Original Research Article

International Journal of Spray and

Combustion Dynamics
0(0) 1–19
Numerical modeling of soot formation in ! The Author(s) 2016
Reprints and permissions:
a turbulent C2H4/air diffusion flame sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/1756827716638814
scd.sagepub.com

Manedhar Reddy Busupally and Ashoke De

Abstract
Soot formation in a lifted C2H4-Air turbulent diffusion flame is studied using two different paths for soot nucleation and
oxidation; by a 2D axisymmetric RANS simulation using ANSYS FLUENT 15.0. The turbulence-chemistry interactions are
modeled using two different approaches: steady laminar flamelet approach and flamelet-generated manifold. Chemical
mechanism is represented by POLIMI to study the effect of species concentration on soot formation. P1 approximation is
employed to approximate the radiative transfer equation into truncated series expansion in spherical harmonics while
the weighted sum of gray gases is invoked to model the absorption coefficient while the soot model accounts for
nucleation, coagulation, surface growth, and oxidation. The first route for nucleation considers acetylene concentration
as a linear function of soot nucleation rate, whereas the second route considers two and three ring aromatic species as
function of nucleation rate. Equilibrium-based and instantaneous approach has been used to estimate the OH concen-
tration for soot oxidation. Lee and Fenimore-Jones soot oxidation models are studied to shed light on the effect of OH
on soot oxidation. Moreover, the soot-radiation interactions are also included in terms of absorption coefficient of soot.
Furthermore, the soot-turbulence interactions have been invoked using a temperature/mixture fraction-based single
variable PDF. Both the turbulence-chemistry interaction models are able to accurately predict the flame liftoff height, and
for accurate prediction of flame length, radiative heat loss should be accounted in an accurate way. The soot-turbulence
interactions are found sensitive to the PDF used in present study.

Keywords
Soot, radiation, soot–turbulence interaction
Date received: 11 October 2014; accepted: 20 June 2015

chemical reactions by detailed mechanisms to resolve

Introduction
Since the beginning of human history, combustion has
a significant role in our daily life. Modern society is
heavily dependent on the transformation of chemical
energy in fossil fuels to thermal energy by combustion
process. However, generation of pollutants such as
nitrogen oxides (NOx) and soot results in adverse
effect on human health. It also results in change in
global environmental pattern due to greenhouse effects.
The design of gas turbine and reciprocating engines are
to be made more efficient and low in pollutant emis-
sions.1,2 Combustion is a complex phenomenon and
generally involves complex chemical reactions along
with heat and mass transfer. Turbulent combustion
involves complex interactions between turbulent trans-
ports, kinetic rate of reactions, radiative heat transfer.
Combustion simulations require representation of
the phenomenon such as extinction and ignition, emis-
sions like NOx, CO and soot.3 Accurate inclusion of
radiative heat transfer is also important; as its fourth
power dependence on temperature makes it a dominant
mode for heat transfer and effects NOx emissions
through mechanism of thermal NO. The incomplete
combustion of hydrocarbons results in the formation
of condensed carbon particles known as soot and an
indicator of poor fuel utilization. In diffusion flames,
Department of Aerospace Engineering, Indian Institute of Technology
Kanpur, Kanpur, India
Corresponding author:
Ashoke De, Department of Aerospace Engineering, Indian Institute of
Technology Kanpur, Kanpur 208016, India.
Email: ashoke@iitk.ac.in.

Creative Commons CC-BY-NC: This article is distributed under the terms of the Creative Commons Attribution-NonCommercial 3.0 License (http://www.
creativecommons.org/licenses/by-nc/3.0/) which permits non-commercial use, reproduction and distribution of the work without further permission provided the
original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/open-access-at-sage).
2 International Journal of Spray and Combustion Dynamics 0(0)
soot formation occurs in the fuel rich side due to low
concentration of oxidizer species.
The soot evolution process has been reviewed by
Haynes and Wagner,4 Bockhorn,5 Kennedy6 and has
been classified into four major sub processes that
includes the formation and conglomeration of polycyc-
lic aromatic hydrocarbons (PAH) (particle inception or
nucleation), surface reaction of particles, coagulation
and oxidation of particles. The formation of particle
like structure by coagulation of PAH is known as
soot particle inception or nucleation. Chemistry of
combustion and soot particle dynamics are linked by
soot particle inception, it controls the number of soot
particles emerging from flames. Three different precur-
sors are proposed to act as the pathway of inception:
polyacetylenes, ionic species and PAHs. Numerous
experimental and modeling studies have concluded
PAH as the principal precursor. These first ring aro-
matics grow by H-Abstraction-Carbon-Addition
(HACA) mechanism to form large PAHs. The growth
of these particles is limited by oxidation of aromatic
species. Oxygen plays a dual role, carbon mass is
reduced by oxidation limiting growth and H-radical
formation required for HACA is promoted. The
number of particles formed depends on the balance of
formation and oxidation. Surface growth is similar to
nucleation but is a heterogeneous process and compli-
cates modeling. Majority of soot mass is accumulated
by soot surface growth. The surface growth of particles
is still uncertain, in general acetylene is considered as
the principal reactant at the surface of gas phase spe-
cies. Frenklach et al.7 and Harris et al.8 described the
effect of acetylene on growth. They have parameterized
surface growth as function of acetylene concentration
and density of active area of soot surface. Coagulation
is the process of collision between initial soot particles
to form larger particles and occurs along with surface
growth. Particle coagulation only changes the soot par-
ticle size distribution. Soot oxidation converts solid
soot particles into gas phase species, acts as a counter-
balance of surface growth. Surface reactions with
molecular oxygen and OH radial lead in the formation
of CO and CO2. Oxidation due to OH is less rigorous
than molecular oxygen as OH exists in super equilib-
rium concentrations.
Parameters that control soot yield can be identified
by accurate prediction of soot formation and oxidation
can save time and cost in design practical systems.
Because of low molecular diffusivity, soot transport is
governed by convection and gradient of temperature.
The transport of soot results in a two-way coupling
between concentration of species, temperature and
soot. The complex formation process of soot results
in large time scales, longer than heat release rate in
flames. Thus, soot cannot be decoupled from the flow
field. The temperature field is effect by through radi-
ation and concentration of gaseous species is effected
by the formation of intermediate products. The effect of
detailed chemical mechanisms and radiative heat trans-
fer models on soot and incorporation of different tur-
bulence-chemistry interaction models is a subject of
research. The complex interactions of soot with turbu-
lence, chemistry and particle dynamics complicates
modeling. Models with varying degree of sophistication
were used in practical systems6 for predicting soot for-
mation rate.
Empirical or semi-empirical soot models were
applied for most of the studies. Brookes and Moss9
and Kronenburg et al.10 have modeled soot formation
in a methane/air flame using an extended flamelet and
conditional moment closure approach, respectively.
The soot predictions were in reasonable agreement
with the experimental measurements. A unsteady
Lagrangian flamelet model was used by Pitsch et al.11
to model soot formation in an ethylene flame. Detailed
chemical kinetics was augmented to describe the first
aromatic species and evolution of soot was incorpo-
rated using the method of moment approach. The
soot predictions were in good agreement with the
experimental data and improved with the consideration
of differential diffusion, the model could not be applied
to complicated geometries as the computation were
restricted by the unsteady Lagrangian flamelet model.
A laminar flamelet model was used by Ma et al.12 to
model soot formation in two ethylene/air flames. Soot
formation was modeled using a two equation model of
Moss.13 The work focused on the implementation
of naphthalene-based soot inception route and effect of
soot surface area on growth rate. They have observed
that naphthalene approach improved soot predictions
and uncertainties reduced with the consideration of
growth as a function of square root of surface area.
A sectional approach was used by Kohler et al.14 to
model soot formation in a unconfined C2H4/air jet
flame, the turbulent-chemistry interactions were imple-
mented by a multivariate assumed PDF. The model was
able to capture the magnitude of maximum soot
volume fraction but discrepancy was observed in the
location and shape. The discrepancies were attributed
to the turbulence model. The objective of the present
study is (a) to study the effect of different turbulence-
chemistry interaction approaches and radiation model-
ing approaches on flow field; (b) to study the effect of
different soot inception routes and oxidation
approaches on soot volume fraction predictions; and
(c) to study the effect of different soot-turbulence inter-
action approaches on soot evolution process. The com-
parison of flow field predictions with experimental data
clearly elucidates the effect of turbulence-chemistry
interactions and radiation on temperature and soot
Busupally and De 3

