Results from the methane diffusion flame and from the two different ethylene
premixed flame experiments are presented and discussed separately. For each
flame type or configuration, the coated–denuded data for all coating types
are considered together.
Soot optical properties from the methane diffusion flame
Observed σabs vs. dp,VED methane soot data
from BC3, BC3+, and BC4. Panels (a) and (b) are λ=405
and 532 nm data, respectively, from the UCD CRD-PAS, and (c) is
λ=630 nm data from the CAPS PMSSA. The nascent and
coated–denuded data have been combined since there is no significant
difference in the absorption cross sections and MACs between the two
datasets. Note the inability of Mie theory to reproduce the observed
σabs,405 nm for all sizes. Vertical solid lines indicating x=0.9, where observations deviate from Mie theory, are provided for
reference. In addition, vertical dashed lines indicate dp,VED=160 nm, above which soot maturity is approximately constant.
MAC values of nascent and coated–denuded soot
Average and median MAC values were determined for BC particles from the
methane diffusion flame with x>0.9 (corresponding to a dp,VED=160 nm at λ=532 nm; Table S3). (The dimensionless size parameter
x=πdp/λ, where dp is particle diameter and
λ is wavelength.) There are no systematic differences in the MAC
values between nascent and nascent–denuded (Df,m=2.16±0.1)
and coated–denuded (Df,m=2.64±0.1) soot at any wavelength for
this range of x despite some degree of collapse for the thickly
coated–denuded particles (Figs. 1 and S2). The average MAC values for x>0.9 were 12.1±1.4 m2 g-1 at 405 nm, 9.1±1.1 m2 g-1 at 532 nm, and 7.1±1.1 m2 g-1 at
630 nm, where uncertainties are 1σ standard deviations of the
measurements. For the reverse-coating experiments, which gave broader BC
per-particle mass distributions compared to forward-coating distributions, we
found no dependence of the derived MAC values on the distribution width. The
collapse was presumably due to the effect of evaporation or condensation of
the coating material and not due to the denuding process alone (Bhandari et
al., 2017). The observed Df,m independence of the MAC (x>0.9) is
consistent with Radney et al. (2014). This contrasts, however, with modeling
studies that use non-Mie-based methods that can account for particle shape
effects, which indicate that lacy soot (with a fractal dimension, as opposed
to mass mobility exponent, of Df=1.8) is more absorbing than
“compact” soot (Df=2.4) (Kahnert and Devasthale, 2011; Scarnato
et al., 2013). In the calculations, the compact soot particles are less
absorbing because the innermost spherules are “shielded” by the outermost
spherules. It is possible that the extent of collapse here was insufficient
to lead to substantial “shielding” in our experiments. Regardless, given
the similarity of the observed nascent and coated–denuded particle
cross sections, they have been recombined into a single dataset in what
follows.
RI values calculated from Mie theory and the RDG
approximation
The σabs have been fit separately using Mie theory and the RDG
approximation to determine optimal, theory-specific effective complex RI
values. The observations and best fits are shown in Fig. 2 at the three
wavelengths (λ=405, 532, and 630 nm), and the derived optimal
wavelength-, flame-, and theory-specific values are reported in Table 1. The
quality of the best fit obtained is dependent upon both theory and wavelength
considered. First, fits performed using Mie theory tend to give reasonably
well-defined minima in the calculated χred2, indicating that
the optimal m and k are unique (Fig. S3). In contrast, fits performed
using the RDG approximation do not give a unique set of m and k but
instead a band of [m,k] pairs that describe the data equally well
(Fig. S4). Since RDG fits are nonunique, optimal k values are reported at
all wavelengths for a fixed value of m=1.80.
There are additional differences between Mie and RDG beyond the uniqueness of
the derived optimal RI. At λ=405 nm, Mie theory provides a poor
fit to the σabs when the fit is performed using data over the
entire size range sampled. In particular, at λ=405 nm the
σabs from Mie theory are overestimated below dp,VED∼120 nm (x∼0.9) and underestimated for larger sizes (Fig. 2a).
