The experiments were conducted in SAPHIR-STAR, a 2 m³ continuously stirred tank reactor (SAPHIR: Simulation of Atmospheric PHotochemistry In a large Reaction chamber; STAR: – STirred Atmospheric flow Reactor). The details about its basic concept of operation have been described previously (Mentel et al., 2009; Baker et al., 2024). For this study the system was operated with a total flow rate of 60 L min⁻¹, of which 32 L min⁻¹ were injected into the reactor, resulting in a residence time of approximately 60 min. Experimental conditions were maintained at 20 °C and at 20 % relative humidity. ⚫OH radicals were produced by photolyzing hydrogen peroxide (H₂O₂) using two 254 nm UV-C lamps (TUV 16W 4P SE, Philips). The photolysis rate was controlled by reducing the photon flux by covering part of the lamps with adjustable bellows. H₂O₂ vapor was introduced into SAPHIR-STAR by bubbling a 0.2 L min⁻¹ nitrogen (N₂) flow through a 30 % w/w H₂O₂ solution (Sigma-Aldrich).

In the α-pinene photooxidation experiments a continuous flow of 0.012 L min⁻¹ from an α-pinene cylinder (with a concentration of ∼ 48.8 ppm) was injected into SAPHIR-STAR resulting in an initial concentration of ∼ 10 ppbv. For pinonaldehyde photooxidation, liquid pinonaldehyde (Orgentis chemicals, 99.7 %) was introduced by use of a syringe pump (Fusion 4000, Chemyx Inc.) at a flow rate of 0.145 µL h⁻¹ resulting in a concentration of 5 ppbv of pinonaldehyde in the reactor. NO was injected into the chamber from a gas cylinder (20.09 ppm NO in N₂ 5.0) via a mass flow controller for initial concentrations before reaction ranging between 0 and 7.5 ppbv.

To investigate the role of the H-abstraction pathway in α-pinene photooxidation we focused on the early stages of the reaction systems before secondary-generation products become important. All precursors were injected into the dark chamber. Once all precursors were well mixed and showed stable concentrations, the UV-C lamps were switched on for ⚫OH production by photolysis. The UV-C lamps were then kept on at a constant setting for one hour to initialize the photooxidation of α-pinene, followed by three hours without irradiation to allow for products to be flushed out and for precursors to recover initial concentrations. This cycle was repeated three times for every system investigated.

Since the first hour of photooxidation is crucial for investigating the relative role of the H-abstraction channel compared to the ⚫OH-addition channel, a cycle of one hour of photolysis followed by three hours without irradiation was repeated three times for each experimental condition listed in Table 1, as mentioned before. The time series of α-pinene, H₂O₂, and C₁₀H₁₇O7⚫ peroxy radicals over a total run of the experiments (14 h) are shown in Fig. S2, together with an example of the experimental procedure. During the final cycle the lamps were kept on continuously for 6 h to allow the system to reach steady state.

All major products from α-pinene photooxidation that were detected by NO3- CIMS, in the absence or presence of NO, were categorized into the following groups: C₁₀ monomers, C₂₀ dimers, and C_<10 fragments. The time series for contributions of each group varies for the three different NO conditions, as shown in Fig. 3a-1, a-2, a-3. A comparison under varying NO conditions is critical to elucidate the role of NO, since the presence of NO, ring opening and isomerization of alkoxy radicals, can be crucial in facilitating HOM formation from H-abstraction channel (Shen et al., 2022). The analysis was focused on the early reaction stages to explore the importance of the H-abstraction channel for HOM formation. The average product distributions of C₁₀ monomers from early stage α-pinene photooxidation without and with NO are illustrated by Kendrik Mass Defect (KMD) plots as a function of oxygen number (Kendrick Mass = oxygen, Fig. 3b-1, b-2, b-3). Each plot features seven distinct families, with compounds in each family sharing the same carbon and hydrogen numbers but differing in oxygen content.

