You are on page 1of 20

Combustion and Flame 140 (2005) 267286

www.elsevier.com/locate/combustflame

Co-oxidation in the auto-ignition of primary reference fuels


and n-heptane/toluene blends
Johan Andrae a, , David Johansson a , Pehr Bjrnbom a , Per Risberg b ,
Gautam Kalghatgi b,c
a Department of Chemical Engineering and Technology, Chemical Reaction Engineering, Royal Institute of Technology,

Teknikringen 42, SE-100 44 Stockholm, Sweden


b Department of Machine Design, Internal Combustion Engines, Royal Institute of Technology, SE-100 44 Stockholm, Sweden
c Shell Global Solutions, P.O. Box 1, Chester CH1 3SH, UK

Received 30 June 2004; received in revised form 29 October 2004; accepted 24 November 2004
Available online 23 December 2004

Abstract
Auto-ignition of fuel mixtures was investigated both theoretically and experimentally to gain further understanding of the fuel chemistry. A homogeneous charge compression ignition (HCCI) engine was run under different
operating conditions with fuels of different RON and MON and different chemistries. Fuels considered were
primary reference fuels and toluene/n-heptane blends. The experiments were modeled with a single-zone adiabatic model together with detailed chemical kinetic models. In the model validation, co-oxidation reactions
between the individual fuel components were found to be important in order to predict HCCI experiments, shocktube ignition delay time data, and ignition delay times in rapid compression machines. The kinetic models with
added co-oxidation reactions further predicted that an n-heptane/toluene fuel with the same RON as the corresponding primary reference fuel had higher resistance to auto-ignition in HCCI combustion for lower intake
temperatures and higher intake pressures. However, for higher intake temperatures and lower intake pressures the
n-heptane/toluene fuel and the PRF fuel had similar combustion phasing.
2004 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
Keywords: HCCI; Homogeneous charge compression ignition; Auto-ignition; Fuel chemistry; Primary reference fuels;
n-heptane; Toluene; Co-oxidation; CHEMKIN

1. Introduction
Homogeneous charge compression ignition
(HCCI) combustion, first studied over 20 years ago
[1,2], is a combustion process which utilizes a (more
or less) homogeneous fuel/air mixture. Combustion is
initiated by auto-ignition of the usually very fuel-lean
* Corresponding author. Fax: +46-8-696-0007.

E-mail address: johana@ket.kth.se (J. Andrae).

mixture. It offers a number of benefits over conventional spark-ignition and diesel engines, such as much
lower NOx emissions, higher combustion efficiency
at part load than its SI counterpart, and zero particulates [3,4]. A disadvantage is the relatively high
emissions of unburned hydrocarbons [5], the sources
being quenching wall boundary layers [6] and crevice
regions [7,8].
Unlike the diesel and spark-ignition engines,
where the combustion is directly controlled by the

0010-2180/$ see front matter 2004 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
doi:10.1016/j.combustflame.2004.11.009

268

J. Andrae et al. / Combustion and Flame 140 (2005) 267286

engine management system, the combustion in HCCI


engines is controlled by its chemical kinetics only [9].
Therefore, HCCI engines will have fuel requirements
that differ from diesel and spark-ignition engines
and must be properly understood. Clearly the critical problems for HCCI engines are the control of
auto-ignition and combustion rate, and increased understanding of the chemical kinetic processes that
lead to auto-ignition is necessary.
Tanaka et al. [10] performed experiments on HCCI
combustion in a rapid compression machine (RCM)
with complex fuels such as cyclic paraffins, olefins,
and aromatics, which exist in petroleum-based fuels.
They concluded that for HCCI combustion, the ignition delay and the burn rate could be independently
controlled using various fuel mixtures and additives.
However, theoretical studies of auto-ignition with
complex fuel mixtures and detailed chemistry have
not been investigated much. The main reason for this
is that the combustion mechanisms have not yet been
well established, especially not for mixtures with aromatic and olefins mixed with paraffins. Simmie [11]
has given a recent review on the development of detailed chemical models for hydrocarbon fuels.
Internal combustion engines burn blends of large
molecular-weight liquid fuels, a class to which the
primary reference fuels (PRF) n-heptane and isooctane belong. For primary reference fuels both very
detailed [1217] and reduced or semi-detailed models
[1821] have been developed.
Klotz et al. [22] developed a mechanism for combustion of toluene-butane blends by merging the neat
mechanisms for the individual fuel components and
validated the mechanism by carrying out experiments
in an atmospheric plug-flow reactor at 1170 K. They
demonstrated that when the chemical interactions between the fuel components were limited to radical
pool effects, in order for a merged kinetic mechanism to predict experimental data for fuel blends, the
blend model should be properly configured to predict
the oxidation processes of the neat fuel components.
Their mechanism was not validated against engine
data with significantly different pressure and time histories compared to a plug-flow reactor.
Ogink [23] has developed a surrogate kinetic
mechanism for gasoline combustion (iso-octane 55%
by volume, toluene 35, and n-heptane 10). This reduced mechanism, consisting of around 100 species
among 500 reactions, has mainly been validated
against shock-tube data and been used in CFD calculations to predict gasoline HCCI combustion.
The pioneering work of Leppard [24] considered
auto-ignition during knock and attributed the sensitivity of a non-PRF fuel to the differences in the lowtemperature chemistry between such fuels and PRF
fuels in the negative temperature coefficient, NTC,

region. Leppard suggested that in the MON test condition, the NTC region dominates the chemistry of
PRF fuels and reactions slow down, making PRF fuels relatively more resistant to knock compared to
fuels containing aromatics and olefins. In the RON
test condition the NTC chemistry becomes less important so that the PRF fuels lose the advantage of
higher resistance to auto-ignition compared to nonPRF fuels. Hence a non-PRF fuel has low MON and
high RONit is relatively more resistant than a PRF
fuel to auto-ignition in the RON test than in the MON
test.
In this work we have studied the auto-ignition of
fuel blends including mixtures of primary reference
fuels and n-heptane/toluene. Emphasis is on detailed
chemistry modeling and a simpler single-zone physical model of the engine is used. The chemistry models
are validated against shock-tube and rapid compression machine data and then used to simulate HCCI
experiments. Our goal is to gain further insight into
the modeling of complex fuel chemistry of fuel blends
that is necessary in order to understand the autoignition process. For that purpose we have especially
investigated the role of co-oxidation between the individual fuel components. An attempt is made to understand why the same non-PRF fuel behaves like different PRF fuels under different conditions in terms of
auto-ignition chemistry.

2. Experimental procedure
The experiments were conducted in a single cylinder engine based on a Scania D12 model, as part of
the Green Car project involving the Swedish government, Volvo, Scania CV, and Shell. The engine
details and fuels studied are given in Tables 1 and 2,
respectively, and the experimental procedure is summarized below. In a typical experiment the engine oil
and water temperatures are brought up to the operating values, 80 C for oil and 90 C for water. The
Table 1
Details for the single-cylinder engine based on a Scania D12
model
Compression ratio
Bore
Stroke
Connecting rod length
Number of valves
Inlet valve opening (IVO)
Inlet valve closing (IVC)
Maximfum inlet valve lit
Exhaust valve opening (EVO)
Exhaust valve closing (EVC)
Maximum exhaust valve lift

16.7
127 mm
154 mm
255 mm
4
367 CAD ATC
581 CAD ATC
14.8 mm
121 CAD ATC
349 CAD ATC
14.8 mm

J. Andrae et al. / Combustion and Flame 140 (2005) 267286

269

intake boost pressure and temperature and the engine speed are fixed and the fuel is introduced until
auto-ignition takes place as indicated by the pressure
signal and the increase in power delivered by the engine. The fuel quantity is adjusted until the required
normalized air fuel ratio, , is achieved. is the inverse of the fuel/air equivalence ratio, . Once the
engine operating characteristics have stabilized at the
required levels a hundred pressure cycles are acquired
and stored for later analysis. This is repeated for different fuels under the same operating conditions

engine speed, intake air temperature and pressure, and


excess air/fuel ratio. Thus different fuels are subjected
to the same pressure and temperature history but they
will show different auto-ignition behaviors because
they have different chemistries.
In this paper we consider four operating conditions
listed in Table 3. Two of these conditions, OP1 and
OP2, have a high inlet pressure but little heating of the
intake air while the other two conditions have a high
intake air temperature with no intake boost. More information of the experiments can be found in [25].

