Combustion and Flame 193 (2018) 440–452

Contents lists available at ScienceDirect

Combustion and Flame

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

Local extinction mechanisms analysis of spray jet flame using high

speed diagnostics
Antoine Verdier, Javier Marrero Santiago, Alexis Vandel, Gilles Godard, Gilles Cabot,
Bruno Renou∗
INSA Rouen, UNIROUEN, CNRS, CORIA, Normandie Universite, Rouen 76000, France

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

Article history: This paper reports an experimental study where flame structure, flow topology and local extinction mech-
Received 14 October 2017 anisms of n-heptane spray flames are investigated. The burner consists of an annular non-swirling co-flow
Revised 23 March 2018
of air that surrounds a central hollow-cone spray injector, leading to a lifted spray flame. The experiments
Accepted 23 March 2018
include measurements of droplet size and velocity by Phase Doppler Anemometry (PDA), flame structure
by High-Speed Planar Laser Induced Fluorescence of OH radical (HS-OH-PLIF) simultaneously recorded
Keywords: with the velocity fields of the reactive flow obtained by High-Speed Particle Image Velocimetry (HS-PIV).
Spray jet flame The poly-disperse spray distribution yields small droplets along the centerline axis while the majority of
Flame structure the mass is located as large droplets along the spray borders. These large droplets associated with high
High-speed OH-PLIF and PIV
velocities have ballistic trajectories and strongly interact with the inner wrinkled partially premixed flame
Flame extinction
front and the outer diffusion flame front. Simultaneous HS-OH-PLIF and HS-PIV images characterize the
dynamics of extinction events in the spray jet flame. In the inner reaction zone, local flame extinctions
are mainly controlled by the shear layer induced by the co-flow and the fuel–air heterogeneities due to
the evaporation of small droplets in the vicinity of the flame front. The large scales of turbulence in the
shear layer play a significant role in the dynamics of these extinctions. It is also found that the large
inertial droplets penetrate the lower part of the inner front reaching the burned gases, where they evap-
orate rapidly. They also disturb the outer reaction zone due to the low droplets temperature and the rich
mixture in the wake of droplet. These new results on local extinction of spray flames and droplet–flame
interactions will also strengthen the CORIA Rouen Spray Burner (CRSB) database for the improvement of
evaporation and combustion models for reacting sprays.
© 2018 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

1. Introduction [1–3] investigated the flame structures in the stabilization zones

New burner designs must move onto configurations with lower
pollutant emissions and higher efficiency. This requires a funda-
mental understanding of the processes involved in two-phase and
lean combustion. Two-phase combustion carries coupled multi-
physical and chemical phenomena which generally take place si-
multaneously within the combustion chambers. To overcome the
difficulties resulting from real geometries, the jet spray flame
configuration is taken as a reference for many experimental in-
vestigations. This canonical configuration includes a central pres-
sure or air-blast fuel injector, surrounded by a co-flow that may
present different thermophysical properties (composition, temper-
ature, etc.). The flame is stabilized downstream of the injection and
presents several kinds of flame structures. Past pioneering works
∗
Corresponding author.
E-mail address: renou@coria.fr (B. Renou).
of a two-phase jet flame above a coaxial air-blast injector. From
quantitative OH Planar Laser Induced Fluorescence (OH-PLIF) mea-
surements, authors suggested that the flame presented a struc-
ture of two diverging and opposed diffusion-like fronts. In their
Jet Spray Flame burner, operating with a pressurized fuel injec-
tor and a large bluff-body, Friedmann et al. [4] showed that the
spray flame also may have a dual reaction zone structure consist-
ing of an outer diffusion flame and an inner partially-premixed
flame. However, changing either the co-flow rate [5,6] or the co-
flow composition [7] affects the flame topology, and local extinc-
tions within the inner reaction zone may occur intermittently.
These results highlighted the variety of flame structures in the
Jet Spray Flame configuration, associated to the burner geome-
try and the fuel droplet generation. To enhance the understand-
ing of the droplets–turbulence–flame interactions and to improve
modeling capabilities of turbulent spray reacting flows, novel ex-
perimental facilities were designed by several research groups to

