Combustion and Flame 143 (2005) 79–96
www.elsevier.com/locate/combustflame
Suppression effects of diluents on laminar premixed
hydrogen/oxygen/nitrogen flames
L. Qiao ∗ , C.H. Kim, G.M. Faeth
Department of Aerospace Engineering, The University of Michigan, Ann Arbor, MI 48109, USA
Received 27 September 2004; received in revised form 10 April 2005; accepted 10 May 2005
Available online 2 August 2005
Abstract
Laminar burning velocities and the flame response to stretch, as characterized by Markstein numbers, were
determined experimentally and computationally for outwardly propagating spherical laminar premixed flames.
The mixtures studied were premixed hydrogen/air/diluent and hydrogen/30% oxygen and 70% nitrogen (by volume)/diluent flames, with the latter condition of interest for pre-external vehicular activity preparation activities
on board manned spacecraft. Other flame conditions were room temperature (300 K), fuel-equivalence ratios of
1.0 and 1.8, pressures of 0.5, 0.7, and 1.0 atm, diluents including helium, argon, nitrogen, and carbon dioxide as
suppression agents, and diluent concentrations of 0–40% (by volume), which implies oxygen indices of 30–10 for
present conditions. Predicted flame behavior was obtained from one-dimensional, spherically symmetric, steady,
and time-dependent numerical simulations with variable-property and multicomponent transport and with detailed hydrogen/oxygen chemical kinetics. Flames studied were sensitive to stretch, yielding unstretched/stretched
laminar burning velocity ratios of 0.6–1.25 for conditions well away from quenching conditions (e.g., Karlovitz
numbers; Ka 0.5). Diluents became more effective (provided greater reductions of the laminar burning velocity
for a given diluent concentration) in the order helium, argon, nitrogen, and carbon dioxide, which reflects their
increased capabilities either to quench the reaction zone by increased specific heats or to reduce flame velocities by
reduced transport rates. The addition of diluents generally decreased Markstein numbers, which made the flames
more susceptible to preferential-diffusion instability. This effect increases flame speeds and tends to counteract
the effect of diluents to reduce laminar burning velocities.
 2005 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
Keywords: Extinction; Flame-stretch interactions; Hydrogen; Fire extinguishing
1. Introduction
Halons have been very successful as chemically active flame suppression agents in applications
where effective and clean control of unwanted fires
is needed; see Drysdale [1] and Tuhtar [2]. Unfor* Corresponding author. Fax: +1 734 763 0578.
E-mail address: [email protected] (L. Qiao).
tunately, halons also contribute to the depletion of
stratospheric ozone that protects the Earth’s surface
from harmful ultraviolet solar radiation. Due to this
undesirable environmental effect, halon manufacture
was stopped in 1994, except for limited production
in some developing countries, under the terms of the
Montreal Protocol [3]. Subsequently, many experimental and computational studies have been undertaken to gain a better understanding of the mechanism
0010-2180/$ – see front matter  2005 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
doi:10.1016/j.combustflame.2005.05.004
80
L. Qiao et al. / Combustion and Flame 143 (2005) 79–96
Nomenclature
D
K
Ka
L
Ma
P
rf
SL
SL∞
t
T
Mass diffusivity
Flame stretch, Eq. (1)
Karlovitz number, KDu /SL2
Markstein length
Markstein number, L/δD
Pressure
Flame radius
Laminar burning velocity based on unburned gas properties
Value of SL at the largest radius observed
Time
Temperature
of flame suppression of the chemically active halons
and their potential replacements; see Safieh et al. [4],
Sheinsohn et al. [5], Walravens et al. [6], Miziolek
and Tsang [7], McIlroy and Johnson [8], Linteris and
Truett [9], Noto et al. [10,11], Casias and McKinnon
[12,13], Linteris et al. [14], Takahashi et al. [15], Saso
et al. [16–18], Kim et al. [19], Faeth et al. [20], and
references cited therein. These studies have shown
that chemically active suppression agents are unusually effective because they interrupt the chemical
pathway of fuel oxidation. Unfortunately, chemically
active agents often generate substances in flame environments that prevent their use in confined spaces that
contain living organisms. Motivated by this observation, the objective of the present investigation was to
study the properties of typical chemically passive suppression agents—diluents that avoid the limitations of
the Montreal Protocol [3] and problems of the generation of substances in flames that are harmful to living
organisms in confined spaces. Effects of diluents on
laminar premixed flames were studied both experimentally and computationally.
An important issue concerning the present study
is the justification for considering only the suppression properties of laminar premixed flames. Although
most practical flames are turbulent, turbulent flames
are difficult to study because experimental conditions
are substantially complicated by the need to describe
and treat a variety of turbulence properties. Another
advantage of laminar flames is that they generally
are tractable for detailed numerical simulations, unlike turbulent flames, enhancing capabilities to use
computations to supplement information about suppression behavior from directly measured properties.
In addition, laminar flames are also relevant to turbulent flames based on widely accepted laminar flamelet
concepts of turbulent flames. Finally, due to the complexities of turbulent flames, it seems unlikely that un-
Xi
Mole fraction of species i
Greek symbols
δD
ρ
ϕ
Characteristic flame thickness, Du /SL
Density
Fuel-equivalence ratio
Subscripts
b
max
u
∞
Burned gas
Maximum observed value
Unburned gas
Unstretched flame condition
derstanding of the suppression of turbulent premixed
flames will precede understanding of the suppression
of laminar premixed flames.
Another issue concerning the present study involves limiting considerations to premixed flames,
even though both premixed and nonpremixed (diffusion) flames are important in practice. This was
done because premixed flames are most relevant to
processes of flame suppression (e.g., even points of
flame attachment in diffusion flames are largely controlled by premixed flame phenomena). In addition,
premixed flames lend themselves to well-defined experimental and computational conditions that simplify the interpretation of both experimental and computational results.
Other limitations used to control the scope of
the present study involved considering combustion
processes involving only hydrogen/oxygen chemical kinetics for outwardly propagating spherical premixed flames. Limiting combustion to hydrogen/
oxygen chemical kinetics is reasonable because these
kinetics are fundamentally important for all combustion processes of hydrocarbons in air, are well known,
and are sufficiently simple so that numerical simulations involving these reactants are computationally
tractable. Furthermore, these reactants provide conservative suppression properties because they generally are the hardest to extinguish among combustibles
of practical interest; see Wieland [21]. In addition,
outwardly propagating spherical premixed flames are
also attractive because they do not involve complex
quenching processes near surfaces. Finally, subsequent discussion will also show that outwardly propagating spherical flames are particularly convenient
for directly measuring and computing the fundamental properties of various suppression agents.
To fix ideas, two general reactant systems were
considered during the present study: (1) hydrogen and
L. Qiao et al. / Combustion and Flame 143 (2005) 79–96
air at pressures of 0.5 and 1.0 atm, to represent a
conservative combustible mixture for suppression at
altitudes typical of cities on Earth, and (2) hydrogen
in an atmosphere consisting of 30% oxygen and 70%
nitrogen (by volume) at a pressure of 0.7 atm, which is
the atmosphere used during external vehicular activity
(EVA) preparation of astronauts on board spacecraft;
see Wieland [21]. Finally, only simple diluents were
considered (e.g., helium, argon, nitrogen and carbon
dioxide) as suppression agents to emphasize effects
of specific heats and transport properties known to
influence diluent performance (see Huggett [22,23])
without having to deal with the complexities of multicomponent flame suppression processes.
A difficulty that is encountered when laminar premixed flames are used to study the suppression properties of diluents is that flame/stretch interactions significantly affect the laminar burning velocities and
structure of laminar premixed flames [24–29] and the
rate of propagation of turbulent premixed flames typical of practical applications [30,31]. To deal with this
problem, diluent performance during the present investigation was determined by using unstretched laminar burning velocities to characterize the intensity
of combustion of the flames and by using Markstein
numbers (Ma) to characterize the sensitivity of the
flames to effects of stretch. Fortunately, outwardly
propagating spherical laminar premixed flames provide a straightforward determination of unstretched
laminar burning velocities and Markstein numbers,
as demonstrated by Aung et al. [32,33] and Kwon
and Faeth [34] for hydrogen laminar premixed flames.
