Direct probing of acylperoxy radicals during ozonolysis of α-pinene: constraints on radical chemistry and production of highly oxygenated organic molecules

Acylperoxy radicals (RO₂) are key intermediates in the atmospheric oxidation of organic compounds and different from the general alkyl RO₂ radicals in reactivity. However, direct probing of the molecular identities and chemistry of acyl RO₂ remains quite limited. Here, we report a combined experimental and kinetic modeling study of the composition and formation mechanisms of acyl RO₂, as well as their contributions to the formation of highly oxygenated organic molecules (HOMs) during ozonolysis of α-pinene. We find that acyl RO₂ radicals account for 67 %, 94 %, and 32 % of the highly oxygenated C₇, C₈, and C₉ RO₂, respectively, but only a few percent of C₁₀ RO₂. The formation pathway of acyl RO₂ species depends on their oxygenation level. The highly oxygenated acyl RO₂ (oxygen atom number ≥6) are mainly formed by the intramolecular aldehydic H shift (i.e., autoxidation) of RO₂, while the less oxygenated acyl RO₂ (oxygen atom number <6) are basically derived from the C–C bond cleavage of alkoxy (RO) radicals containing an α-ketone group or the intramolecular H shift of RO containing an aldehyde group. The acyl-RO₂-involved reactions explain 50 %–90 % of C₇ and C₈ closed-shell HOMs and 14 % of C₁₀ HOMs, respectively. For C₉ HOMs, this contribution can be up to 30 %–60 %. In addition, acyl RO₂ contribute to 50 %–95 % of C₁₄–C₁₈ HOM dimer formation. Because of the generally fast reaction kinetics of acyl RO₂, the acyl RO₂ + alkyl RO₂ reactions seem to outcompete the alkyl RO₂ + alkyl RO₂ pathways, thereby affecting the fate of alkyl RO₂ and HOM formation. Our study sheds lights on the detailed formation pathways of the monoterpene-derived acyl RO₂ and their contributions to HOM formation, which will help to understand the oxidation chemistry of monoterpenes and sources of low-volatility organic compounds capable of driving particle formation and growth in the atmosphere.

1 Introduction

Monoterpenes (C₁₀H₁₆) comprise an important fraction of nonmethane hydrocarbons in the global atmosphere (Guenther et al., 2012; Sindelarova et al., 2014) and make a significant contribution to the secondary organic aerosol (SOA) budget (Pye et al., 2010; Iyer et al., 2021). The presence of a double bond and the large molecular size of monoterpenes favor their oxidation reactivity towards O₃, hydroxyl (OH), and nitrate (NO₃) radicals (Atkinson et al., 1990, 1986; Kurten et al., 2015; Kristensen et al., 2016; Bianchi et al., 2019; Berndt, 2022), as well as the formation of low-volatility products and SOA (Fry et al., 2009, 2014; Zhang et al., 2018; Bianchi et al., 2019; Molteni et al., 2019; Shen et al., 2022). The organic peroxy radicals (RO₂) in the gas-phase oxidation of monoterpenes can undergo autoxidation and form a class of highly oxygenated organic molecules (HOMs) (Jokinen et al., 2014; Mentel et al., 2015; Berndt et al., 2016; Zhao et al., 2018; Bianchi et al., 2019; Bell et al., 2022; Berndt, 2022), which are primarily low-volatility or extremely low-volatility organic compounds (LVOCs and ELVOCs) (Ehn et al., 2014; Bianchi et al., 2019) and thus play a crucial role in SOA formation and growth.

Significant advances have been made in recent years concerning the monoterpene RO₂ autoxidation and its contribution to HOM formation (Ehn et al., 2014; Berndt et al., 2016; Zhao et al., 2018; Xu et al., 2019; Lin et al., 2021; Berndt, 2022; Shen et al., 2022). It is recognized that some monoterpene RO₂ radicals derived from the traditional ozonolysis channel (i.e., isomerization of Criegee intermediates, CIs) and OH addition channel can autoxidize at a rate larger than 1 s⁻¹ and could be an important contributor to HOM formation (Zhao et al., 2018; Xu et al., 2019; Berndt, 2021). Recently, new reaction channels leading to the RO₂ radicals that can undergo fast autoxidation have been proposed. A quantum chemical calculation study indicated that an excited CI arising from α-pinene ozonolysis could undergo ring-breaking reactions and directly lead to a ring-opened RO₂ due to the excess energy, which can autoxidize at a rate of ∼1 s⁻¹ and rapidly form highly oxidized RO₂ with up to 8 oxygen atoms (Iyer et al., 2021). In addition, the minor hydrogen abstraction channel by OH radicals has been proposed as a predominant pathway to HOM formation from OH oxidation of α-pinene under atmospheric conditions (Shen et al., 2022).

RO₂ species can be simply divided into alkyl RO₂ and acyl RO₂ (RC(O)OO) according to whether R is an acyl radical. There are significant differences in the reactivity of these two kinds of RO₂. Firstly, the rate constant of acyl RO₂ with NO is in general slightly higher than that of alkyl RO₂ (Atkinson et al., 2007; Calvert et al., 2008; Orlando and Tyndall, 2012). For example, the reaction rate constants of acyl RO₂, CH₃C(O)O₂, and alkyl RO₂, CH₃CH₂O₂, with NO have been reported to be $\mathrm{20}\times {\mathrm{10}}^{-\mathrm{12}}$ and $\mathrm{9.2}\times {\mathrm{10}}^{-\mathrm{12}}$ ${\mathrm{cm}}^{\mathrm{3}}\mathrm{molec}{.}^{-\mathrm{1}}{\mathrm{s}}^{-\mathrm{1}}$, respectively (Atkinson et al., 2007; Calvert et al., 2008; Orlando and Tyndall, 2012). Besides, acyl RO₂ can react rapidly with NO₂ and form thermally unstable peroxyacyl nitrates (RC(O)OONO₂), which have a lifetime of tens of minutes at room temperature and of days and even months in winter or in the upper atmosphere with lower temperatures (Atkinson et al., 2007; Orlando and Tyndall, 2012). Although alkyl RO₂ radicals can also react with NO₂ and form the alkyl peroxynitrates (ROONO₂), they are extremely unstable and will decompose into RO₂ radicals and NO₂ in less than 1 s (Kirchner et al., 1997; Orlando and Tyndall, 2012). Lastly, the rate constant of cross-reaction of acyl RO₂ ($\mathrm{1.5}\pm \mathrm{0.3}\times {\mathrm{10}}^{-\mathrm{11}}$ ${\mathrm{cm}}^{\mathrm{3}}\mathrm{molec}{.}^{-\mathrm{1}}{\mathrm{s}}^{-\mathrm{1}}$) is significantly higher than that of alkyl RO₂ ($\mathrm{2}\times {\mathrm{10}}^{-\mathrm{17}}$–$\mathrm{1}\times {\mathrm{10}}^{-\mathrm{11}}$ ${\mathrm{cm}}^{\mathrm{3}}\mathrm{molec}{.}^{-\mathrm{1}}{\mathrm{s}}^{-\mathrm{1}}$) (Villenave and Lesclaux, 1996; Tyndall et al., 2001; Atkinson et al., 2007; Zhao et al., 2018). As a result, these two kinds of RO₂ may play different roles in the autoxidation as well as HOM and dimer formation.

