Enhanced Isoprene Secondary Organic Aerosol Formation with C₅-alkene Triols Newly Added to Current Chemical Mechanisms

Abstract

Particulate 2-methyltetrols (2-MT), 2-methylglyceric acid (2-MG), and C₅-alkene triols serve as critical tracers for identifying isoprene-derived secondary organic aerosols (SOAs). Field observations have identified significant concentrations of C₅-alkene triols, yet no air quality models currently account for the formation pathways of these compounds. Simultaneously, the 2-MT concentrations are usually overestimated by 4–5 times compared to field observations due to the lack of the C₅-alkene triols formation pathway. In this study, we expanded the isoprene-SOA scheme and implemented it into the Community Multiscale Air Quality (CMAQ) model with the latest Community Regional Atmospheric Chemistry Multiphase Mechanism version 2 (CRACMM2) mechanism. The missing formation of C₅-alkene triols via the low ${\text{NO}}_x$ isoprene epoxydiols (IEPOX) pathway, the gas-particle partitioning of isoprene-derived SOA tracers, and the high NOx SOA formation pathway are comprehensively considered in this expanded isoprene-SOA scheme. Results indicate that the C₅-alkene triols contribute to 68% of all isoprene-derived SOA tracers, with approximately 50% in the gas phase and the remaining in the particle phase. The 2-MT concentrations are better represented in the new scheme compared to the default CRACMM2 based on the field observations. The overall SOA is improved by $0-4\mu \text{g}/{\text{m}}^3$ in China with the expanded scheme. The isoprene-derived SOA is highly influenced by anthropogenic emissions, especially ${\text{SO}}_x\left({\text{SO}}_2+{{\text{SO}}_4}^{2-}\right)$, and the reduction in ${\text{SO}}_x$ reduces both isoprene aerosol tracers and organosulfates significantly, while a similar reduction in ${\text{NO}}_x$ leads to small increases in these species for both particle and gas phases.

Graphical Abstract

Introduction

Isoprene is the most abundant biogenic volatile organic compound emitted into the atmosphere, with annual global emissions estimated to range from 500 to 750 Tg¹.. Isoprene is highly reactive, as it can be oxidized by hydroxyl radical (HO·)^2,3, ozone (O₃)⁴, and nitrate radicals (NO₃·)⁵. Isoprene oxidation products can undergo multiphase chemical reactions, leading to the generation of semi-volatile and low-volatility organic aerosol products^6–8.

The chemical composition of isoprene-derived secondary organic aerosols (SOA) has been extensively investigated in chamber experiments, field observations, and modeling studies. Among the identified isoprene-derived SOA species, IEPOX-derived compounds, including 2-MT and C₅-alkene triols, have been used as molecular organic tracers to study isoprene-derived SOA composition and formation mechanisms^9,10. Another widely recognized tracer, 2-MG, is derived from methacrylic acid epoxide (MAE)⁷ and hydroxymethyl-methyl-$\alpha$-lactone (HMML)¹¹, both of which are produced from MPAN species under later-generation organic peroxy (RO₂) + NO conditions. This tracer is essential for improving the representation of isoprene oxidation pathways leading to SOA formation, even though its concentration is lower than 2-MT and C₅-alkene triols. However, most chemical transport models and mechanisms (except SAPRC07TIC_ae7i) do not include this high ${\text{NO}}_x$ pathway for isoprene-derived SOA formation.

3-Methylenebutane-1,2,4-triol and 3-methyltetrahydrofuran-2,4-diols, designated as “C₅-alkene triols”, are notably high in field measurements. However, previous chamber and field studies using alternative analytical techniques suggest that observed C₅-alkene triols reactive uptake products are mostly artifacts from the decomposition of methyltetrol sulfate esters and oligomeric sulfated compounds^12,13. Recently, based on nondestructive hydrophilic-interaction liquid chromatography coupled with electrospray ionization interfaced to high-resolution quadrupole-time-of-flight mass spectrometry analysis of chamber-generated SOA, parts of C₅-alkene triols are not analytical artifacts but arise from acid-driven particle-phase IEPOX isomerization¹⁴. These triols have been shown to partition substantially into the gas phase. This process has been demonstrated as an additional important and understudied reactive pathway¹⁴. The observation is consistent with field studies from the Southeastern United States (SE USA), Central Amazonia, and Eastern China, which indicate that isoprene-derived C₅-polyols undergo gas-phase partitioning after being formed within particles through multiphase IEPOX chemistry^15–18. This potential source of isoprene-derived SOA has been overlooked and as a result, variations in field measurements have not been adequately explained, and models of SOA formation and distribution likely require revision since none of the models involves this pathway.

To address these limitations, this study expands the regional isoprene-derived SOA scheme of CRACMM2 implemented in the CMAQ model to simulate the formation of the key isoprene-SOA molecular tracers, C₅-alkene triols. To the best of our knowledge, this is the first modeling study to estimate the formation of C₅-alkene triols and their partitioning. In order to provide a more comprehensive representation of the complete isoprene oxidation mechanism, we incorporates a high-${\text{NO}}_x$ pathway into the model based on previous studies that employed the SAPRC07TIC_aero7i^7,19,20 mechanism. Given that C₅-alkene triols are newly added to the model, sensitivity simulations are conducted to address uncertainties in the rates of particle-phase reactions and the Henry’s Law coefficients $\left(H^{*}\right)$ of their gas-phase precursor. 2-MT, 2-MG, and C₅-alkene triols are treated as semi-volatile compounds with gas-particle partitioning. As the gas-particle phase partitioning equilibrium in the model is governed by $H^{*}$ of these tracers, the sensitivity to $H^{*}$ is assessed in this study as well. We evaluate the model results after constraining the parameters based on partitioning data and observations. Additionally, we explored the model’s response to reductions in ${\text{SO}}_x$ and ${\text{NO}}_x$ emissions. The results of this study will enhance our understanding of isoprene oxidation pathways and SOA formation.

