Formation and temperature dependence of highly oxygenated organic molecules (HOMs) from Δ³-carene ozonolysis

Δ³-carene is a prominent monoterpene in the atmosphere, contributing significantly to secondary organic aerosol (SOA) formation. However, knowledge about Δ³-carene oxidation pathways, particularly regarding their ability to form highly oxygenated organic molecules (HOMs), is still limited. In this study, we present HOM measurements during Δ³-carene ozonolysis under various conditions in two simulation chambers. We identified numerous HOMs (monomers: C₇₋₁₀H₁₀₋₁₈O₆₋₁₄; dimers: C₁₇₋₂₀H₂₄₋₃₄O₆₋₁₈) using a chemical ionization mass spectrometer (CIMS). Δ³-carene ozonolysis yielded higher HOM concentrations than α-pinene, with a distinct distribution, indicating differences in formation pathways. All HOM signals decreased considerably at lower temperatures, reducing the estimated molar HOM yield from ∼ 3 % at 20 °C to ∼ 0.5 % at 0 °C. Interestingly, the temperature change altered the HOM distribution, increasing the observed dimer-to-monomer ratios from roughly 0.8 at 20 °C to 1.5 at 0 °C. HOM monomers with six or seven O atoms condensed more efficiently onto particles at colder temperatures, while monomers with nine or more O atoms and all dimers condensed irreversibly even at 20 °C. Using the gas- and particle-phase chemistry kinetic multilayer model ADCHAM, we were also able to reproduce the experimentally observed HOM composition, yields, and temperature dependence.

1 Introduction

Secondary organic aerosol (SOA), formed through gas-to-particle conversion in the atmosphere, constitute a major contributor to the global submicron aerosol mass (Hallquist et al., 2009). Atmospheric SOA formation has important implications for the climate and for human health (Shiraiwa et al., 2017; Cohen et al., 2017; Shrivastava et al., 2017; Jimenez et al., 2009). The largest precursor of SOA is biogenic volatile organic compounds (BVOCs), which are emitted naturally by vegetation and dominate global VOC emissions (Guenther et al., 2012). In the ambient air, BVOCs can react with a variety of oxidants, such as ozone (O₃), hydroxyl radical (OH), or nitrate radical (NO₃), to produce more functionalized organic products. Highly oxygenated organic molecules (HOMs) are a recently identified group of VOC oxidation products formed through rapid autoxidation processes in the atmosphere (Ehn et al., 2014; Bianchi et al., 2019). Containing six or more oxygen atoms, HOMs are typically highly oxidized and functionalized species with low volatilities, making them crucial in SOA formation through condensation or reactive uptake (Ehn et al., 2014; Bianchi et al., 2019, 2016).

Monoterpenes (C₁₀H₁₆) account for approximately 15 % of the annual global BVOC emissions (Guenther et al., 2012). Previous studies have reported a broad range of SOA yields resulting from monoterpene oxidation, with values spanning from less than 1 % to over 60 % (Saathoff et al., 2009; Warren et al., 2009; Ehn et al., 2014; Hallquist et al., 1999; Kristensen et al., 2020; Thomsen et al., 2022). This variation highlights the significant disparities in monoterpene oxidation mechanisms and the potential of their products to form SOA under different conditions. The most abundantly emitted monoterpene, α-pinene, has been the subject of numerous laboratory and field studies (e.g. Ehn et al., 2014; Berndt et al., 2003; Zhao et al., 2023; Molteni et al., 2019; Tillmann et al., 2010; Kristensen et al., 2020). Many modellers have also employed α-pinene as a representative compound for endocyclic monoterpenes in the regional or global aerosol budget (Boy et al., 2013; Pye et al., 2010). To date, however, the fate of other monoterpenes in the atmosphere remains less understood, and evaluating the variability in their impact on SOA formation continues to be a challenge.

Δ³-carene is a bicyclic unsaturated monoterpene (Scheme 1), distinguished from α-pinene's structure primarily by its three-membered ring. Despite being predicted to have lower emissions than α-pinene at a global scale (Sindelarova et al., 2014), Δ³-carene has been measured in equivalent proportions in ambient air in certain regions (Fry et al., 2013; Kim et al., 2013; Geron et al., 2000; Bäck et al., 2012). SOA yields from the photochemical oxidation of Δ³-carene have been determined to be 2 %–38 % (Hoffmann et al., 1997; Griffin et al., 1999; Lee et al., 2006), and D'Ambro et al. (2022) detected both gas- and particle-phase products from OH oxidation of Δ³-carene and developed a mechanism for the initial stage of carene–OH oxidation with the support of computational chemistry. Recently, experimental and theoretical research on HOM formation from NO₃ oxidation of Δ³-carene has also been carried out (Dam et al., 2022; Draper et al., 2019; Liu et al., 2022; Day et al., 2022), and it has been suggested that the higher SOA yield from NO₃-initiated oxidation of Δ³-carene (15 %–65 %) compared with α-pinene (0 %–16 %) (Hallquist et al., 1999; Fry et al., 2014) might be due to differences in the potential for further radical propagation and oxidation of the first-generation radicals from early unimolecular processes. The ozonolysis of Δ³-carene is also of particular importance as this process can contribute to both SOA and OH formation in the atmosphere. Previous studies have reported that Δ³-carene ozonolysis has similar or slightly higher SOA yields compared to α-pinene ozonolysis under similar conditions (Thomsen et al., 2021; Thomsen et al., 2022). Several studies have measured various less-oxidized products (e.g. organic acids) from Δ³-carene ozonolysis and explored their possible formation pathways (Wang et al., 2019; Glasius et al., 2000; Baptista et al., 2014; Ma et al., 2009). However, very little is currently known about the HOM formation resulting from O₃-initiated Δ³-carene oxidation (Li et al., 2019; Mentel et al., 2015). The only source providing HOM spectra from Δ³-carene ozonolysis is the study by Li et al. (2019).

Scheme 1 Simplified formation mechanism of the different primary RO₂ from ozone- and OH-initiated oxidation of Δ³-carene. POZ: cyclic primary ozonide; CI: Criegee intermediate; VHP: vinyl hydroperoxide.

