 sets of experimental cross sections for
a range of organic molecules, m, and subjected these data to a
multiple linear regression by using the numbers of each atom
type as the independent variables. The resulting equation was
b
ethyne
ethane
propene
propane
26, 30, 31
39, 40, 41
29, 30, 40, 41, 44
39, 40, 41
39, 40, 41
#
#
propadiene
#
propyne
methanol
acetaldehyde
1-butene
29, 31
6.962
29, 41, 44
39, 41, 54, 56
39, 54
39, 41, 54, 56
39, 41, 54, 56
42, 43, 55, 57
42, 43, 55, 57
42, 55, 70
42, 43, 56, 57, 69
78
11.642
10.182
11.642
11.642
9.852
1, 3-butadiene
isobutene
#
2-butene
propanal
#
1
1
1
1
1
1
1
2
2-propanone
9.852
1-pentene#
14.532
17.422
13.042
20.312
21.772
1-hexene
benzene
1-heptene
n-heptane
1
1
41, 56
42, 43, 56, 1000
a
t
Q is the total electron ionization cross sections. S the ratio of ions
collected and the total ions formed. SIM shows the selected ion peaks
for use in selective ion monitoring. Normally, standards were used for
8
identification. Species marked # were identified using the Wiley library.
Q ) 0.082 + niai
(6)
m
∑
b
Vallance et al., 1997.13
i)1
By use of eq 1 and an expression for the total ion current,
where ni is the number of atoms i in the molecule and ai is the
constant contribution coefficient of each atom type i (H, C, N,
O, F, Cl, Br, and I) to Qm, the molecular ionization cross section.
The cross section coefficients used for C, H, and O were 1.43,
IT,m,11 the measured ion current, IS,m, can be expressed as
IS,m ) aI dS Q [X ]
(2)
e
T,m
m
m
0
.73, and 1.10, respectively. Fitch and Sauter11 found a
a is general apparatus properties expressing ion extraction,
current amplification, etc., Ie is the ionization electron current,
d is the ionization path length, Qm is the total electron ionization
cross section for the molecule m, and [Xm] is the molar
concentration of the molecule m.
correlation coefficient of 0.996. Equation 6 was used to calculate
the ionization cross sections of the hydrocarbons produced in
the n-heptane/air flame. Table 2 lists the ionization cross sections
used. The value for krypton was taken from an experiment by
1
3
Vallance et al. We also examined our method for computing
ionization cross sections, using several different multicomponent
gaseous and liquid standard mixtures. The difference compared
to the additivity rule (eq 6) was within (10% for the unsaturated
hydrocarbons investigated. The oxygenates, methanol, MTBE,
and acetone gave about 50% lower response compared to the
additivity rule. Benzene, toluene, and xylene produced about
30%-40% lower response than the additivity rule predicted.
2.8. Temperature Measurements. The temperature of the
burner disk was measured continuously during the experiments.
A small (1 mm i.d.) thermocouple type K was positioned inside
the burner in physical contact with the bottom side of the burner
disk. The flame temperatures were measured using Pt-Pt/Rh10
thermocouples (0.10 mm i.d.) inserted in inert alumina tubes
positioned parallel to the burner surface. Two thermocouples
with different outer tube diameters d (0.5 and 1.0 mm) were
used separately to determine the actual flame temperature. As
expected, the coarse thermocouple showed consistently lower
temperatures in comparison with the thin thermocouple. The
radiation correction for these thermocouples, at a specific height
Flame measurements are normally expressed as molar frac-
tions. The mole fraction, Xm, is related to the molar concentration
1
2
[
Xm] by
X ) (RT/p)[X ]
(3)
m
m
where T is the flame temperature at that height, p is the pressure,
and R is the universal gas constant. By use of eqs 2 and 3, the
measured mole fraction Xm can be expressed by
F I
a S,m
Xm )
(4)
ST,mQm
where Fa ) (RT/p)/(aIed) is function of temperature and
therefore of the height z above the burner.
Finally, a small correction of the computed mole fraction
concentration profiles was needed. During the combustion
process there is a slight increase of 5% in the number of moles.
At the burner surface the total molar flow was 27.0 mmol/min.