Methods

Observational Data

Field observations of SOA organic tracers by our group were conducted at the super monitoring station of Zibo Ecological Environment Bureau (36°50^′N, 118°7^′E; 25 m a.s.l., shown in Figure S1 in Supporting Information), which is located in a suburban area surrounded by Lutai Avenue (100 m south), Lushan Avenue (500 m east), residential housing (1 km north), and a gas station (210 m west). This area is predominantly covered by deciduous broadleaf vegetation. High-time resolution measurements of PM_2.5-bound organic molecular tracers were conducted during the summer period from 6 June to 10 July 2022 using an Aerodyne TAG stand-alone system combined with a gas chromatograph-mass spectrometer (GC-MS) (Agilent GC 7890B/MS 5977B, Agilent, CA, USA). In this field campaign, the major isoprene-derived SOA organic tracers, including 2-MT, 2-MG and C₅-alkene triols were measured online. Isoprene was measured by an online GC-MS system (FID GC5800, Thermo Fisher Scientific). A detailed description of the analytical system and procedures are provided in Supporting Information “Online TAG-GC/MS Observation”. The on-line derivatization step was performed on C₅-alkene triols and the derivatization thermal desorption GC-MS mass spectra associated with extracted ion chromatogram (EIC) traces of m/z 231 are shown in Figure S2. Hourly concentrations of PM_2.5, NO₂, SO₂, and O₃ from national monitoring sites were obtained from the platform released by China National Environmental Monitoring Centre (http://air.cnemc.cn:18007), which are used to perform model performance evaluation.

Model Development and Application

In this study, CMAQ version 5.5 was employed over the China domain at a horizontal resolution of 36 km × 36 km, with 35 vertical layers, from June 1 to June 30, 2022. Emissions were based on the 2019 Multi-resolution Emission Inventory for China (http://www.meicmodel.org). Biogenic emissions were generated using the Model of Emissions of Gases and Aerosols from Nature version 3.2 (https://aqrp.ceer.utexas.edu/projects.cfm)²¹. Additionally, CMAQ-ready emission files were prepared using the Sparse Matrix Operator Kernel (SMOKE) model²². Meteorological input files for CMAQ were generated offline by feeding the Weather Research and Forecasting model version 4.4 output files through the Meteorology-Chemistry Interface Processor (MCIP)²³. CRACMM2²⁴ builds upon the framework of the Regional Atmospheric Chemistry Mechanism, Version 2 (RACM2), enhancing its capabilities by integrating gas- and particle-phase chemistry. It treats the entire pool of atmospheric reactive organic carbon relevant to current emissions, providing a more comprehensive representation of atmospheric processes²⁴. In this work, the CRACMM2 mechanism has been expanded to include the formation of C₅-alkene triols for the first time in atmospheric modeling. It also incorporates 4 gas-phase reactions, 6 heterogeneous reactions, and 16 new species, including MAE and HMML formation in the gas phase (Figure 1). The detailed reactions are described in the Supporting Information, equations S1–S10. Organonitrates from IEPOX and MPAN are less abundant¹⁹, primarily due to low nitrate aerosol concentrations during summer, and thus are not explicitly represented in this model²⁰. Additionally, the model accounts for gas-particle phase partitioning, further enhancing its representation of atmospheric processes.

Figure 1:

Schematic of the isoprene-derived SOA tracers’ formation and partitioning, (a) current CRACMM2 approach, (b) expanded scheme, where the red dashed box shows MPAN-derived SOA formation pathways. Solid green lines indicate gas phase chemistry leading to IEPOX, MAE and HMML. Dashed green bidirectional arrows depict that SOA tracers partitioned to gas. Blue solid lines represent isoprene intermediate gases leading to SOA. The organosulfates derived from IEPOX and MPAN included in the model are not shown here.

Aerosol-Phase Chemistry

CRACMM2 includes 179 gas-phase species and 531 reactions spanning both gas-phase and heterogeneous pathways. It incorporates Automated Model Reduction (AMORE) version 1.2 (v1.2), upgraded from the previous version implemented in CMAQ (AMOREv1.0,²⁵) and GEOS-Chem (AMOREv1.1,²⁶). In contrast to previous AMORE representations^25,26, AMOREv1.2 was specifically designed to enhance ${\text{NO}}_x$ cycling and O₃ predictions compared to previous versions. Additionally, it introduces new SOA precursors, including lumped isoprene tetrafunctional compounds (IPX) and lumped isoprene nitrates (INALD). These new precursors are expected to undergo heterogeneous uptake at rates 2× and 0.5× that of IEPOX, respectively²⁴. The CRACMM2 mechanism only includes isoprene-derived SOA from heterogeneous pathways (no volatility-based pathways)²⁴. Isoprene undergoes reactions with HO, NO, NO₃· and O₃, with epoxides forming exclusively through the HO· reaction. In this study, we define the case labeled as Case0, which incorporates the most optimal model parameters, as outlined in Table 1. This case also utilizes the updated isoprene-derived SOA formation pathways, as illustrated in Figure 1b.

