All compounds were used as received and had the following purities: from Sigma–Aldrich/Fluka (St. Louis, MO): O-tert-Butylhydroxylamine hydrochloride, (TBOX, 99%), terpinolene (90%), toluene (HPLC grade, 99+%), cyclohexane (HPLC grade, 99+%), cyclohexanone (98%), methylglyoxal (40 wt% in water), and glutaraldehyde (50 wt% in water). Water (DI H₂O) was distilled, deionized to a resistivity of 18 MU cm, and filtered using a Milli-Q® filter system (Billerica, MA). Helium (UHP grade), the carrier gas, was supplied by Butler Gas (McKees Rocks, PA) and used as received.

Experiments were carried out at (297 ± 3) K and 1 atmosphere pressure in a 100-L chamber. The chamber/bag was constructed from 5-mil FEP Teflon® film (Welch Fluorocarbon Inc, Dover, NH) and sealed using a W-600T double foot sealer (Sealer Sales, Northridge, CA). Compressed air from the National Institute for Occupational Safety and Health (NIOSH) facility was passed through anhydrous calcium sulfate (CaSO₄, Drierite, Xenia, OH) and molecular sieves (Drierite) to remove both moisture and organic contaminants. This treated dry air was passed through a mass flow controller into a humidifying chamber and was subsequently mixed with dry air to the pre-determined relative humidity (RH) of 50%. The filler system was equipped with a syringe injection port constructed of a heated 6.4-mm Swagelok (Solon, OH) tee fitting with a 10 mm Ice Blue septum (Restek, Bellefonte, PA). This port facilitated the introduction of liquid reactants into the collapsible chamber. (Wells, 2005). Background measurements of the NIOSH facility air showed concentrations of O₃, NO, and NO₂ at less than 1.0, 1.2, and 0.5 ppb, respectively. All reactant mixtures were generated by this system.

Ozone was produced by photolyzing air with a mercury pen lamp (Jelight, Irvine, CA) in a separate Teflon® chamber. Aliquots of this O₃/air mixture were added to the Teflon® reaction chamber using a gas-tight syringe. Ozone concentrations were measured using a Thermo Electron (Waltham, MA) UV photometric ozone analyzer Model 49C. Aliquots of NO were added to reaction chamber from a 100 ppm tank (Butler Gas, McKees Rocks, PA) using a gas-tight syringe. NO and NO₂ concentrations were measured using a Thermo Electron NO_x analyzer Model 49i.

The three carbonyls (cyclohexanone, methylglyoxal, and glutaraldehyde (Table 1)) were used as surrogates for the calibration of all terpinolene + O₃ reaction products. The following retention times were observed: 12.0 min for singly derivatized cyclohexanone (MW = 169), 14.5 min for doubly derivatized methylglyoxal (MW= 214), and 20.0, 20.3, 20.5 min for doubly derivatized glutaraldehyde (MW= 242). Calibration plots for the three carbonyls were generated from integrating the peak areas of the total ion chromatograms (TIC) and are shown in the supplementary information, Fig. S1. The limit of detection (determined from three times the integrated baseline peak area) for the four observed compounds (MG, 4MCH, 6OPH, 36DOH) were: 1.4, 2.0,1.2, and 1.2 ppb, respectively.

Derivatization of non-symmetric carbnyls using TBOX typically resulted in multiple chromatographic peaks due to stereoisomers of the oximes. Due to co-elution, not all stereoisomers were chromatographically distinct. Typically an M+1 ion was observed for the derivatized oxime compounds. Retention times for the dicarbonyl oximes were assigned based on derivatization and analysis of solutions containing each dicarbonyl separately. Identification of multiple peaks of the same oxime compound is relatively simple since the mass spectra for each chromatographic peak of a particular oxime are almost identical. In most cases, the m/z = 57 ion relative intensity for the chromatographic peaks of the oximes was greater than 50% in the mass spectrum and could be effectively used to generate selected ion chromatograms to identify carbonyl compounds.

Four main products from terpinolene ozonolysis are listed below and shown in Table 2. Specific product yields for the reactions: terpinolene + O₃, terpinolene + O₃ + CH, and terpinolene + O₃ + NO are shown in Table 3. A two times the standard error of the regression slope was used to determine yield errors reported for in Table 3.

