You are on page 1of 17

Combustion and Flame 154 (2008) 164180

www.elsevier.com/locate/combustame
Reduction of soot emissions by iron pentacarbonyl in
isooctane diffusion ames
K.B. Kim, K.A. Masiello, D.W. Hahn

Department of Mechanical and Aerospace Engineering, University of Florida, Gainesville, FL 32611, USA
Received 8 August 2007; received in revised form 20 December 2007; accepted 21 January 2008
Available online 3 April 2008
Abstract
Light-scattering measurements, in situ laser-induced uorescence, and thermophoretic sampling with transmis-
sion electron microscopy (TEM) analysis, were performed in laboratory isooctane diffusion ames seeded with
4000 ppm iron pentacarbonyl. These measurements allowed the determination of the evolution of the size, number
density, and volume fraction of soot particles through the ame. Comparison to unseeded ame data provided
a detailed assessment of the effects of iron addition on soot particle inception, growth, and oxidation processes.
Iron was found to produce a minor soot-enhancing effect at early residence times, while subsequent soot particle
growth was largely unaffected. It is concluded that primarily elemental iron is incorporated within the soot parti-
cles during particle inception and growth. However, iron addition was found to enhance the rate of soot oxidation
during the soot burnout regime, yielding a two-thirds reduction in overall soot emissions. In situ spectroscopic
measurements probed the transient nature of elemental iron throughout the ame, revealing signicant loss of
elemental iron, presumably to iron oxides, with increasing ame residence, suggesting catalysis of soot oxidation
via iron oxide species.
2008 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
Keywords: Soot oxidation; Fuel additives; Iron pentacarbonyl; Soot emissions control
1. Introduction
Particulate matter (PM) is the term describing
small particles found in the ambient air, such as dust,
marine-derived particles, liquid droplets, smog com-
ponents, and soot. PM ranges in size from a few
nanometers to tens of micrometers. Because of the
small size of these particles (notably PM2.5, which
consists of particles less than 2.5 m), they are able
to permeate and impact the deepest parts of the lungs,
and are associated with incidences of asthma, chronic
*
Corresponding author. Fax: +1 352 392 1071.
E-mail address: dwhahn@u.edu (D.W. Hahn).
bronchitis, and heart disease. Therefore, various ef-
forts have been made to search for possible solutions
to decreasing the production rates of such ne parti-
cles.
Soot particles, which compose a signicant por-
tion of PM2.5, are rich in amorphous carbon and
polycyclic aromatic hydrocarbons (PAHs), are well
known to be carcinogenic [1], and can play key roles
in the global climate due to their inuence on the
planets radiative transfer. Because soot from com-
bustion processes is a major source of PM, there has
been signicant interest in studying soot formation
mechanisms and methods of soot reduction. While
the reduction of harmful soot emissions from com-
0010-2180/$ see front matter 2008 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
doi:10.1016/j.combustame.2008.01.011
K.B. Kim et al. / Combustion and Flame 154 (2008) 164180 165
bustion processes is desirable, overall combustor per-
formance (notably for propulsion systems) cannot be
compromised, and large-scale combustor modica-
tions may not be practical in many situations. With
such goals and constraints in mind, the reduction of
soot emissions via fuel additives remains an attractive
path for PM emissions control. This paper will focus
on metallic-based additives, namely iron-based com-
pounds.
The use of metallic fuel additives in combustion
applications has a long and at times controversial his-
tory. For example, from the rst half of the previ-
ous century through the 1970s, tetraethyl lead was a
popular additive in gasoline to enhance octane levels
(i.e., reduce engine knock). However, due to envi-
ronmental and health concerns, the EPA ultimately
banned leaded gasoline for on-road vehicles. More
recently, MMT (methylcyclopentadienyl manganese
tricarbonyl) has continued to be used as a fuel additive
for gasoline in many countries, although it remains
of interest to regulatory agencies, given the listing of
manganese as a Clean Air Act metal.
As an alternative to additives containing Pb and
Mn (which target combustion performance), metal-
lic additives have also been explored with the goal
of altering sooting characteristics. Such additives in-
clude the alkali metals Li, Na, K, and Cs and the
alkaline earth metals Ca, Sr, and Ba. In addition, tran-
sition metals are of interest, particularly Mn and Fe.
The history of iron-based additives dates back to the
1960s, with Shayeson demonstrating a reduction in jet
engine exhaust smoke with several additives, includ-
ing iron compounds [2]. A comprehensive summary
of the early additives work, including discussion of
practical combustors and laboratory ames, is pre-
sented by Howard and Kausch, in which Fe and Mn
are noted among the most effective additives under
heavily sooting conditions [3]. In general, additive
performance results vary signicantly, depending on
fuel type, additive concentration, stoichiometry, and
combustor conguration. Interest in the current study
is limited to iron, due primarily to previous research,
summarized below, that suggests an important role
in soot control, as well as the relatively benign na-
ture of iron with respect to environmental and health
concerns. Pertinent literature regarding iron-based ad-
ditives is briey summarized below, with attention
given to sootiron interactions and iron speciation.
Several laboratory studies using laminar, premixed
ames have explored the roles of iron, as either fer-
rocene Fe(C
5
H
5
)
2
or iron pentacarbonyl Fe(CO)
5
,
in hydrocarbon ames. Ritrievi et al. [4] studied
laminar premixed ethylene ames seeded with fer-
rocene, Fe(C
5
H
5
)
2
, for dopant concentrations of
0.0050.14% Fe by weight of fuel, and for ame C/O
ratios of 0.710.83. In studies of a similar premixed
ethylene ame seeded with ferrocene conducted by
Feitelberg et al. [5], iron was added to the fuel in
200 ppm concentrations on a molar basis (0.13% Fe
by weight of fuel). Hahn and Charalampopoulos [6]
studied a premixed propane ame seeded with iron
pentacarbonyl, Fe(CO)
5
, with fuel equivalence ratios
of 2.4 and 2.5. Iron pentacarbonyl was added in con-
centrations of 0.160.32% by weight of iron to the
fuel. While exact results vary somewhat with fuel type
and stoichiometry in the above studies, the consistent
ndings were that the addition of iron tended to in-
crease the amount of soot formed (i.e., soot volume
fraction), including increases in both soot particle
size and number density. These results are all con-
sistent with previous work performed by Bonczyk [7]
and Haynes et al. [8]. The consensus of such studies
is that iron-based compounds nucleate prior to soot
inception, providing increased surface area for soot
formation and growth, ultimately leading to enhanced
soot emissions, although questions were raised as to
the exact nature of the iron-based compounds and fate
of iron species. In efforts to answer these questions,
various means of modeling, extractive sampling, and
chemical analysis were explored in the above stud-
ies to determine the chemical states of iron within the
soot.
The research groups discussed above reported
combinations of Fe and FeO [4] and Fe
2
O
3
[6],
based on extractive sampling and subsequent chemi-
cal analysis, although no in situ measurements have
been made to date. The overall conclusion of the
above studies is that any soot suppression effects must
be limited to the soot burnout (i.e., oxidation) regions
of the ame, which are absent in the premixed ame
congurations. In actual combustors, the nal stage
in the soot emission cycle is generally soot oxidation
or burnout, driven by the diffusion of additional oxy-
gen following primary combustion. In this regime, the
soot particles are partially or completely destroyed,
generally via OH and O
2
oxidation, yielding CO or
CO
2
as products. Therefore, while at-ame burn-
ers are an excellent laboratory tool, further attention
must be given to alternative ame congurations that
are closer to the practical combustor congurations in
which diffusion ames often play a key role.
In a study using an isooctane diffusion ame op-
erating above its smoke point, Bonczyk [9] noted
both soot-enhancing and -suppressing characteristics
of ferrocene added in concentrations of 0.09% Fe by
weight of fuel. At early ame residence times, the
soot particle size and number density all increased,
resulting in enhanced soot production similar to that
in premixed ames. However, these same parameters
were observed to decrease through the soot burnout
regimes of the ame, with the net effect being a re-
duction in soot when compared to unseeded ames.
166 K.B. Kim et al. / Combustion and Flame 154 (2008) 164180
Postame sampling and subsequent chemical analy-
sis revealed that Fe
2
O
3
was the primary condensate
phase. The study concluded that iron might be ini-
tially reduced (via reaction with solid carbon) from
an oxide to elemental iron within the soot particles,
eventually enhancing carbon deposition due to the
catalytic effects of Fe. At later residence times, ele-
mental iron might be oxidized, for example to form
Fe
2
O
3
. The net result of this cycle is the oxidation of
carbon to CO.
Such a method of iron reduction and subse-
quent oxidation is also supported by Zhang and
Megaridis [10], who studied an ethylene diffusion
ame seeded with ferrocene, as well as by Kasper
et al. [11], whose investigation included ferrocene-
seeded methane/argon and acetylene/argon ames. In
particular, Zhang and Megaridis [10,12] performed
detailed chemical microanalysis of sampled soot par-
ticles, verifying the presence of elemental iron within
the soot agglomerate matrix for particles sampled
near the fuel-rich burner axis, while the presence of
both iron and oxygen suggestive of iron oxides was
found in aggregates sampled from other parts of the
ame. They concluded that iron nucleation prior to
soot formation and subsequent incorporation of iron-
rich nuclei into the soot matrix was a critical compo-
nent for soot suppression. A related study of the role
of iron addition in ame inhibition in both premixed
and laminar ames offered additional insight into iron
nucleation [13,14]. The authors reported the forma-
tion of iron-rich particles early in the ame, where
their calculations predicted the presence of Fe, FeO
2
,
FeO, Fe(OH)
2
, and FeOH. Flame residence time, lo-
cal stoichiometry, and temperature were all found to
be inuential parameters, although the former was the
most important factor.
There is not a signicant amount of literature
bridging the gap between practical combustors and
laboratory ames. Toward this end, the effects of fuel
specication and fuel additives on soot formation was
reported in complex ow swirl combustors, including
the effects of ferrocene addition to JP-8, isooctane,
and blends of isooctane [15]. The study reported dis-
parate soot suppressing effects with fuel type, and
further concluded that simple smoke point measure-
ments are not good predictors of additive performance
in a complex ow eld. A more contemporary study
reported similar ndings, namely that the use of the
smoke lamp for assessing the soot-reducing poten-
tial of metal-containing additives is not effective [16].
Iron-based additives (ferrocene and iron naphthenate)
were found to reduce soot emission only when oxy-
gen was available, thereby supporting the hypothesis
of soot-incorporated iron species as oxidative cata-
lysts.
It is important to note that the ndings charac-
terized above are readily summarized by an increase
in the burnout efciency of soot particles in seeded
ames, not by an inhibition of soot formation with the
introduction of iron. In fact, seeded ames are likely
to realize peak soot production levels higher than
those in unseeded ames due to increased surface area
for soot formation in the particle inception regimes, as
quantied in the premixed ame studies noted above.
More recent studies have added to these ndings. Kim
et al. [17] performed a unique set of experiments to
explicitly quantify the oxidation properties of ame-
generated soot in the presence of metallic additives.
They found that the addition of iron (as iron pentacar-
bonyl) reduced the activation energy of soot oxidation
from 162 to 116 kJ/mol. Mass spectrometry of col-
lected soot particles revealed strong signals indicative
of elemental iron, although relatively weak signals of
FeO
+
, Fe
2
O
+
, and Fe
2
O
+
3
were also reported. To
summarize, the consensus of the literature is that the
net reduction of soot is realized by apparent enhance-
ments in soot oxidation throughout the burnout (i.e.,
oxidation) regime, including the catalysis of solid car-
bon oxidation by various combinations and cycles of
Fe and iron oxides [4,6,7,9,18].
An important point is made regarding the addi-
tive studies reviewed above, namely, that all analyses
of soot species were ex situ (i.e., based on extrac-
tive sampling); hence all methodologies are subject
to possible changes in iron states when soot sam-
ples are removed from the true ame environment
prior to analysis. Therefore, a thorough understand-
ing of the role of iron additives on soot emissions
is still somewhat limited by the unknown nature, in-
cluding transient behavior, of the iron species within
the ame environment. The current study involves de-
tailed measurements of the soot inception, growth,
and soot oxidation regimes using diffusion ame con-
gurations, combined with in situ diagnostics, with
the goals of further understanding the roles and mech-
anisms of iron and iron speciation in soot suppression.
2. Experimental methods
2.1. Burner systems
All experiments were performed using prevap-
orized isooctane/oxygen diffusion ames, either un-
seeded or seeded with iron pentacarbonyl, Fe(CO)
5
.
Isooctane (HPLC-grade, Fisher Scientic) and iron
pentacarbonyl (99.5%, Alfa Aesar) were generally
mixed just prior to use to prevent any dissociation
of the iron pentacarbonyl. Overall, isooctane provides
a liquid fuel in which iron pentacarbonyl is readily
soluble, thereby ensuring reliable introduction of the
K.B. Kim et al. / Combustion and Flame 154 (2008) 164180 167
Fig. 1. Concentric diffusion burner schematic (side view and top view). See Table 1 for dimensions of Burner 1 and Burner 2.
additive. Furthermore, the current use of isooctane is
complementary to the detailed work of Bonczyk [9],
although the wide variety of fuel types explored in
the literature (see above) provides no clear consensus
choice for fuel, but rather, the results tend to follow
the ame conguration (i.e., diffusion vs premixed).
Two different concentric tube burners were em-
ployed to investigate the soot inception, growth, and
oxidation regimes with sufcient spatial resolution.
The rst system, designated Burner 1, was specif-
ically designed for studying the soot inception and
growth regimes, while the second system, designated
Burner 2, was designed for exploring the soot burnout
regime. A representative schematic of the burners is
shown from side and top views in Fig. 1, and the exact
dimensions of the two burners, along with ame pa-
rameters, are summarized in Table 1. The liquid fuel
was prevaporized using a vaporization system con-
sisting of a stainless steel tube, packed with stainless
steel balls, maintained at 100

