In recent decades, significant advancements have been made in understanding the oxidation mechanisms of isoprene, due to its key role in atmospheric chemistry and its extremely high abundance in the atmosphere. As the most emitted BVOC globally, isoprene has been widely recognized as a significant SOA precursor, forming specific intermediates such as isoprene-epoxydiol (IEPOX) (Paulot et al., 2009; Nguyen et al., 2014) and methacryloyl peroxynitrate (MPAN) (Nguyen et al., 2015), in addition to highly functionalized low-volatility compounds (Krechmer et al., 2015; D'Arnbro et al., 2017; Xu et al., 2021).

Accurately representing the isoprene chemistry in large-scale atmospheric models is crucial but remains challenging due to the complex reaction mechanisms. MCM, through continuous updates, provides an almost complete compilation of isoprene's degradation pathways. Additionally, Wennberg et al. (2018) conducted a systematic review of current knowledge on isoprene chemistry, resulting in the development of the Caltech isoprene mechanism – a detailed reaction framework capable of dynamically simulating the allylic and peroxy radical systems formed when isoprene reacts with OH radicals (Wennberg et al., 2018). Compared to earlier mechanisms, the Caltech isoprene mechanism introduces significant advancements, including the incorporation of reversible O₂ addition to allyl radicals, the identification of new products from 1,6–H shifts in Z–δ–OH–peroxy radicals, and a reduced yield of C₅-hydroperoxy-aldehydes (HPALD). Furthermore, it includes more intramolecular H-shift processes, such as rapid peroxyl-hydroperoxyl (RO₂–OOH) shifts, which enhance OH recycling rates under low-nitrogen conditions. Despite these advancements, MCM provides a more detailed treatment of isoprene chemistry, including more comprehensive RO₂–RO₂ reactions and improved photolysis processes. The photolysis rate constants in MCM are calculated by integrating light flux over specific wavelengths, enabling accurate representation of photolysis under varying atmospheric conditions. In contrast, the Caltech mechanism calculates photolysis rates using a simplified coefficient, “sun”, which relies on sunrise and sunset times. While effective for outdoor scenarios, this approach does not adequately capture the intricacies of controlled laboratory illumination, making MCM the preferred mechanism for laboratory-based simulations.

To exploit the strengths of both mechanisms, we integrated the MCM and Caltech isoprene mechanisms in our framework. This posed significant challenges due to differences in species naming conventions between the two frameworks. In the MCM, compounds are named systematically based on their chemical structure and assigned a unique identifier, whereas the Caltech mechanism employs a naming system based on the carbon skeleton and functional group positions. As a result, careful mapping of chemical species between the two mechanisms was required to bridge these discrepancies (see Supplement for details). This integrated approach allows us to harness the complementary advantages of both mechanisms for a more robust representation of isoprene chemistry.

Despite these improvements, the descriptions of HOM production from the oxidation of isoprene remain incomplete. To address this gap, we incorporated autoxidation pathways for specific high-yield RO₂ formed during isoprene oxidation, which are crucial intermediates in isoprene oxidation. The key RO₂ radicals considered include those substituted with =O, -OH, and -OOH groups, which originate from isoprene hydroxyperoxyl radicals (ISOPOO) and RO₂ species formed following a 1,6 α-hydroxy H-shift in two Z–δ–ISOPOO isomers. The rapid H-shift reactions of these high-yield RO₂ radicals are critical, as they allow for the production of stable, low-volatility products under typical atmospheric conditions, making them particularly potent for HOM formation.

For these H-shift reactions to compete effectively with bimolecular reactions involving NO, HO₂, and other RO₂, unimolecular reaction rate constants need to reach approximately ∼10-2 s⁻¹ or higher. Reported H-shift rates for these radicals cover a wide range, from 8.2×10-2 to 3.0×105 s⁻¹ at 298 K (Moller et al., 2019), with hydroperoxides exhibiting notably faster rate. While the Caltech isoprene mechanism assumes alkyl radicals predominantly fragment into smaller products, thus limiting HOM formation, we introduced an oxygen addition pathway to alkyl radicals, as suggested in Wang et al. (2018). This modification introduces a competing reaction pathway that can produce new RO₂ species, counterbalancing the dominance of fragmentation. Subsequent bimolecular reactions involving these RO₂ species were implemented based on the work of Jenkin et al. (2019).