In this study, we conducted a series of chamber experiments to investigate the HOM formation from Δ³-carene ozonolysis with and without the presence of an OH scavenger. We present the differences in HOM formation between Δ³-carene and α-pinene ozonolysis concerning potential pathways, molar yield, and composition. Additionally, we estimate the relative volatilities of the detected HOM species to qualitatively assess their contributions to SOA formation. We also explore the impact of temperature on the composition and distribution of HOMs. Finally, the aerosol dynamics and gas- and particle-phase chemistry kinetic multilayer model ADCHAM (Roldin et al., 2019, 2014) was utilized to simulate the HOM formation from Δ³-carene ozonolysis and to be compared to our experimental results.

2 Materials and methods

2.1 Chamber facilities and instrumentation

Δ³-carene ozonolysis experiments were performed in two different chambers: (a) the COALA chamber at the University of Helsinki, Finland, to assess HOM compositions and their potential formation pathways and (b) the AURA chamber at Aarhus University in Denmark to investigate the impact of temperature and relative humidity (RH) on HOM formation.

The COALA chamber is a 2 m³ Teflon reactor maintained at room temperature (25±1 °C) under dry conditions (RH < 1 %). During this campaign, the chamber was run in continuous mode with a total inflow of 40 L min⁻¹ (average residence time: ∼ 50 min). We conducted 11 Δ³-carene ozonolysis experiments under different oxidation conditions (Supplement Table S1). A proton-transfer-reaction time-of-flight mass spectrometer (PTR-TOF 8000, Ionicon Analytik GmbH) was deployed to measure VOC concentrations, while a chemical ionization atmospheric pressure interface time-of-flight mass analyser (CIMS, Tofwerk AG/Aerodyne Research, Inc.) with nitrate (NO${}_{\mathrm{3}}^{-}$) as the reagent ion (hereafter NO₃-CIMS) was employed to probe oxygenated products from Δ³-carene ozonolysis, with a primary focus on HOMs. In this paper, the signal intensity of species detected with NO₃-CIMS is presented as a normalized signal, which refers to the raw signal intensity normalized to the reagent ions, unless specified otherwise.

The AURA chamber is a 5 m³ Teflon chamber situated in a temperature-controlled room (temperature range: −16–26 °C). Throughout the campaign, the AURA chamber was run in batch mode, meaning that all reagents are injected in a single batch at the start of the experiment and that products accumulate progressively. As shown in Table S2, the HOM formation of Δ³-carene ozonolysis was examined under dry conditions (RH < 15 %) at 20 °C (20A and B), 10 °C (10A and B), and 0 °C (0A) and twice under humid conditions (RH 80 %) at 10 °C with two different Δ³-carene loadings (10D: 10 ppb; 10E: 20 ppb). The experiment commenced with the introduction of Δ³-carene into the chamber, marking the time as an experiment time of 0 min. Instruments for both gas-phase and particle-phase measurements were deployed. For HOM measurement, the same type of NO₃-CIMS as that employed in the COALA lab was utilized.

Note that in this paper, in accordance with most previous studies, we term the set of compounds that we observe with the NO₃-CIMS that match the Bianchi et al. (2019) criteria as HOMs. These are species containing six or more oxygen atoms, formed in the gas phase via autoxidation involving peroxy radicals (RO₂) under atmospherically relevant conditions. It is acknowledged, however, that some compounds may not be detected or are detected with lower sensitivity. The schematic of these two chambers is shown in Fig. S1, and more details of the setups and instruments are provided in Supplement Sect. S1. Importantly, the differences between batch- and continuous-mode chamber experiments are described there.

2.2 Δ³-carene ozonolysis chemistry

For the purpose of interpreting our mass spectral observations, we provide a brief overview of the main reaction pathways from the ozonolysis of Δ³-carene. An initial addition of O₃ to the double bond (Scheme 1) results in Criegee intermediates (CIs) that undergo unimolecular isomerization followed by OH loss and O₂ addition, forming “primary” RO₂ with an elemental composition of C₁₀H₁₅O₄. With a suitable structure, RO₂ intramolecular H-shifts and O₂ addition (i.e. autoxidation) can take place, leading to high oxygen content of the formed RO₂ (C₁₀H₁₅O_even). The reaction between Δ³-carene and O₃ also produces OH, as depicted in Scheme 1, and the OH yields were reported to be 0.56–1.1 by previous studies (Wang et al., 2019; Hantschke et al., 2021; Atkinson et al., 1992; Aschmann et al., 2002). OH can react with Δ³-carene as well, where it can either attach to the double bond, resulting in an initial RO₂ as C₁₀H₁₇O₃, or abstract a hydrogen atom to form an RO₂ as C₁₀H₁₅O₂. Subsequent autoxidation is expected to produce RO₂ with the formulas of C₁₀H₁₇O_odd and C₁₀H₁₅O_even from addition and abstraction, respectively. However, it should be noted that for VOCs with double bonds, the OH-abstraction pathway is not typically significant (Atkinson and Arey, 2003), and the dominant source of C₁₀H₁₅O_even will be the ozone reactions.

Highly oxidized RO₂ can terminate to closed-shell compounds, i.e. HOM monomers or HOM dimers, via either unimolecular decomposition (typically Reaction R1) or bimolecular reactions with other RO₂ (Reactions R2–R5) and hydroperoxyl radicals (HO₂) (Reactions R6–R8) in our system.

$$\begin{array}{}\text{(R1)}& {\mathrm{RO}}_{\mathrm{2}}\to {\mathrm{R}}_{-\mathrm{H}}^{\prime }=\mathrm{O}+\mathrm{OH}\end{array}$$

When RO₂ reacts with other RO₂, a tetroxide intermediate is formed and rapidly decomposes to a complex of two alkoxy radicals (RO) and releases an oxygen molecule. The RO complex can then undergo different processes resulting in various products. Firstly, the complex can directly decompose to two RO (Reaction R2), which subsequently generate closed-shell species after HO₂ loss (Reaction R9) or alkyl radicals via channels (Reactions R10, R11). The alkyl radicals related to pathways in Reactions (R10) and (R11) can either terminate unimolecularly or ultimately reform a new RO₂. For example, C₁₀H₁₅O_odd and C₁₀H₁₇O_even can be formed from reactions of C₁₀H₁₅O_even and C₁₀H₁₇O_odd via Reactions (R2), (R10), and (R11). Secondly, the RO complex can decompose to a carbonyl and an alcohol, as simplified in Reaction (R3) (Vereecken and Peeters, 2009). Thirdly, two RO of the complex can recombine into a ROOR' accretion product (i.e. HOM dimer) (Hasan et al., 2020) after intersystem crossing (Reaction R4). To illustrate, the reactions between O₃-initiated RO₂ C₁₀H₁₅O_even and OH-initiated RO₂ C₁₀H₁₇O_odd would produce C₂₀H_30,34O_even and C₂₀H₃₂O_odd dimers, depending on the RO₂ combinations. Finally, the RO complex could also undergo other untypical reactions (Reaction R5), for example, β scission of one RO in the complex (Peräkylä et al., 2023), which are unique for RO with suitable structures.

