Espectroscopia de Emisión de Frentes de Llama en Suspensiones de Aluminio

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Proceedings

of the
Combustion
Institute
Proceedings of the Combustion Institute 31 (2007) 2011–2019
www.elsevier.com/locate/proci

Emission spectroscopy of flame fronts


in aluminum suspensions
a,*
Samuel Goroshin , Jorin Mamen a, Andrew Higgins a, Tim Bazyn b,
Nick Glumac b, Herman Krier b
a
Mechanical Engineering Department, McGill University, Montreal, Que., Canada H3A 2K6
b
Mechanical and Industrial Engineering Department, University of Illinois at Urbana Champaign, Urbana, IL 61801, USA

Abstract

Spatially resolved emission spectra from Bunsen-type flames stabilized in aluminum suspensions in air
and oxygen–argon/helium mixtures were obtained using a mechanical-optical scanning system. A low res-
olution (1.5 nm) spectrometer was used to acquire the broad spectra over the 350–1000 nm range, and a
high-resolution (0.04 nm) instrument was used for observation of AlO molecular bands and non-ionized
atomic aluminum. The temperature of condensed phase emitters in the flame was derived using polychro-
matic fitting of the continuum spectra to Planck’s law. AlO temperature was found by fitting of the theo-
retically calculated shape of the band to experimental data. Peak temperatures of the condensed emitters
were found to be approximately 3250 K in aluminum-air flames and approximately 3350 K for oxygen–
argon/helium flames. Temperatures derived from AlO spectra coincide with the temperature of the con-
densed emitters with measurement accuracy and are only 100–200 C lower than the computed equilibrium
flame temperatures. The radial distribution of the temperature profile of the continuous emitters was found
via Abel deconvolution and recovered the double-front structure of the Bunsen flame cone, with the outer
flame being attributed to a diffusion flame of the fuel-rich products with ambient air. The observation of
atomic aluminum lines seen in emission from the outer flame edge and partial self-absorption from the
inner flame confirms the structure associated with the double-front structure. The implications of these
results for the regime of particle combustion in a dust flame are discussed.
 2006 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

Keywords: Spectroscopy; Aluminum; Flame; Dust; Suspension

1. Introduction minum combustion have been conducted with


relatively large, isolated particles in the
Combustion of aluminum has important appli- 20–200 lm size range. In practical applications,
cations in propulsion, pyrotechnics, explosives, however, aluminum often burns as a dense cloud
and in the prevention of industrial dust explo- consisting of micron-scale particles. The mass
sions. To date, most experimental studies of alu- concentrations of particles in such suspensions
are comparable to the density of gas, and the par-
ticle density may reach millions of particles per
*
Corresponding author. Fax: +1 514 398 7365. cubic centimeter. For example, a stoichiometric
E-mail address: samuel.goroshin@mcgill.ca (S. composition of aluminum in air at normal pres-
Goroshin). sure corresponds to a number density of

1540-7489/$ - see front matter  2006 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
doi:10.1016/j.proci.2006.07.175
2012 S. Goroshin et al. / Proceedings of the Combustion Institute 31 (2007) 2011–2019

