Contrasting impacts of humidity on the ozonolysis of monoterpenes: insights into the multi-generation chemical mechanism

Secondary organic aerosol (SOA) formed from the ozonolysis of biogenic monoterpenes is a major source of atmospheric organic aerosol. It has been previously found that relative humidity (RH) can influence the SOA formation from some monoterpenes, yet most studies only observed the increase or decrease in SOA yield without further explanations of molecular-level mechanisms. In this study, we chose two structurally different monoterpenes (limonene with an endocyclic double bond and an exocyclic double bond, Δ³-carene with only an endocyclic double bond) to investigate the effect of RH in a set of oxidation flow reactor experiments. We find contrasting impacts of RH on the SOA formation: limonene SOA yield increases by ∼100 % as RH increases, while there is a slight decrease in Δ³-carene SOA yield. Although the complex processes in the particle phase may play a role, we primarily attribute the results to the water-influenced reactions after ozone attack on the exocyclic double bond of limonene, which leads to the increment of lower volatile organic compounds under high-RH conditions. However, as Δ³-carene only has an endocyclic double bond, it cannot undergo such reactions. This hypothesis is further supported by the SOA yield enhancement of β-caryophyllene, a sesquiterpene that also has an exocyclic double bond. These results greatly improve our understanding of how water vapor influences the ozonolysis of biogenic organic compounds and subsequent SOA formation processes.

1 Introduction

Secondary organic aerosol (SOA), as an important type of ambient fine particulate matter (PM_2.5: aerosols with aerodynamic diameter ≤ 2.5 µm) (Guo et al., 2014; Huang et al., 2014), has caused a series of negative impacts on human health (Pye et al., 2021), air quality (Zhang et al., 2016) and global climate (Levy II et al., 2013). SOA produced from the oxidation of biogenic volatile organic compounds (BVOCs) is a major component of SOA in heavy forest regions during summer (Sindelarova et al., 2014; Ahmadov et al., 2012) and contributes a large fraction (∼40 %–80 %) of the global OA budget (Cholakian et al., 2019).

Monoterpenes, mostly emitted from coniferous trees, account for ∼11 % of total BVOCs (Sindelarova et al., 2014; Kanakidou et al., 2005). Limonene is one of the most abundant monoterpenes, with an annual emission budget of 11.4 Tg yr⁻¹ (Guenther et al., 2012). Apart from the biogenic source, limonene can also be released from indoor emission, mainly from essential oils (Ravichandran et al., 2018; de Matos et al., 2019; Mot et al., 2022). Limonene has an endocyclic double bond and an exocyclic double bond and is thus more reactive than other monoterpenes towards oxidants such as ozone (O₃), the hydroxyl radical (OH), and the nitrate radical (NO₃) (Chen and Hopke, 2010; Atkinson and Arey, 2003). Δ³-carene is another kind of monoterpene that dominates the monoterpene emission from Scots pine (Bäck et al., 2012). Different from limonene, Δ³-carene contains only one endocyclic double bond, which is similar to most other monoterpenes.

Ozonolysis is an important reaction pathway for limonene and Δ³-carene. Although reactions with OH and NO₃ are faster than that with O₃ for both of the monoterpenes (Atkinson, 1991; Khamaganov and Hites, 2001; Chen et al., 2015; Shaw et al., 2018), the atmospheric concentration of the latter monoterpene is much higher than that of the former (Sbai and Farida, 2019). The contributions of O₃ reactions with limonene and Δ³-carene to tropospheric degradation are 47 % and 24 %, respectively, in the daytime (Ziemann and Atkinson, 2012). In pristine areas where NO₃ concentration is very low, ozonolysis is also the dominant fate for limonene and Δ³-carene in the nighttime. In addition, it has been previously found that the ozonolysis of monoterpenes can produce more extremely low-volatility products than OH-initiated oxidation, which contributes a large fraction of the SOA production (Jokinen et al., 2015). For either limonene or Δ³-carene, the first step of ozonolysis is attacking the endocyclic double bond to form two types of stabilized Criegee intermediates (sCIs) with low energy (Fig. S1 in the Supplement) (Drozd and Donahue, 2011; Chen et al., 2019). The sCI will then trigger a series of chemical reactions, like isomerization, decomposition, and addition reactions. Correspondingly, the major components of Δ³-carene SOA are caric acid, OH-caronic acid, and caronic acid (Ma et al., 2009; Thomsen et al., 2021), while the major components of limonene SOA are limonaldehyde, keto-limononaldehyde, limononic acid, and keto-limononic acid (Pathak et al., 2012; Wang and Wang, 2021).