$$\begin{array}{}\text{(R2)}& {\displaystyle }\begin{array}{rl}{\mathrm{RO}}_{\mathrm{2}}+& {\mathrm{R}}^{\prime }{\mathrm{O}}_{\mathrm{2}}\to {\mathrm{ROOOOR}}^{\prime }\stackrel{-{\mathrm{O}}_{\mathrm{2}}}{⟶}\\ & \mathrm{RO}\mathrm{\cdots }{\mathrm{OR}}^{\prime }\to \mathrm{RO}+{\mathrm{R}}^{\prime }\mathrm{O}\end{array}\text{(R3)}& {\displaystyle }\to \mathrm{ROH}+{{\mathrm{R}}^{\prime }}_{-\mathrm{H}}\mathrm{C}=\mathrm{O}\text{(R4)}& {\displaystyle }\to \mathrm{RO}\mathrm{\cdots }{\mathrm{OR}}^{\prime }\text{(R5)}& {\displaystyle }\to \mathrm{other}\mathrm{products}\end{array}$$

Similarly to the reactions of two RO₂, the reactions of RO₂ with HO₂ can lead to either termination (Reactions R6, R7) (Groß et al., 2014; Praske et al., 2015; Schwantes et al., 2015) or radical propagation (Reaction R8) (Hasson et al., 2005). It is generally expected that HO₂ reactions terminate autoxidation following the reaction channel (Reaction R6). However, other reaction channels also play a significant role for more complex RO₂; e.g. the reaction channel in Reaction (R8) was found to have a high yield for acylperoxy radicals (Groß et al., 2014; Hasson et al., 2005). Therefore, for example, it is theoretically possible that O₃-initiated RO₂ C₁₀H₁₅O_even can form closed-shell monomers C₁₀H₁₆O_x or generate new RO₂ C₁₀H₁₅O_odd via different reactions with HO₂. Note that in this study, the reactions of RO₂ with HO₂ were expected to prevail only when CO was injected into the chamber.

$$\begin{array}{}\text{(R6)}& {\displaystyle }{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{HO}}_{\mathrm{2}}\to \mathrm{ROOH}+{\mathrm{O}}_{\mathrm{2}}\text{(R7)}& {\displaystyle }\to \mathrm{ROH}+{\mathrm{O}}_{\mathrm{3}}\text{(R8)}& {\displaystyle }\to \mathrm{RO}+\mathrm{OH}+{\mathrm{O}}_{\mathrm{2}}\text{(R9)}& {\displaystyle }\mathrm{RO}+{\mathrm{O}}_{\mathrm{2}}\to {{\mathrm{R}}^{\prime }}_{-\mathrm{H}}=\mathrm{O}+{\mathrm{HO}}_{\mathrm{2}}\text{(R10)}& {\displaystyle }\mathrm{RO}\stackrel{\mathrm{scission}}{⟶}\mathrm{products}\text{(R11)}& {\displaystyle }\stackrel{\text{H-shift isomerization}}{⟶}\mathrm{products}\end{array}$$

2.3 ADCHAM modelling

We used the ADCHAM model to simulate the gas-phase chemistry, HOM formation, and SOA formation during the AURA and COALA experiments. The general model setup, aerosol dynamics, and predicted SOA formation during the experiments are described by Thomsen et al. (2024). Here we only describe the new Δ³-carene gas-phase oxidation mechanism that was implemented in ADCHAM. ADCHAM incorporates a comprehensive model for autoxidation and HOM formation for α-pinene, serving as the foundational basis for the carene ozonolysis model. In this study, we updated the ADCHAM model drawing upon prior research on the ozonolysis and OH oxidation of Δ³-carene (D'Ambro et al., 2022; Hantschke et al., 2021; Wang et al., 2019) alongside HOM and SOA data from our campaign. The first-generation reaction rates and branching ratios of the implemented Δ³-carene gas-phase chemistry mechanism are based on the theoretical work on the ozonolysis of Δ³-carene by Wang et al. (2019) and on the experimental work on ozonolysis and OH oxidation of Δ³-carene by Hantschke et al. (2021). According to the theoretical work by Wang et al. (2019), four different Criegee intermediate (CI) conformers are formed during the ozonolysis of Δ³-carene. These CIs undergo prompt unimolecular reactions to form secondary ozonides (SOZs), vinyl hydroperoxides (VHPs), dioxiranes (DIOs), and stabilized CIs (SCIs). The VHPs decompose rapidly and form RO₂ and OH, while the SOZs can isomerize promptly to 3-caronic acid. According to the calculations by Wang et al. (2019), the VHP and OH yield during the ozonolysis of Δ³-carene is ∼ 56 %, the SOZ yield (3-caronic acid yield) is ∼ 24 %, the DIO yield is 16 %, and the SCI yield is 4 %. The theoretically derived OH yield from Wang et al. (2019) is in reasonable agreement with the experimentally derived OH yield of 65 % by Hantschke et al. (2021). Since the existence and fate of dioxiranes are largely unknown (Hantschke et al., 2021), we exclude the proposed reaction pathways leading to these products in the present work. Instead, we assume that the initial CIs exclusively decompose to 65 % VHPs and 35 % SOZs, which result in final first-generation ozonolysis product yields of 65 % RO₂ + OH and 35 % 3-caronic acid.