Under α-pinene NO-free conditions the relative contributions of C₁₀H₁₈O_x, C₁₀H₁₆O_x, and C₂₀H₃₄O_x are 56.1 %, 11.1 %, and 8.8 %, respectively, and they are all closed-shell products related to C₁₀H₁₇Ox⚫. The first generation peroxy radicals with four-membered ring-opening, C₁₀H₁₇O3⚫, can undergo fast unimolecular reactions with rates of 4 ± 2 s⁻¹ (Xu et al., 2019). As shown in Fig. 3b-1, species with the formula C₁₀H₁₇O7⚫ are identified as predominantly HOM peroxy radicals and are detected in the nitrate mass spectra as C₁₀H₁₇O₇(NO3-). This agrees with previous observations by Berndt (2021), which were obtained using four different reagent ions, and calculations by Piletic and Kleindienst (2022). Additional HO2⚫ radicals were produced in our experiments via the reaction of the ⚫OH radical with H₂O₂. The increasing contribution of C₁₀H₁₈O_x (Fig. 3(1-a)) indicates a significant termination reaction of RO2⚫ by HO2⚫. Among the C₁₀H₁₈O_x family, C₁₀H₁₈O₇ shows the highest intensity and is likely produced via termination reactions of C₁₀H₁₇O7⚫ with HO2⚫, which aligns with the dominance of C₁₀H₁₇O7⚫ in the C₁₀H₁₇Ox⚫ family. Compounds with six oxygen atoms but one hydrogen less or one more hydrogen than C₁₀H₁₇O7⚫ (i.e., C₁₀H₁₆O₆ and C₁₀H₁₈O₆) are significant contributors to the total product signal of the C₁₀H₁₇Ox⚫ family. Compounds with the formula C₁₀H₁₆O₆ likely have carbonyl functionalities and are formed either directly from Reaction (R3) or from self-termination reactions of C₁₀H₁₇O7⚫ (Reaction R8) (Iyer et al., 2018; Jenkin et al., 2019). In contrast, C₁₀H₁₈O₆ species are possibly alcohols formed from C₁₀H₁₇O7⚫ via Reaction (R3), or more likely hydroperoxides derived from C₁₀H₁₇O6⚫ via Reaction (R1) (Iyer et al., 2018; Jenkin et al., 2019).

In addition to C₁₀ HOM monomers, C₂₀ accretion products also significantly contribute to HOM, particularly species containing 8, 10, or 12 oxygen atoms (Fig. S8). These have been identified as the most abundant dimers from ⚫OH initiated α-pinene photooxidation system by Berndt et al. (2016). C₂₀H₃₄O_(8,10,12) can be mechanistically explained by recombination of C₁₀H₁₇O7⚫ with C₁₀H₁₇O3⚫, C₁₀H₁₇O5⚫, or another C₁₀H₁₇O7⚫ via Reaction (4). C₁₀H₁₇O(3,5,7)⚫ have previously been reported to dominate the product spectrum (Lee et al., 2023; Berndt, 2021). In summary, the obtained HOM distributions in the base case without NO addition are consistent with previous studies and show mechanistically reasonable results.

When reactions are initiated in the presence of NO, the product distribution in the early reaction stage (when the RO2⚫ + NO regime is dominant) differs significantly from the NO-free case. The C₁₀H₁₇NO_x product family, identified as organic nitrates, accounts for 29.9 % and 29.4 % of the total signal under low-NO and high-NO conditions, respectively, and dominates C₁₀H₁₇Ox⚫-related HOM families. The most abundant species with formulas C₁₀H₁₇NO_(6,7,8) most likely represent organic nitrates formed via reactions of C₁₀H₁₇O(5,6,7)⚫ with NO. As the initial NO concentration increases, contribution from C₁₀H₁₇O6⚫-derived HOM, such as C₁₀H₁₇NO₇ and C₁₀H₁₈O₆, also rises (Figure S9). When the initial NO increased from 0.91 ± 0.05 to 7.4 ± 0.05 ppbv, the ratios of C₁₀H₁₇NO₇ to C₁₀H₁₇NO₈ and C₁₀H₁₈O₆ to C₁₀H₁₈O₇ are enhanced from 0.32 to 0.43 and from 0.63 to 0.78, respectively. At least one alkoxy step (Reaction R7) is necessary to produce peroxy radicals with even number, e.g. C₁₀H₁₇O6⚫. Therefore, the presence NO promotes the occurrence of alkoxy steps in HOM formation. Compared to the NO-free condition, the contributions of accretion products with C₂₀H₃₄O_(9,11) are also promoted. For forming C₂₀H₃₄O_(9,11) via Reaction (R5) an C₁₀H₁₇Ox⚫ peroxy radical with an even oxygen number must be involved, e.g. C₁₀H₁₇O4⚫ and C₁₀H₁₇O6⚫, further highlighting the role of alkoxy steps in HOM formation.

