Chemical characterization of organic compounds involved in iodine-initiated new particle formation from coastal macroalgal emission

Iodine-initiated new particle formation (I-NPF) has long been recognized in coastal hotspot regions. However, no prior work has studied the exact chemical composition of organic compounds and their role in coastal I-NPF. Here we present an important complementary study to the ongoing laboratory and field research on iodine nucleation in the coastal atmosphere. Oxidation and NPF experiments with vapor emissions from real-world coastal macroalgae were simulated in a bag reactor. On the basis of comprehensive mass spectrometry measurements, we reported for the first time a variety of volatile precursors and their oxidation products in gas and particle phases in such a highly complex system. Organic compounds overwhelmingly dominated over iodine in the new particle growth initiated by iodine species. The identity and transformation mechanisms of organic compounds were proposed in this study to provide a more complete story of coastal NPF from low-tide macroalgal emission.

1 Introduction

Coastal new particle formation (NPF) may be driven by daytime low-tide emission of iodine species from macroalgae fully or partially exposed to the air. The phenomenon was reported in hotspot locations of west Europe, Australia and polar regions (O'Dowd et al., 2002; Heard et al., 2006; McFiggans et al., 2010; Whitehead et al., 2009; Sipilä et al., 2016; Allan et al., 2015; Baccarini et al., 2020; Beck et al., 2021). On the southeast coastline of China, we reported intense iodine-initiated NPF based on particle number size distribution and iodine measurements (Yu et al., 2019).

To simulate iodine-initiated NPF (I-NPF) in controlled laboratory conditions, I₂ or CH₂I₂ vapor was usually photolyzed in the presence of ozone to provide nucleation precursors (Burkholder et al., 2004; Jimenez et al., 2003; Monahan et al., 2012; Saunders and Plane, 2005; O'Dowd et al., 2004; Gómez Martín et al., 2020; He et al., 2021; Huang et al., 2022; Gómez Martín et al., 2022). Ashu-Ayem et al. (2012), Monahan et al. (2012), McFiggans et al. (2004), Sellegri et al. (2005) and Sellegri et al. (2016) also investigated NPF from the vapors emitted by real-world macroalgal specimens or seawater in laboratory chamber or apparatus. However, the focus of all the above studies was emission rate, oxidation mechanisms or nucleation pathways of iodine species. For example, positive correlations between particle concentrations and I₂ or CH₂I₂ mixing ratios were usually observed (Burkholder et al., 2004; Jimenez et al., 2003; Sellegri et al., 2005; Monahan et al., 2012). Kinetic studies in a flow tube or a CERN CLOUD chamber proposed the clustering of iodine oxides (I_xO_y) or iodine oxoacids (HIO₃, HIO₂) as nucleation mechanisms on the basis of photoionization time-of-flight mass spectrometry (TOF-MS; Gómez Martín et al., 2020), atmospheric pressure interface TOF (APi-TOF) and nitrate chemical ionization mass spectrometer (nitrate-CIMS) measurements (He et al., 2021; Gómez Martín et al., 2022).

Organic compounds have also been suggested to be involved in coastal NPF (Vaattovaara et al., 2006; Yu et al., 2019). Huang et al. (2022) and Saunders et al. (2010) investigated the effect of uptake of meso-erythritol, glyoxal, dimethylamine and oxalic acid on the growth of iodine oxide nanoparticles. However, no prior work has investigated the exact chemical identity of organic compounds (other than iodomethane) and their role in I-NPF. The role of biogenic terpenes and anthropogenic aromatics in continental NPF has been recognized for a long time (Donahue et al., 2013). Their ozonolysis or photochemistry products have been investigated in depth using electrospray ionization mass spectrometry (ESI-MS) and, more recently, CIMS instruments (Nguyen et al., 2010; Kundu et al., 2017, 2012; Faxon et al., 2018; Wang et al., 2020; Riva et al., 2017; Yan et al., 2020; Ehn et al., 2014). It is very likely that certain volatile organic compounds (VOCs) emitted mutually with iodine or iodinated methane from coastal biota or biologically active sea surface may also be involved in the coastal I-NPF process and promote the growth of iodine particles.

