Combustion and Flame 148 (2007) 100–116

www.elsevier.com/locate/combustflame

Extinction of counterflow diffusion flames with radiative

heat loss and nonunity Lewis numbers
H.Y. Wang a , W.H. Chen b , C.K. Law a,∗
a Department of Mechanical and Aerospace Engineering, Princeton University, Princeton, NJ 08544-5263, USA
b Department of Marine Engineering, National Taiwan Ocean University, Keelung, Taiwan, Republic of China

Received 12 July 2006; received in revised form 30 October 2006; accepted 31 October 2006
Available online 12 December 2006

Abstract
The structure and extinction characteristics of counterflow diffusion flames with flame radiation and nonunity
Lewis numbers of the fuel and oxidant are examined using multiscale asymptotic theory, and a model expressed in
terms of the jump relations and reactant leakages with the proper consideration of the excess enthalpy overlooked
in previous analyses is developed. The existence of the dual extinction limits in the presence of radiative heat loss,
namely the kinetic limit at small Damköhler number (high stretch rate) and the radiative limit at large Damköhler
number (low stretch rate), are identified. It is found that the former is minimally affected by radiative loss, while
a substantial amount of heat loss is associated with the radiative limit. Reactant leakage, however, is the root
cause for both limits. The influence of radiative loss on the extinction Damköhler numbers is found to be through
its effects on the flame temperature, the excess enthalpy, and the reduced extinction Damköhler number. At both
extinction limits, the contribution from the flame temperature is always important and dominant. The contributions
from the other two, however, could be important in some special cases. At small LeF , the contribution from the
reduced extinction Damköhler number is large and even dominant under small radiative loss. The contribution
from the excess enthalpy is important for small LeO and it may be comparable to the contribution from the flame
temperature when radiative loss is small. Thus, overlooking the excess enthalpy in previous analyses may have
resulted in rather large error in the predicted extinction Damköhler numbers, especially the kinetic one.
© 2006 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

Keywords: Diffusion flame; Excess enthalpy; Extinction; Asymptotics

1. Introduction It is shown that the flame structure and response are

The structure and extinction characteristics of dif-
fusion flames in the absence of volumetric heat loss
were first rigorously analyzed using large-activation-
energy asymptotics in a seminal paper by Liñán [1].
* Corresponding author. Fax: +1 609 258 6233.
E-mail address: cklaw@princeton.edu (C.K. Law).
controlled by the system Damköhler number, Da, de-
fined as the ratio of the residence time to the reaction
time, and that there exists a minimum Da, and conse-
quently a maximum flow rate, beyond which steady
burning is not possible due to insufficient time for
adequate reaction to occur to maintain the flame struc-
ture. This extinction mechanism is intrinsic to the
structure of diffusion flames, and hence is operative
under all situations.

0010-2180/$ – see front matter © 2006 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
doi:10.1016/j.combustflame.2006.10.005
 H.Y. Wang et al. / Combustion and Flame 148 (2007) 100–116 101

Nomenclature

BK Pre-exponential factor for the reaction Δ Reduced Damköhler number

rate Δc Minimum value of the reduced Damköh-
cp Specific heat ler number
Da Damköhler number ε Reciprocal of the Zel’dovich number
DaC Collision Damköhler number η Stretched spatial coordinate in the reac-
DaE Extinction Damköhler number tion zone
hi Excess/deficiency enthalpy for species i, κ Planck’s mean absorption coefficient
i = F, O ν Stoichiometric mass ratio of oxidant to
htotal Total excess/deficiency enthalpy in the fuel
reaction zone νi Stoichiometric coefficient of species i,
k Stretch rate of flow i = F, O
Lei Lewis number of species i, i = F, O σ Stefan–Boltzmann constant
q Heat of combustion per unit mass of fuel ω Chemical reaction rate
supplied
ζ Stretched spatial coordinate in the radia-
qR Rate of radiative heat loss per unit vol-
tion zone
ume
Ra Radiative heat loss parameter Overbars, subscripts, and superscripts
Si Leakage function of species i, i = F, O ∼ Unscaled quantities for T and Yi ,
xf Flame sheet location i =F, O
T Temperature ∗ Value for reference state
Tf Flame temperature
−∞ Fuel boundary
Ta Activation temperature
∞ Oxidant boundary
Wi Molecular weight of species i, i = F, O
+ Oxidant side of flame sheet
Yi Mass fraction of species i, i = F, O
− Fuel side of flame sheet
Greek symbols ad Adiabatic state
Γ Relative strength of the radiative loss to ext Extinction condition
the chemical heat release F Fuel
Γc Maximum radiative strength a system K Extinction condition at the kinetic limit
can sustain O Oxidant
γ Heat transfer parameter R Extinction condition at the radiative
δ Order of radiation zone thickness limit

Extinction can be further promoted in the presence
of radiative heat loss [2–4]. Specifically, Sohrab et al.
[4] analyzed the structure and extinction of counter-
flow diffusion flames with flame radiation, recogniz-
ing that radiation is a temperature-sensitive process,
albeit less sensitive than chemical reaction. Conse-
quently, radiative loss is operative within a thin, O(δ),
zone that sandwiches the O(ε) reaction zone, but
is nevertheless embedded within the much thicker
O(1) outer diffusive–convective zone, where ε is the
reciprocal of the Zel’dovich number, and δ satisfies
ε  δ  1. Multiscale asymptotics using δ and ε as
small parameters showed that the radiative contribu-
tion to extinction is through the reduction of the flame
temperature. However, an important qualitative ex-
tinction behavior was not recognized. That is, because
the extent of radiative loss increases with increas-
ing flame volume and thereby flame thickness, and
because flame thickness increases with decreasing
stretch rate and hence increasing system Da, extinc-
tion is expected to be facilitated with increasing Da.
Consequently, in addition to the extinction limit at the
minimum Da, there should exist another extinction
limit at a maximum Da, above which steady burning
is also not possible.
The possible existence of the dual extinction lim-
its at lower and higher Damköhler numbers, hereafter
respectively referred to as the kinetic and radiative ex-
tinction limits, was first suggested and numerically
demonstrated by T’ien [5] for the counterflow dif-
fusion flame in the stagnation region of a condensed
fuel with surface radiation. Theoretically, Chao et al.
[6] first successfully demonstrated, via the multiscale
asymptotic theory of Sohrab et al. [4], the existence
of dual extinction states for droplet combustion with
flame radiation. The analysis further showed that ex-
tinction for both limits is governed by Liñán’s ex-
tinction criterion as a consequence of excessive re-
actant leakage. Subsequently, this multiscale asymp-
totics was employed by Oh et al. [7] for the diffusion
102 H.Y. Wang et al. / Combustion and Flame 148 (2007) 100–116
flame stabilized on a condensed fuel with both flame
and surface radiation and by Liu et al. [8] for the
counterflow diffusion flame with flame radiation and
small deviation of the Lewis numbers of fuel and ox-
idant from unity. Specifically, Liu et al. [8] showed
that larger Lewis numbers lead to a smaller flammable
range of Damköhler number and smaller Lewis num-
bers increase this flammable range. Experimentally,
the existence of the dual extinction limits was ob-
served by Maruta et al. [9] in a counterflow diffusion
flame of methane and air with flame radiation under
microgravity.
The theoretical analyses mentioned above, how-
ever, did not consider the excess/deficiency of the
total enthalpy in the reaction zone, hereafter referred
to as excess enthalpy. Its importance was recognized
by Kim and Williams [10] in their study of coun-
terflow diffusion flame with nonunity Lewis number,
Le. The excess enthalpy arises from the nonconserva-
tive nature of the total enthalpy due to the imbalance
of thermal and mass diffusion as the reactants leak
through the flame as a result of finite-rate chemistry.
It was shown that, although the amount of the ex-
cess enthalpy is small, typically of O(ε), it can lead
to O(1) changes in the reaction rate. Furthermore,
without considering the excess enthalpy, the depen-
dence of Da on the reduced Damköhler number, Δ, is
linear [1], so that the minimum of Da directly cor-
responds to the minimum of Δ, Δc . The physical
extinction limit can be consequently determined from
the reaction-sheet solution using Liñán’s formula [1].
However, with the consideration of excess enthalpy,
this dependence becomes nonlinear, so that Δc does
not necessarily correspond to the minimum of Da,
and as such does not correctly identify the extinction
condition. Recognizing that in addition to nonequid-
iffusion, which is the cause of excess enthalpy in the
analysis of Kim and Williams [10], radiative loss from
flames is the other major source of enthalpy loss in
flames, overlooking the excess enthalpy is expected to
produce O(1) error in the radiative extinction limit, at
which radiative loss is significant. Furthermore, pre-
vious analyses [4,6–8] did not consider adequately
the effect of nonequidiffusion. Since extinction can
be considered to be primarily attributable to exces-
sive heat loss from the reaction zone, which could be
achieved through both conduction and radiation, it is
necessary to consider the effects of nonunity Le and
radiative loss at the same time. Although Liu et al.
[8] considered nonunity Le with O(ε) deviation from
unity, such a small deviation cannot explore all the Le
dependency of the dual extinction limits, as will be
shown later.
In view of the above considerations, one of the pri-
mary objectives of the current study is to perform a
rigorous multiscale asymptotic analysis and develop
a model for distinct and general Lewis numbers of
fuel and oxidant, LeF and LeO , and with proper con-
sideration of the excess enthalpy. This formulation is
then applied to study the dual extinction limits of the
radiative counterflow diffusion flames with nonunity
Lewis numbers. It should be mentioned that Mills
and Matalon [11] studied the extinction of burner-
generated radiative spherical diffusion flames with
nonunity LeF and LeO . However, the radiative loss
was assumed to be of O(ε) quantity and occurs within
a zone of O(1) thickness between the burner surface
and flame, whose location and thickness are arbitrary
given. Thus, this analysis is based on different as-
sumptions from the current study, in which radiative
loss occurs within the radiation zone of O(δ) thick-
ness determined from the temperature sensitivity of
radiation, and hence has a different scope of applica-
tion. Furthermore, Mills and Matalon [11] found that
the Lewis numbers of fuel and oxidant have a signifi-
cant effect on the extinction limits, and steady burning
is not possible when they are sufficiently large. This
substantiates the need to conduct a rigorous extinc-
tion analysis for Lewis numbers sufficiently different
from unity.
We further note that since the formulation devel-
oped in the current study is of a general nature, it
can be applied to other phenomena affected by simul-
taneous radiative loss and mixture nonequidiffusion,
such as the thermal–diffusive instability [12–18] of
radiation-affected diffusion flames. This possible ex-
tension will be discussed subsequently.
2. Formulation
2.1. Governing equations
Fig. 1 shows the counterflow configuration consid-
ered in this study, with the fuel and oxidant streams
approaching from the left- (−∞) and right-hand (∞)
sides, respectively. The chemical reaction is assumed
to be described by a one-step irreversible reaction of
the form
νF F + νO O → products,
Fig. 1. Schematic of a counterflow diffusion flame.
 H.Y. Wang et al. / Combustion and Flame 148 (2007) 100–116 103
 
