Combustion and Flame 156 (2009) 16271640

Contents lists available at ScienceDirect

Combustion and Flame

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / c o m b u s t fl a m e

Simulations of laminar ame propagation in droplet mists

A. Neophytou, E. Mastorakos *
Department of Engineering, University of Cambridge, UK

a r t i c l e

i n f o

Article history:
Received 22 September 2008
Received in revised form 27 December 2008
Accepted 25 February 2009
Available online 23 May 2009
Keywords:
Flame propagation
Flame speed
Spray
Flame structure
Laminar ame speed
n-decane
n-heptane
relight

a b s t r a c t
In order to clarify the conditions conducive to propagation of premixed ames in quiescent sprays, a onedimensional code with detailed chemistry and transport was used. n-Heptane and n-decane, distinguished by their volatility, were studied under atmospheric and low temperature, low pressure conditions. The effects of initial droplet diameter, overall equivalence ratio /0 and droplet residence time
before reaching the ame front were examined. Increasing the residence time had an effect only for
n-heptane, with virtually no evaporation occurring before the ame front for n-decane. The trends were
only marginally correlated with the local gaseous equivalence ratio /eff at the location of maximum heat
release rate. /eff could be as low as 0.4 (beyond the lean ammability limit), but the ame speed could
still be 40% of the gaseous stoichiometric ame speed SL,0. For n-heptane, /eff increased towards /0 with
smaller droplets while high ame speeds occurred when /eff was near 1. This implied that the highest
ame speed was achieved with small droplets for /0 6 1 and with relatively large droplets for /0 > 1.
In the latter case, the oxidiser was completely consumed in the reaction zone and droplets nished evaporating behind the ame where the fuel was pyrolysed. The resulting small species, mainly C2H2, C2H4
and H2, diffused back to the oxidation zone and enhanced the reaction rate there. Ultimately, this could
result in ame speeds higher than SL,0 even with /0 = 4. For n-decane, the same trends were followed but
smaller droplets were needed to reach the same /eff due to the slow evaporation rate. Under low pressure
and low temperature, the effects of pressure and temperature on /eff and the ame speed were competitive and resulted in values close to the ones at atmospheric conditions.
2009 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

1. Introduction
Understanding the propagation of a premixed ame in a cloud
of droplets is relevant for many combustion applications including
spark ignition in internal combustion engines and aviation gas turbines. After the generation and growth of a ame kernel, the subsequent ame propagation will determine whether the ignition is
successful or not [1]. Although realistic systems involve polydisperse sprays and turbulent ows, fundamental understanding of
the factors controlling ame propagation can be provided by
studying monodisperse mists of quiescent droplets, which is the
topic examined in this paper.
Flame propagation in a uniformly-dispersed spray has been
widely studied in the past [29]. As discussed in the literature,
the structure of a laminar unstrained spray ame may be divided
in several zones, as shown in Fig. 1. With analytical methods, Silverman et al. [7] distinguished ve parts. The rst three parts are a
primary evaporation zone with a length l, a heating zone, and a
homogeneous reaction zone. The fuel that has evaporated in
the rst two zones reacts in the homogeneous reaction zone. Be* Corresponding author. Fax: +44 1223 3 32662.
E-mail address: em257@eng.cam.ac.uk (E. Mastorakos).

hind it, the surviving droplets keep evaporating at a high rate. If

oxygen remains, the fuel burns as soon as it evaporates, in the
so-called droplet burning zone. Finally, when there is no more oxidiser, droplets, if any, nish evaporating in a secondary evaporation zone. This ame structure has not been explored in detail
with complex chemistry simulations so far.
Regarding the ame propagation rate, Ballal and Lefebvre [5]
showed that as the initial droplet size became relatively large,
the ame speed decreased. Very small droplets evaporated before
the ame front and resulted in a ame propagation speed similar
to the purely gaseous ame. An important observation was the
existence of an optimum droplet size for a given fuel and overall
equivalence ratio /0 that resulted in enhanced laminar ame propagation compared to the purely gaseous fuel ame. This enhancement has been observed for the whole range of overall equivalence
ratio by Hayashi et al. [10] in a closed volume with laminar ames
of ethanol. For overall lean mixtures, a ame speed enhancement
was shown experimentally for tetralin [2,11], n-decane [9], kerosene [12,13], and ethanol [14]. The phenomenon was attributed
to the wrinkling of the ame by the droplets which increased the
ame surface area and promoted propagation [10,15] and to local
inhomogeneities of the mixture between droplets leading to optimal equivalence ratio (e.g. diffusion ames around droplets)

0010-2180/$ - see front matter 2009 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
doi:10.1016/j.combustame.2009.02.014

1628

A. Neophytou, E. Mastorakos / Combustion and Flame 156 (2009) 16271640

and the residence time of droplets before the ame (which affects
the degree of pre-evaporation) and we examine the importance of
detailed chemistry. Finally, in view of the importance to aviation,
we look at the effect of relight conditions, i.e. low pressure and
low temperature, on the ame speed. Apparently, few investigations have focused on this regime. In order to examine the effects
of fuel volatility, n-heptane and n-decane are used as typical fuels
with high and low volatility respectively. n-Heptane has been employed in academic congurations of ignition [25] while n-decane
is used as a surrogate for kerosene [26].
The numerical model implemented is presented in the next Section, followed by the results and their discussion. The particular
features revealed by the use of detailed chemistry are emphasized.
The paper closes with a summary of the main conclusions.
2. Numerical formulation

Fig. 1. Schematic of the temperature prole across a laminar freely propagating

premixed spray ame with a large SMD. The post evaporation zone exists in a rich
case only. Adapted from Ref. [7].

[4,9,1618]. With an overall rich ame, the existence of an optimum droplet size was related to the incomplete evaporation of
large droplets, which produced a gaseous mixture close to stoichiometry that led to high ame propagation speed. The gaseous stoichiometry at the ame front was denoted effective equivalence
ratio /eff by Hayashi et al. [10]. It was suggested that the optimum
conditions were a combination of droplet size and overall equivalence ratio that resulted in /eff near unity [7,14,16].
The ame speed enhancement has also been predicted numerically with one-dimensional formulations and single-step chemistry for octane [19] for overall rich ames. The explanation given
was that under some conditions the mixture at the ame could become close to stoichiometric. The one-dimensional formulation
does not account for multi-dimensional droplet-scale mixture
inhomogeneities, which is the explanation suggested by experiment, and hence the fact that ame speed enhancement can still
be observed in such simulations suggests that the effective equivalence ratio concept has some merit. However, /eff has not been
discussed yet in the context of detailed chemistry calculations.
Investigations on the minimum energy necessary to ignite
sprays have been reviewed by Aggarwal [20] who pointed out that
for the same reasons, for rich overall stoichiometry, large droplets
may result in a minimum of the ignition energy. Among others, this
was shown experimentally by Danis et al. with n-heptane [21],
Singh and Polymeropoulos with tetralin [22], Dietrich et al. with
n-decane [23] and numerically using single-step chemistry for
n-decane in closed volume by Aggarwal and Sirignano [24]. The
fact that intra-droplet regions in an overall too lean or too rich
spray may be close to stoichiometric also implies a widening of
the ammability limits of ames in sprays [20].
Investigations of the effect of fuel volatility on ame propagation indicate that less volatile fuel leads to a lower ame speed
[16] and a higher ignition energy [20]. This is generally explained
by the low evaporation rate of less volatile fuels resulting in low
fuel concentration in the gas and thus too low /eff. Studies on the
effect of low pressure on spray ames are seldom. Still, Ballal
and Lefebvre [5] provided valuable results for stoichiometric mixtures of diesel oil and air with a Sauter mean drop diameter (SMD)
of 60 lm. They showed that the pressure dependence of the ame
speed varies from p0.3 with pure gas to p0.5 with sprays.
Little work has been done on the numerical study of laminar
monodisperse spray ames with detailed chemistry. Here, we attempt to elucidate the effects of droplet size, equivalence ratio

