Organic Nitrate Aerosol Formation via NO 3 + Biogenic Volatile Organic Compounds in the Southeastern United States

. Gas- and aerosol-phase measurements of oxidants, biogenic volatile organic compounds (BVOC) and organic nitrates made during the Southern Oxidant and Aerosol Study (SOAS campaign, Summer 2013) in central Alabama show that nitrate radical ( NO 3 ) reaction with monoterpenes leads to signiﬁcant secondary aerosol formation. Cumulative losses of NO 3 to terpenes are correlated with 5 increase in gas- and aerosol-organic nitrate concentrations made during the campaign. Correlation of NO 3 radical consumption to organic nitrate aerosol formation as measured by Aerosol Mass Spectrometry and Thermal Dissociation - Laser Induced Fluorescence suggests a molar yield of aerosol phase monoterpene nitrates of 23-44%. Compounds observed via chemical ionization mass spectrometry (CIMS) are correlated against predicted NO 3 loss to BVOC, and show C 10 H 15 NO 5 and 10 C 10 H 17 NO 5 as major NO 3 -oxidized terpene products being incorporated into aerosols. The comparable isoprene product C 5 H 9 NO 5 was observed to contribute less than 1% of the total organic nitrate in the aerosol phase and correlations show that it is principally a gas phase product from nitrate oxidation of isoprene. Organic nitrates comprise 30-45% of the NO y budget during SOAS. Inorganic nitrates were also monitored and showed that during incidents of increased coarse-mode 15 mineral dust, HNO 3 uptake produced nitrate aerosol at rates comparable to that of organic nitrate produced via NO 3 + BVOC.


Introduction
Secondary Organic Aerosol (SOA), formed from the oxidation of volatile organic compounds (VOCs) by ozone (O 3 ), hydroxyl radical (OH), or nitrate radical (NO 3 ), affects visibility as well as regional 20 and global radiative climate forcing (Myhre et al., 2013;Bellouin et al., 2011;Feng and Penner, 2007;Goldstein et al., 2009). Aerosol has been studied as a source for significant risk factors for pulmonary and cardiac disorders (Nel, 2005;Pope and Dockery, 2006). Organic aerosol (OA) contributes a large fraction of the total tropospheric submicron particulate matter (PM, De Gouw, 2005;Heald et al., 2005;Zhang et al., 2007). Biogenic volatile organic compounds (BVOC) are domi-25 nant precursors in SOA formation (Goldstein and Galbally, 2007;Spracklen et al., 2011). SOA is a significant fraction of total aerosol mass in the Southeastern United States (SEUS) (predicted to be 80-90% of the organic aerosol load, Ahmadov et al., 2012, Stocker et al., 2013. Understanding the interaction of anthropogenic pollutants with BVOC is vital to improving our understanding of the human impact on SOA formation (Carlton et al., 2010;Spracklen et al., 2011) and the associated 30 radiative forcing of climate change (Stocker et al., 2013).
Nitrogen oxides (NO x = NO + NO 2 ), common byproducts of combustion, are linked to aerosol formation in the troposphere via daytime and nighttime oxidation mechanisms (Rollins et al., 2012).
Total reactive nitrogen, NO y , consists of NO x , as well as NO x reaction products, including NO 3 , HNO 3 , HONO, alkyl nitrates, peroxynitrates and other particulate nitrates. Alkyl nitrates produced 35 from oxidation of VOC are related to tropospheric ozone generation (Chameides, 1978) and, via low-volatility products, can lead to formation of SOA (Hallquist et al., 2009). Oxidation of NO x to nitric acid (HNO 3 ) can also produce inorganic nitrate aerosol via heterogeneous uptake of HNO 3 onto mineral or sea salt aerosols (Vlasenko et al., 2006) and via co-partitioning with ammonia to form semi-volatile NH 4 NO 3 (Lee et al., 2008). 40 Nitrogen oxides are primarily emitted as NO (Nizich et al., 2000;Galloway et al., 2004;Wayne et al., 1991). NO is oxidized to NO 2 and further to the highly reactive NO 3 radical. NO 3 is especially predominant at night when loss via photolysis and NO reaction are at a minimum (Horowitz et al., 2007;von Kuhlmann et al., 2004;Xie et al., 2013).
