Introduction

Volatile organic compounds (VOCs) are emitted from a range of anthropogenic and biogenic emission sources. After being emitted into the atmosphere, they can undergo a series of oxidation reactions with atmospheric oxidants and then form secondary organic aerosol (SOA). By scattering solar radiation and seeding clouds, SOA affects the earth’s energy budget and climate1.

Laboratory studies have reported SOA formation of individual anthropogenic and biogenic VOCs (AVOCs and BVOCs)2,3,4. For toluene, one of the most typically emitted AVOCs, its SOA yield could reach nearly 30% under low NOx conditions but reduced to around 20% under high NOx conditions5. A recent mechanistic study showed that under relevant polluted conditions with NOx concentration of ~10 ppb, at least 25% of the toluene SOA is formed via the bicyclic, phenolic, and benzaldehyde pathways6. Regarding isoprene, which is the mostly emitted BVOC, the reported SOA yields broadly ranging from <1 to 28.6%7,8,9,10. Under low NOx concentration conditions, the SOA formation from isoprene is dominated by organic peroxy radical (RO2•) chemistry of isoprene hydroxy hydroperoxide (ISOPOOH)8,11, and the reactive uptake of isoprene epoxydiols (IEPOX) that are generated from the oxidation of ISOPOOH by hydroxyl radicals (OH•)12 and subsequent epoxide formation13. When under high NOx conditions, the SOA formation from isoprene is dominated by methacryloylperoxynitrate reaction9,14.

While historical experiments mostly focused on the SOA yields from a single VOC precursor, increasing attention has been paid to the SOA yields from binary or even ternary mixtures of VOCs nowadays. For example, enhanced SOA yields were found in the experiments using myrcene+m-xylene15, myrcene+m-xylene+α-pinene15, β-myrcene+d-limonene16, α-pinene+o-cresol7, and toluene+m-xylene+1,3,5-trimethylbenzene17 compared to the corresponding experiments using a single VOC precursor. However, when isoprene was present, decreased SOA yields were observed in experiments such as isoprene+α-pinene7,18, isoprene+o-cresol7, and isoprene+toluene17,19. For a specific VOC in the mixture, whether its SOA yield is enhanced or reduced depends on its identity and the other VOC(s) in the mixture. Chen et al. 17 found that compared to the experiments using toluene only, the SOA yield of toluene was reduced by 51.4% in the mixture of isoprene+toluene but increased by 83.9% when toluene was mixed with m-xylene and 1,3,5-trimethylbenzene. In addition, Takeuchi et al. 20 found a 50% enhancement in the α-pinene SOA yield but a 20% reduction in the limonene SOA yield in experiments using the mixture of α-pinene+limonene. All of this evidence suggests that SOA yields from VOC mixtures are not simply linearly additive but are instead nonlinear. Such nonlinear effects of VOC mixtures on the SOA formation are complex due to the competition for oxidants between VOCs17,18,19, and the formation of cross-products16,18,20,21.

The interaction of anthropogenic and biogenic emissions impacts the regional environment and climate. Knowledge regarding the SOA formation from the mixture of AVOC and BVOC is still unclear, and the contributions and interactions of individual VOC precursors are not well understood. Therefore, in this study, a series of chamber experiments were conducted using two commonly widespread AVOC and BVOC, toluene and isoprene. Additionally, we provided insights into the formation of OH•, hydroperoxyl radical (HO2•) and RO2•, with the aid of the rapid adaptive optimization model for atmospheric chemistry (ROMAC) developed by Li et al.22 together with the master chemical mechanism (MCM). By studying the SOA yield and gas- and particle-phase composition, we aimed to provide mechanistic insights into how SOA from isoprene and toluene oxidation evolved during the different stages of SOA formation.

