Modeled vs. measured NO3 and O3 exposures
One of the features of the OFR technique is the short residence time
required for conducting high-time-resolution ambient measurements. Combined
with the ability to rapidly change the amount of oxidant injected or
produced in the OFR, this allows for a wide range of oxidation levels to be
studied in a short amount of time (and thus with limited variation of
ambient conditions). In this work, the oxidant concentration was changed
every 20–30 min, covering a range from no added oxidant to maximum
oxidation repeatedly in 2–3 h cycles. In order to interpret the results
over the wide range of oxidant exposure, the amount of exposure must be
quantified. In Palm et al. (2016), OH exposure
was estimated using a model-derived equation
(Li et
al., 2015; Peng et al., 2015) and calibrated using PTR-TOF-MS measurements
of VOC decay in the OFR. In this work, a simple box model was developed and
compared with VOC decay measurements to estimate NO3 and O3
exposures in the OFR.
Modeled vs. measured (a) N2O5, (b) NO2, (c) NO3,
and (d) fraction of ambient MT reacted with NO3 in the output of the
NO3-OFR.
Modeled vs. measured fraction MT reacted by O3 oxidation in the
O3-OFR.
The set of reactions and rate constant parameters included in the modeling
of NO3 exposure are shown in Table S1. Figure 1 illustrates the most
important mixing ratios and reactive fluxes in the OFR with injected
N2O5 under typical conditions. Interconversion between
N2O5 and NO2+NO3 was relatively rapid, which
maintained the system near equilibrium at all times. Wall loss of
N2O5 was estimated to be the main loss of the injected
nitrogen-containing species (84 %), while reaction of NO3 with
biogenic gases (2 %), NO3 wall losses (14 %), and hydrolysis of
N2O5 on particle surfaces (0.2 %) were minor loss pathways.
Figure 2a–c compare the N2O5, NO2, and NO3 mixing
ratios measured in the OFR output with those predicted by the model. The
model is generally consistent with the measurements. The scatter in the
measurements is thought to be due mainly to incomplete and/or variable
mixing of the injected N2O5 flow into the sampled ambient air (see
Sect. S1 for more details), with some contribution from measurement
variability at low ambient MT concentrations. The critical output of this
model for our application is the prediction of the fraction of MT reacted.
Figure 2d shows that the model can reproduce the measured MT decay with an
error (average absolute value of modeled minus measured fraction MT
remaining) of 11 %, providing confirmation that using the model output
NO3 exposure in the subsequent analysis of aerosol mass yields from the
OFR is justified. A similar analysis of SQT decay was not possible, because
ambient SQT concentrations were too small to accurately measure fractional
decays. Also, MBO did not react substantially with NO3 in the OFR,
consistent with the lifetime for reaction of NO3 with MBO that is
approximately 3 orders of magnitude slower than for reaction with MT
(Atkinson and Arey, 2003). This is also representative of the
atmosphere, where MBO will overwhelmingly react with OH or O3 and not
NO3 (Atkinson and Arey, 2003).
Unlike NO3 exposure, the estimation of O3 exposure did not require
a detailed chemical model since the O3 system had no reservoir species
analogous to N2O5. O3 exposure was simply estimated as the
measured O3 concentration in the OFR output multiplied by residence
time. To verify this estimate, the measured fraction of MT that reacted in
the OFR was compared in Fig. 3 to a model prediction calculated using a
simple set of reactions of ozone with the three major MT species (Table S2).
The model is consistent with measurements within an error of 9 % and
shows that a parameterization for mixing of the O3 flow into ambient
air was not needed. In contrast to the slower 10–100 sccm flow of
N2O5, the 0.5 lpm flow of O2+O3 appears to have been
large enough relative to the total OFR flow rate to result in sufficiently
complete mixing. This result suggests that a faster flow of N2O5
could be used in future NO3 oxidation experiments to facilitate better
mixing.
Ambient, measured remaining, and modeled remaining MT from O3
oxidation on 23–24 August (a) and NO3 oxidation on 22–23
August (b) in the OFR. Modeled O3 and NO3 exposures are also shown.
The amount of oxidation was cycled from no added oxidant (no MT reacted) to
maximum oxidation (most or all MT reacted) in repeated 2–3 h cycles.
Time series examples of measured and modeled MT remaining after OFR
oxidation are compared to ambient MT concentrations for both NO3 and
O3 oxidation in Fig. 4. These examples illustrate the dynamic range
from no MT reacted (i.e., when no oxidants were added to the ambient air) to
nearly all MT reacted within the 2–3 h cycles for both oxidants. Further
examples are shown for NO3 oxidation in Fig. S6 and for O3
oxidation in Fig. S7.