Table 1:

New isoprene-derived SOA species considered in the CMAQ model along with their molecular weight, OM/OC ratio, parent hydrocarbon identity, nucleophile that adds to the parent, and the third-order rate constants $K_{i,j}$⁵, where $i$ refers to the nucleophile (e.g., H₂O), and $j$ refers to the acid species involved in the reaction, such as H⁺ or ${\text{HSO}}_4^{-}$, as well as parameters changed in the default CRACMM2 model (parameters used in Case0). IEPOX_OS means IEPOX-derived organosulfates, MPAN_OS means MPAN-derived organosulfates.

Species	MW	OM:OC	Parent Hydrocarbon	Nucleophile	$K_{i,{\text{H}}^{+}}$	$K_{i,{\text{HSO}}_4^{-}}$
2-MT	136	2.27	IEPOX	water	2.5 × 10−4a	1.3 × 10−5
C5-alkene triols	118	2.97	IEPOX	water	7.0 × 10−4b	3.6 × 10−5
IEPOX_OS	216	3.02	IEPOX	sulfate	8.8 × 10−3	2.9 × 10−6
2-MG	120	2.50	MPAN/MAE	water	2.5 × 10−3	1.3 × 10−5
MPAN_OS	200	4.17	MPAN/MAE	sulfate	2.0 × 10−4	2.9 × 10−6

^a)Sensitivity analysis indicated that the default $K_{i,j}$ for the particle-phase reactions ($K_{\text{water},H^{+}}$) are too high, and a value of 2.0 × 10⁻⁴ M⁻²s⁻¹ is more consistent with observation¹⁹. After taking partitioning into consideration, this value was changed to 2.5 × 10⁻⁴ M⁻²s⁻¹.

^b)Newly added to CRACMM2 based on partitioning of C₅-alkene triols and observations. More details are given in the “Particle-phase rate constants” section.

Particle-phase rate constants

The optimized particle-phase rate constants are shown in Table 1, and in this section, we mainly focus on how the rate constants for 2-MT and C₅-alkene triols are selected. $k_{\text{water},H^{+}}$ of 2-MT is set to 9.0×10⁻⁴ M⁻²s⁻¹ in CRACMM2, based on the work of Pye et al.¹⁹. In order to better represent the chemical processes, the $k_{\text{water},H^{+}}$ was split into two components to better represent the chemical processes based on the observations of those two species ratios. The average ratios of 2-MT to C₅-alkene triols were 0.97±1.17, 1.33±1.24, and 3.97±3.08 at 8, 120, and 260 m in altitude respectively²⁷. Some studies also suggested that the loading of 2-MT increased with the enhancement of aerosol acidity²⁸ and found that the relative humidity can affect the ratio of 2-MT to C₅-alkene triols⁸. Recent studies suggested the ratio of 2-MT to C₅-alkene triols decreased with aerosol acidity²⁹. Since this is the first time C₅-alkene triols have been considered in the modeling, the particle-phase rate constants were adjusted based on the ratio of 2-MT to C₅-alkene triols observed in the China Zibo site (i.e., ~1:3). For 2-MT, the $k_{\text{water},H^{+}}$ was set to 2.5 × 10⁻⁴ M⁻²s⁻¹, while for C₅-alkene triols, it was assigned a value of 7.0 × 10⁻⁴ M⁻²s⁻¹, for the $k_{\text{water},H^{+},{\text{SO}}_4^{2-}}$ for C₅-alkene triols we use the same method to determine its value..

IEPOX-derived organosulfates were determined based on the 3rd-order reaction rate parameters in Table 1 and implemented as a heterogeneous uptake process, just like in the case of 2-MT. To quantify the relative abundance of organosulfates versus tetrols (2-MT + C₅-alkene triols), a beta $\left(\beta \right)$ parameter, defined as the ratio of organosulfates to tetrols, can be described by equation 1³⁰.

$$\text{IEPOX}\text{SOA}=\beta \times \text{IEPOX-derived}\text{organosulfates}+(1-\beta )\times \text{tetrols}$$ (1)

Gas-organic phase partitioning

In this study, 2-MT, 2-MG, and C₅-alkene triols are treated as semi-volatile, so they will repartition between gas and organic phases, as shown in Figure 1. The fraction of tracers in the gas phase can be determined using equation 2, and the fraction of tracers in the particle phase is defined as the sum of the fractions in the organic phase and aqueous phase and calculated following equation 3³¹.

$$F_{\text{gas}}=\frac{1}{1+\frac{C_{\text{om}}}{C^{*}}+f_H}$$ (2)

$$F_{\text{P},\text{t}}=1-F_{\text{gas}}$$ (3)

Where $C_{\text{om}}$ represents the concentration of the absorbing organic matter ($\mu \text{g}/{\text{m}}^3$). The saturation mass concentration $C^{*}\left(\mu \text{g}/{\text{m}}^3\right)$ is calculated using equation 4:

$$C^{*}=\frac{{10}^6{p}_L^0\gamma \overline{\left(M{W}_{\text{OM}}\right)}}{RT}$$ (4)

where $p_L^0$ is the subcooled vapor pressure (Pa), $\gamma$ is the activity coefficient of the compound in the absorbing OM phase (assumed to be unity), and $M{W}_{\text{OM}}$ is the mean molecular weight of the OM phase assumed to be 200 g/mol according to Williams et al.³². In this study, the $p_L^0$ of 2-MG (3.76 × 10⁻⁴ Pa), 2-MT and C₅-alkene triols (1.81 × 10⁻⁴ Pa) were adopted based on the work of Zhang et al³¹, which gives $C^{*}$ values of 30.4 and $14.6\mu \text{g}/{\text{m}}^3$, respectively. The enthalpy of vaporization ($\Delta H_{\text{vap}}$) of the three species are 130.0 (2-MG) and 117.7 kJ mol⁻¹ (2-MT and C₅-alkene triols).