The quantum calculations revealed that different functional groups in RO₂ would lead to significantly different intramolecular H-shift rates (Otkjær et al., 2018). The C=O and C=C substituents led to resonance-stabilized carbon radicals and could enhance the H-shift rate constants by more than a factor of 400. The fast aldehydic H-shift rate contributes to a series of acyl radicals (RC(O)) with the radical site at the terminal carbonyl carbon, which further produce the acyl RO₂ with O₂ addition. Many RO₂ formed in the oxidation of monoterpenes have the aldehyde functionality, especially for α-pinene ozonolysis, in which all the primary and many later-generation RO₂ contain at least one aldehyde group (Noziere et al., 2015; Berndt et al., 2018; Li et al., 2019; Berndt, 2022; Zhao et al., 2022). As a result, acyl RO₂ may comprise a considerable fraction of total RO₂ species and contribute significantly to the formation of low-volatility products and SOA in the monoterpene oxidation system. A recent study by Zhao et al. (2022) found that the acyl-RO₂-involved reactions contribute to 50 %–80 % of oxygenated C₁₅–C₂₀ dimers ($\mathrm{O}:\mathrm{C}\ge \mathrm{0.4}$) and 70 % of C₁₅–C₁₉ dimer esters in SOA from α-pinene ozonolysis. However, currently the direct probing of the molecular identities and chemistry of monoterpene-derived acyl RO₂ radicals is rather limited. The role of acyl RO₂ in HOM formation remains to be quantified.

In this study, the molecular identities and formation mechanisms of acyl RO₂ radicals, as well as their contributions to HOM formation in the α-pinene ozonolysis, are investigated. The experiments were conducted in a flow reactor with different concentrations of NO₂, which acted as an efficient scavenger for the acyl RO₂. The molecular composition and abundance of the gas-phase HOMs were measured by a chemical ionization–atmospheric pressure interface–time-of-flight mass spectrometer (CI-APi-ToF) using nitrate as the reagent ions. In addition, kinetic modeling using the Framework for 0-D Atmospheric Modeling (F0AM v4.1) employing the Master Chemical Mechanism (MCM v3.3.1), updated with the latest advances of the RO₂ chemistry, was performed to gain insights into the reaction kinetics and mechanisms of acyl RO₂ species. We find that acyl RO₂ account for a major fraction of highly oxygenated C₇ and C₈ RO₂ and play a significant role in the formation of HOM monomers and dimers with small molecular size. This study will help to understand the role of acyl RO₂ in the formation of low-volatility species from monoterpene oxidation and reduce the uncertainties in the future atmospheric modeling of the formation and impacts of aerosols.

2 Method and materials

2.1 Flow reactor experiments

The α-pinene ozonolysis experiments were carried out under room temperature (298 K) and dry conditions (relative humidity <5 %) in a custom-built flow reactor, which has been described in detail previously (Yao et al., 2019). The α-pinene vapor was generated by evaporating its pure liquid (99 %, Sigma-Aldrich) into a flow of zero air (10.65 L min⁻¹) added to the reactor using an automated syringe pump (TYD01–01 CE, Baoding Leifu Fluid Technology Co., Ltd.). The initial concentrations of α-pinene ranged from 500 ppb to 3 ppm in different experiments. Ozone was generated by passing a flow of ultra-high-purity (UHP) O₂ (150 mL min⁻¹, Shanghai Maytor Special Gas Co., Ltd.) through a quartz tube housing a pen-ray mercury lamp (UV-S2, UVP Inc.), and its concentration (45 and 180 ppb under low and high O₃ conditions, respectively) was measured by an ozone analyzer (Model 49i, Thermo Fisher Scientific, USA). The NO₂, acting as an acyl RO₂ scavenger, was derived from its standard cylinder gas (15.6 ppm, Shanghai Weichuang Standard Gas Co., Ltd.), and its initial concentration ranged from 0 to 30 ppb. To validate the formation mechanisms of acyl RO₂, selected experiments with the addition of NO or cyclohexane were also conducted. NO was derived by its standard cylinder gas (9.8 ppm, Shanghai Weichuang Standard Gas Co., Ltd.), and its initial concentration also ranged from 0 to 30 ppb. The gas-phase cyclohexane (∼500 ppm), acting as an OH scavenger, was generated by bubbling a gentle flow of UHP N₂ through liquid cyclohexane (LC-MS grade, CNW). The total air flow in the flow reactor was 10.8 L min⁻¹, and the residence time was 25 s. The relatively low O₃ concentration and short reaction time in the flow reactor avoid significant production of NO₃ radicals from NO₂ and O₃ and make the NO₃ oxidation contribute only 0.3 %–1.2 % of the total α-pinene oxidation in our experiments. Therefore, the NO₃ chemistry could be neglected in this study. A summary of the experimental conditions is given in Tables S1 and S2 in the Supplement.

The gas-phase RO₂ radicals and closed-shell products were measured by a nitrate-based CI-APi-ToF mass spectrometer (abbreviated as nitrate-CIMS; Aerodyne Research, Inc.), and a long time-of-flight mass spectrometer with a mass resolution of ∼10 000 $\mathrm{Th}/\mathrm{Th}$ was used here. The mass calibration error is below 1.8 ppm. The sheath flow, including a 2 mL min⁻¹ UHP N₂ flow containing nitric acid (HNO₃) and 22.4 L min⁻¹ zero air, was guided through the PhotoIonizer (Model L9491, Hamamatsu, Japan) to generate nitrate reagent ions. The total sample flow rate was 9 L min⁻¹ during the experiments. The mass spectra within the $m/z$ range of 50 to 700 were analyzed using the tofTools package developed by Junninen et al. (2010) based on MATLAB.

To clarify whether there is SOA formation in the experiments, a scanning mobility particle sizer (SMPS, TSI), which consists of an electrostatic classifier (model 3082), a long or nano differential mobility analyzer (model 3081 and 3085 for different particle sizes), and a condensation particle counter (model 3756), was used to monitor the formation of SOA particles. Except in Experiment (Exp) 31 where the reacted α-pinene reached 36.8 ppb and there was low SOA formation with particle mass concentrations of $\mathrm{5.0}\times {\mathrm{10}}^{-\mathrm{4}}$–$\mathrm{5.7}\times {\mathrm{10}}^{-\mathrm{3}}$ µg m⁻³ and number concentrations of 63–395 cm⁻³, no particle formation was observed by SMPS. Therefore, the RO₂ radicals and closed-shell products would be primarily distributed in the gas phase, with their fates negligibly influenced by the low SOA formation under these experimental conditions.

2.2 Kinetic model simulations

Model simulations of RO₂ and HOM formation in selected experiments were performed to constrain the reaction kinetics and mechanisms of acyl RO₂ using F0AM v4.1 (Wolfe et al., 2016), which employs MCM v3.3.1 (Jenkin et al., 2015) updated with the chemistry of RO₂ autoxidation and cross-reactions forming HOM monomers and dimers. Newly added species and reactions to MCM v3.3.1 followed the work by Zhao et al. (2018) and Wang et al. (2021). Considering that the default MCM v3.3.1 does not include highly oxygenated acyl RO₂, we added the possible formation pathways of the potential acyl RO₂ measured in this study to the model based on the mechanisms proposed by Zhao et al. (2022).

