Organonitrate group concentrations in submicron particles with high nitrate and organic fractions in coastal southern California

Abstract

During wintertime measurements in coastal southern California, organonitrate groups accounted for up to 10% of organic mass (OM) in submicron particles. In this study, we report the calibrated absorptivity, the uncertainties in the calibrations, the detection limits for 12 and 24 h ambient sampling, and the multipeak retrieval algorithm for the method developed. Organonitrate groups were observed when both submicron particle-phase nitrate and OM concentrations exceeded 1 μg m−3. These high concentrations were associated with a mixed urban fossil fuel combustion source type that had potential source regions near Riverside and the South Coast Air Basin. The high frequency of these organonitrate observations contrasts with a number of studies of aerosol particles in other regions with more humid conditions, in which organonitrate groups were not detected and submicron sulfate concentrations exceeded those of nitrate. Our results suggest both that organonitrates form and/or exist in significant concentrations during polluted urban conditions and that their lifetime may be limited by hydrolysis in the particle phase.

Full text

Page 1

Author's personal copy
Organonitrate group concentrations in submicron particles with high
nitrate and organic fractions in coastal southern California
Douglas A. Daya, Shang Liua, Lynn M. Russella,*, Paul J. Ziemannb
aScripps Institution of Oceanography, University of California, San Diego, 9500 Gilman Drive, Mail Code 0221, La Jolla, CA 92093-0221, USA
bAir Pollution Research Center and Department of Environmental Sciences, University of California, Riverside, CA 92521, USA
a r t i c l e i n f o
Article history:
Received 14 January 2010
Received in revised form
26 February 2010
Accepted 26 February 2010
Keywords:
Organonitrate
Organic nitrates
Aerosol
FTIR
PMF
a b s t r a c t
During wintertime measurements in coastal southern California, organonitrate groups accounted for up
to 10% of organic mass (OM) in submicron particles. In this study, we report the calibrated absorptivity,
the uncertainties in the calibrations, the detection limits for 12 and 24 h ambient sampling, and the
multipeak retrieval algorithm for the method developed. Organonitrate groups were observed when
both submicron particle-phase nitrate and OM concentrations exceeded 1 mg m?3. These high concen-
trations were associated with a mixed urban fossil fuel combustion source type that had potential source
regions near Riverside and the South Coast Air Basin. The high frequency of these organonitrate
observations contrasts with a number of studies of aerosol particles in other regions with more humid
conditions, in which organonitrate groups were not detected and submicron sulfate concentrations
exceeded those of nitrate. Our results suggest both that organonitrates form and/or exist in significant
concentrations during polluted urban conditions and that their lifetime may be limited by hydrolysis in
the particle phase.
? 2010 Elsevier Ltd. All rights reserved.
1. Introduction
Organonitrate molecules (ReONO2) have been shown to
comprise 10e20% of carbonaceous aerosol mass at urban locations,
of which the organonitrate functional group accounts for 5e10% of
OM (O'Brien et al., 1975; Mylonas et al., 1991; Laurent and Allen,
2004). Organonitrates typically have been observed to have
maximum loadings in submicron-sized particles (Mylonas et al.,
1991; Garnes and Allen, 2002; Laurent and Allen, 2004). Analyt-
ical techniques employed in field studies include separation and
quantification of individual organic molecules by gas chromatog-
raphy mass spectrometry (GCMS) after extraction from filters and
measurements of relative abundance from the spectroscopic
signatures of Fourier Transform Infrared (FTIR) analysis. Recently,
aerosol mass spectrometry has also been investigated as a tech-
nique to measure particulate organonitrate compounds, resulting
in estimates of 0.8e1.6mg m?3(10e19% of OM) or0.2e0.5mg m?3of
organonitrate functional groups during the Study of Organic
Aerosols at Riverside (SOAR) campaign (Farmer et al., 2010). In
another study, Bruns et al. (2010) calculated that a ratio of at least
0.15 for the organic-to-inorganic nitrate ratio contained in aerosol
is required to indicate the presence of organonitrates using an
Aerosol Mass Spectrometer (AMS) and utilizing the differences in
mass fragmentation patterns for ionic and covalently bound nitrate
groups.
Organonitrates are formed in polluted air during the day
through association reactions of NO with organic peroxy radicals
(formed from VOCs via reactions initiated by OH radicals or O3) and
at night through NO3radical-initiated reactions of alkenes (Roberts,
1990; Atkinson, 1997; Atkinson and Arey, 2003; Gong et al., 2005;
Ng et al., 2008; Fry et al., 2009; Lim and Ziemann, 2009;
Matsunaga and Ziemann, 2009). Oxidation of alkanes, alkenes,
and aromatics (the major atmospheric VOCs) leads to a variety of
organonitrate products that are mostly multifunctional. Depending
on the VOC and reaction conditions, these products can include
hydroxynitrates, dihydroxynitrates, carbonylnitrates, and hydro-
peroxynitrates, as well as monofunctional alkyl nitrates. Studies
indicate that the molar yields of first-generation high molecular
weight organonitrates that can potentially form secondary organic
aerosol (SOA) can approach w40e60% for both OH and NO3radical
reactions (Atkinson,1997; Atkinson and Arey, 2003; Fryet al., 2009;
Matsunaga and Ziemann, 2009). Alkyl nitrates, which are the most
volatile organonitrates, are predicted to exist predominantly in the
particle phasewhen carbon numbers are larger than wC20(Lim and
Ziemann, 2009), but this approximate carbon number threshold
will be much lower for multifunctional organonitrates.
* Corresponding author. Tel.: þ1 858 534 4852; fax: þ1 858 534 4851.
E-mail addresses: daday@ucsd.edu (D.A. Day), shl014@ucsd.edu (S. Liu),
lmrussell@ucsd.edu (L.M. Russell), paul.ziemann@ucr.edu (P.J. Ziemann).
Contents lists available at ScienceDirect
Atmospheric Environment
journal homepage: www.elsevier.com/locate/atmosenv
1352-2310/$ e see front matter ? 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.atmosenv.2010.02.045
Atmospheric Environment 44 (2010) 1970e1979