At λ=532 nm, this deviation also occurs at x∼0.9
(dp,VED>160 nm) corresponding to larger size particles. Compared
to 405 nm, the overestimate at small x and underestimate at larger x for
532 nm is smaller. At λ=630 nm, the Mie theory fit compares well
with the observations at x<0.9 (dp,VED∼180 nm) and with
some deviation observed at larger sizes. When the fits are restricted to x<0.9, a reasonable fit using Mie theory is obtained at all wavelengths over
this size range, although there is perhaps a small overestimate at the
smallest sizes. However, these constrained Mie fits extrapolated to larger
sizes (x>0.9) still underestimate the observed absorption. For RDG,
generally good fits are obtained at all wavelengths and across all sizes,
although the RDG-calculated σabs tends to overestimate the
observation at smaller sizes below dp,VED∼100 nm.
It is important to note that the RI values listed in Table 1 are theory- and
property-specific. This means that the Mie-derived RI values are not
appropriate for use with the RDG approximation and vice versa. Additionally,
they must be used assuming a material density of 1.8 g cm-3, since
this value was used to convert mp to dp,VED or
Nspherule. If, for example, a smaller density were used with these
RI values then the particles would have substantially higher MACs. Of note is
that both the real and imaginary RIs from Mie theory are larger than RI
values that are commonly used in global climate models (Bond et al., 2013),
including the RI that is often considered the currently recommended value
(1.95-0.79i) (Bond and Bergstrom, 2006). For reference, using RI =1.95-0.79i the MAC at 532 nm calculated for BC in the small particle limit
(assuming a material density of 1.8 g cm-3) is only
5.1 m2 g-1, but peaks at 7.5 m2 g-1 around
dp,VED=150 nm.
Comparison of measured and calculated MAC values
Another way to look at the extent to which Mie theory or the RDG
approximation can reproduce the observations is to compare the observed and
calculated MAC values as a function of particle size (or x) rather than
the σabs vs. size relationship (as in Fig. 2). Although the
MAC is related directly to the σabs, it is nonetheless useful
to consider the MAC values because they vary over a much narrower range than
do the σabs. The dependence of the MAC on dp,VED and
size parameter for both the observations and the models are shown in Fig. 3.
The observed MAC values generally increase with dp,VED or size
parameter at all wavelengths up to around dp,VED∼160 nm,
above which they plateau and are approximately constant. The ranges (minimum
and maximum) of binned observed MACs are provided in Table S4.
The MAC values observed here (Table S3) are substantially larger than the
value of 5.7±0.8 m2 g-1 at λ=405 nm reported by
Radney et al. (2014), who use a Santoro-type burner (i.e., co-flow diffusion
flame). They did not report any notable size dependence for their MAC values.
Our MAC values (especially for x>0.5) compare well with the value of
8.16 m2 g-1 at λ=532 reported by You et al. (2016)
(extrapolated from 7.89±0.25 m2g-1 at λ=550 nm)
for soot particles generated from the combustion of organic fuel stock over
the dp,VED range ∼80 to ∼210 nm. They also compare
favorably to the range of values 7.2 to 8.5 m2 g-1 reported for
λ=532 nm for dp,VED∼100 nm observed in Saliba et
al. (2016) for particles generated from a cookstove. Some particle size
dependence was reported by Khalizov et al. (2009), who used propane with a
Santoro-type burner, with reported MACs at λ=532 nm of 6.7±0.7 m2 g-1 for dm=155 nm particles and 8.7±0.1 m2 g-1 for dm=320 nm particles. This general
behavior was also observed for soot particles generated from a methane
diffusion flame in Dastanpour et al. (2017), with MAC values reported at
λ=660 nm of ∼5 m2 g-1 for dp,VED=50 nm and ∼7 m2 g-1 for dp,VED=100 nm. One
key reason that differences may exist between studies is that the BC
particles sampled had differing maturity. Soot maturity refers to the extent
to which the BC has a more disordered internal structure with high hydrogen
content (low maturity) vs. a more ordered, graphite-like structure with low
hydrogen content (high maturity) (Johansson et al., 2017). The absorption
cross section for BC likely increases with increasing soot maturity
(López-Yglesias et al., 2014).