The formation and reaction branching ratios of the two α-pinene-derived CIs are updated in the model according to the recent studies (Table S3) (Claflin et al., 2018; Iyer et al., 2021; Zhao et al., 2021; Berndt, 2022). The formation of a ring-opened C₁₀H₁₅O₄-RO₂ species (C10H15O4RBRO2 in Table S3) from α-pinene ozonolysis proposed by a recent study (Iyer et al., 2021), as well as its subsequent autoxidation and bimolecular reactions, is included in the model. The autoxidation rate constant of the ring-opened C₁₀H₁₅O₄-RO₂ is 1 s⁻¹, and a lower limit of its molar yield (30 %) was used according to the recent studies (Zhao et al., 2021; Meder et al., 2023) and our results (see details in Sect. 3.3). We also added the hydrogen abstraction channel of α-pinene oxidation by OH radicals according to a recent study (Shen et al., 2022). The branching ratio of this channel was set to 9 %, with the remaining 91 % being the traditional OH addition pathways. The detailed reaction pathways and rate constants of RO₂ species in this channel followed the work by Shen et al. (2022), except for RO₂ cross-reactions, the rates of which were not reported in that study. As the primary RO₂ radicals (C₁₀H₁₅O₂-RO₂) formed via the hydrogen abstraction by OH radicals are least oxidized with only 2 oxygen atoms, their cross-reaction rate could be relatively low (Atkinson et al., 2007; Orlando and Tyndall, 2012). In the model, this rate constant was set to $\mathrm{1}\times {\mathrm{10}}^{-\mathrm{13}}$ ${\mathrm{cm}}^{\mathrm{3}}\mathrm{molec}{.}^{-\mathrm{1}}{\mathrm{s}}^{-\mathrm{1}}$. For other alkyl RO₂ radicals (including HOM-RO₂), their cross-reaction rate constant is assumed to be $\mathrm{1}\times {\mathrm{10}}^{-\mathrm{12}}$ ${\mathrm{cm}}^{\mathrm{3}}\mathrm{molec}{.}^{-\mathrm{1}}{\mathrm{s}}^{-\mathrm{1}}$ according to Zhao et al. (2018). The dimer formation rates for these alkyl RO₂ are same as their cross-reaction rates.

In flow reactor experiments, the equilibrium formation of ROONO₂ would lead to the consumption of alkyl RO₂ radicals. To account for the influence of this process on the RO₂ budget and HOM formation, we included the reaction of ${\mathrm{RO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{2}}\leftrightharpoons {\mathrm{ROONO}}_{\mathrm{2}}$ in the model, with forward and reverse reaction rate constants of $\mathrm{7.5}\times {\mathrm{10}}^{-\mathrm{12}}$ ${\mathrm{cm}}^{\mathrm{3}}\mathrm{molec}{.}^{-\mathrm{1}}{\mathrm{s}}^{-\mathrm{1}}$ and 5 s⁻¹, respectively (Orlando and Tyndall, 2012). To simplify the parameterization, the forward and reverse reaction rate constants of newly added highly oxygenated acyl RO₂ with NO₂ are the same as default values in MCM v3.3.1. Besides, the cross-reaction rate constants of acyl RO₂ (including acyl RO₂ + acyl RO₂ and acyl RO₂ + alkyl RO₂) forming monomers or dimers were both set to $\mathrm{1}\times {\mathrm{10}}^{-\mathrm{11}}$ ${\mathrm{cm}}^{\mathrm{3}}\mathrm{molec}{.}^{-\mathrm{1}}{\mathrm{s}}^{-\mathrm{1}}$ (Orlando and Tyndall, 2012). Considering that there are large uncertainties in the dimer formation rate of RO₂, a sensitivity analysis was conducted to evaluate its influence on acyl-RO₂-involved HOM formation by varying the rate constant from $\mathrm{1}\times {\mathrm{10}}^{-\mathrm{13}}$ to $\mathrm{1}\times {\mathrm{10}}^{-\mathrm{12}}$ ${\mathrm{cm}}^{\mathrm{3}}\mathrm{molec}{.}^{-\mathrm{1}}{\mathrm{s}}^{-\mathrm{1}}$ for alkyl RO₂ and $\mathrm{1}\times {\mathrm{10}}^{-\mathrm{12}}$ to $\mathrm{1}\times {\mathrm{10}}^{-\mathrm{11}}$ ${\mathrm{cm}}^{\mathrm{3}}\mathrm{molec}{.}^{-\mathrm{1}}{\mathrm{s}}^{-\mathrm{1}}$ for acyl RO₂. The results show that changes in dimer formation rate constants within the above ranges have no significant influence on the contribution of acyl RO₂ to HOM formation (Fig. S1 in the Supplement).

The wall losses of OH, HO₂, and RO₂ radicals, as well as closed-shell HOM monomers and dimers in the flow reactor, were considered using the Knopf–Pöschl–Shiraiwa (KPS) method proposed by Knopf et al. (2015) in the model (Table S4), with an assumption of irreversible uptake of these species on the reactor wall. It is found that the wall loss of OH, HO₂, and RO₂ radicals accounts for 0.08 %–0.14 %, 4.7 %–9.1 %, and 7.3 %–25.5 % of their total production, respectively, with lower values under higher reacted α-pinene concentration conditions. Therefore, the wall loss process would not significantly influence α-pinene oxidation and RO₂ chemistry. The wall losses of closed-shell HOM monomers and dimers account for 18.4 %–34.7 % and 14.2 %–33.1 % of their total production, respectively. It should be noted that the wall losses of typical RO₂ and HOMs have a negligible impact on their responses to the addition of NO₂ (Fig. S2). In addition, with the consideration of the wall loss effects, the contribution of acyl RO₂ to the HOM formation only changed a little (0.02 %–0.5 %). Therefore, the wall losses of RO₂ and HOMs in the flow reactor would not affect the interpretation of the results in this study.

3 Results and discussion

3.1 Molecular composition of acyl RO₂ from α-pinene ozonolysis

The overall formation characteristics of gas-phase RO₂, closed-shell monomers, and dimers with the addition of NO₂ (30 ppb) are shown in Fig. 1 (Exps 8 and 14, Table S1). Since the nitrate-CIMS is only highly sensitive to the highly oxygenated species, we only discuss the production of HOMs with oxygen atoms above 6 here. As for RO₂ and closed-shell monomers (Fig. 1a), the signals of C₇ and C₈ species decrease by more than 50 % with the addition of NO₂, while for C₉ and C₁₀ species, their decreases are relatively small (within 40 %). In addition, we note that there is an unexpected increase in some C₉ and C₁₀ RO₂, and the possible reason will be discussed in detail later.

Figure 1 Mass defect plots of (a) RO₂ and HOM monomers and (b) HOM dimers formed from ozonolysis of α-pinene in the presence of NO₂ measured using a nitrate-CIMS (Exps 8, 14). The circles are colored by the relative changes in signal of RO₂, monomers, and dimers due to the addition of NO₂ (30 ppb). The area of circles is linearly scaled with the cube root of the signal of HOMs formed in the absence of NO₂. The blue lines represent RO₂ radicals.

NO₂ could react rapidly with acyl RO₂ radicals to form RC(O)OONO₂, which has a higher thermal-stability compared to ROONO₂ and can serve as a sink for acyl RO₂ on our experimental timescales. Therefore, a significant decrease in C₇ and C₈ RO₂ and HOMs upon the addition of NO₂ indicates that a major fraction of C₇ and C₈ RO₂ are acyl RO₂. In contrast, the slight decrease in C₉ and C₁₀ HOM monomers shows that the contribution of acyl RO₂ to C₉ and C₁₀ RO₂ is relatively small. However, some of the C₁₀ monomers showed a slight increase with the addition of NO₂, especially for C₁₀H₁₈O_x HOMs. The addition of NO₂ plays a twofold role in dimer formation from α-pinene ozonolysis (Fig. 1b). There is a significant inhibiting effect on C₁₄–C₁₈ dimers, which is due to the large contribution of acyl RO₂ to the total C₇ and C₈ RO₂ that generate such dimers. However, C₁₉ and C₂₀ dimers only show a slight decrease with the addition of NO₂, and some of them are even enhanced. In particular, the enhancement in C₂₀H₃₄O_x is most significant, reaching 30 %.