Table 2
Fuels used and their properties

3. Theoretical discussion

Fuel
A
B
C
D

i-C8 H18
%vol

n-C7 H16
%vol

94
84

6
16
25
35

-CH3
%vol

RON

MON

3.1. MON, RON, and HCCI-combustion


94
84
94.2
83.9

75
65

94
84
82.6
73.2

Table 3
Operating conditions

OP1
OP2
OP3
OP4

Engine
speed, rpm

Intake air
temperature,
C

Intake
pressure,
bar (abs)

900
1200
900
1200

40
40
120
120

2
2
1
1

4.0
5.5
3.5
3.0

Fig. 1 shows the bulk gas temperature plotted


against pressure during compression in four different
HCCI experiments; also shown are the plots corresponding to the RON and MON tests inferred from
information in Ref. [24]. These lines are different because of differences in the engine inlet conditions and
engine design. In the RON and MON tests the engine is run with a slightly rich mixture, the normalized
fuel/air ratio, 1.1, while the HCCI tests are run
with a lean mixture ( 0.3). The mixture temperature will follow such a characteristic polytropic line as
pressure increases but will go above such a line when
auto-ignition reactions start. This deviation will occur at different points depending on the auto-ignition

Fig. 1. Compression temperature versus pressure in different auto-ignition experiments.

270

J. Andrae et al. / Combustion and Flame 140 (2005) 267286

behavior of the fuel/air mixture under the particular


operating condition.
An HCCI engine can be run at a condition such
that fuels of different chemistry will indeed be ranked
for auto-ignition phasing according to their RON rating, e.g., condition OP3 (Fig. 6 in Ref. [25]) or the
MON rating. However engines can also be run at conditions where fuels of different chemistry cannot be
ranked by either RON or MON alone.
For all practical purposes, for both knock and
HCCI, the true auto-ignition quality of a non-PRF fuel
is given by the octane index
OI = RON KS,

(1)

where sensitivity
S = RON MON.

(2)

S is a measure of the difference in auto-ignition chemistry between PRF and the non-PRF fuel being considered, on moving from the MON to the RON test
conditions; for a PRF fuel, S is zero. For a non-PRF
fuel, the OI is the octane number of the PRF fuel with
the equivalent auto-ignition characteristics at the particular condition. K is a constant depending on the
operating conditionsit is not a property of the fuel.
From Fig. 1 it can be seen that for both knock and
HCCI, K decreases as we move to a lower polytrope,
i.e., as the temperature for a given pressure decreases.
K is positive or negative depending on whether the
temperature for a given pressure is higher or lower
than the condition where the auto-ignition quality is
determined by RONin Fig. 1, condition OP3 in
HCCI tests and the RON line for knock tests. The
two lines do not coincide probably because of differences in mixture strengthK is probably a function
of both the operating polytrope and , for > 0.3.
HCCI experiments show that K depends primarily on
the temperature at a given pressure [26,27] but there
is some evidence that K also increases as increases
from 0.3 to 0.4 [26].
HCCI engines can be run at different operating
conditions such that K varies widely. Under condition
OP1, with a high intake pressure, Fuel D, a mixture
of 65% volume toluene and 35% volume n-heptane,
with 84 RON and 73 MON was more resistant to
auto-ignition than iso-octane, PRF100, so that K was
1.6. However an HCCI engine can be run in such
a way that K > 1 (see Fig. 1), by increasing the intake temperature [26]. In such cases, Fuel D would
match a PRF fuel of an octane number less than 73.
Thus compared to PRF 84, Fuel D becomes more or
less resistant to auto-ignition depending on whether
the polytrope is below or above the OP3 polytrope.
In practice, compared to a PRF fuel, a sensitive fuel
(with aromatics, olefins, oxygenates) becomes more

resistant to auto-ignition if the pressure is increased


for a given temperature, and less resistant to autoignition if the temperature is increased for a given
pressure.
If the temperature for a given pressure is even
lower than in the RON test, the non-PRF fuel will
be even more resistant to auto-ignition compared to
a PRF fuel of the same RON at this new conditionit
will match a PRF fuel with an octane number higher
than its own RON; K will be negative. Similarly if
the temperature is increased beyond the MON test
condition, the PRF fuels become even more resistant
to auto-ignition in comparison to sensitive fuels and
the sensitive fuel will have an OI less than its MON;
K will be greater than unity. The effect of pressure,
temperature, and, to a limited extent, mixture strength
on different fuel chemistries must be understood in
order to explain the behavior of non-PRF fuels under
different conditions.
3.2. Co-oxidation reactions
In a mixture of fuels the obvious chemical coupling during the combustion is the radical pool (O,
OH, and H). However, in a mixture consisting of two
or more fuel components that is compressed from a
low temperature until auto-ignition, there is also the
possibility that chemical interactions are taking place
between the different fuel components and their radicals, Examples of co-oxidation reactions are
RH + R1 = R + R1 H

(3)

and
R1 OO + RH = R1 OOH + R ,

(4)

where RH is the fuel, R is alkyl radical after H abstraction, and ROO an alkyl-peroxy radical formed
when the alkyl radical reacts with molecular oxygen.
If co-oxidation reactions are neglected undue restrictions are introduced in the reaction paths for a
mixture of two fuels compared to the reactions of
each of the two fuels alone. This may not be immediately obvious but the following simple example
clearly shows why this is so. Let us assume that two
reactants, A and B, both can react to form C:
2A C;

2,
r1 = k1 CA

(5)

2B C;

2.
r2 = k2 CB

(6)

If we mix A and B we could assume that only


those reactions occur. This would correspond to not
including co-oxidation reactions. We could also assume that A and B can react with each other, corresponding to including co-oxidation reactions:
A + B C;

r3 = k3 CA CB .

(7)

J. Andrae et al. / Combustion and Flame 140 (2005) 267286

If we consider a mixture of A and B with total concentration CF = CA + CB the reaction rate would
become
2 + k C2 + k C C
r = (r1 + r2 + r3 ) = k1 CA
2 B
3 A B

 2
2
2
= k1 + k2 (1 ) + k3 (1 ) CF ,
(8)

where
=

CA
.
CF

(9)

Now let A and B become chemically equal, for


example, assuming that one atom in a B molecule is
from the same element but another isotope than in A.
Then the reaction rate could be written in two ways:
r = k1 CF2 = (r1 + r2 + r3 )


= k1 2 + k1 (1 )2 + k3 (1 ) CF2 .

(10)

Hence




k3 
2
2
2
k1 = k1 + 1 2 + +

,
k1

(11)

and
1 = (2 n) 2 (2 n) + 1,

(12)

where
k
n= 3.
k1

(13)

The only value that can satisfy Eq. (12) for all
values of is n = 2. The explanation of this mathematical fact that we have to count the third reaction
with double reaction rate is that reaction (7) gets reactants from two different populations of molecules
while the two first reactions get both reactant molecules from the same population.
A more comprehensive discussion more directed
to co-oxidation kinetics can be found in [28], which
treats low-temperature autoxidation of hydrocarbons
and olefins. In that context the importance of cooxidation reactions for the chemical kinetics of the
autoxidation of mixtures is well established. More recent work can for example be found in [2931] where
co-oxidation reactions are used in pseudo-detailed
mechanisms used to describe low-temperature oxidation of hydrocarbons and antioxidants.

4. Modeling procedure
A single-zone modeling approach (no heat transfer, crevices, and charge inhomogeneities) is used.
The usefulness of the single-zone assumption is in
providing an estimate of ignition delay time as a function of thermodynamic conditions in the combustion
chamber [32].