Results

SOA yields of isoprene and toluene

Figure 1 shows the wall-loss corrected SOA yields from this study (with stroke color) and the literature (without stroke color). In this study, the SOA yields were estimated using the data at the periods when the top 10th percentile of the produced SOA mass concentrations were observed on the scanning electrical mobility spectrometer (SEMS). We only included the literature data from which the studies were carried out using seed particles, initial NOx concentrations of ≤10 ppb, and similar initial concentrations of VOC precursors as in our study. Our experiments and the selected literature studies were summarized in Table S1 and Table S2, respectively.

Fig. 1: The secondary organic aerosol (SOA) yields of isoprene, toluene and isoprene-toluene mixture.
figure 1

Wall-loss corrected SOA yields as a function of the SOA mass concentrations for isoprene (ISO, a), toluene (TOL, b), and isoprene+toluene mixture (MIX, c) in the presence of acidic (orange) and neutral (blue) seed particles. Markers represent data from this study (circles with stroke) and literature (others without stroke), respectively.

Full size image

When neutral seed particles (i.e., ammonium sulfate, AS; in blue) were used, the isoprene SOA yields observed in our study were 6.3–15.3%, which was higher than those observed by King et al. 10 and Surratt et al. 9, but close to that reported by Liu et al. 8 (Fig. 1a). Note that the isoprene SOA yield reported by King et al. 10 represented the lower limit as no particle wall-loss corrections were applied.

For the experiments using neutral seed particles, the difference in the isoprene SOA yields could be attributed to variations in the amount of H2O2 in use and/or seed particle surface area. In our study, ~26.9 ppm of H2O2 was used, higher than those used in Liu et al. 8 (i.e., 5–15 ppm) and in Surratt et al. 9 (i.e., 4 ppm), but close to that used in King et al. 10 (i.e., 22 ppm). When larger amounts of H2O2 were used for OH• production under photolysis, the HO2-pathway for isoprene photooxidation became more dominant under low-NOx conditions. Consequently, the ISOPOOH-RO2 chemical pathway was favored, and larger isoprene SOA yields were formed8,11,23. In addition, our study was conducted with a seed particle surface area of 1.0–1.5 × 108 nm2 cm–3, which was several times higher than those used in the literature. I.e., 0.31–0.63 × 108 nm2 cm–3 in King et al. 10 and 0.14–0.23 × 108 nm2 cm−3 in Liu et al.8. In the presence of higher seed particle surface areas, oxidized vapors would favor condensing onto seed particles rather than chamber walls.

Regarding the experiments using acidic seed particles (i.e., ammonium bisulfate, ABS; in red), the isoprene SOA yield of this study was 12.4%, lower than that of Surratt et al. (28.6%)9. Previous studies have observed that under low NOx conditions, the isoprene SOA formation was linearly enhanced with increasing seed particle acidity9,24. When seed particles with higher acidity was used, the resulting increases in the reactive uptake of IEPOX, followed by ring-opening and nucleophiles addition, contributed to higher isoprene SOA mass. Different from ABS used in our study, the acidic seed particles used in Surratt et al. 9 were particles containing magnesium sulfate plus sulfuric acid, with higher acidity. Therefore, compared with Surratt et al. 9, the lower isoprene SOA yield in our study was likely attributed to the weaker IEPOX reactive uptake when seed particles with less acidity were used.

As shown in Fig. 1b, the toluene SOA yields in our study were 11.2–14.5% and 16.8–24.5% in the presence of neutral and acidic seed particles, respectively. All the values of SOA yields observed in our study were lower than those in Ng et al.5. The difference in the toluene SOA yields between studies could be primarily attributed to the different seed particle surface areas. A higher seed particle surface area was shown to enhance the toluene SOA formation25. In Ng et al.5, the estimated seed particle surface area was ~2.4 × 108 nm2 cm–3 based on the geometric mean diameter (GMD) of 50 nm, approximately two times higher than that in our study (i.e., 1.0–1.5 × 108 nm2 cm–3).

