Introduction
Organic aerosol is ubiquitous in the atmosphere and has an important
influence on air quality (Finlayson-Pitts and Pitts, 2000; Seinfeld and
Pandis, 2006; R. Y. Zhang et al., 2015), climate (Kanakidou et al., 2005; IPCC,
2013), and human health (Mauderly and Chow, 2008; Shiraiwa et al., 2012).
Secondary organic aerosol (SOA) formed from the oxidation of volatile organic
compounds (VOCs) contributes a substantial fraction (up to 90 %) of
organic aerosol (Zhang et al., 2007). Biogenic VOCs (BVOCs) such as isoprene
(C5H8), monoterpenes (C10H16), and sesquiterpenes
(C15H24) account for ∼ 90 % of global VOC emissions
(Guenther et al., 1995; Goldstein and Galbally, 2007) and are the dominant
contributors to global SOA formation upon reaction with oxidants that are
mainly anthropogenically derived (Kanakidou et al., 2005; Hallquist et al.,
2009).
Sesquiterpenes are an important class of BVOCs, with emissions being
estimated as 9–29 % of those of monoterpenes (Helmig et al., 2007;
Sakulyanontvittaya et al., 2008a). A variety of sesquiterpenes have been
detected in the atmosphere, including β-caryophyllene, α-humulene, longifolene, α-farnesene, and α-cedrene (Helmig
et al., 2007; Sakulyanontvittaya et al., 2008a; Duhl et al., 2008;
Bouvier-Brown et al., 2009). Sesquiterpenes have a high SOA-forming potential
because of their large molecular sizes and, for many of them, the endocyclic
double bond structure, which favors the formation of low-volatility oxidation
products. The results of laboratory chamber studies show high aerosol mass
yields (defined as the mass of organic aerosol formed per mass of precursor
VOC reacted) from sesquiterpene oxidation (Hoffmann et al., 1997; Lee et al.,
2006; Ng et al., 2006; Winterhalter et al., 2009; Chen et al., 2012; Jaoui et
al., 2013; Yao et al., 2014; Tasoglou and Pandis, 2015). For example, an
average 53 % aerosol mass yield was reported for ozonolysis and 55 %
for photooxidation (Jaoui et al., 2013). Field and model studies have shown
that sesquiterpene SOA comprises a significant fraction (6–32 %) of
ambient organic aerosol from local to regional scales (Sakulyanontvittaya et
al., 2008b; Hu et al., 2008; Bouvier-Brown et al., 2009; Ding et al., 2014;
Ying et al., 2015), with its contribution comparable to that of monoterpene
SOA in a variety of environments including rural, suburban, and urban areas
(Hu et al., 2008; Ding et al., 2014).
Although sesquiterpene oxidation contributes substantially to SOA, the
underlying mechanisms of SOA formation and growth are not well understood.
Previous studies of sesquiterpene oxidation have mainly focused on aerosol
mass yields, the identities and formation mechanisms of first-, second-, and
higher-generation low molecular weight (LMW) oxidation products, and their
contributions to the formation of SOA (Hoffmann et al., 1997; Jaoui et al.,
2004, 2013; Lee et al., 2006; Ng et al., 2006; Kanawati et al., 2008; Reinnig
et al., 2009; Winterhalter et al., 2009; Li et al., 2011; Chan et al., 2011;
Chen et al., 2012; Alfarra et al., 2012; Yao et al., 2014). A few studies
have suggested that sesquiterpenes play an important role in new particle
formation. For instance, field observations show that in new particle
formation events, sesquiterpene oxidation products significantly contribute
to initial nucleation and growth (Boy et al., 2008; Bonn et al., 2008). In
addition, Bonn and Moortgat (2003) suggested that the formation of oligomers
was potentially important for nucleation. Lastly, laboratory studies of the
ozonolysis of β-caryophyllene and α-cedrene have reported a
negative influence on particle nucleation of species that can scavenge
stabilized Criegee intermediate (SCI), supporting a key role for
sesquiterpene SCI in forming new particles (Bonn and Moortgat, 2003; Yao et
al., 2014).
α-Cedrene (Fig. 1) is found in air (Duhl et al., 2008), reacts
rapidly with O3 (Richters et al., 2015), and is also an ideal compound
for the study of sesquiterpene oxidation because (1) the single C=C
bond in its structure helps to simplify the oxidation chemistry and the
product distribution and (2) its resemblance to other sesquiterpenes such as
β-caryophyllene and α-humulene in the endocyclic double bond
structure (with a methyl group at one end) may enable, to some degree, the
generalization of SOA formation mechanisms for this class of compounds.
Previous studies of α-cedrene oxidation identified a number of
products with molecular masses below 300 Da in both the gas and particle
phases and proposed reaction mechanisms based on known ozone chemistry
(Jaoui et al., 2004, 2013; Reinnig et al., 2009; Yao et al., 2014).
Preliminary results from our lab (Zhao et al., 2015) subsequently identified
higher molecular weight products in the α-cedrene ozonolysis for the
first time, and the SOA composition suggested that the major mechanisms for
particle formation are likely different than that for the small alkenes
(Sadezky et al., 2008; Zhao et al., 2015).
The structure of α-cedrene.
There have been very few measurements of the phase state of sesquiterpene
SOA. The phase state of SOA has an important influence on a number of
physical and chemical processes of aerosols, such as formation and growth
(Koop et al., 2011; Perraud et al., 2012; Shiraiwa and Seinfeld, 2012;
Renbaum-Wolff et al., 2013), chemical aging (Renbaum and Smith, 2009;
Shiraiwa et al., 2011; Lignell et al., 2014; Chan et al., 2014; Slade and
Knopf, 2014), and water uptake (Mikhailov et al., 2009; Koop et al., 2011;
Bones et al., 2012; Hodas et al., 2015; Pajunoja et al., 2015) and thus
affects their environmental and climate impacts. There is ample evidence that
in many cases, SOA may not be a low-viscosity liquid but rather a highly
viscous semisolid (Virtanen et al., 2010; Cappa and Wilson, 2011; Koop et
al., 2011; Vaden et al., 2011; Saukko et al., 2012; Perraud et al., 2012;
Abramson et al., 2013; Renbaum-Wolff et al., 2013; Kidd et al., 2014a, b;
Bateman et al., 2015). Recently, Saukko et al. (2012) and Pajunoja et
al. (2015) examined the phase state of SOA particles formed from OH and
O3 oxidation of longifolene based on particle bounce measurements and
found that longifolene SOA is solid or semisolid over a wide range of
relative humidities. To better understand the phase state of sesquiterpene
SOA and its implications for various atmospheric processes, more particle
phase state measurements are needed.
Summary of different types of flow reactor and chamber experiments.
Exp
[VOC]
[O3]
[HCOOH] or
RH
Reaction
Particle
Particle
Particle
typea
(ppb)b
(ppm)c
[2-EHN]
(%)
time
size
mass conc.
number conc.
(nm)d, e
(µg m-3)e
(106 cm-3)e
FR1 (2)
138
14
none
< 1
27 s
15.2 ± 0.2
26 ± 2
7.9 ± 0.2
FR2 (2)
138
14
none
< 1
44 s
22.5 ± 0.8
71 ± 8
7.0 ± 0.2
FR3 (2)
63
16
none
< 1
30 s
13.1 ± 0.4
10 ± 1
5.3 ± 0.3
FR4 (5)
138
16
none
< 1
30 s
17.6 ± 0.9
52 ± 3
9.4 ± 0.3
FR5 (2)
275
16
none
< 1
30 s
25.7 ± 1.1
180 ± 16
11.8 ± 0.6
FR6 (2)
138
16
none
75
30 s
16.1 ± 0.9
36 ± 10
9.3 ± 1.7
FR7 (2)
138
16
2 ppm HCOOH
< 1
30 s
22.1 ± 0.4
41 ± 2
2.7 ± 0.1
FR8 (2)
138
16
15 ppm HCOOH
< 1
30 s
22.0 ± 0.8
25 ± 1
1.5 ± 0.1
CH1 (4)
34
1.5
none
< 5
30 min
66 ± 2
73 ± 6
0.34 ± 0.02
CH2 (4)
34
1.5
none
72
30 min
64 ± 2
66 ± 4
0.34 ± 0.03
CH3 (3)
1000
0.12
none
< 5
30 min
111 ± 1
363 ± 37
0.37 ± 0.03
CH4 (3)
215
1.5
none
< 5
60 min
140 ± 3
984 ± 186
0.40 ± 0.02
CH5 (2)
215
1.5
400 ppb 2-EHN
< 5
60 min
127 ± 5
663 ± 44
0.35 ± 0.02
a FR and CH represent flow reactor and chamber experiments,
respectively. The numbers in parentheses after the labels represent the number of times an experiment was repeated.
b The concentrations were calculated from the amount of α-cedrene liquid injected into the flow reactor and chamber and the total
gas flow. c High O3 concentration is needed to get enough reaction to form
particles in less than 1 min in FR experiments.
d Geometric mean diameter; the size distributions of SOA formed in the
flow reactor are given in Figs. S1 and 11.
e All the data are given as average value ± one standard
deviation. SOA mass concentrations are obtained using a particle density
of 1.1 g cm-3 (Yao et al., 2014). The chamber SOA data are given for
30 or 60 min reaction times and were not corrected for the wall loss.