The formation of NO 3 and the associated N 2 O 5 in the atmosphere have been studied in detail 45 (Bertram and Thornton, 2009;Brown and Stutz, 2012;Wagner et al., 2013). The hydrolysis of N 2 O 5 to HNO 3 can be important in the prediction of the tropospheric oxidant burden with respect to the O 3 production, and therefore OH radical production (Dentener and Crutzen, 1993;Evans and Jacob, 2005). However, previous studies in Eastern Texas have found N 2 O 5 uptake into aerosols to be relatively low in the southern United States (TexAQS average γ of 0.003) (Brown 50 et al., 2009;Riemer et al., 2009).
NO 3 is an effective nocturnal oxidizer of BVOC (Atkinson andArey, 2003, 1998;Calogirou et al., 1999;Winer et al., 1984). NO 3 oxidation is especially reactive towards unsaturated, non-aromatic hydrocarbons of which BVOC are major global constituents. NO 3 is less reactive towards aromatic compounds and saturated hydrocarbons, major compounds of anthropogenic VOCs. Nitrate oxida-55 tion of some BVOC compounds, such as β-pinene, lead to rapid production of SOA in laboratory experiments with high yields (Griffin et al., 1999;Jimenez et al., 2009;Zhang et al., 2007;Hallquist et al., 2009;Fry et al., , 2009Boyd et al., 2015). Analysis of previous field studies have characterized the loss of NO 3 to its major daytime sinks, including reaction with NO and photolysis, as well as its loss to BVOC during both daytime and nighttime (Aldener et al., 2006;Brown et al., 60 2005).
Ambient concentrations of alkyl nitrates and peroxynitrates can be quantified using laser-induced 65 fluorescence (Day et al., 2002;Rollins et al., 2010) and mass spectrometry methods (Bahreini et al., 2008;Farmer et al., 2010;Beaver et al., 2012;Fry et al., 2013). Ions and acids (i.e. HNO 3 ) can be quantified using ion chromatography (IC, Makkonen et al., 2012;Trebs et al., 2004) as well as CIMS (Beaver et al., 2012). The combination of these instruments, as well as others discussed below, allow for the determination of a total ambient oxidized nitrogen (NO y ) budget, which enables the 70 interpretation of the importance of nitrogen oxides in SOA formation. Xu et al. (2015a) have reported that organic aerosol from nitrate radical oxidized monoterpenes are strongly influenced by anthropogenic pollutants and contribute to 19-34% of the total OA content (labeled less-oxidized oxygenated organic aerosols, LO-OOA). Monoterpene oxidation products show a large contribution to LO-OOA year-round (Xu et al., 2015b). Another AMS factor specific to 75 reactive uptake of isoprene oxidation products (e.g. IEPOX), Isoprene-OA, is isolated in the warmer summer months in both urban as well as rural areas across the southeastern United States and contributes 18-36% of summertime OA (Hu et al., 2015;Xu et al., 2015a). LO-OOA is seen predominantly during nighttime hours, implying NO 3 oxidation of monoterpenes, and is strongly correlated specifically with the nitrate functionality in organic nitrates (Xu et al., 2015b). It is suggested 80 that during the summer months, increasing nighttime LO-OOA balances with increasing daytime isoprene-OA to give the observed constant OA concentration over the diurnal cycle.
The 2013 SOAS campaign was a comprehensive field intensive in central Alabama near Centreville (CTR), in which concentrations of oxidants, BVOC and aerosol were measured with a particular focus on understanding the effects of anthropogenic pollution on SOA formation. The site was cho-85 sen due to its high biogenic VOC emissions as well as its relatively large distance from anthropogenic pollution ( Figure 1). County-level monoterpene emissions across the US shows the CTR site gives a regional representation of monoterpene emissions in the SEUS (Geron et al., 2000). Furthermore, Xu et al. (2015b) show that the CTR site is representative of more-oxidized and less-oxidized oxygenated organic aerosols (MO-OOA and LO-OOA, respectively) loadings across several monitoring 90 stations in the SEUS. Comparison of annual molar emissions in the SEUS (an 8-state region includ-ing the CTR site) of BVOC (estimated from Geron et al., 2000) to NO x emissions (from 2011 NEI database) suggest that NO x is the limiting reagent in BVOC + NO 3 reactions throughout the region.