T − Tf
where F and O represent the fuel and oxidant, respec- qR = qRf exp , (5)
tively, and νi (i = F, O) the corresponding stoichio- δTf
metric coefficients. Further assuming constant physi- where qRf = 4σ κq 3 Tf4 /(ρcp4 k) and δ = d(ln T )/
cal and chemical properties of the reactants, constant d(ln qR ). With the assumptions of constant density
density, and an optically thin approximation for the and properties made in this study, δ = 0.25.
radiative loss, the appropriate nondimensional gov-
erning equations can be written as 2.2. Asymptotic expansions

∂T ∂T ∂ 2T Following Chao et al. [6], the temperatures T

−x − = ω − qR , (1)
∂t ∂x ∂x 2 in the diffusive–convective, radiation, and reaction
∂YF ∂YF ∂ 2 YF zones are respectively expanded in terms of ε and δ as
−x − Le−1 = −ω, (2) ±
∼ T0± (x, t) + εT1± (x, t)
∂t ∂x F ∂x 2 Tout
∂ 2 YO  
∂YO ∂Y
− x O − Le−1 + δ T2± (x, t) + εT3± (x, t) + O(δ 2 ),
O ∂x 2 = −ω, (3) (6)
∂t ∂x  
TR± ∼ Tf + εΘ1± (ζ ) + δ Θ2± (ζ ) + εΘ3± (ζ )
with the boundary conditions
+ O(δ 2 ), (7)
T = T−∞ , YF = YF,−∞ , YO = 0,
Tin ∼ Tf + ετ1 (η) + ε2 τ2 (η) + O(ε3 ), (8)
as x → −∞,
where ζ = (x − xf )/δ and η = (x − xf )/ε are the
T = T∞ , YF = 0, YO = YO,∞ ,
stretched spatial coordinates in the radiation zone and
as x → ∞, (4) reaction zone, respectively, and xf is the flame sheet
where T is the temperature, Yi the mass fraction of location. The superscripts “+” and “−” denote solu-
species i (i = F, O) and Lei the corresponding Lewis tions on the oxidant and fuel sides of the flame, re-
number, ω = DaC YF YO exp(−Ta /T ) the chemical re- spectively. The reason for expanding Ti only in terms
action rate, qR = 4σ κq 3 T 4 /(ρcp4 k) the rate of radia- of ε is that the reaction zone is very thin so that the
tive heat loss, DaC = BK ρYF,−∞ νO /(WF k) the colli- radiative loss from it can be assumed to be negligible.
Because YF and YO are not directly affected by
sion Damköhler number, and Ta , k, σ , κ, q, ρ, cp , BK ,
radiative loss, their analyses only need to be con-
Wi , respectively denote the activation temperature,
ducted in two zones, namely the reaction zone and
stretch rate of the flow, Stefan–Boltzmann constant
the diffusive–convective zone. Thus, they can be re-
(1.36 × 10−11 kcal/m2 s K4 ), Planck’s mean absorp-
spectively expanded as
tion coefficient, heat of combustion per unit mass of
± ± ± ±
fuel supplied, density, specific heat, pre-exponential Yi,out ∼ Yi,0 (x, t) + εYi,1 (x, t) + ε2 Yi,2 (x, t)
factor for the reaction rate, and molecular weight of
species i. The nondimensional T and Yi are defined as + O(ε3 ), (9)
Yi,in ∼ εyi,1 (η) + ε2 yi,2 (η) + O(ε3 ), (10)
T̃
T= , YF = ỸF , YO = ỸO /ν, where i = F, O. We note that unlike the analysis of
(q/cp )
[6], the outer expansions for YF and YO do not have
where ν = νO WO /νF WF is the stoichiometric mass the O(δ) terms. This is because these terms do not
ratio of the oxidant to the fuel and the over-tilde “∼” have the corresponding matching terms in the reac-
designates the unscaled T and Yi for differentiation tion zone and because of their homogeneous bound-
with their nondimensional counterparts. ary conditions at x → ±∞. Consequently, it can be
With the assumption of large activation energy, shown that they are identically zero throughout the
chemical reaction is confined within a thin zone of diffusive–convective zone.
O(ε) thickness, where ε = Tf2 /Ta and Tf is the flame The above expansions for T are subject to the
temperature. Due to the temperature-sensitive nature matching conditions between the radiation and reac-
of radiation, radiative loss is assumed to be confined tion zones,
⎫
within a radiation zone of O(δ) thickness, with the dΘ1± (0) ⎪
⎪
reaction zone embedded within it, as shown by the Θ1± (0) ∼ τ1 |η→±∞ , ∼ 0, ⎪
⎪
dζ ⎪
⎪
flame structure in Fig. 1. Further recognizing that ±  ⎪
⎪
dΘ (0) dτ  ⎬
the diffusive–convective zone external to the radiation ± 2 1
Θ2 (0) ∼ 0, ∼  ,
zone is of O(1) thickness, we seek a solution for the dζ dη η→±∞ ⎪ ⎪
 ⎪
⎪
flame structure satisfying the relation ε  δ < 1. Fol- ±  ⎪
⎪
± dΘ (0) dτ ⎪
lowing Sohrab et al. [4], the rate of radiative loss, qR , Θ3 (0) ∼ 0, 3
∼ 2
 ,⎪⎭
dζ dη η→±∞
is approximated by an Arrhenius-type function,
(11)
104 H.Y. Wang et al. / Combustion and Flame 148 (2007) 100–116

and the matching conditions between the diffusive– where Da = ε3 DaC exp(−Ta /Tf ) is the reaction
convective and radiation zones, Damköhler number. Integrating Eq. (16) twice and
⎫ applying the relevant matching conditions, (11) and
Tf ∼ T0± (xf ), ⎪
⎪ (13), for τ1 , yi,1 , dτ1 /dη, and dyi,1 /dη yield
⎪
⎪
± ±
Θ1 |ζ →±∞ ∼ T1 (xf ), ⎪
⎪
⎪
⎪
⎪
⎪ dΘ2+ (0)
±  ⎪ 1 1 +
dΘ1

