Combustion and Flame 191 (2018) 408–416

Contents lists available at ScienceDirect

Combustion and Flame

journal homepage: www.elsevier.com/locate/combustflame

Comparative study of laser ignition and spark-plug ignition

in high-speed flows
Takuma Endo∗, Keisuke Kuwamoto, Wookyung Kim, Tomoyuki Johzaki, Daisuke Shimokuri,
Shin-ichi Namba
Department of Mechanical Systems Engineering, Hiroshima University, 1-4-1 Kagamiyama, Higashi-Hiroshima, Hiroshima 739-8527, Japan

a r t i c l e i n f o a b s t r a c t

Article history: Laser ignition and spark-plug ignition were experimentally compared in high-speed ethylene-oxygen mix-
Received 25 September 2017 ture flows of up to approximately 100 m/s. Nd:YAG laser of 12-ns pulse duration and a semi-surface-
Revised 27 October 2017
discharge-type spark plug of 1.8-ms discharge duration were used to conduct the experiments with de-
Accepted 17 January 2018
posited energy of approximately 24 mJ in both cases. The self-emission was observed by a high-speed
camera. The flame-spread behavior and ignition ability were examined in lean-fuel conditions. The study
Keywords: findings revealed that laser ignition was superior to the spark-plug ignition in the aspect of the early-
Ignition stage rapid flame spread, although it showed lower probability of successful ignition than that by the
Laser spark plug near the lean-fuel ignitable limit. These findings suggest that the ignition in high-speed flows
Spark plug
is significantly influenced by the turbulence via the enhancement of heat transport in particular.
High-speed flow
© 2018 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
Ethylene

1. Introduction MIE, higher ignition ability of a laser-induced spark compared to

Currently, the ignition process in most of the spark-ignition
combustors is initiated by gaseous breakdown between a pair
of high-voltage-applied electrodes. Such ignition schemes involve
leaner fuel mixtures that require larger discharge energy, and
higher initial pressure that requires higher voltage to be applied
to the electrodes; both of which result in enhanced erosion of the
electrodes and shorten the lifetime of the spark plug [1].
It has been known since 1963 that gaseous breakdown can also
be realized by focusing a high-power laser beam in a gaseous
medium without a pair of electrodes [2]. In addition, the high-
power laser has been used as an ignition device in laboratory ex-
periments since the 1970s [3,4]. Subsequently, the advancement in
high-power laser technology has led to the recent emergence of
the laser-ignition systems as a new practical spark-ignition tech-
nology [5–9].
The purpose of this study is to compare the characteristics of
laser ignition to those of the conventional spark-plug ignition sys-
tems that are widely used around the world. When considering the
ignition ability of a laser-induced spark, it is known that the mini-
mum ignition energy (MIE) of the laser ignition is higher than that
of the conventional spark-plug ignition for a wide range of equiv-
alence ratios [10]. However, in some lean-fuel conditions of high
∗
Corresponding author.
E-mail address: takumaendo@hiroshima-u.ac.jp (T. Endo).
that of an electrode-induced spark has been demonstrated [10,11].
Recently, the ignition ability of a laser-induced spark was com-
pared with that of an electrode-induced spark using the same de-
posited energy into quiescent lean-fuel propane-air mixtures [12].
This study showed that the ignition ability of a laser-induced spark
was higher than that of an electrode-induced spark near the lean-
fuel ignitable limit. This behavior was attributed not only to the
lack of heat loss to electrodes but also to the large initial flame
kernel. The effective energy contained in the large initial flame ker-
nel was augmented by the rapid heat release from the combustible
mixture sucked into the kernel by a non-spherical inward flow that
was created by the laser-induced spark.
As a result of past investigations, the characteristics of laser ig-
nition in a quiescent gas mixture have become understood to some
degree. However, an actual gas mixture in a practical combustor
is not quiescent when it is ignited. For example, in a reciprocat-
ing engine, the mean and fluctuating velocities of 1 to 10 m/s are
obtained in the clearance volume at the top-center crank posi-
tion [13]. In a scramjet engine, ignition in a combustible gas flow
of approximately 20 0 0 m/s is required [14]. Apart from internal
combustion engines, a high-frequency pulse detonation combustor
developed for thermal spray requires a combustible gas flow of ap-
proximately 100 m/s to be ignited [15].
The influence of flow on the characteristics of spark ignition in-
duced by a pair of high-voltage-applied electrodes has been stud-
ied for more than half a century [16–19]. It has been reported that
both the flow speed itself and the turbulent nature of the flow