At complete combustion, at a distance of 1 mm from the burner,
the molar flow is 28.4 mmol/min. A linear correction of the
mole fraction profiles between these two points was used.
9
z above the burner, depends on d, the outer tube diameter. The
actual flame temperature Tflame(z) corresponds to the temperature
measured if the outer diameter is infinitesimal T0.0(z). T0.5(z) is
the measured temperature using the fine thermocouple, T1.0(z)
is the measured temperature using the coarse thermocouple, and
Tflame(z) is the corrected flame temperature. Tflame(z), the actual
flame temperature, can be computed by
2
.6. Inert Internal Standards. In the experiments 1 ( 0.05
vol % krypton, i.e., XKr ) 0.010, was added as an internal
standard. The use of an inert gas as an internal standard allowed
computation of the mole fraction profiles for the detected
molecules. By use of IS,Kr measured at different heights z, Fa

n-Heptane in Flame
Tflame(z) ≈ T (z) + (T (z) - T (z))
J. Phys. Chem. A, Vol. 103, No. 41, 1999 8225
(7)
of fuel is consumed compared to other types of reactors.
Integration of a miniature flame reactor with an on-line GC/
MS provided detailed chemical information on n-heptane
combustion. The flame zone was narrow; the total flame height
to burnout was 1.25 mm. We detected about 30 different stable
organic intermediates. Figure 1 shows three chromatograms of
the oxidation at 100, 1000, and 1300 µm from the burner
surface. In the present experiment, taking place for 15 h,
spurious electronic and flow disturbances will produce some
outliers randomly. Table 3 shows examples of the reproducibility
for the final run presented in this study. Figure 2 shows the
resulting concentration profile of the n-heptane/air flame.
0.5
0.5
1.0
The cooling in the sample point due to the cooled quartz
microprobe was also investigated. A nude thermocouple,
protruding out of an alumina thermocouple shielding tube, was
positioned vertically in the microprobe holder. The temperature
profiles were measured for three configurations relating to the
cooling induced by the microprobe and its associated cooling
assembly. First, we measured with the thermocouple bead about
4
mm from cooling assembly. This corresponds to the distance
between the tip of the microprobe and the cooling assembly.
Second, we measured the temperature profile with the bead 10
mm away from the cooling assembly. Third, we measured
without the cooling assembly present. The measured temperature
difference is a first-order approximation of the cooling of the
sample introduced at the sampling point.
The species profiles of the C2-C7 alkenes formed; ethene,
propene, 1-butene, 1-pentene, 1-hexene, and 1-heptene share a
common form with broad peaks. Their concentration maxima
were found at a distance around 1000 µm. Figure 3 shows the
profiles for these compounds. Note that the ethene concentration
is divided by 8 and that the propene concentration is divided
by 2 in order to fit the profiles in the same plot. The species
profiles of methane, ethane and, propane have a similar form
with narrow peaks at around 1075 µm. Figure 4 shows the
species profiles of methane, ethane, and propane. Note that the
propane concentration is multiplied by 2. Alkynes and dienes
3. Numerical Modeling
The computations were performed using the Sandia flame
code PREMIX.14 The conditions inside the burner, at the burner
surface, and close to the burner are very difficult to model.
Therefore, the cold boundary conditions are simplified in the
PREMIX flame code and other similar flame codes. For
example, the burner is modeled as a cylinder with an open top
surface without any obstructions. In reality, our burner is a metal
disk with holes, with diameters of 150 µm, occupying about
(
ethyne, propyne, propadiene, and 1,3-butadiene) have their
maxima at around 1075 µm as the alkanes. Figure 5 shows
profiles for alkynes and dienes. Note that the ethyne and 1,3-
butadiene concentration is divided by 2 and that the propyne
concentration is multiplied by 4. Methanol, acetaldehyde,
4
0% of the area. Furthermore, in the flame code there are no
surface reactions included. Consequently, the modeling results
close to the burner are less reliable. This may result in a
distortion of the length scale of the computational profiles
compared to the experimental profiles.