C, which was above the


boiling point for the liquid isooctane (99

C). A nitro-
gen coow stream was introduced into the vaporiza-
tion system at the head of the preheat zone at a rate
of 0.8 l/min to assist in maintaining a steady output.
For Burner 1, the liquid fuel was delivered to the va-
porization system via a peristaltic pump at a rate of
1 ml/min, which corresponds to an isooctane vapor
ow rate of 0.18 l/min at atmospheric pressure. The
mixture of fuel vapor and nitrogen was then passed
through the center tube of the burner. A ow rate of
9 l/min of pure oxygen was passed through the outer
concentric tube, resulting in a laminar, sooting diffu-
sion ame. The resulting ame diameter was about
Table 1
Concentric diffusion burner dimensions and ame parame-
ters
Burner 1 Burner 2
Inner tube ID
(fuel)
7.04 mm 1.5 mm
Outer tube ID 16.56 mm 6.96 mm
Array tubes
(oxygen)
2.54 mm ID 0.3 mm ID
12 holes 9 holes
Flame stabilization Mesh stabilized Unstabilized
Flame height 5.5 cm 31 cm
C
8
H
18
(liquid) 1 ml/min 1.5 ml/min
N
2
coow 0.8 l/min 0.8 l/min
O
2
9.0 l/min 2.6 l/min
Froude number 0.1 11.8
Fe(CO)
5
fuel
seeding
4000 ppm (mass) 4000 ppm (mass)
7.5 mm at the burner lip, increasing to a diameter of
nearly 12 mm at a height of 41 mm above the burner
exit. A stainless steel mesh ame holder was placed
55 mm above the lip of the Burner 1 to promote ame
stability of the buoyancy-driven ame (Froude num-
ber of 0.1), thereby enabling spatially resolved light-
scattering measurements.
For Burner 2, isooctane was prevaporized at a rate
of 1.5 ml/min with a nitrogen coow of 0.8 l/min.
A ow rate of 2.6 l/min of pure oxygen was passed
through the outer concentric tube array. The result-
ing ame was operated above the smoke point, with a
total ame length of 31 cm and an average ame di-
ameter slightly below 1 cm. However, no ame holder
was employed for these experiments, as it was desired
168 K.B. Kim et al. / Combustion and Flame 154 (2008) 164180
Fig. 2. Schematic of the light-scattering conguration (top view).
to record soot proles through the entire ame length.
Therefore, the dimensions of the burner lip were re-
duced to enable the burner to provide a more stable,
momentum-driven ame (Froude number of 11.8)
conducive to measurements throughout the ame oxi-
dation regime. Given the dilution of the fuel ow with
nitrogen, the use of pure oxygen as opposed to air was
found to promote ame stability. The signicant res-
idence time of both ames, combined with the lack
of any shroud ow, suggests considerable diffusion of
ambient air, minimizing any effects due to the use of
pure oxygen. The ame temperatures were recorded
using a 1/16

