Combustion and Flame 138 (2004) 384400

www.elsevier.com/locate/jnlabr/cnf

Characteristics of turbulent nonpremixed jet flames under

normal- and low-gravity conditions
Cherian A. Idicheria, Isaac G. Boxx, Noel T. Clemens
Center for Aeromechanics Research, Department of Aerospace Engineering and Engineering Mechanics,
The University of Texas at Austin, Austin, TX 78712-1085, USA
Received 13 June 2003; received in revised form 19 June 2004; accepted 19 July 2004
Available online 12 August 2004

Abstract
An experimental study was performed with the aim of investigating the structure of transitional and turbulent
nonpremixed jet flames under different gravity conditions. Experiments were conducted under three gravity levels,
viz., 1 g, 20 mg, and 100 g. The milligravity and microgravity conditions were achieved by dropping a jet-flame
rig in the University of Texas at Austin 1.25-s and NASA-Glenn Research Center 2.2-s drop towers, respectively.
The flames studied were piloted nonpremixed propane, ethylene, and methane jet flames at source Reynolds numbers ranging from 2000 to 10,500. The principal diagnostic employed was time-resolved cinematographic imaging
of the visible soot luminosity. Mean and root-mean-square (RMS) images were computed, and volume rendering
of the image sequences was used to investigate the large-scale structure evolution and flame tip dynamics. The
relative importance of buoyancy was quantified with the parameter, L , as defined by Becker and Yamazaki (Combust. Flame 33 (1978) 123149). The results showed, in contrast to some previous microgravity studies, that the
high-Reynolds-number flames have the same flame length irrespective of the gravity level. The mean and RMS
luminosity images and the volume renderings indicate that the large-scale structure and flame tip dynamics are
essentially identical to those of purely momentum-driven flames provided L is less than approximately 23. The
volume renderings show that the luminous structure velocities (i.e., celerities) normalized by the jet exit velocity
3/2
are approximately constant for L < 6, but scale as L for L > 8. The flame length fluctuation measurements
and volume renderings also indicate that the luminous structures are more organized in low gravity than in normal
gravity. Finally, taken as a whole, this study shows that L is a sufficient parameter for quantifying the effects of
buoyancy on the fluctuating and mean characteristics of turbulent jet flames.
2004 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
Keywords: Turbulent nonpremixed flames; Microgravity; Buoyancy

1. Introduction
Becker and Yamazaki [1,2] and Becker and Liang
[3] were among the first to systematically study the
* Corresponding author. Fax: (512)-471-3788.

E-mail address: clemens@mail.utexas.edu

(N.T. Clemens).

effects of buoyancy on the characteristics of turbulent nonpremixed jet flames, such as soot formation,
entrainment, and luminous flame length. They proposed that the effects of buoyancy could be quantified by a nondimensional buoyancy parameter, L
(defined below), which is a measure of the relative
importance of the buoyancy force to source momentum over the entire flame length. They concluded that

0010-2180/$ see front matter 2004 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
doi:10.1016/j.combustflame.2004.07.002

C.A. Idicheria et al. / Combustion and Flame 138 (2004) 384400

the effects of buoyancy on the characteristics of the

flame become negligible when this nondimensional
parameter is less than unity. In their experiments they
lowered L by increasing the Reynolds number; however, this raises the important issue of whether any
observed differences in the flame characteristics are
due to the reduced importance of buoyancy or to the
larger Reynolds number.
In the past couple of decades, microgravity environments have been used to investigate the effects
of buoyancy on a wide range of combustion systems. The microgravity environment offers the advantage that buoyancy effects can be isolated, because gravity can be changed without having to modify the Reynolds number. Bahadori et al. [4] and
Hegde et al. [57] were among the first to investigate
nonpremixed jet flames in the laminar-to-turbulent
regime in normal and microgravity conditions. Their
primary diagnostic was video-rate (30 fps) luminosity
imaging, which was used for flow visualization and
to obtain flame length data over a range of Reynolds
numbers. Their results showed that there are significant differences in the characteristics of normal and
microgravity nonpremixed jet flames. For example, at
a Reynolds number of about 5000 their microgravity
flames were more than twice as long as their normalgravity flames, the latter of which had L 6.5. This
indicates a much stronger dependence of flame length
on L than would be indicated by the results of Becker
and Yamazaki [1]. This raises the issue of whether L
is a sufficient parameter for quantifying the effects of
buoyancy.
With particular focus on the underlying turbulent structure, studies of transitional nonpremixed
jet flames have shown that disturbances originate
at the base of the flame in microgravity and travel
upwards as Reynolds number is increased, whereas
in normal gravity, the disturbances originate near
the flame tip and work their way down [4]. Furthermore, normal gravity studies of turbulent nonpremixed flames have shown that flame tip burnout
dynamics are closely related to the large-scale organization of the jet flame [8], and hence are strongly
affected by buoyancy [9]. To date, there is no consensus as to the nature of the large-scale motions
present in purely momentum-driven round jet flames,
although there is evidence for both axisymmetric
and helical structures [5,8,9]. It has also been suggested that buoyancy can substantially influence the
large-scale structure of even nominally momentumdriven flames, since the low-velocity flow outside of
the flame will be more susceptible to buoyancy effects [10]. Even subtle buoyancy effects may be important because changes in the large-scale structure
have implications for the fluctuating strain rate, which
influences the structure of the reaction zone.

385

There are evident limitations in the range of

conditions that were achieved by Becker and Yamazaki [1,2], by Becker and Liang [3], and in previous microgravity studies [47]. For example, in
Refs. [13], they were not able to obtain values of
L less than 3 and so they could not study flames
that were momentum-dominated (according to their
own criterion) over the full length of the flames,
whereas Refs. [46] investigated only a limited range
of Reynolds numbers (250 < ReD < 5800). This
study aims to improve upon and add to our knowledge
of turbulent nonpremixed jet flames by investigating
a range of Reynolds number and L wider than in
these previous studies and to investigate specifically
the effect of buoyancy on the large-scale luminous
structures observed in the jet flames. We take advantage of an ability to study flames at a range of
Reynolds numbers and under three different gravity
levels, viz., 1 g, 20 mg, and 100 g. The three gravity
levels make it possible to alter the value of L through
two orders of magnitude, while maintaining the same
Reynolds number. The reduced gravity levels are
achieved by using the 1.25-s University of Texas drop
tower facility (UT-DTF) and the 2.2-s drop tower at
NASA Glenn Research Center (GRC). The primary
diagnostic employed was cinematographic imaging
of the flame luminosity. The cinematographic imaging improves upon the video-rate (30-Hz) imaging
used in previous studies of microgravity nonpremixed
jet flames [46] because it enables us to investigate
the evolution and dynamics of large-scale turbulent
structures. Furthermore, the flame length results in
Refs. [46] seem to suggest that L is not sufficient to
quantify the effects of buoyancy, and so a major objective of this work is to address this specific issue and
to determine at what value of L the turbulent structure reaches its asymptotic momentum-dominated
state.