Water is ubiquitous in the atmosphere and can affect the formation mechanism of SOA and its relevant physical and chemical properties (Sun et al., 2013). A number of field measurements have shown that the average molecular weight of the water/organic phase and activity coefficient of condensed organics would be changed due to the change in relative humidity (RH) (Seinfeld et al., 2001; Li et al., 2020). In addition, several laboratory studies have demonstrated that RH can influence the ozonolysis of monoterpenes in different ways. Most of those studies have reported either an inhibitory effect or a negligible effect of high RH on the particle formation (Bonn and Moortgat, 2002; Fick et al., 2002; Zhao et al., 2021; Ye et al., 2018). Nevertheless, a few other studies found that high RH can promote SOA formation from the ozonolysis of limonene (Yu et al., 2011; Gong et al., 2018; Xu et al., 2021), but the reason for this promotion effect remains unclear.

To fully examine the effects of water on SOA formation from the ozonolysis of monoterpenes, especially the related chemical processes, we used an oxidation flow reactor (OFR) to investigate the ozonolysis of limonene and Δ³-carene under different RH conditions in this study. An ultra-high-performance liquid chromatography with quadrupole time-of-flight mass spectrometry (UPLC-Q-TOF-MS) instrument was deployed to analyze the molecular chemical composition of the SOA, which provided insights into the physical and chemical processes influenced by the water content. With these state-of-the-art techniques, we proposed mechanisms that may explain the inhibitory or enhancing RH effects on SOA formation for different monoterpenes.

2 Experimental methods

2.1 Oxidation flow reactor experiments

A series of dark ozonolysis experiments of limonene and Δ³-carene were conducted in a custom-made oxidation flow reactor (OFR). The OFR is a 602 mm long stainless cylinder with a volume of 2.5 L (Fig. S2) (Liu et al., 2014, 2019). A zero-air generator (XHZ2000B, Xianhe, China) was used to generate dry clean air as the carrier gas for the OFR. As shown in Fig. S2, there are four gas paths upstream of the OFR: the first path is the precursor gas channel through which monoterpenes are injected via a syringe pump (ISPLab 01, DuKe, China), the second path is for the flow of 300 sccm dry zero air passing through a mercury lamp (λ=185 nm) to generate O₃, the third path is connected to a water bubbler to generate wet air, and the fourth path is the extra-dry zero air entering the OFR. The RH in the OFR was controlled by adjusting the ratio of the wet and dry zero air flows. A water recycle system was equipped to keep the temperature (T) at around 298 K. The total flow was 0.9 L min⁻¹, resulting in an average residence time of 167 s. The RH and T in the OFR were monitored by a $T/$RH Sensor (HM40, VAISALA, Finland). The concentration of ozone and the consumption of the precursor gas were measured with an ozone monitor (Model 106L, 2B Technologies, USA) and a gas chromatograph with a flame ionization detector (GC-FID 7890B, Agilent Technologies, USA), respectively. The GC was equipped with a DB-624 column (30 m × 0.32 mm, 1.8 µm film thickness), whose temperature was set to ramp up from 100 to 180 ^∘C at a rate of 20 ^∘C min⁻¹, and then held at 180 ^∘C for 2 min. Before each experiment, O₃ was introduced into the OFR to clean it until the background aerosol mass concentration reached <1 µg m⁻³.

