Biogenic volatile organic compounds (BVOCs) constitute the largest flux (88 %) of all non-methane organic compounds to the atmosphere (Goldstein and Galbally, 2007; Guenther et al., 1995, 2012). Isoprene and monoterpenes account for ∼ 60 % of the total global BVOC emissions (Goldstein and Galbally, 2007; Guenther et al., 1995). BVOCs participate in chemical reactions with regulated air pollutants, including NO2 and O3, and with radical species such as OH and NO3, which leads to the formation of low-volatility oxygenated compounds that partition into aerosol particles and represents a source of secondary organic aerosol (SOA) (Atkinson and Arey, 2003; Hallquist et al., 2009; Hatakeyama et al., 1991; Isaksen et al., 2009; Lee et al., 2016; Monks et al., 2009; Perring et al., 2013; Pye et al., 2015; Tuazon and Atkinson, 1990). Globally, the oxidation of BVOCs emitted from forests represents the largest source of SOA and affects climate and air quality (Hallquist et al., 2009).

One specific group of BVOC oxidation products of interest is organic nitrates (RONO2). By sequestering NOx, RONO2 slow the production of tropospheric O3, but it can act as a source of NOx and O3 downwind upon further oxidation and decomposition (Browne and Cohen, 2012; Pusede et al., 2015). These low-volatility, water-soluble compounds are important precursors to and constituents of SOA due to their efficient partitioning into the condensed phase (Biesenthal et al., 1997; Fry et al., 2009, 2014; Perraud et al., 2012; Rollins et al., 2012). Ambient measurements have indicated that up to 23 % of molecules in organic aerosol contain the RONO2 functional group, suggesting RONO2 represent a significant fraction of BVOC-derived oxidation products (Ditto et al., 2020; Rollins et al., 2013). However, the hydrolysis lifetime of organic nitrates in the aerosol phase (∼ 1–3 h) is much shorter than the lifetime of the aerosol itself to deposition (∼ 1 week), which suggests the contribution of organic nitrates to aerosol mass may have been underestimated from the observational data (Romer et al., 2016).

It has been shown that these low-volatility oxidation products readily partition from the gas phase into the particle phase and undergo further reactions (Perraud et al., 2012). Once in the particle phase, organic nitrates can undergo rapid hydrolysis under acid-catalyzed conditions, representing a sink for NOx in the form of HNO3 (Bean and Hildebrandt Ruiz, 2016; Darer et al., 2011; Hu et al., 2011; Liu et al., 2012; Rindelaub et al., 2015; Takeuchi and Ng, 2019). Hydrolysis rates and the relative contribution of particle-phase hydrolysis as a sink for organic nitrates (compared to reaction with O3, OH, photolysis, and deposition) vary significantly and depend on molecular structure (Boyd et al., 2015; Zare et al., 2018). To date, there are few studies of hydrolysis rate constants for monoterpene-derived organic nitrates (Rindelaub et al., 2015). The chemical structure (primary, secondary, and tertiary) can influence the rate of hydrolysis lifetimes of organic nitrates (Bean and Hildebrandt Ruiz, 2016; Boyd et al., 2015, 2015, 2017; Darer et al., 2011; Hu et al., 2011; Liu et al., 2012; Ng et al., 2017; Rindelaub et al., 2016b; Takeuchi and Ng, 2019; Zare et al., 2018), which is sometimes rapid compared to the lifetime of the particles with respect to dry deposition. Thus, more studies are needed to assess the structural dependence of both RONO2 production yields and the hydrolysis rate constants.

Although monoterpenes contribute significantly to the total annual BVOC budget, current models underestimate the impact of RONO2 on SOA mass measured in an ambient environment (Pye et al., 2013). Importantly, RONO2 yields for many BVOCs are unknown. Monoterpenes have been shown to produce more SOA than from isoprene in boreal regions (Lee et al., 2006; Tsigaridis and Kanakidou, 2007). Some of that SOA production from terpenes is now believed to be due to organic peroxides produced from RO2 + RO2 reactions involving the 10-carbon RO2 radicals when NO concentrations are low (Heinritzi et al., 2020). Pratt et al. (2012) showed that, for a temperate mixed forest in Michigan, isoprene-derived nitrates dominate the daytime simulated gas-phase organic nitrates; however, ∼ 20 % is derived from monoterpene nitrates. Of this, nearly one-third results from OH-initiated oxidation of β-ocimene, a tri-olefinic monoterpene of high reactivity that is emitted during the daytime (Pratt et al., 2012). To the best of our knowledge, the production of β-ocimene organic nitrates and aqueous-phase processing of its atmospheric oxidation products have not previously been studied.