2. Experimental program
2.1. Drop-rig
The experiments were conducted using a selfcontained combustion drop-rig in the UT and GRC
drop towers. A schematic of the drop-rig is shown
in Fig. 1. The drop-rig consists of a turbulent jet
flame facility and an onboard image and data acquisition system assembled in the NASA-GRC 2.2-s
drop tower frame. The fuel jet issues from a 1.75-mm
(inner diameter) stainless steel tube, surrounded by
a 25.4-mm-diameter concentric, premixed, methane
air flat-flame pilot (operated near stoichiometric conditions). The pilot flame was used to ignite the main
jet during the drop and also to keep the jet flame

386

C.A. Idicheria et al. / Combustion and Flame 138 (2004) 384400

end (the latter value is due to aerodynamic drag because no drag shield is used). The g-jitter (defined
as peak to peak variation) from these measurements
was typically 3 mg. A Kistler Model 8303-A50 KBeam capacitive accelerometer was used to measure
the deceleration of the drop-rig on impact at the end
of each drop. Impact loading thus measured ranged
from 25 to 30 g. To reduce the effects of outside disturbances while performing the experiments in the
UT-DTF, the sides of the drop-rig were closed with
aluminum sheets.
2.3. 2.2-s drop tower
Fig. 1. Schematic diagram of the drop-rig.

attached. Flame luminosity was imaged using a Pulnix TM-6710 progressive scan CCD camera, capable of operating at 235 or 350 fps, at resolutions of
512 230 and 512 146 pixels, respectively. The
camera was electronically shuttered, with the exposure time depending on flame luminosity (1/235 to
1/2000 s), and was fitted with a 6-mm focal length,
f/16 CCTV lens, chosen to maximize the field of view
(typically 405 mm). The drop-rig was fully automated
through a custom-configured passive back-plane type
onboard computer (CyberResearch Inc). The onboard
computer had no monitor or keyboard due to space
constraints in the rig and was controlled remotely
from a notebook computer. A program developed in
LabVIEW was used for timing and control of the experiment. A more detailed description of the drop-rig
is given in Idicheria et al. [11].
2.2. 1.25-s drop tower
The 1.25-s UT-DTF is 10.7 m tall and has a 2.5-m
square cross-sectional area. The tower is equipped
with a 2-ton capacity electric hoist and a cargo hook
at the end of the hoists chain that acts as the quickrelease mechanism. At the base of the drop tower is
a deceleration mechanism consisting of a container
1.7 m long by 1.1 m wide by 1.8 m deep, filled
with flame-retardant, HR-24 polyurethane foam. The
floor of the container is lined with two 150-mmthick sheets of foam, and the rest of the container is
filled with 150-mm foam cubes. After allowing for
the space taken up by the electric hoist and the deceleration mechanism, the drop tower has a 7.6-m
free-fall section. This allows approximately 1.25 s of
low-gravity time per drop. In order to characterize
the milligravity conditions, data were acquired using a Kistler Model 8304-B2 K-Beam capacitive
accelerometer. These measurements acquired in the
UT-DTF indicate that the gravity levels range from
0 mg at the beginning of the drop to 20 mg by the

The 2.2-s drop tower at NASA-GRC is approximately 24 m tall. The drop-rig is enclosed in a drag
shield to minimize the aerodynamic drag on the experiment. The assembly consisting of the drop-rig and
drag shield is attached to a pneumatic release system
at the top of the tower prior to the drop. At this point,
the drop-rig stands 191 mm from the base of the drag
shield. After the release, the drop-rig falls through the
191 mm inside the drag shield while the whole assembly of drag-shield and drop-rig falls through 24 m.
At the end of the drop the assembly impacts an air
bag and comes to rest. During the 2.2-s drop time microgravity levels of 100 g is attained. Impact levels
during the deceleration are in the range of 1530 g.
3. Experimental conditions
Three different jet fuels were studied (propane,
ethylene, and methane) and experiments were conducted for a range of Reynolds numbers (2000 <
ReD < 10,500). The Reynolds number (ReD =
U0 D/) is based on the jet exit diameter (D), the
jet exit velocity (U0 ), and the kinematic viscosity ()
of the fuel at room temperature. The primary experimental conditions that are referred to in this paper
are shown in Table 1. More cases were also studied to
Table 1
Experimental conditions
L
(1 g)

L
(20 mg)

L
(100 g)

2500
5000
7500
8500

12.0
10.1
8.3
7.9

4.0
2.8
2.3
2.1

0.66
0.49
0.38
0.36

12.4
24.8
37.2
52.2

2500
5000
7500
10,500

8.5
6.6
4.6
3.7

2.5
1.6
1.2
1.0

19.5
24.4

2000
2500

8.0
7.3

2.4
2.1

Fuel

U0
(m/s)

Propane

6.2
12.5
18.7
21.2

Ethylene

Methane

ReD

C.A. Idicheria et al. / Combustion and Flame 138 (2004) 384400

387

Fig. 2. Flame-luminosity image sequences of ethylene jet flames showing the startup transient: (a) normal gravity, ReD = 2500
(t0 = 0.063 s, t = 0.006 s), (b) milligravity, ReD = 2500 (t0 = 0.069 s, t = 0.017 s).

obtain finer resolution in Reynolds number, primarily

for the flame length measurements. The buoyancy
1/3
parameter [1] is defined as L = Ris L/Ds , where
Ris is the source Richardson number (Ris = gDs /U02 )
based on the source diameter Ds = D(0 / )1/2 ,
and source velocity U0 (assuming a top-hat velocity
profile), L is the average luminous flame length, g is
the acceleration due to gravity, 0 is the jet fluid density, and is the ambient density. The experiments
were conducted for three different gravity levels of
1 g, 20 mg, and 100 g; thus, for the different fuels
studied, L varied from 3.7 to 12.0 in normal-gravity,
1.0 to 4.0 in milligravity, and 0.36 to 0.66 in microgravity conditions.

4. Results and discussion

4.1. Instantaneous, mean, and RMS luminosity
Sample time sequences of flame-luminosity images acquired during the startup transient are shown
in Fig. 2. The ethylene jet flames shown developed
in normal-gravity (Fig. 2a) and milli-gravity (Fig. 2b)
conditions and both had a Reynolds number of 2500.
Two times are given in the figure caption; the first is
the elapsed time from the start of a drop (t0 ), and the
second is the time between successive frames (t ).
The luminous structure is seen to be very similar
across gravity conditions in the first two frames but
becomes substantially different as time progresses.

The low-gravity flame is thicker than the normalgravity flame and exhibits disturbances over the full
length of the flame. In contrast, the normal-gravity
flame is thin and seems to be highly laminarized, as
it exhibits fluctuations only at the flame tip. This observation is consistent with the observations of Bahadori et al. [4], who studied laminar and transitional
jet flames in microgravity. They proposed that disturbances in laminar normal-gravity flames start at the
flame tip and propagate to the base, whereas the opposite trend is observed in microgravity. By the last
frames in Fig. 2a, the normal-gravity flame exhibits
disturbances over its entire length, which is something
that continues into the steady-state portion of the run,
as will be shown below (here, the term steady state
means a stationary state that is steady in the mean
properties).
The time-resolved image sequences can be used
to determine the time that is required for the flames to
reach a steady state. This is a critical issue owing to
the short times that are available with the UT drop
tower. In the low-Reynolds-number case of Fig. 2,
the flame tip is seen to reach a steady state in approximately 0.3 s in milligravity, whereas in normal
gravity, it takes only about 0.2 s. The shorter transient
in normal gravity is because the buoyant acceleration
increases the local velocity of the jet and hence reduces the convection time.
The startup transient can be more effectively seen
by considering the time history of the instantaneous
visible flame length. The instantaneous flame length
was located by converting each luminosity image into

