Myrcene (97 %, Aladdin) was further purified through three freeze-pump-thaw cycles before being stored as a liquid in a storage tube. An O2/O3 mixture was generated by passing high-purity O₂ (≥ 99.999 %) at a flow rate of 200 mL min⁻¹ through an ozone generator (Beijing Tonglin Technology Co., Ltd.). The collected O2/O3 mixture was frozen in liquid nitrogen and undergoes several cycles of freeze-pump-thaw to remove the residual impurity gas. O₃ or myrcene, in a 1:100 ratio with Ar, was mixed and stored in a 5 L spherical glass container. The matrix was maintained at 6 ± 1 K within a closed-cycle helium refrigerator (Physike Technology Co., LTD, Qcryo-Scryo-S-300). The characteristic infrared peaks of initial ozonolysis products were detected by using a Fourier transform infrared spectrometer (Bruker Vertex 70v) equipped with a liquid nitrogen-cooled mercury-cadmium-telluride (MCT) detector. Measurements spanned 600–4000 cm⁻¹, employing 32-scan averaging and 0.5 cm⁻¹ resolution. The experimental chamber was maintained under a vacuum of 10⁻⁵ Pa. The deposition of myrcene/Ar and O3/Ar onto the 6 ± 1 K cold window was facilitated by two angled and independent tubes at a rate of 5 mL min⁻¹. This deposition was known as the twin-jet co-deposition mode. The deposition duration was approximately 120 min. To allow limit the diffusion and/or reaction of reactants, these matrices were heated or annealed to 35 K and held for 0.5 h, and then cooled to 6 ± 1 K after which the spectra were recorded. To promote the further occurrence of the reaction and further soften and diffuse the matrix, the matrix was further heated to 45 and 55 K (Yu et al., 2025; Yang et al., 2020). To prevent matrix loss, it was imperative to immediately cool down to 6 ± 1 K after reaching the target temperature and to record the spectra.

The myrcene ozonolysis experiments were performed in a 1.2 m³ atmospheric simulation Teflon chamber maintained at room temperature (24 ± 1 °C) under different RH conditions. Myrcene was introduced into the chamber by passing zero air through a heated three-way U-shaped tube. The n-hexane was conducted in a consistent manner. The addition of approximately 367 ppm of n-hexane resulted in the removal of approximately 98 % of OH radical. The O2/O3 mixture was generated in the same manner as described above. The O2/O3 mixture was introduced in a single injection via a syringe and was not replenished during the experiment. The O₃ concentration was monitored by an O₃ analyzer (Thermo Scientific model 49i) throughout myrcene ozonolysis. The measurement of the myrcene concentration was conducted by means of a gas chromatograph with a quadrupole mass spectrometer (GC-MS, Agilent, 7890, 5977B) equipped thermal desorption instrument (TD). The GC-MS is on an HP-5MS column (30 m × 0.25 mm × 5 µm) with helium as carrier gas and a flow rate of 1.2 mL min⁻¹. The temperature of the chromatographic column was set as follows: the initial temperature was 40 °C and held for 3 min. The temperature was then increased to 140 °C by maintaining a rate of 20 °C min⁻¹; the program was then finished by increasing the temperature to 200 °C at a rate of 25 °C min⁻¹. Organics were quantified in selected ion monitoring (SIM) mode and 41 m/z, 69 m/z and 93 m/z were selected as the characteristic ions detected by mass spectrometry for myrcene. The maintenance of different RH levels was achieved by the implementation of a 10 L min⁻¹ flow of zero air through the water bubbler. The indoor temperature and RH were measured using a hygrometer (Vaisala, HMP3). Throughout each experiment, size distributions and volume concentrations of particles were continuously recorded using a scanning mobility particle sizer (SMPS), which consisted of a differential mobility analyzer (DMA, TSI, Model 3081) and a condensation particle counter (CPC, TSI, Model 3776). The SMPS measured particles every 3 min across a size range of 14.3 to 723.4 nm. A sampling flow of 0.3 L min⁻¹ and a sheath flow of 3.0 L min⁻¹ were used. The yield of SOA was obtained by the ratio of the maximum mass concentration of the corrected particles to the mass concentration of myrcene consumed (Liu et al., 2024; Chen et al., 2022). The specific equation was as follows. 1YSOA=MSOAΔMVOC Here, YSOA represented the SOA yield, MSOA denoted the maximum mass concentration of particle after wall-loss correction during the reaction process, and ΔMVOC referred to the total consumption mass concentration of VOCs throughout the reaction.