Although OH oxidation overwhelmingly dominates the removal of isoprene from the atmosphere, O₃ oxidation accounts for approximately 10 % of isoprene's global loss (Bates and Jacob, 2019). Previous models primarily focused on its contributions to the formation of methyl vinyl ketone (MVK), methacrolein (MACR), CH₂OO (C₁ SCI), and OH, while largely neglecting its potential contribution to HOM formation. Here, we adopted a simplified approach similar to that of Schervish et al. (2024), assuming that only a small fraction of RO₂ radicals are capable of undergoing autoxidation. These RO₂ radicals can subsequently participate in unimolecular or bimolecular reactions, with some contributing to the HOM formation.

Our model extends the first comprehensive Peroxy Radical Autoxidation Mechanism (PRAM) developed by Roldin et al. (2019). PRAM meticulously simulates the autoxidation processes of RO₂ formed from monoterpene oxidation. This mechanism has demonstrated strong agreement with observed HOM concentrations, SOA mass concentrations, and NPF in boreal forest simulations. However, the original PRAM only considered the HOM formation pathway via monoterpene oxidation by O₃ and OH. Nie et al. (2023)later expanded this mechanism by incorporating NO₃-initiated oxidation pathway to the original mechanism, which mainly includes the autoxidation of monoterpenes by NO₃ to form RO₂; the reaction of RO₂ with NO₃ to form RO radicals (though this pathway is negligible under most environmental conditions) and bimolecular termination reactions between NO₂ and specific RO₂ (e.g., acyl RO₂) (Nie et al., 2023).

In this study, we further optimize the key oxidation pathways of monoterpenes, with a particular focus on α-pinene ozonolysis, which serves as a representative case due to its atmospheric relevance and detailed mechanistic understanding. Conventional mechanisms propose that α-pinene ozonolysis begins with the decomposition of the primary ozonide into a carbonyl-substituted Criegee Intermediate (CI). Three isomeric forms of this CI can undergo a 1,4 H-shift to produce vinyl hydroperoxide (VHP). The VHP subsequently decomposes into a vinoxy radical, which reacts with O₂ to form a RO₂ radical. However, under typical conditions, the isomerization of this RO₂ radical is too slow to explain the observed rapid formation of HOMs. To address this, Iyer et al. (2021) proposed a critical solution by introducing a chemically activated ring-opening reaction of one of the vinoxy radicals. This reaction leads to the rapid formation of an endoperoxide and a RO₂ radical with a high oxygen content (containing 8 oxygen atoms). The PRAM mechanism incorporates this modification and further details the progressive increase in oxygen atoms within the molecule through consecutive intramolecular RO₂ H-shifts and O₂ additions during autoxidation. This autoxidation chain ultimately terminates through bimolecular reactions with NO, HO₂, or other RO₂ radicals, forming a variety of products such as alkoxy radicals, closed-shell HOM monomers, or dimers. Once alkoxy radicals are formed, they may undergo further transformations. For example, they can isomerize into hydroxy-substituted alkyl radicals, which subsequently react with O₂ to form new RO₂ species, or they may form closed-shell HOM monomers with additional carbonyl groups. Alternatively, they may decompose into more volatile species. A specific example of these fragment products includes the MCM-modeled RO₂ species C717O₂ (an RO₂), and smaller volatile species like acetone (CH₃COCH₃).

