To understand the current state of SOA modelling in CTMs, we reviewed schemes implemented in several regional models that are used in the US, Europe, and Asia as well as one global model. We review eight SOA schemes implemented in five models, namely the Comprehensive Air Quality with Extensions (CAMx, https://www.camx.com/, last access: 15 February 2024), the Community Multiscale Air Quality (CMAQ, https://github.com/USEPA/CMAQ/, last access: 15 February 2024), GEOS-Chem (https://geos-chem.readthedocs.io/en/stable/, last access: 15 February 2024), WRF-Chem (NOAA, 2022, https://www2.acom.ucar.edu/wrf-chem, last access: 30 May 2026), and CHIMERE (https://www.lmd.polytechnique.fr/chimere/docs/, last access: 15 February 2024). For each model/scheme (hereafter “scheme” for simplicity), we reviewed the official documentation (e.g., user's guide), peer-reviewed publications, and, in some cases, the model source code to understand each SOA parameterization and gather parameter data. A comparative overview of the SOA schemes is presented in Table 1 while comprehensive details regarding their specific parameterizations can be found in Sect. S1 of the Supplement.

The review of these SOA schemes reveals diverse parameterizations ranging from simplified, non-volatile assumptions to complex, multi-dimensional volatility basis sets. CAMx offers two distinct approaches: the SOAP2 scheme (based on Strader et al., 1999) and the 1.5-D VBS (Koo et al., 2014). SOAP2 has non-volatile POA and a two product SOA scheme (Fig. 1) with yields fitted to the aged VBS scheme of Hodzic et al. (2016), effectively treating aging as implicit. The 1.5-D VBS treats POA as semi-volatile and explicitly models gas-phase aging for anthropogenic and intermediate volatility precursors via OH-oxidation (Fig. 2b), although this stepwise aging is disabled for biogenic precursors to prevent aerosol mass overprediction. The CMAQ model also provides alternative schemes: the established AERO7 scheme (Appel et al., 2021) and the newer CRACMM scheme (Pye et al., 2023). AERO7 utilizes a 1-D VBS framework for POA and SOA that incorporates aging primarily through particle-phase processes (Fig. 2d), specifically the oligomerization of anthropogenic and biogenic precursors and the hydrolysis of organic nitrates derived from monoterpenes. CRACMM also utilizes a 1-D VBS framework for POA and SOA and simulates aging through sequential gas-phase oxidation reactions involving functionalization and/or fragmentation (Fig. 2c; Pye et al., 2023). GEOS-Chem includes a “Simple” scheme that treats SOA as non-volatile with fixed yields that are linked to ambient measurements, alongside a “Complex” 1-D VBS scheme without additional aging processes (Fig. 2a; Pai et al., 2020). CHIMERE's 1-D VBS scheme is notable for its comprehensive aging scheme with functionalization, fragmentation, and oligomerization (Fig. 2d) where oxidation products are redistributed across volatility bins (CHIMERE Development Team, 2024). The WRF-Chem MOSAIC scheme employs a 1-D VBS for most VOCs but applies a specific stepwise gas-phase aging mechanism exclusively to IVOCs (Shrivastava et al., 2011).

Despite these structural differences, the schemes share foundational similarities, particularly in the reliance on absorptive partitioning theory by most schemes. With the exception of the GEOS-Chem Simple scheme, which assumes irreversible condensation, all models utilize either a two-product or VBS framework to describe the equilibrium partitioning of semi-volatile organic compounds. However, the treatment of aging remains the most significant source of divergence. Approaches vary from neglecting aging entirely (GEOS-Chem Complex, CAMx SOAP2) to implementing distinct mechanisms such as gas-phase oxidation (CAMx VBS, CMAQ CRACMM) versus particle-phase oligomerization (CMAQ AERO7). Additionally, the representation of IVOCs varies substantially, ranging from omission from the GEOS-Chem Simple scheme, a single lumped IVOC in most schemes, and several lumped IVOCs in the CRACMM scheme.