2 · 106 particles/cm3 for 5 lm particles, with an


average distance between particles of about
80 lm. Even greater particle concentrations are
encountered in propulsion devices that burn at
initially fuel-rich conditions at elevated pressures
(ramjet and hybrid propulsion, underwater pro-
pulsion, etc.) [1,2]. The reaction of particles in
such a dense suspension is influenced collectively
by other particles through global parameters such
as oxygen concentration, average temperature,
and concentration of intermediate gaseous spe-
cies. The particles in a suspension may also be
affected by neighboring particles through distor-
tion of the micro-gradient fields and eventual Fig. 1. Photographs of Bunsen flames in aluminum air
merging of the micro-diffusion flames around the and aluminum–oxygen–helium suspensions stabilized on
particles, similar to the ‘‘clusterization’’ phenome- burner.
non observed in the combustion of volatile sprays
[3]. Such collective effects might lead to quantita-
tive as well as qualitative differences between the velocity, quenching distance etc.) of this alumi-
regimes of combustion for isolated particles and num powder in gaseous oxidizers obtained using
dense particle suspensions. the identical apparatus are reported in Refs.
Taking into account the experimental difficulty [4,5]. In order to clarify the role of nitrogen in alu-
in observing the combustion of individual parti- minum combustion and the effect of gas transport
cles in a dense suspension, the use of a laminar properties on dust flame structure, experiments
flame stabilized in a flow of dusty gas provides a were also performed with synthetic ‘‘air’’ in which
feasible alternative. Unlike the highly transient the nitrogen was replaced with argon or helium.
phenomena associated with particle clouds ignited In air and oxygen–argon mixtures, flames were
by shock waves or in turbulent reacting flows, a stabilized on the conical brass nozzle of the burner
stabilized laminar flame provides a steady-state with an exit diameter of 20 mm. A somewhat
process in which the process of particle combus- smaller nozzle of 18 mm diameter was required
tion is spatially separated, similar to a laminar to prevent flash-back with faster-burning oxy-
flame in gases. Thus, measurement of the spatial gen–helium suspensions. All spectral measure-
variations through a stabilized laminar flame ments were performed at a fixed dust
replaces the necessity of time-resolved measure- concentration of 500 ± 50 g/m3 that corresponds
ments of transient particle combustion. Due to to a fuel-rich mixture with a fuel equivalence ratio
the two-phase nature of the flow, extremely high of / = 1.6 ± 0.2. The laminar flame speed in the
temperatures, and sharp gradients encountered, fuel-rich region has a very weak dependence on
such a study can only be accomplished via non-in- aluminum concentration [5], so operating the
trusive, optical diagnostic techniques. The present burner in this region minimizes fluctuations in
work utilizes spatially resolved emission spectros- the shape and position of the flame during the
copy to examine the structure of a laminar flame 4-second-long scanning procedure (described
in aluminum dust suspensions stabilized on a below).
Bunsen-type burner.
2.2. Spectral scanning system

2. Experimental apparatus A schematic of the scanning system that per-


mitted acquisition of the spatially resolved flame
2.1. Dust burner spectra across the flame cone is shown in Fig. 2.
The flame was imaged using a telescopic system
A Bunsen-type laminar flame was stabilized in (f = 370 mm) mounted on a conventional SLR
a continuous flow of aluminum dust premixed camera with macro T-rings. The camera viewed
with air. The dust dispersion system which gener- the flame via a first surface mirror mounted on a
ates the flow of aluminum suspension is described step motor that rotated in order to sweep the field
in detail elsewhere [4]. Typical laminar flames sta- of view across the flame. Light from the small cen-
bilized on the burner are shown in Fig. 1; the tral point of the magnified (1.5·) image was col-
flame was self-sustained and stable without the lected using a 5-meter-long single strand optical
assistance of a hydrocarbon pilot flame. The alu- fiber mounted at the center of the film plane of
minum used for this study (Ampal Inc., NJ) had the camera and then transmitted to one of two dif-
a narrow particle size distribution and an average ferent spectrometers. A miniature spectrometer
particle size of d32 = 5.6 lm. The particle size dis- (USB-2000, Ocean Optics Inc.) with a fiber diam-
tribution and combustion characteristics (burning eter of 50 lm permits acquisition of spectra in the
S. Goroshin et al. / Proceedings of the Combustion Institute 31 (2007) 2011–2019 2013

Flame The step motor is operated by a square wave


generator that produces a train of a preset number
Protective window
of pulses of adjustable frequency and duration.
Iris aperture The same signal also triggers transfer of data from
100 mm

the spectrometers to the computer-based data


Te lescopic system acquisition system. Real-time data transfer from
250 mm the Ocean Optics spectrometer to the computer
Step allowed acquisition of about 20–25 spectra per
motor Rotating first-surface second. The spatial distance between acquisition
mirror Magnified
flame image
points across the flame was about 250 lm. Thus,
Amplifier about 100 spatial points across the flame were
Optical fiber Reflex acquired in the 4 s required for the sweep. With
50 ms
camera the Hamamatsu detector that allows acquisition
Step-wave of 50 spectra per second, the full scan across the
generator flame can be accomplished in less than 2 s. Scans
of the aluminum-air and aluminum–oxygen–
argon flames were performed at a constant height
Computer data aquisition OceanOptics USB-2000 of 12 mm from the nozzle exit. The shorter flames
or custom UIUC spectrometer in the helium-based mixtures were scanned at a
9 mm height from the nozzle exit. The location
Fig. 2. Schematic of spectral scanning system.
of these scan lines are shown in Fig. 1.