⎪
⎪
⎪ τ1 + yF,1 = η + Θ1+ (0) + Y ,
 ∼ 0, ⎪
⎪ LeF dζ LeF F,1
dζ ζ →±∞ ⎪
⎪
⎪
⎪ (18)
± ⎪
⎬
Θ2 |ζ →±∞ ∼ −∞, 1 dΘ2− (0) 1 −
 (12) τ1 + yO,1 = η + Θ1− (0) +
∂T0± (xf ) ⎪
Y
dΘ2±  ⎪
⎪ LeO dζ LeO O,1
 ∼ ,⎪⎪
dζ ζ →±∞ ⎪
⎪ (19)
∂x ⎪
⎪
⎪
⎪
±
Θ3 |ζ →±∞ ∼ −∞, ⎪
⎪ and the jump relations
⎪
⎪
⎪
⎪
± 
dΘ3 ∂T1 (xf ) ⎪
± ⎪
⎪ dΘ2+ (0) dΘ2− (0)
 
 ∼ ⎪ ∂YF,0

dζ ζ →±∞ ∂x
. ⎭ − = −Le−1
F
dζ dζ ∂x
 
The expansions for YF and YO are subject only ∂YO,0
= −Le−1
O , (20)
to the matching conditions between the diffusive– ∂x
convective and reaction zones:
⎫ Θ1+ (0) − Θ1− (0) = −Le−1 −1
F [YF,1 ] = −LeO [YO,1 ].
±
0 ∼ Yi,0 (xf ), ⎪
⎪ (21)
⎪
⎪
±
∂Yi,0 (xf ) ⎪
⎪
± ⎪
⎪ Integrating Eq. (17) once and applying the relevant
yi,1 |η→±∞ ∼ η + Yi,1 (xf ), ⎪
⎪
∂x ⎪
⎪ matching conditions, (11) and (13), for dτ2 /dη and
 ⎬
± dyi,2 /dη yield the jump relation
dyi,1  ∂Y (x
i,0 f ) (13)
∼ , ⎪
⎪
dη η→±∞ ∂x ⎪
⎪
⎪

dΘ3+ (0)
 
dΘ3− (0)

⎪
 ⎪ xf Θ1+ (0) + − xf Θ1− (0) +
dyi,2 
±
∂ 2 Yi,0 (xf ) ±
∂Yi,1 (xf ) ⎪
⎪
⎪
⎪ dζ dζ
 ∼ 2
η+ .⎪
⎭  
dη η→±∞ ∂x ∂x ∂YF,1
= − xf YF,1 + Le−1
F ∂x
2.3. Jump relations and reactant leakages across the
reaction zone
We aim to derive the jump relations and the re-
actant leakages that serve as the inner boundary
conditions for the outer solutions in the diffusive–
convective zone.
The matching conditions for the O(1) outer solu-
tions in Eqs. (12) and (13) yield the jump relations
+ −
[T0 ] = [YF,0 ] = [YO,0 ] = YF,0 (0) = YO,0 (0) = 0,
(14)
where we have adopted the notation [T0 ] = T0+ (xf ) −
T0− (xf ). Additional jump relations and reactant leak-
ages are to be derived through the asymptotic analysis
of the reaction zone. Substituting the inner expansions
(8) and (10) and the stretched coordinate η into the
governing equations (1)–(3) yields, to O(ε),
d 2 τ1 /dη2 = −DayF,1 yO,1 eτ1 , (15)
d 2 (τ1 + yi,1 /Lei )/dη2 = 0, i = F, O, (16)
and to O(ε2 ),
d 2 (τ2 + yi,2 /Lei )/dη2 = 0, i = F, O, (17)
 
∂YO,1
= − xf YO,1 + Le−1
O . (22)
∂x
So far we have derived the jump relations (14) and
(20)–(22) across the reaction zone. However, for the
current multiscale analysis, the reaction zone is sand-
wiched into the radiation zone so that the temperature
terms in these jump relations are expressed by the ex-
pansion terms in the radiation zone. Since we aim to
derive the jump relations that are to be used as the
inner boundary conditions of the outer expansions,
the terms Θ1± (0), dΘ2± (0)/dζ , and dΘ3± (0)/dζ in
Eqs. (20)–(22) need to be replaced by the outer expan-
sion terms. This can be realized through integrating
the structure equation of the radiation zone and ap-
plying the matching conditions between the radiation
zone and the diffusive–convective zone. Substituting
Eq. (7) and the stretched coordinate ζ into the chemi-
cally frozen form of Eq. (1) yields
d 2 Θ2± /dζ 2 = Ra exp(Θ2± /Tf ), (23)
d 2 Θ1± /dζ 2 = RaΘ1± exp(Θ2± /Tf )/Tf , (24)
d 2 Θ3± /dζ 2 + xf dΘ1± /dζ = RaΘ3± exp(Θ2± /Tf )/Tf ,
(25)
 H.Y. Wang et al. / Combustion and Flame 148 (2007) 100–116 105

where Ra = δqRf . Integrating Eqs. (23)–(25) and ap- hF = A+ T1+ (xf ) + Le−1 +
F YF,1 (xf ),
plying the relevant matching conditions (see Appen-
dix A) yield hO = A− T1− (xf ) + Le−1 −
O YO,1 (xf ) (35)
are the excess/deficiency in the fuel and oxidant en-
dΘ2± (0) ∂T ± (xf )
= A± 0 , (26) thalpies, respectively, evaluated at the reaction sheet.
dζ ∂x This process is standard [1,17] and is not repeated
Θ1± (0) = A± T1± (xf ), (27) here.
  The leakage functions SF and SO are dependent
± ±
dΘ3 (0) 1 1 1 ∂T (xf ) only on Δ and γ . Liñán [1] has shown that for a given
∼ A± − ± T1± (xf ) + ± 1 ,
dζ 2 A A ∂x γ there exists a minimum Δ, Δc , given by
(28)    2  3
Δc = e 1 − |γ | − 1 − |γ | + 0.26 1 − |γ |
where A± = 1 + 2RaTf /(∂T0± (xf )/∂x)2 . Thus,  4 
+ 0.055 1 − |γ | , (36)
the jump relations in terms of the outer solutions
can be obtained by substituting Eqs. (26)–(28) into such that there is no solution for Δ < Δc and there are
Eqs. (20)–(22), respectively, two branches of solutions for Δ > Δc . Their approx-
      imate formulas are given by Cheatham and Matalon
∂T ∂YF,0 ∂YO,0
A 0 = −Le−1 = −Le−1
[17] as
,
∂x F ∂x O ∂x   
d0 Δ−4/3 exp −d1 (Δ − Δc )d2 ,
(29) S1 =  
[AT1 ] = −Le−1 −1 Δ−1/3 q0 + q1 (Δ − Δc )q2 ,
F [YF,1 ] = −LeO [YO,1 ], (30)   
    e0 Δ−4/3 exp −e1 (Δ − Δc )e2 ,
1 ∂T1 ∂Y
= − xf YF,1 + Le−1 S2 =  
F,1
Bxf T1 +
A ∂x F ∂x Δ−1/3 r0 + r1 (Δ − Δc )r2 ,
 