The observations are compared with the calculated MAC values, based on the
fits from Fig. 2. MAC values from the RDG approximation are independent of
x, as the particle MAC is equal to the MAC of the individual spherules
making up the particle. Here, the observed MAC values in the plateau (large
x) regime correspond reasonably well with the MACs as calculated from RDG
values when the optimal RI values are used. The constant RDG MAC value at
λ=532 nm (= 8.8 m2 g-1) is slightly larger than
the often suggested value for atmospheric BC by Bond and Bergstrom (2006) of
7.75±1.2 m2 g-1 (extrapolated from 7.5 at λ=550 nm using 1 for the AAE) and is identical to that reported for soot from
a Santoro-type diffusion burner operating on propane (Zhang et al., 2008).
However, the MACs predicted from RDG overestimate the observed MACs at x<0.90, since the RDG fits are weighted by the greater number of data points
at x>0.9 where MACs are approximately constant.
For Mie theory, calculated MAC values for highly absorbing substances, such
as BC, have a characteristic shape where the MAC is constant up to x∼0.2, increases monotonically by ∼40 % until x∼0.9, and then
decreases rapidly towards larger x. The Mie theory curves calculated here
reproduce the observed values at x<0.90 (especially for λ=532
and 630 nm) but substantially underestimated MACs at x>0.90. This
facilitates an understanding of the Mie model underestimate of
σabs at large x (Fig. 2). Above x∼0.9 the calculated
Mie MAC declines with x for all wavelengths, but the observations indicate
that the MAC is constant. Because x occurs at smaller dp,VED for
shorter wavelengths than for longer wavelengths, the model–measurement
difference in both σabs and MAC is more noticeable at 405 nm
than it is at 532 nm, which is more noticeable than at 630 nm. This is a
consequence of a greater number of data points at x>0.9 at 405 nm,
past the peak in the Mie-calculated MAC.
The reasonable correspondence between the observed and Mie-calculated MAC at
smaller sizes is, however, somewhat surprising given that Mie assumes
spherical particles, yet the particles are not spherical. One potential
reason for the observed dependence of the MAC on particle size is that the
chemical and optical properties of the particles change with size, and the
different chemical composition coincidentally improves agreement with Mie
theory. For the diffusion flame, changes in the particle size distribution
were induced by changing the amount of dilution nitrogen in the sheath flow.
This can influence the maturity of the soot and consequently the soot
absorption (López-Yglesias et al., 2014). The observed increase in MAC
with dp,VED here exhibits some wavelength dependence, which could
reflect differences in the sensitivity of the MAC to maturation. The observed
differences between the MACs observed at the smallest x and the maximum MAC
values were 21 %, 41 %, and 37 % for λ=405 nm, λ=532 nm, and λ=630 nm, respectively. However, one additional
difference between the wavelengths is that at λ=405 nm, there is a
small increase in MAC going from dp,VED∼60 nm
(x405 nm=0.6) to ∼100 nm (x=0.9) after which the MAC
is constant, while at λ=630 nm there is much more of a continuous
increase in the MAC up to larger particle sizes. This could indicate that, at
some point, further changes in the soot maturity (composition) have no
influence at short wavelengths but do at longer wavelengths. As a
complementary explanation, Dastanpour et al. (2017) observed an increase in
primary spherule size with overall particle size for methane-diffusion-flame-generated soot. They attribute the increase in MAC with dp,VED to
changes in the internal structure and/or the degree of graphitization that
occur with changes in spherule size.
The observation of constant MAC values for mature soot (x>0.9 or
dp,VED∼160 nm at λ=532 nm) is an important result
in the context of how BC is commonly treated in climate models. Most climate
models simulate the optical properties of BC using spherical particle Mie
theory. The observations indicate that Mie theory will likely underestimate
the absorption by BC for particles with x>0.9 because, when the particles
are sufficiently absorbing, the attenuation of light by the outer layers of the
(spherical) causes the mass in the center of the particle to not interact
with the electromagnetic field (Bond and Bergstrom, 2006; Kahnert and
Devasthale, 2011). This suggests that the RDG approximation, or even an
assumption of a constant MAC, may provide a more accurate representation of
BC absorption than Mie theory in climate models, at least for uncoated BC.