The gas-aqueous phase distribution factor $f_H$ is calculated by equation 5, where the gas-aqueous phase partitioning equilibrium is described by $H^{*}$:

$$f_H=\frac{C_{\text{aq}}}{C_{\text{gas}}}={10}^{-6}{H}^{*}RT{A}_{\text{LW}}$$ (5)

where $C_{\text{aq}}$ is the tracer concentration in the aqueous phase ($\mu \text{g}/{\text{m}}^3$) and $C_{\text{gas}}$ is the tracer concentration in the gas phase ($\mu \text{g}/{\text{m}}^3$)³¹; $H^{*}$ is the Henry’s law constant (M atm⁻¹); $R$ is the ideal gas constant (atm M⁻¹K⁻¹); $T$ is the ambient temperature (K); and $A_{\text{LW}}$ is the aerosol water content ($\mu \text{g}/{\text{m}}^3$). $H^{*}$ estimated using SPARC online for 2-MT (3.38×10¹⁰ M atm⁻¹), C₅-alkene triols (8.14×10¹⁰ M atm⁻¹), and 2-MG (5.25×10⁸ M atm⁻¹) at 298 K are applied in Case0 of this study. Their uncertainty in the model predictions is discussed in the “Sensitivity Simulations” section.

For 2-MG, $H^{*}$ accounts for the changes in its aqueous solubility due to aerosol acidity, as shown in equation 6:

$$H^{*}=H_{298K}\left(1+\frac{K_a}{\left[H^{+}\right]}\right)$$ (6)

where $K_a$ is the acid dissociation constant of 2-MG (pKa = 3.5, pKa values are not relevant for 2-MT and the C₅-alkene triols),³³ and $\left[H^{+}\right]$ is the concentration of hydrogen ions. The temperature dependence of $H^{*}$ for 2-MT and C₅ is described by the Van ‘t Hoff equation (equation 7):

$$H^{*}\left(T_2\right)=H^{*}\left(T_1\right)\text{exp}\left(\frac{{\Delta }_{\text{sol}}H}{R}\left(\frac{1}{T_1}-\frac{1}{T_2}\right)\right)$$ (7)

where ${\Delta }_{\text{sol}}H$ is the enthalpy of dissolution; and the values of ${\Delta }_{\text{sol}}H/R$ for 2-MT, C₅-alkene triols, and 2-MG are 6×10³ K, estimated by EPI Suite³¹.

Sensitivity Simulations

Sensitivity simulations are performed to examine the effects of changes in model parameters and emissions. We conduct three sensitivity studies for the purpose of investigating the behavior of C₅-alkene triols compounds, examining phase partitioning dynamics and evaluating the impact of ${\text{NO}}_x$ and ${\text{SO}}_x$ emission reductions. The details of all the cases are shown in Table 2.

Table 2:

Simulations conducted in this study.

Scenarios	Objectives	Parameters
Base case	Default CRACMM2 (with AMOREV1.2)	Default parameters
Case0	Updated CRACMM2 (Shown in Figure 1b)	Parameters used in Table 1
Case1	Sensitivities of IEPOX for C5 production	$H^{*}$ of IEPOX × 10
Case2		$H^{*}$ of IEPOX × 0.1
Case3	Sensitivities of $K_{i,j}$ for C5 production	$K_{i,j}$ of C5 × 10
Case4		$K_{i,j}$ of C5-alkene triols × 0.1
Case5	Sensitivities of $H^{*}$ for three tracers partitioning	$H^{*}$ of three tracers × 0.2
Case6		$H^{*}$ of those three tracers × 5
Case7	Isoprene-derived SOA changes to ${\text{NO}}_x$ reduction	${\text{NO}}_x$ reduction 25%
Case8		${\text{NO}}_x$ reduction 50%
Case9	Isoprene-derived SOA changes to ${\text{SO}}_x$ reduction	${\text{SO}}_x$ reduction 25%
Case10		${\text{SO}}_x$ reduction 50%
Case11	Drivers of organosulfates production	$K_{i,j}$ of organosulfates × 0.1
Case12		$K_{i,j}$ of organosulfates × 10

C₅-alkene triols formation sensitivity to $K_{i,j}$ and $H^{*}$

These sensitivity simulations address uncertainties in the rates of particle-phase reaction and the $H^{*}$ (equation 6). The model’s sensitivity to aerosol-phase diffusivity or gas-phase diffusivity (typically 10⁻⁵ m²/s under standard conditions) is minimal unless these values are significantly misestimated—by roughly a factor of 100¹⁹. Since C₅-alkene triols are newly introduced to the model, we performed a sensitivity simulation to test the effect of $H^{*}$ for IEPOX by multiplying factors of 10 and 0.1 to assess its impact on subsequent C₅-alkene triols formation (Case 1 and Case 2). The $H^{*}$ for MAE was not changed, as our focus is on C₅-alkene triols formation. The second sensitivity simulation retains the same $H^{*}$ from Case 0 but alters $K_{i,j}$ of C₅-alkene triols to investigate impacts on predictions of aerosol species. In this simulation, the $K_{i,j}$ values were multiplied by 10 and 0.1 (Case 3 and Case 4). This simulation influences the relative split between tetrols and organosulfates, thereby affecting particle-phase speciation. The sensitivity of parameters to organosulfates is shown in the Supporting Information ‘Driver of organosulfates‘ section.