evolution. The study also examines the ability of an for the unburnt reactants and one for the burnt
empirical approach for prediction of OH radical con- reactants.
centration and its effect soot volume fraction.  
@ @ @ t @~c
ð~cÞ þ ~
ð~vcÞ ¼ þ Sc ð3Þ
Numerical methods @t @xk @xj Sct @xk

The numerical approach using for solving the turbu- The progress variable is defined as equation (4).
lence-chemistry interactions, radiation modeling and  
P
the soot model are described in the following k ak Yk  Yuk Yc
section. c¼ P eq ¼ eq ð4Þ
k ak Y k Yc

The mean reaction rate is modeled as equation (5).

Turbulence-chemistry interactions
X
The fuel and oxidizer can be considered as partially Sc ¼ 0 Ul ð5Þ
premixed at the base, for a lifted turbulent diffusion
flame. The rich and lean regions of the flame can be The flame front
P convolution is defined by the flame
separated by the surface of instantaneous stoichio- surface density .16 The flame surface is approach con-
metric mixture fraction. The burnt and unburnt siders the laminar flame thickness to be larger than the
gases can be separated by an instantaneous flame Kolmogorov eddies, the internal laminar flow is not
front when the mixture of fuel and oxidizer are distorted and turbulence only wrinkles the flame
inhomogeneous and fluctuating. Thus, a partially sheet. The transport equation for flame surface area17
premixed approach is used for this purpose. The is given as equation (6).
position of flame front is determined by solving P
transport equation for the mean reaction progress @ @ X
þ ð~ Þ
variable. The mean mixture fraction and mixture @t @xk
0 P 1
fraction variance transport equations are solved in
@ @ t @  A X
the flow field. Two approaches were used to describe ¼ þ ðP1 þ P2 þ P3 Þ þP4  D
the turbulence-chemistry interactions and explained @xj Sct @xk
in detailed below.
ð6Þ
Steady laminar flamelet approach. The turbulent flame is P
represented as an ensemble of one dimensional laminar where is the flame surface area density, the source
structures in laminar flamelet model.15 The iso-surface term due to turbulence interaction is given by P1, the
of the mixture fraction is used to prescribe the flame source term due to dilatation in the flame is given by
surface within the turbulent flow field. The transport P2,18 the source term due to expansion of burned
equations of the mean mixture fraction and variance gases is given by P3, the source term due to
are given as: normal propagation is given by P4,18 and D is the
! flame area dissipation. The source terms are closed as
@ ~ @  ~ @ t @f~ shown in Table 1. c calculated from c~ as given in
ðf Þ þ uk f ¼ ð1Þ equation (7).
@t @xk @xj t @xk
c~ ub
@ ~ 002 @  ~ 002 c ¼   ð7Þ
ðf Þ þ uk f
@t @xk 1  c~ 1  ub
002 ! 002 !2
@ t @f~ @f~ "
¼ þ Cg t  Cd  f~ 002 ð2Þ
@xj t @xk @xk k

where t , Cg , and Cd are 0.85, 2.86, and 2.0, respect-

Table 1. Source terms for extended coherent flamelet model
ively. The premixed combustion is described by the (by Veynante et al.18).
C-equation approach; the scalar represents the pro-
gress of reaction. In a turbulent flow field, the tem- P1 P2 P3 P4 D
poral and spatial evolution of reaction is described by P P2
21 @ @2 c
a transport equation of C as given in equation (3). C 1:6Kt ~
ðÞ 0:2Ul Ul 0:4Ul
3  @xk c ð1  c Þ @xk 2 ð1  c Þ
varies between zero and one within the flame, zero
4 International Journal of Spray and Combustion Dynamics 0(0)

