Introduction
Particles containing black carbon (BC) are largely anthropogenic and have
the third largest warming effect on the planet, after CO2 and methane
(Bond et al., 2013; Ramanathan and Carmichael,
2008). This makes BC-containing particles a likely target for climate change
mitigation policy (Anenberg
et al., 2012). However, there is high uncertainty regarding the climate
forcing of BC, and thus any potential benefits from BC control technologies
and policies (Jacobson, 2001). The uncertainty stems from the
variable optical, chemical, and physical properties of BC, which affect
particle radiative properties and lifetime, respectively. The properties of
BC-containing particles can change as atmospheric oxidants react with
gas-phase vapors; the products of these reactions can condense onto the
light-absorbing aerosol. Although this condensed material may not absorb
light itself, it can affect the lifetime and cloud-nucleation ability of the
BC-containing particle and increase the internally mixed BC absorption of
sunlight by as much as a factor of 2
(Bond
et al., 2013; Chung and Seinfeld, 2002; Jacobson, 2001; Lack et al., 2009;
Moffet et al., 2008; Ramanathan and Carmichael, 2008).
The burning of non-fossil-fuel biomass, either as biofuel, agricultural
refuse, or in wildfires, is the largest source of light-absorbing
BC-containing particles globally. However, the effect of aging on the
biomass-burning aerosol (BBA) is not well characterized
(Bond et al., 2013). In some cases, co-emitted organic
gases become oxidized and condense on the BBA, simultaneously reducing BBA
atmospheric lifetime and increasing BBA light absorptivity
(Lack
et al., 2009; Mikhailov et al., 2006; Yokelson et al., 2009). In other smoke
plumes, there has been no observed increase in organic matter (OM) with
aging, and in a few cases a decrease in OM has been reported, likely due to
dilution-induced evaporation
(Cubison
et al., 2011; May et al., 2013). In order to identify the contribution of
various chemical processes to the compositional changes in BC-containing
particles, measurements of particle composition must be carried out with a
temporal resolution similar to or faster than the timescale of the chemical
processes. Improved mass measurements of BC and internally mixed material
will enable more accurate model simulations that constrain the climate
forcing by black carbon and other aerosol components.
Particle mass spectrometers provide information about the chemical
composition of atmospheric aerosols with a time resolution comparable to
their transformation processes. In this paper we investigate the response of
two commercial laser-based particle mass spectrometers to the systematic
condensation of secondary organic aerosol (SOA) onto complex biomass-burning
soot particles. One instrument is an infrared (IR) laser vaporization
soot-particle aerosol mass spectrometer (SP-AMS, Aerodyne Research Inc.),
while the other is a laser desorption–ionization single-particle mass
spectrometer (LDI-SP-MS) known as the laser ablation aerosol particle time
of flight mass spectrometer (LAAPTOF, AeroMegt GmbH). Both instruments
provide compositional information regarding the core composition of BBA
particles as well as secondary condensed components on a timescale of 10 min or less. However, each has quantitative challenges related to
particle mixing state and morphology that are the subject of ongoing
research
(Corbin
et al., 2014; Onasch et al., 2015; Spencer and Prather, 2006; Willis et al.,
2014). Herein we evaluate the ability of the two instruments to measure the
composition of BBA composed of both a BC core and SOA coating, as we
simulate atmospheric aging by coating primary BBA particles with complex
secondary organic material in a chamber reactor.
We confirm that significant SOA coatings on BC cause only small changes in
the ion fragmentation patterns of elemental carbon (EC) in the SP-AMS. EC is
defined as ions detected by mass spectrometry that consist of only carbon
atoms. We show that the infrared laser beam in the SP-AMS has a different
effective beam width and therefore different detection efficiency for alkali
metals versus BC. The feasibility of quantitative measurements of OM by the LAAPTOF in complex and realistic particles containing
inorganic salts, refractory black carbon (rBC), and SOA was also investigated. We demonstrate that
there is a positive correlation between the LAAPTOF-measured OM ion signal
and the SOA mass condensed on BBA particles. This is significant given the
especially complex composition of BBA and is highly encouraging for
achieving mass quantitative single-particle measurements of other complex
ambient particle matrices. In this work, we will use the term “SOA” to
describe the condensed-phase organic material that was formed from α-pinene
ozonolysis. We will use “OM” to discuss all condensed organic matter,
including mass that is primary or secondary in nature. Organic ions measured
by mass spectrometry will be identified as OM because in this work we cannot
strictly differentiate between primary and secondary organic material.
Characterization of carbonaceous aerosol by SP-AMS
The SP-AMS is a variant of the conventional aerosol mass spectrometer that
includes an intracavity IR laser that can vaporize
light-absorbing refractory material. A conventional AMS includes a 600 ∘C heater to vaporize aerosol particles in a vacuum before
ionization (Decarlo et al., 2006; Jayne et al.,
2000). Two species of interest are difficult and impossible to detect,
respectively, with a conventional AMS: potassium and rBC. Refractory is an operationally defined term describing any material
not readily vaporized, such as by the 600 ∘C heater in a
conventional AMS. Potassium and rBC are chemical species that are used as
inert, nonvolatile tracers for biomass burning
(Andreae,
1983; Hennigan et al., 2011; Lee et al., 2016). The sensitivity of the
SP-AMS to rBC has been shown to increase with the addition of organic
coatings to rBC-containing particles. Willis et al. (2014) showed that
the increased sensitivity was due to increased overlap between the particle
beam and the IR beam that vaporizes the particles. The particles become more
tightly focused by the aerodynamic lens inlet with increasing coating
thickness and thus sphericity, causing a larger portion of the particle beam
to intersect with the IR laser beam. The fraction of rBC-containing
particles that are vaporized by the IR laser is defined by Onasch et al. (2012) as the shape-dependent collection
efficiency, EIR, and is the determining factor for the
sensitivity of the SP-AMS to rBC. Willis et al. (2014) measured the
EIR for rBC using a beam-width probe with a rBC aerosol
calibration standard (Regal Black, Cabot) coated to varying degrees with
dioctyl sebacate (DOS), a surrogate for primary or hydrocarbon-like OM. They
observed that the EIR was less than unity for uncoated,
collapsed rBC particles such as Regal Black but, for urban ambient
particles and DOS-coated Regal Black, EIR was close to unity.
However, real-world BBA has a different composition than Regal Black, with
particles that are less spherical and that contain inorganic salts (e.g.,
potassium chloride) that may be internally or externally mixed with rBC
(Huffman et al.,
2005; Li et al., 2003; Onasch et al., 2012). We will show that as BBA is
coated with OM, the SP-AMS ion signal response to the mass of potassium and
rBC increases. Furthermore, we will show that the SP-AMS response to
potassium increases more rapidly than that for rBC for equivalent amounts of
OM coating, despite the species being internally mixed. We believe that this
is in part due to the ability of components with low ionization energy to be
ionized directly upon vaporization
(Carbone
et al., 2015; Drewnick et al., 2006). However, most chemical components in
the SP-AMS undergo sequential vaporization followed by 70 eV electron
ionization of the neutral vapors. This allows for extensive but reproducible
molecular fragmentation and subsequent classification and detection of the
resulting ions in the high-resolution time-of-flight mass spectrometer. The
IR laser vaporization enables the subsequent ionization and thus detection
of refractory material in particles containing rBC that strongly absorbs the
IR laser energy. The vapors produced by the IR laser are ionized by the same
electron source used when vaporization is performed only with the 600 ∘C heater. The reproducibility of the ion fragmentation by
electron ionization has led researchers to investigate the possibility of
using the ratios of the various detected EC fragments
(Cx+) to perform source apportionment of rBC. Corbin et al. (2014)
showed that fullerene-rich particles produced a higher ratio of
C4+ to C3+ than particles from higher temperature
combustion did, which did not contain substantial fullerenes. Onasch et al. (2015) confirmed that fragmentation of larger graphitic molecules was a
minor source of C4+ cations in addition to the direct ionization
of vaporized rBC. This suggests that the EC fragmentation
pattern could be used to differentiate between rBC generated from different
types of combustion (e.g., diesel vs. biomass combustion). Here we
investigate the effect of atmospheric aging in the form of OM coatings on
these EC ion ratios.