∂YO,1
= − xf YO,1 + Le−1
where the upper and lower expressions correspond
O , (31)
∂x to the solutions of the lower and upper branches, re-
spectively (see, for example, Fig. 2 in Ref. [14]). For
where B ± = [(A± )2 + 1]/2A± and we have used the γ > 0 and < 0, S1 and S2 correspond to (SF , SO ) and
notation [AT1 ] = A+ T1+ (xf ) − A− T1− (xf ). (SO , SF ), respectively. The coefficients di , ei , qi , and
The jump relations (14) and (29) provide enough ri (i = 0, 1, 2) only depend on γ and are given in the
inner boundary conditions for the O(1) outer so- appendix of Ref. [14]. The heat transfer parameter, γ ,
lutions. However, two additional conditions for the measures the degree of asymmetry of the thermal dif-
O(ε) outer solutions are required to close the prob- fusion across the reaction zone. If γ = 0, the tempera-
lem. They can be supplied by the amounts of leakage ture profile is symmetric and hence heat is conducted
of the fuel and oxidant through the reaction zone, and to the fuel and oxidant sides equally so that SF = SO .
are given as As γ > 0, more heat is conducted to the oxidant side
+ and consequently the oxidant is more completely con-
YF,1 (xf ) = LeF SF (γ , Δ), sumed, so that SF > SO . Similarly, SF < SO as γ < 0.
−
YO,1 (xf ) = LeO SO (γ , Δ) (32) The term htotal = (1 + γ )hO /2 + (1 − γ )hF /2 in
Eq. (34) represents the excess/deficiency of the to-
by transforming the energy equation (15) into the tal enthalpy in the reaction zone. This is the term
Liñán canonical form, where that was overlooked in previous analyses [4,6–8].
   Without considering it, Δ depends linearly on Da
∂T + (xf ) ∂T − (xf ) ∂T −1
γ = − A+ 0 + A− 0 A 0 so that its minimum value, DaE , coincides with that
∂x ∂x ∂x of Δ, Δc . Because γ is determined from the lead-
(33) ing order solutions, the extinction condition, repre-
and sented by DaE , can be accordingly determined di-
  rectly from Δc . However, when the excess enthalpy is
∂T −2 taken into consideration, Eq. (34) becomes a nonlin-
Δ = 4LeF LeO Da A 0
∂x ear relation between Δ and Da because htotal depends
  on the reactant leakages SF and SO , which, in turn,
1+γ 1−γ
× exp hO + hF (34) depend on Δ. Consequently, Δc does not correctly
2 2
identify the extinction condition, DaE , and the deter-
is the reduced Damköhler number, SF and SO are mination of Da from a given Δ requires the O(ε)
the leakage functions of the fuel and oxidant, respec- outer solutions, which substantially complicates the
tively, and solution procedure.
106 H.Y. Wang et al. / Combustion and Flame 148 (2007) 100–116

2.4. Summary of the model and the flame location xf and temperature Tf can be
respectively determined from the jump relation (38)
So far we have derived the jump relations in implicitly as
terms of the outer solutions and the reactant leakages
across the reaction zone, which provide sufficient in- Le−1 2
F YF,−∞ exp(−LeF xf /2)
ner boundary conditions to fully determine the O(1) I − (xf ; LeF )
and O(ε) outer solutions and the flame sheet location.
They are summarized as follows: Le−1 2
O YO,∞ exp(−LeO xf /2)
=− , (45)
To O(1): I + (xf ; LeO )
 
 +   − 
+
[T0 ] = [YF,0 ] = [YO,0 ] = YF,0 −
(0) = YO,0 (0) = 0, ∂T0 (xf ) 2 ∂T0 (xf ) 2
+ 2RaTf + + 2RaTf
(37) ∂x ∂x
     
A
∂T0
= −Le−1
∂Y F,0
= −Le−1
∂YO,0
. Le−1 2
F YF,−∞ exp(−LeF xf /2)
∂x F ∂x O ∂x = . (46)
I − (xf ; LeF )
(38)
To O(ε): Equation (46) indicates that the chemical heat release
is conducted away from the flame to both sides. It is
[AT1 ] = −Le−1 −1
F [YF,1 ] = −LeO [YO,1 ], (39) seen that in addition to being used to heat up the re-
    actants, the heat release now needs to compensate for
1 ∂T1 ∂Y
= − xf YF,1 + Le−1
F,1
Bxf T1 + F
the radiative loss through the term 2RaTf in Eq. (46).
A ∂x ∂x However, the temperature gradient decreases at the
 
∂YO,1 same time, due to the reduction of the flame tem-
= − xf YO,1 + Le−1
O , (40)
∂x perature through radiative loss. Moreover, it is noted
+ − that the total heat release is controlled by the reac-
YF,1 (xf ) = LeF SF (γ , Δ), YO,1 (xf ) = LeO SO (γ , Δ).
tants consumption and hence is fixed. In most cases,
(41)
these two opposite effects on heat transfer by radia-
It is noted that we have not derived the jump re- tive loss are not equal. Therefore, the overall outcome
lations for the O(δ) expansion terms of T . This is of radiative loss is to redistribute the proportions of
because the matching conditions Θ2± (0) ∼ 0 imply heat transfer from the flame to both sides, and as such
that the O(δ) terms can be solved separately on ei- it plays a role similar to that of varying the thermal
ther side of the reaction zone. diffusivities of the reactants.
Applying the homogeneous boundary conditions
2.5. Solutions of counterflow diffusion flame at x → ±∞ and the jump and leakage conditions
(39)–(41), the O(ε) outer solutions, T1± , YF,1 ±
, and
We now apply the model (37)–(41) to the coun- ±
YO,1 , can be fully determined. Here we only show
terflow diffusion flame with nonunity Lewis numbers
them in the form of the excess enthalpies, hF and hO ,
and radiative loss. The O(1) outer solutions subject
because they are the reason the O(ε) outer solutions
to the boundary condition (4) and jump relations (37)
are required. Thus we have
and (38) can be solved as
⎧
⎪ T + (Tf − T−∞ )I − (x; 1)/I − (xf ; 1), hF = CF,1 SF (γ , Δ) + CO,1 SO (γ , Δ),
⎨ −∞
⎪
x < xf ,
T0 (x) = hO = CF,2 SF (γ , Δ) + CO,2 SO (γ , Δ)
⎪
⎪ T∞ + (Tf − T∞ )I + (x; 1)/I + (xf ; 1),
⎩
x > xf , and the total excess enthalpy is then given as
⎧ (42)
 
⎨ YF,−∞ 1 − I − (x; LeF )/I − (xf ; LeF ) ,
htotal = CF SF (γ , Δ) + CO SO (γ , Δ),
YF,0 (x) = x < xf ,
⎩
0, x > xf , where
(43)
 0, x < x ,
 f  Ci = (1 − γ )Ci,1 /2 + (1 + γ )Ci,2 /2, i = F, O,
YO,0 (x) = YO,∞ 1 − I + (x; LeO )/I + (xf ; LeO ) ,  +    
x > xf , MT MT− −1 + MT+
CF,1 = − − Le M
F F −
(44) A+ A− A+
where  −1 MT−  + LeF +  
LeF A− − MF− MO − Le MF
    + O
,
I ± (x; L) = π/2L ∓1 + erf L/2x −1 + −1 −
−LeF MO + LeO MF
 H.Y. Wang et al. / Combustion and Flame 148 (2007) 100–116 107

      
MT+ MT− −1 1 1 ∂T −2
CO,1 = − × exp Ta − A 0 .
A+ A− Tref Tf ∂x
  −1 M −  + −  (47)
LeF A−T − MF− (MO − MO )
× −1 + −1 −
, The extinction Damköhler number Da∗E corre-
−LeF MO + LeO MF sponds to the minimum value of Da∗ ; i.e.,
 +    
MT MT− −1 MT+
CF,2 = − − Le M
F F
+
− ∂SF (γ , Δ)/∂Da∗ → ∞. (48)
A+ A− A+
 +
  We have known that for a fixed system with given
−1 MT
LeF A+ − MF − + LeF + 
MO − Le MF Ra∗ , all the terms in Eq. (47) other than Δ and Da∗
O
+ , can be determined. Thus, Eq. (47) gives a definite re-
−1 + −1 − lation between Da∗ and Δ. Substituting Eq. (47) into
−LeF MO + LeO MF
 +   +  Eq. (48) yields the extinction condition
MT MT− −1 MT MT−  
CO,2 = − − − −
A+ A A+ A ∂h
Δ total −1=0 (49)
∂Δ ext
 MT+ − − 
Le−1 +
F A+ − MF (MO − MO ) from which the reduced Damköhler number at ex-
+
−Le−1 + −1 −
F MO + LeO MF
tinction, Δext , can be solved. Then the extinction
Damköhler number Da∗E can be solved by substitut-
and ing Δext into Eq. (47).
1 exp(−xf2 /2)
MT± = B ± xf + ,
A± I ± (xf ; 1)
3. Results and discussion
exp(−Lei xf2 /2)
Mi± = xf + Le−1 , i = F, O.
i I ± (xf ; Lei ) We use a CH4 /air counterflow diffusion flame to
We note that the total excess enthalpy htotal is a demonstrate the dual extinction characteristics with
linear combination of the reactant leakages, SF and the effects of radiative loss and nonunity Lewis num-
SO , while γ and the coefficients CF and CO are only bers. The system parameters adopted are T̃−∞ =
dependent on the leading order solutions and radia- T̃∞ = 300 K, T̃a = 24,000 K, cp = 0.334 kcal/kg K,
tive loss. Thus, for a fixed system with given radiative and q = 11,990 kcal/kg. The boundary conditions for
loss, the total enthalpy htotal is solely determined by the fuel and oxidant fractions are fixed at ỸF,−∞ = 1
the value of Δ. and ỸO,∞ = 0.23, respectively, throughout this study
unless otherwise specified. Thus, the initial mixture
2.6. Extinction analysis strength, defined as φ = ν ỸF,−∞ /ỸO,∞ , is 17.4 in
the current study.
Since the quantifying parameters ε, Ra, and Da are Fig. 2 shows variations of the flame location, xf ,
functions of the flame temperature, Tf , which varies with the fuel and oxidant Lewis numbers, LeF and
with radiative loss, it is necessary to rescale all the pa- LeO , respectively. It is seen that the flame is always
rameters to a fixed, absolute reference state [6]. Using
the adiabatic flame temperature Tref for ỸF,−∞ = 1,
ỸO,∞ = 0.23, and LeF = LeO = 1 as the reference
state, and denoting the associated quantities by the su-
perscript “∗,” we have
 2 3
Da∗ = Tref /Ta DaC exp(−Ta /Tref ),
4 ρc4 k ,
Ra∗ = 4δσ κq 3 Tref p
so that
 