k = 350–1000 nm wavelength range with a spec- 3. Results


tral resolution of about 1.5 nm. To obtain higher
resolution spectra of AlO bands and Al atomic 3.1. General features of the dust flame spectra
lines, a 400 lm fiber was coupled to a UIUC cus-
tom-built 444-mm-focal length Czerny-Turner Examples of spectra acquired by the Ocean
type spectrometer with a 2400 g/mm holographic Optics spectrometer at different distances from
visible grating and Hamamatsu 7031-1032 CCD the center of the flame cone are shown in Fig. 3.
detector. This system acquired spectra within an In the same figure, cutaways show higher resolu-
adjustable 20 nm wavelength range with 0.04 nm tion spectra of Al lines and the AlO Dm = 1 band
spectral resolution. Before each scan, the spec- obtained with the UIUC-built spectrometer. The
trometers were intensity calibrated with a NIST- most prominent feature of aluminum combustion
traceable Ocean Optics LS-1-CAL tungsten spectra is the blue-green X2R+–B2R+ AlO band
halogen calibration lamp. that has been observed previously in spectra of

500
450
b AlO,
400
0.18 350
300
000 a Al 396.2 nm 250
200
800
394.4 nm 150
100
0.1 35 510 513 516 519
600

Na
400 D = 5.75 mm

200 0.09

394 396
Light intensity

Al
45 D = 0.50 mm

D = 10.00 mm

0
400 500 600 700 800 900
Wavelenth, nm

Fig. 3. Spectra at different distances from the center of the flame cone for aluminum-air obtained with the low resolution
(1.5 nm) spectrometer. Cutaway boxes show Al vapor lines and the AlO Dm = 1 band obtained with the high-resolution
(0.04 nm) spectrometer.
2014 S. Goroshin et al. / Proceedings of the Combustion Institute 31 (2007) 2011–2019

Air O - Ar O - He Here, e(k, T) is the spectral emissivity of the con-


C1 = 0.5954 · 1016 W m2,
2 2
densed emitter,
Normalized light intensity

2
C2 = 1.4388 · 10 m K. For so-called ‘‘grey’’
emitters, where e(k, T) is independent of wave-
length, the left-hand side of Eq. (1) is a linear
function of 1/k, with the slope proportional to
1/T.
As can be seen in Fig. 5a, however, the exper-
imental light intensity data plotted in the form of
Eq. (1) deviate considerably from a straight line,
0 5 10 15 0 5 10 15 0 5 10 15
suggesting that the condensed phase emitters are
Distance from the
flame center, mm
not grey. As was first suggested by Wolfgard
and Parker [8], a strong spectral dependence of
Fig. 4. Intensity profiles plotted versus distance across aluminum flame emissivity can be expected
the flame for air, argon, and helium-based mixtures because the size of the aluminum oxide particles
(k = 600 nm). (which are the major source of the continuum
spectrum in aluminum flames) is smaller
(0.1 lm) than the wavelength of light [9]. The
various aluminum combustion systems [6]. In light emission cross-section for such particles
comparison to the combustion spectra from a sin- decreases with wavelength as 1/k (Rayleigh limit
gle particle or dilute dust cloud [7], however, in [10]). At the same time, the absorption index of
the dense-suspension dust flame, this system is the molten bulk aluminum oxide also decreases
seen superimposed on top of a much stronger con- approximately as 1/k [9] in the 400–1000 nm spec-
tinuous background emission. The ratio of the tral range, suggesting that the spectral emissivity
intensity of AlO lines to the intensity of the con- of an oxide particle is inversely proportional to
tinuum background reaches a maximum in a zone k2. Indeed, with this assumption, a good linear
that can be identified as the beginning of the flame fit is obtained for experimental data plotted in
front and gradually decreases toward the periph- accordance to a modified Eq. (1), as seen in
ery of the flame. Two prominent Al vapor lines Fig. 5b.
(394.40 and 396.15 nm) are seen in emission and The temperature profiles of condensed phase
have a relatively small bandwidth at the periphery emitters across the flame radius, as computed by
of the flame but expand and demonstrate self-ab- the procedure described above, are shown in
sorption in regions close to the flame front (as dis- Fig. 6. Due to the strong dependence of spectral
cussed below). intensity on temperature in Planck’s law
Typical spatial scans of light intensity along (ik,T  exp(24,000/T) at k = 600 nm), the mea-
the radius of the flame cone for air, argon, and sured temperatures can be attributed to the maxi-
helium-based mixtures are shown in Fig. 4 at a mum temperature encountered along the line of
wavelength of k = 600 nm. The aluminum-air site through the flame, even in the case of a mod-
and aluminum–oxygen–argon mixtures show erately optically thick flame.
sharp gradients in intensity at both the inner
and outer edges of the flame, while the helium-
based mixture shows a broad, less steep luminos-
ity profile with a less defined flame boundary.
ln(i ln(i
-10
3.2. Deriving temperature from continuous spectra 1 a -11 b
The continuous part of the flame spectra was 0 -12
used to derive the temperature of the condensed D = 5.60 mm
-13
emitters in the flame. This was achieved using -1
so-called polychromatic fitting of the low resolu- -14
D = 7.40 mm
tion continuum spectra in the 400–900 nm spec- -2
-15
tral range to Planck’s law.
-3 -16
The dependence of light intensity ik,T on wave-
length and temperature emitted by a condensed -4
-17
phase emitter (solid or liquid particles) is -18
described by Planck’s law, which for the range 0.0014 0.00175 0.0021 0.00245 0.0014 0.00175 0.0021 0.00245

