Professional Documents
Culture Documents
a r t i c l e i n f o a b s t r a c t
Article history: This study proposes a detailed soot modeling framework for large-eddy simulation (LES) to accurately
Received 18 May 2023 predict soot formation and particle size distributions (PSD) in turbulent reacting flows. The framework
Revised 7 October 2023
incorporates Flamelet Generated Manifold (FGM) chemistry and a soot model based on the discrete sec-
Accepted 9 October 2023
tional method (DSM) to predict both qualitative and quantitative sooting behavior while keeping the
computational cost affordable. Two elementary modeling strategies are considered in the LES formalism
Keywords: for describing soot formation rates. These strategies rely on an a-priori tabulation of soot formation rates
Discrete sectional method and their run-time computation. The LES formalism is applied to the simulations of a well-characterized,
Flamelet generated manifold non-premixed, turbulent jet flame. A comparative analysis of strategies employed for filtered soot source
Large eddy simulation
term treatment is conducted to investigate their impact on the prediction of soot quantities and the evo-
Turbulent non-premixed flames
lution of soot PSDs. The LES results for the gas phase and soot phase are compared against the available
experimental data. A good prediction of soot evolution is achieved with the two methodologies. The tab-
ulation of soot formation rates leads to a significant reduction in computational cost compared to the
model based on their explicit runtime computation. The LES results reveal that the modeling of filtered
soot source terms has a significant impact on the quantitative prediction of soot formation. The possible
reasons for the observed differences in the soot prediction are discussed. The run-time computation-
based model provides a more consistent treatment of the non-linear interactions between the gas and
soot phases in soot source terms compared to the tabulated soot chemistry approach. On the other hand,
the tabulated soot chemistry model is an interesting and efficient modeling approach for predicting soot
formation in turbulent conditions. Overall, both approaches have their strengths and limitations, and the
choice of approach may depend on the specific needs of the application.
© 2023 The Author(s). Published by Elsevier Inc. on behalf of The Combustion Institute.
This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/)
https://doi.org/10.1016/j.combustflame.2023.113128
0010-2180/© 2023 The Author(s). Published by Elsevier Inc. on behalf of The Combustion Institute. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/)
A. Kalbhor, D. Mira, A. Both et al. Combustion and Flame 259 (2024) 113128
radiation models provide a fairly good prediction of soot forma- pling of the soot model with the turbulent flow. These include:
tion in laminar flames [15] after careful calibration of the models. a) the correlation between reactive scalars and the soot scalar
The progression toward soot prediction in turbulent flames, how- (number density) concerning growth and oxidation terms, and b)
ever, remains a great challenge owing to its complex multi-physical the correlation between number densities in relation to coagula-
nature, characterized by an intricate coupling between flow pa- tion terms. Different strategies have been proposed to model soot-
rameters, gas-phase chemistry, and soot properties. Numerical pre- turbulence interactions in mono-dispersed and moment-based
diction of soot in turbulent conditions necessitates adequate char- models [7,30] and more recently in sectional models [4,11]. Mueller
acterization of the flame behavior as well as different physico- and Pitsch [34] introduced a subfilter PDF model with double-delta
chemical phenomena involved in soot formation. PDF, considering both sooting and non-sooting modes for subgrid-
Turbulent flows are characterized by a wide range of flow scales scale turbulence-soot interaction. This approach was employed in
from energy-containing large scales to the smallest Kolmogorov several other studies [11,35] and subsequently reformulated by
scales [16]. In comparison, the time scales of soot formation are Berger et al. [36], by defining the sooting mode sub-structure as
typically even larger than the integral flow time scales. This dis- a log-normal distribution. Furthermore, Yang et al. [30] proposed
parity makes the prediction of soot formation in turbulent flames a presumed subfilter PDF that explicitly considers soot distribution
a challenging task. Hence, although significant efforts have been in mixture fraction space. Recently, Colmán et al. [37] extended the
made to describe turbulent-chemistry interaction, understanding presumed subfilter PDF model to account for finite-rate soot oxi-
the effects of turbulence on soot formation and their modeling are dation by analyzing the local relative motion of diffusionless soot
still open subjects to be addressed. particles relative to mixture fraction iso-contours. Although closure
The prediction of soot formation in turbulent flames using models for soot-turbulence interactions are important in the de-
the aforementioned soot modeling approaches is preferable to be tailed description of the soot formation process, their impact on
conducted in the context of Direct Numerical Simulations (DNS), the overall prediction of global soot quantities appears less promi-
where a full description of all the temporal and spatial scales as- nent compared to the sensitivity of soot production to model pa-
sociated with turbulence and chemistry are resolved. However, the rameters involved in soot subprocesses [15]. As a result, in some
high computational cost required to resolve the coupled phenom- cases [13,20], interactions between turbulence and soot are ne-
ena associated with the scales of turbulence, chemistry, and parti- glected.
cle dynamics limits the DNS investigation to selected cases [17–19]. Earlier studies on the LES of turbulent sooting flames have been
On the other end, Reynolds Averaged Navier-Stokes (RANS)-based primarily focused on moment-based soot models. El-Asrag and
methods, are more affordable and mainly used at industrially- Menon [38] employed the Linear Eddy Model (LEM) in combina-
relevant conditions [20]. However, RANS approaches hardly provide tion with the moment method for soot prediction with LES. Later,
information on the transient phenomena of the flow, flame dynam- Mueller and Pitsch [7,34] combined the moment-based soot model
ics, and soot formation. Therefore, methods like Large Eddy Simu- with an extended Flamelet/Progress Variable (FPV) [39] model by
lation (LES), which facilitates information on transient features of including lumped-PAH inception kinetics. In this LES approach, the
the flow are more relevant despite their higher computational cost. presumed-PDF method was used to account for subgrid-scale inter-
In fact, LES has been applied very successfully for a wide range actions between turbulence and soot. Xuan and Blanquart [12] pro-
of problems including jet, swirl, or spray flames [4,11,21–26] and posed an alternative approach to the lumped-PAH inception ki-
it generally shows a good trade-off between accuracy and compu- netics, where the filtered transport equations for aromatic species
tational cost. Therefore, the development of reliable soot models were solved with closure based on a PAH relaxation model. This
for LES is essential for the study of soot formation in both funda- aromatic chemistry-turbulence interaction was found to be impor-
mental and industrial configurations. In recent years, LES has been tant in accurately reproducing soot yield in turbulent flames. To
adopted to investigate soot formation in turbulent flames, depart- obtain the information on particle size distribution in LES, the FPV
ing from semi-empirical methods [27–29] and evolving towards presumed-PDF model of Mueller and Pitsch [34] was extended to
more detailed soot modeling approaches [4,7,11,13,30,31]. a sectional method by Rodrigues et al. [11]. The qualitative trends
LES of turbulent sooting flames relies on three aspects: an ac- in soot formation were reproduced in their work while providing
curate description of gas-phase chemistry, modeling of the dynam- detailed information on the evolution of local soot PSD. The decou-
ics of soot particle population, and integration of models for cap- pling of soot and gas-phase species in the LES presumed-PDF ap-
turing interactions between soot-turbulence-chemistry. For reliable proach leads to narrow profiles of the mean soot volume fraction.
prediction of soot formation in turbulent combustion, a detailed Hence, the transported joint PDF (JPDF) method, which allows for
understanding of gas-phase kinetics is necessary to properly cap- a more accurate representation of the turbulence-chemistry inter-
ture the flame structure and formation of gas-phase species partic- actions by directly coupling the full scalar space, is applied to the
ipating in the soot processes (such as polycyclic aromatic hydrocar- sectional model by Tian et al. [31]. Encouraging agreement was ob-
bons (PAHs) or unsaturated hydrocarbons such as acetylene). How- tained with experimental measurements for soot quantities in tur-
ever, using detailed kinetic schemes involving hundreds of species bulent non-premixed flames.
