You are on page 1of 58

PERGAMON

Progress in Energy and Combustion Science 26 (2000) 225282


www.elsevier.com/locate/pecs

Dynamics of flame/vortex interactions


P.-H. Renard, D. Thevenin*, J.C. Rolon, S. Candel
Laboratoire E.M2.C., Ecole Centrale Paris and CNRS, Grande Voie des Vignes, F-92295 Chatenay-Malabry, France
Received 17 March 1999; received in revised form 30 December 1999; accepted 6 January 2000

Abstract
Vortex interactions with flames play a key role in many practical combustion applications. Such interactions drive a large
class of combustion instabilities, they control to a great extent the structure of turbulent flames and the corresponding rates of
reaction, they occur under transient operations or when flames travel in ducts containing obstacles. Vortices of various types are
often used to enhance mixing, organize the flame region, and improve the flame stabilization process. The analysis of flame/
vortex interactions has value in the development of our understanding of basic mechanisms in turbulent combustion and
combustion instability. The problem has been extensively investigated in recent years. Progress accomplished in theoretical,
numerical and experimental investigations on flame/vortex interactions is reviewed in this article. 2000 Elsevier Science Ltd.
All rights reserved.
Keywords: Flame; Vortex; Theory; Experiments; Numerical simulations; Combustion diagrams

Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2. Non-reacting vortex flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1. Structure and occurrence of vortices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2. Generation and evolution of a vortex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.1. Experimental vortices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.1.1. Vortex rings. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.1.2. Vortex dipoles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.1.3. Karman vortex streets. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.1.4. Difficulties in generating a single vortex. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.2. Simulated vortices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.3. Difficulties in flame/vortex interaction studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.3.1. Experimental case. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.3.2. Numerical cases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3. Investigation tools . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1. Configurations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1.1. Flame rolled-up in a single vortex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1.2. Flame in a shear layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1.3. Jet flames . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1.4. Karman vortex street/V-shaped flame interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1.5. Burning vortex ring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1.6. Head-on flame/vortex interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
* Corresponding author. Tel.: 33-141-131056; fax: 33-147-028035.
E-mail address: thevenin@em2c.ecp.fr (D. Thevenin).
0360-1285/00/$ - see front matter 2000 Elsevier Science Ltd. All rights reserved.
PII: S0360-128 5(00)00002-2

227
229
229
230
230
230
233
234
234
235
236
236
236
237
237
237
237
237
237
237
240

226

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

3.2. Formulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.1. First models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.2. Description of non-premixed flame/vortex interactions using the mixture fraction . . . . .
3.2.3. Numerical study of flame/vortex interactions using the full NavierStokes equations . . .
3.3. Experimental setups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.1. Premixed configurations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.1.1. Burning vortex ring. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.1.2. Karman vortex street/V-shaped flame interaction. . . . . . . . . . . . . . . . . . . . . .
3.3.1.3. Jet flame. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.1.4. Head-on laminar flame/vortex ring interaction. . . . . . . . . . . . . . . . . . . . . . . .
3.3.1.5. Single vortex ring/V-shaped flame interaction. . . . . . . . . . . . . . . . . . . . . . . . .
3.3.1.6. Vortex ring/counterflow twin-flame interaction. . . . . . . . . . . . . . . . . . . . . . . .
3.3.2. Non-premixed configurations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.2.1. Reaction front/vortex interaction in liquids. . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.2.2. Jet flames. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.2.3. Burning rings. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.2.4. Vortex ring/counterflow flame interaction. . . . . . . . . . . . . . . . . . . . . . . . . . . .
4. Results for non-premixed flames . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1. Theoretical results for non-premixed flames . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2. Numerical results for non-premixed flame/vortex interactions . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2.2. Flame structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2.3. Thermal expansion effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2.4. Buoyancy effects in jet flames . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2.5. Influence on mixing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2.6. Ignition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2.7. Extinction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2.8. Stretching effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2.9. Consideration of detailed chemistry and transport models . . . . . . . . . . . . . . . . . . . . . . .
4.3. Non-premixed experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3.1. Rollup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3.2. Mixing, diffusion effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3.3. Effects of stretch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3.4. Ignition, extinction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3.5. Lewis number effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3.6. Vorticity generation, baroclinic effects and thermal expansion effects . . . . . . . . . . . . . .
4.3.7. Gravity effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3.8. KelvinHelmholtz instability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5. Results for the premixed flame/vortex interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.1. Theoretical results for the premixed case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2. Numerical results for premixed flame/vortex interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.2. Flame structure and pocket formation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.3. Vorticity generation and dissipation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.4. Ignition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.5. Extinction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.6. Stretching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.7. Consideration of detailed chemistry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3. Premixed experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3.1. Flame propagation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3.2. Influence of vortex characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3.3. Effects of stretch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3.4. Ignition, extinction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3.5. Detailed chemistry effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3.6. Vorticity generation and baroclinic effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3.7. Lewis number effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

240
240
240
244
244
244
244
244
245
245
246
247
247
247
248
249
249
250
250
251
251
251
253
254
255
256
257
257
258
258
258
259
259
259
260
260
260
260
260
260
261
261
261
263
263
263
264
265
266
266
266
267
268
268
268
269

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

5.3.8. Gravity effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .


5.3.9. Time-dependent phenomena . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3.10. Thermal expansion effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6. Turbulent combustion diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.1. Turbulence description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.2. Turbulent premixed combustion diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.3. Turbulent non-premixed combustion diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1. Introduction
Vortices naturally occur in many reacting flows of technological interest. Vortical structures are found for example
in continuous combustors where their production is related
to the streams injected into the chamber and to the developing turbulent motion. Vortex motion is also established in
internal combustion engines as a result of reactant injection
and exhaust processes. In general, concentrated vorticity
constitutes the large-scale structure of the turbulent shear
flows found in combustion systems. Studies in turbulence
carried out during the last 30 years indicate that mixing is
controlled to a great extent by vortex motion and specifically
by the large-scale vortices developing in the highly sheared
regions of the flow. Much experimental evidence indicates
that turbulence may be described as an organized motion at
the largest scale superposed on a fine grain random background of fluctuations in the small scales. This picture has
evolved from experiments carried out on plane shear layers
[1] and has received further confirmation from studies on
jets and wakes (see for example Ref. [2]). These data have
suggested that turbulent combustion could be viewed as a
process dominated by the continuous distortion, extension,
production and dissipation of the flame surface by vortices
of different scales. This conceptual description has given
rise to a variety of flamelet models for turbulent combustion [3]. The elementary interaction between a vortex and a
flame thus appears as a key process in the description of
turbulent reactive flows.
Vortex structures also arise when a flame traveling in a
duct interacts with bluff obstacles, a mechanism which may
lead to significant levels of flame acceleration (Fig. 1(a)). A
starting vortex is also observed when jets or plumes are
formed by a sudden injection or expansion of a mass of
gases into a quiescent medium. The jet features a characteristic mushroom vortex cap (Fig. 1(b)). When dealing with
supersonic flows, it has been known from early experiments
on supersonic combustion that natural mixing was slow and
that combustors could not operate in the supersonic range
without some method of mixing enhancement. One possible
scheme for improving the rate of span-wise mixing relies on
vorticity created by the interaction between weak shock
waves and the hydrogen stream (Fig. 1(c)). This process

227

269
269
269
270
270
270
273
276
277
277

has been extensively investigated in relation with developments of supersonic ramjet combustors [4].
Vortex structures may also appear as a result of flow
instability. There is growing evidence that combustors operating in regimes of oscillation are driven by organized
vortices (Fig. 2). In many cases, the ignition and subsequent
reaction of these structures constitute the sustaining
mechanism by which energy is fed into the oscillation [5].
Vortex roll-up often governs the transport of fresh reactants
into burning regions and this process determines the rate of
reaction in the flow and the amplitude of the pressure pulse
associated with the vortex burn-out.
This panoramic view of practical devices and modes of
operation clearly indicates that vorticity and its dynamical
interaction with combustion are of great practical importance.
Vortex/flame interactions constitute also a fundamental
problem in combustion theory. The problem is generic as
it typifies the more complex situations found in turbulent
flows or in special modes of operation like those found
under combustion oscillation or in pulse combustion
devices. It is also valuable as an example of flame development in a non-uniform configuration. One of the simplest
examples of the effect of flow non-uniformity on a flame is
provided by the positive straining motion obtained for
example in the vicinity of a counterflow stagnation point,
possibly varying the strain-rate in time. This specific
problem has bee investigated most extensively because it
allows basic studies of flame structure, quenching and

Fig. 1. Some typical devices featuring flame/vortex interactions. (a)


Flame acceleration in a duct containing bluff obstacles. (b) Starting
jet and vortex pattern formed by the sudden injection and combustion of premixed reactants stored in a cavity. (c) Mixing enhancement in supersonic combustors by patterns of streamwise vorticity.
The mushroom shapes indicate the interface between the injected
fuel and the oxidizer flow.

228

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

Nomenclature
A
B
cp
d
D
Da
et
F
Fr
h
I
k
K
Ka
ld
lG
lk
lp
lq
lt
Le
m_ i
Mi
Mx
n
N
O
p
P
Pe
Pr
q
q
Q
rc
2R
Re
Ret
Rev
s
Sc
Sl
t
tf
T

Flame surface area


Preexponential factor
Specific heat at constant pressure
Nozzle diameter of the vortex generator
Species diffusion coefficient
Damkohler number tt =tc
Total Energy
Fuel species
Froude number
Enthalpy per unit mass
Hydrodynamic impulse
Turbulent kinetic energy per unit mass
Stretch rate
Karlovitz number tc =tk
Diffusive layer length scale
Gibson length scale
Kolmogorov length scale
Piston stroke
Quenching length scale
Turbulent large structure length scale
Lewis number
Consumption rate of species i
Molecular weight of species i
Mixing index
Normal direction
Number of azimuthal waves on a vortex ring
Oxidizer species
Pressure
Product species
Peclet number
Prandtl number
Heat release per unit mass of fuel
Heat flux
Heat release per unit mole of fuel
Vortex core radius
Vortex ring diameter/vortex pair core
distance
Reynolds number
Turbulent Reynolds number
Vortex Reynolds number
Mass stoichiometric ratio
Schmidt number
Laminar flame speed
Time
Flame time
Temperature

ignition conditions, effects of strain on reaction rates, flame


response and dynamics etc. Flames submitted to constant or
variable strain rate provide a remarkably rich example of the
effect of flow non-uniformity. The flame/vortex interaction
constitutes the next step in the search for simple model

Ta
u0
uc
uk
uG
ui
up
uq
ur
uu
umax
u
v
V
Yi
Z

Activation temperature
Turbulent large structure velocity scale
Convection velocity of a vortex
Kolmogorov velocity scale
Gibson velocity scale
i component of velocity
Piston velocity
Quenching velocity
Radial velocity
Azimuthal velocity
Maximal azimuthal velocity of a vortex
Velocity vector
Volume
Mass fraction of species i
Mixture fraction

Greek symbols
a
Thermal diffusivity
df
Flame thickness
dr
Reaction zone thickness in physical space
e
Turbulent dissipation rate
ei
Strain rate in i direction
es
Strain rate in the plane tangent to the flame
G
Circulation
G0
Initial circulation of a vortex
l
Thermal conductivity
v
Kinematic viscosity
n 0i
Stoichiometric coefficient of reactant i
n 00i
Stoichiometric coefficient of product i
Inverse of the Kolmogorov time
v
r
Density
sr
Reaction zone thickness in mixture fraction
space
tc
Chemical time scale
tk
Kolmogorov time scale
tm
Mechanical time scale
tq
Quenching time scale
tt
Turbulent large structure time scale
t
Viscous stress tensor
w
Equivalence ratio
f
Global mixture ratio
F
Stream function in the frame of the vortex
x
Scalar dissipation rate
c
Stream function in the frame of the
laboratory
vi
ith component of vorticity
v
Chemical molar production rate
v
Vorticity vector

geometries. This case offers new possibilities, as it allows


for example investigations of straining and curvature effects
in the same configuration.
Much has been learned in the recent past from theoretical,
numerical and experimental studies of flame/vortex

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

229

Fig. 2. Observations of vortex structures in unstable combustors. (a) High frequency screech instability [6]. (b) Low frequency instabilities in
a backward facing step geometry [7]. (c) Low frequency instabilities of a dump combustor. (d) Large scale vortices in a premixed shear layer
[8,9]. (e) Vortex driven instability in a multiple jet dump combustor [10]. (f) Vortex driven oscillation in a single jet dump combustor [11]. (g)
Organized vortex motion in a premixed ducted flame modulated by plane acoustic wave [11,12]. (h) Control of a dump combustor using
organized vortices [13]. (i) Vortex motion during the injection phase of a pulse combustor [14].

interactions. Problems which may be examined in this


geometry relate to the structure of the flame when it is
wound up by a vortex, formation of a central core, secondary
vorticity generation, quenching of the reaction zone, ignition dynamics, mixing and combustion enhancement.
This article provides a review of the research effort
devoted to the problem of a flame interacting with a vortex.
This survey is restricted to situations where vortex structures
are isolated (single vortex, vortex pair, vortex rows, vortex
ring) or form a well-organized pattern [15]. It does not touch
upon the more complex situation encountered in turbulence
where the vortex filaments are randomly distributed in
space. The review begins in Section 2 with some background material on vortex dynamics, stability and decay,
structure, and gives some ideas of the difficulties involved
in flame/vortex interaction studies. The third section reviews
a set of theoretical studies of simplified geometries including non-premixed and premixed flames interacting with a
single vortex or a pair of vortices. Numerical calculations
are also surveyed in this section, and the main formulations

used for these theoretical investigations are given. A short


description of the main experimental setups follow (Section
3.3). The most significant results for both non-premixed
(Section 4) and premixed (Section 5) configurations are
then detailed on the theoretical, numerical, and experimental levels. To give an idea of the usefulness of flame/vortex
interactions studies, their use in improving turbulent
combustion diagrams is briefly described in Section 6,
along with the latest versions of these diagrams.

2. Non-reacting vortex flows


2.1. Structure and occurrence of vortices
Before dealing with flame/vortex interactions, it is first
useful to provide some background on vortex structures in
non-reactive streams, give some elements on their experimental and numerical generation and discuss stability
issues.

Fig. 3. Generation of a vortex ring using a piston, visualized by streaklines (from Ref. [22]).

230

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

Fig. 4. Transition from a laminar to a turbulent vortex ring. Visualization of streaklines (from Ref. [22]).

Two-dimensional vortex dipoles (or pairs) and axisymmetrical vortex rings are common in nature. Vortex dipoles
feature circular streamlines with two cores symmetrically
located on the two sides of the propagation axis. Their structure
is well described by a Lamb dipole [16]. Vortex rings are
composed to a first approximation of an inviscid toroidal
core and some fluid entrained around it. The Lamb dipole,
on the contrary, has the shape of a flattened ellipsoid. The
core approximately takes a circular section when the size of
this core diminishes with respect to the torus total diameter.
For both the vortex dipoles and rings, viscous diffusion causes
ring diameter growth, with ambient fluid being entrained into
the core. The total circulation then decays, thereby decreasing
the translation velocity of the vortical structure. This decay
process leads to eventual annihilation of all vorticity initially
contained in the core.
While the two-dimensional vortex pair configurations are
not commonly observed in practice, they are quite common
in many numerical studies. The reason is probably that
direct simulation codes are mostly two-dimensional (it is
well known that direct simulations should be carried out
in three dimensions. However, due to the long times
required for the calculation of reacting flows, many
investigations address only two-dimensional problems
using DNS techniques). One familiar occurrence of vortex
pairs is found in aircraft wakes: two counter-rotating
vortices develop from the wing trailing edges. The flow
induced by these free vortices produces a vertical downwash. Small aircrafts entering in the wake of wide body
jet liners may be exposed to a strong downward motion
and this may lead to serious accidents during landing.
Other examples of vortex pairs are found within thermal
plumes, rising bubbles, and within convective instabilities
such as those occurring in a breaking gravity wave. It is also
possible to experimentally generate the two-dimensional
turbulence featuring a collection of vortex dipoles by
moving a bluff body in a thin sheet of a solution of soapy
water [17,18].
Vortex rings are easier to observe over a broad range of
scales. Some volcanic eruptions produce large dust and ash
rings, while smaller rings are easily produced with the
smoke of a cigarette [19]. Under special meteorological
conditions, one observes small windbursts called microbursts, which are modeled as huge toroidal vortices

impacting on the ground. Some sea mammals blow ring


bubbles and swim through them for amusement [20]; in a
similar manner, cuttle-fish expel a jet of ink when they feel
endangered (this starting jet is preceded by a vortex ring). If
an electric field is applied to ions in superfluid helium II (at
a temperature of about 0.28 K), toroidal vortices appear
with a quantized electric charge. The core of these vortex
[21]. In practical applications, vortex
structures is about 1 A
rings appear when a fluid is suddenly exhausted through a
circular orifice.
2.2. Generation and evolution of a vortex
2.2.1. Experimental vortices
2.2.1.1. Vortex rings. Vortex rings are most commonly
produced by pushing fluid through a nozzle with a circular
cross section. The nozzle may be replaced by an orifice in
the wall or a cylindrical pipe. The vortex sheet induced in
this process, generated by a shedding of the boundary layer
within the orifice, is then rolled up (Fig. 3). In practical setups, the mass of fluid is set in motion with a piston or a
loudspeaker. The volume of fluid injected is perfectly
controlled with a piston, which is not so true with a loudspeaker. In any case, the backward movement of the actuator is to be avoided to prevent flow reversal. This is easy to
achieve with a rigid piston moving forward, but more difficult with a flexible loudspeaker membrane. However, vorticity production with a piston also needs some additional
precautions. The piston must slide perfectly, be rigid and
light enough to minimize the delay between the electrical
signal and the mechanical response. When the piston stops,
negative vorticity is generated, and this must also be taken
into account.
For a given geometry, the vortex structure is controlled to
a great extent by the Reynolds number which may be based
on the average piston velocity up and nozzle diameter d
Rev

up d
n

To define the Reynolds number, one may also use the maximal azimuthal velocity umax
and the vortex ring diameter 2R
u
Rev

2umax
u R
v

Vortex centered on flame

Flame rollup, mixing effects

Flame structure, pockets, core


formation

Marble and Karagozian [65,66]

Marble and Karagozian [65,66]

Rolon, Renard and Thevenin


[110]

Laverdant and Candel


[75,130,132]
Rehm et al. [69]
Ashurst [134,135]
Laverdant and Candel [130]
Rolon, Thevenin and Renard
[108110]

Vortex centered off flame


Head-on collision
between a vortex and a
single flame

Vorticity generation and


dissipation, baroclinic effects,
thermal expansion

Macaraeg and coworkers


[155,157]
Thevenin and Candel [156]

Ashurst [134,135]
Takahashi and Katta [124,158]

Laverdant and Candel [130]

Burning ring

Park and Shin [121]

Reactive shear layer

Riley et al. [147]

Hewett and Madnia [151]


You et al. [165]
Delhaye et al. [76]

Lee et al. [149]

Lee et al. [149]

Laskey et al. [148]


Grinstein et al. [150]
Mueller and Scheffer [120]

Yamashita et al. [62]

Jet flame

Ignition, extinction

Hewett and Madnia [151]


You et al. [165]
McMurtry and coworkers
[138,139]
Grinstein and Kailasanath [141]
Mahalingam et al. [140]
Chen et al. [115]
Katta and coworkers [163,164]

Thevenin, Renard and Rolon


[109,110,123]
Cuenot and Poinsot [168]
Hewett and Madnia [151]
You et al. [165]
Givi and Jou [153]
Ghoniem and coworkers
[152,154]
Laskey et al. [148]

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

Table 1
Points investigated with different kinds of non-premixed flame/vortex interactions. To make this table and the followings more explicit, some authors names have been added

231

232

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

Fig. 5. Generation of a vortex dipole in a stratified fluid. Visualization of streaklines (from Ref. [45]).

It is also convenient to use the circulation G of the vortex


Re v

G
n

The previous definitions of Rev are similar but not identical. The third expression generally yields values about 10
times larger than the first two. At low values of Re, vortex
rings feature a smooth toroidal surface and a laminar velocity profile [23,24]. As the initial strength increases and the
Reynolds number becomes larger, small amplitude
azimuthal waves appear around the torus. A further increase
of the initial circulation leads to shedding of material into
the surrounding fluid. This gives rise to a trailing wake, the
extent of which increases with the initial strength. The toroidal shape is no longer observed after the rollup process.
Instead, a volume of fluid moves forward shedding an appreciable quantity of fluid into a trailing wake, where smaller
scale vortical structures may be distinguished. Eventually, a
well-defined vortex ring emerges and keeps moving
forward. In the vortex bubble, the velocity field displays
the same basic pattern as in the laminar case but with
high-frequency low-amplitude fluctuations superimposed
on top of it, characteristic of turbulent flows [23,24].
At sufficiently low Reynolds numbers, the vortex has a
laminar structure which depends on [25]:
the history of piston velocity;
the shape of the exhaust section which controls the outlet
velocity profile;
the ratio of piston stroke to nozzle diameter;

the ratio of piston diameter to exit diameter should be


considered [26].
For large values of the Reynolds number, the situation
completely changes. The vortex becomes turbulent (Fig.
4). The vortex structure loses the memory of initial and
boundary conditions. After a distance of a few nozzle
diameters, the vorticity distribution only varies with the
vortex speed and dimensions. The toroidal vortex features
two zones [27]. The first one is a rapidly rotating core of
very fine scale turbulence for which the ratio of the torus
radius to the core radius remains constant as the vortex
propagates. The second one is the outer region of the spheroid entrained with the torus that is also turbulent, but the
mean vorticity within it is low and the turbulent scales are
large compared to those characterizing the torus. The latter
region is responsible for fluid entrainment: fluid is carried by
the large-scale corrugations of the outer interface and mixes
with the inner fluid. Most of this fluid is then rejected in the
wake and only a small amount is retained causing the very
slow growth of the vortex ring. The vortex core interface is
only weakly distorted and it entrains fluid at a reduced rate.
The intermediate case is also interesting as stationary
azimuthal waves propagate in the vortex core [28]. These
waves are observed as a transient state during the generation
of a turbulent vortex. First, a vortex annulus is produced by
the rollup of a fluid slug at the exit of the vortex generator.
Then an azimuthal wave mode develops on the ring core.
The number of waves N may be predicted theoretically for a
vortex ring submitted to a weak steady irrotational plane
strain [29]. The straining acts in a plane perpendicular to

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

the vortex axis, yielding:


N

0:34k 1 s1 R
nt 1=2

where k 1 and s1 are dimensionless functions of the magnitude of the strain rate. The value of k 1s1 decreases from 2.51
to 1.65 for typical values of strain rate considered, R is the
torus radius, n the kinematic viscosity and t is a duration
time since a theoretical birth. The wave amplitude grows
until it breaks preferentially, i.e. the breaking does not occur
at all azimuthal locations at the same time. This nonuniformity creates an azimuthal flow along the core that
induces a wave motion. Finally, a turbulent spot is formed.
Further details on turbulent vortex rings are given by
Maxworthy [28].
We have chosen here the Reynolds number as the main
parameter of the vortex ring. It would be equally possible to
investigate the generation of the vortex ring using the hydrodynamic impulse I, which is defined along the z direction (ez
vector) in an axisymmetric case by:
ZZ
It rp
r2 vu r; z; t d2 Aez
5
A

where r is the radial distance and v u is the corresponding


component of the vorticity. The importance of the hydrodynamic impulse is well explained in Ref. [30], and we refer
the interested reader to this book.
As we are essentially interested in combustion applications, only phenomenological aspects of the vortex generation are given below. Shariff reviews some basic knowledge
on general structures of vortex rings [26]. Aspects such as
velocity fields, time evolution of circulation and vorticity
are treated in many Refs. [23,24,28,3133]. The detailed
vorticity distribution is not the key parameter to describe
toroidal vortex dynamics in both laminar and turbulent
cases [34]. The behavior of the ring seems to depend on
the generator geometry: vortex rings ejected from an orifice
entrain less fluid than those ejected from a sharp-edged tube,
and the former also move faster [24].
As vortex rings are quite common, many theoretical
studies have also been carried out. The vortex speed has
been studied extensively [27,31,3538]. The vortex speed
of a viscous thin-core vortex ring is given by Saffman [35].
In the derivation, it is assumed that the vortex speed may be
characterized by the translational velocity of the centroid
associated to the vorticity field yielding

 

G0
8R
uc

0:558
6
ln
4pR
4nt 1=2
where uc is the vortex speed, G 0 the vortex annulus circulation, R the torus radius, and t a virtual birth time of the
vortex. This formulation is not easy to use in practice
because the position of this vorticity centroid is hard to
determine from experiments. Being aware of this limitation,
Wang et al. [38] treated the problem in another way. Their
derivation does not rely on a series expansion of the solution

233

of the (c v ) equation system, but is based on matched


asymptotic expansions giving the velocity of the impulse
centroid. They found that
 



G
8R
G
uc
ln
;
t
7

H
m
4pR
v
4nt 1=2
where Hm is a function of Reynolds number G /v and time t. It
is worth noting that Hm tends to 0.558 when G /v tends to
infinity.
The vortex generation mechanisms have been also extensively investigated in the laminar case [2,3942]. Turbulent
effects were also considered [27]. A widely used description
of vortex formation is the slug flow model first derived by
Saffman [41]. Previous models were assuming that the
vortex ring is formed by the vorticity contained in an
initially circular vortex sheet [39]. Saffman underlined the
fact that such models did not properly account for the
circulation contained in the annulus. He also emphasized
the difference between model assumptions and what is
observed in practice. It was proposed on this basis to substitute a cylindrical vortex sheet to the circular one. From a
calculation first given without error by Maxworthy, one may
estimate the vortex circulation as [28]

tu p2
2

where  denotes the average operator on the time interval of


the piston rise. Experiments carried out by Didden [43]
indicate that

tu p2
A for
2

tup
1
lp

where A is a constant including all the contributions


neglected in the slug flow model and lp is the piston stroke.
2.2.1.2. Vortex dipoles. The experimental generation of
vortex pairs is also based on the brief discharge of fluid.
The orifice must be rectilinear and it should feature a
large aspect ratio (Fig. 5). Trajectory and time evolution
of the physical quantities were studied [18,4446]. Some
of these studies use special techniques to avoid threedimensional effects. Physical mechanics which are used to
suppress the transverse motion and obtain a twodimensional flow are quite diverse. One may use the
gravity in a stratified fluid [45], the Coriolis force in a
rotating homogeneous fluid, the surface tension in a soap
film [17] or a magnetic field in a layer of mercury [18]. In
studies dealing with a constant density fluid, the interest has
been focused on three-dimensional aspects, principally
because two-dimensional vortices undergo a threedimensional instability known as the Crow instability.
This long-wavelength instability has also a shortwavelength counterpart which is related to zero-frequency
Kelvin twist waves, but it has usually a smaller growth rate
than the long-wavelength instability [30].