To test this hypothesis, we conducted oxidation and NPF experiments with vapor emissions from real-world coastal macroalgae in a bag reactor. A suite of mass spectrometric methods including inductively coupled plasma MS (ICP-MS), gas chromatography MS (GC-MS), iodide-CIMS and ESI-orbitrap MS were applied to measure vapor precursors, gaseous products and particulate products during the NPF process. Mass concentrations of total organic carbon (TOC) and total iodine (TI) of new particles were compared to evaluate the relative importance of organics and iodine in new particle growth. The identity and transformation mechanisms of organic compounds were identified to provide a more complete story of coastal NPF from low-tide macroalgal emission. Our study is thus complementary to prior laboratory and field studies of I-NPF but has an emphasis on organics.

2 Experiments

2.1 Experimental apparatus and sample collection

Similar to a potential aerosol mass (PAM) oxidation flow reactor, a bag reactor was designed to provide an oxidizing environment for simulating atmospheric oxidation processes of algae-emitted VOCs. The bag reactor was made from 75 µm thick fluorinated ethylene propylene (FEP) Teflon (1.2 m × 1.5 m, flat dimension). The volume of the bag at full inflation was determined experimentally to be about 200 L. The bag was suspended vertically (Fig. 1) and kept in the dark or directly exposed to room light of a fluorescent lamp. Before each experiment, the bag was purged for several hours to reduce background particle concentrations to below 1 cm⁻³.

Figure 1 Schematic of experimental setup. Solid line: air flows. Dashed lines: sent for offline chemical analysis.

Undaria pinnatifida, a common brown seaweed species found in the Xiangshan Gulf on the east China coast, was collected from local intertidal zone and stored at −10 ^∘C until the experiments. A quantity of 2 kg macroalgae was put in a 20 L Pyrex glass bottle that was filled with ∼ 1 L natural seawater. The specimens were partially exposed to the air to simulate tidal emersion of macroalgae. A flow of particle-free ultra-high-purity (UHP) air blew algae-emitted VOCs out of the bottle and merged with a diluting air flow before entering the bag reactor.

Two types of experiments were conducted. In the three ozonolysis experiments, ozone (O₃) was generated by UHP air flowing through a 5 W 185 nm UV lamp. The O₃ flow was fed just before the bag reactor was fully inflated. Final O₃ concentration in the bag reactor was measured to be ∼ 500 ppbv using an ozone analyzer (Model 49i, Thermo-Fisher Scientific Inc.). RH was estimated to be 10 % in the bag reactor assuming 0.3 L min⁻¹ water-saturated VOC flow was diluted by 2.7 L min⁻¹ dry-air flow. In an additional OH-enhanced experiment, the O₃ $/$ VOC mixture flow was directed through a 254 nm UV light before entering the bag reactor. OH radicals were produced via the reaction O₃ + hv → O₂+O(¹D) and O(¹D) + H₂O → 2OH. Integrated OH exposure time was determined by a SO₂ decay experiment to be 2.4 d in the experimental apparatus assuming an ambient average OH concentration of 1.5×10⁶ molecules cm⁻³ (see Supplement Sect. S1). Other oxidants may include O(³P) that resulted from the quenching of O(¹D) (Li et al., 2015). Because the purpose of this study is to identify gas and particle products of algae-emitted VOCs in the simulated NPF event, a significantly higher oxidation level in the bag reactor than the atmosphere should not change the conclusions in the article. Wall loss, aerosol yield, reaction rate and other kinetic factors in the bag reactor were also not evaluated.

The bag reactor was first operated in a static mode to monitor the time evolution of gaseous products and particle size. In the static mode, the bag was first filled to full inflation with the VOC $/$ O₃ flows. The flows were then shut down; a scanning mobility particle sizer (SMPS; model 3936, TSI Inc., Shoreview, MN, USA) and an Aerodyne iodide-CIMS pulled two flows of 0.3 and 1.8 L min⁻¹ out of the bag, respectively. The SMPS measured the particle number size distribution from 14 to 600 nm.

The bag reactor was then operated in a dynamic mode for a few hours to collect enough particles for offline chemical analysis. In the dynamic mode, the VOC $/$ O₃ flow of 3 L min⁻¹ was fed to the bag continuously, while the SMPS and a vacuum pump (GAST Group Ltd.) pulled sample flows of 0.3 and 2.7 L min⁻¹, respectively, out of the bag reactor. This resulted in an overall residential time of 67 min for the O₃ $/$ VOC mixture in the fully inflated bag. The particles in the 2.7 L min⁻¹ sample flow were collected onto a Zefluor^® PTFE membrane filter mounted in a filter inlet for gases and aerosols (FIGAERO) for iodide-CIMS analysis or, alternatively, onto 47 mm diameter double quartz fiber filter pack mounted in a filter holder for ESI-orbitrap MS, ICP-MS and TOC analysis. The front filter of the double-filter pack collected the particles and also adsorbed some volatile species as positive artifacts, while the back filter placed downstream of the front filter was supposed to adsorb the same amount of volatile species as the front filter.