The first-generation RO₂ is distributed among four distinct isomers. It is hypothesized that one major RO₂ isomer, analogous to the first-generation α-pinene ozonolysis product C109O2 in MCMv3.3.1 and named D3C109O2, predominates with a molar yield (branching ratio) of 96.2 %. The remaining three isomers represent C₁₀ RO₂ that can undergo autoxidation of peroxy radicals, thereby forming HOMs, as detailed in Table S6. In addition, we have included a minor route to C₉ RO₂ that can undergo peroxy radical autoxidation. These C₉ RO₂ are expected to be formed as a second-generation bimolecular reaction product when D3C109O2 reacts with other RO₂ or NO. To capture the observed profound impact of CO on the HOM mass spectrum evolution in COALA, the lower RO₂ + RO₂ reaction rates (on the order of ∼ 5 times lower) for Δ³-carene than in the earlier peroxy radical autoxidation mechanism (PRAM) (Roldin et al., 2019; Nie et al., 2023) were utilized. Otherwise, the RO₂ + RO₂ termination reactions still dominate over the RO₂ + HO₂ reactions after the CO addition in COALA when assigning the MCM generic rate coefficient KRO2HO2 ≈ 2 × 10⁻¹¹ molec.⁻¹ cm³ s⁻¹ for all RO₂ + HO₂ reactions in the model. Additionally, we accounted for different RO₂ + HO₂ reaction pathways (Reactions R6–R8) in this presented Δ³-carene mechanism, while, in the previous monoterpene PRAM (Roldin et al., 2019; Nie et al., 2023), all RO₂ + HO₂ reactions resulted in closed-shell HOM ROOH (C₁₀H₁₆O_x) (Reaction R6).The molar yields of the four RO₂ that can undergo autoxidation are summarized in Table S3. Table S4 provides the Δ³-carene ozone chemistry mechanism, excluding the PRAM, and Table S5 provides the Δ³-carene OH oxidation mechanism. Tables S6 and S7 list all PRAM reactions. The full mechanism, in a format compatible with the kinetic pre-processor, will be provided upon publication in an open-access repository at Zenodo (Roldin, 2024).

3 Results and discussion

3.1 HOM formation and general features

During our experiments in the COALA chamber, numerous HOMs with a broad oxygenation pattern were observed as nitrate adducts in the mass range from 260 Th to 600 Th, as illustrated in Fig. 1a. The spectra observed in our study are similar to those reported by Li et al. (2019), who conducted Δ³-carene ozonolysis under high concentrations (VOC: ∼ 1100 ppb; O₃: ∼ 900 ppb) at room temperature and under dry conditions. Both studies identified predominant monomers as C₁₀H_14,16O_7,9 in both spectra and the most abundant dimer groups as C₂₀H₃₂O${}_{\mathrm{7},\mathrm{9},\mathrm{11}}$ and C₁₉H₃₀O${}_{\mathrm{6},\mathrm{10},\mathrm{11}}$. However, the relative abundances of the dominant species differed slightly. For instance, C₂₀H₃₂O₁₁ exhibited greater abundance in our study, whereas the reverse was true in Li et al. (2019). These disparities could stem from variations in experimental conditions and instrumentation. The HOM distributions from Δ³-carene ozonolysis showed many similarities with the corresponding α-pinene HOMs under identical conditions (Figs. 1b and 2a), as was expected considering their structural similarity. However, clear differences were also observed. Although C₁₀H_14,16O_odd were the dominant monomer groups in both systems, the largest peaks in Δ³-carene ozonolysis had a higher oxygen content (C₁₀H_14,16O₉) than those in α-pinene ozonolysis (C₁₀H_14,16O₇). Furthermore, C₉H_12,14O₉ were only abundant in the Δ³-carene ozonolysis system. The differences in the main dimers were even more pronounced: C₂₀H₃₂O_m and C₁₉H₃₀O_n were the most abundant dimer groups in Δ³-carene ozonolysis, whereas, in the α-pinene system, larger C₁₉H₂₈O_n signals instead of C₁₉H₃₀O_n were observed.

Figure 1 Unit mass resolution (UMR) mass spectra from (a) Δ³-carene and (b) α-pinene ozonolysis in the COALA chamber under the same conditions (VOC: 20 ppb; O₃: 30 ppb). All peaks labelled were detected as a cluster with NO${}_{\mathrm{3}}^{-}$, and the dashed grey lines mark some of the products with the same formulas detected in both Δ³-carene and α-pinene ozonolysis experiments.

Figure 2 Scatter plots of the HOM normalized signal intensity from different experiments. Each marker corresponds to a single detected composition. The subplots depict comparisons between (a) Δ³-carene ozonolysis (Experiment 10) and α-pinene ozonolysis (Experiment 19) in the COALA chamber, (b) Δ³-carene ozonolysis without CO injection (Experiment 10) and with CO injection (Experiment 11) in the COALA chamber, (c) Δ³-carene ozonolysis in the COALA chamber and Experiment 20B in the AURA chamber, (d) Δ³-carene ozonolysis in the AURA chamber at 10 °C (10B) and at 20 °C (20B), and (e) Δ³-carene ozonolysis in the AURA chamber at 0 °C (0A) and at 20 °C (20B). The colour indicates the O atom content in the identified species, and markers distinguish monomers and dimers. The solid lines shown in all subplots are the 1:1 lines.

By grouping the detected HOMs based on C atom and O atom numbers, we can see different contributions of HOM groups between the Δ³-carene and α-pinene ozonolysis systems (Fig. 3). A substantially higher fraction of C₁₀ compounds and a slightly higher fraction of C₉ compounds were observed for Δ³-carene ozonolysis, while C₈ and C₅ compounds were nearly twice as abundant for α-pinene ozonolysis HOMs. The most significant difference within the C₉ and C₁₀ groups in these two systems was attributed to the O₉ species, which contributed more than 2-fold higher to the total HOMs in the Δ³-carene system. Dam et al. (2022) also observed a larger contribution of C₉ species during the NO₃ radical oxidation of Δ³-carene (∼ 27 %) compared to α-pinene (∼ 10 %). However, no significant C₉ signal was detected during OH oxidation of Δ³-carene by D'Ambro et al. (2022). The reason for the larger concentration of C₉ species in Δ³-carene ozonolysis remains unexplained based on our results. In addition, the concentration-weighted number of O atoms for C₁₀ HOMs was similar; however, in the Δ³-carene system, C₉ had a slightly higher concentration-weighted number of O atoms, indicating the C₉ monomers were more oxidized. C₁₇₋₂₀ dimers constituted approximately 33 % and 21 % of the total HOMs from Δ³-carene and α-pinene ozonolysis, respectively. Interestingly, C_19,20 predominated the HOM dimers from Δ³-carene ozonolysis, while the dimer contributions from α-pinene ozonolysis exhibited a steady increasing trend as the number of C atoms increased from 17 to 20. Moreover, due to the higher fractions of O_6,7 species in C_19,20 groups, the concentration-weighted numbers of O atoms in the Δ³-carene system were lower.

Figure 3 Fractions (left-hand y axis) of different HOMs to the total detected HOMs as a function of the C atom and O atom number formed from (a) Δ³-carene and (b) α-pinene ozonolysis in the COALA chamber under the same conditions (VOC: 20 ppb; O₃: 30 ppb). The dots (right-hand y axis) show the average concentration-weighted number of O atoms in each HOM group with the same number of C atoms.