is impractical in LES employing finite-rate chemistry-based meth- Besides flamelet-based methods, several other combustion
ods. As a result, tabulated chemistry methods [7,11], reduced ki- models have been investigated in the LES framework to model soot
netic schemes [32], and globally optimized chemistry [33] are of- formation in turbulent flames. The Conditional Moment Closure
ten used instead. Concerning the soot-phase description, although (CMC) approach was applied to a sectional soot model by Gkan-
quantitative modeling of soot formation in turbulent combustion tonas et al. [24] to study soot formation and predict their size dis-
is still beyond the frontier of current modeling advancements [15], tribution in a lab-scale swirl Rich-Quench-Lean (RQL) combustor.
the state-of-the-art soot models can provide partial solutions to In LES of lifted non-premixed turbulent flames with the sectional
practical problems of interest when used judiciously along with method, Grader et al. [14] used a finite-rate chemistry model,
appropriate predictions of the turbulent flow structures. that requires no assumptions concerning the combustion regime. A
Another challenge in LES of turbulent sooting flames is mod- good prediction of soot evolution was achieved in their study en-
eling the subgrid-scale interactions between soot and turbulence, abling detailed investigations of soot formation and oxidation pro-
as soot chemistry involves non-linear interactions between the cesses. Sewerin and Rigopoulos [40] adapted the stochastic fields
gas and solid phases. In the context of LES, subgrid turbulence- method to the LES of sooting flames, which recently extended to
soot interactions refer to the correlations that arise from the cou- the sectional method by Sun and Rigopoulos [4] to predict the soot
2
A. Kalbhor, D. Mira, A. Both et al. Combustion and Flame 259 (2024) 113128
3
A. Kalbhor, D. Mira, A. Both et al. Combustion and Flame 259 (2024) 113128
= where Cε = 1.8 is a model parameter, and is the LES filter size
ψ ψ (Z, C )P(Z )P(C )dZdC , (9)
(taken as the cube root of the volume of the grid element).
4
A. Kalbhor, D. Mira, A. Both et al. Combustion and Flame 259 (2024) 113128
2.4. Soot modeling gas-phase consumption of precursor species due to soot for-
mation is accounted for during the flamelet calculations step,
2.4.1. Sectional soot model and no additional transport equation is solved for their descrip-
The chemistry and dynamics of soot particles are described us- tion in the LES. Nevertheless, there is also the possibility of ex-
ing the discrete sectional method [59]. In the present model, soot tending the current model to explicitly solve transport equa-
particle volume ranges are divided into a finite number of sections. tions for the precursor species, a technique adopted in several
In each section, i, the governing equation for the soot mass frac- other studies [11,19], allowing for the evolution of slower pre-
tion Ys,i is solved by considering flow convection, diffusion, ther- cursor species in FGM chemistry. For brevity, this approach, in
mophoresis, and chemical rates. The sectional soot transport equa- which the soot source terms are computed on-the-fly is referred
tion can be formulated as: to as FGM-C here.
∂ (ρYs,i ) Additionally, an alternative variant of the FGM-C approach with
+ ∇ · (ρ [u + vT ]Ys,i ) = ∇ · (ρ Ds,i ∇ Ys,i ) + ω˙ s,i
∂t a slightly different formulation for filtered soot source term is
∀ i ∈ [1, nsec ] (19) evaluated. This variant involves tabulating the gas-phase rate
of the filtered chemical soot source term instead of calculat-
where ρ , ρs , u, vT , Ds,i , ω˙ s,i denote gas density, soot density (as- ing it from local gas-phase species concentrations. This partic-
sumed to be equivalent to the density of solid carbon), velocity, ular variant, referred to as FGM-CR, is aimed at understand-
thermophoretic velocity (calculated with Frienlander’s [60] expres- ing the implications of using filtered concentrations of species
sion), soot diffusion coefficient (assumed to be constant for all par- to compute rates. Given that this method is a variant of the
ticle sizes for numerical stability), and sectional source term, re- FGM-C model, it will be addressed separately to enhance clar-
spectively. The soot source terms account for the chemical and ity. A comprehensive analysis of this approach is provided in
physical processes involved in soot formation, including nucleation, Section 4.2.5.
condensation of Polycyclic Aromatic Hydrocarbons (PAHs), surface • The second model concerns the closure of the filtered soot
growth, oxidation, and coagulation. Nucleation is assumed to occur source term through the presumed PDF approach. In a general
through the dimerization of two pyrene (A4) molecules, and PAH form, the filtered soot source term convoluted with presumed
condensation onto soot particles is modeled using the method de- PDF is written as:
scribed by Roy [61]. Surface reactions, including growth and oxida-
1
tion, are modeled using the standard hydrogen-abstraction-C2 H2 - ω˙ s,i = ρ̄ ω˙ s,i (φg , φs )P(φg , φs )dφg dφs . (23)
addition (HACA) mechanism by Appel et al. [62]. Coagulation of ρ
soot particles is described using the model proposed by Kumar Using Bayes’ theorem, the joint PDF can be split into two
and Ramkrishna [63], and the morphological properties (e.g. fractal marginal PDFs for the thermochemical variable and for the
dimension) of soot are not considered for simplicity. As a conse- (φg , φs ) = P
soot such that P (φg )P (φs |φg ). In this model, the soot
quence, the present soot model does have limitations related to its source terms are parameterized through control variables and
omission of aggregation and the potential influence this may have tabulated in the manifold. The marginal PDF P (φg ) is assumed
on both the growth and size distribution of soot. Further details to have β -PDF function for Z and C, while the conditional PDF
about the soot model and its validation can be found in previous of soot variable P (φs |φg ) is treated as a δ -function. This approx-
work [59,64,65]. imation facilitates the partial inclusion of turbulence-soot inter-
actions by accounting for the effect of subgrid-scale fluctuations
2.4.2. Filtered sectional soot equations in the mixture-fraction and progress variables on soot source
For modeling soot formation and evolution in LES, the filtered terms. Besides, the assumption of δ -PDF function (neglecting
equations for sectional soot mass fractions are obtained as: subgrid-scale fluctuations) for P (φg ) is also examined to under-
∂ (ρ̄Ys,i ) νt
stand the impact of PDF integration on the performance of the
+ ∇ · (ρ [u + vT ]Ys,i ) = ∇ · ρ̄ D̄s,i +
∇ Ys,i + ω˙ s,i , tabulated soot chemistry, which will be explained later.
∂t Sct
The approach employed in other studies [7,66] for the transport
∀ i ∈ [1, nsec ]. (20) of slowly evolving species such as PAH and NO is applied for
The thermophoretic velocity
vT is modeled following [11], as: soot. Accordingly, the soot source term for the soot section is
split into production (ω˙ s,i
+
or ω˙ s,i ) and consumption parts (ω˙ s,i
−
prod
∇
T or ω˙ s,i
vT = −0.554ν̄
. (21)
cons ). The consumption part is linearized by soot mass frac-
5
A. Kalbhor, D. Mira, A. Both et al. Combustion and Flame 259 (2024) 113128
Table 1
Modeling approaches for filtered soot source terms.