In our improved model, we incorporated the C₇-fragment due to recent studies indicate that early-formed addition products retain sufficient energy to overcome transition state barriers, leading to the formation of a significant amount of alkyl radicals with an endoperoxide group (EPO). Unlike the traditional view that excess energy dissipates in the next O₂ addition step, Yang et al. (2025a) demonstrated that EPO formation enables rearrangement pathways involving alkoxy radicals with epoxide groups (AOE). Their study identified that cleavage reactions from these intermediates, which yield acetone and C₇ products (AE-C₇), are the fastest. Furthermore, the O₂ addition products (RO₂) formed from AE-C₇ and AE-C₁₀ contain multiple active sites for H-shift, facilitating further autoxidation. These findings provide key insights into the competition between unimolecular H-shift reactions and bimolecular sinks, such as reactions with NO and HO₂. Additionally, the contribution of AE-C₇ and AE-C₁₀ derivatives explains observed peaks in mass spectrometry corresponding to C₇-HOMs.

We also incorporated RO₂–Kb β-cleavage and CH₂O loss to better represent the formation of C₁₉ accretion products, like C₁₉H₂₈O_x and C₁₉H₃₀O_x, as observed in experiments by Berndt et al. (2018). These products are formed through RO₂ self- and cross-reactions involving CH₂O loss. Furthermore, experiments that isolated OH oxidation revealed that these C₁₉ accretion products are not formed through pure OH-mediated processes, but rather arise specifically from O₃-derived RO₂ radicals (Peräkylä et al., 2023; Kenseth et al., 2023). Our updated model explicitly accounts for these distinctions, differentiating the contributions of OH- versus O₃-derived RO₂ radicals in the formation of HOM accretion products.

Prior to our work, HOM formation from sesquiterpenes lacked a dedicated module in atmospheric models. To address this, we developed a comprehensive framework based on the reaction pathways proposed by Richters et al. (2016b), integrating key processes derived from existing knowledge to better represent the oxidation chemistry of sesquiterpenes. Given that β-caryophyllene is the only sesquiterpene included in MCM, our development focuses exclusively on this compound.

The ozonolysis of β-caryophyllene begins with an exothermic reaction between O₃ and the double bonds of sesquiterpene molecules, forming CIs with significant excess energy. These chemically activated CIs can either stabilize through collisions with other molecules or undergo unimolecular reactions. Stable CIs may also engage in bimolecular reactions, initiating a variety of pathways that eventually lead to HOM formation. A critical step involves the isomerization of CIs into vinyl hydroperoxide, which further decomposes by releasing OH radicals, producing alkyl radicals as intermediates. The alkyl radicals rapidly react with O₂, forming the first generation of RO₂ radical.

The fate of these initial RO₂ radicals branch into several pathways. In one major pathway, the RO₂ radical undergoes intramolecular H-shifts followed by O₂ addition, resulting in new RO₂ radicals that resemble the first-generation p-HOM-RO₂ species observed in monoterpene ozonolysis. These products typically include hydroperoxide (-OOH) functionalities. Alternatively, RO₂ radicals can attack the remaining double bond in the sesquiterpene structure. This process leads to the formation of endoperoxides and additional alkyl radicals, which subsequently react with O₂ to form new RO₂ radicals, though these lack the -OOH functional group. Further reaction pathways are also possible. For instance, RO₂ radicals may undergo additional intramolecular H-shifts, resulting in the formation of closed-shell products along with the release of OH radicals; or the subsequent O₂ addition after H-shift forms a new RO₂. Another pathway involves internal RO₂ reactions with the remaining double bond, which can generate cyclic R radicals. These cyclic radicals then react with O₂, producing the next generation of RO₂ radicals.

Epoxide formation is another potential pathway, where an epoxide ring forms within the molecule, followed by cleavage of the acyl alkoxy functionality and the expulsion of CO₂. This step yields alkyl radicals that rapidly react with O₂ to form new RO₂ species. These RO₂ radicals can then enter further autoxidation processes, involving a series of intramolecular hydrogen transfer reactions and successive O₂ additions, to produce higher-generation RO₂ species.