SOA formed from oxidation of ambient air
OA enhancement vs. photochemical age
During BEACHON-RoMBAS, ambient air was oxidized by either OH, O3, or
NO3 in order to study the amount and properties of SOA that could be
formed from ambient precursors. In situ SOA formation from OH oxidation was
the subject of a previous paper
(Palm et al.,
2016). Select results are reproduced here as a comparison to SOA formation
from O3 and NO3 oxidation. Additional new analyses of the chemical
composition of SOA formed from OH oxidation are also included along with
O3 and NO3 oxidation in Sect. 3.3–3.4.
OA enhancement from oxidation of ambient air by O3 (a) and
NO3 (b) as a function of oxidant exposure. Data are colored by
ambient in-canopy MT concentrations and include the LVOC fate correction.
Binned averages for times when ambient MT concentrations were either below
or above 3 µg m-3 (0.66 ppb) are also shown, illustrating the
positive relationship between OA enhancement and MT concentrations at the
higher oxidant concentrations.
In Palm et al. (2016),
SOA formation from OH oxidation in the OFR correlated with ambient MT
concentrations (and implicitly with any other gases that correlated with MT,
such as SQT and possibly terpene oxidation products). Here, Fig. 5 shows the
OA enhancement observed after O3 and NO3 oxidation as a function
of eq. age in the OFR. Similar to OH oxidation, little SOA formation was
observed from O3 or NO3 oxidation when ambient MT concentrations
were low, regardless of the amount of exposure. When MT concentrations were
higher, increasing amounts of SOA were formed with increasing exposure. As
seen in Fig. 5 (and in Fig. 6 below), lower eq. NO3 ages were achieved
when MT concentrations were higher, and higher eq. NO3 ages were
achieved when MT concentrations were lower. This was because the higher MT
concentrations occurred during nighttime, when lower ambient temperatures
shifted the equilibrium towards N2O5 and away from
NO2+NO3 (from the injected N2O5), meaning lower
NO3 exposures were realized in the OFR.
OA enhancement vs. age in eq. d for OH, O3, and NO3
oxidation, separated into daytime (08:00–20:00 LT) and nighttime
(20:00–08:00 LT) data. All data are LVOC fate corrected. OH oxidation
produced several-fold more OA enhancement than O3 and NO3
oxidation. OH-aged OA enhancement data (only for < 5 eq. d aging where the LVOC fate correction could be
applied) are taken from Palm et al. (2016).
Another way to examine the trends in OA enhancement is by separating the
results into daytime and nighttime. Due to diurnal cycles in the emission
rates (that are strong functions of temperature, and also light for some
species), vertical mixing in the boundary layer, and changing rates of
ambient oxidation, the concentration of MT (and other SOA precursors) in
ambient air showed substantial diurnal cycles
(Kim
et al., 2010; Fry et al., 2013; Kaser et al., 2013). Ambient air was
characterized by higher MBO+isoprene (with ambient OH and O3
chemistry) during the day and higher MT+SQT (with ambient O3 and
NO3 chemistry) during the night
(Fry et al., 2013). Due to these
changes, it might be expected that SOA formation in the OFR would also
change diurnally.
OA enhancements vs. eq. age for OH, O3, and NO3 oxidation are
shown together in Fig. 6, split between daytime (08:00–20:00 LT) and
nighttime (20:00–08:00 LT). For all oxidants, more SOA formation was
observed during nighttime. This is consistent with the general increase in
MT and SQT (average of 1.1 and 0.04 ppbv in the canopy during nighttime and
0.4 and 0.03 ppbv during daytime respectively) and related precursor
concentrations in the shallower nighttime boundary layer. This higher SOA
formation during nighttime was not a result of larger temperature-dependent
partitioning to the particle phase at lower nighttime temperatures, as
evidenced by stable values of measured OA enhancement per unit ambient MT
(the dominant measured SOA precursor) across the whole range of ambient
temperatures (shown in Fig. S8). An exploration of the correlation between
maximum SOA formation from each oxidant and all available ambient VOC
concentrations is shown in Fig. S9, illustrating that MT are the best tracer
of SOA production at this forest site. The maximum amount of SOA formed from
OH oxidation was approximately 4 times more than from O3 or NO3
oxidation for both daytime and nighttime over the eq. ages covered in this
work. If the gases that formed SOA from each oxidant were the same, then
this would require the SOA yields from OH oxidation to be more than 4 times
larger than from O3 or NO3 oxidation. The references for SOA
yields from O3 and NO3 oxidation presented herein and for OH
oxidation presented in Palm et al. (2016) show this is
likely not the case. Instead, one possible explanation for this result could
be that a large fraction of SOA-forming gases found in ambient air do not
have C = C bonds (e.g., MT oxidation products such as pinonic acid). Such
molecules would typically not react appreciably with O3 or NO3
over the range of eq. ages achieved in this work but will still react with
OH and may lead to SOA formation. Future O3 and NO3 oxidation
studies should include higher eq. age ranges in order to investigate whether
additional SOA could be formed from ambient precursors at higher ages. This
concept will be discussed further in Sect. 3.2.2.