Most HOM monomers shown in Fig. 1a can be explained by the standard RO₂ chemistry described in Sect. 2.2. For instance, C₁₀H₁₄O_7.9,11 are likely derived from the O₃-initiated RO₂ (C₁₀H₁₅O_even) after unimolecular (Reaction R1) or bimolecular (Reactions R3, R8, and R9) terminations, while C₁₀H₁₆O${}_{\mathrm{6},\mathrm{8},\mathrm{10}}$ can be formed via the same termination reactions from the OH-initiated RO₂ C₁₀H₁₇O_odd. The latter can also form from C₁₀H₁₅O_even terminating by HO₂ (Reaction R6), highlighting the complexity of determining exact mechanisms solely from elemental composition measurements. C₁₀H₁₆O${}_{\mathrm{7},\mathrm{9},\mathrm{11}}$ HOMs might be explained by C₁₀H₁₅O_even terminating via the channel in Reaction (R3) or (R7). In the case of the C₉H_12,14O₉ HOMs, which stood out in the monomer region, they have undergone a fragmentation reaction, which may be associated with an alkoxy decomposition pathway like Reaction (R10). Extremely rapid RO scissions were recently shown to be highly competitive in the α-pinene ozonolysis system (Peräkylä et al., 2023), losing formaldehyde to become C₉ radicals, though in that case they always seemingly ended up forming dimer species. The reason why C₉ monomers were abundantly observed in only Δ³-carene ozonolysis remains unclear.

For all Δ³-carene ozonolysis experiments, we observed C₁₀H₁₅O_8,10 and C₉H₁₃O₁₀ as the three largest signals of radicals, consistent with the detected closed-shell HOMs. Although the COALA chamber is a steady-state chamber, we can examine the rates at which different species appear when we start adding reagents to the chamber. The appearance time for a species was determined when the subsequent signal change exceeded both the mean value and the standard deviation of the background level. Selected time series of closed-shell HOMs and radicals are shown in Fig. 4, and we can see that C₁₀H₁₄O₉ began to rise immediately once C₁₀H₁₅O₁₀ was formed, suggesting that C₁₀H₁₄O₉ was primarily formed from the unimolecular termination pathway (Reaction R1). However, C₉H_12,14O₉ and C₁₀H₁₆O₉, whose formation process was expected to involve at least one step of RO₂ + RO₂ reactions, started to increase ∼ 3 min later. This reinforces our speculation on the potential formation pathways of the most abundant HOM monomers. C₂₀H${}_{\mathrm{30},\mathrm{32},\mathrm{34}}$O_x dimers could originate from reactions between C₁₀H₁₅O_even and C₁₀H₁₇O_odd, depending on the oxidant combinations. Similarly, the formation of C₁₉H₃₀O_n and C₁₉H₂₈O_m (m, n ≥ 6) dimers can result from combinations of the two C₁₀ radicals with the C₉H₁₃O_x radicals that were potentially formed via the alkoxy scissions as described above.

Figure 4 (a) Time series of the selected HOMs and radical in Experiment 10 after the injection of Δ³-carene started. (b) The same data as in panel (a); however, each trace was normalized to its corresponding signal at 20:15 UTC+2 (the final data point in panel a) to display the relative change rate of the selected species. The red line is dashed to highlight that it is the only radical species (C₁₀H₁₅O₁₀) shown in this figure. The solid yellow line represents the normalized sum of the signal intensities of C₉H_12,14O₉. Meanwhile, the solid red line illustrates the sum of the normalized signal intensities of the eight highest dimers. The time resolution of the data shown in this figure is 30 s.

The relative impact of OH on HOM formation was investigated by injecting around 200 ppm CO into the COALA chamber, resulting in over 90 % of OH reacting with CO instead of Δ³-carene, forming a significant amount of HO₂ in the process (Gutbrod et al., 1997). The differences in HOM formation with and without CO presence are illustrated in Figs. S5a and 2b. Evidently, almost all dimer signals decreased significantly after CO addition, as the elevated HO₂ level increased the competitiveness of HO₂ + RO₂ reactions relative to RO₂ + RO₂ reactions. However, the HOM monomers responded differently upon the CO addition (Fig. S6a in the Supplement). C₁₀H₁₄O_odd and C₁₀H₁₆O_odd decreased, while C₁₀H₁₆O_even increased. The latter is easily explained by increased HO₂ termination of the O₃-derived RO₂, while the C₁₀H₁₆O_odd decrease was expected since its major source was believed to be OH-derived RO₂. Interestingly, C₁₀H₁₄O_even and C₉H_12,14O_even monomers, whose formation pathways were also expected to involve bimolecular reactions between RO₂, did not decline. One possible explanation is that C₁₀H₁₅O_even+ HO₂ reactions yield a considerable amount of RO via Reaction (R8). In contrast, no similar trends of C₁₀H₁₄O_even and C₉H_12,14O_even monomers were observed in the α-pinene system (Fig. S6b). All trends observed in the α-pinene system can be explained by the decrease in OH-initiated RO₂ and the reduced likelihood of RO₂+ RO₂ reactions. These findings again emphasize the differences in oxidation pathways between Δ³-carene and α-pinene systems.

Figure 2c displays the scatter diagram of the relationship between Δ³-carene ozonolysis HOMs observed in the AURA chamber at 20 °C and in the COALA chamber at room temperature (25±1 °C). HOM monomers with nine or more O atoms and most HOM dimers with decent signal intensities agreed very well. However, HOM monomers with six to eight O atoms were observed at higher concentrations in the COALA chamber. This discrepancy could result from various factors, including differences in the experimental conditions of the two chambers and variations in the configuration of the NO₃-CIMS used. Moreover, the COALA chamber spectrum was measured during the steady state, while the data from the AURA chamber were collected at an experiment time of 10 min. The HOM monomers with six to eight O atoms are expected to be close to semi-volatile, leading to more complex behaviour in terms of wall interactions and possible accumulation during an experiment. This topic is discussed in more detail in the next section.