Page 2

Author's personal copy
Improved quantification of organonitrate groups in additional
ambient conditions may allow for a more complete budget of
organic particle mass and may provide insight into mechanisms
responsible for SOA formation and sources of other functional
groups. In this work, we extend the FTIR-based identification of
organonitrate groups using a calibration to two laboratory stan-
dards: 2-ethylhexyl nitrate and organonitrate product mixtures
formed in smog chamber reactions. The quantification of ambient
organonitrate group concentrations from this technique provides
further evidence of the atmospheric emission sources and ambient
conditions that lead to organonitrate formation.
2. Methods
2.1. Sample collection and analysis
Measurements were made on submicron particles collected
nearly continuously at the Scripps Institution of Oceanography
(SIO) Pier (32?520N,117?150W) from 23 February 2009 to 30 March
2009. Aerosol particles were sampled from a common inlet located
4 m above the pier (15 m above sea level) at 16.7 standard L min?1.
Sample air passed through a PM1 cyclone to select for submicron
particles. The sample flow was split, and aerosol particles were
collected on two sets of duplicate 37 mm diameter Teflon filters for
quantification of organic functional groups by FTIR spectroscopy.
Particles were collected continuously on one filter set for 12-h
duration, with duplicate filters run in parallel for 24-h samples.
24-h samples were collected from 7 am (local, PST) to 7 am the
following morning; 12-h samples typically spanned the 12-h
periods from 7 am to 7 pm (“daytime”) and 7 pm to 7 am
(“nighttime”). Resulting total flowvolumes for short and long filters
were approximately 6 and 12 m3, respectively. A smaller flow
(0.07 L min?1) from the same inlet was sampled by a Quadrupole
AMS for quantification of size-resolved inorganic and organic mass
fragments in non-refractory submicron particles.
FTIR spectroscopy was used to quantify organic functional group
concentrations for the particles collected on each Teflon filter.
Detailed descriptions of the method, cleanroom conditions, and
standard compound calibrations for the Bruker Tensor 27 spec-
trometerwith DTGS detectorcan be found in Gilardoni et al. (2007),
based on similar techniques calibrated for a different spectrometer
byMaria et al. (2002, 2003). Revisions tothose methods include use
of an automated algorithm for baselining, peak fitting, and inte-
gration, as well as additional calibrations of primary amine and
carboxylic acid functional groups (Liu et al., 2009; Russell et al.,
2009). The organic functional groups typically quantified include
saturated aliphatic (hereafter alkane [CH]) groups, alcohol [COH]
groups (including polyols and other organic hydroxyl-containing
compounds), carboxylic acid [COOH] groups, non-acidic carbonyl
[CO] groups, primary amine [CNH2] groups, and organosulfate
[COSO3] groups; unsaturated aliphatic (hereafter alkene [CH])
groups and aromatic [CH] groups were below detection limit for all
samples, with each of these two functional groups accounting for at
most 3% of OM. The spectra were analyzed for evidence of orga-
nosulfate groups at 876 cm?1, but detectable peaks were identified
in only one sample and are excluded from this analysis. The mass
associated with each mole of absorbing bonds in the identified
functional group is: alkane group (7), alcohol group (23), carboxylic
acid group (45), non-acid carbonyl group (28), amine group (11),
organosulfate group (102), alkene group (13), aromatic group (13),
and organonitrate group (68). Note that for alkane, alcohol, primary
amine, and organosulfate groups, which share their C atom
with another group, half of the C atom mass is assigned to each
group. Identification and quantification of organonitrates are dis-
cussed in Sect. 2.2.
The AMS (Aerodyne, Billerica, MA) was used to measure mass