388

C.A. Idicheria et al. / Combustion and Flame 138 (2004) 384400

Fig. 3. Flame tip time history for propane flame at ReD =

5000.

a binary image by applying a threshold to the pixel

intensities. The threshold was set to be 20% of the
maximum; i.e., pixels with intensity less than 20% of
the maximum were set to black, whereas pixel intensities higher than this were set to white. From this
binary image the perimeter of the visible flame was
computed. The pixel farthest from the jet exit in the
perimeter image was defined as the flame tip. Sample flame-length time histories of propane jet flames
at ReD = 5000, under normal and low-gravity conditions, are shown in Fig. 3. The time scale in Fig. 3
extends over about 1.1 s, which is about the duration
of a drop under milligravity conditions. The propane
flames are shown because they tend to exhibit longer
startup transients than the ethylene flames (owing to
the lower jet exit velocities), and so better exemplify
the limitations of the startup process on the data. It
can be seen that for the case that is presented the
time to steady state is approximately 0.2 s for the
normal-gravity case, but about 0.4 s for the milligravity and microgravity cases. The startup transient of
about 0.4 s corresponded to the maximum transient
that was observed among all of the cases that were
tested in this study. Given this maximum transient,
in all cases at least 0.8 and 1.8 s of steady-state test
time were available under milligravity and microgravity conditions, respectively.
Sample time sequences of flame-luminosity images captured during the steady-state portion of the
run for ethylene and propane flames under normaland low-gravity conditions are presented in Fig. 4.
Only every fourth frame of each data set is shown so
that larger changes in the flame structure can be observed from frame to frame. Figs. 4a and 4b shows
ethylene flames at ReD = 2500 in normal and milligravity conditions, respectively, for which the L
values are substantially different (L = 8.5 in 1 g,

L = 2.5 in 20 mg). A higher Reynolds number

propane flame (ReD = 8500) under normal and microgravity conditions is shown in Figs. 4c and 4d,
which also is characterized by a relatively large difference in L (L = 7.9 in 1 g, L = 0.36 in 100 g).
Fig. 4 is interesting because it enables the comparison
of flames with equal Reynolds number and different
L and also the comparison of flames with similar L
but significantly different Reynolds number.
Figs. 4a and 4b shows that the characteristics of
the low Reynolds number flames at normal and low
gravity are significantly different. For example, turbulent structures travel at a significantly higher velocity under normal gravity than in low gravity. This
can be observed in the time-sequenced images shown
by tracking particular structures, but it is more apparent from viewing the movie sequences. As mentioned above the higher structure velocity (celerity)
in normal gravity is a result of the acceleration induced by the buoyancy forces. Figs. 4a and 4b also
shows that the normal-gravity flame is significantly
thinner than the low-gravity flame, which is an observation that is consistent with previous studies [4,5].
The thicker low-gravity flames are a result of the convection/diffusion competition that governs the growth
rate of low-Reynolds-number jets. In the presence of
gravity, the upward buoyant acceleration increases the
rate of downstream convection and hence reduces the
growth rate. Interestingly, this difference in the width
of the flames persists to higher Reynolds number as
can be seen by comparing Figs. 4c and 4d, for which
ReD = 8500. The persistence of this effect is undoubtedly related to the laminarization that is associated
with high-heat-release flames, particularly in the portion of the flow to the outside (i.e., air side) of the
reaction zone. The difference in the thickness of the
flames will be more apparent when the mean images
are presented, below.
Fig. 5 shows additional instantaneous images of
ethylene and propane flames over a wider range of
ReD and L . Each set of images (i.e., those separated
by vertical lines in Fig. 5) is for the same fuel type
and Reynolds number. All of the cases with L < 1
were taken in microgravity, cases with 1 < L < 3
were taken in milligravity, and cases with L > 3
were taken in normal gravity. Fig. 5a shows ethylene flames at the highest Reynolds number considered
(ReD = 10,500) that develop under normal-gravity
(left) and milligravity (right) conditions. It is seen
that at this highest Reynolds number, the structure
of the flame seems to be very similar regardless of
gravity level. This is perhaps not a surprising result since the L values in normal gravity (L = 3.7)
and milligravity (L = 1.0) are close to the criterion
for momentum-dominated flames of L = 1 proposed
by Becker and Yamazaki [1]. Fig. 5b shows ethyl-

C.A. Idicheria et al. / Combustion and Flame 138 (2004) 384400

389

Fig. 4. Sample time-sequenced luminosity images: (a) normal gravity, ethylene ReD = 2500, t = 0.011 s; (b) milligravity,
ethylene ReD = 2500, t = 0.011 s; (c) normal gravity, propane ReD = 8500, t = 0.017 s; and (d) microgravity, propane
ReD = 8500, t = 0.017 s.

ene flames at a Reynolds number of 5000, and it

appears that there are greater differences in the turbulent structure than at the higher Reynolds number,
which is consistent with the larger difference in L .
Figs. 5c and 5d each show the instantaneous structure
of propane flames under three different gravity conditions. The Reynolds numbers are 8500 and 5000
for Figs. 5c and 5d, respectively, and the normal-,
milli-, and microgravity flames are shown at left, cen-

ter, and right, respectively. It is clear from these figures that the structure of the milli- and microgravity
flames is very similar, which suggests that both the
flames are approximately nonbuoyant. Since the L
values of these milligravity flames range from 23,
this seems to suggest that the criterion for the development of momentum-dominated turbulent structure
may be larger than unity. This issue will be discussed
further below.

390

C.A. Idicheria et al. / Combustion and Flame 138 (2004) 384400

Fig. 5. Sample instantaneous luminosity images: (a) ethylene ReD = 10,500, x/D = 43279, normal (left) and milligravity
(right); (b) ethylene ReD = 5000, x/D = 43279 normal (left) and milligravity (right); (c) propane ReD = 8500, x/D = 76308,
normal (left), milligravity (center), and microgravity (right); and (d) propane ReD = 5000, x/D = 76308, normal (left), milligravity (center), and microgravity (right).

Upon careful viewing of the instantaneous images and movie sequences, a few generalizations can
be made about the luminous turbulent structures that
characterize the low-Reynolds-number flames when
there is a large difference in the magnitude of L .
For example, the most obvious trend seen in the turbulent structures of the transitional flames is that in
low gravity they are more axisymmetric, extend over
a relatively small scale (e.g., about 12 luminous jet
widths), and exhibit a relatively regular spacing. In
normal gravity, the structure is similar to that of the
low-gravity case in the lower portion of the flame, but
farther downstream the flames tend to exhibit a largescale sinuous structure whose wavelength is several
jet widths long. Fig. 6 shows highly simplified cartoons of these differences in the luminous structure of
the normal- and low-gravity transitional flames. The
cartoons are meant to show an exaggerated view of
the differences, but in reality, either type of flame can
exhibit characteristics of the other; i.e., the buoyant
flames can exhibit more axisymmetric structures or
the low-gravity flames can exhibit a sinuous structure. Nevertheless, the differences discussed above
are readily evident upon viewing the time sequences
and approximately describe the gross features of the
two types of flames. Specific examples of these trends
in the instantaneous images can be seen by comparing
high and low L flames in Figs. 4a and 4b, Figs. 5c
and 5d, and the startup sequences in Figs. 2a and 2b.
These differences in the structure also seem to have a
bearing on the flame tip dynamics as will be discussed
below.
Mean and RMS luminosity images were computed
from the time sequences, excluding the startup and
shutdown transient frames. Sample mean luminosity images corresponding to the same conditions as