In the present study, we report the results of a more comprehensive
experimental investigation of ozonolysis of α-cedrene. The phase
state and mechanisms of growth of SOA are examined by probing the evaporation
of a tracer molecule, 2-ethylhexyl nitrate (2-EHN), incorporated into the SOA
during ozonolysis. The structures and formation mechanisms of high molecular
weight (HMW) products, as well as their roles in particle formation and
growth, are elucidated in light of their fragmentation mass spectra, accurate
mass data, size-dependent SOA composition, and the effects of water vapor and
SCI scavenger. The identity and formation mechanisms of some newly observed
LMW (MW < 300 Da) oxidation products are also explored.
Experimental
Experiments on α-cedrene ozonolysis were carried out both in a glass
flow reactor and in static Teflon chambers in the absence and presence of
water vapor or SCI scavengers at 295 ± 1 K. No seed particles or OH
scavengers were used in any of these experiments. Table 1 and Fig. S1 in the
Supplement summarize the conditions and particle characteristics for various
types of flow reactor and chamber experiments.
Flow reactor experiments
The flow reactor (4.6 cm i.d. and 85 cm long) used in this study has been
described in detail previously (Zhao et al., 2015). α-Cedrene was
added to the flow reactor by injecting the pure liquid (Sigma-Aldrich and
Extrasynthese, > 98 %) into a flow of 1.76–2.96 L min-1 of
clean, dry air (Praxair, ultra zero air) using an automated syringe pump
(Pump Systems Inc., model NE-1000). Ozone was generated by passing a flow of
O2 (Praxair, Ultra High Purity, 99.993 %) at 0.24 L min-1
through a pen-ray mercury lamp (model 11SC-2) and subsequently added to the
flow reactor downstream of the α-cedrene inlet. The O3
concentration, determined using a UV-VIS spectrometer (Ocean Optics, HR4000),
was adjusted by changing the UV exposure of the O2 via a movable metal
cover surrounding the lamp. The total flow rate in the reactor was 2.0, 2.9,
or 3.2 L min-1, corresponding to a residence time of 44, 30, or 27 s,
respectively. Some of the experiments were carried out in the presence of
water vapor or formic acid, both of which can act as SCI scavengers. Water
vapor was added by bubbling a flow of clean air through nanopure water
(18.2 MΩ cm) into the flow reactor. The relative humidity of the
airflow (∼ 75 % relative humidity (RH)) in the reactor was measured using a humidity
probe (Vaisala, HMT234). Formic acid (Sigma-Aldrich, ≥ 95 %) was
added to the reactor using the same method as for α-cedrene.
The particle size distributions were measured at the outlet of the reactor
using a scanning mobility particle sizer (SMPS, TSI), which is equipped with
an electrostatic classifier (model 3080), a long differential mobility
analyzer (model 3081), and a condensation particle counter (model 3776). When
the size distribution of SOA was stable, the polydisperse particles were
collected onto a 47 mm PTFE filter (Millipore Fluoropore, 0.45 µm
pore size) at a flow rate of 1.9, 2.8, or 3.1 L min-1, with venting of
the remaining 0.1 L min-1 aerosol flow to the hood. In order to obtain
sufficient particle mass for the analysis, the collection lasted 3–15 h,
depending on the particle mass concentrations. Excellent collection
efficiency (> 99 %) was obtained for particles of all diameters as
established by SMPS measurements downstream of the filter. The filter samples
were extracted immediately with a 2 mL mixture (5 : 3 in v / v) of
methanol (OmniSolv; liquid chromatography–mass spectrometry
(LC-MS) grade) and hexanes (a mixture of hexane isomers;
Fisher Scientific; 99.9 %) under ultrasonication in an ice bath for
30 min. The extracts were then purged gently using a flow of dry N2 at
room temperature to evaporate down to 1 mL. As hexanes are more volatile than
methanol, the remaining solvent in the extracts would be primarily methanol.
The resulting extracts were either analyzed directly using an LCT Premier
electrospray ionization time-of-flight mass spectrometer (ESI-ToF-MS, Waters)
or diluted with water (OmniSolv; LC-MS grade) to yield a final 50 : 50
water : methanol mixed solution, followed by analysis with a Xevo TQS
electrospray ionization triple quadrupole mass spectrometer (ESI-TQ-MS,
Waters). In order to verify that the sonication has no significant impact on
SOA measurements, filter extraction was performed instead, in some
experiments, in the mixture of methanol and hexanes by shaking for 30 min. No
significant difference in product distribution was observed in ESI mass
spectra of the SOA extracted with and without sonication. In separate
experiments, the chemical composition of α-cedrene SOA was also
measured in real time using direct analysis in real-time mass spectrometry
(DART-MS).
Chamber experiments
Chamber experiments were carried out in 450 L Teflon chambers. Two types of
α-cedrene ozonolysis experiments were conducted: (i) under dry or
humid (72 % RH) conditions and (ii) in the absence or presence of gas
phase 2-EHN, which was incorporated into SOA during
ozonolysis and used as a tracer molecule to probe the viscosity of SOA. The
α-cedrene was added using an automated syringe pump to inject a
defined volume of pure liquid into a flow of clean, dry air, which was
directed into the chamber. Water vapor was added by bubbling air through
nanopure water into the chamber. 2-EHN (Sigma-Aldrich, 97 %) was added by
injecting a known volume of liquid into the chamber. After 10 min to allow
for mixing of the gases (or 2 h for the evaporation of the 2-EHN),
ozonolysis was initiated by adding O3 generated by a commercial ozone
generator (Polymetrics, Model T-816) to the chamber.
For experiments CH1–CH3 (Table 1), SOA composition was examined online by a
high-resolution time-of-flight aerosol mass spectrometer (HR-ToF-AMS;
Aerodyne Research Inc.) or by collection onto a PTFE filter at a flow rate of
12 L min-1. Filter sampling started after 30 min reaction time and
lasted about 30 min. Because of the small volume of the chamber (450 L) and
relatively low SOA mass loading in most experiments (see Table 1), in each
experiment the SOA was sampled onto one filter to ensure enough mass for
ESI-MS analysis. The filter was extracted using the same method as for SOA
samples obtained in flow reactor experiments, followed by ESI-MS analysis.
For experiments CH4 and CH5, SOA formed without or with added 2-EHN
at a 60 min reaction time was collected onto a germanium (Ge) attenuated
total reflectance (ATR) crystal using a custom-designed impactor (Kidd et
al., 2014b) at a flow rate of 30 L min-1. The SOA impacted on the ATR
crystal was subsequently probed using attenuated total reflectance Fourier
transform infrared (ATR-FTIR) spectroscopy.
Most experiments were performed with excess O3. In order to evaluate the
influence of excess O3, which could lead to secondary oxidation of
first-generation products, some experiments were performed with excess
α-cedrene (Table 1, experiment CH3). The ESI mass spectra of SOA
formed in the presence of excess O3 or excess α-cedrene are very
similar, suggesting that the excess O3 has no observable influence on
SOA formation. In addition, no significant difference was observed in ESI
mass spectra of SOA formed with or without added 2-EHN, suggesting that the
incorporation of 2-EHN into the SOA does not change its overall composition.
