2nd International Conference “From Scientific Computing to Computational Engineering”

2nd IC-SCCE
Athens, 5-8 July, 2006
© IC-SCCE

ANALYTICAL MODELING OF FLAME KERNEL FORMATION AND

DEVELOPMENT IN SI ENGINES
Hamed Aghajani1 , Nasser S. Mehdizedeh1 , and Sadegh Tabejamaat1
1
Aerospace Engineering Department
Amirkabir University of Technology
Hafez Ave. No. 424 Tehran, Iran
e-mail: seraj@aut.ac.ir

Keywords: Analytical Modeling, Flame Kernel, Spark Ignition Engines.

Abstract. A detailed, one dimensional, time-depended model describing flame kernel development in spark
ignition engines is presented. This model accounts for properties of the ignition system (supplied electrical
energy and power and heat losses to electrodes and chamber walls), of the combustible mixture (pressure,
temperature, equivalence ratio, type of fuel and laminar burning velocity) and of the flow field (mean flow
velocity).

1 INTRODUCTION
Numerous experimentally data and theoretically works, treating the problem of flame kernel formation, have
been published. There were more experimental data available than theoretical work digesting them which,
additionally, was based on heavily simplifying assumptions up to 1970’s. One of the main reasons for this lack
of detailed measurements is the highly non-stationary processes in the very small flame kernel volume (0-10 mm
in diameter) [1]. Published work on flame kernel formation falls into 4 categories:
• PHENOMENOLOGY: Experiments in combustion chambers or test engines were carried out to identify
the interaction between globally measurable values (pressure, geometry, engine speed, etc.) and presumed
reasons (flow field, ignition energy, etc.).
• ELECTRICAL SPARK IGNITION: Effect of spark current and voltage on size and characteristics of
spark plasmas as well as their effects on inflammability.
• REACTION KINETICS: Influence of pressure, temperature and mixture composition on burning velocity
and influence of local losses on reaction rate and heat release.
• MODLING OF FLAME KERNEL FORMATION: Modeling of flame kernel formation requires a
simultaneous treatment of spark physics, chemical kinetics and flow and turbulence interactions.
There has been an interest in modeling the spark ignition process in recent years with the advance of
computer power and also with greater interest in reducing the fuel consumption and exhaust emissions. While
improving fuel injection strategies and in-cylinder flow are still getting the most attention as a way to reduce fuel
consumption in engines, a better understanding of the possibility to future improve the engine with more
advanced ignition systems. Some models simplified the nature of the discharge like Sloane [2-4], who in his 1-D
model increased the breakdown period to 1 µs while normally it is on the order of ten nanoseconds to achieve
acceptable results. Champion and coworkers [5] were the first to report a flame initiation model based on a
critical radius that the kernel needed to reach to be said self sustaining. The model used a one step overall
reaction and a good agreement with experimental results was found. Ko and coworkers [6] developed a similar
model for ignition in quiescent mixture. During the early growth of the flame kernel, a blast wave is proposed to
describe the kernel growth, assuming that thermal diffusion governs the kernel growth. Yashar [7] used KIVA-3
to model a spark ignition model to replace the simpler ignition model found in KIVA-3. His ignition model took
into account flowing through the discharge. The input for the ignition process is a simple current curve as a
function of time. The results showed that the chemical reaction had importance only in the last part of the
ignition process. Akram and coworkers [8-10] presented in a series of papers, two formulations and their
associated equations for a one-dimensional spark discharge. The first formulation is an Eulerian form of the
governing equations, which assumed that the plasma is a linear viscous fluid, while the second formulation is the
Lagrangian form, which assumed that the plasma is an inviscid homogeneous fluid.
 Hamed Aghajani, Nasser S. Mehdizedeh, and Sadegh Tabejamaat.