Characterization of carbonaceous aerosol by LDI-SP-MS
The LAAPTOF is an LDI-SP-MS using a VUV excimer laser pulse to ablate and
ionize individual particles simultaneously. This enables the analysis of a
wider range of particle compositions and types than is possible with the
AMS, even when the IR laser is used. High-time-resolution measurements of
individual particles are achieved; an aerodynamic size and bipolar mass
spectrum can be obtained for each particle. LDI-SP-MS also provides much
greater mass sensitivity compared to the less efficient electron ionization
scheme because the laser ionization produces many more ions than electron
ionization per mole of analyte (Farmer
and Jimenez, 2010; Murphy, 2007; Pratt and Prather, 2012; Sullivan and
Prather, 2005). However, the large number of ions produced per particle can
result in poor MS resolution and/or ion plume chemistry after
desorption–ionization. The molecular fragmentation is also less reproducible
compared to 70 eV electron ionization due to laser shot-to-shot variability,
inhomogeneity of the laser pulse, and varying optical properties of
individual particles. Generation of ions directly from each particle
introduces important matrix effects where the particle's composition and
properties strongly dictate its interaction with the VUV laser pulse, the
distribution of laser energy to the particle's constituents, and the
resulting ion signal and fragmentation
(Gross
et al., 2000; Mansoori et al., 1996; Reinard and Johnston, 2008; Steele et
al., 2005; Thomson et al., 1997; Wenzel and Prather, 2004).
Despite these challenges, LDI-SP-MS can obtain quantitative measurements of
individual particle composition and its evolution. Effective demonstrations
of data analysis strategies for achieving quantitative LDI-SP-MS
measurements include restricting the analysis to particles of a similar
particle type based on their mass spectrum, normalization of ion peak areas
to total ion signal, the use of ion peak area ratios, sensitivity
calibration by comparison to co-located speciated mass measurements, and
signal averaging over numerous individual measurements
(Bhave
et al., 2002; Fergenson et al., 2001; Gross et al., 2000; Healy et al.,
2013; Jeong et al., 2011; Pratt and Prather, 2012; Spencer and Prather,
2006; Sullivan et al., 2009; Sullivan and Prather, 2005). Although the ion
plume chemistry can complicate analysis, the species known to cause
ion plume effects (e.g., water, inorganic salts) are generally present in the
mass spectra and can thereby inform the analysis
(Murphy, 2007; Murray
and Russell, 1994; Neubauer et al., 1998). In this study, we account for
laser shot-to-shot inconsistencies by averaging over particles of similar
size and composition.
Spencer and Prather (2006) used graphite spark-generated EC with condensed unleaded fuel to calibrate the ATOFMS LDI-SP-MS and
found a linear relationship between the ratio of summed select OM cation
signals to summed select EC cation signals, as the mass of OM condensed on
the EC was increased. However, as pointed out by Gysel et al. (2012), EC from graphite spark discharge has different
optical properties than combustion soot. Before such analysis can be used
with confidence on realistic combustion BBA and complex SOA, two key
differences between soot and BBA must be addressed, namely the variable
morphology of BBA and the presence of inorganic salts. Inorganic salts, such
as the potassium chloride observed in BBA, are very readily ionized
(Li et al., 2003;
Reid et al., 2005; Zauscher et al., 2013). They therefore can generate a
dense ion cloud that may reduce MS resolution and can interfere with the
formation/detection of other ions (Murphy, 2007). The presence of
potassium, and the irregular shape and strongly light-absorbing nature of
BBA,
presents an especially complex matrix for LDI-SP-MS. We evaluated the
ability of the LAAPTOF with its 193 nm laser to generate and detect OM ion
signal proportional to the mass of realistic SOA condensed on a BBA
particle.
Experimental methods
Experimental particle generation and conditioning
In Fig. 1 we display the experimental setup for the generation, coating,
and characterization of monodisperse BBA particles. For
these experiments, we burned European white birch bark in a cookstove and
sampled the smoke from the flaming phase 1 m above the fire through 1/2 in. o.d. stainless-steel tubing. We removed the gas-phase organics from the
smoke by pulling polydisperse smoke through a 20 ∘C annular
activated-carbon denuder with a diaphragm pump. The same pump then pushed
the denuded aerosol sample through a series of 85Kr neutralizers and a
differential mobility analyzer (DMA; TSI, model 3081) with a sample flow of
4.0 L min-1 and a non-circulating sheath flow of 18 L min-1 to select a narrow range
of mobility particle diameters. We then injected the monodisperse soot
particles into a 2 m3 Teflon chamber at a rate of 4.0 L min-1 until the
concentration was between 2000 and 14 000 particles cm-3. This
typically took less than 30 min. During this time, we continuously
fueled the fire to maintain flaming conditions. Before soot injection, we
partially filled the chamber with clean filtered dry air. Although the
diaphragm pump generated some small particles (with mobility diameters less
than 50 nm), they were smaller than the selected soot particles (143 to
220 nm mobility diameter) and thus not transmitted through the DMA into the
chamber.
Experimental setup including size-selected biomass-burning aerosol
generation, environmental aging chamber for sequential coating with α-pinene SOA, and subsequent online characterization with particle
instrumentation.
We characterized the nascent soot particles with a suite of particle
instruments described below and confirmed that the particles were nearly
monodisperse; doubly charged particles were always less than 10 % by
number. We then injected precursor gases for SOA formation. We generated
ozone by flowing oxygen through a corona-discharge ozone generator (Azco,
HTU500AC) until the chamber concentration reached ∼ 300 ppb.
We measured the ozone concentration with a UV photometric ozone monitor
(Dasibi, 1008-PC). We injected 0.1 or 0.2 µL aliquots of α-pinene (Sigma Aldrich, > 99 %) through an airtight heated
septum flushed with clean air. We used serial injections of α-pinene
to generate successively thicker SOA coatings on the soot cores. Chamber
experiments with several steps of SOA production typically lasted 3–4 h,
stopping after homogenous nucleation prevented additional coating of soot
seeds.
Particle component mass measurements
We characterized the soot and the SOA coatings using a number of different
methods. We used three instruments to measure the rBC: a Droplet Measurement Technologies single-particle soot photometer
(SP2) (Schwarz et al.,
2010; Stephens et al., 2003), the LAAPTOF mass spectrometer, and an Aerodyne SP-AMS
(Onasch et al., 2012). Extensive descriptions
of the SP2 and SP-AMS instruments are available elsewhere and thus we
provide only a brief overview. For the LAAPTOF we describe the salient
features below; a more detailed discussion of the LDI-SP-MS technique is
available in the form of a review article by Murphy (2007) and
literature descriptions of other closely related instruments
(Marsden et
al., 2016; Pratt et al., 2009; Sullivan and Prather, 2005; Thomson et al.,
2000; Zelenyuk et al., 2009).
The LAAPTOF is a commercially available single-particle mass spectrometer
(AeroMegt GmbH) that uses laser desorption–ionization to generate positive
and negative ions from individual particles that are subsequently detected
by a bipolar time-of-flight mass spectrometer (TOFWerk, AG)
(Gemayel et al.,
2016; Marsden et al., 2016). The particles are focused into the ionization
region by an aerodynamic lens. Normally the particle time of flight is
measured by light scattering between two 405 nm continuous-wave lasers
(OBIS, Coherent Inc.) to determine the velocity of each particle and thus
its vacuum aerodynamic diameter. The particle is then ablated and ionized
with an 8 ns pulse from a VUV 193 nm excimer laser (EX5, GAM Laser, Inc.)
that is triggered immediately by the second particle light scattering event.
The VUV pulse travels coaxially up the particle beam and hits the particle
in the ion extraction region, coincident with and orthogonal to the second
light scattering laser beam. During these experiments, the 405 nm scattering
lasers were not operational, and instead we free-fired the excimer laser at
10 Hz with an average laser pulse energy of 2.0 mJ and an average laser
fluence of ∼ 1.1 × 105 J m-2. We set and periodically
confirmed the VUV laser power using a laser power meter (EnergyMax,
Coherent, Inc.).
The SP2 rapidly heats individual rBC-containing particles
as they pass through an intracavity IR laser beam and then measures the
intensity of emitted thermal radiation resulting from particle
incandescence. The rBC mass of each particle is proportional to this light
intensity. We calibrated the SP2 rBC mass response with fullerene soot that
we size-selected using a DMA. We previously measured the batch-specific
effective density of this soot sample using a Cambustion centrifugal
particle mass analyzer
(Gysel
et al., 2011; Slowik et al., 2007). We measured the particle concentration
during these calibrations using a condensation particle counter (TSI model
3772). Further details are provided by Saliba et al. (2016), where the optical properties of the coated rBC
particles produced in these experiments are presented.
