Aerosol samples were collected at a Korean scientific research station in the Antarctic: King Sejong station (62∘13′ S, 58∘47′ W), which is located at King George Island, Chile (see Fig. S1 of the Supplement). King George Island in the Southern Ocean is 120 km off the coast of Antarctica, and is dominated by pervasive ice caps, with more than 90 % of the island being glaciated. This study examined two aerosol samples S1 and S2 collected on 9 December 2011 in the austral summer and 23 July 2012 in the austral winter, respectively, when the oceanic chlorophyll a levels on the collection days of the samples were quite different, by ∼ 19 times (2.46 vs. 0.13 µg L-1, respectively). The oceanic chlorophyll a levels for water samples collected daily from nearby oceanic water were determined using a fluorometer (TD 700, Turner Design, USA). The detailed description for chlorophyll a determination is given elsewhere (Schloss et al., 2014; Lee et al., 2015). The aerosol samples were collected on Al foil substrates (Sigma-Aldrich, 99.8 % purity) using a three-stage cascade impactor (PM10 Impactor, Dekati Inc.) during daytime at temperatures T=1.1 to 2.1 and -1.9 to -1.5 ∘C and relative humidity RH = 94.1–94.5 and 87.6–92.1 % for samples S1 and S2, respectively. The impactor had aerodynamic cut sizes of 10, 2.5, and 1 µm for stages 1, 2, and 3, respectively, at a 10 L min-1 sampling flow, and individual particles collected at stages 2 and 3 (PM2.5–10 and PM1.0–2.5 fractions with the size ranges of 2.5–10 and 1–2.5 µm, respectively) were examined.

Three-day (72 h) backward air-mass trajectories were obtained using the Hybrid Lagrangian Single-Particle Integrated Trajectory (HYSPLIT) model from the NOAA Air Resources Laboratory's web server (http://www.arl.noaa.gov/ready/hysplit4.html). The back-trajectories for samples S1 and S2 show that the air masses at heights of 500, 1000, and 1500 m above sea level, which originated from the Pacific Ocean, travelled over the Pacific Ocean and passed over Chile, respectively (see Fig. S1 of the Supplement).

Low-Z particle EPMA measurements were carried out by scanning electron microscopy (JSM-6390, JEOL) equipped with an Oxford Link SATW ultrathin window energy-dispersive X-ray (EDX) detector, which has a spectral resolution of 133 eV for Mn Kα X-rays. The X-ray spectra and elemental X-ray maps were recorded using INCA Energy software. An accelerating voltage of 10 kV, a beam current of 0.5 nA, and a typical measuring time of 20 s were used for the X-ray spectral data acquisition using area mode, where the X-ray signals were obtained by scanning electron beam over the entire area of each particle. The net X-ray intensities for the chemical elements were obtained by a nonlinear, least-square fit of the spectra collected using the AXIL program (Vekemans et al., 1994). The elemental concentrations of the individual particles were determined from their X-ray intensities using a Monte Carlo calculation combined with reverse successive approximations (Ro et al., 2003). For the X-ray mapping measurements, an accelerating voltage and beam current are the same as the area-mode measurements, except for a typical measuring time of 30 min. A more detailed discussion of the EPMA measurement conditions can be found elsewhere (Ro et al., 2005, 1999).

The particles collected on Al foil were mounted on the microscope stage of a confocal Raman microspectrometer (XploRA, Horiba Jobin Yvon) equipped with a 100×, 0.9 numerical aperture objective (Olympus). Raman point and mapping measurements were carried out under ambient conditions. Optical images of the particles for relocation were obtained using a video camera. Raman scattering was excited at the 532 nm wavelength using an air-cooled diode laser and detected with a multichannel air-cooled charge-coupled device (CCD) at an 1800 gr mm-1 grating. The excitation laser power delivered to the individual particles was approximately 3 mW using a controlled confocal hole of 300 ∼ 500 µm and a slit, 100 µm in diameter. The spectral ranges of 100–4000 cm-1 were performed with a 5 s acquisition time and five-time accumulation for each measurement. The spectral resolution was 1.8 cm-1 and the spot size of the laser beam at the sample was estimated to be ∼ 1 µm2. The XYZ computer-controlled Raman mapping was performed by obtaining the Raman spectra in point-by-point XY scanning mode with a 1 µm step and a 5 s integration time per pixel. The spectra and images were acquired using Labspec6 software. A more detailed discussion of the RMS measurement conditions for single-particle analysis can be found elsewhere (Sobanska et al., 2012; Eom et al., 2013; Jung et al., 2014).

