Introduction
Approximately 3 billion people live in residences where solid fuels (coal,
wood, charcoal, dung, and crop residues) are combusted for cooking
(Smith et al., 2014). Approximately 57 % of
Indian households report use of wood (49 %) or crop residues (9 %) as
their primary cook fuels, while 8 % report dung as a primary cook fuel
(Census of India, 2011). However, many households will
routinely use two or more of these fuels in combination for their cooking
needs in simple, home-made traditional stoves, or chulhas. These biomass burning
cookstoves have low combustion efficiencies and produce significant
emissions of pollutants, including fine particulate matter (PM2.5).
Epidemiological literature statistically links household air pollution from
solid biomass combustion to acute lower respiratory infections in children;
heart disease, stroke, cataracts, cancers in adults; and low birth weight
for infants of women exposed during pregnancy
(Smith et al., 2014). PM2.5 are small
enough to infiltrate deep into the lungs and penetrate the body's defenses,
and therefore PM2.5 exposure has been commonly used for estimating
risks from both ambient air pollution and cigarette smoke
(Finlayson-Pitts and Pitts, 2000). The degree of adverse health
effects of cookstove smoke likely depends on the chemical composition of the
PM2.5. However, the exact relationship between the chemical composition
and its health effects is largely unknown
(Araujo et al., 2008).
Household cooking is estimated to be responsible for 26–50 % of ambient
PM2.5 in India (Chafe et al.,
2014; Guttikunda et al., 2016; Lelieveld et al., 2015). In this emissions
mixture, carbonaceous particles affect the climate directly by scattering and
absorbing incoming solar radiation and indirectly by acting as cloud
condensation nuclei (Crutzen and Andreae, 1990). In addition
to black carbon (BC), which absorbs solar radiation across the entire
visible spectrum, some molecules in biomass burning organic aerosols (BBOA)
efficiently absorb blue and near-UV solar radiation resulting in
classification of BBOA as brown carbon (BrC) (Laskin et al.,
2015). Modeling studies have shown that in certain geographic areas climate
warming by BrC has the potential to outweigh cooling by scattering organic
aerosols (Feng et al., 2013). South Asia has been
identified as one of these unique regions where emissions from cookstoves
are a significant source of regional BrC (Feng et al.,
2013).
Cookstove emissions have been studied in both laboratory and field
settings. Field studies typically involve observations and measurements
during daily cooking activities in rural village homes. For example,
Xiao et al. (2015)
measured BC and PM2.5 throughout the day for six different houses to
monitor indoor concentrations in the household. In the laboratory, water
boiling test (WBT) protocols are utilized to evaluate stove performance
(Global Alliance for Clean Cookstoves, 2014). The WBT standard
protocols are made up of three phases to represent the stove's combustion
efficiency while cooking: (1) high power, cold start; (2) high power, hot
start; and (3) low power, simmer (Global Alliance for Clean
Cookstoves, 2014). While the WBTs can be carried out under more controlled
conditions, recent studies have found that the WBTs fail to capture periods
of low combustion efficiency in cooking events
(Chen et al., 2012;
Johnson et al., 2008, 2009). This is due to daily cooking activities
involving more than just boiling water
(Johnson et al., 2009). Some cooking
techniques require a smoldering fire, for example the cooking of chapatti, a
traditional Indian flat bread (Johnson
et al., 2009). Alternately, these low combustion efficiency periods may be a
consequence of multitasking around the home
(Johnson et al., 2009). The literature
estimates that emissions of PM2.5 (Roden et al.,
2009) and CO / CO2 ratios
(Johnson et al.,
2008; Kituyi et al., 2001; Ludwig et al., 2003) are underrepresented by the
WBTs relative to field measurements by a factor of 3. There are also
concerns that WBTs cannot be scaled to real cooking events and that climate
models may underrepresent global emissions from biomass burning cookstoves
(Chen et al., 2012;
Johnson et al., 2008, 2009).
A number of studies have characterized the optical properties of cookstove BBOA.
Depending on the measurement approach, different metrics of aerosol
absorption have been reported. In general, methods that take direct
measurements of aerosol particles without extraction report mass-normalized
absorption cross sections of aerosols (MACaerosol). Absorption
measurements with the extracted material report mass-normalized absorption
cross sections of bulk material (MACbulk). In this paper, we use
a subscript “bulk” to help minimize confusion between MACbulk and
MACaerosol. The two can be related if the particle size distribution is
known (Laskin et al., 2015). An advantage of MACbulk is that it can be
used to calculate the imaginary refractive index of the organic material
(Laskin et al., 2015). For particles that are made of material with a real
refractive index of 1.5 and that are small in diameter relative to the
wavelength, MACaerosol∼ 0.7 × MACbulk (Laskin et
al., 2015).
Stockwell et al. (2016) utilized photoacoustic extinctiometers (PAX) to
conduct in situ absorption measurements at 405 and 870 nm, resulting in
particle absorption coefficients from cook fire emissions in Nepal. With a
literature-recommended mass absorption coefficient for light-absorbing
organic compounds
(Lack and Langridge, 2013) and the measured aerosol
absorption by PAX, Stockwell et al., (2016) approximated particle absorption emission factors
(EFs) due to the light-absorbing organic compounds in particles. Organic
carbon (OC)
absorption EFs were 1.5 times higher for the hardwood smoke (EF = 8.40 g kg-1 fuel)
compared to the dung smoke (EF = 5.43 g kg-1 fuel).
Pandey et al. (2016) collected PM2.5 on filters from cook fires in
India, that were fueled by wood, agricultural residues, dung, and a mixture thereof and
reported MACaerosol values. They found that the MACaerosol at
550 nm was a factor of 2.6 higher for wood fuel (1.3 m2 g-1) compared
to dung fuel (0.5 m2 g-1)
(Pandey et al.,
2016). This is consistent with Saleh et al. (2014), who found
that the effective absorptivity of OA in BBOA increases with BC-OA ratio. In
Pandey et al. (2016), they measured EC / OC at 0.0649 for dung, and 0.0826 for wood fuel.
Many organic components of BBOA have been successfully characterized in
previous studies by electrospray ionization high-resolution mass
spectrometry (ESI-HRMS)
(Budisulistiorini
et al., 2017; Laskin et al., 2009; Lin et al., 2012, 2016, 2017; Smith et
al., 2009; Wang et al., 2017; Willoughby et al., 2016). For example,
ESI-HRMS was used to analyze the particle-phase organic constituents of
smoke samples collected during the Fire Lab at Missoula Experiment (FLAME)
campaign (Laskin et
al., 2009; Smith et al., 2009). Fuels utilized in the FLAME studies were
selected to represent North American wildfires and the publications focused
on nonwoody biomass fuels such as detritus and litter as well as ceanothus from the
US Pacific Northwest. Smith et al. (2009)
reported an inventory of species in particle-phase BBOA, with 70 % of the
compounds being reported for the first time.
Laskin et al. (2009) examined the
nitrogen-containing species and observed that a large fraction of the
detected species were N-heterocyclic compounds. Lin et al. (2016) identifies
fuel-specific BrC chromophores in particles collected from FLAME-4 via high-performance liquid chromatography combined with photodiode array
and high-resolution mass spectrometry (HPLC-PDA-HRMS). Two of the four fuels were
woody biomass specific to North America. They found that nitroaromatics,
polycyclic aromatic hydrocarbons (PAHs), and polyphenols were responsible for the light absorption by BBOA
(Lin et al., 2016). Recent papers
investigated the chromophores in BBOA from Lag Ba'Omer, a nationwide bonfire
festival in Israel (Bluvshtein et al., 2017; Lin et
al., 2017). They found nitroaromatics to be the most prominent chromophores
in these samples. Budisulistiorini et al. (2017) similarly identified 41
chromophores from Indonesian peat, charcoal, and fern or leaf burning with a
method relying on chromatographic separation and simultaneous detection by
spectrophotometry and ESI-MS. They identified three types of chromophores:
oxygenated, nitroaromatics, and sulfur-containing
(Budisulistiorini et al., 2017).