-diameter type K thermocouple probe


corrected for radiation loss. The maximum ame tem-
perature was 1850 K, recorded at the height of 1.8 cm,
after which the temperature decreased steadily, drop-
ping to 1200 K at 15 cm (the start of signicant soot
oxidation), and to 1000 K at 22 cm, above which there
is a precipitous drop approaching the ame tip.
For all iron-seeded ame experiments, the iron
pentacarbonyl was added to the liquid isooctane at a
concentration of 4000 ppm by mass (0.11% Fe per
mass of fuel), which was found to be the most ef-
cient value for investigating soot suppression effects
in this study, as described below.
2.2. Optical soot diagnostics
Spatially resolved light-scattering and transmis-
sion measurements were carried out using a fre-
quency-doubled (532-nm) Q-switched (8 ns fwhm)
Nd:YAG laser. The laser beam was vertically polar-
ized with respect to the horizontal scattering plane.
The laser was operated at a pulse repetition rate of
10 Hz, with a pulse energy of 0.3 mJ/pulse, and
with a beam cross-sectional area of 0.0033 cm
2
at
the center of the ame. The corresponding laser u-
ence was equal to 0.09 J/cm
2
, which was determined
in a previous study of pulsed laser/soot interactions
to produce no soot vaporization [19]. Fig. 2 shows
the optical setup for the light-scattering system. The
laser beam rst passed through an aperture to remove
the edges of the Gaussian laser intensity prole be-
fore it was directed through the primary focusing
lens (f =25 cm), such that the focused beam passed
through the center of the ame. After exiting the
K.B. Kim et al. / Combustion and Flame 154 (2008) 164180 169
ame, the beam was terminated in a low-reectance
laser beam dump. The burner apparatus was enclosed
with an opaque Plexiglas box that provided large
openings (20 by 10 cm) along the laser axis and a
5-cm-diameter hole in line with the PMT. The scat-
tered light from the soot particles was collected at
90

with respect to the incident beam and imaged


onto the surface of a photomultiplier tube with unity
magnication using a 100-mm focal length bicon-
vex lens. Apertures (0.5 mm in diameter) were used
both in front of the biconvex lens and in front of the
PMT inlet to further dene the scattering volume. In
addition, a 532-nm narrow-line laser lter and a po-
larizer oriented to pass vertically polarized light with
respect to the horizontal scattering plane were used
to reject stray light, as well as to ensure the mea-
surement of verticalvertical scattering, as discussed
below. Finally, neutral density (ND) lters were used
as needed to ensure PMT signal linearity, which was
checked prior to all measurements. As implemented,
the light-scattering collection system ensured that the
scattered light was collected only from the precise
scattering volume dened within the ame, with a
solid collection angle of about 0.05 sr. The PMT sig-
nal was recorded with a digital oscilloscope (500 Hz,
4 GS/s). A precision high-voltage supply was used
to drive the photomultiplier tube, typically at 650 V.
Individual measurements were recorded as the ensem-
ble average of 300 laser pulses, which corresponds
to a 30-second average at the 10-Hz laser repetition
rate.
Absolute differential scattering coefcients (K

VV
= N

VV
) were measured by performing a calibra-
tion using ultra-high-purity methane gas at ambient
pressure and temperature as a calibration standard.
Corrections for stray light were made by recording the
scattering ratio of the calibration methane signal to a
high-purity nitrogen signal, noting that the ideal ratio
is 2.17 in the absence of any stray light for the current
parameters. For all experiments, stray light calibration
measurements were taken prior to every ame study
to determine an experiment-specic stray light value.
Over the range of experimental measurements, the av-
erage contribution of stray light was about 1/3 of the
methane calibration signal. In addition, ame trans-
mission measurements were recorded using a laser
power meter tted with a narrowband 532-nm laser
line lter. This setup is identical to the scattering sys-
tem shown in Fig. 2, except that the beam dump is
replaced by the laser power meter and aperture. In ad-
dition to the use of a narrowband lter, corrections
were made for ame emission at all heights, although
the contribution was not signicant.
Laser-induced uorescence (LIF) measurements
were performed to assess the chemical state of the
iron species. An optical parametric oscillator (OPO)
was pumped using an injection-seeded, 355-nm
Nd:YAG laser and the resulting laser pulse was fre-
quency-doubled to realize the desired excitation fre-
quency. LIF signals were recorded by collecting emis-
sion along the direction of beam propagation (i.e.,
backscatter) using a pierced mirror and a 50-mm-
diameter collection lens. Collected light was ber-
coupled to a 0.275-m spectrometer and recorded us-
ing an intensied CCD array detector that was syn-
chronized to the laser pulse. A 200-ns intensier gate
centered on the laser pulse (8 ns fwhm) was used
for all LIF experiments, which effectively eliminated
ame emission during the narrow detector gate.
2.3. Soot sampling
To complement the in situ light-scattering mea-
surements, thermophoretic sampling was conducted
over a range of ame heights for Burner 2. The
soot aggregates were extracted from the ame using
thermophoretic sampling, and subsequent transmis-
sion electron microscopy (TEM) analysis was imple-
mented to determine the size and morphology of soot
aggregates. Soot samples were collected directly on
Formvar-carbon coated 150-mesh copper TEM grids
(Electron Microscopy Sciences, Hateld, PA) at 12
different heights along the vertical axis of both the
unseeded and seeded ames using Burner 2. For sam-
pling, TEM grids were afxed to a holder and then
passed through the ame on a horizontally swing-
ing arm. It is necessary that the exposure time of
the grids be long enough to capture an appropriate
amount of soot, but short enough to avoid damaging
the grid or oversampling the aggregates. The sam-
pling times were controlled manually in the present
study. As a result, each sample had a slightly different
residence time in the ame, but care was taken that
all exposure times of the lms were on a timescale of
ms, as based on the total translation time and ame
width.
TEM grids were analyzed using a JEOL 2010F
analytical electron microscope system (point reso-
lution of 0.19 nm). TEM analysis yielded primary
particle diameters, overall agglomerate dimensions,
and numbers of primary particles per agglomerate,
which were subsequently used to calculate fractal di-
mensions. Magnications used for the present mea-
surements ranged from 50,000 to 330,000. For
analysis, individual soot aggregates were randomly
selected at low magnication and then analyzed at the
optimum magnication. Aggregate dimensions and
number of primary particles were manually extracted
from individual TEM micrographs using a light box, a
micrometer, and the calibrated scale bar on each TEM
image.
170 K.B. Kim et al. / Combustion and Flame 154 (2008) 164180
3. Results
3.1. Burner 1: soot formation and growth regime
To spatially resolve the soot formation and growth
processes, light-scattering and transmission measure-
ments were made between the heights of 8.7 and
41.0 mm using Burner 1. Radial measurements were
recorded from the central axis outward to a radius
of 3.2 mm, using 0.64-mm steps. At each ame
position, the absolute differential scattering coef-
cients K

VV
(cm
1
sr
1
) were determined from the
direct calibration procedure. The transmission mea-
surements were reduced using a three-point Abel
inversion routine [20], yielding an extinction coef-
cient (cm
1
) at each radial position. By pairing the
measured extinction and differential scattering co-
efcients, classical light-scattering theory was used
to solve for the equivalent spherical particle diam-
eter and the particle number density. As described
below, thermophoretic sampling was combined with
RayleighDebyeGans (RDG) theory to invert the
light-scattering data for Burner 2. However, the ra-
dially resolved light-scattering measurements per-
formed for Burner 1 were not compatible with ra-
dially resolved soot particle sampling. In addition,
signicant agglomeration is considered less important
in the nucleation and initial growth regimes probed
with Burner 1. With these factors and limitations in
mind, the Burner 1 data were inverted using Mie
theory, utilizing a zeroth-order logarithmic size dis-
tribution and a refractive index m = 2.0 0.35i as
reported for ame-generated soot [21]. In order to
characterize the scattering and extinction properties
of a polydisperse particle system, a suitable distribu-
tion function is necessary. Espenscheid et al. [22] pro-
posed a zeroth-order lognormal distribution (ZOLD)
to characterize a polydisperse system of particles. The
ZOLD function is dened as
(1) p(r) =
exp(
2
o
/2)

