Anthropogenic enhancements to production of highly oxygenated molecules from autoxidation

Significance

Government organizations set standards for permissible levels of atmospheric particulate matter (PM) due to its adverse effects on human health and the environment. Unimolecular reactions that efficiently produce organic PM are suppressed by NO_x, allowing for potential increases in PM mass due to controls on anthropogenic NO_x. Using laboratory experiments and observations of the atmosphere, we assemble a conceptual understanding of how PM from unimolecular organic reactions is affected by NO_x. We demonstrate that the NO_x penalty on PM yield can be offset by reductions in oxidant abundance. As a result, PM from unimolecular reactions is predicted to decrease as NO_x is controlled, consistent with declines in ambient organic aerosol observed in the United States between 1990 and today.

Abstract

Atmospheric oxidation of natural and anthropogenic volatile organic compounds (VOCs) leads to secondary organic aerosol (SOA), which constitutes a major and often dominant component of atmospheric fine particulate matter (PM_2.5). Recent work demonstrates that rapid autoxidation of organic peroxy radicals (RO₂) formed during VOC oxidation results in highly oxygenated organic molecules (HOM) that efficiently form SOA. As NO_x emissions decrease, the chemical regime of the atmosphere changes to one in which RO₂ autoxidation becomes increasingly important, potentially increasing PM_2.5, while oxidant availability driving RO₂ formation rates simultaneously declines, possibly slowing regional PM_2.5 formation. Using a suite of in situ aircraft observations and laboratory studies of HOM, together with a detailed molecular mechanism, we show that although autoxidation in an archetypal biogenic VOC system becomes more competitive as NO_x decreases, absolute HOM production rates decrease due to oxidant reductions, leading to an overall positive coupling between anthropogenic NO_x and localized biogenic SOA from autoxidation. This effect is observed in the Atlanta, Georgia, urban plume where HOM is enhanced in the presence of elevated NO, and predictions for Guangzhou, China, where increasing HOM-RO₂ production coincides with increases in NO from 1990 to 2010. These results suggest added benefits to PM_2.5 abatement strategies come with NO_x emission reductions and have implications for aerosol–climate interactions due to changes in global SOA resulting from NO_x interactions since the preindustrial era.

Ambient particulate matter is responsible for a significant fraction of the global burden of disease (1) and affects climate (2). The organic portion of ambient aerosol forms largely due to the condensation of low vapor pressure (3) or highly soluble (4) gases stemming from atmospheric oxidation of volatile organic compounds (VOCs). During summer, emissions of biogenic VOCs exceed those from anthropogenic activities (5) and monoterpenes are a major class of biogenic VOCs emitted throughout the year (6). Oxidation of monoterpenes results in significant formation of particle mass (7) and is a major source of secondary organic aerosol (SOA) in the southeastern United States (8, 9). While monoterpenes are included as an SOA source in most chemical transport models, many parameterizations (10–12) lack a mechanistic dependence of monoterpene SOA on NO_x and oxidant identity (e.g., OH vs. ozone), and all lack unimolecular RO₂ reactions. This limits the ability of models to predict how changes in emissions will affect ambient concentrations and new particle formation events.

Building a mechanistic understanding of SOA formation pathways is critical for being able to determine historic and future drivers of ambient particulate matter, particularly as the abundance of species, such as nitrogen oxides (NO_x = NO + NO₂), that modulate the efficiency of aerosol formation (8, 13) continues to change. For biogenic VOCs such as isoprene and monoterpenes, a mechanistic description of oxidation can inform what is considered a natural source of particle mass and what is anthropogenically controlled. Furthermore, the extrapolation of laboratory results to long-term organic aerosol trends may not be reliable without a mechanistic understanding of organic aerosol pathways. This issue was recently illustrated for the isoprene system in which a robust, mechanistic representation of SOA based on later generation isoprene epoxydiols resulted in a significantly different magnitude of (11, 14) and trend in (15) isoprene SOA compared with one predicted using an older empirical parameterization without mechanistic underpinnings. Although new information on the monoterpene aerosol system is emerging, it has not yet been incorporated into a model-ready mechanism to allow for similar analyses.