This conclusion is independent of soot maturity, as the falloff in the MAC
with increasing size for Mie theory occurs for all strongly absorbing
particles. Although atmospheric BC particles are predominately generated
through the combustion of fossil fuels or through biomass burning (Bond et al.,
2013), flame-generated BC particles have been shown to be a suitable proxy
for atmospheric BC particles, both in terms of chemical bonding and
structural properties (Slowik et al., 2007; Hopkins et al., 2007). For
example, Hopkins et al. (2007) find that the sp2 content of ethylene and
methane flame soot are similar to diesel soot (63 %, 60 %, and
56 %, respectively) and have similar aromatic content. There is also a
reasonable similarity between SP-AMS mass spectra of flame soot and soot
particles in diesel exhaust or smoke from biomass burning (Onasch et al.,
2015a). The absolute values of the derived RI may be different for diesel or
biomass BC particles, but it can be reasonably assumed that Mie theory does
not reproduce the behavior of atmospheric BC particles.
The discrepancy between Mie theory and the observations is both size and
wavelength dependent. Consequently, the extent to which the true absorption
by BC is underestimated by a given atmospheric model due to the inappropriate use
of Mie theory will depend importantly on the assumed size distribution, both
the position and the width, and the wavelength. Using the effective RI values
determined here, we estimate that absorption is underestimated by around
20 %–40 % when Mie theory is used with reasonable BC size
distributions. The underestimate in absorption from the use of Mie theory
will be even larger if non-theory-specific (typically lower) imaginary RI
values are used, such as that suggested by Hess et al. (1998) or Bond and
Bergstrom (2006), as discussed by Stier et al. (2007). Consider that the
maximum MAC predicted using the Hess et al. (1998) RI (=1.75-0.44i) from
Mie theory is only 3.8 m2 g-1.
Box plots of MACs and MECs as a function of size parameter, with
Δx=0.18. The volume-equivalent diameters are provided at the top of
the plots for reference. Also shown are Mie theory curves for all particles
(black solid lines), Mie theory curves for x<0.90 (gold line), and RDG
curves (dashed line) calculated from the RI values in Table 1. RI fitting was
performed using only the absorption measurements. Panels (a)
and (b) are λ=405 nm data from the UCD CRD-PAS,
(c) and (d) are λ=532 nm data from the UCD
CRD-PAS, and (e) and (f) are λ=630 nm data from
the CAPS PMSSA. The poor match between the calculated Mie theory
curves at λ=405 nm reflects the difficulties in fitting spherical
particle Mie theory to the observed σabs,405 nm over the
entire size range. Although the same particle sizes were sampled over all
wavelengths, the size parameters sampled at each wavelength are different.
Therefore, there are different numbers of boxes for each wavelength. Note
that this figure is directly related to Fig. 2, the difference being that the
y axis values in Fig. 2 (cross sections) have been divided by the
per-particle mass to give the MAC or MEC. Points in each bin range from
N=10 at x1.62 to N=81 at x1.26 for λ=405 nm, N=5 at x0.36 to N=82 at x1.08 at λ=532 nm, and N=8
at x1.08 to N=67 at x0.9 at λ= 630 nm.
MEC and SSA values of nascent and coated–denuded soot
Measurement of bext and the MEC were made in addition to the
babs and MAC measurements (Fig. 3). Above, the RI fitting was
performed using only the absorption measurements, in part because the calculation of extinction using RDG requires additional assumptions regarding the
particle shape. Nonetheless, it is informative to compare the
σext and MEC observations to Mie theory calculations since the
overall climate impacts of BC depend on both absorption and scattering. As
with the MAC values, the observed MEC values also increase with x up to
0.90 (or dp,VED up to 160 nm), after which point they are
relatively constant (Fig. 3). The Mie theory MECs calculated using the RIs
determined by fitting the absorption measurements (Table 1) agree reasonably
well with observations when x<0.9 (using the fits that were constrained
to this range), but, as with absorption, Mie theory underestimates the MEC
above x=0.9. For a given particle size, there is somewhat greater scatter
in the observed MECs than in the MACs. This is likely a result of the
scattering being more sensitive to the shape of the soot particles than is
absorption and of the nascent and coated–denuded particle results being
combined here.