2
-propanone, and propanal were also formed in the flame, and
their profiles are shown in Figure 6. Note that the acetaldehyde
concentration is divided by 2. The spatial averaging uncertainty
due to the probe is estimated to be (50 µm. During the final
experimental run the GC measurements were repeated at the
distances of 500, 700, and 900 µm from the burner. These
measurements illustrate the reproducibility during an experi-
mental series (15 h). Some of these measurements are presented
as double points in Figures 2-7 and also in Table 3.
The reaction mechanism of n-heptane oxidation and the
associated thermochemical and transport data were taken from
the work of Held et al. The mechanism used consists of 266
1
reactions and 41 species. The reaction mechanism has been
validated against flow reactor measurements, ignition delays,
1
and laminar flame speeds. However, it has not been used to
predict the species profiles in laminar premixed flames. Krypton
is present in only a minute inert amount in the combustible
mixture. To simplify the computations, the krypton was treated
as an extra addition of argon.
Table 4 shows the major (20) organic intermediates measured.
Dominating intermediate organic species were in decreasing
order: ethene, propene, ethyne, ethane, methane, 1-butene,
One major problem in the modeling of experimental flames
is the uncertainty of the temperature history experienced by the
sample. Molecules in a free adiabatic flame, a burner-stabilized
flame (without cooling probe), and a burner-stabilized flame
with a cooling probe represent the limiting cases. We have
modeled a burner-stabilized flame with corrections included for
the presence of a cooling probe. In the present study a
corresponding base temperature profile was used. However, as
discussed in many studies, the effect of the flame temperature
on the sampled molecules may be very complex. Therefore,
we also investigated the effect on species profiles of uncertain-
ties in the temperature profile. The computations were performed
using 10 different temperature profiles in the interval (200 K
around the base temperature profile. As a complement, we
modeled another heptane flame, also with φ ) 1 as in our
1
-pentene, propane, and 1,3-butadiene. For the concentration
maxima note the pattern ethene > propene > 1-butene >
-pentene > 1-hexene > 1-heptene. Alkene concentrations were
1
found to be a factor 5-10 or higher than those of the
corresponding alkanes. Significant amounts of propyne, propa-
diene, methanol, and acetaldehyde were also formed. The
concentration of formaldehyde was found to be about 2-3 times
higher than the acetaldehyde concentration. Traces of 2-pro-
panone and propanal were also detected. Below the level of 2
-
03
×
10
mol % traces of higher hydrocarbons in the C4-C7
range were found: 2-alkenes, 3-alkenes, higher alkadienes,
higher alkynes, and benzene. Most of these compounds showed
very distributed profiles and only an approximate identification
could be done for these. Figure 7 shows the species profiles of
8
experiment but burning at 0. 06 atm, using the same reaction
2
-butene, isobutene, and benzene, which are some of the minor
mechanism as above. During the final computations the values
on the GRAD and CURVE parameters were 0.1 and 0.2,
respectively. These values resulted in very dense grids of about
organic intermediates, identified in the flame. Note that the
benzene concentration is multiplied by 4.
1
10 grid points situated mainly in the flame zones.
4.2. Temperature Measurements. Figure 8 shows the
temperature profile of the flame. The temperature on the bottom
side of the burner disk was measured continuously during the
experiments; the temperature was 550 ( 50 K. We believe that
corrected flame temperatures at distances from 700 µm are
4
. Results
4
.1. GC/MS Measurements. The miniature flame reactor
provided very good stability and reproducibility. A minimum

8226 J. Phys. Chem. A, Vol. 103, No. 41, 1999
Ingemarsson et al.
TABLE 3: Reproducibilities of the GC/MS Measurementsa
compound
500:1
500:2
700:1 700:2 900:1 900:2
ethene
0.516 0.510
0.037 0.033
0.082 0.080
0.044 0.042
0.042 0.041
0.026 0.026
0.014 0.014
0.008 0.008
0.649 0.657 0.791 0.794
0.045 0.040 0.055 0.052
0.102 0.104 0.120 0.124
0.055 0.046 0.076 0.060
0.052 0.052 0.063 0.064
0.033 0.035 0.038 0.040
0.017 0.018 0.020 0.021
0.009 0.009 0.010 0.011
0.009 0.010 0.011 0.012
0.010 0.011 0.012 0.012
methane
propene
ethyne
ethane
1
1
-butene
-pentene
propane
1
1
.3-butadien 0.007 0.007
-hexene
0.008 0.008
a
Measured values, in mole percent, at three distances 500, 700, and
9
00 µm. The time between measurements 1 and 2 was about 10 h.