The experimental conditions are shown in Table 1. In these OFR experiments, the precursor (limonene or Δ³-carene) concentration was set to ∼320–340 ppb. A high O₃ concentration of ∼6 ppm was used to realize an equivalent aging time of 0.41 d in the real atmosphere, assuming an average ambient O₃ concentration of 28 ppb (Sbai and Farida, 2019) (see Sect. S1 in the Supplement for the calculation). Under such conditions, most of the precursors were consumed, since the residence time was almost 5 and 3 times the half-life for limonene and Δ³-carene, respectively. Correspondingly, the O₃ consumption for limonene and Δ³-carene was ∼250 and ∼100 ppb, respectively. A series of RH conditions ranging from dry (1 %–2 %) to 60 % with a step of ∼10 % were used to investigate the effects of water content on SOA production and composition (see Table 1). All materials used in the experiments have been described in Sect. S2.

Table 1 Experimental conditions and results.

^a N_{(14.1–735 nm)} means the total particle number concentration from size 13.8 to 723.4 nm. ^b M_{(13.8–723.4 nm)} means the total particle mass concentration from size 13.8 to 723.4 nm. ^c D_(mean) means the particle mean diameter.

2.2 SOA particle analysis

2.2.1 SOA yield

The SOA particle size distribution was measured with a scanning mobility particle sizer (SMPS), which consists of a differential mobility analyzer (DMA) (model 3082, TSI Inc., USA) and a condensation particle counter (CPC) (model 3776, TSI Inc., USA). The samples were measured by SMPS every 5 min with a sampling flow and a sheath flow of 0.3 and 3 L min⁻¹, respectively. The SOA mass concentration was calculated from the volume concentration measured with SMPS and the aerosol density, which was estimated to be 1.25 cm⁻³ for limonene SOA and 1.09 g cm⁻³ for Δ³-carene SOA (Thomsen et al., 2021; Watne et al., 2017).

The SOA yield (Y) for individual organic gas can be calculated as

$$\begin{array}{}\text{(1)}& Y={\displaystyle \frac{\mathrm{\Delta }M}{\mathrm{\Delta }HC}},\end{array}$$

where ΔM is the total mass concentration of SOA, and ΔHC is the mass concentration of reacted precursor (Ng et al., 2007; Odum et al., 1996).

2.2.2 Ultra-high-performance liquid chromatography with quadrupole time-of-flight mass spectrometry analysis

An ultra-high-performance liquid chromatography system (UPLC, UltiMate 3000, Thermo Scientific) coupled with a quadrupole time-of-flight mass spectrometry instrument (Q-TOF-MS, Bruker Impact HD) was used to analyze the molecular-level chemical composition of SOA. First, the SOA particles were collected on the PTFE filters (47 mm diameter, 0.22 µm pore size, Jinteng, China). Next, these filters were dissolved and extracted by 5 mL methanol twice. Extracts were then filtered through PTFE syringe filters (0.22 µm pore size) and were concentrated to near dryness by nitrogen blowing. Finally, the samples were redissolved in a 200 µL solution with 0.1 % ($v/v$) formic acid in a 50:50 methanol $/$ ultrapure water mixture.

The parameters of the LC–MS instrument were set as follows: capillary voltage 4000 V, nebulizer pressure 0.4 bar, dry heater temperature 200 ^∘C, end plate voltage −500 V, and flow of dry gas 4 L min⁻¹. A C₁₈ column (100 Å, 3 mm particle size, 2.1 mm × 50 mm; Waters, USA) was used with a column temperature of 35 ^∘C. The mobile phase was 0.1 % formic acid in methanol (A) and 0.1 % formic acid in ultra-high-purity water (B) with a flow of 200 µL min⁻¹. The injection volume was 5 µL. The mass spectrometer was operated in negative ion mode, and the detection molecular weight range was from $m/z$ 50 to 1500. The temperature ramping program was 0–3 min with 0 %–3 % phase B, 3–25 min with 3 %–50 % phase B, 25–43 min with 50 %–90 % phase B, 43–48 min with 90 %–3 % phase B, and 48–60 min with 3 % phase B.

Figure 1 The effect of RH on the SOA formation: (a) number concentration, (b) mass concentration, (c) SOA yield, and (d) mean diameter.