https://doi.org/10.1016/j.combustflame.2018.03.032
0010-2180/© 2018 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
 A. Verdier et al. / Combustion and Flame 193 (2018) 440–452
perform extensive measurements on the dispersed and carrier
phases.
The Sydney Spray Burner, developed by Masri and co-workers
[8–10], consists of a piloted spray burner where fuel is injected in
the carrier air co-flow by a nebulizer located upstream inside the
spray nozzle. The two-phase flame is stabilized by an annular pi-
lot flame. Authors provided an extensive database on the phenom-
ena of interest in dilute sprays: droplet dispersion, droplet evap-
oration, turbulence–droplets interactions and droplet–flame inter-
actions. The Delft Spray burner was designed by Roekaerts et al.
[11,12] to study spray combustion in Moderate or Intense Low-
oxygen Dilution (MILD) conditions. It consists of a pressure-swirl
atomizer that produces a spray of ethanol droplets issuing in a co-
flow of either air or hot combustion products. Dispersed and car-
rier phase properties were measured by Phase Doppler Anemome-
try (PDA), and a Coherent Anti-Stokes Raman Spectroscopy (CARS)
system was used to evaluate the gas-phase temperature statis-
tics. The Cambridge Spray Burner was updated from the gaseous
bluff-body swirl burner by changing the injection system [13,14].
N-heptane is injected through a pressure swirl hollow cone injec-
tor located within the bluff-body surrounded by a swirled air co-
flow. The burner operated close to blow-off limits and specific op-
tical diagnostics (high-speed OH-PLIF and joint PLIF measurements
of CH2 O and OH) were applied successfully both to investigate
the local flame structure, the reaction zones, and local extinction
holes along the flame sheet. More recently, the Rouen Spray Burner
was proposed as a new configuration to evaluate the droplet–
flame interactions by using original optical diagnostics [15]. The
experimental set-up is composed of an annular non-swirled air co-
flow that surrounds a central hollow-cone spray injector, leading
to a flame stabilized downstream. Both PDA, OH-PLIF and Global
Rainbow Technique (GRT) were applied to quantify the evolutions
of fuel droplet properties (size, velocity and temperature) across
the flame front, leading to a complete database for validating LES
codes.
The previous databases appear to be fundamental tools to val-
idate or to improve LES codes, which are sensitive to the selected
chemical mechanisms (global or detailed chemistry for instance)
and combustion models, and which are able to predict several
types of local flame structure encountered in spray flame. Ma and
Roekaerts [16] performed a 3D LES with Flamelet Generated Man-
ifolds (FGM) of two configurations in the Delft Spray dataset. They
demonstrated that the flame structures are directly impacted by
any changes in the magnitude of different characteristic time scales
of the flow: fuel droplet evaporation, convection and reaction. The
swirling ethanol spray flame near blow-off conditions (Cambridge
burner) was investigated by Guisti et al. [17,18] using the LES/CMC
approach with a simple and detailed chemical mechanisms. They
demonstrated that a detailed kinetic scheme was more appropriate
and able to capture the flame structure and its dynamics, such as
local extinctions and the lift-off height. The recent comparison be-
tween LES and the experimental results of the CORIA Rouen Spray
Burner (CRSB) database [15] was done by Shum-Kivan et al. [19].
The authors demonstrated the ability of LES (AVBP) to accurately
predict the spray jet flame structure. However, although the global
shape is well predicted by the simulation, the position of the lead-
ing edge is underestimated compared to the measurements de-
duced from OH-PLIF. This demonstrates that the simulation of such
geometries is a complex task, since it requires at least an accurate
forecast of the fuel vapor and thermal budget between the droplet
and its gaseous surrounding through the evaporation model. A de-
tailed description of the combustion reactions able to predict the
different modes of combustion is also necessary [20].
Due to the complexity of the multi-physical phenomena, it is
essential to develop and validate numerical models with accurate
experimental data for numerous practical combustion applications.
441
Fig. 1. Detail of the injection system. Dimensions in (mm).
With the improvement of the high repetition rates, high-speed
diagnostics can resolve in time the scalars in turbulent flows.
These efforts have yielded a greater understanding of the turbu-
lent flame dynamics by tracking the evolution in time of transient
phenomena such as local extinction, auto-ignition and turbulence–
chemistry interactions [21–24]. However, an insight on the mech-
anisms that control the flame dynamics and the transient phe-
nomena such as extinction or re-ignition in two-phase flows is
still in progress due to the measurement complexities in spray
combustion. Despite the recent works on the application of high
speed diagnostics in spray combustion [13,25], obtaining accurate
experimental data on transient phenomena such as local extinction
mechanisms in real and representative two-phase flow configura-
tions are still challenging.
In the current study, the CRSB database is completed with tem-
poral and spatial measurements of the flame front and its associ-
ated velocity field. Additional PDA results are also considered and
presented. The objective of this work is to extend understanding of
turbulence–flame interactions in the spray jet flame and droplet–
flame interactions by combining two high-speed optical diagnos-
tics. This paper is organized as follows. Section 2 describes the
burner and the optical diagnostics. The global shape of the flame
and the associated aerodynamics in terms of dispersed and car-
rier phase are presented in the first part of Section 3. Moreover,
the local extinction mechanisms which occur in the reaction zone
will be discussed. A summary of the flame dynamics with transient
phenomena mechanisms is provided in Section 4.
2. Experimental setup
This section is dedicated to the description of the burner, the
operating conditions and the optical diagnostics applied during the
experiments.
2.1. Burner and operating conditions
Figure 1 illustrates the CORIA Rouen Spray Burner (CRSB),
which is based on the geometry of the gaseous KIAI burner [26].
The atmospheric unconfined burner is composed of an external an-
nular non-swirling air co-flow and a pressurized liquid fuel injec-
tor (Danfoss, 1.35 kg h−1 , hollow cone with a angle of 80o ). The
air duct is equipped with 18 radial vanes (non-swirling) that break
any large flow structures remaining in the plenum. The air-flow
is then guided by a convergent into the chamber surrounding the
fuel injector and forming a turbulent air co-flow. An electronic
Coriolis flow controller (Bronkhorst, CORI-FLOW, [0–2] g s−1 ) reg-
ulated the n-heptane mass flow rate and a thermal flow controller
(Bronkhorst, EL-FLOW, [0–15] g s−1 ) was used to control the air
mass flow rate. The inlet conditions were 6 g s−1 (T = 298 ± 2 K)
and 0.28 g s−1 (T = 298 ± 2 K) for the air and fuel, respectively.
442 A. Verdier et al. / Combustion and Flame 193 (2018) 440–452
Fig. 2. Measurement grid for Phase Doppler Anemometry and fields of view for the
different optical diagnostics applied in the spray jet flame. To optimize the spatial
resolution, the field of view of the HS-OH-PLIF is smaller than the OH-PLIF.
Cylindrical coordinates are considered: Uz and Ur represent the
axial and radial velocity respectively.
2.2. Optical diagnostics
The local properties of the flow were measured by phase
Doppler anemometry (PDA) obtaining size-classified velocity data
for fuel droplets and velocity data for the air by seeding the car-
rier phase. The settings were adjusted in order to measure the dis-
persed and carrier phases in reactive conditions. The global flame
structure was investigated by OH-PLIF imaging at a low repetition
rate (10 Hz). High-Speed OH-PLIF (HS-OH-PLIF) and High-Speed
PIV (HS-PIV) were applied simultaneously. The temporal and spa-
tial flame front positions associated with instantaneous velocity
were obtained to study the transient phenomena such as extinc-
tions. Figure 2 shows the different meshes and fields of view for
the different optical diagnostics.
2.2.1. Phase doppler anemometry (PDA)
A two-colour PDA measurement device from DANTEC was used
to perform point measurements of air velocity and fuel droplet
velocity and size with green (514.5 nm) and blue (488 nm) laser
beams. The emission probe used had a focal length of 350 mm and
a beam spacing of 50 mm while the lens in the receiving probe
had a focal length of 310 mm. The receiving probe was placed in
front scattering mode at 50o , close to the Brewster angle, enhanc-
ing the detection of first order refracted light. A mask placed in
the receiver enabled a detection range of 139 μm. The beam cross-
ing generated a probe volume approximated by a cylinder 200 μm
long and with a diameter of 120 μm. Data acquisition was limited
to 30 s or 40,0 0 0 droplets so as to have converged statistics. Sev-
eral parameters such as sphericity or droplet concentration control
the quality of the measurement. High-density regions lead to a de-
crease in validation due to signal rejection when more than one
droplet at a time was present in the measurement volume. In this
study, a decrease on the validation was observed below Z = 10 mm
and it becomes unsatisfactory below Z = 7 mm in the dense region
of the spray. Two different and independent acquisitions enabled
to focus the measurement strategy either on the air velocity or on
the fuel droplet size-conditioned velocity. Indeed, in order to cor-
rectly detect the liquid and the gaseous phase separately, the gain
and voltage of the photomultipliers (PMs) were accordingly ad-
justed. To perform the measurements of the carrier phase in pres-
ence of fuel spray, the air was previously seeded with 2 μm olive
oil droplets and the gain and voltage in the PMs were optimized
(increased) in consequence (configuration: PDA(A)). PDA(A) config-
uration allows to detect the air seeding and small fuel droplets
D10 ≤ 2 μm, which follow perfectly the air velocity due to their
Fig. 3. Optical arrangement for simultaneous HS-PIV and HS-OH-PLIF
measurements.
low stokes number. Both phases were measured in reactive condi-
tions. The size and velocity of the dispersed phase were also mea-
sured (in a different acquisition without seeding oil droplets) in
order to characterize the fuel droplet properties such as size dis-
tribution and classified-size velocities. The PDA configuration and
the gain and voltage of the PMs were adjusted to detect all the
sizes present in the spray (2 ≤ D10 ≤ 80 μm) (configuration: PDA(B)).
Mean velocity and diameter values, presented in the next figures,
were calculated with a minimum of 500 droplets. The velocity of
each droplet was weighted by its transition time in the measure-
ment volume to avoid the well-known velocity bias.
2.2.2. High resolution OH-PLIF (10 Hz)
The global flame structure was investigated by OH-PLIF imag-
ing at low repetition rate (10 Hz). A Nd:YAG.laser operating at
532 nm was used to pump a tunable dye laser (Quantel TDL90).
The excitation wavelength was tuned to the Q1 (5) transition of
 