271

For problem setup and simulation with detailed


chemical kinetics we employ the AURORA software,
which is the perfectly stirred reactor application in
the CHEMKIN collection [33]. The internal combustion engine model of AURORA is set to solve the
equations for a nonisothermal batch reactor (closed
system), where the variation of the volume in a combustion cylinder in an internal combustion engine is
provided by the equations in Heywood [34]. When
a mechanism consists of K species, there is a set
of K + 1 nonlinear ordinary differential equations,
which by integration yields the temperature and mass
fractions. The pressure is calculated with the ideal gas
law.
As a starting point for detailed chemistry modeling of Fuels A and B (see Table 2), the latest
reaction mechanism from the Lawrence Livermore
National Laboratory (LLNL) combustion chemistry
group [15,16] was used. It consists of 4238 reactions among 1034 species, most of them reversible.
The chemical interactions between n-heptane and isooctane are only through the radical pool in the LLNL
mechanism.
To model Fuels C and D (see Table 2) toluene
chemistry is needed. As a first step, the detailed mechanism from [35] and the detailed n-heptane mechanism version 2 from LLNL [16] were merged. The
toluene mechanism in [35] has been validated by experiments of toluene oxidation in a an atmospheric
jet-stirred reactor, by simulating the oxidation of benzene at 0.46 to 10 atm under stirred reactor conditions,
the ignition of benzene-oxygen-argon mixtures, and
the combustion of benzene in flames. The merged
mechanism consisted of 619 species among 3135 reactions, most of which were reversible.
All calculations were run on a Dell 650 Precision
Workstation at 3.06 GHz with 4 GB of RAM. A typical CPU time was for a run with Fuels A and B around
10 min and for Fuels C and D around 5 min. In the calculations the step size for the ODE solver (DASPK)
in AURORA was 0.1 crank-angle degree, in accordance with the gathered experimental data points. The
corresponding CPU times for calculations involving
sensitivity analysis described below depended on the
interval for calculation and size of mechanism, but
were usual within 60 min.
5. Results and discussion
With the assumption that co-oxidation reactions
are taking place as described above, these types of
reactions were added to the initial mechanisms for Fuels AD. First a validation of the mechanisms (mostly
the PRF mechanism as more experimental data are
available) is performed against shock-tube and rapid
compression machine data. Then the experiments in

272

J. Andrae et al. / Combustion and Flame 140 (2005) 267286

Fig. 2. Calculated ignition-delay times for primary reference fuel mixtures at constant volume compared with experimental
shock-tube results from Fieweger et al. [36]. p = 40 bar, = 1.0. The LLNL mechanism with added co-oxidations reactions in
Appendix A improves the prediction for PRF80.

the HCCI-engine are simulated and the impact of cooxidation reactions is discussed.
5.1. Validation of reaction mechanisms
The added co-oxidation reactions for the PRF
mechanism [15,16] together with their rate constants
are listed in Appendix A. Based on [17] and similar
reactions found for n-heptane in the LLNL mechanism (e.g., nc7h16 + c7h15-1 = c7h15-2 + nc7h16)
an activation barrier of 1014 kcal/mol was taken
for the hydrogen abstraction reactions from isooctane by heptyl radicals (reactions 42394254 in
Appendix A). For co-oxidation involving alkylperoxides abstracting hydrogen atoms from iso-octane and
n-heptane (reactions 42554286) a similar activation
energy between 15 and 20 kcal/mol was assumed as
for reactions found in the LLNL mechanism for nheptane and iso-octane (e.g., nc7h16 + c7h15o2-2 =
c7h15-3 + c7h15o2h-2 and ic8h18 + ac8h17o2 =
dc8h17 + ac8h17o2h). Also logical activation energies that accounted for the relative strength of
primary, secondary, and tertiary CH bonds were
adopted in the co-oxidation reactions. Consequently,
as tertiary bonds are weakest, these CH bondbreaking reactions were faster than secondary and
primary ones. Finally co-oxidation reactions involving two other isomers of heptane were assumed;
2,4-dimethylpentane (denoted c7h162-4) and 2,2-

dimethylpentane (denoted neoc7h16) in reactions


42874370 in Appendix A.
Fig. 2 shows model predictions (constant volume
simulations) of shock-tube data from Fieweger et al.
[36] with and without inclusion of co-oxidation reactions in Appendix A. For PRF0 and PRF100 the
mechanism with and without co-oxidation gives basically the same result as expected. The slight deviation
for PRF100 at lower temperatures is due to the formation of neoc7h16 and c7h162-4 from iso-octane decomposition. However, for the PRF80 mixture there
is a significant improvement with co-oxidation reactions included and the experimental data can be predicted well. When simulations were conducted with
reactions 42874370 removed, it could be seen that
those reactions were less significant compared to reactions 42394286 in predicting shock-tube data for
the PRF80 mixture.
In Fig. 3 model calculations with the PRF mechanism with and without co-oxidation reactions are
compared with ignition delay times measured in a
rapid compression machine as a function of research
octane number [10]. The inclusion of co-oxidation reactions in Appendix A improves the prediction, but
the ignition delay times are still too large compared
to the experiments. Tanaka et al. [18] also found better agreement between model predictions and experimental data by including a coupling reaction between
n-heptane and iso-octane. The shortcoming to predict

J. Andrae et al. / Combustion and Flame 140 (2005) 267286

273

Fig. 3. Predicted ignition-delay times for primary reference fuels as a function of octane number in a rapid compression machine.
p = 1 bar, T = 318 K, = 2.5. Filled rings: experimental results from Tanaka et al. [10]. Solid line: LLNL mechanism with
added co-oxidations reactions in Appendix A. Dashed line: LLNL mechanism [16].

rapid compression machine ignition delay times for


leaner mixtures may be due to a n-heptane mechanism that is not perfectly tuned under these conditions [37].
The same type of co-oxidations was assumed for
Fuels C and D as for the PRF Fuels (see Appendix B).
There are no relevant shock-tube data for n-heptane/
toluene blends available in the literature to validate
the mechanism in the same way as for PRF fuels. But
we did perform constant volume calculations for the
merged mechanism with only toluene as reactant, and
found good agreement with shock-tube experimental
data from [38]. When the n-heptane/toluene mechanism was constructed the C1C4 chemistry from nheptane was used. The constant volume calculations
for pure toluene were then a check that the ignition
delay times were not affected after this change to the
original toluene mechanism.
5.2. Comparison between HCCI experiments and
computed results
All HCCI engine calculations started at 99 crankangle degrees before top dead center. This is near
the middle of the compression stroke and before any
chemical reaction has had any influence on the heat
release. Also comparison between experimental data
and model predictions is more reliable compared to
starting at inlet valve close (IVC) 139 crank-angle degrees before top dead center (see Table 1).

The pressure and the average cylinder temperature have been measured in HCCI experiments for
different fuels of different chemistries [25]. However
in engines, the charge is not truly homogeneous [39]
primarily because the fresh charge mixes with hot
residual gases from the previous cycle. Hence autoignition should start at these hot spots [40] whose
temperature would be higher than the average, bulk
temperature. The difference, T , between the temperature of the hot spot and the bulk temperature
should be larger, the cooler the intake charge. Hence
we assume that T at 99 before top dead center,
our initial point in the calculation is 45 K for OP1
and 25 K for OP3. These chosen values for T are
arbitrary but without such an increment on the measured temperature, especially Fuel A (PRF 94) and
Fuel D could not be made to ignite at OP1 and OP2
even when they actually did ignite in experiments.
In Figs. 4 and 5 predicted and experimental pressures are plotted against crank angle for Fuels A, B
and C, D, respectively, for operating condition OP3
(high intake temperature/low intake ressure). The corresponding curves for Fuels B and D at operating condition OP1 (low inlet temperature/high inlet pressure)
are shown in Fig. 6. Both simulations with and without added co-oxidation reactions in Appendices A
and B are shown. Obviously the single-zone model
overpredicts the pressure rise during auto-ignition but
this is of minor importance in this work where the
time of auto-ignition is more relevant.

274

J. Andrae et al. / Combustion and Flame 140 (2005) 267286

Fig. 4. Experimental and calculated pressures for Fuels A and B as a function of crank angle for operating condition OP3 (low
intake pressure/high intake temperature). Initial temperature and pressure at the start of calculations (99 before top dead center)
is 472 K, 1.74 bar for Fuel B and 455 K, 1.37 bar for Fuel A, respectively. ATDC, after top dead center.