https://doi.org/10.1016/j.combustflame.2018.01.029
0010-2180/© 2018 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
 T. Endo et al. / Combustion and Flame 191 (2018) 408–416
have significant influence on the ignition characteristics. Limited
research that examines the characteristics of spark ignition induced
by laser in a flowing medium has been conducted [20–22]. Specif-
ically, to the best of the authors knowledge, no experimental study
has been conducted to compare the characteristics of laser ignition
with those of spark-plug ignition in a flowing medium using the
same deposited energy.
When the ignition of a combustible gas mixture flowing at the
speed of approximately 10 m/s is considered, the initial flame ker-
nel is moved and/or deformed several millimeters in several hun-
dred microseconds. This means that the initial flame kernel expe-
riences movement and/or deformation comparable to its own size
during its growth to the self-propagating flame. In such cases, both
the heat loss to electrodes in a conventional spark-plug ignition
and the flow driven by a laser-induced spark must be largely dif-
ferent from those in quiescent media cases. Therefore, it is impor-
tant to clarify the characteristics of laser ignition in flowing media
in comparison with those of spark-plug ignition using the same
deposited energy.
At the end of Introduction, the difference between laser ig-
nition and spark-plug ignition is briefly described. Basically, a
gaseous medium is ignited by thermal and/or chemical effects of
a tiny plasma, which is usually called a spark. In laser ignition, a
spark is created in a gaseous medium by laser-induced breakdown.
For realizing the laser-induced breakdown in a gaseous medium,
the laser power density of the order of more than 1011 W/cm2
is needed when the duration of a laser pulse is longer than a
nanosecond. This breakdown threshold in a gaseous medium is
usually much higher than that on the surface of a solid material.
Therefore, the laser beam should be focused apart from the sur-
face of the glass window through which the optical access of the
laser beam to the combustion chamber is realized. Otherwise, the
surface of the glass window is damaged. In some cases, the laser
beam is focused on the surface of the inner wall of the combus-
tion chamber, and a spark is created on the solid surface. In this
laser-ignition scheme, the laser power density needed for creating
a spark is much lower than that in the case of the laser-induced
breakdown in a gaseous medium because of much lower break-
down threshold. However, the laser-irradiated solid surface is ab-
lated and a hole is bored there. Accordingly, this scheme is usually
considered different from the ordinary laser ignition. Another char-
acteristic of laser ignition forced by the high breakdown threshold
is the short pulse duration. Because the size of the focal spot of
a laser is usually of the order of 10 μm, the duration of a laser
pulse, whose energy is of the order of 10 mJ, needs to be less than
the order of 100 ns for realizing the laser-induced breakdown in
a gaseous medium. On the other hand, in spark-plug ignition, a
spark is created between a pair of electrodes when the electrical
voltage applied to the electrodes is higher than a certain thresh-
old. Therefore, the spark is created essentially near and on the
solid surface. In spark-plug ignition, the duration of the electrical
pulse applied to the electrodes is arbitrary in principle. However, if
the pulse duration is too short, the electrical current becomes too
large, and accordingly the lifetime of the electrodes becomes too
short due to erosion. Therefore, the duration of the electrical pulse
applied to the electrodes is usually of the order of a millisecond in
the practically-used spark plugs. In summary, the duration of the
energy deposition in laser ignition is shorter than that in spark-
plug ignition more than four orders in magnitude, and a laser-
induced spark is created apart from any solid surface while a pair
of high-voltage-applied electrodes creates a spark near and on the
electrodes.
In this study, the characteristics of laser ignition and conven-
tional spark-plug ignition were compared in high-speed flows of
up to approximately 100 m/s using the same deposited energy.
Ethylene–oxygen gas mixtures were used with a geometrical ar-
409
rangement similar to that of a high-frequency pulse detonation
combustor [15]. The advantage and disadvantage of laser igni-
tion in comparison with conventional spark-plug ignition were dis-
closed. The study findings revealed that the laser ignition exhib-
ited the advantage of rapid flame spread; however, it showed lower
probability of successful ignition than that by the spark plug near
the lean-fuel ignitable limit.
2. Experimental arrangement
Figure 1(a) shows a general view of the experimental setup. The
combustor was a straight rectangular channel the cross-sectional
area of which was the same as that of a circular tube, whose in-
ner diameter was 10 mm, which was the combustor of a high-
frequency pulse-detonation thermal-spray gun [15]. This geomet-
rical arrangement was adopted because of its simpleness. Ethylene
and oxygen were mixed at the right end of the rectangular channel
as head-on colliding jets from nozzles of 3 mm in diameter. The
flow rates of ethylene and oxygen were independently controlled
by both of their supplying pressures and orifice diameters. These
parameters were selected based on the data obtained from the
preliminary experiments in which the flow rates of ethylene and
oxygen were measured as functions of their supplying pressures
and orifice diameters. The preliminary experiments confirmed that
the total flow rate from both gas-supplying pipes was the sum of
the flow rates from the respective gas-supplying pipes under some
conditions. Nitrogen was used to purge the residual gas from the
combustor. In the ignition experiments, the magnetic valves began
to close 5 ms after the beginning of the breakdown so that the ig-
nition and subsequent flame spread occurred in a high-speed flow.
Figure 1(b) shows the experimental setup of the spark-plug ig-
nition. The spark plug used in the experiments was the semi-
surface-discharge type (NGK R847-11), which was driven by an au-
tomotive spark-plug driver: 90919-02240 (TOYOTA). As shown in
Fig. 1(b), only the center electrode of the spark plug, whose di-
ameter and thickness were respectively 1.6 mm and 0.6 mm, pro-
truded from the inner surface of the flow channel. The tempo-
ral waveforms of voltage and current of the electrical spark were
measured in the quiescent atmospheric air in advance using a
high-voltage probe: P6015A (Tektronix) and a current probe: CT-
B0.5 (Magnelab), respectively. The deposited energy was calculated
to be approximately 24 mJ, where the duration of the discharge
was approximately 1.8 ms. In the ignition experiments, the spark
plug was driven under the same conditions as that of the pre-
liminary measurements carried out in the quiescent atmospheric
air, although the temporal waveforms of voltage and current of
the electrical spark in the high-speed gas flows were rather dis-
turbed compared with those in the case of the quiescent atmo-
spheric air. The observed disturbances might be attributed to the
multiple discharges due to the high-speed gas flows [17]. In fact,
self-emission images showing the rebuilding of the discharge path
were observed in some cases, as shown in Fig. 2(b).
Figure 1(c) shows the laser ignition experimental arrangement.
The laser system used in the experiments was a multimode Q-
switched Nd:YAG laser: model LS-2131M-10 (LOTIS TII) with pulse
duration of approximately 12 ns [23]. The laser beam diameter was
expanded from 5 to 40 mm before focusing. An aspherical plano-
convex lens with a focal length of 37.5 mm was used for focusing
the laser beam at the center of the channel. Based on the past ex-
perimental data [23], the effective diameter of the focal area was
estimated to be approximately 12 μm as it was assumed to be pro-
portional to the focusing F number. The absorption efficiency of
the laser pulse was measured in advance as a function of the in-
cident energy, where the channel was filled with ethylene-oxygen
gas mixtures of various equivalence ratios and flow speeds. In the
ignition experiments, the incident energy of the laser pulse was
410 T. Endo et al. / Combustion and Flame 191 (2018) 408–416
Fig. 1. Schematic of experimental setup. (a) General view. (b) Spark-plug ignition. (c) Laser ignition.
measured using a wedge beam splitter and an energy meter: PE
25-C (OPHIR). A half-wave retarder and polarizing beam splitter
were used to adjust the incident energy of the laser pulse to be
31 mJ so that the absorbed energy was 24 mJ.