3.2 Relative volatilities

For Δ³-carene ozonolysis experiments conducted in the AURA chamber, a typical time series of key compounds is shown in Fig. S7b. HOMs were formed via autoxidation and accumulated within the first 10 min following Δ³-carene injection (experiment time of 0 min), but concentrations began to decrease after this due to the increasing condensation sink (CS) caused by the newly formed particles. The lifetime of HOMs in the chamber was on the timescale of minutes, which meant that as long as the source (rate of Δ³-carene oxidation) and loss (condensation onto walls or particles) changed on longer timescales, the HOM concentration could be regarded as being balanced by the instantaneous source and loss terms for most of the experiments. This was at least true for the least volatile HOMs, which do not accumulate in the chamber over time. The ability of HOMs to condense is linked to their volatilities, with dimers generally classified as extremely low-volatility organic compounds (ELVOCs) that can irreversibly condense on particles (Peräkylä et al., 2020). Thus, we used dimer C₂₀H₃₂O₁₁ as a reference to probe the relative volatility of different HOM monomers compared to dimers. All other ELVOCs with a similar formation pathway should behave similarly to the dimers, while more volatile products were expected to accumulate in the chamber as their removal through condensation was less efficient. Hence, in each experiment, the change in the ratio of each HOM to C₂₀H₃₂O₁₁ indicates the relative volatility of that molecule, though it is crucial to recognize that differences in the formation pathways can also influence the ratios.

The signal ratio of each HOM to C₂₀H₃₂O₁₁ (M : C₂₀H₃₂O₁₁, where M represents the signal intensity of HOM M) at each time point was first calculated and then normalized by dividing it by the ratio value at an experiment time of 10 min. This calculated value is referred to as the “normalized ratio” in subsequent discussions. Assuming all dimers are ELVOCs at 20 °C, the normalized ratios of the eight largest dimers exhibited a range of approximately 0.5–1.2 at an experiment time of 70 min, as shown in Fig. S8b. This range provides a reference that suggests potential uncertainties in this method and indicates slightly different formation pathways for some dimers as well. For HOM monomers with nine or more oxygen atoms, the normalized ratios ranged from around 0.9 to 4 (Fig. 5b), comparable to the ratios of the HOM dimers, indicating that those monomers have similar condensation behaviour to dimers but potentially with minor accumulation over time or changes in formation pathways. However, for HOM monomers with eight oxygen atoms, e.g. C₁₀H₁₆O₈, the ratio was 1 order of magnitude higher than that of HOMs with nine or more oxygen atoms (e.g. C₉H₁₂O₉), indicating that HOM monomers with eight oxygen atoms from Δ³-carene ozonolysis are still to some extent semi-volatile. For HOMs with six or seven O atoms, the accumulation over the experiment was considerably higher, in the order of 10–100 times more, suggesting clearly higher volatilities compared to dimers. Interestingly, the ratios of HOMs with six to eight O atoms exhibited a significant decrease with decreasing temperature (Fig. 5b–d), suggesting their lower volatilities at colder temperatures.

Figure 5 (a) Normalized signal intensity of the selected HOMs during the 20B experiment and temporal behaviours of the ratios of different C_9,10 monomers (M) to the reference dimer (C₂₀H₃₂O₁₁) at (b) 20 °C (20B), (c) 10 °C (10B), and (d) 0 °C (0A) in the AURA chamber. The “normalized ratio” on the y axis in panels (b)–(d) was determined by first calculating the ratio M : C₂₀H₃₂O₁₁ at each time point, which was then normalized by dividing it by the ratio value at an experiment time of 10 min. While absolute concentrations may differ for different species due to varying branching ratios, the normalized ratio provides the relative change in M compared to C₂₀H₃₂O₁₁ as a function of time. In cases where the normalized ratio is close to unity for the entire experiment, both the formation and loss rates change similarly, meaning that their condensational loss (i.e. volatility) is equal. Larger normalized ratios indicate accumulation in the chamber over time, which is likely indicative of higher volatility. The shaded areas represent the range of normalized ratios for different HOM groups.

Peräkylä et al. (2020) evaluated the volatilities of HOMs from α-pinene ozonolysis at room temperature and observed similar trends in volatilities as a function of oxygen content, although the transition from semi-volatile to low-volatility products appears to take place between O₈ and O₉ in our Δ³-carene study, whereas their study on α-pinene ozonolysis had the transition between O₇ and O₈. The methods used were different, and our method is purely qualitative, so further work is needed to determine volatilities more quantitatively to assess if Δ³-carene HOMs are indeed slightly more volatile than α-pinene HOMs with the same elemental formulas.

3.3 Temperature impact on HOM formation

Temperature can strongly affect chemical reaction rates, particularly for unimolecular reactions (Rissanen et al., 2014; Kürten et al., 2015). In the AURA chamber, we performed Δ³-carene ozonolysis experiments at three different temperatures, and the detected HOMs are shown in Fig. 6a–c. The dominant peaks in both the HOM monomer and dimer ranges remained consistent across all temperatures, with C₁₀H_14,16O₉ and C₉H_12,14O₉ as the predominant HOM monomers and C₁₉H₃₀O${}_{\mathrm{6},\mathrm{10},\mathrm{11}}$ and C₂₀H₃₂O${}_{\mathrm{7},\mathrm{9},\mathrm{11}}$ as the largest dimers. However, HOM concentrations decreased with decreasing temperature (Fig. 2d and e), in part because the initial VOC oxidation slows down but probably to a much larger extent due to the autoxidation process becoming slower at colder temperatures.

Figure 6 (a–c) Unit mass resolution (UMR) mass spectra from Δ³-carene ozonolysis at 20, 10, and 0 °C and (d) the fractions of different HOM groups (O_X: X ≥ 6) to the total HOMs at 20, 10, and 0 °C in the AURA chamber. The datasets at 20, 10, and 0 °C are from experiments 20B, 10B, and 0A (Table S2), respectively. The signal intensities were multiplied by 2 at 0 °C in panel (c). The colour saturation in panel (d) represents the fraction, with the base colour corresponding to different temperatures shown in panels (a)–(c). All normalized signals were subtracted by the background and averaged over the period from 10 to 20 min after Δ³-carene injection.