234

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

Table 2
Points investigated with different kinds of non-premixed flame/vortex interactions (continued)
Stretch effects

Detailed chemistry effects Buoyancy effects

Vortex centered on flame


Karagozian and Marble [67]
Vortex centered off flame
Ashurst [135]
Head-on collision between a Cuenot and Poinsot [168]
Takahashi et al. [162]
vortex and a single flame
Thevenin, Renard and Rolon Renard et al. [110]
[123,110]
Burning ring
Head-on collision between a Karagozian et al. [112]
vortex and a liquid reaction
Reactive shear layer
Jet flame
Takahashi et al. [169]
Katta and coworkers
[119,163,164]
Roquemore and coworkers
[118,160,161]
Mueller and Scheffer [120]

Theoretical results are also available about the existence


and uniqueness of vortex pairs, their similarity with Lamb
vortices and various other aspects [2,30,45,4755]. From
Eulers equations, an infinite family of stationary solutions
exists for which vz f c where f is an arbitrary integrable
function of c . A particular case is the Lamb dipole
( 2
l F inside a circle of radius R
10
vz
0
outside
where l is a constant and F designates the stream function
in the frame of reference moving with the dipole, i.e. F
c u c x for a dipole moving in the y direction with the
velocity uc. The analytical solution is for a vortex dipole
moving at a velocity uc
8
>
< 2luc J1 lr cos u for r R
11
v J0 lR
>
:
0
for r R

8
>
>
>
<


2u c
J1 lr uc r cos u for r R
lJ0 lR

>
>
R2
>
:
u cos u
r c

12

for r R

where c is in the frame of the laboratory and Jn is the nth


order Bessel function of the first kind. The continuity of
vorticity across the separating stream line imposes that l R
be the first zero of J1 so that

lR 3:831

13

2.2.1.3. Karman vortex streets. As Karman vortex streets are


easy to produce (see for example Fig. 9), they have been
used in flame/vortex interaction experiments and in some
calculations. The vortex street is produced by the flow
around a cylinder at a Reynolds number of about 100. The

Lewis number effects

Katta et al. [124]

Roquemore and coworkers Hancock et al. [119]


[142,118,143146]
Oran and coworkers [63,64] Katta et al. [160]

cylinder slows down the fluid forming a wake featuring a


double shear layer structure. This configuration corresponds
to the superposition of two vortex sheets of opposite signs.
Both layers are subject to the KelvinHelmholtz instability
(which is initiated by a small perturbation of the velocity
field in the vicinity of the cylinder). The two sheets then
degenerate into a double row of vortices rotating in opposite
directions. A peculiarity is that these vortices are alternated
which is a characteristic feature of two-dimensional
instabilities. Linear instability theory predicts that this
configuration of alternated vortices corresponds to the
highest rate of instability amplification. When the
Reynolds number is increased to several thousands so that
the cylinder wake becomes turbulent at small scale, coherent
structures survive but they are less deterministic.
2.2.1.4. Difficulties in generating a single vortex. From an
experimental point of view, the ideal vortex generation
pattern corresponds to the slug flow configuration
described above. Unfortunately, parasitic phenomena
perturb this pattern of vortex formation. These
perturbations are listed by increasing order of importance.
At large times after the beginning of the vortex generation, the velocity is much larger in the center of the outlet
than the piston velocity because of the growing displacement thickness of the internal boundary layers of the
nozzle [43].
At early stages, the vorticity flux into the nascent toroidal
vortex is higher than expected from the slug flow model
because the emerging fluid turns sharply outward near the
corner [43];
Vorticity of opposite sign is generated in the boundary
layer formed on the outer pipe wall at the exit through the
influence of the induced velocity field of the growing
vortex [43].

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

235

Fig. 6. Flame rolled-up in a single vortex structure.

After the piston comes to rest, experiments depict the


formation of an opposite-sign secondary vortex ring
called stop vortex which propagates into the vortex
generator. This vortex absorbs a small amount of the
primary vorticity [32,43].
An instability of the cylindrical vortex sheet may appear.
This is similar to a KelvinHelmholtz instability of a free
shear layer and it generates secondary vortices behind the
primary one [28,56,57]. This instability may also appear
behind a vortex pair too, with the difference that the
secondary vortices are then alternated.
Azimuthal instabilities of the vortex core and transition to
turbulence may perturb the smooth arrangement of
streamlines in the vortex.
Some of these disturbances may be attenuated. For example, Glezer designed his experimental setup so that the
piston would be flush with the exit plane wall [56]. This
arrangement eliminates any strong secondary flow at the
end of the piston stroke and reduces some of the effects of
the image vortex on the primary vortex.
The growth of secondary vortices behind the primary one
is only approximately described with the KelvinHelmholtz
instability theory because the shear layer momentum thickness, velocity profile, and curvature continuously change. It
is also known that there exists a limiting value for the ratio
of piston stroke to nozzle diameter, and vortex rings
generated with ratios exceeding this limiting value do
not absorb all of the discharged fluid mass or vorticity.
This limiting value corresponds to the case where

vorticity in the shear layer of the trailing jet ceases to


flow into the vortex core region (after pinch-off) just
when the piston stops [57].
Finally, it is not possible to scan a large interval of vortex
sizes and velocities because of viscous decay and transition
to turbulence at high Reynolds numbers. It is sometimes
conjectured that disturbances introduced by the instability
of the vortex sheet during the formation of the vortex ring
can accelerate the onset, amplification, and breakdown to
turbulence of the azimuthal instability [56].
2.2.2. Simulated vortices
Fortunately, most of the problems which arise in experiments do not appear in calculations in which the vortex
generation is not simulated. There are at least two ways to
numerically generate a vortex ring or a vortex dipole. One
can first try to simulate the generation process involved in
reality. This allows a better description of what exactly
happens in the experiment but requires a careful calculation
of the sudden expulsion of fluid. The second method consists
of imposing a given vorticity distribution in the numerical
domain. This is less physical but more convenient. Even if
the initial distribution does not exactly obey the Navier
Stokes equations, after a few iterations using these equations, the vorticity field is reorganized and satisfies the equations of motion. The second method does not produce a
realistic vortex ring (i.e. featuring the velocity field existing
in reality) but its characteristics are perfectly defined and the
range of size and strength which may be scanned is much
larger than in practical set-ups.

Fig. 7. Flame/vortex interaction as found in a developing shear layer.

236

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

Fig. 8. Flame/vortex interaction as found in a jet flame.

Because the initial vorticity distribution rapidly adapts to


the physical laws, one may use simple vorticity fields to
generate a vortex pair or a vortex annulus. The simplest
one is the HamelOseen vortex which represents the
viscous decay of an infinite straight line vortex with an
initial circulation G 0. In this special case, the NavierStokes
equations may be solved algebraically. One finds
"
!#
G0
r2
u 0 r; t
1 exp
14
2pT
4nt
where uu is the azimuthal velocity component, r the radial
distance to the initial line vortex, n the kinematic viscosity
and t is the time since the beginning of the decay. This time
is virtual as a vortex line is an idealization of a uniformly
distributed vorticity in a cylindrical tube. The flow field
given by Eq. (14) is used as a starting flow in two-dimensional and axisymmetric cases. Another classical

Fig. 9. Interaction of a V-shaped flame with a Karman vortex street.

formulation is the Lamb-dipole given by Eq. (12). Numerical simulations and experiments show that it is a good
approximation of a vortex pair structure [17,18,46,53,55].
2.2.3. Difficulties in flame/vortex interaction studies
2.2.3.1. Experimental case. The proper generation of a
vortex pair or a ring is, as indicated previously, a difficult
problem. It is also not easy to make it interact with a
well-controlled flame. It is important to define a simple
flame geometry to allow repeatable experiments and simple
interpretation. One may anchor a premixed flame behind an
obstacle (rod, bluff body, etc.) or stabilize it in a counterflow. One may set a non-premixed flame with similar
configurations. Unfortunately, the background flow will
disturb the vortex structure and possibly destroy it before
it interacts with the flame. The influence of this background
flow must be examined to exactly quantify the respective
influences of the flame and of the flow on the vortex structure. The effect of a shear flow is examined in the case of a
vortex pair [58] or in the transverse jet geometry [5961].
Conclusions of these studies may be used to characterize the
relevance of different experimental configurations.
2.2.3.2. Numerical cases. In the premixed case, one has to
stabilize the flame in the computational domain. This can be
carried out by first running a one-dimensional code to
compute the laminar burning velocity. This velocity is
taken as inlet velocity at one domain boundary ensuring
that the flame is stabilized in it.
For the non-premixed case, the flame is first formed by
creating an initial layer of contact. An ignition process is
then used to establish a diffusion flame. The problem is
then that the flame has no fixed thickness (the flame is indeed
becoming thicker with time). Hence, it is not so easy to
extract quantitative information from this kind of flame/
vortex interactions. A possible solution is to simulate a counterflow flame, with the understanding that full simulations
may become quite expensive in this case. Jet diffusion flames
and mixing layers can also be employed for such studies.

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

237

3.1.2. Flame in a shear layer


As a further step, the interaction between flames
developing in a two-dimensional shear layer and the vortical
structures naturally responsible for the growth of this layer
has been investigated (Fig. 7). In this case, the spatial evolution of the flame is of special interest. Several vortical structures are then simultaneously found, and the mutual
interactions between these structures and the flame increase
the complexity of the simulations. This situation comes
closer to experiments. Investigations of coherent motion in
reactive shear layers are quite numerous. We will only
consider cases where the interaction involves a quasiperiodic row of vortices.

Fig. 10. Development of a flame rolled-up in a vortex ring.

3. Investigation tools
3.1. Configurations
As already indicated, a basic issue in flame/vortex interactions studies is to define the configuration which allows
the simplest interpretation. Theoretical studies are restricted
to simple geometries that are not always close to reality, but
they give insight in the basic mechanisms. Numerical and
experimental studies allow investigations of more realistic
cases, even if they are also limited with respect to the attainable parameters. An experimental setup should allow an
easy optical access and a good repeatability. In numerical
calculations, one must ensure that the number of grid points
is sufficient to describe the combustion and fluid dynamics,
that the boundary conditions are compatible and preserve
good numerical stability and that the level of unwanted
reflections is at a minimum. Different configurations, corresponding to various possible applications, are summarized
in the following subsections. A summary is given in Tables
1 and 2 for non-premixed and in Tables 3 and 4 for premixed
flames.

3.1.1. Flame rolled-up in a single vortex


Flames rolled-up in a single vortex structure growing
along the interface between fuel and oxidizer (or fresh
and burnt gases) constitute a fundamental geometry
(Fig. 6). This corresponds to a situation where analytical
studies are possible. These configurations have also been
simulated first for validation purposes. They have also led
to interesting information concerning the dynamics of
flame ignition and spreading. In this configuration, the
velocity field is considered to be solely caused by the
vortex structure, and the computations give the time
evolution of the flame. In the premixed case, the flame
front propagation cannot be avoided and this complicates
the interpretation.

3.1.3. Jet flames


Organized vortex motion is also studied in the case of jet
flames (Fig. 8). As the evolution of large scale structure is
governed by KelvinHelmholtz instabilities, the frequency,
mutual interaction and energy distribution among various
length-scales is controlled by the initial conditions of the
flow. Combustion processes strongly modify the instability
mechanisms but coherent structures similar to those
observed in non-reacting flows may be generated in an
annular jet diffusion flame by forcing it at a preferred
mode frequency. The same instability process that amplifies
the axisymmetric disturbances and generates these structures is amplifying higher modes of instability introducing
the three dimensionality into the flow. In the case of an
axisymmetric configuration, the azimuthal instability
modes which may be amplified and interact with each
other form an infinite set. This makes the analysis of the
modulated jet flame configuration not so easy to study. In
this case again, the influence of the vortical structures naturally present in this configuration on the developing flame
has been investigated. We will only list publications where
the interaction between flames and vortices is a major issue.
3.1.4. Karman vortex street/V-shaped flame interaction
The previous configurations are not easily described in
terms of vortex structures which interact with the flame
front. The interaction takes place while the vortices develop.
The vortex pattern is influenced by the flame presence as it is
being generated. This problem is avoided by considering the
interaction of a branch of a V-shaped premixed flame with a
Karman vortex street (Fig. 9). The vortex street develops
externally and the vortex row subsequently interacts with
the flame.
3.1.5. Burning vortex ring
One early experiment on flame/vortex interactions
concerns the propagation of a flame front in a vortical structure of premixed reactants (Fig. 10). This configuration may
be used in both premixed and non-premixed cases. This
configuration may also be investigated numerically and
this was carried out more recently.

238

Vortex centered on flame


Head-on collision between
a vortex and a single flame

Flame rollup

Flame structure, pockets, core


formation

Vorticity generation and


dissipation, baroclinic effects,
thermal expansion

Peters and Williams [172]


Laverdant and Candel [132,130]

Peters and Williams [172]


Vassilicos and Nikiforakis [181]
Laverdant and Candel [132,130]

Ashurst and McMurtry [182]

Rutland and Ferziger [173,174]


Poinsot et al. [175,176]

Burning ring
Head-on collision between
a vortex and a double
flame
Periodic vortex array

Jet flame

Driscoll and coworkers


[98,100,104,177]
Lee, Santavicca and coworkers
[186,187]
McCormack [81,82]
Cattolica and Vosen [83]
Renard et al. [111]

Ignition, extinction

Rutland and Ferziger [173,174]

Poinsot et al. [175,176]

Louch and Bray [184]


Driscoll and coworkers
[98,101,103]

Jarosinski et al. [94]


Driscoll and coworkers [98,99]

Renard et al. [111]

Ashurst [178]
Dold and coworkers [179,180]
Lee, Santavicca and coworkers
[196]

Helenbrook et al. [188]

Shu et al. [93]


Katta and Roquemore [183]

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

Table 3
Points investigated with different kinds of premixed flame/vortex interactions

Vortex centered on flame


Head-on collision between
a vortex and a single flame

Burning ring
Head-on collision between
a vortex and a double
flame
Periodic vortex array

Jet flame

Stretch effects

Detailed chemistry effects

Gravity effects

Lewis number effects

Driscoll and coworkers [95,99]

Hilka et al. [189]

Driscoll and
coworkers
[102,103]

Rutland and Ferziger [173,174]

Najm and coworkers


[191,192,195]
Lee, Santavicca and coworkers
[186,187]

Katta and Roquemore [190]

Poinsot et al. [175,174]

Najm and coworkers


[191,192,195]
Nguyen and Paul [107]

Driscoll and coworkers [97,103]

Renard et al. [111]

Lee, Santavicca and coworkers


[92,186,187,196]
Helenbrook et al. [188]

Lee, Santavicca and coworkers


[91,196]
Shu et al. [93]
Katta and Roquemore [190]

Katta and Roquemore [190]

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

Table 4
Points investigated with different kinds of premixed flame/vortex interactions (continued)

239

240

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

Fig. 11. Interaction between a flat flame and a counter-rotating vortex pair or toroidal vortex structure.

3.1.6. Head-on flame/vortex interaction


The need to study a very simple configuration is well
fulfilled in the configuration of a head-on interaction
between an initially flat flame and a counter-rotating vortex
pair or ring. This initially flat (or approximately flat) flame is
either a one-dimensional unstrained flame, a counterflow
flame or a V-shaped flame (Fig. 11). The time evolution
of the flame front has been investigated in great detail,
with particular emphasis on flame structure, extinction
limits, pocket formation, effects of vortex size and strength.
Experimental results have sometimes allowed direct
comparisons with numerical calculations.
3.2. Formulations
Several formulations were used to study flame/vortex
interactions from a theoretical and a numerical point of
view. Table 5 summarizes the configurations and topics
investigated.
3.2.1. First models
The flame/vortex interaction was initially described with
analytical techniques. In the case of a diffusion flame rolled
up by a centered vortex as formulated by Marble [65], the
position of the interface is determined by following a flame
element. The local flow field around the local reactive
element is then obtained and the strain rate acting in the
direction tangent to the flame is calculated. Using this strain
rate, one may determine the reaction rate per unit surface.
The size of the core may be estimated by calculating the
mean distance between adjacent flame layers and estimating
the characteristic time required to consume the reactants
separating the successive layers.
The previous method may be assimilated to a Lagrangian
analysis of the motion and subsequent consumption. It

provides the general structure of the flow and may be


extended to other geometrical configurations including
stretched vortices [66], semi-infinite and infinite fuel strips
[67] etc. Similar techniques may be used to study the
premixed flame roll up [68].
The analytical Lagrangian scheme is limited, however, to
rather simple geometries. One possible extension is to use a
numerical version of this approach [6971].
The numerical Lagrangian approach works well except in
regions where the flame sheets come close to each other and
interact strongly. It is therefore preferable to exploit more
standard numerical methods. In the non-premixed flame
case, the mixture fraction formulation is most suitable
when using the infinitely fast chemistry assumption, as
explained in the next section.
3.2.2. Description of non-premixed flame/vortex
interactions using the mixture fraction
Interactions between a non-premixed flame and a vortex
or a collection of vortices are conveniently formulated in
terms of the mixture fraction. This formulation may be
derived for flows in which the density changes with
temperature but it is most valuable when the density
variations associated with the heat release are neglected.
This thermodiffusive approximation is widely used in theoretical investigations [7274,59] and has found many applications in numerical studies of flame vortex interactions
[75,76].
Consider a single-step irreversible chemical reaction

n 0F F n 0O O ! n 00P P

15

which may also be written in the form


F sO ! 1 sP

16

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

241

Table 5
Numerical techniques used in flame/vortex interaction simulations
Non-premixed flame

Premixed flame
Wu and Driscoll [177]
Ashurst [178]
Dold, Kerr and Nikolova [179,180]
Vassilicos [181]
Ghoniem and Givi [152]
Lee and Santavicca [186,187]
Helenbrook et al. [188]

Mixture fraction, infinitely fast chemistry

Mixture fraction, finite rate chemistry

Detailed simulations, simple chemistry

Detailed simulations, detailed chemistry

Marble and Karagozian [6567]


Laverdant and Candel [75,132,130]
Delhaye et al. [76]
Rehm et al. [69]
Macaraeg et al. [155]
Thevenin and Candel [156]
Ashurst [134]
McMurtry and coworkers [138,139]
Mahalingam et al. [140]
Roquemore and coworkers [142146]
Takahashi and Katta [158]
Hewett and Madnia [151]
Katta and coworkers [124,160164]
Thevenin and coworkers [109,110,123]

where s designates the mass stoichiometric ratio


s

n 0O MO
n 0F MF

n 00 M
1 s P0 P
n F MF

Rutland and Ferziger [173,174]


Ashurst and McMurtry [182]
Poinsot, Veynante and Candel [175,176]
Louch and Bray [184]

Katta and coworkers [93,183]


Najm, Wyckoff and coworkers [191,192,195]
Hilka and coworkers [189]

One may introduce in the last expression the heat release per
unit mole of fuel
17

Q DH MO n 0O hO MF n 0F hF MP n 00P hP

24

or the heat release per unit mass of fuel


18

It is assumed that species transport may be described by


Ficks law, thermal diffusion of species may be neglected,
all diffusion coefficients are constant and equal, specific
heats are equal and constant, the Lewis numbers are
constant and equal to one, and the rate of change of pressure
with time 2p=2t may be neglected.
Under these assumptions, the balance of species F, O and
P and the balance of energy written in terms of temperature
take the form
2
rYF 7rvY F 7rD7Y F MF n 0F v_
2t

19

2
rYO 7rvY O 7rD7Y O MO n 0O v_
2t

20

2
rYP 7rvY P 7rD7Y P MP n 00P v_
2t

Laverdant and Candel [132,130]

21

2
rcp T 7rvc p T 7l7T
2t

22

v_ M P n 00P hP MO n 0O hO MF n 0F hF

23

q Dhf hF

MO n 0O
MP n 00P
h
0 hO
MF n F
MF n 0F P

DH=MF n 0F

25
26

The balance of energy becomes


2
rcp T 7rvc p T 7l7T Qv_
2t

27

Consider a flow configuration in which reactants are initially separated by an arbitrary interface. One may examine
for example a flat interface separating the fuel and the oxidizer. In the upper stream y ! ; Y F YF0 ; YO
0; T TF0 while in the lower stream y ! ; YO
YO0 ; YF 0; T TO0 : One may define a global mixture
ratio from conditions prevailing in the two streams of reactants

MO n 0O YF0
Y
s F0
MF n 0F YO0
YO0

28

This quantity measures the ratio of fuel to oxidizer in the


injected streams with respect to the ratio corresponding to
stoichiometric conditions. In fact, f is the inverse of the f
factor of Karagozian and Marble [66]. An examination of
the balance equations and boundary conditions indicates

242

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

that one may form coupling functions that satisfy a balance


equation without source term. As an example the variable

bF

YF
YO

MF n 0F
MO n 0O

29

satisfies an equation of the form


2
rbF 7rvb F 7rD7b F
2t

30

It is more convenient to scale the previous variable and to


define a mixture fraction which takes a value of one in the
fuel stream and of zero in the oxidizer stream: Z 1 at y !
and Z 0 at y ! : One may relate Z to the
coupling function b F by
Z

bF bF
bF bF

31

The mixture fraction also satisfies a convectivediffusive


balance equation
2rZ
7rvZ 7rD7Z
2t

32

Because the previous equation has no source term it is much


easier to handle and solve than one of the balance equations
for the chemical species. Now, it is useful to write Z in terms
of the fuel and oxidizer mass fractions
YF
YO
YO0

MF n 0F
MO n 0O
MO n 0O
Z
YF0
YO0

MF n 0F
MO n 0O

33

34

35

1
1f

Zf

38

On the fuel side (for Z Zf ) there is no oxidizer (YO 0)


and consequently
YF
f1
YF0
Z
f1

YF YF0

39

f 1Z 1
f

40

On the oxidizer side (for Z Zf ) there is no fuel, YF 0


and thus

YO
1
YO0
f1

41

or equivalently
YO YO0 1 f 1Z

42

The product mass fraction may be deduced in a similar


manner. On the fuel side Z Zf


n 00 M
Y
YP P00 P YF0 Z F
43
n F MF
YF0
which may be written as
YP

or equivalently as
YO
Y
00 P
n 0O MO
n P MP
Z
YO0
n 0O MO

37

The previous expressions are particularly useful if one


considers that chemistry is infinitely fast. Under this
assumption which corresponds to the BurkeSchumann
limit, reaction takes place in a thin sheet which separates the
reactants. It is then possible to deduce the chemical variables and the temperature from the mixture fraction Z. This
feature has been used extensively in the modeling of nonpremixed turbulent combustion [74,7779]. It is also quite
useful when one wishes to calculate interactions between
vortices and flames. When chemistry is infinitely fast, fuel
and oxidizer cannot coexist. At the flame sheet YF YO 0
and the value of Z at the flame may be deduced from Eq.
(34). This yields

One may define other coupling functions by forming


linear combinations of mass fractions. The mixture fraction
is thus expressed alternatively as
YF
Y
00 P
n 0F MF
n P MP
Z
YF0
n 0F MF

T
Y
T
0 F O0
Q=cp
n F MF
Q=cp
Z
TF0
Y
T
0 F0 O0
Q=cp
n F MF
Q=cp

or equivalently

and to introduce the global mixture ratio in the previous


expression
YF
Y
f O 1
YF0
YO0
Z
f1

This yields for example

1s
Y 1 Z
f F0

44

On the oxidizer side corresponding to Z Zf one finds


36

One may also form coupling functions with the temperature.

YP 1 sYF0 Z

45

On the flame sheet both expressions yield


YPf

1s
Y
1 f F0

46

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

243

Fig. 12. A typical non-premixed configuration. The flame sheet is rolled-up by streamwise vortices. The maximal value of 7Z is found in the
braids connecting the vortices (from Ref. [80]).