Computation FGM-C Runtime evaluation using tabulated primitive variables (gas-phase species)
Computation FGM-CR Runtime evaluation using tabulated gas-phase rates for soot subprocesses
Tabulation FGM-T Direct a-priori tabulation
−,ox
where ω˙ s,i denotes the sectional soot consumption rate by ox-
idation subprocesses. For computational efficiency, the coagu-
lation process of soot particles is not explicitly solved at run-
time, but the inter-sectional mass transfer due to the coagula-
tion process (or other subprocesses) is included in the flamelet
computation. This approach, in which the soot source term is
tabulated, is referred to as FGM-T from hereon.
6
A. Kalbhor, D. Mira, A. Both et al. Combustion and Flame 259 (2024) 113128
Fig. 2. Illustration of the computational domain and grid (a), and the probability density function of the M parameter in Pope’s criteria for the LES grid (b).
million hexahedral elements. The mesh has been also verified for istry including soot kinetics using the code CHEM1D [71]. In LES
Pope’s criteria [69] in Fig. 2b. More than 90% of the grid cells show simulations, 30 soot sections are transported to describe particle
a ratio of resolved to total kinetic energy (M) larger than 0.8, indi- size distribution, hence the flamelets are also computed with 30
cating the kinetic energy is sufficiently resolved with the selected sections. To cover the composition space from chemical equilib-
mesh. rium to mixing in the flamelet database, first, a series of strained
Synthetic turbulence derived from a Laplacian filter following steady counterflow flamelets are computed by varying the applied
the method proposed by Kempf et al. [70] is employed to define stain rate from lower values (close to chemical equilibrium) until
the inflow condition for the fuel. The mean axial velocity of the the extinction limit. Subsequently, the composition space between
turbulent database is set to a power-law profile with an exponent the extinction limit and the mixing solution is covered by simulat-
of 1/7. Considering the Reynolds number of a pipe flow at the con- ing unsteady quenching flamelets at the extinguishing strain rate.
ditions of the fuel inlet, the turbulent intensity is around 5%. It is The detailed kinetic scheme KM2 of Wang et al. [72], involving 202
worth noting that there is no consensus in the literature regard- species and 1351 reactions, is used for the gas phase chemistry
ing the selection of inlet boundary conditions in this case. Other during the computation of flamelets. This mechanism has been
numerical investigations [4,31] of similar flames have used turbu- extensively validated for soot formation prediction in ethylene-
lence intensity values ranging from 6% to 10%. Unfortunately, due fueled laminar [65,73,74] and turbulent flames [11]. Considering
to the unavailability of velocity measurements, the velocity profile the large Reynolds number of the turbulent jet, turbulent diffusivi-
and Reynolds stresses can not be compared and it was estimated ties are expected to be higher than molecular diffusivities. Hence a
with 5% in this study. The velocity inlet profiles for the pilot and unity-Lewis diffusion model is considered for the species transport
coflow air streams are treated as uniform. The velocity of the air in the flamelet computation. For the mapping of thermochemical
stream is specified as 0.6m/s while the bulk velocity of the pilot variables, the progress variable (Eq. (4)) is defined based on H2 O,
is adjusted to impose the mass flow rate satisfying the experimen- CO2 , CO, O2 , H2 , and A4 species mass fractions with their corre-
tal condition. No-slip adiabatic wall conditions are imposed at the sponding weight factors αH2 O = 0.0555, αCO2 = 0.0228, αCO = 0.0357,
injector boundaries. The pilot inlet is assumed to have a compo- αO2 = −3.13 × 10−4 , αH2 = 0.173, αA4 = 0.0988. The progress vari-
sition close to the equilibrium state of an ethylene-air mixture at able definition is determined using a guess-and-check approach,
an equivalence ratio of 0.9 ( Z = 0.0577, C= 1) and temperature of and shown to preserve the unique mapping of Y in composition
2256K. The fuel and coflow inlet temperatures are maintained at space [75]. Including pyrene (A4) in the progress variable defini-
294K. tion ensures a unique mapping of the tabulated quantities under
For the creation of the FGM database, a series of 1-D non- conditions where the soot models are especially sensitive to the
premixed counterflow flamelets are calculated with detailed chem- local gas phase composition. However, since the A4 species is not
7
A. Kalbhor, D. Mira, A. Both et al. Combustion and Flame 259 (2024) 113128
transported in either of the approaches used for soot source terms tions, the flame reaction zones (indicated by OH contours) show
in the present study, the accuracy of the current FGM-DSM frame- local extinction events within the shear layers. The formation of
work is not significantly impacted by the inclusion of A4 in the PAH (A4) and C2 H2 is predominant in the fuel-rich region. How-
progress variable Y. The thermochemical variables are stored in the ever, a systematic lag can be noticed in the spatial locations of
FGM database with a non-uniform (refined near stoichiometric and A4 formation compared to C2 H2 as the relatively large time scales
equilibrium regions) resolution of 101×11×101×11 grid points in Z, governing A4 formation reflect in its downstream spatial evolution.
SZ , C, and SC space respectively. The production of A4 is observed to initiate mainly after z/d j 50,
Previous studies [11,76] conducted on similar flames have while C2 H2 , a precursor species responsible for surface growth is
demonstrated the importance of considering thermal radiation ef- found to occur closer to the burner exit (z/d j 20). Furthermore,
fects due to the temperature sensitivity of soot formation. The cur- the regions of higher OH concentration, in which soot oxidation
rent LES-FGM-DSM framework can be expanded to include radia- is dominant, are shown to be prominent in the region beyond
tive heat transfer by further augmentation of the flamelet database z/d j 170.
for non-adiabatic conditions and additional parameterization in en-
thalpy space. However, the primary focus of this study is the com- 4.1.1. Gas-phase validation
parative evaluation of two different methods for modeling soot for- For the preliminary validation of gas-phase, computed time-
mation chemistry. Hence, the effects of thermal radiation from gas averaged radial profiles of mean OH mass fractions at different
and soot are neglected for simplicity. downstream heights are shown in Fig. 4 along with the experi-
The simulations are carried out in a Cartesian coordinate sys- mentally measured OH signal. In addition, a comparison of the PAH
tem using the multi-physics code Alya [77], developed at the signal against computed profiles of mean C2 H2 and A4 mass frac-
Barcelona Supercomputing Center (BSC). In the Alya code, a tions at several axial locations is presented in Fig. 4. Overall, the
second-order conservative finite element scheme is used for spatial computed profiles show fair qualitative agreement with the exper-
discretization, while an explicit third-order Runge-Kutta scheme is iments regarding downstream evolution, suggesting that the tur-
employed for the time integration. A low-dissipation scheme based bulent combustion models applied can favorably capture the main
on the fractional step algorithm proposed by Both et al. [46] is features of the flame structure in the gas phase. A slight overpre-
used for continuity and momentum equations under a low-Mach diction of the jet spreading rate is observed in the radial direction
number approximation of reacting flows. The simulations are per- at downstream positions leading to a wider spread in the species
formed using the Hawk cluster equipped with AMD EPYC 7742 profile. This however is found to have a minor effect on the quan-
processors at the High-Performance Computing Center Stuttgart. titative prediction of soot formation.
The temporal statistics for the quantities are performed for a pe-
riod of approximately 250ms after the simulations have reached a 4.2. Characterization of soot formation
statistically steady state.