Figure 2. Concentration profile of n-heptane in a n-heptane/air flame.
For conditions, see Figure 1 caption.
We also estimated that there is a small cooling of the quartz tip
9
due to radiation corresponding to about 25 K. In total, the
presence of the cooled microprobe reduced the temperatures
by at least 200 K at the sampling point. Recently, the cooling
effect of the probe has been discussed and measured for an
investigation of low-pressure n-heptane flames using a sampling
8
cone. The conclusion from their work is that the gas samples
experience a complex temperature history composed of two
effects, a maximum cooling close to the burner and a minimum
cooling far from the burner.
4
.3. Numerical Modeling. The temperature profile at the
sampling points, corrected for cooling due to the cooled
microprobe, was used as a base for the modeling (see Figure
8
1
). Figure 9 shows the computational results, on a grid with
10 grid points, for the alkene intermediates. The locations of
Figure 1. GC/MS chromatograms of the combustion products from a
n-heptane/air premixed laminar flame burning at stoichiometric com-
position and atmospheric pressure. As an internal standard, 1 mol %
krypton was added to the air. Chromatograms are at distances of 100
µm (top), 1000 µm (middle), and 1300 µm (below) from the burner
surface.
the concentration maxima of the species are shifted toward the
burner in comparison with the present experimental study. The
magnitudes of the maxima are with some exceptions within the
uncertainties of the corresponding experimental values. An
increase or decrease of the base temperature profile of 200 K
shifted the location of the maxima about 0.02 cm toward or
away from the burner, respectively. However, the maxima and
especially the ratios of the maxima changed weakly with
changes in the temperature profile. Table 5 shows the compu-
tational maxima using the two temperature profiles shown in
Figure 8.
accurate within (150 K (see section 2.4). A further complication
is introduced by local cooling at the sampling point due to the
microprobe. We have determined the first-order effects of
cooling at the sampling point. The presence of the cooling
assembly was shown to decrease the temperature about 175 K.

n-Heptane in Flame
J. Phys. Chem. A, Vol. 103, No. 41, 1999 8227
Figure 3. Concentration profiles of ethene, propene, 1-butene, 1-pen-
tene, 1-hexene and 1-heptene in a n-heptane-air flame. Note that the
ethene concentration is divided by 8 and that the propene concentration
is divided by 2. For conditions, see Figure 1 caption.
Figure 5. Concentration profiles of ethyne, propyne, propadiene, and
1
,3-butadiene in a n-heptane/air flame. Note that the ethyne and 1,3-
butadiene concentrations are divided by 2 and that propyne is multiplied
by 4. For conditions, see Figure 1 caption.
Figure 4. Concentration profiles of methane, ethane, and propane in
a n-heptane/air flame. Note that the propane is multiplied by 2. For
conditions, see Figure 1 caption.
Figure 6. Concentration profiles of methanol, acetaldehyde, 2-pro-
panone, and propanal in a n-heptane/air flame. Note that the acetal-
dehyde concentration is divided by 2. For conditions, see Figure 1
caption.
8
Dout e´ et al. analyzed a low-pressure premixed laminar flame
of n-heptane at a low pressure of 60 mbar and 0.7 < æ < 2.0
using molecular beam mass spectrometry (MBMS) comple-
mented by GC/MS for stable species. MBMS allows measure-
ments of both stable species and free radicals. They observed a
marked hierarchy in the concentrations of the intermediate
alkenes measured. The maximum mole fractions decreased
strongly with an increase in the carbon number, ethene >
propene > 1-butene > 1-pentene. We modeled one of these
low-pressure flames, also with φ ) 1 as in our experiment, using
the same reaction mechanism as above. In most case, the
experimental and modeling results agreed fairly well regarding
magnitude and locations of major stable organic intermediates.
