You are on page 1of 16

Combustion and Flame 235 (2022) 111712

Contents lists available at ScienceDirect

Combustion and Flame


journal homepage: www.elsevier.com/locate/combustflame

A DNS study of extreme and leading points in lean hydrogen-air


turbulent flames - part II: Local velocity field and flame topology
HsuChew Lee a,b, Peng Dai a, Minping Wan a,b,∗, Andrei N. Lipatnikov c
a
Guangdong Provincial Key Laboratory of Turbulence Research and Applications, Department of Mechanics and Aerospace Engineering, Southern University
of Science and Technology, Shenzhen 518055, China
b
Guangdong-Hong Kong-Macao Joint Laboratory for Data-Driven Fluid Mechanics and Engineering Applications, Southern University of Science and
Technology, Shenzhen 518055, China
c
Department of Mechanics and Maritime Sciences, Chalmers University of Technology, Gothenburg SE-412 96, Sweden

a r t i c l e i n f o a b s t r a c t

Article history: Data obtained in recent direct numerical simulations (Lee et al.) of statistically one-dimensional and pla-
Received 22 March 2021 nar, lean complex-chemistry hydrogen-air flames characterized by three different Karlovitz numbers Ka
Revised 19 August 2021
ranging from 3 to 33 are further analyzed in order to explore local characteristics and structure of (i)
Accepted 21 August 2021
extreme points characterized by the peak (over the computational domain) Fuel Consumption Rate (FCR)
Available online 9 September 2021
or Heat Release Rate (HRR) and (ii) leading points that are also characterized by a high FCR or HRR, but
Keywords: advance furthest into unburned reactants. Results show that, on the one hand, common characteristics of
Hydrogen flame perturbations (curvature, strain and stretch rates, displacement speed) fluctuate significantly in the
DNS extreme or leading, FCR or HRR points and are different in different flames. Moreover, other two-point
Turbulent combustion local quantities such as the local gradients of combustion progress variables or species (e.g., the radical H)
Lewis number mass fractions are different in different flames. Therefore, a common simple configuration of a perturbed
Turbulent consumption speed
laminar flame cannot be used as a catchall model of the entire local structure of zones surrounding the
Diffusive-thermal effects
Leading point concept
discussed points at various Ka. On the other hand, single-point local characteristics (temperature, species
mass fractions, rates of their production) of the FCR extreme points are comparable in all three turbulent
flames and in the critically strained planar laminar flame. In particular, the FCRs in the extreme points
fluctuate weakly and are approximately equal to each other and to the peak FCR in the critically strained
laminar flame. The latter finding implies that (i) the maximum FCR evaluated in the critically strained
laminar flame could be used to characterize, in a first approximation, the local FCR in the extreme or
leading points in turbulent flames, thus, supporting the leading point concept, and (ii) almost the same
extreme FCR can be reached in substantially different local burning structures.
© 2021 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

1. Introduction different conditions over decades, as reviewed elsewhere [5,6], see


also a recent paper [7]. A similar effect was also found (i) in ex-
The concept of global warming motivates rapidly growing in- periments with highly turbulent syngas-air mixtures [8–10], while
terest in utilizing hydrogen as a renewable carbon-free fuel. For the effect is less pronounced in such mixtures, and (ii) in a Direct
that reason, burning of H2 in various engines has been attracting Numerical Simulation (DNS) study of highly turbulent single-step-
significant amount of attention [1–4], with hydrogen flames mov- chemistry flames characterized by various Lewis numbers Le = α /D
ing to the forefront of the research agenda in the combustion field. ranging from 0.34 to 1.2 [11,12]. Here, D and α are molecular dif-
From the fundamental perspective, the most challenging peculiar- fusivity of deficient reactant in a mixture and molecular heat dif-
ity of turbulent combustion of lean hydrogen-air mixtures (when fusivity of the mixture, respectively.
compared to common fossil fuels) consists of abnormally high ra- The abnormally high ratios of UT /SL in lean hydrogen-air mix-
tios UT /SL of turbulent burning velocity UT to the laminar flame tures are commonly attributed [5,6,13–15] to diffusive-thermal ef-
speed SL , well documented in such mixtures by various research fects [16], i.e., to local variations in the mixture composition and
groups in a number of experiments performed under substantially temperature due to imbalance of reactant and heat fluxes to/from
inherently laminar reaction zones stretched and wrinkled by tur-
bulent eddies. While this governing physical mechanism is well

Corresponding author. recognized, prediction of the high ratios of UT /SL is still one of the
E-mail address: wanmp@sustech.edu.cn (M. Wan). major fundamental challenges to the combustion community. In

https://doi.org/10.1016/j.combustflame.2021.111712
0010-2180/© 2021 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
H. Lee, P. Dai, M. Wan et al. Combustion and Flame 235 (2022) 111712

this regard, the so-called leading point concept pioneered by the 2. The aforementioned local increase in ω˙ t,max F or W˙ t,max
T in the
Russian school [13,16] appears to be a promising approach with extreme points stems from an increase in the local tempera-
the potential to predict a strong increase in UT /SL in lean H2 -air ture (solely in the FCR extreme points), equivalence ratio, and
mixtures due to the diffusive-thermal effects. According to the con- mass fractions of the radicals H, O, OH when compared to these
cept, (i) propagation of a premixed turbulent flame is hypothesized quantities evaluated at the same local combustion progress
to be controlled by burning in the leading points that advance fur- variable in the unperturbed laminar flame.
thest into fresh reactants and (ii) the local structure of the lead- 3. While such extreme FCR and HRR points are sometimes ob-
ing points is hypothesized to be strongly perturbed and, hence, is served sufficiently far from the leading edge of the mean flame
strongly affected by the diffusive-thermal phenomena. Thus, tur- brush, zones (leading points) characterized by large local values
bulent burning rate is hypothesized to be controlled by the local of ω˙ tF /ω˙ L,max
F or W˙ tT /W˙ L,max
T , which are sufficiently close to the
combustion rate in the highly perturbed (critically strained [13] or extreme ratios ω˙ F /ω˙ F and W˙ T /W˙ T , respectively, are
t,max L,max t,max L,max
critically curved [17,18]) leading points, rather than by the local observed in the vicinity of the leading edge almost always.
combustion rate in weakly perturbed, inherently laminar flames. 4. The local thermo-chemical structures (species mass fractions,
Consequently, a large magnitude of the diffusive-thermal effects in equivalence ratio, temperature, reaction rates, etc.) of the ex-
turbulent flames can be predicted. treme and leading points are basically similar. On the contrary,
The use of the leading point concept allowed Karpov et al. the local thermo-chemical structures of the FCR and HRR points
[17] to quantitatively predict high turbulent burning rates mea- are substantially different.
sured in very lean hydrogen flames, but these earlier Reynolds- 5. In the FCR extreme or leading points, the major reaction path-
Averaged Navier-Stokes (RANS) simulations were limited to a ways are basically similar to the major reaction pathways in the
single-step chemistry. Over the past decade, the leading point con- laminar flame zone characterized by the peak FCR. On the con-
cept was supported in theoretical studies [19–21] and in recent trary, in the HRR extreme or leading points, the major reaction
RANS simulations [22] that allowed for complex combustion chem- pathways differ significantly from the major reaction pathways
istry. Moreover, the concept was supported by analyzing experi- in the laminar flame zone characterized by the peak HRR.
mental [9,10,23] or DNS [24–26] data. However, certain relevant
fundamental issues have not yet been resolved. In particular, a re- In the present paper, the previous analysis of the thermochem-
cent DNS study [27] shows that, when the normalized rms turbu- ical structure of extreme and leading points and reaction rates in
lent velocity u /SL is increased, (i) the local structure of the leading such points is complemented with investigation of (i) the evolu-
flame kernels does tend gradually to the local structure of the crit- tion and topology of these points and (ii) the local characteristics
ically strained planar laminar flame, (ii) but, nevertheless, there are of flame perturbations, commonly adopted in the literature.
substantial differences at finite values of u /SL (up to 100), explored In the next section, methodology of the present analysis is re-
in the cited work. It is also worth noting that various critically per- ported. Results are discussed in Sect. 3, followed by conclusions.
turbed laminar flames (e.g., the critically strained cylindrical lam-
inar flame, a spherical flame ball [16], etc.) were considered to be 2. Methodology
models of the leading points, with local burning rates in different
critically perturbed laminar flames being substantially different, at 2.1. DNS attributes
least in the case of a single-step chemistry [28].
Thus, while the leading point concept appears to be an attrac- Three-dimensional DNS of freely-propagating statistically planar
tive approach to modeling abnormally high ratios of UT /SL , well turbulent H2 -air premixed flames (the equivalence ratio φ = 0.5,
documented in lean hydrogen mixtures, it requires further study, unburned gas temperature Tu =300 K, and pressure P =1 atm) were
validation, and clarification of some fundamental issues, e.g.: performed using a detailed chemical mechanism (9 species and 22
reactions) by Kéromnès et al. [30] and employing the solver DINO
• Is turbulent burning rate in lean hydrogen mixtures controlled [31], which deals with the low-Mach-number formulation of the
by characteristics of highly perturbed reaction zones, rather Navier-Stokes equations.
than by characteristics of the unperturbed or weakly stretched The computational domain of Lx × Ly × Lz was discretized using
laminar flames? a uniform Cartesian grid of Nx × Ny × Nz cells. Here, Lx = Lz = 2.4
• What is the structure of such zones? mm, Ly = β Lx , Nx = Nz , Ny = β Nx , and the values of Lx , Nx , and
• Are such zones linked with the leading points? β are reported in Table 1. Turbulent inflow and outflow bound-
• Are there different types of leading highly perturbed reaction ary conditions were imposed along the streamwise (y) direction
zones? and periodic boundary conditions were specified along the span-
wise and transverse directions. The inlet flow rms velocity was
This work aims at partially filling this knowledge gap by analyz-
small, but turbulence was generated using the linear velocity forc-
ing DNS data to explore (i) the most perturbed local reaction zones
ing method [32–35] between y = 0.5Lx and y = 8Lx (cases C and
and (ii) the leading reaction zones in lean hydrogen-air premixed
C1) or y = 10Lx (other cases). Characteristics of the forced turbu-
turbulent flames characterized by substantially different Karlovitz
lence are reported in Fig. 2 in Ref. [29].
numbers. In Part I of the present study [29], thermochemical struc-
In each case, inert constant-density turbulence was first sim-
ture of such zones and the local reaction rates are explored. The
ulated for at least 50τt , followed by embedding a steady un-
major results are summarized below:
strained laminar flame into the flow field. Subsequently, the flame
1. The peak (over the computational domain) values ω˙ t,max F and was allowed to evolve for at least another 28τt . Here, τt = L/u is
W˙ t,max
T of the Fuel Consumption Rate (FCR) and Heat Release eddy turnover time and L is an integral length scale reported in
Rate (HRR), respectively, are significantly larger than the peak Table 1. To keep the flame within the forced-flow subdomain, the
FCR, ω˙ L,max
F , and the peak HRR, W˙ L,max
T , respectively, in the un- mean inlet velocity is manually modified with instantaneous in-
perturbed laminar flame, with the effect magnitude increasing crease/decrease of its value in a stepwise fashion when necessary.
weakly (significantly) with Ka for the FCR (HRR, respectively). The simulation conditions are summarized in Table 1, where
Here, superscripts F and T refer to fuel and temperature, re- SL , δLT = (Tb − Tu )/max{|dT /dx|}, and τ f = δLT /SL are the laminar
spectively, subscripts t and L designate turbulent and laminar flame speed, thickness, and time scale, respectively; η = LRet−3/4 is
flames, respectively. the Kolmogorov length scale; Ret = u L/νu , Da = τt /τ f , and Ka =