The C₁₀H₁₇O7⚫ peroxy radicals exhibit a rapid formation rate and a high molar yield in α-pinene photooxidation systems, but their formation shows only a weak dependence on the NO concentration. Although ⚫OH addition to α-pinene yields three structural first-generation isomers (C₁₀H₁₇O3⚫), only the isomer that undergoes an opening of the four-membered ring is capable of subsequent autoxidation (Lee et al., 2023; Piletic and Kleindienst, 2022; Berndt et al., 2016). The ring-opened C₁₀H₁₇O3⚫ can then undergo two rapid unimolecular autoxidation steps, with a rate constant of approximately 4 s⁻¹ for the first step which will lead to the formation of C₁₀H₁₇O5⚫, and a rate constant around 10 s⁻¹ for the subsequent step and formation of C₁₀H₁₇O7⚫ species (Piletic and Kleindienst, 2022; Xu et al., 2019). The fast autoxidation steps of the resulting peroxy radicals imply that the characteristic timescales of their formation are significantly shorter than those of competing bimolecular reactions with NO, RO2⚫, and HO2⚫, whose maximum effective rates in our experiments range from 0.02 to 1.88 s⁻¹ under NO-free, low-NO, and high-NO conditions, respectively. Therefore, biomolecular reactions, particularly involving NO, cannot effectively compete with the rapid autoxidation reactions, which can explain the observed weak NO dependence of the formation of C₁₀H₁₇O7⚫-related HOM.

In addition to the change of C₁₀H₁₇Ox⚫-related HOM, C₁₀H₁₅Ox⚫-related HOM families emerge once NO is introduced into the system (Fig. 3a-2, a-3, b-2, b-3). Under the NO-free condition, neither the C₁₀H₁₅NO_x nor the C₁₀H₁₄O_x family contributes more than 2 %. However, under low-NO conditions their contribution rises to 10.3 % and 4.7 %, respectively, and further increases to 24.9 % and 6.0 % under high-NO conditions. The most abundant peroxy radical in C₁₀H₁₅Ox⚫ family is C₁₀H₁₅O7⚫, as shown in Fig. S9, followed by C₁₀H₁₅O9⚫, which is consistent with the closed shell organic nitrate distribution, where C₁₀H₁₅NO₈ is most prevalent, followed by C₁₀H₁₅NO₁₀. The C₁₀H₁₄O_x family contains products with carbonyl groups and is formed via self-termination reactions of peroxy radicals. In addition to C₁₀ monomers, accretion products with the formula C₂₀H₃₀O_x and C₂₀H₃₂O_x, derived from Reaction (R5), also increase with elevated NO concentrations. Here at least one peroxy radical containing 15 hydrogen atoms must be involved (Berndt et al., 2018). Compared to monomers, formation of accretion products is reduced since in early stages of the reactions the RO2⚫ + R^′O2⚫ pathway is less important compared to the NO + RO2⚫ pathway or the RO2⚫ autoxidation. The intensity of C₁₀H₁₆O₇ increases gradually until it surpasses that of C₁₀H₁₆O₆, which in the absence of NO is the most abundant peak. C₁₀H₁₆O₇ is likely a carbonyl compound (or family), formed either through Reaction (R3), or from self-termination. Both need C₁₀H₁₇O8⚫ as precursors. However, analysis of the closed-shell families C₁₀H₁₈O_x and C₁₀H₁₇NO_x indicates the absence of significant formation of C₁₀H₁₇O8⚫ in presence of NO, which also aligns with Berndt (2021) and Piletic and Kleindienst (2022). C₁₀H₁₆O₇ may also be a hydroperoxide produced via Reaction (R2), or an alcohol produced via Reaction (R3), with C₁₀H₁₅O7⚫ peroxy radicals or C₁₀H₁₅O8⚫ peroxy radicals serving as precursors. Products formed via RO2⚫ + RO2⚫ reactions were produced at significantly slower rates than other species, as will be discussed in Sect. 3.2. Additionally, when C₁₀H₁₅O7⚫ reacts with another RO2⚫ radical, it forms C₁₀H₁₆O₆ or C₁₀H₁₄O₆ rather than C₁₀H₁₆O₇. Thus, Reaction (R3) was considered to not contribute significantly, and the reaction of C₁₀H₁₅O7⚫ with HO2⚫ most likely accounts for the significant increase of C₁₀H₁₆O₇. Similarly, species in the C₁₀H₁₆O_x family that have more than seven oxygen atoms can be attributed to reactions of C₁₀H₁₅Ox⚫ peroxy radicals with HO2⚫.