The direct comparison of simulated SOA concentrations across different schemes through conducting full simulations is time-consuming and uncertain because configuring all models consistently is challenging. Furthermore, differences in the non-SOA physical and chemical processes between models may obscure the distinctions attributable specifically to the SOA schemes themselves. To address this issue, an offline calculation (Huang et al., 2023, 2024) is employed to compare the non-aged SOA yield (i.e., prior to any aging effects) associated with different precursors across various schemes. For a two-product scheme, the non-aged SOA yield (Y) is calculated by combing the gas-particle partitioning theory with the stoichiometric coefficients αi: 1Y=α11+C1*/COA+α21+C2*/COA where COA is the total ambient concentration of organic compounds (i.e., POA + SOA) and α1, α2, C1* and C2* represent the stoichiometric coefficients and the effective saturation concentrations of the above two products, which is obtained by fitting the results of laboratory studies. Similarly, for a four-bin VBS scheme with no aging effects, the SOA yield is calculated as: 2Y=α11+1/COA+α21+10/COA+α31+100/COA+α41+1000/COA where αi is the non-aged oxidation yield for each volatility bin i (i=1,2,3,4). Utilizing either Eq. (1) or Eq. (2), the SOA yields under high- and low-NO_x conditions can be determined at 298 K and total OA concentrations ranging from 0.1 to 50 µgm-3, using the stoichiometric coefficients provided by each scheme (see detailed parameters in Sect. S1 of the Supplement).

The non-aged SOA yield Y is always calculated as mass-based in this study while the stoichiometric coefficients αi could either be expressed in mass-base (gg-1) or molar-base (ppmppm-1). An illustrative calculation is provided in Sect. S2 of the Supplement, with calculation spreadsheet publicly available (see “Data availability” section). The non-aged SOA yields for CMAQ CRACMM are calculated slightly differently (details presented in Sect. S3 of the Supplement), given its special treatment of partially combining gas-phase chemistry and SOA formation. Our analysis included calculations for anthropogenic precursors (benzene, toluene, and xylene), IVOC and biogenic precursors (isoprene, monoterpene, and sesquiterpenes).

Some schemes, such as CAMx two-product and GEOS-Chem Simple, do not account for SOA aging while others adopt varying approaches to represent the aging process (for a comprehensive discussion, refer to Sect. 3). For schemes that include aging effects, we calculated the aged SOA yields for each precursor at a given time t by summing over the particle fraction of all the relevant volatility bins (i) using Eq. (3): 3AgedSOAyields|t=∑n(fparticlei⋅SOA mass|ti) where fparticlei is the particle-phase fraction of each volatility bin (calculated based on Eq. 4) and SOAmass|ti is the bin total SOA mass (gas-phase + particle-phase) at time t; n is the total number of bins. 4fparticlei=11+Ci*/COA

For gas-phase OH-oxidation style aging (e.g., Fig. 2b and c), the SOA mass is stepped through time Δt=t-(t-1) as follows: 5SOAmass|ti=SOAmass|t-1i×fparticlei+fgasi⋅e-kOH*OH*Δt+∑nSOAmass|t-1k×fgask⋅1-e-kOH*OH*Δt×αki

The first term on the right hand side is the SOA mass in volatility bin (i) from the previous time step (t-1) multiplied by the fractions that remain after Δt in the particle-phase (fparticlei) and gas-phase (fgasi⋅e-kOH*OH*Δt; fgasi=1-fparticlei) considering OH-oxidation. The second term is the SOA mass gain from OH-oxidation summed across all n volatility bins. Here, αki is the mass yield coefficient from bin k to bin i and the term fgask⋅(1-e-kOH*OH*Δt) is the gas-phase fraction of SOA in bin k oxidized by OH during Δt.

For particle-phase oligomerization-style aging (e.g. Fig. 2d and the CMAQ AERO7 scheme), the aged SOA yield includes the mass of a non-reactive and non-volatile oligomer bin (OLIG) in addition to the semi-volatile bins: 6AgedSOAyield|t=∑n(fparticlei⋅SOAmass|ti)+SOAmass|tOLIG

The SOA mass (gas-phase + particle-phase) in each volatility bin (i) steps through time following Eq. (7) and the mass of the non-volatile oligomer bin (SOAmass|tOLIG) grows with mass-transfer from semi-volatile bins according to Eq. (8): 7SOAmass|ti=SOAmass|t-1i×(fgasi+fparticlei⋅e-kOLIG*Δt)8SOAmass|tOLIG=SOAmass|t-1OLIG+∑n{SOAmass|t-1i⋅fparticlei⋅(1-e-kOLIG*Δt)⋅βi} kOLIG is the oligomerization rate and βi is the mass yield coefficient from bin i to the non-volatile bin OLIG.