The average effective density of SOA obtained from the myrcene ozonolysis is 1.25 g cm⁻³ (Boge et al., 2013). The SOA was sampled after 3 h of reaction, with a sampling duration of approximately 60 min. The SOA particles generated within the chamber were captured on a 25 mm PTFE filter (Sartorius, 0.45 µm pore size) and subsequently analyzed using an ultra-high performance liquid chromatography with a Quadrupole-Orbitrap mass spectrometer equipped with an electrospray ionization source (UHPLC/ESI-MS, UPLC, UltiMate 3000, Thermo Scientific, ESI-MS, Q-Exactive, Thermo Scientific) equipped with a Hypersil GOLD C18 column (2.1 × 100 mm, 1.9 µm packing size ). The collected particle sample was eluted with 0.5 mL of methanol (Optima™ LC/MS Grade, Fisher Chemical) into a sample bottle. Mass spectrometric analysis utilized positive ion mode, scanning a molecular weight range of m/z 50–750 Da. The elution flow rate was set at 0.2 mL min⁻¹ with a total run time of 4.0 min. An injection volume of 10 µL was used. In positive ion mode, three ionic forms of particulate components were identified, specifically [M+H]⁺, [M+Na]⁺ and [M+NH₄]⁺. Tandem mass spectrometry (MS/MS) was employed to elucidate component structures within the SOA. The isolation width of 1.2 m/z units was applied. The electrospray ionization source was operated at a spray voltage of 3.0 kV and a capillary temperature of 300 °C, with sheath and auxiliary gas flows set to 35 and 10 Arb units, respectively. Both scans were performed at a resolution setting defined at m/z 200, with values of R=70 000 for the MS scan and R=17 500 for the MS/MS scan. Data acquisition and processing were conducted using Xcalibur software (version 3.0).

Additional quantum chemical calculations were performed to compare with the experimentally obtained infrared spectra. The geometries of the myrcene, primary ozonides (POZs) and SCIs were optimized using the hybrid density functional theory B3LYP-D3(BJ) with the aug-cc-pVTZ basis set. Harmonic vibrational frequencies were calculated at the B3LYP/6-311G++(d,2p) calculation level for the comparison with the experimental infrared peak. Various computational levels for CIs were compared, and the one with superior performance was selected accordingly. Please refer to Table S1 in the Supplement for the specific comparison. This method has been proved to be applicable to the relevant calculations of the SCIs system (Chen et al., 2025; Lin et al., 2018; Yu et al., 2025). The above-mentioned related calculations were all performed by using Gaussian 16 software package (Frisch et al., 2016). Molclus 1.1.2 in conjunction with the xtb software package was used to perform a systematic conformational search for myrcene, POZs and SCIs (Lu, 2023). The single-point energy was further calculated at the DLPNO-CCSD(T)/CBS level by using ORCA 5.0 software to obtain the Boltzmann distribution of each conformation more accurately (Neese, 2022).