2
H. Lee, P. Dai, M. Wan et al. Combustion and Flame 235 (2022) 111712

Table 1
Studied cases.

cases SL , m/s δLT , mm u /SL L/δLT Da Ka Ret Le


x/L
x/η Nx β
A 0.58 0.41 2.2 1.1 0.53 3.0 30 0.32 0.082 1.08 64 18
B 0.58 0.41 4.0 1.1 0.29 9.0 56 0.32 0.055 1.13 96 18
C 0.58 0.41 11.8 1.1 0.10 33.0 158 0.32 0.041 1.85 128 16

Table 2 in several points on the same transverse plane, the point charac-
Values of bmax adopted to detect F or T leading points.
terized by the highest rate was selected. Henceforth, local quanti-
A B C ties evaluated in leading points are designated using superscript l p,
lp lp
ω˙ tF (x, t )/ω˙ L,max
F
2.75 3.15 3.25 e.g. cF (t ) and cT (t ) are values of the fuel and temperature-based
W˙ tT (x, t )/W˙ L,max
T
2.25 3.15 5.00 combustion progress variables in leading points. Moreover, lead-
ing points found using the constraints of ω˙ tF (x, t ) > bω˙ L,max
F and
W (x, t ) > bW
˙ T
t
˙ T will be referred as F and T leading points, re-
L,max
(u /SL )3/2 (δLT /L )1/2 are the Reynolds, Damköhler, and Karlovitz spectively.
numbers, respectively; νu is the kinematic viscosity of unburned
mixture;
x is the grid size; the subscripts u and b designate un- 2.3. Local flame perturbations
burnt and burnt mixture, respectively. The equidiffusive flames A1,
A1F, B1, and C1 discussed in Part I [29] will not be addressed in When modeling premixed turbulent combustion, diffusive-
the present paper. thermal effects are commonly addressed by adapting results of
The used computational meshes ensure at least 11 grid points theoretical or numerical studies of perturbed laminar flames, as re-
across the thermal laminar flame thickness δLT , with the flame res- viewed elsewhere [5,6]. Most often results of theoretical studies of
olution being significantly better (over 22 grid nodes per δLT ) in weakly stretched laminar flames [36–39] are invoked for this pur-
case C characterized by the highest Karlovitz number. In all cases, pose, e.g. see Refs. [15,40]. Within the framework of these theories,
the scale η is greater than half of the grid size. flame perturbations are characterized with a stretch rate s˙ , which
Dependencies of the fuel consumption rate ω˙ H2 ,L (cF ) and the (i) is equal to the rate A−1f
dA f /dt of change of the area A f of an
heat release rate W˙ T,L (cT ) in the unperturbed laminar flame on the infinitesimal element of the flame surface and (ii) is defined as fol-
fuel and temperature-based combustion progress variables cF = lows
(YF,u − YF )/YF,u and cT = (T − Tu )/(Tb − Tu ), respectively, are shown s˙ = at + 2Sd κ . (3)
in Fig. 3 in Part I [29]. Here, YF is the fuel mass fraction. The peak
Here,
values of |ω˙ H2 ,L |(cF ) and W˙ T,L (cT ) will be used for normalization
of the counterpart rates extracted from turbulent flames and will ∂ u j ∂ uk
at = −ni n j + (4)
be designated with symbols ω˙ L,max F and W˙ L,max
T , respectively. Hence- ∂ xi ∂ xk
forth, subscript and superscript F or T refer to fuel or temperature, is a strain rate; ui and ni are components of the velocity vec-
respectively. tor u and the unit vector n = −∇ c/|∇ c| normal to the iso-surface
The reader interested in further details of the performed DNS is c (x, t ) =const, respectively;
referred to Part I [29].
−∇ · Jc + Wc
Sd = (5)
2.2. Numerical diagnostics ρ|∇ c|
is a displacement speed, i.e. the speed of that iso-surface with
The local equivalence ratio in extreme or leading points is eval- respect to the local flow; κ = ∇ · n/2 is the local mean curva-
uated as follows ture of that iso-surface; Jc is the molecular flux of the combustion
1 2XH2 + 2XH2 O + XH + XOH + XH O2 + 2XH2 O2 progress variable c through the iso-surface; Wc is the mass rate of
φ= , (1) product creation, i.e. the source term in the transport equation for
2 2XO2 + XH2 O + XO + XOH + 2XH O2 + 2XH2 O2
c (x, t ); and ρ is the local flow density.
where XS designates mole fraction of species S. The aforementioned asymptotic theories [36-39] yield a linear
Unsteady spatially averaged quantities are determined as fol- relation between local burning and stretch rates, whereas Sd de-
lows pends linearly on both at and κ , with the two coefficients (for at
 and κ ) in these linear relations being (i) different from one an-
1
Q (y, t ) = Q (x, t )dxdz, (2) other in a general case and (ii) different for different iso-surfaces
Lx Lz
c (x, t ) =const [41,42]. Since the discussed theories are based on
with the mean spatial profiles of Q (y ) being obtained by averaging an assumption that a ratio of the laminar flame thickness to the
Q (y, t ) over time. smallest length scale of the flow non-uniformities asymptotically
At each instant, extreme points are characterized by the largest vanishes, the aforementioned linear relations are derived under
(over the entire computational domain) magnitude of the instanta- conditions of a weak stretch rate, i.e. τ f s˙  1. For the same rea-
neous FCR or HRR, i.e. |ω˙ H2 | or W˙ T , respectively. For brevity, these sons, a single value of stretch rate, strain rate, or curvature char-
points will be called F or T extreme points (letters F and T refer to acterizes the entire flame, which is infinitely thin when compared
fuel and temperature, respectively), with the extreme rates being to the flow length scales, i.e. these quantities are considered to be
designated with symbols ω˙ t,max
F or W˙ t,max
T , respectively. Following constant along the normal to the flame surface within the studied
arguments provided in Part I [29], leading points are considered thin flame.
to be the most left points characterized by ω˙ tF (x, t ) > bω˙ L,max
F or Due to the constraint of τ f s˙  1, the aforementioned linear re-
W (x, t ) > bW
˙
t
T ˙ T , with b being either equal to unity or to bmax
L,max
lations do not seem to be useful for modeling the local structure of
reported in Table 2. The latter value is bounded by a requirement highly perturbed extreme points [5,6] studied in the present work.
of ω˙ tF (x, t ) > bω˙ L,max
F or W˙ tT (x, t ) > bW˙ L,max
T at least in a single Nevertheless, the fact that the theories show importance of local
point x at each instant t. When one of the above constraints hold flame characteristics such as stretch rate, strain rate, and curvature