The coupling between the gas and the liquid phase implemented in Run1DL [27,28,19], the code that was used here, is
essentially the same as the one presented by Abramzon and Sirignano [29]. The code solves a one-dimensional form of the governing equations with a distance coordinate y. A steady solution with
constant pressure is presented. The governing equations are repeated below for completeness.
2.1. Gas phase
The Eulerian gas phase equations include a source term from
the liquid phase. Assuming a dilute spray allows the treatment of
the spray in a simple statistical manner: the source term for a particular scalar is equal to the scalar production from one droplet
multiplied by the number of droplets per unit volume n. The conservation of mass, species, momentum and energy and the ideal
equation of state in the gas phase are given by

@q @qv

Sm
@t
@y
@qY i @qY i v
@qY i V i


wi Sm;i
@t
@y
@y


@qv @qvv
@p @
@v
l


Sv
@t
@y
@y @y
@y


X
@qcp T @qcp T v
@
@T
k
hi wi Se


@t
@y
@y
@y
i
!
X Yi
p qRT
Wi
i

1
2
3
4
5

The gaseous properties used in those equations are the density of the
gas, q, the one-dimensional velocity v, the pressure p, the mass fraction of species i, Yi, the chemical production of species i, wi, the viscosity l, the constant pressure heat capacity of the gas mixture cp,
the temperature T, the conductivity k, the absolute enthalpy of species i, hi, the ideal gas constant R and the molecular mass of species
i, Wi. The diffusion velocities Vi are calculated by V i  DX ii rX i V c
where Di is the mixture averaged diffusion coefcient of species i
and Vc is a correction velocity used to enforce the constraint
P
p 3
i qY i V i 0. Moreover,  1  n 6 d is the void volume fraction dened as the volume occupied by the gas only divided by the total volume. For our simulations 0 6 1   6 104 which supports the
assumption of dilute spray. The liquid source terms are given by

_
Sm nm


dv d
_ vd
Sv n md
m
dt
_ pd T  T s Lv 
Se nq_ mc

6
7
8

1629

A. Neophytou, E. Mastorakos / Combustion and Flame 156 (2009) 16271640

_ is the mass evaporation rate of the droplet, md is the

where m
instantaneous mass of the droplet, vd is the velocity of the droplet,
q_ is the rate of heat transferred to the droplet, cpd is the heat capacity of the liquid fuel, Ts is the surface temperature of the droplet and
Lv is the latent heat of evaporation at the boiling point. The liquid
source term for species i, Sm,i, is zero except when we consider
the fuel and then Sm,F = Sm.
2.2. Liquid phase

9
10

where d is the instantaneous diameter of the droplet and qd is the

_ is calculated using
density of the droplet. The rate of evaporation m
f and the thin lm properties. The
a modied Sherwood number Sh
Sherwood number and the Nusselt number are modied through
1 B0:7 so as to take into account the
a function FB ln1B
B
change in lm thickness due to Stefan ow [29]. Hence,

f ln1 BM
_ pdqf Df Sh
m
h
i
f 2 1
Sh
1 Red Sc1=3 max1; Red 0:077  1
FBM
qf djv  v d j
Red

lf

BM

Y Fs  Y F1
1  Y Fs

11
12
13
14

where qf is the thin lm density and Df is the binary diffusion coefcient of fuel vapour inside the thin lm. Red is the droplet Reynolds
number, always below 100 in our problem, and lf is the lm viscosity. The mass fraction of fuel at the droplet surface Y F s is found
through the ClausiusClapeyron equation by assuming phase equilibrium at the droplet surface

W F XFs

Y Fs

W F X F s 1  X F s W
 

pref
Lv 1
1
XFs

exp
p
R T ref T s

15
16

where pref and Tref are state reference pressure and temperature, WF
is the molecular mass of the fuel and W is the molecular mass of the
mixture excluding the fuel. Lv is assumed to be constant and equal
to its value at the boiling point. For n-heptane, pref = 1 bar,
Tref = 371.58 K and Lv = 317.927 kJ/kg. For n-decane, pref = 1 bar,
Tref = 447.30 K and Lv = 279.624 kJ/kg.
Inter-droplet interactions and gravity being neglected, the drag
force only affects the droplet movement and the momentum equation reduces to
2

md

d xd
dt

 
3C D q
jv  v d jv  v d
d qd



@T d
1 @
Td
r2
ad 2
r @r
@t
r

with ad the thermal diffusivity of the liquid. The droplet is subject to

the initial condition

and the boundary conditions

@T d
r 0 0
@r 

q_
@T d
d
2
r
2
@r
pd ad qd cp

21
22
d

The surface temperature, Ts, is found by solving Eq. (22) in which q_

is found through [29]

"
_
q_ m


0:38 #
cpd T  T s
T crit  T S
 Lv
BT
T crit  T ref

23

BT 1 BM /  1

24

f a
cp Sh
f
/ d:
:
c pf g
Nu Df

25

g
Nu 2

26

i
1 h
1 Red Pr1=3 max1; Red 0:077  1
FBT

where cpf and af are respectively the lm heat capacity and thermal
diffusivity of the lm mixture. g
Nu is the modied Nusselt number.
The critical temperature Tcrit is 540.15 K for n-heptane and 619.0 K
for n-decane. On the left boundary, droplets have a uniform temperature which is the cold gas temperature and their velocity is the
cold gas velocity. The ame speed is given as the speed of the uid
at the cold boundary. An initial ame speed is provided and iterations are performed until a steady state is established. The procedure for the coupling of the two phases is as follows. First, the
liquid-phase equations are solved using the initial eld of the gas
phase and liquid source terms are calculated. This is done with a
high-order accurate ODE solver whose maximum time-step size is
adapted to be sufciently small for at least two liquid-phase solution points to lie in each gas-phase control volume. This ensures
that in each gas-phase control volume linear interpolation of the liquid-phase source terms on the Eulerian nodes is meaningful and
sufciently accurate. A local source term is attributed to each Eulerian node, without smoothing being added. The gas-phase equations are then solved. The liquid phase is then solved anew based
on the new gas eld. The iterations are repeated until convergence
is achieved. Adaptive gridding based on all variables was used and
grid independence was ensured with about 160 points for all
solutions.
The overall equivalence ratio /0 was determined by the incoming mass uxes of gaseous air, gaseous fuel and liquid fuel. At the
cold boundary, the mass ux of gaseous fuel was set to zero so that
the liquid fuel ux only determined the overall equivalence ratio.
Table 1 summarises the conditions investigated.

Table 1
Range of conditions studied. Case C is referred to as relight.

sion, the conservation equation for the ux of droplets becomes

where 0 denotes the conditions at the cold boundary.

20

17



2=3
Re
24
where C D Re
1 6d is the drag coefcient [29]. In one-dimen-

nud n0 u0

19

T d r; 0 T d0

The liquid phase is computed in a Lagragian manner by tracking

one individual droplet and solving the associated equations for its
mass, momentum and energy. A thin lm assumption is used to
calculate the droplet evaporation and heat exchange with the surroundings [29]. The parameters inside the thin lm, denoted f, are
taken as average between the droplet surface and the gas properties. The variation of the mass of a droplet is given by

dmd
_
m
dt
1 3
md qd pd
6

The internal droplet temperature is calculated by a nite-conductivity model [30]. The temperature within the droplet, assumed
to be a sphere, is given by

18

Case A
Case B
Case C

T (K)

p (bar)

/0

d (lm)

l (mm)

Fuel

298
298
265

1
1
0.4137

0.66
0.6-6
0.66

5140
5140
5140

225
225
225

n-heptane
n-decane
n-decane

A. Neophytou, E. Mastorakos / Combustion and Flame 156 (2009) 16271640

It is important to mention that the evaporated fuel mixes

instantaneously with the gas and that the source term exists everywhere in space where d > 0. This one-dimensional formulation
could be viewed as an integrated form of the problem over a plane
orthogonal to the ow. This precludes inhomogeneities in the mixture fraction due to evaporation and ames surrounding individual
droplets. Hence, we cannot expect the ame speed enhancement
due to local inhomogeneities discussed in Section 1.

