Large contribution of organics to condensational growth and formation of cloud condensation nuclei (CCN) in remote marine boundary layer

Marine low clouds strongly influence global climate, and their radiative effects are particularly susceptible to the concentration of cloud condensation nuclei (CCN). One major source of CCN is condensational growth of pre-CCN particles, and sulfate has long been considered the major condensing species in remote marine boundary layer. While some studies suggested that secondary organic species can contribute to the particle growth, its importance remains unclear. Here we present the first long15 term observational evidence that organics play an important role in particle growth over remote oceans. To the contrary of traditional thinking, sulfate dominated condensational growth for only a small (~18%) fraction of the 62 observed growth events, even fewer than the organic-dominated events (24%). During most (58%) growth events, the major condensing species included both organics and sulfate. Potential precursors of the secondary organics are volatile organic compounds from ocean biological activities and those produced by the air-sea interfacial oxidation. Our results indicate that the condensation of 20 secondary organics contributes strongly to the growth of pre-CCN particles, and thereby the CCN population over remote oceans. https://doi.org/10.5194/acp-2020-625 Preprint. Discussion started: 3 July 2020 c © Author(s) 2020. CC BY 4.0 License.


Introduction
Marine low clouds play an important role in global climate system (Wood, 2012), and their properties and radiative effects are 25 very sensitive to the concentration of cloud condensation nuclei (CCN) (Carslaw et al., 2013;Rosenfeld et al., 2019).
Condensational growth of pre-CCN particles (i.e., particles that are too small to form cloud droplets) (Hoppel et al., 1990;Pierce and Adams, 2006) is one major source of CCN in remote marine boundary layer (MBL) (Pierce and Adams, 2006;Yu and Luo, 2009;Sanchez et al., 2018), and is likely the dominant one in late-spring to fall (Zheng et al., 2018). Over open ocean, dimethyl sulfide (DMS) is the dominant biogenic volatile organic compound (VOC). The major oxidation products of DMS 30 are sulfur dioxide (SO2) and methanesulfonic acid (MSA) (Andreae et al., 1985). Further oxidation of SO2 produces sulfuric acid (H2SO4), which readily condenses onto existing particles and participates in the formation of new particles (Kulmala et al., 2000). It has long been recognized that sulfate produced from DMS oxidation is a major species for particle condensational growth in the remote marine environment (Sanchez et al., 2018). Earlier studies (Willis et al., 2016;Kerminen and Wexler, 1997;Karl et al., 2011) suggest that MSA may also contribute to the growth of pre-CCN particles and thus the formation of 35 CCN. However, the effect of MSA condensation on marine CCN concentration remains unclear. Model simulated effects range from negligible (e.g., a few percent) to significant (~20%) depending on the assumption of MSA volatility and the geographic location (Hodshire et al., 2019).
It has been suggested that in the remote MBL, secondary organics produced from two types of non-DMS VOCs can contribute 40 substantially to particle condensational growth. The first type of VOCs, including isoprene, monoterpenes, and aliphatic amines (Facchini et al., 2008;Dall'Osto et al., 2012;Willis et al., 2017), is related to ocean biological activities, and SOA produced from these VOCs are positively correlated with MSA (Dall'Osto et al., 2012;Willis et al., 2016;Kim et al., 2017;Willis et al., 2017). While the mixing ratios of isoprene and monoterpenes are typically quite low over open oceans (Hu et al., 2013) due to their weak emissions, on rare occasions, elevated monoterpene mixing ratios up to ~100 ppt were observed (Kim 45 et al., 2017), possibly due to enhanced microorganism growth as a result of nutrient replenishment (Kim et al., 2017). The second type of VOCs are produced by the oxidation reactions at the air-sea interface, especially when the sea surface microlayer is enriched in organic surfactants (Mungall et al., 2017;Brüggemann et al., 2018). These water-soluble organics can come from phytoplankton, but can also be from other sources, including other autotrophs and atmospheric depositions (Wurl et al., 2011). Therefore, this type of oceanic VOCs and thus SOA formed may not correlate with MSA (Wurl et al.,50 The Eastern North Atlantic (ENA) atmospheric observatory was established by the Atmospheric Radiation Measurement (ARM) Climate Research Facility (https://www.arm.gov/capabilities/observatories/ena) in October 2013. This remote oceanic site, located on Graciosa Island, Azores, Portugal (39° 5' 30" N, 28° 1' 32" W, 30.48 m above mean sea level) (Mather and Voyles, 2013) straddles the boundary between the subtropics and mid-latitudes in the eastern North Atlantic. The ENA is a region of persistent but diverse marine low clouds, the albedo and precipitation of which are highly susceptible to perturbations 70 of aerosol properties (Wood, 2012;Carslaw et al., 2013). Air masses arriving at this site can originate from North America, Northern Europe, the Arctic, and the Atlantic (Wood et al., 2015;Wang et al., 2016;Zheng et al., 2018). The routine measurements at the ENA site include meteorological parameters, trace gases mixing ratios, and aerosol and cloud properties (Zheng et al., 2018). The relevant routine measurements used in this study are summarized in section 2.3.