To expand our understanding of the chemical evolution and fate of monoterpene-derived RONO2, we investigated the formation of RONO2 produced from the OH-initiated oxidation of β-ocimene in the presence of NOx using a photochemical reaction chamber. These experiments were conducted as a function of chamber relative humidity (RH) to provide insight on aqueous-phase partitioning and processing. The aqueous-phase processing of these species was explored using linear quadrupole ion trap mass spectrometry to determine hydrolysis kinetics. This study aims to quantify the yields of β-ocimene-derived RONO2 and the hydrolysis rate constants to better understand how organic nitrates may impact air quality and climate.

Experiments were conducted in a custom-made 5.5 m3 photochemical reaction chamber consisting of perfluoroalkoxy (PFA)–Teflon walls, PFA-coated endplates, a PFA-coated mixing fan, and UV lamps as described in Chen et al. (1998) and Lockwood et al. (2010). These experiments were conducted as a function of chamber RH, keeping other variables as constant as possible, and at 24(±2) ∘C. Eighteen experiments were carried out under variable humidity conditions, i.e., < 3 % up to 70 % RH, as listed in Table 1. The maximum humidity achievable in this chamber is 70 %, so higher RH is not considered for this study.

Before the experiment, the chamber walls were cleaned by adding ∼ 500 ppb O3, irradiating for 1 h, and flushing with ultra-zero (UZ) air in the dark until the O3 concentration was ∼ 0 ppb. The O3 concentration was monitored using a dual-beam ozone monitor (2B Technologies, model 205), which is calibrated using an ozone calibration source (2B technologies, model 306). The chamber RH was monitored using a LI-COR hygrometer (model 7000). The LI-COR was calibrated from known RH air sampled over a saturated K2SO4 solution at a fixed temperature. Experiments were started only if the chamber particle number concentration was < 10 particles per cubic centimeter.

First, β-ocimene (≥ 90 %, Sigma-Aldrich; (E)-/(Z)-β-ocimene mixture) was introduced to the chamber via injection through a T-shaped heated pyrex inlet, through which ultrapure N2 was used to transfer the injected β-ocimene via 1/4 in. PFA tubing at a flow rate of 5 L min-1. Formaldehyde (37 % v/v in water, Sigma) was injected using the same setup, followed by pure NO (99.5 %, Praxair) injected using the same setup without heat. Formaldehyde photolysis serves as an OH precursor in the presence of NO (Possanzini et al., 2002). Ammonium sulfate ((NH4)2SO4) seed particles were generated using a 10 wt % (NH4)2SO4 aqueous solution and a commercial atomizer (TSI, Inc., model 3076), and subsequently dried by passing through a diffusion dryer prior to entering the reaction chamber (as in Rindelaub et al., 2015). For humid experiments, chamber RH was adjusted by bubbling UZ air through nano-pure water using a commercial bubbler immersed in a temperature-controlled bath. The chamber fan was started during the initial injections, ensuring reactants were well mixed. Following mixing, initial concentrations were measured before the chamber UV lights were activated. After initial concentrations were measured, the fan was stopped to minimize wall losses, the chamber lights were activated (time = 0), and real-time measurements were obtained. A representative experimental time series is shown in Fig. 1, where the area between the blue vertical lines represents the time when the UV lights were switched on.