Da = Da∗ (Tf /Tref )6 exp Ta (1/Tref − 1/Tf ) ,
Ra = Ra∗ (Tf /Tref )4
and Eq. (34) becomes
   Fig. 2. Variations of the flame sheet location, xf , with the
Tf 6
Δe−htotal = Da∗ 4LeF LeO fuel and oxidant Lewis numbers, LeF (LeO = 1) and LeO
Tref (LeF = 1).
108 H.Y. Wang et al. / Combustion and Flame 148 (2007) 100–116
(a)
(b)
Fig. 3. Variations of the coefficient of fuel leakage in the total
excess enthalpy, CF , with (a) LeF (LeO = 1), and (b) LeO
(LeF = 1), for different values of radiative loss, Ra∗ .
located on the oxidant side of the stagnation surface
over the entire ranges of LeF and LeO , which is a
consequence of the stoichiometric nature of diffusion
flames. Furthermore, it is seen that with the decrease
of LeF (LeO ), the flame moves toward the oxidant
(fuel) side and LeF has a much stronger effect on the
flame location than LeO . This is because for the cur-
rent problem the flame is located on the oxidant side
of the stagnation surface. Thus, fuel diffusion has to
overcome convection in the opposite direction to sup-
ply the fuel to the flame. Consequently, fuel diffusion
plays a more important role than that of the oxidant,
so that variation of LeF leads to a much larger shift
of xf .
Fig. 3 shows variations of the coefficient of fuel
leakage in the total excess enthalpy, CF , with LeF
and LeO for different values of Ra∗ , respectively.
Here only the results for CF are presented because
the flame is located on the oxidant side of the stag-
nation surface for the current problem, so that γ is
sufficiently larger than zero and hence SF  SO . It is
seen that CF decreases with the increase of LeF , LeO
(a)
(b)
Fig. 4. Fuel leakage SF as a function of the reduced Damköh-
ler number Δ and Damköhler number Da∗ for LeF = LeO
= 1 and different values of radiative loss, Ra∗ .
and Ra∗ and for LeF > 1, LeO > 1, and Ra∗ > 0,
CF is always negative. Thus, the excess enthalpy is
always negative under these parameters. For flames
with radiative loss, the only possibility for the excess
enthalpy to be positive is that the Lewis numbers are
sufficiently smaller than unity.
Fig. 4 shows the fuel leakage SF as a function
of the reduced Damköhler number Δ (Fig. 4a) and
Da∗ (Fig. 4b) for LeF = LeO = 1 and different val-
ues of Ra∗ . It is seen that there exist minimum val-
ues for Δ and Da∗ , Δc and Da∗E , and there are two
branches of solutions when Δ > Δc and Da∗ > Da∗E ,
respectively. However, because of the nonlinear rela-
tion between Δ and Da∗ , Δc does not coincide with
Da∗E except for the case of zero excess enthalpy when
LeF = LeO = 1 and Ra∗ = 0. It is noted that flame
extinction corresponds to the minimum of Da∗ , Da∗E ,
which is a parameter that can be independently speci-
fied. However, it does not correspond to the minimum
of Δ, Δc , which is not an independently specifiable
parameter because it depends on the value of γ de-
 H.Y. Wang et al. / Combustion and Flame 148 (2007) 100–116
termined from the leading order solutions. As shown
in Fig. 4b, the turning point Δc in the SF ∼ Δ plot,
indicated by the “×” symbol, is located on the up-
per branch of the SF ∼ Da∗ plot. The extinction limit
here, Da∗E , indicated by the “F” symbol, corresponds
to the kinetic extinction limit at high stretch rates.
Near this limit, increasing Da∗ leads to longer res-
idence time and hence more complete burning and
less reactant leakage. Thus, the lower branch is the
physically realistic solution. Fig. 4b also shows that
the SF ∼ Da∗ curve shifts to the right with increas-
ing Ra∗ . This means that for larger Ra∗ a higher Da∗
is required for steady burning in order to compensate
for the effect of radiative loss, and for a fixed Da∗ the
amount of fuel leakage SF with larger Ra∗ is greater
than that with smaller Ra∗ . Consequently, the extinc-
tion limit, Da∗E , increases with Ra∗ . To illustrate the
effect of neglecting the excess enthalpy on the pre-
diction of the extinction limit, the SF ∼ Da∗ curves
for different Ra∗ without considering the excess en-
thalpy are also plotted in Fig. 4b. It is observed that
the fuel leakage at the same Da∗ and the extinction
Damköhler number, Da∗E , are underestimated due to
the omission of the excess enthalpy, and this under-
estimation becomes more significant with increasing
Ra∗ . This is because the excess enthalpy is negative
for the parameters used in this figure (LeF = LeO = 1
and Ra∗ > 0) and therefore tends to weaken the flame.
Thus omitting it would lead to flames that were more
resistant to extinction, with smaller values of Da∗E .
For a given Ra∗ , the extinction Damköhler num-
ber, Da∗E , can be solved from Eqs. (47) and (49).
However, because both Ra∗ and Da∗ depend on the
stretch rate k, the Da∗E solved for a fixed Ra∗ , such
as those in Fig. 4b, is only relevant for studying the
effects of varying the fuel/oxidant system for a coun-
terflow of fixed stretch rate. On the other hand, an
equally relevant question to ask is the effect of varying
the stretch rate for a fixed fuel/oxidant system. Be-
cause of the inverse dependence of both Ra∗ and Da∗
on k, the relative effect of radiative heat loss can be
evaluated by defining a parameter Γ = Ra∗ /Da∗ [6]
that represents the strength of the radiative loss rela-
tive to the chemical heat release. It depends only on
the thermochemical parameters, but is independent of
the stretch rate k. Thus, in additional to Eqs. (47) and
(49), the extinction Damköhler numbers also need to
satisfy Ra∗ /Da∗ = Γ at the same time and hence
can be obtained from the intersection points between
the Da∗E ∼ Ra∗ curves and the Da∗ ∼ Ra∗ /Γ lines.
Fig. 5 shows the variation of Da∗E with Ra∗ for LeF =
LeO = 1 and the Da∗ ∼ Ra∗ /Γ lines for different
values of Γ . It is seen that for a fixed system with
Γ smaller than the critical value Γc ≈ 0.298, there
are two intersection points and hence two extinction
limits at different level of radiative loss. The one at
109
Fig. 5. Variations of the extinction Damköhler number, Da∗E ,
with the radiative loss, Ra∗ , for LeF = LeO = 1 and
the Da∗ versus Ra∗ /Γ lines for different relative radiative
strengths, Γ .
smaller Ra∗ leads to a smaller extinction Damköh-
ler number, namely Da∗E,K , and hence corresponds to
the kinetic extinction limit, whereas the one at larger
Ra∗ corresponds to the radiative extinction limit oc-
curring at a larger Da∗ , namely Da∗E,R . Thus, steady
burning is only possible within a limited range of
Damköhler number, Da∗E,K < Da∗ < Da∗E,R . Further-
more, it is seen that with increasing Γ the two ex-
tinction Damköhler numbers approach each other and
there is no intersection point once Γ exceeds the criti-
cal value, Γc . This trend can be observed more clearly
from the dual extinction Damköhler numbers versus
Γ plot shown in Fig. 6. It is seen that Da∗E,R de-
creases and Da∗E,K increases with the increase of Γ ,
and the two branches of extinction Damköhler num-
bers ultimately merge at Γ = Γc so that burning is
not possible regardless of the stretch rate for Γ > Γc .
Furthermore, it is noted that the merging is mostly
effected by the reduction of Da∗E,R instead of the in-
crease of Da∗E,K . Thus, Da∗E,R is more sensitive to Γ
than Da∗E,K . This is because Da∗E,R is usually much
larger than Da∗E,K and hence increasing Γ implies a
much larger increase of radiative loss at the radiative
limit than at the kinetic limit. Consequently, the extent
of Da∗E,R decrease with increasing Γ is much larger
than the extent of Da∗E,K increase. To illustrate the ef-
fect of missing the excess enthalpy on the dual extinc-
tion limits, the result without considering the excess
enthalpy is also shown in Fig. 6 by the dash curve. It
is seen that while neglecting the excess enthalpy only
has a very small effect on the kinetic extinction limit,
it has a much larger effect on the radiative limit. This
is due to the much larger radiative loss at the radiative
extinction limit. As shown in Fig. 3, the excess en-
thalpy is negative and decreases with the increase of
radiative loss. Thus, the absolute value of the excess
enthalpy at the radiative limit is much larger than that
110 H.Y. Wang et al. / Combustion and Flame 148 (2007) 100–116
Fig. 6. Variations of the dual extinction Damköhler numbers
with the relative radiative strength Γ for LeF = LeO = 1
with and without considering the excess enthalpy.
Fig. 7. Fuel leakage SF as a function of the Damköhler num-
ber Da∗ for LeF = LeO = 1 and different relative radiative
strengths Γ .
at the kinetic limit, and neglecting it as a factor weak-
ening the flame leads to an O(1) overestimation of
Da∗E,R and a slight underestimation of Da∗E,K . For ex-
ample, at Γ = 0.26, missing the excess enthalpy leads
to a 60% overestimation of Da∗E,R and a 13% under-
estimation of Da∗E,K .
Fig. 7 shows variations of the fuel leakage SF with
Da∗ for LeF = LeO = 1 and different values of Γ . It
is seen that the solutions are bounded by the two turn-
ing points, with the one at smaller Da∗ , marked by
“F,” representing the kinetic extinction limit, Da∗E,K ,
and the one at larger Da∗ , marked by “!,” repre-
senting the radiative limit, Da∗E,R . It is shown from
the lower branch solution that with the increase of
Da∗ from Da∗E,K , SF first decreases significantly, then
levels off, and finally increases as Da∗ approaches
Da∗E,R , leading to the second turning point at larger
Da∗ . This trend is due to the fact that radiative loss
is rather small near the kinetic extinction limit and
Fig. 8. Variations of the normalized flame temperature
Tf /Tad with the Damköhler number Da∗ for LeF = LeO =
1 and different relative radiative strengths Γ .
insufficient residence time is the dominant mecha-
nism for the excessive reactant leakage. Thus, increas-
ing Da∗ leads to a longer residence time and con-
sequently more complete burning and less reactant
leakage. However, for a fixed system, increasing Da∗
(decreasing the stretch rate) implies that the radiative
loss, Ra∗ , increases at the same time. Thus, at large
enough Da∗ , the radiative loss eventually becomes
dominant and is expected to lead to substantial re-
duction in the flame temperature, and hence increased
reactant leakage due to the substantial reduction in
the reaction rate. Thus it can be concluded that the
ultimate cause of extinction at the radiative limit is
still the insufficient time for the reaction to complete.
Moreover, it is seen that the fuel leakage SF at the
radiative extinction limit is smaller than that at the ki-
netic limit. This is because near the radiative limit,
the flame is much weaker and hence is susceptible to
extinction with smaller reactant leakages. The upper
branch of the solutions, plotted in dotted lines, shows
the opposite dependence on Da∗ and thus is not phys-
ically realistic. The two branches of solutions form a
closed isola-shape of SF ∼ Da∗ curve with two turn-
ing points instead of only one for the open C-shaped
curve shown in Fig. 4b.
Fig. 8 shows variations of the normalized flame
temperature Tf /Tad with Da∗ for LeF = LeO = 1 and
different values of Γ , where Tad is the adiabatic flame
temperature. It is seen that although the reactant leak-
age decreases and then increases with Da∗ , as shown
in Fig. 7, the flame temperature decreases monoton-
ically because the flame suffers more radiative loss
with larger Da∗ .
We now study effects of the Lewis numbers of
the fuel and oxidant, LeF and LeO , on the extinction
characteristics of the counterflow diffusion flame with
radiative loss. Fig. 9 shows variations of the extinction
Damköhler number, Da∗E , with LeF and LeO for dif-
 H.Y. Wang et al. / Combustion and Flame 148 (2007) 100–116
(a)
(b)
Fig. 9. Variations of the extinction Damköhler number Da∗E
with (a) LeF (LeO = 1) and (b) LeO (LeF = 1) for different
values of radiative loss, Ra∗ .
ferent values of Ra∗ . It is seen that for the adiabatic
flame (Ra∗ = 0), Da∗E increases monotonically with
increasing LeO and LeF , consistent with the results of
Seshadri and Trevino [19]. In the presence of radiative
loss, Da∗E increases monotonically with LeO , whereas
it decreases and then increases with LeF . Specifically,
at small LeF , Da∗E is very sensitive to the radiative
loss, so that even a very small amount of loss, e.g.,
Ra∗ = 2 × 10−4 , is able to induce a very large relative
increase of Da∗E . However, at larger LeF this sensitiv-
ity becomes moderate, leading to the nonmonotonic
dependence of Da∗E on LeF for a fixed Ra∗ , as shown
in Fig. 9a. The ratio of Da∗E to its adiabatic value,
Da∗E,ad , can be obtained from Eq. (47) as
    