2200
1800

2.3. Chemical mechanisms

1400

d=10 m, o=1

1000

d=50 m, =4

d=50 m, =1
o
o

600
200
5

4
5
3
x 10

y(m)

Fig. 2. Temperature variation through n-heptane ames with l = 5 mm and the

indicated equivalence ratio and droplet size.

0.1
0
0.1

Q/(Qgas,max)

For n-heptane, we used a detailed mechanism [31] developed to

describe high-temperature oxidation and pyrolysis (107 species
and 723 reactions). The mechanism has been validated at atmospheric conditions against extensive experimental data including
premixed laminar ame speed, variable pressure ow reactor,
shock tube ignition delay and jet-stirred reactor species proles
with remarkably good agreement [31]. Low-temperature oxidation
chemistry, however, is not included.
For n-decane, the mechanism of Ref. [32] was used. This mechanism comprised 86 species and 641 reactions and includes
high-temperature oxidation and pyrolysis. It was validated by
the developers at atmospheric pressure for laminar ame speed,
oxidation in ow reactor [32] and for ignition delay at high and
low pressure (down to 0.3 bar). It has also been validated for premixed ame extinction strain rate [33]. To our knowledge, no explicit validation was done for reactant temperatures lower than
300 K. We use this mechanism for relight conditions anyway (i.e.
230260 K), in anticipation of the fact that the chemical reactions
responsible for ame propagation occur at high temperatures and
that these do not change much with the lower initial reactant temperature. The thermodynamic properties were calculated by polynomials whose reported range of validity was above 300 K, but
these were also used for the low-temperature conditions for n-decane. This may introduce an uncertainty in the calculated ame
speed, but it is expected that this uncertainty is small due to the
almost linear dependence of these properties with temperature
at low temperatures.
As a preliminary step before performing spray ame simulations, gaseous ame speeds for the conditions reported by the
mechanism developers were calculated with COSILAB at atmospheric conditions and excellent reproduction of the expected values [31,32] was found.

gas, =1
o

T(K)

1630

0.2

gas, =1

0.3

d=10 m, o=1

0.4

d=18 m, =1
o

0.5

d=50 m, o=1

0.6

d=50 m, o=4

0.7
0.8
0.9
1

0.5

1.5

2
3

x 10

y(m)

Fig. 3. Heat release through n-heptane ames with l = 5 mm and the indicated
equivalence ratio and droplet size.

0.25

gas, o=1
d=10 m, o=1

0.2

d=50 m, o=1

d=50 m, o=4

0.15

3. Results
In this Section, we present the dependence of the ame speed
on various factors for the case denoted A in Table 1. The other cases
are qualitatively similar and are mentioned only when signicant
differences exist. Results specic to cases B and C, concerning the
inuence of the fuel and the ambient conditions, are presented in
Sections 3.3 and 3.4.

0.1

C7H16

0.05

0
5

y(m)

4
5
3
x 10

3.1. Typical spray ame structure

Fig. 4. Reactants mass fractions through n-heptane ames with l = 5 mm and the
indicated equivalence ratio and droplet size.

Before we introduce results regarding the ame speed, we compare, at an overall equivalence ratio /0 = 1, the structure of a gaseous n-heptane ame with two types of n-heptane spray ames:
one, with an initial diameter d = 10 lm, is similar to a gaseous
ame while the other, with d = 50 lm, is more characteristic of
spray combustion. In Figs. 25 we present the variations of the
temperature, the heat release and mass fractions of the main species through the simulated ame. The heat release has been normalised by the maximum value of the heat release of the
stoichiometric gaseous ame. We can see in Fig. 2 that the temperature prole is very close to the gaseous case with d = 10 lm while

it differs substantially with big droplets. Moreover, we can observe

in Fig. 3 that the heat release with small droplets is very similar to
that of the gaseous case, that is with a pronounced thin peak. This
is due to the large extent of fuel evaporated before reaching the
ame. The plots of the mass fractions show that the biggest droplets provide little fuel vapour before the ame front. The resulting
lean mixture reacts less intensively and this explains why the peak
of heat release is lower in the case of the big droplets. Nevertheless,
the large droplets keep evaporating after the homogeneous reaction zone. The fuel is then consumed as soon as it is generated in a

1631

A. Neophytou, E. Mastorakos / Combustion and Flame 156 (2009) 16271640

3.2. n-Heptane ame

0.2

0.15

CO

0.1
gas, o=1
d=10 m, o=1

0.05

d=50 m, =1
o

d=50 m, o=4

0
1

5
3

y(m)

x 10

Fig. 5. CO2 mass fraction through n-heptane ames with l = 5 mm and the indicated
equivalence ratio and droplet size.

droplet burning zone. This can be deduced from Figs. 4 and 5,

where Y O2 and Y CO2 , respectively, decreases and increases slowly
over a relatively long length behind the ame front. Here, the oxidation continues until droplets nish evaporating and nally disappear. The reactions are complete with big initial droplets
although the rate at which they occur is slower. The description
presented here agrees well with the structure of a spray ame presented in Section 1.

Fig. 6 presents the variation of the ame speed with the initial
droplet diameter for three different /0 and for different lengths l
of the primary evaporation zone. For /0 = 1, the general trend is
that the ame speed is reduced as the droplet diameter increases.
As the droplet size falls below 30 lm, the ame speed comes close
to the gaseous ame speed at the same overall equivalence ratio.
This agrees well with previous ndings [2,10,16]. It is also evident
that the ame speed is not a monotonic function of the droplet size
and that a maximum exists at a nite droplet diameter. The corresponding optimum diameter increases from 14 lm to 25 lm when
the residence length l varies from 2 mm to 25 mm. This range of
diameters has been interpreted as a transition range below which
the ame behaves like a gaseous ame [11]. Fig. 6 also compares
our data with a theoretical formula based on considerations of
evaporation rates and chemical reaction rates in the framework
of steady ames and /0 6 1. The expression given by [5] is:

"

C 33 Xqd D232
a2
SL a

8C 1 q ln1 BM S2L;g

o=1
0.5

gaseous flame, =0.6

0

0.4

gaseous flame, =1
0

0.35

SL(m/s)

S (m/s)

0.45

2mm
5mm
15mm
25mm
Theory

0.15

27

D32 is the Sauter Mean Diameter at the start of the preheat zone and
BM the transfer number. a and q are respectively the thermal diffusivity and the density of the gas at 1200 K. SL,g is the gaseous ame
speed at the same overall stoichiometry. For a monodisperse spray
containing no fuel vapour, the liquid fraction X is equal to 1 and

o=0.6
0.2

#0:5

0.1

2mm
5mm
15mm
25mm
Theory

0.3
0.25
0.2
0.15

0.05

0.1
0.05

0
0

20

40

60

80

100

120

140

20

40

d( m)

60

80

100

120

140

d( m)

=4
o

0.5
0.45
0.4

S (m/s)

0.35
0.3
0.25

gaseous flame, 0=1

0.2

2mm
5mm
15mm
25mm

0.15
0.1
0.05
0
0

20

40

60

80

100

120

140

d( m)

Fig. 6. n-Heptane ame speed as a function of the initial droplet diameter for different lengths l. For /0 6 1, the data are compared with the theoretical curve prescribed by
Ballal and Lefebvre [5]. (a) /0 = 0.6, (b) /0 = 1 and (c) /0 = 4.