Given this, the dependence of the SSA on particle size is considered
separately for the nascent (more lacy) and coated–denuded (more compact)
particles. The coated–denuded particle SSAs increase with dp,VED,
most noticeably for dp,VED>100 nm, up until dp,VED∼160 nm. Above this size the SSA values are approximately constant at a
value of ∼0.30 (Fig. 4a). (Results at λ=532 nm are
shown in Fig. 4a, but there is a strong correlation between SSA at 532 nm
and at 405 or 630 nm; Fig. S5). In contrast, the nascent SSAs increase
slightly from dp,VED∼50 to 80 nm but above 80 nm are
approximately constant at ∼0.20. This demonstrates that particle
collapse leads to an increase in the SSA for BC, consistent with Radney et
al. (2014). This behavior is also consistent with modeling studies, which
have predicted that compact agglomerates exhibit higher SSA values than lacy
agglomerates, with an absolute increase of ΔSSA ∼0.1 (Scarnato
et al., 2013) or by a factor of 1.2–2.2, depending on the extent of
compaction (China et al., 2015a, b). The increase upon collapse is attributed
to the stronger scattering and electromagnetic coupling between spherules in
compact aggregates. Here, the difference between the SSA for nascent and
coated–denuded soot increases somewhat with particle size, which may result
from changes in soot maturity with size. The SSAs from this study compare
reasonably well to values reported previously at visible wavelengths. For
example, Saliba et al. (2016) report SSA = 0.16 to 0.26 for nascent soot
emitted from a cookstove; Schnaiter et al. (2003, 2006) report SSA = 0.2
to 0.3 for soot from a propane diffusion flame, SSA = 0.18 to 0.25 for
kerosene-derived soot, and SSA = 0.1 to 0.25 for methane diffusion flame
soot; and Sharma et al. (2013) report SSA = 0.18 to 0.25 for soot
generated from a kerosene lamp. However, the values reported here are much
smaller than the value of 0.5 reported in Radney et al. (2014). The
Mie-theory-calculated SSA values are similar to observations, although they show a
somewhat stronger increase with size and seem to plateau at larger SSA values
at large sizes compared to observations.
Box and whisker plots of λ= 532 nm (a) single-scattering albedo (SSA) and (b) absorption Ångström exponent
(AAE) as a function of volume-equivalent diameter produced from the methane
diffusion flame. The black points are coated–denuded data (coating material
is evaporated following coating with DOS or sulfuric acid) and are
potentially collapsed due to coating, and the gray points are nascent or
nascent–denuded data. The black lines are the SSA or AAE predicted by
spherical particle Mie theory, and the dashed black line in (b) is
AAE predicted from the RDG approximation using the effective refractive
indices listed in Table 1. The nascent or nascent–denuded boxes (black) are
shifted by X=0.05. Points in each bin range from N=6 at x0.36 to
N=26 at x1.08 for nascent (or nascent–denuded) and N=5 at
x0.54 to N=9 at x1.26 for coated–denuded points.
AAE of nascent and
coated–denuded soot
The wavelength dependence of absorption has been considered by calculating
the AAE using the measurements at λ=405 and 532 nm (Fig. 4b). The
nascent AAEs at larger particle sizes are slightly larger than the
coated–denuded AAEs, suggesting that particle collapse leads to a slight
decrease in the wavelength dependence of absorption. The average AAE was
1.38±0.36 (N=85) and 1.10±0.37 (N=135) for
nascent and coated–denuded particles, respectively. There is some indication
that AAE decreases with particle size. This may again be the result of the
soot maturity increasing (and the composition changing) with size. The AAE
values from Mie theory, based on the best-fit RI values determined above,
exhibit an increase to x∼0.5 where they peak and then decrease
sharply. This predicted decrease is inconsistent with the observations. The
AAEs calculated using the RDG approximation from the best-fit RI values are
constant.
BC is commonly assumed to have an AAE = 1 (Bergstrom, 1973). The
measurements here are consistent with this expectation for the collapsed
(coated–denuded) particles, but the nascent particles give an AAE that is
somewhat larger than 1. The observed AAE values are similar to results from
previous studies examining either freshly emitted soot particles or soot
particles containing very little organic material (Schnaiter et al., 2003,
2006; Kirchstetter et al., 2004; Bergstrom et al., 2002; You et al., 2016;
Sharma et al., 2013).