Da∗E Tad 6 1 1
= exp Ta −
Da∗E,ad Tf Tf Tad
Δext
× exp(htotal,ad − htotal ) , (50)
Δext,ad
where the subscript “ad” designates the values cor-
responding to the adiabatic state, i.e., Ra∗ = 0. It
111
(a)
(b)
Fig. 10. Variations of the ratio of the extinction Damköhler
numbers Da∗E /Da∗E,ad and the contributions from the effects
of radiative loss on the flame temperature, fT , the excess
enthalpy, fh , and the reduced extinction Damköhler num-
ber, fΔ , with (a) LeF (LeO = 1) and (b) LeO (LeF = 1) for
radiative loss Ra∗ = 2 × 10−4 .
is seen from Eq. (50) that radiative loss modulates
Da∗E through its effects on the flame temperature,
fT = (Tad /Tf )6 exp[Ta (1/Tf − 1/Tad )], excess en-
thalpy, fh = exp(htotal,ad − htotal ), and the reduced
extinction Damköhler number, fΔ = Δext /Δext,ad .
Thus, it is necessary to evaluate their relative contri-
butions to Da∗E at different Lewis numbers and radia-
tive losses.
Fig. 10 shows variations of Da∗E /Da∗E,ad , fT , fh ,
and fΔ with LeF and LeO for Ra∗ = 2 × 10−4 . It
is seen that the relative contributions from fT , fh ,
and fΔ are different at small and large values of LeF
and LeO . Specifically, at small LeF , radiative loss in-
duces a very large increase of Δext , which contributes
mostly to the increase of Da∗E . This contribution de-
creases significantly with increasing LeF so that the
contribution from the temperature term, fT , becomes
112 H.Y. Wang et al. / Combustion and Flame 148 (2007) 100–116
dominant at large LeF . Over the entire range of LeF ,
the contribution from the excess enthalpy, fh , is sec-
ondary. However, it contributes most to the increase
of Da∗E at small LeO , as shown in Fig. 10b. The con-
tribution then decreases significantly with increasing
LeO and becomes a secondary effect when LeO is suf-
ficiently large. The variation of fΔ with LeO shows
the opposite trend to fh . It is small at small LeO but
increases with increasing LeO and becomes compara-
ble to fT at large LeO . Over the entire range of LeO ,
fT plays a consistently important role, although it is
a little smaller than fh at small LeO . The radiative
loss, Ra∗ = 2 × 10−4 , used in Fig. 10 can be treated
as a typical order of value at which kinetic extinction
occurs. Fig. 11 shows variations of Da∗E /Da∗E,ad , fT ,
fh , and fΔ with LeF and LeO for Ra∗ = 0.01, which
can be considered as a typical value at which radiative
(a)
(b)
Fig. 11. Variations of the ratio of the extinction Damköhler
numbers Da∗E /Da∗E,ad and the contributions from the effects
of radiative loss on the flame temperature, fT , the excess
enthalpy, fh , and the reduced extinction Damköhler num-
ber, fΔ , with (a) LeF (LeO = 1) and (b) LeO (LeF = 1) for
radiative loss Ra∗ = 0.01.
extinction occurs. For example, for LeF = LeO = 1
and Γ = 0.1, the kinetic and radiative extinction oc-
cur at Ra∗ = 1.2 × 10−4 and 0.017, respectively. It
is seen from Fig. 11 that the variations of fΔ and fh
with LeF and LeO exhibit the same trend as those for
small radiative loss Ra∗ = 2 × 10−4 , whereas the rel-
ative contribution from the temperature term, fT , is
much larger. This is due to the larger reduction of the
flame temperature under larger radiative loss. Further-
more, Figs. 10a and 11a show that it is the very large
increase of Δext from the adiabatic value at small LeF
that leads to a very large relative increase of Da∗E , as
shown in Fig. 9a.
Because Eq. (36) shows that the minimum of Δ,
Δc , is determined by γ and in general the reduced
extinction Damköhler number, Δext , is close to Δc ,
it is instructive to explore the very high sensitivity of
Δext to radiative loss at small LeF by studying the
variations of γ with LeF . Fig. 12 shows variations
of γ and the corresponding Δext with LeF for dif-
ferent values of Ra∗ . It is seen that for the adiabatic
flame, γ is close to unity at small LeF , implying that
(a)
(b)
Fig. 12. Variations of (a) γ and (b) the reduced extinction
Damköhler number Δext with LeF for LeO = 1 and different
values of radiative loss, Ra∗ .
 H.Y. Wang et al. / Combustion and Flame 148 (2007) 100–116
Fig. 13. Variations of the extinction Damköhler number,
Da∗E , with radiative loss, Ra∗ , for LeF = 1, Γ = 0.1 and
different values of LeO and the Da∗ versus Ra∗ /Γ line with
relative radiative strength Γ = 0.25.
most of the chemical heat release is conducted to the
oxidant side and the flame is nearly adiabatic to the
fuel side. This is because at small LeF the flame lo-
cation is very close to the oxidant stream. It is seen
from Fig. 12b that in this case Δext is close to zero.
However, in the presence of radiative loss, which is
assumed to occur on both sides of the reaction zone
in the current study, the flame always has heat con-
ducted to the fuel side to compensate for the radiative
loss. Thus, γ decreases, leading to an increase of Δext
from its adiabatic value. Furthermore, because the
adiabatic value of Δext is close to zero, even a small
increase of Δext from this value, induced by a very
small amount of radiative loss, implies a very large
relative increase. For example, at LeF = 0.5, with the
radiative loss Ra∗ = 2 × 10−4 , γ decreases from 0.98
to 0.93, whereas Δext increases from 0.053 to 0.18,
which is about 3.4 times larger.
For a fixed system with radiative strength Γ , how-
ever, we have known that the extinction Damköhler
number needs to satisfy Ra∗ /Da∗ = Γ at the same
time. Thus, Fig. 9 alone is not enough to demonstrate
variations of the dual extinction Damköhler numbers
with the Lewis numbers. In the same manner as that in
Fig. 5, Fig. 13 shows variations of Da∗E with Ra∗ for
LeF = 1, Γ = 0.1, and different values of LeO and the
Da∗ ∼ Ra∗ /Γ line with Γ = 0.25. It is seen that due
to the monotonic increase of Da∗E with LeO , increas-
ing LeO leads to an increase of Da∗E,K and a decrease
of Da∗E,R , and hence a smaller flammable range of
Da∗ . Furthermore, it is expected that there are no in-
tersection points between the Da∗E ∼ Ra∗ curve and
the Da∗ ∼ Ra∗ /Γ line if LeO is larger than a critical
value, indicating that steady burning is not possible if
LeO is too large.
Fig. 14 shows variations of the dual extinction
Damköhler numbers with Γ for different values of
113
(a)
(b)
Fig. 14. Variations of the dual extinction Damköhler num-
bers with relative radiative strength Γ for different values of
(a) LeO (LeF = 1) and (b) LeF (LeO = 1).
LeO and LeF , respectively. It is seen that for the same
Γ the flammable range of Da∗ diminishes with the
increase of LeO , leading to a smaller Γc that the sys-
tem can sustain. Thus increasing LeO tends to weaken
the flame monotonically. Although the dependence of
Da∗E on LeF is not monotonic over its entire range,
Fig. 9a shows that, with an increase of LeF , Da∗E de-
creases and increases monotonically for small and
large LeF , respectively. Thus, for the same reason,
it is seen from Fig. 14b that the flammable range of
Da∗ diminishes as LeF increases from 1.5 to 2.5, and