1632

A. Neophytou, E. Mastorakos / Combustion and Flame 156 (2009) 16271640

C3 = C1 = 1 [34]. An effective evaporation constant was employed

keff 8qaq ln1 BM . keff was plotted for various conditions and
d
was evaluated at 1200 K [6]. For n-heptane, keff = 4.7 mm2/s while
for n-decane, keff = 3.8 mm2/s. The good agreement for d P 30 lm
implies that the basic premise of the theory is substantiated from
the detailed chemistry and detailed transport simulations presented here.
At an overall lean equivalence ratio /0 = 0.6, we observe the
same trend. There is an optimum droplet size that increases from
d = 26 lm to d = 34 lm as l is made longer, see Fig. 6a. As mentioned in the Introduction, enhanced ame speed for nite size
was observed in past studies. However, the phenomenon was often
attributed to diffusion ames surrounding droplets and burning at
optimum stoichiometry. The model we use precludes the occurrence of such diffusion ame, see Section 2.2.
Finally, Fig. 6c also depicts the ame speed against the initial
droplet diameter for /0 = 4. A gaseous ame cannot propagate at
/0 = 4, whereas the droplet ame can. Let SL,0 be the gaseous ame
speed at an overall equivalence ratio /0 = 1. We can infer from the
graph that as the droplet diameter decreases from d = 140 lm, the
ame speed increases. It also reaches a value higher than SL,0 for a
droplet diameter between 50 and 90 lm, depending on the degree
of pre-evaporation. Although this is not supported clearly by the
curves in the graph, we can expect very small droplets to behave
like a gaseous ame and thus to result in very low ame speeds.
Following this reasoning, there should be an optimum droplet size
for which the ame velocity is maximum. As we will explain later,
this optimum droplet size should depend on the length l which affects the degree of evaporation. In Section 4.1 we will relate the
ame speed to an effective equivalence ratio /eff calculated inside
the ame, while in Section 4.2 we will discuss an explanation for
the ame acceleration above SL,0.
Fig. 7 shows the variation of the ame speed with /0 for different droplet diameters at a xed length l = 5 mm. For a xed initial
droplet diameter, the laminar ame speed follows the same trend
as the gaseous ame speed. It rst increases with the equivalence
ratio, reaches a maximum and nally decreases with higher fuel
air ratio. Although the last branch is not present on the graph,
we can expect that if the equivalence ratio /0 becomes extremely
high, the amount of fuel evaporated before the ame will be so
large that it will preclude the propagation of the ame. This statement is supported by an analogous result [10] for ethanol with
smaller droplets and a narrower range of stoichiometry. An important feature to notice is that for a given droplet diameter, the
equivalence ratio giving rise to the highest ame speed may be
higher than 1 and that the maximum ame speed may be higher
than the maximum gaseous ame speed. This supports other re-

Flame speeds calculated with n-decane at atmospheric conditions (Flames B) generally show the same trends as presented
above. However, some differences, which may be attributed to
the lower volatility of n-decane, are worth pointing out. In Fig. 9a
we remark again the inverse dependence of the ame speed on
the initial diameter. However, the diameter below which the ame
speed is very close to the gaseous ame speed is around 15 lm,
which is much smaller than for n-heptane. In fact, the reduction
of the ame speed with larger diameters is stronger in the case
of n-decane. This agrees with results reviewed previously. Furthermore, the value of the laminar ame speed is virtually independent

o=1

L,0

0.4
0.35
0.3
0.25
0.2

1
0.8

SL/SL,0

pure gas
5m
18 m
50 m
100 m
140 m

0.45

3.3. n-Decane ame: atmospheric conditions

0.5

S (m/s)

sults reviewed earlier [10,14]. Furthermore, it is interesting to note

that this plot is related to the one given in Ref. [21] regarding minimum ignition energy against overall equivalence ratio for different SMD: the conditions resulting in minimum ignition energy
correspond to the conditions giving highest ame speeds in the
present calculations.
The control parameter used to assess the inuence of the residence time of the droplets in the primary evaporation zone was
the length l of the zone (i.e. the distance between the left boundary of the calculation domain and the preheat zone). Fig. 6 shows
that as the length l is increased the ame speed is enhanced,
especially when the droplet diameter is small. We also observe
that when an optimum diameter exists, its value is markedly affected by l. The trend of ame speed with l is consistent with
experimental data (see, for example, Refs. [5,6]) that show an increase in ame speed with increasing degree of pre-evaporation
X at the preheat zone. The amount of vapour created in the
length l depends on the residence time l/SL, where SL is the ame
speed of the particular ame we are studying. For each droplet
size and fuel studied, a reference evaporation timescale, sevap,
can be estimated from the single-droplet evaporation theory in
stagnant surroundings (e.g. [35]). The ratio (l/SL)/sevap is related
to the amount of fuel evaporated before the preheat zone and
can act qualitatively as a substitute for the vapour fraction X.
For the n-heptane ames with /0 = 1, this timescale ratio is
shown in Fig. 8 for different values of the diameter and compared
with the theoretical curve from Eq. 27. With our results the diameter indicated is that at the beginning of the primary evaporation
zone while in Eq. 27, D32 refers to the diameter at the beginning
of the preheat zone. In spite of that, the theoretical curve yields a
good agreement with our results. Our solution is consistent with
the nding that the presence of fuel vapour in a mist is highly
conducive to ame propagation [5].

30 m

0.6

50 m
0.4

100 m

0.15
0.2

0.1

140 m

0.05
0

0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6

0.25

0.5

0.75

Fig. 7. n-Heptane ame speed as a function of the overall equivalence ratio for
different initial droplet diameters and l = 5 mm.

Fig. 8. Inuence of X and d on n-heptane ame speed based on the theoretical

curve prescribed by Ballal and Lefebvre [5]. Data from the present simulations are
included where the ratio (l/SL)/sevap replaces X; see Section 3.2.

1633

A. Neophytou, E. Mastorakos / Combustion and Flame 156 (2009) 16271640

0.5

0.5

0.45

0.45

gaseous flame, =0.6

o

0.4

gaseous flame, o=1

o=4

0.2

0.3

o=1

0.25

0.15

2mm
5mm
15mm
25mm

o=4

=1

gaseous flame, =1

0.35

S (m/s)

2mm
5mm
15mm
25mm

0.3
0.25

S (m/s)

0.35

gaseous flame, o=0.6

0.4

0.2
0.15

0.1

0.1

=0.6

0.05

o=0.6

0.05

0
0

20

40

60

80

100

120

140

20

d( m)

40

60

80

100

120

140

d( m)

Fig. 9. n-Decane ame speed at various /0 as a function of the initial droplet diameter for different lengths l. The gaseous ame does not exist at /0 = 4. (a) Atmospheric
conditions and (b) relight conditions.