SOA characterization
In both flow reactor and chamber experiments, aerosol samples were directed
through a 10 cm monolith extruded carbon denuder
(Novacarb™; Mast Carbon, Ltd.) to remove the
gas phase species prior to particle collection on the filter. No significant
difference in particle size distribution was observed with or without the
denuder. Blank experiments were also carried out under the same experimental
conditions as the ozonolysis experiments but without adding O3 to the
flow reactor or the chamber.
ATR-FTIR measurements
The ATR crystal with impacted SOA was placed immediately into an ATR cell
(volume ∼ 2 cm3) located in the sampling compartment of a Nicolet
6700 FTIR spectrometer (Thermo Scientific). A flow of dry synthetic air at
100 mL min-1 passed over the sample on the crystal. In some
experiments, this flow first passed through a pen-ray mercury lamp, producing
8 ppm O3, in order to probe the possibility of secondary oxidation of
SOA components by O3. Single-beam spectra at a resolution of
4 cm-1 (128 scans) were collected before and after impaction of SOA.
The absorbance spectra of SOA on the crystal were derived from
log10(S0/S1), where S0 and S1 represent the single-beam
spectra of the clean and SOA-covered crystal, respectively. The infrared (IR) spectra of
SOA were obtained over 20 h of air or O3 exposure to study the
evolution and evaporation of the impacted SOA during this period. In
experiments performed in the presence of 2-EHN, the signal at 1280 cm-1
was followed over time to investigate the evaporation of 2-EHN and determine
its diffusion coefficient throughout the SOA matrix.
ESI-MS measurements
Extracts of both SOA and blank samples were analyzed by ESI-ToF-MS operated
in either positive or negative ion mode. The operating conditions of this
mass spectrometer were described previously (Zhao et al., 2015). Mass spectra
were acquired in the 200–1000 Da mass range. Although ESI often forms
multiply charged ions, the major ions observed throughout the mass spectra
are in the singly charged state, as indicated by the unity spacing of the
isotope peaks (Greaves and Roboz, 2013). Sodium adducts [M + Na]+
were the primary ions observed in the positive ion mode (ESI+) and
deprotonated ions [M - H]- in the negative ion mode (ESI-). Mass
spectra of blanks, in which the peaks are mainly attributed to the impurities
in the solvent and filter and have intensities of 10–30 % of SOA peaks,
were subtracted from those of SOA samples. Accurate mass measurements were
performed using sodium formate, polyethylene glycol, and polyethylene glycol
monomethyl ether as mass calibration standards. In addition, SOA extracts
were analyzed using ESI-TQ-MS to record the fragmentation spectra (MS2)
of selected ions, from which typical structures of reasonable products based
on likely mechanisms were tentatively assigned. The parent ions selected by
the first quadrupole (Q1) enter the collision cell (Q2), where they fragment
via collision-induced dissociation (CID) with argon as the collision gas at
collision energies of 20–30 eV, and the resulting fragment ions are
monitored by the third quadrupole (Q3). In this case, the ESI source was
operated in the positive ion mode under optimized conditions as follows:
capillary voltage 3.2 kV; desolvation gas flow 1000 L h-1;
desolvation gas temperature 500 ∘C; nebulizer gas pressure 7 bar.
AMS measurements
An Aerodyne HR-ToF-AMS was used to analyze the chemical composition and to
examine O : C ratios of polydisperse SOA formed in the chamber. Due to the
small sizes of the SOA formed in the flow reactor, it was impossible to
measure them with the AMS, and thus all AMS results presented hereafter are
exclusively from the chamber studies. A detailed description of this
instrument has been given elsewhere (DeCarlo et al., 2006). The SOA sampled
into the instrument was vaporized at 600 ∘C. High-resolution MS data
were collected in both V mode and W mode and analyzed using the
“Improved-Ambient” method (Canagaratna et al., 2015). Measurements with a
particle filter were carried out before each experiment to aid in
quantification of the CHO+ fragment at m/z 29, which has interference
from gaseous 15NN and can significantly affect elemental analysis.
DART-MS measurements
Direct analysis in real-time mass spectrometry (DART-MS) is an atmospheric
pressure soft ionization method used to examine chemical composition of solid
and liquid samples after thermal desorption. A more detailed description of
this mass spectrometric technique is given elsewhere (Cody et al., 2005; Nah
et al., 2013). In this work, α-cedrene SOA formed in the flow reactor
under dry conditions was measured in real time using a Xevo TQS triple
quadrupole mass spectrometer (Waters) equipped with a commercial DART ion
source (IonSense, DART SVP with Vapur®
Interface). As DART-MS is a surface-sensitive technique, to ensure that the
bulk of particles can be effectively probed, the aerosol stream exiting the
flow reactor, in which the gas phase species were removed using a denuder,
was heated to 160 ∘C before entering into the ionization region. The
DART ion source was operated in either positive or negative ion modes with He
as the reagent gas under the following conditions: He gas flow
3.1 L min-1; He gas temperature 200 ∘C; grid electrode
voltage 350 V. The configuration of the DART ion source interfaced to the MS
is shown in Fig. S2.
Panel (a): a typical ATR-FTIR spectrum of SOA from
ozonolysis of α-cedrene in the chamber (experiment CH4, Table 1).
This spectrum is obtained from log10(S0/S1), where S0 is the
single-beam spectrum of the clean crystal and S1 is that of the
SOA-covered crystal recorded immediately following impaction. Panel
(b, c): typical difference spectra of SOA after 20 h of exposure to
a flow of clean air or 8 ppm O3-containing air, respectively. These
spectra are log10(S1/S20), where S1 is the single-beam
spectrum of SOA-covered crystal collected immediately following impaction and
S20 is that after 20 h of air or O3 exposure. The positive and
negative peaks in the difference spectra represent an increase and decrease,
respectively, of the functional groups in SOA over 20 h of exposure.
Results and discussion
ATR-FTIR measurements
Figure 2a is a typical ATR-FTIR spectrum recorded immediately following
impaction of α-cedrene SOA formed in the chamber in the absence of
2-EHN. The strong C=O band at 1706 cm-1 suggests that aldehydes and
ketones are important SOA components. A shoulder at 1762 cm-1 may
indicate the presence of carboxylic acids, esters, or other species
containing C=O groups, with a more electronegative atom such as
oxygen being attached to the carbonyl carbon (Socrates, 2001; Kidd et al.,
2014a).
Figure 2b is a difference spectrum showing the changes in SOA after exposure
to a flow of clean dry air for 20 h. The positive and negative peaks in the
spectrum represent an increase or decrease, respectively, in the functional
groups in SOA due to air exposure. There is a decrease in peaks at 3416,
1762, 1371, and 1076 cm-1 and an increase in peaks at 1735, 1706, 1410,
and in the 1100–1350 cm-1 region. Note that there is no obvious change
in the C-H peaks around 2957 cm-1, suggesting that evaporation of
organic species from SOA during air exposure is not important. Therefore, the
changes in SOA as indicated by the difference spectrum are due to chemistry
occurring in SOA during air exposure. One of the possible processes is the
decomposition of oligomers that comprise a significant fraction of α-cedrene SOA as discussed later. For example, decomposition of
peroxyhemiacetals and aldol condensation products to their precursors can
lead to the loss of O-H groups and concomitant formation of ketone and
aldehyde C=O groups. The decomposition of oligomers in SOA was supported by
ESI-MS measurements of SOA collected on Teflon filters and then exposed to a
flow of clean dry air; the relative ion intensity of oligomers to LMW
products in the mass spectrum of SOA extracted after 20 h of air exposure is
∼ 15 % lower compared to that of SOA extracted immediately
following collection. The difference spectrum of the SOA film exposed to
clean humid air at 89 % RH for 40 min (Fig. S3) shows a broad band
centered at 3426 cm-1 and a narrow band at 1640 cm-1 resulting
from the stretching and bending vibration of adsorbed water, respectively,
with relative strength of stretching vs. bending of ∼ 2. The absence of
a negative peak at 1640 cm-1 in the difference spectrum after dry air
exposure as shown in Fig. 2b, therefore, suggests that the contribution of
water to the loss in the O-H band at 3416 cm-1 is minor, and the
decrease in this region must be due to a change in SOA components.
Figure 2c shows the difference spectrum of SOA upon exposure to 8 ppm
O3 for 20 h, which is very similar to the difference spectrum of SOA
after 20 h of exposure to clean air (Fig. 2b). This shows that the α-cedrene SOA oxidation products are not reactive toward O3, at least as
detectable by changes in the infrared spectrum.