extends as LeF increases from 0.6 to 0.8. It is noted
that Mills and Matalon [11] predicted monotonic de-
pendence of the flammable range of the Damköhler
number on the Lewis number of the fuel, LeF . This is
because the radiative loss was assumed to be of O(ε)
quantity and hence does not affect the leading order
flame temperature. Thus radiative loss affects extinc-
tion only through the excess enthalpy, which depends
monotonically on LeF .
Fig. 15 shows variations of the dual extinction
Damköhler numbers with LeF for different values of
114 H.Y. Wang et al. / Combustion and Flame 148 (2007) 100–116
Fig. 15. Variations of the dual extinction Damköhler num-
bers with LeF for different values of LeO and relative radia-
tive strength Γ = 0.1.
LeO and Γ = 0.1. It is seen that the flammable range
of Da∗ decreases with increasing LeO and LeF that
sufficiently large. When they are both large enough,
steady burning is not possible for this value of Γ .
With the increase of LeO , the maximum value of LeF
the system can sustain decreases. Fig. 15 indicates
that steady burning for radiative diffusion flames is
only possible within a limited range of Lewis num-
bers. This is because flame extinction is purely ther-
mal in nature under the assumption of a one-step re-
action and since thermal conduction and radiation are
both heat loss mechanisms for flame extinction; either
of them being large enough is able to induce extinc-
tion.
4. Conclusions
The present study has yielded the following spe-
cific contributions and understandings regarding the
subject phenomena. First, we have performed a mul-
tiscale asymptotic analysis for counterflow diffusion
flames with flame radiative heat loss and developed a
formulation for general Lewis numbers of the fuel and
oxidant with the proper consideration of the excess
enthalpy that was overlooked in previous analyses.
This formulation is expressed by the jump relations
in terms of the outer solutions and the reactant leak-
ages through the reaction zone. The reactant leakages
are obtained from solving the canonical form of the
structure equation and hence can take advantage of
the previous results, such as those of Cheatham and
Matalon [17]. The reason that the structure equation
of the reaction zone, with a radiation zone sandwich-
ing it in the current analysis, still can degenerate to
the Liñán canonical form [1] is that the thin reaction
zone is reactive–diffusive in nature. Consequently, its
structure may not be directly affected by the partic-
ulars of the system. Thus, it is possible to perform
a generalized analysis of the reaction zone with the
influences of all the outside processes including ra-
diative loss coming in from the boundary conditions
through matching.
Second, this formulation is then applied to study
the extinction characteristics of the radiative counter-
flow diffusion flame with nonunity Lewis numbers of
the fuel and oxidant. In additional to the dual extinc-
tion limits, namely the kinetic limit at lower Damköh-
ler numbers and radiative limit at higher Damköhler
numbers, identified in previous analyses [5–9,11], the
current study has also gained some additional under-
standing, especially of the effects of nonunity Lewis
numbers. It is found that the kinetic extinction limit
is minimally affected by radiative loss so that extinc-
tion occurs close to the adiabatic flame temperature,
while a substantial amount of heat loss is associated
with the radiative limit, so that the flame temperature
at extinction is significantly reduced from the adia-
batic value. Reactant leakage, however, is ultimately
the root cause for both limits, with the extent of leak-
age being greater for the kinetic limit. The flammable
range of the Damköhler number decreases monoton-
ically with increasing LeO , indicating that increasing
LeO tends to weaken the flame. However, this range
shows nonmonotonic dependence on LeF in that it in-
creases and decreases with increasing LeF when it is
sufficiently small and large, respectively. This non-
monotonic dependence is due to the very high sensi-
tivity of the extinction Damköhler number to radiative
loss at small LeF . The root cause for this phenomenon
is that the flame is located very close to the oxidant
stream at small LeF so that most of the chemical heat
release is conducted to the oxidant side and the flame
is nearly adiabatic on the fuel side. A small amount of
radiative loss can lead the flame to deviate from this
condition and hence can lead to a rather large relative
increase of the extinction Damköhler number. The in-
fluence of radiative loss on the extinction Damköh-
ler numbers is found to be through its effects on the
flame temperature, excess enthalpy, and the reduced
extinction Damköhler number, and their relative con-
tributions are different under different radiative loss
and Lewis numbers of the fuel and oxidant. In most
cases, the contribution from the flame temperature is
the largest and its relative importance increases with
increasing radiative loss. The contributions from the
other two, however, are also important in some cases.
At small LeF , when the flame is located very close
to the oxidant stream, the contribution from the re-
duced extinction Damköhler number is large and even
dominant in the case of small radiative loss. The con-
tribution from the excess enthalpy is important for
small LeO and it may be comparable to that from the
 H.Y. Wang et al. / Combustion and Flame 148 (2007) 100–116
flame temperature when radiative loss is small. Thus,
overlooking the excess enthalpy in previous analyses
may result in rather large error in the predicted extinc-
tion Damköhler numbers, especially the kinetic one.
Moreover, it is found that for a fixed relative radiative
strength, steady burning may not be possible when the
Lewis numbers of the fuel and oxidant are too large.
Finally, it should be noted that the above discussions
are meant for large initial mixture strengths such that
the flame is located in the oxidant side of the stag-
nation surface. The roles of LeF and LeO should be
interchanged for small initial mixture strengths under
which the flame is located in the fuel side of the stag-
nation surface.
Third, the relations (37)–(41), developed in this
study for the counterflow diffusion flame with flame
radiation as a demonstration problem, can be applied
to any one-dimensional configuration with Eq. (41)
revised as
115
crucial element in the stability analyses. For exam-
ple, Kim et al. [18] and Cheatham and Matalon [17]
and Kukuck and Matalon [14] respectively show that
flames exhibit cellular and pulsating instabilities for
positive and negative values of the excess enthalpy.
Acknowledgment
This work was supported by NASA and the Air
Force Office of Scientific Research, under the techni-
cal monitoring of Dr. Kurt Sacksteder and Dr. Julian
Tishkoff, respectively.
Appendix A
Integrating Eq. (23) and applying the matching
conditions, Θ2± |ζ →±∞ ∼ −∞ and Θ2± (0) ∼ 0, yield
    