where c~ is favre averaged progress variable. The turbu- transport equation in terms of un-normalized pro-
lent time scale as is given as equation (8). gress variable (equation (13)) is solved as closure for
FGM.
"
Kt ¼ ½ð1  0 Þ þ 0 k  ð8Þ !
k @Y~ c @  @ @ ~c
Y
 þ  i Y~ c Þ ¼
ðu  eff
D þ S c ð13Þ
@t @xi @xi @xi
where K is the turbulent kinetic energy, " is the
turbulent dissipation, and k intermediate turbulent
net flame stretch (ITNFS). k is given by where S c is modeled from the finite rate kinetic rate as
equation (9). equation (14). is determined from the flamelet library.
Z Z
1
log10 ðk Þ ¼  expððs þ 0:4ÞÞ S c ¼  SFR ðc, fÞPðc, fÞdcdf ¼ S FR ð14Þ
ðs þ 0:4Þ
  0 
u
þ ð1  expððs þ 0:4ÞÞÞ  s  0:11 The thermochemical properties are determined as
Ul
equation (15).
ð9Þ
Z 1 Z 1

where S is defined as equation (10). ~ ¼ c~ ~

b ð f, Þ pð f, Þdfd þ ð1  cÞ u ð f Þ pð f Þdf
0 0
  ð15Þ
lt
s ¼ log10 ð10Þ
0 l

where 0 l is the laminar flame thickness and is modeled

as equation (11)
 has been assumed as a truncated series expansion in
0 l ¼
Ul
The averaged species mass fraction and temperature
for a turbulent flame can be determined as
equation (12).
Z 1 Z 1
~ ¼ c~ ~
b ð f, Þ pð f, Þfd þ ð1  cÞ
0 0
The non-adiabatic steady flamelets for burnt react-
ants have an extra dimension of mean enthalpy for
temperature and mean density. In the presumed PDF
of equation (12), f and are assumed to be statistically
independent and the PDF pð f, Þ can be assumed as
pf ð f Þ p ð Þ. A b-PDF is assumed for pf ð f Þ and double
delta PDF is assumed for p ð Þ.
av ¼
i þ
soot ð17Þ
Flamelet-generated manifold (FGM) approach. In FGM19
model, a turbulent flames scalar evolution is estimated
from the scalar evolution in a laminar flame. A diffu-
sion FGM is used in the present study, they are con-
structed by converting the steady laminar flamelets
soot ¼ b1 soot ½1 þ bT ðT  2000Þ ð18Þ
species fields to progress variable c. ~ With the increase
in strain rate, the chemistry of flamelets departs from
equilibrium and progress variable decrease from unity
to extinction, c~extinction . Instead of the C-equation, a
Radiation modeling
In the present work, the medium is considered to be
optically thick and the radiative heat transfer equation
ð11Þ
spherical harmonics (P1 approximation)20 given by
equation (16).
 
1
r  rGv ¼ Kav ðKbv  Gv Þ ð16Þ
3ðKav  Ksv Þ  Aksv
where the right side of equation (16) is the source of
radiative heat added to energy equation. The absorp-
u ð f Þ pð f Þdf tion coefficient of the product gasses and soot is
ð12Þ given as Kav . In the present work, the absorptive
coefficient is modeled by the weighted sum of gray
gases (WSGG) model. The non-gray medium of is
represented by four fictions gases and details of the
approach can be found in Rakesh et al.21 The effect
on radiative heat transfer by soot formation is
included by the absorption coefficient of soot and
given as equation (17).
The absorption coefficient due to soot particulates is
given as:
where b1 ¼ 1232.4m2/kg and bT ¼ 4.8  104 K1 are
given by Taylor et al.22 and Smith et al.23
Busupally and De 5

Soot modeling The default parameters were used in the present

In the present work, two transport equations in terms study and can be found in Hall et al.1 and Brookes
of volume fraction of soot and concentration of nuclei and Moss.13 The mass fraction of soot precursor has
are used to describe the evolution of soot. been calculated as function of mixture fraction. Two
  approaches were used to describe soot oxidation phe-
@Ysoot t dM nomenon, Lee model considered both OH and O224,25
ÞYsoot ¼ r 
þ r  ð~ rYsoot þ ð19Þ
@t soot dt and Fenimore-Jones considers only OH radical25 for
oxidation of soot. The effect of rate of oxidation on
 
@bnuc  t  1 dN net soot production was studied based on collisional
Þbnuc ¼ r 
þ r  ð~ rb þ efficiency parameter and surface kinetics.
@t nuc nuc Nnorm dt
ð20Þ
OH radical concentration determination. In the present
where N is the number density of soot particle and M is
study, two approaches were used to determine the
the soot mass density; as defined by Brookes and
OH radical concentration. In the first approach, the
Moss13 model to predict soot formation in turbulent
OH radical concentration is determined, as given in
diffusion flames. The source terms of instantaneous
equations (24)26,27 and concentration of O radical is
soot particles and net set productions are given equa-
tions (21) and (22), respectively.

    
dN Xprec P T 24RT 12 12 2
¼ C  NA exp   C  dp N ð21Þ
dt RT T soot NA
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl
ffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
Nucleation Coagulation

      "  2=3 #
dM Xprec P T Xsgs P T 1=3 6M
¼ C  MP  exp  þ C  exp  ðNÞ
dt RT T RT T soot
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
Nucleation SurfaceGrowth
"  2=3        #
pffiffiffiffi 1=3 6M XOH P T!,1 XO2 P T!,2
 Coxid  TðNÞ  C!,1  coll exp  þ C!,2  exp 
soot RT T RT T
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
Oxidation
ð22Þ