of wavelength and temperatures encountered here 1/ nm -1 1/ nm-1


can be written in this form
Fig. 5. Linear fitting of the spectral intensity to Planck’s
  law with the assumption that (a) e(k, T) = constant and
ik;T  k5 C2
ln  2p  C 1 ¼ ð1Þ (b) e(k, T)  1/k2. The slope of the data is proportional
eðk; T Þ kT to 1/T.
S. Goroshin et al. / Proceedings of the Combustion Institute 31 (2007) 2011–2019 2015

3600 Air 21%O +79%Ar 3600 21%O2+79%He


2
Temperature, K

3400 3400

3200 3200

3000 3000

2800 2800

0 2 4 6 8 10 0 2 4 6 8 10 0 2 4 6 8 10

Distance from the flame center, mm

Fig. 6. Temperature distribution across the flame derived from the raw continuous spectra.

The rise in temperature at the outside edge of sity distribution at 600 nm was shown in Fig. 4).
the flame observed in all mixtures in Fig. 6 is The deconvolved spectra were then used to extract
attributed to the existence of a secondary flame temperature at each point in the flame profile
front corresponding to a diffusion flame between using the polychromatic fitting described above.
the vaporized aluminum or aluminum sub-oxides The distributions of temperature and intensity of
produced by the fuel-rich premixed flame and luminosity at 600 nm are shown in Fig. 7 for air,
the ambient air surrounding the burner. It is inter- argon, and helium based mixtures. Comparison
esting to note that this front is not readily visible of the flame temperatures derived from the raw
in the light intensity profile across the flame (Fig. 6) and deconvolved data (Fig. 7) shows that
(Fig. 4), where it appears as a small plateau. the Abel transform does not significantly change
Estimations of the optical depth of the flame as the range of measured temperatures; this result
well as an appearance of the continuous spectra can be taken as indirect confirmation of the small
indicate that the flame in the spectral regions out- optical thickness of the flame outside the range of
side molecular bands and atomic lines is optically molecular and atomic lines.
thin. With this assumption, the measured light
intensities (which are integrated along the line of 3.3. Deriving temperature from AlO molecular
site) can be used to extract the true distribution species
of the intensities of luminosity emitters over the
flame profile using an Abel transform (deconvolu- Temperatures of the gaseous aluminum sub-
tion) procedure for a cylindrically symmetrical oxide AlO were derived by fitting the spectral
object. The three-point Abel transforms [11] were shape of the AlO Dm = 1 molecular band
performed with a polynomial approximation of obtained with the high-resolution UIUC spec-
the radial distribution of luminosity at each of trometer (see Fig. 2) to a theoretically calculated
the 2048 discrete wavelengths measured by the spectra. This particular band of AlO has been
spectrometer (an example of the raw radial inten- chosen due to the sensitivity of its shape to the