C₇H₁₀O₂ (C7-CIs) and C₃H₆O₂ (C3-CIs) were produced from myrcene ozonolysis. The initial ozonolysis mechanism of myrcene has been established as shown in Fig. 1 based on the current studies. The products generated along with C7-CIs and C3-CIs were acetone and 4-vinyl-4-pentental (C7-aldehyde) with yields of 0.27 and 0.73, respectively (Deng et al., 2018). The nascent C7-CIs and C3-CIs, possessing high internal energy, rapidly degraded in part. This process could contribute to the formation of OH radical (Reactions R2 and R3). The unimolecular isomerization pathways of CIs were strongly configuration-dependent. Syn-C7-CIs were mainly isomerized by 1,4-H transfer to form vinyl hydroperoxide (VHP), which decomposed to form OH radical and a vinoxy or β-oxo alkyl radical (Reactions R2 and R4). Anti-C7-CIs mainly proceed via an initial rearrangement (1,3 ring-closure) to form a dioxirane intermediate. Then the dioxirane intermediates could isomerize to form organic acid (Reactions R3 and R5). C₃H₆O₂ isomerized mainly through the formation of VHP. Another portion was collisional stabilized, forming SCIs (Reaction R1) (Criegee, 1975; Hassan et al., 2021; Jr-Min Lin and Chao, 2017; Khan et al., 2018). A portion of the SCIs might undergo unimolecular decomposition reactions, while another portion primarily engaged in bimolecular reactions with trace gases in the atmosphere (Jr-Min Lin and Chao, 2017; Kidwell et al., 2016; Su et al., 2014; Vereecken et al., 2012). The yields of OH radical and SCIs obtained from the myrcene ozonolysis were 0.63 ± 0.09 and 0.30, respectively (Aschmann et al., 2002; Cox et al., 2020; Newland et al., 2018).

Although previous studies had provided some references to the initial ozonolysis mechanism of myrcene, the presence of CIs had only been demonstrated deductively. In this study, MI-FTIR was employed to capture the key intermediates formed during the initial ozonolysis of myrcene. As shown in Fig. 2, some new bands were observed in the spectra obtained by the twin-jet co-deposition. Most of the absorption peaks of the products grew with rising temperature within the 35–55 K interval. This observation indicated that the myrcene initial ozonolysis occurred when the Ar matrix softened and diffused. The new bands were located at 765, 880, 905, 1074, 1177, 1370 and 1720 cm⁻¹. The spectra of a single precursor (Myrcene/Ar or O3/Ar) obtained after annealing at different temperatures as comparison was shown in Fig. S1 in the Supplement.

The characteristic IR vibrational bands of the POZs, formed via initial cycloaddition between myrcene and O₃, were localized on its 1,2,3-trioxolane ring. These included asymmetric O-O-O stretching and C-O stretching vibrations within the five-membered ring structure (Wang et al., 2020). The presence of two different configurations in POZs was attributable to the varying orientations of the central O atom on the five-membered ring. The peaks calculated from the asymmetric stretching vibration in both POZs configurations overlapped at 772 cm⁻¹. Moreover, O-O-O asymmetric stretching vibrations of POZs were commonly observed within the 700–800 cm⁻¹ spectral region as documented in prior literature (Yang et al., 2020; Wang et al., 2020). Thus, the band at 765 cm⁻¹ which obtained from the twin-jet experiment was assigned to the O-O-O asymmetric stretching vibration of POZs. The calculation results indicated that the C-O stretching vibrations of the two POZs were located at positions 1059 and 1165 cm⁻¹. The newly emerged peaks at positions 1074 and 1177 cm⁻¹ corresponded to the C-O stretching vibration of POZs in the twin-jet spectra. Moreover, these POZ-related infrared peaks emerged at 35 K, and their intensity increased with rising annealing temperature until distinct infrared peaks became apparent at 55 K.

The main conformers and the infrared vibration frequencies of C3- and C7-CIs were obtained through quantum chemical calculations. The strongest characteristic infrared vibration peak of CIs was caused by O-O stretching vibration (Su et al., 2013; Lin et al., 2015). The extant research results indicated that the characteristic infrared peak caused by the O-O stretching vibration of C3-CIs in the gas phase located at 887 cm⁻¹ (Wang et al., 2016). In the twin-jet spectra, a distinct new peak was observed at position 880 cm⁻¹ at 55 K, which was likely to be attributed to the C3-CIs. The resulting ∼ 7 cm⁻¹ deviation might be attributable to the Ar matrix effect (Bach, 1999). The calculated O-O stretching vibrations of C7-CIs were located at 882 (syn-C7-CIs) and 910 (anti-C7-CIs) cm⁻¹. According to Table S1 in the Supplement, the calculated IR peaks of CIs were consistently overestimated by over 10 cm⁻¹ relative to experimental values, with the deviation being most pronounced for anti-CIs. Notably, no distinct new peaks appeared below 882 cm⁻¹ in the experiment. This absence suggested that very few stabilized C7-CIs were likely generated during the ozonolysis of myrcene. Instead, most C7-CIs were consumed via unimolecular decay pathways, this conclusion which was also supported by our subsequent analysis.