Optical properties of BC from the ethylene flat-burner flames
Results from BC2: sampling high above the burner surface
During BC2, the ethylene flame was sampled at a height of
∼20.3 cm above the surface. At this height, the particles were
likely reasonably mature, at least relative to sampling that was performed
further into the flame, as was done in BC3+. Particle optical properties
were quantified at λ=405, 532, and 781 nm, with two independent
measurements at 532 nm considered (NOAA and PASS-3). As with the methane
diffusion flame, the data were fit using Mie theory and the RDG approximation
to determine optimal, theory-specific, wavelength-dependent RI values
(Table 2). The soot particles from this flame had, overall, a greater amount
of intrinsic organic carbon associated with them compared to the particles
from the methane diffusion flame. As such, denuding even of the nascent
particles led to changes in the optical properties and particle masses. Thus,
the nascent and denuded particles are considered separately, and we focus on
the denuded particles. The range of particle sizes considered was also
overall smaller than that for the methane diffusion flame.
Theory-specific effective refractive indices for ethylene premixed
flame soot from BC2 and BC3+ retrieved via fitting Mie theory and the RDG
approximation to the σabs observations. Nascent and denuded
experiments are considered separately.
MACpeak
dpVED,peak (nm)/
Number of
Study5
Soot type
Method
λ (nm)
Instrument
n=m+ki
(m2 g-1)
size param.3
data points
BC2
Nascent
Mie1
405
PASS-3
2.21+0.86i
10.45(±2.06)
123/0.95
36
BC2
Denuded
Mie
405
PASS-3
2.19+0.91i
10.68(±1.97)
119/0.92
30
BC2
Nascent
RDG2,4
405
PASS-3
1.80+1.13i
10.06(±2.22)
36
BC2
Denuded
RDG
405
PASS-3
1.80+1.18i
10.32(±2.56)
30
BC2
Nascent
Mie
532
PASS-3
2.13+0.64i
6.92(±0.78)
178/1.07
36
BC2
Nascent
Mie
532
NOAA CRD-PAS
2.39+0.79i
8.02(±0.13)
169/0.99
43
BC2
Denuded
Mie
532
PASS3
1.96+0.83i
7.70(±0.90)
150/0.89
31
BC2
Denuded
Mie
532
NOAA CRD-PAS
2.56+1.11i
9.00(±0.80)
152/0.90
46
BC2
Nascent
RDG
532
PASS-3
1.80+0.78i
6.92(±0.43)
36
BC2
Nascent
RDG
532
NOAA CRD-PAS
1.80+0.85i
6.16(±1.75)
43
BC2
Denuded
RDG
532
PASS-3
1.80+1.08i
7.35(±1.69)
31
BC2
Denuded
RDG
532
CRD-PAS
1.80+1.13i
7.55(±0.68)
46
BC2
Nascent
Mie
781
PASS-3
2.16+0.76i
5.10(±0.40)
250/0.96
31
BC2
Denuded
Mie
781
PASS-3
2.84+0.74i
6.20(±0.20)
239/0.95
36
BC2
Nascent
RDG
781
PASS-3
1.80+0.50i
2.59(±0.33)
36
BC2
Denuded
RDG
781
PASS-3
1.80+0.73i
3.64(±0.81)
31
BC3+
Denuded
Mie
405
UCD CRD-PAS
2.11+1.03i
11.09(±2.82)
110/0.85
27
BC3+
Denuded
RDG
405
UCD CRD-PAS
1.80+2.05i
10.78(±3.11)
27
BC3+
Denuded
Mie
532
UCD CRD-PAS
1.46+0.54i
6.03(±1.12)
118/0.70
22
BC3+
Denuded
RDG
532
UCD CRD-PAS
1.80+0.76i
5.61(±0.46)
22
BC3+
Denuded
Mie
630
CAPS PMSSA
1.78+0.45i
3.93(±0.90)
188/0.93
22
BC3+
Denuded
RDG
630
CAPS PMSSA
1.80+0.55i
3.51(±1.55)
22
1 Uncertainties in MACs are 1σ from the least
χ2 fit. See Figs. S7 and S8. 2 The n values from the RDG
method are nonunique. Therefore, uncertainty estimates from this work are not
available. See text for details. 3 VED and the size parameter (x=πdp/λ) where the peak MAC occurs. 4 There are many
degenerate RI combinations that give similar quality fit to RDG theory. Thus,
a value of 1.80 was chosen for the effective real refractive index. 5 In
BC2 the ethylene flat-burner flame was sampled 20.3 cm between the burner
surface and the sampling inlet, and during BC3 the flame was sampled
ca. 5 cm between the burner surface and the sampling inlet.