3 Results and discussion

3.1 SOA production under different RH conditions

The SOA formation of a representative experiment is shown in Fig. S3. It is found that the formed SOAs are mainly in the size range of 60–200 nm, and the number concentration and mass concentration are relatively stable during the course of the OFR experiment. SOA formation from limonene and Δ³-carene in terms of particle number concentration, particle mass concentration, and SOA yield as a function of RH is illustrated in Fig. 1a–c. We find that all three of the above-mentioned parameters of limonene SOA increase with the increasing RH. The increment of particle mass concentration and SOA yield from the ozonolysis of limonene is ∼100 % higher at wet (60 % RH) than at dry conditions. In contrast, SOA formation from Δ³-carene is suppressed by ∼40 % under high RH. The distinct effects of RH on SOA formation from the ozonolysis of limonene and Δ³-carene found in this study agree with most previous studies (Yu et al., 2011; Jonsson et al., 2006; Bonn et al., 2002; Gong and Chen, 2021; X. Li et al., 2019). As shown in Table 2, Yu et al. (2011) reported a positive correlation between SOA production and RH for the ozonolysis of limonene in the chamber experiments without OH scavenger. Their experimental condition is similar to that in our study regarding the absence of OH scavenger, and, thus, similar results were observed. However, in the presence of OH scavenger, results are quite different. Jonsson et al. (2006) observed a similar enhancement effect of high RH on SOA production with 2-butanol to the OH scavenger, while Bonn et al. (2002) found a negligible or suppressive effect with cyclohexane as the OH scavenger. It should be noted that the OH scavenger not only has the ability to scavenge OH but also produces additional products which may influence the reactions of target precursors. For example, there is no difference between 2-butanol and cyclohexane in the scavenging ability of the OH radical, though 2-butanol will produce more HO₂• than cyclohexane, and, consequently, R• will react with HO₂• to produce more hydroxyl acids and hydroxyl per-acid products, most of which have low volatility, and thus they partition into the particle phase. According to previous studies, the influence of different OH scavengers can vary (Jonsson et al., 2008). This may explain the different findings with and without OH scavenger for limonene SOA. With regard to Δ³-carene, similar results are found in the absence of OH scavenger, namely, high RH has negligible or slightly suppressive effect on SOA production (Bonn et al., 2002; Fick et al., 2002). The same as limonene, the presence of OH scavenger and its different chemical nature can explain the different results found previously (Jonsson et al., 2006; Bonn et al., 2002).

Table 2 Comparison with previous studies on the effect of RH.

^a M means the change trend total particle mass concentration. ^b N means total particle number concentration. ^c N.M. means not mentioned. ^d Positive sign (+) means the mass or number concentration increases with RH. ^e Negative sign (−) means the mass or number concentration decreases with RH.

Figure 2 UPLC/(−) ESI-Q-TOF-MS mass spectra of SOA from limonene ozonolysis. (a) MS under high- and low-RH conditions and (b) the identification of monomers under high-RH conditions.

The enhancement in limonene-SOA production under high RH can be for several reasons from either physical or chemical processes. First, the hygroscopic growth of the particles (i.e., absorption of water content) can lead to higher mass concentration under higher RH, but the enhancement should be at most ∼30 % as the growth factor (GF; the ratio of wet and dry diameter: $D_{\mathrm{wet}}/D_{\mathrm{dry}}$) of limonene SOA is ≤1.1 (Varutbangkul et al., 2006). However, we do not observe an obvious change in the mean diameter when comparing dry and wet conditions (Fig. 1d). In addition, hygroscopic growth should also occur for Δ³-carene SOA, but no obvious enhancement in particle mass is observed (Fig. 1a). Therefore, it is suggested that physical processes regarding hygroscopic growth play a minor role in the enhancement in limonene SOA under high RH. As a consequence, we believe that chemical processes are likely the reason of the enhancement in limonene SOA under high RH. Water can influence chemical processes in the gas phase or in the particle phase. Particle-phase reactions can promote the growth of small particles and, thus, mainly lead to larger particle sizes. As the observed SOA enhancement is mainly from high number concentration particles rather than the large size particles (Fig. 1b and d), it is likely that the water-participated gas-phase reactions are the most possible reasons for the limonene-SOA enhancement. The reaction mechanism is analyzed below based on the mass spectra information on the SOA.