To compare the aging effects of different schemes, we applied Eqs. (3) to (8) for 1 d of aging with an OH concentration of 3×106 molec. cm⁻³ with kOH and/or kOLIG for each scheme when applicable. A time step (Δt) of 0.2 h was used. Any additional calculations and assumptions associated with each scheme are further described below.

We implemented three updated SOA schemes in CAMx and performed 2-layer box modeling of two locations with varied anthropogenic emissions to quantify how differences between schemes can influence predicted SOA concentrations and their response to emission changes. The alternate SOA schemes use the existing CAMx SOAP2 code with updated SOA yield parameters so that model results clearly depend on yield assumptions rather than scheme formulation or coding. Two regions of Texas were selected to capture contrasting emission environments: Dallas-Fort Worth (DFW), a major urban area dominated by anthropogenic emissions (e.g., aromatics and IVOCs), and Tyler (TYL), a rural area in Northeast Texas characterized by high biogenic activity. The box model has a surface layer and a residual layer with time-varying surface layer depth to provide a simple representation of pollutant accumulation, carry-over, and diurnal variation. Simulations were conducted over a 5 d period, utilizing meteorological inputs and initial conditions derived from the Texas Commission on Environmental Quality (TCEQ) 2019 3-D CAMx modeling platform (https://www.tceq.texas.gov/airquality/airmod/data/tx2019, last access: 30 May 2026).

We implemented three new SOA schemes into CAMx that emulate SOA yields produced by the CMAQ AERO7, CMAQ CRACMM and Simple schemes. Each scheme was implemented by updating the yield parameters used by the CAMx SOAP2 scheme. For the CMAQ AERO7 and CRACMM schemes, yield curves were fitted to the respective data (Fig. S3 in the Supplement) for each SOA precursor. These yields were mapped to the volatility bins defined by CAMx SOAP2 to obtain the corresponding molar-based stoichiometric coefficients (Tables S11–S14). Like the GEOS-Chem Simple scheme, the CAMx Simple scheme treats SOA as non-volatile with fixed yields that are based on multi-model averages and work by Seltzer et al. (2021). Detailed descriptions of the fitting procedures, updated Simple yields, and box model configurations are provided in the Sect. S3 of the Supplement.

We further investigate the response of SOA concentrations to varying anthropogenic VOC and NO_x emissions by performing a matrix of 100 simulations for each location and SOA scheme. This approach allows for a comparison of scheme performance across a wider range of atmospheric concentrations and VOC / NO_x ratios. Anthropogenic emissions in the sensitivity runs were based on weekday rates, while the biogenic emissions remained unscaled and varied by date, consistent with the base case simulations. Anthropogenic VOC emissions were scaled from 0.1 to 1.0 (in increments of 0.1), while anthropogenic NO_x emissions were scaled from 0 to 9 (in increments of 1) which caused oxidant production to transition between NO-limited and VOC-limited conditions. An additional simulation with a 50 % reduction in NO_x emissions was conducted to examine SOA response to NO_x abatement, which is a critical consideration for current and future air quality planning.

Non-aged SOA mass yields (gg-1) for all schemes (except CMAQ CRACMM) are summarized in Table 2. Non-aged refers to the yields immediately following precursor oxidation and before any subsequent SOA aging process. Since mass yields can vary with COA, we assume COA of 10 µgm-3 in Table 2, which is relevant to ambient air quality and often used as a reference COA for SOA yield comparisons. The ratio of maximum to minimum non-aged SOA yields for a same precursor shows wide variations across different schemes, with the least variation (factor of 1.8) observed for monoterpene at low NO_x conditions and over three orders of magnitude for IVOC (factor of 3715). These variations can become greater when COA is either increased or decreased from 10 µgm-3 (as assumed for Table 2) and when effects of aging processes on non-aged yields are included (as shown in Sect. 4.2).