RO₂ autoxidation, a unimolecular reaction involving one or more hydrogen shifts and subsequent addition of molecular oxygen, occurs for many hydrocarbon–oxidant systems including OH- and ozone-initiated reactions of VOCs and oxidized VOCs of biogenic and anthropogenic origin (16–22). These intramolecular reactions can proceed rapidly (in seconds), competing with or even outpacing bimolecular reactions with NO or HO₂ (17). Autoxidation can result in highly oxygenated molecules (HOM; O:C ≥ 0.7) after only one reaction with OH or ozone (23). HOM often have low saturation vapor concentrations (C*) and can be classified as extremely low volatility organic compounds (ELVOC; C* < 3 × 10⁻⁴ μg m⁻³) or low volatility organic compounds (LVOC; 3 × 10⁻⁴ < C* < 0.3 μg m⁻³) (24). ELVOC can contribute significantly to new particle formation, and ELVOC, LVOC, and semivolatile organic compounds (SVOC) can contribute to particle mass growth (25, 26).

Initial estimates of HOM from α-pinene monoterpene oxidation indicated that ozonolysis produced these species in greater yield (1.7–6.8% by mole) than oxidation by OH (yields of 0.22–0.88% by mole) (27). More recently, reagent ion-dependent detection efficiencies of HOM were recognized and the HOM yield from α-pinene + OH oxidation is now estimated to be ≥2.4% by mole (20). The resulting role of autoxidation in the α-pinene + OH system in producing SOA mass could be significant as monoterpene reaction with OH is estimated to account for just over half of the daytime monoterpene fate in the southeastern US atmosphere (13), and α-pinene is the most abundantly emitted monoterpene (6, 10). However, NO_x is expected to suppress monoterpene SOA by inhibiting autoxidation (17).

This work reveals that representing autoxidation in models is key to predicting correct amounts of anthropogenic controls on monoterpene SOA. To demonstrate this, a molecular mechanism-based analysis of the coupling between autoxidation, OH abundance, and NO_x was used to describe how NO_x influences prompt HOM formation in the α-pinene + OH system (Materials and Methods). Recent laboratory fast-flow reactor and environmental simulation chamber experiments were used to develop and constrain the detailed molecular mechanism of autoxidation pathways. Insights from the mechanism-laboratory comparisons were used in conjunction with aircraft observations and the Community Multiscale Air Quality (CMAQ) chemical transport model to understand the role of autoxidation and potential for SOA in different chemical environments across the Northern Hemisphere from 1990 to the present day.

Results and Discussion

Mechanistic HOM-RO₂ Formation and Yields.

A critical parameter in the description of HOM formation is the autoxidation rate constant, k_autox, which sets the rate at which α-pinene + OH-derived RO₂ (C₁₀H₁₇O₃) are converted to HOM−RO₂ (peroxy radicals with O:C ≥ 0.7). This rate constant also determines how competitive autoxidation is with traditional bimolecular RO₂ fates such as reaction with NO or HO₂ that do not produce HOM as efficiently. Autoxidation in the α-pinene + OH system, based largely on the work of Berndt et al. (20), is illustrated in Fig. 1. Reaction of α-pinene, at a rate governed by OH oxidant abundance, results in several C₁₀H₁₇O₃ peroxy radical isomers. One of the initial C₁₀H₁₇O₃ isomers can undergo autoxidation to C₁₀H₁₇O₅ peroxy radicals. These peroxy radicals can undergo an additional autoxidation step, also dictated by k_autox, to produce C₁₀H₁₇O₇ HOM-RO₂. When HOM-RO₂ react with HO₂, C₁₀H₁₈O₇ LVOCs are formed and partition almost entirely toward the particle resulting in SOA. At each opportunity for autoxidation, reaction with NO or other peroxy radicals can divert mass away from the path to HOM-RO₂ formation. Simulations of flow tube experiments at short residence times (7.9 s) (20) with the mechanism developed in this work resulted in predicted HOM-RO₂ yields consistent with observations for a k_autox of 0.28 s⁻¹ (20). Specifically, box model simulations using this recommended value resulted in a predicted HOM-RO₂ yield at 7.9 s of reaction time of 3.3% (Table 1 and SI Appendix, Fig. S3), consistent with the observed lower-bound yield of 2.4% (20). As the observed yield is a lower bound, limited by the ability to detect HOM, a higher, but not lower k_autox than the default may also be possible (28) (SI Appendix, Table S7).

Fig. 1.

Schematic of major α-pinene + OH autoxidation pathways leading to HOM SOA. Only one isomer is shown even if many are possible. See SI Appendix, Figs. S1 and S2 and Tables S1–S4 for a complete representation of the mechanism developed in this work.