The goal of the current study is to understand the composition of cookstove
BBOA in additional detail than afforded by previous measurements. We do this
by (1) generating and collecting BBOA from prescribed cooking events carried
out by a local cook and (2) using high-resolution mass spectrometry
techniques to characterize their particle-phase composition. This is part of
a larger study attempting to document the contribution of household solid fuel combustion to air pollution in India.
In this paper we provide an inventory of particle-phase compounds produced
from actual cooking events detected by nanospray desorption electrospray ionization–high-resolution mass spectrometry (nano-DESI-HRMS) and an assessment of
BrC chromophores specific to the biomass type used based on HPLC-PDA-HRMS
analysis. In addition, we compare particle-phase constituents in cook fire
smoke produced from different traditional stoves and fuels.
Experimental methods
Field site
This study was conducted at the SOMAARTH Demographic, Development, and
Environmental Surveillance Site
(Balakrishnan
et al., 2015; Mukhopadhyay et al., 2012; Pillarisetti et al., 2014) run by
the International Clinical Epidemiological Network (INCLEN) in Palwal
District, located approximately 80 km south of New Delhi. SOMAARTH covers 51
villages across three administrative blocks, with an approximate population
of 200 000. Palwal District has a population of approximately 1 million over
∼ 1400 km2; 39 % of households in the district utilize
wood as their primary cook fuel, followed by dung (25 %) and crop residues
(7 %) (Census of India, 2011).
Sample collection
Over 34 days in August–September 2015, PM2.5 samples were collected
from a kitchen in the village of Khatela, Palwal, Haryana, India. Figure 1
shows (a) the kitchen setup, and (b) the stoves (angithi and chulha) and fuels (dung and
brushwood) used. The stoves and fuels were obtained locally, and traditional
meals were prepared by a local cook. The cook was instructed by the
experimenters to prepare a particular standard meal using the selected fuel
and stove. All angithi cookstoves burned dung and were used to prepare buffalo
fodder. Chulha cookstoves burned either brushwood or dung fuels and were used to
prepare a traditional meal of chapati and vegetables for four people. Vegetables were
cooked in a pressure cooker that rests on top of the chulha (Fig. 1b).
Chapatti were cooked in the air space next to the fuel, as is typical for this area.
Brushwood–angithi cook fires were never tested because this combination is not
frequently used in the local households.
The field site and setup for cooking events. (a) The kitchen
setup at the field site. (b) The stoves and fuels used in this study:
angithi, dung-burning chulha, and brushwood-burning chulha.
Figure S2.1 in the Supplement shows a diagram of sample collection. PM2.5 emissions were
sampled via three-pronged probes that hung above the cookstove. Air sampling
pumps (PCXR-8, SKC Inc.) created a flow of BBOA emissions through aluminum
tubing during cooking events. BBOA was captured through cyclone
fractionators (2.5 µm cut point, URG Corporation) and the resultant
flow was taken through a stainless steel filter holder containing a PTFE
filter (Teflon B, SKC Inc., 47 mm). One filter was collected for chemical
analysis (Teflon B), and another filter for gravimetric analysis (Teflon A).
Flows were measured via a mass flowmeter (TSI 4140) before and after each
cooking event to ensure it had not varied more than 10 %. The pumps were
turned on before cooking began so that emissions from the entire cooking
event were captured and turned off when the fire was out. Stove dimensions
and their distance from the probe inlets are detailed in Fig. S3.1. Prior
to analysis, filters were stored in petri dish slides at -80 ∘C
other than during transportation and use. This includes time at the field
site (1–6 h) and transportation back to the United States (24 h).
During these times, samples were stored at ambient temperature.
Nano-DESI-HRMS analysis
PM2.5 collected on PTFE filters was analyzed with an
LTQ-Orbitrap™ high-resolution mass spectrometer (ThermoFisher
Scientific) equipped with a custom built nano-DESI source
(Roach et al., 2010a, b).
Nano-DESI consists of two electrified capillaries with a small (< 1 mm)
droplet, or solvent bridge, forming at the point of their contact. The
nano-DESI solvent mixture (70 % CH3CN/30 % H2O, optimized for
the stability of the nano-DESI source) flows through an electrified
capillary at a flow rate of 0.3–1 µL min-1. The droplet is lowered to
the substrate's surface, where the analyzed material is extracted by the
solvent and immediately sprayed in the mass spectrometer inlet. It has been
shown that the nano-DESI dissolves all extractable material on the filter
surface (Roach et al., 2010a, b). To ensure the material on the filter is not depleted, the droplet is
moved across the filter's surface at roughly 0.2 cm min-1. The spray voltage
was 3.5 kV and the instrument was operated in positive ion mode. The mass
accuracy of the HRMS was calibrated over a wide m/z range with a ThermoFisher
Scientific standard calibration mixture. Two separate mass spectra were
obtained from different portions of the filter to ensure reproducibility.
Only peaks that showed up in both spectra were retained for further
analysis.
Peaks with signal-to-noise ratios of greater than 3 were extracted from the
time-integrated nano-DESI chromatograms using Decon2LS software. Peaks
containing 13C isotopes were excluded from analysis. Sample and
solvent-blank mass spectra peaks were clustered with a tolerance of 0.001 m/z using
a second-order Kendrick analysis with CH2 and H2 base units
(Roach et al., 2011). The spectra were
internally mass-calibrated by assigning prominent peaks of common BBOA
compounds first, and fitting the observed-exact m/z deviation to a linear
regression curve. The m/z correction introduced by the internal calibration was
< 0.001 m/z units, but even at these small levels the correction helped
reduce the ambiguity in the assignments of unknown peaks. We focused on
analyzing peaks with m/z < 350, as peaks above this m/z value were small in
abundance (on average 6 % of total abundance), number of peaks (9 % of
the total number of peaks), and in many cases could not be assigned
unambiguously. Exact masses were assigned using the freeware program Formula
Calculator v1.1 (http://magnet.fsu.edu/~midas/download.html).
The permitted elements and their maximal numbers of atoms were as follows: C (40), H (80), O (35), N (5),
and Na (1). Peaks that could not be assigned
within the described parameters had small abundances and were not pursued
further. There were a few notable exceptions, namely the potassium salt
peaks discussed below. Conversely, potassium-organic adducts were not
observed, presumably due to the higher affinity of organic molecules to
Na+ compared to K+. Permitting sulfur, chloride, and phosphorus
did not increase the fractions of assignable peaks, nor did it change the
assignments for the peaks we report. The double-bond equivalent (DBE) values
of the neutral formulas were calculated using the following equation: DBE = C - H / 2 + N / 2 + 1,
where C, H, and N correspond to the number of carbon,
hydrogen, and nitrogen atoms, respectively.
HPLC-PDA-HRMS
The samples were further analyzed with an HPLC-PDA-HRMS platform
(Lin et al., 2016). To prepare the
samples for analysis half of the PTFE filter was extracted overnight in a
mixture of acetonitrile, dichloromethane, and hexane solvents (2 : 2 : 1 by
volume, 5 mL total), which was empirically found to work well for extracting
a broad range of BBOA compounds (Lin et al., 2017). The
solutions were then filtered with polyvinyl difluoride (PVDF) filter syringes to remove insoluble
particles (Millipore, Duropore, 13 mm, 0.22 µm). The solutions were
concentrated under N2 flow, and then diluted with water and dimethyl
sulfoxide (DMSO) to a final volume around 150 µL. The separation was
performed on a reverse-phase column (Luna C18, 2 × 150 mm, 5 µm
particles, 100 Å pores, Phenomenex, Inc.). The mobile phase comprised of
0.05 % formic acid in LC–MS grade acetonitrile (B) and 0.05 % formic
acid in LC–MS grade water (A). Gradient
elution was performed by the A–B
mixture at a flow rate of 200 µL min-1: 0–3 min hold at 90 % A,
3–62 min linear gradient to 10 % A, 63–75 min hold at 10 % A, 76–89 min
linear gradient to 0 % A, 90–100 min hold at 0 % A, then 101–120 min
hold at 90 % A. The electrospray ionization (ESI) settings were as follows: 5 µL injection
volume, 4.0 kV spray potential, 35 units of sheath gas flow, 10 units of
auxiliary gas flow, and 8 units of sweep gas flow. The solutions were
analyzed in both positive and negative ion ESI-HRMS modes.