Both the SP-AMS and the LAAPTOF use the same aerodynamic lens inlet design
(Huffman et al., 2005; Liu et al., 2007).
The aerodynamic lens these instruments use efficiently transmit particles
with vacuum aerodynamic diameters (dva) between 150 and 700 nm. Particles smaller than 150 nm tend to be removed with the excess gas,
while particles greater than 700 nm may be impacted on the lens' critical
orifices. For the measurements presented here, some particles start smaller
than 150 nm dva and then grow into the ideal lens transmission
regime. This is very important for mass-based measurements like the SP-AMS,
but less so for individual particle analysis for the LAAPTOF. We will
discuss how this affects our results in Sect. 3.3.
The aerodynamic lens focuses particles into a collimated beam for
transmission into the SP-AMS detection region. The SP-AMS uses the same IR
laser as the SP2 to vaporize rBC along with any internally mixed components;
the vapors are then ionized by 70 eV electrons from a tungsten filament and
the ions analyzed by time-of-flight mass spectrometry
(Canagaratna et al., 2015b;
Onasch et al., 2012). The SP-AMS also contains a tungsten heater kept at
600 ∘C for particle thermal desorption, identical to the
conventional AMS design. Unless stated otherwise, we operated the SP-AMS with
both the IR laser and the conventional heater on for all experiments
described here. The largest uncertainty in mass measurements of rBC by the
SP-AMS is the particle collection efficiency, which is determined by the
particle shape and size. The fraction of particles that pass through the
laser beam and are vaporized (EIR) has been shown to be the
largest uncertainty for SP-AMS mass measurements using the IR laser. It is
caused by the incomplete overlap of the particle beam and the IR laser beam
(Huffman
et al., 2005; Onasch et al., 2012; Willis et al., 2014). A perfectly
collimated particle beam would result in all of the rBC-containing particles
passing through the center of the Gaussian IR laser beam energy profile,
where it is most intense. However, the particle beam can spread because
small and/or irregularly shaped particles diverge from an ideal beam
profile. This divergence is due to Brownian diffusion or uneven drag
distribution on the particles as they leave the aerodynamic lens
(Huffman et al., 2005; Liu et al.,
1995). The collection efficiency for nonrefractory material not mixed with
IR laser absorbing rBC (components that promptly evaporate at 600 ∘C) is determined by the particle bounce off of the heater. Note
that the cross-sectional area of the heater is much larger than that of the
IR laser beam. Therefore, particle detection of refractory components such
as rBC in the SP-AMS by IR vaporization is more sensitive to particle shape
than is the detection of nonrefractory components vaporized by the 600 ∘C heater.
We determined the fraction of total particle mass undetected by the SP-AMS
due to diverging, light-absorbing particles missing the IR laser beam. We
did this by measuring the ion signal from refractory species (elemental
carbon or potassium) while sequentially blocking portions of the particle
beam with a thin piece of wire (∅ = 0.41 mm), called the
beam-width probe. We compared the signal with the beam-width probe in place
to the signal observed when the wire was not obstructing any of the particle
beam
(Huffman et
al., 2005; Willis et al., 2014). We show this schematically in Fig. 2 for
the carbon cation signal, Cx+ (where x is a positive
integer). The ratio of the partially blocked signal to the unblocked signal,
hereafter referred to as the attenuation with the beam-width probe,
indicates how narrow the particle beam is. If the measured attenuation is
high, then in the absence of the beam-width probe most of the particles
will pass through the most intense IR laser region and be vaporized. If the
attenuation is low, then some of the particles will miss the laser and the
refractory material will not be detected. The IR laser intensity was
measured by Willis et al. (2014) to have a Gaussian
distribution with a σ≈ 0.18 mm, although this may vary
with IR laser power. As the beam-width probe is wider than the effective
width of the IR laser for rBC, anything less than complete attenuation of
the measured ion signal when the beam-width probe is at the center position
indicates that some of the EC mass on particles containing rBC will not be
detected. Therefore, EIR will be < 1
(Willis et al., 2014).
However, as we will show, vaporization of potassium occurs with a larger
effective IR beam width, and thus EIR is larger for
potassium than rBC.
The particle-beam width probe blocks a fraction of the particles
that would have been vaporized by the IR laser and thus the probe attenuates
the elemental carbon ion signal (Cx+) for rBC-containing particles.
We used six probe positions to establish the particle-beam width. The
measured ion signal is the integrated Gaussian profile shown at the top of
the schematic. The difference between the unblocked (x=-5.0 mm) and
blocked (0.0 mm) green integrated Gaussian profile plots is the attenuation
for that beam width probe blocking position, plotted as the y axis on the
top figure. The top panel shows the attenuation of nascent uncoated soot
(black asterisks), soot with a single coating of SOA (black and light green
crosses), and soot with a second coating of SOA (black and dark green
circles). A wider Gaussian particle-beam shape determined by the beam-width
probe analysis indicates the aerosol has a small diameter and therefore
EIR < 1.
Particle mobility measurements and calculations
During these experiments, there was evidence of subsampling in the aerosol
sampling lines, resulting in different particle concentrations reaching some
instruments. As a result, we do not compare the absolute concentration of
aerosol species measured by the various instruments. Instead, we perform our
analysis based on particle size measurements that are unaffected by the
flow-splitting issues due to the monodisperse aerosol used here.
We classified the size distribution of the BBA
with a scanning mobility particle sizer (SMPS; TSI Inc., model 3081 DMA and
3772 CPC) that measures particle mobility diameter (dmob) size
distributions and with the SP-AMS that measures particle vacuum aerodynamic
diameter (dva) in addition to composition. The SP-AMS measures
the vacuum aerodynamic diameter by accelerating particles into a vacuum and
measuring the time to ion detection after a particle passes through a
rotating 2 % slit chopper. We calibrated the particle time of flight
measurements using polystyrene latex spheres vaporized using the
conventional tungsten heater in the SP-AMS (Jayne et al.,
2000). It is important to note that for the fractal-like nascent soot
particles, the terms dmob and dva do not fully
describe the physical shape of a particle. Rather, they describe the
relationship between the drag force on that particle compared to either a
counterbalancing electrostatic force for dmob or the
acceleration modulated by the particle mass for dva
(DeCarlo et al., 2004).
In Fig. 3 we show aerodynamic size distributions for particles originally
selected with a DMA at 143 nm dmob. Though the nascent
particles were monodisperse in mobility space, the dva
distribution for the nascent soot particles was much wider than the
dmob distribution. This is probably due to the range of
particle masses and shapes that can exist at a given dmob for
highly irregular nascent soot (Zelenyuk et al.,
2008). The dva size range narrowed following the addition of
SOA coatings, likely reflecting the increasing uniformity as the irregularly
shaped particles became coated with SOA. When measuring the mobility size
distribution, we adjusted the sheath flow of the SMPS to ensure that we
captured the complete aerosol size distribution, at times operating at
sheath-to-sample flow ratios as low as 5:1. Although the lower
sheath-to-sample flow ratio reduces the SMPS size resolution, it does not
affect the accuracy of the mode size determination.
Normalized particle size distributions for coating experiments from
two instruments. Traces are mass distributions vs. vacuum aerodynamic
diameter (dva) for two refractory species measured by
the SP-AMS, potassium (purple circles), and rBC (black squares). The filled-in
gray curve is the volume distribution (d3) versus mobility diameter
(dmob) measured by the SMPS. Measurements of the
original uncoated nascent biomass-burning soot particles are shown in the top
panel, while the middle and bottom panel show measurements after we applied
the first and second SOA coatings, respectively. The uncoated particles had
broad mass distributions with modes well below the mobility volume mode,
indicating fractal particles with a wide variability in particle shape. The
mass distribution narrowed and overtook the volume distribution following
condensation of SOA because the particles became more spherical and
homogeneous with respect to shape and composition, with a density greater
than 1.0 g cm-3.