75
From June 2017 to Aug. 2018, the Aerosol and Cloud Experiments in the Eastern North Atlantic (ACE-ENA) campaign (Wang et al., 2016) was conducted in the Azores to investigate the aerosol-cloud interactions in the remote marine boundary layer (MBL). As a key part of this campaign, additional aerosol measurements were carried out at the ENA site, including aerosol size distribution and size-resolved CCN activated fractions (Mei et al., 2013c;Thalman et al., 2017). The instruments and calibration procedures are detailed elsewhere (Zheng et al., 2020), and are briefly described below. The data from these 80 measurements are available at https://www.arm.gov/research/campaigns/aaf2017ace-ena.

Size distribution measurements and mode fittings
Aerosol size distribution was measured by a scanning mobility particle analyzer (SMPS, Model 3938, TSI Incorporated, Shoreview, MN, USA). Dry (RH < 25%) aerosol number size distribution ranging from 10 to 470 nm in particle diameter was measured every 8 minutes. In addition, a condensation particle counter (CPC, Model 3772, TSI Incorporated, Shoreview, MN, 85 USA) was operated side-by-side to measure the total aerosol number concentrations (CN) concurrently. The measured aerosol number size distributions are fitted as a sum of up to three lognormal modes. Based on the fitted mode geometric mean https://doi.org/10.5194/acp-2020-625 Preprint. Discussion started: 3 July 2020 c Author(s) 2020. CC BY 4.0 License. diameters (Dp,n), the fitted modes are classified as the nucleation mode (Dp,n < 20 nm), the Aitken mode (20 < Dp,n <~ 80 nm), the accumulation mode (~80 < Dp,n < ~300 nm), and the sea spray aerosol mode (Dp,n > ~300 nm) (Quinn et al., 2017;Zheng et al., 2018). 90

Size-resolved CCN activated fraction measurements
The size-resolved CCN measurement system (SCCN) consists of a Differential Mobility Analyzer (DMA, TSI Inc., Model 3081) coupled to a CPC (TSI Inc., Model 3010) and a cloud condensation nuclei counter (CCNC, Droplet Measurement Technologies, Boulder, CO) (Frank et al., 2006;Moore et al., 2010;Petters et al., 2007;Mei et al., 2013b). This system measures the activated fraction (i.e., the fraction of particles that activate and form cloud droplets) of size-classified particles 95 as a function of super-saturation (Thalman et al., 2017). During the ACE-ENA campaign, the DMA stepped through 6 dry particle diameters (Dp, SCCN) of 40, 50, 75, 100, 125, and 150 nm. At each Dp, SCCN, the super-saturation level inside the CCNC was varied by changing the flow rate and/or temperature gradient △T. The corresponding supersaturation levels, ranging from 0.07% to 1.34% at 298 K, were calibrated using ammonium sulfate particles following established procedures (Lance et al., 2013;Mei et al., 2013a;Thalman et al., 2017). An entire measurement cycle through the 6 particle diameters took between 100 1~2 h, depending on particle number concentration. Temperature dependence of CCNC supersaturation (Rose et al., 2008;Thalman et al., 2017) and the effect of multi-charged particles (Thalman et al., 2017) are taken into consideration. The particle hygroscopicity parameter under supersaturated conditions, κCCN (Petters and Kreidenweis, 2007), is derived from the activated fraction spectrum and the corresponding particle diameter (Lance et al., 2013;Mei et al., 2013a;Thalman et al., 2017).
Gas-phase SO2 and MSA concentrations are from the Modern-Era Retrospective Analysis for Research and Applications, version 2 (MERRA-2) reanalysis data (Gelaro et al., 2017), at the grid corresponding to the ENA site.