The ATR-FTIR imaging measurements were performed using a Perkin Elmer Spectrum 100 FTIR spectrometer interfaced to a Spectrum Spotlight 400 FTIR microscope. An ATR accessory using a germanium hemispherical IRE crystal, 600 µm in diameter, was used for ATR imaging. The ATR accessory was mounted on the X–Y stage of the FTIR microscope and the IRE crystal was made to come into contact with the sample through a force lever. A spatial resolution of 3.1 µm at 1726 cm-1 (λ=5.79 µm) is achievable (Van Dalen et al., 2007). A 16 × 1 pixel mercury cadmium telluride (MCT) array detector was used to obtain the FTIR images with a pixel size of 1.56 µm. For each pixel, an ATR-FTIR spectrum, ranging from 680 to 4000 cm-1 with a spectral resolution of 4 cm-1, was obtained from eight interferograms, which were co-added and Fourier-transformed. The position of the crystal on the sample was determined using an optical microscope equipped with a light-emitting diode and a CCD camera, which allowed relocation of the same single particles that had been analyzed using RMS before ATR-FTIR imaging. Spectral data processing was performed using Perkin Elmer Spectrum IMAGE software. A more detailed discussion of the ATR-FTIR imaging measurement conditions for single-particle analysis can be found elsewhere (Song et al., 2010, 2013; Jung et al., 2014).

Figure 1 presents typical secondary electron images (SEIs) of the individual particles on two PM2.5–10 (stage 2) samples collected in the austral summer and winter, where the chemical species comprising each particle, determined from X-ray spectral data, is indicated. All the particles on the images are of a marine origin, having major Na and Cl contents with small quantities of C, O, Mg, K, Ca, S, and/or Si. Overall, approximately 600 particles of samples S1 and S2 examined by low-Z particle EPMA were of a marine origin. Na, Mg, Cl, S, C, and O were present in all the particles, whereas K, Ca, and Si were observed more frequently in the summertime sample S1 than in the wintertime sample S2 (93.6 % vs. 79.4 % relative encountering frequencies for K; 93.9 % vs. 75.5 % for Ca; and 70.1 % vs. 0.7 % for Si, respectively, where the relative encountering frequency (in %) for a certain element is defined as the number of particles containing the element divided by the total number of particles analyzed for a sample.). In particular, Si is present exclusively in sample S1, which might be a good indicator of the phytoplankton influence on the nascent SSAs.