In the present study, two models are used to study determined by equation (25),28 known as the equili-
the effect of inception rate on soot formation. In the brium approach.
first approach, the soot precursor was consider as a
linear function of acetylene concentration as given in ½OH ¼ 2:129  102 T0:57 e4595=T ½O1=2 ½H2O1=2 ð24Þ
equation (21). In the second, the soot precursor was
considered as a function of two and three ringed aro- ½O ¼ 3:97  105 T1=2 e31090=T ½O2 1=2 ð25Þ
matic species, given in equation (23) (proposed by Hall
et al.1). In the second approach, the OH radical concentra-
"  # tion is determined from the combustion model.
  
dN NA 2 YC2 H2 2 YC6 H5 WH2
¼ 8  C,1   Soot-turbulence interactions. The effect of turbulence on
dt Inc: MP WC2 H2 WC6 H5 YH2
 soot is included by a single variable PDF in terms of
T,1 temperature or mixture fraction to describe the effect of
 exp  þ 8  C,2
T temporal fluctuations on net soot production.
"   #
NA 2 YC2 H2 2 YC6 H6 YC6 H5 WH2 Z
 
MP WC2 H2 WC6 H6 WC6 H5 YH2 S~soot ¼ Ssoot ðxÞPðxÞdx ð26Þ

T,2
 exp  ð23Þ
T where x is instantaneous temperature or mixture frac-
tion. The solution of soot transport is used to obtain
6 International Journal of Spray and Combustion Dynamics 0(0)

the mean value of independent variables required for velocity profiles are slightly adjusted iteratively to
constructing the PDF. match the experimental flow rates.
In the present work, ANSYS FLUENT 15.031 is used
for all the calculations, the non-gray radiative transfer
Description of test case
model is incorporated using an user defined function
A C2H4-air lifted diffusion flame is considered in the (UDF).
present study. The details of the burner are provided by The turbulence-chemistry interactions are modeled
Kohler et al.29,30 The fuel jet has a diameter of 2 mm using steady laminar flamelet approach (SF) and
and the outer diameter of the pipe is 6 mm. The diam- FGM. The flame thickness using ITNFS treatment is
eter of the coflow is 140 mm, the global equivalence calculated to be 0.00026 for SLF model. Chemistry
ratio is 0.48. The schematic of the burner is shown in has been represented by a detailed POLIMI mechanism
Figure 1(a). The soot measurements for the present containing 82 species and 1450 reactions.32 The absorp-
burner were performed by Kohler et al.14,30 For the tion coefficient of the non-gray medium is calculated by
present computational work, an axisymmetric 2D grid the WSGG method, considering four fictitious gases
has been used. Figure 1(b) shows the computational and the weight functions are calculated from Smiths
domain. Initially, three grids such as 720  220, tables23 as a function of H2O and CO2 partial pressures.
360  110, and 180  55 points are used to study the The pressure-velocity coupling is done using the
grid independence and Figure 2 shows the predictions
using three different grids. As the variation of axial
velocity with the first two grids is negligible, the third
grid is unable to capture the flame lift-off properly and Table 2. Experimental data of C2H4/Air flame.
results in the over prediction of velocity in the axial
direction. That’s why the rest of the computations Variable Jet inlet Coflow inlet
have been performed using the second grid Velocity (m/s) 44 0.29
(360  110). The experimental boundary conditions Temperature (K) 298 312
are shown in Table 1, a fully developed turbulent pro- Re 10,000 –
file was given as inlet condition for velocity, turbulent
Flow rate (g/min) 10.4 320
kinetic energy, and turbulent dissipation. The inlet

Figure 1. (a) Schematic of C2H4/air flame burner and (b) illustrative view of the mesh used (D: diameter of fuel jet).

Figure 2. Centerline plot of axial velocity using FGM model: mesh 1720  220, mesh 2 is 360  110, mesh 3 is 180  55, and symbols
are measurements (Kohler et al.14).
Busupally and De
SIMPLE algorithm. A second-order upwind scheme is
used to discretize all the convective fluxes.
Results and discussion
In this section, the flow field computations with SLF
and FGM approaches are presented and compared with
experimental measurements. The effect of turbulence-
chemistry interactions and radiative heat transfer on
the flame lift off height and length are studied. Then,
the soot predictions with different inception and oxida-
tion approaches are presented and compared with the
experimental measurements; followed by a qualitative
study of the soot source terms. The effect of different
soot-turbulence approaches on evolution of soot has
been comprehensively studied.
Velocity and scalar fields
The contour plot of predicted mean axial velocity is
shown in Figure 3 using Ce1 ¼ 1.6 and Ce2 ¼ 1.92.
The shape and magnitude of the computes results are
in good agreement with the experimental measurement.
The SLF model exhibits slight broadening of velocity
profile after flame ignition, this has discussed in the
subsequent section. In Figure 4, the computed and
Figure 3. Contour of mean axial velocity (a) experimental, (b) SLF model, and (c) FGM model using non-gray radiation model using
Ce1 ¼ 1.6 and Ce2 ¼ 1.92.
7
experimental radial profiles of mean axial velocity are
presented. The turbulence model constants of k-e
model are modified to study the influence on jet spread-
ing. The computations have been performed using
standard (Ce1 ¼ 1.44 and Ce2 ¼ 1.92), modified 1
(Ce1 ¼ 1.6 and Ce2 ¼ 1.92) and modified 2 (Ce1 ¼ 1.44
and Ce2 ¼ 2.0) constants. The SLF model over predicts
centerline velocity near the burner, this can be attribu-
ted to the over prediction of mean temperature. Both
the SLF and FGM model predictions indicate over pre-
diction of jet width, this can be ascribed to the over
prediction of turbulent kinetic energy by the turbulence
model. The excessive spreading of the jet by standard
k-e model is confirmed by Figure 4, the central jet
decelerates rapidly compared with experimental
measurements due to prediction of excessive eddy
viscosity. This leads to under prediction of mean
axial velocity in the downstream direction.
Adjusting the Ce1 from 1.44 to 1.6 has improved
the spreading rate prediction. Modifying the Ce2
from 1.92 to 2.0 has resulted in further increase of
over prediction of spreading rate and under predic-
tion of mean axial velocity near the burner. The
inability of the model in accurately predicting the
spreading rate may be attributed to the round-jet
anomaly, which is quite well known and has been
8 International Journal of Spray and Combustion Dynamics 0(0)