The β-ocimene concentration was quantified using gas chromatography–flame ionization detection (GC-FID; HP 5890 Series II) equipped with a gas injection loop. The GC-FID was calibrated using the same β-ocimene standard injected during the experiments, with gas-phase concentrations prepared in ∼ 200 L PFA–Teflon bags. Seven-point calibration curves with R2>0.995 were obtained and used for β-ocimene quantitation. NO and NO2 concentrations were measured using a custom-built chemiluminescence NOx analyzer (Lockwood et al., 2010). The size-resolved particle mass concentration was determined using a scanning mobility particle sizer (SMPS, TSI, Inc., model 3062) directly connected to the chamber via copper sampling lines. The hydroxy nitrates were quantified using I- chemical ionization mass spectrometry (CIMS) (Xiong et al., 2015, 2016). The inlet of the CIMS was maintained at a high relative humidity via saturated ultra-high-purity N2 carrier flow to eliminate the effects of water vapor on the CIMS sensitivity (Lee et al., 2016; Xiong et al., 2015). A synthesized α-pinene hydroxy nitrate standard was used as a surrogate standard to calibrate the CIMS for β-ocimene hydroxy nitrate determination (Rindelaub et al., 2016b; Slade et al., 2017). We note there may be differences in the sensitivity of the instrument towards individual β-ocimene hydroxy nitrate isomers compared to the α-pinene hydroxy nitrate standard due to differences in polarity and thus binding affinity with I- (Iyer et al., 2016; Lopez-Hilfiker et al., 2016). The synthetic α-pinene hydroxy nitrate standard is a δ-hydroxy nitrate (Rindelaub et al., 2016b), with expected relatively lower polarity and sensitivity for I- adduction compared to its β-analogue, as observed for the β- and δ-hydroxy nitrates of isoprene (Iyer et al., 2016). We expect the β-ocimene hydroxy nitrates consist of both sensitive β-hydroxy nitrates and less sensitive configurations, complicating their quantitation. Thus, further study is warranted, for example quantum chemical calculations and laboratory experiments (Iyer et al., 2016; Lopez-Hilfiker et al., 2016), to examine how the degree of unsaturation and acyclic nature of the β-ocimene hydroxy nitrate isomers affect their binding affinities with I-. O3 measurements were not possible due to interference in the ozone monitor signal by β-ocimene (Walker and Hawkins, 1952). However, for the NO and NO2 concentrations present in the chamber, and the JNO2 value (7×10-4 s-1), the calculated steady-state O3 concentration was < 1 ppb (considering loss via reaction with NO and β-ocimene) and was not an important reactant in the experiments. Photochemistry was terminated by turning off the chamber lights when one-third of the NO concentration remained in the chamber to ensure that β-ocimene oxidation occurred via OH oxidation only, and all RO2 radicals reacted only with NO (Atkinson et al., 2006; Capouet et al., 2004; Rindelaub et al., 2015).

A series of blank and control experiments were conducted to evaluate (i) chamber reactions without β-ocimene or NO, (ii) gas- and particle-phase product loss to chamber walls under different RH conditions, and (iii) dark reactions of NO2 with β-ocimene. Chamber reactions without β-ocimene or NO served as blank experiments and confirmed that all the RONO2 detected during oxidation experiments were produced from the OH-initiated oxidation of β-ocimene in the presence of NO. Additionally, the particle-phase RONO2 yields were corrected for wall loss from a second set of control experiments (kwall SOA=4.3(±0.3) × 10-5 s-1). The gas-phase wall loss rate constant for organic nitrates was determined based on observation of the first-order loss of the CIMS-determined monoterpene hydroxy nitrate (M= C10H17NO4) signal in the dark ([M+1]-; m/z=342; kONg=8.8(±2.2) × 10-6 s-1). These experiments were conducted at varying relative humidities, and the wall loss rate constants (kwall SOA, kONg) were determined from the loss in particle mass concentration and gas-phase concentration over time after the chamber lights were turned off. Blank experiments involving sampling from a cleaned chamber reveal no detectable degassing of organic nitrates from the walls, likely due to hydrolytic loss of adsorbed organic nitrates on the acidic walls (e.g., from uptake of HNO3). All experimental data were corrected for dilution for both the photochemistry experiments, and for the post-experiment sampling, based on the sampling time, flow rate, and makeup gas flow rate utilized during each experiment (kavg dilution=1.4(±0.1) × 10-5 s-1 during the experiments and 4.6(±0.5) × 10-5 s-1 during sampling). To further confirm that all RONO2 were a product of the OH-initiated oxidation in the presence of NO, the third control experiment was conducted by injecting NO2 into the dark chamber (UV lights remained off). The concentration of NO2 and ocimene were monitored over time, and the gas- and particle-phase extracts were analyzed for the presence of organic nitrates.