Measured absorption cross sections vs. volume-equivalent diameter
(panels in the first column) and mass absorption and extinction coefficients
(MACs and MECs; panels in the second and third column, respectively) for
ethylene soot sampled 20.3 cm from burner surface of the McKenna flame as a
function of size parameter (x=πdp/λ) (bottom axes) and
volume-equivalent diameter (top axes). for BC2. Panels (a–c) show PASS-3
data at λ=405 nm, (d–f) show NOAA PAS data at λ=532 nm, (g–i) show PASS-3 data at λ=532 nm, and
(j–l) show PASS-3 data at λ=781 nm. The dashed black and
gray lines are the RDG fits to denuded and nascent data, respectively, and
the solid black and gray lines are Mie fits to denuded and nascent data,
respectively.
The retrieval of effective refractive indices for this flame using Mie theory
resulted in a good fit to σabs vs. dp,VED for all
λ and dp,VED (Fig. 5a, d, g, and j, and Figs. S6–S7). This
difference from the methane diffusion flame is in large part due to the more
restricted size range encountered here, with BC particles only up to
dp,VED=160 nm used. The RDG approximation yielded a reasonably
good fit across all particle sizes for the ethylene particles, although with
some overestimation at smaller sizes. The MAC values tended to increase with
particle size or size parameter, most obviously at 532 nm where the most
data points are available (Fig. 5b, e, h, and k). The range of binned MACs
shown in Fig. 5 is listed in Table S4. The MAC values determined using the
two PAS instruments at 532 nm differ somewhat, with the NOAA PAS MACs
slightly larger than PASS-3 MACs, although the differences are within the
measurement uncertainties. Most likely, this instrument difference stems from
differences in calibration methods.
Box and whisker plots of (a) single-scattering albedo (SSA)
and (b) absorption Ångström exponent (AAE) as a function of
size parameter and volume-equivalent diameter produced from the ethylene
flame during BC2. The SSA was calculated using data from the NOAA PAS, and the
AAE was calculated using data from the λ=405 nm PASS-3 and the
λ=532 nm NOAA PAS data. The black points are coated–denuded data
(coating material is evaporated following coating with DOS or sulfuric acid)
and are potentially collapsed due to coating, and the gray points are nascent–denuded points (evaporation of intrinsic organic matter produced from the
ethylene flame). The black lines are the SSAs predicted using spherical
particle Mie theory using the effective complex refractive index retrieved
from fitting the σabs,532 nm from the NOAA PAS. Here points
are binned with a constant Δx=0.18 for the purpose of comparison
between the different particle treatments. N=6 for nascent or nascent–denuded points, N=5 for DOS coated–denuded points, and Nx=0.54=11, Nx=0.72=8, and Nx=0.9=5 for H2SO4
coated–denuded points.
The observed MAC values tend to be larger for the denuded particles than for
the nascent particles, most likely due to less-absorbing organics present in
nascent soot that contribute ∼25 % of the particle mass (Cross et
al., 2010). The MAC values at λ=405 and 532 nm for the denuded BC2
ethylene flame soot are comparable, within uncertainty, to the MAC values for
the methane diffusion flame soot for the particles with dp,VED∼150 nm. This indicates that the BC from these two flames is similarly
absorbing in nature.
The impact of particle morphology on the SSA and AAE is considered by
comparing the results for nascent–denuded particles with coated–denuded
particles (Fig. 6). Nascent particles are excluded because the presence of
intrinsic organics can increase AAE if the organic material contains brown
carbon and can increase the SSA independent of the underlying BC morphology.
The Df,m values were 2.12±0.06, 2.49±0.07, and 2.17±0.06 for nascent–denuded, sulfuric acid coated–denuded, and DOS
coated–denuded soot, respectively (Cross et al., 2010). The nascent–denuded
and DOS coated–denuded particles have SSA values close to 0, whereas the
sulfuric acid coated–denuded particles have SSA values closer to 0.15 at x>0.5, consistent with particle collapse leading to an increase
in SSA (Fig. 6a). The SSA values for the nascent–denuded particles from this
flame are smaller than the SSA values for the nascent particles from the
methane diffusion flame. The reason for this is not clear but is likely
related to differences in the particle shapes and/or the sizes of the
spherules; the Df,m for the nascent–denuded particles here
(Df,m=2.12) is slightly smaller than for the particles from the
methane diffusion flame (Df,m=2.28) (Fig. S8). The sulfuric acid
coated–denuded particle SSA values from the BC2 ethylene flame are also lower
than the methane diffusion flame coated–denuded SSA values for a given size
(Fig. 6a), despite having similar Df,m values.