4.2.1. Instantaneous fields
4. Results and discussion To investigate the effects of tabulating the source terms and in-
clude the PDF integration on the evolution and distribution of soot,
4.1. Gas-phase characteristics the instantaneous fields of soot volume fraction and soot forma-
+ −
tion rates (split into production ω˙ s and consumption ω˙ s parts) ob-
To describe the main combustion features of the piloted tur- tained with FGM-C and FGM-T approaches are compared in Fig. 5.
bulent jet flame, the instantaneous fields of temperature, and key From the soot volume fraction fields, it is evident that soot is
species involved in soot chemistry are presented in Fig. 3. An iso- mainly formed in the fuel-rich zones beyond z/d j > 50, character-
contour of stoichiometric mixture fraction (
Z = Zst ) characterizing ized by high A4 mass fractions (see Fig. 3). Soot inception is pre-
the flame front is also shown. The flame is found to be stabilized dominant near the flame base, leading to the formation of small-
by the pilot and attached to the burner. Under turbulent condi- sized particles and, consequently, low values of soot volume frac-
Fig. 3. Instantaneous 2-D fields of temperature, OH mass fraction, C2 H2 mass fraction, A4 mass fraction. The stoichiometric mixture fraction is shown with dashed iso-lines.
The instantaneous results correspond to the FGM-C case.
8
A. Kalbhor, D. Mira, A. Both et al. Combustion and Flame 259 (2024) 113128
Fig. 4. Comparison of radial profiles for computed (normalized with maximum) mean mass fractions against measured signals for OH (left panel), PAH (right panel) at
different axial locations.
tion. From Fig. 5, it can be observed that the soot production rate bution of soot consumption rates is found to be qualitatively and
(dominated by surface growth primarily) is concentrated in the quantitatively somewhat similar in both the FGM-C and FGM-T ap-
middle of the flame. Therefore, a high amount of soot volume frac- proaches. Therefore, the noticed discrepancies in the quantitative
tion is noticed in the middle region of the flame, where incepted prediction of soot with FGM-C and FGM-T methods are primarily
particles grow. On the other hand, consumption rates (oxidation) the consequence of differences in the soot production rates.
are prominent at the tip region of the flame, and near the stoi-
chiometric iso-contour where particles are oxidized due to a high 4.2.2. Soot volume fraction fluctuations
concentration of OH. Similar trends are noticed in the previous nu- Scatter plots of soot volume fraction (colored by temperature)
merical studies [11,78]. In the presence of turbulent fluctuations, against mixture fraction at different positions along the axial direc-
the soot formation is highly intermittent, forming sporadic pockets tion are compared in Fig. 7. The scatter represents instantaneous
that detach from the fuel-rich zone and are convected downstream values collected at several time instants. The conditional means of
where soot oxidation takes place. soot volume fraction are included for reference. It can be observed
The FGM-T method results in a higher soot volume fraction in that a large extent of soot is present in fuel-rich mixture fractions
the middle region of the flame compared to FGM-C. Additionally, while below the stoichiometric values, no soot is observed, as it
the FGM-T model predicted soot formation further upstream than is rapidly oxidized on the lean side. At lower heights, soot vol-
the FGM-C model. This is because the soot production rate in FGM- ume fraction samples span a large range of mixture fractions and
T is higher near the flame base (z/d j 50) and in the middle region temperatures. With an increase in axial location, higher values of
(z/d j 100) than compared to FGM-C, causing higher soot concen- soot volume fraction are detected, however, the spread in mixture
tration. For a more quantitative illustration of soot formation char- fraction space tends to gradually decrease. The decreasing branch
acteristics between FGM-C and FGM-T, the scatter plots of the soot of soot volume fraction in mixture fractions beyond 0.2, vanishes
volume fraction and soot formation rates fields are compared in at the downstream position, due to enhanced mixing leading to
Fig. 6 (for the results from Fig. 5). The scatter plots are colored lower mixture fraction values. The qualitative trends of soot vol-
with the temperature. It is clear that soot formation is predom- ume fraction scatter are similar in FGM-C and FGM-T simulations.
inant in the fuel-rich regions spanning 0.1 < Z < 0.4. Within this However, more soot production is evident for FGM-T, especially at
sooting region of composition space, higher values of soot produc- higher mixture fractions than their FGM-C counterparts. Far down-
tion rates are evident for FGM-T as compared to FGM-C, which ex- stream of the flame (z/d j 180), a lower soot volume fraction is
plains the higher fv values obtained by FGM-T. Below the stoichio- found in FGM-T compared to FGM-C.
metric mixture fraction (Zst ), almost all the soot is consumed for The correlation between mixture fraction and soot volume frac-
both methods since soot oxidation is predominant in the compo- tion can be further analyzed through conditional probability den-
sition space close to the stoichiometric value. However, the distri- sity functions (P ( fv |
Z )) of soot volume fractions in Fig. 8. The PDFs
9
A. Kalbhor, D. Mira, A. Both et al. Combustion and Flame 259 (2024) 113128
prod cons
Fig. 5. Instantaneous fields of soot volume fraction f v , soot production rate (ω˙ s ), soot consumption rate (ω˙ s ) for the FGM-C (top panel) and FGM-T (bottom panel)
method. The dashed iso-lines denote Zst .
are examined for three fuel-rich intervals. The P ( fv |Z ) at differ- tion results compared to the experiment are more pronounced at
ent axial positions indicate that fluctuations in soot volume frac- lower heights for the FGM-T approach where soot volume fraction
tion are mainly concentrated within rich regions, identified by values are higher as evident in Fig. 5. Nevertheless, the upstream
2Zst <
Z < 3Zst . Moreover, a substantial amount of soot also exists translation with an underestimation is also noticed in several other
in highly rich regions
Z ≥ 3Zst . The mean values of fv are lower for numerical works [4,79], hence the overall performance of the cur-
FGM-C as compared to FGM-T. For mixture fractions below 2Zst , rent LES-FGM-DSM approaches is quite reasonable in the context
the peak value of P ( fv |
Z ) approaches zero, confirming a minimal of the current state of the art in modeling turbulent sooting flames.
amount of soot in these regions, as soot oxidation is prominent Nevertheless, it is important to note that while subprocesses like
within lean regions. In the FGM-C method, fluctuations in soot soot growth and condensation contribute to the lower soot inter-
volume fraction primarily arise from resolved fluctuations in the mittency in downstream regions, it is also crucial to emphasize the
flamelet independent variables (Z, C) and turbulent transport of impact of the LES resolution and the flapping of the jet on the in-
soot. In FGM-T, besides fluctuations in flamelet independent vari- termittency.
ables, and turbulent transport of soot, fluctuations in the chemical
source term of soot sections are taken into account through the
presumed-PDF integration during tabulation. 4.2.3. Mean soot profiles
The fluctuations in soot volume fraction are often characterized The time-averaged fields of soot volume fraction obtained from
by soot intermittency (Is ). The soot intermittency is defined exper- LES with FGM-C and FGM-T models are compared in Fig. 10a. As
imentally as the probability of observing an instantaneous value can be observed, soot is predominantly restricted to fuel-rich re-
of soot volume fraction lower than 0.03ppm [11]. In Fig. 9 the gions within the stoichiometric mean mixture fraction iso-contour,
experimental probe-resolved and numerical intermittency profiles while the peak soot volume fraction locations are found at approx-
are shown for sampled data at several time instants. The intermit- imately z/d j 125. Because of the soot oxidation (predominantly
tency profiles are favorably captured beyond z/d j > 120 confirming through OH species near stoichiometric conditions) and flow fluc-
the good prediction of soot particle oxidation and turbulent fluctu- tuations, soot particles do not exist over the complete mixture
ations (resolved). The simulation results tend to underestimate the fraction space. For the FGM-T model, substantial soot concentra-
intermittency close to the burner. The discrepancies in the simula- tion is observed at locations close to the burner exit (z/d j 30),
10
A. Kalbhor, D. Mira, A. Both et al. Combustion and Flame 259 (2024) 113128
Fig. 6. Scatter plots of instantaneous soot volume fraction (top panel) and (bottom panel) soot formation rates for FGM-T and FGM-C methods.
