You are on page 1of 7

Available online at www.sciencedirect.

com
Proceedings
ScienceDirect of the
Combustion
Institute
Proceedings of the Combustion Institute 35 (2015) 737–743
www.elsevier.com/locate/proci

The curvature Markstein length and the definition


of flame displacement speed for stationary
spherical flames
G.K. Giannakopoulos a, M. Matalon b, C.E. Frouzakis c,
A.G. Tomboulides a,⇑
a
Department of Mechanical Engineering, University of Western Macedonia, 50100 Kozani, Greece
b
Department of Mechanical Science and Engineering, University of Illinois, Urbana-Champaign, United States
c
Aerothermochemistry and Combustion Systems Laboratory, ETH Zurich, Switzerland

Available online 15 August 2014

Abstract

The separate effects of flame stretch and flame curvature are investigated in the present study consider-
ing a stationary spherical flame configuration that experiences no stretch but has a finite curvature. A the-
oretical expression for the flame displacement speed, derived from asymptotic analysis, is compared with
numerical simulations exhibiting very good agreement between the two. The notion of the curvature Mark-
stein length is introduced, a coefficient which is independent from diffusion and chemical reaction effects, in
contrast to the Markstein length associated with total flame stretch. It is shown that its value depends
strongly on the iso-surface chosen to define the flame, exhibiting large variations in the preheat zone
and more insensitive behavior close to the burned side.
Ó 2014 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

Keywords: Markstein length; Stationary spherical flame; Flame curvature; Flame stretch; Laminar premixed propane
flame

1. Introduction ical way. Rigorous asymptotic studies that


exploited the disparity in length scales between
The dependence of flame speed on flame the flame thickness (characterized by a diffusive
stretch has been studied extensively in the past, length scale) and the representative radius of cur-
theoretically and experimentally. The theoretical vature of the flame surface, led to an explicit
work has primarily followed the original idea of expression for the flame speed with a linear depen-
Markstein [1] who incorporated a dependence of dence on both curvature and hydrodynamic
the flame speed on curvature in a phenomenolog- strain, the combination of which constitutes the
total stretch rate experienced by the flame [2,3].
The flame speed-flame stretch relation is modu-
⇑ Corresponding author. Fax: +30 24610 56631. lated by a coefficient on the order of flame thick-
E-mail address: atompoulidis@uowm.gr (A.G. ness, which mimics the effects of diffusion and
Tomboulides). chemical reactions occurring inside the flame

http://dx.doi.org/10.1016/j.proci.2014.07.049
1540-7489/Ó 2014 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
738 G.K. Giannakopoulos et al. / Proceedings of the Combustion Institute 35 (2015) 737–743