Fig. 6. Cartoon of the luminous flame structure of transitional flames in (a) normal gravity and (b) low gravity.

shown in Fig. 5 are presented in Fig. 7. Each set

of images (i.e., those separated by vertical lines in
Fig. 7) is for the same fuel type and Reynolds number.
These mean images are superior in showing some of
the gross differences that characterize the flames under different conditions. For example, Fig. 7a shows
that the flame with L = 3.7 is very similar in its
mean structure to the one with L = 1. A careful
viewing of all of the images in Fig. 7 supports this
general view that flames with L of order unity and
below are essentially identical in their mean luminosity (flame height and width). As the L of the flames

C.A. Idicheria et al. / Combustion and Flame 138 (2004) 384400

391

Fig. 7. Sample mean luminosity images: (a) ethylene ReD = 10,500, x/D = 43279, normal (left) and milligravity (right); (b)
ethylene ReD = 5000, x/D = 43279 normal (left), and milligravity (right); (c) propane ReD = 8500, x/D = 76308, normal
(left), milligravity (center), and microgravity (right); and (d) propane ReD = 5000, x/D = 76308, normal (left), milligravity
(center), and microgravity (right).

Fig. 8. Variation of normalized flame length with Reynolds

number at different gravity levels.

becomes larger than about 3, the flames become progressively thinner than their momentum-dominated
counterparts.
Fig. 8 shows the variation of the mean visible
flame length (obtained from mean images) normalized by the tube exit diameter for all the cases studied.
Precision uncertainty levels (95% confidence) computed from repeated runs are also shown. The confidence intervals are in the range of 4D to 35D for
all three flames, with higher differences in the lowerReynolds-number cases. It is evident from Fig. 8 that
the ethylene and propane flame lengths exhibit virtually no difference with gravity level at the highest Reynolds-numbers considered. Fig. 8 shows that
the mean flame lengths of the ethylene jet flames
differ by at most 15% across the different gravity levels and over the full Reynolds-number range.

For propane, the variation in length with gravity level

is 20% at the lowest Reynolds-number, whereas at
higher Reynolds numbers the difference is less than
10%. Only low-Reynolds-number methane flames
were tested because at higher Reynolds numbers the
flames were lifted, which was not desirable for the
purposes of this study.
There are some trends in flame-length behavior for
a particular fuel and across gravity conditions that are
evident in Fig. 8. The normal-gravity propane flames
are seen to increase in their normalized mean luminous flame length (L/D) from a value of approximately 200 to 270 as the Reynolds number increases
from 2500 to 8500. The milligravity and microgravity propane flames also show the same trends, increasing mean luminous flame length with increasing
Reynolds number. Substantial differences in the mean
luminous flame length between gravity conditions
are seen only in the low-Reynolds-number propane
flames (e.g., the low-gravity flame is longer than the
normal-gravity flame by 45D at ReD = 2500). However, a majority of the low-gravity propane flames
appear to be longer than their normal-gravity counterparts by about 10%. The normal-gravity ethylene
flame length is seen to increase from L/D of approximately 180 to 200 as the Reynolds number increases from 2500 to 10,500. Differences in flame
length between gravity conditions for ethylene are
seen for Reynolds numbers less than 6000, as the majority of the normal-gravity flames are longer than the
low-gravity flames. However, the flame lengths are
very similar for Reynolds number higher than 6000.
As mentioned previously, the methane flames were
tested under milligravity and normal-gravity conditions only. The methane flames at the two Reynolds
numbers investigated are longer in milligravity than

392

C.A. Idicheria et al. / Combustion and Flame 138 (2004) 384400

in normal gravity by approximately 12 and 5% at

Reynolds numbers of 2000 and 2500, respectively.
Fig. 8 also shows the data of Hegde et al. [6]
for propane flames under normal and microgravity
conditions. Their microgravity data and the current
low-gravity data differ substantially over the entire
Reynolds-number range. The normal-gravity data of
both studies, however, show better agreement, but still
are significantly different in magnitude and trend with
Reynolds number. The computation of visible flame
lengths will depend on the definition of the length,
and therefore absolute differences are not surprising;
however, in contrast to the current findings, the differences in trend between normal and microgravity
flames are seen to be very large even for Reynolds
numbers greater than 4000. Note that similar differences in the flame lengths between normal and microgravity conditions were also observed in methane
and propylene jet flames [4].
The reason for the difference between the measurements of Hegde et al. [6] and the current study
is not known, but one other microgravity study [12]
shows agreement with the current measurements.
Page et al. [12] studied pulsed, turbulent nonpremixed
ethylene/oxygen-enriched-air jet flames in microgravity, but they also included flame length measurements for steady, unpulsed jet flames at a Reynolds
number of 5000. Their normal and microgravity flame
lengths did not exhibit the large difference observed
in Refs. [46], but the normal-gravity flame was actually slightly longer than the microgravity one. This
finding is consistent with the current study where the
normal-gravity ethylene flame at ReD = 5000 was
also observed to be slightly longer than the flame in
the milligravity case.
Despite the agreement of the current results with
those of Ref. [12], the difference from Bahadori
et al. [4] and Hegde et al. [5,6] could mean that the
current setup is generating anomalous results, even
at normal gravity. Therefore, to serve as validation
of the current normal-gravity results, Fig. 9 shows
normal-gravity flame length data taken in the current
study plotted together with the data of Becker and Yamazaki [1] and Mungal et al. [9]. It can be seen that
the present data agree quite well both in trend and in
value with previously published work for the same
fuel and the same range of Reynolds numbers.
To give further confidence in the reliability of
the normal-gravity results, a series of tests were performed to see if the current normal-gravity flames
were sensitive to the particular setup used. First of all,
tests were conducted at normal gravity to see if nonpiloted lifted propane flames differed substantially in
their length from the piloted attached flames. The differences seen in the flame heights for these conditions
were small. Second, tests were conducted with the

Fig. 9. Comparison of current normal-gravity flame length

data with other published data.

burner in various configurations; specifically, normalgravity tests were conducted with the burner inside
and outside the drop rig and with and without the pilot flame housing. In all of these tests, the difference
in observed flames lengths was small. This series of
tests showed that the current normal-gravity flames
were not highly sensitive to how they were generated.
It seems likely, therefore, that the reason for the
observed differences among the various microgravity
studies is that transitional low-gravity flames are particularly sensitive to the boundary conditions under
which they develop. The reason for this proposed increased sensitivity is that under normal gravity conditions, buoyancy-induced fluctuations are the primary
mechanism that triggers the transition to turbulence.
In microgravity, this source of disturbances is removed and therefore it leaves the flame sensitive to
other, possibly much weaker, disturbances. In other
words, a microgravity flame can be sensitive to the
exact nature of the boundary conditionseven when
the same flame under normal gravity would not be
because the disturbances under normal gravity would
be dominated by buoyancy. Under this argument, the
reason for the increased flame lengths of Refs. [46]
is that they exhibit an extended laminar or transitional
region as compared to the current study and Ref. [12].
The effect of buoyancy on the transition to turbulence is well known in laminar flames. For example,
flames that are completely laminar in microgravity
(e.g., [13]) can be highly wrinkled and turbulent in
normal gravity owing to buoyancy-induced vorticity.
In fact, a major advantage of the microgravity environment is that it enables one to study low-strain-rate
laminar flames that would be dominated by buoyant
instabilities in normal gravity.
If this argument is correct, then it would be expected that flame length data obtained in microgravity