Table 1.

Predicted and observed properties of the α-pinene OH-initiated system

	YHOM-RO2	YSOA	SOA O:C	SOA nC	SOA C*
	% by mole	% by mass	mol mol−1	-	μg m−3
MCM v3.3.1	0	<1	0.59	8	104
Regional CTM	0	7	0.52	8	55
HOM mechanism	3.3	17	0.64	10	1.3
Observed	≥2.4	12 ± 4.6	0.68	9	0.08

Gas-phase HOM-RO₂ molar yield (Y_HOM−RO2) at 7.9s for Berndt et al. (20) flow tube conditions as well as SOA mass yield (Y_SOA), mass-weighted mean molar O:C, mass-weighted mean number of carbons per molecule (nC), and mass-weighted effective C* of SOA for OH-dominated SOAFFEE laboratory chamber conditions. Regional chemical transport model (CTM) predictions are based on CMAQv5.2 and OH and O₃ oxidation of monoterpenes (11) at the SOAFFEE-predicted loading of 7 μg m⁻³.

The yield of HOM-RO₂ (assuming pseudo steady state for each peroxy radical) in the absence of bimolecular RO₂ reactions was predicted to be 11%, significantly higher than the few percent observed and predicted at short reaction times. The fraction of α-pinene + OH-derived RO₂ that undergo autoxidation to produce HOM-RO₂, Y_HOM-RO2, can be expressed as:

$$Y_{HOM-R{O}_2}=f{\left(\frac{k_{autox}}{k_{autox}+k_{R{O}_2+NO}\left[NO\right]+k_{R{O}_2+H{O}_2}\left[H{O}_2+R{O}_2\right]}\right)}^2$$ [1]

for steady state conditions and is determined by the assumed fraction of OH-initiated peroxy radicals able to undergo intramolecular H-abstraction reactions that do not self-terminate (f = 11%) (20, 29) as well as the competition with bimolecular reaction (dictated by the autoxidation rate constant). The maximum molar yield of HOM-RO₂ from α-pinene + OH (11%; Fig. 2A and SI Appendix, Fig. S1 and Eq. S1) corresponds to a mass yield of 20% due to the addition of ∼7 oxygens, and the factor of 1.8 increase in molecular weight of C₁₀H₁₈O₇ compared with α-pinene (C₁₀H₁₆). As a result, autoxidation induced by OH reaction with α-pinene is expected to be important for particle mass formation.

Fig. 2.

(A) Molar yield of C₁₀H₁₇O₇ HOM-RO₂ (Y_HOM−RO2) from α-pinene + OH predicted by the updated explicit mechanism under laboratory (SOAFFEE) and atmospherically relevant conditions. Stars (A) represent conditions from the 2013 southern oxidant and aerosol study during the daytime and morning (AM) transition at Centreville, Alabama. The Atlanta plume NO level is based on SENEX aircraft observations (Fig. 3). (B) The Y_HOM−RO2 (black) and production rate of HOM-RO₂ (P_HOM−RO2) (red) for the OH concentration shown in blue following Thornton et al. (52) methods. Yields and production rates in B are shown for default mechanism parameters (k_autox = 0.28 s⁻¹, f = 11%, solid) and parameters corresponding to a lower bound yield (k_autox = 2.8 s⁻¹, f = 2.7%, dashed).

Inclusion of autoxidation in the chemical mechanism used to describe α-pinene oxidation allowed for vastly improved predictions of SOA mass, composition, and volatility in comparison with that observed during chamber studies of α-pinene + OH at low NO_x from the Secondary Organic Aerosol From Forest Emissions Experiment (SOAFFEE; Table 1). Without HOM-RO₂ from autoxidation, an explicit mechanism based on the standard master chemical mechanism (MCM) (30) did not predict meaningful amounts of SOA (mass yield <1%), consistent with previous attempts to use explicit, compound-specific representations of SOA for this system (31). In addition, the SOA that did form based on the standard MCM showed a lower degree of functionalization (O:C = 0.59) and significantly higher volatility (C* of 104 μg m⁻³) than that observed during SOAFFEE [O:C of 0.68 and effective C* of the peak in thermogram signal corresponding to ∼0.08 μg m⁻³ with an order of magnitude uncertainty (32)] (Table 1). The empirical SOA parameterization used in the current (v5.2) CMAQ regional chemical transport model showed similar shortcomings in terms of underestimated yields, low degree of functionalization, and high SOA volatility. This high volatility in CMAQ algorithms implies that monoterpene SOA in models is sensitive to primary organic aerosol emissions via absorptive partitioning, and the low volatility of the observed aerosol implies monoterpene SOA is not strongly influenced by semivolatile partitioning.