spectra of non-refractory organic and inorganic components of
submicron particles, as discussed by Jayne et al. (2000), Jimenez
et al. (2003), and Allan et al. (2004). In this instrument, ambient
aerosol enters through a 100 mm orifice resulting in reduced pres-
sure (w1.2 torr) and is subsequently focused into a narrow beam
(<1 mm diameter) with an aerodynamic lens. Transmission effi-
ciency is expected to be near 100% for 60e600 nm and then
decrease to about zero at w30 nm and w1.5 mm (Jayne et al., 2000;
Jimenez et al., 2003; Zhang et al., 2004). Particles are vaporized by
impaction on a ceramic heater maintained at 600?C. The vapor-
ization source is coupledtoan electron impactionizer (70 eV) at the
entrance to a quadrupole mass spectrometer. Particle size is
measured from the time-of-flight between a chopper that modu-
lates transmission of discrete aerosol packets and chemical detec-
tion at the mass spectrometer.
During this study, AMS operation was alternated between the
“mass spectrum” (MS) mode and the “time-of-flight” (TOF) mode
every 5 min, with 20 m/z scanned in TOF mode. The ionization
efficiency was calibrated weekly using dry 350 nm ammonium
nitrate particles generated with an atomizer and size-selected with
a Differential Mobility Analyzer (DMA), following mass tuning and
optimization of the quadrupole mass spectrometer and electron
multiplier. Analysis of the AMS measurements was accomplished
with small modifications to the ion fragmentation table and “batch
table” based on tests outlined in Middlebrook et al. (submitted for
publication). Since AMS measurements of nitrate and sulfate
include both organic and inorganic nitrate and sulfate, here we
refer to these components as simply nitrate (NO3) and sulfate (SO4)
or AMS nitrate and AMS sulfate rather than exclusively as inorganic
components. However, we expect that most of these components
werecomprised of inorganic compounds (>90%), whichis apparent
in comparing the organonitrate and AMS nitrate measured in this
study. Concentrations were averaged to 1 h and for the time span of
each filter sample collection period. Comparison of total organic
mass (OM) measured by FTIR to non-refractory OM (nrOM)
measured by AMS (no collection efficiency applied) yielded a slope
of 0.72 with a Pearson's correlation coefficient, r, of 0.75 (as shown
in Fig. 1a). All linear regressions reported in this manuscript were
calculated using a reduced major axis regression, thus allowing for
comparable uncertainty in both axes. Given that FTIR OM was less
than AMS nrOM, that the particle mass was less than 50% ammo-
nium sulfate (precluding its use as a surrogate for collection effi-
ciency, as in (Quinn et al., 2006)), and that other independent
chemical comparisons were not available, therewas no information
to justify a non-unity collection efficiency (CE). Consequently,
CE ¼ 1 is used in this analysis. Differences between these two
measures of OM may result from the uncertainties related to cali-
brations of both techniques (?20e25%) including the uncertainty
in the assignment of relative ionization efficiencies for organic AMS
mass fragments (Zhang et al., 2005b) and also to the potential for
underprediction of heteroatoms in both FTIR and AMS techniques
(Russell, 2003; Farmer et al., 2010).
2.2. Organonitrate calibration and quantification
The most common method used to quantify particle-phase
organonitrate groups is FTIR. Peaks located at 1620e1630 cm?1,
1260e1280 cm?1, and 855e860 cm?1have typically been used for
identification and quantification (Mylonas et al., 1991; Allen et al.,
1994;Dekermenjianetal., 1999).
755 cm?1and 700 cm?1have also been identified for some orga-
nonitrate compounds. Both water vapor and amine functional
group absorbances may interfere with the 1630 cm?1peak. The
absorbance peak at 1280 cm?1is well suited for quantification due
Weakerabsorbancesat
D.A. Day et al. / Atmospheric Environment 44 (2010) 1970e1979
1971