Figure 6d indicates quite complex changes in the yields of different HOM groups. In general, the contributions of each HOM monomer group to the total HOMs decreased, while the fractions of most dimer groups increased at colder temperatures, resulting in an increase in HOM dimer-to-monomer ratios from 0.78 at 20 °C to 1.51 at 0 °C, which indicates that the dimers decreased at a slower rate than the monomers. The slower decrease in dimers may be due to a higher formation rate or a lower loss rate. Notably, these data were collected at an experiment time of 10 min when CS and wall loss for HOMs were not yet expected to be significant. Thus, the difference of a factor of ∼ 2 in dimer-to-monomer ratios could primarily be due to the differences in their formation. A plausible hypothesis is that colder temperatures may favour the dimerization pathway of RO₂ + RO₂, allowing the formed complex of two RO to remain bound for a longer time (Reaction R4). Consequently, the probability of RO₂ + RO₂ reactions forming dimers was higher, leading to a less significant decrease in the concentrations of dimers compared to monomers. Simon et al. (2020) also observed an increase in the dimer-to-monomer ratios as temperatures decreased within the temperature range studied for α-pinene ozonolysis. In contrast, Quéléver et al. (2019) reported that HOM dimers from α-pinene ozonolysis decreased at a faster rate than monomers when temperatures dropped. Those experiments in Quéléver et al. (2019) were performed at higher loadings, which might explain the difference. However, it is also possible that the dimer yields are different for the two systems under different temperatures, as the dimer formation mechanism is highly structure-dependent (Hasan et al., 2020; Valiev et al., 2019; Daub et al., 2022a, b; Hasan et al., 2021). Additionally, variations in experimental conditions (e.g. reagent concentrations) and instrumental settings (e.g. voltages) between the two studies represent a notable source of uncertainty that could contribute to these differences. It is also important to note that the colder sample air may cause some changes to the CIMS response, and we cannot rule out the influence of this. However, by the time the air enters the mass spectrometer itself, the air has very likely reached very close to room temperature, given the addition of a room-temperature sheath flow in the CI inlet. Thus, any dramatic changes, e.g. in mass-dependent transmission, are not expected.

In order to estimate HOM molar yields from the Δ³-carene ozonolysis, we need to determine the loss rates of the HOMs. For species with some semi-volatile character, this was not possible; therefore, our yield calculations are limited to HOM monomers with nine or more O atoms and all HOM dimers with more than six O atoms (hereafter referred to as “HOM_O ≥9”). Note that the yield of HOM_O ≥9 should be slightly lower than the total HOM yield, as we do not include all HOMs. The majority of the data from Δ³-carene ozonolysis experiments conducted in the COALA chamber can be explained by HOM_O ≥9 yields of 3 % to 6 % (Fig. S9). Although, as noted in Sect. S2, the absolute calibration of the NO₃-CIMS comes with large uncertainty, we can compare the resulting yield to that of α-pinene ozonolysis in the COALA chamber under identical conditions. From our experiments, we estimate a HOM_O ≥9 yield of 2 % to 4 % for α-pinene ozonolysis, which is in good agreement with previous studies (Jokinen et al., 2015; Ehn et al., 2014). Thus, our results suggest that the Δ³-carene ozonolysis yields around 1.5-fold higher HOM concentrations than α-pinene under identical conditions. We also estimated the HOM_O ≥9 yield for all the Δ³-carene ozonolysis experiments performed in the AURA chamber (Fig. 7), even though the calculation becomes slightly more complicated (Sect. S2). We found that the HOM_O ≥9 yield at 20 °C was ∼ 4 % for 20A and ∼ 3 % for 20B, which was within the range estimated for the experiments in the COALA chamber. However, the yields for the four experiments at 10 °C differ significantly, ranging from 0.2 % to 2.6 %. The large uncertainties of the yields at 10 °C might be attributed to the different settings of the NO₃-CIMS described in Sect. S1. Examining only the experiments conducted with the same instrumental settings (20B, 10B, and 0A; blue circles in Fig. 7), there is a clear decrease from around 3 % to below 1 % in the HOM_O ≥9 yields when the temperature dropped from 20 to 0 °C. This decrease is slightly larger than that observed for α-pinene ozonolysis by Simon et al. (2020) in the CLOUD chamber, in which the total HOM yields declined from 6.2 % at 25 °C to 4.7 % at 5 °C. However, Quéléver et al. (2019) reported a much larger drop (around 50-fold) in the α-pinene system upon a temperature decrease from 20 to 0 °C. Some of these differences may arise from the different conditions of the experiments, in particular the VOC loadings used. At higher loadings, the RO₂ lifetime is shorter, and a change in RO₂ H-shift rates may have a more dramatic impact when the competing reactions are faster. We hope more studies will focus on the temperature-dependent HOM yields in order to better understand these reported differences. We also again emphasize the large uncertainties (at least a factor 3) in our molar yield estimations, though the relative difference between the yields from Δ³-carene ozonolysis and α-pinene ozonolysis (using the same conditions and instruments) is expected to be much smaller.

Figure 7 Estimated HOM_O ≥9 (the sum of HOM monomers with no fewer than nine O atoms and all HOM dimers with more than six O atoms) molar yields in the AURA chamber at different temperatures. Colours represent different operational conditions of the NO₃-CIMS, which correspond to the colours in Fig. S2. The error bar marks the yield range estimated from Δ³-carene ozonolysis experiments in the COALA chamber at room temperature.

3.4 RH impact on HOM formation

Previous studies on the yield and distribution of HOMs have found them not to be affected significantly by RH (Li et al., 2019; Peräkylä et al., 2020), indicating that HOM formation pathways are largely water-independent. We performed two experiments (10D and 10E) in the AURA chamber under high RH conditions (RH of 80 %) with two different VOC loadings (Table S2), which could be used to check the RH impact on HOM formation. The mass spectra indicated that the main peaks of HOMs were similar under both dry and humid conditions (Fig. S10a–c). Although the NO₃-CIMS was also sensitive to water clusters, the detection precision of HOMs with the same $m/z$ as the water clusters (Fig. S10b–c) was slightly hampered. However, the absolute signal intensities of most HOM monomers and dimers at an RH of 80 % were approximately 7 times lower than those with a similar VOC and O₃ under dry conditions (Fig. S10d). Unfortunately, it is difficult to attribute this dramatic drop solely to the elevated RH, as the instrument settings were different between these two experiments (10B and 10D in Table S2), although we applied a correction factor to reduce the influence of the different settings on the HOM detection. High RH can increase the CS for HOMs, thereby reducing their concentrations; however, this effect is expected to be quite low in this case. We observed that the HOM signal increased by a factor of ∼ 3 when we injected twice the amount of Δ³-carene into the chamber at an RH of 80 %, and the rise in HOM dimer signals was even more pronounced than that of monomers (Fig. S10e). Overall, the quality of our data for this comparison is poor, and we cannot state with certainty whether the RH actually had an impact on the HOM formation pathways. Nevertheless, we mention these results here, hoping to prompt further studies to quantify the RH impact on HOM formation from Δ³-carene ozonolysis.