the A2  + (v = 1 ) ← X 2 (v = 0 ) band of OH at λ = 282.665 nm.
The resultant output energy was 30 mJ per shot in the probe
volume. The collection system consisted of an ICCD camera (PI-
MAX 4, Roper Scientific, 1024 × 1024 pixels2 ) equipped with UV
lens (f/2.8). The field of view was 112 × 112 mm2 leading to a mag-
nification ratio of 4.52 pix.mm−1 . A broadband collection strategy
from 308 to 330 nm with a band-pass filter (SCHOTT UG11) was
adopted and a high-pass filter (SCHOTT WG295) was used to re-
duce Mie scattering from the fuel droplets.
2.2.3. Simultaneous high-speed PIV and high-speed OH PLIF
Temporal and spatial flame front positions and associated in-
stantaneous flow fields were obtained by High-Speed OH-PLIF at
10 kHz combined with High-Speed PIV at 5 kHz. A schematic of
the high-speed imaging set-up is shown in Fig. 3. Concerning the
HS-OH-PLIF, a Nd:YAG.laser operating at 532 nm (104 W) was used
to pump a tunable dye laser (Sirah Credo). The excitation wave-

length was tuned to the Q1 (5) transition of the A2  + (v = 1 ) ←

X (v = 0 ) band of OH at λ = 282.671 nm. The resultant output
2
pulse energy was 380 μJ per shot in the probe volume. The beam
was then expanded to 40 mm in height using a set of fused sil-
ica lenses ( f1 = 10 0 0 mm, f2 = −20 mm, f3 = 500 ) mm. Due to the
low energy delivered, the detection system consisted of a CMOS-
camera Photron Fastcam SA5 mounted with an external image in-
tensifier (High Speed IRO, LaVision). The camera was operated at
a repetition rate of 10 kHz with an array of 896 × 848 pix2 and a
magnification ratio of 20.41 pix.mm−1 . The intensifier delay (26 ns)
and gate (500 ns) were set to optimize the signal to noise ratio.
 A. Verdier et al. / Combustion and Flame 193 (2018) 440–452
Fig. 4. (a) Photograph of the spray jet flame. (b) Typical OH-PLIF recording and different flame zones (A, Inner Reaction Zone (IRZ), C, Outer Reaction Zone (ORZ) and S).
Two high-pass optical filters (SCHOTT WG295) were used to re-
duce background noise due to elastic scattering by the droplets.
The camera on-board memory can hold over 7500 frames, corre-
sponding to an acquisition time of 750 ms. The OH-PLIF signal was
collected within the 308–330 nm range using a band-pass filter
(SCHOTT UG11). Moreover, the laser sheet profile was taken into
account and corrected by filling a quartz vessel placed on the test
facility with a homogeneous air–acetone mixture. Image process-
ing tools including non-linear filtering [27] and active contour de-
tection implemented in a level-set method [28] were developed to
extract flame front contours. The method is based on the original
approach formulated by Perona and Malik [29] and has several ad-
vantages: noise is smoothed locally, whereas little or no smooth-
ing occurs between image objects which enhance the local edges
[27,30]. Moreover, non-linear diffusion filter allows preserving the
flame front edges without artificial shifting in position [31].
To obtain instantaneous velocity fields, high speed PIV was
applied at 5 kHz. A double cavity (Nd:YLF Laser Darwin Dual
Quantronix) operating at 527 nm delivers two laser beams sep-
arated by tPIV = 10 μs. Zirconium oxide (ZrO2 ) particles were
used to seed the air co-flow (configuration: HS-PIV-(A)). The Mie
scattering signal from droplets and ZrO2 was collected with a
Phantom V2512 camera (1280 × 800 pix2 ) equipped with a Nikon
f/1.4 50 mm objective leading to a magnification ratio equal to
18.98 pix.mm−1 . In this study, a close attention was paid to the
air entrainment by the air co-flow. To capture the slow air entrain-
ment velocity, the HS-PIV was adjusted (configuration: HS-PIV-(B),
tPIV = 400 μs). The ambient air was seeded with a glycol mist in
order to obtain the air entrainment, which is crucial to explain the
flame structure. The same post-treatment (Dynamics Studio soft-
ware) was applied on images (HS-PIV-(A) and HS-PIV-(B)) with a
cross-correlation algorithm with a window size of 32 × 32 pix2 and
a 50% overlap. A coherent filter was applied to correct isolated
false vectors.
3. Results and discussion
This section is divided into four subsections that correspond to
the results of the global analysis of aerodynamics, flame shape, po-
sition of local extinctions and the study of interactions between
flame, turbulence and droplets.
3.1. Global topology of the spray jet flame
Figure 4(a) shows a photograph of the spray jet flame, where
the global flame appearance can be observed. The global flame
shape and the lift-off height of the flame are governed by air
443
flow rate, spray properties (droplet velocity and size) and global
equivalence ratio, as it was reported in many papers. The axi-
symmetric flame with a classical double structure is illustrated by
the OH-PLIF (Fig. 4(b)), where five different zones (A, IRZ, C, ORZ
and S) are reported. The overall shape of the flame exhibits two
branches corresponding to an inner and an outer reaction zone, IRZ
and ORZ respectively, connected together at the flame leading edge
(S). These two reaction zones are separated by hot gases (zone C)
exhibiting a weak OH-PLIF signal. This flame structure results from
the spray heterogeneity in size [5,12] imposed by the pressure fuel
injector technology. The large droplets spread into the outer part
of the air co-flow corresponding to the ambient air and the small
droplets are mainly located near the centerline in the mixing zone
(A).
3.2. Carrier and dispersed phases properties
Figure 5 presents a map of the two mean velocity components
(axial and radial) of the carrier phase obtained by PDA in con-
figuration PDA(A). Two important features are illustrated in the
mean velocity fields. Consistent with the earlier works [2,7,11,25],
the mean axial co-flow velocity provides two peaks at the exit of
the injector. For high axial stations Z > 25 mm, the bluff-body ef-
fect becomes negligible and the co-flow air peaks merge into a flat
axial velocity profile. Note that the air co-flow jet presents radial
velocities approximately 5 times smaller than the axial velocities
at Z = 10 mm, which rapidly decrease down to zero at Z = 25 mm.
Magnitude of velocity fluctuations is represented in Fig. 6 in
terms of the turbulent kinetic energy k. Isotropy along the two di-
rections perpendicular to the Z-axis is assumed and turbulent ki-
netic energy (k) is calculated following equation:
 
1
k= (uz )2 + 2 × (ur )2 (1)
2
The annular air co-flow discharging into the ambient air cre-
ates a shear layer at the interface between the high air co-flow
velocity and the ambient air inducing turbulence with large scale
structures. This phenomenon is highlighted in the movie given in
Supplementary Material SM1. The quiescent ambient air is accel-
erated radially towards the centerline whereas a portion of the air
jet loses some momentum. Figure 6(b) represents different turbu-
lent kinetic energy profiles for three axial stations. The maximum
value of k (≈ 40 m2 s−2 ) is found at low axial stations, in the region
where the air enters the chamber. The turbulent kinetic energy
does not follow the same pattern as the axial velocity, with peaks
diffusing radially with increasing axial stations. Given that velocity
fluctuations enhance mixing and that all droplets enter the region
444 A. Verdier et al. / Combustion and Flame 193 (2018) 440–452

Fig. 5. Mean components of air velocity flow for reacting conditions with vector field obtained with configuration PDA(A). (a) Mean axial velocity. (b) Mean radial velocity.