Viscosity and phase state of SOA
In order to probe the phase state of α-cedrene SOA, a relatively
volatile organic species, 2-EHN (Psat=1.4 × 10-4 atm at 295 K; Pankow and Asher, 2008), was
included in one type of experiment (CH5, Table 1) so that it could be
incorporated into the SOA as it is formed in the chamber. The evaporation of
2-EHN from the SOA impacted onto the ATR crystal upon exposure to a flow of
clean dry air was followed with time to probe the phase state of SOA.
Figure 3a shows a digital photograph of a typical impaction pattern on the
ATR crystal of α-cedrene SOA formed in the chamber in the presence of
gas phase 2-EHN. The SOA impacts and adheres to the crystal, forming a narrow
film ∼ 1.0 mm in width at the centerline. Similar impaction patterns
were also observed for the SOA formed without added 2-EHN. Figure 3b shows
the SOA number and mass size distributions after 60 min reaction time in the
chamber, as well as the collection efficiency of the impactor as a function
of particle diameter (Kidd et al., 2014b). By dividing the particles into a
number of 10 nm size bins and applying the average particle collection
efficiency at each bin, the SOA mass collected on the crystal is estimated to
be 42 µg using ∑mifi, where mi and fi are the
particle mass and average particle collection efficiency at each size bin,
respectively. Since the particle wall loss in the chamber during impaction
was not considered, this mass should be an upper limit. Assuming a particle
density of 1.1 g cm-3 (Yao et al., 2014), the maximum average
thickness of the SOA film on the ATR crystal is estimated to be L=0.54 µm. This estimate of L relies on the assumption that SOA
impacted on the crystal forms a uniform narrow film. This may underestimate
the film thickness by as much as a factor of 2 if, rather than forming a
thin film, the SOA is collected in separate columns immediately below the
impactor holes. An additional, smaller uncertainty in L results from
variation in the collection efficiency of the impactor over the particle size
range of interest (see Fig. 3b). We estimate the combination of these two to
give an uncertainty in L of about a factor of 2.
Panel (a): digital photograph of a typical impaction
pattern of SOA formed by ozonolysis of α-cedrene in the presence of
2-EHN in the chamber (experiment CH5, Table 1). The photograph shows a
1 cm × 1 cm section of the crystal. The impacted SOA forms a
narrow film ∼ 1 mm in width along the centerline of the crystal. Panel
(b): the number (blue triangles) and mass (red squares) weighted
size distribution of chamber α-cedrene SOA formed in the presence of
2-EHN. Also shown is the collection efficiency of the impactor as a function
of particle diameter (black circle) measured using carboxylate-modified latex
(CML) spheres in a previous study (Kidd et al., 2014b).
The refractive index of α-cedrene SOA is not known, but a value of
∼ 1.5 is reasonable based on literature values for SOA from other
biogenic organic oxidations (Lambe et al., 2013; Kim et al., 2014). The depth
of penetration of the IR beam at 1280 and 1635 cm-1, peaks which
correspond to the absorption bands of organic nitrate as discussed below, is
then calculated to be 0.52 and 0.41 µm (Harrick, 1967),
respectively, for SOA on a Ge crystal. This suggests that the entire depth of
the SOA film on the ATR crystal can be probed reasonably well by the IR beam
in the region of interest.
Figure 4a is a typical ATR-FTIR spectrum of α-cedrene SOA formed in
the presence of 2-EHN in the chamber under dry conditions. The spectrum is
essentially the same as that of SOA formed without 2-EHN (Fig. 2a) except
that there are two new bands at 1635 and 1280 cm-1 resulting from the
vibrations of the nitrate functional group (Socrates, 2001; Bruns et al.,
2010; Perraud et al., 2012) when 2-EHN is present. Figure 4b shows the
difference spectrum of SOA after exposure to a flow of clean dry air for
20 h. In addition to the spectral features similar to the difference
spectrum of 2-EHN-free SOA (Fig. 2b), there is a small loss of nitrate peaks
at 1635 and 1280 cm-1, indicating some evaporation of 2-EHN from SOA
during air exposure. Figure 5 shows the temporal evolution of the integrated
area of the nitrate peak at 1280 cm-1 over 20 h of air exposure. It
can be seen that after 20 h of air exposure, only ∼ 27 % of 2-EHN
evaporated from the SOA. This slow evaporation indicates that 2-EHN is
incorporated in the bulk of the SOA and that the SOA must be a high-viscosity
semisolid rather than a liquid where diffusion would be much more rapid
(Shiraiwa et al., 2011; Koop et al., 2011).
Panel (a): a typical ATR-FTIR spectrum of SOA from ozonolysis of α-cedrene in the presence of gas phase 2-EHN in the chamber (experiment CH5,
Table 1). This spectrum is obtained from log10(S0/S1), where
S0 and S1 are the single-beam spectra of the clean crystal and SOA-covered crystal recorded immediately following impaction. Panel (b): a typical
difference spectrum of SOA after 20 h of exposure to a flow of clean dry
air. This spectrum is log10(S1/S20), where S1 and
S20 are the single-beam spectra of SOA-covered crystal recorded
immediately following impaction and after 20 h of air exposure. The
positive and negative peaks in the spectra represent an increase and
decrease in the functional groups over 20 h of air exposure,
respectively. Panel (c): ATR-FTIR spectrum of SOA formed without 2-EHN following
exposure to a flow of dry air containing 20 ppm 2-EHN for 1 h. This
spectrum is derived from log10(S0/S1), where S0 and
S1 are the single-beam spectra of the clean crystal and SOA-covered
crystal exposed to 20 ppm 2-EHN for 1 h, respectively.
The incorporation of high-volatility 2-EHN into the bulk of high-viscosity
SOA is consistent with a condensation particle growth mechanism
(Finlayson-Pitts and Pitts, 2000; Seinfeld and Pandis, 2006; Riipinen et al.,
2011; Perraud et al., 2012), for which organic species condense kinetically
onto the surface of pre-existing particles and become buried and incorporated
into the bulk by incoming low-volatility organics. It is important to note
that in the ATR-FTIR spectrum (Fig. 4c) of α-cedrene SOA formed
without added 2-EHN in the chamber and then exposed to 20 ppm 2-EHN in the
ATR cell for 1 h, the absorption bands associated with organic nitrate
at 1635 and 1280 cm-1 are negligible. This observation suggests that
the uptake of 2-EHN into or onto previously formed SOA is not important.
Assuming that the evaporation of 2-EHN is determined only by its diffusion in
the film of SOA impacted on the ATR crystal and that the diffusion follows
Fick's Law, for a film with uniform 2-EHN concentration (C0) at t=0
and zero concentration at the surface for t>0, the total fraction (F)
of 2-EHN remaining in the film at time t can be expressed as (Crank, 1975;
Mehrer, 2007)
F=8π2∑j=0∞1(2j+1)2exp-(2j+1)2π2Dt4L2j=0,1,2,… .
In Eq. (1), D is the diffusion coefficient of 2-EHN in the SOA and L is
the thickness of the SOA film on the ATR crystal. A best fit of the
evaporation data of 2-EHN in Fig. 5 to Eq. (1) gives D/L2=5.6×10-5 min-1. With the SOA film thickness L of 0.54 µm
estimated above, the diffusion coefficient (D) of 2-EHN in SOA is
calculated to be 3 × 10-15 cm2 s-1, consistent
with the D values of the order of ∼ 10-15 cm2 s-1
predicted from the timescale for evaporation combined with the estimated
thickness of the SOA layer (Shiraiwa et al., 2011; Koop et al., 2011). The
uncertainty in L of about a factor of 2 described above translates into
an uncertainty in D of about a factor of 4. However, this value for D
is within the range of diffusion coefficients that would be expected in a
highly viscous semisolid matrix, i.e., D=10-10–10-20 cm2 s-1 (Shiraiwa et al., 2011; Koop et
al., 2011), confirming that α-cedrene SOA is a high-viscosity
semisolid. However, because of their small sizes, these particles do not
have enough momentum to bounce as they impact on the ATR crystal, as seen for
larger SOA from α-pinene oxidation in earlier studies (Kidd et al.,
2014a, b). Recently, Saukko et al. (2012) and Pajunoja et al. (2015) reported
the formation of semisolid SOA from the oxidation of the sesquiterpene
longifolene based on particle bounce measurements. However, to our knowledge,
the present work shows the first measurement of the diffusion coefficient of
an organic species in sesquiterpene SOA.