2
o
r
m
exp
_
(lnr lnr
m
)
2
2
2
o
_
,
where r is the particle radius, r
m
is the modal value
of r, and
o
is a dimensionless measure of the width
and skewness of the distribution. The mean radius r
is related to r
m
by
(2) r =r
m
exp
_
3
2

2
o
_
,
and the true standard deviation of the ZOLD is re-
lated to
o
and r
m
by
(3) =r
m
_
exp
_
4
2
o
_
exp
_
3
2
o
__
1/2
.
While the ZOLD was originally presented as an alter-
native to the well-known lognormal distribution, care-
ful examination reveals the ZOLD to in fact represent
the traditional lognormal distribution, only character-
ized by the mode rather than the mean, and by the
standard deviation of the distribution of the log. For
data analysis,
o
was xed at a value of 0.2, which is
close to the value expected for a self-preserving size
distribution (
o
=0.285), and was found to provide a
sufciently broad region of solution space character-
ized by a single value for data inversion. Specically,
the ratio of the differential scattering coefcient to
the extinction coefcient is reduced to a function of
modal diameter alone for a given value of
o
. How-
ever, in general, the functional relationship becomes
multivalued beyond a certain modal diameter, leading
to ambiguity during data inversion. For the present
analysis, the selected complex refractive index and
skewness parameter
o
yielded an acceptable range
of K

VV
/K
ext
with regard to the monotonic solution
range for data inversion.
The optical properties of soot, notably ame-
generated soot, have been the focus of much research
in the combustion community, and, in fact, remain a
contemporary issue. In general, the exact nature of
optical properties can be expected to vary with ame
conditions (e.g., stoichiometry) and fuel type; hence,
researchers are faced with no clear choice when se-
lecting a suitable index value for data inversion. A re-
cent review of light-absorbing carbon provides a de-
tailed treatment of the role of optical properties in ab-
sorption and light scattering [23]. A unied inversion
scheme for determining the optical properties of ag-
gregates has also been reported [24]. The use of par-
ticular index values was explored in a paper by Smyth
and Shaddix, which demonstrates the complexities
and at times ambiguities associated with the selection
of a particular index value [25]. Although the value
selected for the present study, namely 2.0 0.35i,
is consistent with range of recommended values [23]
for the real component, namely 1.9 to 1.95, the imag-
inary component lies in the lower range of recom-
mended values. To quantitatively assess the inuence
of soot optical properties on data inversion, the cur-
rent data were reanalyzed using an alternative value
of m=1.57 0.56i (see Ref. [25]). The overall soot
parameters such as the volume fraction, as discussed
below, were found to increase by about 28% with the
alternative index value; however, the nal percentage
reduction in soot volume fraction found with iron ad-
dition was completely unchanged. In earlier studies,
Charalampopoulos et al. examined the effect of iron
inclusion on the effective optical properties of soot
agglomerates [26,27]. It was found through detailed
analysis using the MaxwellGarnett effective index
model that modest iron inclusion volume fractions
consistent with the current study (see TEM analy-
sis below) did not appreciably alter the scattering
response as compared to homogeneous soot-particle
K.B. Kim et al. / Combustion and Flame 154 (2008) 164180 171
Fig. 3. Measured soot particle modal diameters as a func-
tion of radial position recorded for Burner 1 at a height of
8.7 mm. Data are reported for unseeded and iron seeded
(4000 ppm) ames.
models. More recently, Shu and Charalampopoulos
examined methodologies for measuring the effective
refractive indices of combustion-synthesized iron ox-
ide agglomerates [28], although such an approach is
considered beyond the current scope of work. In con-
cert, the above comments support the conclusion that
no ndings reached in the current study will be al-
tered by the use of alternative soot optical properties
or refractive index models.
The experimentally measured scattering-to-ex-
tinction ratio is related to Mie scattering theory using
the relationship
(4)
K

vv
K
ext
=
N
_

r=0

vv
p(r) dr
N
_

r=0

ext
p(r) dr
,
where the differential scattering cross section (

vv
)
and the extinction cross section (
ext
) are integrated
using the size distribution function p(r), per Eq. (1),
as the weighting function. Note that the overall parti-
cle number density N (cm
3
) cancels from Eq. (4),
making the relationship a function of the modal par-
ticle size and the skewness of the distribution func-
tion. The full Mie solution was used to evaluate the
scattering and extinction cross sections, and the inte-
gration was performed numerically using the trape-
zoid rule. Rather than iteratively solve Eq. (4) for
each extinction-to-scattering ratio, a third-order poly-
nomial curve was t to the function of K

VV
/K
ext
vs modal diameter. This relation was then inverted
to yield a straightforward solution of modal diam-
eter for each experimental K