The temperature may be deduced from Z in a similar


fashion. It is also necessary to distinguish the fuel and oxidizer sides. On the fuel side Z Zf
Q=cp YF0 1 Z
T TO0 ZTF0 TO0 0
n F MF
f

47

Q=cp YF0
TF0 fTO0
0
1f
n F MF 1 f

48

49

which may be written in the alternate form


Tc

YO0 MF n 0F TF0 YF0 MO n 0O TO0 Q=cp YF0 YO0


YO0 n 0F MF YF0 n 0O MO

2YF
2n

m_ F rD7Y F n

Both expressions give the same value for the temperature at


the flame
Tc

m_ F rD

51

Using the relation established for YF in terms of Z one finds

On the oxidizer side Z Zf


Q=cp
T TO0 ZTF0 TO0 0
Y Z
n F MF F0

reactants per unit of flame surface. Consider for example the


consumption rate of fuel

m_ F rD

1f
YF0 7Zn
f

52
53

The normal direction n appearing in this expression is


defined by the gradient of Z
n

7Z
j7Z j

54

and thus
50

The previous expressions completely determine the flame


structure from the sole knowledge of the mixture fraction Z.
The problem is reduced to the solution of the convectivediffusive balance equation (Eq. (32)). The position of the
flame sheet is then determined as the contour where Z
Zf 1=1 f: Mass fractions and temperature are then
obtained from the linear expressions derived above. It is
important to note that one may extract the flame structures
of flames corresponding to different values of the global
mixture ratio f from a single solution of the mixture fraction balance equation. This greatly simplifies the analysis of
the influence of initial compositions.
It is also possible to calculate the rates of consumption of

m_ F rD

1f
YF0 7Z
f

55

This expression has to be evaluated at the flame sheet (i.e.


on the surface Z Z f ). Fig. 12 shows a typical result for a
shear layer. It shows that maximal values of 7Z are found
in the reactive braids connecting the vortices. Similar calculations may be carried out to estimate m_ O : Starting from
m_ O rD

2Y0
2n

56

one finds
m_ O rD1 fYO0 7Z

57

The previous expressions are not independent and one

244

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

may easily check that


m_ O
Y
f O0 s
m_ F
YF0

58

which indicates that reactants are consumed at the flame in


stoichiometric proportions. Expressions (55) and (57) are
applicable at each point belonging to the flame sheet. Both
expressions feature the absolute value of the mixture fraction gradient 7Z indicating that fuel and oxidizer are
consumed in regions where this gradient takes large values.
In contrast, low values of 7Z indicate regions of low reactivity (Fig. 12).
3.2.3. Numerical study of flame/vortex interactions using the
full NavierStokes equations
If a more precise solution is needed, the next step is to use
a low-Mach number formulation, or eventually the full
compressible NavierStokes equations, as
2ru j
2r
0

2xj
2t

59

2rui uj
2tij
2rui
2p

i 13
2xj
2t
2xj
2xj

60

2et puj
2uj tkj
2qj
2et

2xk
2t
2xj
2xj

61

2rYl uj
2rYl VDl;j
2rYl

Wl v_ l l 1 to Ns 62
2xj
2xj
2t
where r represents the density, ui the velocity in direction i,
Yl the mass fraction of species l, and et the total energy.
It is then possible to employ simplified models to describe
the reaction and diffusion processes, usually a single-step
irreversible reaction with Arrhenius reaction rate, Ficks law
for diffusion of species and Fouriers law for heat conduction. It is also possible in this frame to use full reaction
mechanisms and detailed multicomponent diffusion velocities. In the latter case, the computation time is greatly
increased, but the accuracy of the solution is improved.
Both kinds of models have been used to investigate the
flame/vortex interaction problem, as will be seen in later
sections.
3.3. Experimental setups
It is worth describing at this point the different setups
which are used to study flame/vortex interactions. As
experimental investigations are less easy to design and
develop, they are also less numerous and most of them
concern the premixed cases.
3.3.1. Premixed configurations
3.3.1.1. Burning vortex ring. The earliest experiments
carried out in the field of flame/vortex interactions deal

with the spreading of a flame front in a vortical structure


of premixed reactants. The first experiment of this kind
appears to be that of McCormack [81,82]. The vortex
rings were formed by injecting a certain amount of
reactants through a circular orifice by displacing a pneumatically driven piston in a cylindrical cylinder. The
vortex formed in this process propagates through a
combustion chamber and is ignited by the corona
discharge from a sharp wire located downstream.
A similar experiment was designed by Cattolica and
Vosen [83]. The vortex develops as a puff of premixed
reactants is pushed out through a circular orifice separating
a small cylindrical prechamber from the main combustion
vessel. This impulsive jet is produced by a spark-ignited
flame propagating in the prechamber. Reactants cross the
main vessel followed by the flame and they are eventually
overtaken by reaction. A laser sheet was used to excite OH
molecules and a low-light-level vidicon camera to monitor
the resulting fluorescence, in the case of methaneair
mixtures.
The same configuration was later used by Ishizuka et
al. [84] but with the vortex generation principle of
McCormack, which seems to be easier to control
(Fig. 13). Flame velocities were measured by a highspeed video camera, and the maximum tangential velocities in the vortex was obtained with a hot-wire
anemometer.
3.3.1.2. Karman vortex street/V-shaped flame interaction.
To study the interaction of turbulence with combustion,
Namer et al. [85] designed a new experiment involving a
V-shaped premixed flame and a Karman vortex street. The
combustor consisted of a jet, through which premixed
ethylene, air and oil spray was injected. A flame was
anchored on a rod placed in this flow and surrounded
with a coaxial jet of air to prevent perturbations by the
stagnant room air. Vortices were generated by a second
rod placed upstream at various positions with respect to
the flame-holder.
Hertzberg et al. [86] performed Mie scattering on oil
droplets on this configuration. It was found that the inboard
vortex created a large incursion of unburnt gases, while the
outboard vortex generated a smaller one leading to a cusp.
The large incursion of fresh gases due to the inboard vortex
was damped whereas the cusp produced by the outboard one
was convected downstream at the flow velocity.
Esudie and Charnay [8789] employed the same geometry and studied it extensively. They investigated many
cases using laser-Doppler velocimetry to obtain the axial
and transverse velocity fields of the vortex streets and Mie
scattering to image the shape of the flame front.
More recently, Lee, Santavicca and coworkers [9092]
used a similar layout and showed that effects of local flame
stretch could be assessed by direct experimental measurements using particle-image velocimetry (PIV) and flow
visualization (Fig. 14).

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

245

Fig. 13. Experimental setup for studying premixed burning vortex rings (from Ref. [84]).

3.3.1.3. Jet flame. An alternative configuration is the inverse


partially premixed flame established by injecting a fuel-rich
annular jet sandwiched between a central air jet on the inside
and coflowing air on the outside [93]. The burner features
two concentric cylinders with circular plates at both top and
bottom of the burner. Fuel is injected through the outer
cylinder and air through the inner cylinder. The top disk
has a circular exit for air and fuel. The wall of this exit is
pierced by six symmetrically located holes through which
the fuel flows radially inward from the space between the
two cylinders. Air flows into the bottom of the inner cylinder
and out through the circular opening in the top plate. Air and
fuel mix in the exit of the burner and form a partially
premixed flame due to premixed burning inside the
annular jet and diffusion burning in the jet shear layer
(Fig. 15). Large vortical structures propagate around this
reactive layer. They are due to the shear layer instability
associated with the buoyant acceleration of hot gases
outside the flame surface. The diagnostics used on this
configuration were conventional and high-speed videos.
The overall structure of the flame was observed by
seeding the flow allowing an evaluation of the flickering
frequency of the flame.
3.3.1.4. Head-on laminar flame/vortex ring interaction. The
former configurations are interesting but interpretation is
complicated. Indeed, the structure of a burning vortex
ring, especially in the turbulent regime, is not easy to
characterize. In the Karman vortex street case, a periodic
hydrodynamic perturbation does not allow the isolation of
mechanisms occurring in a flame/vortex interaction.
Moreover, the idea that a turbulent flame could be
approximated as a collection of laminar flame elements
interacting with vortices in the large Darmkohler limit
highlighted the need to design new experimental
configurations. Jarosinsky et al. [94] described an
experiment consisting of an upwardly propagating,
premixed flame interacting with a vortex ring traveling

downwards. The setup comprises a vertical flame tube


including two opposing side glass walls, a top section
used as a vortex generator bounded by a loudspeaker and
an orifice plate, and a lower section with an electrode to
provide spark ignition (Fig. 16). Vents are added close to
the orifice plate to ensure that the burner does not blow up
when highly reactive mixtures are used. To perform an
experiment, a mixture of reactants is injected in the tube.
A spark, produced by the electrode, generates a laminar
flame which propagates upwards in the tube. A vortex ring
is generated by the excitation of the loudspeaker at a
selected time so that it interacts with the flame in front of
the glass windows. Schlieren visualization was used to
investigate the flame front dynamics. Driscoll and
coworkers [95] studied a similar configuration with OH
fluorescence imaging and to identify the different regimes
of a premixed flame/vortex interaction [96,97]. A two-color
particle-image velocimetry was also used to observe how a
single toroidal vortex exerts aerodynamic strain on a
premixed flame [98]. The distributions of the
instantaneous tangential strain rate and of the stretch rate
K 1=AdA=dt were measured along a distorted premixed

Fig. 14. Experimental setup for studying Karman vortex street/Vshaped premixed flame interactions (from Ref. [90]).

246

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

Fig. 15. Experimental setup for studying premixed jet flames with vortices (from Ref. [93]).

flame contour, as the flame interacted with a vortex. Mole


fraction profiles of OH were measured as a function of
stretch rate for an unsteady, repeatable process which
includes much of the physics associated with turbulent
flames [99]. The three-dimensional, instantaneous stretch
rate along the flame front was used to compare results to
calculations of steady, planar counterflow flames. Particle
image velocimetry (PIV) was used to obtain a time sequence
of velocity field images and thus compute spacetimeresolved, three-dimensional flame stretch rates and
dilatation rate fields [100,101]. Dilatation rates are
maximum in the flame front, and they may be used to
characterize the flame position. Moreover, the peak value
at each flame segment was referred to as the flame strength
and used as a qualitative indicator of the local peak heatrelease rate. Flame-generated vorticity, which previously
had been predicted but never quantified, was measured.
To precisely determine the influence of the baroclinic
torque, Driscoll and coworkers used the NASA Lewis
Research Center drop tower facility to study the
differences of the flame shapes in one-g and microgravity
conditons [102,103]. Because experimental conditions were
quite challenging for optical diagnostics, only direct flame
chemiluminescence imaging and Mie scattering images
were obtained. Local flame propagation speeds were also
measured from shadowgraph images [104].

3.3.1.5. Single vortex ring/V-shaped flame interaction. The


interaction between a vortex ring and a propagating flame
constitutes an attractive geometry but it has drawbacks
because the unperturbed flame is not easy to produce. The
propagation of the flame strongly depends on the quality of
the reactants mixing which must be carefully controlled.

Fig. 16. Experimental setup for studying head-on premixed laminar


flame/vortex ring collisions (from Ref. [94]).

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

247

Fig. 17. Experimental setup for studying a single vortex ring/Vshaped premixed flame interactions (from Ref. [107]).

The vortex generation produces a pressure wave that


perturbs the flame front before the interaction. Finally, in
microgravity, the flame tends to remain spherical as it is
produced by a spark. So, an interaction with a flat flame is
very unlikely in this case and remains a problem under
normal gravity conditions. Taking these issues into
account, Samaniego designed a novel system which would
have the same advantages as the previous experimental
device and in which the flame would be spatially
stabilized using a rod [105,106]. The flame structure is
two-dimensional and planar contrary to the shape of most
confined propagating flames. A two-dimensional vortex
dipole generator was also developed to preserve the twodimensional nature of the flow. The experimental facility
comprises a vertical duct of square cross-section equipped
with quartz windows for optical access. A mixture of
propane or methane and air is fed into the test section
through a contoured converging nozzle (Fig. 17). The Vshaped flame is stabilized on an electrically heated wire and
a vortex pair is generated by acoustic excitation through a
contoured slot in one of the duct walls. Smoke visualizations
were performed to characterize the cold vortex pair
structure. The flame front dynamics were studied with
Schlieren techniques.
More recently, this configuration was employed by
Nguyen and Paul [107]. They carried out OH and CH planar
laser-induced fluorescence imaging to visualize the flame
front dynamics.
3.3.1.6. Vortex ring/counterflow twin-flame interaction. The
configuration of a V-shaped flame interacting with a single
vortex dipole was designed to answer the problems
encountered
in
the
head-on/propagating
flame
configuration. Unfortunately, new problems arose. The
vortex being two-dimensional is likely to be destroyed by
Crow instabilities and this process is accelerated in a crossflow geometry. Moreover, the dipole needs to be strong
enough to prevent dissipation in the shear layer before

Fig. 18. Experimental setup for studying vortex ring/counterflow


twin premixed flame interactions (from Ref. [111]).

impinging on the flame front. Consequently, the


appearance of secondary vortices is often encountered as
shown in [105] and may complicate the interpretation of
what occurs. These difficulties may be overcome by
adapting the non-premixed configuration of the burner
designed by Rolon [108110] to the premixed case [111].
The apparatus consists of a counterflow burner comprising
two opposite identical nozzles between which two premixed
flames are stabilized symmetrically on both sides of the
stagnation plane. These two flames are exposed to the
same constant relatively low strain rate. The main
advantage of this configuration lays in the flames being
almost flat, well stabilized in the flow and the separation
distance between them being easily controllable. Each
injection nozzle is surrounded by an annular nozzle. The
reactant mixture is injected through the central nozzle
while the outer sheet of nitrogen insulates the reactive
stream from external disturbances. Along the axis of the
lower nozzle, a cylindrical tube is connected to a
cylindrical plenum chamber and generates a single
laminar vortex ring through the displacement of a volume
of mixture by a piston located in this chamber. This is a
particularly interesting geometry allowing investigations
of flame/vortex interactions for much smaller vortices than
those studied in other geometries. It also allows studies of
interactions between colliding flame fronts (Fig. 18). To this
point, only OH and CH spontaneous emission imaging has
been performed.
3.3.2. Non-premixed configurations
3.3.2.1. Reaction front/vortex interaction in liquids. The
earliest experimental configuration analyzing flame/vortex
interaction dynamics in the non-premixed case was a liquid
phase setup involving liquid acid/base reactions [112]. A
water tank with aluminium and Plexiglas walls allows

248

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

interface between the acid and base, initially established at


the exit of the side plates, is stretched and rolled up by the
vortex pair. Flow visualization was achieved with planar
laser-induced fluorescence of a dye mixed with the acid
solution.

Fig. 19. Experimental setup for studying liquid reaction/vortex


interactions (from Ref. [112]).

optical access. The vortex generator is set in the lower


portion of the tank, consisting of two-side plates forming a
convergent two-dimensional nozzle. A thin aluminium plate
separates the base solution from the slightly lighter acid
solution. A vortex dipole can be formed by the introduction
of base liquid pumped from external tanks through the lower
fill valves. This vortical structure of base liquid that fills the
lower part of the tank is pushed into the acid liquid (Fig. 19).
During the vortex propagation into the acid solution, the

3.3.2.2. Jet flames. Jet flames were extensively studied by


Roquemore and coworkers. The burner is composed of a
fuel nozzle located at the center of a coaxial jet assembly
which is mounted in a small vertical combustion tunnel
[113]. The annular air jet has a divergent section and flow
straightener, which consists of a honeycomb and fine mesh
screens, to provide low turbulence surrounding air for the
fuel jet (Fig. 20). Experiments were carried out using Mie
scattering on TiO2 particles providing a marker for the
interface of the cool air and H2O combustion product
outside the flame and the cool fuel/H2O product interface
inside the flame. To achieve this, TiCl4 was added to both
the dry air stream and to the dry propane fuel flow [113
115]. Titanium tetrachlorine reacts with water produced in
the reaction zone and yields particles of TiO2 and HCl
molecules.
In this configuration, vortices are generated from hydrodynamic instabilities. These instabilities only appear at
specific excitation frequencies. It is also interesting to
force other frequencies with a loudspeaker to scan a broader
range of frequencies. This was carried out by Strawa and
Cantwell in mid 1980s [116]. The experiments were carried
out in a facility capable of enclosing a combusting flow at
elevated pressures. The air is injected in the test section
while a nozzle in the center of the section supplies fuel.
The fuel jet was submitted to a periodic fluctuation. By
forcing the jet in a range of frequencies encompassing its
unforced natural frequency, it was possible to produce
a periodic controllable flow. Gutmark et al. studied a
configuration based on the same principle using planar

Fig. 20. Experimental setup for studying jet diffusion flames (from Ref. [113]).

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

249

Fig. 21. Experimental setup for studying burning non-premixed


rings (from Ref. [121]).

laser-induced fluorescence on OH radicals [117]. Hsu et al.


focused on the effect of acoustically driven excitations on
the co-annular jet configuration [118] performing Mie scattering and OH fluorescence. More recently, Hancock et al.
measured temperature using CARS techniques [119]. A
two-dimensional configuration based on those previous
experiments was also built [120].
3.3.2.3. Burning rings. As in the premixed case, some
experimental studies were carried out on burning nonpremixed rings. One study is due to Park and Shin [121].
A jet of fuel is injected in ambient air from a cylindrical
tube. A steady jet diffusion flame is initially set at the exit of
this fuel tube. The fuel flow is stopped and the flame
separates in two parts. The outer flame lifts and burns
away, while the residual flame of the size of the nozzle
diameter enters through the nozzle and travels upstream.
The fuel valve is then opened again with a time delay and
the fuel is ejected again from the nozzle. A fuel vortex ring
is formed at the nozzle exit and is immediately ignited by
the residual flame (Fig. 21). Flame visualization was carried
out with a high-speed Schlieren system and tip maximum
flame temperature was measured with a thermocouple.
Another system of the same type adapted to microgravity
operation was built by Chen and Dahm [122]. The vortex
ring is generated by issuing gaseous fuel from a plenum
through a contoured axisymmetric nozzle. An iris in the
exit plane of the nozzle initially separates the fuel from air
in the test section. At a selected time, the iris is opened to
form a diffusion layer at the fuelair interface, and opens a
solenoid valve to a partial vacuum to draw this diffusion
layer into the nozzle throat until it reaches a spark, which
ignites the diffusion layer. A diffusion flame forms in the
nozzle throat, at which point the microgravity drop tower
package is released. After the flame has flattened, a second
solenoid valve opens to a pressurized fuel cylinder, which
impulsively pushes fuel from the plenum through the nozzle
into a test section containing air at atmospheric pressure. A
vortex ring is formed and wraps the diffusion-reaction layer

Fig. 22. Experimental setup for studying vortex ring/counterflow


diffusion flame interactions (adapted from Ref. [108]).

around itself. These experiments were performed in microgravity conditions in the NASA Lewis Research Center 2.2
s-drop tower facility. A CCD camera was used to collect the
resulting flame luminosity.

3.3.2.4. Vortex ring/counterflow flame interaction. The


head-on interaction between a non-premixed flame and a
vortex cannot be studied in the previous geometries. This
interaction is however of greatest interest in studies of
turbulent combustion. With this in mind, Rolon et al.
devised a new setup including a counterflow and a vortex
generator [108]. A steady non-premixed counterflow flame
of air and hydrogen diluted with nitrogen is first established.
This counterflow is surrounded by two nitrogen curtains to
suppress external disturbances and prevent the appearance
of a diffusion flame around the fuel stream. A vortex ring is
generated from a tube installed in the lower combustor
nozzle and impinges on the flame (Fig. 22). The axial
velocity field was measured by laser Doppler anemometry
(LDA) and direct visualizations of the reaction zone were
conducted. In another contribution, Thevenin et al. made
some visualizations of the vortex structure with Mie
scattering and they used spontaneous emission imaging
to track the flame front motion [109]. Later on, Renard
et al. used the planar laser-induced fluorescence (PLIF) on
OH radicals to follow the flame evolution [110] and
theoretical analyses were also carried out to interpret
the data [123]. One advantage of this configuration is
that it may be used to study the premixed case as well.
Katta et al. used the same setup to study complementary
features of the non-premixed flame/vortex interactions
[124127].
Other research groups explore similar combustion
dynamics through the study of pulsed and continuous microjet injection [128] in a counterflow geometry similar to that
of Rolon and coworkers.

250

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

Fig. 23. Geometrical configurations investigated in non-premixed flame/vortex interaction studies. (a) Vortex centered on a diffusion flame. (b)
Vortex centered off a diffusion flame. (c) Vortex pair impinging on the flame. (d) Vortex stretched along its axis at a constant rate. (e) Vortex
pair interacting with an infinite fuel strip. (f) Vortex pair interacting with a semi-infinite fuel strip. (g) Vortex row centered on the flame, (h)
Flame interacting with organized vortices in a shear layer. (i) Vortex motion in a transverse fuel jet. (j) Forced vortex convected in a jet diffusion
flame.

4. Results for non-premixed flames

follow a similarity rule involving the same parameters

4.1. Theoretical results for non-premixed flames

m_ v =m_ 0 CG 2=3 D1=3

Theoretical studies of flame/vortex interactions were


initiated by an elegant analysis devised by Marble [65] to
understand the basic features of the problem. The configuration investigated comprises a simple plane diffusion flame
initially at rest along the horizontal axis. A vortex with a
circulation G is established at the origin at the initial time
t 0 (Fig. 23(a)).
The vortex induces an aximuthal motion described by Eq.
(14). It is assumed that the chemistry is infinitely fast and
that the heat release does not change the density r . Using the
last assumption, also designated as the thermodiffusive
approximation, it is possible to decouple the fluid mechanics
of the problem from the determination of the flame structure. In fact the problem just defined may be handled analytically as remarkably demonstrated by Marble. Using a
kinematic description of the flame surface roll-up and a
local analysis of the straining flow acting on the reactive
elements composing the flame, Marble was able to describe
the structure of the flow field in the large and to evaluate
some of its main features. The flame comprises a reacted
core and two spiral arms extending from the core to their
original position on the axis. A scaling law for the rate of
growth of the core radius was obtained in the form

64

The previous rules were shown to hold for sufficiently


large Reynolds numbers, the exact condition being that
Re v Sc1=2 50 where Rev G=2pn and Sc n=D: The
related problem of mixing in the field of a viscous core
vortex was also considered by Marble [129] using the
same analytical framework. The geometry is that sketched
in Fig. 23(a) but the species do not react. The vortex
produces a well-mixed core with a radius rc given by the
following scaling rules
r c 0:440G 2=3 D1=3 t1=2 for large m a

65

rc 0:047G 2=3 D1=3 t1=2 m a for ma 4

66

63

using the dimensionless number ma G D =n: It was


shown that mixing could be characterized by a mixing
index Mx. Consider two species 1 and 2 which are initially
unmixed and respectively occupy a fraction j and 1 j of
the total available volume V. Let Y 10 and Y20 designate the
initial mass fractions of the two species. After complete
mixing in the volume V the final mass fractions will be
jY 10 and 1 jY 20 : One may then define the mixing index
by
Z
1
Mx
Y Y dV
67
0
0
j1 jY 1 Y2 V V 1 2

where b is a constant equal to 0.259. The radius increases


like t 1/2 and its rate of change is determined by (G 2/3D 1/3) 1/2.
The augmentation of reactant consumption was estimated to

If the species are totally unmixed, the index equals


zero. Complete mixing will yield Mx 1: The field of
mixing index may be defined in terms of two dimensionless

r 2 b G 2=3 D1=3 t

2=3

1=3

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

parameters

Mx Mx

r
; ma
G 2=3 D1=3 t1=2


68

The mixing index rapidly reaches unity inside the core: at


r rc =2; Mx 0:95: If the two species being mixed
combine according to a slow reaction characterized by
dY1
1
Y1 Y2
tc
dt

69

one may show that the total amount of product increases


according to



G 2=3 D1=3 t c t
t
mP
70
2 ln 1
tc
2tc
b
which yields for large times
mP

G 2=3 D1=3 t
b

71

The analysis of Marble [65] was later extended by Karagozian and Marble [66] to the case of a vortex stretched
along its own axis, a situation which models in some
sense vortex stretching in three dimensional turbulent fields
(Fig. 23(d)). For a constant rate of strain in the z direction,
the flow field is defined by
er
ur z ;
2
"
!#
G
ezr 2
uu
1 exp
; uz ez z
2pr
4n1 expez t
72
When the flame is exposed to this flow field it rolls-up and
forms a burned core. The scaling law for the core radius
becomes
"
  1=2 !#
r2
1 expe z t
D
73

ez
G
G 2=3 D1=3
After a time of order 1/e z, the core radius approaches a
constant value r2 =G 2=3 D1=3 b =ez : The rate of consumption in the presence of the stretched vortex may be compared
to the rate of consumption in a flat flame submitted to the
same strain rate e z in the z direction. It is found that the ratio
of these two quantities is proportional to (G 2/3D 1/3). This
result is identical to that obtained for the vortex developing
in the absence of axial stretch. As noted by Karagozian and
Marble the difference between the two cases lies in the way
products are stored in the flow field. For the two-dimensional vortex, the product core radius grows like t 1/2 and
products occupy a progressively larger area. For the
stretched vortex, the core reaches an asymptotic value
fixed by the rate of strain and products are then convected
out by the straining process.
The previous studies were extended in various directions.
Karagozian and Manda [67] considered the flame distortion
caused by a vortex pair interacting with a two-dimensional

251

fuel strip (Fig. 23(e) and (f)). Two geometries were studied:
in the first, the diffusion flames were bounding an infinitely
long strip of fuel, in the second the vortex pair centers coincide with the corners of a semi-infinite fuel strip. This last
case provides a two-dimensional analog of the flame/vortex
structure formed when a fuel jet is exhausted impulsively
from a nozzle. The flow field is specified by placing a pair of
vortices of circulation ^G 0 at a distance 2R and adding a
velocity u c G=4pR to keep the vortex pair in a fixed
position. The corresponding stream function has the form
" 2
#
G
x y R2
74
c u c y 0 ln
2p
x2 y R2
The calculated flame shapes are conveniently described in
terms of reduced time and spatial scales t tuc =R and r
r=R: It is also useful to introduce a parameter f representing
the fueloxidizer stoichiometric ratio. This parameter
may be related to the more common ratio f sYF0 =YO0
which measures the proportion of fuel to oxidizer in the
injected streams with respect to stoichiometric conditions.
It is easily shown that f 1=f: When f 1 or equivalently f 1 more oxidizer is required than fuel to generate
the combustion products. The limiting reactant will then
depend on the value of this parameter: it is the fuel
when f 1 and the oxidizer when f 1: In the first
case fuel will be depleted. Considering the streamline
annulus formed by the counter-rotating vortex pair, the
limiting reactant will disappear first leaving a region of
combustion products and an excess of the other reactant.
Flame patterns obtained by Karagozian and Manda [67]
are close to those calculated by Laverdant and Candel
[130] in the geometry sketched in Fig. 23(c). The global
mixture ratio f is in this case equal to unity and the
vortex is centered on the flame.
4.2. Numerical results for non-premixed flame/vortex
interactions
4.2.1. Introduction
As we have seen previously, analytical approaches still
require simple configurations and generally imply many
restrictive hypotheses. Numerical simulations have thus
been used to support theoretical and experimental findings.
We will now summarize the main results obtained in these
numerical investigations.
4.2.2. Flame structure
First numerical calculations of non-premixed axisymmetric jet flames used infinitely fast chemistry and a
phenomenological induction parameter to introduce chemical ignition delays [131]. This study employed a FluxCorrected Transport (FCT) algorithm, and revealed that
the active reaction zones appear like a continuous layer
wrapped around the eddies. This is so because the oxidizer
is rapidly consumed inside the vortex cores f 1: The

252

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

Fig. 24. Product mass fraction distribution for a vortex pair with
Rev 50 and f 1; for reduced times 0.005 and 0.013 (from Ref.
[132]).

vortex structures are quickly smeared out because of the


heat release leading to a high dissipation level.
First studies of the roll-up of a non-premixed flame by an
isolated vortex considered a thermodiffusive approximation
(constant density), constant and equal thermodynamic and
transport properties for all species, and an infinitely fast
single-step reaction [75,130,132]. These computations of a
flame rolled-up by a growing viscous vortex investigated the
formation and extension of the vortex core and examined the
flame structure. The velocity field of the viscous vortex was
prescribed by Eq. (14). Comparisons with analytical predictions for the core radius and consumption rate augmentation