The normalized integrated area of -ONO2 peak at 1280 cm-1
as a function of time over 20 h of clean, dry air exposure.
The red line is a best fit (R2 = 0.976) of the evaporation data to
Eq. (1). Error bars represent ±1σ.
ESI mass spectra of SOA formed from ozonolysis of α-cedrene
in the flow reactor (experiment FR4, Table 1) in (a) positive and
(b) negative ion modes. Black and red labels indicate the previously reported
(Jaoui et al., 2004, 2013; Reinnig et al., 2009; Yao et al., 2014) and newly
observed products, respectively. P1 represents the low molecular weight
products, and P2, P3, and P4 denote higher molecular weight products formed
by combinations of two, three, or four of the P1 products, respectively.
SOA composition
Figure 6 shows typical ESI+ and ESI- mass spectra of SOA formed from
α-cedrene ozonolysis in the flow reactor at 30 s reaction time. Ions
in the mass range m/z 220–350 correspond to LMW products derived directly
from oxidation of α-cedrene and retaining much of the structure of
the parent compound (hereafter termed P1 products). Those with
m/z > 420 correspond to HMW products formed using two, three, or four
P1 products (hereafter termed P2, P3, and P4 products, respectively) as
building blocks. We try to avoid using the terms “dimer”, “trimer”, etc.
which may imply simple combinations of smaller species whereas P2, P3, etc.
are clearly complex combinations of different LMW products. As discussed
above, there is no evidence that any of these peaks arise from multiple
charging of higher molecular weight species. The data presented in Fig. 6
suggest that P1 and P2 products comprise a dominant fraction of α-cedrene SOA. Compared to the ESI+ spectrum, the corresponding ions in the
ESI- spectrum are generally 24 Da lower in mass. This is consistent with
the fact that the ions observed in the ESI+ mode are primarily sodium
adducts, while those in the ESI- mode are deprotonated ions with a
predominant contribution from carboxylic acids.
ESI+ mass spectra of polydisperse SOA with geometric mean diameter
of (a) 15 nm and (b) 23 nm formed from ozonolysis of
α-cedrene in the flow reactor at 27 and 44 s reaction times,
respectively, under dry conditions (experiments FR1 and FR2, Table 1). The
size distributions of SOA are shown in Fig. S1a. Note that these experiments
were conducted as a separate series from other experiments. Because ESI-MS
sensitivity changes over long periods (1 month or more), the intensity ratios
of oligomer to P1 products in the mass spectra shown here cannot be directly
compared with other mass spectra obtained in this work. However, those ratios
in the above mass spectra are comparable with each other as they were
acquired on the timescale of days, over which ESI-MS sensitivity is
essentially constant.
Figure 7 shows the ESI mass spectra of polydisperse α-cedrene SOA
(size distributions are given in Fig. S1a) with geometric mean diameters of
15 and 23 nm formed in the flow reactor at the same concentrations of
α-cedrene and O3 but different reaction times (27 and 44 s)
under dry conditions (Table 1, experiments FR1–FR2). Among P2 products, those
at m/z 481–543 are the most abundant ones in the flow reactor SOA and
they have a greater contribution to the smaller particles formed at shorter
reaction times (Fig. 7a). Conversely, P1 products contribute more to larger
particles formed at longer reaction times (Fig. 7b). Figure S4 shows
additional ESI mass spectra of polydisperse α-cedrene SOA particles
of different geometric mean diameters (13, 18, and 26 nm) formed in the flow
reactor at the same reaction time (30 s) but different concentrations of
α-cedrene under dry conditions (Table 1, experiments FR3–FR5). A
similar size-dependent distribution of P1 and P2 products is observed. These
results suggest that P2 products at m/z 481–543 may play an important
role in initial particle formation, while P1 products contribute
significantly to particle growth.
ESI mass spectra of SOA formed from ozonolysis of α-cedrene
in the chamber (experiment CH1, Table 1) in (a) positive and (b) negative
ion modes.
Figure 8 shows typical ESI+ and ESI- mass spectra of α-cedrene
SOA formed in the chamber. The chamber SOA has a geometric mean diameter of
66 nm, larger than those formed in the flow reactor (13–26 nm, depending
on the experimental conditions) due to the longer reaction times and hence
greater extents of reaction. Figure 8 shows that the chamber particles also
have relatively more P1 products, in agreement with the important role of P1
products in particle growth. In addition, the subset of P2 products with
m/z > 543 in the positive ion mode accounts for a greater fraction of
total P2 products in the chamber SOA than in the flow reactor SOA (Figs. 6a
and 7). This suggests that these larger P2 products (i.e., m/z > 543)
are mainly formed at longer reaction times in the chamber and contribute
mainly to particle growth.
It is known that P1 products may undergo ion–molecule reactions to form
noncovalently bound clusters in the ESI source (Müller et al., 2009; Gao
et al., 2010), and the formation of such clusters, if it occurs, is expected
to be positively correlated with the abundance of P1 products. However, as
shown in Figs. 7 and S4, the signal intensity ratio of P2 to P1 products is
inversely dependent on the size of SOA. This suggests that the P2 products
observed in ESI mass spectra are unlikely to be artifacts of in-source
clustering. In addition, HMW products in α-cedrene SOA are also
detected in real time by DART-MS, which employs a different ionization
method. Figure S5 shows typical DART+ (positive ion mode) and DART-
(negative ion mode) mass spectra of flow reactor α-cedrene SOA, which
was heated to 160 ∘C before introduction into the ionization region.
The masses of P1 and P2 products in the DART+ mass spectrum are generally
22–24 Da lower than those in the ESI+ mass spectrum (Fig. 6), consistent
with the ions detected by DART-MS in the positive mode being primarily
[M + H]+, M+, and [M - H]+ (Cody et al., 2005; Nah et
al., 2013). In contrast, essentially the same masses for P1 and P2 products
were observed in DART- and ESI- mass spectra, in both of which the ions
are predominantly [M - H]-. There are some differences between ESI
and DART mass spectra that are likely due to the heating of the incoming
aerosol stream for DART-MS and the higher DART source temperature. Overall,
DART-MS measurements further confirm that the P2 products observed by ESI-MS
are not artifacts from the ion source.
Previous studies have reported an OH yield of 62–67 % from ozonolysis of
α-cedrene (Shu and Atkinson, 1994; Yao et al., 2014). Therefore,
reaction with OH could play a role in α-cedrene oxidation in the
absence of an OH scavenger. We reported in a previous study (Zhao et al.,
2015) ESI-MS spectra in the absence and presence of cyclohexane, which showed
that the relative intensity of P2–P4 peaks in the presence of cyclohexane is
smaller. This suggests that OH oxidation may contribute to the formation of
HMW species.
Possible structures and formation mechanisms of typical P1 and P2 products
(as labeled in Fig. 6) were explored based on their fragmentation mass
spectra (MS2), accurate mass data, previously identified P1 products
(Jaoui et al., 2004, 2013; Reinnig et al., 2009; Yao et al., 2014), and
likely reaction mechanisms of terpene ozonolysis. To the best of our
knowledge, this is the first detailed examination of HMW products initially
reported (Zhao et al., 2015) in this system.
Accurate mass and elemental formulae for P1 and P2 products formed
from α-cedrene ozonolysis measured by ESI-ToF-MS.