Compared with the experiments using neutral seed particles, higher toluene SOA yields were observed in those using acidic seed particles in our study. Cao et al. 26 found that increased proton concentrations, which was indicative of higher seed particle acidity, can lead to higher toluene SOA yields by catalyzing oligomerization (e.g., aldol condensation or hemiacetal) in the particle-phase27,28. However, the acid-catalyzed effect on toluene SOA formation was negligible in Ng et al.5, in which a mixture of AS and sulfuric acid was used as the acidic seed particles. This could be because at low RH ( ~5%), the sulfate particles were very likely present in a solid state when the neutralization extent by ammonia was above 0.729, and the corresponding proton concentration was estimated to be near zero30.

SOA yields of the isoprene-toluene mixture

In the mixture experiments using neutral seed particles, the SOA yields were 11.1–12.9% (Fig. 1c), in between the isoprene and toluene SOA yields (Fig. 1a, b). When acidic seed particles were used, the SOA yields of the mixture were 11.4–12.3%, similar to those in the presence of neutral seed particles. This indicated the minor role of seed particle acidity on the SOA formation from the mixture of isoprene and toluene.

In the mixture experiments, the decays of the two VOC precursors were similar to those in the isoprene-only and toluene-only experiments (Fig. S1). Additionally, there is no clear difference between the observed and expected ratios of changes in isoprene and toluene concentrations, as shown in Fig. S2. These indicate that the decays of toluene and isoprene were not likely to affect each other. As OH• was produced from the photolysis of H2O2 and sufficient for oxidizing VOCs, the scavenger of OH• due to isoprene was expected to be a minor sink19, which agrees with the very similar OH• concentrations in the model simulations between different experiments (Fig. S3).

Based on the consumption of individual VOCs and the corresponding produced SOA mass concentrations in the isoprene-only and toluene-only experiments, we further estimated the linearly additive SOA mass concentrations for the mixture experiments, under the assumption of no cross interactions between isoprene and toluene oxidation products. In the presence of neutral and acidic seed, the measured SOA mass concentrations were 21 and 28% lower than the predicted ones from linear addition, respectively (Fig. S4). These nonlinear effects of VOC mixtures on the SOA formation possibly implied the change in SOA composition.

SOA composition resolved by positive matrix factorization

Positive matrix factorization (PMF) was used to deconvolute the aerosol mass spectrometer (AMS) data into factors for analysis. Based on the Q/Qexp and residuals, a six-factor solution was chosen, and five SOA factors and one background factor were identified (Figs. S5–S7). The mass spectra of different SOA PMF factors are shown in Fig. 2, with the corresponding mass concentrations which have been normalized by sulfates (SO4) shown in Fig. S8. For each SOA factor, its oxygen-to-carbon ratio (O:C), hydrogen-to-oxygen ratio (H:C), and oxidation state (OSC) were estimated using the improved ambient parametrization31.

Fig. 2: The normalized mass spectrum of secondary organic aerosol (SOA) factors.
figure 2

Normalized factor mass spectra of the isoprene- (F-ISO-1, F-ISO-2), toluene- (F-TOL-1, F-TOL-2) and mixture-related (F-MIX) SOA factors with four different m/z ranges in colors (≤50: red; 50–100: cyan; 100–150: green, and >150: deep blue). For readability, the signals in the corresponding m/z ranges were scaled by one, five, fifty, and five hundred times.

Full size image

The five SOA factors were named as F-ISO-1, F-ISO-2, F-TOL-1, F-TOL-2, and F-MIX. F-ISO-1 and F-ISO-2 occurred mainly in the experiments containing isoprene (Fig. S8) and thus represented the oxidation products of isoprene. This was the same case for F-TOL-1 and F-TOL-2 when the experiments contained toluene. In addition to these four factors, a unique factor, F-MIX, was identified only in the experiments that contained both isoprene and toluene. F-TOL-2 and F-TOL-1 were the two most oxidized factors indicated by their O:C or OSC, followed by F-MIX and then F-ISO-1 and F-ISO-2. For F-ISO-1, m/z 53 and 82 constituted a high signal fraction at m/z > 50. They were the two suggested tracers of IEPOX-derived SOA12,32 which was favored to form in the presence of acidic seed particles during isoprene oxidation. Compared to the isoprene-only experiments using neutral seed particles, an enhancement of F-ISO-1 was observed in those using acidic seed particles (Fig. S8). Very likely, F-ISO-1 was associated with the SOA formation from reactive uptake of IEPOX. As for F-ISO-2, the signal contribution of m/z 82 at m/z >50 was negligible; however, a considerable fraction of m/z 91 was observed. The tracer m/z 91 has been identified in previous chamber studies of SOA derived from ISOPOOH-RO2 chemistry8,11. Thus, F-ISO-2 was associated with SOA formation through the ISOPOOH-RO2 chemical pathway.