The HPLC-PDA-HRMS data were acquired and first analyzed using Xcalibur 2.4
software (Thermo Scientific). Possible exact masses were identified by LC
retention time using the open source software toolbox MZmine version 2.23
(http://mzmine.github.io/) (Pluskal et al.,
2010). Formula assignments were obtained from their exact m/z values using the
Formula Calculator v1.1.
Representative nano-DESI mass spectra collected for (a) dung–angithi (b) dung–chulha
and (c) brushwood–chulha cook fires. Relative abundance is plotted
against m/z. Peaks are colored by their elemental makeup, CxHyNw
(red), CxHyOzNw (purple), CxHyOz (blue),
potassium salts (green), and unassigned (black). The pie charts illustrate
the fraction of count-based, normalized peak abundance that is attributed to
each elemental category.
MACbulk and AAE
Selected filter halves were extracted as described in Sect. 2.4.
Absorption spectra of the extracts were collected with a dual-beam UV-Vis
spectrophotometer (Shimadzu UV-2450). MACbulk values were calculated
from the following equation:
MACbulk(λ)=A10(λ)⋅ln(10)b⋅Cmass,
where A10 is the base-10 absorbance, b is the path length of the
cuvette (m), and Cmass is the mass concentration of the extracted
organic material in (g m-3). The largest uncertainty in
MACbulk came from uncertainty in Cmass of the
extract. First, the overall mass of PM2.5 on the filter had to be
estimated from another filter collected specifically for gravimetric
analysis. The PM2.5 mass on the chemical analysis filter was calculated
from the mass on the gravimetric analysis filter after accounting for
different flows through the two filters (See Fig. S2.1). This calculation
assumed the same PM2.5 collection efficiency for both filters. The
particle mass distribution on the filter was assumed to be uniform, and the
extraction efficiency of PM2.5 mass was estimated to be 50 % by
comparing the weights of filters before and after the extraction.
Uncertainties incorporate flow rate measurements (10 % relative error)
and extraction efficiency of PM2.5 mass (40 % relative error).
Absorption angstrom exponents (AAE) were calculated for both samples by
fitting the log(MACbulk) vs. log(λ) to a linear function
over the wavelength range of 300 to 700 nm. It should be noted that there
are many methods for measuring optical properties of PM2.5 particles,
and the method used here provides MACbulk and AAE of extractable
organic bulk material. The advantages and limitations of other methods are
explained in Laskin et al. (2015).
Results and discussion
Nano-DESI-HRMS analysis of cookstove particles
Representative nano-DESI mass spectra from the three major types of
cook fires sampled are shown in Fig. 2. It is clear from the mass spectra
in Fig. 2 that the three combinations of fuel–stove types lead to distinct
particle compositions. We compare the particle composition of the three
major cook fire types by averaging the percentage of
CxHyNw, CxHyOz, and
CxHyOzNw peaks in the nano-DESI spectra from multiple
samples. Samples used and a summary of the following discussion is detailed
in Table S1.1. The overwhelming majority of detected species by nano-DESI in
dung cook fire smoke PM2.5 was attributed to CxHyNw,
compounds that contain only carbon, hydrogen, and nitrogen atoms. The
average count-based fractions from CxHyNw species were
79.9 % ± 4.4 and 82.1 % ± 1.0 % for dung–chulha and
dung–angithi experiments, respectively, but only 23.8 % ± 7.8 % for
brushwood–chulha experiments. All nitrogen-containing compounds in the smoke
PM2.5 come from the nitrogen content in the fuels
(Coggon et al., 2016), which is likely higher for
dung.
Stockwell
et al. (2016) reported the nitrogen content of yak dung as 1.9 % by weight,
while it is found to be lower for woods such as black spruce (0.66 % by
weight and ponderosa pine (1.09 % by weight)
(Hatch et al., 2015). It should
be noted that another study of fuels in India found the nitrogen content was
roughly the same for brushwood (1.4 ± 0.3 % by weight) and dung
(1.4 ± 0.1 % by weight) (Gautam et al.,
2016), so additional characterization of fuel composition in the future is
desirable. In contrast to dung fuel, PM2.5 from brushwood cook fire
smoke contained higher fractions of CxHyOz species.
Specifically, the count-based fraction assigned as CxHyOz
species was 43.1 % ± 14.6 % in brushwood–chulha cook fires compared to only
4.1 % ± 0.9 and 3.2 % ± 3.3 % for dung–chulha and
dung–angithi experiments, respectively. Many of the CxHyOz formulas
were consistent with species reported previously as lignin-pyrolysis
products (Collard and Blin, 2014;
Simoneit et al., 1993). Fractions of CxHyOzNw did not
correlate well with fuel–stove variables and ranged from 4.1 to 34.4 %
in the analyzed samples.
Potassium
(Hosseini
et al., 2013; Sullivan et al., 2014) and levoglucosan
(Jayarathne
et al., 2017; Simoneit et al., 1999) are well-established flaming and
smoldering BB tracers, respectively. Gas-phase chlorine species have been
observed in BBOA previously
(Lobert
et al., 1999; Stockwell et al., 2016). Therefore it is not surprising that
inorganic salt peaks containing potassium and chlorine were observed in more
than half of dung cook fires (8 out of 14) and all brushwood cook fires. These
peaks were pursued apart from the original analysis because the peak
abundance was very large in many mass spectra, and they served as convenient
internal m/z calibration points. These mass spectra all contained
K2Cl+ as the most prominent salt peak and K3Cl2+
was also present in a few mass spectra. Isotopic variants of these salts,
namely with either 37Cl or 41K (24 or 6.7 % natural
abundance) instead of 35Cl or 39K (76 or 93.3 % natural
abundance), were also found. The resolving power of the HRMS instrument is
insufficient to distinguish the isotopic shifts from Cl and K (Δ
mass37Cl-35Cl= 1.997 Da, Δ mass41K-39K= 1.998 Da), but
one or both of the isotopes were consistently present in all mass spectra
containing potassium ions. Adducts corresponding to a replacement of K by Na
were also detected. The observed potassium signals may have depended not
only on the potassium content of the fuel but also on the amount of flaming
combustion (combustion efficiency), the specific food items cooked, or the
stove material itself. Inorganic salts were observed in all chulha cook fire
PM2.5 samples regardless of fuel type and were absent in all angithi cook fire
PM2.5 samples. On average, chulha stoves have a higher combustion
efficiency (dung–chulha 90.7 % ± 0.6 %, dung–angithi
87.5 % ± 1.8 %)
consistent with more flaming combustion and therefore more potassium
emissions. The chulha and angithi stoves produced meals for people and animal fodder,
respectively. Also, the chulha was made mainly from brick with a covering of
local clay, whereas the angithi was only made from clay. With the presently available data
it is impossible to determine whether the potassium salts originated from
flaming combustion, originated from the chulha material, or are the result of different food
items cooked.