Kinetic model simulations show that the concentration of alkyl RO₂ decreases by 1 %–20 % with the addition of 30 ppb NO₂ under different reacted α-pinene conditions (Exps 1–28). Considering that the acyl RO₂ could be rapidly consumed by NO₂, if the signal reduction of a RO₂ specie significantly exceeds 20 % with 30 ppb NO₂ addition, we presume it has significant contribution from acyl RO₂. As a result, a total of 10 acyl RO₂ were identified according to the changes of RO₂ signal as a function of initial NO₂ concentration, which include C₇H₉O₆, C₇H₉O₇, C₈H₁₃O₆, C₈H₁₃O₈, C₈H₁₃O₉, C₈H₁₃O₁₀, C₉H₁₃O₉, C₉H₁₇O₇, C₉H₁₇O₉, and C₁₀H₁₅O₇. Figure 2 shows the averaged normalized acyl RO₂ signals measured as a function of the added NO₂ concentration under different experimental conditions (Exps 1–28). Similarly, since the nitrate-CIMS is only highly sensitive to products with high oxygen content, we only observed acyl RO₂ with oxygen atoms above 6. Consistent with the significant decrease in C₇ and C₈ species with the addition of NO₂ in Fig. 1a, C₇ and C₈ acyl RO₂ decrease by more than 50 % with the increase in NO₂ concentration (Fig. 2a and b). For C₉ acyl RO₂, the C₉H₁₇O₇-RO₂ and C₉H₁₇O₉-RO₂ also decrease dramatically with increasing NO₂, and the decrease in C₉H₁₃O₉-RO₂ is relatively smaller (Fig. 2c). In addition, C₁₀H₁₅O₇-RO₂ also shows a small decrease (Fig. 2d), with a reduction of only 30 % at 30 ppb NO₂. The relatively small reduction in the abundance of some of these RO₂ radicals indicates the presence of alkyl RO₂ radicals with the same chemical formulas. Along with the marked reduction in acyl RO₂ signals, the production of highly oxygenated RC(O)OONO₂ species such as C₉H₁₃O₉NO₂, C₉H₁₇O₇NO₂, and C₁₀H₁₅O₇NO₂ with the addition of NO₂ was observed (see the spectra in Fig. S3). However, we note that although some RC(O)OONO₂ such as C₈H₁₃O₆NO₂ and C₈H₁₃O₈NO₂ are expected to be formed with NO₂ addition, they could not be unambiguously detected by the nitrate-CIMS due to the overlapping of their peaks with strong alkyl RO₂ peaks (C₉H₁₅O₈-RO₂ and C₉H₁₅O₁₀-RO₂) in this study.

Figure 2 Averaged normalized signal of the measured acyl RO₂ as a function of the added NO₂ concentration under different experimental conditions (Exps 1–28).

Figure 3 shows the contribution of acyl and alkyl RO₂ to the highly oxygenated C₇–C₁₀ RO₂. Acyl RO₂ contribute 66.9 %, 94.3 % and 31.7 % to the total C₇, C₈, and C₉ RO₂ signals, respectively. By contrast, the only C₁₀ acyl RO₂ measured in this study is C₁₀H₁₅O₇, which contributes to only 0.4 % of the total C₁₀ RO₂. It should be noted that there might be other C₁₀ acyl RO₂ that were not observed due to the interferences from the alkyl RO₂ with the same chemical formulas, which respond differently to the addition of NO₂ than acyl RO₂ do (see details in the following discussion). Considering that some RO₂ formulas such as C₁₀H₁₅O₇ may have contributions from both acyl RO₂ and alkyl RO₂, we assumed the decrease of RO₂ signal with the addition of NO₂ as the signal of acyl RO₂. Besides, it is obvious that the normalized signal basically decreases to the lowest value when the initial NO₂ concentration reaches 10 ppb (Fig. 2), indicating that most of the acyl RO₂ are depleted at this NO₂ concentration. In addition, the decreasing extents of some acyl RO₂ are different for different reacted α-pinene concentrations, with lower decreasing extent for higher reacted α-pinene concentrations (Fig. S4). This difference might be due to the promoted cross-reactions of acyl RO₂ as well as their precursor RO₂ at higher α-pinene concentrations, which are competitive with the reactions leading to acyl RO₂ formation as well as the acyl RO₂ + NO₂ reactions.

Figure 3 Contributions of acyl and alkyl RO₂ to the highly oxygenated C₇–C₁₀ RO₂ measured by a nitrate-CIMS.

In addition to the changes of acyl RO₂ signal, we also show the changes of normalized alkyl RO₂ signal with the increasing initial NO₂ concentration in Fig. S5. Although ROONO₂ formed by the reaction of alkyl RO₂ with NO₂ is thermally unstable and would decompose quickly to release RO₂, it would still reach a formation–decomposition equilibrium in the system, thus consuming a small amount of alkyl RO₂. However, it can be seen from Fig. S5 that during 25 s of reaction in the flow reactor, a large part of alkyl RO₂ has an increasing trend with the increase in NO₂ concentration. We speculate that a portion of ROONO₂ could decompose back to RO₂ and NO₂ in the nitrate chemical ionization inlet where the sample gases were diluted instantly and the equilibrium of ROONO₂ was disturbed, resulting in the release of a large amount of RO₂.

To verify our speculation, the decomposition of ROONO₂ in the chemical ionization inlet was simulated based on the dilution ratio (1:3.5) and residence time (200 ms) in the inlet. As shown in Fig. S6, more than 40 % of ROONO₂ decomposes back to RO₂ and NO₂ in the chemical ionization inlet, which would inevitably lead to an increase in RO₂ concentration. As the C₁₀H₁₅O₈NO₂ has a significant contribution from the relatively stable RC(O)OONO₂ arising from the ring-opened acyl C₁₀H₁₅O₈-RO₂ reported by Iyer et al. (2021), its decomposition is relatively small (∼21 %). It should be noted that the RO₂ measured here is only a part of total RO₂ produced and that a large amount of RO₂ has already reacted to form closed-shell products as well as ROONO₂ in the flow reactor. Taking Exp 14 as an example (30 ppb NO₂), the simulated concentrations of RO₂ and ROONO₂ are 1.3 and 1.9 ppb, which approximately accounts for 27.1 % and 39.6 % of the total production of RO₂, respectively. Therefore, the decomposition of ROONO₂ could indeed result in an increase in the RO₂ signal. It should also be pointed out that because of the very short residence time in the chemical ionization inlet, such an increase in the RO₂ concentration would not significantly impact HOM formation.

To confirm the reliability of our results, we examined the changes in the signals of RO₂ and closed-shell products as a function of reacted α-pinene in the absence of NO₂ (Sect. S1 and Fig. S7 in the Supplement), and the results are consistent with previous studies (Zhao et al., 2018). In addition, we repeated Exps 15–21 on another nitrate-CIMS, and a similar increase in alkyl RO₂ signals with the addition of NO₂ was observed on that instrument (Fig. S8).

3.2 Formation mechanisms of acyl RO₂ during α-pinene ozonolysis

It has been recently suggested that there are three main pathways that directly lead to the formation of monoterpene-derived acyl RO₂ (Zhao et al., 2022): (i) the autoxidation of RO₂ containing aldehyde groups (Reaction R1), (ii) the cleavage of a C–C bond of RO containing an α-ketone group (Reaction R2), and (iii) the intramolecular H shift of RO containing an aldehyde group (Reaction R3). In addition, the secondary OH oxidation of aldehyde products can also produce acyl RO₂ radicals. However, in the present study, the secondary OH oxidation is expected to be insignificant due to an excess of α-pinene compared to O₃. Indeed, kinetic model simulations incorporating the secondary OH chemistry show that the contribution of secondary OH oxidation to acyl RO₂ formation is negligible even under high O₃ conditions (see details in Sect. S2 and Fig. S9).