zone, that has become known as the Markstein concerned with diffusion flames, examining the
length. Explicit expressions for the Markstein dependence of the flame standoff distance and
length for two-component mixtures, with temper- flame extinction on the burner dimension and
ature-dependent diffusivities, and reactions that other physicochemical properties, in the presence
proceed with different reaction orders, were and absence of radiative heat loss [15–18]. Fewer
derived in [4]. studies were concerned with premixed flames
The dependence of flame speed on flame [19,20].
stretch, for weakly stretched flames, has also been The more elaborate case of a spherically
verified experimentally in various configurations, expanding flame, where the flame is curved and
and numerous studies have been carried out for stretched, is the subject of a separate study [21].
measuring the Markstein length (e.g., [5–9]). The
linear dependence of flame speed on flame stretch
was exploited experimentally to develop a meth- 2. Flame stretch and flame displacement speed
odology for extracting the laminar flame speed
from measurements of stretched flames through A general definition for flame stretch rate was
linear extrapolation to zero stretch, c.f., [10,11]. given by Karlovitz [22] as the Lagrangian time
An interpretation of the asymptotic results, based derivative of an element of the flame surface area A
on a control volume analysis, was presented in 1 dA
[12] and used to extract global flame parameters. K¼ ð1Þ
A dt
The Markstein length as determined from the
asymptotic analysis, is uniquely defined in the limit and can be expressed [23,4] as a combination of
when the ratio of the thermal thickness to the rep- the flame surface curvature j and the underlying
resentative curvature tends to zero. Since real hydrodynamic strain K S it experiences, namely
flames have a finite thickness and measurements K ¼ SLj þ K S
are typically taken at a specific reference location
inside the flame, large deviations in the Markstein Curvature is equal to j ¼ r  n, where n is the unit
length, itself known to within an amount propor- normal vector of a flame iso-surface, taken to be
tional to the flame thickness, may be encountered positive (negative) when the flame is convex
between various measurements. The same can be (concave) to the reactants. Hydrodynamic strain
said about numerical simulations, where the deter- depends on velocity v and is given by
mination of the flame location and the flame dis- K S ¼ n  E  n, where E ¼ rv þ ðrvÞT is the
placement speed are based on a specific iso- strain rate tensor; K S affects the local reactivity
surface of temperature or mass fraction within through thermo-diffusive effects.
the flame zone. To remove such ambiguities Mat- Since the flame is stationary and the diverging
alon and co-workers [13,14] provided an explicit flow through the flame is radial, a stationary
expression for the mass flux through the flame spherical flame sustained by a source flow experi-
which enables the determination of the displace- ences no stretch. The flame is curved, with a
ment speed and Markstein length at any location curvature j ¼ 2=R that depends on the flame
within the preheat zone, provided the linear location r ¼ R. As will be clarified below, the
dependence of flame speed on stretch remains stationary flame experiences normal strain
valid. This expression allows for a better compar- K S ¼ 2S L =R that balances the curvature terms
ison of experiments and/or simulations with such that K ¼ 0. In contrast, the outwardly prop-
theory. agating spherically-symmetric flame is stretched
In this paper we provide a systematic and con- _
and the stretch rate K ¼ 2R=R based on definition
sistent comparison between theory and simula- (1) decreases as the flame grows larger (here, the
tions for stationary spherical premixed flames, dot denotes differentiation with respect to time).
showing excellent agreement between the two. In that case, it can be verified [21] that the flame
The stationary flame is not stretched but has a experiences normal strain K S ¼ 2ðr  1ÞS L =R,
finite curvature, and the displacement speed in which, when combined with the curvature contri-
this case at any location reduces to a dependence bution jS L ¼ ð2=RÞS L , yields K ¼ 2rS L =R ¼
on curvature only. In this setup, the combustible _
R=2R, where r ¼ qu =qb is the thermal expansion
mixture is supplied steadily through a porous and the subscripts u and b denote the unburned
spherical burner and the flame is stabilized at a and burned states, respectively.
specific location within the radially diverging flow. In most numerical simulations of premixed
The characteristics of burner-stabilized stationary flame propagation, the flame displacement speed
spherical flames was the subject of numerous the- (FDS), defined as the speed of a scalar iso-surface
oretical investigations, primarily because of their relative to the flow of fresh reactants, is used to
fundamental and practical importance, but has represent the flame motion. For example, using
not been a popular choice for experimental studies temperature as the progress variable, the FDS
perhaps because of practical difficulties in stabiliz- can be computed from the overall energy conser-
ing the flame. Most of the investigations were vation for the mixture [24] as
G.K. Giannakopoulos et al. / Proceedings of the Combustion Institute 35 (2015) 737–743 739
" #
1 XN XN
two separate Markstein lengths can be identified,
Sd ¼  hi x_ i þ r  ðkrT Þ  q cp i Y i Vi  rT one associated with total stretch and the other
qcp jrT j i¼1 i¼1
with curvature. For a spherically expanding flame,
ð2Þ j and K are proportional to each other and
where all quantities are evaluated on the selected decrease with increasing radius such that both
isotherm. In this expression hi and x_ i are the spe- tend to zero at the same rate as R ! 1, leading
cific enthalpy and production/consumption rate to a single Markstein length for any selected iso-
of species i; k is the mixture-averaged thermal con- surface within the flame zone. In contrast, the
ductivity and cp i ; Y i and Vi are the specific heat stretch rate in a stationary spherical flame is zero
capacity, mass fraction and diffusion velocity of but its curvature approaches a finite value
species i, respectively, for a mixture consisting of j ¼ 2=R. Thus, the theoretical expression for
N species. The first term on the right hand side the flame speed reduces to
represents the effect of the heat release rate and Z r
e kðxÞ
the last two terms account for heat conduction S f ¼ S L  Lc jS L ; with Lc ¼ lf dx
and heat transfer through species diffusion, h x  1