Fig. 6. (a) Turbulent kinetic energy (k) of the air in reacting conditions, calculated
from Eq. 1. (b) Turbulent kinetic energy profiles for different axial stations.

of maximum k directly after atomization, the small droplet trajec-

tories will be strongly influenced by the airflow velocities. This in-
teraction segregates the droplets spatially, transporting the small
droplets (with low Stokes number) to the center while the large
droplets are directed to the spray borders. The fuel spray shows
a complex heterogeneous distribution, with droplet sizes ranging
from 2 to 60 μm. Profiles of the mean diameter (D10 ) and Sauter
mean diameter (SMD, D32 ), at four axial stations, Z = 16, 25, 35 and
45 mm, are represented in Fig. 7(a) and (b). The mean diameter
profiles grow (from 5 to 50 μm) radially from the center towards Fig. 7. (a) Fuel droplet Sauter Mean Diameter (SMD) D32 in reacting condition. (b)
Fuel droplet mean diameter D10 in reacting condition. Error bars indicate the RMS
the borders and vary significantly with Z due to the evaporation
of the droplet size distribution. (c) PDF of droplet size fitted with Rosin-Rammler
process. Droplets are expelled from the atomizer under a cone distribution.
with an 80o opening, following a direction that will force them
to suddenly interact with the air co-flow. The central region con-
tains a very low quantity of droplets due to the centrifugal move- To illustrate the behavior of the droplets, the mean and RMS
ment imposed by the swirling liquid injection inside the atomizer. of the axial and radial velocity components of different droplet
The smallest droplets resulting from the liquid jet breakup are re- size-classes at four axial stations, Z = 16, 25, 35 and 45 mm,
sponsive to the air velocity field and follow the flow due to their are shown in Figs. 8 and 9. The class width was chosen to be
associated small Stokes numbers. Furthermore, in this region, the 10 μm. The behavior of the 60–70 size group is very close to
very small droplets represent only a small quantity of the liquid that of the 50–60 group but, since they contain less droplets,
mass leading to a lean heterogeneous mixture that is close to the they are not presented. Values for the last two groups (30–40,
lean flammability limit [19]. However, larger droplets with higher 50–60) are not shown in the central region because less than
Stokes numbers are less affected by the aerodynamic conditions 500 droplets in 30 s were detected per point. The carrier phase
and follow ballistic trajectories. It is also worth noting that the velocity profiles are also superimposed to evaluate slip velocity
number distributions of droplet size are well represented by Rosin- statistics. Figure 8 shows that droplets are ejected from the nozzle
Rammler distributions everywhere, as illustrated in Fig. 7(c). with strong radial velocities. At the first axial station, the three
 A. Verdier et al. / Combustion and Flame 193 (2018) 440–452 445

Fig. 8. Mean components of fuel droplet velocity in reacting conditions separated in four size-classes obtained with PDA (B) configuration. o represent the [0–10] μm group,
the [20–30] μm group, the [30–40] μm and represent the [50–60] μm. Black lines represent the Air velocities profiles.

Fig. 9. RMS of fuel droplet velocity in reacting conditions separated in four size-classes. o represent the [0–10] μm group, the [20–30] μm group, the [30–40] μm and
represent the [50–60] μm. Black lines represent the Air RMS velocities profiles.

groups have appreciable radial velocities (≈ 10 m s−1 ) but farther
downstream small droplets adapt their velocity to the airflow
velocities. The large droplets follow more ballistic trajectories and
continue to have radial centrifugal velocities until Z = 35 mm.
When overlapped to the air stream, axial mean velocities of small
droplets (groups 0–10 μm) are similar to that of the air whereas
large droplets continue moving at smaller velocities until farther
downstream. When fuel droplets exit the high air velocity region,
they find a quiescent area where they can decelerate. Concerning
the velocity fluctuations, small droplets are more affected by
the shear layer and the coherent structures generated by the air
co-flow turbulence (Fig. 9). Near the radial station R = −10 mm,
in the region where the turbulent kinetic energy is maximum and
for axial stations where large turbulent scales are fully developed,
the RMS for radial velocities is maximum.
3.3. Flame structures
3.3.1. Inner reaction zone (IRZ)
An instantaneous snapshot of OH-PLIF at 10 Hz in a vertical
cross section is reported in Fig. 4(b) and illustrates the Inner Re-
action Zone topology. The IRZ exhibits a highly wrinkled struc-
ture placed on the shear layer formed between the issuing air jet
and the surrounding air. It was shown numerically that IRZ ex-
hibits a double (premixed and non-premixed) structure [19]. In-
deed, the IRZ consists first in a premixed flame fed by the cen-
tral part of the spray (Zone A) where the mixing between small
droplets (D10 < 20 μm) and the air co-flow generates a very lean
region close to the flammability limit of n-heptane. Large amount
of fuel vapor is found between the Inner and Outer reaction zones,
where medium to large droplets (D10  20–60 μm), which account
for the major part of the injected fuel mass, evaporate fast result-
ing in a fuel rich mixture. Consequently, the remaining air from
the lean premixed flame then reacts with the fuel vapor that dif-
fuses from the fuel-rich, hot region (zone C) in a diffusion flame
very close to the lean premixed flame. The IRZ is then located in a
very lean region of the flow (for its inner flame) and subjected to
strong velocity fluctuations. Consequently, the IRZ is strongly dis-
turbed (momentum and mass transfer) and wrinkled by local ve-
locity gradients, and frequent events of transient phenomena such
as local extinctions are observed. These local extinctions are illus-
trated by holes in High-Speed OH-PLIF video given as supplemen-
446 A. Verdier et al. / Combustion and Flame 193 (2018) 440–452

√
Fig. 10. (a) Spatially averaged ORZ position (black lines obtained with OH-PLIF measurements) and air entrainment streamlines with the norm of velocity (|U | = U2 + V 2)
obtained with HS-PIV-(B) configuration). (b) Schematic representation of ORZ : Generic structure of a laminar diffusion flame.

tary material 2 and will be investigated in more details in the fol-

lowing section.