Contributions from peroxy radicals containing more than seven oxygen atoms to the C₁₀H₁₇Ox⚫ family are negligible, which is consistent with the number of oxygen atoms in the observed closed-shell products. This phenomenon is in accordance with previous measurement studies and calculations (Berndt, 2021; Lee et al., 2023). In the NO-free condition, compounds with more than seven oxygen atoms account for only 0.1 % of the C₁₀H₁₈O_x family, whereas for the low-NO and high-NO cases, compounds with more than eight oxygen atoms contribute 5 % and 6 %, respectively, to the C₁₀H₁₇NO_x family. Although the fate of the C₁₀H₁₇O7⚫ family remains unclear, the product distributions indicate that it is unlikely for them to undergo further steps of autoxidation. However, the C₁₀H₁₅Ox⚫-related HOM are more oxidized than C₁₀H₁₇Ox⚫-related HOM. For example, species containing more than eight oxygen atoms in C₁₀H₁₅NO_x account for 43.6 % and 46.7 % in low-NO and high-NO cases, respectively, of the entire C₁₀H₁₅NO_x family. Obviously C₁₀H₁₅Ox⚫ isomers with x≥7 are formed, which exhibit faster unimolecular reaction rates than the accessible isomers of the C₁₀H₁₇Ox⚫ peroxy radicals.

The distribution of the less oxygenated compounds was investigated based on the mass spectra detected by amine-CIMS, which are shown in Fig. S10. Consistent with nitrate-CIMS, C₁₀H₁₅Ox⚫-related less oxygenated compounds also emerge with elevated NO concentrations, particularly those in the C₁₀H₁₄O_x and C₁₀H₁₅NO_x families. The dominant compounds in the C₁₀H₁₅NO_x family are C₁₀H₁₅NO₄, C₁₀H₁₅NO₆, and C₁₀H₁₅NO₈, which should be derived from Reaction (R6) involving peroxy radicals with the formula C₁₀H₁₅O3⚫, C₁₀H₁₅O5⚫, and C₁₀H₁₅O7⚫, respectively. Among the major compounds in the C₁₀H₁₅NO_x family detected by amine-CIMS only highly oxidized species with the formular C₁₀H₁₅NO₈ can also be detected in significant amounts by nitrate-CIMS. The same phenomenon can also be seen for C₁₀H₁₇Ox⚫-related compounds. In amine-CIMS, the first-generation peroxy radical, C₁₀H₁₇O3⚫, and its associated closed-shell products play a dominant role in the combined C₁₀H₁₇Ox⚫, C₁₀H₁₈O_x, and C₁₀H₁₇NO_x families, although the C₁₀H₁₇O7⚫ related products are dominant in the mass spectrum obtained by nitrate-CIMS. For example, C₁₀H₁₇NO₄ and C₁₀H₁₈O₃ are major peaks, which are produced mechanistically through termination reactions of C₁₀H₁₇O3⚫ by NO and by HO2⚫ radicals, respectively. C₁₀H₁₇NO₄ contributes 56.6 % and 57.5 % to the C₁₀H₁₇NO_x family in low-NOand high-NO conditions, respectively. C₁₀H₁₈O₃accounts for 70.6 %, 77.6 %, and 69.3 % in NO-free, low-NO, and high-NO cases, respectively.

Only ring-opened C₁₀H₁₇O3⚫ can undergo fast autoxidation reactions to produce HOM, and they account only for approximately 25 % of the sum of C₁₀H₁₇O3⚫ peroxy radicals derived from the ⚫OH addition pathway (Xu et al., 2019, Vereecken et al., 2007). However, the nitrate-CIMS technique ionizes specifically highly oxygenated organics (Ehn et al., 2014), resulting in the detection of only a narrow range of highly oxidized compounds derived from four-membered ring-opened C₁₀H₁₇O3⚫. The amine-CIMS is efficient in detecting a large range of products including less oxygenated compounds (O < 7) (Berndt, 2021). C₁₀H₁₆O₂(C₃H₇NH3+) is found to be the strongest peak among the entire mass spectrum in amine-CIMS, indicating the production of pinonaldehyde. Pinonaldehyde can be formed from the formation of alkoxy radicals via reactions of four-membered ring-retained C₁₀H₁₇O3⚫ peroxy radicals and NO, which then undergo a fast opening of the six-membered ring, shown in Fig. S1 (Rolletter et al., 2019). The inclusion of amine-CIMS extends the detection of products, especially less oxygenated ones, which could indicate the fate of four-membered ring-retained C₁₀H₁₇O3⚫ peroxy radicals.