The gas- and particle-phase products were separated and collected for offline analysis by sampling through a XAD4 resin-coated eight-channel annular denuder (URG-200) followed by a 47 mm PTFE particle filter. The denuder and filter were separately extracted in tetrachloroethylene (≥ 99.9 %, Sigma-Aldrich) immediately following a chamber experiment and stored in a refrigerator at 5 ∘C. The extracts were then analyzed for total organic nitrate concentration using Fourier transform infrared spectroscopy (FTIR, Thermo Nicolet 6700) using a 1.00 cm liquid cell, as described by Rindelaub et al. (2015). A quality control experiment was conducted to determine if degradation of RONO2 occurred when the samples were stored in the refrigerator, and no change in concentration was observed over a 24 h period. Extracts were analyzed by FTIR within 24 h of extraction and stored in a refrigerator when not in use to minimize any degradation or additional reactions in situ post-extraction. The organic nitrate concentrations were determined using the asymmetric -ONO2 stretch at 1640 cm-1 (Nielsen et al., 1995), using an absorption coefficient of 12 900 L mol-1 cm-1. An uncertainty of 7 % was applied to the FTIR measurements to account for the uncertainty in the absorption coefficient when using a proxy (ethyl hexyl nitrate in this study) for -ONO2 quantitation (Carrington, 1960). After the FTIR analysis, samples were concentrated to near-complete dryness with ultra-high-purity nitrogen for use in high-resolution mass spectrometry analysis for structure elucidation and hydrolysis studies.

A series of quality control experiments were performed to evaluate the gas-phase organic nitrate collection efficiency by the denuder, particle transmission efficiency through the denuder, and particle collection efficiency on the filter, as described in Slade et al. (2017). To determine collection efficiency by the denuder, ∼ 200 L PFA–Teflon bags of known concentrations of ethyl hexyl nitrate were prepared and the concentrations were measured using GC-FID before entering the denuder and immediately after. The collection efficiency of the denuder was determined to be 100(±9) %. The filter collection efficiency (96(±19) %) was determined by measuring the particle mass concentration before and after the filter. Extraction efficiencies from both the denuder (99(±9) %) and filter (100(±2) %) were determined by measuring the concentration of RONO2 in 50 mL aliquots of serial extracts, for actual experiments. To account for sample loss during concentration, different aliquots of the gas- and particle-phase extracts were analyzed at varying points during the solvent evaporation process. As a result, we correct for losses of the organic nitrates due to evaporation during solvent concentration, which contributes a 14 % uncertainty to the final concentration. That uncertainty is included in the overall concentration uncertainty discussed below.

To better understand the oxidation products, denuder and filter extracts were analyzed for their chemical composition via ultra-high-performance liquid chromatography with electrospray ionization time-of-flight tandem mass spectrometry (UPLC-ESI-ToF-MS/MS, Sciex 5600+ TripleToF with Shimadzu 30 series pumps and autosampler) (Slade et al., 2017). The separation was achieved using reverse-phased liquid chromatography (Phenomenex Kinetex EVO C18 column, 100 Å, 100 mm × 2.1 mm, 5 µm) following the gradient elution described in Surratt et al. (2008), and MS analyses were completed using negative and positive electrospray ionization (ESI) modes with multiple-reaction monitoring in the W reflectron geometry mode. The mass resolution (m/Δm) of this mass spectrometer is 25 000 at m/z 195 using an accumulation time of 50 ms. The denuder and filter extracts were reconstituted in a 1:1 v/v solvent mixture of HPLC-grade methanol with 0.1 % acetic acid and HPLC-grade water. The gradient elution was performed at 0.3 mL min-1 with a binary mobile phase system: (A) 0.1 % acetic acid in water and (B) 0.1 % acetic acid in methanol. The 12 min gradient elution was as follows: the concentration of eluent B was 0 % for the first 2 min, linearly increased to 90 % from 2 to 10 min, held at 90 % from 10 to 10.2 min, and then decreased back to 0 % from 10.2 to 12 min (Surratt et al., 2008).