3.3.2. Outer reaction zone (ORZ)

Fuel droplets both in non-reactive zones (spray exit, mixing
zone) and in reactive zones (zones C and ORZ) can be easily iden-
tified by Mie scattering in Fig. 4(b). The ballistic trajectories of
large inertial droplets lead to the presence droplets in zone C,
where the high temperature increases the evaporation rate gener-
ating a gaseous fuel reservoir. Despite there are few big droplets
(D10 > 30 μm) coming to the outer region of the spray, the droplet
evaporation rate implies the production of fuel vapor until reach-
ing an equivalence ratio higher than φ = 1.8 [19]. In addition,
fuel thermal decomposition and incomplete hydrocarbon combus-
tion will produce CO [16], CO2 and H2 O, and intermediate species
Fig. 11. Positions of the flame base.
such as soot precursors (PAH), as evidenced by the yellow color
of the flame. Consequently, the fuel mass fraction YF is obviously
far less than unity in zone C, and the mixture fraction falls be-
tween ξ st < ξ < 1. The air entrainment is shown in Fig. 10(a), where
streamlines and the velocity magnitude are illustrated. The stream-
lines are uniform far away from the air jet, and diverge close to the
ORZ with a stagnation point located in average at R = 37 mm and Z
= 52 mm. At this stagnation point, flow trajectories are separated:
one move towards the leading edge and the second one towards
the top of the flame. It is worth noting that the quiescent sur-
rounding air is accelerated towards the air co-flow exit and reaches
1 m s−1 just below the flame leading edge. The entrained air re-
acts with the gaseous fuel reservoir in a diffusion-like outer flame
front (ORZ), where a stoichiometric line is present and located be-
tween the rich (partially burned gases) and lean (fresh gases) sides.
Compared to the IRZ, the ORZ is less wrinkled, more stable, thicker
and characterized by a smoother OH gradient. Figure 10(b) sum-
Fig. 12. Mean OH-PLIF with extinction locations. Red inverted triangles represent
marizes the ORZ structure.
the local flame extinction positions in the IRZ, whereas blue triangles concern the
local extinction positions in the flame leading edge. Green ellipse mark a region,
3.3.3. Stabilization of the flame base (S) where large inertial droplet cross the flame leading edge and interact with the ORZ.
The stabilization mechanism of a lifted spray jet flame is con- (For interpretation of the references to color in this figure legend, the reader is
ditioned by the droplet convection-vaporization, the flow aerody- referred to the web version of this article).
namics and the chemical reaction budgets. For this specific op-
erating condition, the flame is lifted and stabilized around R =
15.5 ± 3 mm and Z = 24 ± ± 3 mm. Each square in Fig. 11 This preliminary description of the different zones of the Jet
corresponds to the leading edge location extracted from a set of Spray Flame illustrate the strong interactions between large inertial
500 OH-PLIF images. This position presents a favorable condition to droplets and the leading edge revealing also some local extinctions
sustain a stable reaction: the turbulence is low (Fig. 6) and there is in the IRZ and on the flame leading edge. The time-scales of tur-
enough fuel vapor available fed by the evaporation of large inertial bulence and chemistry are small and require accurate experimen-
droplets. tal data through high-speed measurements to analyze the transient
 A. Verdier et al. / Combustion and Flame 193 (2018) 440–452 447

Fig. 13. Extinctions in the IRZ : flame–turbulence interactions. HS-OH-PLIF images overlaid with the fluctuation velocity field and the strain rate (SM3) - time separation =
0.2 ms.

phenomena that occur in a spray jet flame. This is presented in the
following sections.
3.4. Position of local extinctions
In this work, OH-PLIF signal is used as a marker of the reaction
zones since disruptions in continuous OH fronts correlate well
with local flame front extinctions [32]. Here, the different zones
of local flame extinction are marked in Fig. 12 where they are
superimposed with the mean OH-PLIF image. These markers
represent the position (R, Z) of local flame extinction appearance
extracted from time series of HS-OH-PLIF images. One can notice
three different zones of local extinctions associated with three dif-
ferent flame structures. First, local flame extinctions are observed
in the IRZ (red inverted triangles) in axial regions from R = 10
and R = 15 mm for different heights above 27 mm. These local
extinctions occurs the most frequently in a region where the flame
front is submitted to intense local strain. Occasionally, some local
extinctions occur close to the flame leading edge (blue triangles).
This zone is characterized by large droplets (D10 > 30 μm) with a
mean velocity equal to 10 m s−1 . At this location, the local turbu-
lence intensity is moderate, and the extinction is therefore mainly
controlled by droplets–flame interactions. The third zone concerns
the ORZ, where the large initial droplets expelled from the pres-
448 A. Verdier et al. / Combustion and Flame 193 (2018) 440–452

Fig. 14. Schematic representation of the extinction mechanisms in the inner reaction zone.

Fig. 15. Profile at R = 16 mm and Z = 30 mm. (a) Extinction occurrence temporal analysis. (b) Extinction occurrence frequency analysis with three runs.

sure injector cross the flame leading edge and are still present
when they interact with the ORZ. The perturbations between large
droplets and ORZ (which are not extinctions) are observed in the
green delimited region in Fig. 12.
In summary, three regions of transient phenomena are observed
in the spray jet flame and defined here for subsequent analysis
with more details:
(i) Extinctions in the IRZ far away from the flame leading edge:
flame–turbulence interactions.
(ii) Extinctions in the IRZ close to the flame leading edge:
droplets–flame interactions.
(iii) Perturbations in the ORZ: droplets–flame interactions.
3.5. Flame–turbulence interactions
Figure 13 presents a sequence of flame extinctions in the IRZ
illustrated by OH-PLIF images superimposed with the correspond-
ing velocity fields obtained by HS-PIV. The OH signal is associated
to both reaction zones (premixed and non-premixed) as it was
numerically demonstrated in [19] producing OH radicals in IRZ.
However, these reaction zones are very close together and it
is not experimentally possible to discriminate OH signals from
their origin. A clear disruption in continuous OH front in the IRZ
means that OH signal is neither observed from premixed nor
non-premixed reaction zones. Note that the velocity vectors are
overlaid with the instantaneous strain rate of flow and computed
 1
as σ = ( Ei2j ) 2 , where Eij is estimated by Ei j = 12 (∂ j vi + ∂i v j ), the
2D strain rate tensor. The maximum value of σ is around 50 0 0 s−1
and appears intermittently in the flow region interacting with the
IRZ. These large scales of velocity fluctuations are created by the
velocity gradients in the shear layer between the co-flow and the
ambient air (Supplementary Materials, SM3). Figure 13 highlights
that the flame front thinning and wrinkling are caused by a high
 A. Verdier et al. / Combustion and Flame 193 (2018) 440–452 449

Fig. 16. HS-OH-PLIF images - time separation = 0.1 ms. Extinction in the leading edge due to the droplet–flame interactions.