The modeling results showed analogous sensitivities to changes
in the temperature profile as above. Table 6 shows the
normalized computational results. In contrast to the model,
8
Dout e´ et al. found that 1-pentene peaked at 0.05 cm from the
burner. Then 1-butene, propene, and ethene peaks appeared at
distances of 0.10, 0.20, and 0.245 cm from the burner,
respectively. In comparison, the peaks of the same alkenes in
the model are very compressed with positions between 0.130
and 0.165 cm.

8228 J. Phys. Chem. A, Vol. 103, No. 41, 1999
Ingemarsson et al.
Figure 7. Concentration profile of 2-butene, isobutene, and benzene
in a n-heptane/air flame. Note that the benzene concentration is
multiplied by 4. For conditions, see Figure 1 caption.
Figure 8. Flame temperature profile of the n-heptane/air flame. The
line is an extrapolation of the temperature between 0 and 550 µm. For
conditions, see Figure 1 caption.
TABLE 4: Organic Compounds Measured in the
n-Heptane/Air Flame at æ ) 1 and 1 atma
distance
(µm)
exptl max
(mol %)
name
ethene
propene
ethyne
ethane
methane
1075
1000
1025
1075
1075
900
0.990
0.127
0.103
0.080
0.069
0.040
0.021
1
1
-butene
-pentene
900
1000
1075
900
0.016
0.014
0.012
methanol
acetaldehyde
propyne
1075
1075
875
1000
925
1075
1075
925
1050
850
0.0091
0.0078
0.0057
0.0042
0.0037
0.0023
0.0013
0.00099
0.00036
propadiene
1
-heptene
isobutene
-butene
propanal
-propanone
benzene
2
2
-
5
5. 0 × 10
a
Distances are estimated to be accurate within (50 µm. Species
concentrations are uncertain by a factor of 1.5. Direct calibration
indicated that the concentration maxima for the oxygenates should be
a factor 2 higher. No corrections are made to the concentrations because
of the calibration of the additivity rule (see section 2.7).
Figure 9. Computational results, on a grid with 110 grid points, for
the alkene intermediates.
calibration procedures were used. Therefore, the uncertainties
could be minimized and checked subsequently.
5. Discussion
First, on-line GC/FID (aided by GC/FTIR for identification)
as a complementary instrumental method was used. However,
accurate absolute concentration measurements using a direct
(inert) internal standard are not possible with GC/FID. The
results, using the different GC methods, such as ratios of
concentration maxima relative to ethene, were within (25% in
most cases. Second, to produce absolute concentration measure-
ments, krypton was added as an internal standard. Use of an
internal inert standard is the most reliable approach possible
for quantification of signals to concentrations. The reason is
that there are drifts in the experimental system and especially
the GC/MS instrument. Use of a carbon mass balance does not
First, we want to discuss the major concentration uncertainties
involved in the present experiment. Second, we want to discuss
our results in relation to other experiments and theoretical
mechanisms.
5.1. Experimental Concentration Uncertainties. The present
experimental setup is quite complex, which motivates a sum-
mary and discussion of the major uncertainties involved
regarding concentration measurements. Several preliminary
experimental series were produced, yielding consistent results.
Most important, overlapping instrumental and evaluational
methods regarding instrumental methods, internal standards, and

n-Heptane in Flame
J. Phys. Chem. A, Vol. 103, No. 41, 1999 8229
TABLE 5: Organic Compounds in the n-Heptane/Air Flame
at æ ) 1 and 1 atma
ibilities and uncertainties in concentration measurements are not
normally well described or documented. In comparing these
investigations, one has to understand some underlying similari-
ties. The distance from a premixed laminar burner, the time in
a flow reactor, and the temperature scale in a jet-stirred reactor
experiment reflect the progress of the total reaction. Of course,
there are also important differences. For example, the burner
experiments are performed at higher temperatures, and transport
effects affect the burner results. However, it is interesting to
find out if there are features common for all of the investigations.