Another finding of these studies is that their flame
structure predictions suggested that mainly H and to a
lesser degree OH radical production and transport are
important aspects of preferential-diffusion/stretch interactions. This is not surprising, however, due to the
well known proportionality between laminar burning
velocities and H radical concentrations of hydrogen
laminar premixed flames, first pointed out by Padley
and Sugden [35] based on the laminar burning velocity measurements of Jahn that are cited in Lewis
and von Elbe [36] and subsequently noted by others
for hydrogen premixed flames for various conditions,
e.g., Kim et al. [19], Butler and Hayhurst [37], and
references cited therein. Another aspect of the findings of Kim et al. [19] is that changes of flame conditions that tended to reduce the concentrations of H
and OH radical concentrations in the reaction zone
also tended to make these flames more susceptible to
preferential-diffusion instability as measured by reduced values of the Markstein number. This behavior
has the potential to increase flame speeds (the absolute flame velocity in laboratory coordinates) due
to the creation of flame surface area by the distortion or wrinkling of the flame surface as a result of
81
the action of preferential-diffusion instability. Thus,
the application of flame suppression agents to premixed flames involves two counteracting effects. Suppression agents have the capability to reduce H and
OH radical concentrations in the reaction zone, which
leads to corresponding reductions of laminar burning velocities which tend to reduce flame intensities and thus suppress the flame. Furthermore, the
same capability of suppression agents to reduce H
and OH radical concentrations in the reaction zone
also enhances the potential for the development of
preferential-diffusion-induced flame surface instabilities that increase flame speeds, which tends to increase flame intensities and thus reduces effects of
flame suppression.
2. Experimental methods
2.1. Apparatus
Experimental methods were similar to past work
and will be described very briefly; see Aung et al. [32,
33] and Kwon and Faeth [34] for more details. The
experiments were conducted in a spherical windowed
chamber having an inside diameter of 360 mm and
an internal volume of 0.024 m3 . Optical access was
provided by two 100-mm-diameter quartz windows
mounted opposite one another along a horizontal line
passing through the center of the chamber. The chamber was capable of operation over a pressure range
extending from complete vacuum up to a maximum
of 34 atm.
The reactant mixture was prepared within the
chamber by adding gases at appropriate partial pressures to reach the total initial pressure of the reactant
mixture for a test (0.5, 0.7, and 1.0 atm for the present
test range). The reactant gases were mixed using a
small metal fan located inside the chamber with the
fan-induced motion allowed to decay before ignition
(5–10 min for mixing and at least 30 min for decay);
given these conditions, motion picture shadowgraphs
did not indicate any distortion of the flame surface
or convection of the flame kernel from its position
centered on the spark kernel. After combustion was
complete, the chamber was vented to the laboratory
exhaust system and then purged with dry air to remove condensed water vapor prior to refilling for the
next test.
The combustible mixture was spark-ignited at the
center of the chamber using electrodes extending
from the top and bottom of the chamber. One electrode was fixed, whereas the other electrode could be
moved with a micrometer having a positioning accuracy of 10 µm. The tips of the electrodes were fine
tungsten wires having diameters of 250 µm and free
82
L. Qiao et al. / Combustion and Flame 143 (2005) 79–96
lengths of 40 mm. The spark gap was varied in the
range 0.5–2.0 mm, with the larger gaps used to ignite
flames having relatively small laminar burning velocities that required relatively large ignition energies.
The spark energy was supplied by a high-voltage capacitor discharge circuit having a variable capacitance
(100–7000 pF) and voltage (0–10 kV) and a discharge
time of roughly 5 µs. Spark gaps and spark energies
were adjusted by trial so that they were close to minimum ignition energies (5–20 mJ, with the larger values used for flames having relatively small laminar
burning velocities) to minimize effects of initial flame
acceleration due to excessive spark energies.
2.2. Instrumentation
The flames were observed using high-speed shadowgraph motion picture photography. The shadowgraph system consisted of a 100-W mercury short-arc
lamp (ARC; HSA-150 HP) with the light collimated
by a pair of f6 parabolic reflectors having 1220-mm
focal lengths. The flame images were recorded using
a 16-mm motion picture camera (Hycam; Model K20
34E) operating at speeds of 4000–8000 pictures per
second. Kodak Hawkeye surveillance film on a daylight loading spool (SP-430), with perforations on two
sides, was used for the photographs. The framing rate
of the camera was sensed electronically so that ignition occurred only when the proper framing rate was
reached. The framing rate and the ignition pulse were
recorded using a digital oscilloscope (LeCroy 9400A)
so that the film records could be synchronized.
2.3. Data reduction
Present measurements were limited to flames having diameters larger than 10 mm, to avoid ignition
disturbances, and smaller than 60 mm, to limit the
volume of burned gas to less than 0.5% of the total chamber volume so that the chamber pressure
remained constant within 0.7% throughout the observed period of flame propagation. Laser velocimeter
measurements for this test arrangement (but with a
slightly smaller chamber), due to Kwon et al. [31],
indicated that velocities within the unburned gas varied as expected for outwardly propagating unconfined
spherical flames for the range of flame sizes considered during the present investigation.
Similar to past measurements of laminar premixed
flame properties [32–34], determinations of flame
properties were limited to conditions where δD /rf <
2% so that effects of flame curvature and transient effects associated with the thickness of the flame were
negligible, as discussed by Tseng et al. [38]. Next,
laminar burning velocities were generally greater than
150 mm/s, so that the intrusion of effects of buoyancy due to Earth’s gravity was negligible as shown
by Ronney and Wachman [39]; this behavior was confirmed using the present flame photographs which
indicated negligible effects of flame distortion, or motion of the origin of the flame, due to buoyancy in
the period when present observations were made. Furthermore, effects of radiative heat losses were small
(less than 1% of the rate of thermal energy release
of reaction within the test flames) based on earlier
estimates of these losses for hydrogen flames at similar conditions due to Aung et al. [33] that were carried out as discussed by Siegel and Howell [40]. This
assessment also agrees with an earlier evaluation of
effects of radiation for hydrogen flames at similar
conditions due to Dixon-Lewis [41]. Finally, effects
of the spark energy were small compared to the energy release due to combustion in the region where
the flames were observed (the energy release due to
combustion was generally greater than 20 times the
spark energy for flames having diameters larger than
10 mm for present test conditions). Under these assumptions, Strehlow and Savage [26] showed that the
laminar burning velocity and flame stretch are given
by the following quasi-steady expressions:
SL = (ρb /ρu ) drf /dt,
K = (2/rf ) drf /dt.
(1)
The density ratio appearing in Eq. (1) was found
from McBride et al. [42], assuming adiabatic constant-pressure combustion with chemical equilibrium
in the combustion product gases and the same concentrations of elements in the unburned and burned gases.
This is only a convention that follows past practice
[32–34], however, because it ignores preferentialdiffusion effects that modify local element mass
fractions and energy transport and cause ρb /ρu to
differ from plane adiabatic flame conditions. This
convention is convenient, however, because a single density ratio relates all flame speeds at a given
reactant mixture condition. In addition, this convention retrieves the correct flame displacement velocity, drf /dt, for given unburned mixture conditions
and degree of flame stretch. Finally, based on past
numerical simulations of stretched hydrogen premixed flames, the present assumptions used to find
ρb /ρu are quite reasonable for the conditions considered during the present study. In particular, values
of ρb /ρu for stretched hydrogen flames agree within
10% with those for unstretched (plane) flames when
δD /rf < 2%; see Kwon [43].
Final results were obtained by averaging the measurements of four to six tests at each condition. Experimental uncertainties were estimated as described by
Tseng et al. [38] and references cited therein. The resulting experimental uncertainties (95% confidence)
are as follows: SL less than 9%, Ka less than 21%,
L. Qiao et al. / Combustion and Flame 143 (2005) 79–96
SL∞ less than 10%, and | Ma | less than 25% for
| Ma | > 1 and less than 25%/| Ma | for | Ma | < 1.
2.4. Data correlation
The measurements were analyzed to find laminar
flame properties (Markstein numbers to characterize
the sensitivity of the flame to effects of stretch and
unstretched laminar burning velocities to characterize the intensity of combustion and the capability of a
particular diluent to reduce flame intensity). As mentioned earlier, present considerations were limited to
thin flames (δD /rf < 2%) for conditions where effects of ignition disturbances, flame radiation, and
variations of element concentrations on ρb /ρu were
small. Then, a convenient relationship between the
laminar burning velocity and the flame stretch can be
obtained by combining an early proposal of Markstein
[25] and the local conditions hypothesis of Kwon et
al. [31] to yield the expression
SL∞ /SL = 1 + Ma Ka,
(2)
where values of SL and the Karlovitz number, Ka
(the dimensionless flame stretch = KδD /SL ), were
found from Eq. (1) as already discussed. For these
definitions, δD is based on the stretched laminar burning velocity and the mass diffusivity of the fuel in
the unburned (and in the present case, unsuppressed)
gas, as conventions. The decision to use the Du for
unsuppressed flames as the Du for the suppressed
flames also was made to provide an absolute evaluation of SL∞ so that the effectiveness of various
diluents could be directly compared. The small stretch
limit is also of interest to connect present results at
finite levels of stretch to the conditions of classical
asymptotic theories of laminar premixed flame propagation at negligibly small levels of stretch, as follows
(see Aung et al. [33]):
SL /SL∞ = 1 − Ma∞ Ka∞ ,
| Ka∞ | 1.