Alabama is home to a number of power plant facilities that are large point sources of NO x capable of being carried long distances. Alabama's non-interstate roadways also have large emissions 95 of NO x , though a majority of the emissions come from urban areas. Although the NO x emissions have been steadily dropping since 1998, they are still substantial (2.70 million tons in reported for SEUS in 1999to 1.75 million tons in 2008, Blanchard et al., 2013. Frequent controlled biomass burning events (crop burning, Crutzen and Andreae, 1990), as well as vehicular sources (Dallmann et al., 2012) also contribute to local NO x emissions and PM concentrations (a full analysis of 100 contributions can be found at the EPA National Emissions Inventory, http://www.epa.gov/ttn/chief /net/2011inventory.html).
In the present study, we investigate the production of SOA species from NO 3 reaction with monoterpenes. NO 3 loss to BVOC is calculated and compared to AMS and TD-LIF measurements of aerosol organic nitrates, as well as individual product nitrates measured by CIMS. We compare 105 this NO 3 loss to BVOC to an alternate fate of NO x , heterogeneous HNO 3 uptake to produce inorganic nitrate aerosol, which is considered in detail in a companion paper (Allen et al., 2015). These two pathways from NO x to nitrate aerosol shown in Scheme 1 alternated dominance during SOAS.

Experimental
Measurements for the SOAS campaign took place near the Talladega National Forest, 6 miles south-110 west of Brent, AL (32.9029 N, 87.2497 W), from June 1 -July 15, 2013. The forest covers 157,000 acres to the northwest and southeast of Centerville, AL. Figure 1 shows a map of the site location as well as nearby point sources of anthropogenic NO x and SO 2 . The site is in a rural area representative of the transitional nature between the lower coastal plain and Appalachian highlands (Das and Aneja, 2003). Wind direction varied during SOAS allowing for periods of urban influence from 115 sources of anthropogenic emissions located near the sampling site, including the cities of Montgomery, Birmingham, Mobile and Tuscaloosa (Hidy et al., 2014). The closest large anthropogenic Two cavity ringdown spectrometers (CRDS) were used to determine ambient mixing ratios of 120 NO x , O 3 , NO y , NO 3 and N 2 O 5 (Wild et al., 2014;Wagner et al., 2011). CRDS is a high sensitivity optical absorption method based on the decay time constant for light from an optical cavity composed of two high reflectivity mirrors. NO 2 is measured using its optical absorption at 405 nm in one channel, and O 3 , NO and total NO y are quantitatively converted to NO 2 and measured simultaneously by 405 nm absorption on three additional channels. NO 3 is measured at its characteristic 125 strong absorption band at 662 nm. N 2 O 5 is quantitatively converted to NO 3 by thermal dissociation and detected in a second 662 nm channel with a detection limit of 1 pptv (30 s, 2 σ) for NO 3 and 1.2 pptv (30 s, 2 σ) for N 2 O 5 (Dubé et al., 2006;Wagner et al., 2011).
Thermal Dissociation Laser-Induced Fluorescence (TD-LIF, PM 2.5 size-cut) (Day et al., 2002;Farmer et al., 2010;Rollins et al., 2010) was used to measure total alkyl nitrates (ΣANs), total peroxy 130 nitrates (ΣPNs) and aerosol phase ΣANs (Rollins et al., 2012). High-resolution time-of-flight aerosol mass spectrometry (HR-ToF-AMS, hereafter AMS, DeCarlo et al., 2006;Canagaratna et al., 2004, PM 1 size-cut) was used to measure submicron organic and inorganic nitrate aerosol composition using the nitrate separation method described in Fry et al. (2013). Organic nitrates in the particle phase (pRONO 2 ) decompose prior to ionization on the AMS vaporizer to NO + 2 organic fragments, 135 hence pRONO 2 cannot be quantified directly from AMS data. The contribution of pRONO 2 to total particulate nitrate was calculated using the method first discussed in Fry et al. (2013) and is briefly summarized here. This method relies on the different fragmentation patterns observed in the AMS for organic nitrates vs NH 4 NO 3 , specifically the ratio of the ions NO + 2 to NO + . Since this ratio depends on mass spectrometer tuning, vaporizer settings and history, Fry and coauthors proposed 140 to interpret the field ratio of these ions in relation to the one recorded for NH 4 NO 3 (which is done routinely during in-field calibrations of the instrument). Using such normalized ratios, most field and

145
(2013), interpolating linearly between pure ammonium nitrate and organic nitrate. It should be noted that a) the relative ionization efficiency (RIE) for both types of nitrate is assumed to be the same (since similar neutrals are produced) and b) that the organic part of the molecule will be quantified as OA in the AMS. Therefore, while only equivalent NO 2 pRONO 2 can be reported from AMS measurements, this makes the technique well suited for comparison with the TD-LIF method. These 150 measurements correlate well to one another, but the magnitudes differ by a factor of approximately 2-4 for unknown reasons, with TD-LIF being larger than AMS (see Supplemental Information).