3
H. Lee, P. Dai, M. Wan et al. Combustion and Flame 235 (2022) 111712

may be useful for exploring highly perturbed flames in turbulent However, inspection of Table 3 puts such a correlation into ques-
flows. Indeed, approaches to modeling turbulent combustion that tion. For instance, in case C, the rms of s˙ is significantly larger than
invoke results of numerical or theoretical research into critically its mean value, whereas magnitude of the oscillations of the lo-
perturbed laminar flames deal either (more often) with a highly cal HRR is much less than its mean value, see the aforementioned
strained planar laminar flame [9,10,13,22,23,27,43–47] or with a Fig. 6b. Moreover, there is a substantial probability of finding neg-
highly curved spherical laminar flame [17,18], e.g. a flame ball dis- ative stretch rates that are commonly associated with a decrease
covered theoretically by Zel’dovich [16]. In particular, within the in burning rate for lean hydrogen-air mixtures, but the normalized
framework of the leading point concept, both highly strained pla- local HRR is significantly larger than unity at all instants. Further-
nar and highly curved spherical laminar flames were considered to more, the mean normalized strain rate documented in the T ex-
be models of the local structure of leading flame kernels within a treme points in case C is close to 100, i.e. much higher than the
premixed turbulent flame brush, see Refs. [9,10,13,22,23,27,46,47] critical strain rate that extinguishes the planar stationary laminar
and [17,18], respectively. flame. The former rate is very large even in cases A and B. Accord-
Based on the above brief overview of the use of results of in- ingly, transient effects1 are likely to play a very important role in
vestigations of perturbed laminar flames for modeling of premixed the extreme T points.
turbulent combustion (the reader interested in a detail review of Second, similar trends are also observed in the T leading points,
the subject is referred to Refs. [5,6,15]), the following issues appear but (i) the rms of s˙ is smaller than its mean value, (ii) the curva-
to be of interest: (i) whether or not there is a relation between ture is mainly positive in case A, but is negative in cases B and
local characteristics of highly strained planar laminar flames and C, (iii) the strain rate is smaller when compared to the T extreme
the extreme or leading points in turbulent flames and (ii) whether points, but (iv) the stretch rate is larger.
or not local flow-induced perturbations in the extreme or leading Third, in the F extreme points, (i) the local curvature is almost
points can be reduced to a single strain rate, curvature, or stretch always positive and its magnitude is increased with increasing Ka,
rate. These two issues are addressed in the next section. (ii) the local strain rate is predominantly negative in cases A and
To explore the former issue, stationary planar laminar flames B, but is predominantly positive in case C, (iii) on the contrary,
subject to various strain rates were simulated by running code the local stretch rate is predominantly positive in cases A and B,
Cantera 2.4 [48], using a fine mesh (up to 1 0 0 0 cells at the highest but is predominantly negative in case C. Moreover, the time av-
strain rates), the same chemical mechanism [30], and varying the eraged absolute values of the normalized strain and stretch rates
strain rate at up to a critical value at,cr such that further increase are about 10 in case C, with the corresponding rms values be-
in at > at,cr extinguishes the flame. For the studied lean hydrogen- ing much higher. Accordingly, the F extreme points can be subject
air mixture, at,cr = 16 1/ms or at,cr δLT /SL = 11. The choice of the to very large instantaneous strain or stretch rates, thus, implying
strained planar flames was motivated by a wide use of this sim- again that transient effects play an important role.
ple problem for modeling the local structure of leading points in Fourth, trends observed for the F leading and extreme points
premixed turbulent flames [9,10,13,22,23,27,46,47]. are similar, with the mean values of the discussed quantities being
smaller in the leading points in cases A and B. However, in case C,
3. Results and discussion (i) the positive mean curvature is larger in the leading points, (ii)
the mean strain rate remains negative, but small in these points,
3.1. Local flame perturbations in extreme and leading points and (iii) the magnitude of the negative mean stretch rate is signif-
icantly higher in the leading points.
Major characteristics of local flame perturbations in the ex- Statistics of leading points detected using bmax and b = 1 show
treme and leading points are shown in Figs. 1 and 2, respectively, similar trends, which, however, are less pronounced in the latter
where t ∗ = t/τt designates the normalized time. These characteris- case. For instance, the mean curvature or stretch rate are smaller
tics oscillate around time-averaged values, with the magnitude of in such F points or the mean strain and stretch rates are smaller in
such oscillations in the F extreme points being large when com- such T points. Moreover, for each considered quantity, signs of its
pared to weak fluctuations in the local FCR, see Fig. 6a in Part I mean values are the same for b = 1 and bmax , with two exceptions:
[29] or Fig. 13, which will be discussed later. This significant dif- (i) normalized mean curvature is positive (1.1 ± 0.6 ) in the T lead-
ference implies the lack of a correlation between the extreme FCRs ing points detected using b = 1 in case B and (ii) normalized mean
and the local flame perturbation characteristics such as curvature, strain rate is positive (0.4 ± 5) in the F leading points detected us-
strain or stretch rate. Indeed, statistical analysis of the DNS data ing b = 1 in case B, but the rms values of at are significantly larger
did not reveal such a correlation in the F extreme or leading points. than its mean values in both (b = 1 and bmax ) F leading points in
It is also of interest to note that the signs of κ are opposite in the F case B or C.
(positive) and T (negative) extreme points in all three cases. Such a Finally, it is worth noting two interesting observations rele-
qualitative difference is also observed in the F and T leading points vant to the (i) F and (ii) T extreme points. First, while the ex-
in cases B and C. treme FCR does not correlate with the local curvature, strain or
Statistical characteristics (the first two moments) of the dis- stretch rate, as already noted, there is a persistent relevant trend
cussed oscillations of various local quantities, normalized using the in the F extreme points. Fig. 3 reports the y-profiles of the strain
laminar flame speed SL and thickness δL = 1/max|∇ cF | = 0.5 mm, rate at (y ) extracted at three instants characterized by the lowest
are reported in Table 3, which also includes probabilities P (X > 0 ) value of |1 + ny (t )| in the F extreme points. Here, ny (t ) is the y-
of finding positive values of these quantities. The leading points component of the unit normal vector n and it is very close to −1
were detected using bmax reported in Table 2. in the selected instants. Accordingly, these profiles of at (y ) illus-
Table 3 shows the following trends. First, in the T extreme trate spatial variations in the strain rate in the direction normal
points, the local curvature is almost always negative, the local to the F extreme points. All these profiles show a rapid decrease
strain rate is always positive, and the local stretch rate is predom- in the strain rate at (y ) upstream (y < 0) of the F extreme points
inantly positive, with the absolute values of these quantities be-
ing significantly increased with increasing Karlovitz number (from 1
As predicted theoretically by Clavin and Joulin [49] and confirmed in subse-
case A to case C). At first glance, these observations imply a corre- quent experimental and numerical studies reviewed elsewhere [5], due to transient
lation between these perturbation characteristics and a significant effects, a laminar flame can survive under the influence of very strong strain or
increase in the local HRR with Ka, reported in Fig. 6b in Part I [29]. stretch rate if the rate rapidly changes in time.

4
H. Lee, P. Dai, M. Wan et al. Combustion and Flame 235 (2022) 111712

Fig. 1. Evolution of normalized curvature (top row), strain rate (second row), stretch rate (third row), and density-weighted displacement speed S˜d ≡ ρ Sd /ρu (bottom row) in
the extreme F (solid lines) and T (dashed lines) points in flames A (left column), B (middle column), and C (right column). The curvature and displacement speed characterize
the local cF (x, t ) and cT (x, t ) fields in F and T points, respectively.

(y = 0) when the distance |y| between the considered point y and Here, Fick’s law is used as the simplest model of the molecular flux
the extreme point is decreased (such a trend is not observed very Jc and Dc is the molecular diffusivity of the combustion progress
close to the F extreme point in flame C at t ∗ = 24.24). In other variable. In the T extreme points, the iso-surface cT (x, t ) =const is
words, the product n · ∇ at is positive in the F extreme points at highly and negatively curved, see Table 3. Accordingly, the curva-
the selected instants. Statistics of this product sampled from the ture term is positive and results in increasing Sd . Fig. 4a implies
F extreme points at all instants yields (δL2 /SL )n · ∇ at = 1.7 ± 1.3, that this effect is of more importance when compared to the local
4.2 ± 3.8, and 28 ± 31 in flames A, B, and C, respectively, with the variations in the two other (reaction and normal diffusion) terms
probability of n · ∇ at > 0 being equal to 0.91, 0.89, and 0.83, re- on the right hand side of Eq. (6).
spectively. Such significant spatial variations of the strain rate sug- The lack of a similar correlation in the F extreme points, see
gest that results of asymptotic theories that address large-scale Fig. 4b, could be explained as follows. Comparison of the signs
perturbations of laminar flames [36-39] are hardly applicable to of the mean values of Sdex (t ) · κFex (t ), at,F
ex (t ), and s˙ ex (t ), reported
F
modeling the local structure of the F extreme points, at least, un- in Table 3, indicates that stretch rate s˙ is mainly controlled (i) by
der conditions of the present study. the curvature contribution Sd κ in the F extreme points (the signs
Second, Fig. 4a shows that the local displacement speed in the of Sdex · κFex and s˙ ex
F
are the same, but opposite to the sign of at,F ex )

T extreme points correlates well with the local curvature κ of the and (ii) by the strain rate in the T extreme points. Since the lo-
iso-surface cT (x, t ) =const. Such correlations were found neither cal stretch rate characterizes the local change of iso-scalar surface
for strain/stretch rates nor in the F extreme points, see Table 4 and area and, accordingly, |∇ c|, the correlation between the local cur-
Fig. 4b. The correlations reported in Fig. 4a are associated with the vature and |∇ cF | in the F extreme points is expected to be bet-
last (tangential diffusion or curvature) term in the following well- ter pronounced than the correlation between κ and |∇ cT | in the
known decomposition [50-52] of displacement speed T extreme points. This is confirmed by a positive and large corre-
lation coefficient r (T3 , |∇ c| ) extracted from the DNS data in the F
Wc n · ∇ ( ρ Dc n · ∇ c ) extreme points and reported in the bottom row in Table 5 (note
Sd = + −Dc ∇ · n . (6)
ρ|∇ c| ρ|∇ c|    that the curvature term T3 includes the sign minus). Consequently,
      T3 if the magnitude of the negative (in the F extreme points) term T3
T1 T2

5
H. Lee, P. Dai, M. Wan et al. Combustion and Flame 235 (2022) 111712

Table 3
Mean and rms values of various local characteristics of flame perturbations in the F and T extreme and leading points, normalized using SL and δL , as well as probabilities
of finding positive values of these characteristics.