Uncertainty of $H^{*}$ for tracer’s phase partitioning

The gas-aqueous phase partitioning equilibrium in the model is governed by $H^{*}$ according to equation 2–7. Estimates of $H^{*}$ for 2-MT, 2-MG, and C₅-alkene triols can vary over several orders of magnitude, depending on the methods used^34,35. Among these estimates, the SPARC online tool is considered to provide the most reliable values, predicting $H^{*}$ for 2-MT as 3.38 × 10¹⁰ M atm⁻¹, for 2-MG as 5.25 × 10⁸ M atm⁻¹, and for C₅-alkene triols as 8.14 × 10¹⁰ M atm⁻¹ at 298 K. These values are used in Case 0 in this study.³⁶ To evaluate the impact of the associated uncertainties, additional simulations were conducted in which $H^{*}$ was scaled by factors of 0.2 and 5 (Case 5 and Case 6). These two cases provide insights into the uncertainty of the fraction of these three tracer species in the gas and particle phase.

${\text{NO}}_x$ and ${\text{SO}}_x$ reduction

To assess the impact of emission reductions on isoprene-derived SOA and the formation of SOA through different pathways, two additional sets of simulations were conducted. These scenarios included a 25% and 50% reduction in ${\text{NO}}_x$ emissions (Case 7 and Case 8) and a 25% and 50% reduction in total ${\text{SO}}_x$ emissions (Case 9 and Case 10). The simulations were performed for June 2022, using the Case 0 model parameterization for heterogeneous uptake processes.

Results and discussion

Comparisons with Observations

The evaluation of isoprene (Figure S3a), PM_2.5 (Figure S3b), O₃ (Figure S3c) and total organic carbon (OC) (Figure S4) at the Zibo site (Figure S1) is presented in the Supporting Information ‘Model Performance Evaluation‘ section. The spatial distributions of PM_2.5, O₃, SO₂, and NO₂ are illustrated in Figure S5. Additionally, a comparison of the evaluation results between the updated CRACMM2 and the default CRACMM2 mechanisms for PM_2.5 (Figure S6) and O₃ (Figure S7) across 500 monitoring sites in five regions of China is provided.

The 2-MT, C₅-alkene triols, and 2-MG were collected every other hour in June 2022 at the Zibo site³⁷, and the data were paired in time with the model results. Model evaluation of 2-MT, 2-MG, and C₅-alkene triols at the Zibo site are shown in Figure 2a. The time series of the three organic molecular tracers are shown in Figure 2b–d. It can be seen that the model accurately captures the high values, with 2-MT and C₅-alkene triols peaking from the 16th to the 20th, and 2-MG exhibiting high values from the 22nd to the 25th. Additionally, the model also successfully captures the low values observed from the 4th to the 15th. The newly added C₅-alkene triols formation pathway achieves a pearson’s correlation coefficient (R) value of 0.71, with a normalized mean bias (NMB) value of −26%. An $R$ value of 0.4 and an NMB of −35% are achieved for 2-MT. Previously, the modeled 2-MT concentrations were several times higher than observations. After the model update, 2-MT concentrations align more closely with observations. After we added the formation of 2-MG pathway via high ${\text{NO}}_x$, the modeled concentration is lower than the observation, but the model was able to capture 2-MG concentrations relatively well, with a NMB of −18%.

Figure 2:

(a) Model evaluation of aerosol phase 2-MT, 2-MG and C₅-alkene triols. Error bars indicate the standard deviation of daily concentrations, filled dots represent the average tracer concentrations. Time series of (b) 2-MG, (c) C₅-alkene triols and (d) 2-MT during the modeling period.

With the updated CRACMM2, the isoprene-derived SOA concentration is 75.1 ng/m³, compared to 80.4 ng/m³ in the default CRACMM2. While the total amount remains relatively unchanged, the composition and formation pathways of isoprene-derived SOA have been further refined and made more scientifically robust. Additionally, the newly incorporated C₅-alkene triols formation pathway and partitioning of tracers help bridge the gap between observations and simulations, further improving the representation of isoprene oxidation and SOA formation mechanisms. Since C₅-alkene triols is newly added to the model, the uncertainty of precursor’s $H^{*}$ and $K_{i,j}$ of C₅-alkene to all the tracers formation (Figure S8) is analyzed in Supporting Information “Uncertainty Analysis of isoprene-derived SOA” part.