Levoglucosan was present in 3 out of 8 dung–chulha cook fires, 4 out of 6
dung–angithi cook fires, and 4 out of 11 brushwood–chulha cook fires. We expect
levoglucosan to be found in BBOA from all fires, based on other studies
(Jayarathne
et al., 2017), and we therefore conclude that levoglucosan peak must have
been suppressed in the nano-DESI source by the more ionizable components of
the mixture. By extension of the same logic, ions corresponding to other
carbohydrates, and more broadly to lignin-derived CxHyOz species, were likely suppressed by this technique, and therefore a
significant fraction of BBOA constituents may be absent in this inventory.
Due to the variability in observing levoglucosan we conclude that for ESI-MS
studies levoglucosan serves as a marker rather than a tracer for digested
biomass burning and woody biomass burning.
Particle-phase biomass burning markers
An inventory was compiled of compounds that were reproducibly observed in samples from
three different cooking events using the same fuel–stove combination. Samples were chosen for the inventory by considering the measured
fuel moisture content and meal cooked, with the goal of comparing samples
from similar cook fires (see Table S1.1 for sample details). Peaks that
did not appear in mass spectra of all three samples were discarded to ensure
reproducibility and help filter out noise peaks from the nano-DESI source.
The remaining peak abundances were first normalized to the largest peak
abundance, then the three mass spectra were averaged. Since the absolute peak
abundances varied in individual spectra, only approximate relative
abundances are reported here grouped into three logarithmic bins, denoted as
LOW (< 1 %), MEDIUM (1–9.9 %), and HIGH (10–100 %). This analysis
was completed for the emissions from each of the three types of
fuel–stove combinations studied in this work.
List of common compounds found in all PM2.5 samples regardless
of fuel or stove type. Tentative molecular structure assignments are listed
when the compound has previously been identified in the chemical biomass
burning literature, supported by the references in the last column.
Count-based, normalized peak abundances are designated LOW
(< 1 %), MEDIUM (1–9.9 %), and HIGH (10–100 %). All
species were detected as protonated ions.
Chemical
Relative
Observed
Calculated
formula of
DBE
average
Tentative
References
m/z
m/z
neutral species
abundance
assignment(s)
species
111.091
111.092
C6H10N2
3
MEDIUM
Smith et al. (2009)
121.064
121.065
C8H8O
5
MEDIUM
123.091
123.092
C7H10N2
4
MEDIUM
124.075
124.076
C7H9ON
4
MEDIUM
125.107
125.107
C7H12N2
3
MEDIUM
Smith et al. (2009)
133.075
133.076
C8H8N2
6
MEDIUM
Laskin et al. (2009)
134.071
134.071
C7H7N3
6
MEDIUM
Laskin et al. (2009)
137.059
137.060
C8H8O2
5
HIGH
Anisaldehyde
Simoneit et al. (1993);
Smith et al. (2009)
137.106
137.107
C8H12N2
4
MEDIUM
Smith et al. (2009)
138.090
138.091
C8H11ON
4
LOW
139.122
139.123
C8H14N2
3
MEDIUM
Smith et al. (2009)
147.091
147.092
C9H10N2
6
MEDIUM
151.074
151.075
C9H10O2
5
MEDIUM
Vinylguaiacol
151.122
151.123
C9H14N2
4
MEDIUM
153.138
153.139
C9H16N2
3
HIGH
159.091
159.092
C10H10N2
7
MEDIUM
Laskin et al. (2009)
160.075
160.076
C10H9ON
7
MEDIUM
Laskin et al. (2009)
161.059
161.060
C10H8O2
7
MEDIUM
161.106
161.107
C10H12N2
6
MEDIUM
162.102
162.103
C9H11N3
6
LOW
163.074
163.075
C10H10O2
6
MEDIUM
165.138
165.139
C10H16N2
4
MEDIUM
167.069
167.070
C9H10O3
5
HIGH
Veratraldehyde
Simoneit et al. (1993)
167.153
167.154
C10H18N2
3
MEDIUM
173.106
173.107
C11H12N2
7
MEDIUM
174.090
174.091
C11H11ON
7
MEDIUM
175.074
175.075
C11H10O2
7
MEDIUM
175.122
175.123
C11H14N2
6
MEDIUM
177.053
177.055
C10H8O3
7
MEDIUM
177.090
177.091
C11H12O2
6
MEDIUM
177.101
177.102
C10H12ON2
6
LOW
177.137
177.139
C11H16N2
5
LOW
Laskin et al. (2009)
179.069
179.070
C10H10O3
6
MEDIUM
Coniferaldehyde
179.153
179.154
C11H18N2
4
MEDIUM
181.169
181.170
C11H20N2
3
HIGH
183.090
183.092
C12H10N2
9
HIGH
183.184
183.186
C11H22N2
2
MEDIUM
186.090
186.091
C12H11ON
8
MEDIUM
Laskin et al. (2009)
187.122
187.123
C12H14N2
7
MEDIUM
188.106
188.107
C12H13ON
7
MEDIUM
189.101
189.102
C11H12ON2
7
MEDIUM
Laskin et al. (2009)
189.137
189.139
C12H16N2
6
MEDIUM
191.069
191.070
C11H10O3
7
MEDIUM
191.117
191.118
C11H14ON2
6
LOW
191.153
191.154
C12H18N2
5
LOW
193.085
193.086
C11H12O3
6
MEDIUM
193.169
193.170
C12H20N2
4
MEDIUM
197.106
197.107
C13H12N2
9
MEDIUM
Continued.
Chemical
Relative
Observed
Calculated
formula of
DBE
average
Tentative
References
m/z
m/z
neutral species
abundance
assignment(s)
species
199.122
199.123
C13H14N2
8
LOW
200.106
200.107
C13H13ON
8
MEDIUM
201.137
201.139
C13H16N2
7
MEDIUM
202.085
202.086
C12H11O2N
8
MEDIUM
Laskin et al. (2009)
203.117
203.118
C12H14ON2
7
MEDIUM
203.153
203.154
C13H18N2
6
MEDIUM
205.085
205.086
C12H12O3
7
MEDIUM
207.184
207.186
C13H22N2
4
MEDIUM
209.079
209.081
C11H12O4
6
MEDIUM
209.200
209.201
C13H24N2
3
MEDIUM
211.095
211.096
C11H14O4
5
HIGH
Syringylethanone/
Simoneit et al. (1993)
trimethoxyphenylethanone
211.121
211.123
C14H14N2
9
MEDIUM
213.137
213.139
C14H16N2
8
MEDIUM
Laskin et al. (2009)
214.121
214.123
C14H15ON
8
MEDIUM
215.153
215.154
C14H18N2
7
MEDIUM
216.100
216.102
C13H13O2N
8
MEDIUM
217.132
217.134
C13H16ON2
7
MEDIUM
217.168
217.170
C14H20N2
6
MEDIUM
219.100
219.102
C13H14O3
7
MEDIUM
227.153
227.154
C15H18N2
8
MEDIUM
229.132
229.134
C14H16ON2
8
MEDIUM
229.168
229.170
C15H20N2
7
MEDIUM
230.116
230.118
C14H15O2N
8
MEDIUM
231.147
231.149
C14H18ON2
7
LOW
232.095
232.097
C13H13O3N
8
MEDIUM
235.095
235.096
C13H14O4
7
MEDIUM
241.168
241.170
C16H20N2
8
MEDIUM
243.147
243.149
C15H18ON2
8
MEDIUM
243.184
243.186
C16H22N2
7
LOW
244.131
244.133
C15H17O2N
8
MEDIUM
246.111
246.112
C14H15O3N
8
MEDIUM
249.110
249.112
C14H16O4
7
MEDIUM
Figure 3 summarizes how reproducibly detected PM2.5 compounds are
organized in the inventory. First, we provide a list of compounds common to
the emissions from all three
types of cook fires including: dung–chulha,
dung–angithi, and brushwood–chulha (Sect. 3.3, Table 1). Then, we discuss compounds
exclusively found in the brushwood–chulha cook fire emissions (Sect. 3.4, Table S4.1).