2.2 Chemical analysis

Before the ozonolysis experiments, the algae-emitted VOCs in the bag reactor were collected by a 6 L pre-evacuated stainless-steel canister (Entech Instruments, Inc., Simi Valley, CA, USA) and analyzed using a quadrupole GC-MS system (model TH-300B, Wuhan Tianhong Instruments Co. Ltd., Wuhan, China). The algae-emitted VOCs, as well as their gaseous and particulate products, were also measured by the FIGAERO-iodide-CIMS. Iodide-adduct chemical ionization is well suited for measuring oxygenated or acidic compounds with minimal fragmentation. More details of the GC-MS and FIGAERO-iodide-CIMS measurements can be found in Sect. S2. The theory and design of the two instruments were described by Wang et al. (2014) and Lopez-Hilfiker et al. (2014).

The particles collected on quartz fiber filters were sent for offline quantification of TOC and TI, as well as non-target analysis of organic compounds using ESI-orbitrap MS. The front and back filters were treated, separately, following the procedure as below: the filter was ultrasonicated twice with 10 mL water and acetone nitrile solvent mixture ($v:v=\mathrm{1}:\mathrm{1}$). Ultrasonication time and power were 20 min and 150 W. The extract was filtered by a 0.2 µm PTFE syringe filter and evaporated in a rotary evaporator to 0.5 mL. After being centrifuged for 30 min at 12 000 rpm, the supernatant was collected for TI analysis by Agilent 1100 HPLC-7900 ICP-MS (Agilent Technologies, Santa Clara, CA, USA) and TOC analysis by a TOC analyzer (Model TOC-5000A, Shimadzu, Japan). TI or TOC in the particles was obtained by subtracting the amount on the back filter from that on the front filter. Nontarget analysis of organic compounds in the supernatant was conducted using a Q Exactive hybrid Quadrupole-Orbitrap mass spectrometer (Thermo Scientific, Bremen, Germany). The supernatant was directly infused by a syringe pump and ionized in negative ESI source. All the ions in the $m/z$ range from 50 to 500 Th were scanned with a mass resolution of 70 000. The chemically sound CHO molecular formulas were computed with a mass tolerance of ±2 ppm for these ions. Only the compounds that existed solely in the front filter or with ion intensity in the front filter higher than that in the back filter by a factor of 3 were regarded as the organic compounds in the particle phase (Wang et al., 2017).

3 Results and discussion

No particles formed in the absence of room light or O₃ (see Supplement Fig. S1). Therefore, light was on throughout the experiments reported in the article. In the static mode experiments, we could not observe gas-phase products until 48 min after O₃ injection. New particles larger than 14 nm were observed only 58 min after O₃ injection. Afterwards, new particles began to grow to form a typical banana-shape particle size spectrum (Fig. 2a). This relatively long waiting time was likely due to the buildup of O₃ and oxidation products. Time zero of Fig. 2 was thus set as the time when gaseous products first appeared.

Figure 2 Time evolution of particle number size distribution (a) and relative intensities of gaseous molecules and radicals (b–f); the fractions of organic compounds grouped by O and C atom numbers in the selected time points (g–h) in a typical ozonolysis experiment (static mode). Time zero was set as the time when gaseous products first appeared.

3.1 Macroalgal emission

It is of particular interest to know what VOCs are emitted from coastal macroalgae. They are potential precursors of iodine particle nucleation and growth. The canister sampling followed by GC-MS analysis showed that the top nine non-CHO compounds with highest TIC (total ion chromatogram) peak areas (Table 1) are C₅ alkanes, C₁₀ α-pinene, and halogenated C₁, C₃ and C₅ alkanes. The top 10 CHO compounds are C₂–C₆ alcohols and carbonyls with saturated or unsaturated carbon chain.

Table 1 Major volatile organic compounds emitted by macroalgae as potential NPF precursors, sorted by TIC peak area measured by GC/MS or MS peak intensity measured by the iodide-CIMS.