Alkoxy radicals derived from reactions involving C₁₀H₁₇Ox⚫ peroxy radicals and NO can undergo an H-shift, β-scissions, or a HO2⚫ loss step; however, none of these reactions are expected to form C₁₀H₁₅Ox⚫ peroxy radicals directly. For example, neither H-migration nor β-scission will lead to the loss of a hydrogen atom without a carbon-carbon bond cleavage, and the resulting alkyl radicals will react efficiently with O₂ to regenerate C₁₀H₁₇Ox⚫ radicals and thus continue the autooxidation cycle. The HO2⚫ loss channel is expected to be a minor pathway for the considered alkoxy radicals, because intramolecular H-migration and β-scission are more favorable, as previously shown by Vereecken and Peeters (2009, 2010). If the HO2⚫ loss channel would happen, it would lead to the formation of carbonyl and HO2⚫, losing one hydrogen atom forming C₁₀H₁₆O_x products. While the C₁₀H₁₆O_x products could, in principle, react with ⚫OH radical again to produce C₁₀H₁₅Ox⚫ intermediates, this process would occur on the timescale of secondary chemistry and is therefore unlikely to contribute significantly during the early reaction stages. Pinonaldehyde was dominant among C₁₀H₁₆O_x compounds, but its oxidation cannot become a significant source of C₁₀H₁₅Ox⚫-related HOM, which will be discussed in Sect. 3.2.

In conclusion, the product distributions from α-pinene photooxidation during early reaction stages were investigated. Compared to NO-free conditions, the presence of NO enhanced the formation of organic nitrates and promoted alkoxy formation steps, which lead to significant formation of C₁₀H₁₅Ox⚫-related compounds. These were detected by both nitrate-CIMS and amine-CIMS. The initial peroxy radicals leading to C₁₀H₁₅Ox⚫ related products were identified as C₁₀H₁₅O3⚫, which cannot arise from ozonolysis of α-pinene since α-pinene was dominantly oxidized by ⚫OH radicals rather than O₃, as shown in Fig. S7. Also, C₁₀H₁₅O4⚫ peroxy radicals are mechanistically expected to be first generation peroxy radicals in ozonolysis. Potential mechanisms that could be responsible for the emergence of C₁₀H₁₅Ox⚫-related HOM will be discussed in the following section.

In α-pinene oxidation systems three possible pathways can lead to monomers with 15 hydrogen atoms: either the H-abstraction channel by ⚫OH radicals (Shen et al., 2022) , ozonolysis (Berndt, 2022), or oxidation of pinonaldehyde (Rolletter et al., 2020). However, as mentioned in Sects. 2.2 and 3.1, ozone was not present at the beginning of each transient cycle and accounted for less than 1 % of α-pinene reaction loss. Thus ozonolysis cannot explain the significant formation of C₁₀H₁₅Ox⚫-related products, in particular ⚫OH-initiation yields more HOM than ozonolysis (Berndt, 2022). To explore potential formation pathways of C₁₀H₁₅Ox⚫-related compounds, FCM clustering was applied to the time series of 156 ions detected by nitrate-CIMS and amine-CIMS. Selection criteria for the ions were contributions of more than 0.1 % to the total ion signal in nitrate-CIMS or more than 1 % to the total ion signal in amine-CIMS.

A four-cluster solution was identified as the optimal solution. Corresponding results are presented in Fig. 4. Figure 4a shows the time series of the four cluster centers and the contributing ions. Compounds in cluster 1 form immediately after reactions are initiated by switching on the UV-C lamps, followed by cluster 2 and cluster 3, while compounds in cluster 4 show the longest formation time. The cluster centers peak at approximately 160, 210, 290, and 850 s, respectively. As shown in Fig. 4b, the major monomers in the C₁₀H₁₈O_x family and the major dimers (C₂₀H₃₀O_x, C₂₀H₃₂H₁₂, C₂₀H₃₄O₈) are all found in cluster 4, which exhibits the slowest formation rate among four clusters. These compounds originate from biomolecular termination reactions (Reactions R1 and R5), with their production rates largely controlled by the concentration of HO2⚫ and RO2⚫ radicals. NO is already in a steady state in the dark chamber, while HO2⚫ and RO2⚫ only begin to accumulate after the UV-C lamps are switched on. Therefore, the RO2⚫ fate is immediately affected by NO. Consequently, under high-NO conditions, RO2⚫ + NO reactions as well as autoxidation play a dominant role for the RO2⚫ in early reactions, and only these two reactions are discussed in following sections.

Compounds in cluster 1 and cluster 2 initially increase dramatically before decreasing again, and cluster 1 exhibits a steeper decline than cluster 2. The bulk chemical properties and the average carbon oxidation state (OSc‾) of each cluster are calculated (Kroll et al., 2011; Wu et al., 2021) and plotted as a function of average carbon (nC) or nitrogen atoms (nN) (Fig. S11). The early-generation cluster 1 exhibits the highest OSc‾ and nN but the lowest nC, indicating that it mainly consists of highly oxidized nitrogen-containing compounds and fragments (C < 10). This characteristic distinguishes cluster 1 from the other clusters. It is determined by the emerging of ⚫OH radicals at elevated NO concentration in the early stage.