Case-A Case-B Case-C

quantity point mean rms P(X>0) mean rms P(X>0) mean rms P(X>0)

Extreme points
κ F 1.78 0.29 1.00 2.60 0.55 1.00. 3.22 1.58 0.98
T −0.40 0.31 0.08 −1.40 0.60 0.00 −3.74 1.06 0.00
at F −3.33 1.51 0.01 −3.33 3.22 0.13 8.80 20.0 0.65
T 10.8 1.92 1.00 25.5 6.03 1.00 99.0 20.9 1.00
S˜d F 1.47 0.14 1.00 1.21 0.25 1.00 −0.48 0.73 0.23
T 1.89 0.18 1.00 2.70 0.36 1.00 5.52 0.76 1.00
s˙ F 18.3 2.98 1.00 22.8 6.59 0.99 −10.6 45.3 0.49
T 6.88 2.88 0.97 7.98 8.81 0.83 21.8 35.3 0.77
Leading Points (b = bmax )
κ F 1.42 0.30 0.997 2.06 0.56 1.00 3.93 1.57 1.00
T 0.39 0.31 0.900 −0.62 0.46 0.02 −2.86 1.05 0.03
at F −2.35 1.66 0.076 −2.34 3.66 0.26 −0.93 11.2 0.43
T 5.35 2.41 0.997 19.1 4.90 1.00 77.7 26.1 1.00
S˜d F 1.39 0.17 0.997 1.11 0.25 1.00 −0.64 1.07 0.28
T 1.41 0.17 0.997 1.54 0.24 1.00 4.39 0.71 1.00
s˙ F 13.1 2.32 0.997 16.8 4.14 1.00 −24.3 58.9 0.36
T 8.12 1.92 0.991 12.5 7.04 0.95 36.3 29.3 0.87
Leading Points (b = 1.0)
κ F 1.15 0.29 0.997 1.52 0.49 1.00 3.10 1.31 1.00
T 1.05 0.40 0.994 1.07 0.65 0.97 −0.33 1.32 0.43
at F −0.91 2.23 0.366 0.46 5.23 0.48 −1.35 15.8 0.43
T 1.44 2.44 0.739 6.22 6.64 0.80 34.13 25.0 0.95
S˜d F 1.14 0.22 0.997 0.85 0.39 0.96 −1.38 1.64 0.18
T 1.41 0.17 0.997 0.92 0.41 0.99 1.21 0.77 0.98
s˙ F 6.83 2.18 0.983 7.24 6.06 0.90 −43.37 76.2 0.28
T 8.12 1.92 0.991 9.55 5.62 0.95 29.61 24.6 0.93

Table 4
Pearson’s correlation coefficients between the displacement speed and curvature, strain rate, or stretch rate in the F and T extreme points.

Case-A Case-B Case-C

Correlation, r F T F T F T

r ( sd , κ ) −0.2671 −0.8488 −0.3772 −0.9290 −0.576 −0.8498


r (sd , at ) 0.2542 0.4349 0.2920 0.4036 0.3021 0.1889
r (sd , s˙ ) 0.5246 −0.7055 0.7041 −0.7770 0.8029 −0.6951

Table 5 strain rate, which mainly controls the local stretch rate, as noted
Pearson’s correlation coefficients between various terms in Eq. (6).
earlier.
Case-A Case-B Case-C

Correlation, r F T F T F T 3.2. Topology and evolution of leading points


r (T3 , T1 ) −0.86 −0.40 −0.93 0.27 −0.91 0.74
r (T3 , T2 ) −0.65 0.35 −0.62 0.24 −0.51 0.03 The fact that the curvature of the T leading points is nega-
r (T3 , |∇ c| ) 0.79 0.48 0.92 0.04 0.89 −0.51 tive in flames B and, especially, C appears to be surprising, be-
cause curvature of a leading point should always be positive for
purely topological reasoning. It is worth emphasizing, however,
is increased, the gradient |∇ cF | is decreased and, hence, the mag- that the T leading points were detected based on the local value
nitude of the positive term T1 is increased, thus, counterbalancing of HRR, whereas the local curvature was evaluated for the iso-
the increase in the negative T3 . The first row in Table 5 shows that surface cT (x, t ) =const. If the local curvature is defined as follows:
such a negative correlation between the terms T1 and T3 is well κHRR = ∇ · nHRR , where the unit vector nHRR = −∇ W˙ tT /|∇ W˙ tT | is
pronounced in the F extreme points. This negative correlation can normal to the iso-surface W˙ tT (x, t ) =const, then, the local curva-
also be easily seen in Fig. 5. Since the nominator in the reaction ture is positive, see Fig. 6, as this should be for a leading point.
term T1 varies weakly in the F extreme points, as shown in Fig. 6a Moreover, the curvature of the latter isosurface is very large and is
in Part I [29], the negative r (T1 , T3 ) results solely from the positive increased by Ka.
r (T3 , |∇ c| ), see the bottom row in Table 5. Due to the discussed These results imply that the iso-surfaces cT (x, t ) =const and
mutual cancelation of variations of the terms T1 and T3 , see Fig. 5, W˙ tT (x, t ) =const are substantially different in the vicinity of the
and because the correlation between the diffusion terms T2 and T3 T leading points. Indeed, red dashed lines in Fig. 7 show that a
is less pronounced due to variations of the nominator in the for- scalar product of the unit normal vectors nT = −∇ cT /|∇ cT | and
mer term, the DNS data are highly scattered in Fig. 4b. A similar nHRR differs significantly from unity, with the effect being more
mutual cancelation of variations of the terms T1 and T3 does not pronounced in the highly turbulent flame C. On the contrary,
occur in the negatively curved T extreme points, because the two the scalar products nF · nT , nF · nF CR , and nF · nφ (not shown to
terms have the same sign there. Consequently, in these extreme make Fig. 7 readable) are always close to unity in the T lead-
points, |∇ cT | (i) correlates poorly with the curvature term T3 , see ing points, indicating similarities of the iso-surfaces cF (x, t ) =const
the last row in Table 5, and (ii) is significantly affected by the local and cT (x, t ) =const, ω˙ tF (x, t ) =const, or φ (x, t ) =const, respec-

6
H. Lee, P. Dai, M. Wan et al. Combustion and Flame 235 (2022) 111712

Fig. 2. Evolution of normalized curvature (top row), strain rate (second row), stretch rate (third row), and density-weighted displacement speed (bottom row) in the extreme
F (solid lines) and T (dashed lines) points in flames A (left column), B (middle column), and C (right column). The curvature and displacement speed characterize the local
and fields in F and T points, respectively.

tively. Here, nF = −∇ cF /|∇ cF |, nF CR = −∇ ω˙ tF /|∇ ω˙ tF |, and nφ = W˙ tT (x, t )/W˙ L,max


T > bmax is found in this cube at the next instant.
−∇ φ /|∇ φ|. Readers interested in further details of the tracking algorithm em-
Results reported in Figs. 6 and 7 and similar results obtained ployed in this work are referred to Supplemental Materials. This
in the T extreme points indicate that the local flame structure is tracking algorithm offers an opportunity to observe disappearance
highly perturbed in the considered points. As illustrated in Fig. 8, of preceding leading points due to implosion of unburned gas, see
the local structures of the T leading points are very different at the bottom row in Fig. 10, but such implosions typically occur after
different instants. Sometimes, the local structure resembles a flame the change of the leading points. Implosions of unburned mixture
cusp, see the second column, or implosion of a pocket of unburned occur not only in the vicinity of preceding T leading points, but
mixture enveloped by a flame, see the right column. Such images also in other flame zones, see the top right corner in images in the
imply that the T leading points could disappear due to local flame top row in Fig. 10.
collisions. However, analysis of a number of subsequent images Black and white crosses in Fig. 8 indicate that the F and T lead-
shows that a typical scenario of the T leading point change consists ing points are not neighboring points (note that the z-coordinates
in moving another point characterized by W˙ tT (x, t )/W˙ L,max T > bmax of these two points are also different in each image). By applying
further to the left when compared to the preceding leading point, various diagnostics techniques to processing the DNS data, we did
e.g. see images in the left and middle rows in Fig. 9. While there not reveal a link between coordinates of the F and T leading points.
is an unburned mixture finger in the vicinity of the earlier lead- For instance, the distance between the leading F and T points is al-
ing point (t ∗ = 28.41, left column), the finger does not disappear most always larger than the distance between points characterized
quickly and still survives at t ∗ = 28.55, see the right column. lp lp
by cF and cT in the unperturbed laminar flame. Moreover, while
Note that the right images in Fig. 9 have been obtained by a T leading point is typically closer to the inlet boundary when
tracking the motion of the earlier leading point (t ∗ = 28.41) after lp lp
compared to a F leading point, i.e., yT (t ) < yF (t ), the probability
the change of the T leading points at t ∗ = 28.43, see middle col- lp lp
of finding yT (t ) > yF (t ) does not vanish.
umn. For this purpose, at each instant, the leading or preceding
One might assume that processes peculiar to the T leading
leading point is placed in the center of a cube whose side length
points occur in other zones also and, in particular, in the vicin-
is equal to 15
x = 0.57δL and the most left point characterized by