3600 0.0025 0.001


3600 0.0012

Air 21%O2-79%Ar 21%O2-79%He


Light intensity (arbitrary units)

3400
3400 0.002 3400
0.0008 0.001
Temperature, K

3200
3200 0.0015 3200

0.0006 0.0008

3000
3000 0.001
3000

0.0004 0.0006

2800
2800 0.0005
2800

0.0002 0.0004

2600
2600 0
2600

3 4 5 6 7 8 9 10 11 3 4 5 6 7 8 9 10 11 3 4 5 6 7 8 9 10 11
Distance from the flame center, mm

Fig. 7. True distributions of light intensity and temperature of condensed sources across the flame as derived from Abel
deconvolved continuous spectra.
2016 S. Goroshin et al. / Proceedings of the Combustion Institute 31 (2007) 2011–2019

3400 34 00
5
34 00
5
5
Air 21%O2+79%Ar 21%O2+79%He

4 4
4
3300 33 00 33 00
Temperature, K

Optical depth
3 3 3

3200 32 00 32 00

2 2 2

3100 31 00 31 00

1 1 1

3000 30 00
0 0
30 00 0
0 2 4 6 8 10 12 0 2 4 6 8 10 12 0 2 4 6 8 10 12
Distance from the flame center, mm

Fig. 8. Temperature distribution and optical depth (unscaled) across the flame as derived from raw AlO spectra.

temperature over the range encountered in alumi- mixtures and were attributed to non-ionized alu-
num combustion. The theoretical shape of AlO minum vapor. The shape of the lines, however,
spectra is reconstructed following the procedure changes across the flame profile and was also dif-
of Arnold et al. [11] using the spectroscopic con- ferent for air and helium mixtures. Fig. 9 illus-
stants of Coxon and Naxakis [12], and the band trates the changes in the appearance of the
intensities of Partridge et al. [13]. Five indepen- aluminum lines as observed with the high-resolu-
dent parameters (temperature, optical path length, tion spectrometer at different points in an alumi-
a general intensity scaling parameter, and two lin- num-air flame. The thin, 1.5- to 2-mm-thick
ear parameters describing the non-reasonant zone (A) that is close to the outside edge of the
background) were optimized to produce the fit flame emits aluminum lines that show no evidence
spectrum. A detailed description of this procedure of self-absorption. The absence of self-absorption
can be found in Ref. [7]. The resulting tempera- in a gradient media usually corresponds to a rising
tures and optical thicknesses derived from the temperature profile in the direction towards the
spectra acquired across the profile of the flame observer and correlates with the existence of the
are shown in Fig. 8. external diffusion flame seen in the temperature
While the observed AlO spectra are the result measurements. The broad 5- to 6-mm-thick inter-
of integrated emissions along the line of view nal zone (B) shows evidence of self-absorption at
through the flame, the single emission temperature line peaks indicating the change in the tempera-
that was fit to the data is probably close to the ture gradient. In this zone, temperature decreases
maximum temperature encountered along the line in the direction toward the observer (see tempera-
of view in the moderately optically thick flames in ture profile in Fig. 7). Thus radiation emitted by
air and argon mixtures. Justification for this the hotter layers closer to the flame front has to
assumption is similar to that provided above for pass through relatively cold media at the periph-
the continuous spectra and is based on the strong ery of the zone, leading to partial self-absorption
dependence of the band intensity on temperature
(see Ref. [7] for details). For the flame in helium
mixtures, however, where the optical thickness is 300