Measured vs. predicted SOA formation for O3 and NO3
oxidation in an OFR. The measured SOA formation includes the LVOC fate
correction and includes all ages greater than 0.7 eq. d for O3-PAM and
greater than 0.3 eq. d for NO3-PAM. Predicted SOA formation was
estimated by applying published chamber SOA yields to the mass of VOCs
predicted by the model to be oxidized in the OFR (see Sect. 2.3 for
details).
Whereas a net loss of OA was observed at > 10 eq. days of OH
aging due to heterogeneous oxidation
(shown in Fig. 7 of Palm et al., 2016), a similar net loss of OA at the highest eq. ages of
O3 and NO3 oxidation was not observed. Since the highest eq. ages
for both O3 and NO3 oxidation were approximately 5 days, it is
unclear whether O3 or NO3 heterogeneous oxidation would lead to net
loss of ambient OA at substantially higher ages. Future experiments could be
designed to achieve higher ages in order to investigate this effect.
Measured vs. predicted OA enhancement
When ambient air is sampled into an OFR, any gases or particles present in
that air are subject to oxidation. Measurement of the resultant SOA
formation is a top-down measure of the total SOA formation potential of that
air as a function of eq. age of oxidation. In other words, an OFR can be
used to determine the relative concentrations of SOA-forming gases present
in ambient air at any given time. To provide context to the measurements in
the OFR, a bottom-up analysis can be carried out by applying laboratory SOA
yields to the measured ambient SOA-forming gases that are entering the OFR.
The measured SOA formation after oxidation by O3 and NO3 is shown
vs. the SOA predicted to form from measured precursor gases in Fig. 7. The
measured SOA formation includes all ages greater than 0.7 eq. d for
O3-PAM and greater than 0.3 eq. d for NO3-PAM, where most or all
of the VOCs have reacted. For both oxidants, the data are scattered along
the 1:1 line of equal measured and predicted SOA formation. This is in
contrast to the analysis for OH oxidation in Palm et al. (2016), where a factor
of 4.4 more SOA was formed from OH oxidation than could be explained by
measured VOC precursors. As shown in that analysis, the additional
SOA-forming gases in ambient air were likely S/IVOCs, where the SOA
formation from S/IVOCs was 3.4 times larger than the source from VOCs. This
conclusion was supported by unspeciated measurements of total S/IVOC
concentrations (classified by volatility). SOA yields from S/IVOCs or any
other sources are not required to explain SOA formation from O3 or
NO3. This suggests that the majority of S/IVOCs in this ambient forest
air generally did not contain C = C bonds and therefore did not typically
react with O3 or NO3 to produce SOA on atmospherically relevant
timescales. This is consistent with expectations based on laboratory and
ambient studies of MT and SQT oxidation products. Typical oxidation products
include compounds such as pinic acid, pinonic acid, pinonaldehyde,
caronaldehyde, and nopinone, none of which contain C = C double bonds
(e.g., Calogirou et
al., 1999b; Yu et al., 1999; Lee et al., 2006). As an example, the reaction
rates of pinonaldehyde with OH, O3, and NO3 are 3.9 × 10-11, < 2 × 10-20,
and 2.0 × 10-14 cm3 molec-1 s-1 respectively
(Atkinson et al., 2006). These rates
correspond to eq. lifetimes of 4.7 h, > 579 days, and 29 days,
respectively, showing that pinonaldehyde will typically only react with OH
in the atmosphere or in the OFR under the conditions in this study.