A high-speed camera: FASTCAM SA4 (Photron) was used in the
ignition experiments to observe the self-emitted visible radiation
from the reacting and/or heated gas. The observation was carried
out at the frame rate of 5 × 104 fps (frame per second) in order to
quantify the behavior of the ignition and subsequent flame spread.
On the other hand, the frame rate of 5 × 103 or 1 × 104 fps was
used for the observations to assess the probability of successful ig-
nition. The successful ignition was defined as the case in which the
flame propagated along the whole combustion chamber. Multiple
trials were carried out to determine the probability of successful
ignition for each experimental condition. The number of trials was
40 except for the cases of “almost quiescent gas” explained later,
in which the number of trials was 10.
3. Results
Figure 2 shows the observed self-emission images, where the
origin of the time axis corresponds to the beginning of the break-
down, Uav is the average flow speed of the unburned gas, φ is
the equivalence ratio, FL shows the result of laser ignition, and SP
shows the result of spark-plug ignition. In this study, “medium-
speed cases” correspond to the experimental conditions with Uav =
40 − 46 m/s and “high-speed cases” correspond to the experimen-
tal conditions with Uav = 82 − 114 m/s, although both cases fall
under the high-speed flows category considering the movement
and/or deformation of an initial flame kernel as explained in the
Section 1 “Introduction”. Additionally, “almost quiescent cases” are
the cases in which ethylene and oxygen were supplied to the
combustor under the conditions of medium-speed flow rates for
the designated equivalence ratio first, then the magnetic valves
were closed 150 ms before the ignition. Thus, the “almost quies-
cent cases” are the cases that involved some remaining turbulence
although the average flow speed was zero.
It should be noted that the time range shown in Fig. 2(a) is one
order of magnitude larger than those in Fig. 2(b) and (c). Specif-
ically, the flame spread in the flowing gas was much more rapid
than that in the “almost quiescent cases”. In addition, the flame
spread in the case of the laser ignition occurred earlier than that
in the case of the spark-plug ignition at the same flow speed. In
order to quantify the behavior of the flame spread, one half of the
emitting region width: (xR − xL )/2 was plotted against time, whose
origin corresponds to the beginning of the breakdown, in Fig. 3. As
shown in Fig. 3, in the flowing gas, the flame spread was more
 T. Endo et al. / Combustion and Flame 191 (2018) 408–416
Fig. 2. Self-emission images. (a) Almost quiescent cases. (b) Medium-speed cases.
(c) High-speed cases. FL and SP stand for “Focused Laser” and “Spark Plug”,
respectively.
rapid in the laser ignition than in the spark-plug ignition. This
characteristic presents the advantage of laser ignition in flowing
gas. Figure 3(a), which shows the results in the almost quiescent
cases, expresses the larger burned-gas region in the laser ignition
compared with that in the spark-plug ignition. This is qualitatively
consistent with the past experiment [12].
411
Fig. 3. Flame-spread behaviors in almost quiescent cases (a), medium-speed cases
(b), and high-speed cases (c).
Fig. 4. Probability of successful ignition.
Fig. 5. Propagation speed Su,L and thickness δ LF of the stationary laminar flame.
Figure 4 shows the probability of successful ignition as a func-
tion of the equivalence ratio φ . The laser ignition showed lower
probability of successful ignition compared with the spark-plug ig-
nition in flowing gas near the lean-fuel ignitable limit, which rep-
resents the disadvantage of laser ignition in flowing gas.
4. Discussion
Initially, the propagation speed and the thickness of the sta-
tionary laminar flame for the combustible gas mixtures used in
the experiments were calculated using the Flame Speed model
of Chemkin-Pro [24] and the reaction model GRI-mech 3.0 [25].
Figure 5 shows the results of the calculations, where Su, L and δ LF
denote the propagation speed and the thickness of the station-
412 T. Endo et al. / Combustion and Flame 191 (2018) 408–416
Fig. 6. Flame-propagation speeds in almost quiescent cases (a), medium-speed
cases (b), and high-speed cases (c).
ary laminar flame, respectively. For the calculation of δ LF , used
was the formula δLF = (Tb − Tu )/|∇ T |max in which Tb and Tu were
the temperatures of the burned and unburned gases, respectively,
and |∇ T|max was the maximum magnitude of the temperature
gradient.
In order to quantitatively discuss the behavior of the flame ac-
celeration, the flame propagation speeds are deduced from the re-
sults shown in Fig. 3 assuming quasi-steady planar flame propa-
gation. When a flame propagates at the speed Su relative to the
unburned gas with mass density of ρ u , the mass of the burned gas
created per unit area per unit time is ρ u Su , which corresponds to
the volume of vb ρu Su = (ρu /ρb )Su , where vb = ρb −1 is the specific
volume of the burned gas. In the present experiments, assuming a
pair of planar combustion waves propagating along the axis of the
combustor in the opposite directions, d(xR − xL )/dt corresponds to
2(ρ u /ρ b )Su , and their relation can be expressed as follows:
ρu 
d xR − xL
 of that in the spark-plug ignition after the beginning of the break-
Su = . (1)
ρb dt 2 down. The fact that the duration of the laser pulse (approximately
In fact, as shown in Fig. 2, the flame did not propagate at the same
velocity in both directions even if the flow speed of the unburned
gas was taken into account. Therefore, the speed Su defined by
Eq. (1) is the representative speed characterizing the flame-spread
behavior. Considering the conservation laws of mass, momentum,
and energy in the coordinate frame moving with the combustion
wave in steady state, the density ratio between the unburned and
burned gases can be written as follows:
1     
2 γb 2 γb 2 1
ρu γb −1
2
γu + M u − γb −1 γu + M u
2 2
− 4 γγb +1 Mu 2 γu2−1 + Mu 2 + 2 aqu 2
b −1
= ,
ρb 2 γγb +1 Mu 2
b −1
where γu , au , γb , q, Mu = Su /au denote the specific-heat ratio of
the unburned gas, the sound speed of the unburned gas, the ef-
fective specific-heat ratio of the burned gas, the specific heat re-
lease by combustion reduced to the absolute zero of temperature,
and the propagation Mach number of the combustion wave, re-
spectively. The values of γ u and au can be calculated easily if the
perturbation by the precursor compression waves is ignored. The
value of γ b can be estimated using NASA CEA [26] assuming iso-
baric combustion. The value of q can be obtained from the results
of Su, L and ρ u /ρ b of the laminar-flame calculations stated above re-
Fig. 7. Evaluation of the size and position of the initial flame kernel in a laser-
ignition case (a) and a spark-plug ignition case (b).
garding Fig. 5 using the following equation.
  S 2 ρ
au 2 2γb 1 u,L u γb + 1  Su,L 2
q= + −
2 γb − 1 γu au ρb γb − 1 au
 ρ 2  S 2
u 2 u,L
× − + (3)
ρb γu − 1 au
Namely, Eq. (2) shows that the density ratio ρ u /ρ b is a function of
the propagation speed of the combustion wave Su . Accordingly, the
left-hand side of Eq. (1) is a function of Su only, and the results
shown in Fig. 3 can be converted to Su by solving Eq. (1) numer-
ically. Figure 6 shows the results of the data reduction. Note that
in Fig. 6(a), the ordinate is one order of magnitude smaller and
the abscissa is one order of magnitude larger than Fig. 6(b) and
(c). As shown in Fig. 6(b) and (c), the flame-propagation speeds in
the flowing gas were much faster than the corresponding laminar-
flame speeds, and the flames in the flowing gas were accelerated
in a significantly early stage. As shown in Fig. 6(b) and (c), the be-
havior of the flame acceleration is similar between the laser igni-
tion and the spark-plug ignition once the significant flame accel-
eration begins. However, the time at which the significant flame
acceleration begins is earlier in the laser ignition than the spark-
plug ignition. Specifically, the time at which the significant flame
acceleration begins in the laser ignition was approximately a half
12 ns) was five orders of magnitude shorter than that of the spark-
plug discharge (approximately 1.8 ms) is thought to be one of the
major causes of the early flame acceleration observed in the laser
ignition. The shorter duration of the energy deposition in the laser
ignition results in the faster development of the initial flame ker-
nel. As shown in Fig. 2, the burning region in the laser ignition
spreads faster and reaches the side wall of the channel earlier. Af-
ter the burning region spreads over the whole cross-section of the
channel, the flame is significantly stretched by the gradient of the