Fig. 5. Experimental and calculated pressures for Fuels C and D as a function of crank angle for operating condition OP3 (low
intake pressure/high intake temperature). Temperature and pressure at the start of calculations (99 before top dead center) is
472 K and 1.74 bar, respectively.

Comparing Fuel B in Fig. 4 (OP3) and Fig. 6


(OP1) shows that adding the co-oxidation reactions

does not change ignition delay at OP3 but decreases


ignition delay at OP1. After addition of co-oxidation

J. Andrae et al. / Combustion and Flame 140 (2005) 267286

275

Fig. 6. Experimental and calculated pressures for Fuels B and D as a function of crank angle for operating condition OP1 (high
intake pressure/low intake temperature). Temperature and pressure at the start of calculations (99 before top dead center) is
415 K and 3.34 bar, respectively.

reactions the change of reactivity of Fuel B from OP1


to OP3 agrees better with experiments than before.
For Fuels C and D the inclusion of co-oxidation reactions significantly improves the model prediction
for OP3 (see Fig. 5). In fact, Fuel C did not even
ignite without adding the co-oxidation reactions in
Appendix B. The kinetic models with co-oxidation
reaction give similar ignition delay for Fuel B and
Fuel D at OP3.
Fig. 6 shows that adding co-oxidation reactions to
Fuel D decreases the ignition delay at OP1. Both with
and without co-oxidation reaction we get that Fuel B
has about 5 CAD less ignition delay than Fuel D at
OP1 (see Fig. 6).
Figs. 7 and 8 show the heat-release rates (J/deg)
for Fuels B and D for conditions OP3 and OP1, respectively. The modeled heat-release rates have been
multiplied by 0.1 and 0.2 for OP3 and OP1, respectively, a consequence of using a single-zone model
that overpredicts the heat release rate. The simulated
and experimental ignition delays are close to each
other at OP3 and that in both experiments and simulations Fuel B ignites more than 5 CAD before Fuel D at
OP1. However, the simulated ignition delay increased
between OP3 and OP1 for Fuel D while it was mainly
unchanged for Fuel B. On the other hand, the experimental ignition delay was mainly unchanged for
Fuel D while it decreased for Fuel B.

Moreover, the mechanisms cannot predict the


trends in comparing the low-temperature heat release
for the fuels at OP1. The model predicts that lowtemperature heat release starts earlier and becomes
higher for Fuel D compared to Fuel B, while the experiments predict that the heat release starts earlier
for Fuel B than for Fuel D with similar amounts of
energy released by low-temperature chemistry (see
Fig. 8). The double peaks for the main heat release in
Fig. 8 are predicted by the model and are not a result
of bad resolution.
Thus, by adding co-oxidation reactions to the kinetic models, we could predict that the difference in
ignition delay between Fuel B and D, evident at condition OP1, disappears for condition OP3. This was
not possible without co-oxidation reactions. However,
the changes of ignition delays for each fuel did not
agree well with experimental data. This may have to
do with the difficulty in determining the most appropriate starting temperatures for the simulations due
to inhomogeneities of the charges, including hot-spot
formation, in the HCCI experiments.
The results for OP2 and OP4 not shown in figures
were in accordance with the results discussed above
for OP1 and OP3. In addition to trends regarding
combustion phasing with and without co-oxidation
reactions, the model could in line with experiments
predict that Fuel C did not ignite for OP1 and OP2
(high intake pressure/low intake temperature).

276

J. Andrae et al. / Combustion and Flame 140 (2005) 267286

Fig. 7. Experimental and calculated heat release rate (J/deg) for Fuels B and D as a function of crank angle for operating
condition OP3 (low intake pressure/high intake temperature). Temperature and pressure at the start of calculations (99 before
top dead center) is 472 K and 1.74 bar, respectively.

5.3. Conversion of individual fuel components


Fig. 9 shows the conversion of the individual fuel
components for Fuel B calculated for conditions OP1
and OP3. Different trends can be seen for OP1 and
OP3. As could be expected from Fig. 4 showing the
pressure, for OP3 there is no difference in the results
with or without co-oxidation reactions. One likely
explanation is that OP3 is in the negative temperature coefficient (NTC) region [24,41] for this type
of fuel and that the effect of new reaction paths is
compensated by the effect of those parts in the mechanism that gives NTC behavior. For OP1 on the other
hand, the difference is evident. Fuel B experiences
cool flame behavior and around 40% of n-heptane
(nc7h16) is consumed. On turning off the reactions in
Appendix A the cool flame response is much weaker,
although present. This indicates the importance of cooxidation between iso-octane (ic8h18) and n-heptane
(nc7h16) for OP1.
As already noted in the theoretical discussion
above, the increased conversion of both fuel components is simply the effect of new reaction paths being
added for both components. Obviously this general
effect is important since it decreases ignition delays
significantly resulting in better agreement with experiments.
Fig. 10 shows a bar plot of the fractional contribution of the production of c7h15-2 radicals from 99 to

20 crank-angle degrees before top dead center, Fuel B


and OP1. The bars are calculated by normalizing the
integral of rate of production (mol/(cm3 s)) for the
most important reactions. As would be expected, cooxidation reactions with iso-octane and n-heptane are
important prior to auto-ignition.
Figs. 11 and 12 show the conversion of the individual fuel components for Fuel D calculated for
conditions OP3 and OP1, respectively. The initiation
of toluene is the formation of benzyl (PHCH2 ) and
hydroperoxy radicals by the reaction of toluene with
oxygen:
TOLUEN + O2 = PHCH2 + HO2 .

(14)

Reaction (14) can begin at relatively low temperatures since the CH bonds of the methyl group
in toluene have unusually low bond dissociation energy [41]. As the radical pool builds by the lowtemperature oxidation of n-heptane, the reaction
TOLUEN + OH = PHCH2 + H2 O

(15)

starts to dominate the toluene consumption and benzyl production. This in turn will inhibit n-heptane oxidation as toluene competes with n-heptane for OH
radicals. Benzyl radicals, the product in reaction (15),
are less reactive compared to heptyl radicals produced by n-heptane and OH. Note that there is a rapid
partial conversion at an early CAD, and then there
is a plateau in toluene conversion combined with a

J. Andrae et al. / Combustion and Flame 140 (2005) 267286

277

Fig. 8. Experimental and calculated heat release rate (J/deg) for Fuels B and D as a function of crank angle for operating
condition OP1 (high intake pressure/low intake temperature). Temperature and pressure at the start of calculations (99 before
top dead center) is 415 K and 3.34 bar, respectively.

Fig. 9. Calculated conversion for individual fuel components of Fuel B (ic8h18 and nc7h16) as a function of crank angle for
OP3 and OP1.

slow n-heptane conversion until the final conversion


of toluene is taking place at a much later CAD. The
plateau in toluene conversion may be an effect of the

thermal stability of the benzyl radicals due to electron


delocalization [41] resulting in an equilibrium condition.

278

J. Andrae et al. / Combustion and Flame 140 (2005) 267286

Fig. 10. Contribution of different reactions to overall rate of production for the c7h15-2 radical: Fuel B, OP1, Interval 9920
crank angles before top dead center.

Fig. 11. Calculated conversion for individual fuel components of Fuel D (nc7h16 and c6h5ch3) as a function of crank angle for
OP3.