Fig. 7. Comparison of scatter plots of soot volume fraction colored by temperature along with conditional means of f v |
Z (dashed lines) at different axial locations for
FGM-C (a) and FGM-T (b) approaches.
11
A. Kalbhor, D. Mira, A. Both et al. Combustion and Flame 259 (2024) 113128
Fig. 8. Comparison of soot volume fraction PDF conditioned on the mixture fraction at different axial locations along the flame for FGM-C (a), and FGM-T (b).
are compared against the measurements in Fig. 10b. The RMS for
computed soot volume fraction is calculated as:
2
fvRMS = fv − fv 2 . (26)
The qualitative trends in the experimental data are reasonably re-
produced in the simulations. The normalized RMS profiles of soot
volume fraction show very good agreement with measurements
for both approaches, however, the magnitude of the soot volume
fraction is underpredicted. The peaks observed in the soot volume
fraction in the lean region away from the jet (r/d j 5) are not well
captured in the simulations. Contrary to measurements, soot al-
most ceases to exist beyond z/d j 180 in the computed results.
Fig. 9. Comparison of soot intermittency Is profiles along the centerline for FGM-C
The underprediction of soot volume fraction in far downstream re-
and FGM-T approaches (lines) with measurements (symbols). gions could be attributed to either the overprediction of OH oxi-
dation rates or the underprediction of the overall flame length in
while the soot formation is somewhat delayed in FGM-C simu- simulations.
lations. For a more quantitative illustration, the radial profiles of The axial profiles of mean and RMS soot volume fraction along
computed mean and RMS soot volume fraction at several heights the centerline axis are compared in Fig. 11. In the present turbulent
Fig. 10. Time-averaged fields of soot volume fraction for FGM-T, and FGM-C closure models (a), and a comparison between experimental (symbols) and numerical (line) data
for mean and normalized RMS of soot volume fraction profiles at several axial heights (b).
12
A. Kalbhor, D. Mira, A. Both et al. Combustion and Flame 259 (2024) 113128
Fig. 11. Comparison between experimental (symbols) and numerical (line) profiles of mean and RMS of soot volume fraction at the centerline for FGM-T and FGM-C ap-
proaches.
jet flame, the overall prediction of soot formation along the center- high computational efficiency and low computational cost. This as-
line is mainly controlled by surface growth (in the middle region) pect will be addressed in Section 4.3.
and oxidation (due to OH) as shown in [4]. A reasonable agreement As previously mentioned, the FGM-T model accounts for the
between simulated and measured soot volume fraction profiles is subgrid-scale chemistry-soot-turbulence interactions, as the tabu-
obtained, but the peak value of fv is under-predicted by a factor lated source terms are integrated with the presumed β -PDF func-
of two in the simulations. The prediction of RMS fluctuations of tion. To investigate the impact of the presumed-PDF model on soot
soot volume fraction is similar in magnitude to the mean with ap- prediction with FGM-T, additional simulation is carried out by ne-
proximately a factor 2 lower. This underprediction is aligned with glecting the influence of subgrid-scale fluctuations on tabulated
the state-of-the-art results from this flame, where a comparable soot source terms by using a δ -PDF. In Fig. 12, the axial profiles of
under/over prediction is found with other approaches [4,11,12,31]. mean and RMS soot volume fraction along the centerline are com-
From the present LES results, it can be inferred that the FGM- pared for the two PDF functions. The soot volume fraction profiles
DSM formalism can fairly capture the soot volume fraction distri- are only marginally influenced by the exclusion of the presumed-
bution under turbulent conditions. Besides, for both approaches, PDF treatment accounting for subgrid-scale interactions on soot
the position of the peak soot volume fraction is slightly shifted source terms. Such a response of soot formation can be elucidated
upstream compared to experiments, indicating that the soot con- by analyzing the spatial distribution of soot volume fraction and
sumption is predicted early in simulations. Nevertheless, an early mixture fraction variance (which characterize the influence of tur-
formation of soot is evident for the FGM-T approach compared to bulence fluctuations on scalar mixing). The time-averaged fields of
FGM-C. soot volume fraction (with δ -PDF) and mixture fraction variance
In the current LES, soot mass fractions slowly increase while are shown in Fig. 13. As expected, the high variance of mixture
going downstream from the inlet toward a statistically steady state fraction is primarily found in the shear layers generated by strong
from the initial no-soot condition. Therefore, capturing the chemi- velocity gradients between the main jet and coflow. In the base re-
cal trajectories of soot evolution (from the gas phase to the steady gion of the flame, close to the burner, the mixture fraction variance
state) becomes crucial in the accurate prediction of soot formation. is significant. However, soot formation is not predominant in this
In the FGM-C approach, soot subprocesses are explicitly computed region. On the contrary, the soot formation zone is mainly spanned
using the local soot mass fractions. This yields a better qualitative in the mid-flame region, where mixture fraction gradients are low
description of unsteady soot evolution for the LES. On the contrary, (leading to low variance). As a result, the subgrid-scale fluctuations
in FGM-T soot source terms are calculated and stored for soot mass only slightly affect the soot source terms. Consequently, overall
fractions in a steady-state flamelet, therefore, the chemical trajec- soot formation is found to be only marginally affected by subgrid-
tories concerning the formation of soot from the gas phase to the scale turbulent fluctuations in the present flame.
steady state are not explicitly retained in the FGM tabulation strat-
egy. Naturally, an extension of FGM to incorporate the chemical 4.2.4. Time-averaged particle size distributions
trajectories of soot formation would require augmentation of the The coupling of LES with the sectional method provides infor-
database with unsteady flamelets (at every level of scalar dissipa- mation on the spatio-temporal evolution of the soot PSD. Hence,
tion rate), and an additional controlling variable to entirely param- the calculated time-averaged PSDs at different axial locations along
eterize the reaction progress of soot, making manifold generation the flame centerline are plotted in Fig. 14 for FGM-T and FGM-
more complex. Therefore, this aspect is left out of the scope of the C. The evolution of the PSD along its trajectory in the flame is
present work. Moreover, the non-linear dependency of soot pro- strongly correlated to the particle history characterized by a suc-
duction rates on soot variables is not included in the present FGM- cession of chemical and collisional processes associated with soot
T formulation. Hence, the direct look-up of soot production rates formation. The time-averaged soot PSDs feature mainly unimodal
may lead to their overestimation. As a result, higher soot volume shapes for FGM-C and FGM-T approaches. The number density of
fractions are noticed at lower axial positions for FGM-T compared larger-sized particles increases during this process as nucleated
to FGM-C. Nevertheless, the results for both approaches are overall soot particles grow primarily through surface reactions. For FGM-T,
in good agreement with the current state-of-the-art of this flame the shift of PSD towards large diameters is observed as compared
and the FGM-T approach shows high potential for LES due to the to FGM-C, which translates into higher soot volume fractions pre-
13
A. Kalbhor, D. Mira, A. Both et al. Combustion and Flame 259 (2024) 113128
Fig. 12. Comparison between experimental (symbols) and numerical (line) profiles of mean and RMS of soot volume fraction at the centerline for FGM-T case with β and δ
PDF integration applied to tabulated soot rates.