The formation rate of each cluster can be interpreted kinetically as “typical” chemical formation rate (Wu et al., 2024). The compounds C₁₀H₁₅NO₇(NO3-) and C₁₀H₁₅NO₈(NO3-) are assigned to cluster 1 with the fastest formation rate; thus, they are likely first-generation products. Pinonaldehyde (C₁₀H₁₆O₂(C₃H₇NH3+)), a possible precursor, can only be found in cluster 3, which has a much slower formation rate. This clearly shows that pinonaldehyde oxidation cannot explain the production of C₁₀H₁₅NO₇(NO3-) or C₁₀H₁₅NO₈(NO3-). C₇H₉NO₈(NO3-), which is detected in high amount in nitrate-CIMS and also attributed to cluster 1, goes through a fragmentation step and is possibly formed via decomposition of C₁₀H₁₅Ox⚫, followed by further autoxidation and termination reactions by NO (Shen et al., 2022). Therefore, H-abstraction is probably the reason for the significant formation of C₁₀H₁₅Ox⚫-related HOM, and not the oxidation of pinonaldehyde. The fragment C₇H₁₁NO₈(NO3-) is formed faster than the organic nitrate monomer C₁₀H₁₇NO₈(NO3-), even though both are initialized by the OH-addition channel. This relation differs from that of C₇H₉NO₈(NO3-) and C₁₀H₁₅NO₈(NO3-).

As shown in Fig. S12, the summed abundance of C₇ families (C₇H₍₁₀,₁₂₎O_x, C₇H₍₉,₁₁₎NO_x) increases accordingly with the increase of the NO injection, accounting for 0.01 %, 0.3 % and 2.4 % of HOM compounds under NO-free, low-NO and high-NO conditions, respectively. It indicates increasing alkoxy radical decomposition with increasing NO. C₇ fragments are formed via dissociation of C₁₀ alkoxy radicals derived from bimolecular reactions of C₁₀ peroxy radicals with NO. Thus, the presence of NO promotes C₇ fragments by enhancing the production of alkoxy radicals (Vereecken and Peeters, 2000; Berndt, 2021). The most abundant families, C₇H₉NO_x and C₇H₁₁NO_x, likely originate from decomposition of C₁₀H₁₅Ox-1⚫ and C₁₀H₁₇Ox-1⚫, respectively. Under high-NO conditions, the ratio of C₇H₉NO_x to C₁₀H₁₅NO_x reaches 0.22, substantially higher than the ratio of C₇H₁₁NO_x to C₁₀H₁₇NO_x (0.05). This observation suggests that either alkoxy radical formation or subsequent decomposition is more efficient in the H-abstraction pathway than in the ⚫OH-addition pathway.

Evidence for alkoxy radical H-migration is provided by the enhanced formation of C₁₀H₁₇NO₇ and C₂₀H₃₄O₁₁ under high-NO conditions (Fig. S9). According to the oxygen-parity framework, C₁₀H₁₇Ox⚫ radicals formed solely through autoxidation are expected to contain an odd number of oxygen atoms, whereas an alkoxy H-migration step changes the oxygen parity from odd to even (Kang et al., 2025). Thus, the observed increase in C₁₀H₁₇NO₇ and C₂₀H₃₄O₁₁ suggests the formation of C₁₀H₁₇O6⚫ peroxy radicals through at least one alkoxy radical isomerization step.

To initialize the HOM chain via ⚫OH-addition pathway, four-membered rings must open from chemically activated tertiary radicals, which is independent of NO (Xu et al., 2019). In contrast, HOM formation through the H-abstraction pathway requires NO. The initially formed peroxy radicals (C₁₀H₁₅O2⚫) are unlikely to undergo autoxidation directly; therefore, their reaction with NO to form alkoxy radicals (C₁₀H₁₅O⚫), followed by the ring-opening step producing C₁₀H₁₅O3⚫ radicals, is necessary to initiate the HOM formation chain.