VV
/K
ext
ratio. Fig. 3
shows the measured modal particle diameters for both
the unseeded and iron-seeded ames as a function
of radial positions for a height of 8.7 mm above the
burner tip. In general, the current modal diameter
values in the range of 20 to 80 nm are consistent
with soot particle/agglomerate sizes from hydrocar-
bon ames inverted with light-scattering theory. The
light-scattering results reveal an increase in soot parti-
cle size at the lower ame heights with iron addition,
including a factor of 2 increase in particle size near
the center axis, as shown in Fig. 3. The increase in
soot particle size at lower ame heights is concluded
to arise from the presence of additional surface area
for soot nucleation and growth, which results fromthe
nucleation of iron species (Fe or Fe
x
O
y
) in advance
of soot nucleation. Hence, soot inception and initial
growth processes are advanced by the availability of
what are essentially seed nuclei that provide addi-
tional surface area. However, such gains are eventu-
ally offset as the soot particles are altered via surface
growth at higher ame heights, thereby enveloping
the iron-rich nuclei within soot particles. Toward the
higher regions of the ame, the growth rates of the un-
seeded and seeded ames were found to converge, as
the initial increase in surface area with iron addition
is lost as compared to the unseeded ame following
signicant soot growth.
In addition to soot particle size, it is useful to ex-
amine the soot particle number density and the corre-
sponding soot loading (i.e., volume fraction) between
ame conditions. Solution of Eq. (4) for the soot par-
ticle modal diameter readily allows calculation of the
overall soot particle number density N from the rela-
tion N =
ext
/K
ext
. With the particle modal diameter
and number density known, the particle volume frac-
tion is determined from
(5) f
v
=
4
3
Nr
3
m
exp
_
15
o
2
_
,
where r
m
and
o
are ZOLDparameters described pre-
viously. Equation (5) is realized by the product of the
particle number density and the weighted integral of
the single-particle volume using the size distribution
p(r) as the weighting function. Fig. 4 shows the mea-
sured soot particle volume fractions (i.e., total soot)
at the greatest height measured in Burner 1, namely
41 mm above the surface. The Fig. 4 data reveal a
slight increase in soot volume near the burner axis,
although such differences are well within the exper-
imental uncertainty. Specically, Abel inversion can
compound data inversion errors, notably those asso-
ciated with the innermost data point; hence consider-
able uncertainty can be expected for the central node,
as reected by the relatively large error bar corre-
sponding to this location in Fig. 4. Examination of the
rate of change of the integrated soot volume fractions
with respect to ame height (i.e., soot growth rate)
revealed continuous soot growth throughout the mea-
172 K.B. Kim et al. / Combustion and Flame 154 (2008) 164180
Fig. 4. Measured soot particle volume fractions as a function
of radial position recorded for Burner 1 at a height of 4.1 cm.
Data are reported for unseeded and iron-seeded (4000 ppm)
ames.
surement region of Burner 1. Recall that the Burner 1
conguration was designed to probe the soot incep-
tion and growth regimes with high resolution; hence it
was expected that no signicant soot oxidation would
be observed.
Overall, it is concluded that the roles of the iron
additive are primarily limited to iron nucleation and
subsequent incorporation into the soot particulates
within the inception and growth regimes. TEM analy-
sis, reported below, will further dene the role of iron
incorporation. The exact nature of the iron chemical
state remains to be determined, and additional discus-
sion will be done in combination with the soot oxida-
tion region measurements following the next section.
3.2. Burner 2: soot oxidation regime
The investigation of the soot oxidation (i.e., burn-
out) regime was performed using the Burner 2 con-
guration, which provided sufcient residence time
to observe soot burnout. Because the strong radial
dependence of soot properties gives way to a nearly
uniform radial prole at relatively great ame heights
(as veried experimentally), combined with the en-
hanced spatial variability of the unstabilized ame,
only axial (i.e., = 0) light-scattering measurements
were recorded.
The unstabilized Burner 2 conguration provided
access to ame residence all the way to the ame
tip, which was necessary to examine the oxidation
regime to the full extent. This ame was also use-
ful for selecting an optimal concentration of iron
pentacarbonyl by examining the smoke point depen-
dence. Specically, the smoke point was determined
by adjusting the oxygen ow rate, for a xed fuel
Fig. 5. Measured smoke point (expressed as fuel equivalence
ratio) as a function of Fe(CO)
5
concentration recorded for
Burner 2. Error bars represent one standard deviation.
ow rate and iron pentacarbonyl concentration, un-
til the point at which visible smoke at the ame tip
rst occurred. These measurements were repeated for
iron pentacarbonyl concentrations ranging from 0 to
20,000 ppm, with the results presented in Fig. 5.
An increased fuel equivalence ratio (i.e., increasingly
fuel-rich) at the smoke point corresponds to a reduced
propensity to smoke; hence the Fig. 5 data demon-
strate that the addition of iron pentacarbonyl provides
benets regarding the breakthrough of visible soot
from the ame tip (i.e., smoke). Furthermore, the data
reveal a breakpoint near a seeding concentration of
4000 ppm iron pentacarbonyl, beyond which addi-
tional seeding concentrations provide only marginal
change in the measured smoke point. The Fig. 5 data
also provide insight into the relative roles of diffu-
sion and surface reaction, suggesting the importance
of the former. Given the complexity of the current ex-
periments, it was desirable to explore a single optimal
value of iron pentacarbonyl seeding rather than per-
form a broad, parametric study; hence, 4000 ppm was
selected as the nal value for the iron pentacarbonyl
seeding concentration.
In a manner similar to that followed before, ab-
solute differential scattering coefcients were re-
corded between a height of 9.4 cm above the burner
tip and a height of just over 25 cm, which was es-
sentially the ame tip. Fig. 6 presents the average
values of the differential scattering coefcients along
with the standard deviation (N = 10 observations
over multiple days) for the unseeded and iron-seeded
ames. The differential scattering coefcients tend to
peak in both cases near a height of about 14 cm above
the burner surface, and then steadily decrease with ad-
K.B. Kim et al. / Combustion and Flame 154 (2008) 164180 173
Fig. 6. Measured differential scattering coefcients as a
function of height above burner recorded for Burner 2.
Data are reported for unseeded and iron-seeded (4000 ppm)
ames. Error bars represent one standard deviation.
ditional height (i.e., increasing residence time). This
behavior indicates that the soot prole has transi-
tioned from the soot growth regime to the oxidation
regime above this height, which is characteristic of
diffusion ames. Regarding the comparison of the
unseeded and the iron-seeded ames within the soot
growth regime (i.e., lower ame heights), the scatter-
ing coefcients are essentially identical (within ex-
perimental variability) for both seeded and unseeded
ames. Such behavior is in excellent agreement with
the spatially resolved Burner 1 data, which like-
wise probed the soot formation and growth regime,
supporting the conclusion that iron particulates are
quickly incorporated into soot agglomerates, giving
way to typical (i.e., unseeded) soot growth and soot
yields. Meanwhile, the prominent deviation between
the two ames in scattering coefcient is observed
in the soot oxidation zone, increasingly so at greater
ame heights approaching the ame tip. The Fig. 6
data support the conclusion that iron addition acts
within the soot oxidation regime of the ame. It is
noted that soot processes are generally considered to
occur in concentric regions between the ame center
and edge zone [29], which is consistent with the radi-
ally resolved measurements of Burner 1 (see Fig. 3).
However, for Burner 2, radial scattering measure-
ments did not reveal any signicant radial structure
throughout the oxidation zone (>15 cm); hence the
current in situ measurements and extractive sampling
data are considered representative of the relevant soot
processes (e.g., oxidation).
To support quantitative analysis of the light-
scattering measurements, soot particles were sampled
at axial locations using thermophoretic sampling, as
described above, and subsequently analyzed using
TEM. Typical soot particle micrographs are presented
in Fig. 7 for lower and upper ame positions, for both
the unseeded and iron-seeded ames. The TEM im-
ages were analyzed to extract the relevant parameters,
including agglomerate size and fractal dimensions,
and the primary soot particle size and primary particle
number density (i.e., particles per agglomerate). For
each ame height investigated, 375 primary soot par-
ticles were randomly selected from a number of soot
agglomerates, to provide a statistical average of the
primary particle diameters. Consistent with the results
from Burner 1, the soot primary particle parameters
were similar in both size and morphology at the lower
ame heights. Specically, the average primary soot
particle diameters at the lower heights (averaged be-
tween 9.4 and 11.7 cm above burner) were equal to 31
and 29 nm in the unseeded and iron-seeded ames, re-
spectively. The relative standard deviations averaged
to 6.5% for these data; hence, the difference is on the
order of the experimental uncertainty. In contrast, at
the greatest residence time investigated (H =25 cm),
the primary particle diameters were reduced to an
average of 24 and 20 nm in the unseeded and iron-
seeded ames, respectively. The overall decrease in
particle size with increasing ame height (i.e., ame
residence) is consistent with signicant soot oxidation
in both ames. However, the 37% greater reduction
in nal soot particle size in the iron-seeded ame as
compared to the unseeded ame is characteristic of
enhanced soot oxidation rates and overall soot sup-
pression. Overall, the primary particle diameters, as
measured from the TEM micrographs, depict the tran-
sition from soot growth to soot oxidation along the
length of the ame. As described above, no radial
ame resolution was realized for the current soot sam-
pling. It is difcult to assess the overall uncertainly
associated with the radially averaged soot sampling;
however, the overall uniformity of the observed soot
agglomerate TEM data (e.g., 6.5% RSD for primary
soot particles) provides some quantitative assessment
of variability. To further quantify the overall differ-
ences between the two ame conditions, it is useful to
examine the total soot volume fractions.
As the soot particles undergo a combination of sur-
face growth and signicant agglomeration with ame
residence time, their morphology evolves to a frac-
tal structured aggregate. Using either Rayleigh scat-
tering theory or Mie scattering theory itself is not
a reliable application for the large-sized and open-
structured soot agglomerates. Such particles are more
accurately modeled using the RayleighDebyeGans
(RDG) scattering theory. The RDG theory has been
widely used to interpret light-scattering data from ag-
gregates, notably ame-generated soot [3033]. The
174 K.B. Kim et al. / Combustion and Flame 154 (2008) 164180
Fig. 7. TEMimages of sampled soot particles. H =10 cm for unseeded (top left) and iron-seeded (top right) samples. H =25 cm
for unseeded (bottom left) and iron-seeded (bottom right) samples.
major assumptions for RDG applicability include that
the soot agglomerates are fractal-like objects with a
mass fractal dimension of less than 2 and that the
primary particles are sufciently small so that they
satisfy the Rayleigh scattering theory.
The morphological fractal features of soot ag-
glomerates can be characterized by a power-law re-
lationship between the number of primary particles in
an aggregate and its projected area on the TEM im-
age,
(6) N
p
=k
f
(R
g
/d
p
)
D
f
,
where N
p
is the number of primary particles per ag-
gregate, k
f
is the fractal prefactor, R
g
is the radius of
gyration of an aggregate, d
p
is the primary particle
diameter, and D
f
is the mass fractal dimension im-
plying the openness of the soot aggregate. Both the
parameters N
p
and d
p
were directly determined us-
ing transmission electron microscopy (TEM) analysis
of thermophoretically sampled soot aggregates. The
radius of gyration then can be calculated using a cor-
relation [32], namely,
(7) (LW)
1/2
/(2R
g
) =1.17,
where L is the geometric mean of maximum length,
and W is the maximum projected width normal to L.
With these parameters known, the fractal properties
are obtained using a linear regression method with a
least-squares approach. In other words, when N
p
is
plotted as a function of R
g
/d
p
in logarithmic scale
for a set of aggregates, the fractal dimension de-
scribes the slope, and the fractal prefactor determines
the magnitude of the least-squares straight-line t to
the data. The analysis of hundreds of individual ag-
glomerates revealed no statistical difference in fractal
dimension between the unseeded (1.84 0.08) and
seeded ames (1.80 0.14), with the overall average
fractal dimension equal to 1.82 (6% RSD) for all data.
K.B. Kim et al. / Combustion and Flame 154 (2008) 164180 175
This value is in excellent agreement with a number of
values reported for ame-generated soot, which range
from 1.54 to 1.85 [3033].
Using the primary soot particle parameters in com-
bination with RDG theory enables calculation of the
average agglomerate differential scattering cross sec-
tion. The differential scattering cross section (cm
2
/sr)
of a single soot agglomerate is dened as
(8)