Results from the updated mechanism indicated that prompt SOA formation from the α-pinene + OH system, at or above 285 K, is essentially entirely dependent upon autoxidation. The autoxidation mechanism developed here predicted an SOA yield of 17% consistent with the observed yield (12 ± 4.6%) and used an internally consistent representation of molecular composition with lower volatility (mass-weighted C* ∼1 μg m⁻³), which has not typically been achieved with explicit chemical mechanisms (30, 33–37). Furthermore, both observations and the updated mechanism showed significant amounts of species with 10 carbon backbones and high O:C (0.5–0.7) in the particle (SI Appendix, Fig. S5). Although sufficient time for autoxidation existed in the 3.6-h chamber residence time of SOAFFEE, predicted HOM-RO₂ yields were 8% (Fig. 2A) instead of the theoretical maximum of 11% due to the presence of elevated levels of HO₂ radicals (∼100 ppt) that competed with autoxidation through bimolecular reactions. Even if a lower bound on autoxidation yield was assumed (f = 2.65% and k_autox = 2.8 s⁻¹; SI Appendix, Table S7), the autoxidation mechanism still accounted for 29% (range: 21–47%) of the SOA formation during SOAFFEE, further indicating a substantial role for autoxidation.

Atmospheric Implications.

As α-pinene is a major component of monoterpene emissions in midlatitude regions (6), and of SOA even in isoprene dominated regions (8), this system is a proxy for the behavior of a large fraction of biogenic SOA and potentially for other systems that can undergo autoxidation. The ability of the detailed autoxidation mechanism developed here to reproduce several features of α-pinene + OH system observed in the laboratory broadly supports the combination of mechanistic parameters and allows for an assessment of autoxidation-derived SOA responses to anthropogenic perturbations on local and regional scales. Here, we demonstrate that autoxidation results in aerosol that is enhanced in the presence of NO_x, is prevalent in the current ambient atmosphere, and produces aerosol that is mitigated by controls on anthropogenic NO_x.

HOM formation in the presence of NO_x.

Production rates of HOM-RO₂ scale with the yield of HOM-RO₂ (Eq. 1) and the availability of oxidants, allowing for HOM concentrations to increase with increasing NO_x. Given that HOM are formed promptly at significant yields from the reaction of α-pinene + OH, the production rate of autoxidation-derived HOM, and therefore SOA, can be approximated on short timescales (hours) as:

$$P_{HOM-R{O}_2}=k_{\alpha -pinene+OH}\left[\alpha -pinene\right]\left[OH\right]Y_{HOM-R{O}_2}.$$ [2]

Based on the mechanism and combination of k_autox and f evaluated in this work, we find that although elevated NO_x monotonically suppresses the relative importance of autoxidation as an RO₂ fate (Fig. 2A and Eq. 1) as expected (17), the well-known nonlinear effect of NO_x on OH radical concentrations (38) offsets this suppression and leads to enhancements in the absolute formation rate of HOM (Fig. 2B and Eq. 2) as NO_x increases. These enhancements in absolute HOM-RO₂ production will translate into localized enhancements in biogenic SOA due to the low volatility of resulting products and role of autoxidation in SOA formation demonstrated in this work. The nonlinear response of OH concentrations to increasing NO_x does lead to a turnover point (∼1 ppb of NO in Fig. 2B), above which, further increases of NO suppress OH and therefore also local HOM production. However, unless NO_x is very high (>5 ppb in Fig. 2B), the production of HOM-RO₂ is facilitated by NO_x in monoterpene-rich regions. The NO concentration at which HOM production peaks is different from that where OH concentrations peak and is ultimately a function of the autoxidation rate constant with higher k_autox values (such as k_autox = 2.8 s⁻¹) resulting in a yield penalty from NO_x emission reductions (Eq. 1) that is easier to overcome and peak HOM-RO₂ production at an NO level closer to that of peak OH production.