Fig. 17. HS-OH-PLIF images - time separation = 0.1 ms. Perturbations in the outer reaction zone (ORZ) due to the droplet–flame interactions.

strain rate (σ > 50 0 0 s−1 ) in IRZ region. From t = 0.0 ms to t
= 0.4 ms the OH signal in the flame front becomes thinner and
thinner. The extinction event is observed on the HS-OH-PLIF im-
ages at t = 0.4 ms, which correspond to a position R = 12 mm and
Z = 35 mm. The local extinction seems to appear when the flame
front is too fine to support a high stretch rate. Then, with the local
extinction in the IRZ, a gap is produced and causes a hole separat-
ing the inner flame front in two reaction zones. The hole and the
local IRZ are convected downstream due to the mean air velocity
field. The hole propagation rate is not constant and depends on the
local turbulent flow and composition. Note that in this configura-
tion, and in the OH-PLIF cross section, 11 local extinctions in the
IRZ are observed for 10 0 0 instantaneous OH-PLIF images. More-
over, with these operating conditions (air: 6 g s−1 ; fuel: 0.28 g s−1 )
a local extinction in the IRZ cannot cause a global extinction of the
flame.
Figure 14 presents a schematic representation of the
turbulence–flame interactions in the IRZ. Close to the injector
when the co-flow is discharging in the quiescent air, large coher-
ent and intense turbulence structures are created by the mean
velocity gradient. They are large enough to go far out into the
co-flow and into zone (C) and then to induce a strong large scale
mixing between fuel droplets, vapor and air. Then, this local
heterogeneous and turbulent mixture interacts with the reaction
zone and disturbs the flame front by wrinkling processes. In
the hot region close to the IRZ the kinematic viscosity changes
approximately by a factor of 20 from fresh to burned gases. Even
if the flow is laminarized by the higher viscosity close to the
450 A. Verdier et al. / Combustion and Flame 193 (2018) 440–452
IRZ, the flame is disturbed by the flow aerodynamics. When local
strain rate associated with local equivalence ratio exceeds a critical
value, local flame extinction occurs and the extinguished part
is convected downstream by the flow field. Indeed, recent study
[33] demonstrates that during flame–vortex interactions, the flame
can locally extinguish when the flow exceed a critical extinction
strain rate. The main difference with local extinction with gaseous
flames [34,35] is on the presence of small fuel droplets near the
reaction zone which can enhance fuel heterogeneities or quickly
evaporate when crossing the flame front. Franzelli et al. [36] has
shown that local extinctions occurring in a spray jet flame may
also be induced by local fuel depletion.
To validate the predominant aerodynamic effect on local extinc-
tion mechanisms in the IRZ, caused by local velocity gradients and
lean mixtures, an extinction frequency analysis is applied to dif-
ferent HS-OH-PLIF measurements. An analysis region with a probe
located in the mixing region in the vicinity of the reaction zone
(R = 14 mm and Z = 30 mm) is used to determine the temporal ap-
pearance of extinctions. In Fig. 15(a), Df represents the distance be-
tween the probe and the flame front position, extracted from OH-
PLIF images. The specific value D f = 0 corresponds to the position
of the instantaneous flame front when the IRZ is located at the
same position as the probe, whereas the negative and positive val-
ues of Df imply a flame-probe distance located in the fresh and
burnt gases, respectively. The values of D f > 10 mm correspond to
the distance between the probe and the ORZ. This occurs when
there is a local extinction in the IRZ. The temporal evolution of
Df has a so-called telegraphic shape, alternating D f > 10 mm and
D f  0 mm values. The power spectral density function (PSD) of
the temporal evolutions of Df are reported in Fig. 15(b). A strong
peak around 80 Hz is clearly visible, repeatable, indicating pe-
riodic flame extinction phenomenon. These measurements have
been performed three times separated by several weeks. It can be
observed that for the red and blue lines, the PSD show a strong
peak around 80 Hz whereas for the third acquisition (green) a peak
at around 80 Hz is still observed but with a more broadened shape.
The frequency analysis of the large-scales of turbulence seems
relevant for the understanding of the extinction mechanisms and it
is now discussed. From high-speed PIV measurements, a temporal
velocity signal is extracted at R = 13 mm and Z = 30 mm located
in the shear layer (region A). The integral of the auto-correlation
function of these velocity fluctuations is then determined and is
equal to 12.2 ms. The latter represents the temporal integral length
scale of the turbulence at this location, where local flame extinc-
tions occur. It represents a measurement of the longest time that
the velocity fluctuation remains autocorrelated being, hence, rep-
resentative of coherent large structures in the flow. In this case it
is equivalent to a frequency of 82 Hz. This value is similar to the
flame extinction occurrence underlying that flame extinction in IRZ
is mainly controlled by flame/turbulence interactions. Indeed, in
the mixing zone of the flame (zone A), the lean mixture is very
close to the flammability limit [19] and the local flame strain rate
induced by the most energetic length scales of the shear layer will
exceed locally and periodically the extinction strain rate of the IRZ.
3.6. Droplet–flame interactions
3.6.1. Extinctions in the leading edge
As already seen in the previous section, the small droplets
mainly stay in the cold central zone, where they evaporate slowly
and never cross the flame front. On the contrary, large inertial
droplets (D10 > 30 μm) move to the external part of the spray cross-
ing the flame base and reaching the hot region. Figure 16 presents
a local extinction event in the leading edge caused by a large
droplets. At t = 0.2 ms, a large droplet (yellow circle) crosses the
leading edge. 0.1 ms after, a hole is produced corresponding to a
Fig. 18. Schematic representation of the droplet–flame interactions.
local flame extinction. A detailed analysis of the OH-PLIF images
was performed to identify the number of extinctions during a set
of measurement. In the plane of measurement, 91 local extinctions
events were detected for 10 0 0 instantaneous OH-PLIF images or for
100 ms.
Previous numerical studies [37,38] suggest that the droplet–
flame interaction mechanisms, in a partially premixed flame,
should be related to a thermal disturbance and to a modification of
the local mixture. Indeed, Nakamura and Akamatsu [37] show that
the energy exchange between droplets and gaseous phase may dis-
turb the reaction rate. The temperature of droplets is much lower
than of the gaseous phase, acting as a sink term for the chemical
reactions. The intermittent passage of large droplets in the leading