Spatial Distribution of Isoprene-Derived SOA

The model predicted isoprene-derived SOA chemical compositions over China (Figure 3a–e) and precursors (Figure 3f) are shown in Figure 3. The overall spatial distribution of all the isoprene-derived SOA is consistent except the IEPOX-derived organosulfates. 2-MT and C₅-alkene triols are major isoprene-SOA species in summer in China, especially in southern area. C₅-alkene concentrations are much higher than other tracers, which is consistent with field observations³⁷. Pseudo-first-order particle-phase rate constants (Figure S9a) range from 0.02 to 0.03 s⁻¹ in forested regions, such as northern and western China, where H⁺ concentrations (Figure S9c) are also elevated during the modeling period. This leads to higher isoprene aerosol concentrations (Figure 3) in these areas. However, the highest concentrations of isoprene-derived aerosols are observed over the Sichuan Basin (SCB), where both isoprene (Figure 3f) and IEPOX + MAE gas-phase SOA precursor (Figure S9f) concentrations are high. Our model results indicate substantial formation of IEPOX-derived organosulfates (Figure 3b) with average concentrations ranging from 25 to 150 ng m⁻³ (100 ng m⁻³ on average at the surface and 84 ng m⁻³ on average aloft in the troposphere as reported in a previous study¹⁹). In contrast, the MPAN-derived organosulfates are predicted to have concentrations of less than 1 ng m⁻³ in the Base case simulation (Figure 3e). The average concentration is 10 ng m⁻³ in Yorkville, Georgia in a previous study³⁸. The spatial distribution of MPAN-derived organosulfates is similar to that of sulfate (Figure S10c), with high values observed in the SCB region. However, it differs significantly from the spatial distribution of IEPOX-derived organosulfates. The “Drivers of Organosulfates formation” section in the Supporting Information discusses the formation mechanisms of organosulfates and it indicates that $K_{i,j}$ of organosulfates is the main driver of its concentration. Due to the limited duration of sampling campaigns, only a qualitative evaluation of the organosulfates concentrations is performed.

Figure 3:

Model mean concentrations for June 2022 of (a) 2-MT, (b) IEPOX-derived organosulfates, (c) C₅-alkene triols, (d) 2-MG, (e) MPAN-derived organosulfates, and (f) Isoprene for Case0.

The IEPOX-derived species, 2-MT and C₅-alkene triols, dominate primarily due to the relative abundance of aerosol water (Figure S9b) compared to other nucleophiles. The spatial distribution of C₅-alkene triols is similar to that of 2-MT, but their concentrations are significantly higher, which is also consistent with the observations³⁷. This may be attributed to the higher $K_{i,j}$ of C₅-alkene triols. The $\beta$ is 0.47 closely aligns with previous laboratory findings, where $\beta =0.4$ was used in the simplified GAMMA model³⁹. This value corresponds to the most concentrated bulk solution studied by Eddingsaas et al (0.1 M H₂SO₄ with 3.0 M Na₂SO₄)³⁹.

Aerosol formation from IEPOX generally exceeds that from MAE, primarily due to the greater availability of gas-phase precursors (Figure S10a and b), the sum of gas-phase IEPOX+MAE and the higher uptake efficiency (Figure S9d and e). While both IEPOX and MAE have identical particle-phase reaction rates, the key difference lies in their uptake coefficients, which can be mostly attributed to their $H^{*}$ values. The much lower $H^{*}$ for MAE (1.2 × 10⁵) compared to IEPOX (3.0 × 10⁷) results in slower dissolution of precursor epoxides that is likely lowering the uptake coefficient for MAE.Research findings^19,31,40,41 consistently indicate that IEPOX-derived 2-MT generally dominates over MPAN-derived 2-MG, particularly during the summer. The model reproduces this trend, with a higher abundance of 2-MT in southeastern China.

In the model, the dominance of 2-MT over 2-MG can be explained by the fate of RO₂, which is governed by competitive pathways. The gas-phase formation of the precursors IEPOX and MAE/HMML is determined by the reaction of HO· and the fate of RO₂. When ${\text{NO}}_x$ is high, the reaction of RO₂ with NO proceeds more rapidly than RO₂ with hydroperoxy radicals (HO₂·), favors the formation of MAE/HMML. Conversely, RO₂ reacts preferentially with HO₂·, leading to increased IEPOX formation. Differences in precursor production and their relative gas-phase abundances can also matter. IEPOX formation begins with isoprene reacting with HO·, followed by HO₂· to form hydroxy hydroperoxides, and NO to form $\beta$-hydroxy alkylnitrates. Subsequent reactions with HO· produce IEPOX with mass-based yields up to 44%. In contrast, MAE is formed through a pathway in which isoprene peroxy radicals react with NO to produce methacrolein, which then reacts to produce MPAN and eventually MAE⁷. However, carbon is lost through competing pathways. Only when the rate of RO₂ + NO exceeds that of RO₂ + HO₂· are the precursors to 2-MG expected to surpass IEPOX, a condition likely to occur in ${\text{NO}}_x$-rich areas (e.g., urban areas) or during winter²⁵.