Boundary layer height was measured using a CHM 15k-Nimbus and method employs photon counting of back-scattered pulse of near-IR light (1064 nm During the SOAS campaign, we monitored reactant and product species indicative of NO 3 + BVOC, which may partition into the aerosol phase and consequently serve as a source of first generation SOA. NO 3 reaction with biogenic alkenes forms organic nitrates (R1).
NO 3 and N 2 O 5 (which exists in equilibrium with NO 2 + NO 3 ) in the region were consistently low 185 during the campaign. The resulting NO 3 mixing ratio was below the detection limit of the cavity ringdown instrument (1 pptv) for the entire campaign. Calculated steady-state N 2 O 5 was validated against observed measurements (see below) and NO 3 predicted from the steady-state approximation was used for all calculations involving NO 3 radical mixing ratios. Using the rate constant for NO 2 + O 3 (Table 1), the production rate of the nitrate radical (P(NO 3 )) is given by: The calculated loss rate of NO 3 , τ (NO 3 ), to reactions with individual BVOC, NO and photolysis.
(j NO3 , modeled for clear sky from MCM, (Saunders et al., 2003)) is j NO3 values were calculated from solar zenith angles and NO 3 photolysis rates (Saunders et al., 195 2003). The values were then adjusted for cloud cover by taking measured solar radiation values (Atmospheric Research and Analysis, Inc., W/m 2 ) and normalizing their peak values to those of the modeled photolysis data. Peak modeled j NO3 values were 0.175 s −1 for clear sky at the daily solar maximum. After normalizing, typical values of j NO3 were 0.110 s −1 during daytime.
Using equations 1 & 2, a steady-state predicted NO 3 mixing ratio (NO 3,SS ) can be calculated: NO 3,SS can then be used to calculate steady-state predicted N 2 O 5 from the N 2 O 5 equilibrium (Table   1) and measured NO 2 Sander et al., 2011, see Table 1). Com-205 parison of the predicted N 2 O 5 to the measured N 2 O 5 mixing ratios for the campaign demonstrates that both timing and magnitude of predicted N 2 O 5 peaks match observations ( Figure S1). Predicted steady-state N 2 O 5 tracked observations when the latter were available and propagation of the error of calculated N 2 O 5 shows peak measured values fall within uncertainty bounds of the predicted ( Figure S2a); therefore, NO 3,SS is hereafter used as the best estimate of NO 3 to calculate production 210 rates of BVOC-nitrate products. NO 3,SS peaks at 1.4 ppt ± 0.4 ppt. Propagation of errors in rate constants in the NO 3,SS calculation ( Figure S2b) shows that the error spans or is close to a mixing ratio of 0 for NO 3 during the entire campaign when data was available. Correlation of measured N 2 O 5 vs predicted N 2 O 5 shows that during periods of high N 2 O 5 , we overestimate the concentration by a factor of two ( Figure S1). Furthermore, propagation of error in the NO 3,SS calculation ( Figure S2b) 215 shows that the error encompasses the measured NO 3 during the entire campaign when data was available, showing that predicted NO 3,SS is consistent with the lack of detection of NO 3 by CRDS.
A substantial fraction (30-45%) of the NO y budget is comprised of organic nitrates (ΣANs + ΣPNs, Figure S3). Measurements of gas phase and aerosol phase alkyl nitrates show that a substantial fraction of the organic nitrates are in the aerosol phase (30% when aerosol phase AMS is 220 compared to TD-LIF total ΣANs vs 80% when comparing aerosol phase ΣANs to TD-LIF total ΣANs at 5 am CDT) when total ΣANs concentration builds up (Figures 2 & 3). The average diurnal cycle shown in Figure 3 also shows that TD-LIF measured ΣANs are almost completely in the aerosol phase at night, but only about 50% in the aerosol phase during the day. During peaks in NO 3,SS , we see corresponding spikes in the ΣANs concentrations as well as organic nitrate concen-225 tration from AMS, all of which occur during nighttime periods ( Figure 2). This is consistent with organic nitrates formed by NO 3 + BVOC rapidly partitioning into the aerosol phase.