edge induces a locally high strain rate due to the droplet motion,
which can reduce the reaction rate [39–41]. When crossing the
flame, fuel droplet temperature increases up to a welt-bulb tem-
perature [15], which remains lower than the fuel boiling tempera-
ture. In addition, the high burned gas temperature leads to a strong
fuel evaporation rate and a large values of gaseous fuel mass frac-
tion YF gas around the droplets. The local extinction in the leading
edge can be explained by the droplet temperature in this region
(Tdroplet Tflame ) and the mixture around the droplet, which is very
rich and outside the flammability limits. It is worth noting that
the local flame extinction due to droplets crossing a flame front is
not explicitly taken into account in most models (closure models
in RANS or subgrid models in LES).
3.6.2. Perturbations in the ORZ
Figure 17 presents HS-OH-PLIF images separated by 0.1 ms,
where two big droplets (red circles) cross the zone C with a slip
velocity of ≈ 7 m s−1 . Further downstream, droplets that have sur-
vived the leading edge move towards the ORZ, where they disturb
the flame front. Indeed, after the passage of the drops, a local per-
turbation located in the wake can be easily identified. The movie
of this droplet–flame interaction is available in Supplementary Ma-
terials (SM2).
Figure 18 presents a schematic representation of the droplet–
flame interaction. As described in the previous section, large
droplets cross the leading edge and continue to evaporate in
burned gases in zone C. The strong thermal exchange leads to a
fast evaporation regime resulting in an increase of the gaseous fuel
mass fraction and, hence, a droplet diameter reduction. However,
the largest droplets can reach the ORZ and induce local pertur-
bations in the reaction zone. In the burned gases (zone C), lot of
species are present and imply gaseous fuel mass fraction YF < 1.
However, when considering fuel droplets within the ORZ, the pure
fuel vapor generated by the evaporation leads to an increase of the
fuel mass fraction until to reach YF close to one near the liquid in-
terface. The YF fluxes towards the reaction zone are then strongly
 A. Verdier et al. / Combustion and Flame 193 (2018) 440–452
modified in the vicinity of the droplets and should locally mod-
ify the location of the stoichiometric line (ξ St ) in the ORZ. The
droplet evaporation decreases also the surrounding gas tempera-
ture within the ORZ due to the heat latent absorption until reach-
ing the wet-bulb temperature of the droplet. This latter was mea-
sured to around 330 K in zone C by Verdier et al. [15] and may
locally quench the flame. The droplet–flame mechanism in the ORZ
is then related to a cooling effect (strong gradient between the
droplet and the flame temperature Tdroplet Tflame ) and to a lo-
cal increase of the gaseous fuel mass fraction around the droplet,
leading to the appearance of a wake behind the droplets traveling
across the ORZ. In most cases, the droplets disturb the flame front
but no extinctions can be observed, different from what is seen in
the premixed flame zone (IRZ). Furthermore, the perturbation time
is shorter than the local extinction duration in IRZ. In the ORZ, the
turbulence intensity is very low and combustion occurs in diffu-
sion regime, which prevents the amplification of the perturbation
caused by the droplets.
4. Conclusion
A detailed experimental study of turbulence–droplet–chemistry
interaction events in an n-heptane spray jet flame has been pre-
sented. The sequences of the simultaneous high-speed measure-
ments allow to capture the transient phenomena, such as the
mechanisms governing the local extinctions events with sufficient
spatial and temporal resolution. The flame exhibits a double struc-
ture with inner and outer reaction zones, where fuel droplets are
still present and interact with the reaction zones. Each flame zone
exhibits different mechanisms of droplet–turbulence–flame inter-
actions.
• Flame/Turbulence extinction mechanism. Periodic local flame ex-
tinctions are observed in the IRZ downstream the leading edge
for axial stations above Z = 27 mm. These local extinctions oc-
cur in a region where the lean partially vaporized premixed
flame front is submitted to intense local strain rate induced by
the shear layer. This extinction mechanism is mainly associated
to flame–turbulence interactions in the gas phase since most of
the small droplets within this region have been rapidly evapo-
rated at the entrance of the reaction zone.
• Flame/Droplet extinction mechanism. The mechanism related to
the droplet–chemistry interactions occurs at the flame leading
edge and occasionally produces a local flame extinction. In-
deed, this zone is characterized by low turbulence intensities
and large droplets crossing the flame front with high velocity.
These droplets are characterized by a low temperature thus act-
ing as a heat sink for the flame front, which finally extinguishes
mainly due to the cooling effect.
• Flame/Droplet perturbation mechanism. The last mechanism as-
sociated to the droplet–flame interaction mechanism occurs
in the ORZ. The largest droplets reaching the ORZ evaporate
quickly in the burnt gases region producing large amount of
fuel vapor surrounding the droplet. They interact with the dif-
fusion flame front where a strong decrease of the OH is ob-
served around the droplet during its trajectory. If no local ex-
tinction is observed, a strong modulation of OH signal indicate
a local decrease of the flame heat release rate.
All these experimental results provide new insight on the
experimental evidences of droplets/flame/turbulence interaction
mechanisms in a spray jet flame, and an original database suit-
able for numerical modeling to predict non-stationary events such
as local flame extinction in spray flames.
451
Acknowledgments
The authors gratefully acknowledge the financial support from
the Agence Nationale de la Recherche with the grant TIMBER
ANR-14-CE23-0 0 09. Dr. Malbois and Salaun are also warmly ac-
knowledged for their technical support on data processing tools.
Supplementary material
Supplementary material associated with this article can be
found, in the online version, at 10.1016/j.combustflame.2018.03.
032.
References
[1] P. Goix, C. Edwards, A. Cessou, C. Dunsky, D. Stepowski, Structure of a
methanol/air coaxial reacting spray near the stabilization region, Combust.
Flame 98 (3) (1994) 205–219.
[2] A. Cessou, D. Stepowski, Planar laser induced fluorescence measurement of hy-
droxyl radical in the stabilization stage of a spray jet flame, Combust. Sci. Tech-
nol. 118 (4–6) (1996) 361–381.
[3] D. Stepowski, P. Goix, A. Cessou, Simple description of the combustion struc-
tures in the stabilization stage of a spray jet flame, At. Sprays 9 (1) (1999)
1–27.
[4] J.A. Friedman, M. Renksizbulut, Investigating a methanol spray flame inter-