3.2 Molecular analysis of SOA particles

The UPLC/ESI-Q-TOF-MS was used to examine the SOA molecular composition under high- and low-RH conditions. As shown in Fig. 2a, the mass spectra of limonene SOA are divided into four groups: monomeric group ($<m/z$ 300), dimeric group ($m/z$ 300–500), trimeric group ($m/z$ 500–700), and tetrameric group ($m/z$ 700–1000), corresponding to products containing one, two, three, and four oxygenated limonene units, respectively (Bateman et al., 2009). Most of the SOA molecules are monomers (>60 %) (Fig. 2b) and dimers (∼25 %), while trimers and tetramers contribute very small fractions (<10 % and ∼3 %) (Table S1 in the Supplement). Correspondingly, the distribution of Δ³-carene SOA can be divided into four groups (Fig. S4), comparable to that of limonene SOA. Most of the SOA molecules are monomers (∼70 %) and dimers (∼25 %), while trimers and tetramers contribute to smaller proportions (∼2 % and <1 %, respectively) (Table S2). Although the SOA mass concentration increases by ∼100 % under high-RH conditions, the relative intensities of MS peaks do not significantly change with varying RH conditions. In other words, we did not observe an obvious change in the overall MS patterns, and the fractions of the four groups only slightly differed under different RH conditions; e.g., the fraction of monomers was 62 % under dry conditions and 66 % under wet conditions. However, if we take a closer look, the intensities and contributions of specific peaks are quite different with varying RH. For example, the relative intensity of C₁₀H₁₆O₂, a possible first-generation product (Gong et al., 2018), decreases by ∼20 % with increasing RH from dry to 60 % (Table S3). This is likely due to the multi-generation reactions influenced by water vapor concentration, as discussed below with the proposed reaction mechanism of limonene ozonolysis.

Figure 3 Proposed formation mechanism for SOA formation from eCI₁ oxidation under high RH. The compounds in blue and in boxes are identified using UPLC/(−) ESI-Q-TOF-MS.

The proposed reaction mechanism of limonene ozonolysis is shown in Figs. 3 and 4. The initial step in the reaction of O₃ with limonene is the attack of the endocyclic double bond to form eCI₁ and eCI₂ (excited Criegee intermediates, with branching ratios of 0.35 and 0.65, respectively). In the context of eCI₁, several complex reactions occur, with the most dominant reaction being the generation of hydroxyl radicals (OH) and a reaction pathway known as sCI₁. The sCI₁ pathway can proceed through three distinct reactions, as depicted in Fig. 3. The first pathway is the reaction with H₂O, alcohol, or carboxylic acid to form a carboxylic acid species with hydroxyl, which would subsequently lose a molecule of water to form limononaldehyde or lose a molecule of hydrogen peroxide to form limononic acid (Grosjean et al., 1992; X. Li et al., 2019). The second and third pathways involve reactions of sCI₁ with carboxylic acids and carbonyls, respectively, leading to the formation of anhydrides and secondary ozonides. Additionally, the generated OH radicals can react with limonene, giving rise to another alkyl radical, C₁₀H₁₇O•. These alkyl radicals react with O₂ and form peroxy radicals (RO₂•). The atmospheric fate of produced RO₂• in the absence of NO_x includes the reaction with RO₂• or HO₂• (Atkinson and Arey, 2003) and the unimolecular H shift. The RO₂• + HO₂• route mainly forms hydroperoxide (ROOH), and the minor fraction forms alcohols and carbonyls (Atkinson and Arey, 2003). The products of bimolecular reactions between RO₂• and RO₂• are alcohols, carbonyls, alkoxy radicals, peroxides, and ROOR dimers (Hammes et al., 2019; Peng et al., 2019). The H shift of RO₂• can form second-generation R• and trigger a main generation channel of highly oxidized molecules (HOMs); i.e., R• would go through a process of repeated oxygen addition and hydrogen-atom shift to form HOMs with high $\mathrm{O}/\mathrm{C}$ ratios of >0.7–0.8 (Molteni et al., 2018; Bianchi et al., 2019).

Figure 4 Proposed formation mechanisms for SOA formation from eCI₂ and exocyclic double-bond oxidation under high RH. The compounds in blue and in boxes are identified using UPLC/(−) ESI-Q-TOF-MS.