Here, we further investigated the formation mechanisms of acyl RO₂. Figure 4 shows the reaction schemes leading to the formation of example acyl RO₂ radicals. The detailed formation mechanisms of acyl RO₂ measured in this study are shown in Fig. S10. The formation of acyl RO₂, especially those having the small molecular size (C₇–C₉), requires the production and subsequent decomposition (or ring-opening process) of RO radicals. Taking C₈H₁₃O₆-RO₂ as an example (Fig. 4), two steps of RO formation and decomposition following the primary C₁₀H₁₅O₄-RO₂ lead to the ring-opened C₈H₁₃O₄-RO₂ that can undergo rapid aldehydic H shift to form the acyl RO₂.

$$\begin{array}{}\text{(R1)}& {\displaystyle }{\mathrm{RO}}_{\mathrm{2}}\stackrel{\text{autox}}{⟶}\text{acyl }{\mathrm{RO}}_{\mathrm{2}}\text{(R2)}& {\displaystyle }{\mathrm{RO}}_{\mathrm{2}}\stackrel{+{\mathrm{RO}}_{\mathrm{2}}}{⟶}\mathrm{RO}\stackrel{\mathrm{C}\text{–}\mathrm{C}\text{cleavage}}{⟶}\text{acyl }{\mathrm{RO}}_{\mathrm{2}}\text{(R3)}& {\displaystyle }{\mathrm{RO}}_{\mathrm{2}}\stackrel{+{\mathrm{RO}}_{\mathrm{2}}}{⟶}\mathrm{RO}\stackrel{\mathrm{H}\text{shift}}{⟶}\text{acyl }{\mathrm{RO}}_{\mathrm{2}}\end{array}$$

Figure 4 Three different formation pathways of acyl RO₂ during ozonolysis of α-pinene. The acyl RO₂, C₉H₁₃O₄ and C₁₀H₁₅O₅, formed via pathways 2 and 3, respectively, were not detected by the nitrate-CIMS in this study due to their relatively low oxygenation level.

To verify the formation mechanisms of acyl RO₂, we added NO in some experiments (Exps 33–56) to see how acyl RO₂ respond to the increasing NO concentration. As shown in Fig. 5, the changes of C₇ and C₈ acyl RO₂ show an opposite trend with the increasing NO and NO₂ concentration, except for C₈H₁₃O₈-RO₂. NO can react with RO₂ to form RO radicals and promote the formation of RO₂ that requires the involvement of RO radicals in their formation. In addition to C₈H₁₃O₆-RO₂ discussed above, the formation of C₇H₉O₇-RO₂ and C₈H₁₃O₉-RO₂ needs two and four steps of the RO formation following C₁₀H₁₅O₄-RO₂ (Fig. S10), respectively. Therefore, the increase in RO concentration due to the addition of NO would promote the production of these acyl RO₂. These results prove that the RO radicals indeed play an important role in the acyl RO₂ formation, while for C₈H₁₃O₈-RO₂, its signal decreases substantially with the addition of NO up to 3 ppb, similar to the trend observed with the addition of NO₂. After reaching the minimum at 7 ppb NO, the signal of C₈H₁₃O₈-RO₂ tends to increase with the further increase in NO concentration. Given that C₈H₁₃O₈-RO₂ is likely to directly come from the autoxidation of C₈H₁₃O₆ acyl RO₂ (see Fig. S10), the rapid consumption of C₈H₁₃O₆-RO₂ by NO and NO₂ (formed by O₃ oxidation of NO) may outcompete its autoxidation process, thus leading to a decrease in C₈H₁₃O₈-RO₂ signal. Besides, it can be seen that the increasing extent in C₈H₁₃O₆-RO₂ is also relatively small before the NO concentration reaches 3 ppb (Fig. 5c), indicating that the promotion effect of NO on C₈H₁₃O₆-RO₂ formation is not that strong at this concentration.

Figure 5 Averaged normalized signal of typical acyl RO₂ as a function of initial NO or NO₂ addition (Exps 1–28 and 33–56).

It is interesting to note that most of the measured highly oxygenated acyl RO₂ are formed by the autoxidation of aldehydic RO₂, and only the C₈H₁₃O₉-RO₂ is formed by the H shift of the RO radical (Fig. S10). The measured signal of acyl RO₂ from the autoxidation pathway accounts for 96 % of all highly oxygenated acyl RO₂ signals. Considering that the acyl RO₂ with small molecular size are generally the ring-opened RO₂, the autoxidation rate constant of their precursor RO₂ is expected to be relatively high (e.g., 1 s⁻¹) (Iyer et al., 2021). Taking a RO₂ cross-reaction rate constant of $\mathrm{1}\times {\mathrm{10}}^{-\mathrm{12}}$ ${\mathrm{cm}}^{\mathrm{3}}\mathrm{molec}{.}^{-\mathrm{1}}{\mathrm{s}}^{-\mathrm{1}}$ (Zhao et al., 2018) and a model-predicted total RO₂ concentration of 1.7 ppb (Exp 8), the simulated contributions of autoxidation and cross-reactions to the total RO₂ reaction are 96.0 % and 4.0 %, respectively. Considering a RO₂ cross-reaction rate constant that is 10 times larger (i.e., $\mathrm{1}\times {\mathrm{10}}^{-\mathrm{11}}$ ${\mathrm{cm}}^{\mathrm{3}}\mathrm{molec}{.}^{-\mathrm{1}}{\mathrm{s}}^{-\mathrm{1}}$), the simulated contributions of RO₂ autoxidation and cross-reactions would be 70.4 % and 29.6 %, respectively. These simulations suggest that the autoxidation of aldehydic RO₂ plays a dominant role in the formation of the highly oxygenated acyl RO₂. Although the acyl RO₂ with low oxygen content were not measured in this study, all acyl RO₂ containing oxygen atoms less than 6 seem to be derived from the cleavage of the C–C bond or H shift of RO containing an α-ketone or aldehyde in the currently known reaction mechanisms (Figs. 4 and S11).

Recently, Shen et al. (2022) found that the hydrogen abstraction by OH radicals during α-pinene oxidation plays an important role in HOM formation. In such mechanisms, the primary RO₂ reacts with NO and forms RO radicals, which could undergo rapid ring-breaking reactions to form a series of ring-opened C₁₀H₁₅O_x-RO₂, which contains aldehyde functionality and can easily autoxidize to C₁₀ acyl RO₂. In the absence of NO, the cross-reactions of RO₂ can also produce RO radicals. However, only a few C₁₀ acyl RO₂ were detected in this study, and they contribute less than 1 % of the total C₁₀ RO₂ signal. This phenomenon could be due to the fact that the primary RO₂ (C₁₀H₁₅O₂) formed by the hydrogen abstraction by OH radical are least oxidized with only 2 oxygen atoms, which are expected to have a relatively low cross-reaction rate constant (Orlando and Tyndall, 2012; Berndt et al., 2018). As a result, the formation of ring-opened C₁₀H₁₅O_x-RO₂ via cross-reactions of the primary C₁₀H₁₅O₂-RO₂ may not be important. As shown in Fig. 6, when the cross-reaction rate constants of C₁₀H₁₅O₂-RO₂ are considered to be $\mathrm{1}\times {\mathrm{10}}^{-\mathrm{13}}$ ${\mathrm{cm}}^{\mathrm{3}}\mathrm{molec}{.}^{-\mathrm{1}}{\mathrm{s}}^{-\mathrm{1}}$, the simulated contribution of the H-abstraction pathway to the HOM formation is less than 3 % under both low-reacted (2.4 ppb) and high-reacted (9.6 ppb) α-pinene conditions. It should be note that the cross-reaction rate constants of the less-oxygenated RO₂ could be even lower (Orlando and Tyndall, 2012); therefore the contribution of this pathway to HOM formation could be ignored when NO is absent.