Page 3

Author's personal copy
to its sharpness, isolation, and greater (or equal) intensity
compared to other peaks.
Fig. 2 shows spectra obtained using the Bruker Tensor 27 FTIR
spectrometerto analyze laboratorystandardsandambient particles
collected on Teflon filters. Three frequencies previously attributed
to the strongest organonitrate absorption peaks are located at
1629.6, 1278.6, and 860.9 cm?1. The 2-ethylhexyl nitrate (2-EHN)
spectrum was produced by atomizing liquid 2-EHN (Aldrich, 97%
purity) while sampling with a Teflon filter at 1 L min?1. The three
distinctive peaks at 1629 cm?1, 1279 cm?1, and 864 cm?1were
observed. The isobutyl nitrate (IBN) spectrum in Fig. 2b was
produced by applying a drop of liquid IBN (Aldrich, 96%) directly to
a Teflon filter. The samples were immediately scanned since rapid
vaporization of IBN occurred, which resulted in interference at the
higher frequencies due to minimal purging of the detection
chamber with nitrogen gas. Three distinct organonitrate peaks
were observed. The spectrum of oleic acid þ NO3reaction products
(Fig. 2c) was obtained from a filter sample collected after reacting
oleic acid aerosol particles with NO3radicals (formed by thermal
dissociation of N2O5) in a smog chamber at UC Riverside. The re-
action products are primarily b-hydroxynitrates and b-carbon-
ylnitrates, both of which contain a carboxylic acid group from the
oleic acid (Docherty and Ziemann, 2006). The smog chamber
reaction products show distinct organonitrate peaks at 1630 cm?1,
1273 cm?1, and 854 cm?1, and a peak at 1710 cm?1attributed to
carbonyl groups. The spectra in Fig. 2e,f are from the products of
smog chamber reactions in which 1-tetradecene and n-pentade-
cane were reacted separately with OH radicals in the presence of
NOx. Organonitrate products formed from the 1-tetradecene reac-
tion are b-hydroxynitrates and dihydroxynitrates (Matsunaga and
Ziemann, 2009) whereas those formed from the n-pentadecane
reaction are alkyl nitrates, 1,4-hydroxynitrates, and a variety of
other multifunctional mononitrates and dinitrates (Lim and
Ziemann, 2009). Organonitrates in SOA filter samples collected
from eachreaction werequantified
(Docherty and Ziemann, 2006) using 2-EHN as the calibration
standard for the n-pentadecane reaction and authentic multifunc-
tional hydroxynitrate products for the 1-tetradecene reaction
(Matsunaga and Ziemann, 2009). Organonitrate peaks in these
samples were observed at 1621 cm?1,1274 cm?1, and 861 cm?1. The
locations of the peaks for the spectra from the 1-tetradecene and n-
pentadecanereactions were
collected at the Scripps pier on 1 March 2009 (Fig. 2d) shows the
samethree absorptionpeaks
magnitudes.
To quantify the absorbance associated with each peak, an
automated algorithm for baselining and peak-fitting the three
largest organonitrate peaks was tested using the standard and
smog chamber spectra. The pre-sampling FTIR absorbance spec-
trum was subtracted from the post-sampling FTIR absorbance
spectrum. Prior to this subtraction, both spectra were scaled by the
relative Teflon absorbance to improve the comparison of the
1280 cm?1peak given its proximity tothe strong Teflon absorbance
(1240e1250 cm?1). A baseline was applied using a third-order
polynomial fit to two low-absorbance regions at 700e780 cm?1
and 1860e2200 cm?1. Fits for baselining and peak-fitting were
calculated using non-linear least squares fitting. Local baselines for
each of the three prominent organonitrate absorbance regions
were determined using different methods. The 860 cm?1peak
baseline was fit with a second-order polynomial between two
surrounding regions at 805e815 cm?1and 925e960 cm?1. The
1280 cm?1peak baseline was set to a constant value determined by
the local minimum between 1290 and 1320 cm?1. The 1630 cm?1
peak was baselined using a third-order polynomial and two regions
bracketing both the carbonyl and organonitrate group absorbances
(1500e1550 cm?1and 1850e1950 cm?1). Single Gaussian func-
tions were fit to the 860 cm?1and 1280 cm?1peaks and double
Gaussian functions were fit to overlapping peaks at 1630 cm?1
(organonitrate group) and 1715 cm?1(carbonyl group).
The rangesused to constrain the peak locations and widths were
determined using the 2-EHN standard and smog chamber samples
and implemented in the algorithm for application to ambient
by spectrophotometry
indistinguishable. The sample
withdifferentrelativepeak
abc
Fig.1. Comparison of a) OM (FTIR) vs. Organics (AMS), b) OM (FTIR) w24 h vs. w12 h samples, and c) organonitrates (FTIR): w24 h vs. w12 h samples. Fitted slopes (forced through
zero) are 0.72, 0.98, and 1.05 with Pearson's correlation coefficients (r) of 0.75, 0.94, and 0.96, respectively.
a
b
c
d
e
180016001400
Wavenumber (cm−1)
12001000800600
f
n
o i t p r o
s
b
A
Fig. 2. FTIR spectra of a) 2-ethylhexyl nitrate, b) isobutyl nitrate, c) oleic acid þ NO3
reaction products, and d) ambient sample collected at SIO pier, e) n-pentadecane þ OH/
NOxreaction products, and f) 1-tetradecene þ OH/NOxreaction products. Reactions
were conducted in a smog chamber and particulate products were analyzed. Vertical
lines indicate three frequencies attributed to the strongest organonitrate absorption
peaks(1629.6,1278.6,and860.9cm?1).Frequencyassociatedwithcarbonyl(1715cm?1)
and two weaker organonitrate peaks (755 cm?1and 700 cm?1) are shown.
D.A. Day et al. / Atmospheric Environment 44 (2010) 1970e1979
1972