(2)
mean flow velocity, and the rapid flame acceleration begins. Al-
though the beginning of the flame acceleration in the spark-plug
ignition cases was delayed compared with that in the laser igni-
tion cases, it should be noted that the flame acceleration occurred
in the spark-plug ignition cases in significantly shorter time scale
than the discharge duration of approximately 1.8 ms.
In order to discuss the influence of flow on the flame-kernel de-
velopment, the size and position of the initial flame kernel are esti-
mated using the early-stage images obtained in the “almost quies-
cent cases”, as shown in Fig. 7. The influences of the turbulent flow
 T. Endo et al. / Combustion and Flame 191 (2018) 408–416
in the channel are discussed for the initial flame kernels shown in
Fig. 7.
First, the Reynolds number (Re) of the unburned-gas flow is
calculated based on the hydraulic diameter [27] of the rectangu-
lar channel used in the experiments: DH = 4A/L = 8.86mm, where
A is the cross-sectional area of the channel and L is the wetted
perimeter of the cross-section of the channel. Hereafter, the in-
fluence of the turbulence on the flame-kernel development is dis-
cussed based on the replacement of the actual rectangular channel
by the circular pipe with an inner diameter of DH = 8.86mm. The
calculated values of Re are as follows.
 where Re∗ = u∗ DH /νu was the turbulent Reynolds number and the
Re = (2.4 − 2.8) × 104 for the medium − speed cases
(4)
Re = (5.0 − 6.8) × 104 for the high − speed cases
According to Ref. [28], flow in a channel is, in general, turbulent
when Re > 5 × 103 . Although the entrance length of approximately
40DH is required to form a fully developed turbulent flow when
the fluid flows smoothly into the channel [28], the flow in the
channel is considered as fully developed turbulent flow in the en-
tire region because the ethylene and oxygen are mixed as head-
on colliding jets at the entrance of the channel in the present
experiments.
Based on the calculated values of Re, the Fanning friction factor
[28]: f = |τw |/( 12 ρuUav 2 ), where |τ w | is the magnitude of the shear
stress on the wall and ρ u is the mass density of the unburned gas,
can be calculated using  the Prandtl’s universal
 law of friction for
smooth pipes [27]: 1/ 4 f = 2.0log10 (Re 4 f ) − 0.8. The calculated
values of f are as follows.