Figure 6 Contributions of the H-abstraction pathway by OH radicals (yellow) and OH addition and ozonolysis pathways (blue) to the formation of typical HOMs under low-reacted (a) and high-reacted (b) α-pinene conditions simulated by the kinetic model. The cross-reaction rate constant was set to $\mathrm{1}\times {\mathrm{10}}^{-\mathrm{13}}$ ${\mathrm{cm}}^{\mathrm{3}}\mathrm{molec}{.}^{-\mathrm{1}}{\mathrm{s}}^{-\mathrm{1}}$ for the primary C₁₀H₁₅O₂-RO₂ and $\mathrm{1}\times {\mathrm{10}}^{-\mathrm{12}}$ ${\mathrm{cm}}^{\mathrm{3}}\mathrm{molec}{.}^{-\mathrm{1}}{\mathrm{s}}^{-\mathrm{1}}$ for the more oxygenated RO₂.

In the presence of cyclohexane as an OH scavenger (Fig. S12, Exp 32), the measured signals of C₁₀H₁₇O_x-RO₂ formed via the OH addition channel and the corresponding C₁₀H₁₈O_x HOMs decrease by more than 70 %, while the C₁₀H₁₅O_x-RO₂ and its related closed-shell products decrease by less than 15 %, in good agreement with the measurements in previous studies (Zhao et al., 2018). As the C₁₀H₁₆O₈ HOM could come from both C₁₀H₁₅O_x-RO₂ and C₁₀H₁₇O_x-RO₂, its reduction is at a medium level. The significantly smaller decrease in the signals of C₁₀H₁₅O_x-RO₂ and its corresponding closed-shell products as compared to those of C₁₀H₁₇O_x-RO₂ and the related closed-shell products further illustrates that the H abstraction by OH has a minor contribution to HOM formation in the absence of NO.

Figure 7 shows the changes in measured signal of C₁₀H₁₅O₈-RO₂ and C₁₀H₁₅O₁₀-RO₂ as a function of initial NO concentration (Exps 33–56). It should be noted that due to the existence of O₃ in our experiments, these two RO₂ could come from both O₃ and OH reactions with α-pinene, and NO could be rapidly oxidized to NO₂ by O₃. The normalized signals of C₁₀H₁₅O₈-RO₂ and C₁₀H₁₅O₁₀-RO₂ increase firstly under low NO conditions, which is similar to the change of acyl RO₂ as shown in Fig. 5. This increase could be for two reasons: (1) the promoted formation of C₁₀H₁₅O₈ and C₁₀H₁₅O₁₀ acyl RO₂ from the H-abstraction channel by NO addition and (2) the equilibrium decomposition of ROONO₂ formed by the two alkyl RO₂ from ozonolysis of α-pinene in the chemical ionization inlet (see Sect. 3.1). As mentioned above, the ring-opened C₁₀H₁₅O_x-RO₂ formed from the H-abstraction channel contain aldehyde functionality and can autoxidize rapidly. The F0AM simulations show that the C₁₀H₁₅O₈ and C₁₀H₁₅O₁₀ acyl RO₂ formed from the H-abstraction channel contribute to 68 % and 56 % of the total C₁₀H₁₅O₈-RO₂ and C₁₀H₁₅O₁₀-RO₂ with the addition of 10 ppb NO, respectively. Therefore, the initial increases of these two RO₂ with increasing NO concentration are likely mainly due to the enhanced formation of C₁₀H₁₅O₈ and C₁₀H₁₅O₁₀ acyl RO₂. When the NO concentration increases to a high level, there is more NO and NO₂ in the system, which promotes the consumption of acyl RO₂. As a result, C₁₀H₁₅O₈-RO₂ exhibits a decreasing trend, and the increasing extend of C₁₀H₁₅O₁₀-RO₂ becomes much smaller.

Figure 7 Averaged normalized signal of the measured C₁₀H₁₅O₈- and C₁₀H₁₅O₁₀-RO₂ as a function of the added NO concentration (Exps 33–56).

3.3 Contributions of acyl RO₂ to the formation of gas-phase HOMs

With the addition of NO₂, the distribution of gas-phase products in the α-pinene ozonolysis changes significantly (see Fig. 1), and the consumption of acyl RO₂ by NO₂ plays an important role. NO₂ influences the formation of HOM monomers mainly in three ways. Firstly, NO₂ could react rapidly with acyl RO₂ and form RC(O)OONO₂, thus inhibiting the formation of HOMs with the involvement of acyl RO₂. Secondly, as mentioned above, although ROONO₂ is thermally unstable, their formation–decomposition equilibrium still consumes a small amount of alkyl RO₂, resulting in a decrease in HOM formation. Thirdly, NO₂ can consume a part of HO₂ radicals (Fig. S13), thus inhibiting the RO₂+HO₂ reaction pathway.

Figure 8 shows the normalized signal of C₇–C₁₀ HOM monomers as a function of initial NO₂ concentration. The signals of C₇, C₈, and some of C₉ HOMs decrease significantly with increasing NO₂ concentration due to the relatively large contribution of acyl RO₂ to the total C₇–C₉ RO₂. The C₇ HOMs decrease by more than 50 % when the NO₂ concentration reaches 30 ppb, while C₈ HOMs decrease by more than 70 %, and some of them even decrease by 90 %. The C₉ HOMs decrease by 30 %–60 %, and the species with relatively large decrease are mostly acyl RO₂-related HOMs. For C₁₀ HOMs, although there is also an obvious decrease in their formation with the addition of NO₂, most of them have a smaller decreasing extent compared to the C₇–C₉ HOMs due to the low contribution of acyl RO₂ to the C₁₀ RO₂. It is worth noting that a few C₁₀ HOMs increase initially with the addition of NO₂ up to 10 ppb, suggesting that there might be some processes that promote the formation of their precursor RO₂ radicals and thus offset the inhibiting effect of NO₂.

Figure 8 Averaged normalized signal of the measured C₇–C₁₀ HOMs as a function of the added NO₂ concentration (Exps 1–28).

As mentioned above, the addition of NO₂ has the most significant influence on the formation of small HOM monomers. Combined with the large contribution (67 %–94 %) of acyl RO₂ to the total C₇ and C₈ RO₂ (Fig. 3), it can be considered that the reduction in the formation of C₇ and C₈ HOM monomers with NO₂ addition is overwhelmingly due to the consumption of acyl RO₂ by NO₂. As a result, acyl RO₂ were found to have a contribution of 50 %–90 % to C₇ and C₈ HOM monomer formation during α-pinene ozonolysis. Since acyl RO₂ also have a considerable contribution (32 %) to the total C₉ RO₂, an upper limit (30 %–60 %) of their contribution to C₉ HOMs could be derived with the assumption that the decrease of C₉ HOMs with the addition of NO₂ is also mainly due to the consumption of C₉-acyl RO₂ by NO₂. By contrast, acyl RO₂ account for a very small fraction (0.4 %) of the total C₁₀ RO₂, and their contribution to C₁₀ HOMs cannot be quantified based solely on the experimental measurements given that the equilibrium reaction between alkyl RO₂ and NO₂ can also affect the formation of HOMs. Therefore, we used F0AM to simulate the contribution of acyl RO₂ to C₁₀ HOM formation according to the acyl RO₂ measured in this study. It should be noted that the HOMs from the acyl RO₂ and its subsequent RO₂ (formed from acyl RO₂ reactions) are all considered to be acyl-RO₂-related HOMs in the model.

Figure 9 Simulated average contribution of acyl and alkyl RO₂ to C₇–C₁₀ HOM formation from ozonolysis of α-pinene under typical experimental conditions (Exps 1, 8, 15, and 22).