Page 4

Author's personal copy
spectra. Also, an additional peak was fit in the 1500e1550 region to
account for primary amine absorbance, and another one was fit in
the 805e960 region (constrained to 876 cm?1) to account for any
organosulfate absorbance that might bias the organonitrate peak
quantification at that frequency. Table 1 summarizes the results of
the fitting algorithm for the standards, smog chamber, and ambient
sample spectra.
Absorptivities for each of the three organonitrate peaks were
determined by a calibration curve using a range of mass loadings of
2-EHN. Samples were weighed before and after FTIR scanning to
quantify evaporation during analysis. Fig. 3 shows the fitted peak
areas as a function of organonitrate group mass loading. Error bars
on the x-axis span the range of mass observed before and after each
scan. Fits of the integrated peak area of the absorbance to the molar
amount of organonitrate groups in the 2-EHN samples yielded
average absorptivities of 13.3 ? 2.7, 6.5 ? 1.4, and 8.8 ? 1.7
cm?1mmol?1for the 1630 cm?1, 1280 cm?1, and 860 cm?1peaks,
respectively. For the smog chamber samples, the absorptivity at
1630 and 1280 cm?1are lower by 20% and 60%, respectively, and at
860 cm?1the absorptivity is greater by 10%.
We use the 860 cm?1peak for quantification of organonitrate
functional groups for ambient samples, since the absorptivity at
this peak is similar between both the 2-EHN and the smog chamber
samples. This choice also avoids the uncertainty resulting from
interference of the amine absorbance near 1630 cm?1and the
Teflon absorbance near 1280 cm?1. Detection limits werecalculated
to be the peak area that exceeds the peak areas of the blank filters
collected as back-up filters for each sample at a 95% confidence
level. For 12-h and 24-h samples, that detection limit corresponded
to 0.01 and 0.005 mg m?3, respectively. The uncertainty for ambient
organonitrate concentrations presented here is the greater of half
of this detection limit or the uncertainty in the measured absorp-
tivity (?25%). Reanalysis of eleven samples for organonitrate
concentrations after 11 months of storage yielded concentrations at
70% of the initial values indicating that large evaporative or
chemical losses between sampling and analysis are unlikely. Other
functional groups did not show significant decreases following
reanalysis. Typical sample histories included one day of equilibra-
tion at 55% RH and 20?C in the cleanroom environment prior to
analysis, followed by storage at ?4?C.
3. Results
Fig. 4 and Table 2 show the variation and averages of the FTIR
and AMS measurements of submicron particle components during
the campaign. OM concentrations were 1.6 ? 1.4 mg m?3, with
daytime averages slightly higher and more variable than during
nighttime. Alkane functional groups comprised the largest fraction
of OM throughout most of the campaign, on average 40%
(0.72 ? 0.85 mg m?3). Alcohol and carboxylic acid functional groups
represented the other major contributions to OM at 25% of OM
each, with averages and standard deviations of 0.29 ? 0.19 mg m?3
and 0.40 ? 0.39 mg m?3, respectively. On average, amine, organo-
nitrate, and non-acid carbonyl functional groups (primarily
ketones, aldehydes, and esters) each comprised <10% of OM.
Organonitrate groups were observed to vary over a large range of