f = (6.0 − 6.2) × 10−3 for the medium − speed cases
(5)
f = (4.9 − 5.2) × 10−3 for the high − speed cases

Using the calculated values of f, the friction velocity u∗ = |τw |/ρu
[27] can be calculated as follows.
 region. The temperature increase at the stagnation point was esti-
u∗ = 2.2 − 2.5m/s for the medium − speed cases
(6)
u∗ = 4.2 − 5.6m/s for the high − speed cases
The friction velocity u∗ indicates the order of magnitude of the ve-
locity fluctuation in the turbulent flow [29,30]. The friction veloc-
ity u∗ has the same order of magnitude as the calculated laminar-
flame speed Su, L shown in Fig. 5. The dissipation rate of the kinetic
energy of the turbulent motion per unit mass ɛ was evaluated by
the relation: ε = 0.202k3/2 /DH , where k = (3/2 )u∗ 2 was the rep-
resentative kinetic energy of the turbulent motion per unit mass.
The factor of 0.202 was quoted from Ref. [31]. As a result, ɛ/cp, u ,
where cp, u was the specific heat at constant pressure of the un-
burned gas, was estimated to be less than 10 K/s. In the present
experiments, the transit time of the flowing gas through the ex-
perimental apparatus was shorter than 0.1 s. Therefore, the tem-
perature increase due to the dissipation of the turbulent kinetic
energy is negligible. In addition, as shown in Fig. 4, the probabil-
ity of successful ignition did not vary sharply near the ignitable
limit as the equivalence ratio varied in high-speed flow. The lo-
cal equivalence ratio at the ignition point possibly fluctuated with
time due to turbulence. As a result, the local equivalence ratio at
the ignition point at the instant of ignition might not be exactly
the same as the global equivalence ratio. Consequently, the proba-
bility of successful ignition obtained from 40 trials became scatter-
ing as shown in Fig. 4. Estimating the time scale of the fluctuation
of the local equivalence ratio at the ignition point by DH /u∗ , the re-
sult is a few milliseconds. Therefore, taking account of the lifetime
of the created plasma, the results of the probability of successful
ignition can scatter to some degree. Using the calculated values of
u∗ , the total thickness of the viscous sublayer and the buffer layer
can be obtained by [28]: 30(ν u /u∗ ), where ν u is the kinematic vis-
413
cosity of the unburned gas, and is calculated as follows.

30(νu /u∗ ) = 0.18 − 0.20mm for the medium − speed cases
30(νu /u∗ ) = 0.079 − 0.10mm for the high − speed cases
(7)
When these values are compared with the size and position of the
initial flame kernel shown in Fig. 7, the initial flame kernel can
be considered as created in the turbulent core. Finally, the Kol-
mogorov microscale η and the Taylor microscale λ were calculated
using η = 1.28Re∗ −3/4 DH and λ = 6.36Re∗ −1/2 DH , respectively [32],
factors of 1.28 and 6.36 were quoted from Ref. [33]. The calculated
results are as follows.

Re∗ = (1.3 − 1.6) × 103 for the medium − speed cases
(8)
Re∗ = (2.5 − 3.4) × 103 for the high − speed cases

η = 46 − 51μm for the medium − speed cases
(9)
η = 25 − 32μm for the high − speed cases

λ = 1400 − 1600μm for the medium − speed cases
(10)
λ = 970 − 1200μm for the high − speed cases
It is expected that dissipative eddies consist of vortex tubes with a
diameter of the order of the Kolmogorov microscale η and a length
or spacing of the order of the Taylor microscale λ [34,35]. In the
case of the present experiments, the relation: η < δ LF < λ holds as
shown in Eqs. (9) and (10) and Fig. 5. In this case, the flame is
categorized as the “severely wrinkled flame” according to the text-
book by Glassman [36].
In Fig. 2(b) and (c), it can be seen that the flame spread to the
upstream was faster than that to the downstream in some time
mated to be less than 10 K for the incoming unburned gas flow.
Therefore, there seems no possibility of auto ignition by stop-
ping of the incoming flow by the flame. Although the detailed
mechanism for this phenomenon is not clear at present, the fol-
lowing qualitative explanation may be possible. The rapid flame
spread drives compression waves toward both the downstream and
the upstream. In the downstream, a large exhaust chamber ex-
ists near the observed region as shown in Fig. 1(a). Accordingly,
the compression waves propagating downstream decay in the ex-
haust chamber by the diverging effect, and rarefaction waves prop-
agate from the exhaust chamber toward the combustion region.
As a result, the compression of the unburned gas ahead of the
flame propagating downstream may be negligible. On the other
hand, in the upstream, the pressure release occurring in the down-
stream does not occur, and therefore the effects of the compres-
sion waves accumulate to some degree in the unburned gas ahead
of the flame propagating upstream. By this compression, the flame
propagating upstream becomes gradually thinner compared with
the flame propagating downstream. The present experimental con-
ditions correspond to the parameter region of “severely wrinkled
flame.” When the severely wrinkled flame becomes thinner, it be-
comes easier for the unburned gas pockets to be brought into the
burned gas. Consequently, the micro explosions of the unburned
gas pockets behind the flame front occur in the upstream side
more frequently than the downstream side. By these micro ex-
plosions, the rapid extension of the combustion region to the up-
stream begins in some time region.
The experimentally observed flame speeds were not saturated
yet. As a reference, the Chapman–Jouguet deflagration speed Su, CJ ,
which is the upper limit of the propagation speed of a quasi-
steady flame accelerated smoothly, is calculated by the following
414 T. Endo et al. / Combustion and Flame 191 (2018) 408–416

formula:



+ −
Su,CJ = au −
2 2
  2   (11)
γb − γu 2 γb − 1 q γb
= + , = ,
γu (γu − 1 ) au 2 γu
as Su,CJ = 51.6 m/s for the case shown in Fig. 6(b) and Su,CJ =
52.4 m/s for the case shown in Fig. 6(c). The obtained values of
Su, CJ are slightly higher than, but remarkably close to, the ex-
perimentally observed flame speeds. In the calculations of Su, CJ ,
the values of au under the conditions of 1 atm and 293.15 K
were used ignoring the perturbation by the precursor compres-
sion waves. Therefore, the calculated values of Su, CJ may rather
be underestimations. However, it was suggested that the ob-
served flames were accelerated by turbulence up nearly to the
upper limit of the propagation speed of a quasi-steady flame
accelerated smoothly. Although the deflagration-to-detonation
transition is difficult to occur in the present experiments because
of small equivalence ratio, the observed flame speeds almost cor-
respond to the final stage of the flame acceleration before the
deflagration-to-detonation transition.
Next, Prandtl’s mixing-length theory is used to calculate the Fig. 8. Evaluated spatial distributions of ux , dux /dy, αT /αL for the cases when
transport coefficients and the average velocity distribution of the φ = 0.216 and Uav = 40 m/s (a) and when φ = 0.205 and Uav = 89 m/s (b).
turbulent flow in the pipe [29,37]. Using the Prandtl’s mixing
length l, the eddy kinematic viscosity ν T is expressed as νT =
l 2 |dux /dy|, where ux (y ) is the ensemble average of the flow veloc- In addition to the velocity gradient, the thermal diffusivity is
ity along the axis of the pipe and y is the distance from the side critical for the initial flame-kernel development. Based on the as-
wall of the pipe. The Prandtl’s mixing length l is empirically given sumption that the turbulent Prandtl number equals unity, the eddy
as a function of y by the following formula for fully-developed tur- thermal diffusivity α T is given as αT = l 2 |dux /dy| [37]. On the other
bulent flows in smooth pipes [27]. hand, according to the study by Cardin et al. [20], the heat loss
  2  4  
from an initial flame kernel in a laminar case is governed by the
y y DH following thermal diffusivity α L :
l = RH 0.14 − 0.08 1 − − 0.06 1 − RH =
RH RH 2 λb
αL = , (16)
(12) c p,b ρu
In order to connect l with ux (y ), the balance of the forces acting where λb , cp, b , andρ u are the thermal conductivity of the burned
on a control volume of a thin disk with a radius of r, thickness of
, gas, the specific heat at constant pressure of the burned gas, and
located at the center of a circular pipe, is written as π r 2 |d p/dx|
= the mass density of the unburned gas, respectively. Thus, the effect
2π r
|τ |, where |τ | is the magnitude of the shear stress at radius r. of the turbulence on the development of an initial flame kernel
Similarly, the balance of the forces acting on a control volume of a via heat loss can be evaluated using the ratio α T /α L , the evaluated
thin disk with a radius of RH , thickness of
, located at the center values of which are shown in Fig. 8.
of a circular pipe, is written as π RH 2 |d p/dx|
= 2π RH
|τw |, where Figure 7 shows that the initial flame kernel is created in the
|τ w | is the magnitude of the shear stress on the side wall of the region of approximately 0.5 ≤ y/RH ≤ 1 in the laser ignition and
pipe. Based on these force-balance relations, the following relation 0 ≤ y/RH ≤ 0.15 in the spark-plug ignition, where RH = 4.43 mm.
is obtained considering y = RH − r. Thus, the effects of the velocity gradient and heat loss on the de-
r y
  velopment of initial flame kernel will be discussed based on the
|τ | = |τw | = |τw | 1 − (13) above-mentioned values on the positions of the initial flame ker-
RH RH
nels and the evaluation results plotted in Fig. 8 where Fig. 8(a) and
Therefore, from the above relation and the expression of the shear
(b) correspond to the typical conditions of “medium-speed cases”
stress: |τ | = ρu νT |dux /dy| = ρu l 2 (dux /dy )2 , the following equation
and “high-speed cases”, respectively.
is finally obtained assuming that dux /dy > 0.