(see previous section) show a very good agreement, with the


core growing at a faster pace for increasing vortex Reynolds
number Rev G=2pn: Varying the global mixture ratio,
the flame is displaced towards the defect reactant, and the
core of burnt products is mostly surrounded by oxidizer
(respectively, fuel) in the lean (respectively, rich) case.
Calculations were also carried out in situations which had
not been treated analytically but which were easily calculated numerically. It was shown for example that a vortex
centered at a distance from the flame (Fig. 23(b)) did not
produce a reacted core of the kind found in the centered
case. The mass fraction of products did not reach a constant
value and one of the reactants was entrained in this region.
First computations of the interaction with a propagating
vortex pair were also carried out (Fig. 24). Two reacted
cores are formed and entrained by the vortices. Behind the
vortices, the flame is pinched and exhibits a region of
reduced product mass fraction. This phenomenon is
observed in the analytical study of Karagozian and Manda
[67] and in experiments carried out by Southerland et al.
[133]. In all cases, the Reynolds number of the vortex was
limited to values lower than 100.
Similar two-dimensional simulations using infinitely fast
chemistry have also been carried out using the vorticity
transport equation and variable density in the low-Mach
number limit [134] to examine the evolution of the flame
front, using a Gaussian distribution of vorticity centered at
the origin. The influence of the initial distance between
flame and vortex on flame roll-up was specifically investigated. The amount of flame tip rotation mainly depends on
the magnitude of diffusion, increasing rapidly with the
Peclet number Pe G=D: Up to relatively high values of
Pe, the stoichiometric contour is only partially wrapped
around the vortex, forming a quasi-steady flame tongue.
Increasing heat release, thus causing more volumetric
expansion, mainly pushes apart the flames that are back to
back, but does not qualitatively change the flame pattern.
Indeed, thermal expansion effects appear to play a minor
role in this case. Generation of vorticity due to the misalignment of density and pressure gradients along the flame is
also observed, but again does not appear to play a significant
role in the dynamics of the interaction for the parameters
considered in this work. A region with low scalar dissipation
is observed around the flame tip, while most of the flame
surface is submitted to a relatively uniform scalar dissipation associated with the swirling flow. At the flame tip, the
compressive strain rate is tangent to the reaction sheet,
significantly diminishing normal gradients. This point is
therefore predicted to become the hottest part of the flame
and this may have an impact on the local chemical evolution
[135].
Simple two-dimensional incompressible simulations of
a non-premixed flame rolled-up in a vortex may also be
found for a single-step Arrhenius type reaction, but
neglecting the temperature dependency [69], corresponding to a consumption rate of the form v_ BY F YO : The

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

velocity field is again given by Eq. (14). A simulation


was carried out for Rev 100; Sc 10 and unity global
mixture ratio f . The structure of the flame winding up
and expanding with the growing vortex is investigated,
and it is observed that reaction is complete in the vortex
core. Displacement of the peak reaction-rate for varying
global mixture ratios is found as expected. It appears
that the structure of the flame tends towards a flame
sheet solution for larger values of the Damkohler
number.
Using a thermodiffusive approximation and a single-step
infinitely fast irreversible reaction [76], calculations of the
two-dimensional reactive shear layers have been used to
identify combustion regimes. It is observed that the flame
front is rolled-up by the vortices when the flame location is
close to the vortex centers. Higher values of the strain rate
are found in the vortex braids, increasing flame surface area,
which remains nevertheless bounded due to flame surface
destruction near the vortex cores, where successive reaction
sheets come close together and interact. Varying the
global mixture ratio f from 0.08 to 16 clearly shows
that the flame is displaced away from the reactant which
is in excess. Results of simulation were used to examine
the distribution of fuel consumption along the flame
sheet. In regions where the flame is isolated (i.e. at a
sufficient distance with respect to other flame elements),
the rate of consumption was found to be close to that of
a plane flame element strained at a value equal to the
local rate of strain


1 f Des 1=2
m_ F YF0
exph2
75
f
2p

253

reaction rate and convecting cores in which products are


accumulated and the reaction rate is reduced (Fig. 12).
The study of Delhaye et al. confirms that chemical reactions
in shear layers are controlled by the competition between a
mechanism of production of flame surface by the rate of
strain with local enhancement of the reaction rate by the
strain rates acting on the flame sheet and a mechanism of
annihilation of flame surface by mutual interaction of tightly
wound reactive sheets. These mechanisms also prevail in
more complicated turbulent flames as suggested initially
by Marble and Broadwell [136] and their modeling has
allowed detailed calculations of many types of turbulent
flames of the premixed, non-premixed and partially
premixed types.
The structure of a non-premixed flame during the head-on
interaction with a moving pair of counter-rotating vortices
has been analyzed using two-dimensional full numerical
simulations with detailed chemistry and multicomponent
transport models. These simulations rely on the compressible multicomponent NavierStokes equations. The vortex
pair is simulated in the computations using a stream function
corresponding to two incompressible non-viscous vortices,
i.e. for each vortex:
!
r2
78
c C exp
2r2c

77

Grid spacings below 20 mm are used in both directions,


with sixth-order spatial accuracy. At the beginning of the
interaction, the heat released by the chemical reactions
ahead of the vortex pair increases strongly. Later on, due
to the high strain-rate induced by the vortices in this simulation (diameter of 2 mm, rotational velocity umax
of
u
36 m s 1), extinction is observed as the vortex pair passes
through the flame (Fig. 25), but an intense reaction still
prevails on the vortex skirt and in the trail [109].
The integrated heat release over the whole computational
domain is increased by about 50% when vortices interact
with the flame compared to the case without vortices, showing the potential use of vortex structures to increase combustion intensity.

Expression (75) yields remarkably good estimates for the


reaction rate for isolated flame branches. In regions where
the flame sheets come close together and interact, the reaction rate differs markedly from that given by Eq. (75). When
the flame is wound-up by the vortex cores numerical calculations indicate that the consumption levels are reduced to a
great extent. Estimates obtained from Eq. (75) do not follow
this trend because the underlying strained flame model
assumes that the reactive sheet is isolated. The flame interactions, which are neglected in the model control the rate of
conversion when the flame surface is tightly wound.
Products accumulate in these regions and gradients of reactants are reduced to a great extent. When the flame interacts
with the large scale vortices the reactive structure features
isolated braids and correspondingly large values of the

4.2.3. Thermal expansion effects


Numerical simulations with a two-dimensional low-Mach
number formulation have been used to investigate the influence of heat release on the development of a temporally
growing reacting mixing layer [138]. A pseudo-spectral
method with AdamsBashforth time-stepping scheme is
used to solve the system of equations, and 64 64 grid
points are employed. The chemistry is modeled by a
single-step reaction with a production term depending
only on species concentrations and not on temperature.
Fluctuations are added to the initial solution in order to
feed natural instabilities. The Reynolds number of the simulation is 250, the Peclet number is 100 and a slow reaction is
considered. As a result, the reaction zone is quite thick.

where

h erf 1

1f
1f


76

Here e s designates the local strain rate which may be


calculated by considering the projection of the velocity
gradient on the plane tangent to the flame sheet

es nn : 7v 7v

254

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

Fig. 25. Example of computed results for a non-premixed diluted-hydrogen/air flame interacting with a pair of counter-rotating vortices. We
have here f 0:6; R 1:5 mm and rc 1 mm: The OH mass fraction is shown for six different interaction times, and two isolevels of vorticity
are also plotted in solid lines to identify the instantaneous position of the vortices. A local extinction of the flame front is observed for these
parameters (from Ref. [137]).

Comparisons between cases without and with moderate heat


release show that the main effect of heat release is to significantly reduce the rate of product generation in the layer, but
one should recall that this effect would be less effective for a
temperature-dependent reaction rate. The layer thickness
also decreases with increasing heat release, and the vorticity
at the center of the vortex structures is greatly reduced,
leading to a much more diffuse vorticity field. This work
was later on extended to three-dimensional simulations
[139], with similar conclusions. Two-dimensional coherent
structures are still identifiable, but turbulent kinetic energy
is again lower when heat release is present. The highest
reaction rate is found in the regions featuring the highest
strain rates, i.e. near the vortex braids.
The stabilizing effect of heat release for low-speed,
coflowing axisymmetric jets was also examined using a
single-step Arrhenius reaction and a low-Mach number
formulation [140]. Forced simulations, both cold and reacting are performed by adding small perturbations at the
inflow, for a jet Reynolds number of 2000. Large-scale
mixing of fuel and oxidizer by the vortices is visible in the
non-reacting simulations. Increasing heat release leads to a
strong decrease in the growth rate of the instabilities, and the
frequency corresponding to the most amplified mode is
shifted to lower frequencies. At this relatively low Reynolds
number, the roll-up of vortex rings progresses to a lesser
extend in the reacting case.
The effect of heat release on a three-dimensional
compressible, reactive mixing layer was also investigated
using a Flux-Corrected-Transport (FCT) formulation [141],

using a one-step, irreversible, Arrhenius chemical reaction


rate and semi-complex transport models. It was shown, in
accordance with previous studies, that energy release
reduces the shear layer growth and the amount of chemical
product formed.

4.2.4. Buoyancy effects in jet flames


Calculations of a buoyancy-dominated axisymmetric fuel
flow surrounded by coflowing air have been carried out with
the QUICKEST scheme [142,143], using infinitely fast
chemistry and unity Lewis number. Results for jet Reynolds
numbers somewhat higher than 100 and Peclet numbers
between 150 and 200 show a fair agreement between
computations and corresponding experimental results. The
influence of gravity was demonstrated by suppressing the
gravity term. As expected, a longer stable flame with no
moving vortex is obtained in this case. The impact of grid
resolution was assessed in a later work [118]. The species
concentration gradients are so sharp that an extremely fine
grid is needed, in particular to observe extinction, even if the
buoyancy-induced vortex structures are correctly described
with a coarse grid. Adding a small perturbation at the jet
inflow, smaller vortices also develop inside the flame
surface, causing small-scale wrinkles, but their overall
impact on the combustion process is negligible for lowspeed jets [144]. For the same case without buoyancy, the
flame has no outside vortical structures and becomes very
thick compared to the laboratory flame [145]. Buoyancy is
not only the main factor to generate the larger, outer

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

vortices, but also for maintaining the coherence of the inner


vortices over a long distance downstream [146].
4.2.5. Influence on mixing
Three-dimensional full simulations of incompressible
flows using a single-step reaction and neglecting heat
release have been carried out to investigate temporally
developing turbulent mixing layers [147], using pseudospectral methods. Maximum Reynolds number was about
50. A pseudo-turbulent velocity field was superimposed
over the initial solution, using parameters obtained from
experimental results. Vortex growth and roll-up was
observed, leading to an increased mixing of the reactants.
Using an infinitely fast reaction instead of a finite-rate
expression led only to a 15% increase in the product thickness. A good agreement was obtained with results from
similarity theory.
Numerical simulations of the Euler equations were
carried out to simulate a jet diffusion flame [148], using
a fourth-order Barely-Implicit-Correction-FCT method.
Diffusion was modeled with Ficks law, and reaction with
a single, global reaction using the Parametric Diffusion
Reaction model (an extension of the flame sheet model
allowing coexistence of fuel and oxidizer at the same
point, as reaction is considered to occur over a finite volume
and time interval). For low jet speeds, growth and shedding
of vortical structures was observed outside the flame
surface, affecting in particular the transport of oxygen
towards the flame. For higher jet velocities, vortices form
also inside the flame surface. These inner vortices enhance
the transfer of fuel towards the flame surface, while the
larger outer structures mainly have an influence on flame
surface area, flame location and heat transport. These structures increase the amount of reaction due to a more efficient
convection of unburnt reactants, but can also quench the
flame by supplying an excess of cold gases.
Results of full numerical simulations have been analyzed
to investigate the role of coherent structures in incompressible, temporally growing, reacting free shear flows
[149], using a single-step isothermal chemical reaction.
The initial shear layer solution is perturbed with its linearly
most unstable two-dimensional mode plus low-level threedimensional random noise to trigger natural instabilities. It
is found that vortex roll-up and pairing entrains the
unreacted species on both sides into the chemical reaction
zone. The concentration gradients remain quite high in the
braids but tend to vanish in the vortex cores. Vortex roll-up
and pairing results in the stretching of the reaction zone,
increasing product formation, this effect being much higher
for lower species diffusivity. Due to this stretching effect,
most of the reaction takes place in the vortex braids.
A monotonically integrated large-eddy simulation
(MILES) approach has been combined with a FluxCorrected Transport (FCT) solver to investigate the interaction of vortices in an excited axisymmetric reacting jet with
coherent structures [150]. An azimuthal perturbation is

255

introduced at the jet inflow. The interaction between


the vortex rings and the imposed perturbation results in
a deformation of the flame which triggers process of selfinduction and vortex interactions. The mixing pattern is
deeply modified leading to fuel-rich and fuel-lean
regions, affecting the combustion pattern and eventually
the energy release.
Recent full numerical simulations of an axisymmetric
reacting vortex ring may also be found [151]. A one-step
Arrhenius reaction is used to mimic the combustion of a
typical hydrocarbon in air, Ficks law is employed for diffusion processes and the Lewis number is set equal to one. The
vortex ring is generated by impulsing a jet of cold fuel in a
quiescent ambient at a much higher temperature, leading to
self-ignition after some delay. Third-order accuracy is
achieved for the derivative computations, and a fixed refined
grid of 513 513 points is used. By adjusting the temperature ratio Tambient/Tair, ignition occurs either during the formation of the ring or later. For low values of this temperature
ratio, a weak reaction initially surrounds the vortex ring.
Later, a strong reaction zone appears in the rear of the vortex
wake, and propagates in two directions to ignite the mixture
at the interface and also reaches the core region, accelerating
through it until it reaches the front edge of the vortex bubble.
Both reaction fronts merge later on to completely consume
the oxidizer in the core. For higher values of Tambient/Tair, rollup of oxidizer in the core region becomes inefficient because
the reaction starts immediately, and the reaction is confined
at the outer limit of the vortex ring, with very little reaction
inside the core. This shows the positive influence of partial
premixing and of a delayed ignition in this case. The influence of chemical reaction on the vorticity is also investigated. It is found that chemical reaction leads to a faster
decay of the total circulation of the flow. Baroclinic production of vorticity changes strongly for the reacting case due to
flame structure modification and resulting alteration of the
pressure and density fields by the flame. This may lead to an
increase or a decrease in the resulting vorticity, depending
on the exact geometry of the reaction front. Other effects are
identified that play a significant role in the evolution of
vorticity: different stretching levels of the vortex ring in
the azimuthal direction are observed, and this stretching of
vortex filaments increases the local magnitude of the
vorticity due to the conservation of angular momentum;
thermal expansion effects have also a significant impact
on the vorticity field, with a reduction of the total circulation
by a factor greater than 2; finally, changes in the viscosity
due to the temperature increase induced by reaction act as a
strong vorticity sink. In all cases, reaction affects the core
structure, leading to large distortions, slower vortex propagation speeds and a larger radial expansion. The influence of
the chemical reaction on the evolution of the strain-rate was
also assessed. It is observed that the magnitude of the strainrate is reduced with chemical reaction due to the high dilatation level resulting from heat release and due to a substantial increase in viscosity.

256

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

Fig. 26. Reduced heat-release plotted for the case with moderate
initial temperatures (air temperature 800 K), and vortex Reynolds
number Rev 160: Results are presented for times 0.4, 0.5, 0.8 and
1 ms. A diffusion ignition mode is observed in this case (from Ref.
[156]).

4.2.6. Ignition
Using a Lagrangian formulation with vortex elements to
discretize the vorticity field and scalar elements to represent
species or temperature, a two-dimensional incompressible
reacting mixing layer was investigated with two different
reaction models [152]. Heat release is considered to be
negligible, and a single step irreversible reaction is
employed to represent chemical processes. Reactants are
entrained into the vortex cores during roll-up and chemical
reaction proceeds. In a first computation, the reaction rate
v_ BYF YO is independent, while in the second computation, the reaction rate is defined by an Arrhenius reaction
rate:


v_ BYF YO exp

Ta
T


79

In both cases, the Reynolds number is equal to 10 000, the


Peclet to 4000, with a velocity ratio of 0.33 and a unity
Lewis number. In the case of a temperature-independent
reaction rate, products are formed immediately downstream
of the splitter plate. With the Arrhenius reaction rate, an
ignition delay is observed, which can be reduced by increasing the initial temperature. The high level of scalar dissipation in the vortex braids leads to a reduction of the chemical
activity, while product formation is maximum in the vortex
cores. Similar computations using a spectral-element
method and a temperature-dependent Arrhenius reaction
rate may also be found [153]. The reaction rate approaches
zero at the braids of the vortices due to the high scalar
dissipation level. Decreasing the Damkohler number, the
effects of finite-rate kinetics become more pronounced.
These works were later on extended [154] to propose
another version of the flamelet model, describing flame
elements in a Lagrangian frame as convected flamelets with
variable strain-rates. Interactions between flames in the
vicinity of the vortex cores are taken into account. It is found
that combustion is proceeding intensely within the braids and
around the outer periphery of the eddies, where the strain-rate
is high, due to the delay between straining and the flame
response. A high temperature is obtained in the vortex cores,
corresponding to the interaction of several flamelets.
The work of Laverdant, Candel and Ashurst was the starting point of later investigations [155,156], using the thermodiffusive approximation (constant density) and a single-step
irreversible finite-rate reaction to study the time-dependent
roll-up of a diffusion flame in the velocity field induced by a
vortex. The velocity field of the viscous vortex growing with
time is again given by Eq. (14). In the first paper [155], a
detailed theoretical analysis with asymptotic expansion is
followed by numerical solutions to investigate the location
of the first ignition spot. It is observed that the vortex
Reynolds number Rev has little influence on the ignition
time, that ignition occurs deep in the region of the initially
hotter reactant for small Rev but tends to travel towards the
viscous core center for higher values of Rev. Numerical
solution of the full system of equations for Rev 28 leads
to the observation that a hot spot initially appears within the
viscous core and rapidly develops into an isolated, almost
circular flame which grows as time increases. Later on, two
diffusion flames appear in the outer region and move
towards the viscous core, eventually merging with the first
flame. The influence of a superimposed shear flow on this
flame/vortex interaction was also investigated using a lowMach number formulation [157]. It is found that, in that
case, velocity oscillations appear, resulting in mixing
enhancement, viscous heating and multiple ignition points.
The identification of three different ignition modes was
the subject of the numerical study carried out by Thevenin
and Candel [156]. In particular, the influence of the vortex
Reynolds number Rev and of the oxidizer temperature TO
was investigated. It is found that a purely non-premixed
ignition takes place at low initial temperatures (Fig. 26),

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

Fig. 27. Computed flame and flow field during the interaction.
Temperature fields are shown and instantaneous locations of particles, which were injected from the air and fuel nozzle, are also
displayed. An annular extinction is observed for these parameters
(from Ref. [124]).

due to the fact that the vortex core does not reach a sufficient
temperature for ignition, limiting the reaction processes to
the vortex braids. In contrast, a premixed ignition mode of
the vortex core is found for high initial temperatures,
while a mixed regime with simultaneous premixed and
non-premixed flames is observed for intermediate temperatures. The influence of the vortex Reynolds number was also
investigated, revealing that the premixed ignition mode is
favored for high values of Rev, due to the high scalar dissipation level in the braids, delaying the ignition of the diffusion flames. In general, consumption of the core takes place
in a mixed mode, the diffusion flame consuming the outer
layers of the core, while a premixed flame consumes at a
later time but at a faster pace the inner, well-mixed part of
the core.
4.2.7. Extinction
The case of a pair of coaxial vortex rings moving into or
away from a stable jet flame, its axis coinciding with that of
the rings was treated numerically [158] with the method of
Katta and Roquemore [144]. If the Peclet number defined as
Pe dd vr =D (with vr the radial ejection velocity and d d the
diffusive layer thickness) is small enough Pe p 100; the
vortex system just displaces the flame surface, while for
Pe 100; the vortex ring cuts through the high-temperature
layer with minimal flame movement, creating a thin diffusive-thermal layer, analogous to a strained counterflow
flame.
Detailed axisymmetrical simulations of the extinction
process for a counterflow hydrogen/air flame interacting
with a vortex ring have been carried out with detailed
chemistry and transport models [124]. Two different
extinction processes are identified, one taking place on the
centerline and one occurring at a distance from the centerline (Fig. 27). The latter occurs when the vortex ring is
forced at a moderate speed and the flame deforms

257

significantly before extinction. Analysis reveals that extinction does not result from a high strain-rate value in this case,
but is a consequence of preferential diffusion and flame
curvature effects.
The extinction of a non-premixed flame during the headon interaction with a moving pair of counter-rotating
vortices was analyzed using full numerical simulations
with full chemistry and multicomponent transport models
[123]. The strain-rate induced by this fast vortex pair on the
flame is found to be quite high, exceeding 2000 s 1 during a
long time and reaching peak values above 20 000 s 1. This
large strain-rate is responsible for extinction, as the amount
of mixing between fuel and oxidizer in the extinction region
remains fairly constant during the interaction. No depletion
of either fuel or oxidizer is observed and, in fact, the mixing
is improved by the vortex pair, as the reduction in the reaction rates leads to leakage of reactants through the weaker
flame elements.
The influence of the direction of propagation of the vortex
ring interacting with a H2/air counterflow diffusion flame
has also been investigated with full numerical simulations
and detailed chemistry [159]. These computations show
opposite effects produced by flame curvature of opposite
orientations by placing a vortex on either the air or the
fuel side. Extinction is only observed when the flame curvature is convex towards the air stream.
4.2.8. Stretching effects
The simple-chemistry simulations of jet diffusion flames
[144] were later extended to include detailed chemical
kinetics and non-unity Lewis number effects [160]. It is
found that the size and shape of the outer vortical structures
is basically unaffected by the fast-chemistry and unityLewis assumptions. The convective motions of the outer
vortices induce high stretch levels along the flame surface
by squeezing the flame at some locations and bulging it at
others. Due to Lewis number effects, the temperature of the
stretched flamelets is increased, leading to noticeable fluctuations in the flame temperature. In addition, preferential
diffusion effects tend to complicate the picture, by locally
changing the effective equivalence ratio. Computations of
H2/air jet flames with eleven species (including NO) and 40
reactions were carried out to assess preferential diffusion
and stretching effects [161]. Vortices were introduced on
the fuel side by perturbing the inflow condition, whereas
the air-side vortices were naturally present in the simulation,
both vortex structures resulting in local flame stretch. For
fuel-side vortices, when the flame is stretched, a local
increase in temperature is observed, and the temperature
diminishes when compression is taking place. Combined
curvature and preferential diffusion effects for Le 1 determine the flame response. In contrast, for the air-side
vortices, stretching diminishes the temperature while
compression increases the temperature. This is a consequence of the fact that, for non-unity Lewis numbers, preferential diffusion effects become important, and the response

258

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

Fig. 28. Example of experimental results for the interaction between a diluted-hydrogen/air flame and a toroidal vortex ring, for f 0:5 and
uc 1:31 m=s: The instantaneous OH concentration obtained by Planar Laser-Induced Fluorescence is shown here for six different interaction
times (from Ref. [110]).

of the flame during interaction with a vortex depends on the


exact location of the vortex structure and of the flame
surface. The production of NO was also lowered during
interaction with an air-side vortex, due to temperature
reduction.
4.2.9. Consideration of detailed chemistry and transport
models
Earlier simple chemistry simulations of radially moving
vortex rings interacting with a stable jet flame [158] were
later improved to include detailed chemical kinetics and
non-unity Lewis number [162]. For high Peclet numbers,
the unsteady extinction mechanism was investigated in
some detail. It is shown that, in a first step, molar production
and heat release rates increase significantly while the width
of the thermal layer decreases. Subsequently, leakage of fuel
and oxidizer through the flame zone becomes important and,
after a while, combustion can no longer be sustained as the
radical pool is depleted and the flame is extinguished.
The impact of thermal diffusion in hydrogen jet diffusion
flames was assessed later on [119]. The addition of a thermal
diffusion term improves the agreement between experiment
and calculations. Representations of thermal diffusion are
achieved by a rather crude scheme in which the usual hydrogen diffusion coefficient is multiplied by a factor of 1.7. A
detailed investigation of the chemical processes associated
with local extinction may also be found [163], and the influence of three different reaction schemes was examined [164]
in the case of a buoyancy-dominated methane jet diffusion
flame. It is found that a semi-detailed mechanism with 17
species and 52 reactions is sufficient to correctly reproduce
the features of the jet diffusion flame and the associated
vortices for a wide range of jet velocities. The peak temperatures are overestimated by about 50 K. On the contrary, the
CPU time is reduced by a factor of 5 with the simplified
chemistry. It is also observed that inner vortices dissipate
downstream due to entrainment of viscous combustion
products from the flame surface. The interaction of these
vortices with the flame front leads to a high degree of wrinkling and to important variations of flame temperature and

species distributions as previously observed for hydrogen jet


flames.
Detailed studies on the radical species may be used to
delineate extinction regimes as shown in calculations of a
non-premixed flame interacting with a moving pair of counter-rotating vortices using full numerical simulations with
detailed chemistry and multicomponent transport models
[110]. The vortex pair is simulated in the computations
using a stream function corresponding to an incompressible
pair of non-viscous vortices, as given by Eq. (78). For this
hydrogen/air reaction, represented with 9 species and 37
reactions, it is observed that chemical radicals fall more or
less in two families. In particular, OH, O and H tend to
behave like the temperature, while HO2 and H2O2 follow
more closely the behavior of the local heat release rate. If
one considers the heat release rate integrated across the
flame front, used as a measure of the burning intensity of
the flame, H and O appear to be the best tracers for this
quantity, but they are difficult to measure experimentally.
On the contrary, the OH radical behaves like this quantity
and is very easy to measure. It may therefore be used to
detect the boundary between burning and extinction
regimes.
4.3. Non-premixed experiments
4.3.1. Rollup
Because the location of a diffusion flame depends on
transport processes in the diffusive layers, it is very sensitive
to rollup phenomena. A vortex is able to wind a nonpremixed flame [110]. The flame may be strongly lengthened and rolled-up if it is weak enough and the residence
time of the vortex in the flame is long enough. This rollup
leads to an increase of the global heat release rate which
may be counterbalanced by local extinction due to excessive
strain and curvature induced by the vortex (Fig. 28).
Vortex dynamics have an influence on combustion but the
latter has also a significant influence on fluid dynamics. Park
and Shin studied the developing process of a diffusion flame
in a vortex [121]. They found that the presence of the flame