7
H. Lee, P. Dai, M. Wan et al. Combustion and Flame 235 (2022) 111712

Fig. 3. Profiles of normalized strain rate along the y-axis, which is locally normal to the F extreme points (y = 0) at the selected instants specified in legends. Results
obtained from flames A, B, and C are plotted in the left, middle, and right columns, respectively.

Fig. 4. Correlation between the local curvature of the iso-surface (a) cT (x, t ) =const or (b) cF (x, t ) =const and the local displacement speed in the extreme (a) T and (b)
F points in flames A (circles and red solid lines), B (squares and blue dashed lines), and C (pentagons and green dotted lines). Straight lines show least square fits, with
Pearson’s correlation coefficients being reported in legends . (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of
this article.)

Fig. 5. Evolution of terms T1 (red solid lines) and T3 (black dashed lines) in Eq. (6) in the F extreme points in flames A (left column), B (middle column), and C (right
column) . (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

ity of the F leading points, thus, affecting burning in the latter Finally, it is worth noting that changes of the T leading points
points. As discussed in Part I [29], there are the following three occur more often when compared to the F leading points. For in-
peculiar features of the T leading points: (i) the local HRR (FCR) is stance, during the evolution of flame C, the changes of the latter
higher (lower, respectively) than W˙ L,max
T (ω˙ L,max
F , respectively), (ii) (former) points occurred 16 (37, respectively) times during an in-
signs of the rates of production of the radicals H and O are op- terval of about 35τt . In flame B, such changes occurred 22 (61, re-
posite (negative and positive, respectively, see Fig. 19b in Part I spectively) times during an interval of about 65τt . In flame A char-
[29]), and (iii) curvature of the temperature iso-surface is nega- acterized by the lowest Karlovitz number, there were 7 (8, respec-
tive, see dashed lines in Figs. 2b and 2c. Accordingly, zones where tively) changes during an interval of about 20τt .
the features (i)-(iii) held were sought in the vicinity and upstream
of the F leading points, i.e., in a rectangular whose right side con- 3.3. Thermo-chemical structure of strained laminar flames and
tained the F leading point in the center and whose axial length extreme points
was equal to n
x. However, the probability of finding such zones
was very low even for n = 4. Thus, statistical evidence that burn- Figure 12 shows that the computed peak local FCR and HRR
ing structures similar to the T leading or extreme points exist very are monotonously increased in strained laminar flames, with the
close to the F leading points and, therefore, could affect them, has maximum values of the FCR and HRR being reached at the criti-
not yet been found. Nevertheless, images reported in Fig. 11 show cal strain rate at = at,cr ≈ 16 1/ms close to combustion extinction.
that such burning structures exist sufficiently close to the F leading While the maximum HRR in the strained laminar flames is signif-
points sometimes. Accordingly, an issue of eventual influence of re- icantly (by a factor of about two) less than the HRRs in the ex-
actions and heat release in zones similar to the T leading points on treme T points in the highly turbulent flame C, see Fig. 13d, the
combustion in the F leading points requires further target-directed maximum FCR reached in the critically strained laminar flame is
study. comparable with the FCRs in the extreme F points in all three tur-
bulent flames, see Fig. 13c. It is also of interest to note that ratios

8
H. Lee, P. Dai, M. Wan et al. Combustion and Flame 235 (2022) 111712

Fig. 6. Evolution of normalized (using δL ) curvature of iso-surface of W˙ tT (x, t ) =const in (a) extreme and (b) leading T points. The leading points were detected using bmax
reported in Table 3.

Fig. 7. Evolution of scalar products of various unit normal vectors specified in legends in T leading points in flames (a) A and (b) C. The leading points were detected using
bmax reported in Table 3.

of ω˙ t,max
F /ω˙ L,str
F [cF = cFex (t )] plotted in Fig. 13a are close to ratios from W˙ t,max
T /max{W˙ L,str
T (c )}, thus, implying that, in the critically
T
of ω˙ t,max
F /max{ω˙ L,str
F (cF )}
reported in Fig. 13c. Thus, the peak FCR strained flames, the maximum HRR is reached at cT that differs
is reached at almost the same cF in the critically strained laminar substantially from cTex (t ) in flame C. Furthermore, mean normalized
flame and in the extreme F points in turbulent flames. It is worth mass fractions of H, O, OH, HO2 , and H2 O2 differ significantly from
remembering that cFex (t ) weakly fluctuates in the vicinity of 0.7 in unity in the right diagram in Fig. 15, as well as mean normalized
the turbulent flames, see Fig. 14a in Part I [29]. rates of production/consumption of H, OH, and HO2 in the right
Values of other single-point (i.e., quantities that do not involve diagram in Fig. 16b. Finally, the rates of production/consumption
spatial derivatives) thermo-chemical characteristics of the F ex- of H2 O2 have opposite signs in the T extreme points in flame C
treme points in the studied turbulent flames are also close to the and in the point characterized by the peak HRR in the critically
values of the counterpart characteristics conditioned to the peak strained laminar flame, see right gray rectangular in Fig. 16b. Ac-
FCR in the critically strained laminar flame, e.g. see (i) evolution cordingly, single-point thermo-chemical characteristics of the crit-
of the local temperature or equivalence ratio, shown in Figs. 14a- ically strained laminar flame are less suitable for describing the T
14c or 14d-14f, respectively, (ii) mean mass fractions of various extreme points when compared to the F extreme points.
species, reported in Fig. 15, or (iii) rates of production/consumption Similarities between single-point thermo-chemical characteris-
of various species, plotted in Fig. 16a. While differences between tics of the F extreme points and critically strained laminar flames,
the former and latter quantities are increased with the Karlovitz observed in Figs. 13-16, appear to be surprising, because common
number for certain characteristics, e.g., see (i) Figs. 14d and 14f, characteristics of local flame perturbations (curvature, strain and
(ii) mass fractions of various species with the exception of H and stretch rates), evaluated in the F extreme points and plotted in
OH in Fig. 15, or (iii) rates of production/consumption of H and O solid lines in Fig. 1, are very different from the counterpart char-
in Fig. 16a, the trend is weakly pronounced when compared to the acteristics of a steady strained laminar premixed flame. Moreover,
T extreme points discussed below. the density-weighted displacement speed S˜d conditioned to the
In flames A and B, single-point thermo-chemical charac- peak FCR in the critically strained laminar flame is about 0.8SL ,
teristics of the T extreme points are also close to the val- whereas the mean S˜d (t )/SL evaluated in the F extreme points are
ues of these characteristics conditioned to the peak HRR in equal to 1.39 and −1.11 in flames A and C, see Table 3. Further-
the critically strained laminar flame, e.g., see curves plotted more, the mean value of the gradient |∇ cF | (or |∇ cT |) evaluated in
in black and red lines in Figs. 13b and 13d or mass frac- the F (or T) extreme points and normalized using |∇ cF | (or |∇ cT |)
tions and rates of production/consumption of various species conditioned to the peak FCR (or HRR) in the critically strained lam-
in Figs. 15 and 16b, respectively. However, in flame C, the ra- inar flame are equal to 0.93 and 0.73 (or 0.89 and 1.52) in flames
tio W˙ t,max
T /max{W˙ L,str
T (c )} fluctuates around two, see blue curve
T A and C, respectively.
in Fig. 13d. Moreover, comparison of blue curves in Figs. 13b Instantaneous profiles of |∇ cF |(y ) and |∇ cT |(y ) extracted at in-
and 13d shows that W˙ t,max T T [c = cex (t )] differs substantially
/W˙ L,str T T stants characterized by the minimum (and very small) value of

9
H. Lee, P. Dai, M. Wan et al. Combustion and Flame 235 (2022) 111712

Fig. 8. 2D images of cT (x, t )-fields (top row) and W˙ tT (x, t )/W˙ L,max
T
-fields (bottom row) obtained from flame C at t ∗ = 17.5, (first column), 21.2 (second column), 36.6 (third
column), and 50.1 (fourth column). White crosses show the T leading points. Black crosses indicate x and y coordinates of the F leading points, which belong to other (when
compared to the T leading points) planes z =const. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this
article.)