Derivation of the hygroscopicity parameter of condensing species 115
Continuous growth of Aitken mode particles is identified from the aerosol size distribution time series. As a result of the condensational growth, aerosol chemical compositions and thus the hygroscopicity of Aitken mode particles are expected to https://doi.org/10.5194/acp-2020-625 Preprint. Discussion started: 3 July 2020 c Author(s) 2020. CC BY 4.0 License. evolve with time during the growth events. As potential condensing species (Table S1) have contrasting hygroscopicity parameters, the variation of hygroscopicity parameter during the growth events can therefore be used to infer the major condensing species. 120

Matching aerosol size modes with the hygroscopicity measurements
Here we detail the procedure to correlate the aerosol size distribution with the SCCN measurements. The same procedure is also applied to correlate the aerosol size distribution with the HTDMA measurements. The CCN activated fraction spectrum was measured at 6 fixed sizes (Dp, SCCN). As the size of Aitken mode particles evolves continuously during the growth events, we first determine if the hygroscopicity of the growing Aitken mode can be captured by the measurement at one of the six Dp, 125 SCCN using the following two criteria. The first criterion is that Dp, SCCN (e.g., 40 nm, 50nm, or 75 nm) is within one geometric standard deviation (σ) of the Aitken mode diameter, i.e., Dp, n σ -1 < Dp, SCCN < Dp, n σ (Fig. 1a). For example, at time t0, Dp, SCCN (40 nm) is within one σ range of the Aitken mode diameter (i.e., dark blue shaded area in Fig. 1a), and the κCCN value measured at 40 nm is considered representative of the Aitken mode (solid blue curve in Fig. 1a). In contrast, at a later time t1, the Aitken mode grew to larger sizes (dash blue curve in Fig. 1a), and 40 nm became smaller than Dp, n σ -1 (light blue shaded area in Fig.  130 1a). Therefore, κCCN measured at 40 nm no longer represents the hygroscopicity of the Aitken mode at t1. The second criterion is that particle concentration at Dp, SCCN is dominated by the Aitken mode only (Fig. 1b), i.e., over 95% of the particles at the measured Dp, SCCN is contributed by the Aitken mode. As an example, both the Aitken mode (blue curve) and the accumulation mode (red curve) contribute to the number size distribution at Dp, SCCN (black dash line, Fig. 1b). Although Dp, SCCN is within one σ of the Aitken mode diameter, the contribution of Aitken mode is less than 95% at this size (orange curve in Fig. 1b). 135 Therefore, measurement at Dp, SCCN is not deemed as representative of the Aitken mode due to the substantial contribution from accumulation mode particles. Only data points that meet both criteria are selected, as illustrated in Fig. 1c. Figure 2a gives an example of the time series of Aitken mode diameter and paired κCCN value during a growth event.

Derivation of the hygroscopicity of condensed species (κc) during growth events
The derivation is applied to condensational growth events when there are sufficient number (> 6 points) of κ measurements 140 that satisfy both criteria described in M2.1. For each condensational growth event selected (e.g., Fig. 2a), the average hygroscopicity parameter of condensing species, κc, is derived based on the following three assumptions. Here, κ represents either the hygroscopicity derived from SCCN (i.e., κCCN) or HTDMA (i.e., κGF) data.
The first assumption is that the change in particle volume (diameter) is due to the condensational growth only, namely: 145 where V is the particle volume and Dp is the particle diameter. Hereinafter we use X1 and X0 to denote the corresponding particle property X after and before the condensational growth, respectively, and Xc refers to the property X of the condensed species. The second assumption is that the aerosol κ follows the volume-weighted mixing law (Petters and Kreidenweis, 2007): https://doi.org/10.5194/acp-2020-625 Preprint. Discussion started: 3 July 2020 c Author(s) 2020. CC BY 4.0 License.
The third assumption is that the growth rate is identical for particles of the same size, and thus the relative position of any given particle in the accumulative size distribution is maintained throughout the growth. Let CDF0 and CDF1 denote the particle cumulative size distributions before and after the particle growth, and Dp0 and Dp1 represent particle diameters before and after particle growth, respectively. The number of particles smaller than Dp1 following particle growth should be the same as the number of particles smaller than Dp0 prior to the growth event (Fig. 2b): 155 For each particle size (i.e., Dp1) measured during the growth events, the original particle size (i.e., Dp0) is derived from Eq. (3).
The volume fraction of condensed species, fV, cond, is given by: By combining Eq. 1-4, we have: 160 Both κ1 and fV, cond are from the measurements as described above. Therefore, κc and κ0 can be derived from the linear fitting of κ1 vs. fV, cond for each growth event (e.g., Fig. 2c), where κ0 is the intercept, and κc is the sum of slope and intercept. The method described here was applied to both SCCN and HTDMA measurements, and κc derived are referred to as κc,CCN and κc,GF hereinafter, respectively. 165 Figure 3 shows two examples of the identified growth events, with the dominant condensing species being sulfate and organics, respectively. While the measured κCCN of the Aitken mode particles (i.e., pre-CCN particles that are below ~ 80 nm) are similar (~0.45) at the start of both events, the variations of κCCN with growing particle size show opposite trends. For the July case ( Fig. 3a,b), κCCN increased with the volume fraction of condensed species (fV,cond, Fig. 3b), indicating that the hygroscopicity 170 of the condensed species, κc,CCN, exceeds that of the original particles. The derived κc,CCN value is 0.7, which is typical of sulfates (Table S1). In contrast, during the September growth event (Fig. 3c,d), κCCN decreased as the particles grew. The derived κc,CCN value is ~0.3, indicating organics as the dominant condensing species. We note that κc,CCN is derived from the volume-weighted mixing law (Petters and Kreidenweis, 2007) (i.e., ideal Zdanovskii, Stokes, and Robinson (ZSR) mixing).