As mentioned above, the formation of ring-opened C₁₀H₁₅O₄-RO₂ reported by Iyer et al. (2021) is included in the model, and its autoxidation produces a ring-opened acyl C₁₀H₁₅O₈-RO₂. When we consider the upper limit of the yield of ring-opened C₁₀H₁₅O₄-RO₂ (89 %) in the model and assume that the other primary RO₂ with the cyclobutyl ring autoxidize at a very slow rate (0.01 s⁻¹), the simulated acyl C₁₀H₁₅O₈-RO₂ would contribute to ∼80 % of the total C₁₀ RO₂. However, we could not see a large decrease in the measured signal of C₁₀H₁₅O₈-RO₂ and its related HOM monomers with the addition of NO₂. Similarly, a recent study by Zhao et al. (2022) found that the C₁₀H₁₅O₈-related monomers and dimers in α-pinene SOA did not significantly decrease with NO₂ addition. There might be three reasons for the discrepancy between the simulations and measurements. Firstly, the yield of the ring-opened C₁₀H₁₅O₄-RO₂ might be significantly smaller than 89 % (Zhao et al., 2021; Meder et al., 2023). Secondly, the autoxidation rate of other primary C₁₀H₁₅O₄-RO₂ with the cyclobutyl ring could be significantly larger than 0.01 s⁻¹. Thirdly, the ring-opened C₁₀H₁₅O₈-RO₂, a highly functionalized acyl RO₂ radical with an –OOH group, may be able to undergo very fast intramolecular H-scrambling reactions to form a peroxy acid as suggested by Knap and Jørgensen (2017), which would compete with the NO₂ reaction and result in a lower reduction in its signal upon NO₂ addition (see details in Sect. S3).

To examine the contributions of acyl RO₂ to C₁₀ HOM production, we updated the branching ratios and autoxidation rates of the primary RO₂ during α-pinene ozonolysis in the model according to the recent studies (Kurten et al., 2015; Claflin et al., 2018; Zhao et al., 2021; Berndt, 2022) (Table S3) and used a lower limit (30 %) of the ring-opened C₁₀H₁₅O₄-RO₂ yield reported by Iyer et al. (2021). As displayed in Fig. 9, the simulated acyl RO₂-related HOMs contribute to 14 % of the total C₁₀ HOMs, which is slightly smaller than the measured decrease in C₁₀ HOMs with the addition of NO₂. This discrepancy could be for two reasons. Firstly, the decrease in HOMs can partly result from the consumption of alkyl RO₂ and HO₂ radicals by the addition of NO₂. Secondly, as mentioned above, there might be other C₁₀ acyl RO₂ that were not observed in this study due to the decomposition of the ROONO₂ from the alkyl RO₂ with the same formulas. The contributions of acyl RO₂ to the formation of C₇–C₉ HOMs were also simulated (Fig. 9). For C₇ and C₈ HOMs, the model predicts a contribution of 52 %–98 % from acyl RO₂, which is consistent with the measurements (50 %–90 %). However, the simulated contribution of acyl RO₂ to C₉ HOMs is over 99 %, which is not consistent with the measurements (Fig. 8c). Recent studies indicated that the CI radicals from α-pinene ozonolysis did not form the alkyl C₉H₁₅O₃-RO₂ (C96O2 in default MCM v3.3.1) (Kurten et al., 2015; Zhao et al., 2021; Berndt, 2022). As a result, this primary C₉ alkyl RO₂ was not considered in the model, and most of C₉ RO₂ considered are acyl RO₂ or from acyl RO₂ reactions. In view of the significantly lower measured (less than 30 %–60 %) than simulated (over 99 %) contribution of acyl RO₂ to C₉ HOMs, we speculate that a small part of CI radicals might be able to form the C₉H₁₅O₃-RO₂, which could further react to form highly oxygenated alkyl C₉ RO₂.

A sensitivity analysis of the alkyl C₉H₁₅O₃-RO₂ yield was conducted to see its influence on the contribution of acyl RO₂ to the total C₉ HOMs. The model simulations show that when the yield of this C₉ RO₂ from one of the CIs ranges between 0.5 % and 2 %, the contribution of acyl RO₂ to the total C₉ HOMs ranges from 27.5 % to 59.8 % (Fig. S15), which is almost consistent with the measurements. This result indicates that a small part of CIs could generate the C₉ alkyl RO₂. We note that the small production of C₉H₁₅O₃-RO₂ from CIs has no significant influence on the yield of C₁₀H₁₅O₄-RO₂ and the subsequent acyl RO₂. As shown in Fig. S16, as the C₉H₁₅O₃-RO₂ yield increases from 0 % to 3 %, the simulated concentrations of C₁₀H₁₅O₄-RO₂ exhibit negligible to small (5 %) changes. As the C₉H₁₅O₃-RO₂ is considered to only produce highly oxygenated alkyl RO₂ in the model, it results in a decrease in the contribution of acyl RO₂ to the total C₉ HOMs. However, the contributions of acyl RO₂ to total C₇, C₈, and C₁₀ HOMs are almost unchanged.

The cross-reaction rate constant of acyl RO₂ is generally larger than that of alkyl RO₂ (Atkinson et al., 2007; Orlando and Tyndall, 2012), and the fast cross-reaction may lead to an important contribution to the HOM dimer production. The responses of dimer formation to increasing concentration of initial NO₂ during α-pinene ozonolysis are given in Fig. 10. The C₁₄–C₁₈ dimers decrease by up to 50 %–95 % with the increase in NO₂ concentration up to 30 ppb (Fig. 10a–e). The rapid cross-reaction rate of acyl RO₂, as well as their dominant contribution to the small RO₂ species, makes acyl RO₂ an important contributor to the formation of these dimers. The consumption of acyl RO₂ by NO₂ greatly inhibits the bimolecular reactions involving acyl RO₂, resulting in a rapid decrease in the signal of the corresponding dimers. Considering the predominance of acyl RO₂ in small RO₂ and their high reaction rate with NO₂ compared to the alkyl RO₂, it can be concluded that the cross-reactions involving acyl RO₂ contribute to roughly 50 %–95 % of the C₁₄–C₁₈ dimer formation.

Figure 10 Averaged normalized signal of the measured C₁₄–C₂₀ dimers as a function of the added NO₂ concentration (Exps 1–28).

For C₁₉ dimers, due to the relatively smaller contribution of acyl RO₂ to C₉ and C₁₀ RO₂, their signal decreases only by 10 %–40 %, and this reduction has contributions from both acyl and alkyl RO₂. For C₂₀ dimers, their signal changes with the addition of NO₂ can be discussed according to the number of hydrogen atoms in the molecules. Firstly, the signals of C₂₀H₃₀O₇ and C₂₀H₃₀O₉ decrease by 40 %–60 % with the addition of 30 ppb NO₂, indicating a significant contribution of acyl RO₂ such as C₁₀H₁₅O₅-RO₂ (acyl RO₂ in default MCM v3.3.1) and C₁₀H₁₅O₇-RO₂ in their formation, while other C₂₀H₃₀O_x dimers decrease by ∼30 %. The C₂₀H₃₂O_x dimer series also exhibits a small reduction (less than 20 %) with the addition of NO₂. However, the C₂₀H₃₄O_x series shows an unexpected increase with the addition of NO₂ up to 10 ppb and almost remains unchanged with the further increase in NO₂ concentration. Given that the cross-reaction rate constant of acyl RO₂ can be orders of magnitude higher than that of counterpart alkyl RO₂ (Atkinson et al., 2007; Orlando and Tyndall, 2012), the rapid consumption of acyl RO₂ by NO₂ would preserve the alkyl RO₂ that tend to react with acyl RO₂ at a fast rate in the absence of NO₂, which to some extent would elevate the concentration of alkyl RO₂ in the system and thus promote the less competitive alkyl RO₂ + alkyl RO₂ reactions to form C₂₀H₃₄O_x dimers. The slight increase in some C₁₀H₁₈O_x HOMs with the addition of NO₂ up to 10 ppb could also be for this reason.