Observations in the Atlanta urban plume (Fig. 3) illustrate facilitation of HOM in the presence of elevated NO. Gas-phase concentrations of C₁₀H₁₈O₇ and other HOM compounds (C₁₀H_14,16,18O_7–8 and C₁₀H_15,17NO_8–9; see SI Appendix, Fig. S11), likely from monoterpene autoxidation (Fig. 1), were enhanced over regional background concentrations by a factor of 3 to 5 in the urban plume from Atlanta, Georgia, during the summer of 2013. This enhancement in HOM coincided with a depletion of monoterpenes and occurred in the presence of NO. The NO levels in the Atlanta plume (100–300 ppt) were not so high as to shut down autoxidation (Fig. 2A), and enhanced ozone in the plume suggests that the primary OH source would be enhanced as well and lead to increased rates of oxidation. If gas-particle equilibrium is assumed with a typical HOM saturation concentration of 7.3 × 10⁻³ μg m⁻³ (SI Appendix, Table S6), the measured HOM vapor concentration implies the corresponding C₁₀H₁₈O₇ SOA enhancement accounted for 30% (range: 10–40%) of the total OA increase in the Atlanta urban plume. These data provide evidence that anthropogenic enhancements to biogenic VOC derived HOM exist in the present-day atmosphere and illustrate the importance of understanding multiple roles of NO_x in the production of low volatility compounds that lead to anthropogenically controlled biogenic SOA (39).

Fig. 3.

Concentrations of NO, ozone, monoterpenes (MT), organic aerosol (OA), and gas-phase C₁₀H₁₈O₇ species measured downwind of the Atlanta, Georgia metropolitan area (A) on June 12, 2013 aboard the NOAA WP-3 aircraft. Pink vertical (B–D) shading indicates in-plume conditions characterized by enhanced NO, CO (generally >190 ppb), ozone, and organic aerosol (B and C). HOM vapor concentrations (D) are reported as 1-min averages with shading for the 25th to 75th percentile of the 1-Hz data.

Significant HOM yields due to autoxidation.

The rural southeastern United States was predicted to experience near the maximum possible fraction of RO₂ undergoing autoxidation based on observed (Fig. 2A) and CMAQ air quality model predicted (Fig. 4) radical abundances. Specifically, Fig. 2A indicates typical daytime conditions in the southeastern US atmosphere as captured by the 2013 southern oxidant and aerosol study (SOAS) in Centreville, Alabama (30 ppt HO₂, 40 ppt NO) and conditions during the morning when NO was maximum (300 ppt NO, 5 ppt HO₂) (13, 40, 41). These observed ambient conditions and those for the urban Atlanta plume resulted in predicted HOM-RO₂ yields from α-pinene + OH just under the maximum with slightly lower values during the morning transition hours due to higher NO. High resolution simulations for the present day and 20 y of simulations covering the Northern Hemisphere showed a similar result for the rural United States with molar HOM-RO₂ yields of 9% for Centreville over the past two decades (Fig. 4 A and B).

Fig. 4.

(A) Estimated molar yield of first generation C₁₀H₁₇O₇ HOM-RO₂ from α-pinene OH-initiated C₁₀H₁₇O₃ peroxy radicals (Y_HOM−RO2) for present-day conditions in the contiguous United States; (B) trend in Y_HOM−RO2 for years 1990–2010; (C) production rate of C₁₀H₁₇O₇ HOM−RO₂ (P_HOM−RO2) in the Northern Hemisphere (year 2010); and (D) trend in P_HOM−RO2 for years 1990–2010. Locations marked in A and C, and featured in the time series include: Centerville, Alabama, United States (CTR); Atlanta, Georgia, United States (ATL); Hyytiälä, Finland (FIN); Sanming, Fujian, China (SAN); and Guangzhou, Guangdong, China (GNZ). Predictions are averaged over daytime, summer conditions.

Fig. 4 also indicates variability in the predicted fraction of RO₂ converted to HOM-RO₂ via autoxidation across the United States and Northern Hemisphere as a result of the spatial distribution of NO emissions, which suppress the relative fraction of RO₂ undergoing autoxidation (Eq. 1) (17). Model calculations indicated that for the past two decades, autoxidation has been an increasingly important fraction of the RO₂ fate in the Atlanta region (Fig. 4B) because of decreasing NO (SI Appendix, Fig. S10), while Hyytiälä, Finland, experienced just under the maximum possible autoxidation (11% by mole yield of HOM-RO₂) for decades. Note that once NO levels are sufficiently low, the yield of HOM-RO₂ will not be as sensitive to the NO level (Fig. 2).