respectively. The last term is typically an order ð5Þ


of magnitude smaller than the other terms and is where Lc is the “curvature Markstein length”,
identically zero if the specific heat for all species which varies with the position within the flame
is equal. For a meaningful comparison between zone, i.e., the isovalue chosen to define the flame.
values of the FDS defined at different locations, Figure 1 shows typical values of the Markstein
S d is often normalized by the ratio of the local number Lc =lf at different locations throughout
density q to the density of the fresh mixture qu the flame zone identified by the temperature ratio
to obtain the density-weighted flame displacement T =T u . Note that Lc ! 1 when T ! T u , and
speed eS d ¼ qS d =qu . Lc ¼ 0 for T ¼ T b , for any functional dependence
As noted in the introduction, an explicit kðxÞ.
expression for the mass flux through the flame
was derived in [4] which enabled the determina-
tion of an expression for the flame displacement 3. Asymptotic solution
speed at any location within the flame preheat
zone [14]. The density-weighted flame displace- Stationary spherical flames. The spherical flame
ment speed can then be written in the form configuration adopted in the present study is
  Z r Z r 
L kðxÞ kðxÞ shown in Fig. 2(a). We consider a flame stabilized
e
S f ¼ S L  lf K  dx  dx in the diverging flow from a point source from
lf h xðx  1Þ 1 x
Z r  which the combustible mixture is ejected, as
kðxÞ shown in Fig. 2(a). Assuming spherical symmetry,
þ jS L dx ð3Þ
h x  1 mass conservation implies that the total mass flow
where S L is the laminar flame speed, lf ¼ Dth =S L is rate
the flame thickness with Dth the thermal diffusivity m_ ¼ 4pqr2 m ð6Þ
of the mixture, kðxÞ is the functional dependence
of thermal conductivity on temperature and where m is the flow velocity in the radial direction
 Z r r. Hence, the mass crossing a sphere of radius r
r kðxÞ bðLeeff  1Þ
L¼ dx þ
r1 1 x 2ðr  1Þ 20
Z r   
r  1 kðxÞ λ=T
 ln dx lf ð4Þ 1/2
1 x1 x 15 λ=T
λ=1
where b is the Zel’dovich number and Leeff the
mixture effective Lewis number; details on the pre-
Lc / lf

10
vious expressions can be found in [14,21]. The
equation of state implies that qT ¼ qu T u , such
that r ¼ T b =T u , where T b is the adiabatic flame
5
temperature (the mixture molecular weight is
assumed constant). The integrands in (3) are func-
tions of the dimensionless temperature h ¼ T =T u , 0
such that x ¼ 1 represents the unburned tempera- 0 1 2 3 4 5 6 7 8
Temperature T / Tu
ture, x ¼ r the adiabatic flame temperature, and
x ¼ h any temperature in the preheat zone, i.e.,
Fig. 1. Theoretical curvature Markstein number Lc =lf
with 1 < h < r.
as a function of flame temperature for the most common
Of interest here are the terms multiplying assumptions of the temperature dependence of k on
flame stretch and curvature. It is evident that temperature.
740 G.K. Giannakopoulos et al. / Proceedings of the Combustion Institute 35 (2015) 737–743

Burned gas

r
r−R = 0

Unburned
gas

(a) (b)

Fig. 2. (a) Stationary flame configuration and (b) schematic view of the computational mesh used in the simulations.