In addition to the eCI₁ route, the eCI₂ pathway is also responsible for the generation of various products (Fig. 4). Since the reaction of the hydroxyl radical (OH) attacking limonene is already depicted in Fig. 3, our main emphasis in Fig. 4 is on the pathways involved in the generation of SCI. First, sCI₂ reacts with H₂O and decomposes to limononaldehyde and H₂O₂. Additionally, sCI₂ could experience an O₂ addition, •OH loss, and isomerization to produce two types of RO₂•, which can undergo the similar reactions as the RO₂• formed from the sCI₁ route, and the major products are also shown in Fig. 4.

Since limonene and Δ³-carene both have an endocyclic double bond, similar reactions to those mentioned above can occur for the ozonolysis of Δ³-carene (Fig. S5), and most corresponding formulas in Fig. S5 can be identified in Table S4. However, the reactivity of limonene towards O₃ is expected to be higher owing to its exocyclic double bond. As shown in Fig. 4, the attack of O₃ of the exocyclic double bond mainly leads to sCI₃ (highlighted in red), with the unpaired electrons outside the ring (Leungsakul et al., 2005). sCI₃ can react with H₂O to form a carbonyl called keto-limonene. It should be noted that this reaction can occur not only for limonene but also for all the products that retain the exocyclic double bond. As a result, the compounds that are colored in blue in Figs. 3 and 4 can undergo further reactions to generate products with an additional carbonyl (see the boxes in Figs. 3 and 4). Furthermore, their molecular formulas shown in Table S5 have been identified using the Q-TOF-MS. This mechanism can explain the decrease in the relative intensity of C₁₀H₁₆O₂ well from high RH to low RH and the increase in the relative intensity of C₉H₁₄O₃ from low RH to high RH (Table S3).

With such progress, we cannot rule out the possibility that relative humidity (RH) may influence the generation of other free radicals (Ma et al., 2009), thereby impacting the formation of secondary organic aerosols (SOAs), such as OH-radical reactions (Bonn et al., 2002; Fick et al., 2002). However, molar OH-radical yields were reported as 0.65±0.10 (Hantschke et al., 2021), 0.86±0.11 (Aschmann et al., 2002), and 0.56 to 0.59 (Wang et al., 2019) for Δ³-carene, while for limonene, the reported yields were 0.67±0.10 (Aschmann et al., 2002) and 0.76±0.06 (Herrmann et al., 2010). It seems that the OH radicals produced from limonene and Δ³-carene are quite similar within the range of uncertainties. Therefore, the increased ozone consumption by limonene seems primarily attributed to the presence of its exocyclic double bond.

3.3 Processes leading to the increase or decrease in SOA formation

Based on the results and mechanisms shown above, we present evidence that high humidity enhances limonene-SOA formation. First, the presence of water vapor enhances the formation of carbonyls from the reaction of exocyclic double bond, and the oligomerization of these carbonyls generates more dimers including hemiacetal (or acetal) formation and aldol condensation (Zhang et al., 2022; Kroll et al., 2005; Jang et al., 2003). As shown in Table S6, the intensity of dimers generating from multi-carbonyls under high RH is higher than that under low RH, and 54 out of the total 187 dimers were exclusively observed for limonene under high humidity conditions, contributing to a corresponding intensity of ∼19 % (Table S7). These dimers can be classified as low-volatility organic compounds (LVOCs; $\mathrm{3}\times {\mathrm{10}}^{-\mathrm{4}}$ < C₀ < 0.3 µg m⁻³) and extremely low-volatile organic compounds (ELVOCs; C₀ < $\mathrm{3}\times {\mathrm{10}}^{-\mathrm{4}}$ µg m⁻³) (Fig. 5a) and thus promote the nucleation and new particle formation in different ways. This finding is similar to that from a previous study showing that high RH can promote dimer formation from the ozonolysis of α-pinene (Kristensen et al., 2014). Second, we find that high RH can also promote the formation of HOMs, although the mechanism remains unclear. As shown in Table S3, many HOMs proposed from the mechanism are detected under high-RH conditions but not detected under low-RH conditions, including both monomers and dimers. Many HOMs have low volatilities and, thus, can also promote new particle formation. Overall, the promoted dimer and HOM formation may greatly enhance the new particle number concentration under high-RH conditions (Fig. 6).