(3)
Several other proposals to represent effects of
flame stretch on laminar burning velocities have been
made; see Taylor [44], Dowdy et al. [45], Brown et
al. [46], Karpov et al. [47], and Bradley et al. [48].
The approach used in Eq. (2), however, is particularly convenient because the Markstein number has
proven to be relatively constant for particular reactant
mixture conditions over wide ranges of the Karlovitz
number based on both measurements and detailed numerical simulations of laminar premixed flames [30–
34]. Thus, SL∞ and Ma provide convenient and concise measurements of laminar premixed flame burning rates and response to stretch, as discussed by
Aung et al. [33]. See Aung et al. [33] for a discussion
of other advantages of the present characterization of
premixed-flame/stretch interactions.
83
2.5. Test conditions
Experimental conditions are summarized in Tables 1 and 2. Experimental conditions for H2 /air/diluent flames, summarized in Table 1, seek to be representative of human habitation conditions on Earth
at various altitudes, as follows: reactant mixtures at
room temperature (298 ± 0.5 K), fuel-equivalence ratios of 1.0 and 1.8, pressures of 0.5 and 1.0 atm, and
diluent concentrations of 0–40% (by volume) with
helium, argon, nitrogen, and carbon dioxide as suppression agents. The fuel-equivalence ratios of 1.0
and 1.8 were chosen due to their relevance to flame
suppression with stoichiometric conditions being repTable 1
H2 /air/diluents laminar premixed flame test conditionsa
Diluents
XD
ρu /ρb
SL∞ (mm/s)
Kamax
Ma
0.07
0.06
0.06
0.06
0.09
0.05
0.08
0.14
0.22
0.09
0.13
0.11
0.19
0.09
0.14
0.18
0.40
0.2
1.5
1.3
1.3
0.9
1.1
0.5
−0.2
−0.3
0.0
−0.3
−0.7
−1.0
0.0
−0.5
−0.7
−0.6
p = 1.0 atm, φ = 1.8, Du = 72.9 mm2 /s
–
0.0
6.30
2900
0.1
5.89
2340
N2
0.2
5.44
1830
N2
N2
0.3
4.71
1400
0.4
4.48
830
N2
CO2
0.1
5.48
2040
0.2
4.78
1330
CO2
CO2
0.3
4.18
760
0.4
3.65
240
CO2
0.04
0.03
0.04
0.05
0.08
0.05
0.05
0.09
0.12
3.5
3.2
3.1
2.8
2.6
2.7
2.1
1.4
1.2
p = 0.5 atm, φ = 1.0, Du = 145.8 mm2 /s
–
0.0
6.89
2020
0.1
6.54
1600
N2
N2
0.2
6.14
1320
0.3
5.67
920
N2
N2
0.4
5.13
640
0.1
6.20
1320
CO2
CO2
0.2
5.60
950
CO2
0.3
4.98
480
0.4
4.36
270
CO2
0.09
0.11
0.13
0.22
0.36
0.13
0.26
0.29
0.49
1.7
1.0
0.7
−0.2
−0.4
0.5
−0.2
−0.6
−0.6
p = 1.0 atm, φ = 1.0, Du = 72.9 mm2 /s
–
He
He
He
He
Ar
Ar
Ar
Ar
N2
N2
N2
N2
CO2
CO2
CO2
CO2
0.0
0.1
0.2
0.3
0.4
0.1
0.2
0.3
0.4
0.1
0.2
0.3
0.4
0.1
0.2
0.3
0.4
6.89
6.63
6.50
6.19
5.79
6.63
6.50
6.19
5.79
6.56
6.15
5.67
5.13
6.23
5.61
4.98
4.36
2140
1960
1730
1430
1170
1770
1290
1100
760
1650
1170
860
480
1330
770
400
180
a Initial mixture temperature of 298 ± 0.5 K.
84
L. Qiao et al. / Combustion and Flame 143 (2005) 79–96
Table 2
H2 /30% O2 and 70% N2 /diluents laminar premixed flame
test conditionsa
Diluents
XD
ρu /ρb
SL∞ (mm/s)
p = 0.7, φ = 1.0, Du = 104.1 mm2 /s
–
0.0
7.45
3350
He
0.2
7.21
2700
He
0.4
6.67
1880
Ar
0.2
7.21
2320
Ar
0.4
6.67
1380
0.2
6.92
2100
N2
N2
0.4
5.97
1150
0.2
6.44
1570
CO2
CO2
0.4
5.06
480
Kamax
Ma
0.04
0.04
0.04
0.06
0.09
0.08
0.10
0.07
0.21
2.9
2.4
2.1
2.2
0.6
0.6
−0.3
0.7
−0.3
a Initial mixture temperature of 298 ± 0.5 K.
resentative of flame attachment conditions in nonpremixed flames, whereas φ = 1.8 represents conditions where the unstretched laminar burning velocities
of hydrogen/air mixtures reach a maximum [32–34]
and conditions that are most difficult to extinguish
for premixed flames of these reactants. The values
of ρu /ρb in Table 1 were found from McBride et
al. [42], as discussed earlier. These measurements involved unstretched laminar burning velocities of 180–
2900 mm/s, Karlovitz numbers of 0–0.5, and Markstein numbers of −1.0–3.5.
Experimental conditions for H2 /30% O2 and 70%
N2 (by volume)/diluent flames, summarized in Table 2, involve EVA-preparation conditions for spacecraft, as follows: reactant mixtures at room temperature (298 ± 0.5 K) and a pressure of 0.7 atm, a
fuel-equivalence ratio of unity, and diluent concentrations of 0–40% (by volume) with helium, argon, nitrogen, and carbon dioxide as suppression agents. These
measurements involved unstretched laminar burning
velocities of 480–3350 mm/s, Karlovitz numbers of
0–0.21, and Markstein numbers of −0.3–2.9.
3. Computational methods
3.1. Numerical simulations
Computational methods for the present flames
were similar to those of Aung et al. [32,33] and Kwon
and Faeth [34]. The outwardly propagating spherical laminar premixed flames were simulated using
the unsteady one-dimensional laminar flame computer code, RUN-1DL, developed by Rogg [49]. This
algorithm allows for mixture-averaged multicomponent diffusion, thermal diffusion, variable thermochemical properties, and variable transport properties.
The CHEMKIN package [50–53] was used as a preprocessor to find the thermochemical and transport
properties for RUN-1DL. Transport properties were
found from the transport property database of Kee et
al. [51]. Thermochemical properties were found from
the thermodynamic data base of Kee et al. [50], except for HO2 , where the recommendations of Kim et
al. [54] were used. Before computing flame properties, all transport and thermodynamic properties were
checked against original sources. Similar to the measurements, effects of radiation were small due to the
relatively large flame velocities of hydrogen flames
for present conditions and were ignored. Flame propagation was allowed to proceed sufficiently far so that
effects of initial conditions were small, similar to the
measurements. Other limitations used to control experimental uncertainties, e.g., δD /rf < 2%, etc., were
also applied to the predictions. The computational
grid in space and time was varied to ensure numerical
accuracy within 1%, estimated by Richardson extrapolation of SL . Finally, the numerical simulations
were analyzed similar to the measurements, taking
the flame position to be the point where gas temperatures were the average of the temperatures of the
hot and cold boundaries. Due to the present stringent
flame thickness limitations, however, the present results were not affected significantly by the criterion
used to define the flame position.
Separate numerical simulations were carried out
for unstretched (plane) flames using the steady onedimensional laminar premixed flame code, PREMIX,
due to Kee et al. [53]. Other properties of these calculations and the limits of numerical accuracy were
similar to those using the RUN-1DL algorithm. This
code was mainly used to predict the structure of unstretched flames.
3.2. Chemical kinetic mechanism
Aung et al. [32,33], Kwon and Faeth [34], and
Kim et al. [19] carried out extensive evaluations of
available detailed hydrogen/oxygen chemical kinetic
mechanisms proposed by Kim et al. [54], Yetter et al.