With co-oxidation reactions it is, according to


the model, possible for benzyl radicals to attack nheptane resulting in higher net fuel conversion since
these reactions consume a relative stable benzyl rad-

ical while producing a reactive heptyl. The following reactions of heptyl will then produce OH that in
turn will convert more toluene into benzyl, according to reaction (15), leading to the higher initial con-

J. Andrae et al. / Combustion and Flame 140 (2005) 267286

279

Fig. 12. Calculated conversion for individual fuel components of Fuel D (nc7h16 and c6h5ch3) as a function of crank angle for
OP1.

version of both fuels that can be seen in Figs. 11


and 12.
Note that the partial n-heptane conversion is more
pronounced at OP1 than at OP3. Again this may be
due to compensation of some of that effect by expected NTC behavior at OP3 conditions by some parts
of the n-heptane oxidation mechanism [24,41].
Fig. 13 shows a bar plot of the fractional contribution of the production of benzyl (PHCH2 ) from
99 to 28 crank-angle degrees before top dead center,
Fuel D and OP1. The bars are calculated by normalizing the integral of rate of production (mol/cm3 s) for
the most important reactions. Fig. 13 supports the fact
that reaction (15) above is the dominant path for benzyl production when the radical pool builds up, while
reactions 14 in Appendix B dominates the consumption of benzyl.
5.4. Sensitivity analysis
To further examine the fuel chemistry in the lowtemperature region prior to auto-ignition, a sensitivity
analysis was carried out.
The system of ordinary differential equations that
describe the physical problem are of the general form
d
= F (, t, a),
dt

(16)

where , in this case, is the vector of temperature


and mass fractions. The parameter a represents the

rate expressions for the gas-phase reactions. The firstorder sensitivity coefficient matrix is then defined as
wj,i =

,
i

(17)

where the indices j and i refer to the dependent variables and reactions, respectively.
Differentiating Eq. (17) with respect to the preexponentials Ai yields
dwj
F j
F
=
wj,i +
.
dt

(18)

In DASPK, the sensitivity equations are solved simultaneously with the dependent variables of the solution itself. The Jacobian matrix F / in Eq. (18)
is exactly the one that is required by the backwardsdifferentiation formula method in solving the original
model problem, so it is readily available for sensitivity
computation. The raw sensitivity coefficients are normalized in the form of logarithmic derivates to make
them more useful, e.g., for species mass fractions


ln Yk 
i Yk 
(19)
=
.
ln i F
Yk i F
Although the gas-phase species solution variables
are mass fractions, the sensitivity coefficients are
computed in terms of mole fractions as follows,



Kg

i Xk 
1 Yk 
i Yk 

=
i W
,
Xk i F
Yk i F
W i F
j =1 j
(20)

280

J. Andrae et al. / Combustion and Flame 140 (2005) 267286

Fig. 13. Contribution of different reactions to overall rate of production for the benzyl (PHCH2 ) radical: Fuel D, OP1, interval
9928 crank angles before top dead center.

Fig. 14. Normalized first-order sensitivity coefficients of the OH radical with respect to preexponentials for Fuel B at OP1. 20
CA ATDC, T = 826 K, p = 46.5 bar.

where Xk are the mole fractions, Wj are the species


molecular weights, and W is the mean molecular
weight of the mixture.

Fig. 14 shows a bar plot of normalized first-order


sensitivity coefficients of the OH radical with respect
to preexponentials for Fuel B at OP1. The sensitiv-

J. Andrae et al. / Combustion and Flame 140 (2005) 267286

281

Fig. 15. Normalized first-order sensitivity coefficients of the OH radical with respect to preexponentials for Fuel D at OP1. 28
CA ATDC, T = 751 K, p = 31.3 bar.

ity calculation was done between 22 and 20 crankangles before top dead center and the position shown
in Fig. 14 is 20 before top dead center. It can be noted
that co-oxidation reactions are among the most sensitive for building up the radical pool.
In Fig. 15 shows a bar plot of normalized firstorder sensitivity coefficients of the OH radical with
respect to preexponentials for Fuel D at OP1. The sensitivity calculation was done between 30 and 28 crank
angles before top dead center and the position shown
in Fig. 15 is 28 before top dead center. It can be
noted that co-oxidation reactions 14 in Appendix B
are among the most sensitive for building up the radical pool as they transform benzyl radicals to the more
reactive heptyl radicals. This also supports the discussion above that co-oxidation reactions lead to an
increased OH production and a faster fuel conversion.
5.5. An attempt at chemical explanation of the
co-oxidation effects
The ignition delays of Fuels B (PRF84) and D
(toluene/n-heptane, RON84) in the HCCI experiments are almost equal at OP3 (low intake pressure,
high intake temperature) while Fuel B ignites much
earlier than Fuel D at OP1 (high intake pressure, low
intake temperature). Thus their ignition delays do not
correspond to their RON or MON. This raises the
question if co-oxidation reactions could be important

for this behavior. The following discussion summarizes our attempt to answer this question.
n-heptane has 10 secondary carbonhydrogen
bonds and 6 primary ones. Iso-octane has one tertiary carbonhydrogen bond, two secondary carbon
hydrogen bonds, and 15 primary carbonhydrogen
bonds. Toluene has three primary carbonhydrogen
bonds and five carbonhydrogen bonds on the aromatic nucleus. Due to the increasing bond dissociation energies free radicals abstract hydrogen most
easily from the tertiary carbonhydrogen bonds, less
easily from the secondary bonds, and least easily from
the primary bonds in iso-octane and n-heptane. The
abstraction of hydrogen from the aromatic nucleus
in toluene is very difficult and may almost be neglected while abstraction of hydrogen from the three
primary carbonhydrogen bonds would be the easiest case according to the bond dissociation energy
that is 5 kcal/mol less for a primary bond in toluene
than for a tertiary carbonhydrogen bond in a paraffin.
However, the benzyl radicals formed after hydrogen
abstraction from toluene are thermally stable due to
electron delocalization and their reaction with oxygen
at auto-ignition temperatures is thermodynamically
less favored. They will preferably take part in radicalradical reactions [41]. The radicals formed from isooctane and n-heptane, on the other hand, will have
numerous reaction paths available. However, some of
the reaction paths in n-heptane oxidation and to a less

282

J. Andrae et al. / Combustion and Flame 140 (2005) 267286

extent in iso-octane oxidation result in a moderation


or decrease of the oxidation rate, within the so-called
NTC temperature region [41].
If we compare an iso-octane/n-heptane mixture
with a toluene/n-heptane mixture the addition of cooxidation reactions to the kinetic models would mean
that we add steps where free radicals generated from
n-heptane, such as heptyl, hydroperoxy, and alkoxy
radicals, may abstract hydrogen from iso-octane in
the first case and from toluene in the second case.
Also radicals generated from iso-octane and from
toluene may abstract hydrogen from n-heptane. However, the number of reaction paths for radicals formed
by hydrogen abstraction is much higher in the isooctane case than in the toluene case. Furthermore
the benzyl radicals formed by hydrogen abstraction
from toluene have a limited power to initiate oxidation chain reactions due their thermal stability. Thus
we would expect that if we add co-oxidation reactions
to the kinetic models outside the NTC region the isooctane/n-heptane mixture would appear to increase
more in reactivity than the toluene/n-heptane mixture.
Such an effect would occur at high intake pressure
and low intake temperature (OP1) due to favorable
conditions for low-temperature chemistry since OP1,
unlike OP3, is outside the NTC region. When adding
co-oxidation reactions at low intake pressure and high
intake temperature (OP3), the conditions are in the
NTC region and the iso-octane/n-heptane mixture
would not increase in reactivity while the n-heptane/
toluene would do so to some extent due to less influence of the NTC effects. Such an explanation would
agree well with Leppards work [24] as discussed in
the introduction.
Obviously the experimental results in the HCCI
tests concerning Fuels B and D are in line with this
argument if some of the deviations between simulations and experiments could be explained by charge
inhomogeneity and the formation of hot spots. The
experimental results show that the two fuels were
more equal in terms of ignitions delays at OP3 while
Fuel B had much shorter ignition delays than Fuel D
at OP1. In the simulations this behavior could only be

seen if co-oxidation reactions were used in the kinetic


models.
6. Conclusions
Auto-ignition of fuel mixtures including primary
reference fuels and n-heptane/toluene blends has been
modeled with detailed chemical kinetics together with
a single-zone adiabatic model, and the results compared with experimental data. The most important
findings are summarized below.
The addition of co-oxidation reactions to the kinetic model improved the prediction of shock-tube
data for PRF80 and ignition delay times in rapid compression machines as a function of octane number.
By adding co-oxidation reactions, analogous to
the corresponding reactions in autoxidation of hydrocarbons, a significant improvement of the prediction
of the general trend of auto-ignition phasing in HCCI
combustion of fuel mixtures was found.
The model could, in agreement with the experiments, predict that iso-octane/n-heptane and toluene/
n-heptane with research octane number 84 (note that
the motor octane number for such a toluene/n-heptane
mixture is 73) have a similar auto-ignition behavior
at low intake pressure and high intake temperature
while iso-octane/n-heptane auto-ignites earlier than
toluene/n-heptane at high intake pressure and low intake temperature. This could only be predicted by the
use of co-oxidation reactions.
Acknowledgments
Financial support from the Swedish Research
Council under Contract No. 2002-5807, the Swedish
government, Volvo, Scania CV, and Shell under the
Green Car project is gratefully acknowledged. We
also thank Professor Hans-Erik ngstrm for his support, Eric Lycke in the engine laboratory at KTH for
his help, and guest researcher Matthew Bardsley, Imperial College, for his help in conducting calculations
within a summer project. Finally we thank the reviewers for valuable comments that significantly improved
the quality of our paper.