In addition to the early intermediates, acetone was identified as a major product co-produced with C7-CIs in myrcene ozonolysis. The spectral bands of 1370 and 1720 cm⁻¹ appeared in the twin-jet spectra belong to acetone (Han and Kim, 1996). The generation of acetone also indirectly demonstrated the formation of C7-CIs. The peak at 905 cm⁻¹ coincided in temperature with the C3-CIs peaks and suggested its assignment to the =CH₂ wagging vibration of the C7-aldehyde. MI-FTIR experiments unequivocally verified that myrcene ozonolysis proceeded via the Criegee mechanism. Furthermore, both C3- and C7-CIs were generated during this process.

To explore the contribution of CIs to the SOA generation, the myrcene ozonolysis experiments were conducted under different conditions (with or without an OH scavenger and at varying RH levels) in a 1.2 m³ smog chamber (Table 2). The unimolecular decay of both excited CIs and SCIs could produce OH radical. Therefore, SOA formation included contributions from both SCIs and OH radical derived in myrcene ozonolysis. As shown in Table 2, Exp. 1–2 added n-hexane (HA) as an OH radical scavenger to highlight the contribution of SCIs-derived pathways. Exp. 3–5 were conducted with progressively increasing RH to modulate SOA formation by introducing direct reactions between water and SCIs. The volume concentration of the particles rose sharply within approximately 30 min and reached its maximum value within 180 min (Fig. S2). The dominant size range of SOA expanded from 50–250 nm during myrcene ozonolysis (Fig. 3). The suppression of particle formation induced by the OH scavenger and elevated RH was conspicuously demonstrated. Following the 367 ppm HA addition, the particle yield from myrcene ozonolysis decreased by approximately 72 %. As RH increased from dry conditions to ∼ 20 % and subsequently to 50 %, the particle yield gradually declined by 17 % to 40 % (Table 2). As shown in Fig. 3a, after the addition of HA, no significant reduction in the particle size distribution was observed within the first 30 min. This might be attributed to the rapid reactions of SCIs, indicating that gas-phase reactions of SCIs played an important role in particle nucleation. Between 90 and 180 min of reaction, the particle size distribution decreased markedly. This reduction was due to the scavenging of OH radicals, which inhibited the formation of OH-RO₂ pathway. Bimolecular reactions and the autoxidation of OH-RO₂ played a crucial role in the growth of SOA (Baker et al., 2024). As the RH gradually increased from ∼ 20 % to 50 %, the growth of particle size was inhibited across different time periods. This phenomenon occurred because reactions between water and SCIs might prevent these compounds from converting into larger, less volatile products. Therefore, OH radical generated from CIs and SCIs also contributed to the growth of particle size distribution and enhanced SOA yield.

The O₃ concentration listed in the table is the value recorded after the O₃ analyzer stabilized for 20 min, which corresponds to the 20 min mark after the initiation of the experiment.

UHPLC-ESI-MS was employed to analyze the chemical compositions of SOA collected offline. CIs were important SOA precursors during the ozonolysis process. In this study, the contribution mechanisms of CIs to SOA formation were proposed in myrcene ozonolysis.