As ambient relative humidity (RH) at the sampling times was higher than 87.6 % and the efflorescence RHs (ERHs) of the inorganic sea salt components (e.g., ERHs of NaCl and CaSO4 are ∼ 45–47 and ∼ 80–90 %, respectively; Gupta et al., 2015; Schindelholz et al., 2014; Xiao et al., 2008), the SSAs would be collected as aqueous droplets at the time of collection. Once exposed at a low RH, e.g., by being either handled under the dry ambient conditions or placed in the vacuum chamber of SEM, they would crystallize fractionally, resulting in their heterogeneous mixing states, as shown in Fig. 1, having bright and crystalline solids, segregated and somewhat dark regions, and elongated rods (indicated by the yellow arrows in Fig. 1), which are more distinctive for the summertime particles. The fractional crystallization of SSAs has also been reported (Ault et al., 2013a; Hara et al., 2013, 2014). To determine the chemical species of the crystalline solids, dark regions, and rods, elemental X-ray and molecular Raman mapping measurements were performed on the same individual SSA particles. Figure 2 presents the SEIs and molecular Raman and elemental X-ray map images of two typical summertime and wintertime SSA particles. As Raman-inactive NaCl and MgCl2 species cannot generate Raman signals, Raman mapping was performed to determine the spatial distributions of CaSO4 (using a Raman signal in the 1000–1020 cm-1 range), Na2SO4 (using a Raman signal in the 985–995 cm-1 range), and organic species (using a Raman signal in the 2800–3000 cm-1 range). X-ray mapping images of Na, Mg, Ca, Cl, S, C, and O are overlaid in different colors on the SEIs. Molecular Raman images look broader than elemental X-ray images as the spatial resolution of Raman mapping (∼ 1 µm) is larger than that of X-ray mapping (∼ 0.1 µm). Especially, Raman images for organic species look more spread than C X-ray map images as the low energy C X-rays generated from underneath are not often detected due to the strong absorption by solid particles sitting above. Nonetheless, the combined Raman and X-ray map image data of Fig. 2a clearly indicate that the upper bright solid (region 1, notated on the SEI of Fig. 2a) of the summertime SSA particle is composed of NaCl, the bottom-right region 3 is a mixture of MgCl2 and organic species (having a somewhat dark appearance due to the low secondary electron yield of organic species), and the two elongated rods are of a mixture of CaSO4 and Na2SO4. The wintertime SSA particle in Fig. 2b is composed of NaCl (at region 1) and the mixture of MgCl2 and organic species (at region 2). As C and O are overlapping in their X-ray maps of Fig. 2, the organic species appear to contain a significant amount of oxygen. Figure 3 shows the X-ray spectra and elemental atomic concentrations obtained from the entire regions of the summertime and wintertime particles using area-mode X-ray data acquisition. The summertime particle contains more C, O, Si, S, and Ca than the wintertime particle. As the amount of sulfate (by assuming all the sulfur exists as sulfate) for the summertime particle is larger than that of Ca, the sulfate first crystallized as CaSO4, and the remaining sulfate crystallized as Na2SO4, resulting in the formation of elongated rods composed of a mixture of CaSO4 and Na2SO4. For the wintertime particle, CaSO4 was observed weakly in the upper-right region because of the low sulfate content.

Table S1 in the Supplement shows the mean elemental concentrations for a total of ∼ 600 individual particles in PM1.0–2.5 and PM2.5–10 fractions of the summertime and wintertime samples, obtained by low-Z particle EPMA. As all the particles analyzed in these samples are of a marine origin, the mean atomic concentrations of Na and Cl are largest (ranging in 25.2–28.3 and 24.8–29.2 %, respectively), followed by high C and O concentrations (18.8–27.1 and 17.3–19.5 %, respectively), compared to those of Mg, Si, S, K, and Ca, which are in the range 0.0–2.9 %. Based on the mean elemental weight concentrations, the C and O contents were smaller based on the mean atomic concentrations, even though they were still considerable (9.6–14.6 and 12.0–13.6 %, respectively). On the other hand, the organic contents on a molecular basis would be smaller than the elemental C contents, but the molecular organic content could not be estimated because the organic molecular species in SSAs have not been identified clearly (Ault et al., 2013b; Laskina, 2015; Quinn et al., 2015). An interesting observation was that all the supermicron Antarctic SSAs, both in the summertime and wintertime samples, were a mixture of inorganic and organic species.

To better examine the chemical compositional contrast between samples S1 and S2, Table 1 lists the mean elemental concentration ratios to Na for individual particles together with those for bulk seawater. The atomic concentration ratios of C, O, Si, S, and Ca; Cl; and Mg and K for the summertime sample were higher and lower than and similar to those of the wintertime sample, respectively (also see Fig. S2, which clearly shows different distributions of individual particles having specific elemental concentration ratios between the summertime and wintertime samples), indicating that C, O, Si, S, and Ca; and Cl are enriched and depleted in the summertime sample, respectively. In addition, those enriched and depleted elements have higher and lower concentration ratios than the bulk seawater ratios, respectively.