The AAE values (using the λ=405 and 532 nm pair) for
nascent–denuded and coated–denuded particles are generally similar when
compared over the same size range (Fig. 6b). However, the sulfuric acid
coated–denuded particle AAE values may be slightly smaller than for the
nascent–denuded or DOS coated–denuded particles; all are close to unity at x∼0.8. This indicates that changes in morphology do not lead to
substantial changes in AAE. The AAEs for the coated–denuded particles
decrease strongly with dp,VED, with the mean AAE ∼ 1.6 for
the smallest particles (x=0.5) and an AAE mean of ∼0.5 for the largest
particles (x=0.9). Values below 1 contrast with the methane data but
have been observed in a few previous studies (Clarke et al., 2007; Hadley et
al., 2008; Lack et al., 2008). While such a decrease with size is consistent
with Mie theory predictions, given the MAC results, it seems more likely that
the size dependence of the AAE is related to changes in the soot maturity
with particle size. To the extent that the larger particles, which tend to
have larger MAC values, are reflective of more mature soot, this suggests that
mature soot from this flame type has an AAE < 1.
Results from BC3+: sampling near the burner surface
During BC3+, the particles were sampled at variable heights above the
burner surface to select for particles in different size ranges, but most
often they were sampled from relatively close to the burner surface (5.1 cm)
compared to BC2 sampling conditions (20.3 cm). It is likely that these
different sampling conditions gave rise to particles with different chemical
properties. Multiple studies have shown changes in soot maturity and soot
optical properties as a function of sampling height in ethylene premixed
flames, though at a distance significantly closer to the flame front than
sampled here (Olofsson et al., 2015; Migliorini et al., 2011). The particles
during BC3+ had very little, if any, intrinsic organic carbon, unlike the
BC2 particles that were ∼25 % organic by mass. However, the small
organic content for BC3+ was likely a consequence of the use of a hot
sampling line prior to dilution. Consequently, the optical properties of both
nascent and denuded ethylene soot particles from BC3+ (sampled close to the
burner) differ substantially from the BC2 particles (sampled well above the
burner). Average MAC values over various size ranges are listed in Table S5.
In general, the MACs for BC3+ ethylene particles are similar to BC2
particles at λ=405 nm, whereas the λ=532 nm MAC for
BC3+ particles are smaller than those for BC2 particles. (Measurements at
λ=30 nm were not made during BC2, nor were measurements at 781 nm
made during BC3+.) At dp,VED>70 nm, the λ=405 nm MAC
values were approximately constant with increasing dp,VED
(Fig. S9). The behavior is consistent with methane diffusion flame
observations, but the constant MAC seems to occur at a lower dp,VED.The number of data points available for the BC3+ ethylene
particles is limited, making conclusions regarding the size dependence of
properties somewhat tenuous. The AAE for BC3+ ethylene particles are
reasonably independent of particle size (Fig. S10). The average value of
AAE405 nm–532 nm=2.01±0.21 for x>0.5, which is higher
than observed for the methane diffusion flame for this range of x (=1.18±0.35). These observations indicate that differences in sampling and soot
maturity result in different optical properties. Previous studies have
observed differences in optical properties, chemical composition, and primary
spherule size for different flame sampling heights (Bladh et al., 2011;
Olofsson et al., 2015; Migliorini et al., 2011). Absorption by less mature
soot appears to decrease more rapidly with wavelength than for more mature
soot, such that the MAC values in the mid-visible (e.g., λ=532 and
630 nm) are lower for less mature soot. These wavelength-dependent optical
results appear to match trends observed previously using active remote-sensing techniques to characterize particles within flames (Olofsson et al.,
2015; Migliorini et al., 2011). The extent to which this conclusion can be
generalized will require further investigation.