While both the measured and predicted SOA formation shown in Fig. 7 are
consistent with each other, two main caveats limit the strength of the
conclusions that can be drawn from this particular study. First, the amount
and dynamic range of SOA formed from O3 and NO3 oxidation were
relatively small, as were the total ambient aerosol concentrations. This
caused the SMPS+AMS measurement noise and variability to be larger
relative to the total aerosol measurements than they would be for higher
aerosol concentrations. Also, as only a small amount of new SOA was formed,
the aerosol condensational sink remained relatively low for all
measurements. According to the LVOC fate model, on average only 31 and
36 % of LVOCs condensed to form SOA during O3 and NO3 oxidation
respectively (see Fig. S3). This required a correction of approximately a
factor of 3 to correct measured SOA formation to what would occur in normal
atmospheric conditions.
Scatter plots of µmol m-3 O and µmol m-3 H
added per µmol m-3 C added from OH oxidation of ambient air in
the OFR. Slopes are fit to the photochemical age ranges of 0.1–0.4
(avg. = 0.18) days, 0.4–1.5 (avg. = 0.9) days, 1.5–5 (avg. = 2.7) days, and 5–15
(avg. = 10) days, showing that the atomic O : C (H : C) ratios of the SOA mass
formed in those ranges were 0.55 (1.60), 0.84 (1.44), 1.13 (1.36), and 1.55
(1.22) respectively. At higher ages, heterogeneous oxidation led to loss of
C and H and little to no loss of O.
H : C and O : C ratios of SOA formed from oxidation of ambient air
Analysis of ambient high-resolution AMS spectra can be used to estimate the
elemental composition of OA
(Aiken
et al., 2008; Canagaratna et al., 2015). When SOA is formed in the OFR, the
OA that is sampled in the OFR output is a sum of preexisting ambient OA and
any SOA produced from oxidation. At sufficiently high eq. ages, the sampled
OA will also include the effects of heterogeneous oxidation. The amount of
O, C, and H atoms added by oxidation can be calculated by subtracting the
ambient elemental concentrations from those measured after aging. The
amounts of each element added by oxidation can be used to determine the O : C
and H : C elemental ratios of the SOA that is formed in the OFR.
The amounts of O and H vs. C added from OH oxidation are shown in Fig. 8.
Slopes were fit to the data with positive net addition of C in order to
determine the O : C and H : C of the SOA formed for the eq. photochemical age
ranges of 0.1–0.4 (avg. = 0.18) days, 0.4–1.5 (avg. = 0.9) days, 1.5–5
(avg. = 2.7) days, and 5–15 (avg. = 10) days. The elemental O : C (H : C) ratios of
the SOA mass formed in those ranges were 0.55 (1.60), 0.84 (1.44), 1.13
(1.36), and 1.55 (1.22). For data with ages of longer than several eq. days,
O was added coincident with loss of C (i.e., negative x intercept), which is
likely due to heterogeneous oxidation leading to fragmentation/evaporation
of preexisting OA. This conclusion is reinforced by the evidence that for
eq. OH ages greater than several days, heterogeneous oxidation resulted in a
net loss of C when ambient MT concentrations were low (Fig. S10), but not
for lower eq. ages. Similarly, George and Abbatt (2010)
suggested that the lifetime of ambient OA with respect to heterogeneous OH
oxidation is approximately 2 to 3 days. Therefore, the change in
amounts of O, C, and H after several eq. days of oxidation will be a mix of
heterogeneous change to preexisting OA and addition of new SOA. These
effects of heterogeneous oxidation (i.e., x and y intercepts) are likely to
be approximately the same for all data within each given age range, meaning
the slopes fitted above are independent of the heterogeneous processes and
contain information about the elemental changes associated with the
formation of varying amounts of SOA within each age range.
Scatter plots of µmol m-3 O and µmol m-3 H
added per µmol m-3 C added from O3 oxidation of ambient air
in the OFR. Data are colored by eq. d of O3 exposure. The slopes show
that the atomic O : C (H : C) ratio of the SOA mass formed was 0.50 (1.61). The
slopes did not change with increasing photochemical age.
Analogous to Fig. 8, the amount of O and H vs. C added from O3 and
NO3 oxidation is shown in Figs. 9–10. The SOA added from O3
oxidation had O : C and H : C ratios of 0.50 and 1.61. The SOA added from
NO3 oxidation had O : C and H : C ratios of 0.39 and 1.60. This O : C value
of 0.39 for NO3 oxidation includes only the O atoms that were bound to
the C backbone of the organic molecules and excludes the two O atoms that
are bound only to N in the -ONO2 (nitrate) functional group
(Farmer et al., 2010). If all O atoms in the
nitrate functional group are included, the O : C of this added SOA mass was
0.44. Inclusion of only the carbon-bound oxygen of the nitrate functional
group is more reflective of the carbon oxidation state and is also what is
typically reported for AMS O : C measurements (since the organic -NO2
moiety is measured in the AMS as total nitrate and typically not separated
from inorganic nitrate).