By systematically integrating these reaction pathways into our model, we developed a robust framework to simulate HOM formation from sesquiterpene ozonolysis. The inclusion of detailed autoxidation chemistry, along with pathways involving both -OOH and non-OOH functional group formation, ensures a more comprehensive representation of the sesquiterpene oxidation process and its contribution to atmospheric HOM and SOA formation.

In real atmospheric conditions, VOC mixtures produce a variety of RO₂ radicals that can react with each other to form RO, closed-shell monomers, or dimers. Given the vast number of RO₂ species in detailed chemical mechanisms, explicitly representing all possible cross-reactions is impractical. To address this, MCM uses a simplified approach, which assumes that all RO₂ radicals interact uniformly within a “RO₂ pool” at a collective rate. This is implemented using the parameter Σ[RO₂], which represents the summed concentration of all RO₂ species. Within this framework, the total rate of all possible cross-reactions for a particular RO₂ radical is approximated as a pseudo-unimolecular reaction with a rate coefficient of k×Σ[RO₂]. While this simplification reduces computational complexity, it overlooks the specific contributions of individual RO₂ combinations, particularly in processes like dimer formation. The CRI-HOM model addresses dimerization by treating it as a simplified reaction in which one RO₂ radical produces half of the total dimer product. Although efficient, this approach fails to account for differences in how specific RO₂ combinations influence product distributions.

In our improved model, we redefine dimerization as a specific bimolecular reaction between distinct RO₂ species. To reduce complexity, we focus on the interactions between two key types of RO₂: those capable of autoxidation and those that cannot undergo autoxidation (as represented in MCM). Autoxidizable RO₂ radicals, due to their higher degree of functionalization, exhibit faster dimerization rates Berndt et al., 2018). Non-autoxidizable RO₂ radicals, which are generally present at higher concentrations, can serve as reaction partners in these bimolecular reactions. Reactions between two autoxidizable RO₂ are not explicitly represented, due to their extremely low concentrations, which makes their contribution to dimer formation negligible. Likewise, reactions between two non-autoxidizable RO₂ are not treated as explicit accretion product formation pathways due to due to their products are probably not HOM. Instead, these reactions remain represented within the generic RO₂ loss framework of the RO₂ pool, where they contribute to closed-shell monomer formation through the pseudo-unimolecular reaction scheme. In this way, the total RO₂–RO₂ reaction flux among all RO₂ species is still accounted for, while only those combinations most relevant for HOM dimer formation are treated explicitly.

RO₂ + RO₂ rate coefficients are assigned according to the overall oxidation state and molecular size of the RO₂ radicals, using the number of oxygen atoms as a proxy for the degree of functionalization. This approach is consistent with the parameterization described above, where more highly oxygenated and generally larger RO₂ radicals are assumed to exhibit higher reactivity in RO₂ + RO₂ reactions. SIM-HOM uses RO₂ + RO₂ reaction rates leading to closed-shell monomer products in the range of 1×10-12 to 1.5×10-11 cm³ s⁻¹ and RO₂ + RO₂ reactions leading to HOM dimers in the range of 1×10-14 to 1.5×10-12 cm³ s⁻¹. This choice reflects the generally smaller branching fraction of the accretion channel relative to the formation of monomer products. As a result, the effective branching ratio toward accretion products in the model is typically on the order of ∼10 %, consistent with recent experimental and theoretical studies (Murphy et al., 2025; Berndt et al., 2018). Because these dimers arise from explicit bimolecular reactions, the identities of these products are determined by the specific pair of reacting RO₂ precursors. This approach balances computational efficiency with improved accuracy, capturing the variability in RO₂ reactions and their impact on product distributions, particularly in mixed VOC systems.