of the length l of the primary evaporation zone. This suggests that

no control of the ame speed can be done by extending the primary evaporation zone consistent with the little evaporation
achieved with n-decane due to its low saturated vapour pressure.
Fig. 10a shows the variation of the n-decane ame speed with
the equivalence ratio for different droplet diameters at a xed
length l = 5 mm. The same trends seen with n-heptane are followed, but with the heavy fuel the ame speed does not exceed
SL,0. Fig. 11 compares our results with the theoretical curve presented in Section 3.2 and experimental results collected by [5]
for heavy fuel oil. Our results are realistic and show that pure ndecane represents well, at least qualitatively, the inuence of droplet size on ame propagation in commercial fuel.
3.4. n-Decane ame: relight conditions
Fig. 12 shows the gaseous ame speed versus /0 at atmospheric
condition and two relight conditions, (1) p = 41.37 kPa, T = 265 K
and (2) p = 30 kPa, T = 230 K. The former corresponds to the typical
inlet stagnation conditions of a combustor at an altitude around
9000 m, while the latter corresponds to the values of the standard
atmosphere at the same altitude. Both are sometimes used by gas
turbine engineers to denote the conditions inside the combustor if
the ame has extinguished and must be relighted. It is readily seen
that although relight conditions impose low temperatures detri-

mental to the reaction rates, the reduction of the ame speed is

not large (less than 20%). This suggests that the negative effect of
the low temperature is overcome by the benets of the reduction
in pressure, with the typical dependence p0.5 in pure gas [36]. Calculations that we performed at 298 K showed a pressure dependence p0.22 in fair agreement with the exponent 0.3 found by
[5] for Diesel oil.
Now, we consider spray ames with p = 41.37 kPa, T = 265 K
(Flames C). Fig. 9b shows that the trends described for atmospheric
conditions remain unchanged. The effect of the residence length l
on the ame speed is also presented in Fig. 9b, which indicates that
l has very little effect, consistent with the very small inuence of l
in the case of n-decane at atmospheric conditions. Finally, it is
noteworthy that the ame speed variation with overall equivalence ratio /0 at relight conditions is quantitatively comparable
to the one at atmospheric conditions, Fig. 10b. As with normal conditions, we notice that the ame speed at rich /0 can exceed the
corresponding gaseous ame speed. However, the spray ame
speed for n-decane does not exceed SL,0; that phenomenon seems
to be specic to n-heptane, which is more volatile.
4. Discussion
In this Section, we give some additional interpretations for
the trends introduced previously. First, we discuss the effective

0.5

0.5

0.45

0.45

0.4

0.4
0.35

L,0

0.25
0.2
0.15
0.1
0.05

pure gas
5m
10 m
30 m
50 m
100 m

0.3

S (m/s)

SL(m/s)

0.35

L,0

pure gas
5m
15 m
30 m
50 m
100 m

0.3
0.25
0.2
0.15
0.1
0.05

0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6

Fig. 10. n-Decane ame speed as a function of the overall equivalence ratio for different initial droplet diameters and l = 5 mm. (a) Atmospheric conditions and (b) relight
conditions.

A. Neophytou, E. Mastorakos / Combustion and Flame 156 (2009) 16271640

S /S

L,0

1634

Numerical
Theory
Experiment

1.1
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0

20

40

60

80

100

120

140

d( m)
Fig. 11. Comparison between the theoretical curve, experiments with heavy fuel oil
(SL,0 = 0.41 m/s) [5] and our results with n-decane at an /0 = 1 and l = 2 mm.

0.5
P=100kPa,T=298K
P=41.37kPa,T=265K
P=30kPa,T=230K

0.45
0.4

S (m/s)

0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
0

0.5

1.5

Fig. 12. Flame speed of gaseous n-decane at three different conditions of pressure
and initial temperature.

equivalence ratio /eff, and then we examine in detail the ame

structure. An explanation for the increase of the ame speed above
SL,0 for rich sprays is discussed, as well as the optimum diameter
that maximises SL.
4.1. Effective equivalence ratio
As suggested by previous studies [10,7,14], the equivalence ratio in the gas at the front of the reaction zone should explain, at
least partly, the different ame speeds observed under various conditions. The idea is that the closer the effective equivalence ratio
/eff is to unity, the faster the reaction rates and the higher the ame
speed. We can expect /eff to be inuenced, on the one hand by the
preheat zone where the evaporation rate is accelerated, and on the
other hand by the length l of the primary evaporation zone where
droplets evaporate in the cold gas.
The effective equivalence ratio /eff is here dened as the equivalence ratio, based on the mass fraction of the carbon element in
the gas, at the location where the heat release becomes a maximum. For purely gaseous ames, this location is virtually identical
to the location where the temperature prole has an inection
point and can hence be identied as the inner layer equivalence
ratio [37]. For droplet ames, however, the temperature is affected
by the evaporation and hence a small shift between the peak heat
release and the inection point of the temperature prole has been
observed. The equivalence ratio at the peak heat release provides a
more realistic representation of the effective combustion fuelair
ratio. Note that the carbon-to-oxygen ratio in the gas varies very
sharply in space due to the evaporation and differential diffusion
and hence /eff should be treated as an approximate concept.

Fig. 13 shows /eff for the n-heptane ames as a function of the

initial droplet diameter for the lean, stoichiometric and rich cases
and for different lengths l. We observe that as the initial SMD is reduced, the effective equivalence ratio increases towards the overall
equivalence ratio /0 due to the higher evaporation rate. In fact, /eff
is brought to a value very close to /0 when d 6 30 lm, which is in
the transition range. Furthermore, we can notice that as the length
l is increased, /eff increases as well. Under atmospheric conditions,
n-heptane droplets have a high evaporation rate and produce substantially more fuel over a longer primary evaporation zone. Those
trends are clearly correlated with the previous results on the ame
speed, Fig. 6. By noticing that when /eff is closer to one, the ame
speed tends to be larger, we can infer that /eff inuences to some
extent the propagation rate. For example, under lean or stoichiometric conditions, as the droplet diameter goes below 10 lm, almost
all the liquid fuel has evaporated before reaching the ame and
thus /eff is equal to /0. This results in a ame speed close to, but
slightly lower than, the gaseous laminar speed, probably because
some of the combustion energy has been spent into evaporating
all the liquid.
Furthermore, we saw in Section 3.2 that, for /0 = 4, the ame
speed can be high in a spray with large droplets although the overall equivalence ratio is well above the rich ammability limit. We
observe in Fig. 13 that the conditions which involve high ame
speeds correspond to the ones resulting in /eff close to 1. This occurs for d higher than 30 lm. Although this is not supported by
the graphs, very small droplets should result in very large /eff,
which is not conducive to ame propagation. Therefore, at rich
equivalence ratio, moderately large droplets are benecial to ame
speed. Still, too large diameters result in too lean effective equivalence ratio and thus slower ames. By comparing Figs. 13 and 6, we
observe that the highest ame speed in the rich case corresponds
to 0.7 6 /eff 6 0.9. We can expect that the optimum conditions
are a combination of diameter, length l and /0 such that /eff is close
to one, consistent with analytical [7,38] and experimental [10]
conclusions. Since the extent of evaporation is affected by l, the
optimum diameter varies with l.
Fig. 14 presents /eff versus SMD for n-decane ames at relight
and atmospheric conditions. Since n-decane is a heavy fuel with
a low saturated vapour pressure, its evaporation rate under atmospheric conditions is very weak and this explains the small inuence of l on /eff and in turn of l on the ame speed. This also
explains why, under lean and stoichiometric condition, the ame
speed is more easily reduced by an increase in diameter with
n-decane than with n-heptane. The evaporation rate is so slow that
much smaller droplets are needed to reach the same /eff. Finally, it
accounts for the fact that the laminar ame has a high value for
diameters down to 35 lm in the rich case /0 = 4. Even if the spray
is rich, the low evaporation rates help keeping /eff below one.
The effect of the residence length l on /eff for the n-decane ame
at relight conditions is shown in Fig. 14b. It is apparent that all the
curves merge into a single one implying that l has very little effect.
We have already noted the poor inuence of l in the case of n-decane at atmospheric conditions. Our results show that this is even
more pronounced at relight conditions, and imply that the vapour
generated by the heavy fuel in the low temperature primary evaporation zone is negligible. Hence, most of the fuel that burns in the
homogeneous reaction zone evaporates in the preheat zone. We
could expect /eff to be much lower at relight conditions than at
atmospheric conditions, but interestingly, this is not the case as
shown in Fig. 14. In the preheat zone, where the droplet temperature approaches the boiling temperature, the low pressure has a
very positive effect on the evaporation rate since it increases the
transfer number B through the ClausiusClapeyron equation.
Thanks to this phenomenon, the low pressure also has a positive
effect on the ame speed, as long as /eff 6 1, since it increases

1635

A. Neophytou, E. Mastorakos / Combustion and Flame 156 (2009) 16271640

o=0.6

o=1

1.2
1

1.2

2mm
5mm
15mm
25mm

2mm
5mm
15mm
25mm

0.8

eff

eff

0.8

0.6

0.6

0.4

0.4

0.2

0.2

0
0

20

40

60

80

100

d( m)

120

140

20

40

60

80

d( m)

100

120

140

o=4
1.2

2mm
5mm
15mm
25mm

eff

0.8
0.6
0.4
0.2
0
0

20

40

60

80

100

120

140

d( m)

Fig. 13. Effective equivalence ratio for n-heptane as a function of the initial droplet diameter for different lengths l. (a) /0 = 0.6, (b) /0 = 1 and (c) /0 = 4.