concentrations from 0.008 to 0.43 mg m?3, or 0.8 to 9% of OM.
Oxygen-to-carbon ratios were on average 0.57 ? 0.18, showing
moderately higher ratios during nighttime. Concentrations of
alkane and carboxylic acid functional groups showed slightly larger
concentrations during daytime, while alcohol functional groups
were lower. Fig. 4 (top) shows the time series of the organic func-
tional groups (short, w12-h samples only), demonstrating the large
variability in OM, which ranged from 0.1 to 9 mg m?3. The
concentrations tended to vary on 3e5 day (synoptic) timescales
with relatively small differences between daytime and nighttime
samples.
Inorganic and organic compounds measured by AMS are
summarized inTable 2. On average, nrOM constituted nearly half of
the total AMS non-refractory submicron particle mass (41 ? 14%,
2.1 ? 1.7 mg m?3). Sulfate was the second largest contribution at
37 ? 19% (1.5 ? 0.8 mg m?3). Nitrate and ammonium were
comparable, representing on average 11e12% of the AMS mass
(0.69 ? 1.0 mg m?3and 0.60 ? 0.43 mg m?3). Nitrate was quite
variable, however, with concentrations varying by a factor of 100,
corresponding to relative contributions of 1.5% to more than 30%.
On average, for the 36-day campaign, ammonium, sulfate, nitrate,
and nrOM showed no significant differences between nighttime
and daytime concentrations or mass fractions. The time series of
these components show changes on 3e5 day timescales, largely
tracking changes in the functional group composition.
In ordertoidentify the types of sources that contributed tothese
changing organic mixtures, we used positive matrix factorization
(PMF) of both the FTIR mass-weighted spectra (e.g. Russell et al.,
2009) and the AMS mass spectra (e.g. Ulbrich et al., 2009) to
separate source contributions to the particle components (Alfarra
et al., 2004). Different FPEAK rotation parameter values (?0.2, 0,
0.2, and 0.4) resulted in the same factors; therefore, FPEAK ¼ 0 was
used to represent the solution for both analyses.
3.1. FTIR PMF analysis
For the FTIR spectra from 1600 to 3600 cm?1, two- to six-factor
solutions with FPEAK were investigated. The five-factor solution
Table 1
Summary of peak-fitting results and absorptivity quantification. Means and standard deviations (parenthesis) for each sample set are shown for three peak locations (cm?1),
peak widths (cm?1; s20.5), and relative absorptivities.
Isobutyl nitrate (IBN) 2-Ethylhexyl
nitrate (EHN)
Oleic acid þ NO3
1-Tetradecene or
n-pentadecane (þOH/NOx)
1621.3 (2.0)
14.6 (1.2)
1274.2 (0.7)
8.8 (0.5)
861.0 (2.9)
25.3 (5.1)
2.6 (0.28)
2.2 (0.50)
6
10.9 (2.5)
4.1 (0.9)
9.8 (0.8)
Pier Winter 2009
Location 1
Width 1
Location 2
Width 2
Location 3
Width 3
Area 1:Area 2
Area 3:Area 2
Sample size
Absorptivity 1 (area mmol?1)
Absorptivity 2 (area mmol?1)
Absorptivity 3 (area mmol?1)
1630.2 (2.3)
43.4 (0.3)
1283.9 (0.6)
20.6 (0.3)
864.7 (0.1)
35.0 (0.3)
2.05 (0.04)
1.71 (0.11)
2
e
e
e
1629.0 (1.5)
17.8 (2.8)
1278.5 (0.5)
9.9 (2.2)
863.8 (1.1)
19.4 (0.8)
2.13 (0.21)
1.22 (0.08)
12
13.3 (2.7)
6.5 (1.4)
8.8 (1.7)
1629.6 (e)
22.6 (e)
1273.3 (e)
15.6 (e)
854.3 (e)
21.3 (e)
2.16 (–)
1.21 (e)
1
e
e
e
1634.5 (1.0)
18.7 (3.3)
1280.2 (2.1)
10.6 (2.5)
854.1 (0.1)
22.2 (7.2)
1.93 (0.98)
0.91 (0.52)
114
e
e
e
D.A. Day et al. / Atmospheric Environment 44 (2010) 1970e1979
1973