Previous investigations of terpene ozonolysis have resulted in the observation of multiple reaction products (Atkinson, 2003; Atkinson and Arey, 2003; Hakola et al., 1994; Harrison and Wells, 2013; Ma and Marston, 2009). These observations have been made using various analytical techniques such GC-FTIR, GC-FID, and GC-MS with and without using derivatization agents (e.g. PFBHA, BF₃/methanol). Common oxidation products observed include: mono/di-carboxylic acids and mono/dicarbonyl compounds; however, no tricarbonyl species have been observed previously (Hakola et al., 1994; Harrison and Wells, 2013; Ma and Marston, 2009; Reissell et al., 1999). This may be due to the high molecular weights of the derivatized tricarbonyls (i.e. masses that exceed 750 amu if PFBHA (adds 195 amu per carbonyl) were used for derivatization) or decomposition of the derivatized or underivatized tricarbonyl during GC analyses. Although tricarbonyls have not been observed from terpinolene ozonolysis, the possibility of forming tricarbonyl species has been predicted and detected from other terpene oxidation reactions. The Master Chemical Mechanism v3.3 (University of Leeds, UK) predicts a number of tricarbonyls from the ozonolysis of the terpenes: limonene, α-pinene, and β-pinene. Additionally, Carslaw modeled limonene oxidation by OH• and O₃ and identified the formation of several multi-carbonyl species (Carslaw, 2013). Furthermore, recent gasphase results of limonene ozonolysis with TBOX as the derivatization agent identified the tricarbonyl (3-acetyl-6-oxoheptanal) reaction product which was also predicted by Carslaw’s limonene/OH• model (Wells and Ham, 2014). Based on this data, the use of TBOX as a derivatization agent could aid in improving the carbon mass balance from terpene ozonolysis reactions.

Methylglyoxal (MG) is an oxidation product observed from terpinolene ozonolysis that is most likely generated through primary and secondary reaction processes. It is possible that MG is a primary product formed during O₃ addition to the exocyclic carbon–carbon double bond to form 4MCH. However, in this proposed scenario one would expect the yield of MG (0.10 ± 0.01) and 4MCH (0.32 ± 0.04) to be similar (Table 3). This suggests that MG further decomposes on separation from the ring or it is formed as the ring itself decomposes from ozone addition to the endocyclic carbon–carbon double bond. Interestingly, when OH• is scavenged in the terpinolene + O₃ system the yield of MG (0.14 ± 0.02, Table 3) is larger than the yield from terpinolene + O₃ suggesting that OH radicals may remove MG (k_{OH + MG} = 15 × 10⁻¹² cm³ molecule⁻¹ s⁻¹) (Atkinson, 2003) or OH radicals react with a MG precursor. The addition of 20 ppb (4.92 × 10¹¹ molecules cm⁻³) NO to the reaction system had no significant effect on the overall yield of MG (Table 3).

Ozone addition to the exocyclic double bond leads to the formation of 4MCH as previously described (Harrison and Wells). The 0.33 ± 0.04 yield of 4MCH (Table 3) from the terpinolene + O₃ (OH• scavenged) reaction, reported here, coincides with the previously measured yields: 0.28 ± 0.06 (Harrison and Wells, 2013), 0.40 ± 0.06 (Hakola et al., 1994), 0.40 ± 0.08 (Reissell et al., 1999), and 0.50 ± 0.05 (Ma and Marston, 2009). The high yield reported here and previously for this reaction was expected as O₃ addition to exocyclic double bonds and subsequent reactions are favored. The similar yields of 4MCH for the terpinolene + O₃ (0.32 ± 0.04) and terpinolene + O₃ (OH• scavenged, 0.33 ± 0.04) reactions indicate that OH• does not contribute to its formation and thus 4MCH is a primary reaction product. This was unexpected as the terpinolene + OH• rate under the experimental conditions is quite fast (2.25 × 10⁻¹⁰ cm³ molecule⁻¹ s⁻¹) (Aschmann et al., 2002) which would likely result in the formation of 4MCH or 6OPH. This suggests that in this experimental system OH• may preferentially react with something else, such as terpinolene, over 4MCH (AOP-Win calculated k_{OH + 4MCH} = 9 × 10⁻¹¹ cm³ molecule⁻¹ s⁻¹) or terpinolene. The addition of 20 ppb (4.92 × 10¹¹ molecules cm⁻³) NO to the reaction system had no significant effect on the overall yield of 4MCH (Table 3).

The dicarbonyl (6OPH) can be formed through O₃ or OH• (generated from Criegee intermediate decomposition) addition to the endocyclic double bond as previously described (Harrison and Wells, 2013). Based on the observed results, OH• addition is the primary route to 6OPH formation as a 67% reduction in yield was observed in comparing the terpinolene + O₃ and terpinolene + O₃ (OH• scavenged) experimental systems; 0.05 ± 0.01 vs 0.017 ± 0.004, respectively. Hakola et al. reported the yield of 6OPH (≤0.02) for the terpinolene + O₃ (OH• scavenged) reaction which is comparable to the yield reported (under OH• scavenged conditions) here, 0.017 ± 0.004 (Hakola et al., 1994). Additionally, comparing the observed yield of 4MCH (0.33 ± 0.04) to 6OPH (0.017 ± 0.004) further supports the notion that for terpinolene exocyclic O₃ addition is a more favored reaction pathway. The addition of 20 ppb (4.92 × 10¹¹ molecules cm⁻³) NO to the reaction system had no significant effect on the overall yield of 6OPH (Table 3).