Within Sect. 3.5 we discuss compounds unique to the dung–chulha (Table S4.2) and
the dung–angithi (Table S4.3) cook fire experiments, as well as the
compounds they had in common (Table 2).
An overview of the particle-phase compounds inventory based on the
results of molecular characterization using nano-DESI-HRMS. Each area of the
Venn diagram contains the bolded number of reproducibly detected formulas in
blue, as well as the Table that lists peaks for each category. Merging all of
the tables listed here provides a complete inventory of compounds detected
in this study.
List of compounds found exclusively in the emissions from dung
cook fires, regardless of stove type. The labels for the peak abundances are
the same as in Table 1. All species were detected as protonated ions unless
otherwise noted.
Observed
Calculated
Chemical
Relative
m/z
m/z
formula of
DBE
average
neutral
abundance
species
124.099
124.099
C7H12N2∗
3
MEDIUM
135.080
135.080
C9H10O
5
LOW
135.092
135.092
C8H10N2
5
MEDIUM
136.076
136.076
C8H9ON
5
LOW
137.071
137.071
C7H8ON2
5
MEDIUM
138.115
138.115
C8H14N2∗
3
LOW
141.138
141.139
C8H16N2
2
LOW
145.076
145.076
C9H8N2
7
MEDIUM
146.060
146.060
C9H7ON
7
MEDIUM
146.084
146.084
C9H10N2∗
6
LOW
149.071
149.071
C8H8ON2
6
LOW
149.107
149.107
C9H12N2
5
LOW
151.086
151.087
C8H10ON2
5
LOW
152.107
152.107
C9H13ON
4
LOW
152.130
152.131
C9H16N2∗
3
LOW
155.154
155.154
C9H18N2
2
LOW
160.099
160.099
C10H12N2∗
6
LOW
162.091
162.091
C10H11ON
6
LOW
163.086
163.087
C9H10ON2
6
MEDIUM
163.123
163.123
C10H14N2
5
MEDIUM
164.107
164.107
C10H13ON
5
LOW
169.076
169.076
C11H8N2
9
HIGH
169.170
169.170
C10H20N2
2
MEDIUM
171.091
171.092
C11H10N2
8
MEDIUM
172.075
172.076
C11H9ON
8
MEDIUM
175.086
175.087
C10H10ON2
7
MEDIUM
176.070
176.071
C10H9O2N
7
LOW
176.107
176.107
C11H13ON
6
LOW
176.118
176.118
C10H13N3
6
LOW
178.086
178.086
C10H11O2N
6
LOW
184.075
184.076
C12H9ON
9
MEDIUM
185.107
185.107
C12H12N2
8
MEDIUM
187.086
187.087
C11H10ON2
8
MEDIUM
188.118
188.118
C11H13N3
7
LOW
189.091
189.091
C12H12O2
7
LOW
190.086
190.086
C11H11O2N
7
MEDIUM
190.133
190.134
C11H15N3
6
MEDIUM
191.081
191.082
C10H10O2N2
7
LOW
192.102
192.102
C11H13O2N
6
LOW
193.133
193.134
C11H16ON2
5
LOW
195.091
195.092
C13H10N2
10
MEDIUM
195.185
195.186
C12H22N2
3
HIGH
197.201
197.201
C12H24N2
2
MEDIUM
198.091
198.091
C13H11ON
9
MEDIUM
198.102
198.103
C12H11N3
9
LOW
199.086
199.087
C12H10ON2
9
MEDIUM
200.118
200.118
C12H13N3
8
LOW
Continued.
Observed
Calculated
Chemical
Relative
m/z
m/z
formula of
DBE
average
neutral
abundance
species
201.102
201.102
C12H12ON2
8
MEDIUM
202.122
202.123
C13H15ON
7
LOW
202.133
202.134
C12H15N3
7
LOW
204.101
204.102
C12H13O2N
7
LOW
204.149
204.150
C12H17N3
6
MEDIUM
205.097
205.097
C11H12O2N2
7
LOW
205.133
205.134
C12H16ON2
6
MEDIUM
205.169
205.170
C13H20N2
5
MEDIUM
206.117
206.118
C12H15O2N
6
LOW
207.112
207.113
C11H14O2N2
6
LOW
207.149
207.149
C12H18ON2
5
LOW
209.107
209.107
C14H12N2
10
MEDIUM
209.128
209.128
C11H16O2N2
5
LOW
211.144
211.144
C11H18O2N2
4
MEDIUM
212.106
212.107
C14H13ON
9
MEDIUM
212.118
212.118
C13H13N3
9
LOW
214.086
214.086
C13H11O2N
9
MEDIUM
215.117
215.118
C13H14ON2
8
MEDIUM
216.149
216.150
C13H17N3
7
LOW
217.085
217.086
C13H12O3
8
LOW
217.097
217.097
C12H12O2N2
8
MEDIUM
218.103
218.104
C10H11ON5
8
LOW
218.117
218.118
C13H15O2N
7
LOW
218.165
218.165
C13H19N3
6
LOW
219.112
219.113
C12H14O2N2
7
MEDIUM
219.149
219.149
C13H18ON2
6
LOW
219.185
219.186
C14H22N2
5
MEDIUM
221.080
221.081
C12H12O4
7
LOW
221.128
221.128
C12H16O2N2
6
MEDIUM
221.201
221.201
C14H24N2
4
MEDIUM
223.122
223.123
C15H14N2
10
MEDIUM
223.216
223.217
C14H26N2
3
MEDIUM
224.107
224.107
C15H13ON
10
LOW
225.102
225.102
C14H12ON2
10
MEDIUM
225.138
225.139
C15H16N2
9
MEDIUM
226.122
226.123
C15H15ON
9
MEDIUM
227.117
227.118
C14H14ON2
9
MEDIUM
228.101
228.102
C14H13O2N
9
MEDIUM
228.138
228.138
C15H17ON
8
MEDIUM
230.164
230.165
C14H19N3
7
LOW
231.112
231.113
C13H14O2N2
8
MEDIUM
232.133
232.133
C14H17O2N
7
LOW
233.128
233.128
C13H16O2N2
7
LOW
233.164
233.165
C14H20ON2
6
LOW
233.201
233.201
C15H24N2
5
MEDIUM
235.216
235.217
C15H26N2
4
MEDIUM
237.138
237.139
C16H16N2
10
MEDIUM
239.117
239.118
C15H14ON2
10
MEDIUM
239.153
239.154
C16H18N2
9
MEDIUM
241.133
241.134
C15H16ON2
9
MEDIUM
Continued.
Observed
Calculated
Chemical
Relative
m/z
m/z
formula of
DBE
average
neutral
abundance
species
242.117
242.118
C15H15O2N
9
LOW
243.112
243.113
C14H14O2N2
9
LOW
244.096
244.097
C14H13O3N
9
LOW
245.128
245.128
C14H16O2N2
8
MEDIUM
245.164
245.165
C15H20ON2
7
MEDIUM
247.143
247.144
C14H18O2N2
7
LOW
247.216
247.217
C16H26N2
5
MEDIUM
249.232
249.233
C16H28N2
4
MEDIUM
251.153
251.154
C17H18N2
10
LOW
253.133
253.134
C16H16ON2
10
LOW
255.112
255.113
C15H14O2N2
10
LOW
255.148
255.149
C16H18ON2
9
LOW
255.185
255.186
C17H22N2
8
LOW
258.112
258.112
C15H15O3N
9
LOW
259.143
259.144
C15H18O2N2
8
LOW
259.180
259.180
C16H22ON2
7
LOW
269.127
269.128
C16H16O2N2
10
LOW
283.143
283.144
C17H18O2N2
10
LOW
∗ species detected as an ion-radical.
The numbers of reproducibly detected formulas are shown in Fig. 3 in blue.