We implemented several key updates to the model, focusing on the autoxidation pathway, RO₂–RO₂ interaction, fragmentation and termination processes. Together, these enhancements provide a more accurate and comprehensive representation of atmospheric oxidation, enabling better simulation of VOC oxidation and HOM formation:

Expanded Autoxidation Pathways: we refined the autoxidation scheme for isoprene, monoterpenes, and sesquiterpenes by incorporating recent experimental and theoretical advancements. Key updates include (1) High-yield RO₂ autoxidation reactions for isoprene oxidation; (2) Extended formation and autoxidation pathways for C₇-RO₂ species in monoterpene oxidation; (3) Integration of new autoxidation and HOM formation processes for sesquiterpenes, building on the MCM framework.

To assess the model's performance under varying environmental conditions, we calculated the relative error of HOM concentrations by normalizing the difference between observed and simulated concentrations. Only species with concentrations exceeding 5×104 cm⁻³ (the CI-APi-ToF detection limit) were considered. As shown in Fig. 2a, the relative errors (defined as the difference between modeled and measured HOM concentrations normalized by the measurements) are illustrated as box plots. They remain close to 0 under most conditions, indicating strong agreement between model and measurements. However, in the mixed system of three VOCs (IP + MT + SQT), particularly at low VOC concentrations, larger discrepancies appear. This may be attributed to photolysis of oxidation products by UVH lamps, which is not explicitly considered in detail, leading to deviations in HOM predictions. The number of selected data points (gray dashed line) also varies across different conditions, influencing error estimates.

Figure 2b compares the simulated and observed HOM concentrations across different VOC systems. The total HOM concentration varies significantly across different precursors. The isoprene system produces relatively lower HOM concentrations compared to monoterpene and sesquiterpene systems, consistent with the expected differences in oxidation pathways and HOM formation efficiency. The monoterpene system exhibits the highest HOM concentrations, particularly at higher precursor concentrations, followed by the sesquiterpene system. In mixed systems, including the isoprene-monoterpene system (Mix I) and the three-VOC mixture (Mix II), the total HOM concentration increases compared to the isoprene-only system, reflecting the contribution of monoterpenes and sesquiterpenes. The model successfully reproduces the general trends, capturing the HOM concentrations in isoprene, monoterpene, and sesquiterpene systems, as well as in mixed systems. However, overestimation is observed in some cases, particularly in low VOC concentration scenarios, as discussed earlier, likely due to incomplete representation of photolysis processes. Despite these minor deviations, the overall variations in HOM production across different VOC regimes are well reproduced.

In the pure isoprene system, the isoprene was continuously injected to maintain a concentration between 4 and 5 ppb, while O₃ was approximately 40 ppb, under both dark and illuminated conditions. We compared the simulation results from our improved module, MCM, the Caltech isoprene mechanism (Caltech), and their combination (MCM + Caltech). All models showed strong performance in predicting OH concentration (Fig. 3c).

Across different experimental conditions, our model exhibits significant advantages in capturing the distribution of highly oxygenated oxidation products compared to other models. For OOMs containing 5 oxygen atoms, all four models predicted concentrations notably higher than the experimental measurements (Fig. 3a and b), likely due to the reduced sensitivity of the NO3- CI-APi-ToF mass spectrometer towards OOMs with five or fewer oxygen atoms (Riva et al., 2019). For OOMs with six or seven oxygen atoms, the MCM model significantly overestimated their concentrations, while the Caltech model underestimated them. In contrast, our model closely matched the measured data, indicating superior accuracy in simulating these oxidized species. More importantly, for OOMs containing eight or more oxygen atoms, our model was the only one capable of capturing their formation, underscoring its ability to accurately represent the complex pathways of HOM formation during isoprene oxidation.