3.5 HOM simulation by ADCHAM

A modified peroxy radical autoxidation mechanism was developed specifically to replicate the observed HOM formation during the AURA and COALA Δ³-carene ozonolysis experiments. In Fig. 8, we compare the modelled and measured HOM concentrations during the COALA experiments where 200 ppm CO was added after approximately 5 h. The model is able to predict the observed absolute concentrations of HOM monomers and dimers and how the general patterns in the HOM observations change upon CO addition. Unimolecular termination (Reaction R1) of C₁₀H₁₅O₈ may explain why the C₁₀H₁₄O₇ concentration does not decrease substantially during the CO addition. However, the model cannot explain why the C₁₀H₁₆O₇ signal does not decrease appreciably during the CO addition. The model verifies that increasing HO₂ concentration upon CO addition can explain the observed decreasing concentrations of C₁₀H_14,16O₉, C₉H_12,14O₉, and HOM dimers and increasing concentrations of C₁₀H₁₆O_8,10 and C₉H₁₄O₁₀ (Fig. S11). The substantial decrease in C₁₀H_14,16O₉ but moderate increase in C₁₀H₁₆O₁₀ upon CO addition is captured by the model if assuming that only a minor fraction (∼ 25 %) of the C₁₀H₁₅O₁₀+ HO₂ reactions result in C₁₀H₁₆O₁₀ products via Reaction (R6). For all other PRAM RO₂+ HO₂ reactions, the ROOH formation via Reaction (R6) was regarded as the only production pathway.

Figure 8 Evaluation of modelled HOM concentrations during an ozonolysis Δ³-carene experiment in COALA (Experiment 10: VOC 20 ppb; O₃ 30 ppb), with the addition of CO (Experiment 11: VOC 20 ppb; O₃ 30 ppb; CO 200 ppm). Panel (a) shows the modelled and measured mass spectrum before CO addition, and panel (b) shows these values after CO addition. Panel (c) shows the modelled and measured total HOM monomer ($m/z$ 312–384 Th) and HOM dimer ($m/z$ 385–600 Th) concentrations, and panels (d), (e), and (f) show the concentrations of the major closed-shell HOM monomer species. For panels (c)–(f), CO was injected at a time of ∼ 5 h.

The absolute HOM_O ≥9 yields and their temperature dependencies agree reasonably well with the observations in AURA (Fig. 9). HOM_O ≥9 accounts for < 5 %, 5 %–9 %, and 12 %–15 % of the modelled total SOA mass in AURA at 0, 10, and 20 °C, respectively (Fig. S12). The HOM_O ≥9 SOA fraction depends on both the temperature and the VOC loading. The model also demonstrates how the HOM yields become higher and less sensitive to temperature in the chamber when the VOC concentration decreases to more typical atmospheric levels (< 1 ppbv), as observed in previous CLOUD chamber experiments (Simon et al., 2020; Nie et al., 2023).

Figure 9 Modelled HOM_O ≥9 molar yields for conditions with different fixed Δ³-carene concentrations and an ozone concentration of 30 ppb. The dashed line with error bars represents measured HOM_O ≥9 molar yields with 10 ppb initial Δ³-carene under dry conditions. At low (atmospherically relevant) Δ³-carene concentrations of ≤ 1 ppb, the HOM_O ≥9 yields decrease by less than 25 % between +25 and 5 °C, while, at Δ³-carene concentrations ≥ 10 ppb, the HOM_O ≥9 yields decrease by more than 50 %.

4 Conclusions

HOM formation from O₃-initiated Δ³-carene oxidation was investigated in two simulation chambers. Our findings reveal that ozonolysis of Δ³-carene yields HOM monomers (C₇₋₁₀H₁₀₋₁₈O₆₋₁₄) and dimers (C₁₇₋₂₀H₂₄₋₃₄O₆₋₁₈). The detected HOMs could mostly be explained by RO₂ from O₃-initiated (C₁₀H₁₅O_even) or OH-initiated (C₁₀H₁₇O_odd) oxidation, followed by autoxidation and different termination reactions. Our study also identified that HOM monomers with nine or more O atoms and all dimers typically condense onto particles irreversibly. However, HOM monomers with six to eight O atoms behaved more similarly to semi-volatile organic species, maintaining a noticeable gas-phase concentration. The HOM_O ≥9 yield at room temperature was estimated to be higher than that of α-pinene ozonolysis under the same conditions, with our best estimate being in the range of 3 %–6 %.

We observed that HOM concentrations decreased considerably at lower temperatures. This observation is consistent with previous studies on α-pinene ozonolysis (Quéléver et al., 2019; Simon et al., 2020), though the extent of the decrease varies considerably. The ADCHAM model, featuring a modified peroxy radical autoxidation mechanism, predicted the decrease observed in our study. However, further research is warranted to understand the causes of discrepancies observed across different studies. In addition, our study found that the HOM spectra were similar at three different temperatures (20, 10, and 0 °C) and were dominated by C₁₀H_14,16O₉ and C₉H_12,14O₉ in the monomer range and C₁₉H₃₀O${}_{\mathrm{6},\mathrm{10},\mathrm{11}}$ and C₂₀H₃₂O${}_{\mathrm{7},\mathrm{9},\mathrm{11}}$ in the dimer range. However, all dimers decreased at a slower rate than monomers, resulting in an increasing HOM dimer-to-monomer ratio from 0.78 to 1.51 when the temperatures decreased from 20 to 0 °C, which aligns with the results reported by Simon et al. (2020) for α-pinene ozonolysis but contrasts with the findings of Quéléver et al. (2019). The ADCHAM model managed to replicate the HOM formation, with the simulated composition, yield, and temperature dependence all agreeing reasonably well with our observations. We also found a sharp decrease in HOM concentrations at high RH (80 %), but, due to large instrumental uncertainty during the high-RH experiments, further work is required to verify the validity of this observation.

Taken together, our experimental results provide valuable insights into the Δ³-carene ozonolysis process in the atmosphere. The characterization of HOM oxidation products and estimation of yield help to further elucidate their potential impact on SOA formation. Additionally, the comparison between the results of Δ³-carene and α-pinene ozonolysis highlights the influence of different monoterpene precursors on the formation, distribution, and properties of HOMs, consequently affecting the properties of SOA. Thus, current models that group all monoterpenes together and represent them by α-pinene may lead to inaccuracies in the predicted SOA concentrations and their ultimate impact on the climate.