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

in the vortex delayed the vortex sheet rollup and that the
spreading of this vortical structure after rollup is faster due
to exothermic expansion. A visual investigation of the
development of a non-premixed flame inside a burning
ring has been carried out recently [165].
4.3.2. Mixing, diffusion effects
The importance of mixing phenomena in non-premixed
flames is illustrated in a jet flame configuration. The peak
OH mole fraction may decrease significantly with axial
distance in the upstream region of the unperturbed flame
[120]. A simplified model to account for reactant dilution
by products shows a decrease in OH levels comparable with
experimental observations. On this basis together with some
other considerations, Mueller and Schefer conclude that
products formed at upstream locations are convected downstream, parallel to the flame surface in the coflowing reactant stream configuration. This promotes dilution of the
reactants by product species that reduce the scalar dissipation rate on the stoichiometric surface. Thus, OH concentrations are particularly sensitive to product dilution through
the scalar dissipation rate.
Another configuration that enhances mixing is the configuration of Underwood and Waitz who placed a lobed
mixer in the flow [166]. They observed that this device
enhanced the initial mixing rate by a factor between 6 and
12 and that this mixing rate in the vicinity of the mixer was
less sensitive to heat release than in the flat plane case. In the
far field where the streamwise vortices had decayed, the
decrease in mixing rate with heat release was approximately
proportional to the decrease in mixing rate observed for the
flat plate.
4.3.3. Effects of stretch
Strain rate has an important influence on reaction rate.
This feature is studied by Karagozian et al. [112]. Conclusions reached in this study must be considered bearing in
mind that some of the fundamental diffusion processes in
gases Sc 1 cannot be well represented by a liquid reaction Sc 100. In particular, all the mechanisms depending on heat release are not found in the liquid reaction
experiment. The actual variation in the diffusion layer thickness predicted for a gaseous reaction, resulting from variations in the local strain rate of the diffusion flame, is not
observed in the liquid experiments, however, because of the
lower effective diffusivities of reactants in liquid phase.
Nevertheless, the effects of positive strain on the augmentation of the fast reaction process are at least well described: a
clear intensification of reaction was observed in high strain
rate regions induced by the vortex. It also appeared that
reacted cores essentially composed of reaction products
were formed in the vortex core.
In the experiment of Karagozian et al. no flame thinning
due to positive stretch was observed. In a more realistic
configuration, Mueller and Schefer reported observations
of significant thinning and thickening of the OH layer due

259

to periodic hydrodynamic perturbations [120]. Positive


strain rates resulted in thinning of the OH layer by up to
50%, while negative strain provokes a 100% thickening of
the OH layer. Of course, this observation must be interpreted bearing in mind that OH is also present in hot
gases because of H2O dissociation. Thus, OH present in
burnt gases may stagnate close to the reaction zone because
of very low convective fluxes. So, the definitive conclusion
that such large increases in OH layer thickness are not
predicted by the laminar flamelet model, must be first
checked by measuring the actual flame thickness with a
different diagnostic method.
Using a WolfhardParker slot burner, the effect of
acoustically forced two-dimensional vortices on a
methaneair diffusion flame was assessed. Significant
variations in the OH layer thickness are observed. In
particular, negative strain produces a doubling of the
flame thickness.
In a simpler configuration, Renard et al. described the
effects of strain rate and curvature [110]. The main conclusion is that the strain rate, curvature, and preferential diffusion operate in combination. Strain rate lengthens the flame
front, increases the heat release rate and possibly leads to
local extinction. Curvature and preferential diffusion diminish chemical reaction intensity especially at the base of the
vortex-induced wrinkle.
4.3.4. Ignition, extinction
The same configuration was used to investigate the
processes involved in a local extinction [110]. In some
cases, local extinction is due to an excessive local strain
rate. Indeed, no complete extinction was observed in highly
curved regions of the flame front. It always occurs at the top
of the vortex ring. This was confirmed by a theoretical
analysis made by Thevenin et al. assuming that the leading
stagnation point of a toroidal vortex may be considered as a
counterflow [123]. In Linans analysis of diffusion flames
giving an order of magnitude of quenching strain rate for a
given flame, it was assessed that the stretch (which is essentially dominated by strain rate at the wrinkle top) experienced by the flame was larger than this quenching limit for
extinction cases and smaller in the other case [167]. It could
be argued that a flame/vortex interaction is an unsteady
phenomenon whereas the previous theoretical analysis
assumes that the flame is steady. Using an asymptotic analysis derived by Cuenot and Poinsot, a characteristic response
time for the flame may be computed and was found to be
much smaller than the interaction time [168]. This also
underlined the importance of the vortex speed in the extinction mechanisms [110]. As the vortex strain rate is proportional to the maximum azimuthal velocity (for a given
vortex ring diameter) and this velocity is of the same
order of magnitude as the vortex speed, the residence time
of the vortex and the strain rate are strongly linked. If the
vortex circulation is high, the flame is quenched more readily than when the circulation is low. In the latter case,

260

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

extinction is replaced by extension of the flame front by


straining and roll-up. These observations were confirmed
in other configurations [118,169].
A visual investigation of the development of a nonpremixed flame inside a burning ring has also been carried
out [165]. The ignition and extinction processes of the flame
were investigated in more detail.
Other experiments have been recently carried out to
investigate the extinction process for a methanol spray
flame interacting with a toroidal vortex [170], showing
that strain-rates at extinction are consistently larger for a
gaseous flame compared to the spray counterpart.
4.3.5. Lewis number effects
Very few experiments on non-premixed flame/vortex
interactions specifically deal with Lewis number effects.
In the experiment of Hancock et al. [119], thermal diffusion
apparently increases the H2 mass flux to the flame zone,
leading to a broadening of the temperature profile and an
increase of the peak temperature in the flame front. Flame/
vortex interactions tend to increase the local temperature
above the steady-state undriven flame temperature. The
increase is observed at the bulge where the flame is positively
stretched and curvature is concave toward the fuel side.
In the counterflow configuration [108], a specific quenching pattern may be observed. When the vortex is forced
toward the flame at a moderately high speed, quenching
develops in an annular manner away from the stagnation
line. Numerical simulations confirm this point [124]. During
extinction, the strain rate in the annular region is equal or
even lower than that of a stagnation point flame being extinguished. Curvature is low there and strain rate is lower than
the quenching level. In contrast, the flame curvature is not
negligible in the annular region where the flame is starting to
quench. Therefore, this extinction process may stem from
the combined effect of preferential diffusion and flame
curvature.
4.3.6. Vorticity generation, baroclinic effects and thermal
expansion effects
It is usually considered that vorticity is generated by baroclinic effects and decays by thermal expansion. Contrary to
the premixed case, no velocity measurement was carried out
on diffusion flame/vortex interactions to investigate vorticity generation. Chen et al. made a theoretical analysis
of experimental results on jet flames [115]. It appeared
that three regimes of vorticity generation may be distinguished. In the fuel lean regime or outside the luminous
flame, vorticity production due to baroclinic torque is
balanced by expansion induced dissipation. Inside the luminous flame, both the volumetric expansion and baroclinicity
behave like sink terms in the vorticity balance, except in a
sub-regime close to the flame sheet where the volumetric
expansion acts as a source term. In the post flame region,
vorticity generated by the baroclinic torque may be
increased or decreased by volumetric expansion that may

be a source term in the core region or a sink term in the outer


region. The vorticity production in the post flame regime is
qualitatively similar to that found in a buoyant jet, and
conversely, opposite effects are expected for a negatively
buoyant jet.
4.3.7. Gravity effects
Few experiments were carried out under microgravity
conditions. Chen and Dahm studied a diffusion flame
embedded in a vortex ring and they confirmed the main
features of theoretical predictions of Karagozian and
Manda [67] and Manda and Karagozian [171]. They did
not focus on the differences between the microgravity and
the one-g cases.
The most common process associated with buoyancy
effects is flickering. In their experiments on a jet flame,
Davis et al. [142] showed that both flame flickering and
double-peaked temperature profiles (which is a peculiarity
of their experimental configuration) are closely associated
with the buoyancy-induced vortices outside the flame
surface in the column of hot gases surrounding the flame.
Finally, important three-dimensional effects were observed
to eventually appear as the flame bulges and vortex structures move upward.
The significant role of buoyancy effects on the development of a non-premixed flame inside a burning ring has been
observed in [165].
4.3.8. KelvinHelmholtz instability
It is well known that jet flows feature azimuthal instabilities. These modes interact so that large-scale structures are
coherent only in an average sense [117]. Of course, the
influence of these azimuthal instabilities becomes more
significant as the Reynolds number of the flow is increased.
The three-dimensionality is initiated in the braid region
where it affects the local combustion, inducing local flame
extinction and the formation of random discrete flamelets.
The amplification of higher modes of instability through the
generation of three-dimensional substructures leads to the
eventual decay of large-scale structures to small scale ones.
This results in enhanced mixing of reactants and an extension of reaction layers leading to a global increase of the
reaction rate.
5. Results for the premixed flame/vortex interaction
5.1. Theoretical results for the premixed case
The theory of premixed flame roll-up by vortices is less
developed. The configurations which have been considered
are sketched in Fig. 29. With the exception of the study due
to Peters and Williams [172], most investigations of
premixed flame/vortex interactions are numerical or experimental and they will be reviewed later in this article. Peters
and Williams show that for the initial geometry of Fig.

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

261

Fig. 29. Geometrical configurations investigated in premixed flame vortex interaction studies. (a) Vortex centered on a premixed flame. (b)
Vortex centered off a premixed flame. (c) Vortex pair interacting with a premixed flame.

29(a), the roll-up gives rise to a reacted core as in the nonpremixed case. The core radius is obtained from simple
considerations or from a more elaborate asymptotic reasoning. In both cases, this radius is estimated by neglecting the
effect of heat release (the thermodiffusive approximation).
The core radius grow according to
2=3
r p2=3 G 1=3 S1=3
l t

80

In this expression Sl designates the laminar flame speed. By


normalizing the radius and time with the flame thickness
df a=Sl and flame time tf a=S2l ; where a is the thermal
diffusivity, one obtains
r =df 2=p1=3 Pe1=3 t=tf 2=3

81

where the Peclet number is Pe RePr and Rev G=2pn


and Pr n=a are the Reynolds and Prandtl numbers. The
scaling law (Eq. (80)) differs from that established for the
diffusion flames. For the same value of the vortex circulation G the ratio of the core radii may be estimated by
!1=6
r diff
D

tf =t1=6
82
r premix
Sl2 t
The premixed core grows faster than the diffusion flame
core but their ratio (Eq. (82)) remains close to unity for
typical interaction times. For tf t 100tf ; this ratio
remains between 1 and 0.46.

the formation of a vortex core after the initial wrinkling of


the flame, revealing first details of the flame structure (Fig.
30). The velocity field was given by Eq. (14). Flame selfpropagation was not important in that case because of the
low flame-speed. Due to the limited value of the vortex
Reynolds number Rev 100; no extinction was observed,
but a delayed burning of the fresh gases entrained inside the
vortex core was found.
Direct simulations using a unity Lewis number hypothesis, a low-Mach number formulation and a single-step reaction have later on been employed to investigate head-on
flame/vortex interactions [173,174]. Simplified cases
where either the flame or the vortex is frozen were first
considered to simplify the analysis. The velocity specified
initially was purely tangential and given by:
"
#
G0r
r2
vu r 2 exp 2
83
2rc
rc
When the flame moves over a frozen vortex, the formation
of pockets of unburnt gases is only observed for low values
of the Damkohler number, Da 0:5; with Da rc2 St =Ddf
while, for Da 5; the flame remains almost unperturbed. In
the intermediate range 0:5 Da 5; flame surface is
changed by convection but remains a simply connected
sheet.

5.2. Numerical results for premixed flame/vortex


interactions
5.2.1. Introduction
As indicated previously, analytical approaches are
restricted to simple configurations and generally imply
many restrictive hypotheses. Numerical simulations have
thus been used more often for the investigation of flame/
vortex interactions. We will now summarize the main
results obtained with these numerical investigations.
5.2.2. Flame structure and pocket formation
Initial studies were based on the thermodiffusive approximation (constant density), constant and equal thermodynamical and transport properties for all species, and a singlestep global reaction with an Arrhenius reaction rate
[130,132]. The equivalence ratio was considered to be
low, leading to the hypothesis that the oxidizer
concentration was constant. These computations of a
flame rolled-up by a growing viscous vortex concerned

Fig. 30. Premixed flame rolled up in a vortex (visualization of


temperature field) (from Ref. [132].

262

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

Fig. 31. Time history of flame perimeter (indicator of reactant consumption rate) which shows the five physical processes occurring typically
during a premixed flame/vortex interaction (from Ref. [177]).

Full numerical simulations have also been used to investigate quenching processes during the head-on interaction
between a premixed flame and a counter-rotating vortex pair
[175,176], employing a single-step irreversible reaction,
constant non-unity Lewis number and ad hoc heat losses.
Four typical configurations were observed: (1) local quenching of the front; (2) formation of a pocket of fresh gases in
the burnt stream without quenching; (3) a wrinkled flame
front; or (4) a negligible global effect. More details are given
in a later section.
Using the Simple Line Interface Calculation (SLIC),
representing the flame as an infinitely thin interface moving
normal to itself, the interaction between a premixed flame
and a counter-rotating vortex pair was investigated [177].
The main simplifying hypotheses were a frozen velocity
field, a constant flame propagation speed and no thermal
expansion. These computations were used to describe the
main features of a premixed flame/vortex interaction:
attachment of the flame to the vortex, initial roll-up of the
flame, burn-through from the rear of the vortex, formation of
pockets just behind the vortices and observation of a remaining cusp in the initial flame front (Fig. 31). The attachment
of the flame to the vortex pair is only observed when the
counter-rotating vortices have a rotational velocity greater
than Sl uc ; where uc is the convection velocity of the
vortex pair. The velocity field corresponding to the vortex
is again based on an Oseen vortex pair, keeping the typical
radius of the vortex ring R constant. A stronger vortex is

observed to cause a greater degree of flame roll-up, while a


larger convection velocity provides less time for the interaction and yields therefore less roll-up, the latter effect being
in general more important.
Limits for the formation of pockets of unburnt gas were
also investigated in this work. Pockets form only for a given
range of convection velocities. For very low vortex speeds,
the flame has more time to burn inwards, resulting in smaller
(or no) pockets, while for large convection velocities, a very
small portion of the flame front is pulled along with the
moving vortex and there are no pockets. The case of vortex
pairs exerting compressive strain was also studied, showing
that these vortices lead to a reduced roll-up and a weaker
interaction, due mainly to a shorter residence time.
Computations of the motion of a premixed flame through
and between swirling eddies was exploited to investigate the
evolution of the apparent flame speed [178], using the Gequation formulation and neglecting heat release. The
collection of eddies, used to simulate a turbulent velocity
field, generates a root mean-square velocity u 0 ; with quiet
zones between the eddies. Formation of pockets of unburnt
gases appears only for values of u 0 =Sl exceeding 2.5.
The discovery of a new class of NavierStokes solutions
representing steady periodic stretched vortices was
employed to study flame propagation through periodic
vortices with no diffusive decay [179], using the G-equation
with constant flame speed and constant density. Flame
movement is enhanced where the flow around the vortex

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

cores is in the same direction as the flame propagation, while


cusps tend to form where different directions of propagation
cross each other. Noting K and Sl the non-dimensional
stretch rate of the vortices and laminar flame speed, it is
observed that these stretched periodic vortices lead to a
linear increase in effective flame speed for sufficiently
small values of K =Sl ; while the flame becomes trapped
between the vortices for large values of K =Sl ; being unable
to escape the suction effects. This same configuration was
later on combined with a ray-tracing technique [180] in
order to identify the fastest flame path through the periodic
array of vortices. The maximum enhancement of flame
speed is found to grow with the strength of the vortices.
The flame is relatively unaffected by vortices with typical
velocities much below Sl. For large vortex strength, the
flame is trapped in the vortex cores.
The G-equation was also employed in two dimensions to
study the detachment of pockets of unburnt fuel during the
interaction of a premixed flame with a single viscous-core
vortex [181], using a constant laminar flame speed and
neglecting heat release. The detachment scale lD is found
to be nearly proportional to the Gibson scale S3l =e (see
Section 6 for a definition of this scale), pockets appearing
for vortices larger than lD, while smaller vortices only
wrinkle the flame.

5.2.3. Vorticity generation and dissipation


The creation of new vorticity caused by misalignment
between the density and pressure gradients during the rollup of a flame by a single vortical region was investigated
using a combined low-Mach number/discrete vortex twodimensional formulation and global reaction with an Arrhenius reaction rate [182]. The density variation across the
flame is restricted to a ratio of 5. The vorticity transport
equation describes thermal expansion effects and baroclinic
generation of vorticity. Calculations indicate that predominant vorticity production takes place with a rotation sign
opposite to the initial value, so that the baroclinic effect
creates a vortex dipole from the initial monopole. A
significant production of vorticity only occurs when the
vortex rotation time is fast compared to the residence time
of the flame inside the vortical domain, i.e. rc =Sl q
max
rc =umax
q Sl :
u ; or uu
In full simulations using a unity Lewis number hypothesis, and a single-step reaction [173,174], vorticity generation by the baroclinic torque appears to reach high levels.
When both the flame and the vortex are free to evolve
simultaneously, the vortex is compressed when crossing
the flame. Production of vorticity depends strongly on the
intensity of heat release and on flame curvature.
Flame/vortex interactions in a premixed jet flame were
investigated using axisymmetric simulations taking into
account detailed chemistry (9 species for H2 air combustion) [183]. The high viscosity of combustion products leads
to an intense dissipation of vorticity, and the presence of

263

vortices does not enhance the burning rate, compared to the


case without vortex structures.
An inverse partially premixed flame established by injecting a CH4 air annular jet between a central air jet and
coflowing air on the outside was also investigated [93].
For large values of the Froude number (Fr 15; comparing
the square of the jet exit velocity to the product of the jet
diameter multiplied by the gravity g), the effect of buoyancy
are completely negligible. As the Froude number is
decreased, vortices roll-up periodically due to the Kelvin
Helmholtz instability. Still, for Fr 3; the vortices only
interact with the hot plume downstream of the flame. For
Fr 0:5; a weak flame/vortex interaction is observed, leading to flame pinch-off and flicker. When Fr is further
reduced, the flame flickering frequency is increased, and
real flame/vortex interactions begin to appear.
Full simulations have also been used to investigate the
evolution of vorticity during the head-on collision between a
vortex pair and a premixed flame, superimposing different
mean pressure gradients and a moderately low turbulence
intensity in the background, such that the flame was not
ruptured [184]. Three different vorticity regimes are identified, one allowing full passage of vorticity across the flame
front, one preventing it completely and one intermediate
regime. In all cases, it is found that the incoming vortex
structure is essentially destroyed, and that new vortical
motions are created inside the flame by the baroclinic
torque. Exact results depend strongly on the flame topology
and on the pressure gradient.
5.2.4. Ignition
The ignition of a turbulent premixed jet was investigated
with a random vortex method amended by a simple flame
propagation algorithm [152]. The flame was treated as a thin
interface, which is displaced by advection due to the velocity field, self-advancement at a normal (constant) burning
speed and volumetric expansion, modeled by a line of
sources. Temperature ratio across the flame front was set
equal to 4, and the jet Reynolds number was 10 000. Two
cases are considered. In the first one, flame initiation is
obtained by ignition at the center of the jet orifice. Islands
of unreacted fluid are observed, due to a high degree of
entrainment of reactants into the products at the center of
the jet. In the second case, combustion is initiated uniformly
in the jet. The flame front then establishes at the outer
boundary of the jet and its structure is dominated by
large-scale eddies. Comparisons between calculations with
and without flames show strong effects of volumetric expansion due to the heat release, manifested by the swelling and
spreading of the vortex patterns, reducing the coherence
between them.
5.2.5. Extinction
Full numerical simulations were used to investigate
quenching processes during the head-on interaction between
a premixed flame and a counter-rotating vortex pair

264

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

Fig. 32. Reaction rate fields at four instants for an interaction leading to pocket formation (Le 1:2; 2R=lf 5 and umax
u =Sl 12) (from Ref.
[175]).

[175,176], employing a single-step irreversible reaction,


constant non-unity Lewis number and a simplified model
for heat losses. These computations employ a pair of incompressible non-viscous vortices corresponding to the stream
function (for one vortex) specified by Eq. (78). The flame
fronts are found to be much more resistant to quenching than
expected from classical theories. The global stretch rate is
an essential parameter, but curvature, viscous dissipation
and transient dynamics also affect the process (Fig. 32).
Small-scale vortices are dissipated by viscosity too rapidly
to significantly influence the flame. Systematic calculations
carried out for vortices of different sizes and a range of
strengths were employed to develop new diagrams for
turbulent premixed combustion, as described in a later
section.
The influence of radiative losses during interaction of a
vortex ring with a lean methaneair flame has been recently
investigated using full numerical simulations with a skeletal
mechanism (16 species) to describe the chemistry [185].
Different levels of radiative heat losses have been tested.

These computations show that both a sufficiently high


strain-rate and heat losses are simultaneously required for
local quenching. As expected, increasing strain-rate and/or
heat losses promote quenching, but extinction never occurs
instantaneously. Even for the strongest vortices, there is a
minimum time required for quenching the flame.
5.2.6. Stretching
Kinematic displacement of the flame front was used in
Ref. [186] to calculate the structure of a premixed flame
front interacting with an isolated vortex or a vortex pair.
The flame front was modeled as a two-dimensional propagating surface convected by the fluid motion while advancing at a constant laminar flame velocity, and the vortices
were frozen during the interaction. The rotational velocity
induced by the vortex was specified using Eq. (83). Pocket
formation is observed to be much easier for a counter-rotating vortex pair than for an isolated vortex. During the interaction with a single vortex, peaks of positive and negative
flame stretch appear near the vortex center, while the outer

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

265

bending effect is also observed in this model problem with


increasing values of u 0 =Sl : The effect of stretch on the mean
burning velocity is insignificant when the vortex scale is
much larger than the flame thickness. On the contrary, for
small vortices, stretch causes a noticeable decrease in the
mean burning velocity. Strain rates are observed to be quite
small over the whole flame surface area. Extinction generally occurs at the cusp for Lewis numbers less than one,
while it becomes more probable at the flame tip for Le 1:

Fig. 33. Maximum flame stretch during interaction with a single


vortex for various parameters (from Ref. [186]).

regions are mainly controlled by the curvature effects. The


peak of positive flame stretch slowly grows due to the positive flame curvature that causes outward propagation of the
flame front. This increases the flame area, which decreases
later on after reaching its maximum. On the contrary, the
negative peak grows indefinitely until a cusp is formed. The
importance of adding a curvature contribution to determine
maximum flame stretch when varying umax
u =Sl was emphasized (Fig. 33). The interaction duration scales with the
inverse of the maximum flame stretch. During the interaction with an array of identical vortices, flame curvature is
found to depend more sensitively on vortex size than on
vortex velocity, with mean flame curvature scaling as the
inverse of vortex size. Vortex arrays generate considerably
higher mean flame stretch than an isolated vortex of the
same strength. It is also observed that the coupling between
the effects of adjacent vortices leads to very different flame
stretch statistics than simply summing up individual isolated
contributions. Finally, the maximum flame stretch K was
found to be correctly estimated by K umax
u =rc Sl =2rc :
The spatial extent of flame stretch profiles scales as rc2 K=Sl
for a wide range of umax
u =Sl ratios [187].
The G-equation and the incompressible flow assumption
were used in Ref. [188] to calculate flame propagation and
stretching in an array of modified Burgers vortices. In practice, the interaction with two successive vortices is found
sufficient to predict the effect of an infinite row of co-rotating vortices. In order to take flame stretch into account,
strain and curvature are computed at each point and the
corresponding local burning velocity is then determined
and used to propagate the front. Because of numerical
problems coming from cusp formation, a limiting value
for the burning velocity is chosen, depending on the Lewis
number. Because regions with negative curvature collapse
into cusps, positive curvature dominates the overall distribution of curvature over the flame surface. The classical

5.2.7. Consideration of detailed chemistry


Direct simulations of the head-on interaction between a
pair of counter-rotating vortices and a lean premixed
methaneair flame were carried out using detailed (52 reactions) and reduced (four global reactions) chemical schemes
for comparison purposes [189]. It is found that the
(computed) luminous emission does not closely follow the
heat release rate. Comparing the profiles of different chemical radicals, significant deficiencies are locally observed for
the evolution of CO, CO2 and H2O using the reduced
scheme, mainly because transport of intermediate species
is neglected. On the contrary, the CPU time is reduced by
one order of magnitude compared to that required by the full
reaction scheme.
Flame/vortex interactions in a premixed jet flame were
investigated using axisymmetric simulations with detailed
chemistry (nine species for H2 air combustion) [190]. As
the Lewis number is below unity for a fuel-lean hydrogen
flame, a decreasing temperature along the length of the jet
flame is observed. To investigate this point, three identical
calculations are carried out assuming constant homogeneous
Lewis numbers equal to 0.5, 1 and 2 successively. The flame
with Le 0:5 appears to be similar to the detailed chemistry
and transport model case, but with more buoyancy-induced
vortices developing outside the flame surface. Temperature
is quite uniform all along the flame front for Le 1; and
increases for Le 2: The large buoyancy-induced toroidal
vortices lead to a strong entrainment of ambient air into the
combustion products, and thus increase the production of
thermal NO.
Studies of the head-on collision between a stoichiometric
methaneair flame and a counter-rotating vortex pair using
a low-Mach number approximation and detailed chemistry
(skeletal C1 mechanism with 27 reversible reactions) were
also carried out [191]. The authors find that the total flame
stretch-rate is largely dominated by strain-rate, the flame
thickness being greatly decreased in regions with large positive tangential strain-rates. They also observe that flowinduced wrinkling and stretching significantly affect the
reaction layers, leading to large variations of the local reaction rates. For example, an overall decrease in hydrocarbon
consumption rates is found in the cusp region, in relation
with a high production rate and a low consumption rate of
the H radical. This is not observed using simpler reaction
mechanisms. Diffusion effects are found to be dominant
over reaction and convection in the cusp region, which is

266

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

Fig. 34. Observation of a wrinkled flame during premixed flame/


vortex ring interaction (CH spontaneous emission, w 0:7; Rev
440; umax
u =Sl 7:3; 2R=df 3:63). From Ref. [111].

consistent with previous correlations in terms of curvature.