Such features would reflect central properties in the oxidation
of n-heptane, independent of the specific physical condition and
specific reactor.
cooled probe
no probe
dist
max
dist
max
name
ethene
propene
ethyne
ethane
methane
(µm)
(mol %)
(µm)
(mol %)
481
481
563
444
506
456
447
513
450
544
556
719
0.783
0.165
0.184
0.097
0.108
0.088
0.060
0.028
0.015
0.024
0.004
356
356
444
331
388
331
331
388
331
413
444
519
0.697
0.160
0.142
0.113
0.102
0.111
0.063
0.028
0.015
0.028
0.002
1
1
1
1
-butene
-pentene
,3-butadiene
-hexene
acetaldehyde
propadien
benzene
1
Held et al. investigated lean and rich oxidation of n-heptane
-
6
-6
in a flow reactor at 3 atm and a temperature interval between
5 × 10
3 × 10
9
30 and 980 K. The intermediate organic species consisted
a
1
Model predictions based on the mechanism by Held et al.
Temperature profiles used in the model were taken from Figure 8.
primarily of small alkenes in decreasing quantity as the carbon
number increased. The largest alkene detected was 1-hexene
(limit, 1-5 ppm); errors and reproducibilities were not given.
C4-C6 species formed early in the reaction, ethane formed in
the middle of the reaction, and the other C1-C3 species formed
relatively late. An exception to the rule was 1,3-butadiene, which
TABLE 6: Organic Compounds in the n-Heptane/Air Flame
at æ ) 1 and 0. 06 atma
experiment
model
dist
max
dist
max
1
formed late in the reaction. Interestingly, Held et al. found that,
name
ethene
propene
ethyne
(cm)
(mol %)
(cm)
(mol %)
in comparison with experiment, for æ ) 0.79 the model
produced a factor of 2.5 higher maximum of 1-pentene than
the experiment. For 1-butene and 1,3-butadiene the model
produced a factor of 1.5 higher maxima than the experiment.
0.950
0.200
0.300
0.250
0.100
1.360
0.450
0.564
unread
0.156
0.163
0.244
0.188
0.138
0.119
0.131
0.188
0.250
0.131
0.231
0.298
2.029
0.480
0.358
0.288
0.234
0.305
0.161
0.073
0.072
0.041
0.008
methane
b
1
-butene
0.152
7
Dagaut et al. investigated oxidation of n-heptane in a jet-stirred
ethane
nd
0.061
0.034
b
reactor (550-1150 K, 10 atm, 0.3 e æ e 1.5). For the alkenes
the highest maximum mole fractions were found for ethene >
propene > 1-butene > 1-pentene > 1-hexene. Other intermedi-
ate hydrocarbons at high concentrations were methane, ethane,
and 1,3-butadiene. C4-C7 species formed at low temperatures,
ethane formed at intermediate temperatures, and the other C1-
C3 formed at higher temperatures. The dominating oxygenates
found were formaldehyde, acetaldehyde, methanol, propanal,
and 2-propanone.
1
1
-pentene
,3-butadiene
0.050
0.250
b
acetaldehyde
-hexene
nd
nd
1
propadiene
benzene
0.250
0.300
0.007
7 × 10
-
4
-5
1 × 10
a
8
b
Experimental results by Dout e´ et al. using MBMS.
Calibrated
by GC. MBMS gave higher values. Model predictions using the mech-
1
8
anism by Held et al. and the temperature profile are from Figure 1c.
There are four central features found in our study and
supported by the experimental investigations discussed above.1
They exhibit the same following key features as our study. First,
there is a marked hierarchy in the maxima of the species profiles
in the concentrations of the intermediate alkenes: ethene >
propene > 1-butene > 1-pentene >1-hexene > 1-heptene.
Second, the concentrations of the alkenes are significantly higher
than the corresponding alkanes. Third, C4-C7 intermediates
peaks early during the oxidation of n-heptane and C1-C3
hydrocarbons late. Fourth, as an exception to the rule, 1,3-
butadiene forms late. Strikingly, the key features above are true
over a wide range of æ (equivalence ratio) and pressure.
work. Differential diffusion leads to changes in the elemental
mass fraction for carbon atoms throughout a premixed laminar
flame.