PM2.5 from dung cook fires had a higher observed chemical complexity
(i.e., had more observed peaks) than PM2.5 from brushwood cook fires.
Further, there were more observed peaks in PM2.5 from dung–angithi cook fires
compared to dung–chulha cook fires. There were 93 compounds reproducibly detected
in the brushwood–chulha cook fire PM2.5 samples compared to 212 and 262 for
dung–chulha and dung–angithi cook fires, respectively. There were five compounds the
chulha cook fires had in common, with two of them being the potassium salt peaks
described earlier. There was one compound (C14H16O3) shared
by only dung–angithi and brushwood–chulha. Because of the small number of these peaks,
they will not be discussed in this paper. In the following sections, we
discuss compounds that are common in all cook fires, as well as unique
compounds.
A summary of the inventory in terms of the count-based, normalized
peak abundances. (a) Contribution of PM2.5 compounds to each elemental
formula category for those found in all cook fires and those found in all
dung-burning cook fires. (b) The compounds by cookstove type classified as
compounds common to all cook fires in grey, compounds common to all dung
cook fires in brown, and unique compounds in orange.
Figure 4 summarizes the BBOA inventory described in more detail in Sect. 3.3–3.5,
i.e., compounds common to dung–chulha, dung–angithi, brushwood–chulha cook fires;
compounds found exclusively in the emissions from brushwood–chulha cook fires; and
species that are unique to dung cook fires. Figure 4a compares the fraction
of count-based, normalized abundance in each elemental category. PM2.5
compounds shared among all samples of this study are diverse. In terms of
count-based abundance, compounds emitted from all dung-burning cook fires are
largely nitrogen-containing. From Fig. 4b, the common compounds make up
the vast majority (97 %) of detected compounds from the brushwood–chulha
cook fires. Similarly for the dung cook fires, the common cook fire compounds
(grey) and dung cook fire compounds (brown) make up 95 % or more of the
mass spectra abundance as shown in Fig. 4b. Therefore, the common
compounds (Table 1) and dung compounds inventories (Table 2) contain the
bulk of the PM2.5 species in terms of count-based abundance.
Compounds common to dung–chulha, dung–angithi, brushwood–chulha cook fires
Table 1 provides a complete list of eighty reproducibly detected compounds
that were common to emissions from all cook fires. These common compounds
make a large contribution to the mass spectra for every cook fire type
(Fig. 4), with MEDIUM being the most common relative abundance given in
Table 1. More than half of the abundance (59 %) was due to the
nitrogen-containing compounds (CxHyNw or
CxHyOzNw), as shown in Fig. 4a. ESI detection likely
biases the elemental make up of smoke PM2.5, as nitrogen-containing
species are more easily ionized compared to sugars and lignin-derived
compounds (Laskin et al., 2010; Wan and Yu,
2006). Nevertheless, a large overlap in the CxHyNw and
CxHyOzNw species was observed.
The common compounds make up a large fraction for all cook fire types. This
is especially true for the sample from brushwood–chulha cook fires where their
fraction is ∼ 86 % in number. Many of these
CxHyOz species have elemental formulas consistent with
typical lignin and cellulose-derived products such as anisaldehyde,
veratraldehyde, vinylguaiacol, syringylethanone, trimethoxyphenylethanone,
etc. reported previously in the literature
(Laskin et
al., 2009; Simoneit et al., 1993; Smith et al., 2009). These tentative
molecular assignments are listed in Table 1 alongside their elemental
formulas.
Approximately 20 % of the common compounds (17 out of 80 formulas) have
been also identified in earlier studies reporting molecular characterization
of PM2.5 samples collected from burning of one or more of the following
fuels: Alaskan duff, ponderosa pine duff, southern United States pine
needles, or ceanothus fuels
(Laskin et al.,
2009; Smith et al., 2009). Many of these fuels are nonwoody and all are
undigested biomass, very different kinds of biomass from those used as
cookstove fuels in this study and in this region of India. This suggests
that perhaps 20 % of the compounds listed in Table 1 might be reproducibly
detected in BBOA samples using ESI-MS, regardless of biomass type. The
overlap is not surprising as all biomass is composed of three polymers:
lignin, hemicellulose, and cellulose
(Collard and Blin, 2014).
Compounds found exclusively in the emissions from brushwood–chulha cook fires
Table S4.1 lists the compounds observed exclusively in the samples from
brushwood–chulha cook fires. Many of them correspond to lignin-derived products
that have been previously identified in BBOA by gas chromatography methods,
as indicated in Table S4.1
(Lee
et al., 2005; Simoneit, 2002; Simoneit et al., 1993; Smith et al., 2009).
Lignin is an essential component of wood, comprising roughly a third of its
dry mass
(Collard
and Blin, 2014; Simoneit, 2002). Lignin is generally composed of
p-coumaryl, coniferyl, and syringyl alcohol units. During pyrolysis, the
coumaryl, vanillyl, and syringyl moieties are preserved and
are found in smoke. More generally, the lignin pyrolysis products found in
smoke contain a benzene ring often with hydroxy and/or methoxy substituents.
Based on these previous observations, and the assumption that these are
lignin pyrolysis products, tentative molecular structures were assigned to
CxHyOz compounds. It is likely that some
CxHyOz molecular species specific to the emissions from the
brushwood burning were not detected in this study due to their low
ionization efficiency.
Species unique to dung smoke PM2.5
Overall, the chemical composition of PM2.5 samples of dung-burning
emissions was observed to be far more complex than the samples from the
brushwood-burning cook fires. Table 2 lists the 115 compounds found
exclusively and reproducibly in the dung-fueled samples. These compounds are
largely CxHyNw, as shown in Fig. 4b. Only a few of the
elemental formulas, C8H16N2, C11H8N2, and
C13H11ON, have been reported previously
(Laskin et al.,
2009; Smith et al., 2009).
In addition to the common dung compounds listed in Table 2, there were
compounds detected exclusively in the emissions from either dung–chulha cook fires
(Table S4.2) or dung–angithi cook fires (Table S4.3). All of these compounds are
nitrogen-containing, and none have been reported previously as BBOA
compounds, to the best of our knowledge. However, in this section, we
combine all compounds found in dung-burning cook fire PM2.5,
presented in Tables 2, S4.2, and S4.3, and discuss their possible
molecular character.
Figure 5 shows the double bond equivalent (DBE) as a function of the carbon
number of compounds detected in all investigated samples. The DBE versus C
dependence for classes of compounds with different degrees of unsaturation,
including: terpenes (red), polyenes (orange), polycyclic aromatic
hydrocarbons (yellow shaded) are also shown to aid the classification
of the compounds observed in the PM2.5 samples. A total of 30 of the 193
formulas fall in the PAH region of the plot, suggesting that they have
aromatic structures (Fig. 5a). Figure 5b compares the DBE values of the
molecular components detected in the emissions exclusive to
brushwood–chulha cook fires (Table S4.1) and the common compounds from all studied
samples (Table 1). In general, the DBE increases with carbon number for the
compounds common to all cook fires. Only 8 of the 87 compounds fall
directly in the PAH region. There are more aromatic structures specific to
the dung smoke compared to the compounds detected in all cook fires.
Double bond equivalent (DBE) as a function of the carbon number
for (a) a combined set of compounds detected from all dung cook fires (brown
circles) and (b) compounds that all cook fires have in common (grey diamonds as well
as compounds exclusively found in brushwood (blue circles). Markers
representing one or multiple species are sized by their LOW, MEDIUM, and
HIGH designations. The curves illustrate theoretically where terpenes (red)
and polyenes (green) would fall. Similarly, the yellow-shaded region shows
where PAHs would appear, including: cata-condensed PAHs with 0, 1, and 2
heterocyclic nitrogen atoms and circular PAHs.
Detected nitrogen compounds with high DBE values are likely N-heterocyclic
PAH compounds. Fig. 6 displays possible structures for the select detected
nitrogen-containing compounds with a high DBE.