On the contrary, a simultaneous decrease in production and
consumption rates of the H radical is found on the centerline, where strain-rate is high but curvature effects are negligible. On the centerline, extinction is coupled to radical
depletion, in particular of H, O and OH, due to the movement of the flame towards the products, leading to an
increased temperature and reduced reactant concentrations.
The evolution of flame burning rates was studied in the same
configuration [192]. It is shown that the peak dilatation rate,
CH, OH , C2 ; CH are not appropriate, OH and CO2 are
acceptable but not perfect, HCO being the best flame marker
and the closest to the burning rate.
An inverse partially premixed flame established by injecting a CH4 air annular jet between a central air jet and
coflowing air on the outside [93] was used to compare
results obtained with a global one-step chemistry and a
semi-detailed model for methane/air combustion including
17 species. Reactions occur much faster with the global
scheme, leading to a reduction in flame height and changes
in flame dynamics. Local extinctions are only captured with
the detailed mechanism. Trying to adjust the chemical
kinetics parameters of the global scheme to obtain the
correct flame structure was not successful, leading to the
conclusion that a detailed representation of the chemistry
is important for predicting stability (ignition/extinction
limits) and pollutant emissions. Using the detailed scheme,
combustion is observed to take place in two parts. First,
methane is decomposed in H2 and CO, which is followed
by conversion of these fuels into combustion products.
5.3. Premixed experiments
5.3.1. Flame propagation
Earliest measurements of flame propagation speeds were
performed by McCormack [81,82]. The flame propagation
speed in a vortex ring of premixed propane and air was
found to be significantly higher than that observed in a

quiescent gaseous mixture. The speed appeared to approximately increase as a linear function of the vortex strength.
McCormack concluded that the turbulent behavior in the
flow could not fully account for this increase in flame propagation speed. In a similar configuration, Ishizuka et al.
showed that over the range of maximum tangential velocities available, the flame velocity was approximately proporand its
tional to the maximum tangential velocity umax
u
proportionality factor was about unity [84,193].
Sinibaldi et al. measured the local displacement speed Sl
and values were found to vary over a wide range (by a factor
of 7.5) [104]. The measured values of Sl did not agree with
the steady-state theory of stretched flames. Instead, in most
regions, results were following the same trends as those
found at Bunsen tips by Echekki and Mungal [194] and in
the calculations of Najm and Wyckoff [195]. The measured
displacement speeds Sl were more sensitive to changes in
flame curvature than theory predicts. Using a modified
theory including only flame curvature and no effect of strain
rate allows an agreement with the measured results at early
interaction times, but differences appear at later times due to
unsteady effects.
5.3.2. Influence of vortex characteristics
Spectral diagrams are now used to identify regimes of
combustion and establish combustion diagrams (see Section
6 on this particular application). This complements the
phenomenological analyses used earlier. The vortex size
influence is a key-point in these analyses. As previously
predicted by Poinsot et al. [176], the experiments carried
out by Roberts et al. [9597] confirmed the existence of
four distinct regimes: the no-effect regime, the wrinkled
flame regime, the pocket formation regime, and the quenching regime. Experiments provided the boundaries between
these regimes for methane and propane but they were not
able to scan the same range of ratios of the vortex size to the
flame thickness. There is also a difference in definition of
vortex size: the vortex core diameter 2rc is used in Refs.
[9597] while the outer dimension of the dipole 2R rc
is used in Ref. [176]. The range of sizes studied in Refs.
[9597] is between 6 and 32 instead of 15.3 for 2rc =df :
Experiments indicate that small vortices are less likely to
disturb the flame front than large vortices having the same
vortex strength umax
u =Sl which is consistent with the simulations of Poinsot et al. [176]. This stems from the fact that
small vortices have a lower Reynolds number than large
vortices and therefore decay quicker because of viscous
forces eventually before they can effectively interact
strongly with the flame. These observations were only
made on methaneair flames, the vortices for propaneair
flames being too big to exhibit this kind of behavior. More
recently, Renard et al. [111] were able to examine regimes
which could not be reached with the setup used in Refs.
[9597]. Using propaneair flames, they confirmed that
small vortices are less likely to wrinkle the flame front or
even to create pockets (Fig. 34).

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

267

Fig. 35. Single vortex quenching curve from Renard et al. [111]. Hollow symbols correspond to a propane flame, f 0:6; Le 1:87; sl
14 cm s1 ; df 0:55 mm [95]. Dotted hollow symbols are our measurements for different values of f . Filled symbols are the simulations of
Poinsot et al. [176]. Circles correspond to the no-effect region, diamonds to the wrinkled flame region, squares to the pocket formation region
and triangles to the quenching region. The solid line is the Kolmogorov line Re 1 (vortices are dissipated by viscosity if Re 1). The
dashed line represents Vrr 1 (a pocket may be formed only if Vrr 1). The long-dashed line corresponds to Pr 1 (the vortex is
dissipated before it reaches the flame if Pr 1). The dashed-dotted line is the line Ka 1 (Ka 1 is necessary to have local flame
quenching). The dotted line is the quenching limit deduced from DNS calculations of Poinsot et al. The thin solid line is the no-effect limit.

These experiments as well as previous simulations indicate that the thin flame regime of combustion extends over a
larger range of turbulence intensities than previously
assumed. One may then conclude that a significant fraction
of turbulent fluctuations are too weak to wrinkle the reactive
layer and that a fraction of the small-scale turbulence can be
neglected in simulations of turbulent flames. It also follows
that micromixing models, which assume that small vortices
control turbulent combustion dynamics, are inadequate.
Experiments on periodic flame/vortex interactions also indicate that smaller vortices are less effective in generating
flame area [196]. The flame area increases as a relatively
weak function of umax
u =Sl as shown by experiments on a
Karman vortex street interacting with a V-shaped flame.
This result may not be general. Fig. 35 indicates that the
translational velocity of the vortex with respect to the reactants influences the flame distortion, pocket formation, and
quenching processes.
Taking a closer look at the detailed structure of the vortical flow field during the interaction, it is noticeable that the
maximum strain rate position depends on the characteristic
size of the vortex. Experiments reported in Ref. [98] indicate
that the largest strain rates are reached at a distance form the
centerline where the flame propagates close to the vortex
core boundary. For smaller vortices, the maximum strain
rate is induced on the centerline. It follows that the effects
of strain rate on a flame cannot be modeled as a self-similar
process, a standard practice in flamelet models of turbulent
combustion.

The vortex size also affects the distribution of the angle


between the flame surface normal direction and the direction
of propagation. The distribution peaks near 0 for small
values of the vortex intensity umax
u =Sl ; while an increase in
umax
=S
results
in
broadening
of
the flame orientation distriu
l
bution and a shift toward positive angle values stemming
from the increased distortions in the flame front and
increases in the effective flame propagation speed, respectively. If the vortex size is increased and the Lewis
number is decreased below 1, the flame orientation distribution is again broadened and shifted but the effect is not as
pronounced.
5.3.3. Effects of stretch
Stretch effects dominate flame wrinkling to the extent that
pockets may be formed. For a toroidal vortex, it is known
that the flow field near the centerline is similar to a stagnation point flow, and this flow exerts extensional strain on the
flame. In experiments reported in Ref. [95], the flame perimeter is typically increased by a factor of 3 during vortex
crossing. This increase is followed subsequently by a
decrease due to the burn-through of narrow regions of reactants. Also, the reaction rate is enhanced due to flame curvature when the flame front is convex towards the reactants for
Lewis numbers less than 1, in agreement with laminar
stretched flame theory [96,97].
Another important feature noticed in experiments is that
while strain rates and curvature exhibit large variations in
the direction normal to the flame, their sum is approximately

268

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

constant in cold reactants, through the preheat zone, and


even past the location of peak volumetric expansion [99].
Stretch including strain rate and curvature effects is
responsible for an increase of the flame front surface area
but this does not mean that the global consumption is
increased in the same proportion. Indeed, Lee et al.
measured the peak OH LIF intensity for hydrogen-air and
propane-air flames [196]. They found that for hydrogenair
flames this peak could depart from the values corresponding
to a plane unstrained flame by 20150%. In these experiments, the flame curvature ranged from 1.5 to 0.7 mm 1.
For propaneair flames measured variations of OH LIF
were within ^20% of the value at zero curvature. One
may conclude that the local flame speed should not be
taken as a constant in turbulent premixed flames. This is
specifically not adequate for hydrogen flames, while it is
more reasonable for typical hydrocarbon flames.
Measurement of the averaged peak OH LIF intensity
[196] indicate that this quantity is nearly linear with respect
to the variation in flame curvature from 1.2 to 0.8 mm 1.
A more careful analysis showed that the average flame
stretch measured in an individual flame is found to be independent of the average strain rate and linearly related to the
average curvature. This corroborates the stretched laminar
flame theory used for turbulent premixed flames, which
assumes that the local flame speed is a linear function of
the flame stretch and a Markstein length [92]. However, the
correlation between the peak OH LIF intensity and curvature tends to level off for positive curvatures larger than
0.5 mm 1 as umax
u =Sl increases, indicating that the response
of the flame to large flame stretch may be non-linear at high
umax
u =Sl [91]. The flame curvature probability density functions (pdf) for flames interacting with Karman vortex streets
show a slight shift toward positive flame curvature stemming from the large area of positively curved flame
elements that develop downstream along the V-shaped
flame. If the vortex size is decreased, the flame curvature
tends to increase and thus the pdfs broaden, while umax
u =Sl
and the Lewis number have relatively small effects on the
flame curvature pdfs.
5.3.4. Ignition, extinction
Quenching associated with stretching may be investigated
by considering flame/vortex interaction [94]. According to
the KlimovWilliams criterion (see also Section 6 for a
further analysis), quenching in a turbulent flow field occurs
for Karlovitz numbers greater than one. The vortices that
induce quenching have a typical scale equal to the flame
thickness. This criterion is not confirmed by numerical
simulations [176]. The quenching boundary deduced from
simulations is shown to lie at a distance from the Klimov
Williams boundary (Fig. 35). Roberts et al. [97] experimentally verified this point and retrieved the trends predicted by
direct numerical simulations. Quantitative values differ
because the calculations were two-dimensional and assumed
larger heat losses than those prevailing in the experiment

[176]. Simulations indicate that heat losses have a major


effect on flame extinction. For example, decreasing the
burnt gas temperature in hot stream by 15% for a steady,
planar counterflow flame reduces the quenching strain rate
by more than a factor of two. In practice, this effect is not as
strong as in computations. Mueller et al. find that typical
heat losses in their experiment cause less than a 10% change
in OH mole fractions for moderate strain rates (i.e. up to half
the extinction value) [99].
It is often assumed in flamelet models of turbulent
combustion that the flame locally behaves like a steady,
planar counterflow flame. This description was evaluated
in Ref. [98] by measuring the local rates of stretch. The
data indicate that unsteady effects do not influence quenching, at least for the conditions of their experiment. Measured
extinction stretch rates were found to be close to values
obtained for corresponding steady, planar counterflow
flames.
5.3.5. Detailed chemistry effects
The chemical state in a given flame is always difficult to
characterize because the number of species which are accessible to measurements is reduced. It is also not easy to obtain
quantitative data from fluorescence measurements because
the fluorescence signal depends on local temperature and on
the quenching processes due to collisional deexcitation.
These issues are well emphasized in the work of Nguyen
and Paul [107]. From simultaneous images of OH and CH
fluorescence taken during a flame/vortex interaction, these
authors were able to show that the CH distribution was
broken in places where OH was still continuous. One may
then wonder if OH is a good indicator of flame extinction
processes. It may be considered on one hand that the detection of OH is ambiguous because some OH may be due to
the dissociation of water in hot gases. On the contrary, it was
shown in other non-premixed experiments that if OH
decreases below a certain level, then this corresponds to a
quenched state [110]. Experimental results indicate however
that the OH response to the vortex action does not follow the
CH changes. The peak OH concentration may increase or
decrease, depending on the flame stoichiometry. In the
premixed case, the structure of CH and OH profiles does
not evolve consistently as in a strained laminar flame
submitted to a similar transient strain rate.
5.3.6. Vorticity generation and baroclinic effects
Vorticity generation and dissipation in turbulent reacting
flows has fundamental implications for combustion intensity. A quick analysis of this problem suggests that turbulence is damped across a premixed flame front in relation to
the volumetric expansion and the increase of viscosity. In
certain cases, however, flames can generate vorticity due to
the action of a baroclinic torque [98]. This particular point is
investigated in Ref. [101]. In the case of large eddy scales,
low turbulence levels, and products flowing upward, flamegenerated vorticity may exceed that of the initial vortex ring.

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

This additional vorticity is created by a torque associated


with a pressure gradient. This effect acts as a stabilizing
mechanism in the case of a downward-propagating flame.
It suppresses wrinkles independently of whether the corrugations point towards the reactants or products. This effect is
also observed in the absence of an impinging vortex. A
flame propagating downward under normal gravity conditions is almost flat whereas in microgravity, it keeps a
spherical shape [103]. Theoretical estimates indicate that
the baroclinic torque mechanism scales with Froude
number. The induced velocity was evaluated to be more
than twice the laminar burning velocity, which explains its
important effect.
5.3.7. Lewis number effects
Pocket formation and quenching depend on the Lewis
number as predicted in Ref. [176] and confirmed experimentally in Ref. [111]. For Lewis numbers greater than one,
thermodiffusive effects are stabilizing and thus limit pocket
formation by increasing the flame speed where the flame
front is convex with respect to fresh gases. Regarding
extinction processes, Roberts et al. noticed that lean
propaneair and ethaneair flames Le 1 need much
larger Karlovitz numbers to achieve quenching than lean
methaneair flames Le 1 [97]. The same conclusions
were reached in Ref. [111]. This may be attributed to thermodiffusive instabilities which appear on the flame front for
Le 1 and tend to promote the elongation action of the
vortex. This is confirmed in experiments for both microgravity and one-g conditions carried out by Sinibaldi et
al., where the degree of wrinkling of a methaneair flame
is larger than that of a propaneair flame [103]. Similar
observations were made in a periodic flame/vortex interaction for H2/He/air flames Le 0:21; CH4/air flames Le
1; and C3H8/air Le 1:79 [91,196]. The OH LIF intensity
increases when the local flame curvature becomes positive
(negative) for thermodiffusively unstable (stable) flames. In
addition to this phenomenon, it was also found that flame
area increases more rapidly with umax
u =Sl for H2/He/air
flames than for CH4/air or C3H8/air flames [91].
Flame curvature probability density functions were found
to be nearly symmetric with respect to a zero mean and
insensitive to Lewis number. The flame front is oriented
more randomly as Lewis number decreases and umax
u =Sl
increases. An explanation of this behavior could be that
even though the local flame curvature may have a strong
influence on the local structure of non-unity Lewis number
flames, these fluctuations cancel in the mean due to the
linear relationship between local burning rate and curvature
for the most probable values of curvature (between 0.5
and 0.5 mm 1) and due to the symmetry and zero mean of
the curvature distribution. As a consequence, the main effect
of turbulence for non-unity Lewis number configurations is
to wrinkle the reactive layer, to produce flame surface area,
and slightly increase the mean burning rate per unit surface
area through flow straining effects.

269

5.3.8. Gravity effects


It is not easy to study flame/vortex dynamics under microgravity conditions. This requires specific facilities and
experimental setups. Sinibaldi et al. used the NASA Lewis
Research Center 2.2s drop tower [102,103] and found that
the amplitude of the wrinkle in the flame caused by the same
single vortex is 1.52 times larger under microgravity
conditions than that observed in the corresponding one-g
case. This suggests that turbulent flames are wrinkled and
propagate significantly faster in microgravity and may
increase hazard of fire in space stations. Moreover, it was
found that the laminar flame speed Sl for a lean propane
air mixture in microgravity is 1.82.2 times larger than
that of a flame moving downward at one-g, which is
consistent with theoretical predictions, which state that
buoyancy causes flames moving downward within tubes
to move slower than those propagating upward. As
explained in a previous section, this is due to the fact that
new vorticity is created by buoyancy which suppresses
wrinkles in one-g cases. When this mechanism is absent,
the flames wrinkle more freely.

5.3.9. Time-dependent phenomena


Fast transients induce time-dependent effects. This is
made evident in the measurements of the temporal evolution
of reaction intensity reported in Ref. [100]. In regions where
positive stretch rates are 210 times greater than the
extinction value for a steady flame, flame strength
remains above 90% of its unstretched value for 1.2
laminar flame time, after which it drops quickly as
extinction occurs. This extinction time delay highlights the
existence of delays between hydrodynamic perturbations
and the flame response complicating the prediction of
flame behavior.

5.3.10. Thermal expansion effects


Gas expansion within the flame tends to completely
attenuate the initial vortex if the flame is not severely
stretched as shown in Ref. [101]. In the case investigated,
viscous dissipation within the flame is not strong because
vortices are three to four times larger than the flame thickness. One must remember that viscosity is more important if
smaller vortices are considered as pointed out in Ref. [111].
The strong damping by volumetric expansion should occur
for vortices of all sizes. A sufficiently strong vortex might
survive flame crossing if it weakens the flame down to the
extinction limit so that the attenuation of vorticity by gas
expansion becomes negligible.
The measurement of the local heat release rate would be
interesting for comparison with numerical results, but is
quite difficult. The possibility of using the product of simultaneous PLIF measurements of OH and CH2O during interaction of a V-shaped methaneair premixed flame with a
line-vortex pair has been assessed in a recent work [197].

270

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

6. Turbulent combustion diagrams

and

Studies of flame/vortex interactions provide detailed


information on the complicated process which take place
when a flame spreads in a turbulent flow field. The identification and classification of regimes of combustion
progressed to a great extent mainly because of simulations
and experiments on flame/vortex configurations.
Early attempts to classify turbulent flames in different
regimes in diagrams are due to Barre`re [198], Bray [199],
Borghi [200], Williams [74] and Peters [3] for premixed
flames and to Bilger [78] and Peters [201] for non-premixed
flames. These analyses are essentially phenomenological.
The nature of combustion is more complicated and initial
ideas that had to be revised to take into account some less
obvious aspects. To introduce more recent idea and identification of regimes, it is convenient to summarize the basic
parameters allowing the classification of combustion
regimes.

tt

u 02
l
t0
e
u

The turbulent Reynolds number Ret characteristic of the


large scale eddies is defined by
Ret

u 0 lt
n

6.1. Turbulence description

u 0i u 0i
2

84

and
u0

2k
3

1=2

85

where designates a mass average operator. The turbulence dissipation rate e is given by
* 0
+
2u i 2u 0i
86
en
2xj 2xj
The length and time scales corresponding to the large eddies
are
lt

u0 3
e

87

89

Turbulence is composed of a continuous spectrum of


eddy length and velocity scales. The macroscales are the
biggest ones. At the other end of the range of sizes, the
Kolmogorov scales feature a characteristic velocity uk, a
time scale t k and a length scale lk. The size of the Kolmogorov eddies is of the order of the thickness of a viscous
layer developed during a time t k. Assuming that the Kolmogorov eddies are isotropic and that all the energy contained
in the macroscale eddies is dissipated at this level, one
obtains the following set of expressions
uk en1=4

A flame is influenced by turbulence into which it is


spreading and one may therefore characterize the turbulence
field in the reactants ahead of the flame. The interaction
between turbulence and combustion is two-ways: the turbulent field could be changed by the flame; and a turbulent
flame may even propagate into non-turbulent reactants
[202,203]. Conditions prevailing in the upstream flow
ahead of the flame are generally taken as reference
conditions. In what follows, reasoning relies on the Kolmogorov cascade to describe turbulence fluctuations and to
express relations between the spatial and time scales. The
turbulence intensity may be characterized by the large scale
velocity fluctuation u 0 (or equivalently by the turbulent
kinetic energy per unit mass k) [204]

88

lk

tk

n3
e

90

!1=4

 1=2
n
e

91

92

The Reynolds number Rek associated to the Kolmogorov


scales is equal to 1, which is a particularity very often used
in the derivation of combustion diagrams. To be more
precise about the size ls of smallest motions, it has been
argued that a reasonable definition of ls is the wavelength
corresponding to the centroid of the dissipation spectrum
[205]. Using a standard model of the energy spectrum in
high-Reynolds number turbulence, one obtains
ls 13lk

93

6.2. Turbulent premixed combustion diagram


In the premixed case, the flame thickness d f and the laminar burning velocity Sl are easy to define as these quantities
are intrinsic properties of the flame. Considering that large
scales of turbulence are characterized by the length scale lt
and the velocity scale u 0 , regimes of combustion may be
described in terms of these four physical quantities. One
may then draw a diagram in terms of length scale (lt/d f)
and velocity scale (u 0 /Sl) ratios. This choice is not unique.
One may use alternatively u 0 /Sl and the Reynolds number
[199] or u 0 /Sl and lt/d f [74] or the Reynolds and the Damkohler numbers [206]. Other parameters might also be important (in particular the density ratio between fresh and burnt
gases and the Lewis number, a characteristic ratio for heat
losses, etc.). As the turbulent flows have the peculiarity to
show a spectral distribution of vortex sizes and strengths, it
is useful to consider the smallest vortex sizes corresponding

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

to the Kolmogorov scales. The size lk and the velocity uk of


these eddies are given by Eqs. (90) and (91).
One may then introduce the Damkohler and the Karlovitz
numbers to characterize the competition between chemical
and fluid dynamical processes. The turbulent Damkohler
number is defined as
t
Da t
94
tc
and the turbulent Karlovitz number is
t
Ka c
tk

95

It is also convenient to express these numbers in terms of the


usual combustion diagram coordinates
Da

lt Sl
df u 0

96

and

d
d u
Ka 4 f f k
Sl
lk Sl

u0
Sl

!3 
2

lt
df

 1
2

97

where
1
4

tk

 1=2
e
n

98

is the inverse of the Kolmogorov time and describes the


straining by the Kolmogorov eddies. The approximation in
Eq. (97) stems from the use of the ZeldovitchFrank
Kamenetskiis estimate of the flame thickness

df 4

l=cp Tref
;
ru Sl

99

where l and cp are typical values of the thermal conductivity


and the specific heat of the unburnt gas mixture at a reference temperature (here, Tref 300 K and r u is the unburnt
gas density. The turbulent Reynolds, Karlovitz and
Damkohler numbers are linked by the relation
Ret Da2 Ka2

100

The Karlovitz number compares the flame thickness to


the Kolmogorov scale (using the fact that the Reynolds
number of Kolmogorov eddies is equal to one and that the
Prandtl number is close to one for gases)
 2
df
Ka
101
lk


uk
Sl

2

102

Two characteristic scales are defined for the flow structure,


which are the bounds of the vortex spectrum and only one
for the flame. Nevertheless, it is well known that in typical
industrial applications, the flame front is very thin regarding
the main transient species (the Damkohler number is high)

271

but quite thick for some species and especially pollutants


which are of great importance for industrial purposes. This
is one of the limitations of this approach.
When the Damkohler number is much greater than one,
the flame front is thin with respect to the turbulent eddies
and its internal structure is not affected by turbulent fluctuations. In this flamelet regime, the thin flame is wrinkled by
turbulence. In the classical theory, the flamelet regime
prevails as long as the Kolmogorov time scale is larger
than the chemical characteristic time. The boundary of
this zone is defined by the KlimovWilliams criterion [207]
Ka 1

103

This criterion was initially used to distinguish the flamelet


regime Ka 1 and a domain where the flame structure is
modified by turbulence Ka 1: It is then more suitable to
describe the region corresponding to Ka 1 as the flamesheet region than as a reaction-sheet regime [74] as the
preheat zone is also concerned. When Ka exceeds unity,
two subdomains may be distinguished.
Da 1 and Ka 1: The chemical reaction is slower
than turbulent mixing. All the eddies of the turbulence
spectrum are able to disrupt the inner reaction regions.
Then, the reaction rate is controlled by chemistry, while
turbulent eddies mix the reactants. This well stirred reactor regime hardly even occurs in combustion processes,
but is rather common in chemical engineering.
Da 1 and Ka 1: In contrast to the previous case,
chemical reactions are faster than turbulent fluctuations.
Then, the reaction rate is essentially controlled by turbulent mixing. Kolmogorov vortices are smaller than the
flame thickness and may penetrate into the flame sheet
causing local quenching. The notions of flame structure
and laminar-burning velocity are less meaningful. The
chemical time scale however remains meaningful. Inner
reaction zones smaller than the flame thickness persist
and quenching of these inner flame layers by stretch
constitutes an important physical process [201]. It is
then possible to define a quenching length scale lq
which corresponds to the size of the largest eddy within
the inertial range which may quench a thin reaction zone
lq et3q 1=2

104

where t q is the quenching time scale. The corresponding


velocity scale uq is introduced
lq
d
f
uq
Sl

105

It is easy to show that


3
lq
Da 2
lt

106

and also
3
lq
Ka 2
lk

107

272

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

Fig. 36. Premixed combustion diagram based on a DNS spectral diagram (from Ref. [176]).