F-TOL-1 and F-TOL-2 were dominated by m/z 28, 44, and 55, similar to the toluene SOA mass spectra in Chhabra et al. 33 and Liu et al. 34 in which SOA was formed via the photooxidation of toluene under low NOx conditions. Compared to the isoprene SOA factors, F-TOL-1 and F-TOL-2 had higher signal fractions at m/z >100. Given the hard ionization in the AMS, the m/z fragments at m/z >100 were likely to stem from the high-molecular-weight organics, which were potentially formed from oligomerization35,36. F-TOL-1 was less oxidized, and its formation rate slowed over time, whereas F-TOL-2 was more oxidized, and its formation rate increased with time (Figs. 3 and S9). Hence, F-TOL-1 and F-TOL-2 represented early- and later-generation products in toluene SOA, respectively. Based on this evidence, it is plausible that compounds grouped into F-TOL-2 were formed via the gas-phase oxidation and/or oligomerization in the particle-phase from those compounds grouped into F-TOL-1.

Fig. 3: The time series of chamber data in isoprene-only and toluene-only experiments.
figure 3

The time series of the volatile organic compound (VOC) decay (a), secondary organic aerosol (SOA) mass concentration (b), geometric mean diameter (GMD) (c), SOA factors (d), and their ratios (e) in the isoprene- (ISO, left) and toluene-only (TOL, right) experiments using ammonium sulfate (AS). Note that the SOA factors were normalized by sulfate.

Full size image

For F-MIX, most of the total signals were attributed to m/z 28, 29, 43, and 44. A considerable signal was observed at m/z >100, again indicating the possible formation of high-molecular-weight organics via oligomerization. Compared with the four other SOA factors, there were no distinct m/z fragments in the mass spectra of F-MIX. In addition, its f43, f44, and elemental ratio were between those of the isoprene and toluene factors. The mass spectra of F-MIX seemingly represented the combination of isoprene- and toluene-related factors.

Possible mechanism of isoprene and toluene SOA formation

In all isoprene-only experiments (Figs. 3 and S9), almost all the isoprene was consumed after 2 h of photooxidation, and the isoprene SOA formation stopped, followed by decreases in the isoprene SOA mass concentrations and GMD. In this case, the isoprene SOA was mainly formed via the gas-particle partitioning of oxidized vapors with sufficiently low volatilities4. The observed decreases in the SOA mass concentration and GMD were plausibly due to the loss of particle-phase constituents via fragmentation reactions followed by volatilization37.

The ratio between f91 and f82 (f91:f82) has been used to indicate the relative importance of ISOPOOH-RO2 chemical pathway and reactive uptake of IEPOX during isoprene SOA formation11. In the ISOPOOH oxidation experiments, Riva et al. 11 found that the estimated f91:f82 were 2–4 and ~0.7 in the presence of neutral and acidic seed particles, respectively. Similarly, we found that the f91:f82 were ~3.0 and ~0.45 in the experiments using neutral and acidic seed particles, respectively. In both studies, the low f91:f82 in the experiments using acidic seed particles was attributed to the enhanced reactive uptake of IEPOX. Note that in our study, such an enhancement in the IEPOX pathway was also evidenced by the relative increase in F-ISO-1 per sulfate mass in the experiments using acidic seed particles, compared to those using neutral seed particles (Figs. S8, S9). Regardless of the acidity of the seed particles used in the experiments, the mass concentration of the compounds attributed to F-ISO-2 was several times larger than that of F-ISO-1. Under the condition with sufficient amounts of OH• from the photolysis of H2O2 and low NO concentration (<1 ppb), the reaction regime was dominated by HO2•. This favored the formation of compounds of low volatility via the ISOPOOH-RO2 chemistry pathway8, which were plausibly grouped into F-ISO-2.