C.A. Idicheria et al. / Combustion and Flame 138 (2004) 384400

393

Fig. 10. Sample RMS luminosity images: (a) ethylene ReD = 10,500, x/D = 43279, normal (left), and milligravity (right);
(b) ethylene ReD = 5000, x/D = 43279 normal (left), and milligravity (right); (c) propane ReD = 8500, x/D = 76308,
normal (left), milligravity (center), and microgravity (right); and (d) propane ReD = 5000, x/D = 76308, normal (left), milligravity (center), and microgravity (right).

would exhibit much more scatter than equivalent data

obtained in normal gravity. For example, whether a
jet flame is piloted or not, or enclosed or free, may
have a greater impact on the flow development under
low-gravity conditions; therefore, slight differences
in the flow configuration between the current work,
Hegde et al. [5,6], and Page et al. [12] may lead to
large differences in the flame heights observed under
low-gravity conditions. Given this possibility, the differences between the experimental configuration of
the current study and those of Refs. [47] and [12]
are documented below. The jet flames in Refs. [47]
were unpiloted and enclosed in a cylindrical chamber
(with a volume of 0.087 m3 ), whereas in the current study the jet flame issued into the quiescent air
inside the drop-rig (with an unoccupied volume of
0.24 m3 ). In general, the enclosure around a flame
can have an effect because it allows recirculation of
products into the oxidizer stream, and therefore can
change the overall stoichiometry and density ratio.
However, Refs. [47] state that the flame lengths were
the same irrespective of the run time of the flame
they investigated, and so it appears that confinement
was not an issue. Another important configuration difference across the experiments is geometry near the
jet exit. In the current study, the flames were piloted
with a 25-mm concentric laminar premixed flame,
but otherwise the jet-exit region was unobstructed.
In Refs. [47], the flame was unpiloted, and a base
plate was used that was located 5 to 10 mm below the
nozzle exit. According to the authors of Refs. [47],
the presence of this plate could have impeded the entrained air near the nozzle exit, and this could have led
to lift-off and blowout at moderately low Reynolds
number. With regard to the experimental configuration used in Ref. [12], the flames were enclosed and

stabilized by an igniter (which was always present)

and issued into a weak co-flow. It is not known if these
differences are those that are most responsible for the
differences in the flame lengths, but the fact that such
differences in geometry are present give future researchers specific issues to consider when designing
experiments that will be used to study transitional microgravity jet flames.
The RMS fluctuations of the flame luminosity time
sequences were computed to determine if the trends
that are observed in the mean images are also seen in
fluctuating quantities. It is well known that soot luminosity depends on many factors and so cannot be
related in a simple manner to a particular property of
the soot, such as the soot volume fraction. As a consequence, if the RMS fluctuations for two cases are
different, it could be due to differences in the soot
properties, temperature or the underlying fluid mechanics. Nevertheless, the RMS fluctuations can provide useful information because we are interested in
detecting differences in the normal- and low-gravity
flames, regardless of the underlying mechanism. The
RMS luminosity is useful toward this end because
it provides a more sensitive measure of the potential differences (as do all higher order statistics) than
the mean luminosity. Fig. 10a shows RMS images for
the ethylene flames at ReD = 10,500 in normal and
milligravity. In these images black corresponds to a
maximum and white corresponds to a minimum RMS
value. The flames have noticeable similarities, but
clear differences are also apparent, such as the lower
peak RMS values on the centerline of the L = 1.0
flame. Furthermore, more drastic differences can be
seen when flames with a larger difference in L are
(Fig. 10b). Figs. 10c and 10d compare the RMS luminosity for the propane flames (ReD = 8500 and

394

C.A. Idicheria et al. / Combustion and Flame 138 (2004) 384400

5000). In Fig. 10c it is seen that the fluctuations are

nearly identical for the L = 2.1 and L = 0.38 cases;
however, both differ substantially from the L = 7.9
case. A similar trend is observed for the image set
of Fig. 10d. Interestingly, regardless of fuel type or
Reynolds number, the low-L flames all have qualitatively similar RMS contours; i.e., the fluctuations
peak near the periphery of the flame and remain low
even at the flame tip. This observation is consistent with expectations of a momentum-dominated jet
where the largest scalar fluctuations occur at the outer
edges of the jet where the intermittency is largest [14].
4.2. Flame tip dynamics
The time-resolved data, such as shown in Fig. 4,
enable us to investigate fluctuations in the instantaneous flame length and the dynamics of the flame tip
burnout process. Careful examination of the images
reveals that in normal gravity (or more correctly, high
L ), the flame structure elongates near the flame tip
and often tears away from the main body and then
burns out. Examples of this feature can be seen in the
third and ninth images from the left in Fig. 4a and the
fourth and tenth images from the left in Fig. 4c. The
low-gravity image sequences (Figs. 4b and 4d) indicate different flame tip behavior because the structures in low-gravity conditions are more compact and
thicker near the flame tip and the tearing of flame
structures from the flame body is not as common. In
low gravity, the luminous structures more typically
convect downstream and burn out as a whole.
The characteristics of the flame tip fluctuations
can also be seen by considering the time histories of
the instantaneous luminous flame length as shown
in Fig. 11. These data are for propane flames at
varying Reynolds number and were generated by
computing the instantaneous luminous flame length.
It is expected that the flame tip fluctuation fre-

Fig. 11. Instantaneous flame tip location for propane flames

at various L .

quency will scale with the local large-scale timescale /Uc (with the local width and Uc the
centerline velocity) [9,15], but in the current study
the local velocity is not known for all conditions.
Since /Uc x 2 /(U0 D) (D/U0 )(x/D)2 , then
/Uc (D/U0 )(L/D)2 for a turbulent momentumdominated flame of length L. For the same fuel (and
hence stoichiometry), then L/D will be nearly constant and the large-scale time (/Uc ) will scale as
D/U0 ; therefore, the time axis has been scaled by the
characteristic time scale D/U0 . This scaling should
be sufficient for removing the effect of differences in
the local convection velocity on the flame tip fluctuations for flames that are momentum-dominated and of
the same fuel type. These plots show that the flame tip
fluctuations are very similar for L values of 2.8 and
below, which indicates that the fluctuations are associated with the same type of large-scale motions in
all of the momentum-dominated cases. The L = 7.9
case seems to exhibit higher frequency fluctuations,
and this is clearly the case at L = 10.1 also. Since
the time-scale normalization used does not account
for buoyant acceleration, these higher-frequency fluctuations are clear evidence of the effect of buoyancy on the flame tip dynamics. Careful inspection
of Fig. 11 reveals some interesting trends in the nature of the flame tip fluctuations. For example, at the
lower values of L the flame-tip time histories exhibit
ramp-like characteristic, whereby the flame length
gradually increases and then abruptly decreases. Similar ramp-like oscillations in the flame length were
observed in Ref. [9] and in the liquid-phase, acidbase flames in Ref. [15]. The liquid-phase flames
were purely momentum-driven, and they exhibited
a particularly high degree of quasi-periodicity [15].
The movie sequences acquired in the current study
show that the ramp-like behavior is associated with
the flame tip burnout characteristics. In particular, the
movies show that for momentum-dominated flames,
a large-scale luminous structure will form near the
flame tip, travel downstream, and then the entire
structure will burn out in a relatively uniform manner. It is the burnout of the entire structure that causes
the flame length to abruptly decrease. In Ref. [15] it
is argued that the rapid burnout of the flame tip structure indicates that the entire structure is mixed to a
relatively uniform composition. In some cases, the
flame tip seems to burnout starting from its upstream
edge, which was also observed in liquid flames [15].
In Ref. [15], this upstream-to-downstream mode of
burnout was attributed to the entrainment motions,
which sweep ambient fluid into the structure from the
upstream side and so it is this side that reaches stoichiometric proportions first.
Although ramp-like structures can at times be seen
in the high L traces of Fig. 11, they are not as