All the eddies smaller than lq would also be able to


quench the reactive layers and they contribute to enhance
mixing. Then, lq defines the size of a local well-stirred
reactor. The corresponding regime of combustion
involves distributed reactive layers.
The flamelet regime corresponding to Ka 1 may be
subdivided in two regions:
u 0 =Sl 1: In this wrinkled flamelet regime, the velocity
u 0 , which is of the same order of magnitude as the
characteristic velocity of integral scale vortices, is not
sufficiently large to induce merging and canceling
between several flame fronts. The laminar propagation
of the flame is predominant, and turbulence and combustion interact weakly allowing simplified analyses in the
high activation energy limit.
u 0 =Sl 1: Multiple flame sheets are found in the corrugated flamelet regime. As u 0 Sl uk ; the largest
vortices are strong enough to distort the reactive zone
and induce flame interactions. Conversely, the smallest
vortices do not wrinkle the flame front as they have a
turnover velocity smaller than the burning velocity.
Among the many eddies which interact with the flame,
one may single out those corresponding to a velocity fluctuation equal to the laminar flame speed [201]. The size lG of
these eddies should be greater than lk. According to the
Kolmogorov similarity rules, the size and velocity of these
eddies are such that
u3G elG

108

Now, if uG is equal to Sl, the eddy size designated by Peters

as the Gibson scale lG is given by


lG

S3l
e

109

This analysis was checked by Poinsot et al. using flame/


vortex interaction calculations [176]. Systematic simulations for various vortex sizes and velocities were synthesized in a spectral diagram describing the outcome of
interactions between a single vortex pair and a flame front.
The principle of this classification differs from the
previously described turbulent combustion diagram. Here,
one considers the flame response to an isolated vortex and
not to the multi-scale eddies composing a turbulent flow.
The result obtained is reproduced in Fig. 35.
This kind of diagram may be used to identify premixed
turbulent combustion regimes as a function of turbulence
characteristics on the basis of Kolmogorovs cascade. For
each point of the turbulent combustion diagram, a spectral
diagram may be drawn where Kolmogorovs cascade is
characterized by a segment of slope 1/3 ended by the integral and Kolmogorov scales. If this segment intersects the
quenching curve, then the point considered belongs to the
quenching limit of the turbulent combustion diagram. By
drawing segments for different values of integral scale
length and velocity, the quenching curve may be determined. Fig. 36 shows the result of this approach. It is
obvious from the diagram that the flamelet domain is
much wider than generally expected from the Klimov
Williams criterion. Clearly vortices which quench the
flame front have to be one or two orders of magnitude larger
and faster than Kolmogorov scales. When this result is used
to reconstruct a new diagram for the regimes of turbulent
combustion, one finds that quenching is not likely to occur

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

273

Fig. 37. Premixed combustion diagram (from Ref. [209]).

in many practical cases and that a flamelet description may


be safely used. Spectral diagrams give also a cut-off scale
below which the flame does not respond to incoming perturbations.
The previous analysis was recently refined [208,209] by
noting that the flame structure should be described in terms
of two scales: the flame thickness d f and the reactive layer
thickness d r (in typical flames, dr 0:1df : If the size of the
turbulent eddy is larger than d r but smaller than d f, it will be
able to disturb the preheat zone but not the reaction zone.
One may then determine the order of magnitude of the
Karlovitz number corresponding to the case where turbulence distorts the preheat zone
 2
 2
dr
d
Kar
100 f 100Ka
110
lk
lk
This leads to a new classification of turbulent combustion
regimes.
Ka 1: The flame sheet may be divided in two subregimes.
u 0 =Sl 1: The velocity fluctuations do not cause flame
merging (wrinkled flame regime). The laminar propagation of the flame predominates, allowing simplified
analyses in the high activation energy limit.
u 0 =Sl 1: Turbulence is sufficiently strong to corrugate the flame (corrugated flame regime). The largest
vortices are strong enough to distort the reactive zone
and create some flame front interactions.
1 Ka 100: In this thickened thin flame regime, the
vortices thicken the preheat zone but do not modify the
reactive layer which retains the characteristics of laminar
combustion.

100 Ka: The well-stirred reactor regime is such that


both reaction and preheat zones are perturbed by vortices
and no flame structure may be identified.
This yields the diagram shown in Fig. 37. This diagram
indicates that many practical combustion devices operate in
the flamelet and the distributed reaction zone regimes. In
using the diagram, one should remember that the analysis
relies on intuitive considerations that only yield orders of
magnitude. Other limitations are listed below.
The analysis is based on the Kolmogorov cascade
description of turbulence. Real turbulent flows do not
follow this behavior. For this reason, different combustion regimes probably coexist in a given flame.
Limits in the diagram only define the approximate values
of parameters which induce a change in regime.
Unsteady effects are neglected. It was proved however
that a flame withstands transient strain rates greater than
the steady state quenching level.
All the physical quantities used in the diagram are not
easy to define in realistic reactive flows as the analysis
assumes a single step chemistry process. Real flames are
more complicated as the kinetics feature a range of time
scales.
6.3. Turbulent non-premixed combustion diagram
The identification of combustion regimes in nonpremixed cases is complicated by the lack of characteristic
length and velocity scales. The flame structure is imposed
by the velocity field and the mixing rate. One may still
attempt to consider that the flamelet regime prevails when
the reaction zone is thinner than the dissipative scales as

274

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

proposed by Bilger [78] but this criterion is difficult to use in


practice. The choice of dimensionless groups is an open
question. As the flame has no typical length scale, one
may choose instead a chemical time scale and compare it
with a characteristic turbulent time. This defines a Damkohler number. This number may be derived for example from
activation-energy asymptotics. It is then natural to choose as
a second dimensionless group the Reynolds number. One
may then establish a turbulent combustion diagram in terms
of Re and Da. Some authors prefer to work in terms of
mixture fraction fluctuations-Damkohler number. Indeed,
the mixture fraction field is a key variable in non-premixed
flame structure, as the width of the reaction zone s r is independent of the flow field in mixture fraction space, not in
physical space [201].
In a coordinate system moving with the flame front, in
which x and z coordinates locally define the stoichiometric
mixture surface, the flame extends towards the y direction.
The reaction zone thickness in physical space d r is related to
that in mixture fraction space s r by [201]

dr 

sr

2Z
2y st

111

The scalar dissipation rate


g
x~ D7Z 2

112

may be estimated in terms of the mixture fraction fluctuations Z 00 . One assumes that in a non-premixed turbulent
flame, the mean thickness of the mixing layer is of the
same order as the integral scale of turbulent motion
~ 1
7Z

ek3=2
e~

113

The diffusive layer thickness ld adjacent to the reactive


zones may be evaluated by
0
11=2
1
A
ld @
114
g
7Z 00 2
Now, the scalar dissipation x~ quantifies the intensity of
turbulent micromixing which operates at scales of the
order of magnitude of ld. As a consequence, the gradient
of the mean value Z~ may be neglected in comparison with
that of the fluctuation Z 00 [210]
g00 2
g 00 r D7Z
~ 2 rD7Z 00 2 27Z
~ r D7Z
r x~ rD7Z

115
Assuming that the mean scalar dissipation rate may be
modeled in the case of isotropic turbulence by




2Z 2
2Z 2
r x~ 3rD
3r D
116
2y
2y av
and using the fact that very often, the last term of Eq. (115)

is evaluated by a simple linear relaxation model


002
e~ 002
Zf
g
r D7Z 00 2 C D
CD Zf
tt
k~

117

This yields the following estimate for the reaction zone


thickness
!1=2
sr
3Dk~
d~ r
118
CD e~
00 2 1=2
Zg

~ e~ 1=2 defines the typical diffusion layer


The quantity Dk=
thickness when it is submitted to the large scale strain rate
~ Accordingly, reaction zones are separated and indie~ =k:
vidual non-premixed flamelets may exist if the mixture fraction variance is high whereas reaction zones are connected
when this is not the case. In addition to the previous analysis, Peters [201] gives a turbulent non-premixed diagram
similar to the turbulent premixed diagram. Another analysis
of the same type is outlined by Borghi [208]. Alternate
analyses are outlined by Libby and Williams [211], Bray
and Peters [212] and Vervisch [213].
Considering Eq. (114), the typical diffusion layer thickness is estimated by
 1=2
D
ld
119
x~
and the ratio between turbulence and diffusion scales is then
given by
 1=2

1=2
 1=2 ~3=2
k
k~
lk
x~
f
002
002 Re

Zf

Z
t
ld
D
e~
De~ 1=2
120
where the Schmidt number is taken equal to one and the
Reynolds number is
Ret

u 0 lt
k~

n
De~ 1=2

121

The ratio of the Kolmogorov scale to the diffusion layer


scale is given by
0
11=2
002
lk
Zf
@
A

122
ld
Re1=2
t
For a single step reaction nF F nO O ! P; one may show
that the reaction layer thickness is related to the diffusion
layer thickness by

dr ld Da1=ntot

123

where ntot nO nF 1 [213]. The ratio of Kolmogorov


scale to reaction layer thickness is then given by
0
11=2
f
002
lk
Z
@ 1=2 A Da1=ntot
124
dr
Ret
The Damkohler number is defined as a ratio of a mechanical
to a chemical time. The mechanical time is estimated as the

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

275

Fig. 38. Non-premixed combustion diagram (from Ref. [213]).

inverse of the scalar dissipation calculated at the stoichiometric surface

tm

1
xst

125

and the laminar Damkohler number is given by


Da

1
tc xst

126

or alternatively
Da

tt 1
tc Zf
002

127

It is convenient to express the mixture fraction fluctuation


in terms of a segregation factor S. This gives an indication
on the level of mixing: if S is close to 1, the flow is
composed of fuel and oxidizer pockets; if S is close to 0,
the flow is almost perfectly mixed. The Damkohler number
thus becomes [213]
1
tt 1
Da 
 tc S
Z1 Z

128

A high level of mixture fraction fluctuations leads to a


decrease of Da and even to a local extinction. This is in
agreement with the fact that high values of Zf00 induce a
thin mixing zone, leading to small values of ld.
To distinguish the flamelet regime from the distributed
reaction zone regime, one may consider the ratio of the
Kolmogorov scale to the reaction layer thickness. When
this ratio is greater than unity, the flamelet regime prevails.
Expression (124) then indicates that
!ntot =2
Re1=2
t
Da
129
002
Zf

One may then distinguish several regimes according to


Vervisch [213] who proposes a diagram synthesizing
those of Libby and Williams [211] and Bray and Peters
[212].
f
002 ntot =2 : This condition defines the flamelet
Da Re1=2
t =Z
regime. The segregation factor S is close to one, reaction
zones develop between unmixed reactant pockets and
flame element are well defined. If S 0:5; the flamelets
merge.
f
002 ntot =2
Da Re1=2
: The flame becomes sensitive to
t =Z
micro-mixing unsteadiness and one should distinguish
two sub-regimes. If S 0:5; the mean reaction zone is
thickened by the turbulent field. If S 0:5; local extinctions appear and grow as S comes close to unity.
This reasoning yields the diagram shown in Fig. 38. In using
this classification, one should remember that the analysis
relies on intuitive considerations and uses approximations.
Turbulence is described in terms of the Kolmogorov
scales.
Limits between regimes define a trend and should not be
interpreted as real frontiers between regimes.
The criterion separating flamelet and non-flamelet
combustion needs further evaluation.
Unsteady effects are neglected.
The analysis relies on a single chemical time scale while
real flames feature a range of time scales. This makes
the physical quantities used in this diagram not easy to
evaluate.
As already seen in the premixed case, flame/vortex calculations or experiments may be used to asses the validity of
combustion diagrams. They may be used to identify regimes

276

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

Fig. 39. Non-premixed combustion diagram based on a DNS spectral diagram (from Ref. [214]).

and determine the boundaries of the different modes of


combustion. Calculations reported in Ref. [214] were used
to build a spectral diagram. Fig. 39 indicates that the thin
flamelet regime prevails as long as the Damkohler is above a
certain value. Below this value, the flame does not respond
rapidly enough to fast temporal evolution of the scalar dissipation rate associated with the turbulent eddies. Another
limit is obtained when the ratio of the vortex size to the
initial flame thickness (the one before the interaction) is
of the order of one. Smaller vortices induce a delayed
flame response and additional curvature effects. No extension to a non-premixed combustion diagram is available at
this time.
7. Conclusion
The present article reviews past and recent progress in the
field of flame/vortex interactions. Flame/vortex interactions
are defined as an intermediate configuration between plane
strained flames and turbulent flames and they may be used to
study basic mechanisms of combustion dynamics. The
means employed to investigate the physical processes are
of three kinds.
Some theoretical analyses were carried out and have
provided considerable insight in a variety of processes
(flame structure, reacted core spreading, mixing enhancement, etc.). The problem is however complex enough and
this limits the application range of theoretical reasoning.
Experimental studies have provided a considerable
amount of data in a wide variety of geometries. However

the investigation is limited by operational requirements


and available diagnostics. Experiments are also limited
by flame stability phenomena, reproducibility of the
interaction, optical access, and quality of the vortical
structure.
Full numerical simulations have been used extensively
and have provided a large set of results. Simulations
are restricted on terms of computation domain size and
computational time requirement. Boundary conditions
must be treated with precautions to prevent parasitic
reflections. Two-dimensional simulations are appropriate
for these configurations, but detailed chemistry is often
needed to be able to carry out comparisons with experimental measurements.
Studies carried out in the field of flame/vortex interactions provide valuable descriptions of dynamical
effects on flames, extinction, ignition, flame front
merging, mixing, baroclinic torque, Lewis number
effects, gravity effects, and thermal diffusion effects.
Results also guide the identification of combustion
regimes and they have been used to improve turbulent
combustion diagrams.
The head-on collision of a vortex ring and a flame is
best documented as it is the simplest to interpret. But
configurations like burning rings are quite interesting
and deserve further attention. Other more complicated
configurations are now being considered in which the
flame or the vortex could be formed with a spray. This
would allow analyses of group combustion effects in a
well-controlled situation.

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

Acknowledgements
The authors wish to thank the Delegation Generale pour
lArmement for the PhD fellowship of P.H. Renard. This
work is part of a collaboration with the Air Force Research
Laboratory of the Wright Patterson Air Force Base of
Dayton (USA), the PCI laboratory of the University of
Bielefeld (Germany) and the LAERTE of ONERA (France).

References
[1] Brown GL, Roshko A. On density effects and large structure
in turbulent mixing layers. J Fluid Mech 1974;64:775816.
[2] Cantwell BJ. Viscous starting jets. J Fluid Mech 1986;173:
15989.
[3] Peters N. Laminar flamelet concepts in turbulent combustion.
In: 21st Symposium (International) on Combustion,
Munchen, Germany, 1986. The Combustion Institute,
Pittsburgh. p. 123150.
[4] Marble FE. Gasdynamic enhancement of nonpremixed
combustion. In: 25th Symposium (International) on Combustion, Irvine, CA, 1994. The Combustion Institute, Pittsburgh.
p. 112.
[5] Candel S. Combustion instabilities coupled by pressure
waves and their active control. In: 24th Symposium (International) on Combustion. Sydney, Australia, 1992. The
Combustion Institute, Pittsburgh. p. 127796.
[6] Rogers DE, Marble FE. Mechanism for high frequency
oscillations in ramjet combustors and afterburners. Jet
Propu 1956;26:45662.
[7] Keller JO, Vaneveld L, Korschelt D, Hubbard GL, Ghoniem
AF, Daily JW, Oppenheim AK. Mechanism of instabilities in
turbulent combustion leading to flashback. AIAA J 1982;
20(2):25462.
[8] Keller JO, Daily JW. The effects of highly exothermic
chemical reaction on a two-dimensional mixing layer.
AIAA J 1985;23(12):193745.
[9] Ganji AR, Sawyer RF. Experimental study of the flow field
of a two-dimensional premixed turbulent flame. AIAA J
1980;18(7):81724.
[10] Poinsot T, Trouve A, Veynante D, Candel S, Esposito E.
Vortex-driven acoustically coupled combustion instabilities.
J Fluid Mech 1987;177:26592.
[11] Yu KH, Trouve A, Candel S. Combustion enhancement of a
premixed flame by acoustic forcing with emphasis on the role
of large scale vortical structures. In: AIAA 29th Aerospace
Sciences Meeting, Reno, NV, 1991. AIAA 91-0367.
[12] Blackshear PL. Growth of disturbances in a flame generated
shear region. Technical Report 1360, NACA, 1956.
[13] Yu KH, Trouve A, Daily JW. Low frequency pressure
oscillations in a model ramjet combustor. J Fluid Mech
1991;232:4772.
[14] Barr PK, Keller JO, Bramlette TT, Westbrook CK. Pulse
combustor modeling: demonstration of the importance of
characteristic times. Combust Flame 1990;82(3/4):25269.
[15] Coats CM. Coherent structures in combustion. Prog Energ
Combust Sci 1996;22:427509.
[16] Lamb H. Hydrodynamics. Cambridge, UK: Cambridge
University Press, 1932.

277

[17] Couder Y, Basdevant C. Experimental and numerical study


of vortex couples in two-dimensional flows. J Fluid Mech
1986;173:22551.
[18] Nguyen Duc J-M, Sommeria J. Experimental characterization of steady two-dimensional vortex couples. J Fluid Mech
1988;192:17592.
[19] Guyon E, Hulin J-P, Petit L. Hydrodynamique physique.
Savoirs Actuels. InterEditions/Editions du CNRS, 1991.
[20] Lundgren TS, Mansour NN. Vortex ring bubbles. J Fluid
Mech 1991;224:17796.
[21] Rayfield GW, Reif F. Quantized vortex rings in superfluid
helium. Phys Rev A 1964;136(5):1194208.
[22] Didden N. Untersuchung laminarer, instabiler Ringwirbel
mittels Laser-Doppler-Anemometrie. Mitteilungen aus dem
MPI fur Stromungsforschung und der AVA 64, Max Planck
Institut, 1977.
[23] Sallet DW, Widmayer RS. An experimental investigation of
laminar and turbulent vortex rings in air. Z Flugwiss 1974;
22(6):20715.
[24] Dziedzic M, Leutheusser HJ. An experimental study of
viscous vortex rings. Exper Fluids 1996;21:31524.
[25] Lavrentiev M, Chabat B. Effets hydrodynamiques et mode`les
mathematiques. Editions Mir, Moscow, Russia. Translated
from Russian edition, 1980.
[26] Shariff K. Vortex rings. Annu Rev Fluid Mech 1992;24:235
79.
[27] Maxworthy T. Turbulent vortex rings. J Fluid Mech 1974;
64(2):22739.
[28] Maxworthy T. Some experimental studies of vortex rings. J
Fluid Mech 1977;81(3):46595.
[29] Saffman PG. The number of waves on unstable vortex rings.
J Fluid Mech 1978;84(4):62539.
[30] Saffman PG. Vortex dynamics. Cambridge, UK: Press
Syndicate of the University of Cambridge, 1992.
[31] Maxworthy T. The structure and stability of vortex rings. J
Fluid Mech 1972;51(1):1532.
[32] Weigand A, Gharib M. On the evolution of laminar vortex
rings. Exper Fluids 1997;22:44757.
[33] Heeg RS, Riley N. Simulations of the formation of an
axisymmetric vortex ring. J Fluid Mech 1997;339:
199211.
[34] Wakelin SL, Riley N. On the formation and propagation of
vortex rings and pairs of vortex rings. J Fluid Mech 1997;
332:12139.
[35] Saffman PG. The velocity of viscous vortex rings. Stud Appl
Math 1970;49(4):37180.
[36] Moore DW. The velocity of a vortex ring with a thin core of
elliptical cross section. Proc R Soc London A 1980;370:407
15.
[37] Moore DW. The effect of compressibility on the speed of
propagation of a vortex ring. Proc R Soc London A 1985;
397:8797.
[38] Wang C-T, Chu C-C, Chang C-C. Initial motion of a viscous
vortex ring. Proc R Soc London A 1994;446:58999.
[39] Taylor GI. Formation of a vortex ring by giving an impulse to
a circular disk and then dissolving it away. J Appl Phys
1953;24(1):104.
[40] Sallet DW. Impulsive motion of a circular disk which causes
a vortex ring. Phys Fluids 1975;18(1):10911.
[41] Saffman PG. On the formation of vortex rings. Stud Appl
Math 1975;54(3):2618.

278

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

[42] Pullin DI. Vortex ring formation at tube and orifice openings.
Phys Fluids 1979;22(3):4013.
[43] Didden N. On the formation of vortex rings: rolling-up and
production of circulation. Z Angew Math Phys 1979;30(1):
10116.
[44] Auerbach D. Experiments on the trajectory and circulation of
the starting vortex. J Fluid Mech 1987;183:18598.
[45] Voropayev SI, Afanasyev YD, Filippov IA. Horizontal jets
and vortex dipoles in a stratified fluid. J Fluid Mech
1991;227:54366.
[46] Laursen TS, Rasmussen JJ, Stenum B, Snezhkin EN. Formation of a 2D vortex pair and its 3D breakup: an experimental
study. Exper Fluids 1997;23:2937.
[47] Moore DW, Saffman PG. Axial flow in laminar trailing
vortices. Proc R Soc Lond A 1973;333:491508.
[48] Saffman PG. The structure and decay of trailing vortices.
Arch Mech 1974;26(3):42339.
[49] Sheffield JS. Trajectories of an ideal vortex pair near an
orifice. Phys Fluids 1977;20(4):5435.
[50] Pierrehumbert RT. Family of steady, translating vortex pairs
with distributed vorticity. J Fluid Mech 1980;99:12944.
[51] Swaters GE. Viscous modulation of the Lamb dipole vortex.
Phys Fluids 1988;31(10):27457.
[52] Cantwell BJ, Rott N. The decay of a viscous vortex pair.
Phys Fluids 1988;31(11):321324.
[53] Hesthaven JS, Lynov JP, Nielsen AH, Rasmussen JJ,
Schmidt MR. Dynamics of a nonlinear dipole vortex. Phys
Fluids 1995;7(9):22209.
[54] Burton GR. Uniqueness for the circular vortex-pair in a
uniform flow. Proc R Soc London A 1996;452:234350.
[55] Nielsen AH, Rasmussen JJ. Formation and temporal evolution of the Lamb-dipole. Phys Fluids 1997;9(4):98291.
[56] Glezer A. The formation of vortex rings. Phys Fluids
1988;31(12):353242.
[57] Gharib M, Rambod E, Shariff K. Universal time scale for
vortex ring formation. J Fluid Mech 1998;360:12140.
[58] Garten JF, Arendt S, Fritts DC, Werne J. Dynamics of counter-rotating vortex pairs in stratified and sheared environments. J Fluid Mech 1998;361:189236.
[59] Karagozian AR, Nguyen TT. Effects of heat release and
flame distortion in the transverse fuel jet. In: 21st Symposium (International) on Combustion, Munchen, Germany,
1986. The Combustion Institute, Pittsburgh. p. 12719.
[60] Karagozian AR, Nguyen TT, Kim CN. Vortex modeling of
single and multiple dilution jet mixing in a cross flow. J
Propu Power 1986;2(4):35460.
[61] Karagozian AR. An analytical model for the vorticity associated with a transverse jet. AIAA J 1986;24(3):42936.
[62] Yamashita H, Kushida G, Takeno T. A numerical study of
the transition of jet diffusion flames. Proc R Soc London A
1990;431(1882):30114.
[63] Ellzey JL, Laskey KJ, Oran ES. Dynamics of an unsteady
diffusion flame: effects of heat release rate and viscosity. In:
Kahl AL, Leyer JC, Borisov AA, Sirignano WA, editors.
Dynamics of deflagrations and reactive systems. Part I:
flames, Progress in astronautics and aeronautics, vol. 131.
Washington, DC: AIAA, 1989. p. 179.
[64] Kaplan CR, Oran ES, Kailasanath K, Ross HD. Gravitational
effects on sooting diffusion flames. In: 26th Symposium
(International) on Combustion. The Combustion Institute,
Pittsburgh, 1996. p. 13019.