When neutral seed particles were used, the IEPOX formed from the oxidation of ISOPOOH might not effectively partition from the gas-phase to the particle-phase via reactive uptake. This was possibly due to the low particle acidity9 and newly formed organic coating on the particle surface38,39. Instead, the IEPOX was very likely to quickly react with OH• in the gas and particle phases40. When acidic seed particles were used, the high particle acidity could promote the reactive uptake of IEPOX, contributing to the isoprene SOA formation. Overall, during the 4-h photooxidation timescale, F-ISO-2, which very likely represents ISOPOOH-RO2 chemistry pathway, still dominated the isoprene SOA formation (Figs. 3d and S9d), with an average contribution of 94.8% and 76.4% in mass in the experiments using neutral and acidic seed particles, respectively.

The similarities of gas-phase compounds measured by the iodide chemical ionization mass spectrometer (CIMS) with the SOA factors were calculated41. The better the similarity, the smaller θ, with five different levels of similarities as excellent (0–5), good (5–10), many (10–15), limited (15–30), and poor ( >30). Here, we estimated the saturation vapor pressure of the gas-phase compounds measured by the CIMS according to Graham et al. 42 and then classified the compounds into different volatility classes according to Donahue et al. 43. Only those compounds grouped into the classes of intermediate-volatility and semi-VOCs were chosen for the analysis, as their condensation was governed by the gas-particle partitioning equilibrium in the experiments44. As shown in Fig. S10a, in the presence of neutral seed particles, many gas-phase compounds had good or many similarities with F-ISO-1. Nearly all compounds with OSC <1 had excellent and good similarities with F-ISO-2, including C4H6O3, C4H8O3, C5H8O4, and C5H12O5, which were formed via the non-IEPOX pathway8,11,13,23. In the experiment using acidic seed particles, most of the gas-phase compounds have better similarities with F-ISO-1 compared to those using neutral seed particles. Likely, in the presence of acidic seed particles, these compounds could partition into the particle phase in a similar way as IEPOX8. Furthermore, almost identical similarities were observed between the gas-phase compounds and F-ISO-2 in the presence of the two different seed particles.

Unlike isoprene SOA, the mass concentration of toluene SOA continuously increased with the reaction time. This increasing trend was also observed in the two toluene-related SOA factors (i.e., F-TOL-1 andF-TOL-2) (Figs. 3 and S9). In the toluene-only experiments, F-TOL-1 and F-TOL-2 dominated the toluene SOA formation in the reaction timescale of before and after 2 h (Fig. 3e), respectively.

Regardless of experiments using neutral and acidic seed particles, nearly all gas-phase compounds had high degrees of similarity with F-TOL-1 (Fig. S10b). However, only a few compounds, especially those with OSC of 1 or higher, showed many or good similarities with F-TOL-2. As indicated by the OSC shown in Fig. 2, F-TOL-2 was more oxidized than F-TOL-1, and its formation was likely associated with the highly oxidized compounds. Considering the weak sensitivity of iodide CIMS towards the highly oxidized compounds45, it might be difficult to detect most of them in our study.

In comparison to experiments using neutral seed particles, the toluene SOA yields were enhanced in the presence of acidic seed particles, which was due to the acid-catalyzed reactions in the particle phase. However, the time evolution in the ratio of F-TOL-1 and F-TOL-2 (Figs. 3 and S9) was very similar between experiments using neutral and acidic seed particles. This suggests that for each toluene factor, seed particle acidity played a similar role in the associated particle-phase reactions which contributed to the formation of the factor. Given that the AMS operated with a hard ionization scheme, it is difficult to directly identify compounds belonging to each toluene factor according to the AMS ion fragments. However, it is still possible to speculate the groups for compounds that contributed F-TOL-1and F-TOL-2 by comparing the elemental ratios between these factors and those compounds reported in literature46,47,48,49,50,51,52,53. As shown in Fig. S11, F-TOL-1 likely contained the C4–C5 carbonyls and carboxylics with ring-containing compounds, while F-TOL-2 possibly consisted of the mixture of C2–C5 oxo-carboxylics and ring-containing compounds.