Producta
ESI+ mode ([M + Na]+)
ESI- mode ([M - H]-)
Observed
Elemental
Calculated
Absolute
Relative
Observed
Elemental
Calculated
Absolute
Relative
accurate
formula
exact
mass error
mass error
accurate
formula
exact mass
mass error
mass error
mass (Da)
mass (Da)
(mDa)
(ppm)
mass (Da)
(Da)
(mDa)
(ppm)
P1-245
245.1525
C14H22O2Na
245.1517
0.8
3.3
–b
P1-247
247.1326
C13H20O3Na
247.1310
1.6
6.4
223.1341
C13H19O3
223.1334
0.7
3.1
P1-259
259.1681
C15H24O2Na
259.1674
0.7
2.7
–b
P1-261
–b
237.1494
C14H21O3
237.1491
0.3
1.3
P1-275
275.1628
C15H24O3Na
275.1623
0.5
1.8
251.1646
C15H23O3
251.1647
-0.1
-0.4
P1-277
277.1422
C14H22O4Na
277.1416
0.6
2.2
253.1447
C14H21O4
253.1440
0.7
2.8
P1-289
289.1431
C15H22O4Na
289.1416
1.5
5.2
265.1476
C15H21O4
265.1440
3.6
13.6d
P1-291
291.1577
C15H24O4Na
291.1572
0.5
1.6
267.1607
C15H23O4
267.1596
1.1
4.1
P1-305
305.1370
C15H22O5Na
305.1365
0.5
1.7
–b
P1-307
307.1516
C15H24O5Na
307.1521
-0.5
-1.8
283.1554
C15H23O5
283.1545
0.9
3.2
P1-321
321.1308
C15H22O6Na
321.1314
-0.6
-1.9
297.1345
C15H21O6
297.1338
0.7
2.4
P1-323
323.1465
C15H24O6Na
323.1471
-0.6
-1.9
299.1492
C15H23O6
299.1495
-0.3
-1.0
P2-481
481.3298
C29H46O4Na
481.3294
0.4
0.8
–b
P2-497
497.3262
C29H46O5Na
497.3243
1.9
3.8
473.3271
C29H45O5
473.3267
0.4
0.9
P2-511c
511.3337
C30H48O5Na
511.3399
-6.2
-12.1d
487.3086
C29H43O6
487.3060
2.6
5.3
P2-513
513.3203
C29H46O6Na
513.3192
1.1
2.1
489.3215
C29H45O6
489.3216
-0.1
-0.2
P2-527
527.3317
C30H48O6Na
527.3348
-3.1
-5.8
503.3387
C30H47O6
503.3372
1.5
3.0
P2-543
543.3309
C30H48O7Na
543.3298
1.1
2.0
519.3320
C30H47O7
519.3322
-0.2
-0.4
a Labels “P1-xxx” and “P2-xxx” denote P1 and P2 products having a
nominal mass of xxx for their sodiated ions, respectively.
b Accurate mass was not measured because of the very low ion
intensity or the strong interference from other ions such as impurities.
c Different formulae for P2-511 were identified in ESI+ and ESI-
modes, which correspond, respectively, to aldol condensation products
(P2-511-3 and P2-511-4) and esters (P2-511-1 and P2-511-2) as shown in Fig. 10c.
d The relatively large mass errors likely result from unknown
interferences for P1-289 in ESI- mode and P2-511 in ESI+ mode. The given
formulae are the closest to the observed masses but may not be correct
because of the interferences.
P1 products
The characterization of the molecular structures and formation mechanisms of
P1 products from the ozonolysis of α-cedrene has been the emphasis of
earlier studies (Jaoui et al., 2004, 2013; Reinnig et al., 2009; Yao et al.,
2014). A number of multifunctional P1 products were tentatively identified
using different mass spectrometric and derivatization techniques in those
studies. Figure S6 shows the MS2 spectra (ESI+) as well as the proposed
structures of the products at m/z 245–291 measured in the current
experiments. As reported by Tolocka et al. (2004), fragmentation of sodium
adducts is difficult, so the product ion intensities are relatively weak.
However, fragmentation is seen to occur mainly via loss of 18 Da (H2O,
indicating the presence of an aldehyde, carboxylic, or hydroxyl group), loss
of 30 Da (CH2O, indicating the presence of a formyl group), loss of
32 Da (CH4O, indicating the presence of a hydroxymethyl group), loss of
44 Da (CO2 or C2H4O, indicating a carboxylic or acetyl
group), and loss of 46 Da (CH2O2, indicating a carboxylic group).
The fragmentation pathways of the functional groups as shown here generally
agree with those reported in the literature (Tolocka et al., 2004; Hall and
Johnston, 2012), where ESI-MS in positive ion mode was used to characterize
the structure of reaction products from ozonolysis of α-pinene. The
structures assigned to the products at m/z 245–291 are consistent with
those reported in the previous studies (Jaoui et al., 2004, 2013; Reinnig et
al., 2009; Yao et al., 2014). In many cases, the MS2 spectra can be due
to multiple isomeric structures. Table 2 shows accurate mass data for these
P1 products. Their measured accurate masses are for the most part in
excellent agreement with the elemental formulae for their proposed
structures, the exception being P1-289 in the negative ion mode.
The proposed potential structures of the newly observed products
corresponding to the [M + Na]+ ions of m/z 305, 307, 321, and 323.
Label “P1-xxx” represents a P1 product, the sodiated ion of which has a
nominal mass of xxx. Labels “P1-xxx-1” and “P1-xxx-2” indicate different
isomeric structures for the product “P1-xxx”. MW: molecular weight. Note
that there may be additional isomeric structures for these products.
In addition to the products reported previously (Jaoui et al., 2004, 2013;
Reinnig et al., 2009; Yao et al., 2014), some products corresponding to the
[M + Na]+ at m/z 305–323 not previously identified were detected
by ESI-MS. Figure S7 shows the parent ion spectra of these sodiated ions at a
collision gas energy (CE) of 6 eV that is equivalent to the CE for the MS
scan. These spectra show negligible contributions from higher molecular
weight ions at low CE. This strongly suggests that m/z 305–323 ions are
molecular ions rather than fragments from larger products. However, as shown
in the insets of Fig. S7, significant parents are observed for these ions
when the CE is increased to 20 eV. This is reasonable since these P1
products can be precursors to HMW products, which are expected to fragment
readily back to the precursors at higher CEs. In addition, it is likely that
the fragmentation of HMW products formed from other P1 products gives
fragments with the same m/z values of 305–323. These P1 products were also
observed by DART-MS in the m/z 281–301 region primarily as
[M + H]+ with some contribution from [M]+ in the positive ion
mode or [M - H]- ions in the negative ion mode (Fig. S5), consistent
with DART-MS studies of single compounds (Nah et al., 2013).
Figure 9 shows the potential structures of these newly observed P1 products
proposed based on their ESI-MS2 spectra (Fig. S8). The fragments
resulting from the loss of water and multiple oxygenated functional groups
such as carboxylic, carbonyl, and hydroxyl groups were observed. It was found
in our previous study that the formation of the products corresponding to
m/z 305–323 ions continues in the presence of cyclohexane as an OH
scavenger (Zhao et al., 2015). This suggests that the newly observed products
are not formed by OH oxidation but rather by O3 oxidation. Figure S9
illustrates some potential formation mechanisms for these newly observed
products. The Criegee intermediates (CIs) form peroxy (RO2) radicals via
the vinyl hydroperoxide (VHP) channel. The conversion of RO2 to alkoxy
(RO) radicals and the subsequent intramolecular H abstraction and O2
addition lead again to RO2 radicals, which can further undergo similar
reactions to form RO2 radicals with higher oxygen content. Termination
reactions with RO2 or HO2 radicals lead to the formation of the
newly observed products.
While accurate mass data for these products (Table 2) agree well with the
proposed structures, there may be additional structures and reaction pathways
in the formation of the P1 products at m/z 305–323 via the VHP channel of
CIs that contribute to these peaks. For example, because of their large
carbon skeleton, RO radicals may have multiple isomerization pathways to form
different structures with the same elemental composition.
Secondary ozonides formed through intramolecular reactions of SCI were
observed as the major gas phase products from ozonolysis of α-cedrene
(Yao et al., 2014) and β-caryophyllene (Winterhalter et al., 2009),
with their formation being significantly suppressed by the addition of water
vapor. In the present study, although a sodiated ion with m/z 275 and
elemental composition of C15H24O3Na, consistent with the mass
and formula of the intramolecularly formed secondary ozonide (SOZ), was
observed in ESI+ mass spectra of α-cedrene SOA, as will be
discussed in Sect. 3.4, the relative intensity of this ion in the mass
spectra does not decrease at all at 75 % RH and is still pronounced with
high concentrations of formic acid (15 ppm) added as an SCI scavenger. This
indicates that the ion at m/z 275 is unlikely to be the SOZ. Similarly,
Yao et al. (2014) did not observe the SOZ in the particle phase using
high-performance liquid chromatography–mass spectrometry (HPLC-MS) and gas
chromatography–mass spectrometry (GC-MS). A possible explanation is that the
intramolecularly formed SOZ has a relatively high vapor pressure
(1.2 × 10-6 atm at 295 K) and therefore a low potential to
partition to the particle phase.