C₁₀H₁₈O₅ and C₁₀H₁₈O₄, with relative abundances (RA) of 76 % and 73 % respectively, ranked among the top 3 in the absence of any scavengers at RH < 0.5 % as shown in Fig. 4a. The formation of these two intense peaksbelonged to the contribution of OH radical oxidation processes. The OH radical was almost entirely originated from the CIs unimolecular decomposition reactions. The OH radical and O₂ reacted sequentially with myrcene to form C₁₀H₁₇O₃ radical (as shown in Fig. 1, Reaction R6). C₁₀H₁₇O₃ radical then underwent autoxidation channel which was highly likely to occur to produce C₁₀H₁₇O₅ radical (Jokinen et al., 2014). The reaction of C₁₀H₁₇O₅ radical with HO₂ yielded C₁₀H₁₈O₅ which was detected as the main product. Meanwhile, C₁₀H₁₇O₅ radical could also react with any RO₂ radical to yield C₁₀H₁₈O₄. The relative peak intensity of C₇H₁₀O₂ was slightly lower than that of C₁₀H₁₈O₅ and C₁₀H₁₈O₄, and its molecular formula was consistent with C7-CIs. It was inferred that it might originate from C7-OH produced during the unimolecular decomposition of C7-CIs. C7-CIs underwent unimolecular decomposition via the VHP pathway, producing OH radical while simultaneously generating R radical. The R radical subsequently reacted with O₂ to form C₇H₉O₃. C₇H₉O₃ could react with any RO₂ radical to form C₇H₁₀O₂ (C7-OH). The OH radical yield from myrcene ozonolysis was generally high, which also confirmed that the larger CIs generated during this process tend to react via unimolecular decay pathways (Cox et al., 2020).

C₁₀H₁₇O₅ radical, as the main RO₂ radical, could undergo sequential oligomerization. Several peaks representing ordered oligomers with C₃H₆O₂ as chain units were found. The chemical formula of C₃H₆O₂ was identical to C3-CIs and confirmed to existent by the matrix isolation experiment. The proposed pathway was shown in Reaction (R7). The rate constants for the reaction of RO₂ radicals with SCIs were much higher than that for ROOH (Zhao et al., 2017; Chao et al., 2024; Chhantyal-Pun et al., 2020; Caravan et al., 2024). C₁₀H₁₇O₅ radical reacted with one or two C3-CIs, followed by reaction with HO₂ radical resulting in chain termination (C₁₀H₁₇O5+n C3-CIs + HO₂, n=0–2, Sequence 1). However, the relative intensities of this set of oligomer signals were relatively low. This was likely because C₁₀H₁₇O₅ radical more readily reacted with HO₂ or RO₂ to form C₁₀H₁₈O₅ and C₁₀H₁₈O₄, thereby reducing the likelihood of its reaction with C3-CIs in the gas phase. These oligomer sequences exhibited similar distribution patterns of major fragment ion peaks in MS/MS spectra in Fig. S3a. Therefore, these oligomers with C3-CIs as the chain units had similar structural units in their chemical composition. Moreover, the corresponding structural assignments for its major fragment peaks were provided in Fig. S3b. The fragments C₅H7+ (m/z 67.06), C₆H₉O2+ (m/z 113.06), and C₉H₁₃O⁺ (m/z 137.10) revealed that the oligomers in Sequence 1 all feature a carbon chain skeleton characterized by conjugated double bonds. The C₃H₇O⁺ (m/z 59.05)/C₃H₇O2+ (m/z 75.05) fragment ions originated from cleavage of terminal -OH or -OOH functionalized tertiary carbon moieties in the oligomers. For C7-CIs, no significant oligomers with C7-CIs as chain units were observed when reacting with C₁₀H₁₇O₅ radical. Although the mass spectral peak corresponding to C₁₇H₂₈O₇ was observed and its molecular formula aligned with the expected product of C₁₀H₁₇O5+ C7-CIs. But C₁₇H₂₈O₇ could also originate from C₁₀H₁₇O5+ C₇H₁₁O₄ (the formation pathway as shown in Scheme S1). Moreover, C₁₇H₂₈O₇ was classified as a low-volatility organic compound (LVOCs) (Donahue et al., 2012; Donahue et al., 2011), which readily partitioned into the particle phase. But its low abundance implied C₁₇H₂₈O₇ did not appear as a dominant product. Consequently, the contribution of C7-CIs oligomerization to myrcene-derived SOA formation might be considered negligible. In addition, an ordered oligomer sequence formed by C₇H₁₀O2+n C3-CIs was also identified. The corresponding compounds in the sequence are C₇H₁₀O₂ (m/z 127.075), C₁₀H₁₆O₄ (m/z 201.112), C₁₃H₂₂O₆ (m/z 292.175) and C₁₆H₂₈O₈ (m/z 371.168), namely Sequence 2 (The peaks with relative abundance of less than 1 % were ignored). The proposed pathway of Sequence 2 was as shown in the Reaction (R8). Sequence 2 had a greater degree of oligomerization and contributed more to particle formation than Sequence 1. This indicated that C₇H₉O₃, which produced by C7-CIs unimolecular reaction, more readily reacted with C3-CIs in the gas phase. The oligomerization reactions within the Sequence 2 all terminated with the reaction of RO₂ radical (C₇H₉O3+n C3-CIs + RO₂). Figure S4 presented the MS/MS spectrum of Sequence 2 along with the proposed structures corresponding to the major fragment peaks. The presence of fragments such as C₅H7+ (m/z 67.06), C₇H₇O⁺ (m/z 95.05), and C₆H₉O2+ (m/z 113.06) indicated that the oligomers in Sequence 2 possessed a conjugated double-bond skeleton. This suggested that this sequence was initiated by C7 compounds. The fragment at m/z 71.05 (C₄H₇O⁺) was generated via α-cleavage adjacent to a carbonyl group in the oligomer backbone. This fragmentation was typical of compounds containing conjugated carbonyl motifs and further supported the presence of a C7-based conjugated skeleton in the sequence. The C₃H₇O⁺ (m/z 59.05) fragment ions originated from cleavage of terminal -OH functionalized tertiary carbon moieties in the oligomers. The formation of this ion fragment was similar to that in Sequence 1. In the MS/MS spectra of C₁₀H₁₆O₄, we observed that besides the fragment peaks annotated in Fig. S4, there existed fragment ion peaks with higher intensities. This suggested that C₁₀H₁₆O₄ was formed through multiple pathways (such as the RO2+ RO₂ reaction) (Peräkylä et al., 2023; Frandsen et al., 2025). Consistently, no discernible oligomerization signals were detected for reactions using C₇H₉O₃ as the starting reactant with C7-CIs. R7