much larger, the screening of high temperature 250


A
radiation by the colder external layers may be 200

considerable. This might explain the somewhat 150

lower temperature measured in helium flames in R, mm 100

comparison to argon mixtures, while thermody- 15


50
350

B
namics predicts that the equilibrium flame temper-
300

250

atures should be identical for these two mixtures. A


200

Due to the optical thickness of the flames in the 10


B
150

AlO spectral bands, no attempts have been made 100


550

to recover the true spatial distribution of AlO 500 C


C 450

emitters across the flame profile using the Abel 5 400

deconvolution procedure. 350

300

250

0
3.4. Observing atomic Al vapor lines
200

Fig. 9. Appearance of atomic aluminum lines in differ-


Strong lines appearing at 394.40 and ent zones of the aluminum-air flame (flame cone viewed
396.15 nm were detected within flames in all three from above in schematic).
S. Goroshin et al. / Proceedings of the Combustion Institute 31 (2007) 2011–2019 2017

of the atomic aluminum lines. In the thin 1- to 1.5- helium flames can be attributed to the greater
mm-thick zone (C) at the inner edge of the flame flame speed in helium mixtures (approximately
front, the central part of the aluminum lines can three times greater than flame speed in air- and
be seen in absorption, with the intensity dripping argon-based mixtures [5]) that results in the tem-
below the background of the continuous spectra, perature profile being spatially stretched out.
indicating that the temperature of the condensed The greater thermal diffusivity of helium mixtures
emitters is greater than the temperature of atomic may also contribute to the shallower temperature
Al. This zone was not observed in oxygen–helium gradients.
flames, which in general showed weaker self-ab- The experimental observations of optical
sorption, consistent with the observation of small- thickness of AlO in the flames shown in Fig. 8 also
er temperature gradients in the helium flames correlate with thermodynamic calculations.
(Fig. 7). According to equilibrium calculations, the con-
centration of AlO sub-oxides in the products is a
strong function of the flame temperature, and
4. Discussion AlO concentration doubles in argon and helium
based mixtures as compared to flames with air.
4.1. Comparison of results with equilibrium Roughly the same ratio between optical thickness
calculations of AlO in air and argon is observed experimental-
ly (Fig. 8). The larger optical thickness of AlO
The results of equilibrium calculation of the observed in helium flames in comparison to argon
flame temperatures and concentrations of major mixtures is due to the smaller temperature gradi-
species for aluminum-air and aluminum–oxygen– ent in the helium flame that results in greater
argon (or helium) suspensions at an average fuel AlO average concentrations.
equivalence ratio of / = 1.6 are shown in Table Finally, presence of large quantities of free
1. The calculated flame temperatures demonstrate aluminum vapor in a relatively cold region of
very good agreement with the experimental data. aluminum-air flame indicated by the strong self-
The maximum flame temperatures derived from absorption of aluminum atomic lines is also in
the continuum spectra are typically 100–150 C agreement with thermodynamic calculations that
lower than the thermodynamically computed adi- predict a very little bonding of aluminum with
abatic flame temperatures (see Fig. 7). The differ- nitrogen. This also might suggest that the compo-
ence is somewhat higher (200–250 C) for the sition of the condensed aluminum oxide formed is
temperatures derived from AlO spectra shown in close to stoichiometric.
Fig. 8. Taking into account that both measure-
ments of temperature are estimated to have an 4.2. Regimes of particle combustion in flame
error of about 2%, the values coincide within
experimental accuracy. Unlike solid aluminum Identification of the transition from a diffusive-
oxide at low temperature, liquid aluminum oxide ly to a kinetically controlled combustion mecha-
particles at high temperature are strong emitters nism as the aluminum particle size is reduced, as
[9]. Thus, strong radiative heat loss is probably has been predicted by several researchers [7,14],
the main cause of the lower than adiabatic flame was a major initial motivation for the current
temperatures and relatively sharp cooling of the work. The Damköhler number that characterizes
products observed in air and argon mixtures. the ratio of kinetic to diffusive resistance and
The weaker temperature gradients observed in defines the particle combustion regime is propor-
tional to the particle diameter and inversely pro-
portional to the oxygen diffusivity of the
Table 1 mixture: Da  do/D [14]. Thus, in terms of the
Calculated equilibrium flame temperatures and concen- Damköhler number, varying the mixture diffusiv-
tration of major species in fuel-rich aluminum flames ity is equivalent to varying the particle size (pro-
(U = 1.6) vided that it does not affect kinetics). In respect
Air 21% O2 + 79% Ar (He) to kinetics and thermodynamics, oxygen–argon
and oxygen–helium mixtures having the same
T (K) 3417 3566
oxygen content are equivalent. Due to the differ-
O 0.6 2.4
AlO 2.3 4.5 ence in oxygen diffusivity, however, the Damköh-
Al2O 7.8 7.2 ler number in helium is 2.7 times smaller than in
Al2O2 1.4 1.7 the argon mixture. Our previous work has indicat-
Al 6.0 6.3 ed that the experimentally measured ratio of flame
Al2O3 74.5 64.4 propagation velocities in helium and argon-based
Concentrations of gaseous components are given in % of mixtures was smaller than predicted by theory
the total pressure and concentration of condensed alu- assuming particles burn in the diffusive regime
minum oxide is shown in % from the maximum stoi- and larger than predicted if particle combustion
chiometric value in cold products. is controlled by reaction kinetics in both mixtures
2018 S. Goroshin et al. / Proceedings of the Combustion Institute 31 (2007) 2011–2019