BVOC measurements show large mixing ratios of isoprene throughout the entire campaign (daytime peaks above 8 ppb), followed by αand β-pinene (peak nighttime mixing ratios of 0.5-1 ppb, Figure S4). Using the measured mixing ratios of VOC, and their reaction rates with NO 3 , predicted 230 NO 3 losses are calculated and compared to organic nitrate aerosol. Figure 4 shows the diurnally averaged NO 3 losses for the entire campaign period (June 1 -July 15, 2013). Daytime losses include photolysis and reaction with NO. Approximately half the daytime losses are due to reaction of NO 3 with BVOC (Note, this does not necessarily imply that NO 3 reaction is a substantial loss process from the perspective of BVOC; during the day, P(HO x ) exceeds P(NO 3 ) by a factor of 10-70 235 at SOAS, so OH will typically dominate.) However, from the standpoint of NO 3 lifetime, previous forest campaigns have assumed NO 3 + monoterpene reactions to be important only during the night and that photolysis and NO losses were the dominant NO 3 sinks during the day (Geyer et al., 2001;Warneke et al., 2004). In this study, we predict significant losses of NO 3 to isoprene and monoterpenes during daylight hours.

240
To assess heterogeneous losses of N 2 O 5 to particles, an uptake rate coefficient of N 2 O 5 into deliquesced aerosols is estimated using PM surface area (S A , nm 2 /cm 3 ), the molecular speed of Conditions of high relative humidity in the SEUS necessitated a higher γ of 0.02 as the uptake uptake to be very small over the campaign despite high relative humidity. When PM 2.5 concentration was at its highest in mid-July, the calculated uptake rate coefficient was calculated at 1.6 ×10 −3 s −1 in mid July, representing less than 1% of the loss of NO 3 . 250 3.1.1 Calculation of NO 3 loss to BVOC Using literature NO 3 + BVOC rate coefficients and calculated NO 3,SS , we calculate instantaneous NO 3 loss rates ((NO 3,loss ) inst ) for the campaign.
BVOC mixing ratios from GC-MS and rate constants shown in Table 1 were used to calculate the 255 time-integrated nitrate loss to reactions with BVOC.
Specifically, time loss of NO 3 radical to reaction with BVOC ((NO 3,loss ) integ ) were calculated dur- Under the assumption of a constant nighttime boundary layer height and an approximately uniform, area wide source that limits the time rate of change due to horizontal advection (i.e., a nighttime box), the time integrals of RONO 2 produced provide estimates of the evolution of RONO 2 concentrations at night (this assumption was verified using CO to minimize first order effects of dilution from changes in the boundary layer (Blanchard et al., 2011)). Time periods of CIMS RONO 2 or 270 aerosol buildup were chosen to determine time intervals for calculation of (NO 3,loss ) integ when data was available.
(NO 3,loss ) integ is the calculated time integral of the reaction products of NO 3 with individual or combined mixing ratios of BVOC and ∆t is the time step between each calculated value of (NO 3,loss ) inst,i . Data are averaged to 30 minute increments, a time step sufficient to resolve the 275 observed rate of change. Figure 5 shows an example of the resulting calculated integrated NO 3 losses from (7) to both isoprene and summed monoterpenes. These nightly loss values are correlated with organic nitrate gas-and aerosol-phase measurements and linear fits and correlation coefficients were calculated to aid in the interpretation of gas-and aerosol-phase organic nitrate formation. Note that these peak times occur during nighttime hours when the boundary layer is shallow ( Figure S5). To derive an estimated SOA mass yield from these correlations, we propose the following rough 295 calculation. Conversion of the reported molar yield to an SOA mass yield requires assuming 1:1 reaction stoichiometry of NO 3 with monoterpenes (MW = 136 g mol −1 ) and estimating the average molecular weight (250 g mol −1 ) of the condensing organic nitrates. Using the range of molar yields determined here (23-44%), this conversion gives an SOA mass yield range from 42% to 81%. These apparent aggregated yields of SOA from NO 3 + monoterpene are higher than one might expect 300 from laboratory-based yields from individual monoterpenes, particularly since NO 3 + α-pinene SOA yields are essentially zero (Fry et al., 2014;Hallquist et al., 1999;Spittler et al., 2006) and α-pinene average molecular weight of the condensing species is unknown (we do not include sesquiterpene oxidation products and higher molecular weight BVOC products as reported by Lee et al. (2015), with which we would calculate larger SOA mass yields), this comparison is not straightforward, but it appears that the aggregate SOA yield suggests higher ultimate SOA mass yields than simple chamber experiments dictate, perhaps suggesting that post-first generation products create more condensable 310 species.