Page 5

Author's personal copy
was chosen since it reproduced the measured OM best among all
the solutions (with a slope of 1.05 and an r ¼ 0.93) and resulted in
factors with smoothly-varying FTIR spectra with spectral features
similar to those reported for known organic components of the
atmosphere. The five factors were divided into two groups of
correlated factors, and the factors in each group have similar
potential source regions. A two-factor solution was generated from
recombination of factors in each group, yielding two independent
factors that were identified as a Combustion Factor and a Marine
Factor, due to their correlation with inorganic components (sulfate
and nitrate) and potential source locations, as described below. We
interpret the identification of only two factors to mean that the
different types and locations of sources that influenced the organic
composition during the campaign were too similar to be fully
resolved with the limited number of samples and campaign
duration.
Compositions and correlations of these factors are summarized
in Table 2. The Combustion Factor was on average 1.4 mg m?3and
comprised 80% of PMF OM (on a sample-by-sample basis). The
Marine Factor was smaller, averaging 0.26 mg m?3and comprising
the other 20% of the PMF OM. The OM reconstructed from PMF
factors tracks the measured OM well and showed a large temporal
variability. Fig. 4 shows the time series of the two factors identified
as fractions of OM resolved by PMF. The Combustion Factor was
approximately half alkane (49%, 0.69 mg m?3) with carboxylic acid
comprising the other dominant fraction (27%, 0.38 mg m?3). Amine
and alcohol functional groups were minor fractions (7% each). The
Marine Factor was dominated by alcohol functional groups, which
comprised 81% (0.21 mg m?3) of the OM assigned to this factor, and
alkane and amine groups were 6% and 13%, respectively (cf. Table 2
and Fig. 4). The chemical composition of these factors are compa-
rable to other Combustion and Marine Factors that have been
ab
Fig. 3. a) 2-Ethyl-hexyl nitrate absorptivity calibration: FTIR peak area (absorbance cm?1) vs. organic nitrate functional group applied to filter. Calculated slopes provide absorptivity
used to calculate organic nitrate filter loading for ambient samples. b) Absorptivity calibration for smog chamber aerosol formed from OH/NOxreactions of 1-tetradecene or
n-pentadecane. Fitted slopes for 2-ethyl-hexyl nitrate calibration for 1630 (squares), 1280 (triangles), and 860 (circles) peaks is 13.3, 6.5, and 8.8 with Pearson's correlation
coefficients (r) of 0.85, 0.86, and 0.90, respectively. For 1-tetradecene or n-pentadecane products, slopes were 10.9, 4.1, and 9.8, with r values of 0.86, 0.83, and 0.97 (fits forced
through zero). In panel b, 1-tetradecene samples are denoted by a “þ” overlaid on the markers.
Fig. 4. (top panel) FTIR functional groups concentrations (w12-h samples), (middle panel) AMS concentrations averaged onto FTIR sample timestamp, and (bottom panel) FTIR PMF
factors as a fraction of total OM resolved by PMF. Pies graphs showing the chemical composition for the PMF factors are also shown in the bottom panel (see top panel for legend).
D.A. Day et al. / Atmospheric Environment 44 (2010) 1970e1979
1974