,7,8
1
5
Third, the determination of the final concentrations of the
measured species was based on the use of the additivity rule
for computing ionization cross sections. The good agreement
with direct calibrations indicates that this was a good choice.
Fourth, the effect of nonlinearity was checked for several
representative species including argon, krypton, methane,
propane, heptane, and carbon dioxide. The major reason for
nonlinearity is a temporary high pressure in the ion source. Only
carbon dioxide, present at very high maximum concentrations,
among the products showed a significant nonlinearity. For the
HP 5989A MS engine, using powerful differential pumping,
linearity of more than 3 orders of magnitude is expected.
We believe that the major uncertainty stems from integration
of chromatographic peaks, especially small flat peaks disturbed
by baseline noise. Overlapping of neighboring peaks is another
factor increasing the concentration uncertainty. The results from
the complementary methods discussed all produced consistent
results. The concentration results presented here are uncertain
by about a factor of 1.5 for the major intermediates.
5.3. Differences between Computational and Experimental
Concentration Maxima. In the present study we have inves-
tigated a flame model based on the semiempirical mechanism
1
developed by Held et al. The most important results are
summarized in Table 7. The model describes the produced
amount of major organic intermediates comparatively well.
However, important species not present in the mechanism are
propane, propyne, methanol, isobutene, 2-butene, and 1-heptene.
Therefore, the model will have a reduced flexibility. Further-
more, compared to the present flame experiment the model
produced significantly higher maxima for some species: 1-butene,
1,3-butadiene, and 1-pentene. Similar results, regarding 1,3-
butadiene and 1-pentene, for the low-pressure flame experiments
5.2. Comparison with Other Reactor Experiments. Below
we have cited experimental investigations of oxidation of
n-heptane in different reactors.1
,7,8
Comparisons with other
8
experimental studies are hampered by the fact that reproduc-
measured by Dout e´ et al. at æ ) 1 were found. Furthermore,

8230 J. Phys. Chem. A, Vol. 103, No. 41, 1999
Ingemarsson et al.
TABLE 7: Organic Compounds in the n-Heptane/Air
Flames at æ ) 1a
ments, with regard to 1-butene, 1-pentene and 1,3-butadiene
may be explained by these simplifications. For the case of 1,3-
butadiene, Held et al. stated that the discrepancy found in their
1
expt
1 atm
model
1 atm
expt
0.06 atm
model
0.06 atm
name
ethene
propene
ethyne
oxidation experiment was probably due to an incomplete
oxidation mechanism. However, the combined effect of the
uncertainties of the rate constants in a flame model may also
lead to distortions of the computed maxima. This is best
illustrated by comparing two different heptane mechanisms. For
example, for φ ) 1.3 the computed flame velocities are 29.4
and 45.7 cm/s using the flame models developed by Held et
1
1
1
1
0.128
0.104
0.070
0.040
0.081
0.021
0.014
0.008
0.012
0.004
5 × 10
0.210
0.236
0.139
0.112
0.124
0.076
0.036
0.030
0.019
0.005
6 × 10
0.331
0.415
unread
0.118
nd
0.045
0.025
nd
nd
0.005
5 × 10
0.237
0.176
0.142
0.115
0.150
0.079
0.036
0.035
0.020
0.004
5 × 10
methane
1
-butene
ethane
1
1
-pentene
,3-butadiene
1
3
al. and Lindstedt and Maurice, respectively. This is a very
substantial difference indicating significant uncertainties regard-
ing central parameters in one or both of the flame models.
acetaldehyde
-hexene
1
propadien
benzene
-
5
-6
-4
-6
Acknowledgment. We are very grateful to Prof. Dryer and
Dr. Kosakov at Princeton University for sending us their
mechanism and preliminary results regarding heptane flames.
This work was in part supported by the Combustion Engine
Research Center, Chalmers University of Technology, Bycosin
AB, Karlstad and Kemiindustrins forskarskola, Association of
Swedish Chemical industries, Stockholm.
a
Normalized maximum values for organics with respect to ethene,
based on Tables 4-6. Present experiment at 1 atm and experiment at
8
0
.06 atm by Dout e´ et al. using MBMS. Model predictions are based
1
on the mechanism by Held et al.