The iodide-CIMS is more sensitive to more oxygenated or acidic compounds and thus complementary to the GC-MS measurement. The 76 organic precursors detected by the iodide-CIMS before ozone addition were characterized by C${}_{\mathrm{1},\mathrm{2},\mathrm{3},\mathrm{6}}$ and O₂₋₃ formulas (Fig. 3a). The top seven compounds with highest ion intensities were CH₂O₂, C₂H₄O₂, C₃H₆O₃, C₆H₁₀O₃, C₂H₆O₂, C₄H₈O₂ and C₆H₁₂O₃, which accounted for 82.5 % of total ion intensity. They are C₁–C₆ mono-carboxylic acids, hydroxyl carboxylic acids or oxo-carboxylic acids with two to three oxygen atoms (Table 1). Their carbon atom numbers are in general consistent with the VOCs detected by GC-MS.

Figure 3 Oxygen and carbon atom number distributions of potential VOC precursors (a), gaseous products (b) and particulate products measured by the iodide-CIMS (c), as well as the particulate products measured by ESI-orbitrap MS (d) in a typical ozonolysis experiment (dynamic mode). Hatched bars indicate the fractions of organic formulas observed in both gas and particle phases by the iodide-CIMS.

Figure 4 Integrated ion intensities of inorganic molecules and radicals in the gas phase (static mode) and particle phase (dynamic mode) measured by the iodide-CIMS in a typical ozonolysis experiment. The ions were color-coded according to their elemental composition. For each ion cluster, the parent neutral molecule is on the left-hand side of the middle dot, while the clustering ion I⁻ or NO${}_{\mathrm{3}}^{-}$ is on the right-hand side. Those without a clustering ion are shown as bare anions.

Relatively high signals of NO${}_{\mathrm{3}}^{-}$ and HNO₃I⁻ were observed before the addition of ozone to the bag reactor. They were likely HNO₃ or nitrate vaporized from algal specimens or natural seawater. Because NO${}_{\mathrm{3}}^{-}$ and HNO₃I⁻ were also observed in the particle phase during the NPF (Fig. 4), we assume HNO₃ was also an important precursor of particle formation.

3.2 Gaseous products

3.2.1 Gaseous inorganic molecules and radicals

Being different from the nitrate-CIMS, our iodide-CIMS did not detect nucleating clusters of iodine oxides or oxyacids after the addition of ozone. Instead, dozens of new inorganic molecules or radicals were observed as clusters with I⁻, NO${}_{\mathrm{3}}^{-}$ or deprotonated ions in the gas or particle phase (Fig. 4). We grouped these species by elemental composition and investigated their role in the NPF by observing how their gaseous ion intensities evolved during the NPF event in the bag reactor (Fig. 2b–f).

1. Cl, I, Cl₂ and ClI in the gas phase. The intensities of I and Cl increased ca. 10 min before 14 nm particles appeared and decreased as the particles grew up (Fig. 2b). Based on prior work of Burkholder et al. (2004), Jimenez et al. (2003) and O'Dowd et al. (2004), we suggested the photolysis of CH₂Cl₂, CHBrCl, CH₃I and C₃H₇I was the source of halogen atoms (e.g., CH₃I + hv→ CH₃ + I), although we could not exclude the photolysis of other precursors like I₂ and HOI that are invisible to GC-MS and the iodide-CIMS. There was a time lag of 20–25 min between the appearances of Cl and I and those of ClI and Cl₂, which probably resulted from anion exchange reactions of Cl ⋅ I⁻ and I ⋅ I⁻ with Cl atoms.

2. IO₂, IO and ClIO in the gas phase. These species showed a similar time evolution to I and Cl atoms (Fig. 2c). They could be from the reactions between I, ClI and O₃ (Saiz-Lopez et al., 2014). Sequential oxidation and aggregation reactions might have formed other halogen oxides (Gómez Martín et al., 2013), but they might not be detectable by the iodide-CIMS.