_ When the
per unit time is constant and equal to m. based on the Gauss–Lobatto–Legendre quadra-
mixture is ignited, a spherical flame is stabilized ture points. The discretized equations are inte-
after an initial transient at a distance r ¼ R. grated in time with a highly-efficient parallel
Within the context of the hydrodynamic theory code based on nek5000 [27] and a high-order split-
[4], the flame is treated as a discontinuity separat- ting scheme for low-Mach number reactive flows
ing burned from unburned gas and the flame posi- [28]. A semi-explicit integration scheme is used
tion is determined from the flame speed relation for the continuity and momentum equations,
S f ¼ S L  LK. The velocity on the unburned side while the energy and species equations are solved
at r ¼ R is thus given by (6) with q ¼ qu . Since the implicitly using CVODE [29]. The Chemkin trans-
flame is stationary (R ¼const.) and experiences no port and chemistry libraries are used for the eval-
stretch, mjR ¼ S L determines the flame position as uation of the transport properties and chemical
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi reaction rates. The reference length and the refer-
R ¼ m=4pq_ uSL: ð7Þ ence velocity used for non-dimensionalization
Mass conservation across the flame surface in the reported simulations are d ¼ ðT b  T u Þ=
implies that the velocity on the burned side of ð@T =@xÞmax and S L , respectively and are calculated
the flame is mjRþ ¼ m=4pq 2 based on the planar adiabatic flame solution
_ b R . The velocity field
is therefore given by using PREMIX ([30]). For all other variables the
( m_ properties of the unburned mixture are used as
4pqu r2
for r < R reference quantities.
m¼ m_
ð8Þ Chemistry is described by the single-step global
4pq r2
for r > R
b
reaction C3H8 + 5O2 ! 3CO2 + 4H2O with reac-
from which it can be verified that, indeed, tion rate r ¼ A expðEa =RT Þ½C3 H8 ½O2 . For the
K S ¼ 2S L =R at the flame sheet, as noted above. conditions considered, the reaction orders were
Spherically expanding flames. The spherically taken as unity while the activation energy is
expanding flame is discussed in detail in [4], which Ea ¼ 40:2 kcal=mol and the pre-exponential fac-
should be consulted for further details. tor A ¼ 5:4  1015 cm3 =mole  s [31]. The compu-
tational domain, which spans Rin < r < Rout as
4. Numerical simulations shown schematically in Fig. 2(b), is discretized
using 12,000 spectral elements. For all variables,
In order to verify the theoretical expression for axisymmetric boundary conditions (BC) are
the flame speed, a series of numerical simulations applied along the axis and zero-Neumann BC
were performed investigating the stationary spher- along the outflow boundary at Rout ¼ 100d. The
ical flame configuration of Fig. 2(a), considering a inner boundary at Rin ¼ 10d corresponds to the
stoichiometric premixed propane-air mixture at surface of a porous sphere that supplies the com-
temperature 298 K and pressure 1 atm. The bustible gas. Dirichlet BC are then specified at
time-dependent conservation equations are solved r ¼ Rin for the velocity (m ¼ min ) and temperature
in their low-Mach number form and are discret- (T ¼ T u ) and adiabatic BC for the mass fractions
ized in space using the spectral element method of all species. Mass conservation (6) implies that
2
[25,26]. The gaseous mixture is assumed to be min ¼ m=4pq
_ u Rin .
ideal. The solution and geometry are expressed The values min =S L ¼ 4; 9 and 16 are chosen to
as N th -order tensor product Lagrange polynomials investigate three cases with different supply condi-
G.K. Giannakopoulos et al. / Proceedings of the Combustion Institute 35 (2015) 737–743 741

tions, corresponding to different flame radii at predicted by the theory as the integral in (5) is
steady state. Using (7), the flame is stabilized at equal to zero only when h ¼ r (i.e., in the burned
R ¼ Rin ðmin =S L Þ1=2 , and for Rin =d ¼ 10, the flame side), resulting in eS f ¼ S L . It also should be noted
radii at steady state for the three cases considered that during the transient the stretch rate for the
are R=d ¼ 20; 30 and 40. In all cases, the initial flame sustained by a point source is at least an
condition is a fully-burned mixture for R=d > 15 order of magnitude smaller than its value in the
connected smoothly via a hyperbolic tangent pro- outwardly propagating flame and becomes identi-
file to the fresh reactants at R=d < 15. Figure 3 cally zero only when the flame reaches steady
shows the simulation results for min =S L ¼ 4. The state.
temporal variations of flame radius, curvature Figure 5 shows the density weighted FDS as a
and stretch rate are shown in Fig. 3(a). As can function of flame curvature normalized by the
be observed, the flame propagates outwards until steady-state value (jst ) for each case min =S L ¼ 4; 9
it is stabilized at the radial location R=d ¼ 20. The and 16 and for the temperature iso-surfaces of
stretch rate tends to zero, while the curvature T  ¼ 1:01 and T  ¼ 7:50, corresponding to iso-
approaches a finite value jjj ¼ 2d=R ¼ 0:1 (the surfaces close to the unburned and the burned
absolute value is shown in the figure). At steady sides respectively. It can be observed that in all
state, the propagation speed balances the flow cases, when evaluated on the burned side the
velocity at the flame location, as can be seen in FDS approaches the value of S L (as Lc ! 0 in this
Fig. 3(b). region). This is not the case on the cold side of the
flame where the FDS at steady state is different for
each min (namely jst ).
5. Results and discussion In order to provide a quantitative comparison
between the results obtained from the simulations
Figure 4(a) and (b) shows the density-weighted with those predicted from the asymptotic theory,
FDS as a function of the stretch rate for an out- Fig. 6(a) reports the calculated density-weighted
wardly propagating flame and a stationary flame FDS e S d using (2) along with the theoretical flame
sustained by a point source during its transition
to steady state for the same mixture conditions. speed e S f using (5) for the three stationary flames
The density-weighted FDS is calculated using (2) considered in this study. The limiting values of
and at various temperature iso-contours inside e
S d in this figure were calculated using the regres-
the flame. We note that for the outwardly propa- sion lines for selected temperature iso-surfaces at
gating flame at the limit of zero stretch, all speeds the limit of zero stretch and steady-state curvature
approach the value of 1 (i.e., the laminar flame for each case (see Fig. 4(b)). The numerical results
speed S L ). In contrast, for the flame sustained by were further processed for the calculation of the
a point source the FDS corresponding to different curvature Markstein length. The FDS e S d at
temperature iso-surfaces approaches different lim- steady state is plotted in Fig. 6(b) as a function
its as stretch rate goes to zero (at steady state) and of the (steady-state) curvature jlf jst j. The result-
only the iso-surfaces very close to the burned side ing slope is the curvature Markstein number
(i.e., T  P 7:5, in this case the maximum flame Lc =lf for each temperature iso-surface, which is
temperature is T b ¼ 8:02) approach the value of compared in Fig. 7 to the corresponding theoreti-
S L (in a density-weighted sense). This behavior is cal values using (5). The agreement between