As autoxidation proceeds, higher oxygenated peroxy radicals, such as C₁₀H₁₇Ox⚫ and C₁₀H₁₅Ox⚫, are formed from C₁₀H₁₇O3⚫ and C₁₀H₁₅O4⚫, respectively. These peroxy radicals can further react with NO, yielding either the corresponding alkoxy radicals (C₁₀H₁₇Ox-1⚫ and C₁₀H₁₅Ox-1⚫) or organic nitrates (C₁₀H₁₇NO_x+1 and C₁₀H₁₅NO_x+1). However, the branching ratio between these two channels remains poorly constrained, contributing substantial uncertainty to the role of NO in HOM formation. Once formed, the fate of alkoxy radicals strongly depends on alkoxy radical structure (Vereecken and Peeters, 2010, 2009). Regardless of whether the precursor RO2⚫ radicals originate from the ⚫OH-addition or H-abstraction pathway, their subsequent chemistry can proceed through alkoxy radical intermediates. Therefore, the fate of these alkoxy radicals ultimately determines the impact of NO on HOM formation.

Overall, NO does influence HOM formation through alkoxy-radical chemistry. By promoting the conversion of RO2⚫ radicals to RO⚫ radicals, NO initializes pathways for H-migration and decomposition that would otherwise be less important. The extent to which these processes enhance or suppress HOM formation depends strongly on the structures of both the peroxy and alkoxy radicals, introducing substantial uncertainty into the overall role of NO.

To assess the robustness of the clustering results, additional cluster analysis was performed using a three- and five-cluster solution, as shown in Figs. S13 and S14, respectively. Consistent conclusions are obtained regardless of the number of clusters specified. In particular, the key compounds derived from the H-abstraction channel, such as C₁₀H₁₅NO₈(NO3-) and C₇H₉NO₈(NO3-), were consistently assigned to the first cluster with the fastest formation rate. The stability of the clustering results was further evaluated by performing 50 clustering runs with random initializations and using the four-cluster solution. The distributions of the cluster assignments were then analysed, as shown in Fig. S15, with particular attention given to the key compounds discussed above, including C₁₀H₁₅NO₈(NO3-), C₇H₉NO₈(NO3-), and C₁₀H₁₆O₂(C₃H₇NH3+). The results demonstrate the high stability of the cluster solution, as these compounds were consistently assigned to the same cluster across all runs. For example, C₁₀H₁₅NO₈(NO3-) was assigned to the first cluster in all runs. Therefore, the clustering results and insights derived from them are highly robust and reliable.

To assess the potential contribution of pinonaldehyde oxidation to the formation of C₁₀H₁₅Ox⚫-related HOM, pinonaldehyde photooxidation experiments were conducted under NO-free and high-NO conditions, analogous to conditions applied in the α-pinene systems (Table 1). Time series of α-pinene and its oxidation product pinonaldehyde during the first 30 min of the α-pinene (high-NO) experiment are shown in Fig. 5a-1. The corresponding turnovers of α-pinene and pinonaldehyde are presented in Fig. 5a-1. Similarly, the concentration of pinonaldehyde and its turnover during the pinonaldehyde (high-NO) experiment are shown in Fig. 5a-2. ⚫OH concentrations were estimated using model simulations.

During the early stage, the average pinonaldehyde turnover was 8.9 × 10⁶ molecules cm⁻³ s⁻¹ under pinonaldehyde (high-NO) condition, approximately 6 times higher than the turnover (1.5 × 10⁶ molecules cm⁻³ s⁻¹) under α-pinene (high-NO) condition. To facilitate comparison between the two experiments, we calculated the pinonaldehyde turnover for both experiments and compared it to obtain the ratio between the α-pinene (high-NO) and pinonaldehyde (high-NO) experiments (hereafter referred to as the pinonaldehyde turnover ratio). Signals measured in the pinonaldehyde experiment were subsequently normalized by this ratio to account for differences in the pinonaldehyde turnover between the two systems. The distributions of major C₁₀H₁₅Ox⚫-related products observed during α-pinene and pinonaldehyde experiments are presented in Fig. 5b. Major products belonging to the C₁₀H₁₄O_x and C₁₀H₁₅NO_xfamily detected by both nitrate-CIMS (N) and amine-CIMS (A) are included. The data shown in Fig. 5b represents averages over the first 320 s following the initiation of oxidation, corresponding to the early reaction stage, during which RO2⚫ + NO accounts for more than 90 % of the total bimolecular RO2⚫ loss rate. In addition, all product signals here as well as all data used for comparison in the following paragraphs have been normalized by the pinonaldehyde turnover.