Heterogeneous oxidation was not expected to be a factor for the O3 and
NO3 ages used in this work. This assumption was reinforced by the fact
that no net loss of C was observed for these amounts of oxidation, even when
ambient MT concentrations (and OA enhancement) were low, as shown in Figs. S11–12. This assumption is also consistent with previous research on
lifetimes of OA components with respect to O3 and NO3
heterogeneous oxidation. For instance, several aldehydes were found to have
a relatively long lifetime equivalent to approximately 2–8 days for
NO3 heterogeneous oxidation when calculated using 1 pptv ambient
NO3 (Iannone et al., 2011). Ng et al. (2017)
summarized that reactive uptake of NO3 into particles is slow for most
molecules, with the exception of unsaturated or aromatic molecules, which
were unlikely to be major components of the ambient OA in this remote forest
(Chan
et al., 2016). Although the lifetime of pure oleic acid (which contains a
C = C bond) particles with respect to heterogeneous O3 oxidation can be
as short as tens of minutes (Morris et
al., 2002), lifetimes for oleic acid in atmospheric particle organic
matrices can be tens of hours to days (Rogge
et al., 1991; Ziemann, 2005). Furthermore, the uptake coefficients for
O3 to react with saturated molecules are typically 1–2 orders of
magnitude slower than for unsaturated molecules
(de Gouw and Lovejoy, 1998). In summary, this
previous research suggests that heterogeneous oxidation by O3 or
NO3 may be important at higher eq. ages, but not for those achieved in
the present work.
Scatter plots of µmol m-3 O and µmol m-3 H
added per µmol m-3 C added from NO3 oxidation of ambient
air in the OFR. The amount of O added is shown without including the O from
the -NO2 group, since those O atoms do not affect the oxidation state
of C. The slopes show that the atomic O : C(H : C) ratio of the SOA mass formed
was 0.39 (1.60). The slopes did not change with increasing NO3
exposure. Contrary to Figs. 8–9, data are not colored by NO3 exposure.
The ranges of NO3 exposure achieved during daytime vs. nighttime were
unequal (Figs. 5–6, S12), obscuring any trend of OA enhancement vs. eq. age.
To put the O : C and H : C values of the SOA formed in the OFR in perspective,
Van Krevelen diagrams of H : C vs O : C ratios for OA measured after OH,
O3, and NO3 oxidation are shown compared to concurrent
measurements of ambient OA in Fig. 11a–c and summarized together in Fig. 11d. The effect of heterogeneous OH oxidation on preexisting aerosol is also
shown as a line with a slope of -0.58. This line was fitted to the H : C vs.
O : C of all OH-aged data where a net loss of C was observed (i.e., SOA
formation was not observed and heterogeneous oxidation dominated). Generally
speaking, less-oxidized (“fresh”) OA will lie in the upper left portion of
a Van Krevelen plot, with higher H : C values and lower O : C values.
Conversely, more-oxidized (“aged”) OA will move towards the lower right,
with lower H : C values and higher O : C values
(Heald
et al., 2010; Ng et al., 2011). Shown in Fig. 11, the SOA formed from
O3, NO3, and the lowest amount of OH aging (0.1–0.4 eq. days) was
found at the upper left of the range occupied by ambient OA. As OH aging
increased to higher ranges, the values of H : C decreased and the values of
O : C increased, already moving beyond the local ambient range after 0.9 eq.
days. At the higher ages, the H : C of the SOA formed lies at higher H : C
values than those of the total OA measured after OH aging, which are closer
to the trend of heterogeneous oxidation in the Van Krevelen space. This
shows that SOA formed via gas-phase OH oxidation processes in an OFR has a
higher H : C than the OA that results from heterogeneous oxidation, while both
processes lead to similar increases in O : C. The net movement in the Van
Krevelen space can be considered as starting at the ambient H : C and O : C and
moving along two vectors: one vector along the heterogeneous oxidation line
and another towards the H : C and O : C values of the new SOA formed in the gas
phase, where the length of those two vectors is weighted by the amount of
OA resulting from each process. When little SOA is formed, the H : C and O : C
measured after oxidation lie along the heterogeneous oxidation line. When
high amounts of SOA are formed, the H : C and O : C after oxidation shift to
higher H : C values, lying closer to the curve defined by the H : C and O : C of
SOA mass added in the OFR at the different age ranges (see Fig. S13).