For hydrolysis studies, three experiments were selected representing dry, mid-RH, and high-RH conditions. The denuder and filter extracts were combined to form a bulk (gas- and particle-phase) β-ocimene nitrate solution that was then concentrated by solvent evaporation. The mixture was then reconstituted in 2.0 mL of an aqueous buffer solution at pH =2.5, 4.1, and 7.0. To facilitate a direct comparison to results obtained in Rindelaub et al. (2015), the same buffer solutions were used, listed in Table 2. The solution was agitated using a magnetic stir bar, and 50 µL was extracted every 2 min. The extract was then mixed with 50 µL of methanol (99.9 %, Fisher Chemical). The resulting solution was stirred and immediately analyzed (to avoid any impact of continued hydrolysis) via direct-infusion ESI into a Thermo Scientific LTQ XL Linear Ion Trap mass spectrometer operated in negative ionization mode. The hydrolysis rate constants were determined from the first-order decay of various organic nitrate-related functional groups monitored by MS2 fragmentation experiments. Briefly, fragmentation of the isolated molecular ion m/z 214.1 (C10H16NO4, expected first-generation oxidation product) was monitored at the instrument setting of collisional dissociation energy of 35 eV. The m/z 46 (NO2-) and 62 (NO3-) mass fragments were used to quantify relative concentrations of C10H16NO4 species as a function of the hydrolysis time. The exponential decrease in peak area of m/z 46 (NO2-) was used to quantify the observed hydrolysis rate constant, which is effectively the weighted average for the different structural isomers present in the solution, weighted by their relative concentrations. The m/z 62 (NO3-) fragment was used as a confirmation ion.

The production of organic nitrates serves as an important sink for gas-phase NOx. The first-generation organic nitrate products are of relatively high vapor pressure and, for the polyunsaturated terpene β-ocimene, are subject to further gas-phase oxidative loss, to produce smaller but more oxidized products (Bateman et al., 2011). While the main products of that oxidation will be hydroxy nitrooxy carbonyl compounds of lower carbon number, and thus perhaps higher vapor pressure, the yield of dihydroxy dinitrates will be significant. The latter compounds will be of much lower vapor pressure; for example that for compound B shown in Fig. 2 is 2.4 × 10-12 atm at 20 ∘C (Pankow and Asher, 2008), low enough that it should rapidly undergo uptake into the aerosol phase, followed by hydrolysis to produce a 10-carbon tetrol and NO3-. However, the extent to which the gas-phase oxidation of the first-generation olefinic hydroxy nitrates re-releases NOx needs to be studied from laboratory experiments with the pure compounds. The organic nitrates will undergo hydrolysis in the particle phase, transforming from a hydroxy nitrate to a diol. The nitrooxy group will be released as the nitrate ion, which can impact regional nitrogen cycling by either remaining in the particle phase (at certain pHs) or being released as gas-phase nitric acid (at very high particle acidities). In either case, NOx has been permanently removed from the system (Galbavy et al., 2007; Rindelaub et al., 2015; Romer Present et al., 2020; Suarez-Bertoa et al., 2012; Zafiriou and True, 1979). Therefore, knowledge of the hydrolysis rate constants, especially for isoprene nitrates, is badly needed. The relatively large organic nitrate yield (38 ± 9 %) in combination with the rapid aqueous-phase hydrolysis tends to make organic nitrate production an important sink for NOx in terpene-impacted forest environments. In addition, the light-dependent emission rate of β-ocimene suggests that in dense forest environments, like that of the University of Michigan Biological Station (Pratt et al., 2012), there could be steep vertical gradients in organic nitrate concentrations between canopy and ground, exceeding that of the other terpene nitrates modeled in Schulze et al. (2017). It is thus necessary to further assess the organic nitrate yield values, hydrolysis rate constants, and major oxidation products for a range of terpenes to fully understand the impact of their oxidation on the fate and distribution of NOx.

These results suggest that β-ocimene hydroxy organic nitrates may be an important sink for gas-phase NOx in forest environments. Further work should be done to examine the structural dependence of hydroxy nitrate hydrolysis kinetics, for example the difference between β-hydroxy and δ-hydroxy nitrates. Previous work has determined thermodynamic stabilities of structurally different organic nitrates (primary, secondary, tertiary) based on acid-catalyzed hydrolysis, where tertiary organic nitrates are less stable than primary or secondary organic nitrates at atmospherically relevant pHs, suggesting potential differences in hydrolysis rates between β-hydroxy and δ-hydroxy nitrates (Hu et al., 2011). Differences between the hydrolysis rates of β-ocimene organic nitrates from NO3 and OH-initiated oxidation should be explored, as significant differences in the hydrolysable fractions have been shown between organic nitrates formed from NO3 and OH-initiated oxidation of both α- and β-pinene (Takeuchi and Ng, 2019), and daytime NO3 concentrations can be relevant in forests like that of University of Michigan Biological Station (Pratt et al., 2012; Schulze et al., 2017). It is necessary to understand the SOA components produced during this oxidation to better predict aqueous-phase processing. Further work on the condensed-phase chemistry of alcohols and olefinic oxygenated compounds in the aerosol phase is necessary.