Constraining the major condensing species in remote MBL
Organic surfactants may facilitate CCN activation by lowering surface tension of growing droplets (Ovadnevaite et al., 2017). 175 In scenarios when particles contain organic surfactants, particle hygroscopicity κCCN may be greater than the simple volume average of participating species. As a result, the derived κc,CCN value based on the volume-weighted mixing law may be overestimated, therefore leading to an underestimation of the contribution of organics to the particle condensational growth.

Monthly distributions of the dominant condensing species 185
The monthly distribution of the identified growth events and the dominant condensing species are shown in Fig. 4. Relatively more events were observed during the summer seasons due to favorable synoptic conditions. In summer, there is a stronger influence by the Azores High while the influence from mid-latitude cyclones and the corresponding wet scavenging are much weaker (Zheng et al., 2018). The distribution of the event categories shows that, contrary to the conventional thinking, NH4HSO4/H2SO4 dominated the condensational growth during only 18% of the growth events. This is less than the events 190 dominated by organics at 24%. The majority (58%) of the growth events exhibit intermediate κc,CCN values, suggesting that (NH4)2SO4 or a mixture of organics and sulfate are responsible for the particle condensational growth.
To further constrain the condensing species for the intermediate κc,CCN category, we compare the κc,CCN value with the hygroscopicity under sub-saturated conditions (κc,GF), which is derived from measured particle hygroscopic growth (section 195 3). For (NH4)2SO4, the difference between κc,CCN and κc,GF is relatively small (within 20%) (Petters and Kreidenweis, 2007), while the difference is usually substantially larger (Wex et al., 2009;Rastak et al., 2017;Petters et al., 2009;Pajunoja et al., 2015;Ovadnevaite et al., 2011;Massoli et al., 2010) for organic species. The large difference has been attributed to the solution non-ideality (Petters et al., 2009), the formation of hydrogels (Ovadnevaite et al., 2011), and the solubility and phase states (Pajunoja et al., 2015;Rastak et al., 2017). One example of the intermediate κc,CCN category is shown in Fig. S1. For this case, 200 the derived κc,CCN and κc,GF values are 0.59 and 0.45, respectively (Fig. S1). The difference is close to the measurement uncertainty (i.e., 20%), and therefore the major condensing species for this example is classified as (NH4)2SO4. growth. In addition, chemical composition of sub-micron non-refractory aerosol (NR-PM1; aerodynamic diameters below 1 μm) indicates an ammonium-poor condition over the ENA (color bar in Fig. 5), typical of remote marine environment (Adams et al., 1999). Therefore, sulfate is not fully neutralized as (NH4)2SO4. These evidences suggest that during most of the intermediateκc,CCN events, the condensed species are a mixture of sulfates and organics instead of dominated by (NH4)2SO4.
condensational growth for the low-κc,CCN category. Together, these two categories represent a total of ~80% of the growth events and occurred throughout the year.