1.2

0=1

=4

0.8

1.2

2mm
5mm
15mm
25mm

=4
0

0.8

2mm
5mm
15mm
25mm

0=1

eff

eff

0.6

0.6

=0.6
0

0.4

0.4

=0.6
0

0.2

0.2

0
0

20

40

60

80

100

120

140

d( m)

20

40

60

80

100

120

140

d( m)

Fig. 14. Effective equivalence ratio for n-decane as a function of the initial droplet diameters for different lengths l. (a) Atmospheric conditions and (b) relight conditions.

/eff. As an overall result, the temperature and the pressure have

two competitive effects. On one hand, the low temperature decreases the ame speed and reduces the amount of fuel evaporated, /eff. On the other hand, the low pressure increases the
ame speed and enhances the evaporation rate and thus /eff. This
twofold effect of the pressure agrees with the nding of Ballal
and Lefebvre [5], who observed that for lean mixtures the ame
speed has a stronger pressure dependence in a spray compared
to a pure gas.

In Fig. 15 we plotted the ame speed normalised by SL,0 against

/eff for all the data collected. Although the parameter /eff is useful
in explaining different phenomena observed, there is no complete
correlation between the ame speed and /eff. The large scatter observed implies that other phenomena affect the ame speed. For
instance, for the same /eff and d, if /0 is different, the ame speeds
are different. It is also evident that for a given /eff, the ame speed
is practically always above the gaseous ame speed at the same value of equivalence ratio. Another interesting observation is that the

1636

A. Neophytou, E. Mastorakos / Combustion and Flame 156 (2009) 16271640

1.4

1.4

1.2

1.2

SL/SL,0

SL/SL,0

1
0.8
0.6

1
0.8
0.6

0.4

0.4

0.2

0.2

0
0

0
0.2 0.4 0.6 0.8

1.2 1.4 1.6 1.8

0.2 0.4 0.6 0.8

eff

1.2 1.4 1.6 1.8

eff

1.4
1.2

SL/SL,0

1
0.8
0.6
0.4
0.2
0
0

0.2 0.4 0.6 0.8

1.2 1.4 1.6 1.8

eff

Fig. 15. Normalised ame speeds as a function of /eff. (a) n-Heptane, (b) n-decane atmospheric and (c) n-decane, relight. Line: gas. Symbols: spray.

ame speed is nite at very low /eff, even beyond the lean ammability limit. Although the governing equations used here do not
fully account for ammability limits as there is no heat loss mechanism, the code converges for gaseous ames only in a range of
equivalence ratios and gives very low speeds (SL/SL,0 < 0.2) at
/0  0.5 and /0  2. The spray ames, in contrast, show high ame
speeds for /eff as low as 0.4. This widening of the ammability limits in sprays has been attributed to inter-droplet inhomogeneities,
where locally the mixture may reach ammable values [2,17,9]
and hence sustain ame droplet-to-droplet propagation. In the
next section, we provide an alternative explanation for this phe-

x 10

YOH

3
gas, =1
o

d=10 m, =1

d=50 m, =1
o

d=50 m, =4

0
1

y(m)

5
3

x 10

Fig. 16. OH mass fraction through n-heptane ames with l = 5 mm and the
indicated equivalence ratio and droplet size.

nomenon, that also accounts for the enhanced ame speed in

sprays above the maximum gaseous ame speed (i.e. the fact that
SL/SL,0 can be above unity in Fig. 15).
4.2. Structure of the ame and inuence of chemistry
In this Section, we present some details on the chemical aspects
of the spray ames, as revealed from the present detailed chemistry calculations. All the examples are presented for n-heptane and
l = 5 mm, but the conclusions are valid for other fuels and lengths.
Three main structures of spray have been observed which are
variants of Fig. 1. In the rst one, almost all the droplets evaporate
before reaching the ame front and /eff is close to /0. The ame is
very similar to a gaseous ame and we observe a thin reaction
zone, as depicted in Figs. 24, 5, 17 for a diameter d = 10 lm and
/0 = 1.
On the same graphs, we can notice a second structure in the
case d = 50 lm and /0 = 1 where the effective equivalence ratio is
very low. We have already commented this structure in Section
3.1: little oxidiser is consumed in the rst reaction zone and the
rest is consumed in a droplet burning zone over a length of a few
millimetres. This is illustrated by Fig. 16, which shows that YOH
keeps increasing over a long length, denoting a long oxidation
zone. Hence, although /eff is low, the droplet burning zone is wide
and provides heat release. This explains why the ame speed may
be relatively high even with /eff as low as 0.4.
Finally, a third structure has been observed for rich spray ames
where 0.6 6 /eff 6 1 and where the droplets still have a signicant
size at the inner layer of the reaction zone. As a result, the oxidiser
is entirely consumed at the reaction zone and in a short droplet
burning zone while droplets still exist behind it. There, instead of

1637

A. Neophytou, E. Mastorakos / Combustion and Flame 156 (2009) 16271640

3

x 10

gas, o=1

d=50 m, =1

d=10 m, =1

YC

0.03

gas, =1

2 2

0.02

d=10 m, =1

d=50 m, =4

d=50 m, o=1

d=50 m, =4

0.01

1
0
1

0
1

y(m)

5
3

y(m)

x 10

x 10

0.05

0.04

gas, =1

C H

0.03

d=10 m, =1

C H

d=10 m, =1

2 4

gas, =1

x 10

3 6

d=50 m, =1
o

d=50 m, =1

d=50 m, =4

d=50 m, =4
o

0.02

0.01

0
1

0
1

x 10

y(m)

y(m)

5
3

x 10

x 10

gas, =1
o

d=10 m, o=1
d=50 m, =1

Y1C

d=50 m, o=4

4 8

0
1

y(m)

5
3

x 10

Fig. 17. Mass fraction of intermediate species through n-heptane ames with l = 5 mm and the indicated equivalence ratio and droplet size. (a) H2, (b) C2H2, (c) C2H4, (d) C3H6
and (e) 1  C4H8.

being oxidised, the fuel evaporated is pyrolysed. To illustrate this,

we compare the three main structures: the gaseous case at stoichiometry, the ame with a long droplet burning zone (d = 50 lm and
/0 = 1), and the third structure represented by d = 50 lm, /0 = 4. In
Fig. 17, mass fractions of hydrogen, acetylene, ethylene and higher
hydrocarbons show that in the third structure products of pyrolysis and subsequent reactions keep being formed after the main
heat release zone.
The above arguments are substantiated by examination of the
rates of direct oxidation and of pyrolysis for the fuel. Fig. 18 shows
the rates of the elementary reactions C7H16 + O , C7H15 + OH and

C7H16 + OH , C7H15 + H2O; these elementary reactions have the

highest rate of all n-heptane H abstraction reactions by OH and
O. In the same graph, the rates of oxidation of smaller hydrocarbons (CH2O + OH , HCO + H2O and HCO + O2 , CO + HO2) are presented. The rst reaction is sometimes considered responsible for a
signicant part of the heat release in hydrocarbon combustion
[39]. Finally, the rates of the pyrolysis reactions with the highest
values, C7H16 , C5H11 + C2H5 and C7H16 , pC4H9 + nC3H7, are depicted in Fig. 19. From these graphs, we notice that: in the gaseous
case, all reactions are concentrated in a small region and they become negligible after 0.7 mm; in the case d = 50 lm and /0 = 1,