C.A. Idicheria et al. / Combustion and Flame 138 (2004) 384400

dominant as at lower L . At L = 10.1, the structures in the time traces are jagged, but more symmetric than at lower L . This apparent difference in
the ramped structures seems to suggest that the high
L flames also deviate from the mode of burnout described above. In particular, the large-scale structures
near the flame tip are stretched out by the buoyancy
forces into the sinuous structures described above,
and apparently the entrainment motions create less
uniformly mixed structures that burnout more gradually. In addition to the difference in the ramp-like time
traces, careful observation of the movie sequences indicates that the luminous structures at the flame tip in
the momentum-dominated flames seem to be more organized, or coherent, than the ones that exhibit strong
buoyancy effects. The more regular flame length fluctuations in low-gravity seem to be related to the more
regularly spaced structures as illustrated in Fig. 6. The
flame tip fluctuations shown in Fig. 11 also suggest
a lower degree of organization with increasing buoyancy, since the fluctuations seem to be more random at
high L . The observation that the liquid-phase flames,
which are momentum-dominated, exhibit a high degree of periodicity, even at higher Reynolds numbers,
seems to add support to this hypothesis. This issue
will be discussed further below, but it should be noted
that in Ref. [9] it was remarked that the flame tip
fluctuations seemed to be organized across the same
range of L as considered here. In fact, it seems that
their low and high L cases all exhibit the ramp-like
burnout characteristics and arguably exhibit the same
degree of organization. Since their data were taken at
higher Reynolds numbers than in the current study it
is possible that this is the reason for the apparent discrepancy.
4.3. Volume rendering
Volume rendering of jet flame image sequences
was used to investigate further the characteristics of
the large-scale luminous structures. In this imageprocessing technique, discussed in Ref. [9], the twodimensional (x, y) images are stacked along the time
axis (t ) as shown in Fig. 12. A three-dimensional
volume (x, y, t ) of the jet flame edge is then generated using image processing. This rendered volume
enables qualitative and quantitative comparisons of
features such as large-scale structure evolution and
celerity. The celerity is the absolute velocity of a luminous structure measured in the laboratory frame of
reference and is not necessarily a convection velocity,
because a luminous structure can theoretically have
a different speed than the local flow velocity. A simulated light source, usually to the left of the stacked
images, provides illumination of the rendered surface
and shadowing for depth perception. The advantage

395

Fig. 12. Illustration of volume-rendering technique.

of the volume rendering technique is that the largescale structuresvisualized as wrinkles or bands in
the renderingscan be readily tracked over their entire lifetimes. The slope of each band in the volume
rendering is equal to the celerity of the luminous
structure. In these renderings, higher celerity structures will exhibit bands that have larger slopes. In the
current study, the renderings were computed using a
Pentium III machine equipped with 1 GB of RAM and
a commercial software package called SlicerDicer.
Using this technique, Mungal et al. [9] found the
celerity of luminous structures to be about 12% of
the jet exit velocity irrespective of the buoyancy parameter (up to L = 9) and fuel type. This observation
that the celerity is constant is intriguing because the
fluid velocities decay with downstream distance, and
it might be expected that the luminous structures velocities should decrease also. Mungal et al. [9] suggest the reason for the constant celerity is that the
stoichiometric mixture fraction surface, on which the
flame resides, is similar in shape to a constant velocity
surface, and so the luminous structures remain associated with nearly constant velocity fluid.
Sample renderings for ethylene and propane are
shown in Fig. 13. The renderings (Figs. 13a13d) are
shown from the side view and so the y-direction is
into the page. The wrinkles represent luminous structures that travel up the flame with increasing time. The
faster the structures move downstream, the larger will
be the slope of the wrinkles. The flame length variations are seen by the spiky top surface of the renderings. Figs. 13a and 13b show the rendering of ethylene flames at ReD = 2500 for L values of 8.5 and 2.5.
Fig. 13b shows the entire duration of the 1.25-s drop,
including startup (t = 0) and impact. The impact of
the drop rig into the deceleration system is marked by
the time when the flame length becomes very large.
The movie sequences show this large flame length is
associated with the creation of a large super-buoyant,
mushroom-like flame that is generated by the 1530 g
deceleration.
A comparison of Figs. 13a and 13b shows that
there are significant differences between the two

396

C.A. Idicheria et al. / Combustion and Flame 138 (2004) 384400

Fig. 13. Sample volume renderings: (a) ethylene, ReD = 2500, normal gravity (L = 8.5); (b) ethylene, ReD = 2500, milligravity
(L = 2.5); (c) ethylene, ReD = 7500, normal gravity (L = 4.6); (d) ethylene, ReD = 7500, milligravity (L = 1.2); (e) propane,
ReD = 5000, normal gravity (L = 10.1); (f) propane, ReD = 5000, milligravity (L = 2.8); and (g) propane, ReD = 5000,
microgravity (L = 0.49).

cases. It can be clearly seen that the flame tip fluctuates at a higher frequency in normal-gravity than in
milligravity. Also, the wrinkles in the normal-gravity
case have higher slopes than those for the milligravity case implying higher celerities in normal-gravity
than in milligravity. Renderings for a higher Reynolds
number of 7500 are presented in Figs. 13c (L = 4.6)
and 13d (L = 1.2). The large differences seen at the
lower Reynolds number are not readily apparent in
these renderings, and the superbuoyant flame is less
prominent in the milligravity case; however, subtle
differences in the flame tip oscillation frequencies are
still visible on careful viewing.

Figs. 13e13g show renderings for propane at a

Reynolds number of 5000 at three different gravity
levels, rotated by 25 about the y-axis. Owing to the
high density of propane, at this Reynolds number, the
jet exit velocity is relatively low and so these flames
take longer to reach a steady state in low-gravity conditions. Fig. 13f shows that this relatively long startup
transient is seen to take up about one-third of the
drop time. Comparing the slopes of the bands between Figs. 13e13g, it is apparent that the buoyancy parameter has a dominant effect on the luminous structure celerities for the propane flames also.
The normal-gravity case (Fig. 13e) exhibits wrinkles

C.A. Idicheria et al. / Combustion and Flame 138 (2004) 384400

that seem to have a finer spacing and which exhibit

larger slopes than the milligravity and microgravity
cases (Figs. 13f and 13g). The similarity in slopes between Figs. 13f and 13g indicate the negligible effect
of buoyancy when L changes from 2.8 to 0.49, but
the time at which the flow is stationary is so short that
the L = 2.8 (milligravity) case is not very convincing
in this regard.
The nearly constant slope of the wrinkles in
all of the renderings indicates that the structures
move downstream at approximately a constant velocity, in agreement with previous observations in jet
flames [9]. Occasional pairing of the structures can
also be seen as a coalescence of the wrinkles in the
renderings. Although this might not be readily apparent to the reader, after looking at many such renderings, and after watching the movies, we can conclude
that the pairing of the structures is more dominant in
the strongly buoyant cases. In other words, the luminous structures in the momentum-dominated flames
seem to have longer lifetimes, or to maintain their
identity longer, than in the buoyant flames. Perhaps
a related observation is that the difference in the nature of the flame tip fluctuations, as discussed above,
can also be seen in these renderings. For example, a
comparison of Figs. 13e and 13g shows that the variations in the flame length at normal gravity appear to
be much larger than in microgravity.