Van Krevelen diagrams of H : C vs. O : C ratios of OA after oxidation
by (a) NO3, (b) O3, and (c) OH along with concurrent ambient ratios.
The H : C and O : C ratios of the new SOA mass formed in the OFR (i.e., the
slopes from Figs. 8–10) are shown for each oxidant (diamonds) and are
summarized in (d) compared with all ambient measurements. For data where no
net C addition was observed after OH oxidation, the slope along which
heterogeneous OH oxidation transforms the ambient OA is shown (purple dashed
line).
While these two vectors describe the possible oxidation processes in the
OFR, there may be other vectors (e.g., from condensed-phase chemistry or
reactive uptake) occurring in the atmosphere. As documented in Hu et al. (2016), SOA
formation processes that require reactive uptake or within-particle
non-radical chemistry (such as uptake of isoprene epoxydiols to form
IEPOX-SOA) on timescales longer than the several minute residence time in
the OFR are not captured with the OFR method used in this work. This is
because the rate of reactive uptake and non-radical particle-phase chemistry
do not speed up proportionally to increased OH and HO2 (or O3 or
NO3). However, to our knowledge the only precursor for which reactive
uptake of epoxides has been shown to be a major pathway is isoprene, which
was a very minor precursor at this site (Karl et al.,
2012). The formation of epoxides during MBO oxidation has been proposed to
play at role during BEACHON-RoMBAS
(Zhang
et al., 2012). However, recent results suggest that formation of epoxides
during MBO oxidation is not important in the atmosphere
(Knap et al., 2016). Thus, at this time it is not
clear whether any important SOA-forming processes in this environment are
missed by the OFR setup, and this question should be investigated in future
studies.
The H : C of the least-oxidized SOA formed in the OFR from all oxidants was
near 1.6. As discussed in Palm et al. (2016), SOA formation
from OH oxidation in the OFR correlated with MT, and the S/IVOC sources of
SOA may have been MT oxidation products or other related biogenic gases.
Biogenic terpenes are composed of isoprene units, meaning they all have H : C
of 1.6. Therefore, the SOA formed from the lowest eq. ages in the OFR was
consistent with oxidation processes that add roughly 4–6 O atoms without
removing net H atoms. Addition of -OH or -OOH functional groups after -H
abstraction by OH radicals results in addition of O without loss of H and
is consistent with the RO2+HO2 reaction conditions that are
expected during OH oxidation in the OFR
(Kroll
and Seinfeld, 2008; Ortega et al., 2016). OH can also add to a C = C bond,
which could lead to addition of H atoms after oxidation. O3 and
NO3 are expected to react with MT almost exclusively by addition to a
C = C bond, which leads to addition of O without initial removal of H atoms
(Atkinson and Arey, 2003). However, previous research has shown
that many precursor gases, including aromatic molecules with initial H : C
close to 1, can form SOA with H : C close to 1.6
(Chen
et al., 2011; Chhabra et al., 2011; Canagaratna et al., 2015; Hildebrandt
Ruiz et al., 2015). Therefore, H : C alone cannot provide direct evidence
about the specific identities of precursor gases in ambient air. The SOA
from O3, NO3, and 0.1–0.4 eq. days OH aging had H : C values
similar to typical semivolatile oxidized organic aerosol (SV-OOA), while
the H : C of SOA from 0.4 to 1.5 eq. days or longer OH aging resembled low-volatility oxidized organic aerosol (LV-OOA); these two types of SOA have
been identified in ambient air at many locations
(Jimenez
et al., 2009; Canagaratna et al., 2015).
The relative timescales of oxidation and condensation in the OFR also need
to be considered in order to properly interpret the H : C and O : C of the SOA
mass formed in the OFR. In the atmosphere, once a molecule is oxidized to an
LVOC that is able to condense onto a particle, lifetimes for condensation
onto aerosols are on the order of several minutes
(Farmer
and Cohen, 2008; Knote et al., 2015; Nguyen et al., 2015). This is typically
much shorter than the lifetimes for subsequent reaction with OH, O3, or
NO3 of tens of minutes to several hours or longer, so condensation will
likely occur prior to further oxidation. In OFR oxidation experiments, the
lifetime for subsequent oxidation of LVOCs is shortened proportional to the
increase in oxidant concentration. However, the condensation lifetime does
not scale with oxidant concentration and remains roughly constant. At
sufficiently high oxidant concentrations, LVOCs can be subjected to further
oxidation steps that they would not be subjected to in the atmosphere prior
to having a chance to condense to form SOA. To compare SOA formation in the
OFR vs. ambient air, these relative timescales are considered here as a
function of both oxidant type and amount of oxidant exposure.