As the [C] / [Na] ratios for both samples were high compared to the bulk seawater [C] / [Na] ratio, even the supermicron Antarctic SSAs contain significantly enriched organic species. The [C] / [Na] ratios of sample S1 were higher than those of sample S2, suggesting that the higher organic matter is related to the higher phytoplankton activities, and those for particles in the PM1.0–2.5 fractions of samples S1 and S2 (1.12 and 0.83, respectively) were higher than PM2.5–10 fractions (0.87 and 0.70, respectively), indicating that the smaller particles contain more organic species, which is consistent with other observations reporting more organics in the smaller SSAs (Quinn et al., 2015).

The [O] / [Na] ratios of sample S1 are higher than those of sample S2, and those for particles in the PM2.5–10 fractions of samples S1 and S2 (0.77 and 0.68, respectively) are higher than the PM1.0–2.5 fractions (0.71 and 0.66, respectively). Similar observations were made for S and Ca, for which the elemental concentration ratios were somewhat higher in sample S1 and in larger size fractions (see Table 1). In addition, the frequencies of encountering particles having higher [S] / [Na] or [Ca] / [Na] ratios than bulk seawater were significantly higher in the summertime sample and in the larger size fractions (see the encountering frequency data for S and Ca in Table 1), indicating that O, S, and Ca are interrelated to common sources, which is also supported by the observation of elongated CaSO4 rods in the Raman and X-ray mapping measurements. The enriched S and O in the S1 sample appear to be due to the elevated nss-SO42- levels. In the austral summer (November–March) of the Antarctic, higher solar radiation levels and temperatures than the other seasons tend to enhance the phytoplankton activities (as supported by its high chlorophyll a level for sample S1), which enhances the production and emission of oceanic dimethyl sulfide (DMS) (Wagenbach et al., 1998; Preunkert et al., 2008). The volatile DMS in the atmosphere undergoes complex sequences of gas-phase oxidation reactions, generating a range of sulfur-containing products, such as dimethyl sulfoxide (DMSO), methanesulfonic acid (MSA), SO2, and H2SO4 (Gaston et al., 2010). These oxidized products can condense onto preexisting particles, resulting in the formation of nss-SO42--containing SSAs. As CaSO4 can efflorescence at very high RH, the nss-SO42- can combine easily with Ca, as observed in Fig. 1, where the CaSO4 rods are observed more frequently in sample S1.

Si is observed for the summertime particles, and more abundantly ([Si] / [Na] = 0.03 vs. 0.01) and frequently (encountering frequency = 93.4 % vs. 47.5 %) in the PM1.0–2.5 fraction than in the PM2.5–10 fraction. As Si is observed mostly in sample S1 and more in the smaller size fraction, it appears to be from fragments of silica cell walls of diatoms, a major group of algae and a common type of phytoplankton in the oceans (Litchman and Klausmeier, 2008; Alpert et al., 2015). In the winter, the reduced diatom activities would decrease the emission of Si species into the atmosphere, resulting in the scarce observation of Si in sample S2.

In the SSAs of samples S1 and S2, only Cl is depleted compared to bulk seawater ([Cl] / [Na] = 1.00 and 1.03 for samples S1 and S2, respectively, vs. 1.16 for seawater), and the Cl depletion is somewhat higher for the summertime SSAs than the wintertime and for PM1.0–2.5 fractions than PM2.5–10 fractions, suggesting that Cl was liberated by the reactions of NaCl and/or MgCl2 with nss-SO42- and/or CH3SO3-, which are more abundant in the summer, with more depletion for smaller SSAs having a higher surface-to-volume ratio and higher reactivity.

Based on low-Z particle EPMA analysis, the C, O, Si, S, and Ca levels were elevated for the summertime SSAs on a single-particle basis. This quantitative elemental X-ray analysis provides useful information on their morphology, elemental chemical compositions, and mixing states of individual Antarctica SSAs. On the other hand, as low-Z particle EPMA has a limitation on molecular speciation and hydrogen detection, the RMS and ATR-FTIR imaging techniques were applied in combination for the analysis of the same individual SSAs to investigate their Raman- and IR-active organic and inorganic molecular species. Raman and ATR-FTIR techniques are useful because their spectra of organic and inorganic compounds are quite specific, depending on their chemical species, phase, crystallinity, and neighboring environment. In particular, the complicated vibrational spectral patterns observed in the fingerprint region (< 1500 cm-1) in the Raman and FTIR spectra can be critically useful for the positive or negative identification of specific organic compounds with the same phase and crystallinity. In addition, the differences in their spectra owing to their different signal generation mechanisms (i.e., scattering vs. absorption of energy) and different selection rules would make the two fingerprint techniques rather complementary (Jung et al., 2014).