2 PHENOMENOLOGY
Electrical spark discharge is characterized with four phases, that in the following we have described them
briefly.
1) Pre-breakdown phase: As voltage is applied to the anode (Fig 1), the electric fields in its vicinity attract
drifting electrons, which are accelerated by the field. The electrons accelerate in the electric field, and
may reach high enough kinetic energy that once collides with an atom; the atom becomes ionized and
loses an electron. The region of ionized gas and electrons is called a streamer. As the streamer progresses
to the cathode with help from the electromagnetic field, the number of electrons and ions grows
exponentially. The electron density in this phase is only function of E (electric field) and streamer
diameter (Fig 2) is function of applied voltage and gap width. The affect of ambient pressure and
temperature on streamer diameter or electron concentration is not available. The chemical reactions,
taking place during the pre-breakdown phase, are non-thermal and the provided energy is accumulated as
electrostatic energy.
2) Breakdown phase: The pre-breakdown phase is followed by the breakdown itself for which three
definitions have been presented:
− Threshold process which, by applying a sufficiently strong electric field, a non-conducting
material transforms into a conductor. Accordingly, breakdown occurs instantaneously when a
streamer closes the gap[11].
− Maly defined the breakdown phase from the moment the gap is closed to the point, where the
voltage has dropped to 100 Volts. By this definition, Maly covered the closing of the gap and
also the transient electrical reaction of the ignition system [12].
− This phase starts from the moment when the gap is bridged up to the moment when thermal
equilibrium between the electron temperature and the gas temperature is achieved in the plasma
channel. The advantage of this definition is that it covers all non-equilibrium part of the
breakdown where the chemistry is both thermal and non-thermal [13].
So far, only breakdown in air has been modeled and the non-thermal mechanisms are not well defined in
the literature. Kravchik, Sher and coworkers [14] have published papers on modeling of spark channel.
The first one connected only on the breakdown phase in the spark channel. This model calculated the
temperature of the gas inside the plasma channel from the energy deposition. Then they calculated the
expansion of the plasma channel and hence the new diameter of the plasma. From this expansion they
calculated a new plasma temperature. They obtained a plasma temperature of 45000 K, which corresponds
to an expanded plasma temperature of 27000K. This temperature was then used in their spark ignition
model as an initial condition. The temperature was then be used as an initial condition in their spark
ignition model.
3) Arc Phase: This phase is characterized by Maly [12] as having a current in excess of 100mA and an
−3
electron density of 1017 e.cm . Basically, high but falling currents and low voltages characterize the arc
phase. Most of cathode erosions take place in this phase. In a definite value of spark gap current transition
from arc to glow phase occurs. The current in the gap is due to electron emission from the cathode. This
electron emission is governed by following two basic mechanisms:
− Thermionic emissions (arc phase)
− Bombardment of positive ions on the cathode (glow phase)
4) Glow phase: This phase is the final phase of the spark. During glow phase, the mechanism of the emission
of electrons from the cathode surface changes. The glow phase begins when the dominance of the
thermionic emissions of electrons ends. The bombardment of positive ions on the electrode surface now
becomes the dominant mechanism. The glow phase is characterized by low currents and high voltages
[15].

3 NOMENCLATURE
.
ρu Unburned gas density Q ht Heat loss rate to the electrodes
.
Ak Flame kernel surface area Q ch Reaction heat release rate
.
sl Laminar flame speed Q spk Electrical energy transfer rate
sp Plasma velocity Uk Internal energy of burned gas
Vk Kernel volume φ Equivalence Ratio
.
Q Net heat transfer rate to the kernel
 Hamed Aghajani, Nasser S. Mehdizedeh, and Sadegh Tabejamaat.

4 MODEL DESCRIPTON
In this study a thermodynamic system of ignition kernel is applied to analyze the flame kernel growth rate.
The system is schematically shown in (Fig 3). The following assumptions are made in the calculations [1,16,17]:
• The calculation starts at 1µs after the breakdown period. The ignition kernel is assumed to be spherical. Its
radius is assumed to be 0.5mm at this time, which is obtained from experiments [12].
• The pressure is uniform both inside outside of the kernel.
• The temperature inside the kernel is uniform.
• The general gas law is applicable.
• The flame kernel volume is much smaller than the combustion chamber volume.
• The ignition flame kernel is very thin and it separates the burned and unburned gas. The temperature and
all reactive scalars jump from their value in unburned mixture to the corresponding equilibrium values.
• For simplicity only mean values for thermodynamic properties are used.
Increase in flame kernel mass is written as:
(
m& k = ρ u Ak sl + s p ) (1)
Measurements in engines show that the flame kernels are influenced by turbulence as early as 0.1 ms after spark
onset. Laminar and turbulent velocity of plasma is shown in (Fig 4).
Using mk = ρ kVk the mass burning rate can be written as:
d ( ρ k Vk ) dρ k
= ρ u Ak (sl + s p )
dVk
= ρk + Vk (2)
dt dt dt
and utilizing the ideal gas law:
1 dρ k 1 dp 1 dTk
= − (3)
ρ k dt p dt Tk dt
By substituting equation (2) in (3):
dVk ρ u  1 dTk 1 dP 
= Ak (sl + S p ) + Vk  −  (4)
dt ρk  Tk dt P dt 
By normalizing the volume increase to the flame kernel surface area, the mean expansion velocity of flame
kernel can be obtained as:
drk ρ u V  1 dTk 1 dP 
dt
=
ρk
( )
sl + S p + k 
Ak  Tk dt
−
P dt 
 (5)