[55], Mueller et al. [56], Marinov et al. [57], Wang
and Rogg [58], and Frenklach et al. [59,60] based
on their measurements of the properties of hydrogen
outwardly propagating laminar premixed flames. The
chemical kinetic mechanism of Mueller et al. [56]
was found to provide the best comparison between
measurements and predictions for hydrogen flames
involving nitrogen, argon, and helium as suppression
agents, fuel-equivalence ratios of 0.6–4.5, pressures
of 0.3–3.0 atm, and volumetric oxygen concentrations
in the nonfuel gases of 0.21–0.36. These conditions
were generally similar to present flame conditions;
therefore, the numerical simulations of flames reported here were limited to the Mueller et al. [56] hydrogen/oxygen chemical kinetic mechanism. Similar
to the earlier evaluation of Mueller et al. [56] chemi-
L. Qiao et al. / Combustion and Flame 143 (2005) 79–96
cal kinetic mechanism by Kwon and Faeth [34], this
mechanism was simplified because C/H/O and N/O
chemistry were not important for present conditions
and could be deleted from the mechanism. The final reduced chemical kinetic mechanism involved 11
chemical species and 19 reversible reactions. This reaction mechanism does not provide information about
the elementary reactions involving helium and their
reaction coefficients; therefore, the same reactions
and their reaction coefficients as those used for argonsuppressed hydrogen flames were chosen for the simulations of the helium-suppressed flames due to the
thermodynamic similarities of argon and helium.
4. Results and discussion
4.1. Flame stability and evolution
Three kinds of flame surface instabilities were
observed during present experiments: preferentialdiffusion instability (observed only when Ma < 0),
hydrodynamic instability (observed for all values
of Ma), and buoyant instabilities (observed only when
laminar flame speeds or corresponding laminar burning velocities were small). Shadowgraph photographs
of flame surfaces after distortion by these instabilities
for outwardly propagating spherical flames appear in
Kim et al. [19] and Kwon et al. [31]. The presence of
preferential-diffusion instability could be identified
by irregular (chaotic) distortions of the flame surface
relatively early in the flame propagation process and
as noted earlier only when Ma < 0. Fortunately, flame
surfaces remained smooth at small flame radii even
for conditions that involved preferential-diffusion instability so that laminar burning velocities could be
measured for a time even at these conditions. Hydrodynamic instability could be identified by the development of a somewhat regular cellular disturbance
pattern on the flame surface, very similar to the observations of Groff [61]; fortunately, these instabilities
were observed only for flame diameters larger than
60 mm so that they did not affect the present measurements limited to flame diameters smaller than
50 mm. Finally, buoyant instabilities were observed
when laminar burning velocities were small and were
readily detected by distortion of the flame surface
from a spherical shape when vertical planes of the
flame were observed (which was the case for the
present experimental arrangement). In addition, the
flame boundary was also deflected upward from the
location of the spark kernel when effects of buoyant
instability were important. As noted earlier, however,
effects of buoyancy were small as long as laminar
burning velocities were larger than 150 mm/s, which
involved velocities well below the present test range.
85
Finally, no measurements reported here were made at
conditions where any of these instabilities were observed.
4.2. Burning velocity/stretch interactions
Measurements at finite flame radii involve finite
values of flame stretch through Eq. (1); therefore,
the laminar burning velocity at the largest flame radius observed still differs from the fundamental unstretched laminar burning velocity of a plane flame,
SL∞ . Thus, values of SL∞ were found from Eq. (2)
/S , where S by plotting SL∞
L
L∞ is the value of the
laminar burning velocity at the largest flame radius
observed, as a function of Ka, similar to past work
[32–34]. As will be seen subsequently, this yielded
linear plots so that extrapolation to Ka = 0 yielded
/S
SL∞
L∞ and thus SL∞ as summarized in Tables 1
and 2. Given SL∞ , plots of SL∞ /SL as a function of
Ka could be constructed for various reactant mixtures
and pressures, as prescribed by Eq. (2). Examples
of plots of this type for hydrogen flames at various
conditions, based on both measurements and predictions, are illustrated in Figs. 1–4; results at other test
conditions were qualitatively similar to the results illustrated in Figs. 1–4. Finally, it is evident that the
variations of SL∞ /SL as a function of Ka are all linear for the results illustrated in Figs. 1–4. This implies
that Ma is a constant that can be determined from the
constant slopes of the plots of SL∞ /SL as a function
of Ka for each flame condition that was studied. These
values of the Markstein number are also summarized
Fig. 1. Measured and predicted laminar burning velocities
as functions of Karlovitz number and the concentration of
nitrogen diluent for premixed stoichiometric hydrogen/air
flames at NTP.
86
L. Qiao et al. / Combustion and Flame 143 (2005) 79–96
Fig. 2. Measured and predicted laminar burning velocities
as functions of Karlovitz number and the concentration of
carbon dioxide diluent for premixed stoichiometric hydrogen/air flames at NTP.
Fig. 4. Measured and predicted laminar burning velocities as
functions of Karlovitz number and the concentration of carbon dioxide diluent for premixed stoichiometric hydrogen
flames at room temperature and spacecraft EVA-preparation
conditions.
Table 3
H2 /air laminar premixed flame property measurementsa
Source
ρu /ρb SL∞ (mm/s) Kamax Ma
Present study
6.30
Kwon and Faeth [34] 6.30
Aung et al. [32,33]
6.30
2900
2860
2610
0.04
0.06
0.08
3.5
2.4
3.7
a Unsuppressed flames having φ = 1.80 at an initial mixture pressure and temperature of 1 atm and 298 ± 3 K;
Du = 72.9 mm2 /s.
Fig. 3. Measured and predicted laminar burning velocities
as functions of Karlovitz number and the concentration of
nitrogen diluent for premixed stoichiometric hydrogen/air
flames at room temperature and spacecraft EVA-preparation
conditions.
in Tables 1 and 2 for all conditions tested during the
present investigation.
Given the preceding description of the way that the
unstretched laminar burning velocities and Markstein
numbers of outwardly propagating laminar premixed
flames were found, it is of interest to compare present
measurements with earlier results. This comparison
could be carried out only for premixed H2 /air flames
at a fuel-equivalence ratio of 1.8 with the reactants at
room temperature and pressure (NTP or 298 ± 3 K
and 1 atm) where present test conditions and those of
Kwon and Faeth [34] and Aung et al. [32,33] overlap. The results of all three studies are summarized in
Table 3. Values of Kamax differ for the three studies
but this occurs due to the somewhat arbitrary selection of the range of flame radii to be used to find SL∞
and Ma. On the other hand, the fundamental measured
quantities, SL∞ and Ma, are seen to agree among the
three studies within the ranges of experimental uncertainties (95% confidence) that were specified earlier.
This behavior was typical of other comparisons of
present and earlier results that could be made, as will
be seen subsequently.
The plots of laminar burning velocity as a function of stretch illustrated in Figs. 1–4 involve results
for hydrogen/air flames at NPT for a fuel-equiv-
L. Qiao et al. / Combustion and Flame 143 (2005) 79–96
alence ratio of unity with nitrogen and carbon dioxide as suppression agents (Figs. 1 and 2) and for
hydrogen/EVA-preparation atmospheres for a fuelequivalence ratio of unity with nitrogen and carbon dioxide as suppression agents (Figs. 3 and 4).
Measurements on these plots are indicated by open
symbols for stable preferential-differential conditions (Ma 0) and closed symbols for unstable
preferential-diffusion conditions (Ma < 0). For unstable conditions, the measurements are limited to
values of Ka significantly greater than zero because
flame surfaces became wrinkled due to preferentialdiffusion instability for radius values within the normal range of measurements.
The first notable observation from Figs. 1–4 is
that effects of flame/stretch interactions are important
for present test conditions; for example, over all the
present results, SL∞ /SL varied in the range 0.6–1.25
for Ka < 0.49, which does not involve a close approach to quenching conditions (which would involve
Ka ≈ 1 from Law [28]) where effects of Ka on SL
are expected to be large. Next, the linear relationship
between SL∞ /SL and Ka clearly is satisfied for all
measurements and predictions illustrated in Figs. 1–4
which correspondingly implies constant Markstein
numbers for each flame condition, providing a convenient and concise way to summarize flame/stretch
interactions for the present flames. Notably, this behavior has been observed for all outwardly propagating flame conditions studied thus far (see [19,20,
30–34] and references cited therein). Furthermore, the
progressive addition of diluents to the flames illustrated in Figs. 1–4 causes the slopes of the plots of
SL∞ /SL as a function of Ka to become progressively
more negative, with the exception of a few conditions
having large concentrations of diluent. In a number of
cases, this implies that stable flames to preferentialdiffusion/stretch interactions at small concentrations
of diluent become unstable flames to preferentialdiffusion/stretch interactions at large concentrations
of diluent. Due to increased flame speeds as a result of
increased flame surface area caused by wrinkling, this
behavior clearly tends to counteract the ability of diluents to reduce combustion rates by reducing laminar
burning velocities and tends to reduce the effectiveness of diluents to some extent.