Purcell et al. (2007) found that pyridinic PAH
compounds were readily ionized from standard mixtures of N-heterocyclics in
positive-ion ESI. This gives us more confidence in our observation of
C13H9N, tentatively acridine, and C11H8N2,
tentatively β-carboline, which have pyridinic nitrogen atoms and
likely have high ionization efficiencies. The peak abundances of these
compounds are significant, with medium and high designations, respectively.
C12H9ON cannot have a pyridinic nitrogen and is tentatively
assigned as phenoxazine.
Possible structures of N-heterocyclic PAHs found in dung cook fire
emissions. C13H9N was detected reproducibly in dung–chulha emissions
only, while C12H9ON and C11H8N2 were reproducibly
detected in all dung cook fires.
Kendrick analysis identifies homologous series of structurally related
compounds that share a core formula and differ in the number (n) of
additional CH2 units (Hughey
et al., 2001). 172 of the 193 detected compounds from the dung-burning
cook fire emissions can be grouped into 43 homologous series based on the
Kendrick mass defect plot, as shown in Fig. 7. There are 15 homologous
series and 5 independent formulas that make up the 61 total
CxHyNw (red) compounds. This suggests that there are at least
20 distinct types of structures that made up the observed
CxHyNw species. Similarly, there are 30 homologous series
for CxHyOzNw (purple) formulas and 12
CxHyOzNw formulas yielding at least 42 distinct types
of structures for this formula category. There are no homologous series from
CxHyOz species, presumably because only a few members of
this group can be detected by ESI-based methods in the PM2.5 from the
dung cook fires. From this analysis, there are at least 66 unique types of
structures in the 193 compounds detected from dung-burning cook fire
emissions. This Kendrick analysis suggests that some of the observed
N-heterocyclic PAHs have alkyl substituents. For example, phenoxazine and
β-carboline (Fig. 6) serve as the core molecules in the homologous
series CnH2n-15ON and CnH2n-14N2, respectively
(Fig. 7).
The CH2 Kendrick mass defect plot for compounds emitted only
from dung stoves. The marker color determines the compound category for
CxHyNw compounds (red), CxHyOz (blue), or
CxHyOzNw (purple). Marker shape indicates the stove(s)
that reproducibly produced the compound: chulha and angithi
(•), angithi (□), or chulha (+).
Homologous series are identified with dotted horizontal lines, suggesting that they
have similar structures.
Light-absorbing properties and chromophores from cookstove
emissions
Figure 8 shows MACbulk values, which represent bulk absorption
coefficient normalized by mass concentration of organic solvent extractable
components. The MACbulk values were determined assuming that 50 %
of the particle mass could be extracted from the filter. Error bars account
for the uncertainties in the extraction efficiency (relative error 40 %),
and flows during sample collection (relative error 10 %). MACbulk
values for the samples from the brushwood burning are roughly twice that of
dung between 300 and 580 nm. Assuming higher EC / OC for wood compared to dung as
reported in Jayarathne et al. (2017), the results are consistent with Saleh
et al. (2014), who predict higher effective OA absorbance for higher BC-OA ratios.
Comparing MACbulk (m2 g-1) for organic solvent
extractable material from brushwood–chulha (blue) and dung–chulha (red) samples. Shaded
regions represent errors due to extraction efficiency and sampling flow
rates.
HPLC-PDA chromatogram showing BrC chromophores detected in the
emission samples from (a) brushwood and (b) dung cook fires. Highly absorbing
molecules and their corresponding PDA retention times are given above the
peak.
MACbulk values at 400 nm were 1.9 ± 0.8 and
0.9 ± 0.4 m2 g-1 for the samples from brushwood–chulha and
dung–chulha cook fires, respectively. For comparison, Kirchstetter and colleagues
reported MACbulk of 2.9 m2 g-1 at 400 nm for the BrC in
biomass smoke samples (Kirchstetter et
al., 2004). Chen and Bond measured MACbulk values at 360 nm of nearly
2.0 m2 g-1 for methanol extracts of particles resulting from oak
pyrolysis, and nearly 2.5 m2 g-1 for pine wood pyrolysis
(Chen and Bond, 2010). Our MACbulk value at 360 nm for
brushwood was larger at 3.7 ± 1.5 m2 g-1, possibly due to a
more efficient extraction of a broader range of chromophores by the utilized solvents. The pyrolysis temperature and wood composition could also
contribute to this difference. Our MACbulk value at 360 nm for dung was
lower compared to our brushwood sample at 1.8 ± 0.8 m2 g-1.
This could be a combined result of the likely lower pyrolysis temperature
and difference in the biomass composition (Chen and Bond,
2010).
While the MACbulk values are smaller for the dung–chulha cook fires, the
PM2.5 emission factors (a detailed analysis of the emission factors
will be reported in a follow up paper) are more than a factor of 2.5 higher
for dung–chulha fires (21.1 ± 4.2 g kg-1 fuel) compared to
brushwood–chulha fires (7.3 ± 1.8 g kg-1 fuel). The product MACbulk × EF
can be used to estimate the contribution of smoke to the
absorption coefficient for the per unit mass of the fuel burned. At 400 nm,
MACbulk × EF = 19.0 ± 9.2 m2 kg-1 fuel and
13.9 ± 6.8 m2 kg-1 fuel for dung–chulha fires and
brushwood–chulha fires, respectively. For particles that are small in diameter
relative to the wavelength, MACaerosol∼ 0.7 × MACbulk
(Laskin et al., 2015). Based on this we can estimate
MACaerosol × EF = 13.3 ± 6.5 m2 kg-1 fuel and
9.7 ± 4.8 m2 kg-1 fuel for dung–chulha fires and
brushwood–chulha fires, respectively. The values are somewhat higher than the “EF
Babs 405 just BrC” values reported by Stockwell et al. (2016) at 405 nm,
which were 8.40 m2 kg-1 fuel and 5.43 m2 kg-1 fuel
for hardwood cooking smoke and dung cooking smoke, respectively. However,
both the present results and the data from Stockwell et al. (2016) show
that the dung-based and wood-based fuels make comparable contributions to
the absorption coefficient of the smoke for same amount of fuel consumed.
The AAE values for the extractable organics in brushwood and dung samples
are 7.5 and 6.8, respectively. Our brushwood AAE fits into the lower end of
the AAE range for extracted organics presented in Chen and Bond, 6.9 to 11.4
(Chen and Bond, 2010). Typical AAE values cited in the
literature for BrC in BBOA are in a range of 2–11
(Kirchstetter et al., 2004;
Laskin et al., 2015). The AAE of the entire cooking aerosol (with the
contribution of the insoluble BC included) should be lower. For example,
Stockwell et al. (2016) reported in situ measurements of AAE of 3.01 and 4.63
for brushwood and dung cooking particles, respectively.
The list of retention times, absorption peak maxima, and chemical
formulas of the BrC chromophores detected in the brushwood smoke sample.
Tentative assignments are given based on compounds previously identified in
the lignin pyrolysis literature.
LC retention
λmax
Nominal molecular
Chemical
Tentative
time (min)
(nm)
weight (amu)
formula(s)
assignment
6.26
383
192
C9H8N2O3
7.15
392
141
C7H8O3
10.55
305
183
C9H10O4
Homovanillic acid/syringealdehyde
13.29
265
155
C8H10O3
Syringol
14.44
305
169
C8H8O4
Vanillic acid
183
C9H10O4
Homovanillic acid/syringealdehyde
15.57
299
181
C10H12O3
Ethyl-3-methoxybenzoate
167
C9H10O3
Veratraldehyde
16.95
313, 334
186
C11H7NO2
17.25
331
165
C9H8O3
162
C9H7NO2
18.13
341
209
C11H12O4
18.32
229, 337
179
C10H10O3
19.78
305, 330
194
C10H10O4
Ferulic acid
24.11
290, 330
259
C15H14O4
28.07
334
184
C8H9NO4
29.24
330
198
C13H10O2
230
C13H10O4
33.81
340
227
C14H10O3
The list of retention times, absorption peak maxima, and chemical
formulas of the BrC chromophores detected in the dung smoke sample.