dp dT
During the ignition process, we assume ≅ 0 and ≅0.
dt dt
The energy balance for the control volume surrounded by ignition kernel surface is given by:
.
 dE  .
 dV 
Q=  − (hu ) m k + p  (6)
 dt  CV  dt  CV
Unburned gas velocity can be neglected compared with the unburned gas enthalpy.
dE
The changes in flame kernel internal energy, ( ) CV , is:
dt
 dE  d (m k k)
U dU k
 = = mk + U k m& k (7)
 dt  dt dt
Moreover, the change in volume work can be written as:
2 ρ
 dV 
p
dt
2 dr
 = 4πrk p k = 4πrk p u s l + s p
dt ρk
( ) (8)
  CV
Combining equations (7) and (8), equation (6) can be rearranged as:
ρ 
+ U k m& k − hu m& k + Ak p  u (sl + s p )
dU k
Q& spk + Q& ch − Q& ht = mk (9)
dt  ρk 
If no spark discharge energy is put into the kernel and no heat is lost by the electrodes, equation (9) becomes:
. dU a . . ρ
Q ch = m k + U a m f − hu m f + Ak p u s l (10)
dt ρk
where U a is the burned gas specific internal energy at adiabatic flame temperature.
 Hamed Aghajani, Nasser S. Mehdizedeh, and Sadegh Tabejamaat.

The mass burning rate is only related to the laminar flame velocity, by the following equation:
.
m f = ρ u Ak sl (11)
Subtracting equation (10) from equation (9) following equation is obtained:
. .  dU k dU a  . .
 . .
 ρ
Q spk − Q ht = m k  −  + U k m k − U a m f − hu  m k − m f  + 4πrk2 p u s p (12)
 dt dt    ρk
According to the assumption that the gas temperature, within the kernel, equals to the adiabatic flame
temperature, so U a = U k . Furthermore by using
. .
mk − m f = 4πrk2 ρu s p (13)
equation (12) is written as:
. .  ρ 
Q spk − Q ht = Ak  ρ u (U k − hu ) + p u  s p (14)
 ρk 
. .
Q spk − Q ht is the net electrical energy transfer, which is absorbed by the kernel and can be estimated by:
. . .
Q spk − Q ht = Q spk .η eff (15)

Electrical energy transfer efficiency ( η eff ) is affected by the plug geometry. Most of the ignition process occurs
in the glow discharge phase and η eff is relatively low ( ≈ 30% ) [18].
Finally, the plasma velocity is obtained as:
Q& totalη eff
sp =
 ρ  (16)
4πrk2  ρ u (U k − hu ) + p u 
 ρk 
The laminar flame speed is calculated from Metghalchi’s relation as [19]:
α1 −α 2 (φ −1) − β1 + β 2 (φ −1)

(
sl = s1 − s 2 (φ − s3 )
2
) T 
× 

×
Pk 2 
5 
× (1 − 2.1R ) (17)
 298   1.013 ×10 
Laminar burning velocity of methane decreases with increase of pressure and decrease of mixture equivalence
ration, for which, the trends are shown in Fig 5. In the above equation R is the residual gas fraction, T is obtained
from Eq.18, and s1 , s 2 , s 3 , α 1 , α 2 , β1 , β 2 are model coefficients listed in table 1.

Table 1: Laminar flame speed coefficients of Metghalchi and Keck [19,20]

Methane Propane Gasoline
S1 36.92 34.22 26.32
S2 140.51 138.65 84.72
S3 1.11 1.08 1.13
α1 2.18 2.18 2.18
α2 0.8 0.80 0.8
β1 0.16 0.16 0.16
β2 0.22 0.22 0.22
In equation 17, T can be obtained as:
1
1−
P  γ
(18)
T = Tk1  k1 

 Pk 2 
where Tk 1 and Pk 1 are temperature pressure of the ignition, Pk 2 is the instantaneous cylinder pressure, and γ is
the ratio of the specific heats.

5 CONCLUSION
Validation of the model has been demonstrated regarding Maly's experimental results, which are shown in
Fig 6. Ignoring the initial step, there is good agreement between the results. As mentioned before, primary flame
 Hamed Aghajani, Nasser S. Mehdizedeh, and Sadegh Tabejamaat.

kernel radius has been considered 0.5mm in our calculations. The developed zero-dimensional thermodynamic
model predicts, correctly, linear trend from the beginning to the finishing radius of flame kernel radius. In Fig 7
variation of flame kernel mass can be observed. Flame development and kernel radius variations with respect to
the pressure are shown in Fig 8. Accordingly, this simple model may be used in different engine simulation
codes for simulation of early flame kernel development during ignition delay time.