ω˙ s,i = ω˙ s,ij (φg , φs ), (27)
j
= kgj (φg )ζi j (φg )ij (φs ) ∀ i ∈ [1, nsec ], (28)
j
14
A. Kalbhor, D. Mira, A. Both et al. Combustion and Flame 259 (2024) 113128
15
A. Kalbhor, D. Mira, A. Both et al. Combustion and Flame 259 (2024) 113128
Fig. 15. Time-averaged fields of soot volume fraction for FGM-C, and FGM-CR closure models (a), and a comparison between experimental (symbols) and numerical (line)
data for mean and normalized RMS of soot volume fraction profiles at several axial heights (b).
has certain limitations, for example, soot-independent treatment of with a presumed PDF (probability density function) approach tends
the soot production term, and a lack of information on the gas-to- to yield good quantitative soot prediction, making it a promising
soot history effects in flamelets. Therefore, in simulation applica- tool to study soot production in light of industrial applications
tions where the unsteady evolution of soot quantities is not of in- with LES. However, it was found that tabulating soot chemistry has
terest, the FGM-T approach is more suitable for predicting the soot limitations in capturing the transient evolution of soot. Therefore,
formation. Especially, in simulations of practical combustion sys- the tabulated soot source term approach requires further improve-
tems, the FGM-T is an excellent strategy to gain an understanding ments to meet limitations regarding capturing soot history effects.
of soot formation and information on size distribution, at afford- On the other end, the model that solves for the complete set of
able computational cost. On the other hand, for more fundamental soot formation subprocesses provides a better qualitative descrip-
and parametric studies, the FGM-C method is recommended as it tion of soot formation and evolution. It is further observed that
accounts for the non-linear interactions between the soot and gas the separation of soot and gas phase terms only marginally im-
phase, without any modeling assumptions, and can be more reli- pacts soot formation prediction for the run-time soot source term
able for highly transient cases. computation model in turbulent conditions.
In summary, the present study demonstrates that the sectional
5. Concluding remarks soot model coupled with FGM chemistry is a feasible and efficient
approach to predict soot production in turbulent flames. The LES
This study presents two strategies based on the discrete sec- framework developed in this work for the FGM-DSM coupling does
tional method coupled with FGM tabulated chemistry in the not consider subgrid-scale interactions between soot and turbu-
context of LES for the prediction of soot formation in turbulent lence. Including subgrid-scale soot-turbulence interaction models
non-premixed flames. The performance of the two strategies (e.g. Ref. [7,11]) in the LES would be an important extension to in-
for modeling soot source terms is assessed on a turbulent vestigate their impact on overall soot prediction. In addition, future
non-premixed jet flame with a focus on the prediction of soot for- work could consider incorporating developing new models or ex-
mation and particle size distributions. The LES results for the gas tending the linearization model used in the tabulated soot chem-
phase and soot phase are compared against the available experi- istry approach to account for the non-linear effects of the soot pro-
mental data. Despite some discrepancies, the results for the soot duction rate. Moreover, the authors acknowledge that the modeling
phase show reasonable agreement with the experimental results. aspects related to thermal radiation, preferential diffusion, and soot
The LES study suggests that the soot source term closure in agglomeration hold significant potential for enhancing the accuracy
FGM-DSM coupling substantially influences the quantitative and of the proposed soot modeling framework and are thus considered
qualitative prediction of soot formation in turbulent conditions. essential components for future research work.
The observed differences in soot prediction for the investigated
soot source-term models can be attributed to their capabilities in Novelty and significance
capturing unsteady chemical trajectories of soot formation, which
evolve at a slower time scale than flame propagation. The compu- Applications of detailed soot models, such as the discrete sec-
tationally efficient model based on tabulated soot formation rates tional method, have become increasingly important with emission
16
A. Kalbhor, D. Mira, A. Both et al. Combustion and Flame 259 (2024) 113128
regulations on soot particle size. However, several studies that em- [10] K. Netzell, H. Lehtiniemi, F. Mauss, Calculating the soot particle size distri-
ploy sectional methods for soot prediction in turbulent conditions bution function in turbulent diffusion flames using a sectional method, Proc.
Combust. Inst. 31 (1) (2007) 667–674.
rely on the explicit computation of soot reaction rates, which suf- [11] P. Rodrigues, B. Franzelli, R. Vicquelin, O. Gicquel, N. Darabiha, Coupling an
fer from high computational costs with an increase in the num- LES approach and a soot sectional model for the study of sooting turbulent
ber of soot sections. To address this, the present work offers a non-premixed flames, Combust. Flame 190 (2018) 477–499.
[12] Y. Xuan, G. Blanquart, Effects of aromatic chemistry-turbulence interactions on
computationally efficient tabulated soot chemistry-based sectional soot formation in a turbulent non-premixed flame, Proc. Combust. Inst. 35 (2)
soot modeling framework for the LES approach. The compara- (2015) 1911–1919.
tive analysis presented in this study demonstrates the good pre- [13] F. Ferraro, S. Gierth, S. Salenbauch, W. Han, C. Hasse, Soot particle size distribu-
tion reconstruction in a turbulent sooting flame with the split-based extended
dictive capabilities of the proposed method despite several mod-
quadrature method of moments, Phys. Fluids 34 (7) (2022) 075121.
eling approximations. Overall, the proposed LES framework of- [14] M. Grader, C. Eberle, P. Gerlinger, Large-eddy simulation and analysis of a soot-
fers an attractive modeling choice for reducing the computational ing lifted turbulent jet flame, Combust. Flame 215 (2020) 458–470.
[15] S. Rigopoulos, Modelling of soot aerosol dynamics in turbulent flow, Flow Tur-
cost of sooting flame calculations in industry-relevant conditions
bul. Combust. 103 (3) (2019) 565–604.
while maintaining good predictive capabilities. The proposed tabu- [16] T. Poinsot, D. Veynante, Theoretical and numerical combustion, RT Edwards,
lated soot chemistry model, therefore, has the potential to signif- Inc., 2005.
icantly advance the design and development of clean combustion [17] F. Bisetti, G. Blanquart, M.E. Mueller, H. Pitsch, On the formation and early
evolution of soot in turbulent nonpremixed flames, Combust. Flame 159 (1)
systems. (2012) 317–335.
[18] A. Attili, F. Bisetti, M.E. Mueller, H. Pitsch, Formation, growth, and transport of
Declaration of Competing Interest soot in a three-dimensional turbulent non-premixed jet flame, Combust. Flame
161 (7) (2014) 1849–1865.
[19] A. Wick, A. Attili, F. Bisetti, H. Pitsch, DNS-driven analysis of the
The authors declare that they have no known competing finan- flamelet/progress variable model assumptions on soot inception, growth, and
cial interests or personal relationships that could have appeared to oxidation in turbulent flames, Combust. Flame 214 (2020) 437–449.
[20] D. Aubagnac-Karkar, J. Michel, O. Colin, P. Vervisch-Kljakic, N. Darabiha, Sec-
influence the work reported in this paper. tional soot model coupled to tabulated chemistry for diesel RANS simulations,
Combust. Flame 162 (8) (2015) 3081–3099.
CRediT authorship contribution statement [21] S. Yang, J.K. Lew, M.E. Mueller, Large eddy simulation of soot evolution in tur-
bulent reacting flows: strain-sensitive transport approach for polycyclic aro-
matic hydrocarbons, Combust. Flame 220 (2020) 219–234.