Fig. 9. 2D images of W˙ tT (x, t )/W˙ L,max


T
-fields (top row) and cT (x, t )-fields (bottom row) at t ∗ = 28.41 (left column), 28.43 (middle column), and 28.55 (right column). Black
crosses in top left and middle images show the T leading points, whose x-coordinates are equal to 1.52 and 0.69 mm, respectively. Images in the left and right columns have
been obtained for the same x = 1.52 mm. Flame C.

|1 + nex
y (t )| are plotted in lines in Fig. 17. In flame C, two of four (pentagons). In flames A and B, see Figs. 18a and 18b, respectively,
profiles of |∇ cF |(y ) along the normal to the F extreme point, see variations of ∇ YH along the normal to the F extreme point differ
Fig. 17c, are sufficiently close to the profile of |∇ cF |(y ) in the crit- significantly from ∇ YH (y ) in the critically strained or unperturbed
ically strained laminar flame (pentagons). In flames A and B, see (squares) laminar flames. In the vicinity of the T extreme points,
Figs. 17a and 17b, respectively, variations of |∇ cF | along the normal see Figs. 18d-18f, the former (strained) laminar-flame profiles of
to the F extreme point differ significantly from |∇ cF |(y ) in the crit- ∇ YH (y ) are closer to the turbulent ones in all three flames. Never-
ically strained or unperturbed (squares) laminar flames, with the theless, only in case B, profiles of ∇ YH (y ) along the normal to the
latter laminar-flame profiles being closer to the turbulent ones. On T extreme point are close to the profile of ∇ YH (y ) in the critically
the contrary, profiles of |∇ cT |(y ) along the normal to the T ex- strained laminar flame, see Fig. 17e. In flames A and C, the tur-
treme point in flames A and B are sufficiently close to the profile bulent and laminar profiles of ∇ YH (y ) are different, with the peak
of |∇ cT |(y ) in the critically strained laminar flame, see Figs. 17d value of ∇ YH (y ) being smaller (larger) in flame A (C, respectively)
and 17e, but such profiles are significantly different in flame C, see when compared to the critically strained laminar flame.
Fig. 17f.
Some of these trends are also observed for gradients of mass 4. Discussion: implications for modeling and a new challenge
fractions of major radicals, e.g. see Fig. 18 where such results are
reported for H. In flame C, one instantaneous profile of ∇ YH (y ) Similar to results discussed in Part I [29], results reported in
along the normal to the F extreme point, see Fig. 18c, is close the present paper neither reject nor prove the leading point con-
to the profile of ∇ YH (y ) in the critically strained laminar flame cept. On the one hand, Figs. 1- 4 and Table 3 indicate that quan-

10
H. Lee, P. Dai, M. Wan et al. Combustion and Flame 235 (2022) 111712

Fig. 10. 2D images of cF (x, t )-fields (left column), cT (x, t )-fields (middle column), and W˙ tT (x, t )/W˙ L,max
T
-fields (right column) at t ∗ = 28.83 (top row) and 30.03 (bottom row).
Flame C.

Fig. 11. 2D images of cF (x, t )-fields (first column), cT (x, t )-fields (second column), W˙ tT (x, t )/W˙ L,max
T
-fields (third column), and ω˙ tF (x, t )/ω˙ L,max
F
-fields (fourth column) obtained
from flame C at t ∗ = 27.83, (top row) and 28.03 (bottom row). Crosses show the F leading points.

tities used commonly for characterizing perturbations (curvature, flame. This is not surprising, in particular, because transient effects
strain or stretch rate) of laminar premixed flames vary significantly appear to play a pivotal role in the extreme or leading points, as
in the F or T extreme points with time, with their mean values implied by large (“supercritical”) instantaneous values of the lo-
being substantially affected by Ka. Accordingly, the local structure cal strain or stretch rate, cf. Figs. 1 and 2 with Fig. 12. At first
of extreme or leading F or T points is not reduced to the struc- glance, these results put utility of the leading point concept into
ture of a single simply perturbed (strained or/and curved) laminar question.

11
H. Lee, P. Dai, M. Wan et al. Combustion and Flame 235 (2022) 111712

to the peak FCR in the stationary, planar, critically strained laminar


flame. However, spatial gradients of species mass fractions in the
vicinity of the F extreme points differ significantly from the coun-
terpart gradients in the aforementioned laminar flame, see Figs. 17
and 18.
The fact that the computed extreme FCRs ω˙ t,max F are always
(at different Ka and at different instants) close to the peak FCR
ω˙ str,max
F in the stationary, planar, critically strained laminar flame,
see Fig. 13c, appears to be beneficial for modeling. More specif-
ically, it suggests the use of the latter rate ω˙ str,max
F as an input
parameter for unsteady CFD (not only RANS, but also Large Eddy
Simulation, LES) research into premixed turbulent combustion. This
idea is consonant with the basic idea of the leading point concept,
i.e. substitution of the laminar flame speed SL with a local con-
sumption velocity ucr c,L
evaluated in a critically perturbed laminar
flame [5,6,13]. Accordingly, it would be of great interest to eval-
uate the local consumption velocity uF,ex c,t in the F extreme points
Fig. 12. Peak local fuel consumption (solid line) and heat release (dashed line) rates and compare the obtained values with ucr evaluated in the sta-
computed for stationary, planar, strained laminar premixed flames and normalized c,L
using the peak rates in the unperturbed laminar flame vs. normalized strain rate.
tionary, planar, critically strained laminar flame. Due to differences
in spatial gradients shown in Figs. 17 and 18, differences between
uF,ex cr
c,t /uc,L and unity are likely to be larger than small differences
On the other hand, Figs. 13 and 14 show that major single-point (Fig. 13c) between ω˙ t,max
F /ω˙ str,max
F and unity. Nevertheless, if the
thermo-chemical characteristics (FCR, equivalence ratio and tem- former differences are sufficiently small, the substitution of SL with
perature) of the F extreme (or leading, see Fig. S1 in Supplemen- ucr
c,L
in equations of a turbulent combustion model could be justi-
tal Materials) points (i) oscillate weakly, (ii) are almost the same fied, at least to the leading order. These issues will be addressed in
for three different Karlovitz numbers, and (iii) are close to the future papers.
counterpart quantities conditioned to the peak FCR in the station- Moreover, the highlighted fact (ω˙ t,max
F /ω˙ str,max
F ≈ 1 in Fig. 13c)
ary, planar, critically strained laminar flame. Moreover, Figs. 15 and implies that, for each H2 /O2 /N2 /I mixture (here, I is an inert dilu-
16 show that the mean mass fractions of various species and the ent such as He, Ar, etc.), there is an extreme perturbation of the
mean rates of their creation/consumption, evaluated in the F ex- local FCR, which can be reached in different ways, e.g., in rapidly
treme (or leading, see Figs. S2 and S3 in Supplemental Materials) evolving extreme points in various turbulent flames or in the sta-
points are also almost the same for three different Karlovitz num- tionary, planar, critically strained laminar flame. This hypothesis
bers and are also close to the counterpart quantities conditioned could be supported in Figs. 20b, 20d, and 20f in Part I [29], which

Fig. 13. Evolution of ratios of (a) ω˙ t,max


F
/ω˙ L,str
F
[cF = cFex (t )], (b) W˙ t,max
T
/W˙ L,str
T
[cT = cTex (t )], (c) ω˙ t,max
F
/max{ω˙ L,str
F
(cF )}, and (d) W˙ t,max
T
/max{W˙ L,str
T
(cT )} evaluated in the F (left
column) or T (right column) extreme points in turbulent flames A (black solid lines), B (red dashed lines), and C (blue dotted lines). The rates ω˙ L,str F
(cF ) and W˙ L,str
T
(cT ) are
calculated at the critical strain rate. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

12
H. Lee, P. Dai, M. Wan et al. Combustion and Flame 235 (2022) 111712

Fig. 14. Evolution of normalized temperature (top row) and equivalence ratio (bottom row) in extreme F (black solid lines) or T (red dashed lines) points in flames A (left
column), B (middle column), and C (right column). The normalization is done using the value of T or φ in point characterized by the peak FCR or HRR in the critically
strained laminar flame. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 15. Time-averaged mass fractions of various species in the F or T extreme points, normalized using mass fractions of the same species in points characterized by the
peak FCR or HRR, respectively, in the critically strained laminar flame.