There are three potential mechanisms for the formation of the newly observed tricarbonyl (36DOH). In the terpinolene + O₃ experiments, OH• radicals generated via Criegee decomposition can add to the endocyclic carbon–carbon double bond of 4MCH (Fig. S3). Subsequent reactions with O₂ and alkoxy radicals lead to formation of the stabilized tricarbonyl. This seems like a reasonable pathway since the 36DOH yield was reduced by 73% when OH• was scavenged. Secondly, it is possible that 36DOH may have been formed through secondary O₃ and/or OH• addition to the carbon–carbon double bond of 6OPH. This pathway seems unlikely for two reasons: 1) this potential reaction mechanism would suggest a yield of 36DOH (0.021 ± 0.003) (Table 3) that would be dependent on 6OPH (0.017 ± 0.004) yield, yet it can be seen that 36DOH and 6OPH yield measurements are of similar magnitude suggesting two independent formation routes and 2) since 6OPH concentrations are small (>7.4 × 10¹⁰ molecule cm⁻³ (3 ppb)) in this experimental system it is unlikely that the OH• addition rate to 6OPH would effectively compete with other possible OH• addition/H-abstraction pathways. Lastly, 36DOH could be made via O₃ addition to the exocyclic double bond of terpinolene and as the Criegee intermediate begins to break apart, an adjacent H from either of the terminal methyl groups may be extracted. The newly formed OH• moiety can possibly add to the endocyclic carbon–carbon double bond (Kumar et al., 2014a) instead of entering the gas phase. Subsequent breaking of the endocyclic double bond leads to a biradical which further reacts with O₂ to form 36DOH (Fig. S3). This pathway is supported by the observation that the 36DOH yield is not zero (0.021 ± 0.003) even when OH• is scavenged. There has been recent work on the modeling of Criegee intermediates or carboxylic acid catalyzed Criegee intermediate reaction that could support this alternate mechanism (Kumar et al., 2014a, b, c). The addition of 20 ppb (4.92 × 10¹¹ molecules cm⁻³) NO to the reaction system had no significant effect on the overall yield of 36DOH (Table 3).

Recent modeling of VOC oxidation by O₃, OH•, and NO₃• in the presence of NO_x highlighted the significance of NO on the VOC oxidation rate (Waring and Wells, 2015). In the model, the addition of NO resulted a reduced impact of ozonolysis on VOC oxidation due to scavenging of O₃ by NO. Additionally, NO reacts with alkyl peroxy radicals (RO₂•) to form RO•, which can stabilize to form carbonyls, or through addition and isomerization form alkyl nitrates (Finlayson-Pitts and Pitts, 2000).

In a low NO/high VOC environment such as the terpinolene system described here, the current data suggests that the formation of RONO₂ is more likely because this species is the primary terminator in the NO to NO₂ cycle for larger organic compounds (Perring et al., 2013). Evidence suggesting organic nitrate formation can be seen in Fig. 3 as MG, 6OPH and 36DOH concentrations decrease as NO concentration increases at a fixed O₃ concentration. Initially, 4MCH concentration decreases upon initial NO addition to the reaction systems, but increases as NO concentration increases. This could be due to the formation of additional OH• by: HO2+NO→OH·+NO2     KHO2+NO    =9.6×10−12cm3s−1 (Bohn and Zetzsch 1997) where OH• can add to the exocyclic double bond of terpinolene to form 4MCH. It is unlikely that O₃ could react with NO to eventually form NO₃ as O₃ is more likely to react with terpinolene (1.7 ppm, reaction rate = 0.08 s⁻¹) than NO (100 ppb, reaction rate = 0.045 s⁻¹). While some O₃ is likely lost due to reaction with NO, the predominate O₃ loss is through reaction with terpinolene. It is interesting to note that MG concentration, which could be formed from the exocyclic OH• addition decomposition, decreases as NO concentration increases. This may suggest that the observed reduction in MG concentration is the result of the formation of a multifunctional RONO₂ such as a hydroxy nitrate which would not be detected by TBOX derivatization. When OH• is scavenged, the MG concentration doubles even in the presence of 20 ppb NO, which may also suggest the formation of a hydroxy nitrate since both OH• and NO might be needed to form this species. The increase in 4MCH in the OH• scavenged terpinolene/O₃/NO system is not as easily explained. The gas-phase concentration of 4MCH may be affected by the presence of secondary organic aerosols (SOA). There may be a slight fraction of 4MCH partitioning into SOA. It has been suggested that SOA concentrations may be reduced in the presence of NO_x, in this case NO, which could result in an increase in 4MCH concentration (Donahue et al., 2005). The removal of two loss pathways for 4MCH (partitioning into particulate matter and reaction with OH• radicals) would be possible in the terpinolene/O₃/NO/cyclohexane system.

While the derivatization method described here has several advantages, there can be some limitations. These include: multiple molecular structures for carbonyls with the same molecular weight, no detection of other oxidation products such as carboxylic acids, alcohols and organic nitrates, and possible difficulty extracting multi-functional oxidation products such as organic nitrate/carbonyls and/or carboxylic acid/carbonyls. The use of water as the impinger solvent, though convenient, may also prevent capture of more hydrophobic oxidation products during collection (e.g. organic nitrate moiety) and/or prevent extraction of hydrophilic compounds due to enhanced solubility (e.g. carboxylic acid moiety).