Possible mechanism of mixture SOA formation

In the mixture experiments, the time series of the total SOA mass concentration was different from that in isoprene- and toluene-only experiments. The peak total SOA concentration occurred later than in the isoprene-only experiments, and the subsequent decrease was also more gradual (Fig. 4b). It could be found that the evolution of total SOA was governed by the formation of isoprene-related SOA within the first 2 h and then by the formation of toluene- and mixture-related SOA for the remaining experiments (Fig. S12). The mixture-related SOA factor (F-MIX) reached its maximum mass concentration at around 3 h, which accounted for 30% of the total SOA mass. As reported in previous studies on SOA formation from VOC mixtures, gas- and particle-phase cross-products have been directly observed in other experiments using VOC mixtures, including isoprene+α-pinene18,21, limonene+α-pinene20, toluene+xylene17, β-myrcene+d-limonene16, and tetramethylethylene+1,3,5-trimethylbenzene54. Although direct measurement of cross-products was not available in our study, there is some plausible evidence to indirectly support the hypothesis that F-MIX arose from the cross-reactions between oxidation products from isoprene and toluene.

Fig. 4: The time series of chamber data in isoprene-toluene mixture experiment.
figure 4

The time series of the volatile organic compound (VOC) decay (a), secondary organic aerosol (SOA) mass concentration (b), geometric mean diameter (GMD) (c), SOA factors (d), and their ratios (e, f) in the mixture experiments (green or blue) using ammonium sulfate (AS). The grey points in (e) and (f) indicate the SOA factors ratios in single VOC precursor experiments. Note that the SOA factors were normalized by sulfate.

Full size image

A handful of gas-phase compounds from the oxidation of isoprene and toluene had many and even good similarities with F-MIX (Figs. S13 and S14). This suggests the formation of F-MIX was associated with both isoprene- and toluene-oxidation products. In the mixture experiments, the cross reactions between RO2• from isoprene- (RisoO2•) and toluene (RtolO2•) oxidation could potentially happen in the gas phase. The cross-reaction rate constant between RisoO2• and RtolO2• was estimated to be 18.3 × 10–12 cm3 molecule–1 s–1, using the eq. 20 in Jenkin et al. 55. Based on the RO2• concentrations from the model simulation (Table S3), we found that the estimated reaction rate of the cross reaction between RtolO2• and RisoO2• was 6.7 × 106 molecules cm–3 s–1, which nearly was one or two orders of magnitude faster than the self-reaction rates of RtolO2• and RisoO2•. In addition, we can not exclude the possibility of cross-reactions in the particle phase. It is because isoprene-11,13 and toluene-46,47,50,53 oxidation products contain a great number of functional groups, including hydroxyl, carbonyl, and oxocarbonyl, and they are available for the cross-reactions in the particle phase20.