Finally, the qualitative agreement between measured and predicted burning velocity/stretch interactions, using the hydrogen/oxygen chemical kinetic
mechanism of Mueller et al. [56], is reasonably good.
This is particularly promising because the measurements used to develop the hydrogen/oxygen chemical
kinetic mechanism of Mueller et al. [56] did not involve any direct consideration of flame/stretch interactions. This evaluation of predictions will continue
during subsequent consideration of Markstein num-
87
bers and unstretched laminar velocities, which provide more direct and complete comparisons of measurements and predictions than is possible for the results illustrated in Figs. 1–4.
4.3. Markstein numbers
Markstein numbers are independent of Karlovitz
numbers for present conditions and are summarized
in Tables 1 and 2 as a function of reactant conditions.
A portion of these results, involving measured and
predicted Markstein numbers as a function of diluent
concentrations, are plotted in Figs. 5–7.
Measurements and predictions of Markstein numbers as a function of diluent concentrations for H2 /air
flames having φ = 1 at NTP are illustrated in Fig. 5,
considering helium, argon, nitrogen, and carbon dioxide as suppression agents. With the exception of helium, values of the Markstein number generally become progressively more negative as the concentration of diluent increases, with some tendency for
this decrease to become small at large concentrations of the more effective nitrogen and carbon dioxide suppression agents. In these cases, preferentialdiffusion instability is promoted as the flames become
more suppressed. Results for helium as a suppression
agent differ from this behavior, however, because the
large thermal conductivity of the fast-diffusing helium molecules tends to promote preferential quenching of the reaction zone and thus stability of the
flames to preferential-diffusion/stretch interactions.
Finally, the qualitative and quantitative agreement between predictions and measurements is reasonably
good in Fig. 5, providing potential for the predictions
Fig. 5. Measured and predicted Markstein numbers as functions of the concentration of helium, argon, nitrogen, and
carbon dioxide diluents for premixed stoichiometric hydrogen/air flames.
88
L. Qiao et al. / Combustion and Flame 143 (2005) 79–96
Fig. 6. Measured and predicted Markstein numbers as functions of the concentration of nitrogen and carbon dioxide diluents for premixed hydrogen/air flames at a fuelequivalence ratio of 1.8 and NTP.
to help explain the complex effects of preferentialdiffusion/stretch interactions that influence the behavior of suppressed laminar premixed flames.
Measurements and predictions of Markstein numbers as a function of diluent concentrations for H2 /air
flames having φ = 1.8 at NTP are illustrated in Fig. 6,
considering nitrogen and carbon dioxide as suppression agents. In this case, the experiments involve fuelrich conditions where H2 /air flames are intrinsically
stable based on classical models of flame instability due to effects of preferential diffusion proposed
by Manton et al. [24] and Markstein [25], namely,
that laminar premixed flames are unstable to effects
of preferential diffusion at conditions where the fastdiffusing component (H2 in the present instance) is
deficient (at fuel-lean conditions in the present instance). The subsequent effect of adding nitrogen and
carbon dioxide as suppression agents, however, is
similar to results at other conditions where the addition of a diluent tends to shift the Markstein number
toward more negative (unstable) values. Finally, the
comparison between measurements and predictions
in Fig. 6 is excellent.
Measurements and predictions of Markstein numbers as a function of diluent concentrations for
H2 /EVA-preparation conditions for φ = 1 and room
temperature are illustrated in Fig. 7, considering helium, argon, nitrogen, and carbon dioxide as suppression agents. These results are qualitatively similar
to the results for H2 /air mixtures at NTP illustrated
in Fig. 5: the addition of diluents generally causes
Markstein number to decrease, helium as a suppression agent differs from the rest due to its capability to
quench the reaction zone at stretched conditions as a
result of the fast-diffusion and high heat transfer rates
Fig. 7. Measured and predicted Markstein numbers as functions of the concentration of helium, argon, nitrogen, and
carbon dioxide diluents for premixed stoichiometric hydrogen flames at room temperature and spacecraft EVApreparation conditions.
of helium, and the agreement between measured and
predicted values of the Markstein numbers is excellent.
4.4. Unstretched laminar burning velocities
In the following, measured values of laminar burning velocities will be limited to stretch-corrected results that yield unstretched laminar burning velocities.
The measured values of unstretched laminar burning
velocities are summarized in Tables 1 and 2. Plots of
measured and predicted values of unstretched laminar burning velocities as functions of the concentrations of various diluents for some typical reactant
conditions appear in Figs. 8–11 (the second independent variable on these figures, oxygen index, will be
discussed later). These results are plotted for H2 /air
flames having φ = 1 at NTP in Fig. 8, considering helium, argon, nitrogen, and carbon dioxide as suppression agents. In addition to the present measurements
and predictions, the measurements of Kim et al. [19]
for nitrogen and carbon dioxide as suppression agents
at these conditions are shown on the plot. Notably, the
agreement between present measurements and those
of Kim et al. [19] and the agreement between present
predictions and measurements are seen to be excellent. These results indicate that all diluents cause the
unstretched laminar burning velocities to decrease as
the concentrations of diluent are increased and that
the suppression effectiveness of diluents (taken as the
reduction of the unstretched laminar burning velocity for a particular diluent concentration) increases in
the order helium, argon, nitrogen, and carbon dioxide. Both these behaviors can be explained for argon,
L. Qiao et al. / Combustion and Flame 143 (2005) 79–96
Fig. 8. Measured and predicted unstretched (Ka = 0) laminar burning velocities as functions of the concentrations of
helium, argon, nitrogen, and carbon dioxide diluents for premixed stoichiometric hydrogen/air at NTP.
Fig. 9. Measured and predicted unstretched (Ka = 0) laminar burning velocities as functions of the concentrations of
nitrogen and carbon dioxide diluents for premixed hydrogen/air flames at a fuel-equivalence ratio of 1.8 and NTP.
nitrogen, and carbon dioxide according to Huggett
[22,23] as a result of the increase of the specific
heat of the nonfuel gases per unit oxygen concentration. This causes a corresponding reduction of temperatures within the reaction zone of the flames with
the associated reduction of laminar burning velocities
following in accord with classical phenomenological
theories of premixed laminar flame propagation; see
Law [28]. Helium as a suppression agent is a predictable exception to this behavior; its specific heat
effect is identical to that of argon as a suppression
agent but this effect is counteracted by its increased
heat and mass transfer rate capabilities that tend to increase unstretched laminar burning velocities for the
helium-containing flames to some extent, compared
89
Fig. 10. Measured and predicted unstretched (Ka = 0) laminar burning velocities as functions of the concentrations
of nitrogen and carbon dioxide diluents for stoichiometric
premixed hydrogen/air flames at room temperature and pressures of 0.5 and 1.0 atm.
Fig. 11. Measured and predicted unstretched (Ka = 0) laminar burning velocities as functions of the concentrations of
helium, argon, nitrogen, and carbon dioxide diluents for premixed hydrogen flames at room temperature and spacecraft
EVA-preparation conditions.
to argon-containing flames, based on classical phenomenological theories of premixed flames.
The third-body reaction H + O2 + M = HO2 + M
is important as a chain-terminating reaction. It competes with the branching reaction H + O2 = OH + O
at temperatures less than ∼900 K [63]. Therefore, for
weakly propagating flames where the temperature is
low, this third-body reaction has a dominant effect in
the H2 –O2 chemistry. It is also found that the thirdbody reactions, especially H + OH + M = H2 O +
M, are important for laminar flame speed propagation
only when the pressure is high [64]. Pressures considered here, however, are low, 0.5, 0.7, and 1 atm, and
90
L. Qiao et al. / Combustion and Flame 143 (2005) 79–96
the flame, even with 40% CO2 as diluent, is still far
away from the flammability limit. Therefore, in the
present investigation, the third-body reaction efficiencies are not as important as the impact of heat capacity
and transport properties of the diluents.
Measured and predicted unstretched laminar burning velocities are plotted as a function of diluent concentration for H2 /air flames having φ = 1.8 at NTP
in Fig. 9, considering nitrogen and carbon dioxide as
suppression agents. In general, the unstretched laminar burning velocities at φ = 1.8 in Fig. 9 are larger
than those at φ = 1.0 in Fig. 8, which is well known
behavior because the laminar burning velocities of
H2 /air mixtures at NTP reach a maximum at φ = 1.8
[32–34]. Other trends in Fig. 9 are similar to those
in Fig. 8: the comparison between measurements and
predictions is excellent, and diluents progressively reduce unstretched laminar burning velocities as their
concentrations increase with suppression effectiveness additionally increasing in the order nitrogen and
carbon dioxide.