Recent studies have demonstrated that isoprene-derived highly oxygenated organic molecules (IP-HOMs) play a key role in new particle formation (NPF) in the upper troposphere (Curtius et al., 2024; Shen et al., 2024; Zha et al., 2024). Our model effectively simulates the formation of HOMs, underscoring its relevance for investigating the mechanisms of NPF and subsequent particle growth. However, the observed HOM spectrum in our chamber experiments differs from that of the atmosphere due to weak OH recycling, a consequence of the absence of NO_x and the predominantly dark or low-light experimental conditions. Atmospheric OH⚫ levels during daytime typically remain above 10⁶ cm⁻³, even in the presence of isoprene, sustaining secondary oxidation processes. As a result, atmospheric isoprene oxidation predominantly produces C₅H₁₂O_x, C₅H₁₁NO_x and C₅H₁₀N₂O_x via second-generation OH oxidation from ISOPOOH and isoprene hydroxy nitrate (Curtius et al., 2024; Shen et al., 2024; Zha et al., 2024; Xu et al., 2021). In contrast, our experimental spectrum is dominated by C₅H_8–10O_x (Fig. S2), derived from the mono- and bimolecular reaction of RO₂ formed directly via isoprene oxidation. A comparison of C₅H₁₂O_x concentrations with other C₅ compounds (Fig. S3) further confirms that second-generation oxidation plays a minimal role in our experiments. Despite these differences, our model integrates all relevant oxidation pathways, providing a more comprehensive representation of isoprene oxidation chemistry.

Figure 4 presents the predicted oxidation products spectrum from our improved model alongside experimental measurement under different conditions. A detailed comparison with the experimental data demonstrates that our model achieves high accuracy in reproducing monoterpene oxidation processes. In particular, it effectively captures the formation of RO₂ radicals with 10 carbon atoms (C₁₀–RO₂) and accurately simulates their subsequent reaction pathways. Fragment simulation has traditionally been a challenge in numerical modeling. In our improved model, we incorporated C₇ fragments formation by incorporating recent findings on early-stage product cleavage (Yang et al., 2025a). The model successfully reproduces the isomerization of C₁₀–RO₂, leading to carbon skeleton cleavage and C₇-RO₂ formation, which undergoes autoxidation and termination to produce C₇–HOMs. Furthermore, C₇–RO₂ produces C₅–RO₂ through RO pathway, accurately simulating key fragment products such as C₅H₆O₇, validating the model's capability in capturing essential oxidation pathways.

Beyond accurately representing the main monomer and fragmentation product distributions, the model also exhibits significant improvements in simulating dimer formation. In particular, the enhanced formation of C₇–RO₂ has led to an improved prediction of C₁₇H₂₆O₉ concentration, which corresponds to the most abundant dimer observed in our measurements. Additionally, the incorporation of RO₂–Kb β-cleavage and CH₂O loss processes have contributed to the relatively high concentrations of C₁₉H₂₈O₉ and C₁₉H₂₈O₁₁ dimers. Despite these advancements, the model still slightly underestimates these dimer concentrations compared to experimental observations. Several factors may contribute to this discrepancy. First, the actual yield of RO₂–Kb is higher than estimated in MCM (Meder et al., 2025), leading to an underrepresentation of key dimer-forming precursors. Second, RO₂–Kb cleavage products may have a higher efficiency in forming accretion products compared to other RO₂, an aspect not fully accounted for in the current model. Third, additional RO₂ species beyond those currently considered may undergo similar cleavage reactions, contributing to dimer formation through pathways not yet incorporated (Peräkylä et al., 2023). Addressing these uncertainties by refining the branching ratios and reaction rate constants of dimerization pathways could help resolve these discrepancies and further improve the model's predictive capabilities.

Figure 5 shows that our newly developed sesquiterpene-HOM model is in strong agreement with the experimental data. During the validation process, we meticulously compared the model predictions with experimental results obtained under controlled conditions, ensuring that key chemical products – including fragmentation products, monomers, and dimers – were accurately represented. The model accurately captures the formation and transformation of these products, confirming its reliability in simulating sesquiterpene oxidation.