The lowest range of OH aging for which O : C and H : C values were measured was
0.1–0.4 (avg. 0.18) eq. d, which is 2.4–9.6 (avg. 4.3) eq. h of oxidation.
Typical terpenes have lifetimes for reaction with OH on the order of tens of
minutes to several hours in the atmosphere (Atkinson and Arey,
2003), which is similar to this lowest eq. age range in the OFR. Typical
terpene oxidation products have lifetimes ranging from 3.9 h
(caronaldehyde;
Alvarado et al., 1998) to 4.7 h
(pinonaldehyde; Atkinson et al., 2006) to
11–13 h
(nopinone;
Atkinson and Aschmann, 1993; Calogirou et al., 1999a) to a computationally
estimated 18–21 h (pinic and pinonic acid;
Vereecken and Peeters, 2002). As a rough approximation, this suggests that
the SOA formed in the OFR is likely a result of approximately one or at most
a few oxidation steps occurring to the molecules that enter the OFR (which
may have already experienced one or more oxidation steps in the atmosphere
prior to entering the OFR). The aging in this range strikes a balance
between achieving enough oxidation to react all incoming precursors at least
once while not reacting them an unrealistic number of times in the gas phase
before allowing sufficient time for condensation. In the next age range of
0.4–1.5 (avg. 0.9) eq. d of OH aging, in which the maximum OA enhancement
occurred, some primary precursors are likely starting to be oxidized
multiple times inside the OFR prior to condensation, while some oxidation
products will still be oxidized only ∼ 1–2 times. The SOA
formed in this range may represent SOA formed from multiple generations of
chemistry. At higher ages in the OFR, the aerosol is likely mainly modified
by heterogeneous oxidation, with a small contribution from condensation of
highly oxidized products. This OA at the highest ages resembles ambient OA
found in remote locations
(Jimenez et al., 2009;
Chen et al., 2015). Indeed, OFRs have previously been used to study
heterogeneous oxidation processes
(George et al., 2008; Smith et
al., 2009).
For O3 and NO3 oxidation, the oxidants will react only with C = C
double-bond-containing gases. The major MT and SQT species at this field
site all contain only a single C = C bond (isoprene and minor MT and SQT
species contain two). Subsequent reaction lifetimes of oxidation products
with these oxidants will likely be longer than the lifetime for condensation
onto particles. For example, the lifetimes for pinonaldehyde with respect to
O3 and NO3 oxidation are > 579 days, and 29 days
respectively (Atkinson et al., 2006).
Therefore, we can approximate that multiple generations of oxidation are not
dominant for SOA formation when investigating O3 or NO3 oxidation
in the OFR at this site. This is consistent with previous chamber SOA
formation experiments that suggested that first-generation oxidation
products dominate SOA formation from O3 oxidation of a variety of
biogenic compounds with a single C = C bond rather than products of later
generations of oxidation (Ng et al.,
2006). The SOA formed via O3or NO3 oxidation in the OFR is
likely formed from reaction with primary VOCs and a small subset of their
reaction products that still contain C = C bonds, such as the α-pinene oxidation product campholenic aldehyde (Kahnt
et al., 2014). This SOA should be representative of typical atmospheric SOA
formation processes.
Particulate organic nitrate (pRONO2) formation from NO3
oxidation of ambient air
In addition to estimating the elemental composition of OA, the AMS can also
be used to estimate the amount of inorganic vs. organic nitrate in submicron
aerosols
(Farmer et
al., 2010; Fry et al., 2013). The ratio of NO2+ to NO+
fragment ions produced by thermal decomposition on the AMS vaporizer and
electron impact ionization depends on the type of nitrate. NH4NO3
typically produces a ratio of approximately 0.3–1, while particulate organic
nitrate (pRONO2), in which the -ONO2 functional group is
covalently bonded to the carbon backbone (R) through an oxygen atom,
typically produces a ratio ∼ 2–3 times lower
(Fry
et al., 2009; Bruns et al., 2010; Farmer et al., 2010; Liu et al., 2012; Day
et al., 2017). The measured NO2+ to NO+ ratio is a linear
combination of these two chemical components. Using this principle, the
NO3 measured by the AMS was split into the estimated fractions of
NH4NO3 and pRONO2 according to the method described in Fry et al. (2013). For the instrument in this work, ratios of 0.3 and 0.13 were
used for the NO2+ to NO+ ratios of NH4NO3 and
pRONO2 respectively (Fry et
al., 2013).