As a nascent soot particle becomes coated with SOA, the particle grows and
gains organic mass. Thus dmob, dva, and the
organic mass per particle all increase. We calculated the SOA mass per
particle (mSOA) from mobility and mass measurements following
the method described in
Slowik et al. (2004),
with some differences as described here. In addition to dmob
and dva, we measured the single-particle black carbon
mass (mrBC) with the SP2. The SP2 also confirmed that
> 97 % of the nascent particles were composed of significant
amounts of black carbon. We assumed that the black-carbon density
(ρrBC) was 1.8 g cm-3
(Park et al., 2004). For thickly coated particles,
with mSOA ≫ mrBC,
we assumed that the particles were effectively spherical and determined the
density of the SOA (ρSOA) to be 1.3 g cm-3 by taking
the ratio of dmob and dva. Knowing the two
densities, we then iteratively solved for the particle dynamic shape factor
(χ), volume equivalent diameter (dve), and the
average mass of SOA per particle (mSOA) using Eqs. (1), (2),
and (3) below, taking into account the Cunningham slip correction factor
(Cc).
dmob=dveCcdmobχCc(dve)dva=dveχ(ρ0)mrBC+mSOAmrBCρrBC+mSOAρSOAmrBCρrBC+mSOAρSOA=π6⋅dve3
We neglected the contribution of ammonium, sulfate, and nitrate to particle
mass and volume as they are small (< 10 %) relative to rBC and
SOA. Unless specified otherwise, SOA mass per particle is calculated in this
manner. The combustion of solid biomass fuel generates particles with shapes
and compositions that vary more widely than those produced from controlled,
well-mixed internal-combustion engines
(Reid et al., 2005;
Schwarz et al., 2008). Relative to engine exhaust soot, biomass-burning
particles are typically larger and contain more non-carbonaceous components,
such as inorganic salts
(Bond et
al., 2006; Li et al., 2003; Reid and Hobbs, 1998). Electron microscopy
studies of aerosol particles have shown that biomass burning can result in a
wide range of particle compositions and morphologies
(Li et al., 2003; Pósfai and Buseck,
2010). For this work, based on our measurements it appears that most of
the particles initially consisted of mostly rBC, with trace primary organic
material and inorganic material, including potassium salts. The particles
maintained their initial core composition mass as SOA was condensed onto
them. This made the particles increasingly homogeneous in terms of
composition and shape. Single-particle measurements by the SP2 showed that
> 97 % of the particles detected by light scattering contained
> 0.7 fg of rBC. In cases where the uncoated, non-BC-containing
particles were too small to be detected by the SP2 via light scattering, the
fraction of rBC-containing particles was monitored after the particles were
grown with SOA to detectable sizes. The fraction of rBC-containing particles
did not change, except in cases where there was substantial and obvious new
particle formation. We suspect that the largest variability in particles was
with regard to the amount of potassium in a particle and with respect to
particle shape. We do not have an estimate of variability in individual
particle potassium content. Variability in particle shape is observed by the
broad distribution in vacuum aerodynamic diameters. As the particles became
coated, and therefore more uniform in shape and composition, the vacuum
aerodynamic diameter distribution narrowed. The coating of fractal-like soot
with organics has also been shown to cause structural collapse of the
particle, potentially affecting its light absorption cross section
(Cross et al., 2010;
Ghazi and Olfert, 2013; Zhang et al., 2008).
Results and discussion
Effect of SOA condensation on soot-particle shape
Figure 4 shows the calculated dynamic shape factor, χ, for
particles as they were coated with SOA. Nascent soot particles are very
fractal-like, with χ > 1.6. This is within the
range of previously observed soot from fuel-rich combustion
(Slowik et al., 2004).
For context, Schwarz et al. (2008) measured the thickness of
coatings on ambient BBA using an SP2. The instrument was onboard an aircraft
that transected a biomass-burning plume. They determined aged ambient
particles with a rBC core mass equivalent diameter of 200 nm (BC mass of 8.4 fg) had a coating thickness of 79 nm ∼ 1 h after emission
from the biomass-burning source. For comparison, we assume the BBA particles
discussed in Schwarz et al. (2008)
are coated with SOA with a density of 1.3 g cm-3. The SOA coating mass
would then be 25.7 fg of OM mass per particle and would result in a mostly
spherical particle. Although we cannot say the particle coating observed by
Schwarz et al. (2008) consisted entirely of OM mass, the volume equivalent
of any secondary component such as sulfate would also result in a spherical
particle shape, with a SOA to rBC mass ratio greater than > 3.2
(Ghazi and Olfert, 2013). As shown before, because the
particle shape and size influences the particle-beam profile and the beam
width at the IR laser, the shape factor influences the total rBC signal
detected by the SP-AMS.
Dynamic shape factor (χ) of soot particles became more
spherical (χ→1) with increasing SOA mass. Nascent particles (black)
were coated sequentially, and after 5 fg of SOA per particle had condensed
the particles were mostly spherical in shape. Additional coatings served to
increase the diameter of the now effectively spherical particles. Differently
colored and shaped symbols indicate different initial selected soot core mobility
diameters: red triangles were initially 142 nm, teal squares were 188 nm,
and purple diamonds were 220 nm.
Monodisperse biomass soot particles to which two successive
additions of SOA were applied. Fractal-like, monodisperse soot particles with
initial mobility diameter of 143 nm grew via condensation of successive SOA
coatings following discrete injections of α-pinene vapor (indicated
by vertical lines) into a chamber containing ozone. Particle growth is
evident in the volume-weighted SMPS mobility size distribution (blue-black,
top) and the SP-AMS measured OM (green closed circles) and EC (black open
squares) ion signal as a function of time. SP2 rBC mass measurements (red
dotted line) reveal the steady decay of soot-particle mass concentration due
to chamber wall loss, followed by greater loss when we purged the chamber at
t > 1.5 h.
Sequential coating of monodisperse soot in a smog chamber
In Fig. 5 we display a time series from a particle coating experiment for
initially monodisperse particles with dmob=143 nm. All of
the experiments followed this general pattern. We show the particle mobility
size distribution from the SMPS, the rBC mass from the SP2, and the SP-AMS
signal for both rBC and organic material (OM). Prior to coating
(t < 0), the SP2 and SP-AMS rBC signals decreased as particles were lost to the
chamber walls. After each α-pinene vapor injection, which drove SOA
coating, the particle mobility diameter increased and there was also an
increase in the SP-AMS OM signal. After the second injection (t = 0.6 h), the
SP-AMS rBC signal increased while the SP2-measured rBC mass
concentration continued to decay, as expected due to particle wall loss.
This difference demonstrates that the increase in SP-AMS rBC signal with
thicker SOA coatings was almost certainly due to an increase in the IR laser
beam particle collection efficiency (EIR) of the SP-AMS and
not an increase in the actual rBC mass present in the chamber.
Stick integrated high-resolution mass spectra from the SP-AMS for
nascent (top) and thickly SOA-coated soot (bottom). Mass spectra were
collected with the IR laser on and have been normalized by total rBC mass as
measured by the SP2 to account for particle wall loss. Peak bar colors
correspond to the assigned chemical components for each unit m/z ion peak,
based on analysis of the high-resolution mass spectra. The nascent soot
spectrum is rich in refractory black carbon, inorganic ions, and organic
fragments. The spectrum from the SOA-coated soot, however, is
dominated by the OM from the secondary organic aerosol. However, increased
sensitivity to larger EC fragments from rBC in the coated particles is
obvious.
Particle composition from mass spectrometry
In Fig. 6 we show speciated high-resolution mass spectra from the SP-AMS
for nascent and thickly coated biomass-burning particles, detected with the
IR laser on. The coating was SOA formed from α-pinene ozonolysis.
High-resolution peak fitting was done using PIKA version 1.15
(Decarlo et al., 2006). Highly resolved ions were classified
into a species and are displayed at unit mass resolution according to the
fragmentation table therein. With respect to this dataset, the fragmentation
table was only used to adjust the apportioning of CO2+ and
C1+ ions. To account for CO2+ in the gas phase vs.
particle phase, the fragmentation table was adjusted using HEPA filter
gas-only measurements. To account for C1+ that may result from the
fragmentation of nonrefractory OM components, the fragmentation table
specifies that the amount of C1+ attributed to rBC is limited to 0.625⋅C3+,
the ratio observed for the rBC calibrant, Regal Black
(Onasch et al., 2012). Recent studies have
begun to investigate the degree to which nonrefractory OM contributes to
larger EC ion fragments (Cx+, where
x > 1) using IR laser vaporization. This has important
implications for quantitative measurements of rBC using the SP-AMS and also
for source apportionment based on the ion fragmentation pattern of rBC
(Corbin et al., 2014).