Fig. 16. Time-averaged rates of production/consumption of various species in (a) F or (b) T extreme points, normalized using the rates evaluated in points characterized by
the peak FCR or HRR, respectively, in the critically strained laminar flame.

show that the extreme FCR is mainly controlled by the two reac- strained flames, highly curved expanding cylindrical and spherical
tions H2 +OHH + H2 O and H2 +OH+OH. The rates of these re- flames, or more complicated flame configurations, e.g., see papers
actions are likely to peak at certain optimal combinations of mass by Mikolaitis [53,54], Uemichi and Nishioka [55], or Zhou et al.
fractions of H2 , OH, and O, with these optimal combinations be- [56].
ing reached at certain optimal local equivalence ratio. Note that As far as the T extreme or leading points are concerned, they
since the activation temperatures for these reactions are moderate characteristics (i) fluctuate with higher magnitudes when com-
(about 3 500 and 3 150 K, respectively, when compared to 7 650 K pared to the F points, (ii) depend on the Karlovitz number, e.g.
for the branching chain reaction O2 +HOH+O and 6 700 K for the see the right column in Fig. 13 or the top row in Fig. 14, and (iii)
propagation chain reaction H2 O+OOH+OH [30]), sensitivity of can significantly differ from characteristics of the stationary, planar,
the discussed reaction rates to the local temperature variations is critically strained laminar flame, e.g. see a ratio of HRRs plotted in
also expected to be moderate. These hypotheses should be further blue dotted line in Fig. 13d. Accordingly, contrary to the extreme
examined, e.g. in complex-chemistry simulations of various highly FCR, the present analysis does not reveal a characteristic of the
perturbed laminar flames, such as stationary cylindrical critically T extreme or leading points, which could be useful for advancing

13
H. Lee, P. Dai, M. Wan et al. Combustion and Flame 235 (2022) 111712

Fig. 17. Variations of |∇ cF | (top row) or |∇ cT | (bottom row) along the normal to the unperturbed (squares) or critically strained (pentagons) laminar premixed flame or
along the normal to the F (top row) or T (bottom row) extreme point (lines). All profiles are shifted such that the FCR (top row) or HRR (bottom row) peaks at y = 0. Profiles
extracted from turbulent flames A (left column), B (middle column), and C (right column) are instantaneous, with relevant instants t ∗ being specified in legends.

Fig. 18. Variations of ∇ YH along the normal to the unperturbed (squares) or critically strained (pentagons) laminar premixed flame or along the normal to the F (top row)
or T (bottom row) extreme point (lines). All profiles are shifted such that the FCR (top row) or HRR (bottom row) peaks at y = 0. Profiles extracted from turbulent flames A
(left column), B (middle column), and C (right column) are instantaneous, with relevant instants t ∗ being specified in legends.

the leading point concept. Therefore, the concept should (if any) 5. Conclusions
be adapted to model the FCR-based turbulent burning velocity UtF ,
but not the HRR-based UtT . Nevertheless, since UtF ≈ UtT , e.g., in the Data obtained in recent DNS [29] of statistically one-
present DNSs, it is not necessary to apply the concept to both tur- dimensional and planar, lean complex-chemistry hydrogen-air
bulent burning velocities. It is sufficient to model UtF by adapting flames characterized by three different Karlovitz numbers Ka rang-
the leading point concept, whereas an approach to directly mod- ing from 3 to 33 were further analyzed in order to explore local
eling the influence of diffusive-thermal effects on UtT has not yet characteristics and structure of (i) extreme points characterized by
been put forward, to the best of the present authors knowledge. the peak (over the computational domain) Fuel Consumption Rate
This situation is somehow similar to modeling the speed of a fully (FCR) or Heat Release Rate (HRR) and (ii) leading points that were
developed turbulent flame whose mean thickness is statistically also characterized by a high FCR or HRR, but advanced furthest
stationary. This speed could be evaluated either by exploring the into unburned reactants. The following numerical results are worth
speed of the flame leading edge or by integrating the mean FCR emphasizing.
or HRR along the normal to the mean flame brush, with the re- First, no correlation between the local FCR or HRR in the ex-
sult should be the same. Accordingly, to solve the problem, a more treme or leading points and the conventional local characteristics
elaborated method may be selected among alternative approaches. of flame perturbations (curvature, strain or stretch rate) is found.

14
H. Lee, P. Dai, M. Wan et al. Combustion and Flame 235 (2022) 111712

Almost the same FCR (or almost the same HRR) can be observed at AL acknowledges support provided by Combustion Engine Research
significantly different (not only in magnitudes, but even in signs) Center (CERC).
local values of flame curvature, strain rate, or stretch rate. More-
over, the local strain rate in an extreme or leading point can be Supplementary materials
much higher than the extinction strain rate for the stationary, pla-
nar, strained laminar flame, thus, indicating pivotal importance of Supplementary material associated with this article can be
transient effects. Furthermore, other two-point local characteris- found, in the online version, at doi:10.1016/j.combustflame.2021.
tics of the extreme points, e.g., the local gradients of combustion 111712.
progress variables or species (e.g., the radical H) mass fractions, are
References
different in different flames. Therefore, conventional simple flame
configurations do not seem to be appropriate for accurately model- [1] G.A. Karim, Hydrogen as a spark ignition engine fuel, Int. J. Hydrogen Energy
ing all characteristics of the extreme or leading points. This is not 28 (2003) 569–577.
surprising bearing in mind highly unsteady and multidimensional [2] S. Verhelst, T. Wallner, Hydrogen-fueled internal combustion engines, Prog. En-
ergy Combust. Sci. 35 (2009) 490–527.
nature of the extreme or leading points. [3] H.L. Yip, A. Srna, A.C.Y. Yuen, S. Kook, R.A. Taylor, G.H. Yeoh, P.R. Medwell,
Second, a linear correlation between the local displacement Q.N. Chan, A review of hydrogen direct injection for internal combustion en-
speed and the curvature of the temperature iso-surface is well pro- gines: towards carbon-free combustion, Appl. Sci. 9 (2019) 4842.
[4] G.D.Y. Choubey, W. Huang, L. Yan, H. Babazadeh, K.M. Pandey, Hydrogen
nounced in the HRR-based extreme points. In the FCR-based ex- fuel in scramjet engines - a brief review, Int. J. Hydrogen Energy 45 (2020)
treme points, the correlation is significantly weaker, because fluc- 16799–16815.
tuations of the reaction and tangential-diffusion components of the [5] A.N. Lipatnikov, J. Chomiak, Molecular transport effects on turbulent flame
propagation and structure, Prog. Energy Combust. Sci. 31 (2005) 1–73.
displacement speed mitigate one another due to well pronounced [6] A.N. Lipatnikov, Fundamentals of Premixed Turbulent Combustion, CRC Press,
correlation between |∇ cF | and the tangential-diffusion term. Boca Raton, Florida, 2012.
Third, in the vicinity of the HRR-based leading points, the tem- [7] S. Yang, A. Saha, W. Liang, F. Wu, C.K. Law, Extreme role of preferential diffu-
sion in turbulent flame propagation, Combust. Flame 188 (2018) 498–504.
perature and HRR iso-surfaces differ significantly. In particular,
[8] S. Daniele, P. Jansohn, J. Mantzaras, K. Boulouchos, Turbulent flame speed
a scalar product of unit vectors normal to these iso-surfaces in for syngas at gas turbine relevant conditions, Proc. Combust. Inst. 33 (2011)
a HRR-based leading point is often significantly less than unity, 2937–2944.
[9] P. Venkateswaran, A. Marshall, D.H. Shin, D. Noble, J. Seitzman, T. Lieuwen,
whereas a scalar product of unit vectors normal to the cF . and
Measurements and analysis of turbulent consumption speeds of H2 /CO mix-
FCR iso-surfaces is always close to unity in the FCR-based lead- tures, Combust. Flame 158 (2011) 1602–1614.
ing points. Moreover, curvatures of the temperature and HRR iso- [10] P. Venkateswaran, A. Marshall, J. Seitzman, T. Lieuwen, Pressure and fuel ef-
surfaces have opposite signs in the HRR-based leading points. Thus, fects on turbulent consumption speeds of H2 /CO blends, Proc. Combust. Inst.
34 (2013) 1527–1535.
the local structure of the HRR-based extreme or leading points is [11] N. Chakraborty, A.N. Lipatnikov, Effects of Lewis number on conditional fluid
highly complicated and differ significantly from the local structure velocity statistics in low Damköhler number turbulent premixed combustion:
of the FCR-based extreme or leading points. a direct numerical simulation analysis, Phys Fluids 25 (2013) 045101.
[12] N. Chakraborty, I. Konstantinou, A.N. Lipatnikov, Effects of Lewis number on
Fourth, single-point local characteristics (temperature, species vorticity and enstrophy transport in turbulent premixed flames, Phys Fluids
mass fractions, rates of their production) of the FCR extreme points 28 (2016) 015109.
are comparable in all three studied turbulent flames and in the sta- [13] V.R. Kuznetsov, V.A. Sabelnikov, Turbulence and Combustion, Hemisphere Publ.
Corp., New York, 1990.
tionary, planar, critically strained laminar flame. In particular, the [14] H. Kido, M. Nakahara, A model of turbulent burning velocity taking the pref-
FCR exhibits weak fluctuations around mean values, which are ap- erential diffusion effect into consideration, JSME Int. J. 41 (1998) 666–673.
proximately equal in the three flames characterized by significantly [15] D. Bradley, P.H. Gaskell, X.J. Gu, A. Sedaghat, Premixed flamelet modelling: fac-
tors influencing the turbulent heat release rate source term and the turbulent
different Karlovitz numbers. Moreover, the extreme FCR is always
burning velocity, Combust. Flame 143 (2005) 227–245.
(in the three flames and at different instants) close to the peak [16] Ya.B. Zel’dovich, G.I. Barenblatt, V.B. Librovich, G.M. Makhviladze, The Mathe-
FCR in the critically strained laminar flame. These findings imply matical Theory of Combustion and Explosions, Consultants Bureau, New York,
1985.
that (i) the maximum FCR evaluated in the stationary, planar, crit-
[17] V. Karpov, A. Lipatnikov, V. Zimont, A test of an engineering model of pre-
ically strained laminar flame could be used to characterize, in a mixed turbulent combustion, Proc. Combust. Ins. 26 (1996) 249–257.
first approximation, the local FCR in the extreme or leading points [18] V.P. Karpov, A.N. Lipatnikov, V.L. Zimont, Flame curvature as a determinant of
in a turbulent flame, in line with the leading point concept, and preferential diffusion effects in premixed turbulent combustion, in: W.A. Sirig-
nano, A.G. Merzhanov, L. De Luca (Eds.), Advances in Combustion Science: In
(ii) almost the same extreme FCR can be reached in substantially Honor of Ya. B. Zel’dovich, vol. 173, Reston, VA: AIAA, 1997, pp. 235–250.
different local burning structures. These hypotheses resulting from [19] V.A. Sabelnikov, A.N. Lipatnikov, Transition from pulled to pushed premixed
the present work definitely require further assessment and devel- turbulent flames due to countergradient transport, Combust. Theory Modelling
17 (2013) 1154–1175.
opment. [20] V.A. Sabelnikov, A.N. Lipatnikov, Transition from pulled to pushed fronts in
premixed turbulent combustion: theoretical and numerical study, Combust.
Flame 162 (2015) 2893–2903.
Declaration of Competing Interest [21] K.Q.N. Kha, V. Robin, A. Mura, M. Champion, Implications of laminar flame fi-
nite thickness on the structure of turbulent premixed flames, J. Fluid Mech.
787 (2016) 116–147.
The authors declare that they have no known competing finan- [22] S. Verma, F. Monnier, A.N. Lipatnikov, Validation of leading point con-
cial interests or personal relationships that could have appeared to cept in RANS simulations of highly turbulent lean syngas-air flames with
influence the work reported in this paper. well-pronounced diffusional-thermal effects, Int. J. Hydrogen Energy 46 (2021)
9222–9233.
[23] W. Zhang, J. Wang, Q. Yu, W. Jin, M. Zhang, Z. Huang, Investigation of the fuel
effects on burning velocity and flame structure of turbulent premixed flames
Acknowledgement
based on leading points concept, Combust. Sci. Technol. 190 (2018) 1354–1376.
[24] S.H. Kim, Leading points and heat release effects in turbulent premixed flames,
This work has been supported in part by NSFC (Grant Nos. Proc. Combust. Inst. 36 (2017) 2017–2024.
[25] H.L. Dave, A. Mohan, S. Chaudhuri, Genesis and evolution of premixed flames
91752201 and 51976088), the Shenzhen Science and Technol-
in turbulence, Combust. Flame 196 (2018) 386–399.
ogy Program (Grant No. KQTD20180411143441009), Department [26] A.N. Lipatnikov, N. Chakraborty, V.A. Sabelnikov, Transport equations for reac-
of Science and Technology of Guangdong Province (Grant Nos. tion rate in laminar and turbulent premixed flames characterized by non-unity
2019B21203001 and 2020B1212030 0 01), and by project no. LCH- Lewis number, Int. J. Hydrogen Energy 43 (2018) 21060–21069.
[27] A. Amato, M. Day, R.K. Cheng, J. Bell, D. Dasgupta, T. Lieuwen, Topology and
2019011 under the Joint Program of Shenzhen Clean Energy Re- burning rates of turbulent, lean, H2 /air flames, Combust. Flame 162 (2015)
search Institute and SUSTech through contract CERI-KY-2019–003. 4553–4565.