1
in a flow reactor Held et al. themselves found significant higher
computational maxima compared to their experimental concen-
tration maxima, for 1-butene, 1-pentene, and 1,3-butadiene for
the lean case at æ ) 0.79. The flame model also produced a
factor of 10 less benzene compared to the present experiment.
References and Notes
(1) Held, T. J.; Marchese, A. J.; Dryer, F. L. Combust. Sci. Technol.
8
For the low-pressure experiments at æ ) 1 the model produced
1997, 123, 107-146.
(2) Curran, H. J.; Gaffuri, P.; Pitz, W. J.; Westbrook, C. K. Combust.
Flame 1998, 114, 149-177.
a factor of 100 less benzene than the experiment. Normalization
against ethene revealed that for the stoichiometric conditions
evaluated, the mechanism behaved very similarly, within a wide
range of temperatures of (200 K for two very different pressures
(
3) Lindstedt, R. P.; Maurice, L. Q. Combust. Sci. Technol. 1995, 107,
3
17-353.
(4) Bollig, M.; Pitsch, H.; Hewson, J. C.; Seshadri, K. Twentysixth
Symposium (International) on Combustion; Combustion Institute: Pitts-
burgh, PA, 1996; pp 729-737.
1
and 0.06 atm (see Table 7). For a recent explanation of such
15
scaling effects with pressure, see Pope et al. The breakdown,
resulting in subsequent formation of a sequence of gradually
smaller alkenes, is very compressed in the model in comparison
with the experimental results obtained at 0.06 atm by Dout e´ et
(5) Come, G. M.; Warth, V.; Glaude, P. A.; Fournet, R.; Battin-Leclerc,
F.; Scacchi, G. Twentysixth Symposium (International) on Combustion;
Combustion Institute: Pittsburgh PA, 1996; pp 755-762.
(
(
6) Hamins, A.; Seshadri, K. Combust. Flame 1987, 68, 295-307.
7) Dagaut, P.; Reuillon, M.; Cathonnet, M. Combust. Sci. Technol.
8
al.
1
994, 95, 233-260.
8) Dout e´ , C.; Delfau, J. L.; Akrich, R.; Vovelle, C. Combust. Sci.
Technol. 1997, 124, 249-276.
9) Fristrom, R. M.; Westenberg, A. A. Flame Structure; McGraw-
Hill Inc.: New York, 1965.
The central features of high-temperature oxidation of n-
heptane are as follows.1 H abstraction of n-heptane leads to
formation of four distinct n-heptyl radicals 1-heptyl, 2-heptyl,
(
,2
(
3
-heptyl, and 4-heptyl. Site-specific abstraction rate constants
(
10) Ingemarsson, A° .; Nilsson, U.; Nilsson, M.; Pedersen, J.; Olsson, J.
control the formation rates of 1-alkenes. Because of the ethene,
propene, 1-butene, 1-pentene, and 1-hexene will form in
decreasing amounts. The smaller alkenes such as ethene will
also form via other reactions paths. In the semidetailed mech-
anism used in the present study, this complex initial oxidation
of n-heptane is compressed in a few empirical reactions.
Furthermore, many important organic intermediates are not
present in the present simplified mechanism. The high compu-
tational concentration maxima found, compared to the experi-
Chemosphere 1998, 36, 2879-2889.
(11) Fitch, W. L.; Sauter, A. D. Anal. Chem. 1983, 55, 832-835.
(12) Kee, R. J.; Miller, J. A.; Jefferson, T. H. Report SAND80-8003;
Sandia National Laboratories: Livermore, CA, 1980.
13) Vallance, C.; Harris, S. A.; Hudson, E. J.; Harland, P. W. J. Phys.
(
B 1997, 30, 2465-2475.
(14) Kee, R. J.; Grcar, J. F.; Smooke, M. D.; Miller, J. A. A Fortran
Program for Modeling Steady Laminar One-Dimensional Premixed Flames;
Report SAND85-8240; Sandia National Laboratories: Livermore, CA, 1985.
(15) Pope, C. J.; Shandross, R. A.; Howard, J. B. Combust. Flame 1999,
116, 605-614.