Since nitrate product buildup occurs over multiple hours ( Finally, because this yield is based on total ambient monoterpene concentrations, it incorporates nitrate radical loss to α-pinene, which is known to produce very modest yields of SOA (0-10%) from 320 NO 3 reaction (Fry et al., 2014;Spittler et al., 2006). This suggests effective overall SOA yields from other BVOC must be large.

Organic Nitrate Product Analysis
Observations of NO 3,SS compared to TD-LIF and AMS ( Figure 2) suggest aerosol organic nitrates are dominated by nighttime NO 3 + BVOC, rather than other known nitrate-producing reactions (e.g.

325
RO 2 + NO), which would dominate during the daytime and would not coincide with peaks in [NO 3 ].
Researchers at University of Washington describe the observation of particle phase C 10 organic nitrate concentrations peaking at night during SOAS (Lee et al., 2015), consistent with high SOA yield from NO 3 + monoterpenes. Observed C 10 organic nitrates include many highly oxidized molecules, suggesting that substantial additional oxidation beyond the first-generation hydroxynitrates occurs 330 (Lee et al., 2015). Specific first generation monoterpene organic nitrate compounds were identified and measured in the gas-and aerosol-phases (Lopez-Hilfiker et al., 2014;Beaver et al., 2012). Using the (NO 3,loss ) integ calculations, another correlation analysis is conducted to identify key gasand aerosol-phase products of NO 3 oxidation. Observed buildups in gas-and aerosol-phase organic nitrate concentrations from each CIMS are scattered against predicted (NO 3,loss ) integ to monoterpenes ( Figure 7). The generally good correlations suggest that all of the molecular formulae shown here have contributions from NO 3 oxidation. Comparisons of observed R 2 values and slopes for each of these correlation plots may then provide some mechanistic insight. For example, the species with larger R 2 (C 10 H 17 NO 5 ) may indicate a greater contribution to these species from NO 3 radical chemistry. If we assume the same sensitivity across phases in the cases where the same species is 340 observed (Figure 7a/b and d/e), we can estimate the relative amount in each phase by the ratio of the slopes. This would suggest that C 10 H 15 NO 5 partitions more to the particle phase than C 10 H 17 NO 5 .
Although the gas phase monoterpene nitrate product correlations display substantial scatter, likely due to their multiple possible sources and rapid partitioning to the aerosol phase, we can use the calibrated mixing ratios measured by the CIT-CIMS to calculate approximate lower limit molar 345 yields for C 10 H 15 NO 5 (0.4%), C 10 H 17 NO 5 (3%), and C 10 H 17 NO 4 (3%) from NO 3 , based on the slope of correlations shown in panels c, f and h. We estimate these to be lower limits, because no losses of these species during the period of buildup is taken into account in this correlation analysis.
The median particulate fraction of C 5 H 9 NO 5 (particle phase/total) observed by the UW-CIMS was less than 1%, and C 5 H 9 NO 5(p) comprised less than 1% of total particulate organic nitrate (Lee 350 et al., 2015). Those C 5 species that are observed in the particle phase constitute less than 12% of total particulate organic nitrate mass (as measured by the UW-CIMS, Lee et al., 2015 Supplemental Information), and are more highly oxidized molecules, inconsistent with first-generation NO 3 + isoprene products. This suggests that most (especially first-generation) isoprene nitrate products remain in the gas phase. The correlation of gas phase first-generation isoprene nitrate concentrations 355 with NO 3 loss again provides evidence about the oxidative sources of these molecules (Figure 8).