Figure 5 (a) Distribution of the limonene SOA in the two-dimensional volatility basis set (2D-VBS) space. (b) Partitioning coefficients of limonene monomers and dimers under low- and high-RH conditions.

Figure 6 Schematic diagram of the possible mechanisms for the enhancement of limonene SOA.

High particle number concentration generally provides more surface areas for semi-volatile organic compounds (SVOCs; 0.3 < C₀ < 300 µg m⁻³) to condense on, which results in a higher condensation sink (CS). In the OFR, the fates of SVOCs include condensing on aerosol, getting lost on the wall, and reacting with OH radicals to form functionalization and/or fragmentation products (Palm et al., 2016; K. Li et al., 2019). The promoted condensation by higher CS leads to a higher fraction of SVOCs getting into the particle phase rather than getting lost on the wall or becoming smaller fragments staying in the gas phase, thus promoting SOA formation (K. Li et al., 2019). Furthermore, the transformation from a C–C double bond to carbonyl shown in Figs. 3 and 4 decreases the volatility of molecules, which can largely influence the gas–particle partitioning of the monomeric compounds (Fig. 5b). For example, the C₀ values of C₁₀H₁₆O₂ and C₁₀H₁₆O₃ are 90 701 and 19 968 µg m⁻³, corresponding to partitioning coefficients of 0.01 and 0.05, respectively (Fig. 5b and Table S3), with an SOA mass concentration of ∼1000 µg m⁻³ under dry conditions. When they are converted to carbonyls C₉H₁₄O₃ and C₉H₁₄O₄, the values of C₀ become 45 556 and 6479 µg m⁻³, corresponding to partitioning coefficients of 0.02 and 0.13, respectively (Fig. 5b and Table S3), with the same SOA loading. This enhancement in partitioning coefficient can largely promote the condensation of SVOCs and, thus, enhance the SOA mass concentration. In addition, the enhanced SOA formation can further influence the equilibrium; e.g., the partitioning coefficient of C₁₀H₁₆O₃ increases from 0.05 to 0.10 when SOA mass concentration increases from ∼1000 µg m⁻³ under dry conditions to ∼2000 µg m⁻³ under wet conditions (Fig. 5b and Table S3). The distribution of saturation vapor pressure for monomers and dimers identified by MS has also been shown in Fig. 5a. As can be seen from this figure, around 50 % monomers are categorized as SVOCs, thus having the large fraction in the particle phase when converting from dry to wet conditions. Overall, it is likely that the different fate and partitioning of SVOCs largely enhance the amount of SVOCs in the particle phase (Fig. 6).

Concluding the analysis above, high humidity promotes the SOA formation from the ozonolysis of limonene in two steps: nucleation of new particles and condensation of SVOCs on them (Fig. 6). While our study highlights significant changes in gas-phase chemistry, we cannot exclude the possibility of concurrent reactions occurring in the condensed phase. These two steps are closely related to the multi-generation reactions of the exocyclic C = C bond, which are unlikely to happen for the ozonolysis of Δ³-carene. Interestingly, Gong and Chen (2021) have found that high RH can inhibit the SOA formation from the first-generation oxidation of limonene ozonolysis but enhance the SOA formation from the second-generation oxidation (Gong and Chen, 2021); their results agree well with the results and analysis shown here. In contrast, X. Li et al. (2019) found negligible change in dimers and HOMs in limonene-O₃ system when changing RH from 0 % to 60 %. The discrepancy is mainly attributed to the different experimental conditions. The ozone exposure in this study is ∼18 times higher than in X. Li et al. (2019), while the limonene concentration in this study is only ∼30 % of that in their study. These two conditions both favor the multi-generation reactions that occurred at the exocyclic double bond of limonene and its products. Thus, we believe this leads to the different results regarding the formation of HOMs and dimers.