Example time series of OA, NH4, and NO3 (split into
pRONO2 and NH4NO3) aerosol measurements after NO3
oxidation in the OFR, compared to ambient aerosol, NO2+ to
NO+ ratio, model-derived eq. age of NO3 oxidation, MT
concentration, and RH measurements. Production of both NH4NO3 and
pRONO2 was observed at different times, which appears to depend on
changes in experimental conditions.
A two-night example of both ambient and NO3-radical aged aerosol on
20–22 August is shown in Fig. 12. In ambient air, the majority of NO3
aerosol was organic. After oxidation in the OFR, different behavior was seen
on the two nights shown. On the first night, mainly inorganic nitrate was
produced, as evidenced by the higher NO2+ to NO+ ratio, the
formation of NH4 aerosol, and the relatively small amount of SOA
formed. On the second night, pRONO2 was produced, as evidenced by the
lower NO2+ / NO+ ratio, a lack of NH4 aerosol formation,
and substantial SOA formation. The organic nitrate formation and SOA
formation also roughly tracked the ambient MT concentrations.
These two distinct behaviors in the NO3-OFR were likely controlled by
ambient RH. There was a competition between thermal dissociation of injected
N2O5 to produce NO3+NO2 (favored at high temperatures
and low RH) and the hydrolysis of N2O5 on wetted OFR walls to
produce HNO3 (favored at low temperatures and high RH). When hydrolysis
occurred rapidly, there was a sharp decrease in N2O5
concentrations. The NO3 radical concentrations were also greatly
reduced, and thus fewer NO3 radicals were available to react with
ambient gases (e.g., MT) to produce pRONO2. HNO3 reacted with NH3
in ambient air or evaporating from OFR surfaces to produce NH4NO3.
The results shown in Fig. 12 illustrate this behavior, with NO3 radical
exposure being reduced while NH4NO3 was produced during the first
night. Despite the presence of similar MT concentrations on both nights,
little SOA was produced on the first night. Future applications could
include heating of the OFR slightly above ambient temperatures in order to
prevent hydrolysis of N2O5 on the OFR walls. Inhibiting
NH4NO3 formation artifacts would be especially critical for data
interpretation if measuring aerosol enhancements with only non-chemical
instruments such as an SMPS.
Organic -ONO2 mass added vs. OA added from OH, O3, and
NO3 oxidation in an OFR. No pRONO2 formation was observed (or
expected) from OH or O3 oxidation under the experimental conditions.
The slope of 0.10 from NO3 oxidation is consistent with previous
chamber measurements (shown in grey), which range from approximately
0.1 to 0.18 (Fry et al., 2009, 2011; Boyd et al., 2015).
Despite this complex chemistry, information about the chemical composition
of pRONO2 formed from real atmospheric precursors can still be derived
from times when conditions favored pRONO2 formation. Shown in Fig. 13
is the mass of organic -ONO2 added vs. SOA added from oxidation by
each of the three oxidants. Substantial formation of pRONO2 was
observed only for NO3 radical oxidation, and not for O3 or OH
oxidation. This was expected, since ambient NOx concentrations were
generally low
(0.5–4 ppb; Ortega et al., 2014), and the NO3 oxidation experiment was the
only one with an added source of reactive nitrogen. The slopes of Fig. 13
represent the ratio of -ONO2 to the rest of the organic molecules in
pRONO2. In this study, the slope after NO3 radical oxidation was
0.10, which is similar to the range of 0.1–0.18 found in previous chamber
studies of NO3 oxidation of terpenes
(Fry
et al., 2009, 2011; Boyd et al., 2015). To put this in context, if every SOA
molecule formed in the OFR contained a single -ONO2 group (with its
mass of 62 g mol-1), then the molecular mass of the full pRONO2
molecules would be an average of 620 g mol-1 (giving the slope of
62 g mol-1/620 g mol-1= 0.10 in Fig. 13). Alternatively, if all
molecules are assumed to have a mass of 200 or 300 g mol-1, then 32
or 48 % of the molecules, respectively, would contain a -ONO2
functional group (assuming no molecules contain more than one -ONO2
group). Again, this result is roughly consistent with previous research. For
the fraction of OA composed of pRONO2 in NO3+β-pinene
SOA, Fry et al. (2009) estimated 32–41 %
(assuming an average molecular weight of 215–231 g mol-1), Fry et al. (2014) estimated 56 % (assuming 214 g mol-1),
and Boyd et al. (2015)
estimated 45–68 % (assuming 200–300 g mol-1).