Among the eight schemes we evaluated, the CAMx SOAP2, GEOS-Chem Simple, and GEOS-Chem Complex schemes do not incorporate explicit SOA aging processes. However, the SOAP2 yields are derived from a VBS parameterization that includes aging (Hodzic et al., 2016) and may therefore be considered pre-aged (Emery et al., 2024). The remaining schemes account for SOA aging using one or more of three mechanisms: gas-phase OH-oxidation, condensed-phase oligomerization, and condensed-phase hydrolysis. Gas-phase OH-oxidation aging is typically parameterized as functionalization reactions that generate less volatile products (e.g. Fig. 2b) and/or fragmentation reactions that produce more volatile products (e.g. Fig. 2c). This mechanism is adopted in CAMx VBS (for AVOC and IVOCs only), CMAQ CRACMM, WRF-Chem MOSAIC and CHIMERE VBS, with implementation being specific to each scheme. Condensed-phase oligomerization aging (e.g. Fig. 2d) is characterized as a first-order particle-phase reaction, usually assuming a lifetime of 30 h, leading to non-volatile product formation independent of oxidant concentrations. CMAQ AERO7 applies this process for SOA formed from all precursors except monoterpenes. For SOA derived from NO₃ oxidation of monoterpenes, AERO7 instead applies condensed-phase hydrolysis, yielding non-volatile products with a lifetime of ∼3 h.

To compare the aging effects on SOA yields across different schemes, we assumed a 24 h exposure to an OH concentration of 3×106 molec. cm⁻³ (equivalent to 2.6×1011 molec. s⁻¹ cm⁻³) for the OH-oxidation aging process or a 24 h of condensed-phase oligomerization/hydrolysis when calculating the aged SOA yields (details presented below). Table 3 summarizes the aged SOA yields (when applicable) at 298 K with a COA of 10 µgm-3 as simulated by each scheme. Figures 6 and 7 further compare the non-aged and aged SOA yields for each scheme and SOA precursor, while Fig. 8 separately illustrates the aging effects on IVOC-derived SOA across different schemes.

Evaluating SOA yields for seven precursor types across eight modeling schemes results in 56 potential characterizations of how NO_x levels influence SOA formation. Tables S18 and S19 present the ratios of non-aged and aged SOA yields under high and low NO_x conditions. For non-aged SOA yields, 11 of the 56 cases are missing and 14 cases are designed to exhibit no NO_x dependence, leaving 31 meaningful comparisons. Among these, 29 cases show lower SOA yields under high NO_x conditions. For aged SOA yields, 5 cases are missing and 14 are designed with no NO_x effect, leaving 37 meaningful comparisons, of which 34 also show lower SOA yields under high NO_x conditions. These results indicate a general trend of higher SOA yields under low NO_x conditions, although a few exceptions are observed. In some cases, the NO_x effect is reversed – that is, SOA yields are higher under high NO_x conditions than under low NO_x conditions. This is seen for BENZ in the CAMx SOAP2 scheme, TERP in CMAQ CRACMM, and SESQ in the GEOS-Chem Complex scheme. The NO_x effect on terpene-derived SOA in CRACMM is particularly noteworthy: the model predicts an eightfold increase in SOA yields under high NO_x conditions compared to low NO_x. This is significant given that terpenes are key SOA precursors in many forested regions, such as the Eastern US, where anthropogenic NO_x emissions may change due to ongoing urban development (which could increase NO_x levels) or the implementation of emission control technologies (which may reduce them). Experimental studies, including those by Sarrafzadeh et al. (2016) and Wildt et al. (2014), have consistently found that terpene-derived SOA yields are higher under low NO_x conditions, a result that aligns with most of the evaluated schemes but contrasts with the predictions made by CRACMM.

The box model simulations reveal significant discrepancies in predicted SOA concentration and composition among the selected schemes at both the urban (DFW) and rural (TYL) locations (Fig. 9). The total SOA concentrations can vary by a factor of 2–3 between schemes even under identical meteorological and emission inputs. Overall, SOA concentrations are higher at DFW than at TYL. At both locations, all schemes exhibit a consistent diurnal profile characterized by SOA accumulation throughout the day and night, followed by a sharp decline in the early morning (beginning around 06:00 LST) caused by expansion of the planetary boundary layer (PBL). At DFW, the temporal trends are similar across the four schemes; however, the magnitude varies, with Simple and AERO7 predicting the highest concentrations, while CRACMM predicts the lowest. At TYL, the inter-model spread is narrower than at DFW. Notable differences in diurnal dominance emerge: SOAP2 and CRACMM predict the highest concentrations overnight – a pattern distinct from DFW – while Simple and CRACMM produce the highest values during daytime hours. Figure 9c–d shows the maximum and minimum average SOA concentrations. At DFW, Simple and CRACMM predict the highest and lowest total SOA concentrations, respectively. Conversely, at TYL, CRACMM predicts the highest average concentration, while AERO7 predicts the lowest.