3. INO₂, ClNO₂ and INO₃. INO₂ and ClNO₂ were detected in the gas phase with similar time evolution with halogen atoms and halogen oxides (Fig. 2d). INO₃ was found in both gas and particle phases. INO₂ and INO₃ were usually thought to form upon the reactions I + NO₂ + M → IONO + M and IO + NO₂ + M → IONO₂ + M in the atmosphere (Saiz-Lopez et al., 2012), which seems to be unlikely in our bag reactor because NO₂ was not added. Considering NO${}_{\mathrm{3}}^{-}$ was ubiquitous in the bag reactor of our experiment, it is likely that INO₂ and INO₃ formed via I₂O₂ + NO${}_{\mathrm{3}}^{-}$ → IO${}_{\mathrm{3}}^{-}+$IONO and I₂O₃ + NO${}_{\mathrm{3}}^{-}$ → IO${}_{\mathrm{3}}^{-}$ + IONO₂. These reaction pathways have been supported by theoretical calculation and flow tube mass spectrometry experiments (Gómez Martín et al., 2022, 2020). ClNO₂ was likely to form upon similar reaction between Cl₂O₂ and NO${}_{\mathrm{3}}^{-}$ in the bag reactor.

4. HIO, HIO₂ and HIO₃. HIO₃ seems to be the end product of the above intermediates because its gas-phase ion intensity kept on increasing during new particle growth (Fig. 2e). Based on this fact, we assume that HIO₃ could be from I₂O₅ + H₂O → 2HIO₃ or I₂O_y + NO${}_{\mathrm{3}}^{-}$ → IO${}_{\mathrm{3}}^{-}$ + INO_y. On the other hand, HIO₃ was not detected in the particle phase by the iodide-CIMS, which is contrary to the offline analysis of the quartz filter by HPLC-ICP-MS showing that total iodine was mostly dominated by IO${}_{\mathrm{3}}^{-}$ peak. We speculate that HIO₃ might have been dehydrated to I₂O₅ under thermal desorption temperature up to 180 ^∘C in FIGAERO. The signals of IO⁻, IO${}_{\mathrm{2}}^{-}$ and HIONO${}_{\mathrm{3}}^{-}$ (corresponding to HIO and HIO₂) were found in the particle phase but not in the gas phase. He et al. (2021) proposed HIO₂ formation via I⁻ + H₂O + O₃ → HIO₂ or I₂O₂ + H₂O →HIO + HIO₂. With limit experimental evidence of our work, the exact formation pathways of HIO_x remain to be explored in future.

5. CH₃SO₃H, S${}_{\mathrm{2}}^{-}$, S${}_{\mathrm{3}}^{-}$ and SO${}_{\mathrm{3}}^{-}$. We observed methane sulfonic acid (CH₃SO₃H, MSA) in both gas and particle phases. Gaseous MSA increased in the beginning but decreased after new particles appeared (Fig. 2f). Apparently, our measurement suggested MSA contributed to the growth of new particles, but it is unknown if it also participated in nucleation. We suggested S${}_{\mathrm{2}}^{-}$, S${}_{\mathrm{3}}^{-}$ and SO${}_{\mathrm{3}}^{-}$ ions observed in the particle phase were thermal decomposition products of MSA.

We note that CH₃I vapor was added as an ion source reagent to the ion molecule reactor (IMR) of the iodide-CIMS. It is likely that this extra CH₃I in the IMR might obscure the interpretation of the observed iodine containing clusters. We believed that the ion source reagent CH₃I should have relatively small interference with inorganic iodine compounds from the bag reactor, on the basis of two facts: (1) the ion source reagent CH₃I was added directly from the permeation tube into the IMR. Without photolysis, the ion source reagent CH₃I in the IMR should not become a source of I and I_xO_y. (2) The concentration of the ion source reagent CH₃I and its potential products should be quite constant as long as O₃ was present in the IMR, which was not supported by the variable signals of I, ClI, IO₂, IO, ClIO, HIO₃, INO₂ and INO₃ in Fig. 2.