Gas-Particle Partitioning

The gas-particle partitioning results at the Zibo observational site during June 2022 are shown in Figure 4. For 2-MT, 81.7% was present in the particle phase, which is consistent with previous studies³¹. In the case of 2-MG, the particle-to-gas phase ratio was 51.2%, exceeding the 31% reported by Zhang et al³¹, which can be attributed to the $H^{*}$ of 2-MG. For C₅-alkene triols, we refer to Frauenheim’s chamber study, which estimated approximately 50% partitioning between the particle and gas phases¹⁴. The uncertainties associated with $H^{*}$ for three tracers in model predictions at the Zibo site are illustrated in Figure 4. Reducing the $H^{*}$ of the SOA species by 80% (Case 5) results in a 30% decrease in the particle-phase fraction of 2-MT, although its total concentration shows only a slight increase, while for C₅-alkene triols, it reaches 60%. For 2-MG, the particle-phase concentration varies by approximately ±50% in response to fivefold changes in $H^{*}$. In addition, increasing the $H^{*}$ of SOA species by a factor of 5 (Case 6) results in an approximately 30% increase in the fraction of C₅-alkene triols partitioned into the particle phase, accompanied by a slight decrease in the total concentration of C₅-alkene triols. For 2-MT, the change is smaller (17.9%). Previous study found that 2-MT has higher sensitivity to $H^{*}$ than 2-MG since 2-MT is primarily lost through gas-phase reaction and has a shorter lifetime against reaction with HO^· than 2-MG³¹. Additional studies are needed to reduce this uncertainty.

Figure 4:

Gas-particle partitioning of (a) C₅-alkene triols, (b) 2-MT, and (c) 2-MG in Base0 and with $H^{*}$ scaled by a factor of 5 and 0.2.

SOA Formation Pathway

The observed and modeled scatter plot is shown in Figure S4a, which demonstrates that OC in default CRACMM2 is generally underpredicted with a mean bias (MB) of $1.2\mu {\text{g}}^{-3}$ without the heterogeneous uptake pathways for MAE + HMML and C₅-alkene triols added. The inclusion of aerosol contributions to those pathways improves the model’s prediction of total OC (Figure S4b). In terms of diurnal profiles, observations from the Southeastern Aerosol Research and Characterization study show a relatively flat profile for OC¹⁹, while earlier versions of CMAQ³⁶ tend to exhibit a pronounced nighttime peak driven by primary organic aerosols, monoterpene + nitrate radical aerosols, and additional influences from sesquiterpene-derived SOA. The new isoprene-derived SOA simulation in this study produces a diurnal profile with a nighttime peak (Figure 5a), which aligns more closely with earlier CMAQ versions³⁶ and is mainly driven by low volatility, semi-volatile, and intermediate volatility organic compounds (L/S/IVOC) SOA (44.4%). For monoterpene SOA, the contribution to total SOA levels is 13.5% (Figure 5a) which is lower than 24% found by Ying et al⁴², as the sesquiterpene SOA in CRACMM2 is allocated to L/S/IVOC SOA. The contributions of AGLY and ASOAT to SOA are approximately 10%, with ASOAT showing a similar contribution to previous aromatic SOA estimates (13.2%). For AGLY, previous studies have also indicated that SOA formation from the reactive uptake of glyoxal on particles is notable, with AGLY accounting for more than 20% of isoprene-derived SOA⁴². The isoprene derived SOA, which includes C₅-alkene triols and MAE + HMML pathways contributes to an overall SOA increase of 23.5%. The updated CRACMM2 leads to an overall SOA increase, as shown in the spatial distribution across China in Figure 5. SOA in eastern China is increased by $0-4\mu g/m^3$. This finding is further supported by the PM_2.5 evaluation with lower MB and NMB across five regions of China (Figure S6) by accounting for the enhanced SOA contribution from the updated mechanism, which indicates that the updated mechanism enhances isoprene-derived SOA. It highlights the importance of isoprene oxidation in atmospheric organic aerosol budgets.

Figure 5:

(a) Composition of SOA at the Zibo site as predicted by CMAQ, June 2022, as a function of local time. Aerosol includes secondary oxygenated L/S/IVOCs (L/S/IVOC SOA), a lumped category of nonvolatile SOA species which include SOA from HC3, HC5, SLOWROC, ROCIOXY and aromatics (ASOAT), SOA produced in clouds from glyoxal and methylglyoxal (AORGC), SOA from reactive uptake of glyoxal on particles (AGLY), monoterpene SOA (AHOM + AELHOM), semi-volatile organic peroxide reacting with HO· produced from aromatics (AOP3), and all isoprene IEPOX + MPAN SOA. The vertical dashed line marks 9 pm (local time). and (b) SOA between Base case - Case0.

Impacts of Emission Reductions on SOA Mitigation

${\text{NO}}_x$ reduction

For a 25% reduction in ${\text{NO}}_x$ (Case7), the changes in 2-MG (Figure 6a) are minimal, remaining below 1 ng/m³. IEPOX-derived aerosol species increase (Figure 6b, c, d) as the isoprene RO₂· fate is shifted away from the NO pathway to reaction with HO₂·. HO₂· (Figure S12b) is increased in BTH and YRD regions, HO· (Figure S12a) decreases across China following ${\text{NO}}_x$ reduction (Figure S12c). The overall SOA changes are shown in Figure S13a. For anthropogenic SOA (ASOA), the reduction in ${\text{NO}}_x$ leads to an increase in YRD, BTH, and SCB regions, as shown in Figure S13b. For biogenic SOA (BSOA), a general decreasing trend is observed, as illustrated in Figure S13c. Case7 results in a small increase in organosulfate (2.2%), followed by 2-MG (8%). At the Zibo site, 2-MT and C₅-alkene triols increase by approximately 20%. When the ${\text{NO}}_x$ reaches a 50% reduction, this change becomes nearly double as shown in Figure 10 for both particle and gas phase.