Figure 4. Radial profile of mean axial velocity, (I) SLF model and (II) FGM model: solid lines are with gray radiation, dashed lines are
with non-gray radiation, and symbols are measurements.

reported in plenty of literatures.33 The modified k-e
model (Ce1 ¼ 1.6 and Ce2 ¼ 1.92) is able to capture
the axial velocity profiles reasonably well in com-
parison with the experimental measurements and
has been used for the rest of the computations.
Kohler et al. 14 have observed an under prediction
close to the centerline of the burner, which was pos-
sibly due to the usage of bulk velocity as velocity
inlet instead of a turbulent profile, and there results
tend to improve along the downstream directions.
Considering the experimental uncertainties, the pre-
dictions can be considered to be in good agreement
with the experimental data. The velocity profiles are
not effect by the non-gray radiation approach.
Figure 5 shows the contour of calculated mean tem-
perature. The liftoff height is calculated based on the
gradient of temperature along the radial direction.
Notably, the lift-off height is observed to be 25 mm for
Busupally and De
Figure 5. Contour of mean temperature: (a) SLF with gray radiation model, (b) SLF with non-gray radiation model, (c) FGM with gray
radiation model, and (d) FGM with non-gray radiation model.
SLF model, while it is 24 mm for FGM model. Both the
models are able to accurately capture the experimental
lift off height of 26 mm,30 indicating the ability of these
models to capture turbulence-chemistry interactions
accurately. The non-gray radiation approach does not
significantly affect the liftoff height. The radial profiles
of mean temperature are presented in Figure 6. Figure 6
indicates an over prediction in the centerline tempera-
ture near the burner by SLF model and improves along
the downstream direction. The SLF model neglects local
flame extinction, leading to over prediction of flame
temperature near the burner; this has resulted in the
over prediction of velocity about the centerline near
the burner inlet. For the present study, burning flame-
lets are only used, this doesn’t capture the local flame
extinction and as the flamelet model neglects the curva-
ture effects on the flame structure, this results in a
sudden change in scalar dissipation resulting in rapid
rise in temperature near the regions of ignition. The
over prediction in temperature also results in excessive
thermal expansion, as observed in Figure 3(b). The
FGM model has been able to accurately predict the cen-
terline temperature. Both the models indicate over pre-
diction of flame width as already mentioned; predicted
temperature profiles rise slower compared with the
experimental data. The non-gray radiation approach
does not indicate significant effect on temperature near
the burner inlet, the effect of radiation increases along
the downstream direction. For the SLF model, with the
inclusion of non-gray radiation approach, the flame
length decreased by 130 mm to 430 mm and the
9
centerline maximum temperature reduced to 1850K
from 2100K. The global maximum temperature
shifted upstream by 35 D in the upstream direction
and 4 D away from the centerline; the magnitude
reduced from 2150 K to 1950 K.
For the FGM model, with the inclusion of non-gray
radiation approach, the flame length decreased by
90 mm to 490 mm and the centerline maximum tem-
perature reduced to 1950 K from 2150 K. The
global maximum temperature shifted upstream by
20 D in the upstream direction and 2 D away
from the centerline; the magnitude reduced from
2250 K to 2000 K. The SLF model predictions are
closer to the experimental measurements of 410 mm.
The computed and experimental fuel distributions
are compared in Figure 7, in the region before to com-
bustion. The experimental data indicate a decrease in
fuel fraction at 7 mm, the computed fuel fraction with
both SLF and FGM indicate a decrease at 17 mm
about the centerline. The over prediction can be attri-
bute to the inability of turbulence model in capturing
the spreading rate. The observations are similar to that
of Kohler et al.14 The radial and axial profiles of fuel
fraction are shown in Figure 8. At X ¼ 5 mm, the mag-
nitude and shape are in good agreement with the
experimental data. At X ¼ 15 mm, the magnitude of
fuel fraction is over predicted indicating under predic-
tion of turbulent shear stress and acceleration of central
jet. The over prediction reduces along the downstream,
as observed at X ¼ 20 mm. A similar comment can be
made on the central line profile, it tends to improve in
10
Figure 6. Radial and axial profile of mean temperature, (I) SLF model and (II) FGM model: solid lines are with gray radiation, dashed
lines are with non-gray radiation, and symbols are measurements.
the downstream direction due to increase in turbulent
shear stress. The contours of predicted and experimen-
tal OH concentration are presented in Figure 9. Similar
to Kohler et al.,14 the experimental OH concentration is
normalized to 2500 ppm for easier comparison with
predicted data.
The prediction with SLF model indicates a slight over
prediction in magnitude of OH concentration and
increase in periphery distribution compared with
International Journal of Spray and Combustion Dynamics 0(0)
experimental data. The prediction with FGM indicate
a slight under prediction in terms of magnitude and a
similar comment can be made for the distribution of the
OH concentration in the radial direction as in case of
SLF model. Figure 10 displays the radials profiles of OH
concentration. The SLF model indicates slight under
predictions at x ¼ 25 mm, the predictions improve
along the downstream direction and indicate slight
over prediction at x ¼ 100 mm. OH concentration at
Busupally and De 11

Figure 7. Contour of C2H4 mole fraction (a) experimental, (b) SLF model, and (c) FGM model using non-gray radiation model.

Figure 8. Radial and axial profiles of C2H4 mole fraction using non-gray radiation approach: symbols are measurements.

x ¼ 25 mm was considerable under predicted by FGM prediction by SLF and FGM model can be attributed
model and the under prediction continued in the down- to the over prediction of scalar dissipation and slower
stream direction, with a broader profile than the experi- finite rate kinetics, respectively. The location of radial
mental prediction. The disparities in OH radical maximum was accurately captured by both the models.
12 International Journal of Spray and Combustion Dynamics 0(0)

Figure 9. Contour of OH mole fraction (a) experimental, (b) SLF model, and (c) FGM model using non-gray radiation model.