Sources of the condensing organics 215
Given the importance of secondary organics to particle condensational growth, the potential sources of the condensing organics are investigated by examining the air mass origins (SI S1). Here we classify the origin of air mass during the growth events into four types: (1) continental air masses from North America or Europe, (2) the Arctic, (3) the subtropical, and (4) the midlatitude Atlantic. Note that an air mass is denoted as continental if it passed over the North America or Europe, so the noncontinental types represent the air masses that had stayed over oceans or clean continental areas (i.e., Arctic region) for at least 220 10 days (SI S1).
Growth events of mid-latitude Atlantic or Arctic type were observed exclusively from May to September, a period that coincides with the phytoplankton blooms in mid-latitude Atlantic or Arctic, but not the subtropics (Sapiano et al., 2012). For these events, κc,CCN is anti-correlated with MSA/SO2 ratio (Fig. 6a), which is from MERRA-2 reanalysis data (section 2.3). As 225 fixed yields of SO2 and MSA from DMS oxidation are assumed in MERRA-2 data (Chin et al., 2000;Randles et al., 2017), a lower MSA/SO2 ratio suggests other SO2 sources in addition to DMS oxidation contribute to these events. These other sources could include volcanic emissions and combustion products from international shipping (Randles et al., 2017). As MSA is a tracer of biogenically derived SOA in marine environment (Seinfeld and Pandis, 2016), the anti-correlation also indicates that the condensed organics are likely SOA produced from VOCs emitted from ocean biological activities (e.g., phytoplankton 230 blooms). The value of κc,CCN is not correlated with the NR-PM1 organic/sulfate ratio (Fig. 6b), suggesting different sources of the condensed species in pre-CCN and the accumulation mode particle composition.
Among the remaining growth events, only four of them are subtropical cases, which occurred outside the bloom periods.
During the other events, air masses were potentially influenced by continental emissions (Fig. S2). For these events, κc,CCN is 235 instead positively correlated with MSA/SO2 ratio (Fig. 6c), indicating that secondary organics formed from phytoplanktonemitted VOCs likely played a minor role in the observed particle condensational growth. The κc,CCN value generally decreases with increasing NR-PM1 organic / sulfate ratios (Fig. 6d), suggesting that the formation of SOA led to increased organic fraction for both pre-CCN and accumulation mode particles. Possible sources of the condensed organics include SOA generated from long-range transported continental VOCs and VOCs released by the sea-surface microlayer oxidation that are 240 not directly related to phytoplankton emissions.
As continentally emitted VOCs are removed by oxidation during long-range transport, it is expected that in-situ SOA production from these VOCs is low and plays a minor role in particle condensational growth over the remote oceans (Kelly et https://doi.org/10.5194/acp-2020-625 Preprint. Discussion started: 3 July 2020 c Author(s) 2020. CC BY 4.0 License. al., 2019;D'Andrea et al., 2013). On the other hand, aromatic compounds were detected in pre-CCN particles in clean air 245 masses at a coastal site , indicating potential contribution of SOA from anthropogenic VOCs with long lifetime. Oxidation reactions at the air-sea interface can produce VOCs, which lead to subsequent SOA formation (Mungall et al., 2017;Brüggemann et al., 2018). This VOC source is present all-year round, even during winter when there is little biological activity in the ocean (Brüggemann et al., 2018). Therefore, the secondary organics produced via this pathway can contribute to the growth of pre-CCN particles outside the biologically active seasons of the ocean. 250

Conclusions
In summary, we show that during all seasons, secondary organics play an important role in the condensational growth of pre-CCN particles, and by extension, the formation of CCN in the remote marine boundary layer. The secondary organic species likely derive from a variety of precursors, including VOCs produced from marine biogenic activity, continentally emitted VOCs with long lifetime that survive the long-range transport, and VOCs formed by oxidation at the air-sea interface. Current 255 global models typically assume that sulfates dominate the particle growth over remote oceans, and therefore may substantially underestimate the formation of CCN by condensational growth in remote marine boundary layer.
Author contributions. J.W. and G.Z. designed the study. J.W. G.Z., C.K., J.U., and T.W carried out the measurements, G.Z. and J.W. conducted the analysis and wrote the manuscript with contributions from all authors. 265 Competing interests. The authors declare no competing interests. Figure 1. Matching aerosol Aitken mode with hygroscopicity measurements at fixed particle diameters. The black dash lines indicate the selected particle size (i.e., Dp, SCCN) at which the hygroscopicity parameter κ is derived. The shaded areas indicate one σ range from the fitted lognormal mode diameter, Dp, n. (a) Mode diameter and hygroscopicity κCCN of the growing Aitken mode. The black circles are fitted mode diameter, Dp, n, and the shaded area indicate the one σ range of the fitted mode. Black line shows the increasing trend of Dp, n, which is used to identify growth events. (b) Derivation of the original particle dimeter (Dp0) at the beginning of the growth event from particle diameter after growth (Dp1) using cumulative particle 475 size distributions. (c) Derivation of κc,CCN through linear fitting of κCCN versus fV, cond.