As shown in Fig. 4b, the peaks corresponding to C₁₀H₁₈O₄ and C₁₀H₁₈O₅ disappeared after the addition of the OH radical scavenger, as expected. The C₇H₁₀O₂ peak remained, which further demonstrating that C₇H₁₀O₂ originated from SCIs-derived products. Compared to the mass spectrum without the scavenger, the contribution of the oligomers in Sequence 2 markedly increased. Following the introduction of the scavenger, the peak intensities corresponding to compounds with higher degrees of oligomerization (C₁₃H₂₂O₆ and C₁₆H₂₈O8) within the sequence were significantly enhanced. Moreover, compared to the spectrum acquired in the absence of the scavenger, a more highly oligomerized compound (C₁₉H₃₄O10) was detected. Water served as the dominant removal pathway for SCIs in the atmosphere. Correspondingly, the contribution of Sequence 2 to SOA formation decreased progressively with increasing RH as shown in Fig. 4c and d, especially substances with a higher degree of oligomerization in the sequence (C₁₃H₂₂O₆ and C₁₆H₂₈O8). So, the Sequence 2 was primarily contributed by SCIs-derived. R8

Meanwhile, some compounds were found to have the same number of C and H atoms, differing only in the number of O atoms in parallel oligomer sequences, namely, C₇H₁₀O₂₋₅, C₁₀H₁₆O₄₋₇, C₁₃H₂₂O₆₋₉ and C₁₆H₂₈O₈₋₁₁ (Sequence 2N) as shown in Fig. 5a. Compounds with identical carbon and hydrogen atom counts exhibited closely resembling major fragment ions in their MS/MS spectra as shown in Fig. 5b. This suggested that compounds with identical carbon and hydrogen atom counts exhibited highly consistent formation pathways. The process leading to progressive oxygenation (i.e., increasing oxygen atom content) was the RO₂ autoxidation mechanism (Jokinen et al., 2014; Liu et al., 2023). Simultaneously, Sequence 2N was the oligomer series formed by oligomerization with C3-CIs as chain units. This indicated that SCIs oligomerization and RO₂ autoxidation mechanistically interplayed in the formation of Sequence 2N. The proposed formation mechanism of Sequence 2N was illustrated in Fig. 5c. RO₂ underwent oligomerization with n SCIs to form the oligomer RO₂-n SCIs, while concurrently undergoing autoxidation to generate R'O₂. R'O₂ might subsequently underwent analogous cycles of SCIs oligomerization and autoxidation. This radical propagation cascade ultimately terminated via reactions with RO₂ or HO₂ radical. SCIs inherently contained at least two oxygen atoms in their structure due to the presence of terminal -COO groups. The RO₂ autoxidation inherently involved reaction with O₂, which rapidly elevated the oxygen content in products. Consequently, the synergistic interplay between SCIs oligomerization and RO₂ autoxidation facilitated rapid formation of high-molecular-weight species and highly oxygenated molecules (HOMs), which substantially contributed to particle nucleation and growth.