Regarding the mass spectra of F-MIX, it did not contain any distinct m/z markers but resembled the combinations of mass spectra from isoprene- and toluene-related SOA factors. It indicated that the monomers that contributed to the formation of F-MIX were the same as those in isoprene- and toluene-only experiments, and therefore the ion fragments became similar20. Furthermore, in experiments using neutral seed particles, we observed increases in the ratios of F-ISO-1/F-ISO-2 and F-TOL-1/F-TOL-2 (Fig. 4) and weaker similarities of the gas-phase compounds with F-ISO-2 and F-TOL-2 (Figs. S13 and S14), compared to single VOC experiments. It could be found that in the mixture experiments, the SOA yields of F-ISO-2 and F-TOL-2 were lower than those in the single VOC experiments (Fig. S15). Therefore, the changes in the factor ratios and similarities were driven by the suppression of the formation of F-ISO-2 and F-TOL-2. Correspondingly, these oxidation intermediates which were contributed to F-ISO-2 and F-TOL-2 formation interacted with each other and contributed to forming F-MIX. Interestingly, similar factor ratios for F-ISO-1/F-ISO-2 and F-TOL-1/F-TOL-2 and signal fractions of F-MIX were observed between the mixture experiments using neutral and acidic seed particles (Figs. 4 and S16). This suggests the minor role of particle acidity in the SOA formation from the mixture of isoprene and toluene. It was because the intermediates of isoprene- and toluene-oxidation were involved in the cross-reactions to form F-MIX, fewer of them were available for the acid-catalyzed reactions compared to the experiments using single VOC precursors. Overall, all the evidence mentioned above indicates that the formation of F-MIX was attributed to the cross-reactions between the intermediates of isoprene- and toluene-oxidation, and those intermediates originally possibly contributed to the formation of F-ISO-2 and F-TOL-2.

Discussion

With the aid of chamber photooxidation experiments, our study reported the SOA formation from the mixture of isoprene and toluene. For the first time, our study provided compositional insights into how individual SOA factors contributed to total SOA formation and how they evolved with reaction time. In the mixture experiments, two isoprene-related factors, F-ISO-1 and F-ISO-2, dominated the total SOA formation during the early stage, contributing ~45% in mass. However, in the later stage of SOA formation, two toluene-related factors, F-TOL-1 and F-TOL-2, became dominant, with an average contribution of ~25% in mass. Apart from these four factors, we identified a unique factor, F-MIX, which only occurred in the mixture experiments and represented the cross-reaction between the products from isoprene and toluene oxidation. With a ~30% contribution to the total SOA mass, the presence of F-MIX highlighted the significant role of cross-product formation during SOA formation from the studied VOC mixture. Compared to the predicted total SOA mass concentrations from linear addition, the measured ones reduced by 21 and 28% in the presence of neutral and acidic seed particles, respectively. The formation of F-MIX led to the nonlinear relationship of SOA formation between single VOC precursors and mixture systems.

Beyond the change in chemical composition17,19,56,57, the interactions between AVOCs and BVOCs also altered the particle volatility7 and phase behaviors of SOA particles58. For example, Voliotis et al. 7 found that the volatility of SOA particles in α-pinene+o-cresol was between those measured in single precursor experiments. Mahrt et al. 58 showed that when different SOA types were mixed together, liquid-liquid phase separation would be observed if the difference in the average O:C ratio of SOA types was larger than 0.47. These findings are likely also applicable to our study, considering the formation of F-MIX and the difference in the O:C ratios between isoprene-related and toluene-related factors.

Based on our study and previous literature7,16,19,20, we believe that the complex interaction between VOCs and their oxidation products widely exists in the real atmosphere. Previous studies have reported the coexistence of AVOCs and BVOCs in the urban or vicinity of urban59,60,61, or in the forest and remote area62,63,64,65. Hence, although the presence of anthropogenic emissions can enhance the formation of biogenic SOA66,67,68, our findings demonstrate that the complex interplay between BVOCs and AVOCs can also lead to a reduction in the total SOA mass formation. Thus, when studying SOA formation in a real atmosphere, these two opposite effects must be carefully evaluated to ensure an accurate understanding of the processes involved. This will be crucial for advancing our understanding of environmental chemistry, improving atmospheric models, and informing policy development.

Methods

Chamber experiments

In this study, an 8 m3 Teflon chamber was used for photooxidation experiments. The chamber was placed inside a temperature-controlled room at 25 °C. The \({J}_{{{NO}}_{2}}\) value is 0.18 min–1, under irradiation of 10 UV-B lamps mounted on two sides. Details of the smog chamber have been described previously69.