agg
=N
2
p

p
S(q),
where the differential scattering cross section of the
primary soot particle within an agglomerate,

p
, is
assumed to satisfy the criteria for Rayleigh scatter-
ing theory, namely that the size parameter (d
p
/)
is much less than unity. The new parameter, S(q),
shown in Eq. (8) is denoted as the angular scattering
form factor, given by
(9) S(q) =C(qR
g
)
D
f
,
provided that the product qR
g
is less than unity. In
these expressions, q is the modulus of the scattering
vector (cm
1
), dened as
(10) q =
4

sin(/2).
The measured differential scattering coefcient
(cm
1
sr
1
) is then a product of the calculated dif-
ferential scattering cross section of the soot agglom-
erates (Eq. (8)) and the overall soot agglomerate num-
ber density, N
agg
, namely,
(11) K

vv,agg
=N
agg

agg
.
Based on the above analysis, the TEM data were used
to calculate the average differential scattering cross
section of the soot agglomerates at each ame height.
This value was then combined with the measured dif-
ferential scattering coefcient and Eq. (11) to yield
the soot agglomerate number density as a function of
ame height for both the unseeded and iron-seeded
ames. The overall soot volume fraction, f
v
, is then
calculated from the relation
(12) f
v
=
_

6
d
3
p
_
N
p
N
agg
,
where N
p
and d
p
are the average number of primary
particles per aggregate and the primary particle diam-
eter, respectively, as directly measured per the TEM
analysis.
The volume fraction data as a function of ame
height are presented in Fig. 8 for the unseeded and
iron-seeded ames. Error bars represent the standard
error based on the formal propagation of experimen-
tal error (as standard deviation) related to each of the
measured fundamental parameters, namely the differ-
ential scattering coefcient, angular scattering form
Fig. 8. Measured soot volume fractions as a function of
height above burner recorded for Burner 2. Data are re-
ported for unseeded and iron-seeded (4000 ppm) ames.
The dashed line represents the nominal transition from soot
growth to soot oxidation. Error bars represent the standard
error.
factor, number of primary particles per aggregate, and
primary particle diameter. The Fig. 8 data reveal two
distinct regions with regard to the comparison of the
unseeded and iron-seeded ames. In the lower regions
of the ame, below about 15 cm, the volume fractions
of the two ames are comparable within experimental
uncertainty (i.e., error bars). This is in agreement with
the Burner 1 results, and is the region where the iron
is incorporated within the soot aggregates during the
primary growth regime, thereby exerting little inu-
ence after inception. However, over the last 5 cm of
ame residence, the iron-seeded ame is character-
ized by a marked reduction in soot volume fraction.
Specically, at a height of 25 cm, which is near the
ame tip, the soot volume fraction is reduced from
3.52 10
7
in the unseeded ame to a value of
1.22 10
7
in the iron-seeded ame, correspond-
ing to a 66% reduction in soot emissions. The Fig. 8
data support the conclusion that the soot-suppressing
role of the iron additive functions primarily in the soot
burnout regime, although additional discussion is re-
quired as to the exact mechanisms of action.
3.3. Chemical speciation of iron species
To elucidate the processes through which the iron
addition apparently enhances the oxidation of soot,
it is necessary to understand the chemical state of
the iron species within the soot aggregates, as dis-
cussed above. Several additional measurements were
performed toward this goal. First, in the Burner 1
conguration, rust-colored residue was noted on the
176 K.B. Kim et al. / Combustion and Flame 154 (2008) 164180
Fig. 9. Raman spectra of iron oxide (Fe
2
O
3
) reference pel-
let (upper spectrum) and of the red deposit observed on the
ame holder of Burner 1 after iron-seeded operation (lower
spectrum).
surface of the ame holder with iron seeding of the
fuel. Confocal micro-Raman spectroscopy (LabRam
Innity, Jobin Yvon) was used to analyze the state of
the residual iron on the ame holder. The excitation
wavelength was 632.8 nm, generated from a He:Ne
continuous wave (cw) laser. For comparison, refer-
ence Fe
2
O
3
Raman spectra were taken from bulk iron
oxide powders pressed into a 3-cm-diameter disc of
approximately 2 mm thickness.
The ame holder used during a seeded ame ex-
periment was placed under the 100 objective of the
micro-Raman spectrometer and a Raman spectrum
was recorded from the area corresponding to the rust-
colored residue. Spectra were recorded using a 5-s
integration time averaged over eight individual acqui-
sitions. The resulting ame holder spectrum was con-
sistent with the reference Fe
2
O
3
spectra, with both
shown in Fig. 9. The Raman analysis was consistent
and repeatable when recorded immediately after re-
moving the ame holder from the ame, or when
recorded after several days. In addition, no trace of
other iron oxides (FeO or Fe
3
O
4
) was observed, nor
were any such iron oxide signals recorded from the
ame holder following the unseeded ame experi-
ments. It is noted that an additional peak centered at
662 cm
1
is also apparent from the Raman spectra
in Fig. 9. The 662 cm
1
peak is not characteristic of
soot or carbon, and is attributed to the ame holder
substrate, as veried using control experiments.
To further assess the state of iron within the sam-
pled soot aggregates, high-resolution TEM in combi-
nation with EDS analysis was used to probe individ-
ual agglomerates. As can be observed in Fig. 7, as
well as additional high-resolution images, the TEM
images from the iron-seeded soot particles reveal
small inclusions (less than 10 nm) throughout the pri-
mary soot particles. Energy-dispersive spectroscopy
(EDS) was used to probe these individual inclusions
with nanometer spatial resolution. EDS revealed the
presence of both iron and oxygen from many of the
inclusions, with Fe:O ratios (molar) that varied from
about 1:0.5 to 1:4. About 40% of the measured inclu-
sions revealed only a strong iron signal, with essen-
tially no oxygen signal recorded. No clear trend was
observed with regard to ame height. Analysis of the
soot structure in regions away from these inclusions
(i.e., presumably pure soot) revealed no detectable
iron or oxygen signals. Therefore, the TEM/EDS
analysis fully supports the nucleation of iron-rich par-
ticles and subsequent incorporation of these seed nu-
clei (5 to 10 nm size) into the soot agglomerates,
although no clear or consistent chemical species as-
sociated with the iron-rich inclusions was found.
Finally, in an attempt to gain information as to
the chemical nature of iron and possible transfor-
mations within the ame, laser-induced uorescence
(LIF) measurements were performed as an alterna-
tive in situ probe. To optimize the LIF signal for iron
species within a ame environment, preliminary stud-
ies were performed in an iron-pentacarbonyl-seeded
CO diffusion ame, which is known to produce iron-
rich agglomerates without the presence of interfer-
ing soot [34]. Using such a ame, three primary
excitation bands for LIF of elemental iron were ex-
plored, namely 295.39 nm (704.034,547.2 cm
1
),
296.69 nm (033,695.4 cm
1
), and 297.31 nm
(704.034,328.7 cm
1
). Signicant uorescence was
observed at 372.8, 376.4, and 378.8 nm for the rst
wavelength, 373.5 nm for the second wavelength, and
375.0 and 375.8 nm for the third wavelength, respec-
tively. The strongest LIF scheme for iron was found to
be excitation of the resonant 296.69-nm transition and
observation of the 373.5-nm emission line (6928.3
33,695.4 cm
1
). Representative uorescence spectra
are shown in Fig. 10 for two different ame heights. It
was also noted that no emission signal was recorded
on or near the 373.5-nm spectral region when the
laser was tuned 0.5 nm off of the 296.69-nm reso-
nant line, ensuring that no spurious signals or other
emission sources (e.g., broadband uorescence) were
present other than those signals arising from elemen-
tal iron atoms. Using this scheme, LIF measurements
were recorded as a function of ame height (0 to
25 mm) using Burner 2 for the iron-seeded ame con-