3.2.2 Gaseous organic products

After ozone addition, a gradual transformation from C₁–C₃ precursors to C₅–C₈ gaseous products was observed during the NPF process (Fig. 2h). In the meanwhile, the oxygen atom number of the compounds increased from two–three to four–seven (Fig. 2g). The formation of compounds with more carbon atoms than the parent VOCs is unlikely in the gas phase, except bimolecular reactions of stabilized Criegee intermediates (SCIs) that typically form upon alkene ozonolysis. Similar to isoprene ozonolysis (Riva et al., 2017; Inomata et al., 2014), we propose the SCI addition mechanism can also explain the transformation observed in our case: (1) C₄ SCIs formed upon the ozonolysis of CHO precursors with C=C double bonds (e.g., those observed by GC-MS in Table 1), and (2) the insertion of C₄ SCIs into carboxylic acid precursors (e.g., those observed by the CIMS in Table 1) produced oligomeric hydroperoxides. An example is shown in Scheme 1 for the reactions of most abundant ethyl vinyl carbinol (C₅H₁₀O), ozone and formic acid (CH₂O₂), but the same mechanism is also applicable for ethyl vinyl ketone (C₅H₈O) and other abundant C₂–C₅ carboxylic acids and hydroxyl carboxylic acids. As a result, a series of gaseous oligomeric hydroperoxides, C₅H₁₀O₅, C₆H₁₀O₅, C₆H₁₂O₅, C₇H₁₂O₆, C₇H₁₄O₆, C₈H₁₄O₅, C₈H₁₆O₆, C₈H₁₆O₅ and C₉H₁₆O₆, were observed with high intensity by the iodide-CIMS.

Scheme 1 A proposed addition reaction involving stabilized Criegee intermediates in the gas phase.

3.3 Particulate products

3.3.1 Relative mass contribution of organic carbon and iodine to new particles

In the dynamic-mode experiments, O₃ in the bag reactor was kept at its maximum concentration of 200 ppbv. With a prolonged residential time of 67 min, the particles grew to 102±23 nm, which was measured by the SMPS at the outlet of the bag reactor. The TOC and TI measurements show that organic compounds contributed more particle mass than iodine, with a TOC $/$ (I + TOC) ratio of 96.1±2.9 % (Table 2).

Table 2 Particle number concentration (N), mean diameter (D_p), total organic carbon (TOC) and total iodine (TI) of new particles with a residential time of 67 min in the bag reactor in the ozonolysis experiments and OH-enhanced experiment (dynamic mode). Those of 10–56 nm new particles collected by a nano micro-orifice uniform deposit impactor (nano-MOUDI; MSP, Inc.) during an I-NPF event at a coastal site of Ningbo, China (Yu et al., 2019), are also listed.

In the OH-enhanced experiment (dynamic mode), more particulate products were generated with enhanced oxidation capacity: TI in the particles increased by a factor of 10.8, TOC increased by a factor of 2.7 and particle number concentration increased by a factor of 7.4. On the other hand, particle size decreased to 73 nm, and the TOC $/$ (TI + TOC) ratio decreased to 92.9 % (Table 2). These differences indicate that more iodine nuclei were produced with enhanced oxidation capacity, probably via OIO + OH → HOIO₂ (Plane et al., 2006) and O(³P) + CH₃I → IO + CH₃ (Teruel et al., 2004). Competitive uptake of condensing organic vapors onto these iodine nuclei limited the growth of individual new particles. Nevertheless, organic compounds overwhelmingly dominated over iodine in the mass contribution to new particle growth.

The significant organic contribution observed in the laboratory condition is generally consistent with the TOC $/$ (I+TOC) ratio of 98.2 % in 10–56 nm new particles collected during a coastal I-NPF event in China (Yu et al., 2019), although TOC and TI during the field event are 2 orders of magnitude lower than those in the bag reactor (Table 2). The mean diameter of new particles was observed to be only 16 nm during the field event. But those small new particles are expected to grow into cloud condensation nuclei (CCN) active sizes, given longer residence time and uptake of more condensing vapors in the atmosphere (He et al., 2021).

3.3.2 Particulate organic products

In the end of a typical ozonolysis experiment (dynamic mode), 100 and 364 new formulas were observed in the gas and particle phases, respectively, including 73 semivolatile organic compounds (SVOCs) that appeared in both gas and particle phases. Those SVOCs accounted for 81 % and 20 % of total ion intensities of gaseous and particulate products, respectively. Being different from unimodal atom number distributions of gaseous products (C_max= 7 and O_max= 5, Fig. 3b), particulate products were characterized by distinct bimodal or trimodal distribution of carbon number (C_max= 8, 14 and 16, Fig. 3c) and oxygen number (O_max= 4 and 8), implying possible dimer formation via accretion reactions in the particle phase.