The addition of different scavengers also affected the formation of Sequences 2N. Figure 6 illustrated the proportions of compounds with various oligomerization degrees of oligomerization in Sequence 2N under different scavenger conditions. The dominant compounds in Sequences 2N were C₁₃H₂₂O₆₋₉ which contained two C3-CIs as chain units under dry conditions, with or without added OH radical (Fig. 6a and b). Furthermore, following the addition of OH radical, the proportions of both C₁₃H₂₂O₆₋₉ and C₁₆H₂₈O₈₋₁₁ in Sequences 2N gradually increased (Fig. 6b). This indicated that the SCIs oligomerization reaction significantly contributed to the particle formation after the addition of the OH scavenger. This was consistent with the result observed in Sequences 2. As the RH increased, the main compound in Sequences 2N became C₁₀H₁₆O₄₋₇, which contained only a single C3-CIs as a chain unit (Fig. 6c and d). The proportions of both C₁₃H₂₂O₆₋₉ and C₁₆H₂₈O₈₋₁₁ in Sequence 2N decreased with increasing RH. This indicated that the reaction of water with C3-CIs directly affected the generation of oligomers with higher degrees of oligomerization. Additionally, the fraction of C₇H₁₀O₂₋₅ decreased significantly with increasing RH. This indicated that besides reacting with C3-CIs, water also scavenged C7-CIs to some extent. Therefore, the contribution of the SCIs oligomerization mechanism to SOA formation were suppressed under elevated RH conditions.

Sequence 1, 2 and 2N demonstrated ordered oligomers formed with C3-CIs as chain units. However, ordered oligomeric sequences (RO2+n C7-CIs + HO2/RO2, n≥2) with C7-CIs as the chain unit were not discovered. Based on the current formation mechanism of oligomers, C3-CIs and C7-CIs differed in their contribution to particle formation in myrcene ozonolysis. For smaller molecular-sized CIs (C3-CIs), the primary pathway for partitioning into the particle involved oligomerization with RO₂ radical and other species. The C7-CIs had higher molecular masses than C3-CIs. The products from unimolecular and bimolecular reaction of C7-CIs partitioned more readily to particles which reduced the viability to function as chain units in oligomerization processes (Donahue et al., 2012; Donahue et al., 2011). Furthermore, C7-RO₂ generated from unimolecular decomposition of C7-CIs acts as a chain-initiating precursor that reacts with C3-CIs, thereby contributing to particle nucleation. These findings implied divergent dominant pathways for SOA formation mediated by SCIs across varying molecular dimensions in VOCs ozonolysis. The C10-CIs with high molecular weights generated during the ozonolysis of α-pinene and limonene might also be more likely to incorporate into the particle via unimolecular reaction mechanisms rather than through oligomerization. Our findings further demonstrated that when evaluating SCIs contribution pathways to SOA, the molecular size of SCIs might be prioritized especially during monoterpenes ozonolysis. The molecular size of SCIs might lead to differences in their primary mechanisms of contributing to SOA formation. In the myrcene ozonolysis system, the branching ratio might favor the formation of C3-CIs over C7-CIs, which further enhanced the propensity of C3-CIs toward oligomerization. Currently, no literature has elucidated the contribution mechanisms of CIs of different sizes to SOA formation from the perspective of molecular size. This study presented this concept for the first time.