Response surface plots for 24 h average SOA concentrations derived from the matrix of simulations with varied anthropogenic NO_x and VOC emissions are presented in Figs. S4–S5. At DFW, total SOA concentrations generally decrease as anthropogenic NO_x emissions increase relative to the base case (scaling factors 2–9) for all schemes, with the notable exception of CRACMM. CRACMM predicts negligible changes or a slight increase in SOA as NO_x rises from low to mid scaling factors and a decrease is observed only when NO_x emissions are increased by about a factor of 7 or greater. CRACMM also predicts the lowest SOA concentrations across all scaling factors. AERO7 and SIMPLE are very similar in both response surface shape and magnitude. For each scheme, SOA concentrations decrease as the anthropogenic VOC emissions decrease. At TYL, the response surfaces for AERO7, Simple, and SOAP3 are similar, with SOA concentrations remaining relatively constant at NO_x scaling factors greater than 1. In contrast, CRACMM demonstrates a much stronger response, with SOA mass increasing concurrently with NO_x emissions. At this biogenic-dominated site, SOA concentrations in all schemes are minimally sensitive to variations in anthropogenic VOC emissions.

Understanding how SOA responds to NO_x reductions is critical for near-term air quality planning, as many regulatory strategies (e.g., cleaner vehicles, energy transition) produce substantial NO_x emissions abatement (Crippa et al., 2016; EPA, 2017; Li et al., 2024). Table 4 summarizes the impact of a 50 % reduction in NO_x emissions on total SOA, ASOA, and BSOA. In all schemes, reducing NO_x leads to increased ASOA concentrations, with the most pronounced increase predicted by SOAP2. SIMPLE predicts the smallest increase in ASOA at DFW, while CRACMM predicts the smallest increase at TYL. At DFW, this NO_x reduction also drives an increase in BSOA concentrations across most schemes, with the notable exception of CRACMM. Consistent with the ASOA results, SOAP2 predicts the largest increase in BSOA. The distinct behavior in CRACMM is driven by its monoterpene SOA parameterization; contrary to other schemes and experimental evidence (Lane et al., 2008; Sarrafzadeh et al., 2016; Zhao et al., 2018), CRACMM predicts decreasing yields under lower NO_x conditions. Consequently, the substantial reduction in BSOA predicted by CRACMM at DFW results in a net decrease in total SOA, a trend opposite to that observed in the other schemes. At TYL, all schemes predict a decrease in both BSOA and total SOA, with the most significant reductions observed in the CRACMM simulation. The dominance of biogenic emissions at TYL compared to DFW is reflected in the significantly higher BSOA concentrations. The differing SOA responses to NO_x reduction between the two sites are attributable to the distinct biogenic emission regimes producing different VOC / NO_x emission ratios.

CTMs employ diverse approaches to simulate SOA, from simple schemes that treat SOApre as non-volatile to more complex VBS schemes that utilize multiple basis sets to represent different types of precursors. Variability in how SOA aging is treated further adds to the overall diversity across schemes. In our view, such diversity is valuable from a research perspective, given that the underlying processes driving SOA formation remain uncertain and, in many cases, poorly characterized. The variation in SOA yields across different schemes reflects the extent of these uncertainties. The differences in scheme formulation, coupled with the large numbers of parameters employed in some schemes, pose practical challenges for applying multiple schemes to standardized scenarios. Addressing these challenges may require innovative approaches. Nonetheless, comparisons under standardized conditions are essential for achieving meaningful quantitative inter-comparisons. Evaluating SOA yields under standard conditions and plotting SOA yield curves (i.e., yield vs. COA) are effective strategies for identifying similarities and differences among schemes. However, it is important to note that the results presented here may not fully capture the ranges of conditions encountered in three-dimensional atmospheric simulations.