Appendix A
Rate constants for added co-oxidation reactions between n-heptane (nc7h16) and iso-octane (ic8h18) to the LLNL mechanism
[15,16]
4239.
4240.
4241.
4242.
4243.
4244.
4245.

ic8h18 + c7h15-1 = nc7h16 + ac8h17


ic8h18 + c7h15-1 = nc7h16 + bc8h17
ic8h18 + c7h15-1 = nc7h16 + cc8h17
ic8h18 + c7h15-1 = nc7h16 + dc8h17
ic8h18 + c7h15-2 = nc7h16 + ac8h17
ic8h18 + c7h15-2 = nc7h16 + bc8h17
ic8h18 + c7h15-2 = nc7h16 + cc8h17

9.00E+11
2.00E+11
1.00E+11
6.00E+11
9.00E+11
2.00E+11
1.00E+11

0.0
0.0
0.0
0.0
0.0
0.0
0.0

13500.0
11200.0
9000.0
13500.0
14500.0
11200.0
10000.0
(continued on next page)

J. Andrae et al. / Combustion and Flame 140 (2005) 267286

283

Appendix A (continued)
4246.
4247.
4248.
4249.
4250.
4251.
4252.
4253.
4254.
4255.
4256.
4257.
4258.
4259.
4260.
4261.
4262.
4263.
4264.
4265.
4266.
4267.
4268.
4269.
4270.
4271.
4272.
4273.
4274.
4275.
4276.
4277.
4278.
4279.
4280.
4281.
4282.
4283.
4284.
4285.
4286.
4287.
4288.
4289.
4290.
4291.
4292.
4293.
4294.
4295.
4296.
4297.
4298.
4299.
4300.
4301.
4302.
4303.

ic8h18 + c7h15-2 = nc7h16 + dc8h17


ic8h18 + c7h15-3 = nc7h16 + ac8h17
ic8h18 + c7h15-3 = nc7h16 + bc8h17
ic8h18 + c7h15-3 = nc7h16 + cc8h17
ic8h18 + c7h15-3 = nc7h16 + dc8h17
ic8h18 + c7h15-4 = nc7h16 + ac8h17
ic8h18 + c7h15-4 = nc7h16 + bc8h17
ic8h18 + c7h15-4 = nc7h16 + cc8h17
ic8h18 + c7h15-4 = nc7h16 + dc8h17
ac8h17o2 + nc7h16 = ac8h17o2h + c7h15-1
ac8h17o2 + nc7h16 = ac8h17o2h + c7h15-2
ac8h17o2 + nc7h16 = ac8h17o2h + c7h15-3
ac8h17o2 + nc7h16 = ac8h17o2h + c7h15-4
bc8h17o2 + nc7h16 = bc8h17o2h + c7h15-1
bc8h17o2 + nc7h16 = bc8h17o2h + c7h15-2
bc8h17o2 + nc7h16 = bc8h17o2h + c7h15-3
bc8h17o2 + nc7h16 = bc8h17o2h + c7h15-4
cc8h17o2 + nc7h16 = cc8h17o2h + c7h15-1
cc8h17o2 + nc7h16 = cc8h17o2h + c7h15-2
cc8h17o2 + nc7h16 = cc8h17o2h + c7h15-3
cc8h17o2 + nc7h16 = cc8h17o2h + c7h15-4
dc8h17o2 + nc7h16 = dc8h17o2h + c7h15-1
dc8h17o2 + nc7h16 = dc8h17o2h + c7h15-2
dc8h17o2 + nc7h16 = dc8h17o2h + c7h15-3
dc8h17o2 + nc7h16 = dc8h17o2h + c7h15-4
c7h15o2-1 + ic8h18 = ac8h17 + c7h15o2h-1
c7h15o2-2 + ic8h18 = ac8h17 + c7h15o2h-2
c7h15o2-3 + ic8h18 = ac8h17 + c7h15o2h-3
c7h15o2-4 + ic8h18 = ac8h17 + c7h15o2h-4
c7h15o2-1 + ic8h18 = bc8h17 + c7h15o2h-1
c7h15o2-2 + ic8h18 = bc8h17 + c7h15o2h-2
c7h15o2-3 + ic8h18 = bc8h17 + c7h15o2h-3
c7h15o2-4 + ic8h18 = bc8h17 + c7h15o2h-4
c7h15o2-1 + ic8h18 = cc8h17 + c7h15o2h-1
c7h15o2-2 + ic8h18 = cc8h17 + c7h15o2h-2
c7h15o2-3 + ic8h18 = cc8h17 + c7h15o2h-3
c7h15o2-4 + ic8h18 = cc8h17 + c7h15o2h-4
c7h15o2-1 + ic8h18 = dc8h17 + c7h15o2h-1
c7h15o2-2 + ic8h18 = dc8h17 + c7h15o2h-2
c7h15o2-3 + ic8h18 = dc8h17 + c7h15o2h-3
c7h15o2-4 + ic8h18 = dc8h17 + c7h15o2h-4
ic8h18 + xc7h15 = c7h162-4 + ac8h17
ic8h18 + xc7h15 = c7h162-4 + bc8h17
ic8h18 + xc7h15 = c7h162-4 + cc8h17
ic8h18 + xc7h15 = c7h162-4 + dc8h17
ic8h18 + yc7h15 = c7h162-4 + ac8h17
ic8h18 + yc7h15 = c7h162-4 + bc8h17
ic8h18 + yc7h15 = c7h162-4 + cc8h17
ic8h18 + yc7h15 = c7h162-4 + dc8h17
ic8h18 + zc7h15 = c7h162-4 + ac8h17
ic8h18 + zc7h15 = c7h162-4 + bc8h17
ic8h18 + zc7h15 = c7h162-4 + cc8h17
ic8h18 + zc7h15 = c7h162-4 + dc8h17
ic8h18 + xc7h15o2 = ac8h17 + xc7h15o2h
ic8h18 + yc7h15o2 = ac8h17 + yc7h15o2h
ic8h18 + zc7h15o2 = ac8h17 + zc7h15o2h
ic8h18 + xc7h15o2 = bc8h17 + xc7h15o2h
ic8h18 + yc7h15o2 = bc8h17 + yc7h15o2h

6.00E+11
9.00E+11
2.00E+11
1.00E+11
6.00E+11
9.00E+11
2.00E+11
1.00E+11
6.00E+11
3.15E+13
2.10E+13
2.10E+13
1.05E+13
3.15E+13
2.10E+13
2.10E+13
1.05E+13
3.15E+13
2.10E+13
2.10E+13
1.05E+13
3.15E+13
2.10E+13
2.10E+13
1.05E+13
1.21E+14
1.21E+14
1.21E+14
1.21E+14
3.06E+13
3.06E+13
3.06E+13
3.06E+13
1.53E+13
1.53E+13
1.53E+13
1.53E+13
8.07E+13
8.07E+13
8.07E+13
8.07E+13
9.00E+11
2.00E+11
1.00E+11
6.00E+11
9.00E+11
2.00E+11
1.00E+11
6.00E+11
9.00E+11
2.00E+11
1.00E+11
6.00E+11
1.21E+14
1.21E+14
1.21E+14
5.60E+13
5.60E+13

0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0

14500.0
14500.0
11200.0
10000.0
14500.0
14500.0
11200.0
10000.0
14500.0
17500.0
15500.0
15500.0
15500.0
17500.0
15500.0
15500.0
15500.0
17500.0
15500.0
15500.0
15500.0
17500.0
15500.0
15500.0
15500.0
18500.0
18500.0
18500.0
18500.0
15500.0
15500.0
15500.0
15500.0
14500.0
14500.0
14500.0
14500.0
18500.0
18500.0
18500.0
18500.0
13500.0
11200.0
9000.0
13500.0
14500.0
11200.0
10000.0
14500.0
14500.0
11200.0
10000.0
14500.0
18500.0
18500.0
18500.0
15500.0
15500.0