[65] Marble FE. Growth of a diffusion flame in the field of a


vortex. In: Casci C, Bruno C, editors. Recent advances in
the aerospace sciences, New York: Plenum Press, 1985. p.
395413.
[66] Karagozian AR, Marble FE. Study of a diffusion flame in a
stretched vortex. Combust Sci Tech 1986;45:6584.
[67] Karagozian AR, Manda BVS. Flame structure and fuel
consumption in the field of a vortex pair. Combust Sci
Tech 1986;49:185200.
[68] Peters N, Williams FA. The asymptotic structure of stoichiometric methaneair flames. Combust Flame 1987;68:185
207.
[69] Rehm RG, Baum HR, Tang HC, Lozier DC. Finite-rate
diffusion-controlled reaction in a vortex. Combust Sci Tech
1993;91:14361.
[70] Miralles-Wilhelm F, Rangel R, Sirignano WA. An analysis
of molecular mixing in a vortical structure: bias in PDF
measurement. In: AIAA 27th Aerospace Sciences Meeting,
Reno, NV. 1989. AIAA 89-0482.
[71] Meiburg E. Lagrangian simulation of diffusion flames.
Combust Sci Tech 1989;31:169.
[72] Buckmaster J, Ludford GSS. Theory of laminar flames,
Cambridge monographs on mechanics. Cambridge, UK:
Cambridge University Press, 1982.
[73] Clavin P. Dynamic behavior of premixed flame fronts in
laminar and turbulent flows. Prog Energ Combust Sci
1985;11:159.
[74] Williams FA. Combustion theory. 2nd ed. Reading, MA:
Addison-Wesley, 1985.
[75] Laverdant A, Candel S. Numerical analysis of a diffusion
flame-vortex interaction. Combust Sci Tech 1988;60:79.
[76] Delhaye B, Veynante D, Candel S, Ha Minh H. Simulation
and modeling of reactive shear layers. Theoret Comput Fluid
Dyn 1994;6:6787.
[77] Bilger RW. Turbulent jet diffusion flames. Prog Energ
Combust Sci 1976;1(2/3):87109.
[78] Bilger RW. The structure of turbulent non-premixed
flames. In: 22nd Symposium (International) on Combustion,
Seattle, WA, 1988. The Combustion Institute, Pittsburgh.
p. 47588.
[79] Peters N. Laminar diffusion flamelet models in nonpremixed turbulent combustion. Prog Energ Combust Sci
1984;10:31939.
[80] B. Delhaye, Etude de flammes de diffusion turbulentes
simulations directes et modelisation. PhD thesis, Ecole
Centrale Paris, ECP 1994-48. 1994.
[81] McCormack PD. Combustible vortex ring. Proc R Ir Acad A
1971;71(6):7383.
[82] McCormack PD, Scheller K, Mueller G, Tisher R. Flame
propagation in a vortex core. Combust Flame
1972;19:297303.
[83] Cattolica RJ, Vosen SR. Two-dimensional fluorescence
imaging of a flamevortex interaction. Combust Sci Tech
1986;48:7787.
[84] Ishizuka S, Murakami T, Hamasaki T, Koumura K, Hasegawa R. Flame speeds in combustible vortex rings. Combust
Flame 1998;113:54253.
[85] Namer I, Bill RG, Talbot JL, Robben F. Density fluctuations
in a flame in a Karman vortex sheet. AIAA J
1984;22(5):64754.
[86] Hertzberg JR, Namazian M, Talbot L. Laser tomographic

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

[87]

[88]

[89]
[90]

[91]

[92]

[93]

[94]

[95]

[96]

[97]

[98]

[99]

[100]

[101]

[102]

study of a laminar flame in a Karman vortex street. Combust


Sci Tech 1984;38:20516.
Escudie D, Charnay G. Experimental study of the interaction
between a premixed confined laminar flame and coherent
structures. In: Fifth Symposium on Turbulent Shear Flows,
Ithaca, NY. Berlin: Springer, 1987. p. 34760.
Escudie D. Stability of a premixed laminar V-shaped flame.
In: Kuhl AL, Bowen JR, Leyer J-C, Borisov A, editors.
Dynamics of reactive systems. Part I: flames, Progress in
astronautics and aeronautics, volume 113. Washington,
DC: AIAA, 1988. p. 21539.
Escudie D. Experimental study of the interaction flame frontvortices. ERCOFTAC bulletin 1991:147.
Lee T-W, North GL, Santavicca DA. Surface properties of
turbulent premixed propane/air flames at various Lewis
numbers. Combust Flame 1993;93:44556.
Lee JG, Lee T-W, Nye DA, Santavicca DA. Lewis number
effects on premixed flames interacting with turbulent
Karman vortex streets. Combust Flame 1995;100:1618.
Nye DA, Lee JG, Lee T-W, Santavicca DA. Flame stretch
measurements during the interaction of premixed flames and
turbulent Karman vortex streets using PIV. Combust Flame
1996;105:16779.
Shu Z, Aggarwal SK, Katta VR, Puri IK. Flamevortex
dynamics in an inverse partially premixed combustor: the
Froude number effects. Combust Flame 1997;111:27695.
Jarosinski J, Lee JHS, Knystautas R. Interaction of a vortex
ring and a laminar flame. In: 22nd Symposium (International) on Combustion, Seattle, WA, 1988. The Combustion
Institute, Pittsburgh. p. 50514.
Roberts WL, Driscoll JF. Laminar vortex interacting with a
premixed flame: measured formation of pockets of reactants.
Combust Flame 1991;87:24556.
Roberts WL, Driscoll JF, Drake MC, Ratcliffe JW. OH fluorescence images of the quenching of a premixed flame during
an interaction with a vortex. In: 24th Symposium (International) on Combustion, Sydney, Australia, 1992. The
Combustion Institute, Pittsburgh. p. 16976.
Roberts WL, Driscoll JF, Drake MC, Goss LP. Images of the
quenching of a flame by a vortex-to quantify regimes of
turbulent combustion. Combust Flame 1993;94:5869.
Driscoll JF, Sutkus DJ, Roberts WL, Post ME, Goss LP. The
strain exerted by a vortex on a flame-determined from velocity field images. Combust Sci Tech 1994;96:21329.
Mueller CJ, Driscoll JF, Sutkus DJ, Roberts WL, Drake MC,
Smooke MD. Effect of unsteady stretch rate on OH chemistry during a flamevortex interaction: to assess flamelet
models. Combust Flame 1995;100:32331.
Mueller CJ, Driscoll CJ, Reuss DL, Drake MC. Effects of
unsteady stretch on the strength of a freely propagating flame
wrinkled by a vortex. In: 26th Symposium (International) on
Combustion, Napoli, Italy, 1996. The Combustion Institute,
Pittsburgh. p. 34755.
Mueller CJ, Driscoll JF, Reuss DL, Drake MC, Rosalik ME.
Vorticity generation and attenuation as vortices convect
through a premixed flame. Combust Flame 1998;112:342
58.
Sinibaldi JO, Driscoll JF, Mueller CJ, Tulkki AE. Flame
vortex interactions: effects of buoyancy from microgravity
imaging studies. In: AIAA 35th Aerospace Sciences Meeting
and Exhibit, Reno, NV, 1997. AIAA 97-0669.

279

[103] Sinibaldi JO, Mueller CJ, Tulkki AE, Driscoll JF. Suppression of flame wrinkling by buoyancy: the baroclinic stabilization mechanism. AIAA J 1998;36(8):14328.
[104] Sinibaldi JO, Mueller CJ, Driscoll JF. Local flame propagation speeds along wrinkled, unsteady, stretched premixed
flames. In: 27th Symposium (International) on Combustion,
Boulder, CO, 1998. The Combustion Institute, Pittsburgh. p.
82732.
[105] Samaniego J-M. Generation of two-dimensional vortices in a
cross-flow. In: CTR annual research briefs. Stanford University, Stanford, CA, 1992. p. 43141.
[106] Samaniego J-M. Stretch-induced quenching in flamevortex
interactions. In: CTR annual research briefs. Stanford
University, Stanford, CA, 1993. p. 20518.
[107] Nguyen Q-V, Paul PH. The time evolution of a vortexflame
interaction observed via planar imaging of CH and OH. In:
26th Symposium (International) on Combustion, Napoli,
Italy, 1996. The Combustion Institute, Pittsburgh. p. 35764.
[108] Rolon JC, Aguerre F, Candel S. Experiments on the interaction between a vortex and a strained diffusion flame.
Combust Flame 1995;100:4229.
[109] Thevenin D, Rolon JC, Renard P-H, Kendrick DW,
Veynante D, Candel S. Structure of a non-premixed flame
interacting with counterrotating vortices. In: 26th Symposium (International) on Combustion, Napoli, Italy, 28
July2 August 1996. The Combustion Institute, Pittsburgh.
p. 107986.
[110] Renard P-H, Rolon JC, Thevenin D, Candel S. Investigations
of heat release, extinction, and time evolution of the flame
surface, for a non-premixed flame interacting with a vortex.
Combust Flame 1999;117:189205.
[111] Renard P-H, Rolon JC, Thevenin D, Candel S. Wrinkling,
pocket formation and double premixed flame interaction
processes. In: 27th Symposium (International) on Combustion, Boulder, CO, 37 August 1998. The Combustion Institute, Pittsburgh. p. 65966.
[112] Karagozian AR, Suganuma Y, Strom BD. Experimental
studies in vortex pair motion coincident with a liquid reaction. Phys Fluids 1988;31:186271.
[113] Chen L-D, Roquemore WM. Visualization of jet flames.
Combust Flame 1986;66:816.
[114] Chen L-D, Seaba JP, Roquemore WM, Goss LP. Buoyant
diffusion flames. In: 22nd Symposium (International) on
Combustion, Seattle, WA, 1988. The Combustion Institute,
Pittsburgh. p. 67784.
[115] Chen L-D, Roquemore WM, Goss LP, Vilimpoc V. Vorticity
generation in jet diffusion flames. Combust Sci Tech
1991;77:4157.
[116] Strawa AW, Cantwell BJ. Visualization of the structure of a
pulsed methaneair diffusion flame. Phys Fluids
1985;28(8):231720.
[117] Gutmark EJ, Parr TP, Hanson-Parr DM, Schadow KC.
Vortex dynamics in diffusion flames. In: Sixth Symposium
on Turbulent Shear Flows, Toulouse, France, 1987.
[118] Hsu KY, Chen L-D, Katta VR, Goss LP, Roquemore WM.
Experimental and numerical investigations of the vortex
flame interactions in a driven jet diffusion flame. In: AIAA
31st Aerospace Sciences Meeting and Exhibit, Reno, NV,
1993. AIAA 93-0455.
[119] Hancock RD, Schauer FR, Lucht RP, Katta VR, Hsu KY.
Thermal diffusion effects and vortexflames interactions in

280

[120]

[121]

[122]

[123]

[124]

[125]

[126]

[127]

[128]

[129]

[130]

[131]

[132]
[133]

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282
hydrogen jet diffusion flames. In Twenty-Sixth Symposium
(International) on Combustion, Napoli, Italy, 1996. The
Combustion Institute, Pittsburgh. p. 108793.
Mueller CJ, Schefer RW. Coupling of diffusion flame structure to an unsteady vortical flowfield. In Twenty-Seventh
Symposium (International) on Combustion, Boulder, CO,
1998. The Combustion Institute, Pittsburgh. p. 110512.
Park J, Shin HD. Experimental investigation of the developing process of an unsteady diffusion flame. Combust Flame
1997;110:6777.
Chen SJ, Dahm WJA. Diffusion flame structure of a laminar
vortex ring under microgravity conditions. In: 27th Symposium (International) on Combustion, Boulder, CO, 1998. The
Combustion Institute, Pittsburgh. p. 257986.
Thevenin D, Renard P-H, Rolon JC, Candel S. Extinction
processes during a non-premixed flame/vortex interaction.
In: 27th Symposium (International) on Combustion, Boulder,
CO, 37 August 1998. The Combustion Institute, Pittsburgh.
p. 71926.
Katta VR, Carter CD, Fiechtner GJ, Roquemore WM, Gord
JR, Rolon JC. Interaction of a vortex with a flat flame formed
between opposing jets of hydrogen and air. In: 27th Symposium (International) on Combustion, Boulder, CO, 1998. The
Combustion Institute, Pittsburgh. p. 58794.
Gord JR, Donbar JM, Fiechtner GJ, Carter CD, Katta VR,
Rolon JC. Experimental and computational visualization of
vortex-flame interactions in an opposed-jet burner. In: International Conference on Optical Technology and Image
Processing in Fluid, Thermal and Combustion Flow,
Yokohama, Japan, 1998. Visualization Society of Japan
and SPIE.
Fiechtner GJ, Carter CD, Katta VR, Gord JR, Donbar JM,
Rolon JC. Regimes of interaction between a non-premixed
hydrogen/air and an isolated vortex. In: AIAA 37th Aerospace Sciences Meeting and Exhibit, Reno, NV, 1999.
Gord JR, Fiechtner GJ, Renard P-H, Carter CD, Rolon JC.
Characterizing the interaction of a vortex with a laminar
opposed-jet flame. In: 33rd National ASME, AIChE, ANS,
and AIAA Heat Transfer Conference, Albuquerque, NM,
1517 August, 1999.
Takagi T, Yoshikawa Y, Yoshida K, Komiyama M,
Kinoshita S. Studies on strained non-premixed flames
affected by flame curvature and preferential diffusion. In:
26th Symposium (International) on Combustion, Napoli,
Italy, 1996. The Combustion Institute, Pittsburgh. p. 1103
10.
Marble FE. Mixing, diffusion and chemical reactions of
liquids in a vortex field. In: Moreau A, Turq P, editors.
Chemical Reactivity in Liquids, New York: Plenum Press,
1988. p. 58196.
Laverdant A, Candel S. Computation of diffusion and
premixed flames rolled up in vortex structures. J Propu
Power 1989;5(2):13443.
Oran ES, Boris JP, Kailasanath K, Grinstein FF. Numerical
simulation of unsteady mixing layers. In: 10th International
Colloquium on the Dynamics of Explosions and Reactive
Systems, Berkeley, CA, 1985.
Laverdant A, Candel S. Interaction of diffusion and premixed
flames with a vortex. Rech Aerosp 1988;3:1328.
Southerland KB, Porter III JR, Dahm WJA, Buch KA. An
experimental study of the molecular mixing process in an

[134]

[135]

[136]

[137]

[138]

[139]

[140]

[141]

[142]

[143]

[144]

[145]

[146]

[147]

[148]

[149]

[150]

axisymmetric laminar-vortex ring. Phys Fluids A 1991;


3(5):138592.
Ashurst WT. Vorticity generation in a non-premixed flame
sheet. In: Dervieux A, Larrouturou B, editors. Numerical
combustion, Lecture notes in physics, vol. 351. Berlin:
Springer, 1989. p. 3.
Ashurst WT, Williams FA. Vortex modification of diffusion
flamelets. In: 23rd Symposium (International) on Combustion, Orleans, France, 1990. The Combustion Institute, Pittsburgh. p. 54350.
Marble FE, Broadwell JE. The coherent flame model of
nonpremixed turbulent combustion. Project Squid TRW-9PU, TRW, 1977.
Thevenin D, Renard P-H, Rolon JC, Candel S. Dynamics of
the flame front during a non-premixed flame/vortex interaction. In: Seventh International Conference on Numerical
Combustion, York, UK, 30 March1 April 1998. SIAM,
Philadelphia, PA. p. 96.
McMurtry PA, Jou W-H, Riley JJ, Metcalfe RW. Direct
numerical simulations of a reacting mixing layer with chemical heat release. AIAA J 1986;24(6):96270.
McMurtry PA, Riley JJ, Metcalfe RW. Effects of heat release
on the large-scale structure in turbulent mixing layers. J Fluid
Mech 1989;199:297332.
Mahalingam S, Cantwell BJ, Ferziger JH. Effects of heat
release on the structure and stability of a coflowing chemically reacting jet. In: AIAA 27th Aerospace Sciences Meeting, Reno, NV, 1989. AIAA 89-0661.
Grinstein FF, Kailasanath K. Chemical energy release,
spanwise excitations, and dynamics of transitional, reactive,
free shear-flows. In: AIAA 29th Aerospace Sciences Meeting, Reno, NV 1991. AIAA 91-0247.
Davis RW, Moore EF, Roquemore WM, Chen L-D, Vilimpoc V, Goss LP. Preliminary results of a numericalexperimental study of the dynamic structure of a buoyant jet
diffusion flame. Combust Flame 1991;83:26370.
Davis RW, Moore EF, Chen L-D, Roquemore WM, Vilimpoc V, Goss LP. Numerical/experimental study of the
dynamic structure of a buoyant jet diffusion flame. Theoret
Comput Fluid Dyn 1994;6:11323.
Katta VR, Roquemore WM. Role of inner and outer structures in transitional jet diffusion flames. Combust Flame
1993;92:27482.
Katta VR, Goss LP, Roquemore WM. Numerical investigations of transitional H2/N2 jet diffusion flames. AIAA J
1994;32(1):8494.
Katta VR, Goss LP, Roquemore WM. Simulation of vortical
structures in a jet diffusion flame. Int J Num Meth Heat Fluid
Flow 1994;4:41324.
Riley JJ, Metcalfe RW, Orszag SA. Direct Numerical simulations of chemically reacting turbulent mixing layers. Phys
Fluids 1986;29(2):40622.
Laskey KJ, Ellzey JL, Oran ES. Numerical study of an
unsteady diffusion flame. In: AIAA 27th Aerospace Sciences
Meeting, Reno, NV, 1989. AIAA 89-0572.
Lee C, Metcalfe RW, Hussain F. Large scale structures in
reacting mixing layers. In: Seventh Symposium on Turbulent
Shear Flows, Stanford, CA, 1991. p. 33143.
Grinstein FF, Gutmark EJ, Parr TP, Hanson-Parr DM.
Streamwise and spanwise vortex interaction in a circular
reacting jetan experimental and computational study.

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

[151]
[152]
[153]

[154]

[155]

[156]
[157]

[158]

[159]

[160]

[161]

[162]

[163]

[164]

[165]

[166]

[167]

[168]

[169]

Tenth Symposium on Turbulent Shear Flows 1995;2:


1318.
Hewett JS, Madnia CK. Flamevortex interaction in a
reacting vortex ring. Phys Fluids 1998;10(1):189205.
Ghoniem AF, Givi P. Lagrangian simulation of a reacting
mixing layer at low heat release. AIAA J 1988;26(6):6907.
Givi P, Jou W-H. Direct numerical simulations of a twodimensional reacting, spatially developing mixing layer by
a spectral-element method. In: 22nd Symposium (International) on Combustion, Seattle, WA, 1988. p. 63543.
Petrov C, Ghoniem AF. Numerical simulation of reacting
flow with multi-step kinetics. In: AIAA 35th Aerospace
Sciences Meeting and Exhibit, 1997. AIAA 97-0291.
Macaraeg MG, Jackson TL, Hussaini MY. Ignition and
structure of a laminar diffusion flame in the field of a vortex.
Combust Sci Tech 1992;87:36387.
Thevenin D, Candel S. Ignition dynamics of a diffusion flame
rolled-up in a vortex. Phys fluids 1995;7(2):43445.
Macaraeg MG, Jackson TL, Hussaini MY. Ignition dynamics
of a laminar diffusion flame in the field of a vortex embedded
in a shear flow. Combust Sci Tech 1994;102:23153.
Takahashi F, Katta VR. Numerical experiments on the
flamevortex interactions in a jet diffusion flame. J Propu
Power 1995;11(1):1707.
Lee JC, Frouzakis CE, Boulouchos K. Numerical study of a
opposed jet H2/air diffusion flame/vortex interactions. In:
17th International Colloquium on the Dynamics of Explosions and Reactive Systems, Heidelberg, Germany, 2530
July 1999.
Katta VR, Goss LP, Roquemore WM. Effect of non-unity
Lewis number and finite rate chemistry on the dynamics of a
hydrogen-air jet diffusion flame. Combust Flame 1994;
96:6074.
Katta VR, Roquemore WM. On the structure of a stretched or
compressed laminar flameletinfluence of preferential
diffusion. Combust Flame 1995;100:6170.
Takahashi F, Katta VR. Unsteady extinction mechanisms of
diffusion flames. In: 26th Symposium (International) on
Combustion, Napoli, Italy, 1996. The Combustion Institute,
Pittsburgh. p. 115160.
Katta VR, Hsu KY, Roquemore WM. Local extinction in an
unsteady methaneair jet diffusion flame. In: 27th Symposium (International) on Combustion, Boulder, CO, 1998. The
Combustion Institute, Pittsburgh. p. 11219.
Katta VR, Roquemore WM. Simulation of dynamic methane
jet diffusion flames using finite rate chemistry models. AIAA
J 1998;36(11):204454.
You YH, Lee DK, Shin HD. Visual investigation of a vortex
ring interacting with a nonpremixed flame. Combust Sci
Tech 1998;139:36583.
Underwood DS, Waitz IA. Effect of heat release on streamwise vorticity enhanced mixing. J Propu Power
1996;12(4):63845.
Linan A. The asymptotic structure of counterflow diffusion
flames for large activation energies. Acta Astron
1974;1:100739.
Cuenot B, Poinsot T. Asymptotic and numerical study of
diffusion flames with variable Lewis number and finite rate
chemistry. Combust Flame 1996;104:11137.
Takahashi F, Schmoll WJ, Trump DD, Goss LP. Vortex
flame interactions and extinction in turbulent jet diffusion

[170]

[171]

[172]

[173]

[174]
[175]

[176]

[177]

[178]
[179]
[180]

[181]
[182]

[183]

[184]

[185]

[186]

[187]
[188]

281

flames. In: 26th Symposium (International) on Combustion,


Napoli, Italy, 1996. The Combustion Institute, Pittsburgh.
p. 145152.
Santoro VS, Kyritsis DC, Gomez A. Extinction behaviour of
either gaseous or spray counterflow diffusion flames interacting with a laminar toroidal vortex. In: 17th International
Colloquium on the Dynamics of Explosions and Reactive
Systems, Heidelberg, Germany, 2530 July 1999.
Manda BVS, Karagozian AR. Effect of heat release on diffusion flamevortex pair interactions. Combust Sci Tech 1988;
61:10119.
Peters N, Williams FA. Premixed combustion in a vortex. In:
22nd Symposium (International) on Combustion, Seattle,
WA, 1988. The Combustion Institute, Pittsburgh. p. 495
503.
Rutland CJ, Ferziger JH. Interaction of a vortex and a
premixed flame. In: AIAA 27th Aerospace Sciences Meeting, Reno, NV, 1989. AIAA 89-0127.
Rutland CJ, Ferziger JH. Simulations of flamevortex interactions. Combust Flame 1991;84:34360.
Poinsot T, Veynante D, Candel S. Diagrams of premixed
turbulent combustion based on direct simulation. In: 23rd
Symposium (International) on Combustion, Orleans, France,
1990. The Combustion Institute, Pittsburgh. p. 6139.
Poinsot T, Veynante D, Candel S. Quenching processes and
premixed turbulent combustion diagrams. J Fluid Mech
1991;228:561606.
Wu M-S, Driscoll JF. Numerical simulation of a vortex
convected through a premixed laminar flame. Combust
Flame 1992;91:31022.
Ashurst WT. Flame propagation through swirling eddies, a
recursive pattern. Combust Sci Tech 1993;92:87103.
Dold JW, Kerr OS, Nikolova IP. Flame propagation through
periodic vortices. Combust Flame 1995;100:35966.
Kerr OS, Dold JW. Flame propagation around stretched periodic vortices investigated using ray-tracing. Combust Sci
Tech 1996;118:10125.
Vassilicos JC, Nikiforakis N. Flameletvortex interaction
and the Gibson scale. Combust Flame 1997;109:293302.
Ashurst WT, McMurtry PA. Flame generation of vorticity:
vortex dipoles from monopoles. Combust Sci Tech 1989;
66:1737.
Katta VR, Roquemore WM. Vortexflame interactions in
premixed jet flames. In: AIAA 33rd Aerospace Sciences
Meeting, Reno, NV, 1995. AIAA 95-0871.
Louch DS, Bray KNC. Vorticity and scalar transport in
premixed turbulent combustion. In: 27th Symposium (International) on Combustion, Boulder, CO, 1998. The Combustion Institute, Pittsburgh. p. 80110.
Patnaik G, Kailasanath K. A computational study of local
quenching in flamevortex interactions with radiative losses.
In: 27th Symposium (International) on Combustion, Boulder,
CO, 1998. The Combustion Institute, Pittsburgh. p. 7117.
Lee T-W, Santavicca DA. Flame front geometry and stretch
during interactions of premixed flames with vortices.
Combust Sci Tech 1993;90:21129.
Lee T-W. Scaling of vortex-induced flame stretch profiles.
Combust Sci Tech 1994;102:3017.
Helenbrook BT, Sung CJ, Law CK, Ashurst WT. On
stretched-affected flame propagation in vortical flows.
Combust Flame 1996;104:4608.

282

P.-H. Renard et al. / Progress in Energy and Combustion Science 26 (2000) 225282

[189] Hilka M, Veynante D, Baum M, Poinsot T. Simulation of


flame-vortex interactions using detailed and reduced chemical kinetics. 10th Symposium on Turbulent Shear Flows
1995;2:1924.
[190] Katta VR, Roquemore WM. Numerical studies on the structure of two-dimensional H2/air premixed jet flame. Combust
Flame 1995;102:2140.
[191] Najm HN, Wyckoff PS. Premixed flame response to unsteady
strain-rate and curvature. In: Fall Meeting of the Western
States Section of the Combustion Institute, 1996.
[192] Najm HN, Paul PH, Mueller CJ, Wyckoff PS. On the
adequacy of certain experimental observables as measurements of flame burning rate. In: Spring Meeting of the
Western States Section of the Combustion Institute, 1997.
[193] Ishizuka S, Hamasaki T, Koumura K, Hasegawa R. Measurements of flame speeds in combustible vortex rings: validity of
the back-pressure drive flame propagation mechanism. In:
27th Symposium (International) on Combustion, Boulder,
CO, 1998. The Combustion Institute, Pittsburgh. p. 72734.
[194] Echekki T, Mungal MG. Flame speed measurements at the
tip of a slot burner: effects of flame curvature and hydrodynamic stretch. In: 23rd Symposium (International) on
Combustion, Orleans, France, 1990. The Combustion
Institute, Pittsburgh. p. 45561.
[195] Najm HN, Wyckoff PS. Premixed flame response to unsteady
strain-rate and curvature. Combust Flame 1997;110:92112.
[196] Lee T-W, Lee JG, Nye DA, Santavicca DA. Local response
and surface properties of premixed flames during interactions
with Karman vortex streets. Combust Flame 1993;94:146
60.
[197] Paul PH, Najm HN. Planar laser-induced fluorescence
imaging of flame heat release rate. In: 27th Symposium
(International) on Combustion, Boulder, CO, 1998. The
Combustion Institute, Pittsburgh. p. 4350.
[198] Barre`re M. Mode`les de combustion. Revue Generale de
Thermique 1974;148:295308.
[199] Bray KNC. Turbulent flows with premixed reactants. In:
Libby PA, Williams FA, editors. Topics in applied physics,
vol. 44. New York: Springer, 1980. p. 115.
[200] Borghi R. On the structure and morphology of turbulent
premixed flames. In: Casci C, Bruno C, editors. Recent

[201]

[202]

[203]
[204]
[205]
[206]
[207]
[208]
[209]
[210]

[211]

[212]

[213]

[214]

advances in the aerospace sciences, New York: Plenum


Press, 1985. p. 11738.
Peters N. Length scales in laminar and turbulent flames. In:
Oran ES, Boris JP, editors. Numerical approaches to
combustion modeling, Progress in astronautics and aeronautics, volume 135. Washington, DC: AIAA, 1991. p. 15582.
Wright FH, Zukoski EE. Flame spreading from bluff-body
flame holders. In: Eighth Symposium (International) on
Combustion, Pasadena, CA, 1962. The Combustion Institute,
Pittsburgh. p. 93343.
Sivashinsky GI. On self-turbulization of a laminar flame.
Acta Astron 1979;6:56991.
Hinze JO. Turbulence. 2nd ed. New York: McGraw-Hill,
1975.
Tennekes H, Lumley JL. A first course in turbulence.
Cambridge, MA: MIT Press, 1972.
Pope SB. Turbulent premixed flames. Annu Rev Fluid Mech
1987;19:23770.
Klimov AM. Laminar flame in a turbulent flow. Zhur Prikl
Mekh I Tekhn Fiz 1963;3:4958.
Borghi R, Destriau M. Combustion and flames. Paris, France:
Technip, 1998.
Peters N. The turbulent burning velocity for large scale and
small scale turbulence. J Fluid Mech 1999;384:10732.
Vervisch L, Poinsot T. Direct numerical simulation of nonpremixed turbulent flames. Annu Rev Fluid Mech 1998;30:
65591.
Libby PA, Williams FA. Fundamental aspects and a review.
In: Libby PA, Williams FA, editors. Turbulent reactive
flows, London, UK: Academic Press, 1994. p. 261.
Bray KNC, Peters N. Laminar flamelets in turbulent flames.
In: Libby PA, Williams FA, editors. Turbulent reactive
flows, London, UK: Academic Press, 1994. p. 63113.
Vervisch L, Veynante D. Turbulent combustion. In: von
Karman Institute Lecture Series, Rhode-Saint-Gene`se,
Belgium, 1999. von Karman Institute.
Cuenot B, Poinsot T. Effects of curvature and unsteadiness in
diffusion flames. Implications for turbulent diffusion
combustion. In: 25th Symposium (International) on Combustion, Irvine, CA, 1994. The Combustion Institute, Pittsburgh.
p. 138390.

You might also like