The pinonaldehyde turnover ratio during the first 320 s exceeded 5 and was substantially higher during the first 100 s. This large difference arose because pinonaldehyde was present at the start of the pinonaldehyde system and was therefore immediately available for reaction with ⚫OH. In contrast, under the α-pinene system, pinonaldehyde had to be formed through α-pinene oxidation before it could undergo further reactions, resulting in a much lower pinonaldehyde turnover during the early stages of the experiment. For C₁₀H₁₅Ox⚫-related HOM detected by nitrate-CIMS the summed signals of the C₁₀H₁₄O_x and C₁₀H₁₅NO_x families under the α-pinene system exceed those of pinonaldehyde system by factors of approximately 20 and 10, respectively. Although pinonaldehyde oxidation contributes to the formation of C₁₀H₁₅Ox⚫-related HOM under high-NO conditions, this contribution is minor and can explain approximately 5 % of the HOM yield observed from α-pinene oxidation. Similarly, for less oxygenated products measured by amine-CIMS, the summed concentrations of the C₁₀H₁₄O_x and C₁₀H₁₅NO_x families were 98 times higher in the α-pinene system, and 6.7 times higher in the pinonaldehyde system.

For highly oxygenated compounds detected by nitrate-CIMS, species containing more than 10 oxygen atoms account for 11 % of the C₁₀H₁₄O_x family and for 13 % of the C₁₀H₁₅NO_x family under α-pinene high-NO condition, whereas their contributions are much lower for pinonaldehyde high-NO conditions (3 % and 5 %, respectively). The initial NO concentrations (Table 1) for two conditions are comparable, and consequently autoxidation rates can be the reason responsible for the differences in HOM distributions. Therefore, for peroxy radicals with more than 9 oxygen atoms the autoxidation chain of C₁₀H₁₅Ox⚫ peroxy radicals in pinonaldehyde photooxidation experiments are more easily terminated by NO than for peroxy radicals resulting from α-pinene photooxidation systems. The difference in the product distributions is significant between the two systems. In the pinonaldehyde system, the intensity of the C₁₀H₁₄O_x family is lower than that of the C₁₀H₁₅NO_x family, indicating that the branching towards Reaction (R8) undergoing self-termination is lower than the branching towards Reaction (R6).

For less oxygenated compounds detected by amine-CIMS, the most abundant species in C₁₀H₁₅NO_x family, C₁₀H₁₅NO₄,derived from the reaction of C₁₀H₁₅O3⚫ peroxy radicals with NO in the α-pinene system, does not appear in the pinonaldehyde system. Primary peroxy radicals with a formula of C₁₀H₁₅O4⚫ are formed when ⚫OH abstracts a hydrogen atom, mainly from the aldehyde group in pinonaldehyde (Rolletter et al., 2020; Fantechi et al., 2002), as a prerequisite for autoxidation. C₁₀H₁₅O3⚫ form when ⚫OH radicals abstract a hydrogen atom from α-pinene molecules, followed an alkoxy step (Reaction R7), a 6-membered ring opening, and an O₂ molecule addition (Shen et al., 2022). Here, the appearance of C₁₀H₁₅NO₄ could be explained by termination reactions between C₁₀H₁₅O3⚫ and NO in the α-pinene system, when the H-abstraction pathway indeed exists. The species detected in amine-CIMS with the formula C₁₀H₁₅NO₆ are significantly higher than other compounds in the C₁₀H₁₅NO_x family in the pinonaldehyde system. These species are less important than C₁₀H₁₅NO₄ in the α-pinene system. If an alkoxy radical is formed by the reaction of C₁₀H₁₅O4⚫ with NO, followed by H-shift and an O₂ addition, C₁₀H₁₅O5⚫ peroxy radicals will be produced and can be terminated by NO to produce C₁₀H₁₅NO₆ species. This is a potential pathway which could be responsible for the significant formation of C₁₀H₁₅NO₆ in the pinonaldehyde system.

In summary, clustering analysis shows that C₁₀H₁₅NO_x HOM species exhibit the fastest formation rate among all HOM compounds. They are also faster than first-generation product generation (e.g., pinonaldehyde) in the OH-addition channel under α-pinene high-NO conditions. The intensity of C₁₀H₁₅Ox⚫ related HOM (C₁₀H₁₄O_x and C₁₀H₁₅NO_x) in the α-pinene photooxidation system is still 11 times higher than those under the pinonaldehyde system although the pinonaldehyde turnover is corrected. In addition, product distributions observed in the α-pinene system differ substantially from those in the pinonaldehyde system, regardless of whether they are detected by nitrate-CIMS or amine-CIMS. Therefore, these findings demonstrate that the H-abstraction pathway, rather than the oxidation of pinonaldehyde, accounts for the formation of C₁₀H₁₅Ox⚫ related HOM in the α-pinene system.