A key achievement of the model is its accurate prediction of C₁₅H₂₂O₉, the most abundant HOM molecule under these experimental conditions and a critical ELVOC contributing to NPF (Dada et al., 2023). This highlights the model's ability to capture the dominant oxidation pathways of sesquiterpenes. To comprehensively describe the reaction mechanism, we incorporated an epoxide formation step, in which CO₂ is expelled from the acyl alkoxy functional group of the intermediate, forming an alkyl radical that rapidly reacts with O₂ to form an RO₂ radical. This modification successfully explains the observed C₁₄ fragmentation product, further improving the model's accuracy in reproducing experimental spectrum. However, while the model successfully reproduces the observed peaks for most monomers and fragments, it fails to predict C₁₀H₁₄O₁₀, which is detected at non-negligible concentrations in experiments. This discrepancy arises from the absence of reported formation pathways for this compound, suggesting that additional reaction mechanisms may need to be explored.

Regarding dimer formation, the model incorporates detailed RO₂ cross-reaction pathways. While the model provides a reasonable overall prediction, it overestimates the concentrations of lower-molecular-mass dimers while underestimating the most abundant dimers, such as C₂₉H₄₄O₁₂ and C₂₉H₄₆O₁₄. This suggests that refinements in the reaction rates governing dimerization processes-particular the relative contribution of RO₂–RO₂ from fragment RO₂ are necessary to achieve better agreement with observations. Additionally, the inclusion of alternative dimerization channels, such as those involving secondary oxidation, may be required to better represent the observed dimer distribution.

Overall, the strong agreement between model predictions and experimental data indicates that our model effectively captures the fundamental oxidation mechanism governing sesquiterpene-derived HOM formation. By incorporating detailed reaction pathways and refining key parameters, the model not only reproduces observed concentration but also provides mechanistic insights into HOM formation. These advancements establish our sesquiterpene-HOM module as a valuable tool for simulating complex chemistry under atmospheric conditions.

In this section, we compared the observed and simulated HOM concentrations in mixed systems under various conditions, including different VOC combinations, changes in VOC concentrations and the introduction of UV light (see Figs. S4 and S5). Across all cases, the simulations overall matched the observed HOM distributions, accurately reproducing key peaks and maintaining the expected carbon number distributions. Additionally, we analyzed the variation in HOM concentrations with different carbon numbers for four representative mixing conditions: (1) a mixture of isoprene and monoterpenes (Mix I-MT600), (2) an increased monoterpene concentration (Mix I-MT1200), (3) the introduction of UV light (Mix I-MT1200, Illuminated), and (4) the subsequent addition of sesquiterpenes (Mix II-High, Illuminated).

In the initial isoprene–monoterpene mixture, oxidation led to a diverse range of HOMs from both C₅– and C₁₀–RO₂ pathways (Fig. S4a). Increasing the monoterpene concentration enhanced HOM formation, particularly for monoterpene-derived species, like C_6–10– and C_16–20–HOMs (from Mix I-MT600 to Mix I-MT1200, Fig. 6b and d). The introduction of UV light shifted the HOM distribution by promoting isoprene oxidation, leading to an increase in C_1–5- and C_1–15-HOMs (from Mix I-MT1200 to Mix I-MT1200, illuminated, Fig. 6a and c) while reducing monoterpene-derived termination products, in particular dimers with more than 15 carbon atoms. However, the model underestimated C_11–15–HOM concentrations, likely due to an underrepresentation of cross-reaction between isoprene-derived and monoterpene-derived RO₂. The addition of sesquiterpenes further altered HOM distribution, notably increasing C_11–15–HOM (Fig. 6c), suggesting sesquiterpenes exhibit a higher propensity for autoxidation and further amplifying these product channels. However, the significant decrease in C_16–30–HOM concentration observed in three-VOC mixture, which was not reflected in the simulations (Fig. 6d), could be attributed to the overestimated rate of dimer formation involving sesquiterpenes. This shift was likely driven by the competitive consumption of available RO₂ radicals, which are reflected in both observations and models. Our model's ability to simulate key features – such as the redistribution of HOM classes due to competitive RO₂ chemistry – demonstrates its robustness in simulating the oxidation of complex VOC mixtures. This agreement highlights the model's capability to provide mechanistic insights into HOM formation under varied atmospheric conditions.