In China, the contribution of HOM to PM_2.5 is unknown, but α-pinene + OH may serve as an archetype system with lessons for anthropogenic VOCs [e.g., alkylbenzenes (21)] that also undergo autoxidation. Emissions of nitrogen oxides in China have increased since 1990 (42) and predictions for both urban (Guangzhou) and suburban (Sanming) regions showed decreasing fractions of RO₂ undergoing autoxidation. Even in Guangzhou, with its higher NO levels and bimolecular RO₂ reactions outpacing autoxidation in the present day, autoxidation was still predicted to lead to significant HOM-RO₂ yields on the order of 5% by mole (9% by mass). Monoterpene SOA from OH oxidation predicted by chemical transport models would show the opposite trend in yield with time compared to that shown here for Guangzhou and Atlanta (Fig. 4B) due to current model formulations that assume semivolatile SOA and therefore dependencies of yield on primary organic aerosol abundance (43).

NO_x emission reduction cobenefits for SOA production from autoxidation.

The dual role of NO_x in competing with autoxidation and influencing oxidant concentrations does not allow for direct extrapolation of autoxidation efficiency (e.g., HOM yields from Fig. 4B and Eq. 1) to regional SOA formation rates because absolute HOM-RO₂ production rates (P_HOM−RO2) also depend on oxidant and precursor hydrocarbon abundance (Eq. 2). The spatial distribution of predicted HOM-RO₂ production (Fig. 4C) reflects the abundance of monoterpenes and OH and illustrates that increases in OH due to higher NO_x can offset decreases in the autoxidation efficiency such that HOM-RO₂ production rates are enhanced in polluted regions otherwise rich in monoterpene emissions (Fig. 4D). Model results for urban locations (Guangzhou and Atlanta) showed locally increased OH and ozone concentrations as well as HOM-RO₂ production compared with their suburban/rural counterparts (e.g., Sanming and Centreville) due to higher anthropogenic emissions.

In addition, the influence of NO_x on oxidant abundance was found to drive the predicted 1990–2010 change in HOM-RO₂ for Guangzhou and Atlanta regions (Fig. 4D and SI Appendix, Fig. S10). Changing the mechanism yield of HOM-RO₂ able to undergo autoxidation to C₁₀H₁₇O₇ HOM-RO₂ (value of f; Eq. 1) would change the ability of autoxidation to contribute to ambient HOM-RO₂ in an absolute sense, but not in terms of trends over 20 y. The conclusion that oxidants drive the 20-y trend in HOM-RO₂ production is insensitive to the rate of autoxidation for k_autox > 0.28 s⁻¹. Taking these NO_x–oxidant–autoxidation couplings into account, we predicted localized increases in HOM-RO₂ production rates in China between 1990 and 2010 (Fig. 4D), at the same time autoxidation decreased relative to bimolecular RO₂ fates (Fig. 4B).

An oxidant-driven decrease in HOM production of about 20% was predicted over the 1990–2010 period for the Atlanta region (Fig. 4D). Other work suggests organic aerosol has declined about 40% in the United States for this same time period (43). If monoterpenes account for half of ambient organic aerosol (8), then changes in monoterpene SOA due to the NO_x–oxidant–autoxidation couplings discussed herein could be responsible for a quarter of the recent downward trend in organic aerosol in the southeastern United States. As NO levels continue to decline and reaction with HO₂ becomes a major HOM−RO₂ fate, the resulting aerosol-phase peroxides could also have implications for public health (44) even if particulate matter as a whole declines.

Conclusions

Autoxidation is an effective pathway to substantial amounts of low volatility organic compounds and recent trends indicate anthropogenic NO_x controls will be an effective way to mitigate localized enhancements in particulate matter from monoterpene autoxidation in the present-day atmosphere. Furthermore, due to the low volatility of aerosol from autoxidation, anthropogenic controls on primary organic aerosol emissions are not expected to significantly influence SOA from autoxidation, in contrast to most current chemical transport model representations of monoterpene-derived aerosol. Although changes in OH levels between preindustrial and present day remain uncertain (45), our data suggest that the autoxidation efficiency of monoterpene peroxy radicals allows for increasing NO_x to enhance biogenic SOA formation on regional scales due to the resulting enhancements in OH and ozone. Only in highly polluted urban cores do calculations suggest suppression of autoxidation by NO is significant enough to outweigh the locally enhanced monoterpene reactions that can lead to autoxidation.