 
As shown in Fig. 4, the ignition ability of the laser was slightly
dux u∗ y y inferior to that of the spark plug in “medium-speed cases”, and
= 1− 0≤ ≤1 (14)
dy l RH RH the former was considerably inferior to the latter in “high-speed
cases”. In general, an initial flame kernel in a high-speed flow may
Based on the above discussion, the velocity distribution in the pipe
be extinguished by excessive stretching of the flame kernel as a
is calculated by the following formulas.
⎧ result of excessive velocity gradient or by excessive heat loss from
⎨ux =
u∗ 2
νu y for 0 ≤ y ≤ 5 νuu∗ the flame kernel as a result of excessive thermal diffusivity. As
 u∗  
ux = 5u∗ ln νu y + 1 − ln (5 ) for 5 νuu∗ < y ≤ 30 νuu∗ (15) shown in Fig. 8, the velocity gradient in the region corresponding
⎩u = u |   
for 30 νuu∗ < y ≤ RH to the initial flame kernel created by the spark plug is much larger
y dux
x x y=30 νu + 30 νu dy
dy
u∗ u∗ than that in the region corresponding to the initial flame kernel
The first row in Eq. (15) shows the velocity distribution in the vis- created by the laser. However, the ignition ability of the spark
cous sublayer [28], the second row shows the velocity distribution plug was superior to that of the laser in both of “medium-speed
in the buffer layer [28], and the third row shows the velocity dis- cases” and “high-speed cases”. This means that the large velocity
tribution in the turbulent core. The evaluated velocity distributions gradient near the wall did not have critical influences on the de-
are shown in Fig. 8. velopment of the flame kernels in the present experiments. In
 T. Endo et al. / Combustion and Flame 191 (2018) 408–416
contrast, as shown in Fig. 8, the ratio of the thermal diffusivi-
ties α T /α L in the region corresponding to the initial flame ker-
nel created by the spark plug is smaller than that in the region
corresponding to the initial flame kernel created by the laser in
both of “medium-speed cases” and “high-speed cases”. Further-
more, the difference in the ratio α T /α L between the regions cor-
responding to the initial flame kernels created by the spark plug
and the laser is more remarkable in “high-speed cases”. These are
consistent with the behavior of the experimentally-observed prob-
ability of successful ignition shown in Fig. 4. Accordingly, the heat
loss from the laser-created initial flame kernel was significantly en-
hanced by the turbulence, and therefore the probability of success-
ful ignition achieved by the laser ignition was low compared with
that obtained by the spark-plug ignition especially in “high-speed
cases”. In fact, in the cases of ignition failure, the flame kernel
could not be observed at all although the initial plasma created
by the laser or the spark plug could be observed. That is, although
the sensitivity and spatiotemporal resolutions of the camera were
not perfect, Pign = Pker in the present experiments, where Pign de-
notes the probability of establishing the whole flame and Pker de-
notes the probability of initiating a kernel [38]. In other words, the
extinctions occurred just after the energy deposition, and accord-
ing to the terminology by Mastorakos [39], the short mode of ig-
nition failure was observed in the present experiments. The ob-
served short mode of ignition failure supports the above interpre-
tation that the ignition failure was governed by not the velocity
gradient near the side wall but the heat loss enhanced by turbu-
lence, because the flame stretch by the velocity gradient near the
side wall could occur only after initiating a kernel.
5. Conclusion
The process of ignition and subsequent flame spread in high-
speed flows of up to approximately 100 m/s was experimentally
investigated using laser ignition and spark-plug ignition with the
same deposited energy in both cases. The laser ignition was su-
perior to the spark-plug ignition in the aspect of rapid spread
of initial flame kernel. The most plausible reason for this advan-
tage of the laser ignition in high-speed flows is the short energy-
deposition duration by the laser, which was five orders of magni-
tude shorter than that by the spark plug. Specifically, the energy-
deposition power by the laser was five orders of magnitude greater
than that by the spark plug. However, the detailed mechanism
of this flame-spread behavior has not yet been fully understood.
On the other hand, the laser ignition showed lower probability of
successful ignition compared to the spark-plug ignition near the
lean ignitable limit. This disadvantage of the laser ignition in high-
speed flows can be attributed to the fact that unlike the spark-
plug ignition, the initial flame kernel created by the laser was apart
from the wall of the flow channel. Consequently, the heat loss from
the initial flame kernel created by the laser was more enhanced by
the eddy thermal diffusivity in a turbulent flow. In addition, the
difference in the plasma composition between the laser-induced
breakdown and the spark-plug-induced breakdown may have some
influences on the flame kernel evolution. However, the atomic and
chemical processes in the laser-produced plasma whose peak tem-
perature is several million kelvins are beyond the scope of this
work.
Acknowledgments
This work was supported by JSPS KAKENHI Grant number
JP17H03482; and Paloma environmental technology development
foundation. The authors are grateful to Yuhei Takenaka for his as-
sistance in laser operation during the experiments.
415
Supplementary materials
Supplementary material associated with this article can be