 ±   ±
Buf T1 −
1 ∂T1
= − uf YF,1 − LeF −1 ∂YF,1 dΘ2± ∂T0 (xf ) 2 Θ2
A ∂x ∂x =∓ + 2RaTf exp
dζ ∂x Tf
 
∂Y (A.1)
= − uf YO,1 − Le−1
O,1
O ,
∂x and
where uf is the flow speed at the flame sheet. dΘ2± (0) ∂T0± (xf )
Furthermore, although this formulation is applied = A± . (A.2)
dζ ∂x
exclusively to the extinction analysis in the current
study, it is also applicable to the analyses of thermal– Multiplying Eq. (24) by dΘ2± /dζ , integrating once,
diffusive instabilities in a manner similar to those in and applying the matching conditions (12) for dΘ1± /
dζ , dΘ2± /dζ , and Θ2± as ζ → ±∞ yield
Cheatham and Matalon [17] and Kukuck and Mat-
alon [14]. For example, we introduce small perturba-
 ±
tions in the form dΘ1± (0) dΘ2± (0) Θ2
    = RaΘ1± exp . (A.3)
(T , YF , YO ) = T̄ , ȲF , ȲO + εα u(x), v(x), w(x) dζ dζ Tf

× exp(ik1 y + σ t), (51) Substituting Eq. (23) into (A.3) to replace the term
Ra exp(Θ2± /Tf ) with d 2 Θ2± /dζ 2 , integrating once,
where the overbar “−” designates steady state solu- and applying the matching condition (12) for Θ1± and
tions, y is the transverse direction, k1 the wavenum-
Θ2± as ζ → ±∞ yield
ber in this direction, σ a complex number whose real
 
part identifies the growth rate of the perturbation, α
± T1± (xf ) dΘ2±
a small parameter, and u, v, and w the perturbations Θ1 =
for T , YF , and YO , respectively. Substituting Eq. (51) ∂T0± (xf )/∂x dζ

into the governing equations (1)–(3) and Eqs. (37)–   ±
 2RaTf Θ2
(41) yields the governing equations and the jump and = T1± (xf ) 1 + ± exp
(∂T0 (xf )/∂x) 2 Tf
leakage conditions for the perturbations, from which
the dispersion relation relating the growth rate to the (A.4)
parameters describing the combustion system, such as and
the Damköhler number and the Lewis numbers of the
fuel and oxidant, is formed. Because this model is Θ1± (0) = A± T1± (xf ). (A.5)
able to predict the dual extinction limits of radiative Integrating Eq. (25) once in the same manner yields
diffusion flames, it is expected to predict the neutral
!
stability boundaries near both the kinetic and radiative dΘ2± dΘ3± dΘ1± dΘ2±
extinction limits. = −xf dζ
dζ dζ dζ dζ
Finally, it should be emphasized that although the  ±
Θ2
inclusion of excess enthalpy does not lead to qual- + RaΘ3± exp + C±, (A.6)
itative differences in the extinction results, it is a Tf
116 H.Y. Wang et al. / Combustion and Flame 148 (2007) 100–116

where C ± are integral constants. Substituting [2] U. Bonne, Combust. Flame 16 (1971) 147–159.
Eq. (A.4) into (A.6) to replace the term dΘ2± /dζ by [3] M. Sibulkin, A.K. Kulkarni, K. Annamalai, Proc. Com-
 ∂T0± (xf )/∂x  ± bust. Inst. 18 (1981) 611–617.
± Θ1 and applying the matching condi- [4] S.H. Sohrab, A. Liñán, F.A. Williams, Combust. Sci.
T1 (xf )
tion (12) for Θ1± , Θ2± , and dΘ3± /dζ as ζ → ±∞
Technol. 27 (1982) 143–154.
[5] J.S. T’ien, Combust. Flame 65 (1986) 31–34.
lead to
[6] B.H. Chao, C.K. Law, J.S. T’ien, Proc. Combust. Inst.
 
∂T ± (xf ) ∂T1± (xf ) 1 23 (1990) 523–531.
C± = 0 + xf T1± (xf ) [7] T.K. Oh, J.S. Lee, S.H. Chung, Int. J. Heat Mass Trans-
∂x ∂x 2
fer 37 (1994) 2893–2900.
and [8] F. Liu, G.J. Smallwood, O.L. Gulder, Y. Ju, Combust.

 Flame 121 (2000) 275–287.
dΘ3± 1 ±
 2RaTf exp(Θ2± /Tf ) [9] K. Maruta, M. Yoshida, H. Guo, Y. Ju, T. Niioka, Com-
= − xf T1 (xf ) 1 +
dζ 2 (∂T0± (xf )/∂x)2 bust. Flame 112 (1998) 181–187.
[10] J.S. Kim, F.A. Williams, J. Eng. Math. 31 (1997) 101–
 ±
T1± (xf ) ± Θ2 118.
+ ± RaΘ 3 exp [11] K. Mills, M. Matalon, Proc. Combust. Inst. 27 (1998)
(∂T (xf )/∂x) 2 Tf
0 2535–2541.
∂T1± (xf )/∂x + 12 xf T1± (xf ) [12] R.H. Chen, G.B. Mitchell, P.D. Ronney, Proc. Com-
+ " . bust. Inst. 24 (1992) 213–221.
2RaTf exp(Θ2± /Tf )
1+ ± 2
[13] M. Füri, P. Papas, P.A. Monkewitz, Proc. Combust. Inst.
(∂T0 (xf )/∂x) 28 (2000) 831–838.
Then, applying the matching conditions, Θ2± (0) ∼ 0 [14] S. Kukuck, M. Matalon, Combust. Theory Model. 5
(2001) 217–240.
and Θ3± (0) ∼ 0, yields [15] C.H. Sohn, J.S. Kim, S.H. Chung, K. Maruta, Combust.
  Flame 123 (2000) 95–106.
dΘ3± (0) 1 1
±
1 ∂T (xf )
∼ A± − ± T1± (xf ) + ± 1 . [16] M. Miklavčič, A.B. Moore, I.S. Wichman, Combust.
dζ 2 A A ∂x Theory Model. 9 (2005) 403–416.
(A.7) [17] S. Cheatham, M. Matalon, J. Fluid Mech. 414 (2000)
105–144.
[18] J.S. Kim, F.A. Williams, P.D. Ronney, J. Fluid Mech.
References 327 (1996) 273–301.
[19] K. Seshadri, C. Trevino, Combust. Sci. Technol. 64
[1] A. Liñán, Acta Astronaut. 1 (1974) 1007–1039. (1989) 243–261.