0.2 20 10

19 8
0.15

18 6
0.1
17 4

0.05 Stretch rate Temperature


Curvature 16 2 Velocity
Flame radius
0 15 0
0 5 10 15 20 25 30 10 20 30 40 50 60 70 80 90 100

(a) (b)
Fig. 3. (a) Temporal evolution of the (dimensionless) radius r=d, stretch rate dK=S L and curvature jdjj, and (b)
temperature and flow velocity profiles in the radial direction (solid: temperature, dashed: flow velocity) at steady state
(min ¼ 4S L ).
742 G.K. Giannakopoulos et al. / Proceedings of the Combustion Institute 35 (2015) 737–743

(a) (b)
Fig. 4. FDS e
S d as a function of stretch rate lf K=S L for selected temperature iso-surfaces inside the flame zone for (a) an
outwardly propagating spherically flame, (b) a stationary spherical flame (min ¼ 4S L ).

Fig. 5. FDS e S d as a function of curvature for selected Fig. 7. Comparison between theory (solid curves) and
temperature iso-surfaces inside the flame zone and simulations (symbols) of the curvature Markstein num-
different inflow velocities. Curvature is normalized by ber Lc =lf .
the steady-state value jst for each inflow velocity.

theory and simulations is remarkable. The kðT Þ ¼ T 0:87 , which approximately corresponds
theoretical curves in both Figs. 6(a) and 7 were to the detailed transport behavior observed in
obtained using a functional dependence for the the simulations. To compare with the asymptotic
mixture thermal conductivity on temperature theory, the numerical results were appropriately

(a) (b)
Fig. 6. (a) Comparison of the density-weighted FDS between theory (solid curves) and simulations (symbols) for
different inflow velocities; (b) FDS from simulation as a function of flame curvature jlf jj for selected temperature iso-
surfaces inside the flame zone.
G.K. Giannakopoulos et al. / Proceedings of the Combustion Institute 35 (2015) 737–743 743