We measured refractory and nonrefractory material from nascent soot using the
SP-AMS with the IR laser on. High-resolution analysis of nonrefractory
organic material showed significant contribution from aliphatic ions,
C4H7+ (m/z+55), C4H9+ (m/z+57),
C5H9+ (m/z+69), and C6H9+
(m/z+81). This is consistent with other
AMS measurements of fresh, flaming-phase biomass-burning emissions
(Corbin
et al., 2015; Cubison et al., 2011; Hennigan et al., 2011). We also observed
the highly oxygenated ions C2H3O2+ (m/z+60) and
C3H5O2+ (m/z+73), which are common tracer ions for
biomass burning. However, C2H3O2+ was less than 0.5 %
of the total organic signal for nascent soot, much less than what has
previously been reported in BBA measurements
(Aiken
et al., 2010; Corbin et al., 2015; Cubison et al., 2011; Hennigan et al.,
2010, 2011; Lee et al., 2010). This is likely due to low cellulose content in
the bark that we burned, resulting in less formation of the anhydrosugars
including levoglucosan compared to burning wood (Branca et al., 2007). The
average oxidation state of carbon (OS‾c=2O : C – H : C) for the organic fragments (including HO+,
H2O+, and CO2+) was -0.56 ± 0.25. This low
oxidation state is consistent with primary organic aerosol
(Canagaratna et
al., 2015a; Kroll et al., 2011). We also
observed refractory material, including elemental carbon series
(C1+–C9+) and some metals including potassium, rubidium, and
zinc. The alkali metals have very low ionization energies, and thus may
become ionized by heating in the IR laser. This ionization mechanism is
independent of the 70 eV electrons that usually ionize neutral vapors in the
SP-AMS. A broad abnormal ion peak shape of the metals indicates that they
underwent single-step thermal ionization in the IR beam, rather than
conventional two-step thermal vaporization with subsequent electron
ionization
(Allan
et al., 2004; Carbone et al., 2015; Corbin et al., 2015; Drewnick et al.,
2006). Condensation of SOA from α-pinene ozonolysis increased the
signal from most organic fragments. Especially notable was the signal
increase for singly oxygenated organic fragments
C2H3O+ (m/z+43), C3H3O+ (m/z+55),
C4H7O+ (m/z+71), and C5H7O+ (m/z+83). Reduced
fragments such as C5H9+ (m/z+69) also increased with SOA
coating. After a thick coating (with an SOA : BC mass ratio of
approximately 9) the average oxidation state of carbon was 1.15 ± 0.04.
This is consistent with the α-pinene ozonolysis SOA formed in smog
chambers
(Chhabra et al., 2010; Shilling et
al., 2008); however, the SOA mass spectra we obtain show relatively less
fragmentation than similar α-pinene SOA mass spectra obtained with a
conventional AMS using the 600 ∘C heater. This suggests that a
significant fraction of the particles were vaporized by the IR laser
(> 97 % of the BBA contained sufficient black carbon that
could be measured by the SP2) and that this produced marginally less
fragmentation; the much lower f44 (fraction of total ion signal at m/z+ 44) is also consistent with this hypothesis. Canagaratna et
al. (2015b) used near-threshold VUV
ionization with a SP-AMS to confirm that ionization via the IR laser resulted
in significantly less fragmentation for pure components, relative to when
vaporized by the 600 ∘C heater.
Attribution of rBC in the SP-AMS mass spectrum would be straightforward if
graphitic material had a consistent fragmentation pattern and if all
elemental carbon fragments (the Cx+ family) arose only from rBC.
However, studies using high-resolution transmission electron microscopy
(HR-TEM) have shown that rBC can have varying degrees of disorder that result
from formation conditions (Vander Wal and Tomasek, 2004). Onasch et
al. (2015) showed the IR laser in the SP-AMS may cause restructuring of the
rBC due to annealing. Annealing has been observed in other graphitic particle
systems using HR-TEM and a pulsed laser (Vander Wal and Choi, 1999; Vander
Wal and Jensen, 1998). It is also well known that organic matter produces
C1+ and C2+ fragments following electron ionization (Alfarra,
2004; Corbin et al., 2014), but we have evidence that SOA either produces
larger Cx+ fragments or changes the rBC fragmentation
pattern. In Fig. 7 we show the ion peak area ratios for
C4+ / C3+ and C6-9+ / C3+ for
size-selected soot particles as they became coated with SOA. The SOA mass per
particle on the x axis is the mass calculated from particle size
measurements, described in Sect. 2.3. Both ratios increase consistently with
coating mass, with a larger slope for smaller cores (with smaller
mrBC). This is also apparent in Fig. 6, where black
sticks representing EC from rBC are evident for m/z≥72 after SOA
coating, while they were not detected in the nascent soot. We propose three
possible explanations for these trends: (1) SOA coating changes the
fragmentation pattern of EC to reduce the C3+ signal, thereby
enhancing the apparent ratio of other EC ions relative to it; (2) SOA
generates significant signal for C>3+; or (3) the abundant signal at
the same nominal masses as C>5+ causes the HR peak fitting to
incorrectly attribute some signal to C>5+. During HR peak fitting,
the user decides whether to include an ion based on the residual signal. The
residual signal is the amount of measured signal that is not reproduced by
the fitted HR ion species. Although the addition of the higher Cx+
fragments appear to reduce the residual signal, the peak fitting without them
is still very good (residual < 0.05 %). Furthermore, the signal
at C1+–C5+ is much higher and better resolved in the peak
fitting due to fewer available peaks to fit. Although the causes of these
trends are not clear, the trends themselves indicate that soot source
apportionment by the SP-AMS might be most meaningful for rBC that has been
thermally denuded to remove any coatings as this would remove any effect of
OM on the Cx+ ratios measured from the rBC.
SP-AMS measurements of carbon ion family (Cx+) peak area
ratios versus SOA coating mass on three different BBA soot core sizes,
indicated by symbol colors. SOA mass determined from SP-AMS measurements of
all identified OM mass (as in Fig. 4). The ratio of C4+ and C∑6-9+ compared to C3+ increases with organic coating due
to either contributions to the Cx+ family from nonrefractory OM or a
decrease in relative fragmentation of EC to C3+.
We detected small amounts of nonrefractory species other than OM, amounting
to less than 10 % of the total mass. Ammonium sulfate condensed onto the
particles after they were injected into the chamber. This was probably caused
by the formation of sulfuric acid from oxidation of SO2(g) and
subsequent neutralization by ammonia in the dry chamber. We also detected
nitrate ions after addition of ozone to the chamber, as well as after
injections of α-pinene, suggesting the formation of some
organonitrates (Farmer et al., 2010; Zhang et al., 2006). The
NO+ / NO2+ ratio was 2.2, as opposed to 1.45 for ammonium
nitrate calibration particles. We observed chloride with both SP-AMS
vaporizer modes (IR laser on and off), while the potassium signal was much
larger when we operated the SP-AMS with the laser on. This is further
evidence that the potassium is internally mixed with black carbon and that
it underwent one-step thermal ionization in the IR beam
(Corbin
et al., 2015; Drewnick et al., 2006; Lee et al., 2016). Sulfur dioxide,
ammonia, nitrogen oxides, hydrogen chloride, and chloride salts are
known common primary emissions from biomass burning
(Levin
et al., 2010; Li et al., 2003; Reid et al., 2005; Stockwell et al., 2015;
Zauscher et al., 2013).
Positive polarity mass spectra and ion assignments for two
representative particles analyzed by the LAAPTOF. The elemental carbon
fragment series (Cx+) dominates the nascent soot spectra in the
bottom, while the contribution from oxidized organic matter is increased in
the top panel for a soot particle that was coated with α-pinene SOA.
Figure 8 shows single-polarity mass spectra obtained by the LAAPTOF for
individual particles before and after coating with SOA. The characteristic
ion series for elemental carbon included C1+, C2+, and
C3+. We also observed anions C1- and C2- for some
particles, but not as consistently. We observed NO+ in both coated and
uncoated particles. We believe that this may be the result of NOx and
O3 combining in the chamber and forming NO3 radicals. These
NO3 radicals may then react with water or organic vapors to form
HNO3 or organonitrates, respectively. Either species could then condense
onto the existing particles and fragment to NO+ in the mass
spectrometers. K+ was readily identified by its isotopic abundance at
m/z+39 and +41 (Bahadur et al., 2010; Healy et al., 2012; Silva et al.,
1999). OM fragments tentatively identified as CO+ at m/z+28 and
C2H3O+ at m/z+43 were measured from both nascent and
oxidized particles. We also observed sulfur and sulfate ions (S+,
SO+, and HSO4-), likely fragments of sulfates. Zn+ was
observed and identified by its isotopic fingerprint.