Figure 10. Radial profile of OH mole fraction using non-gray radiation approach: symbols are measurements.

fraction using Moss–Brookes and Moss–Brookes–Hall

Soot predictions
The axial profile of soot volume fraction using different
inception, oxidation and turbulence-chemistry inter-
action effects are shown in Figure 11. From
Figure 11(a) and (b), it can inferred that soot volume
approaches increases by a factor of 2 for the SLF
model and decrease by a factor of 8 with the FGM
model. The decrease in soot with FGM model can be
attributed to the models inability in accurately predict-
ing the concentration of two and three ring aromatics.
Busupally and De
Figure 11. Axial profile of soot volume fraction along centerline: solid lines are with OH equilibrium approach and solid symbols are
measurements.
It can be deduced that there is considerable increase in
soot volume fraction with the two-three ringed aro-
matic inception route when compared with that of acet-
ylene-based route for the SLF model. The Moss–
Brookes and Moss–Brookes–Hall approaches indicate
decrease in soot volume fraction by a factor of 1.5
with the instantaneous OH approach, indicating that
equilibrium approach is under predicting the OH con-
centration. The location of maximum soot volume frac-
tion occurs downstream by 50 D when compared with
experimental data, this can be ascribed to the over pre-
diction of flame length. Figure 11(c) and (d) displays
decrease in soot volume fraction with the Fenimore-
Jones oxidation model and are considerable under pre-
dicted when compared with the experimental data for
the FGM approach. The decrease in soot volume frac-
tion with the Fenimore-Jones model is due to higher
oxidation scaling than when compared with Lee
model. The SLF model along with Moss–Brookes–
Hall and instantaneous approach, gives a reasonable
good estimate and the location of maximum soot
volume fraction shifts in the upstream direction.
Soot nucleation: The soot source terms axial profiles are
presented in Figure 12. It can be inferred that soot
nucleation occurs at lower mixture fraction for the acet-
ylene-based inception route and occurs at a higher mix-
ture fraction with the two-three ringed aromatic-based
route (Moss–Brookes–Hall model), for the FGM, the
maximum nucleation occurs at 0.27 in the mixture
fraction space. The magnitude also increases by a
13
factor of 1000 with first aromatic-based approach.
The SLF and FGM models do not indicate considerable
effect on nucleation with acetylene-based route. The
nucleation with two-three ringed aromatic-based
approach is relatively wider in the mixture space with
SLF model than that of FGM model, resulting in the
substantial under prediction in soot volume fraction.
The narrow band in mixture fraction space with FGM
model is due to the inability of the model in accurately
estimating the mass fractions of higher order hydrocar-
bons. From Figure 12, it can be observed that different
oxidation and OH concentration approaches have not
significantly effected nucleation, the net effect of oxida-
tion on nucleation is observed to be negligible.
Soot surface growth: Surface growth increases in magni-
tude and widens with two-three ringed aromatic
approach for the SLF model. For the FGM model, sur-
face growth decreases with Moss–Brookes–Hall model,
this can be attributed to the lack of active sites of higher
order actives required for the heterogeneous reactions.
Furthermore, it can also be observed that for the FGM
model and OH instantaneous approach significantly
reduction in soot surface growth, this is due to the
under prediction of active sites for heterogeneous reac-
tions. The slight decrease due to OH instantaneous
approach is due to over prediction of oxidizer species.
The oxidation scaling parameter is 0.015 and 1; the
collision efficiency is 0.13 and 0.04 for the Lee and
Fenimore-Jones models, respectively. The decrease in
surface growth with Fenimore-Jones model is due to
14
Figure 12. Axial profile of soot source terms along centerline: (I) with SLF model and (II) with FGM model.
the high oxidation scaling parameter compared
with Lee, the high collision efficiency of Lee model
could not compensate for the low scaling parameter
as the diameter of the soot particles are large.
Thus, the parameters require additional tuning the
accurately capture the oxidation of active sites of sur-
face growth.
Soot oxidation: From Figure 12, it can be inferred that
oxidation increases in magnitude with the Fenimore-
Jones model, it also shifts to the space of high mixture
fraction. The width of oxidation rate in mixture frac-
tion space is not significantly affected. The oxidation
International Journal of Spray and Combustion Dynamics 0(0)
rate is considerably under-predicted by the FGM
model. Similar to surface growth, oxidation also
decreases with the first aromatic species route with
the FGM model.
The total soot source term, the balance of soot
nucleation, surface growth and oxidation are presented
in Figure 13. It can be noted that majority of the sur-
face growth occurs in the vicinity of the shear layer and
slightly away from the center line. The net oxidation
shifts upstream and increases in magnitude with the
Fenimore-Jones oxidation model. Figure 13(a) and (c)
indicates shift in surface growth in upstream direction,
a similar comment can be made on Figure 13(b) and
Busupally and De 15

Figure 13. Contour of total soot source term with Moss–Brookes–Hall model (a) SLF with Lee oxidation model, (b) SLF with
Fenimore-Jones oxidation model, (c) FGM with Lee oxidation, and (d) FGM with Fenimore-Jones oxidation model.

Figure 14. Axial profile of soot volume fraction along centerline using SLF and Moss–Brookes–Hall model: solid symbols are
measurements (a) Equilibrium approach and (b) Instantaneous approach.