Measured and predicted unstretched laminar burning velocities are plotted as a function of diluent
concentration for H2 /air flames having φ = 1.0 at
room temperature and pressures of 0.5 and 1.0 atm
in Fig. 10, considering nitrogen and carbon dioxide
as suppression agents. Similar results are found in
Figs. 8 and 9. Finally, the effect of pressure on laminar burning velocities is relatively small for the results
illustrated in Fig. 10, which is well-known behavior
for H2 /air flames at room temperature and pressures
in the range 0.35–1.0 atm; see Aung et al. [33]. Even
though SL∞ is not strongly affected by pressure for
the range of conditions illustrated in Fig. 10, however, it should be recalled that rates of chemical energy release for a particular laminar burning velocity
and reactant temperature are directly proportional to
the concentration of the reactant mixture and thus the
pressure. As a result, fire severity decreases with decreasing pressure. In addition, reduced pressures tend
to increase suppression agent concentrations in the reactant mixture for a given mass release of suppression
agents, helping to promote extinction. Finally, chemical kinetic considerations near flammability limits
also point toward improved capabilities to extinguish
premixed flames at reduced pressures, although this
property was not studied during the present investigation.
Measured and predicted unstretched laminar burning velocities are plotted as a function of diluent concentration for premixed hydrogen flames at room temperature and EVA-preparation conditions in Fig. 11.
In this case, all four diluents—helium, argon, nitrogen, and carbon dioxide—have been compared as
suppression agents. The suppression effectiveness increases in the order helium, argon, nitrogen, and car-
bon dioxide for the reasons that have already been
discussed in connection with Fig. 8. The enriched
oxygen concentration of EVA-preparation conditions
compared to a conventional air environment (oxygen
concentrations of 30% by volume compared to 21%
by volume) causes the unstretched and unsuppressed
laminar burning velocity in the EVA-preparation atmosphere to be larger than that in air at NTP, e.g.,
3350 mm/s compared to 2140 mm/s. On the other
hand, the reduced pressure of the EVA-preparation atmosphere reduces the mass burning rate; therefore,
the fire intensity is only roughly 10% larger in the
EVA-preparation atmosphere than the mass burning
rate in air at NTP. The reduced pressure of the EVApreparation environment also tends to increase the
effectiveness of particular mass discharges of diluents compared to systems operated at NTP. Thus,
present results do not clearly establish potential advantages for either normal or EVA-preparation environments with respect to fire suppression performance.
The oxygen index, which is the concentration of
oxygen (in %) by volume in the nonfuel gases, is a
single-valued function of diluent concentration for the
conditions of Figs. 8–11 and is shown as an independent variable on these figures. For present hydrogen
flames, oxygen indices reach values as small as 10
with no sign of approach to extinction conditions,
whereas hydrocarbon flames typically reach flammability limits for oxygen indices of 12–15. For example, Westbrook [62] has proposed an approximation
to find conditions where laminar premixed flames extinguish at a laminar burning velocity of 50 mm/s; in
contrast, present flames at oxygen indices of 10 exhibit unstretched laminar burning velocities greater
than 180 mm/s and are still well away from the extinction conditions suggested by the Westbrook [62]
criterion. This behavior is not unexpected, however,
due to the well-known difficulties of suppressing hydrogen flames; see Lewis and von Elbe [36].
The use of EVA-preparation atmospheres definitely increases oxygen indices at a particular diluent
concentration, as illustrated in Fig. 11, which is reflected by the increased laminar burning velocities in
EVA-preparation atmospheres compared to those in
air at NTP. On the other hand, the reduced pressure
of the EVA-preparation atmosphere reduces the measure of the fire hazard, taken as the degree of hazard,
by a factor of 1/P 1/2 according to Huggett [22]. Further study is required, however, to establish whether
suppression is more difficult in EVA-preparation atmospheres at room temperatures due to the simultaneous increase of the oxygen concentration and decrease
of the pressure.
L. Qiao et al. / Combustion and Flame 143 (2005) 79–96
Fig. 12. Predicted structure of an unstretched (Ka = 0) premixed stoichiometric hydrogen/air flame with no diluent
present at NTP.
91
Fig. 13. Predicted structure of an unstretched (Ka = 0) premixed stoichiometric hydrogen/air flame with a 40% by volume helium diluent at NTP.
4.5. Flame structure
As mentioned earlier, measurements and predictions of unstretched laminar burning velocities were
in reasonably good agreement, including effects of
variations of fuel-equivalence ratio, pressure, ambient oxygen concentration, and presence of diluents;
therefore, the predictions were exploited to gain a better understanding of the effects of diluents on laminar
burning velocities. The approach involved numerical
simulations of plane (unstretched) H2 /air flames in
the presence of various diluents.
Typical predicted structures of plane unstretched
H2 /air flames at a fuel-equivalence ratio of unity and
NTP are illustrated in Figs. 12–16. Results in Fig. 12
provide the baseline flame structure when no diluent
is present. Figs. 13–16 provide similar results for diluent concentrations of 40% (by volume) for helium,
argon, nitrogen, and carbon dioxide, in turn. All these
results are based on the hydrogen/oxygen chemical
kinetics mechanism of Mueller et al. [56]. In each
figure, the top graph provides profiles of the temperature and the stable species (H2 , O2 , H2 O) concentrations, whereas the bottom graph provides profiles
of radical species (H, OH, O, HO2 , and H2 O2 ) concentrations, all as functions of distance through the
flame. It should be noted that the origins of the length
scales in these figures are arbitrary and do not cor-
Fig. 14. Predicted structure of an unstretched (Ka = 0) premixed stoichiometric hydrogen/air flame with a 40% by volume argon diluent at NTP.
92
L. Qiao et al. / Combustion and Flame 143 (2005) 79–96
Fig. 15. Predicted structure of an unstretched (Ka = 0) premixed stoichiometric hydrogen/air flame with a 40% by volume nitrogen diluent at NTP.
Fig. 16. Predicted structure of an unstretched (Ka = 0) premixed stoichiometric hydrogen/air flame with a 40% by volume carbon dioxide diluent at NTP.
respond to the central ignition point. In addition, the
scales of the concentrations of radical species coordinates become more expanded going from Fig. 12
to Fig. 16 so that the plots remain readable as maximum radical concentrations decrease with increasing
suppression agent effectiveness. The results show that
the maximum concentrations of the radicals HO2 and
H2 O2 are roughly two orders of magnitude smaller
than the concentrations of the radicals H, OH, and O;
therefore, the latter tend to dominate reactive effects in the present flames. Comparing the stable
species concentrations for a flame not having a diluent present (Fig. 12) to those in flames having a
diluent present (Figs. 13–16), indicates expected reductions of the reactant concentrations (H2 and O2 )
and product concentrations (H2 O) due to the dilution
caused by the diluents having initial concentrations
of 40% (by volume). Another effect that is evident
is the preferential diffusion of the fast-diffusing reactant, H2 , compared to the slow-diffusing reactant,
O2 , for a plane flame; this can be seen from the increase of the concentration of O2 near the cold boundary of the flame before the concentration of O2 decreases once again upon approach to the reaction zone
of the flame. The next major trend is the progressive reduction of the final flame temperature from
2250 K (for no diluents), to 1750 K (for the diluents He and Ar), to 1600 K (for the diatomic diluent
N2 ), and finally to 1350 K (for the triatomic diluent CO2 ). This behavior is solely due to the progressive increase of the specific heat of these diluents in
the order He and Ar (the same), N2 , and CO2 . On
the other hand, the increased thermal diffusivity of
He compared to Ar has no effect on the final flame
temperature because these flames are all adiabatic.
For the present stoichiometric flames, the radical H
generally has the largest maximum concentrations in
the flames, with OH having somewhat smaller maximum concentrations, e.g., roughly 1/4–1/3 as large
as H, and with the other radicals all having significantly smaller concentrations. In addition, the maximum concentration of H in the flames progressively
decreases in the order no diluent, helium, argon, nitrogen, and carbon dioxide as suppression agents.
Similarly, but not shown here, the maximum concentration of H in the flame for a particular reactant
mixture progressively decreases with increasing concentrations of suppression agents. Based on the findings of Kwon and Faeth [34], for flames having hydrogen and oxygen as reactants, it is expected that
this reduction of the maximum concentration of H
should cause a corresponding reduction of the laminar burning velocity of these flames. The potential
for this behavior will be considered in the next section.