In each experiment, H2O2 was introduced to the chamber by passing zero air through a gently heated glass flask with the 500 μl H2O2 solution (50%, Sigma-Aldrich). Assuming no loss during injection, the concentration was estimated to ~26.9 ppm based on the amount of H2O2 and the chamber volume. Afterward, toluene (99%, Sigma-Aldrich), isoprene (99%, Sigma-Aldrich), or their mixture was injected into the chamber. In addition, d9-butanol (98% Cambridge Isotope Laboratories Inc.) was introduced into the chamber and used as a tracer of OH•. Seed particles were generated by atomizing an aqueous solution of ammonium sulfate (AS) or ammonium bisulfate (ABS) of 10 g L–1 using a collision nebulizer (CH Technologies) and then passing through a diffusion dryer. Afterward, the dried AS or ABS particles were introduced into the chamber until an initial well-mixed seed particle number concentration reached 5000–6000 cm–3 with a geometric mean diameter (GMD) of 80–90 nm. Each photooxidation experiment lasted for 4 h with all UV lights on. Details about each experiment are summarized in Table S1.

After each experiment, the chamber was flushed with humidified zero air at a relative humidity of 40% overnight. Following this cleaning procedure, the particle number concentration could be reduced to ≤50 cm–3, as measured by a condensation particle counter (CPC, A20, Airmodus). The NO concentration <1 ppb, as measured by a NOx analyzer (Ecotech, ML9841, Australia).

Instruments

A suite of real-time instruments was used to measure gas vapors and aerosol particles. The concentrations of isoprene, toluene, and d9-butanol were monitored using a selected ion flow tube mass spectrometer (SIFT-MS, Voice200 ultra, Syft Technologies Ltd.)70. Additionally, the oxidized gas vapors were measured using a high-resolution time-of-flight chemical ionization mass spectrometer (HR-ToF-CIMS, Aerodyne Research Inc.) with iodide (I−) as the regent71. The dried particle size distribution (5–500 nm) was monitored with a scanning electrical mobility spectrometer (SEMS 2100, Brechtel), and the aerosol bulk composition was analyzed with a compact time-of-flight aerosol mass spectrometer (c-ToF-AMS, Aerodyne Research Inc.)72.

The particle wall loss correction was applied to the particle measurement data from SEMS44. In the wall loss experiments, we injected either AS or ABS into the chamber by passing through a silica gel diffusion drying tube, with the seed number concentration between 4000 and 6000 cm–3. Based on the first-order decay of the particle number concentrations within 6 h under dark conditions, we obtained the particle wall loss rates of AS and ABS seed particles for different size bins. The loss of aerosol particles in the size bin j was calculated using Eq. 1:

$${P}_{{loss},\,\Delta t,\,j}={P}_{t,j}\times \exp ({k}_{{wall},j}\times \Delta t)$$
(1)

where \({P}_{{loss},\,\Delta t,j}\) is the loss of aerosol particles in the size bin j within time \(\triangle t\). \({P}_{t,j}\) is the suspended particle number concentration in the size bin \(j\) at time t, \({k}_{{wall},j}\) is the particle wall loss rate of size j. Note that the particle wall loss rates of AS and ABS seed particles were only applied for the experiments that used AS or ABS as seed particles, respectively.

Model simulation

We used the ROMAC and the latest version of MCM (v3.3.1; http://mcm.leeds.ac.uk/MCM/) to study the formation of OH•, HO2•, and RO2•. ROMAC is a 0-D box model simulating atmospheric chemical kinetics problems, which involves >17,000 chemical reactions and around 5900 species. Detailed descriptions of ROMAC can be found at Li et al.22. Here, the input temperature, relative humidity, and pressure were set to 298 K, 40%, and 101.3 kPa, respectively. In addition, the initial concentration of H2O2 and NOx, as well as the observed concentrations of toluene and isoprene, were used to constrain the model runs. Table S3 summarizes the averaged concentrations of radicals (i.e., OH•, HO2•, and RO2•), chosen reaction rate constants between radicals, and estimated reaction rates.