As previously stated, for these experiments the instrument was operated in
UV laser free-fire mode, without the aid of light scattering modules to more
efficiently detect particles and trigger the excimer. As a result, only 454 individual
mass spectra were collected by the LAAPTOF across the four
experiments. Of 454 particles, 31 contained only potassium and inorganic
species, with no detectable elemental carbon; 4 of 454 particles contained
S+ or SO+. Of the 454, 11 had a negative ion mass spectrum in
addition to a positive ion mass spectrum; 5 of these negative ion spectra
contained only HSO4-, 2 were only C1-, and the remaining
were a richer spectrum of elemental carbon with larger signal from even
carbon number anions, consistent with previous observations
(Bloomfield et al.,
1985; Onasch et al., 2015). Of 454 particles, 104 contained Zn+ and 281 particles contained NO+.
SP-AMS sensitivity to refractory species
We determined the ion response of the SP-AMS to biomass-burning particles as
a function of SOA coating mass using two ions produced from refractory
biomass-burning material – K+ and C3+ – normalized by the rBC
mass concentration measured by the SP2 (rBCSP2). In Fig. 9 we
show the K+ and C3+ SP-AMS ion signal per rBCSP2
mass during progressive SOA coating experiments. The plotted points are
10 min averages for four batch chamber experiments using different initial
soot core sizes. We determined the SOA mass per particle using the method
described in Sect. 2.3. Different colored and shaped traces indicate the
mobility mode diameter of the initial soot core injected into the chamber,
and the black dots indicate the initial nascent soot, prior to SOA coating.
We used the SP2-measured rBC mass to correct for particle wall loss in the
chamber because the rBC mass measurement of the SP2 is insensitive to
particle shape and coating thickness for particles with at least 0.7 fg of
rBC present (Schwarz et al., 2010). The smallest particle mode measured in
these experiments was 1.1 fg of rBC, well above detection limits for the
SP2. All of the wall-loss-corrected ion signals in Fig. 9 were also
normalized to the uncoated soot conditions. This was done for three reasons.
First, the inherent variability in biomass burning limits how much one can
expect uncoated soot particles to have similar amounts of K or rBC. However,
within a given coating experiment, the average composition of the soot-particle core is guaranteed to be the same. Second, it allows for ready
comparison of the data in Figs. 9 and 10, which compares the changes in the
two ions directly. Finally, the normalization is a small change relative to
the effect induced by coating the soot particles with SOA, which is the focus
of this work. Prior to normalization, initial wall-loss-corrected values for
C3+ (K+) for nascent conditions agreed within 30 %
(20 %), compared to the sometimes 300 % (600 %) change due to
coating the particles with SOA.
SP-AMS-measured biomass-burning ion signals for C3+ (top)
and K+ (bottom) for three mobility-selected core particle sizes versus
four different metrics. These signals are corrected for particle wall loss
using SP2-measured rBC, and then normalized to values obtained from the
uncoated nascent soot. Symbols are colored/shaped by their initial soot core
mobility diameter (SP2-measured rBC mass per particle), prior to SOA coating,
where red triangles indicate 143 nm (1.2 fg), teal squares indicate 187 nm (2.0,
2.3 fg, two replicate experiments), and purple diamonds indicate 220 nm
(3.9 fg). The nascent particles are indicated by black dots, and the lines
connect them with the data points following subsequent SOA coatings. The left
panels (a, e) display the normalized ratio of the SP-AMS-measured
C3+ and K+ ion signals, respectively, to the SP2-measured black
carbon mass concentration as a function of the mass of SOA per particle. The
SP-AMS-measured ion to SP2-measured rBC mass ratio is normalized to the
uncoated soot values. Panels (b, f) display the
attenuation of the SP-AMS ion signal caused by the beam-width probe in the
center of the particle beam; greater attenuation indicates a more collimated,
narrow particle beam, as expected for particles that are larger and/or more
spherical. Panels (c, g) display the mode mobility
diameter of the particles that produced either the C3+ or K+ ion
signal. The vertical dashed line indicates the lower size cutoff for the
aerodynamic lens. Panels (d, h) display the particle
dynamic shape factor; particles start as less spherical (χ > 1.0) and move towards sphericity (χ=1) as more SOA
mass is added. The measurement uncertainties are indicated by the vertical
error bars and represent the standard deviation of 1 min AMS integration
time from the 10 min averages presented by each symbol.
Potassium is a useful marker for refractory BBA material because it has a
high signal-to-noise ratio, it is not produced by fragmentation of
nonrefractory OM, and it is a nonvolatile unreactive conserved tracer. By
turning off the IR laser we confirmed that very little of the potassium
signal resulted from particles that hit the conventional 600 ∘C
heater (< 1 % of total K+ signal with IR laser on).
Therefore, the potassium we observed with the IR laser on was internally
mixed with rBC. We selected C3+ at m/z+36 as an ion of interest
because of its large contribution to the total rBC Cx+ family signal
(Fig. 4) and the very low contribution from OM. The C3+ signal from
OM is less than 0.08 % of the total OM signal when the IR laser is off
and particles are vaporized by the heater. One important difference between
the two measured species, K+ and C3+, is the method of ion
formation within the SP-AMS. rBC undergoes the conventional process of IR
vaporization followed by 70 eV electron ionization of neutral vapors.
Potassium, however, has a very low ionization energy and at high temperatures
can undergo one-step thermal ionization, without interacting with 70 eV
electrons to become ionized
(Corbin
et al., 2015; Drewnick et al., 2006; Svane et al., 2004; Zandberg, 1995).
Furthermore, rBC particles approach ∼ 4000 K in the IR laser
before vaporizing, while potassium may vaporize or ionize at temperatures less
than 1500 K (Svane et al., 2004).
Panels a and e in Fig. 9 reveal an increasing SP-AMS response to both
potassium and rBC as the SOA coating grew. Panels b and f show the
attenuation of the K+ or C3+ ion signal normalized to
SP2-measured rBC that resulted from moving the beam-width probe into the
center of the particle beam. The attenuation increased as the SOA coating
thickened because the particle beam narrowed. The increased response of the
SP-AMS with successive coatings of SOA is thus due to an increase in the
SP-AMS IR-beam particle collection efficiency, EIR. However,
the attenuation of K+ rises steeply, whereas the attenuation of
C3+ steadily continues to rise more shallowly with coating
thickness.
The IR laser beam cross section that vaporizes particles has a roughly
Gaussian intensity profile. The difference we observe between C3+ and
K+ attenuation suggests that the effective IR beam for vaporizing rBC,
described by Willis et al. (2014), is narrower than the effective beam for
thermal ionization of potassium. That is, rBC must pass through a region of
higher laser energy density near the center of the IR beam to be vaporized,
ionized by the electron source, and then detected,
relative to the lower laser energy density that is required for thermal ionization of potassium.
If an internally mixed particle containing potassium
and rBC passes through the center of the laser, two processes will take
place. rBC will become vaporized and ionized, and potassium will be thermally
ionized. Since turning off the IR laser reduced the potassium signal to
< 1 % of that when the laser was on, we know that the observed
changes in K signal result from rBC particles internally mixed with some K.
Thus, our measurements show that particles may pass through the potassium
thermal ionization region of the IR beam but miss the smaller rBC
vaporization region.
The ability of an aerodynamic lens, such as that used on the SP-AMS, to
effectively focus a particle depends on particle morphology. Panels c and g
in Fig. 9 illustrate the increasing response of the SP-AMS to coated rBC as a
function of vacuum aerodynamic diameter. The increasing response is likely
due to the product of the aerodynamic lens transmission efficiency and the
overlap between the particle beam and the IR laser, EIR.