L. Qiao et al. / Combustion and Flame 143 (2005) 79–96
93
4.6. Radical behavior
The flame structure results of Figs. 12–16, and
similar results presented earlier by Kim et al. [19]
and Kwon and Faeth [34], indicate that the H radical has the highest concentrations of all the radicals
in premixed hydrogen and oxygen flames. In addition, OH radical concentrations are the next largest
concentration after H for φ 1, ranging up to same
order of magnitude of maximum concentration of H
and OH for φ as small as 0.6. This behavior is significant due to the strong correlation between the SL and
the maximum concentration of H in the reaction zone
of premixed hydrogen and oxygen flames observed by
Padley and Sugden [35] and Butler and Hayhurst [37].
Based on the flame structure results just discussed in
connection with Figs. 12–16, it appeared likely that
a similar correlation would also be observed for the
present suppressed flames; this possibility is considered next.
Similar to past work, the most robust correlation between laminar burning velocities and radical
concentrations for laminar premixed hydrogen and
oxygen flames was obtained by using the maximum
H + OH mole fraction in the flames, which generally
was obtained at the condition where the mole fraction of H was a maximum. The resulting correlation
is illustrated in Fig. 17, where the value of SL∞ is
plotted as a function of the maximum H + OH mole
fraction, computed as just described. All the results
illustrated in Fig. 17 are for premixed hydrogen and
oxygen flames at room temperature and include findings from Kwon and Faeth [34], Kim et al. [19], and
the present investigation. The ranges of experimental
conditions for these investigations are summarized in
Table 4. The original correlation of Padley and Sugden [35] along these lines is not included in Fig. 17
because their plotted results were limited to SL as
a function of the maximum H mole fraction in the
flames. Another difficulty about the measurements of
Padley and Sugden [35] is that the extent of flame
stretch is unknown for these results.
Clearly, there is a rough correlation between the
SL∞ and the maximum H + OH mole fraction in
Fig. 17. Laminar burning velocities as functions of the maximum H + OH mole fraction in the reaction zone for hydrogen flames having various concentrations of diluents at
room temperature; see Table 4 for the range of experimental
conditions.
the flames for the results illustrated in Fig. 17. Notably, SL∞ and radical mole fractions were varied
in a number of ways for these results: dilution by
various concentrations of chemically passive suppression agents (He, Ar, N2 , and CO2 ), dilution by various concentrations of a chemically active suppression agent (CF3 Br), variation of fuel-equivalence ratios, variation of the concentration of oxygen in the
nonfuel gases, modest variations of pressure (0.7–
1.0 atm), and variation of the degree of flame stretch;
see Table 4 for a complete summary of the ranges of
the experimental flame conditions. Finally, the degree
of scatter of the correlation of SL∞ as a function of
the maximum H + OH mole fraction in Fig. 17 is
particularly small for the recent studies of unstretched
suppressed laminar premixed flames due to Kim et al.
[19] and the present investigation. These results provide a best-fit correlation between the SL∞ and the
maximum H + OH mole fraction, which is shown on
the plot, as
SL∞ (mm/s) = 260 + 36,600 (XH + XOH )max ,
(4)
Table 4
Test conditions for hydrogen premixed flame studiesa
Source
Diluentsb
φ
O2 /(O2 + N2 )
(% vol.)
P (atm)
Kac (–)
XD d (% vol.)
Kwon and Faeth [34]
Kim et al. [19]
Present investigation
–
N2 , CO2 , CF3 Br
He, Ar, N2 , O2
0.6–4.5
0.6–1.8
1.0 and 1.8
21–36
21
21 and 30
1.0
1.0
0.7–1.0
0.0–0.50
0.0
0.0
0
0–2
0–40
a
b
c
d
Initial mixture temperature 298 ± 5 K.
He, Ar, N2 , and CO2 are chemically passive suppression agents, whereas CF3 Br is a chemically active suppression agent.
Ka = 0 denotes unstretched flames.
XD = 0 denotes unsuppressed flames.
94
L. Qiao et al. / Combustion and Flame 143 (2005) 79–96
where this expression considers only the unstretched
(plane) laminar premixed flames from Kim et al. [19]
and the present investigation.
5. Conclusions
The effects of flame stretch and the concentrations and types of diluents on the laminar burning
velocities of hydrogen premixed flames were studied both experimentally and computationally. The
experiments involved unsteady outwardly propagating laminar premixed spherical flames and steady
plane laminar premixed flames, similar to past work
in this laboratory, e.g., Tseng et al. [38], Aung et
al. [32,33], and Kwon and Faeth [34]. Experimental
and computational conditions considered premixed
hydrogen/air/diluent and hydrogen/30% oxygen and
70% nitrogen (by volume)/diluent flames, with the
latter condition of interest for EVA-preparation activities onboard manned spacecraft. The other flame conditions were as follows: room temperature (298 K);
fuel-equivalence ratios of 1.0 and 1.8; pressures of
0.5, 0.7, and 1.0 atm; chemically passive gaseous suppression agents (diluents) including helium, argon,
nitrogen, and carbon dioxide; and diluent concentrations of 0–40% (by volume), which is equivalent
to oxygen indices of 30–10 for present flame conditions. Predicted flame behavior considered variable
transport and thermodynamics properties, multicomponent transport, and the detailed hydrogen/oxygen
chemical kinetic mechanism of Mueller et al. [56].
The major conclusions of the study are as follows.
(1) Effects of flame/stretch interactions for both
measurements and predictions of suppressed flames
could be correlated based on the local-conditions hypothesis according to SL∞ /SL = 1 + Ma Ka to obtain a linear relationship between SL∞ /SL and the
Karlovitz number. This behavior implies a constant
Markstein number for given reactant conditions, similar to earlier findings for unsuppressed flames.
(2) Effects of flame stretch on laminar burning velocities were substantial, yielding values of SL∞ /SL
in the range 0.60–1.25, for Ka < 0.5, which does
not involve a close approach to quenching conditions
where Karlovitz numbers typically have values on the
order of unity [28]; corresponding Markstein numbers
were in the range −1.0 to 3.5.
(3) Measured and predicted unstretched laminar
burning velocities and Markstein numbers were in
reasonably good agreement using the hydrogen/oxygen chemical kinetic mechanism of Mueller et al. [56].
(4) The chemically passive suppression agents–
diluents increase in effectiveness (based on reduction of the unstretched laminar burning velocity for
a given concentration of diluent (in % by volume)) in
the order helium, argon, nitrogen, and carbon dioxide
which mainly reflects their progressively increasing
specific heats and progressively decreasing mass and
thermal transport properties.
(5) Predictions showed that the unstretched laminar burning velocities of the present flames were
strongly correlated with the maximum H + OH mole
fraction in the reaction zone for variations of these
concentrations due to effects of chemically passive
suppression agents, similar to an early proposal of
Padley and Sugden [35] for unsuppressed and unstretched hydrogen/air flames, the recent observations
of Kwon and Faeth [34] for a variety of unsuppressed
and stretched and unstretched flames involving H2
and O2 as reactants, and the recent observations of
Kim et al. [19] for a variety of stretched and unstretched flames involving H2 and O2 as reactants
that were subjected to a chemically active suppression
agents (Halon 1301).
(6) Finally, there is a consistent tendency for
the addition of suppression agents, either chemically
active or chemically passive, to reduce the Markstein number for a given reactant mixture at the
same time that the unstretched laminar burning velocity is reduced, causing unsuppressed flames that
are stable to effects of preferential-diffusion/stretch
interactions (positive Markstein numbers) to become suppressed flames that are unstable to effects
of preferential-diffusion/stretch interactions (negative Markstein numbers) in some instances. Thus,
the tendency of suppression agents to reduce laminar burning velocities (and thus act to reduce the
severity of unwanted fires) is counteracted to some
extent by the tendency of suppression agents to reduce Markstein numbers (and promote flame instabilities that tend to increase the severity of unwanted
fires).
Acknowledgments
This research was sponsored by NASA Grants
NCC3-661, NAG3-1878, NAG3-2040, and NAG32404 under the technical management of F. Takahashi
of the NASA Glenn Research Center. The authors
thank Dr. Elaine Oran at NRL for her help and encouragement in the revising process of this paper.
References
[1] D. Drysdale, An Introduction to Fire Dynamics, Wiley,
New York, 1985, pp. 24 and 222.
[2] D. Tuhtar, Fire and Explosion Protection: A System
Approach, Wiley, New York, 1989, p. 127.
L. Qiao et al. / Combustion and Flame 143 (2005) 79–96
[3] Montreal Protocol on Substances That Deplete the
Ozone Layer, Final Act, United Nations Environmental Program (UNEP), Montreal, Canada, 1981.
[4] H.Y. Safieh, J. van Dooren, P.J. van Tiggelen, Proc.
Combust. Inst. 19 (1982) 117.
[5] R.S. Sheinsohn, J.E. Penner-Hahn, D. Indritz, Fire Saf.