1638

A. Neophytou, E. Mastorakos / Combustion and Flame 156 (2009) 16271640

400

700

gas, o=1
d=18 m, =1
o

w(mol/m3/s)

300

d=50 m, o=4
200

w(mol/m3/s)

d=50 m, o=1

gas, o=1

600

d=18 m, =1

500

d=50 m, o=1

d=50 m, o=4

400
300
200

100

100
0

0
2

12

17

12

17

x 10

y(m)

1600

x 10

y(m)

gas, o=1
d=18 m, =1

2800

gas, o=1

2400

d=50 m, =1

2000

d=50 m, =4

d=18 m, =1

o
o

d=50 m, =4

o
o

w(mol/m /s)

w(mol/m3/s)

d=50 m, =1

1200

800

1600
1200
800

400

400
0

0
2

12

17

12

17
4

x 10

y(m)

x 10

y(m)

Fig. 18. Rates of elementary reactions through n-heptane ames with l = 5 mm and the indicated equivalence ratio and droplet size. (a) C7H16 + O , C7H15 + OH,
(b) C7H16 + OH , C7H15 + H2O, (c) CH2O + OH , HCO + H2O and (d) HCO + O2 , CO + HO2.

200

200

gas, o=1
d=18 m, =1
o

d=50 m, o=1

100

50

150

w(mol/m3/s)

d=50 m, o=4

w(mol/m /s)

150

gas, =1

100

d=18 m, o=1
d=50 m, o=1

50

d=50 m, o=4

0
2

y(m)

12

17
4

x 10

y(m)

12

17
4

x 10

Fig. 19. Pyrolysis reaction rate through n-heptane ames with l = 5 mm and the indicated equivalence ratio and droplet size. (a) C7H16 , C5H11 + C2H5 and (b)
C7H16 , pC4H9 + nC3H7.

reactions involving oxygen atoms extend beyond 1 mm. In this

structure, the reaction CH2O + OH , HCO + H2O indicates that
combustion occurs over a long length in a droplet burning zone;
for the case d = 50 lm, /0 = 4, reactions involving oxidant are insignicant after 1 mm implying that very little oxidiser remains and
the droplet burning zone is short. This is consistent with the sudden decrease of OH seen in Fig. 16. After 1 mm, all reactions seem
to be negligible except pyrolysis reactions. These reactions keep
occurring over a long length because the droplets are large and sur-

vive far beyond the ame. The temperature and heat release
proles presented in Figs. 2 and 3 are consistent with the endothermic nature of pyrolysis: the temperature prole presents a
negative slope behind the ame front, coincident with the change
of the heat release sign.
In Fig. 20 we present the net formation rate of some products of
the decomposition of fuel: H2 and C2H2, both of which are highly
reactive. In the case d = 50 lm, /0 = 4, the production rates of these
species are very high in the pyrolysis zone because large amounts

1639

A. Neophytou, E. Mastorakos / Combustion and Flame 156 (2009) 16271640

30

20

20

w(mol/m /s)

w(mol/m /s)

10
10
0
gas, =1
o

10

d=18 m, =1
o

0
gas, =1
o

d=18 m, o=1

10

d=50 m, =1

d=50 m, =1

20

d=50 m, =4

d=50 m, =4

30

12

y(m)

17

20

12

17
4

y(m)

x 10

x 10

Fig. 20. Formation rate of some products of fuel decomposition through n-heptane ames with l = 5 mm and the indicated equivalence ratio and droplet size. (a) C2H2 and
(b) H2.

of fuel are being evaporated and subsequently pyrolysed. The rates

are virtually zero in the other cases after 1 mm. It can be inferred
that the existence of a pyrolysis zone leads to a build up of very
reactive intermediate species, see Fig. 17. The fact that the ame
speed of these species (H2, C2H2 and C2H4) is high, if they are considered as separate fuels, implies the possibility of acceleration of
the n-heptane ame because we may expect that the high positive
gradient of the mass fractions of these species induces a back diffusion towards the oxidation zone where the reaction rates are enhanced. Hence, we can roughly consider that very rich spray ames
are qualitatively equivalent to hydrogen- and acetylene- and ethylene-enriched lean or stoichiometric higher hydrocarbon ames.
In order to nally relate explicitly the formation of these species
to the high ame speed observed in this case, we carried out a
sensitivity analysis of ame speed for the cases presented in
Fig. 21. It was found that, for the gaseous ame, the most sensitive
reactions were the following: H + O2 , O + OH; OH + CO , H +
CO2; H + OH + M , H2O + M; HCO + M , H + CO + M and OH
CH3 () CH2 H2 O. It was also found that the same elementary
reactions are the most sensitive in the spray ames, but also that
as the droplet size increases, the normalized sensitivity goes down.
In the graph, we also included the elementary reactions H2 + OH ,
H2O + H; C3H6 + OH , aC3H5 + H2O; C2H5 + O2 , HO2 + C2H4, which
are very sensitive in the stoichiometric gaseous case and for

d = 50 lm, /0 = 4. Many of these important elementary reactions

involve species generated from pyrolysis. Fig. 22 shows that the
reaction rates of some of these key reactions are intensied in
the rich spray ames. Therefore, the presence of a pyrolysis zone
improves the ame propagation rate. This would explain the very
high ame speed observed in the case d = 50 lm, /0 = 4. This ame
structure is achieved in rich sprays where all the oxidiser is consumed in a relatively short reaction zone, and with surviving droplets in the post ame region.
Finally, we discuss here the existence of an optimum diameter
for /0 = 1 and /0 = 0.6. As observed in Fig. 6a and b, SL increases
with d until about 30 lm and then decreases as d increases further.
The decrease from the optimum as d increases can be understood
due to the decreasing /eff, see Figs. 13a and 13b. The initial increase
from very small d cannot be explained simply through /eff, because
/eff is relatively independent of d, but can be understood by the following argument. For a gaseous
premixed ame, simplied theory
p
R
predicts the relation SL  a Qdy, which states that the ame
speed is increased by a higher thermal diffusivity and a higher integrated heat release Q [36]. Although not explicitly presented for
the sake of brevity, the transport properties evaluated by the code
show that in the preheat zone, the thermal diffusivity of the gas is
higher with the optimum diameter than with a smaller diameter.
This is a rst factor contributing to the existence of an optimum
diameter. In Fig. 13, we notice that the value of /eff associated with
the optimum diameter is very close to or equal to /0. In the case

10000

H+O O+OH
2

3
w(mol/m /s)

8000
6000

H +OHH O+H
2

4000
HCO+MH+CO+M

2000
0
0.2

0.7

1.2

y(m)

Fig. 21. Normalised sensitivity coefcient of the velocity on reaction rates. The
particular reactions selected have the highest degree of sensitivity.

1.7
3
x 10

Fig. 22. Rates of the indicated elementary reaction, selected as reactions with high
degree of sensitivity through n-heptane ames with l = 5 mm and the indicated
equivalence ratio and droplet size. Lines: gas, /0 = 1. Circles: d = 50 lm, /0 = 4.