397

(a)

4.4. Celerity measurements

Fig. 14a shows a plot of the ratio of the luminous
structure celerity to jet exit velocity as a percentage,
Us /U0 (%), versus the buoyancy parameter, L . The
normal-gravity flames (high L values) are associated with higher celerity, which can be attributed to
the buoyant acceleration. This suggests that luminous
structure celerity is in fact buoyancy dependent, contrary to the findings of Mungal et al. [9]. It should
be noted, however, that since Mungal et al. [9] studied higher Reynolds number jet flames, it is possible
that the disagreement is due to a Reynolds number
effect. For L values less than about 6, the celerity
is independent of the gravity level and fuel type. In
this regime, there is reasonable agreement with the
findings of Mungal et al. [9]. The bars shown on
each data point represent the standard deviation of
the celerities and therefore quantify the variation of
measured values. It is interesting to note that highL cases have higher deviations, which imply that
the structures have a wider distribution of celerity.
However, the deviations become smaller with decreasing L which suggests greater organization (or
repeatability) of the structure celerity. This conclusion of greater organization is consistent with the
lack of merging of the luminous structures described

(b)
Fig. 14. Celerity of large-scale structures as a percentage of
the jet exit velocity: (a) linear plot; (b) loglog plot.

above, and the more regular fluctuations of the flame

tip that were observed under low-gravity conditions.
This observation of a higher degree of organization
for momentum-dominated flames is a new one, because it is usually assumed that buoyancy increases
the large-scale organization of turbulent flames (e.g.,
the large billowing structures observed in oil-well or
pool fires) [16]. Although the pure buoyancy-driven
limit may indeed exhibit strong organization, it appears that the first effect of buoyancy is to reduce the
organization by disrupting the hydrodynamic instability of the momentum-dominated jet.
The loglog plot (Fig. 14b) shows that for L > 8,
3/2
the celerity values are consistent with a L scaling
law. We can derive this result from a simplified momentum equation analysis. It is assumed that if the
structure follows the local velocity at the stoichiometric contour then the celerity should be equal to

398

C.A. Idicheria et al. / Combustion and Flame 138 (2004) 384400

tum. Following these assumptions, Eq. (1) reduces to

FB = m(x)U

c.

(2)

The buoyancy force that is exerted on the flame, modeled as an inverted cone as discussed above, is
FB


1
2 xg f .
12

(3)

Owing to the reduced density in a flame, we have

( f ) and (3) simplifies to
FB

1
2 xg .
12

(4)

The momentum at the downstream location, x, is approximated as follows:

2
m(x)U

c f Uc

Fig. 15. Schematic diagram of the control volume used in

the celerity scaling analysis.

the local centerline fluid velocity at the stoichiometric

flame length. Becker and Yamazaki [1] use a quasi-1D momentum analysis to show that in the buoyancy3/2
dominated limit the entrainment rate scales as x .
We use this same procedure to show how the local
velocity scales at the flame tip under these same conditions. Consider the simplified geometry and control
volume of a jet flame issuing into quiescent ambient
fluid as shown in Fig. 15. Let the jet fuel of density
0 exit the nozzle into the ambient of density
from a tube of diameter D with a velocity U0 and
a mass flow rate of m
0 . Assume the jet flame to be
an inverted cone of width and height x, and that
the density at each x-location can be approximated as
an appropriate average density of f (i.e., a mixingcup density [1]). Furthermore, the jet entrains ambient fluid with a mass flow rate m
e , but assume that
this entrained fluid has no initial momentum in the xdirection and so it does not contribute to the momentum balance. Owing to the presence of heat release the
jet will experience a buoyancy force, FB , as shown in
the schematic. At a particular downstream location x,
let the mass flow rate and velocity be given by m(x)

and Uc (x), respectively. Also, to further simplify the

problem, assume that the downstream velocity profile
is a constant and denote it as Uc as shown in Fig. 15.
Applying the momentum principle in the x direction
gives

m
0 U0 + FB m(x)U
c = 0.

(1)

Now consider the case where the flame is buoyancydominated, in which case the buoyancy-induced momentum is much larger than the initial source momen-

2
4

(5)

and substituting Eqs. (4) and (5) in Eq. (2) gives

2
1
2 xg f Uc2 .
12
4

(6)

Note that Eq. (6) is not a function of the source conditions (U0 or Ds ) because the source momentum
was assumed to be negligible. However, because the
celerity is normalized by U0 , we introduce the source
parameters into Eq. (6) to obtain the relation for the
normalized centerline velocity,



 


x

gDs
Uc 2
.

U0
Ds
f
U2

(7)

Now, Ris gDs /U02 and x (Ris )1/3 (x/Ds ) and,

hence, Ris = x3 (x/Ds )3 . Using these relations in
Eq. (7) yields
 1 

1/2
Uc
3/2 x
x
.
U0
Ds
f

(8)

Following the nomenclature of Tacina and Dahm [17],

we define a modified source diameter, D + , which like
Ds in nonreacting jets, is able to collapse velocity
and mixture fraction decay data in turbulent flames.
For our purposes, we define D + Ds ( /f )1/2 ,
which differs somewhat from that of [17] and was
used because it was found to work better for scalingmixture-fraction data measured in the current facility [18]. With this definition of D + we can write


x 1
Uc
3/2
x
.
U0
D+

(9)

To obtain a scaling in terms of the flame length parameters, x is replaced with L in Eq. (9). Furthermore, it
is assumed that the celerity (Us ) will scale with the local centerline velocity (Uc ) and therefore at the flame

C.A. Idicheria et al. / Combustion and Flame 138 (2004) 384400

tip we have


L
Us
3/2
L
U0
D+

1

3/2

L .

(10)

Equation (10) shows that the normalized celerity

3/2
near the flame tip will approximately scale as L
provided the flame is buoyancy-dominated. Fig. 14
shows the celerity data plotted with a line that follows
3/2
the L scaling. It is seen that this scaling seems to
be appropriate for L > 8 or so. Note that Eq. (10)
suggests that the celerity will depend on L/D + and
L but the effect of the former term will be small in
Fig. 14 if L/D + is approximately constant. Specifically, the L/D + value for the flames in the current
study were measured to be approximately 90 [18].
This suggests that the normalized celerity will be a
function of L only.
A similar analysis can be used to explore the
scaling of celerity at the momentum-dominated limit
(L 0). The normalized centerline velocity of a
momentum dominated jet flame is found to scale
as [17]
Uc

U0


x 1
.
D+

(11)

At the flame tip it is again assumed that the celerity scales with the centerline velocity and hence for
a momentum-dominated flame