According to the noticeable increasing trend in C₂₀H₃₄O_x as compared to other C₂₀ dimers, we speculate that acyl RO₂ react faster with C₁₀H₁₇O_x alkyl RO₂ than with C₁₀H₁₅O_x alkyl RO₂. Therefore, when the acyl RO₂ are depleted, the preservation of C₁₀H₁₇O_x-RO₂ is more significant, and the promotion of their cross-reactions to form C₂₀H₃₄O_x is more evident. It is also possible that the reaction of NO₂ with C₁₀H₁₇O_x alkyl RO₂ is less efficient compared to the reaction with C₁₀H₁₅O_x alkyl RO₂, so more C₁₀H₁₇O_x than C₁₀H₁₅O_x is available for dimer formation in the presence of NO₂.

To further prove the above two speculations, we performed sensitivity analyses for the reaction rates of C₁₀H₁₇O_x-RO₂ using F0AM. Figure 11a shows the changes in C₂₀H₃₄O_x dimers with NO₂ addition at different C₁₀H₁₇O_x-RO₂ + NO₂ reaction rates under the conditions of Exps 8–14. As the reaction rate varies from $\mathrm{1}\times {\mathrm{10}}^{-\mathrm{13}}$ to $\mathrm{1}\times {\mathrm{10}}^{-\mathrm{12}}$ ${\mathrm{cm}}^{\mathrm{3}}\mathrm{molec}{.}^{-\mathrm{1}}{\mathrm{s}}^{-\mathrm{1}}$, the increasing trend of C₂₀H₃₄O_x dimers versus the added NO₂ concentration is weakened, and the simulation is more deviated from the measurements. When the reaction rate increases to $\mathrm{7.5}\times {\mathrm{10}}^{-\mathrm{12}}$ ${\mathrm{cm}}^{\mathrm{3}}\mathrm{molec}{.}^{-\mathrm{1}}{\mathrm{s}}^{-\mathrm{1}}$, the C₂₀H₃₄O_x dimers decrease significantly with increasing NO₂, which is in striking contrast to the measurements. Figure 11b presents the sensitivity analysis results for the cross-reaction rate constants of acyl RO₂ + C₁₀H₁₇O_x-RO₂. As this rate constant varies from $\mathrm{1}\times {\mathrm{10}}^{-\mathrm{12}}$ to $\mathrm{1}\times {\mathrm{10}}^{-\mathrm{10}}$ ${\mathrm{cm}}^{\mathrm{3}}\mathrm{molec}{.}^{-\mathrm{1}}{\mathrm{s}}^{-\mathrm{1}}$, the increasing trend of C₂₀H₃₄O_x versus the NO₂ concentration is more pronounced and more consistent with the measurements. These sensitivity analyses support our speculation that the C₁₀H₁₇O_x alkyl RO₂ may be different from other alkyl RO₂ radicals in terms of the reaction efficiency with NO₂ and acyl RO₂ species, which leads to different responses of C₂₀H₃₄O_x dimers to NO₂ addition compared to other C₂₀ dimers. These results also suggest that the presence of acyl RO₂ could affect the fate and contribution of alkyl RO₂ to HOM formation in atmospheric oxidation systems given the different reactivity of acyl RO₂ from alkyl RO₂.

Figure 11 Sensitivity of C₂₀H₃₄O_x dimer production to (a) the reaction rates of NO₂ with C₁₀H₁₇O_x-RO₂ and (b) the cross-reaction rate of acyl RO₂ with C₁₀H₁₇O_x-RO₂ considering a C₁₀H₁₇O_x-RO₂+NO₂ reaction rate of $\mathrm{1}\times {\mathrm{10}}^{-\mathrm{12}}$ ${\mathrm{cm}}^{\mathrm{3}}\mathrm{molec}{.}^{-\mathrm{1}}{\mathrm{s}}^{-\mathrm{1}}$.

4 Conclusions

In this study, the molecular identities, formation mechanisms, and contributions of acyl RO₂ to the formation of HOMs during ozonolysis of α-pinene are investigated using a combination of flow reactor experiments and detailed kinetic model simulations. Based on the marked decrease in RO₂ signal as a function of initial NO₂ concentration, a total of 10 highly oxygenated acyl RO₂ are identified during α-pinene ozonolysis. The acyl RO₂ contribute to 67 %, 94 %, and 32 % of C₇, C₈, and C₉ highly oxygenated RO₂ but only 0.4 % of C₁₀ highly oxygenated RO₂, respectively, when NO is absent. Three main pathways are identified for the formation of monoterpene-derived acyl RO₂: (i) the autoxidation of RO₂ containing aldehyde groups, (ii) the cleavage of a C–C bond of RO containing an α-ketone group, and (iii) the intramolecular H shift of RO containing an aldehyde group. The autoxidation of aldehydic RO₂ formed involving multiple RO decomposition or ring-opening steps plays a dominant role in the formation of the highly oxygenated acyl RO₂ radicals (oxygen atom number ≥6), while the less-oxygenated acyl RO₂ (oxygen atom number <6) are mainly derived from the other two pathways.

The acyl-RO₂-involved reactions explain 50 %–90 % of C₇ and C₈ HOM monomers and 14 % of C₁₀ HOMs, respectively. For C₉ HOMs, this contribution can be up to 30 %–60 %. For the HOM dimers, acyl-RO₂-involved reactions contribute 50 %–95 % to the formation of C₁₄–C₁₈ dimers. Owing to the higher cross-reaction rate constant of acyl RO₂ compared to alkyl RO₂, the acyl RO₂ + alkyl RO₂ reaction would outcompete the alkyl RO₂ + alkyl RO₂ reaction. Therefore, the rapid consumption of acyl RO₂ by NO₂ in the experiments (as well as in polluted atmospheres) would make the alkyl RO₂ that are supposed to react with acyl RO₂ retained, which to some extent elevates the concentration of alkyl RO₂ in the system and thus promotes the reaction of alkyl RO₂ + alkyl RO₂ to form dimers such as C₂₀H₃₄O_x. The contribution of H abstraction of α-pinene by OH radical to the formation of acyl RO₂ and HOMs is found to be negligible in the absence of NO. This is because the primary C₁₀H₁₅O₂-RO₂ radicals formed in such pathways are least-oxidized and thus have relatively low cross-reaction efficiency to produce RO radicals, which are the key intermediates for the formation of acyl RO₂ and HOMs in that channel. However, in the presence of NO, the formation of highly oxygenated acyl RO₂ via the H-abstraction pathway is demonstrated, consistent with previous studies (Shen et al., 2022).

In this study, acyl RO₂ species are identified according to a dramatic decrease in their signals with the addition of NO₂. It should be noted that the presence of NO₂ could also inhibit the formation of alkyl RO₂ species involving acyl RO₂ reactions. If there are any contributions of alkyl RO₂ to acyl RO₂ identified in this study, the influence of such alkyl RO₂ species on HOM formation would reflect an indirect effect of acyl RO₂. However, given that the formation of most of the acyl RO₂ identified in this study can be reasonably explained by the proposed mechanisms and verified by their responses to the addition of NO, the acyl RO₂ identified here are expected to have no significant contributions from alkyl RO₂. Currently, the reaction kinetics of monoterpene-derived acyl RO₂ are still poorly understood. Considering the important contribution of acyl RO₂ to HOM formation, further kinetic studies are needed to get more specific rate constants for their autoxidation and cross-reactions, thereby deepening our understanding of the role of acyl RO₂ in HOM and SOA formation under atmospheric conditions.