15
H. Lee, P. Dai, M. Wan et al. Combustion and Flame 235 (2022) 111712

[28] A.N. Lipatnikov, J. Chomiak, Lewis number effects in premixed turbulent com- [42] H.L. Dave, S. Chaudhuri, Evolution of local flame displacement speeds in tur-
bustion and highly perturbed laminar flames, Combust. Sci. Technol. 137 bulence, J. Fluid Mech. 884 (2020) A46.
(1998) 277–298. [43] R.G. Abdel-Gayed, K.J. Al-Khishali, D. Bradley, Turbulent burning velocities and
[29] H.C. Lee, P. Dai, M. Wan, A.N. Lipatnikov, A DNS study of extreme and leading flame straining in explosions, Proc. R. Soc. London A 391 (1984) 393–414.
points in lean hydrogen-air turbulent flames - part I: local thermochemical [44] K.N.C. Bray, Methods of including realistic chemical reaction mechanisms in
structure and reaction rates, Combust. Flame, in press. turbulent combustion models, in: J. Warnatz, W. Jager (Eds.), Complex Chem-
[30] A. Kéromnès, W.K. Metcalfe, K.A. Heufer, N. Donohoe, A.K. Das, C.-J. Sung, ical Reaction Systems. Mathematical Modelling and Simulation, Springer, Hei-
J. Herzler, C. Naumann, P. Griebel, O. Mathieu, M.C. Krejci, E.L. Petersen, delberg (1987), pp. 356–375.
W.J. Pitz, H.J. Curran, An experimental and detailed chemical kinetic modeling [45] D. Bradley, A.K.C. Lau, M. Lawes, Flame stretch rate as a determinant of turbu-
study of hydrogen and syngas mixture oxidation at elevated pressures, Com- lent burning velocity, Phil. Trans. R. Soc. London A 338 (1992) 359–387.
bust. Flame 160 (2013) 995–1011. [46] V.L. Zimont, A.N. Lipatnikov, To computations of the heat release rate in tur-
[31] A. Abdelsamie, G. Fru, T. Oster, F. Dietzsch, G. Janiga, D. Thévenin, To- bulent flames, Dokl. Phys. Chem. 332 (1993) 592–594.
wards direct numerical simulations of low-Mach number turbulent reacting [47] V.L. Zimont, A.N. Lipatnikov, A numerical model of premixed turbulent com-
and two-phase flows using immersed boundaries, Comput. Fluids 131 (2016) bustion, Chem. Phys. Reports 14 (1995) 993–1025.
123–141. [48] D. Goodwin, N. Malaya, H. Moffat, R. Speth, Cantera: An object-Oriented Soft-
[32] T. Lundgren, Linearly Forced Isotropic Turbulence, Tech. Rep., Minnesota Uni- ware Toolkit for Chemical Kinetics, Thermodynamics, and Transport Processes,
versity of Minneapolis, 2003. Caltech, Pasadena, CA, 2009.
[33] C. Rosales, C. Meneveau, Linear forcing in numerical simulations of isotropic [49] P. Clavin, G. Joulin, High-frequency response of premixed flames to weak
turbulence: physical space implementations and convergence properties, Phys. stretch and curvature: a variable density analysis, Combust. Theory Modelling
Fluids 17 (2005) 095106. 1 (1997) 429–446.
[34] P.L. Carroll, G. Blanquart, The effect of velocity field forcing techniques on the [50] J.H. Chen, H.G. Im, Correlation of flame speed with stretch in turbulent pre-
Karman–Howarth equation, J. Turbul. 15 (2014) 429–448. mixed methane-air flames, Proc. Combust. Inst. 27 (1998) 819–826.
[35] C. Wang, M. Ge, Applying resolved-scale linearly forced isotropic turbulence in [51] N. Peters, P. Terhoeven, J.H. Chen, T. Echekki, Statistics of flame displacement
rational subgrid-scale modeling, Acta Mech. Sin. 35 (2019) 486–494. speeds from computations of 2-D unsteady methane-air flames, Proc. Combust.
[36] P. Pelce, P. Clavin, Influence of hydrodynamics and diffusion upon the stability Inst. 27 (1998) 833–839.
limits of laminar premixed flames, J. Fluid Mech. 124 (1982) 219–237. [52] T. Echekki, J.H. Chen, Analysis of the contribution of curvature to premixed
[37] M. Matalon, B.J. Matkowsky, Flames as gas dynamic discontinuities, J. Fluid flame propagation, Combust. Flame 118 (1999) 308–311.
Mech. 124 (1982) 239–260. [53] D.W. Mikolaitis, The interaction of flame curvature and stretch, part 1: the
[38] A.G. Class, B.J. Matkowsky, A.Y. Klimenko, A unified model of flames as gasdy- concave premixed flame, Combust. Flame 57 (1984) 25–31.
namic discontinuities, J. Fluid Mech. 491 (2003) 11–49. [54] D.W. Mikolaitis, Stretched spherical cap flames, Combust. Flame 63 (1986)
[39] A.P. Kelley, J.K. Bechtold, C.K. Law, Premixed flame propagation in a confining 65–111.
vessel with weak pressure rise, J. Fluid Mech. 691 (2012) 26–51. [55] A. Uemichi, M. Nishioka, Combustion mechanism of ultralean rotating counter-
[40] K.N.C. Bray, R.S. Cant, Some applications of Kolmogorov’s turbulence research flow twin premixed flame, Combust. Theory Modelling 19 (2015) 57–80.
in the field of combustion, Proc. R. Soc. London A 434 (1991) 217–240. [56] Z. Zhou, Y. Shoshin, F.E. Hernández-Pérez, J.A. van Ojien, L.P.H. de Goey, Effect
[41] G.K. Giannakopoulos, A. Gatzoulis, C.E. Frouzakis, M. Matalon, of pressure on the lean limit flames of H2 -CH4 -air mixture in tubes, Combust.
A.G. Tomboulides, Consistent definitions of “Flame Displacement Speed” Flame 183 (2017) 113–125.
and “Markstein Length” for premixed flame propagation, Combust. Flame 162
(2015) 1249–1264.

16

You might also like