Regarding Δ³-carene, the mechanisms and processes are almost opposite to those of limonene. First, water vapor reacts with sCI₁ or sCI₂ to promote the formation of α-hydroxyalkyl–hydroperoxides (Fig. S5). Their subsequent products without second ozonolysis of exocyclic double bond have higher volatility and may most likely prevail in the gas phase. In addition, it has been found that α-hydroxyalkyl–hydroperoxides preferentially decompose into aldehydes and H₂O₂ (Kumar et al., 2014; Chen et al., 2016), i.e., 3-caronaldehyde for Δ³-carene, which has higher volatility than the products from other reaction pathways. Correspondingly, the number and relative intensity of HOMs and dimers detected under high-RH conditions are both lower than those under low-RH conditions (Table S8). Furthermore, out of a total of 178 dimers, 63 dimers were exclusively identified under low-RH conditions (Table S7). As a result, high RH shows an inhibitory effect on the SOA formation from Δ³-carene ozonolysis.

To investigate the multi-generation reactions of limonene under low-concentration conditions, we conducted low-concentration limonene ozonolysis experiments, and the results are shown in Fig. S6. In these experiments, the limonene and O₃ concentrations were 20.5 ppb and 5.7 ppm, respectively. According to the experimental results, the number concentration of SOA formed from limonene ozonolysis increased by approximately 1.4 times under high RH, which is similar to the increase observed under high-loading conditions. The mass concentration increased by approximately 1.3 times at a precursor concentration of 20.5 ppb. The relatively small increase in mass concentration compared to the high-concentration conditions may be attributed to the less pronounced distribution of SVOCs at low mass concentrations. This result suggests that the enhancement effect on limonene SOA by high RH is still valid for low precursor concentrations.

To further confirm the assumption that water-influenced multi-generation reactions of the exocyclic double bond enhance the SOA formation, we conducted two comparative analyses: firstly, we examined the ozonolysis of the endocyclic double bond in limonene, leaving the exocyclic double bond unreacted. This was done by applying a low O₃ concentration (∼67 ppb), since the reaction of O₃ with an endocyclic double bond is ∼30 times faster than the reaction of O₃ with an exocyclic double bond (Shu and Atkinson, 1994). Interestingly, when limonene was oxidized at only the endocyclic double bond, we observed a slight decrease in both the number and mass concentrations as the RH increased (Fig. S7). This is similar to the results obtained for Δ³-carene, which contains only one endocyclic double bond. Secondly, we compared the ozonolysis of structurally similar β-caryophyllene, which has an exocyclic C–C double bond that can undergo further reactions (Fig. S8). As expected, we observe a large enhancement in SOA formation under high-RH conditions (Table S9 and Fig. S9). This implies that monoterpenes, sesquiterpenes, and other BVOCs with two unsaturation double bonds may follow similar reaction mechanisms during ozonolysis and thus have a RH dependency in SOA production.

4 Conclusions

In this study, the effect of humidity on SOA production from the ozonolysis of two monoterpenes (limonene and Δ³-carene) was investigated with an OFR. Contrasting impacts of RH on the SOA formation were observed: limonene-SOA yield increases by ∼100 % when RH changes from ∼1 % to ∼60 %, while Δ³-carene-SOA yield slightly decreases. By analyzing the chemical composition of SOA with ESI-Q-TOF-MS, we find that the multi-generation reactions of the exocyclic C–C double bond are likely the driving force of the enhancement in limonene SOA. The presence of water promotes the formation of carbonyls from the reaction of exocyclic double bond and further favors the formation of dimers and HOMs. This leads to promoted new particle formation and subsequent condensation of SVOCs. These reactions also lower the volatilities of the SVOCs and further promote the gas–particle partitioning. Moreover, this hypothesis is supported by a similar behavior of the ozonolysis of β-caryophyllene (sesquiterpene with an exocyclic double bond) in SOA enhancement under high-RH conditions. However, since aerosol dynamics of small clusters and particles are very complex, we do not rule out a series of reactions that may occur in the particle phase. The results in this study suggest that multi-generation reactions play an important role in SOA formation from the ozonolysis of BVOCs, which are significantly influenced by humidity. This impact is largely dependent on the molecular structure of the SOA precursors (e.g., with or without the exocyclic double bond), thus highlighting the importance of considering the molecular structure of monoterpenes in modeling and field studies of biogenic SOA.