Figure 6:

Changes in (a) MPAN-derived 2-MG, (b) IEPOX-derived C₅ tetrols, (c) IEPOX-derived 2-MT, and (d) IEPOX-derived organic sulfate due to a 25% reduction in ${\text{NO}}_x$ emissions compared to Case0 simulation for June 2022.

Figure 10:

Predicted percent changes in concentrations of C₅-alkene triols, 2-MT, 2-MG at the Zibo site due to ${\text{NO}}_x$ and ${\text{SO}}_x$ reductions for June 2022: (a) gas phase and (b) particle phase. IEPOX-derived organosulfates are considered only in Particle phase.

We also investigated the impact of ${\text{NO}}_x$ reductions on the formation of different systems (Figure 7). The contribution of SOA formed in clouds from glyoxal and methylglyoxal (AORGC) is small (Figure 5), with minimal changes observed. SOA from reactive uptake of glyoxal on particles is also reduced (AGLY), which is consistent with the research of McNeill et al.³⁰ and Pye et al.⁴³. Similarly, SOA from aromatic, furan, as well as AHOM show a decline in concentration. SOA derived from AOP3 (primarily originated from aromatics and oxygenated L/S/IVOCs) and L/S/IVOC SOA (predominantly anthropogenic in origin) exhibits an increased trend in YRD, BTH, and SCB regions. Based on the changes in the four main isoprene-derived SOA components following 25% and 50% reductions (Case7) in ${\text{NO}}_x$, as shown in Figure S14, we concluded that the variations in IEPOX organosulfates and 2-MG are relatively minor. In contrast, the other two major isoprene-derived SOA components show a pronounced increase with decreasing ${\text{NO}}_x$ levels.

Figure 7:

Changes in (a) AGLY, (b) AHOM, (c) AORGC, (d) L/S/IVOC SOA, (e) ASOAT, and (f) AOP3, due to a 25% reduction in ${\text{NO}}_x$ emissions compared to Case0 for June 2022.

${\text{SO}}_x$ reduction

The majority of sulfate and nearly all SO₂ emissions in China are of anthropogenic origin. The formation of epoxide-derived aerosols, including the dominant isoprene aerosol species (such as tetrols), strongly depends on aerosol acidity and is therefore largely controlled by ${\text{SO}}_x$ emissions. Both IEPOX- and MPAN-derived aerosols are substantially reduced (Figure 8) due to lower sulfate concentrations (Figure S15a), which slows the rates of particle-phase reactions and reduces uptake efficiency. The conversion of epoxides to the aerosol phase is further hindered by a decrease in particle surface area¹⁹. In a sensitivity run with a 25% reduction in ${\text{SO}}_x$ emissions (Case9), the decrease in tetrols at the Zibo site (Figure 10) is close to the 35% reduction reported by Pye et al.¹⁹, but significantly lower than the 70% reduction observed by Budisulistiorini et al.⁴⁴. For IEPOX-derived organosulfates, in Budisulistiorini et al.’s study, a 25% reduction in ${\text{SO}}_x$ emissions resulted in a similar 70% decrease in both IEPOX-derived organosulfates and tetrols. our findings indicate that organosulfates decreased by 5%. These differences may be attributed to variations in aerosol loadings across different study periods (2006 in Pye et al., 2013 in Budisulistiorini et al., and 2022 in our study) as well as the lower sulfate reaction rate constant used in Budisulistiorini et al.’s work. Additionally, spatial variations in SOA reduction across the domain warrant further investigation. With a 50% reduction in ${\text{SO}}_x$ emissions, the SOA reduction rate is nearly twice as large as that observed with a 25% reduction, for both the gas and particle phases (Figure S17 and Figure 10).

Figure 8:

Changes in (a) MPAN-derived 2-MG, (b) IEPOX-derived C₅ tetrols, (c) IEPOX-derived 2-MT, and (d) IEPOX-derived organic sulfate due to a 25% reduction in ${\text{SO}}_x$ emissions compared to Case0 simulation for June 2022.

From Figure 9, AORGC and ASOAT show a minor increase due to the reduction in ${\text{SO}}_x$ while all other SOA pathways exhibit a decrease. This is primarily attributed to the reduction of HO, as shown in Figure S15b for 25% ${\text{SO}}_x$ reduction S15a and c. A 25% reduction in ${\text{SO}}_x$ results in a noticeable decrease in isoprene-derived SOA, as shown in Figure 8, with C₅-alkene triols showing the largest decrease, followed by 2-MT. The reductions in organosulfates and 2-MG are relatively smaller compared to those of C₅-alkene triols and 2-MT. Consistent with these findings, Figure S16 shows that both ASOA and BSOA decrease following ${\text{SO}}_x$ reduction. Based on the changes observed in the four primary isoprene-derived SOA pathways under 25% and 50% reductions in ${\text{SO}}_x$ (Case9 and Case10, respectively) (Figure S17), we concluded that ${\text{SO}}_x$ reduction has a stronger impact on isoprene-derived SOA formation than an equivalent ${\text{NO}}_x$ reduction. This finding is consistent with previous studies^19,44.

Figure 9:

Changes in (a) AGLY, (b) AHOM, (c) AORGC, (d) L/S/IVOC SOA, (e) ASOAT, and (f) AOP3, due to a 25% reduction in ${\text{SO}}_x$ emissions compared to Case0 for June 2022.