Materials and Methods

Mechanism Development.

Additional pathways and products (185 reactions, 115 species) were added to MCM v3.3.1 to represent intramolecular H-shift, cyclization, and O₂ addition with a focus on α-pinene + OH chemistry leading to the formation of highly oxygenated peroxy radicals and coproducts. Autoxidation pathways followed those proposed by Berndt et al. (20) and started with the double-bond-retaining, OH-initiated C₁₀H₁₇O₃−RO₂ able to undergo H-shift as indicated by Vereecken et al. (29). The resulting C₁₀H₁₇O₅−RO₂ could further autoxidize (leading to C₁₀H₁₇O₇−RO₂). Thus, formation of HOM-RO₂ required two sequential autoxidation steps each with rate constant k_autox and resulted in a squared dependence of HOM-RO₂ yield on [NO]. k_autox represented the effective autoxidation rate constant and was applied to five different structures with O:C ratios of 0.3–0.5. Although the mechanism developed here performs well for the metrics examined (Table 1), future work should further constrain the autoxidation rate constant (including its dependence on structure and temperature) and overall product distribution for both OH and ozone reaction under a variety of laboratory and ambient conditions. The full mechanism is available in the SI Appendix.

Box Model Framework.

The mechanism developed here was implemented in F0AM v3.1 (46) with Washington Aerosol Module (WAM) extension (22). Compounds with saturation concentrations less than 500 μg m⁻³ were dynamically partitioned between the gas and aerosol phases using vapor pressures determined by EVAPORATION algorithms (47, 48).

Laboratory Data.

The F0AM-WAM model with detailed chemistry was applied to two laboratory systems: Berndt et al. (20) flow tube experiments (SI Appendix, Fig. S3) and the steady state chamber of the SOAFFEE at Pacific Northwest National Laboratory (parameters in SI Appendix, Table S5). F0AM-WAM simulations specifically focused on conditions where OH oxidation was predicted to account for 91% of the α-pinene reacted. H₂O₂ and α-pinene were continuously injected into the chamber and OH was generated by photolysis. Aerosol mass spectrometer data had a 38% uncertainty (49). A high-resolution time-of-flight chemical ionization mass spectrometer (HR-ToF-CIMS) with Filter Inlet for Gases and Aerosols (FIGAERO) utilizing iodide-adduct ionization measured gas and particle-phase composition (32). Loss of vapors and particles to walls occurred in both the observations and model for SOAFFEE.

Regional Predictions.

Specific product yields and rate constants from the mechanism were used with two sets of archived hourly oxidant and monoterpene fields from CMAQ to estimate the yield and production of HOM-RO₂. The 12 km by 12 km horizontal resolution contiguous US output for June 1–July 15, 2013, from CMAQ v5.2 (doi:10.5281/zenodo.1167892) were obtained from the work of Xu et al. (9) (Carbon Bond 6 revision 3 chemistry; 2011 EPA National Emission Inventory with year specific emissions when available). CMAQ v5.0 (doi:10.5281/zenodo.1079888) July predictions for the Northern Hemisphere from 1990 to 2010 were obtained from the work of Xing et al. (42) (Carbon Bond 5 chemistry; Emission Database for Global Atmospheric Research [EDGAR] version 4.2 emissions). All means were calculated when OH was in the top 50th percentile.

Aircraft Observations.

Observations for Atlanta were obtained from the Southeast Nexus (SENEX) field study on June 12, 2013, (SENEX flight number 3) aboard the National Oceanic and Atmospheric Administration (NOAA) WP-3 aircraft. From 11:00 to 13:30 local time on June 12, the wind was from the northwest (mean speed of 5.0 m/s). Details on the instrumentation used to generate observations can be found in Warneke et al. (50) and Lee et al. (51). CIMS signals were converted to abundance using the kinetic limit ionization sensitivity (lower limit on concentration).

Code and Data Availability.

F0AM is available at https://github.com/AirChem/F0AM.

WAM is available at https://www.atmos.washington.edu/∼thornton/washington-aerosol-module.

CMAQ is available at https://github.com/USEPA/CMAQ.

SENEX data are available at https://data.eol.ucar.edu/master_list/?project=SAS.

SOAFFEE data, SENEX HOM data, and CMAQ output have been deposited in the Environmental Protection Agency Science Hub repository (https://catalog.data.gov/harvest/about/epa-sciencehub) using doi:10.23719/1502525.