[5]. This result might suggest a transition in the difficult. Thus, combination of several optical
combustion regime when argon is replaced by investigative techniques such as interferometry,
helium. PLIF, etc. would be needed to fully resolve the
For fuel-lean flames, with fuel concentrations flame structure and directly measure local temper-
significantly below stoichiometric, the diagnostic ature gradients, thereby permitting the particle
of the combustion regime would be straightfor- combustion regime to be identified.
ward. In this case, experimentally measured tem-
peratures considerably in excess of the
equilibrium flame temperature would unambigu-
Acknowledgments
ously indicate the existence of ‘‘hot spots’’
(micro-diffusion flames) associated with diffu-
The authors gratefully acknowledge support of
sion-controlled particle combustion. For the mod-
this work at UIUC provided by the Office of Na-
erately rich aluminum flames used in the present
val Research under direction of Dr. Judah Gold-
study, if micro-diffusion flames exist, they would
wasser and by the Canadian Space Agency at
not exceed the flame temperature that is already
McGill University under direction of Dr. Marcus
close to the maximum, stoichiometric value. In
Dejmek.
addition, depletion of oxygen in fuel-rich flames
further decreases the temperature difference
between the micro-diffusion flame and surround-
ing gas, implying that temperature inhomogeniety References
in fuel-rich flames are small and are difficult to
observe even in diffusion-controlled particle com- [1] S. Goroshin, A.J. Higgins, M. Kamel, in: 39th
bustion. Nevertheless, the higher temperature of AIAA/ASME/SAE/ASEE Joint Propulsion Con-
the solid phase in comparison to gas observed in ference and Exibit, AIAA Paper 2001-3919.
a thin zone close to the reaction front in alumi- [2] A.G. Egorov, E.D. Kal’nei, A.P. Shaikin, Combut.
num-air flames, indicated by absorption of contin- Expl. Shock Waves 37 (5) (2001) 516–522.
[3] K. Annamalai, W. Ryan, Prog. Energy Combust.
uum background spectra by aluminum vapor
Sci. 18 (1992) 221–295.
lines, suggests the existence of such microflames [4] S. Goroshin, M. Bidabadi, J.H.S. Lee, Combust.
in air. The absence of this phenomenon in helium Flame 105 (1996) 147–160.
mixtures points, accordingly, to its temperature [5] S. Goroshin, I. Fomenko, J.H.S. Lee, Proc. Com-
homogeneity that is characteristic of kinetic bust. Inst. 26 (1996) 1961–1967.
combustion. [6] P. Bucher, R.A. Yetter, F.L. Dryer, T.P. Parr,
D.M. Hanson-Parr, E.P. Vicenzi, Proc. Combust.
Inst. 26 (1996) 1899–1908.
5. Conclusion [7] N. Glumac, H. Krier, T. Bazyn, R. Eyer, Combust.
Sci. Technol. 177 (2005) 485–511.
[8] H.G. Wolfhard, W.G. Parker, Proc. Phys. Soc. A
The current work demonstrates the ability of 62 (8) (1949) 523–529.
emission spectroscopy to extract important infor- [9] D.L. Parry, M.Q. Brewster, J. Thermophys. Heat
mation on two-phase flame structure. The mea- Transf. 5 (2) (1990) 142–149.
sured condensed and gas phase temperatures and [10] C.F. Bohren, D.R. Huffman, Absorption and
observations of gaseous species in flames correlate Scattering of Light by Small Particles, Wiley, New
well with thermodynamic predictions. Emission York, 1983, p. 544.
spectroscopy, however, is by no means a universal [11] J.O. Arnold, E.E. Whiting, G.C. Lyle, J. Quant.
tool for the study of the dust flame structure. The Spectrosc. Radiat. Transf. 9 (6) (1969) 775–798.
[12] J.A. Coxon, S. Naxakis, J. Mol. Spectrosc. 111 (1)
extremely high flame temperatures encountered in
(1985) 102–113.
aluminum suspensions result in the formation of [13] H. Partridge, S.R. Lengsfield, B. Liu, J. Quant.
large quantities of intermediate combustion prod- Spectrosc. Radiat. Transf. 30 (5) (1983) 449–462.
ucts, the recombination of which produces long [14] R.A. Yetter, F.L. Dryer, in: H.D. Ross (Ed.),
‘‘tails’’ that make flames optically thick and identi- Microgravity Combustion: Fire in Free Fall, Aca-
fication of the reaction (non-equilibrium) zone demic Press, New York, 2001, p. 448.