dition. The LIF signal was quantied by integrating
over the full width of the 373.5-nm emission line,
and then normalizing the signal to the data recorded
over the rst ve ame heights (0 to 4 mm), with
the results shown in Fig. 11. The data reveal a rather
steady but slow decay in elemental iron, as measured
K.B. Kim et al. / Combustion and Flame 154 (2008) 164180 177
Fig. 10. Laser-induced uorescence spectra recorded from
the iron-seeded (4000 ppm) ame using Burner 2. The up-
per spectrum was recorded 3.7 cm above the burner, and the
lower spectrum was recorded 18.5 cm above the burner sur-
face. The excitation line was 296.7 nm (033,695.4 cm
1
)
for both measurements. The prominent peak at 373.5 nm
corresponds to the 6928.333,695.4 cm
1
atomic iron tran-
sition.
Fig. 11. Intensity of 373.5-nm laser-induced uorescence
intensity as a function of height above burner for the
iron-seeded (4000 ppm) ame using Burner 2. The excita-
tion line was 296.7 nm. Error bars represent one standard
deviation.
by the LIF signal, through the soot nucleation and
growth regime. This is followed by a rapid and signif-
icant drop in iron signal throughout the soot oxidation
regime, where the signal was observed to decrease by
more than two orders of magnitude between a ame
height of 15 cm and the ame tip.
Prior to detailed discussion of this trend, a few
comments are noted concerning the LIF data. With
LIF, one must always consider quenching effects,
especially in the presence of a variation in concomi-
tant species and temperature, as in the current ame
regimes. To independently corroborate the loss of the
LIF signal with increasing ame height, additional
transmission measurements were performed using
an iron hollow-cathode lamp, noting that while LIF
is subject to quenching, absorption spectroscopy is
free from this effect. Using this technique, absorption
measurements using the strong Fe resonant transition
at 271.9 nm were recorded as a function of ame
height. The transmission measurements revealed a
trend identical to that in the LIF measurements,
namely, low and consistent transmission through the
soot inception and growth regime (up to a height of
14 cm), followed by a rapid increase in transmis-
sion to near unity (i.e., nonabsorbing) at the ame
tip. Therefore, the presence of a strong LIF signal at
low heights is the result of the presence of elemen-
tal iron throughout the lower ame heights prior to
the onset of soot oxidation. In contrast, the marked
demise of the LIF signal through the soot oxidation
regime (from 15 cm to the ame tip) is the result
of the loss of elemental iron, presumably to an iron
oxide species, and not the result of preferential signal
quenching.
Additional LIF measurements were performed
to determine the presence of any FeO within the
iron-seeded ames. Several resonant lines of FeO,
as reported by Mavrodineanu and Boiteux [35],
were investigated, including both 590.3-nm (870
17,800 cm
1
) and 564.7-nm (87018,570 cm
1
) ex-
citation schemes with corresponding uorescence at
621.9 nm for the former and 627.9 and 593.5 nm for
the latter. Using both of these schemes, no FeO LIF
was detected, leading to the conclusion that FeO was
not present in a signicant quantity (i.e., above the
detection limit) within the iron-pentacarbonyl-seeded
ame. It is noted, however, that the use of backscatter
for collection of the LIF signals in combination with
the unstabilized ame does not enable probing of the
ame with any signicant radial resolution.
4. Discussion of results
The current ame studies provide additional in-
sight into the mechanisms of soot suppression with
iron-based fuel additives. Detailed light-scattering
measurements and TEM analysis of sampled soot par-
ticles were performed throughout the soot inception,
growth, and oxidation regions, revealing two distinct
regimes for iron addition regarding the interactions
with sooting characteristics. The data strongly sup-
port the role of iron-rich nucleus particles (5 nm in
size) acting as early nucleation sites for subsequent
178 K.B. Kim et al. / Combustion and Flame 154 (2008) 164180
soot inception and growth. This is consistent with ob-
servations made in previous studies of iron-seeded
premixed ames for several other fuels [58,18]. As
the soot particles undergo surface growth through-
out the soot growth regime, the TEM images suggest
that these iron-rich nuclei become completely sur-
rounded by soot, so that the resulting primary soot
particles are characterized by the appearance of typ-
ical ame-generated soot, but with the addition of
discrete iron-rich clusters or inclusions. Soot volume
fraction proles at the end of the growth regime in
the iron-seeded ame were found to be typical of the
unseeded ames, as the inuence of the iron-rich in-
clusions on the overall soot surface growth processes
was found to be negligible. Clearly the signicant
inuence of iron addition is not within the soot in-
ception and growth regimes, other than incorporation
of iron species, but rather is manifest within the soot
oxidation or burnout regimes. However, the exact na-
ture of the chemical state of the iron species is an
important question.
The present study suggests several answers re-
garding the evolution of iron species within the
iron-seeded ames. The ex situ Raman spectroscopy
analysis of the ame holder for the Burner 1 experi-
ments is consistent with the transition from elemental
iron to iron oxide species (e.g., Fe
2
O
3
) within the
latter regions of iron-seeded ames, although several
important comments are noted with regard to the ex-
act species. The measurements only correspond to the
stabilized ame-holder conguration and to the soot
actually deposited and aged on the ame holder. This
provides considerable residence time to age and there-
fore oxidize iron species, and therefore provides no
direct evidence as to the state of iron species within
the inception, growth, and oxidation regimes. In addi-
tion, thermodynamic equilibrium calculations suggest
that FeOmay be a dominant species under the broader
range of ame conditions, notably at higher tempera-
tures, although it is difcult to model the exact oxygen
mole fractions given the diffusion ame congura-
tion. Also of importance is the difculty of ensuring
that equilibrium conditions are in fact reached given
the dynamic nature of the ame species proles and
the rather short residence time of iron species, namely
about 50 ms. With these comments in mind, and the
overall fuel-rich nature of the ame based on inlet
ow rates of fuel and oxygen, it is quite probable that
elemental iron is a signicant component of the soot
particles resulting from the iron-seeded ames prior
to entering the soot oxidation regime.
The presence of Fe at early times is in full agree-
ment with the in situ LIF measurements of the present
study, and also is consistent with recent work by Kim
et al. [17], which reported evidence of considerable
elemental iron in soot sampled from an ethylene dif-
fusion ame. The rather marked demise of the ele-
mental iron LIF signal through the oxidation regime
in the current study may be explained by the forma-
tion of iron oxide species (FeO, Fe
2
O
3
, or Fe
3
O
4
).
At earlier residence times, the iron is largely protected
from oxidation by encapsulation within the soot parti-
cles and by the competition of hydrocarbon oxidation
for available oxygen. However, at longer residence
times, (1) the oxygen concentrations are signicantly
increased as additional oxygen diffuses into the ame
and within the soot particles, (2) the signicant oxi-
dation of carbon (i.e., soot) further exposes the iron
to additional oxygen, and (3) the longer time scales
may allow the approach to equilibrium conditions. In
the aggregate, the above comments suggest a transi-
tion from elemental iron to iron oxide species and the
subsequent interaction of these new iron species with
soot throughout the soot burnout regime.
The presence of signicant elemental iron entering
the soot oxidation zone combined with the increased
rate of soot oxidation (as compared to the unseeded
ame) throughout the burnout region provides direct
evidence of the soot reduction role of iron addition.
The direct oxidation of soot by iron oxide species,