(d). The total growth reduces significantly for the FGM
with Fenimore-Jones oxidation model.
Soot-turbulence interactions: The effect of soot-turbulence
interaction on axial profile soot volume fraction is pre-
sented in Figure 14. The net soot produced slightly
increases with temperature and mixture fraction-based
soot-turbulence interaction approach. The equilibrium
approach is able to predict the location of maximum
soot volume fraction but the magnitude has been over-
predicted by a factor of 3. However, the instantan-
eous approach is able to predict the magnitude of soot
volume fraction about the centerline but the location
predicted to be upstream by 50 D when compared
with the experimental predictions.
The shift in the upstream direction is due to over
prediction of OH concentration. The over prediction
of soot inception rate about the centerline results in
early evolution of soot by 100 D, as observed by
Kohler et al.14 and Blacha et al.34 They were also
able to capture the magnitude of maximum soot
volume fraction. They have attributed the early evo-
lution of soot to the avoiding of soot fluctuations, a
similar comment can be made for the present results
as we have also utilized a presumed shape PDF for
modeling the turbulence chemistry interactions. Thus,
the turbulence-chemistry interactions require add-
itional considerations. The radial profiles of soot
volume fraction are presented in Figure 15. A similar
tend as predicted in the axial profiles can be observed.
At x ¼ 63 mm, the profiles indicate slight over predic-
tion of maximum soot volume fraction. The over pre-
diction increases at x ¼ 163 mm and 213 mm due to
over predication of nucleation. At x ¼ 263 mm, there
is under prediction in the radial maximum soot
volume fraction.
16 International Journal of Spray and Combustion Dynamics 0(0)

Figure 15. Radial profiles of soot volume fraction using SLF and Moss–Brookes–Hall model: solid symbols are measurements.

Figure 16. Contour of soot volume fraction using SLF and Moss–Brookes–Hall model (a) experimental, (b) without STI, (c) with
temperature-based STI, and (d) with mixture fraction-based STI.

It can be observed that soot formation occurs in a
thinner region than the experimental measurements,
while the predictions increase considerable with the
mixture fraction-based approach than when compared
with temperature-based approach. The computed
results are in reasonable agreement with the experimen-
tal data. The computed and experimental soot volume
fractions are compared in Figure 16. Formation of soot
occurs slightly earlier than experimental data. With the
inclusion of soot-turbulence interactions, the soot
volume fraction increases in the downstream direction
and also widens in the radial direction. But both the
axial and radial locations of maximum soot volume
fraction are not properly reproduced. The contours of
soot source terms are shown in Figure 17. From Figure
17(a) and (b), the mixture fraction-based approach has
predicted as increase in nuclei concentration and coagu-
lation compared to without STI and temperature-based
STI. This can be due to the mixture fraction PDF abil-
ity in accurately capturing the increase in number of
nuclei, due to increase in flame surface area. The tem-
perature PDF does not capture this effect, might be due
to the under prediction of rms temperature, leading to
decrease in net soot production rate. The soot surface
 Busupally and De

Figure 17. Contour of soot source terms using SLF and Moss–Brookes–Hall model (a) without STI, (b) with temperature-based STI, and (c) with mixture fraction-based STI.
17
18
growth and oxidation source terms by both the
approaches indicate increase when compared with the
prediction without turbulence. The increase with tem-
perature-based PDF approach is less than the mixture
fraction PDF, as discussed earlier. The RANS-based
approaches cannot capture temporal fluctuations
accurately, thus the STI require additional consider-
ations. Over the past couple of decades, soot modeling
has undergone several strides in predicting the evolu-
tion of soot in combustion systems. The present models
are able to accurately provide a qualitative description,
sometimes quantitative as well, for soot evolution. The
interactions between turbulence, chemistry and soot
need to be accurately modeled for improving quantita-
tive predictions. Higher order PAH (Poly Aromatic
Hydrocarbons) may be used for modeling soot nucle-
ation. The parameters effecting oxidation of soot have
to be studied and role of oxyl radical on oxidation has
to be further investigated.
Conclusions
In this work, soot formation in a C2H4/air lifted turbu-
lent diffusion flame has been studied using two turbu-
lence-chemistry interactions (SLF and FGM)
approaches. The velocity predictions are found to be
in good agreement with the experimental data. The lift-
off height and flame length are reasonably predicted by
both SLF and FGM models, while the radiative loss is
accurately accounted by the non-gray radiation model.
Soot production has been examined using two nucle-
ation approaches, two OH concentration approaches
and two oxidation approaches with SLF and FGM
models. With SLF model, the two and three ring aro-
matic approach produce a good representation of
nucleation, whereas surface growth is significantly
affected by OH concentration approaches, as it turns
out that equilibrium approach under-predict the OH
concentration. Fenimore-Jones oxidation model
reduces soot volume fraction considerably, indicating
the dominance of OH radicals on soot oxidation.
With FGM, the soot predictions are considerably over
predicted, due to the inability of the model in capturing
the mass fractions of higher hydrocarbons accurately.
The mixture fraction-based STI approach indicate
better estimate of soot-turbulence interactions when
compared with the temperature-based approach. The
SLF model with Moss–Brookes–Hall model and
Fenimore-Jones oxidation model and mixture fraction
STI approach is able to capture the magnitude of max-
imum soot volume fraction and location and is esti-
mated to be 50 D in the upstream direction.
Considerable improvement is found in location of max-
imum soot volume fraction when compared with pre-
vious predictions. Further considerations are required
International Journal of Spray and Combustion Dynamics 0(0)
in turbulence-chemistry interactions to accurately pre-
dict percusses and oxidizing species. The infrequent
soot events also require additional considerations to
capture the soot-turbulence interactions.
Acknowledgments
The authors would like to acknowledge the IITK computer
center (www.iitk.ac.in/cc) for providing the resources to per-
form the computation work, data analysis and article prep-
aration. The authors would like to thank Dr. Klaus Peter of
DLR, Germany for providing the experimental data of C2H4/
air flame.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with
respect to the research, authorship, and/or publication of this
article.
Funding
The author(s) received no financial support for the research,
authorship, and/or publication of this article.
References
1. Hall RJ, Smooke MD and Colket MB. Physical and
chemical aspects of combustion: A tribute to Irvine glass-
man. Amsterdams: Gordon & Breach, 1997, p.189.
2. Mehta R. Detailed modeling of soot formation and turbu-
lence-radiation interactions in turbulent jet flames. PhD
dissertation, Pennsylvania, USA: Pennsylvania State
University, 2008.
3. Wang L. Detailed chemistry, soot, and radiation calcula-
tions in turbulent reacting flows. 2004.
4. Haynes B and Wagner H. Soot formation. Prog Energy
Combust Sci 1981; 7: 229–273.
5. Bockhorn H (ed.) Soot formation in combustion. Berlin:
Springer-Verlag, 1994.
6. Kennedy I. Models of soot formation and oxidation.
Prog Energy Combust Sci 1997; 23: 95–132.
7. Frenklach M, Clary DW, Gardiner WC, et al. Detailed
kinetic modeling of soot formation in shock-tube pyroly-
sis of acetylene. Symp Combust 1985; 20: 887–901.
8. Harris SJ, Weiner AM and Blint RJ. Formation of small
aromatic molecules in a sooting ethylene flame. Combust
Flame 1988; 72: 91–109.
9. Brookes S and Moss J. Predictions of soot and thermal
radiation properties in confined turbulent jet diffusion
flames. Combust Flame 1999; 116: 486–503.
10. Kronenburg A, Bilger R and Kent J. Modeling soot for-
mation in turbulent methane–air jet diffusion flames.
Combust Flame 2000; 121: 24–40.
11. Pitsch H, Riesmeier E and Peters N. Unsteady flamelet
modeling of soot formation in turbulent diffusion flames.
Combust Sci Technol 2000; 158: 389–406.
12. Ma G, Wen JZ, Lightstone MF, et al. Optimization of
soot modeling in turbulent nonpremixed ethylene/air jet
flames. Combust Sci Technol 2005; 177: 1567–1602.
Busupally and De 19