The potential structures for the P2 products corresponding to the
[M + Na]+ ions of (a) m/z 481, (b) m/z 497, (c) m/z 511, (d) m/z 513,
(e) m/z 527, and (f) m/z 543 formed from α-cedrene ozonolysis. Label
“P2-xxx” represents a P2 product, the sodiated ion of which has a nominal
mass of xxx. Label “P2-xxx-n” (n=1–4) indicates different potential
structures for the product “P2-xxx”. Note that there may be additional
isomeric structures for these products.
P2 products
Table 2 gives the accurate masses and elemental formulae of the most abundant
P2 products as labeled in the ESI mass spectra (Fig. 6a). Figure 10 shows
proposed potential structures of these products based on their accurate mass
data and ESI-MS2 spectra (Fig. S10) (see Supplement for the detailed
discussion). These structures include aldol condensation products (formed
from the reaction of two carbonyl compounds; Reaction R1), peroxyhemiacetals
(formed from the reaction of a carbonyl compound and an organic
hydroperoxide; Reaction R2), and esters (formed from the reaction of a
carboxylic acid with an alcohol; Reaction R3), with the building blocks being
the P1 products typically observed in the SOA.
Aldol condensation products and peroxyhemiacetals are the most commonly
identified structures. Of 17 proposed structures for these P2 products, 10
are aldol condensation products and 5 are peroxyhemiacetals. Except for
products P2-481 and P2-497, which are identified as aldol condensation
products, other P2 products may have contributions from multiple structures.
For example, product P2-511 may have contributions from both aldol
condensation products and esters, and products P2-513, P2-527, and P2-543 may
have contributions from both aldol condensation products and
peroxyhemiacetals. Although no detailed data concerning the formation of such
HMW products during sesquiterpene ozonolysis are available in the literature,
a number of laboratory and field studies have found that aldol condensation
products, peroxyhemiacetals, and esters are the major HMW products detected
in the SOA formed from ozonolysis of monoterpenes (e.g., α-pinene and
β-pinene) (Hoffmann et al., 1998; Tolocka et al., 2004; Docherty et
al., 2005; Müller et al., 2008, 2009; Heaton et al., 2009; Gao et al.,
2010; Yasmeen et al., 2010; Hall and Johnston, 2012; Kristensen et al., 2013,
2014; Witkowski and Gierczak, 2014; X. Zhang et al., 2015). Formation of such
products has traditionally been thought to occur via acid-catalyzed reactions
(Reactions R1–R3) in the condensed phase (Kroll and Seinfeld, 2008;
Hallquist et al., 2009; Yasmeen et al., 2010; Ziemann and Atkinson, 2012).
However, some recent studies suggested that such a process is not favored for
esterification in SOA (Birdsall et al., 2013; DePalma et al., 2013;
Kristensen et al., 2014; X. Zhang et al., 2015) and that gas phase formation
followed by rapid gas to particle conversion and condensed phase
rearrangement of labile HMW diacyl peroxides may potentially explain the
formation of esters observed in SOA (X. Zhang et al., 2015). Formation of
peroxyhemiacetals has also been suggested to be thermodynamically favorable
in the gas phase (DePalma et al., 2013), although whether the kinetics are
sufficiently fast is not known. It is noted that there may be some
decomposition of HMW products over time, for example, over 20 h of exposure
to clean air as discussed earlier.
The effects of water vapor and SCI scavenger
The reaction of α-cedrene with ozone has a very high SCI yield
(> 88 %) (Yao et al., 2014). If SCI reactions are key to SOA formation
in this system, the addition of SCI scavengers such as water vapor and formic
acid should have significant influence on the formation of SOA.
Figure 11 shows the size distributions of SOA formed from ozonolysis of
α-cedrene in the flow reactor under dry conditions, at 75 % RH,
or in the presence of 15 ppm formic acid. SOA with very similar number
concentrations is formed with or without added water vapor, while
significantly fewer, but larger, particles are formed in the presence of
formic acid. Recent kinetics measurements of the CH2OO CI reaction with
water vapor (monomer and dimer) give rate constants of
kdimer = (4.0–6.5) × 10-12 cm3 molecule-1 s-1
(Chao et al., 2015; Lewis et al., 2015) and
kmonomer = 3.2 × 10-16 cm3 molecule-1 s-1
(Berndt et al., 2015). The rate constant for the reaction of CH2OO with
formic acid is much larger,
kHCOOH = 1 × 10-10 cm3 molecule-1 s-1
(Welz et al., 2014). The reactivity of Criegee intermediates has been shown
to depend on structure for a number of reasons, including different extents
of pressure stabilization of excited CI, different reactivities for the
syn- and anti-forms, and different natures of the groups
attached to the central CI carbon (Ryzhkov and Ariya, 2004; Anglada et al.,
2011; Donahue et al., 2011; Vereecken et al., 2012; Taatjes et al., 2013;
Welz et al., 2014). While the α-cedrene CI is clearly very different
than CH2OO, if the rate constants for the latter are representative in a
relative sense, the ratio of the rates for reaction of the α-cedrene
CI with formic acid and water dimer relative to water monomer are
∼ 240 : 18 : 1 under our experimental conditions
(3.7 × 1014 HCOOH cm-3,
4.8 × 1017 H2O cm-3 and
5.1 × 1014 (H2O)2 cm-3, based on a water
equilibrium constant of 0.0536 atm-1; Ruscic, 2013). Thus, it is not
surprising that formic acid has a much greater impact than water.
Furthermore, the large effects of formic acid further support the importance
of the α-cedrene CI in SOA formation. These impacts of water and
formic acid are in agreement with those reported by Yao et al. (2014), who
reported a similar quenching effect of added acetic acid on particle
formation in the α-cedrene ozonolysis but relatively little impact
of water vapor.
The size distributions of SOA formed from ozonolysis of α-cedrene in the flow reactor under dry conditions (purple squares,
experiment FR4), at 75 % RH (green triangles, experiment FR6), or in the
presence of 15 ppm formic acid (red diamonds, experiment FR8). Error bars represent ±1σ.
ESI+ mass spectra of polydisperse SOA formed from ozonolysis
of α-cedrene in the flow reactor (a) under dry conditions
(experiment FR4), (b) at 75 % RH (experiment FR6), and (c) with 15 ppm
formic acid (experiment FR8).
Figure 12a and b show the ESI mass spectra of α-cedrene SOA formed in
the flow reactor without and with added water vapor, respectively. There are
no significant differences in product distribution with and without added
water vapor. In addition, ESI-MS2 measurements show that typical P1 and
P2 products as labeled in both mass spectra have very similar MS2
spectra, suggesting that the product composition is not impacted by the
presence of water. Measurements of the chamber SOA by AMS (see Fig. S11) also
did not show significant differences in the particle mass spectra upon water
addition. On the other hand, Fig. 12c shows the ESI mass spectra of SOA
formed in the presence of 15 ppm formic acid. The formation of HMW products
is significantly reduced as are the number of peaks in the P1 product region.
However, P1 products at m/z 259, 275, and 291, as well as P2 products, for
example, at m/z 527, 543, and 559, are still evident. All of these P1
products have very similar MS2 spectra (and thus likely similar
structures) to those formed in the absence of formic acid. In contrast, P2
products at m/z 527 and 543 show significantly different MS2 spectra
compared to those obtained without added formic acid.
Yao et al. (2014) reported an increase in the formation of α-cedronaldehyde in the presence of acetic acid, consistent with an increase
in the present study in the relative abundance of the P1 product at
m/z 259 assigned to α-cedronaldehyde when formic acid was added
(Fig. 12c). The reaction of α-cedrene SCI with formic acid is
expected to form α-formyloxy hydroperoxide (MW 298 Da), which is
believed to contribute to SOA growth. While the corresponding sodium adduct
at m/z 321 is not observed in the mass spectrum, this is not surprising as
hydroperoxides are likely to undergo decomposition during SOA sampling and
analysis (Witkowski and Gierczak, 2013).