normalized by the representative flame thickness References


used in the theoretical formulation, lf instead of
d (for the conditions considered in this study, [1] G. Markstein, Nonsteady Flame Propagation,
lf ¼ 5:07  103 cm and d ¼ 3:17  102 cm). McMillan Publication, New York, 1964.
[2] M. Matalon, B.J. Matkowsky, J. Fluid Mech. 124
(1982) 239–259.
6. Conclusions [3] P. Pelce, P. Clavin, J. Fluid Mech. 124 (1982) 219–237.
[4] M. Matalon, C. Cui, J. Bechtold, J. Fluid Mech.
The objective of the present work is to provide 487 (2003) 179–210.
a systematic and consistent comparison between [5] C.K. Wu, C.K. Law, Proc. Combust. Inst. 20 (1985)
theory and simulations for stationary spherical 1941–1949.
premixed flames. In particular, an expression for [6] G. Searby, J. Quinard, Combust. Flame 82 (1990)
the flame displacement speed – derived from 298–311.
[7] L.K. Tseng, M.A. Ismail, G.M. Faeth, Combust.
asymptotic theory – was evaluated using numeri- Flame 95 (1993) 410–426.
cal simulations in a flame configuration which is [8] U. Müller, M. Bollig, N. Peters, Combust. Flame
not subject to stretch but has finite curvature. 108 (1997) 349–356.
The results demonstrated a very good agreement [9] E. Varea, V. Modica, A. Vandel, B. Renou,
between the asymptotic theory and the simula- Combust. Flame 159 (2) (2012) 577–590.
tions. It was shown that only when the flame cur- [10] K.T. Aung, M.I. Hassan, G.M. Faeth, Combust.
vature and the total stretch rate are proportional Flame 109 (1997) 1–24.
to each other (as in the case of spherically-expand- [11] S. Kwon, L.K. Tseng, G.M. Faeth, Combust. Flame
ing flames), a single Markstein length emerges. 90 (1992) 230–246.
[12] C. Sun, C.J. Sung, L. He, C.K. Law, Combust.
This is not the case in the stationary flame config- Flame 118 (1999) 108–128.
uration, where a distinction is made between the [13] J. Tien, M. Matalon, Combust. Flame 84 (1991)
Markstein length associated with total stretch rate 238–248.
and the curvature Markstein length. The latter [14] J. Bechtold, M. Matalon, Combust. Flame 127
proved to be independent of Lewis and Zel’dovich (2001) 1906–1913.
numbers, which implies that its notion does not [15] K. Mills, M. Matalon, Combust. Sci. Technol. 129
involve any diffusion and chemical reaction (1997) 295–319.
effects. Moreover, it was found that only iso-sur- [16] K. Mills, M. Matalon, in: Twenty-Seventh Sympo-
faces close to the burned side of the flame give sium (International) on Combustion/The combustion
Institute, pp. 2535–2541.
results which are not sensitive to the chosen iso- [17] S. Yoo, E. Christiansen, C. Law, Proc. Combust.
value. In this region, the curvature Markstein Inst. 29 (2002) 29–36.
length is zero and the flame displacement speed [18] Q. Wang, B. Chao, Combust. Flame 158 (2011)
is equal to the laminar flame speed S L . Otherwise, 1532–1541.
large variations in the calculated/measured Mark- [19] J. Qian, J.K. Bechtold, C.K. Law, Combust. Flame
stein length and the FDS should be expected. This 110 (1997) 78–91.
is also true for spherically expanding flames as [20] S. Yoo, D. Zhu, C.K. Law, Rev. Sci. Instrum. 77
shown in [21]. Finally, it is worth pointing out (2006) 075102.
that the stationary flame configuration examined [21] G.K. Giannakopoulos, A. Gatzoulis, C.E. Frouza-
kis, M. Matalon, A.G. Tomboulides, Combust.
in this study allows the determination of the Flame (submitted for publication).
unstretched flame speed S L without the need for [22] B. Karlovitz, D. Denniston, D. Knapschafer, F.
extrapolation to zero stretch, which often leads Wells, Proc. Combust. Inst. 4 (1951) 613.
to inconsistencies. [23] M. Matalon, Combust. Sci. Technol. 31 (1983) 169–181.
[24] T. Poinsot, D. Veynante, Theoretical and Numerical
Combustion, Edwards, 2001.
Acknowledgments [25] A.T. Patera, J. Comput. Phys. 58 (1984) 468–488.
[26] M.O. Deville, P.F. Fischer, E.H. Mund, High-order
This work has been co-financed by EU-ESF Methods for Incompressible Fluid Flows, Cambridge
and Greek NF through the Operational Program University Press, 2002.
“Education and Lifelong Learning” of the NSRF- [27] P.F. Fischer, J.W. Lottes, S.G. Kerkemeier,
Research Funding Program: Heracleitus II nek5000 Web page, 2008.
(GKG). It was also partially supported by the [28] A.G. Tomboulides, J.C.Y. Lee, S.A. Orszag, J. Sci.
EU-FP7-SST-2008-RTD-1 under grant 233615- Comput. 12 (1997) 139–167.
LESSCCV (AGT), the National Science Founda- [29] G.D. Byrne, A.C. Hindmarsh, Int. J. High Perform.
Comput. Appl. 13 (1999) 345–365.
tion under grant CBET 1067259 (MM), and the [30] R.J. Kee, J.F. Grcar, M.D.S. Miller, J.A. Miller,
Swiss National Science Foundation under grant PREMIX: A Fortran Program for Modeling Steady
200021-149500 (CEF). MM would like to thank Laminar One-Dimensional Premixed Flame, Tech-
the Aerothermochemistry and Combustion Sys- nical Report, SAND85-8240, 1985.
tems Laboratory of ETHZ for the hospitality in [31] J. Kim, K.N. Kim, S.H. Won, O. Fujita, J.
the spring of 2014 during which part of this work Takahashi, S.H. Chung, Combust. Flame 145
was completed. (2006) 181–193.

You might also like