Tentative assignments are given based on compounds previously identified in
the lignin pyrolysis literature.
LC retention
λmax
Nominal molecular
Chemical
Tentative
time (min)
(nm)
weight (amu)
formula(s)
assignment
8.5
295
167
C8H9NO3
9.09
282,300
166
C9H10O3
168
C8H8O4
10.59
252, 289, 393
182
C9H10O4
Homovanillic acid/syringealdehyde
12.22
282
122
C7H6O2
Benzoic acid
14.44
306
168
C8H8O4
Vanillic acid
182
C9H10O4
Homovanillic acid/syringealdehyde
164
C9H8O3
15.57
300
174
C10H12O3
Ethyl-3-methoxybenzoate
166
C9H10O3
Veratraldehyde
16.35
286
174
C11H10O2
18.28
290, 330∗
162
C10H10O2
19.5
323∗
220
C12H12O4
19.72
331∗
194
C10H10O4
Ferulic acid
20.85
352∗
188
C12H12O2
24.54
299, 308
178
C10H10O3
25.28
290, 320
218
C12H10O4
29.17
332
198
C13H10O2
230
C13H10O4
29.6
358∗
213
C13H9O3
∗ signifies a shoulder, rather than a clear peak
We now focus on identifying selected chromophores that contribute to the
high MACbulk we observe for cookstove PM2.5. Two cook fires using
dung and brushwood fuels were selected for a more detailed analysis of the
light-absorbing molecules (BrC chromophores). The dung cook fire utilized an
angithi cookstove to prepare buffalo fodder. The brushwood cook fire was used to
prepare a traditional meal of rice and lentils with a chulha. More detailed sample
information is provided in Table S1.3. The samples were analyzed using
HPLC-PDA-ESI–HRMS platform following the methods described elsewhere
(Lin
et al., 2015, 2016, 2017). The identified chromophores and their PDA
chromatograms are illustrated in Fig. 9, and the retention times and peaks
in the absorption spectra are listed in Tables 3 and 4 for the emissions
from brushwood and dung cook fires, respectively.
The BrC chromophores for both brushwood and dung samples are largely
CxHyOz compounds (Tables 3 and 4). We conclude that
lignin-derived BrC chromophores account for the majority of the extracted
light-absorbing compounds in both samples. We also found a few
nitrogen-containing BrC chromophores (e.g., C9H7NO2 and
C8H9NO3) in both the brushwood and dung samples. The woody
and digested biomasses shared three strongly absorbing chromophores,
C8H8O4 (tentatively vanillic acid), C10H12O3
(tentatively ethyl methoxybenzoate), and C13H10O2, as well as
comparably weaker-absorbing chromophores.
C10H10O3 is another strong absorber of near-UV radiation
that was found in both samples. In the brushwood-derived PM2.5, C10H10O3
elutes at 18.3 min (λmax= 337 nm), while in the dung smoke
sample, it is not observed until 24.5 min
(λmax= 299, 308 nm). These are clearly different
chromophores with the same chemical formula, possibly coniferaldehyde and
methoxycinnamic acid. C9H8O3 is a similar case, in which
the same chemical formula appears at different retention times in the
selected ion chromatograms (SICs) for brushwood- and dung-derived
PM2.5. In the brushwood-derived PM2.5 sample,
C9H8O3 coelutes with C9H7NO2 at 17.3 min
(Table 3). In the dung PM2.5 sample C9H8O3 coelutes
with C8H8O4 and C9H10O4 at 14.4 min (Table 4).
The C9H8O3 formula could correspond to coumaric acid for
either retention time. Because the compound coelutes with other potential
chromophores, we refrained from assigning a proposed structure to the
chemical formula.
There were light-absorbing molecules specific to brushwood-derived
PM2.5 (Table 3) that could account for higher MACbulk values
compared to the dung-derived PM2.5. C8H9NO4 is a
possible nitroaromatic compound with its absorbance peaking around 335 nm.
C8H10O3, tentatively syringol, is closely related to syringic
acid, a lignin monomer. The formula was also detected in the dung-derived
PM2.5 sample, but the absorption was lower by approximately a factor
of 20 is therefore not considered a main chromophore.
There were strongly absorbing BrC chromophores in the PM2.5 generated
by burning dung fuel that eluted in the first couple of min of the
sample run (See Fig. 9b). These early eluting chromophores were likely
polar compounds that were not retained well by the column and thus could not
be assigned. The challenges with assigning co-eluting chromophores in BBOA
were previously noted by Lin et
al. (2016). For both PM2.5 samples, most of the chromophores eluted in
the first 30 min of the run shown in Fig. 9. Compounds eluting in the
range of 30 to 60 min were also satisfactorily separated, but these were
weakly absorbing. The nonpolar PAH compounds absorbing in UV-Vis range are
not ionized by the ESI source and subsequently not detected by HRMS
(Lin et al., 2016). It is possible
that additional light-absorbing molecules essential to dung smoke were
strongly retained by the column and eluted after 60 min.
Absorption spectra recorded in tandem with the mass spectra provide
additional constraints on the assignments. For example, at 15.6 min
C10H12O3 and C9H10O3 coeluted in both BBOA
samples. These compounds were given the tentative assignments of
ethyl-3-methoxybenzoate and veratraldehyde, respectively. The UV-Vis
absorbance of ethyl-3-methoxybenzoate shown in Fig. 10 provides a
reasonable match for the recorded PDA spectra for both samples at a
retention time of 15.6 min. Veratraldehyde, which is derived broadly from
lignin, has an absorption spectrum that peaks at 308 nm in aqueous solution
(Anastasio et al., 1997). Therefore, both
ethyl-3-methoxybenzoate and veratraldehyde contribute to the spectrum
observed by the PDA detector, although they cannot be completely separated
with this HPLC protocol.
UV-Vis absorption spectra from the PDA analysis of cookstove BBOA
samples. The blue and red curves represent the background-subtracted
absorbance at retention time of 15.57 min for brushwood-derived PM2.5 and dung-derived PM2.5, respectively. The reference absorption
spectrum of ethyl-3-methoxybenzoate (green) was reproduced from the NIST
Chemistry WebBook database (Talrose et al.,
2017). The structure of ethyl-3-methoxybenzoate is pictured.
For many formulas, multiple structural isomers were observed in SICs with
peaks appearing at more than one retention time. This behavior has been
observed for other types of BBOA samples, described in
Lin et al. (2016), and is inherent
to lignin's nature, such that a single CxHyOz chemical formula
can correspond to multiple possible structural isomers. There are several
cases in which chemical formulas show up multiple times in Tables 3–4. An
example from the brushwood PM2.5 (Table 3) is C9H10O4
which elutes at 10.6 and 14.4 min. C9H10O4 has been
previously found in lignin pyrolysis BBOA in the forms of homovanillic acid
and syringealdehyde (Simoneit et al., 1993).
C8H8O4 and C9H10O3 are additional
examples of the similar occurrence in the sample of dung-derived PM2.5,
as they both appear twice in the SICs as shown in Table 4. One peak
corresponding to C8H8O4 is very likely to be vanillic acid
(Simoneit, 2002; Simoneit
et al., 1993). C9H10O3 could be either veratraledehyde
or homoanisic acid, both have been observed from lignin pyrolysis
(Simoneit et al., 1993). Collectively, these results indicate
that many of the lignin-like chromophores have multiple structural isomers,
some of which have likely been observed before
(Simoneit, 2002; Simoneit
et al., 1993).