13. Brookes S and Moss J. Predictions of soot and thermal 32. Ranzi E, Frassoldati A, Grana R, et al. Hierarchical and
radiation properties in confined turbulent jet diffusion comparative kinetic modeling of laminar flame speeds of
flames. Combust Flame 1999; 116: 486–503. hydrocarbon and oxygenated fuels. Prog Energy Combust
14. Köhler M, Geigle K-P, Blacha T, et al. Experimental Sci 2012; 38: 468–501.
characterization and numerical simulation of a sooting 33. Pope SB. An explanation of the round jet/plane jet anom-
lifted turbulent jet diffusion flame. Combust Flame aly. AIAA J 1978; 16: 279–281.
2012; 159: 2620–2635. 34. Blacha T, Di Domenico M, Köhler M, et al. Soot model-
15. Peters N. Turbulent combustion. United Kingdom: ing in a turbulent unconfined C2H4/air jet flame. 49th
Cambridge University Press, 2000. AIAA aerospace sciences meeting including the new hori-
16. Poinsot T and Veynante D. Theoretical and numerical zons forum and aerospace exposition. January 2011,
combustion, 2nd ed. Philadelphia, PA: Edwards, 2005. pp.1–10 ,USA.
17. Candel SM and Poinsot TJ. Flame stretch and the bal-
ance equation for the flame area. Combust Sci Technol
1990; 70: 1–15.
18. Veynante D and Vervisch L. Turbulent combustion mod- Appendix 1
eling. Prog Energy Combust Sci 2002; 28: 193–266. Notation
19. van Oijen JA and de Goey LPH. Modelling of premixed
laminar flames using flamelet-generated manifolds. a2,i emissivity weighting factor
Combust Sci Technol 2000; 161: 113–137. as characteristic strain rate
20. Siegel R and Howel Jr. Thermal radiation heat transfer. bnuc normalized radical nuclei concentration
4th ed. New York, NY: CRC press, 2002. c~ progress variable
21. Yadav R, Kushari A, Verma AK, et al. Weighted sum of ki absorption coefficient of gas
gray gas modeling for nongray radiation in combusting
kT turbulent time scale
environment using the hybrid solution methodology.
ksoot absorption coefficient of soot
Numer Heat Transf Part B Fundam 2013; 64: 174–197.
22. Taylor PB and Foster PJ. Some gray gas weighting coef- f mixture fraction
ficients for CO2-H2O-soot mixtures. Int J Heat Mass Sc reaction progress source term
Transfer 1975; 18: 1331–1332. T Temperature
23. Smith TF, Shen ZF and Friedman JN. Evaluation of Ul flame speed along the surface
coefficients for the weighted sum of gray gases model. Yk mass fraction of kth species
J Heat Transfer 1982; 104: 602–608. Ysoot soot mass fraction
24. Lee KB, Thring MW and Beér JM. On the rate of com-
bustion of soot in a laminar soot flame. Combust Flame
1962; 6: 137–145.
25. Fenimore CP and Jones GW. Oxidation of soot by Greek symbols
hydroxyl radicals. J Phys Chem 1967; 71: 593–597.
 thermal diffusivity of unburnt reactants
26. Baulch DL, Cobos CJ, Cox RA, et al. Evaluated kinetic
data for combustion modelling. J Phys Chem Ref Data
i absorption coefficient of ith gray gas
1992; 21: 411. 0 density of fresh gas
27. Westbrook CK and Dryer FL. Chemical kinetic model- soot turbulent Prandtl number of soot
ing of hydrocarbon combustion. Prog Energy Combust transport
Sci 1984; 10: 1–57.  representative scalar
28. Westeenberg AA. Kinetics of NO and CO in lean, pre- scalar dissipation rate
mixed hydrocarbon-air flames. Combust Sci Technol st scalar dissipation rate at f ¼ fst
1971; 4: 59–64.
29. Köhler M, Boxx I, Geigle KP, et al. Simultaneous planar
measurements of soot structure and velocity fields in a
turbulent lifted jet flame at 3 kHz. Appl Phys B 2011; 103: Subscripts
271–279.
30. Köhler M, Geigle KP, Meier W, et al. Sooting turbulent b burnt reactants
jet flame: characterization and quantitative soot measure- eq chemical equilibrium
ments. Appl Phys B 2011; 104: 409–425. u unburnt reactants
31. ANSYS Inc, Ansys fluent 15.0 user’s guide. Csnosburg,
PA, USA.