J. 15 (1989) 437.
[6] B. Walravens, F. Batten-LeClerc, G.M. Gome, F.
Baronnet, Combust. Flame 103 (1995) 339.
[7] A.W. Miziolek, W. Tsang, Halon Replacement: Technology and Science, American Chemical Society,
Washington, DC, 1995, p. 322.
[8] A. McIlroy, L.K. Johnson, Combust. Sci. Technol. 116
(1996) 31.
[9] G.T. Linteris, L. Truett, Combust. Flame 105 (1996) 15.
[10] T. Noto, V. Babushok, A. Hamins, W. Tsang, A.
Miziolek, Proc. Combust. Inst. 26 (1996) 1377.
[11] T. Noto, V. Babushok, A. Hamins, W. Tsang, Combust.
Flame 112 (1998) 147.
[12] C.R. Casias, J.T. McKinnon, Combust. Sci. Technol. 116 (1996) 289.
[13] C.R. Casias, J.T. McKinnon, Proc. Combust. Inst. 27
(1998) 2731.
[14] G.T. Linteris, D.R. Burgess Jr., V. Babushok, M.
Zachariah, W. Tsang, P. Westmoreland, Combust.
Flame 113 (1998) 164.
[15] T. Takahashi, T. Inomata, T. Abe, H. Fukayama, E.
Hayashi, T. Ono, Combust. Sci. Technol. 131 (1998)
187.
[16] Y. Saso, N. Saito, C. Liao, Y. Ogana, Fire Saf. J. 26
(1996) 303.
[17] Y. Saso, D.L. Zhu, H. Wang, C.K. Law, N. Saito, Combust. Flame 114 (1998) 457.
[18] Y. Saso, Y. Ogawa, N. Saito, Combust. Flame 118
(1999) 489.
[19] C.H. Kim, O.C. Kwon, G.M. Faeth, J. Prop. Power 18
(2002) 1059.
[20] G.M. Faeth, C.H. Kim, O.C. Kwon, Clean Air 4 (2003)
115.
[21] P.O. Wieland, Designing for Human Presence in Space:
An Introduction to Environmental Control and Life
Support Systems, NASA Reference Publication 1324,
Marshall Space Flight Center, AL, 1994.
[22] C. Huggett, Aerospace Med. 40 (1969) 1176.
[23] C. Huggett, Combust. Flame 20 (1973) 140.
[24] J. Manton, G. von Elbe, B. Lewis, J. Chem. Phys. 20
(1952) 153.
[25] G.H. Markstein, Non-Steady Flame Propagation, Pergamon, New York, 1964, p. 22.
[26] R.A. Strehlow, L.D. Savage, Combust. Flame 31
(1978) 209.
[27] P. Clavin, Prog. Energy Combust. Sci. 11 (1985) 1.
[28] C.K. Law, Proc. Combust. Inst. 22 (1988) 1381.
[29] C.M. Vagelopoulos, F.N. Egolfopoulos, C.K. Law,
Proc. Combust. Inst. 25 (1994) 1341.
[30] K.T. Aung, M.I. Hassan, O.C. Kwon, L.-K. Tseng,
G.M. Faeth, Combust. Sci. Technol. 173 (2002) 61.
[31] S. Kwon, L.-K. Tseng, G.M. Faeth, Combust. Flame 90
(1992) 230.
[32] K.T. Aung, M.I. Hassan, G.M. Faeth, Combust.
Flame 109 (1997) 1.
95
[33] K.T. Aung, M.I. Hassan, G.M. Faeth, Combust.
Flame 112 (1998) 1.
[34] O.C. Kwon, G.M. Faeth, Combust. Flame 124 (2001)
590.
[35] P.J. Padley, T.M. Sugden, Proc. Combust. Inst. 7 (1958)
235.
[36] B. Lewis, G. von Elbe, Combustion Flames and Explosions of Gases, third ed., Academic Press, New York,
1987, p. 395.
[37] C.J. Butler, A.N. Hayhurst, Combust. Flame 115
(1998) 241.
[38] L.-K. Tseng, M.A. Ismail, G.M. Faeth, Combust.
Flame 95 (1993) 410.
[39] P.D. Ronney, H.Y. Wachman, Combust. Flame 62
(1985) 107.
[40] R. Siegel, J.R. Howell, Thermal Radiation Heat Transfer, second ed., McGraw–Hill, New York, 1989, p. 613.
[41] G. Dixon-Lewis, Combust. Sci. Technol. 34 (1983) 1.
[42] B.J. McBride, M.A. Reno, S. Gordon, CET93 and
CETPC: An Interim Updated Version of the NASA
Lewis Computer Program for Calculating Complex
Chemical Equilibrium with Applications, NASA TM
4557, 1994.
[43] O.C. Kwon, A Theoretical and Experimental Study of
Preferential-Diffusion/Stretch Interactions of Laminar
Premixed Flames, Ph.D. thesis, Univ. of Michigan, Ann
Arbor, MI, 2000.
[44] S.C. Taylor, Burning Velocity and Influence of Flame
Stretch, Ph.D. thesis, Univ. of Leeds, Leeds, UK, 1991.
[45] D.R. Dowdy, D.B. Smith, S.C. Taylor, A. Williams,
Proc. Combust. Inst. 23 (1990) 325.
[46] M.J. Brown, I.C. McLean, D.B. Smith, S.C. Taylor,
Proc. Combust. Inst. 26 (1996) 875.
[47] V.F. Karpov, A.N. Lipatnikov, P. Wolanski, Combust.
Flame 109 (1996) 436.
[48] D. Bradley, R.A. Hicks, M. Lawes, C.G.W. Sheppard,
R. Woolley, Combust. Flame 115 (1998) 126.
[49] B. Rogg, RUN-1DL: The Cambridge Universal Laminar Flame Code, Department of Engineering, TR
CUED/A-THERMO/TR39, Univ. of Cambridge, Cambridge, UK, 1991.
[50] R.J. Kee, G. Dixon-Lewis, J. Warnatz, M.E. Coltrin,
J.A. Miller, A FORTRAN Computer Code Package for
the Evaluation of Gas-Phase, Multicomponent Transport Properties, Report SAND86-8246, Sandia National Laboratories, Albuquerque, NM, 1992.
[51] R.J. Kee, F.M. Rupley, J.A. Miller, The CHEMKIN
Thermodynamic Data Base, Report SAND87-8215B,
Sandia National Laboratories, Albuquerque, NM,
1992.
[52] R.J. Kee, F.M. Rupley, J.A. Miller, CHEMKIN II: A
Fortran Chemical Kinetics Package for the Analysis
of Gas Phase Chemical Kinetics, Report SAND898009B, Sandia National Laboratories, Albuquerque,
NM, 1993.
[53] R.J. Kee, J.F. Grcar, M.D. Smooke, J.A. Miller, A
FORTRAN Program for Modeling Steady Laminar
One-Dimensional Premixed Flames, Report SAND858240, Sandia National Laboratories, Albuquerque,
NM, 1993.
[54] T.J. Kim, R.A. Yetter, F.L. Dryer, Proc. Combust.
Inst. 25 (1994) 759.
96
L. Qiao et al. / Combustion and Flame 143 (2005) 79–96
[55] R.A. Yetter, F.L. Dryer, H. Rabitz, Combust. Sci. Technol. 79 (1991) 97.
[56] M.A. Mueller, T.J. Kim, R.A. Yetter, F.L. Dryer, Int. J.
Chem. Kinet. 31 (1999) 113.
[57] N.M. Marinov, C.K. Westbrook, W.J. Pitts, in: S.H.
Chan (Ed.), Transport Phenomena in Combustion,
vol. I, Taylor & Francis, Washington, DC, 1996, p. 118.
[58] W. Wantz, B. Rogg, Combust. Flame 94 (1993) 271.
[59] M. Frenklach, C.T. Bowman, G.P. Smith, W.C. Gardiner, World Wide Web location, http://www.me.
berkeley.edu/grimech/, Version 2.1, 1995.
[60] M. Frenklach, C.T. Bowman, G.P. Smith, W.C. Gardiner, World Wide Web location, http://www.me.
berkeley.edu/grimech/, Version 3.0, 1999.
[61] E.G. Groff, Combust. Flame 48 (1982) 51.
[62] C.K. Westbrook, Combust. Sci. Technol. 34 (1983)
201.
[63] J.V. Michael, M.C. Su, J.W. Sutherland, J.J. Carroll,
A.F. Wagner, J. Phys. Chem. A 106 (2002) 5297–
5313.
[64] J. Li, Z. Zhao, A. Kazakov, F.L. Dryer, Int. J. Chem.
Kinet. 36 (2004) 1–10.