Table 2 shows relative encountering frequencies of the organic and inorganic species for ∼ 250 individual Antarctic SSAs. The encountering frequency of certain chemical species was determined by counting the number of individual SSAs containing the species, regardless of its content, as the Raman and ATR-FTIR spectral data were used for qualitative molecular speciation. Based on X-ray analysis, C and O were present in all the analyzed Antarctic SSAs. Indeed, organic salt species were detected for all the particles of samples S1 and S2, showing that organic species are ubiquitously present, even in supermicron SSAs. As shown in Table 2, organic salt species were categorized into three groups containing (i) MgAla, (ii) MgFAs, and (iii) mixtures of the two organic salts. The Raman and IR active inorganic salts were always observed together with organic salt species, so that the relative encountering frequencies of inorganic species are shown in each organic group.

All the particles of sample S1 contained only MgAla together with other inorganic species. In particular, CaSO4 and Na2SO4 are mixed almost internally with MgAla (for PM1.0–2.5 and PM2.5–10 fractions, the encountering frequencies of CaSO4 were 98.3 and 92.9 %, respectively, and those of Na2SO4 were 98.3 and 88.6 %, respectively), indicating that SO42- is mostly in the form of a CaSO4 and Na2SO4 mixture. Although a N X-ray signal was not detected, probably due to the small amount of NO3- present in the Antarctic SSAs, Mg(NO3)2 and NaNO3 were frequently observed in samples S1 and S2 using Raman and ATR-FTIR techniques. The nitrate in seawater can be generated by the photoammonification process, which transforms dissolved organic nitrogen (DON) to labile inorganic nitrogen, mainly ammonium (NH4+) (Kitidis et al., 2006; Aarnos et al., 2012; Xie et al., 2012; Rain-Franco et al., 2014; Paulot et al., 2015), followed by the microbial oxidation of ammonium into nitrate (NO3-) by nitrifying bacteria (Carlucci et al., 1970; Hovanec and Delong, 1996; Smith et al., 2014; Tolar et al., 2016). As the photoammonification depends on solar radiations, the ammonium and nitrate production would be enhanced in the summer with a higher solar radiation level. Indeed, as shown in Table 2, nitrates are more frequently observed in summertime sample S1 than wintertime sample S2. For the PM1.0–2.5 and PM2.5–10 fractions of sample S1, the overall encountering frequencies of Mg(NO3)2 are 51.7 and 77.1 %, respectively, and those of NaNO3 are 0.0 and 38.6 %, respectively, where the NO3- moiety was observed more in the PM2.5–10 fraction. The reason why the NO3- moiety is more abundant in the PM2.5–10 fraction is unclear. The encountering frequencies of SiO2 are 46.6 and 27.1 % in the PM1.0–2.5 and PM2.5–10 fractions of sample S1, respectively. SiO2, which would be from fragments of silica cell walls of diatoms, appears to be in colloidal form because SiO2 species are not water-soluble and were observed more in the PM1.0–2.5 fraction than in PM2.5–10. A small number of Mg methanesulfonate was observed only in the PM1.0–2.5 fraction of sample S1. Higher phytoplankton activities in the summer enhance the production and emission of oceanic DMS, resulting in the production of MSA, which is a strong acid that can exist as an anion in seawater and is observed as Mg salts in SSAs, even though its encountering frequency is not high compared to other sulfates.