C 5 H 9 NO 5 (panels a and b) shows the strongest correlation with (NO 3,loss ) integ to isoprene among all the individual molecules (R 2 = 0.54 for UW and 0.70 for CIT), suggesting that this compound is a product of NO 3 oxidation. The better correlations of these C 5 species than observed in Figure 7 may be due to slower gas phase losses of organic nitrates relative to the semi-volatile C 10 species.

360
Using the calibrated mixing ratios from CIT for C 5 H 9 NO 5 , we calculate an approximate lower limit molar yield of 7%. The C 5 H 9 NO 4 and C 4 H 9 NO 5 isoprene products (panels c and d) show poorer correlation with (NO 3,loss ) integ to isoprene (R 2 =0.11 and 0.35, respectively), suggesting that these products are not (exclusively) a NO 3 + isoprene product, and may instead be a photochemically or ozonolysis produced organic nitrate, via RO 2 +NO. 365 We note that the two CIMS for which data is shown in Figures 7 and 8  Partitioning of semivolatile ammonium nitrate into aerosol represented a small fraction of aerosol contribution throughout the campaign based on AMS and MARGA data (Allen et al., 2015). A more important route of NO x conversion to nitrate aerosol occurred via HNO 3 heterogeneous reaction on the surface of dust or sea salt particles (Scheme 1). This process, which was observed to be especially 375 important during periods of high mineral or sea salt supermicron aerosol concentrations, is described in detail in a companion paper (Allen et al., 2015). Briefly, we observe that while concentrations of organic and inorganic nitrate aerosol are generally comparable ( Figure S3 and Figure 3), the inorganic nitrate is more episodic in nature. Periods of highest NO − 3 concentration as measured by the MARGA were observed during two multi-day coarse-mode dust events, from June 9 to 15 and 380 June 23 to 30, while organic nitrates have a more regular diurnal pattern indicative of production from locally-available reactants, with most of the organic nitrate present in the condensed phase ( Figure 3).
In order to estimate the fluxes of NO x loss to aerosol via the two pathways shown in Scheme 1, we calculate the reactive losses of NO 2 to organic nitrate (limiting rate is taken to be with the included terpenes α-pinene, β-pinene, limonene and camphene) and to inorganic nitrate via heterogeneous HNO 3 uptake (Allen et al., 2015). A substantial fraction of the suface area is in the transition regime, so HNO 3 uptake is reduced due to diffusion limitations. To account for this, a Fuchs-Sutugin correction is applied (Seinfeld and Pandis, 2006): with S a is surface area, R p is the radius, D g is the diffusivity of HNO 3 in air (0.118 cm 2 s −1 ) and α is estimated at 0.1 for an upper limit.
Since we have seen that the organic nitrates are present predominantly in the condensed phase, we take this comparison to be the relative rate of production of organic nitrate aerosol vs. inorganic nitrate aerosol (Figure 9), and we see that over the summer campaign, the rates are comparable in 395 magnitude, but peak at different times. This analysis suggests that substantial nitrate aerosol (peak values of 1 µg m −3 hr −1 , with average rates 0.1 µg m −3 hr −1 for both inorganic and organic nitrate rates) is produced in the SEUS by both inorganic and organic routes (depicted in Scheme 1), converting local NO x pollution to particulate matter. We note that this calculation accounts only for the production rates of these two types of nitrate aerosol and does not account for subsequent chemistry 400 that may deplete one faster than the other; hence, relative mass concentrations are not necessarily expected to correlate directly to these relative production rates.

Implications of NO 3 oxidation on SOA formation in the SEUS
The importance of the NO 3 + BVOC reaction SOA has only recently been recognized (Beaver et al., 2012;Fry et al., 2013;Rollins et al., 2012). Pye et al. (2010) showed that including NO 3 radical 405 oxidation increased predicted SOA yields from terpenes by 100% and total aerosol concentrations by 30% (Pye et al., 2010). The results of this study underscore the importance of NO 3 in SOA 0.8 Tg yr −1 (9 × 10 5 tons yr −1 ) of PM 2.5 in 2011. We can estimate the fraction of the NO x emitted that is converted to PM using several assumptions. NO 2 is estimated to contribute 50% of the NO y budget ( Figure S3), so we multiply the NO x emission by 0.5 to account for half of the instantaneous NO x residing in the atmosphere as other NO y species at any given time. An average lifetime of 16 hours for O 3 + NO 2 reaction was calculated (1/k[O 3 ]) and, with an average nighttime length of 420 9 hours, we estimate about 55% of NO 2 is converted to NO 3 overnight. Using the average molar organic nitrate aerosol yield of 30% determined in this study and an estimated molecular weight of 250 g mol −1 for oxidized product (terpene hydoxynitrate with two additional oxygen functional groups, Draper et al., 2015), we convert from molar yield to mass yield of organic nitrate aerosol.