(continued on next page)

284

J. Andrae et al. / Combustion and Flame 140 (2005) 267286

Appendix A (continued)
4304.
4305.
4306.
4307.
4308.
4309.
4310.
4311.
4312.
4313.
4314.
4315.
4316.
4317.
4318.
4319.
4320.
4321.
4322.
4323.
4324.
4325.
4326.
4327.
4328.
4329.
4330.
4331.
4332.
4333.
4334.
4335.
4336.
4337.
4338.
4339.
4340.
4341.
4342.
4343.
4344.
4345.
4346.
4347.
4348.
4349.
4350.
4351.
4352.
4353.
4354.
4355.
4356.
4357.
4358.
4359.
4360.
4361.

ic8h18 + zc7h15o2 = bc8h17 + zc7h15o2h


ic8h18 + xc7h15o2 = cc8h17 + xc7h15o2h
ic8h18 + yc7h15o2 = cc8h17 + yc7h15o2h
ic8h18 + zc7h15o2 = cc8h17 + zc7h15o2h
ic8h18 + xc7h15o2 = dc8h17 + xc7h15o2h
ic8h18 + yc7h15o2 = dc8h17 + yc7h15o2h
ic8h18 + zc7h15o2 = dc8h17 + zc7h15o2h
ac8h17o2 + c7h162-4 = ac8h17o2h + xc7h15
ac8h17o2 + c7h162-4 = ac8h17o2h + yc7h15
ac8h17o2 + c7h162-4 = ac8h17o2h + zc7h15
bc8h17o2 + c7h162-4 = bc8h17o2h + xc7h15
bc8h17o2 + c7h162-4 = bc8h17o2h + yc7h15
bc8h17o2 + c7h162-4 = bc8h17o2h + zc7h15
cc8h17o2 + c7h162-4 = cc8h17o2h + xc7h15
cc8h17o2 + c7h162-4 = cc8h17o2h + yc7h15
cc8h17o2 + c7h162-4 = cc8h17o2h + zc7h15
dc8h17o2 + c7h162-4 = dc8h17o2h + xc7h15
dc8h17o2 + c7h162-4 = dc8h17o2h + yc7h15
dc8h17o2 + c7h162-4 = dc8h17o2h + zc7h15
ic8h18 + nc7h15 = neoc7h16 + ac8h17
ic8h18 + nc7h15 = neoc7h16 + bc8h17
ic8h18 + nc7h15 = neoc7h16 + cc8h17
ic8h18 + nc7h15 = neoc7h16 + dc8h17
ic8h18 + oc7h15 = neoc7h16 + ac8h17
ic8h18 + oc7h15 = neoc7h16 + bc8h17
ic8h18 + oc7h15 = neoc7h16 + cc8h17
ic8h18 + oc7h15 = neoc7h16 + dc8h17
ic8h18 + pc7h15 = neoc7h16 + ac8h17
ic8h18 + pc7h15 = neoc7h16 + bc8h17
ic8h18 + pc7h15 = neoc7h16 + cc8h17
ic8h18 + pc7h15 = neoc7h16 + dc8h17
ic8h18 + qc7h15 = neoc7h16 + ac8h17
ic8h18 + qc7h15 = neoc7h16 + bc8h17
ic8h18 + qc7h15 = neoc7h16 + cc8h17
ic8h18 + qc7h15 = neoc7h16 + dc8h17
neoc7h16 + ac8h17o2 = nc7h15 + ac8h17o2h
neoc7h16 + bc8h17o2 = nc7h15 + bc8h17o2h
neoc7h16 + cc8h17o2 = nc7h15 + cc8h17o2h
neoc7h16 + dc8h17o2 = nc7h15 + dc8h17o2h
neoc7h16 + ac8h17o2 = oc7h15 + ac8h17o2h
neoc7h16 + bc8h17o2 = oc7h15 + bc8h17o2h
neoc7h16 + cc8h17o2 = oc7h15 + cc8h17o2h
neoc7h16 + dc8h17o2 = oc7h15 + dc8h17o2h
neoc7h16 + ac8h17o2 = pc7h15 + ac8h17o2h
neoc7h16 + bc8h17o2 = pc7h15 + bc8h17o2h
neoc7h16 + cc8h17o2 = pc7h15 + cc8h17o2h
neoc7h16 + dc8h17o2 = pc7h15 + dc8h17o2h
neoc7h16 + ac8h17o2 = qc7h15 + ac8h17o2h
neoc7h16 + bc8h17o2 = qc7h15 + bc8h17o2h
neoc7h16 + cc8h17o2 = qc7h15 + cc8h17o2h
neoc7h16 + dc8h17o2 = qc7h15 + dc8h17o2h
ic8h18 + nc7h15o2 = nc7h15o2h + ac8h17
ic8h18 + oc7h15o2 = oc7h15o2h + ac8h17
ic8h18 + pc7h15o2 = pc7h15o2h + ac8h17
ic8h18 + qc7h15o2 = qc7h15o2h + ac8h17
ic8h18 + nc7h15o2 = nc7h15o2h + bc8h17
ic8h18 + oc7h15o2 = oc7h15o2h + bc8h17
ic8h18 + pc7h15o2 = pc7h15o2h + bc8h17

5.60E+13
2.80E+13
2.80E+13
2.80E+13
8.07E+13
8.07E+13
8.07E+13
6.30E+13
6.30E+13
6.30E+13
6.30E+13
6.30E+13
6.30E+13
6.30E+13
6.30E+13
6.30E+13
6.30E+13
6.30E+13
6.30E+13
9.00E+11
2.00E+11
1.00E+11
6.00E+11
9.00E+11
2.00E+11
1.00E+11
6.00E+11
9.00E+11
2.00E+11
1.00E+11
6.00E+11
9.00E+11
2.00E+11
1.00E+11
6.00E+11
1.17E+14
1.17E+14
1.17E+14
1.17E+14
2.60E+13
2.60E+13
2.60E+13
2.60E+13
2.60E+13
2.60E+13
2.60E+13
2.60E+13
3.90E+13
3.90E+13
3.90E+13
3.90E+13
2.52E+14
2.52E+14
2.52E+14
2.52E+14
2.60E+13
2.60E+13
2.60E+13

0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0

15500.0
14500.0
14500.0
14500.0
18500.0
18500.0
18500.0
17500.0
15500.0
16500.0
17500.0
15500.0
16500.0
17500.0
15500.0
16500.0
17500.0
15500.0
16500.0
13500.0
11200.0
9000.0
13500.0
14500.0
11200.0
10000.0
14500.0
14500.0
11200.0
10000.0
14500.0
14500.0
11200.0
10000.0
14500.0
17500.0
17500.0
17500.0
17500.0
15500.0
15500.0
15500.0
15500.0
15500.0
15500.0
15500.0
15500.0
17500.0
17500.0
17500.0
17500.0
18500.0
18500.0
18500.0
18500.0
15500.0
15500.0
15500.0

(continued on next page)

J. Andrae et al. / Combustion and Flame 140 (2005) 267286

285

Appendix A (continued)
4362.
4363.
4364.
4365.
4366.
4367.
4368.
4369.
4370.

ic8h18 + qc7h15o2 = qc7h15o2h + bc8h17


ic8h18 + nc7h15o2 = nc7h15o2h + cc8h17
ic8h18 + oc7h15o2 = oc7h15o2h + cc8h17
ic8h18 + pc7h15o2 = pc7h15o2h + cc8h17
ic8h18 + qc7h15o2 = qc7h15o2h + cc8h17
ic8h18 + nc7h15o2 = nc7h15o2h + dc8h17
ic8h18 + oc7h15o2 = oc7h15o2h + dc8h17
ic8h18 + pc7h15o2 = pc7h15o2h + dc8h17
ic8h18 + qc7h15o2 = qc7h15o2h + dc8h17

2.60E+13
1.30E+13
1.30E+13
1.30E+13
1.30E+13
1.68E+14
1.68E+14
1.68E+14
1.68E+14

0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0

15500.0
14500.0
14500.0
14500.0
14500.0
18500.0
18500.0
18500.0
18500.0

kf = AT n exp(E/RT ).
A units mol cm s K; E units cal/mol.
Appendix B
Rate constants for added co-oxidation reactions between n-heptane (n-C7H16) and toluene (-CH3) to merged neat mechanisms
for individual fuels [16,35]

1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.