Abhijit Kalbhor: Conceptualization, Methodology, Visualization, [22] D. Mira, O. Lehmkuhl, A. Both, P. Stathopoulos, T. Tanneberger, T.G. Reichel,
Investigation, Formal analysis, Writing – original draft. Daniel C.O. Paschereit, M. Vázquez, G. Houzeaux, Numerical characterization of a pre-
mixed hydrogen flame under conditions close to flashback, Flow Turbul. Com-
Mira: Conceptualization, Visualization, Resources, Writing – review
bust. 104 (2) (2020) 479–507.
& editing. Ambrus Both: Validation, Writing – review & editing. [23] B. Franzelli, L. Tardelli, M. Stöhr, K. Geigle, P. Domingo, Assessment of LES of
Jeroen van Oijen: Conceptualization, Funding acquisition, Supervi- intermittent soot production in an aero-engine model combustor using high-
sion, Writing – review & editing. -speed measurements, Proc. Combust. Inst. (2022).
[24] S. Gkantonas, M. Sirignano, A. Giusti, A. D’Anna, E. Mastorakos, Comprehensive
soot particle size distribution modelling of a model Rich-Quench-Lean burner,
Acknowledgments Fuel 270 (2020) 117483.
[25] J. Benajes, J.M. García-Oliver, J.M. Pastor, I. Olmeda, A. Both, D. Mira, Analysis of
local extinction of a n-heptane spray flame using large-eddy simulation with
The research leading to these results has received fund- tabulated chemistry, Combust. Flame 235 (2022) 111730.
ing from the European Union’s Horizon 2020 Programme un- [26] H. Bao, A. Kalbhor, N. Maes, B. Somers, J. van Oijen, Investigation of soot for-
der the ESTiMatE project, grant agreement No. 821418, and the mation in n-dodecane spray flames using LES and a discrete sectional method,
Proc. Combust. Inst. (2022).
AHEAD PID2020-118387RB-C33 and SAFLOW TED2021-131618B- [27] B. Franzelli, E. Riber, B. Cuenot, M. Ihme, Numerical modeling of soot pro-
C21 projects from the Ministerio de Ciencia e Innovación. DM duction in aero-engine combustors using large eddy simulations, Turbo Expo:
acknowledges the grant from the Ministerio de Ciencia e Inno- Power for Land, Sea, and Air, vol. 56697, American Society of Mechanical En-
gineers, 2015. p. V04BT04A049
vación, Ayudas para contratos Ramón y Cajal (RYC) 2021: RYC2021-
[28] J. Consalvi, F. Nmira, Transported scalar PDF modeling of oxygen-enriched tur-
034654-I. The authors gratefully acknowledge the Partnership for bulent jet diffusion flames: soot production and radiative heat transfer, Fuel
Advanced Computing in Europe (PRACE) grant Soot Aero for com- 178 (2016) 37–48.
[29] L. Pachano, C. Xu, J.M. García-Oliver, J.M. Pastor, R. Novella, P. Kundu, A two-e-
putational resources and technical support in the numerical simu-
quation soot-in-flamelet modeling approach applied under spray a conditions,
lations. Combust. Flame 231 (2021) 111488.
[30] S. Yang, J. Lew, M. Mueller, Large eddy simulation of soot evolution in turbu-
References lent reacting flows: presumed subfilter PDF model for soot–turbulence–chem-
istry interactions, Combust. Flame 209 (2019) 200–213.
[1] EEA, Regulation (EC) no 715/2007 of the European parliament and of the coun- [31] L. Tian, M. Schiener, R. Lindstedt, Fully coupled sectional modelling of soot
cil of 20 June 2007 on type approval of motor vehicles with respect to emis- particle dynamics in a turbulent diffusion flame, Proc. Combust. Inst. 38 (1)
sions from light passenger and commercial vehicles (Euro 5 and Euro 6), 2022. (2021) 1365–1373.
[2] Committee on aviation environmental protection report (doc 10126). ed. [32] A. Felden, E. Riber, B. Cuenot, Impact of direct integration of analytically re-
by international civil aviation organization, 2023, (https://store.icao.int/en/ duced chemistry in LES of a sooting swirled non-premixed combustor, Com-
committee- on- aviation- environmental- protection- report- doc- 10126). bust. Flame 191 (2018) 270–286.
[3] S. Rigopoulos, Population balance modelling of polydispersed particles in reac- [33] H.M. Colmán, N. Darabiha, D. Veynante, B. Fiorina, A turbulent combustion
tive flows, Prog. Energy Combust. Sci. 36 (4) (2010) 412–443. model for soot formation at the LES subgrid-scale using virtual chemistry ap-
[4] B. Sun, S. Rigopoulos, Modelling of soot formation and aggregation in turbulent proach, Combust. Flame 247 (2023) 112496.
flows with the LES-PBE-PDF approach and a conservative sectional method, [34] M.E. Mueller, H. Pitsch, Large eddy simulation subfilter modeling of soot-tur-
Combust. Flame 242 (2022) 112152. bulence interactions, Phys. Fluids 23 (11) (2011) 115104.
[5] M. Balthasar, M. Kraft, A stochastic approach to calculate the particle size [35] S.T. Chong, V. Raman, M.E. Mueller, P. Selvaraj, H.G. Im, Effect of soot model,
distribution function of soot particles in laminar premixed flames, Combust. moment method, and chemical kinetics on soot formation in a model aircraft
Flame 133 (3) (2003) 289–298. combustor, Proc. Combust. Inst. 37 (1) (2019) 1065–1074.
[6] M. Mueller, G. Blanquart, H. Pitsch, Hybrid method of moments for mod- [36] L. Berger, A. Wick, A. Attili, M.E. Mueller, H. Pitsch, Modeling subfilter soot-tur-
eling soot formation and growth, Combust. Flame 156 (6) (2009) 1143– bulence interactions in large eddy simulation: an a priori study, Proc. Combust.
1155. Inst. 38 (2) (2021) 2783–2790.
[7] M. Mueller, H. Pitsch, LES model for sooting turbulent nonpremixed flames, [37] H.M. Colmán, A. Attili, M.E. Mueller, Large eddy simulation of turbulent non-
Combust. Flame 159 (6) (2012) 2166–2180. premixed sooting flames: presumed subfilter PDF model for finite-rate oxida-
[8] A. D’anna, J. Kent, A model of particulate and species formation applied to tion of soot, Combust. Flame (2023) 112602.
laminar, nonpremixed flames for three aliphatic-hydrocarbon fuels, Combust. [38] H. El-Asrag, S. Menon, Large eddy simulation of soot formation in a turbulent
Flame 152 (4) (2008) 573–587. non-premixed jet flame, Combust. Flame 156 (2) (2009) 385–395.
[9] F. Gelbard, Y. Tambour, J. Seinfeld, Sectional representations for simulating [39] C.D. Pierce, P. Moin, Progress-variable approach for large-eddy simulation of
aerosol dynamics, J. Colloid Interface Sci. 76 (2) (1980) 541–556. non-premixed turbulent combustion, J. Fluid Mech. 504 (2004) 73–97.
17
A. Kalbhor, D. Mira, A. Both et al. Combustion and Flame 259 (2024) 113128
[40] F. Sewerin, S. Rigopoulos, An LES-PBE-PDF approach for predicting the soot [61] S.P. Roy, Aerosol-dynamics-based soot modeling of flames, Pennsylvania State
particle size distribution in turbulent flames, Combust. Flame 189 (2018) University, 2014 Ph.D. thesis.
62–76. [62] J. Appel, H. Bockhorn, M. Frenklach, Kinetic modeling of soot formation with
[41] J. van Oijen, L. de Goey, Modelling of premixed laminar flames using flamelet– detailed chemistry and physics: laminar premixed flames of C2 hydrocarbons,
generated manifolds, Combust. Sci. Technol. 161 (1) (20 0 0) 113–137. Combust. Flame 121 (1–2) (20 0 0) 122–136.