Comments

Igor Rahinov, Tel-Aviv University, Israel. As far as I Reply. Abel deconvolution does indeed produce in
know the edge point in the Abel reconstruction usually some cases non-physical jumps before the flame edge.
tends to be inaccurate. Can it be the reason for some In such cases, the deconvolved signals were truncated pri-
artifact that causes this edge temperature ‘‘jump’’ I or to the flame front. The deconvolved results reported in
saw in your graphs? the paper were obtained with different deconvolution
S. Goroshin et al. / Proceedings of the Combustion Institute 31 (2007) 2011–2019 2019

algorithms and are believed to be largely independent on d


the particular Abel deconvolution technique used.
Trygve Skjold, University of Bergen, Norway. Would
d you expect the measured laminar burning velocities to be
scale dependent due to radiative heat loss?
Arthur Fontijn, Rensselaer Polytechnic Institute,
USA. Al and ALO have well known spectra. You also Reply. We have found in our previous work [1] that
show numbers for Al2O2 and Al2O3. These have no burning velocity derived from aluminum Bunsen-type
known electronic spectra as far as I recall. How did flames actually increases with the decrease in flame
you handle these species? In case you used a continuum diameter (nozzle size) and with increase of the flow rate
for Al2O3, how certain can you be that there are no and therefore the flame length. The relative radiation
interfering continua; did you check this? heat loss is larger for smaller and longer flames, and thus
one would expect an opposite outcome if radiation heat
Reply. The numbers shown in Table 1 for concentra- loss dominates the scaling of burning velocity. Probably
tions of condensed product (Al2O3) and gaseous species other effects, such as flame curvature, have greater effect
(AlO, Al2O, and Al2O2) are the result of the thermody- on flame propagation speed for such small-size flames
namic equilibrium calculations not experimental mea- than radiation heat loss. However, for considerably larg-
surements. In the experiment, we observed only the er flames (tens of centimeters to meters) radiation might
known molecular bands belonging to AlO. Thermody- play a more significant role in the flame propagation
namic calculations of the flame equilibrium also show mechanism, and thus for such large flames the scaling
that no other condensed species besides Al2O3 are tendency might be different.
formed in the flame in significant numbers. This was
experimentally confirmed in our previous work ([1] in pa-
per) with the analysis of the condensed product collected Reference
from the Bunsen aluminum flame at different fuel equiv-
alence ratios ([1] in paper). Experimental data show that [1] S. Goroshin, M. Kolbe, J. Lee, J. Burning velocity
submicron aluminum oxide comprises more than 90% of measurements in aluminium-air suspensions using
the total mass of the condensed product and formation of stabilized dust flames, in: Proceedings of the 18th
aluminum nitrides is insignificant. Thus, as Al2O3is the International Colloquium on the Dynamics of Explo-
dominant condensed phase product, there are not be- sions and Reactive Systems, Seattle, 2001 (http://
lieved to be any interfering continuous emitters. www.engr.washington.edu/epp/icders/).

You might also like