Comparison of Fig. 12c to a shows that HMW products (P2–P4) formed with
added formic acid are less important than those produced in the absence of
formic acid. Combined with the smaller number concentration (Fig. 11), the
data suggest that HMW products must be important in new particle formation.
As fewer particles are formed with the addition of formic acid, more P1
products are available for each particle to grow to larger sizes.
Mechanisms
The size-dependent composition of α-cedrene SOA and the effect of SCI
scavengers on particle formation suggest that HMW products (P2–P4) play an
important role in the initial stages of particle formation. This is
consistent with earlier suggestions in simpler systems that higher molecular
weight products of alkene ozonolysis are primarily responsible for
nucleation, while many different products from low to high molecular weight
can contribute to growth (Lee and Kamens, 2005; Sadezky et al., 2008; Winkler
et al., 2012; Zhao et al., 2013; Ehn et al., 2014; Kidd et al., 2014a; Zhao
et al., 2015; X. Zhang et al., 2015). Bonn and Moortgat (2003) estimated an
upper limit for the saturated vapor pressure (Psat) of nucleating
species produced from sesquiterpene ozonolysis to be
1.2 × 10-13 atm. Donahue et al. (2013) also suggested that
organic vapors with Psat < 10-13 atm may contribute to
particle nucleation. The Psat of typical P1 products, as well as
the P2 products (e.g., ester and peroxyhemiacetal) are estimated by averaging
the predictions (see Table S1 in the Supplement) from two group contribution
methods – SIMPOL.1 (Pankow and Asher, 2008) and EVAPORATION (Compernolle et
al., 2011) – and are shown in Fig. 13. The Psat for the P2 products are
3–6 orders of magnitude lower than the suggested nucleation threshold,
consistent with their important contributions to initial particle formation.
The newly observed products (P1-321 and P1-323) also have vapor pressures
lower than the nucleation threshold and thus may in principle also contribute
to nucleation. This is also expected to be the case for the HMW P3 and P4
products. However, without information on their chemical composition and
structures, data on P3 and P4 could not be included in Fig. 13. In contrast,
the smaller, more volatile P1 products will mainly contribute to particle
growth, as suggested by their relatively larger abundance in ESI mass spectra
of larger particles (Figs. 7 and S4).
Saturated vapor pressures of typical P1 and P2 products in α-cedrene SOA estimated by taking the average of the predictions from
SIMPOL.1 and EVAPORATION. The dashed line indicates the upper limit of
saturated organic vapor pressure for nucleation (Bonn and Moortgat, 2003; Donahue et al., 2013).
Compared to P2 products at m/z 481–543, those with m/z > 543 are
significantly less important in the flow reactor SOA (geometric mean diameter
13–26 nm) but contribute a greater fraction to total P2 products in the
chamber SOA (geometric mean diameter 66 nm). This suggests that these larger
P2 products (i.e., m/z > 543) play an important role in particle growth
and may be formed mainly via condensed phase reactions of P1 products in the
SOA.
Elemental analysis of SOA from chamber experiments using AMS results in an
average O : C ratio of 0.34 ± 0.03 (1σ) and an H : C ratio
of 1.51 ± 0.02. These values are within the range given for P1 products
detected by ESI-MS (Table 2) and are consistent with our observation that P1
products predominantly contribute to particle growth. Along with the
potential mechanisms discussed above to explain the formation of observed P1
and P2 products and their contribution to particle formation and growth,
another mechanism considered for particle formation and growth is the
production of extremely low-volatility organic compounds (ELVOCs), which was
proposed to occur via sequential intramolecular hydrogen abstraction and O2
addition of RO2 radicals (Vereecken et al., 2007; Crounse et al., 2013;
Ehn et al., 2014; Rissanen et al., 2015) and shown to play an important role
in particle formation and growth from monoterpene oxidation (Zhao et al.,
2013; Ehn et al., 2014; Jokinen et al., 2015). Because of the high oxygen
content in ELVOCs, particles formed from these compounds are expected to have
high O : C ratios. The ELVOC mechanism cannot be ruled out as being
involved in initial particle formation during α-cedrene ozonolysis.
However, based on the O : C ratios measured for chamber SOA, it is
unlikely to be a dominant contributor to total particle mass in this system.
Measurements by AMS are not possible for the flow reactor experiments as the
particles are too small to be efficiently transmitted into the AMS.
It is clear from the impact of formic acid on particle formation that the SCI
plays a key role in forming the HMW products and new particles. In the case
of the trans-3-hexene ozonolysis (Sadezky et al., 2008; Zhao et al.,
2015), the composition of the SOA clearly showed evidence for oligomer
formation from the sequential reaction of RO2 radicals with SCI, leading
to ESI mass spectra that showed the repeat unit of oligomers corresponding to
SCI. A search for similar products in the α-cedrene reaction was not
successful, indicating that while this may contribute, other mechanisms are
more important (Zhao et al., 2015). This is not surprising, given the number
of potential reaction paths for the SCI from α-cedrene (e.g., Fig. S9)
and the large number of low-volatility products that can quickly contribute
to growth and the SOA composition once nucleation has occurred.
The incorporation of 2-EHN into SOA and its very slow evaporation back out
(Figs. 4 and 5) are consistent with a condensation type of growth mechanism
(Finlayson-Pitts and Pitts, 2000; Seinfeld and Pandis, 2006) in which organic
species are irreversibly taken up by the particle surface and thus their
incorporation into particles depends on the collision frequency of the gas
with the particle surface, and the magnitude of the uptake coefficient
(Perraud et al., 2012). Such a growth mechanism is characteristic of highly
viscous SOA, in which incorporated organic species undergo very slow
evaporation because of the diffusion limitation, in contrast to an
equilibrium mechanism that applies for low-viscosity liquid particles. This
result also suggests that growth will occur not only via low-volatility
products but also via uptake of higher-volatility species such as the smaller
P1 products identified here.
Summary
The present study examines the phase state, composition, and mechanisms of
formation and growth of SOA from ozonolysis of α-cedrene. The SOA is
characterized to be a high-viscosity semisolid, with an estimated diffusion
coefficient of 3 × 10-15 cm2 s-1 for 2-EHN which
was incorporated into the SOA during ozonolysis. High molecular weight
products, tentatively identified mainly as aldol condensation products,
peroxyhemiacetals, and esters, comprise a significant fraction of SOA. The
size-dependent distribution of these products in the SOA as well as their
positive correlation with new particle formation suggests that at least some
of them are responsible for initial particle formation. In contrast, lower
molecular weight (P1) products mainly contribute to particle growth via a
kinetic condensation mechanism.
Bonn and Moortgat (2003) have suggested that sesquiterpene ozonolysis could
be one source of new particle formation in the boreal forest. Evidence for
the role of sesquiterpenes in atmospheric new particle formation has been
recently presented by field observations in the boreal forest in Finland
(Bonn et al., 2008) and in the Front Range of the Colorado Rocky Mountains
(Boy et al., 2008). These studies have proposed possible nucleation
mechanisms involving low-volatility products such as intermolecularly formed
SOZ and organosulfates from sesquiterpene oxidation. The results of the
present study suggest that the formation of HMW products during ozonolysis
may serve as an important mechanism for such new particle formation.
Mechanisms of ozonolysis of alkenes and, in particular, the pathways that lead
to SOA formation are highly dependent on the size and structure of the parent
alkene. However, in all cases stabilized Criegee intermediates play a key
role. For example, SOA generated from ozonolysis of small alkenes such as
trans-3-hexene is primarily composed of oligomers formed from the
sequential addition of SCI to RO2 radicals (Sadezky et al., 2008; Zhao
et al., 2015). For larger alkenes, such as α-cedrene, the SCIs react
via multiple pathways, leading to a variety of low-volatility products (e.g.,
P2 and newly observed P1 products) that can nucleate on their own to form new
particles. There is yet a third group of alkenes with intermediate molecular
sizes such as monoterpenes, for which the major first-generation ozonolysis
products do not have low enough volatilities to nucleate on their own
(Hallquist et al., 2009), and therefore the ELVOC mechanism, despite the small
yields (a few percent, Jokinen et al., 2015), may play a more dominant role
(Zhao et al., 2013; Ehn et al., 2014). The dependence of phase, composition,
and mechanisms for particle formation and growth should be taken into account
in atmospheric models of the formation and impacts of SOA on visibility,
human health, and climate.