However, it appears that the largest increase in particle sensitivity occurs
when the particle dva increases beyond 200 nm. The
lower size cutoff of the aerodynamic lens is 150 nm, as discussed in
Sect. 2.2 and illustrated by a black dotted line in Fig. 9c and g. Any
changes in sensitivity for dva > 150 nm
are therefore minimally affected by particle lens transmission. The particles
grow as more SOA mass is condensed, and the larger particles are focused more
efficiently by the aerodynamic lens towards the center of the IR laser,
resulting in a larger instrument ion signal response with both increasing
particle size and increasing SOA mass. The other factor that influences
effective focusing of the particles in an aerodynamic lens is particle shape,
reported as a calculated dynamic shape factor, χ. As seen in panels d
and h, the shape factor decreases towards unity (χ=1 for spheres) as
the particles become more thickly coated with SOA. Although the particles
become nearly spherical after a few coatings, they still need to grow to a
sufficiently large diameter to be successfully focused into the IR laser beam
of the SP-AMS. Neither particle shape nor diameter alone is sufficient to
describe EIR. Even at dva=250 nm
there is a factor of 2 difference in response to particles with different
dynamic shape factors. However, the largest increase in instrument response
occurs after the particles are mostly spherical (χ < 1.2),
which as we discussed in Sect. 2.3 may describe the state of rBC particles in
ambient biomass-burning plumes after less than 1 h of aging. This increase
in instrument signal is driven by growth from condensation of additional SOA.
Panels d and h demonstrate this by plotting the refractory ion signal ratios
versus the dynamic shape factor (χ) calculated as described in
Sect. 2.3.
Figure 10 shows the same normalized K+ and C3+ signals from
Fig. 9, accounting for particle wall loss and normalized to the ion response
under nascent conditions. The differently shaped and colored traces show the
relative enhancement of ion signals, measured for the same SOA coating
conditions. The greater increase in the K+ signal compared to the
increase in the C3+ ion signal for the same coatings is attributed
to two distinct ion formation mechanisms that the two species experienced.
As the IR laser has a Gaussian intensity profile, the different mechanisms
occur in different regions of the IR beam. Evidently, a higher intensity is
necessary for the vaporization of rBC, but a lower intensity can still drive
thermal vaporization/ionization of alkali metals, such as potassium and
sodium. The Gaussian IR beam thus has a greater effective width for alkali
metals than for rBC.
Correlation of refractory ion signals from the SP-AMS during SOA
coating experiments. Colored symbols represent the relative enhancement of
K+ and C3+, as in Fig. 9. All four traces, indicating different
coating experiments, begin at (1,1) and generally increase with increasing SOA coating. The increase in
measured K+ compared to C3+ ion signal for the same coatings
shows that there are two different effective laser widths for the processes:
thermal ionization of K+ and two-step vaporization-ionization of rBC.
Figure 11 displays the wall-loss-corrected K+ and Cx+ signal
measured by the SP-AMS as a function of wall-loss-corrected OM
signal. We corrected the SP-AMS signal for wall losses using the rBC mass
measurement from the SP2, and we normalized the signals from each experiment
to the signals for the nascent soot particles. We smoothed the traces using
a three-point boxcar moving average to clarify trends. The OM signal is not
quantitative due to unknown relative ionization and collection efficiencies
for organics detected with both the IR laser and heater operating
simultaneously. However, it provides a high-time-resolution relative metric
of the amount of condensed SOA. It shows that even at our highest achieved
coating thickness the signals from both K+ and Cx+ continued
to increase; neither signal saturated. This is different from what is
expected based on the results of Willis et al. (2014). When they coated
Regal Black with OM in the form of DOS, they observed that both OM ions and
EC ions reached a maximum enhancement after coating the Regal Black with a
thick coating of DOS (OM mass : EC mass > 3.) Although we cannot
rule out the effect of EIR, the continued increase merits
investigation of other causes for increased ion signals with additional SOA
coating.
Wall-loss-corrected SP-AMS signal from rBC Cx+
(panel a) and K+ (panel b) have been normalized to
their nascent values (before SOA coating) and are shown as a function of
wall-loss-corrected SP-AMS OM signal. Marker shapes and sizes indicate the
initial soot core mobility diameter for a given experiment. One-minute
averages show that the evolution of the particle sensitivity is continuous
and, even at our thickest SOA coating, we continue to see an increase in
instrument sensitivity to both rBC and K refractory material.
The continuous increase in K+ ion signal may potentially be explained by
changes in the thermal ionization efficiency, described by the Saha–Langmuir
equation (Zandberg and Ionov, 1971). This states that the probability of
thermal ionization of a species will increase if the surface it vaporizes
from has a higher work function or reaches a higher temperature. One
explanation is that the highly oxygenated SOA possesses a higher work
function than the rBC, and thus as the particle is coated it may generate
potassium ions more efficiently. However, it seems unlikely that
potassium would vaporize before all the
SOA, as is seen in similar, albeit ambient pressure, IR laser systems
(Moteki
and Kondo, 2007; Schwarz et al., 2010; Stephens et al., 2003). An alternative
explanation is that thicker SOA coatings may cause the particle to penetrate
deeper into Gaussian profile of the IR laser before the
potassium vaporizes and it subsequently
vaporizes from a hotter surface, thus generating more ions thermally.
With respect to the increasing Cx+ signal with increasing coating
thickness, a potential explanation includes the fragmentation of SOA to
contribute significantly to Cx+ mass. As stated in Sect. 3.4,
laser-off measurements of the SOA showed that C3+ accounted for
0.08 % of the SOA mass. With an OM mass increase of 10 times the rBC
present in a particle, and assumed relative ionization efficiencies for rBC
and OM of 0.2 and 1.4, respectively, the perceived increase in EC that would
be attributed incorrectly would be 1.1 %. This is much smaller than the
observed relative increase in C3+ for OM : rBC > 3,
∼ 63 % for the change in the last two data points for
dmob=220 in Fig. 9a. Furthermore, with the increased
fraction of particles being vaporized by the IR laser, it has been shown that
fragmentation would decrease, thereby decreasing the contribution of SOA to
C3+ (Canagaratna et al., 2015b).
Alternative explanations may include instrument differences and variability
in IR laser beam width. This illustrates that particle-beam width is an
important metric for quantitative measurements of BBA, even with a high
SOA : rBC mass ratio > 9 and particle vacuum aerodynamic
diameters greater than 400 nm.
LAAPTOF quantification of OM on externally mixed soot particles
Single-particle analysis provides valuable insight into the physical and
chemical evolution of biomass-burning plumes as they are diluted by
entrainment of background air and undergo transformation processes during
transport. Single-particle analysis also facilitates the determination of
contributions from biomass-burning particles to aerosol loading for source
apportionment and can reveal changes in the mixing state of biomass-burning
particles as they age
(Chen et
al., 2014; Moffet et al., 2008; Silva et al., 1999; Zauscher et al., 2013).
Although LDI, such as that used by the
LAAPTOF, is often regarded as a semi-quantitative method, there are numerous
examples demonstrating that it can be mass quantitative for constrained
systems where similar particle matrices are studied
(Healy
et al., 2013; Jeong et al., 2011; Mansoori et al., 1994; Spencer and
Prather, 2006; Sullivan et al., 2007, 2009). Quantitative mass measurements
using (LDI-SP-MS) are difficult because of the particle matrix effects that
influence how much energy is absorbed by a given particle and subsequently
the production and fate of the generated ions
(Murphy, 2007; Steele et al., 2005;
Sullivan and Prather, 2005). Calibration of the LDI-SP-MS ion-signal
response to particles of known composition is also required to yield
mass-quantitative component measurements. The SOA-coated soot experiments we
present here provide a unique opportunity to characterize the response of
the LAAPTOF to realistic biomass-burning cores coated with complex and
realistic biogenic SOA, where the masses of rBC and OM are well constrained.