-CH2 + nc7h16 = -CH3 + c7h15-1


-CH2 + nc7h16 = -CH3 + c7h15-2
-CH2 + nc7h16 = -CH3 + c7h15-3
-CH2 + nc7h16 = -CH3 + c7h15-4
c7h15o2-1 + -CH3 = -CH2 + c7h15o2h-1
c7h15o2-2 + -CH3 = -CH2 + c7h15o2h-2
c7h15o2-3 + -CH3 = -CH2 + c7h15o2h-3
c7h15o2-4 + -CH3 = -CH2 + c7h15o2h-4
-CH2 + c7h15o2-1 = c7h15o-1 + -CH2 O
-CH2 + c7h15o2-2 = c7h15o-2 + -CH2 O
-CH2 + c7h15o2-3 = c7h15o-3 + -CH2 O
-CH2 + c7h15o2-4 = c7h15o-4 + -CH2 O

7.50E+11
5.00E+11
5.00E+11
2.50E+11
3.00E+12
3.00E+12
3.00E+12
3.00E+12
3.00E+12
3.00E+12
3.00E+12
3.00E+12

0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0

8000
7000
7000
7000
12000
12000
12000
12000
12000
12000
12000
12000

kf = AT n exp(E/RT ).
A units mol cm s K; E units cal/mol.

References
[1] S. Onishi, S.H. Jo, K. Shoda, P.D. Jo, S. Kato, Society
Automotive Engineers, SAE-790501, 1979.
[2] P.M. Najt, D.E. Foster, Society of Automotive Engineers, SAE-830264, 1983.
[3] M. Christensen, B. Johansson, P. Einewall, Society of
Automotive Engineers, SAE 97- 2874, 1997.
[4] M. Christensen, A. Hultqvist, B. Johansson, Society of
Automotive Engineers, SAE 1999-01-3679, 1999.
[5] A. Hultqvist, M. Christensen, B. Johansson, Society of
Automotive Engineers, SAE 2000-01-1833, 2000.
[6] J. Andrae, P. Bjrnbom, L. Edsberg, L.-E. Eriksson,
Proc. Combust. Inst. 29 (2002) 789795.
[7] M. Christensen, B. Johansson, A. Hultqvist, Society of
Automotive Engineers, SAE 2001-01-1893, 2001.
[8] D.L. Flowers, S. M Aceves, J. Martinez-Frias, R.W.
Dibble, Proc. Combust. Inst. 29 (2002) 687694.
[9] C.K. Westbrook, Proc. Combust. Inst. 28 (2000) 1563
1577.
[10] S. Tanaka, F. Ayala, J.C. Keck, J. B Heywood, Combust. Flame 132 (2003) 219239.
[11] J.M. Simmie, Prog. Energy Combust. Sci. 29 (2003)
599634.
[12] C.K. Westbrook, J. Warnatz, W.J. Pitz, Proc. Combust.
Inst. 22 (1988) 893901.

[13] H.J. Curran, W.J. Pitz, C.K. Westbrook, C.V. Callahan,


F.L. Dryer, Proc. Combust. Inst. 27 (1998) 379387.
[14] H.J. Curran, P. Gaffuri, W.J. Pitz, C.K. Westbrook,
Combust. Flame 114 (12) (1998) 149177.
[15] H.J. Curran, P. Gaffuri, W.J. Pitz, C.K. Westbrook,
Combust. Flame 129 (3) (2002) 253280.
[16] http://www-cms.llnl.gov/
combustion/combustion_home.html, 2004.
[17] P.A. Glaude, V. Conraud, R. Fournet, F. Battin-Leclerc,
G.M. Cme, G. Schacchi, P. Dagaut, M. Cathonnet, Energy Fuels 16 (2002) 11861195.
[18] S. Tanaka, F. Ayala, J.C. Keck, Combust. Flame 133
(2003) 467481.
[19] H.S. Soyhan, P. Amnus, F. Mauss, C. Sorusbay, Society of Automotive Engineers, SAE 1999-01-3484,
1999.
[20] H.S. Soyhan, F. Mauss, C. Sorusbay, Combust. Sci.
Technol. 174 (2002) 7391.
[21] S.S. Ahmed, G. Moreac, T. Zeuch, F. Mauss, Efficient Lumping Technique for the Automatic Generation
of n-Heptane and Iso-Octane Oxidation Mechanism,
Preprints of SymposiaAmerican Chemical Society,
Division of Fuel Chemistry, 2004, pp. 265266.
[22] S.D. Klotz, K. Brezinsky, I. Glassman, Proc. Combust.
Inst. 27 (1998) 337344.
[23] R. Ogink, PhD thesis, Chalmers Institute of Technology, Gothenburg, Sweden, 2004.

286

J. Andrae et al. / Combustion and Flame 140 (2005) 267286

[24] W.R. Leppard, Society of Automotive Engineers, SAE


902137, 1990.
[25] G.T. Kalghatgi, P. Risberg, H.-E. ngstrm, Society of
Automotive Engineers, SAE 2003-01-1816, 2003.
[26] P. Risberg, G. T Kalghatgi, H.-E. ngstrm, Society of
Automotive Engineers, SAE 2003-01-3215, 2003.
[27] G.T. Kalghatgi, R.A. Head, Society of Automotive Engineers, SAE 2004-01-1969, 2004.
[28] L. Reich, S.S. Stivala, Autoxidation of Hydrocarbons
and PolyolefinsKinetics and Mechanisms, Decker,
New York, 1969, pp. 101109.
[29] S. Zabarnick, Ind. Eng. Chem. Res. 32 (1993) 1012
1017.
[30] S. Zabarnick, Energy Fuels 12 (1998) 547553.
[31] N.J. Kuprowicz, J.S. Erwin, S. Zabarnick, Fuel 83
(2004) 17951801.
[32] W.L. Easley, A. Agarwal, G.A. Lavoie, Society of Automotive Engineers, SAE 2001-01-1029, 2001.
[33] R.J. Kee, F.M. Rupley, J.A. Miller, M.E. Coltrin, J.F.
Grcar, E. Meeks, H.K. Moffat, A.E. Lutz, G. DixonLewis, M.D. Smooke, J. Warnatz, G.H. Evans, R.S.
Larson, R.E. Mitchell, L.R. Petzold, W.C. Reynolds,

[34]
[35]
[36]
[37]
[38]
[39]

[40]
[41]

M. Caracotsios, W.E. Stewart, P. Glarborg, C. Wang, O.


Adigun, W.G. Houf, C.P. Chou, S.F. Miller, Chemkin
Collection, Release 3.7.1, Reaction Design, San Diego,
CA, 2003.
J.B. Heywood, Internal Combustion Engine Fundamentals, McGrawHill, New York, 1988, p. 43.
P. Dagaut, G. Pengloan, A. Ristori, Phys. Chem. Chem.
Phys. 4 (2002) 18461854.
K. Fieweger, B. Blumenthal, G. Adomeit, Combust.
Flame 109 (1997) 599619.
J. Herzler, L. Jerig, P. Roth, Proc. Combust. Inst. 30
(2004), in press.
A. Burcat, C. Snyder, T. Brabbs, NASA TM,
vol. 87312, 1986.
A. Hultqvist, M. Christensen, B. Johansson, M. Richter,
J. Nygren, J. Hult, M. Aldn, Society of Automotive
Engineers, SAE 2002-01-0424.
X.J. Gu, D.R. Emerson, D. Bradley, Combust.
Flame 133 (2003) 6374.
R.W. Walker, C. Morley, in: R.G. Compton, G. Hancock, M.J. Pilling (Eds.), Comprehensive Chemical Kinetics, vol. 35, Elsevier, New York, 1997.

You might also like