[42] J. van Oijen, A. Donini, R. Bastiaans, J. ten Thije Boonkkamp, L. de Goey, [63] S. Kumar, D. Ramkrishna, On the solution of population balance equa-
State-of-the-art in premixed combustion modeling using flamelet generated tions by discretization–I. A fixed pivot technique, Chem. Eng. Sci. 51 (8) (1996)
manifolds, Prog. Energy Combust. Sci. 57 (2016) 30–74. 1311–1332.
[43] D. Carbonell, A. Oliva, C.D. Perez-Segarra, Implementation of two-equation soot [64] C. Hoerlle, Modelling of soot formation based on the Discrete Sectional
flamelet models for laminar diffusion flames, Combust. Flame 156 (3) (2009) Method: CO2 effects and coupling with the FGM technique, Universidade Fed-
621–632. eral do Rio Grande do Sul, Porto Alegre, Brazil, 2020 Ph.D. thesis.
[44] A. Kalbhor, D. Mira, J. van Oijen, A computationally efficient approach for soot [65] A. Kalbhor, J. van Oijen, Effects of hydrogen enrichment and water vapour dilu-
modeling with discrete sectional method and FGM chemistry, Combust. Flame tion on soot formation in laminar ethylene counterflow flames, Int. J. Hydrogen
255 (2023) 112868. Energy 45 (43) (2020) 23653–23673.
[45] A. Vreman, An eddy-viscosity subgrid-scale model for turbulent shear flow: [66] M. Ihme, H. Pitsch, Modeling of radiation and nitric oxide formation in turbu-
algebraic theory and applications, Phys. Fluids 16 (10) (2004) 3670–3681. lent nonpremixed flames using a flamelet/progress variable formulation, Phys.
[46] A. Both, O. Lehmkuhl, D. Mira, M. Ortega, Low-dissipation finite element strat- Fluids 20 (5) (2008) 055110.
egy for low mach number reacting flows, Comput. Fluids 200 (2020) 104436. [67] J. Zhang, C.R. Shaddix, R.W. Schefer, Design of “model-friendly” turbulent
[47] N. Peters, Laminar flamelet concepts in turbulent combustion, Proc. Combust. non-premixed jet burners for C2+ hydrocarbon fuels, Rev. Sci. Instrum. 82 (7)
Inst. 21 (1) (1988) 1231–1250. (2011) 074101.
[48] J. van Oijen, Flamelet-generated manifolds: development and application to [68] International sooting flame (ISF) workshop, 2022, (https://www.adelaide.edu.
premixed laminar flames, Eindhoven University of Technology, Eindhoven, the au/cet/isfworkshop/).
Netherlands, 2002 Ph.D. thesis. [69] S.B. Pope, Ten questions concerning the large-eddy simulation of turbulent
[49] R. Bilger, S. Stårner, R. Kee, On reduced mechanisms for methane-air combus- flows, New J. Phys. 6 (1) (2004) 35.
tion in nonpremixed flames, Combust. Flame 80 (2) (1990) 135–149. [70] A. Kempf, M. Klein, J. Janicka, Efficient generation of initial-and inflow-condi-
[50] N. Malik, T. Løvås, F. Mauss, The effect of preferential diffusion on the soot tions for transient turbulent flows in arbitrary geometries, Flow Turbul. Com-
initiation process in ethylene diffusion flames, Flow Turbul. Combust. 87 (2) bust. 74 (1) (2005) 67–84.
(2011) 293–312. [71] L. Somers, The simulation of flat flames with detailed and reduced chemical
[51] A. Attili, F. Bisetti, M.E. Mueller, H. Pitsch, Effects of non-unity lewis number models, Eindhoven University of Technology, Eindhoven, the Netherlands, 1994
of gas-phase species in turbulent nonpremixed sooting flames, Combust. Flame Ph.D. thesis.
166 (2016) 192–202. [72] Y. Wang, A. Raj, S. Chung, A PAH growth mechanism and synergistic effect on
[52] P. Domingo, L. Vervisch, D. Veynante, Large-eddy simulation of a lifted PAH formation in counterflow diffusion flames, Combust. Flame 160 (9) (2013)
methane jet flame in a vitiated coflow, Combust. Flame 152 (3) (2008) 1667–1676.
415–432. [73] Y. Wang, A. Raj, S. Chung, Soot modeling of counterflow diffusion flames of
[53] A. Vreman, B. Albrecht, J. Van Oijen, L. De Goey, R. Bastiaans, Premixed and ethylene-based binary mixture fuels, Combust. Flame 162 (3) (2015) 586–596.
nonpremixed generated manifolds in large-eddy simulation of Sandia flame D [74] Y. Wang, S. Chung, Strain rate effect on sooting characteristics in laminar coun-
and F, Combust. Flame 153 (3) (2008) 394–416. terflow diffusion flames, Combust. Flame 165 (2016) 433–444.
[54] L. Ma, X. Huang, D. Roekaerts, Large eddy simulation of CO2 diluted oxy-fuel [75] A. Kalbhor, J. van Oijen, An assessment of the sectional soot model and FGM
spray flames, Fuel 201 (2017) 165–175. tabulated chemistry coupling in laminar flame simulations, Combust. Flame
[55] Y. Hu, R. Kurose, Large-eddy simulation of turbulent autoigniting hydrogen 229 (2021) 111381.
lifted jet flame with a multi-regime flamelet approach, Int. J. Hydrogen Energy [76] F. Liu, J.-L. Consalvi, F. Nmira, The importance of accurately modelling soot and
44 (12) (2019) 6313–6324. radiation coupling in laminar and laboratory-scale turbulent diffusion flames,
[56] J.C. Massey, Z.X. Chen, N. Swaminathan, Modelling heat loss effects in the large Combust. Flame (2022) 112573.
eddy simulation of a lean swirl-stabilised flame, Flow Turbul. Combust. 106 (4) [77] M. Vázquez, G. Houzeaux, S. Koric, A. Artigues, J. Aguado-Sierra, R. Arís,
(2021) 1355–1378. D. Mira, H. Calmet, F. Cucchietti, H. Owen, et al., Alya: multiphysics engineer-
[57] A. Both, High-fidelity numerical simulations of reacting flows with tabulated ing simulation toward exascale, J. Comput. Sci. 14 (2016) 15–27.
chemistry, Universitat Politèchnica de Catalunya, Spain, 2023 Ph.D. thesis. [78] Y. Xuan, G. Blanquart, A flamelet-based a priori analysis on the chemistry tab-
[58] A. Both, O. Lehmkuhl, D. Mira, M. Ortega, Low-dissipation finite element ulation of polycyclic aromatic hydrocarbons in non-premixed flames, Combust.
strategy for low mach number reacting flows, Comput. Fluids 200 (2020) Flame 161 (6) (2014) 1516–1525.
104436. [79] H. Colmán, A. Cuoci, N. Darabiha, B. Fiorina, A virtual chemistry model for
[59] C. Hoerlle, F. Pereira, Effects of CO2 addition on soot formation of ethylene soot prediction in flames including radiative heat transfer, Combust. Flame 238
non-premixed flames under oxygen enriched atmospheres, Combust. Flame (2022) 111879.
203 (2019) 407–423.
[60] S. Friendlander, Smoke, dust and haze: fundamentals of aerosol dynamics, Ox-
ford University Press, New York, USA, 20 0 0.
18