1640

A. Neophytou, E. Mastorakos / Combustion and Flame 156 (2009) 16271640

d = 18 lm, /0 = 1, this implies that a large part of oxidiser is consumed in the homogeneous reaction zone while droplets still exist
behind it. Furthermore, Fig. 19 shows that behind the homogeneous reaction zone, the fuel evaporated by droplets is intensively
pyrolysed eventually resulting in non-negligible generation of H2,
see Fig. 20. Although this is not a proper pyrolysis zone because
combustion reactions still occur, see Fig. 18, we can infer that a
pool of intermediates such as H2 is created. Those reactive species
can then diffuse back and enhance the reaction rates, as in the
d = 50 lm, /0 = 4 ame. To corroborate this statement, we calculated the integrated heat release for the case d = 18 lm, /0 = 1
and found a value of 5.32  107 W/m3 which was higher than both
cases. The
the d = 10 lm, /0 = 1 and the gaseous stoichiometric
p
R
a
overall result
was
that
the
ratio
between
Qdy
for
d = 18 lm,
p
R
/0 = 1 and a Qdy for d = 10 lm, /0 = 1 was 1.07, while the ratio
of the ame speeds was 1.06. This close agreement supports the
fact that incomplete evaporation at the reaction zone may be benecial to the ame propagation.

ditions because the low pressure compensated for the detrimental

effect of low temperature.
Further numerical work on ame propagation should focus on
three-dimensional unsteady situations where the droplet-to-droplet ame propagation mechanism can be simulated. The present
data show that the impact of detailed chemistry on the propagation speed of spray ames is very important and provides an additional explanation for the high ame propagation speeds observed
in sprays under some conditions.
Acknowledgments
This work has been partly funded by Rolls-Royce Group plc. We
thank Dr. H. Ernst of Rotexo for his assistance with the RUN1DL
code and Prof. R.W. Bilger for useful discussions on the effective
equivalence ratio concept.
References

5. Conclusions
Calculations of one-dimensional laminar ames in sprays of
n-heptane and n-decane under atmospheric and relight conditions
have been presented. The effects of droplet diameter, overall
equivalence ratio and residence time before the ame zone were
examined. The trends were only marginally correlated with the
effective equivalence ratio /eff, based on the total gaseous carbon
mass fraction at the location of maximum heat release rate. /eff
was close to the total equivalence ratio /0 for small droplets and
long residence times. High ame speeds were more likely to occur
when /eff was close to unity. This implies that the maximum ame
speed is achieved with small diameters and long residence time
under lean conditions. Under rich conditions, the ame speed
was greater for relatively large droplet sizes that caused /eff to
be near unity. This was in agreement with previous studies.
Three different structures of a spray ame were observed. In the
rst one, droplets evaporated completely before the ame front
and the structure was identical to a gaseous ame. In the second,
droplets evaporated very little so that /eff was low. Little oxidiser
reacted in the homogeneous reaction zone and the rest was consumed in a long droplet burning zone providing heat release. The
third ame structure occurred for rich sprays where /eff was close
to unity and the droplets large enough. In this case, the oxidiser
was consumed in a relatively short reaction zone. The fuel evaporated behind it by the surviving droplets was pyrolysed and the
reactive species generated (mainly H2, C2H2, C2H4) diffused back
towards the oxidation zone and enhanced the reaction rates. This
could result in enhanced ame speed, above the maximum gaseous
ame speed even for /0 = 4.
The same trends were followed with the less volatile fuel,
n-decane. However, the low evaporation rate generally induced a
lower /eff than with n-heptane at atmospheric conditions. At high
altitude relight conditions, the residence time had no effect on /eff
and the ame speed. Despite the low temperature, the value of /eff
was not markedly affected by the relight conditions thanks to the
low pressure that enhanced the evaporation rate in the preheat
zone. The ame speed was comparable to that at atmospheric con-

[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
[37]
[38]
[39]

A.H. Lefebvre, Gas Turbine Combustion, Taylor and Francis, 1998.

J.H. Burgoyne, L. Cohen, Proc. Roy. Soc. Lond. 225 (1954) 375392.
J.A. Browning, T.L. Tyler, W.G. Krall, J. Ind. Eng. Chem. 49 (1957) 142147.
C.E. Polymeropoulos, Combust. Sci. Technol. 9 (1974) 197207.
D.R. Ballal, A.H. Lefebvre, Proc. Combust. Inst. 18 (1981) 321328.
G.D. Myers, A.H. Lefebvre, Combust. Flame 66 (1986) 193.
I. Silverman, J.B. Greenberg, Y. Tambour, Combust. Flame 93 (1993) 97118.
F. Atzler, M. Lawes, Burning velocities in droplet suspensions, in: ILASS-Europe
98, Manchester, 68 July, 1998.
Y. Nunome, S. Kato, K. Maruta, H. Kobayashi, T. Niioka, Proc. Combust. Inst. 29
(2002) 26212626.
S. Hayashi, S. Kumagai, T. Sakai, Combust. Sci. Technol. 15 (1976) 169177.
K. Chan, C. Jou, Fuel 67 (1988) 12231227.
Y. Mizutani, T. Nishimoto, Combust. Sci. Technol. 6 (1972) 110.
C.E. Polymeropoulos, S. Das, Combust. Flame 25 (1975) 247257.
H. Nomura, M. Koyama, H. Miyamoto, Y. Ujiie, J. Sato, M. Kono, S. Yoda, Proc.
Combust. Inst. 28 (2000) 9991005.
J.B. Greenberg, I. Silverman, Y. Tambour, Combust. Flame 113 (1998) 271273.
S.K. Aggarwal, W.A. Sirignano, Combust. Flame 62 (1985) 6984.
C.E. Polymeropoulos, Combust. Sci. Technol. 40 (1984) 217232.
M. Mikami, H. Oyagi, N. Kojima, Y. Wakashima, M. Kikuchi, S. Yoda, Combust.
Flame 146 (2006) 391406.
M. Zhu, B. Rogg, Meccanica 31 (1996) 177193.
S.K. Aggarwal, Prog. Energy Combust. Sci. 24 (1998) 565600.
A.M. Danis, I. Namer, N.P. Cernansky, Combust. Flame 74 (1988) 285294.
A.K. Singh, C.E. Polymeropoulos, Proc. Combust. Inst. 21 (1988) 513519.
D.L. Dietrich, N.P. Cernansky, M.B. Somashekara, I. Namer, Proc. Combust. Inst.
23 (1991) 13831389.
S.K. Aggarwal, W.A. Sirignano, Proc. Combust. Inst. 20 (1985) 17731780.
T. Marchione, S.F. Ahmed, E. Mastorakos, Combust. Flame 156 (2009) 166
180.
P. Dagaut, M. Cathonnet, Prog. Energy Combust. Sci. 32 (2006) 4892.
Rotexo-Softpredict-Cosilab, GmbH and Co. KG Bad Zwischenahn (Germany),
Cosilab Collection, Version 2.1.0, 2008. Available from <www.SoftPredict.com>.
N.-H. Chen, B. Rogg, K.N.C. Bray, Proc. Combust. Inst. 24 (1992) 15131521.
B. Abramzon, W.A. Sirignano, Int. J. Heat Mass Transfer 32 (1989) 16051618.
W.A. Sirignano, Prog. Energy Combust. Sci. 9 (1983) 291322.
M. Chaos, A. Kazakov, Z. Zhao, F.L. Dryer, Int. J. Chem. Kinet. 39 (2007) 399
414.
Z. Zhao, J. Li, A. Kazakov, F.L. Dryer, S.P. Zeppieri, Combust. Sci. Technol. 177
(2005) 89106.
K. Kumar, C.-J. Sung, Combust. Flame 151 (2007) 209224.
D.R. Ballal, A.H. Lefebvre, Proc. Combust. Inst. 18 (1981) 17371746.
S.R. Turns, An Introduction to Combustion: Concepts and Applications, second
ed., McGraw-Hill, New York, 2000.
I. Glassman, Combustion, Academic Press, 1996.
J. Gottgens, F. Mauss, N. Peters, Proc. Combust. Inst. 24 (1992) 129135.
T.H. Lin, C.K. Law, S.H. Chung, Int. J. Heat Mass Transfer 31 (1988) 10231034.
P.H. Paul, H.N. Najm, Proc. Combust. Inst. 27 (1998) 4350.