found, in the online version, at doi:10.1016/j.combustflame.2018.01.
029.
References
[1] J.D. Dale, M.D. Checkel, P.R. Smy, Application of high energy ignition systems
to engines, Prog. Energy Combust. Sci. 23 (1997) 379–398.
[2] P.D. Maker, R.W. Terhune, C.M. Savage, Optical third harmonic generation, in:
P. Grivet, N. Bloembergen (Eds.), 3rd International Congress on Quantum Elec-
tronics, Dunod, Paris, 2, Columbia University Press, New York (1964), pp.
1559–1576.
[3] J.H. Lee, R. Knystautas, Laser spark ignition of chemically reactive gases, AIAA
J. 7 (1969) 312–317.
[4] F.J. Weinberg, J.R. Wilson, A preliminary investigation of the use of focused
laser beams for minimum ignition energy studies, Proc. R. Soc. Lond. A 321
(1971) 41–52.
[5] P.D. Ronney, Laser versus conventional ignition of flames, Opt. Eng. 33 (1994)
510–521.
[6] T.X. Phuoc, Laser-induced spark ignition fundamental and applications, Opt.
Laser. Eng. 44 (2006) 351–397.
[7] J. Tauer, H. Kofler, E. Wintner, Laser-initiated ignition, Laser Photon. Rev. 4
(2010) 99–122.
[8] M.H. Morsy, Review and recent developments of laser ignition for inter-
nal combustion engines applications, Renew. Sust. Energ. Rev. 16 (2012)
4849–4875.
[9] G. Dearden, T. Shenton, Laser ignited engines: progress, challenges and
prospects, Opt. Expr. 21 (2013) A1113–A1125.
[10] C. Xu, D. Fang, Q. Luo, J. Ma, Y. Xie, A comparative study of laser ignition
and spark ignition with gasoline-air mixtures, Opt. Laser Technol. 64 (2014)
343–351.
[11] J.D. Dale, P.R. Smy, R.M. Clements, Laser ignited internal combustion engine –
an experimental study, SAE Paper 780329, SAE, (1979).
[12] T. Endo, Y. Takenaka, Y. Sako, T. Johzaki, S. Namba, D. Shimokuri, An exper-
imental study on the ignition ability of a laser-induced gaseous breakdown,
Combust. Flame 178 (2017) 1–6.
[13] J.B. Heywood, Internal combustion engine fundamentals, McGraw-Hill, New
York, USA, 1988 Ch. 9.
[14] S. Brieschenk, S. O’Byrne, H. Kleine, Laser-induced plasma ignition studies in a
model scramjet engine, Combust. Flame 160 (2013) 145–148.
[15] T. Endo, R. Obayashi, T. Tajiri, K. Kimura, Y. Morohashi, T. Johzaki, K. Matsuoka,
T. Hanafusa, S. Mizunari, Thermal spray using a high-frequency pulse detona-
tion combustor operated in the liquid-purge mode, J. Thermal Spray Technol.
25 (2016) 494–508.
[16] I. Kimura, S. Kumagai, Spark ignition of flowing gases, J. Phys. Soc. Jpn. 11
(1956) 599–604.
[17] C.C. Swett Jr., Spark ignition of flowing gases, NACA-TR-1287, Lewis Flight
Propulsion Lab, Cleveland, OH, United States, 1956.
[18] D.R. Ballal, A.H. Lefebvre, The influence of flow parameters on minimum
ignition energy and quenching distance, Symp. (Int.) Combust. 15 (1975)
1473–1481.
[19] T. Horstmann, W. Leuckel, B. Maurer, U. Maas, Influence of turbulent flow con-
ditions on the ignition of flammable gas/air-mixtures, Process Saf. Prog. 20
(2001) 215–224.
[20] C. Cardin, B. Renou, G. Cabot, A.M. Boukhalfa, Experimental analysis of laser-in-
duced spark ignition of lean turbulent premixed flames: New insight into ig-
nition transition, Combust. Flame 160 (2013) 1414–1427.
[21] J. Griffiths, M.J.W. Riley, A. Borman, C. Dowding, A. Kirk, R. Bickerton, Effect
of flow velocity and temperature on ignition characteristics in laser ignition of
natural gas and air mixtures, Opt. Laser. Eng. 66 (2015) 132–137.
[22] D. Bradley, C.G.W. Sheppard, I.M. Suardjaja, R. Woolley, Fundamentals of high-
-energy spark ignition with lasers, Combust. Flame 138 (2004) 55–77.
[23] T. Endo, Y. Takenaka, Y. Sako, T. Honda, T. Johzaki, S. Namba, Reinjection of
transmitted laser light into laser-produced plasma for efficient laser ignition,
Appl. Optics 55 (2016) 1132–1137.
[24] http://www.ansys.com/products/fluids/ansys- chemkin- pro.
[25] http://combustion.berkeley.edu/gri-mech/.
[26] S. Gordon, B.J. McBride, Computer program for calculation of complex chemical
equilibrium compositions and applications, 1311, NASA Reference Publication,
1996.
[27] H. Schlichting, Boundary-layer theory, 7th ed., McGraw-Hill, New York, USA,
1979 Ch. 20.
[28] D.W. Green, R.H. Perry (Eds.), Perry’s chemical engineers’ handbook, Mc-
Graw-Hill, New York, USA, 2008 Sec. 6 (Incompressible flow in pipes and chan-
nels).
[29] H. Schlichting, Boundary-layer theory, 7th ed., McGraw-Hill, New York, USA,
1979 Ch. 19.
[30] L.D. Landau, E.M. Lifshitz, Fluid mechanics, 2nd ed., Butterworth-Heinemann,
Oxford, UK, 1987 Sec. 42.
[31] A.Y. Abdalla, B.B. Ali, D. Bradley, S.B. Chin, Stratified combustion in recirculat-
ing flow, Combust. Flame 43 (1981) 131–143.
416 T. Endo et al. / Combustion and Flame 191 (2018) 408–416

[32] H. Tennekes, J.L. Lumley, A first course in turbulence, MIT Press, Cambridge,
USA, 1972 Ch. 1.
[33] R.G. Abdel-Gayed, D. Bradley, A two-eddy theory of premixed turbulent flame
propagation, Phil. Trans. R. Soc. Lond. A 301 (1981) 1–25.
[34] K. Yamamoto, I. Hosokawa, A decaying isotropic turbulence pursued by the
spectral method, J. Phys. Soc. Jpn. 57 (1988) 1532–1535.
[35] H. Tennekes, Simple model for the small-scale structure of turbulence, Phys.
Fluids 11 (1968) 669–671.
[36] I. Glassman, Combustion, 3rd ed., Academic Press, San Diego, USA, 1996 Ch. 4.
[37] R.B. Bird, W.E. Stewart, E.N. Lightfoot, Transport phenomena, 2nd ed., John Wi-
ley & Sons, New York, USA, 2002 Sec. 13.3.
[38] E. Mastorakos, Ignition of turbulent non-premixed flames, Prog. Energy Com-
bust. Sci. 35 (2009) 57–97.
[39] E. Mastorakos, Forced ignition of turbulent spray flames, Proc. Combust. Inst.
36 (2017) 2375–2391.