ESI-Orbitrap MS differs from the FIGAERO-iodide-CIMS in its extraction method (ultrasonic solvent extraction from quartz fiber filter vs. thermal desorption from PTFE membrane filter), ionization source (electrospray ionization vs. iodide-adduct chemical ionization) and MS-resolving power (70 000 vs. 4500). The result showed that ESI-orbitrap MS and the FIGAERO-iodide-CIMS detected 336 and 364 organic formulas, respectively, in the particle phase. A total of 167 organic formulas were commonly observed by both methods, which accounted for 87 % and 54 % of total ion intensity of organic formulas by the two methods, respectively (Fig. S2). As shown in Fig. 3c and d, the FIGAERO-iodide-CIMS had better sensitivity toward the organic compounds with more oxygen atoms (e.g., O ≥ 8) and carbon atoms (e.g., C ≥ 10). As a result, bimodal carbon and oxygen atom number distributions were observed by the FIGAERO-iodide-CIMS but not ESI-orbitrap MS.

The measurement by ESI-orbitrap MS provided more insights about the formation mechanism of particulate products. We compared the 336 formulas detected by ESI-orbitrap MS in our study with the 414 formulas of isoprene ozonolysis secondary organic aerosol (SOA) products (Nguyen et al., 2010) and 922 formulas of α-pinene ozonolysis SOA products (Putman et al., 2012) measured by the ESI-orbitrap MS. It was found that 72 % of the formulas in this study can also be found in isoprene SOA, but only 39 % can be found in α-pinene SOA. This seems to imply that some similar alkene ozonolysis reactions occurred in our system and isoprene ozonolysis.

For such a highly complex system full of various algae-emitted precursors, it is impossible to simply propose a reaction mechanism to explain the formation of all particulate products, nor to list all reactions occurring in the bag reactor. On the basis of particle-phase oligomer chemistry (Seinfeld and Pandis, 2016), especially the well-understood isoprene ozonolysis SOA chemistry (Nguyen et al., 2010; Inomata et al., 2014; Riva et al., 2017), we suggest that a variety of accretion reactions without uniform oligomerization pattern (e.g., esterification, aldol condensation, hemiacetal reactions, peroxyhemiacetal formation and SCI reactions) transformed O_max= 4 and C_max= 8 multifunctional monomers (like alcohols, carbonyls, hydroperoxides and carboxylic acids) to O_max= 8 and C_max= 14 or 16 dimers. As an example, we used two simplified reaction equations to illustrate addition-type cross-oligomerization between C₆ and C₈ monomers and self-oligomerization of C₈ monomers, respectively:

$$\begin{array}{ll}{\displaystyle }& {\displaystyle }{\mathrm{C}}_{\mathrm{6}}{\mathrm{H}}_{\text{6–12}}{\mathrm{O}}_{\text{3–6}}+{\mathrm{C}}_{\mathrm{8}}{\mathrm{H}}_{\text{10–16}}{\mathrm{O}}_{\text{3–6}}\cdot \to {\mathrm{C}}_{\mathrm{14}}{\mathrm{H}}_{\text{16–26}}{\mathrm{O}}_{\text{6–12}}\\ {\displaystyle }& {\displaystyle }{\mathrm{C}}_{\mathrm{8}}{\mathrm{H}}_{\text{10–16}}{\mathrm{O}}_{\text{3–6}}+{\mathrm{C}}_{\mathrm{8}}{\mathrm{H}}_{\text{10–16}}{\mathrm{O}}_{\text{3–6}}\cdot \to {\mathrm{C}}_{\mathrm{16}}{\mathrm{H}}_{\text{20–32}}{\mathrm{O}}_{\text{6–12}},\end{array}$$

in which the C₆ , C₈, C₁₄ and C₁₆ formulas in the equations are among the most abundant ones observed in the particle phase by the iodide-CIMS.

4 Conclusions

Using a suite of mass spectrometers, we reported, for the first time, the chemical compositions of volatile precursors emitted by real-world coastal macroalgae and their gaseous and particulate oxidation products. In the presence of room light and ozone, the photolysis of halogenated C${}_{\mathrm{1},\mathrm{3},\mathrm{5}}$ alkanes ends up as HIO₃ and INO₃. It was most likely HIO₃ initiated NPF and provided nuclei for the further condensation of other products like MSA and CHO compounds. Gas-phase SCI reactions and particle-phase accretion reactions transformed C₁–C₆ and O₂–O₃ precursors gradually to particulate products with C_max= 8, 14 and 16 and O_max= 4 and 8. As a result, organic carbon was found to overwhelmingly dominate over iodine in the mass contribution to the new particle growth. Although our instruments did not allow for the detection of nucleating clusters of iodine oxides or oxyacids, our study provided important complementary information to the ongoing laboratory and field research on coastal I-NPF.