L 1
Us

.
(12)
U0
D+
Equation (12) shows that the normalized celerity is
(obviously) independent of L and will have a constant value if L/D + is constant. Fig. 14 shows relatively good agreement with this scaling law because
the celerities are independent of L for L < 6, and
seem to exhibit similar values over this same range
of L .
The analysis above shows that the celerity seems
to scale with the local mean velocity, but whether it
has the same value as the local mean velocity is another issue. To explore this further consider the measured centerline velocity decay in a turbulent nonreacting jet [14], which is given by
 1
x
Uc
= 6.2
.
U0
Ds

(13)

Assuming that the velocity decay in a momentumdominated reacting jet can be obtained by substituting
Ds with D + [17] in (13) gives


x 1
Uc
= 6.2
.
U0
D+

(14)

399

Furthermore, assuming the celerity is the same as the

centerline velocity at the flame tip and that the normalized flame length L/D + is approximately 90 [18],
Eq. (14) will predict a constant normalized celerity
(Us /U0 ) of approximately 7%. Fig. 14 shows that
the mean celerities measured in this study range from
about 8 to 18% at the low-L limit, and those of
Ref. [9] were measured to be 12%. Both of these studies, therefore, suggest that the luminous structures
travel faster than the local mean fluid velocity. The
reason why the celerity is different from the local fluid
velocity is not known but it is possible that the luminous structures exhibit a wave-like behavior, with a
wave speed that differs from the local fluid velocity.
For example, consider an essentially steady laminar
flame surface that is located in a region of low-speed
flow, but which surrounds a column of fast-moving jet
fluid. If a velocity perturbation were to be introduced
into the high-speed jet fluid, then this disturbance
would travel downstream at the local jet fluid velocity.
As the disturbance moves downstream it would cause
a bulge in the laminar flame surface, which would
have the same velocity as the disturbance. The bulge
in the flame surface would have a larger velocity than
the local fluid velocity. We do not know if this discussion correctly describes the physics of the flow, but at
least it emphasizes the point that although the celerity
may scale with the local fluid velocity, there is really
no obvious reason why it should be equal to it.

5. Conclusions
The characteristics of turbulent nonpremixed jet
flames were studied at Reynolds numbers ranging
from 2000 to 10,500 and at three levels of gravity,
viz., 1 g, 20 mg, and 100 g. The flames were piloted with a small concentric premixed methaneair
flame to keep them attached to the flame base for all
Reynolds numbers considered. Time-resolved (cinematographic) imaging of the natural soot luminosity
was used to investigate the mean and RMS luminosity, flame tip dynamics, and evolution of large-scale
structures. The relative importance of buoyancy over
the entire length of the flame was quantified with the
Becker and Yamazaki [1] buoyancy parameter, L .
The mean flame luminosity data show that the normal and low-gravity flames exhibited approximately
the same flame lengths for all Reynolds numbers
tested. This result is different from some previous
studies in the literature that have shown large differences in flame lengths between normal and microgravity flames. It is conjectured that the reason
for this difference is that the microgravity flames in
the previous studies may have exhibited an extended
laminar/transitional region owing to the absence of

400

C.A. Idicheria et al. / Combustion and Flame 138 (2004) 384400

turbulence-induced vortical perturbations. This emphasizes the importance of documenting the boundary
conditions under which the flames develop when conducting microgravity studies. Furthermore, the mean
and RMS luminosity, and flame tip fluctuations suggest that the structure of the large-scale turbulence
reaches its momentum-driven asymptotic state for
values of L less than about 23. Volume renderings of image time-sequences show that the largescale luminous structure celerity depends on the value
of L . In particular, the celerity was found to be nearly
constant for momentum-dominated flames (L < 6),
3/2
but to scale as L in the buoyancy-dominated limit
(L > 8). It is argued that the celerity should scale
with the local fluid velocity, although not necessarily be equal to it, and a simple momentum-equation
analysis supports this view. Taken as a whole, the
results of this study indicate that L is sufficient to
quantify the effects of buoyancy on both the mean luminosity and different measures of the fluctuations,
provided the flame is turbulent.
Another interesting finding of this work is that
the visible flame tip time histories, volume renderings, and movie sequences, support the view that the
luminous structures of the jet flames are better organized, or coherent, when the flames are momentumdominated than when they are influenced by buoyancy. This result contradicts the view that buoyant
instabilities should cause the flame-structures to become more coherent. Although this latter view may be
true at the buoyancy-dominated limit, it appears that
as buoyancy effects first become nonnegligible, the
buoyant acceleration disrupts the KelvinHelmholtz
instability of the jet, and this causes reduced coherence of the turbulent structures.

Acknowledgments
This research was supported under cooperative
agreement NCC3-667 from the NASA Microgravity
Sciences Division. We thank our technical monitor,
Dr. Zeng-Guang Yuan of NCMR, for his hard work in
facilitating the NASA GRC 2.2-s drop tower experiments. Furthermore, we acknowledge useful discus-

sions with Dr. Uday Hegde regarding the effects of

boundary conditions on microgravity flames.

References
[1] H.A. Becker, S. Yamazaki, Combust. Flame 33 (1978)
123149.
[2] H.A. Becker, S. Yamazaki, Proc. Combust. Inst. 16
(1977) 681.
[3] H.A. Becker, D. Liang, Combust. Flame 32 (1978)
115137.
[4] M.Y. Bahadori, D.P. Stocker, D.F. Vaughan, L. Zhou,
R.B. Edelman, Modern Developments in Energy, Combustion and Spectroscopy, Pergamon, Oxford, 1995,
p. 49.
[5] U. Hegde, L. Zhou, M.Y. Bahadori, Combust. Sci.
Technol. 102 (1994) 95100.
[6] U. Hegde, Z.G. Yuan, D.P. Stocker, M.Y. Bahadori, in:
Proceedings of Fifth International Microgravity Combustion Workshop, 1999, p. 259.
[7] U. Hegde, Z.G. Yuan, D.P. Stocker, M.Y. Bahadori,
AIAA Paper 2000-0697, 2000.
[8] M.G. Mungal, J.M. ONeil, Combust. Flame 78 (1989)
377389.
[9] M.G. Mungal, P.S. Karasso, A. Lozano, Combust. Sci.
Technol. 76 (1991) 165185.
[10] W.M. Roquemore, L.D. Chen, L.P. Goss, W.F. Lynn,
Lecture Notes in Engineering, vol. 40, Springer-Verlag,
Berlin/New York, 1989, p. 49.
[11] C.A. Idicheria, I.G. Boxx, N.T. Clemens, AIAA Paper
2001-0628, 2001.
[12] K.L. Page, D.P. Stocker, U.G. Hegde, J.C. Hermanson,
H. Johari, in: Proceedings of Third Joint Meeting of US
Sections of the Combustion Institute, 2003.
[13] S.-J. Chen, W.J.A. Dahm, Proc. Combust. Inst. 27
(1998) 25792586.
[14] C.J. Chen, W. Rodi, in: C.J. Chen (Ed.), Vertical Turbulent Buoyant JetsA Review of Experimental Data,
Pergamon, London, 1980.
[15] W.J.A. Dahm, P.E. Dimotakis, AIAA J. 25 (1987)
12161223.
[16] E.E. Zukoski, B. Cetegen, T. Kubota, Proc. Combust.
Inst. 20 (1984) 361366.
[17] K.M. Tacina, W.J.A. Dahm, J. Fluid Mech. 415 (2000)
2344.
[18] C.A. Idicheria, I.G. Boxx, N.T. Clemens, in: Proceedings of the Spring 2004 Technical Meeting of the Central States Section of The Combustion Institute, 2004.