Evaluating seven precursor types under both high and low NO_x conditions yields 14 distinct comparisons. Across these comparisons, the ratio of maximum to minimum non-aged yields (max / min) ranges from 1.8 to >1000 with a median max / min of 4.2 (Table 2). Among the eight schemes examined, three (CAMx SOAP2, GEOS-Chem schemes) do not include explicit SOA aging processes. Four schemes account for aging in a subset of precursor types and/or NO_x-conditions, while only one (CMAQ CRACMM) includes aging for all precursors. Aging mechanisms considered by these schemes include gas-phase OH-oxidation of evaporated SOA, particle-phase oligomerization, hydrolysis, and photolysis. The impacts of aging on SOA yields vary by scheme and precursor (Table S20): in 67 of the 98 evaluated cases (defined as one scheme/precursor/NO_x-condition combination, CRACMM excluded), aging has no effect; in 31 cases, it increases SOA yields. Considering the aging effects, the ratio of max / min aged yields ranges from 5.0 to >70, with a median value of 8.3 (Table 3). The relative rankings of precursors by their initial/aged SOA yields differ across schemes (Tables S22 and S23), indicating that different aging schemes can lead to divergent conclusions regarding the relative importance of specific SOA precursors – a consideration with potential implications for policy guidance. For instance, discrepancies in yields lead to different precursor rankings (e.g., the relative importance of aromatics vs. IVOCs). A model that underpredicts the SOA potential of IVOC emissions might disproportionately focus policy strategies on traditional VOCs (e.g., from petroleum-based solvents) and potentially lead to ineffective widespread controls that miss the critical contribution of IVOCs.

SOA aging remains an area in need of improved representation, with careful attention required to ensure consistent underlying assumptions. Notably, only the two CAMx schemes incorporate condensed-phase SOA photolysis, despite growing evidence that both anthropogenic and biogenic SOA can undergo substantial photolytic depletion (Hodzic et al., 2016; Baboomian et al., 2020), although a portion of SOA appears recalcitrant to such degradation (O'Brien and Kroll, 2019).

The SOA yields from IVOCs show wider variation (from negligible to 1.0 gg-1) than for other anthropogenic precursors (Tables 2 and 3), partly due to different assumptions across schemes. For example, schemes such as WRF-Chem MOSAIC and CHIMERE VBS predict very low non-aged yields from IVOC, based on the assumption that several generations of oxidation are required before forming condensable products. Even after 1 d of aging, IVOC SOA yields remain highly variable, ranging from 0.02 to 1.20 gg-1. Although IVOC are generally classified based on volatility, factors such as high molecular weight or the presence of polar functional groups can shift compounds into the IVOC volatility range (Pankow and Asher, 2008). As a result, volatility and SOA yield are not necessarily well correlated (Donahue et al., 2011). Improving model representations of IVOC-derived SOA yields will require more detailed differentiation of IVOC emissions into multiple subtypes, as illustrated by the CRACMM scheme (Pye et al., 2023). A unified classification or “lumping” scheme for IVOC would be particularly advantageous, allowing multiple models to utilize a common emissions framework and enabling more direct comparisons of IVOC SOA yields. Improving the representation of oxygenated VOCs with reduced volatility – such as glycols and glycol ethers – within gas-phase chemical mechanisms can also support improved differentiation of IVOC-related SOA formation (Yarwood and Tuite, 2024; Yu et al., 2023). More generally, a yield-based lumping approach for IVOCs (e.g., categorizing them into low, medium, or high yield classes) may be more practical to implement than strictly chemically-based schemes.

Laboratory experiments play a crucial role in guiding SOA model development and constraining key model parameters, particularly yields. However, these experiments are subject to operational and design limitations, including the need to account for chamber wall effects (Zhang et al., 2014) and to achieve atmospherically relevant concentration ranges (Peng et al., 2022; Kenagy et al., 2024). The role of autoxidation reactions in SOA formation further complicates the design of atmospherically relevant experiments, as discussed in detail by Kenagy et al. (2024). For instance, studying a reaction mechanism that includes RO₂ radical autoxidation at a rate of 0.1 s⁻¹ requires that the effective rates of competing bimolecular reactions, particularly RO2+ NO, be reduced to 0.1 s⁻¹ or lower. This necessitates NO mixing ratios below approximately 5 ppb, which are now typical of photochemically active urban environments such as Los Angeles (Praske et al., 2018). Many SOA chamber experiments designed to investigate high NO_x conditions exceed 5 ppb NO, thereby preventing autoxidation. Some chamber experiments, such as those by Sarrafzadeh et al. (2016), have been specifically designed to achieve atmospherically relevant NO (and other radicals) concentrations, making their results particularly valuable for SOA model development. In contrast, oxidation flow reactors face greater challenges (Peng et al., 2019) than chamber experiments in studying SOA formation due to their amplification of radical concentrations, which significantly shortens RO₂ lifetime and effectively suppresses autoxidation reactions. Wennberg (2023) has suggested shifting from the conventional terminology of high/low NO_x to high/low NO to emphasize the critical role of NO concentration in determining RO₂ radical fate.