A significant portion of SSAs of sample S2 contain only MgAla (overall 76.6 and 33.9 % for PM1.0–2.5 and PM2.5–10 fractions, respectively) (see Table 2). Considering the encountering frequencies of MgAla mixed with MgFAs (21.9 and 54.8 % for PM1.0–2.5 and PM2.5–10 fractions, respectively), MgAla is also almost ubiquitous in sample S2 (overall 98.5 and 88.7 % for PM1.0–2.5 and PM2.5–10 fractions, respectively). MgFAs mixed internally with MgAla were encountered significantly in sample S2 (overall 23.5 and 66.1 % for PM1.0–2.5 and PM2.5–10 fractions, respectively). For the PM1.0–2.5 and PM2.5–10 fractions, the encountering frequencies of CaSO4 are 98.5 and 88.6 % overall, respectively, whereas those of Na2SO4 are 26.6 and 8.0 %, respectively, indicating that SO42- is mostly in the form of CaSO4. For the PM1.0–2.5 and PM2.5–10 fractions, the overall encountering frequencies of Mg(NO3)2 are 43.8 and 75.8 %, respectively, and those of NaNO3 are 12.5 and 27.4 %, respectively, where the NO3- moiety was also observed more in the PM2.5–10 fraction. SiO2 was encountered much less frequently, 7.9 and 3.2 % in the PM1.0–2.5 and PM2.5–10 fractions, respectively, compared to those of sample S1 (i.e., 46.6 and 27.1 %, respectively). The observation of a higher encountering frequency of SiO2 in sample S1 is consistent with that of X-ray analysis, where the detection of the Si X-ray signal was 70.1 and 0.7 % for samples S1 and S2, respectively.

The relative encountering frequency data for the organic and inorganic species of samples S1 and S2 clearly show their different chemical compositional features. MgAla-containing SSAs are predominant for samples S1 and S2. The MgFAs were not observed in sample S1, but were observed in sample S2, mostly as internal mixtures with MgAla. As alanine is water-soluble and anions of fatty acids are surfactants, they would be present mostly at the bulk seawater and SSML/sea surface, respectively, before becoming airborne. Therefore, alanine- and fatty acid-containing SSAs are expected to be airborne through jet- and film-drop production during bubble busting, resulting in the generation of supermicron and submicron SSAs, respectively (de Leeuw et al., 2011; Quinn et al., 2015). In this study, supermicron SSAs were investigated for which MgAla is almost ubiquitous in samples S1 and S2, indicating that the supermicron SSAs were generated as jet drops. As MgFAs were observed mostly together with MgAla in sample S2, the MgFA-containing SSAs originating from film drops might agglomerate with MgAla-containing supermicron SSAs in the air.

In a recent mesocosm experiment, the organic matter in SSAs generated from the wave braking of natural seawater was monitored for 29 days after adding nutrients at the beginning of the experiment during which two phytoplankton blooms occurred (Wang et al., 2015). The aliphatic-rich organic matter level in the nascent SSAs was enhanced during the first bloom, whereas the oxygen-rich organic matter level increased at the early period of the experiment before the first bloom and remained somewhat constant thereafter, including the second bloom period. The MgAla and MgFAs observed in this study are the aliphatic-rich and oxygen-rich organic matters in their work, respectively, because the Raman spectra of MaAla and MgFAs are the same as those for oxygen- and aliphatic-rich organic matters and the O / C atomic ratios of alanine, palmitic, and stearic acids are 0.67, 0.13, and 0.11 (in their work, O / C > 0.5 for oxygen-rich organic matters and < 0.25 for aliphatic-rich organic matters). In this study, the summertime Antarctic SSAs contain oxygen-rich organic matter, such as MgAla, whereas the wintertime SSAs contain aliphatic-rich organic matter, such as MgFAs as well as oxygen-rich organic matter. The aliphatic-rich organic matter was observed only during the first bloom in the mesocosm experiment, whereas supermicron MgFA-containing SSAs were encountered only in the wintertime sample S2 collected during no bloom event, suggesting that the chemical features of organic matter in nascent SSAs cannot be correlated consistently with the phytoplankton activity. As microalgae can produce more lipid and less protein under environmental stress, such as limited nutrients and low temperature (Wu et al., 2011; Yu et al., 2009; Olson and Ingram, 1975), MgFAs, which were biodegraded from lipid, may be observed more frequently in the wintertime oligotrophic Antarctic Ocean with a lower temperature.