REFERENCES
[1] Herweg, R., Maly, R.R. (1992), “A Fundamental Model for Flame Kernel Formation in S. I. Engines”,
Proceedings of SAE, paper 922243.
[2] Sloane, T.M. (1989), “Ignition and Flame Propagation Modeling with an Improved Methane Oxidation
Mechanism”, J. Combustion Science and Technology, Vol.63, pp.287-313.
[3] Sloane, T.M. (1990), “Numerical Simulation of Electric Spark Ignition in Atmospheric Pressure Methane-Air
Mixtures”, J. Combustion Science and Technology, Vol.73, pp.367-381.
[4] Sloane, T.M. (1992), “Numerical Simulation of Electric Spark Ignition in Methane-Air Mixtures at Pressure
above One Atmosphere”, J. Combustion Science and Technology, Vol.86, pp.121-133.
[5] Champion, M., Deshaies, B., Joulin, G. and Kinoshita, K. (1986), “Spherical Flame Initiation: Theory Versus
Experiments for Lean Propane-air Mixtures”, J. Combustion and Flame, Vol.65, pp.319-337.
[6] Ko, Y., Arpaci, V.S., Anderson, R.W., (1991), “Spark Ignition of Propane-Air Mixtures Near the Minimum
Ignition Energy: Part I. An Experimental Study”, J. Combustion and flame, vol.83, pp. 75-87.
[7] Yaşar, O. (2001), “A New Ignition Model for Spark-Ignited Engine Simulations”, J. Parallel Computing,
Vol.27, pp.179-200.
[8] Akram, M., Lundgren, E. (1996), “The Evolution of Spark Discharges in Gases: 1. Macroscopic Models”,
Journal of Physics D: Applied Physics, Vol.29, pp. 2129-2136.
[9] Akram, M. (1996), “The Evolution of Spark Discharges in Gases: 2. Numerical Solution of One-Dimensional
Models”, J. of Physics D: Applied Physics, Vol.29, pp.2137-2147.
[10] Akram, M. (1996), “Two Dimensional Model for Spark Discharge Simulation in Air”, AIAA Journal, V.34,
N.9, pp. 1835-1845.
[11] Raizer, Y.P. (1997), Gas Discharge Physics, 2nd edition. Springer-Verlag, Germany.
[12] Maly, R., Vogel, M. (1978), “Initiation and Propagation of Flame Fronts in Lean CH4-Air Mixtures by the
Three Modes of the Ignition Spark”, Proceedings of 17th Symposium on Combustion, pp.821-831.
[13] Seers, P. (2003), Spark Ignition: An Experimental and Numerical Investigation, PHD Dissertation, the
University of Texas, Austin.
[14] Kravchik, T., Sher, E., Heywood, J.B. (1995), “From Spark Ignition to Flame Initiation”, J. Combustion
Science and Technology, Vol.108, pp.1-30.
[15] Meek, J.M., Craggs, J.D. (1953), Electrical Breakdown of Gases, Oxford at the Clarendon Press, London.
[16] Fan, L., Reitz, R.D. (2000), “Development of an Ignition and Combustion Model for Spark-Ignition
Engines”, Proceedings of SAE, Paper 2000-01-2809.
[17] Fan, L., Li, G., Han, Z. and Reitz, R.D. (199), “Modeling Fuel Preparation and Stratified Combustion in a
Gasoline Direct Injection Engine”, Proceedings of SAE, paper 1999-01-0175.
[18] Heywood, J.B. (1988), Internal Combustion Engine Fundamentals, Mc Graw-Hill, London.
[19] Metghalchi, M., Keck, J.C. (1980), “Laminar Burning Velocity of Propane-Air Mixtures at High
Temperature and Pressure”, J. Combustion and Flame, Vol.38, pp.143-154.
[20] Metghalchi, M., Keck, J.C. (1982), “Burning Velocities of Mixtures of Air with Methanol, Isooctane and
indolent at High Pressure and temperature”, J. Combustion and Flame, Vol.48, pp.191-210.

Fig 1. Electrodes schematic

 Hamed Aghajani, Nasser S. Mehdizedeh, and Sadegh Tabejamaat.

Fig 2. Streamers propagation with time

Fig 3. Thermodynamics of flame kernel system

Fig 4. Plasma laminar and turbulent flame velocity in a SI premixed engine

40
Laminar Burning Velocity(cm/s)

30

20 1
0.9

10 0.8
0.7

0
1 2 3 4 5 6
P(atm)

Fig 5. Laminar burning velocity for different equivalence ratios

 Hamed Aghajani, Nasser S. Mehdizedeh, and Sadegh Tabejamaat.

R0=0.5mm
2

R adius of Ignition K ernel m m

1.5

0.5
0 1 2 3
T(sec) -4
x 10

Fig 6. Radius of ignition kernel: left (experimental results of Maly), right (model results)
-3 M0=0.0001 gr
x 10
6
Mass of Ignition Kernel (gr)

0
0 1 2 3
T(sec) -4
x 10

Fig 7. Ignition kernel mass variation by time

R0=0.5mm
4.5

3.5
Radius of Ignition Kernel mm

2.5

2 P=1atm

1.5
P=6atm

0.5
0 1 2 3
T(sec) -4
x 10

Fig 8. Ignition kernel radius variation in time for various pressures