We show the average OM signal per particle measured by the LAAPTOF in
Fig. 12. We observed a large degree of molecular fragmentation; most of the
OM signal appeared at m/z+28, which we attribute to CO+. All data
points represent the average of at least five particles with a minimum of
100 Hz of total EC ion signal at m/z+24, +36, +48, and +60,
representing C2-5+. This excludes any homogeneously nucleated SOA
particles and low-signal particles from the analysis. Of the
454 total particles detected, 160 particles met the above criteria. A linear regression fit
of all LAAPTOF OM ion signal as a function of SOA mass per particle results
in a R2=0.72 (OM signal (Hz) = 75(fg SOA) + 167). This may be
due to different light absorption and laser energy distribution properties of
different sized soot cores. Narrowing the analysis to the experiment with
187 nm soot cores results in a highly linear fit with a R2=0.998 (OM
signal (Hz) = 120(fg SOA) + 179). SOA from a single precursor can
still contain hundreds or thousands of organic compounds, with a broad range
of molecular weights, degrees of functionalization, and optical properties
(Zhang et al., 2015). Despite the complex nature of the SOA coating and
BBA core, we observed a strong linear relationship
between the amount of SOA condensed on the particle and the OM signal from
the LAAPTOF. It is important to note that although other LDI-SP-MS studies
have found a good correlation between OM signal and reference measurements,
this work shows that an even better correlation can be obtained if the
analysis can be informed by the chemical information provided by the mass
spectra. In this case we show that, with some improvements in excimer laser
homogeneity, it may be possible for an algorithm to isolate rBC-containing
particles and identify the necessary function to convert OM signal into SOA
mass per particle.
Averaged OM signal (m/z+28; CO+) from nascent and
SOA-coated soot particles measured by the LAAPTOF during SOA coating
experiments vs. SOA mass per particle. Error bars indicate the standard
error. There is a strong positive correlation despite the complex
biomass-burning aerosol matrix. Open marker colors/shapes indicate the
initial dmob of the size-selected soot core, and black
markers indicate nascent, uncoated soot particles.
Figure 13 shows the same OM ion signal measured by the LAAPTOF plotted in
Fig. 12, now normalized by the concurrent LAAPTOF EC ion signal. This
parallels the analysis by ATOFMS of spark-generated EC particles coated by
the condensation of diesel fuel presented by Spencer and Prather (2006), but
using more complex and realistic particle EC and OM components. EC fragments
used included C2-5+ but excluded C1+ because we have observed
that OM can also be charred to C1+ and thus produce an interference
to the quantification of EC at m/z+12. The OM : EC ion ratio used here
is thus CO+ : C2-5+. We use the ratio of OM / EC ions to
account for laser shot-to-shot variability, where more ions may be generated
due to increased laser fluence or increased absorption of laser energy. We
sampled biomass-burning particles that contained potassium salts, which
ionize readily, as well as strongly light-absorbing rBC
(Gross et al., 2000). This represents the
complex composition of realistic aged BBA. Initial particles were composed
mostly of black carbon, with initially variable shapes, and with an unknown
distribution of potassium salts per particle. Despite the variability in the
soot core composition of individual particles, a positive correlation
between the amount of SOA mass per particle and the LAAPTOF ion signal from
oxidized organics was still observed from these mixed rBC and inorganic salt
particles, when averaged over the many particles we sampled.
LAAPTOF-measured OM ion signal (m/z+28; CO+) normalized by
LAAPTOF EC ion signal (sum of m/z+24, +36, +48, +60; C2-5+)
for uncoated and SOA-coated soot particles. Error bars indicate the standard
error. Here the OM : EC ratio is shown as a function of the ratio of
mobility-derived SOA mass to SP2-measured rBC. The LAAPTOF EC signal serves
as an internal standard to normalize for the actual amount of laser energy
the particle absorbed, which can change with increased rBC mass or SOA
coating, and particle size and shape. Marker colors/shapes indicate the
initial size-selected soot core dmob for each
experiment.
Other LDI-SP-MS instruments that use longer ionization wavelengths, such as
266 nm used in the ATOFMS, do not typically fragment oxidized organics to
CO+; instead their major oxidized OM ion fragment is observed at m/z+43, presumably from C2H3O+ (Moffet and Prather, 2009; Spencer
and Prather, 2006). We operated the LAAPTOF's excimer laser at a moderate
pulse energy of 2.0 mJ, as a tradeoff between increased sensitivity while
avoiding excessive decomposition or fragmentation of OM into EC Cn+
fragments. In the work described in Spencer and Prather (2006), graphitic
soot particles with no condensed organics had an OM : EC ion signal ratio
of ∼ 0, and those coated with a mass of condensed fuel equal to the EC
mass had an OM : EC ion signal ratio of 2.25. For a similar OM : EC mass
ratio ∼ 1 in our experiments, we saw a more modest OM : EC ion
ratio of ∼ 0.35. This difference may be caused by differences in
instrument operation and chemical composition of the refractory EC core and
OM coating. The LAAPTOF uses an exciplex VUV laser at 193 nm (compared to
the “softer” 266 nm Nd:YAG laser in the ATOFMS) and we observe more
extensive molecular fragmentation of the OM. Thus, we quantified OM with a
single ion (CO+), compared to the various ions used by Spencer and
Prather (i.e., m/z+27, +29, +37, +43). A softer ionization might
have resulted in less charring of SOA to C1+ in the LAAPTOF and thus
increased the slope of the LAAPTOF-measured OM : EC ions versus mass
fraction of SOA. Spencer and Prather used condensed diesel fuel vapors as
their source of OM, while we used SOA from the ozonolysis of α-pinene. Therefore, the OM used in the prior study was unoxidized primary
organic aerosol, while our OM was highly oxidized complex secondary organic
aerosol.
Conclusions
We investigated the response of two particle mass spectrometers to
biomass-burning particles with carefully controlled amounts of organic
matter and well characterized particle properties. Our analysis revealed a
variety of particle morphologies and compositions, leading to a broad
distribution of vacuum aerodynamic diameters at a single selected mobility
diameter. As these particles of various compositions and shapes became
coated with SOA, measurements using the beam-width probe of the SP-AMS
revealed that the effective IR laser beam width for thermal ionization of
potassium is larger than for vaporization of rBC. Future measurements using
the SP-AMS would do well to quantify the particle-beam and effective
IR laser beam overlap for the species of interest using beam width probe
measurements. The fraction of rBC-containing particles that experience
sufficient laser fluence to vaporize the nonrefractory coating should
always be greater than or equal to the fraction of particles whose
refractory material is also vaporized. The use of optical components to
change the IR laser from a Gaussian to a uniform flat-top energy profile
would greatly simplify the analysis of rBC-containing soot particles by the
SP-AMS by ensuring that all particles passing through the laser experience
the same amount of energy. This would eliminate the difference given above
regarding the fraction of rBC-containing particles whose nonrefractory
versus refractory components are detected.
Although these measurements explored thick SOA coatings and large particle
sizes, we did not observe a plateau in instrument response to potassium or
black carbon as the coating was increased. These findings have important
implications for obtaining quantitative mass measurements and can help to
better inform the analysis and interpretation of SP-AMS measurements of the
emissions and aging of BBA
(Corbin
et al., 2015; Dallmann et al., 2014; Fortner et al., 2012; Lee et al., 2016;
Massoli et al., 2015). We would like to add, however, that although changes
in dva may result in varying EIR, SP-AMS
measurements of rBC have been shown to correlate with reference measurements
(Fortner
et al., 2012; Willis et al., 2014). This suggests that the changes in
secondary aerosol mass required to cause large changes in EIR
for rBC did not happen on the timescale of those measurements.
Mass spectral analysis with the SP-AMS also revealed that increased SOA
coatings on the biomass-burning soot changed the relative abundance of
EC clusters. Specifically, as the particles became more
thickly coated with SOA, the ratio of C>3+ to C3+
increased. The degree of change for the ratio of C4+ to
C3+ was smaller than the precision suggested by Corbin et al. (2014)
for identifying soot-particle source types using ratios of the
elemental carbon ion family, Cx+. However, the observed ratio of
C4+ to C3+ was inconsistent with that previously
observed from quenched-combustion flame soot
(Corbin et al., 2014; Maricq, 2014). This
reinforces the need for thermal denuding of ambient soot samples before
attempting source analysis using EC ratios.
Despite the challenges presented by complex particle composition and shape
for laser desorption ionization single-particle mass spectrometry, there is
a strong correlation between the average OM ion signal measured by the
LAAPTOF and the SOA mass per particle. Although this quantitative
relationship has been shown previously for LDI-SP-MS analysis of graphite
spark discharge soot coated by diesel fuel condensation
(Spencer and Prather, 2006), this is the first time it has been
explored with realistic combustion soot in the presence of inorganic
components and complex realistic SOA. This opens the way for more
quantitative single-particle measurements using techniques such as
laser-beam homogenization and particle type informed ion sensitivity
calibrations. Additional work is required to investigate the response of
single-particle mass spectrometry to other atmospherically relevant
core–shell combinations.