such as the following,
C
solid
+ FeO CO + Fe, (13)
can be considered unimportant for several reasons.
First, the data suggest a transition from elemental iron
to some other iron species (e.g., iron oxides) within
the oxidation region, not the transition to elemental
iron. Secondly, the amount of soot reduction achieved
in the present study is approximately two orders of
magnitude greater than could be obtained by direct
oxidation, as based on an overall mass balance as-
suming that all iron atoms are incorporated in soot
and can participate. Discounting of direct oxidation
leads to the consideration of catalytic soot oxidation
by iron species through reactions such as the follow-
ing processes,
C
solid
+ O
2
+ Fe CO
2
+ Fe, (14)
C
solid
+ OH + Fe CO + H + Fe. (15)
These reactions would occur on or near the surface
of the iron-rich nuclei dispersed throughout the soot
particles, noting that the transition metals such as iron
have been widely studied as catalyst materials. In ad-
dition to oxygen, the role of OH as a soot oxidizer
may be important in the presence of iron catalysts,
where, for example, the following surface reactions
may be responsible for the creation of OH,
2Fe(s) + O
2
O(s) + O(s), (16)
Fe(s) + H H(s), (17)
O(s) + H(s) OH, (18)
K.B. Kim et al. / Combustion and Flame 154 (2008) 164180 179
where Fe(s) denotes a free iron-surface site, and O(s)
and H(s) indicate surface species. Finally, OH radi-
cals are desorbed from the surface of the iron and
are available to oxidize the solid carbon via Reac-
tion (15). In addition to catalysis by elemental iron,
it is perhaps more reasonable to assume oxidation of
elemental iron and the subsequent catalysis by iron
oxide species:
(19) aFe +bO
2
Fe
x
O
y
,
(20) C
solid
+Fe
x
O
y
+O
2
CO
2
+Fe
x
O
y
.
Such a transition is consistent with the increase in
available oxygen in the latter ame regions and is in
agreement with the decreased elemental iron LIF sig-
nals, as noted above. Hence the results most likely
indicate iron oxide formation concomitant with en-
hanced soot oxidation.
The relative importance of Reaction (19), as well
the overall partitioning of iron species, was further
explored using STANJAN (Version 3.89) to assess
equilibrium mole fractions as a function of ame tem-
perature. The code was run based on the input fuel,
oxidizer, metal additive, and diluent ow rates, with
no attempts to correct for species diffusion of ambi-
ent air with ame residence time. For temperatures
near the maximum ame temperature (1800 K), the
dominant iron-containing species are predicted to be
Fe(OH)
2
and FeO (vapor and liquid), which are con-
sistent with the ndings of Rumminger and Linteris
[13,14], as well as recent work by Guo and Kennedy
using iron-pentacarbonyl-seeded hydrogen diffusion
ames [36]. At the present recorded ame temper-
atures of 1000 to 1200 K, corresponding to the
ame oxidation region (>15 cm), predicted domi-
nant species include FeO and Fe
3
O
4
, with the latter
rapidly increasing with decreasing temperature. At
even lower temperatures, the oxide Fe
2
O
3
becomes
dominant, which may explain the presence of this par-
ticular oxide as the species recorded directly on the
ame holder. The failure to nd evidence of FeO via
LIF measurements, notably at early ame heights, is
signicant in terms of the equilibrium calculations,
and is perhaps evidence of lack of equilibrium at early
heights as discussed above.
The combination of experimental measurements,
namely the LIF data, and the equilibrium calculations,
suggests the importance of iron oxide formation (e.g.,
Fe
3
O
4
and Fe
2
O
3
) with increasing ame residence
time in the acceleration of soot oxidation via Reaction
(20) within the iron-seeded ames. While the role of
elemental iron is considered important during soot in-
ception and early growth, the observed rapid demise
of Fe during the region of enhanced soot oxidation
suggests that the primary role of iron is as a precursor
to formation of catalyzing iron oxide species.
5. Conclusions
In summary, the soot-suppressing role of iron pen-
tacarbonyl suggests the importance of iron species,
notably elemental iron, at early ame residence times,
where iron-rich nuclei provide surface area for soot
inception and growth, and ultimately lead to iron-rich
inclusions within the ame-generated soot particles.
The actual soot-suppressing mechanisms are then
manifest in the soot oxidation regime, whereby the
current LIF measurements support a transformation
of elemental iron to iron oxide species with increas-
ing ame residence. These iron oxide species then
presumably catalyze the oxidation of soot, thereby
resulting in reduced soot emissions (66% in the cur-
rent studies) from the overall ame. Overall, the ex-
act ame conguration, combustor residence times,
and combustor fuel type and stoichiometry are all
expected to play key roles in the soot-suppressing ef-
fects realized with iron-based additives such as iron
pentacarbonyl. The specic roles of iron species nu-
cleation and subsequent incorporation into the soot
particle matrix, as well as chemical transformations,
are fundamental steps for practical realization of ben-
ecial soot reduction with metallic fuel additives.
Acknowledgment
This work was supported in part by the U.S. Navy
Ofce of Naval Research, Physical Sciences S&T Di-
vision, through Contract N00014-0210838.
References
[1] K. Katsouyanni, G. Pershagen, Cancer Causes Con-
trol 8 (1997) 289291.
[2] M.W. Shayeson, SAE Report #670866, 1967.
[3] J.B. Howard, W.J. Kausch, Prog. Energy Combust.
Sci. 6 (1980) 263276.
[4] K.E. Ritrievi, J.P. Longwell, A.F. Sarom, Combust.
Flame 70 (1987) 1731.
[5] A.S. Feitelberg, J.P. Longwell, A.F. Sarom, Combust.
Flame 92 (1993) 241253.
[6] D.W. Hahn, T.T. Charalampopoulos, Proc. Combust.
Inst. 24 (1992) 10071014.
[7] P.A. Bonczyk, Combust. Sci. Technol. 59 (1988) 143
163.
[8] B.S. Haynes, H. Jander, H.Gg. Wagner, Ber. Bunsen-
ges. Phys. Chem. 84 (1980) 585592.
[9] P.A. Bonczyk, Combust. Flame 87 (1991) 233244.
[10] J. Zhang, C.M. Megaridis, Combust. Flame 105 (1996)
528540.
[11] M. Kasper, K. Sattler, K. Siegmann, U. Matter, H.C.
Siegmann, J. Aerosol Sci. 30 (1999) 217225.
180 K.B. Kim et al. / Combustion and Flame 154 (2008) 164180
[12] J. Zhang, C.M. Megaridis, Proc. Combust. Inst. 25
(1994) 593600.
[13] M.D. Rumminger, G.T. Linteris, Combust. Flame 123
(2000) 8294.
[14] M.D. Rumminger, G.T. Linteris, Combust. Flame 128
(2002) 145164.
[15] G.S. Samuelsen, R.L. Hack, R.M. Himes, M. Azzazy,
University of California Irvine Report #ESL TR-83-
17, 1983.
[16] N.D. Marsh, I. Preciado, E.G. Eddings, A.F. Sarom,
A.B. Palotas, J.D. Robertson, Combust. Sci. Tech-
nol. 179 (2007) 9871001.
[17] S.H. Kim, R.A. Fletcher, M.R. Zachariah, Environ. Sci.
Technol. 39 (2005) 40214026.
[18] D.H. Cotton, N.J. Friswell, D.R. Jenkins, Combust.
Flame 17 (1971) 8798.
[19] G.D. Yoder, P.K. Diwakar, D.W. Hahn, Appl. Opt. 44
(2005) 42114219.
[20] C.J. Dasch, Combust. Flame 31 (1992) 11461152.
[21] S. Chippett, W.A. Gray, Combust. Flame 31 (1978)
149159.
[22] W.F. Espenscheid, M. Kerker, E. Matijevic, J. Phys.
Chem. 68 (1964) 30933097.
[23] T.C. Bond, R.W. Bergstrom, Aerosol Sci. Technol. 40
(2006) 2767.
[24] G.C. Shu, T.T. Charalampopoulos, Appl. Opt. 39
(2000) 67136724.
[25] K.C. Smyth, C.R. Shaddix, Combust. Flame 107 (1996)
314320.
[26] D.W. Hahn, T.T. Charalampopoulos, J. Phys. D Appl.
Phys. 26 (1993) 18511858.
[27] D.W. Hahn, T.T. Charalampopoulos, H. Chang, Appl.
Opt. 31 (1992) 65196528.
[28] G.C. Shu, T.T. Charalampopoulos, Appl. Opt. 42
(2003) 39573969.
[29] R.J. Santoro, T.T. Yeh, J.J. Horvath, H.G. Semerjian,
Combust. Sci. Technol. 53 (1987) 89115.
[30] S.S. Krishnan, K.C. Lin, G.M. Faeth, J. Heat Trans-
fer 123 (2001) 331339.
[31] H.W. Kim, M. Choi, Aerosol Sci. 34 (2003) 1633
1645.
[32] .. Kyl, G.M. Faeth, T.L. Farias, M.G. Carvalho,
Combust. Flame 100 (1995) 621633.
[33] C.M. Sorensen, Aerosol Sci. Technol. 35 (2001) 648
687.
[34] M.T. Cheng, G.W. Xie, M. Yang, D.T. Shaw, Aerosol
Sci. Technol. 14 (1991) 7481.
[35] R. Mavrodineanu, H. Boiteux, Flame Spectroscopy,
Wiley, New York, 1965.
[36] B. Guo, I.M. Kennedy, Aerosol Sci. Technol. 41 (2007)
944951.

You might also like