acting with an annular air jet using phase-doppler interferometry and planar
laser-induced fluorescence, Combust. Flame 117 (4) (1999) 661–684.
[5] S.K. Marley, E.J. Welle, K.M. Lyons, W.L. Roberts, Effects of leading edge en-
trainment on the double flame structure in lifted ethanol spray flames, Exp.
Thermal Fluid Sci. 29 (1) (2004a) 23–31.
[6] S.K. Marley, K.M. Lyons, K. Watson, Leading-edge reaction zones in lifted-jet
gas and spray flames, Flow Turbul. Combust. 72 (1) (2004b) 29–47.
[7] G. Cleon, D. Honore, C. Lacour, A. Cessou, Experimental investigation of struc-
ture and stabilization of spray oxyfuel flames diluted by carbon dioxide, Proc.
Combust. Inst. 35 (3) (2015) 3565–3572.
[8] S.H. Starner, J. Gounder, A.R. Masri, Effects of turbulence and carrier fluid on
simple, turbulent spray jet flames, Combust. Flame 143 (4) (2005) 420–432.
[9] A.R. Masri, J.D. Gounder, Turbulent spray flames of acetone and ethanol ap-
proaching extinction, Combust. Sci. Technol. 182 (4–6) (2010) 702–715.
[10] J.D. Gounder, A. Kourmatzis, A.R. Masri, Turbulent piloted dilute spray flames:
flow fields and droplet dynamics, Combust. Flame 159 (11) (2012) 3372–3397.
[11] H. Correia Rodrigues, M.J. Tummers, E.H. van Veen, D.J.E.M. Roekaerts, Spray
flame structure in conventional and hot-diluted combustion regime, Combust.
Flame 162 (3) (2015a) 759–773.
[12] H. Correia Rodrigues, M.J. Tummers, E.H. van Veen, D.J.E.M. Roekaerts, Effects
of coflow temperature and composition on ethanol spray flames in hot-diluted
coflow, Int. J. Heat Fluid Flow 51 (2015b) 309–323.
[13] D.E. Cavaliere, J. Kariuki, E. Mastorakos, A comparison of the blow-off be-
haviour of swirl-stabilized premixed, non-premixed and spray flames, Flow,
Turbul. Combust. 91 (2) (2013) 347–372.
[14] R. Yuan, J. Kariuki, A. Dowlut, R. Balachandran, E. Mastorakos, Reaction zone
visualisation in swirling spray n-heptane flames, Proc. Combust. Inst. 35 (2)
(2015) 1649–1656.
[15] A. Verdier, J. Marrero Santiago, A. Vandel, S. Saengkaew, G. Cabot, G. Grehan,
B. Renou, Experimental study of local flame structures and fuel droplet prop-
erties of a spray jet flame, Proc. Combust. Inst. 36 (2) (2017) 2595–2602.
[16] L. Ma, D. Roekaerts, Numerical study of the multi-flame structure in spray
combustion, Proc. Combust. Inst. 36 (2) (2017) 2603–2613.
[17] A. Giusti, M. Kotzagianni, E. Mastorakos, LES/CMC simulations of swirl-sta-
bilised ethanol spray flames approaching blow-off, Flow, Turbul. Combust. 97
(4) (2016) 1165–1184.
[18] A. Giusti, E. Mastorakos, Detailed chemistry LES/CMC simulation of a swirling
ethanol spray flame approaching blow-off, Proc. Combust. Inst. 36 (2) (2017)
2625–2632.
[19] F. Shum-Kivan, J. Marrero Santiago, A. Verdier, E. Riber, B. Renou, G. Cabot,
B. Cuenot, Experimental and numerical analysis of a turbulent spray flame
structure, Proc. Combust. Inst. 36 (2) (2017) 2567–2575.
[20] L. Ma, B. Naud, D. Roekaerts, Transported pdf modeling of ethanol spray in
hot-diluted coflow flame, Flow, Turbul. Combust. 96 (2016) 469–502.
[21] I. Boxx, C.M. Arndt, C.D. Carter, W. Meier, High-speed laser diagnostics for the
study of flame dynamics in a lean premixed gas turbine model combustor, Exp.
Fluids 52 (3) (2010) 555–567.
[22] M. Stohr, R. Sadanandan, W. Meier, Experimental study of unsteady flame
structures of an oscillating swirl flame in a gas turbine model combustor, Proc.
Combust. Inst. 32 (2) (2009) 2925–2932.
[23] C. Abram, B. Fond, A.L. Heyes, F. Beyrau, High-speed planar thermometry and
velocimetry using thermographic phosphor particles, Appl. Phys. B 111 (2)
(2013) 155–160.
[24] C.D. Slabaugh, A.C. Pratt, R.P. Lucht, Simultaneous 5 kHz OH-PLIF/PIV for the
study of turbulent combustion at engine conditions, Appl. Phys. B 118 (1)
(2015) 109–130.
452 A. Verdier et al. / Combustion and Flame 193 (2018) 440–452
[25] M.S. Mansour, I. Alkhesho, S.H. Chung, Stabilization and structure of n-heptane
flame on CWJ-spray burner with kHZ SPIV and OH-PLIF, Exp. Thermal Fluid Sci.
73 (2016) 18–26.
[26] M. Cordier, A. Vandel, G. Cabot, B. Renou, A.M. Boukhalfa, Laser-induced spark
ignition of premixed confined swirled flames, Combust. Sci. Technol. 185 (3)
(2013) 379–407.
[27] H. Malm, G. Sparr, J. Hult, C.F. Kaminski, Nonlinear diffusion filtering of images
obtained by planar laser-induced fluorescence spectroscopy, J. Opt. Soc. Am. A
17 (12) (20 0 0) 2148–2156.
[28] S. Osher, J.A. Sethian, Fronts propagating with curvature-dependent speed: al-
gorithms based on hamilton-jacobi formulations, J. Comput. Phys. 79 (1) (1988)
12–49.
[29] P. Perona, J. Malik, Scale-space and edge detection using anisotropic diffusion,
IEEE Trans. Pattern Anal. Mach. Intell. 12 (7) (1990) 629–639.
[30] G. Hartung, J. Hult, R. Balachandran, M.R. Mackley, C.F. Kaminski, Flame front
tracking in turbulent lean premixed flames using stereo PIV and time-se-
quenced planar LIF of OH, Appl. Phys. B 96 (4) (2009) 843–862.
[31] P. Xavier, A. Vandel, G. Godard, B. Renou, F. Grisch, G. Cabot, M.A. Boukhalfa,
M. Cazalens, Investigation of combustion dynamics in a cavity-based combus-
tor with high-speed laser diagnostics, Exp. Fluids 57 (4) (2016) 50.
[32] J. Hult, U. Meier, W. Meier, A. Harvey, C.F. Kaminski, Experimental analysis of
local flame extinction in a turbulent jet diffusion flame by high repetition 2-D
laser techniques and multi-scalar measurements, Proc. Combust. Inst. 30 (1)
(2005) 701–709.
[33] M.K. Geikie, K.A. Ahmed, Lagrangian mechanisms of flame extinction for lean
turbulent premixed flames, Fuel 194 (2017) 239–256.
[34] P.H. Renard, D. Thevenin, J.C. Rolon, S. Candel, Dynamics of flame/vortex inter-
actions, Progr. Energy Combust. Sci. 26 (3) (20 0 0) 225–282.
[35] A.M. Steinberg, J.F. Driscoll, Straining and wrinkling processes during turbu-
lence–flame interaction measured using temporally-resolved diagnostics, Com-
bust. Flame 156 (12) (2009) 2285–2306.
[36] B. Franzelli, A. Vie, M. Ihme, Characterizing spray flame/vortex interaction: a
spray spectral diagram for extinction, Combust. Flame 163 (2016) 100–114.
[37] M. Nakamura, F. Akamatsu, R. Kurose, M. Katsuki, Combustion mechanism of
liquid fuel spray in a gaseous flame, Phys. Fluids 17 (12) (2005) 123–301.
[38] H. Nomura, M. Hayasaki, Y. Ujiie, Effects of fine fuel droplets on a laminar
flame stabilized in a partially prevaporized spray stream, Proc. Combust. Inst.
31 (2) (2007) 2265–2272.
[39] X. Mercier, M. Orain, F. Grisch, Investigation of droplet combustion in strained
counterflow diffusion flames using planar laser-induced fluorescence, Appl.
Phys. B 88 (1) (2007) 151–160.
[40] D.H. Wacks, N. Chakraborty, Flame structure and propagation in turbulent
flame–droplet interaction: a direct numerical simulation analysis, Flow, Turbul.
Combust. 96 (4) (2016) 1053–1081.
[41] D.H. Wacks, N. Chakraborty, E. Mastorakos, Statistical analysis of turbulent
flame–droplet interaction: a direct numerical simulation study, Flow, Turbul.
Combust. 96 (2) (2015) 573–607.