This assumes that NO x is the limiting reagent for SOA production from this chemistry; as noted in 425 the introduction, comparison of regional NO x and BVOC emissions rates supports this assumption.
Finally, using the summed NEI NO x emissions data for the SAMI states, we calculate a source estimate of 0.6 Tg yr −1 of NO 3 -oxidized aerosol. Adding this to the NEI primary PM 2.5 emissions estimate of 0.8 Tg yr −1 gives a total 1.4 Tg yr −1 , showing that NO 3 initiated SOA formation would contribute a substantial additional source of PM 2.5 regionally, nearly doubling primary emissions.

Conclusions
The contribution of NO 3 + BVOC to SOA formation is found to be substantial in the terpenerich SEUS. An estimated 23-44% of nitrate radical lost to reaction with monoterpenes becomes 440 aerosol phase organic nitrate. Predicted nitrate losses to isoprene and to monoterpenes are calculated from the steady-state nitrate and BVOC mixing ratios and then time integrated during evenings and nights as RONO 2 aerosol builds up. Correlation plots of AMS, TD-LIF, and CIMS measurements of gas-and aerosol-phase organic nitrates against predicted nitrate losses to monoterpenes indicate that NO 3 + monoterpenes contribute substantially to observed nitrate aerosol. Two specific C 10 445 structures measured by CIMS are shown to be NO 3 radical products by their good correlation with cumulative (NO 3,loss ) integ ; their semi-volatile nature leads to their variable partitioning between gas-and aerosol-phase. Calibrated gas phase mixing ratios of selected organic nitrates allow estimation of lower limit molar yields of C 5 H 9 NO 5 , C 10 H 17 NO 4 , C 10 H 17 NO 5 from NO 3 reactions (7%, 3%, and 3% respectively). The fact that these molar yields of monoterpene nitrates are substantially 450 lower that the aggregated aerosol phase organic nitrate yield may suggest that further chemical evolution is responsible for the large SOA yields from these reactions, consistent with Lee et al. (2015).
The NO 3 + BVOC source of nitrate aerosol is comparable in magnitude to inorganic nitrate aerosol formation, and is observed to be a substantial contribution to regional PM 2.5 .  β-pinene + NO3 → Products 2.51 × 10 −12 Calvert et al. (2000) Camphene + NO3 → Products 6.6 × 10 −13 Calvert et al. (2000) Myrcene + NO3 → Products 1.1 × 10 −11 Calvert et al. (2000) Limonene + NO3 → Products 1.22 × 10 −11 Calvert et al. (2000) * Reaction rate constants are reported as: k(T) = A e −(Ea/R)/T , in units of (cm 3 molecule −1 s −1 ) ** Equilibrium constants are reported as: K eq = A e B/T , in units of (cm 3 molecule −1 ) Scheme 1 Generalized reaction fate for NO 2 in the troposphere. Oxidation of NO 2 from atmospheric oxidants leads to two possible paths.     range from 2% for myrcene to -40% for β-pinene (Calvert et al., 2000). Uncertainties in rate constants of BVOC + NO3 range from ± 30% for α-pinene to up to a factor of two for isoprene (Calvert et al., 2000); NO measurements had ± 35% uncertainty, BVOC measurements ± 20%, and photolysis ± 20% based on solar radiation measurement uncertainty.   Figure 8. Gas phase CIMS data correlated to predicted isoprene + NO3, during periods of buildup of these C5 and C4 nitrates as measured by each CIMS. Panels a & b show C5H9NO5, which is well correlated to predicted isoprene + NO3 suggesting this is a NO3 gas phase product, with the calibrated mixing ratios measured by CIT enabling estimation of an approximate lower limit molar yield of 7%. Panel c shows that C5H9NO4 is poorly correlated to isoprene + NO3 suggesting that this product comes (at least in part) from another oxidative source (e.g. RO2+NO). Panel d, C4H7NO5, also shows a poorer correlation than panels a & b, suggesting it is not exclusively a product of NO3 oxidation, or has rapid losses.