The Svalbard archipelago includes all landmasses between 74 and 81∘ north and 10 and 35∘ east (Fig. 1). The islands cover 63 000 km2. Ny-Ålesund town is situated on the west coast of the largest island, Spitsbergen, and is 1200 km from the North Pole. It is a central platform for Arctic research.

The observation site is based at the Chinese Arctic Yellow River Station (78∘55′ N, 11∘56′ E) (Chi et al., 2015; Geng et al., 2010). The site is about 2 km away from the Zeppelin observatory station (78.9∘ N 11.88∘ E) run by the Ny-Ålesund Science Managers Committee. On the west coast of the island of Spitsbergen, Ny-Ålesund is a Norwegian research and monitoring infrastructure, hosting national and international research projects and programs. The Norwegian Polar Institute (NPI) runs the Sverdrup Research Station at the coast and Zeppelin Observatory at the Mountain 475 m a.s.l. (above sea level), and Sweden, Germany, France, Italy, Japan, China, England, The Netherlands, South Korea, and India are the other countries to have established long-term programs in Ny-Ålesund (https://www.esrl.noaa.gov/psd/iasoa/stations/nyalesund, last access: 8 July 2019). Two to three samples were regularly collected at 09:00, 16:00, and 21:00 (local time) of each day, with a total of 46 samples during 7–23 August 2012.

A sampler containing a single-stage impactor with a 0.5 mm diameter jet nozzle (Genstar Electronic Technology, China) was used to collect individual particles by the air flow rate at 1.5 L min-1. Aerosol particles were collected onto copper TEM grids coated with carbon film. This sampler has a collection efficiency of 31 % at 100 nm aerodynamic diameter and 50 % at 200 nm, assuming the density of the particles is 2 g cm-3. Sampling times varied from 20 min to 2 h depending on the loading of particles. After collection, each sample was placed in a sealed dry plastic tube and stored in a desiccator at 20±3 % RH for analysis. Ambient laboratory conditions (17 % RH–23 % RH and 19–21 ∘C) are effective at preserving individual hygroscopic aerosol particles and reducing changes that would alter samples and subsequent data interpretation (Laskina et al., 2015). During the sampling period, meteorological data at the sampling site including pressure (P), relative humidity (RH), temperature (T), wind speed (WS), and wind direction (WD) were recorded every 5 min using a pocket weather meter (Kestrel 4500, Nielsen-Kellermann Inc., USA). Sample information including local sampling date and time and various meteorological conditions are listed in Table 1.

Individual particle samples were examined by a JEOL JEM-2100 transmission electron microscopy operated at 200 kV with energy-dispersive X-ray spectrometry (TEM/EDS). TEM can observe the mixing structure of different aerosol components within an individual particle on the substrate because electron beams transmit through the specimen to form an image. EDS spectra are acquired for a maximum time of 30 s to minimize potential beam damage and collect particle X-ray spectra with sufficient intensity. TEM grids are made of copper (Cu) and covered by a carbon-reinforced substrate, so Cu is excluded from the quantitative analyses of the particles. Because of the substrate contribution, C content in the TEM grid coated by carbon film might be overestimated in EDS spectra of individual particles.

The distribution of aerosol particles on TEM grids was not uniform, with coarser particles occurring near the center and finer particles on the periphery. Therefore, to ensure that the analyzed particles are representative, five areas were chosen from the center to the periphery of the sampling spot on each grid. Through a labor-intensive operation, 2002 aerosol particles with diameter <10 µm in 21 samples were analyzed by TEM/EDS (Table 1). To check the elemental composition of individual particles, EDS was manually used to obtain elemental spectra of individual particles. In the clean Arctic air, there are simple particle types including sea salt, sulfate, soot, OM, and mineral. Because soot particles have chain-like aggregation, it is not necessary to check their elemental composition. Sea salt particles display spherical or square shapes and are stable under the electron beam in TEM (Chi et al., 2015). Sulfate particles are spherical but flat on the substrate and produce unstable bubbles under the electron beam (Buseck and Posfai, 1999). TEM observations can also identify sulfate particles or sulfate with OM coating. TEM/EDS analysis is very time-consuming. Thus, we did not check the composition of every single particle analyzed. Instead, we randomly checked the elemental composition of 20–30 particles in each sample (Table 1). EDS spectra of 575 particles were manually selected and saved in the computer for elemental composition analysis. Particles examined by TEM were dry at the time of observation in the vacuum of the electron microscope. In our study, the effects of water and other semi-volatile organics were not considered as they evaporated in the vacuum.

Elemental mapping and line profiles of selected individual aerosol particles were also obtained from the EDS scanning operation mode of TEM (STEM). The STEM information clearly display elemental distribution in the targeted individual particles which cannot be provided by the above EDS analysis. Based on preliminary individual analysis, we further chose the typical samples containing abundant sulfate with OM coating for the STEM analysis. High-resolution elemental distribution in individual particles provides a detailed mixing structure of sulfate and OM in individual particles.

The iTEM software (Olympus soft imaging solutions GmbH, Germany) is an image analysis platform for electron microscopy. In this study, it was used to manually or automatically obtain area, perimeter, and equivalent circle diameter (ECD) of individual particles. In these analyzed samples, we found there were abundant sulfate particles (∼30 % by number) in the 11 samples collected during 9–15 August 2012. In other samples, there are more sea salt particles with few particles. Based on the TEM observations, we selected the samples containing more sulfate particles for further microscopic analyses (see below).

Because the sulfate particles in different samples collected in the Arctic had similar elemental composition and mixing state from the TEM observations, we selected three samples (Table 1) for nanoscale secondary ion mass spectrometry (NanoSIMS) analysis (CAMECA Instruments, Geneviers, France). A micro-cesium source was used to generate Cs+ primary ions, with an impact energy of 16 kV for sample interrogation. The primary beam was stepped across the sample to produce element-specific quantitative digital images. The Cs+ primary ion beam was used to obtain 16O-, 12C14N-, 14N16O-, 32S-, 35Cl-, and 16O23Na- ions in this study. The NanoSIMS analysis can perform ion mapping of particles with nanometer spatial resolution over a broad range of particle sizes (Fig. S1 in the Supplement). Because the substrate of TEM grid is carbon, CN- is adopted to represent OM in individual particles (Chi et al., 2015; Ghosal et al., 2014). S- is used to infer the presence of sulfates in individual particles (Li et al., 2017). A total of 32 sulfate-containing particles were analyzed by the NanoSIMS.

We used scanning electron microscopy (SEM) (Zeiss, ULTRA 55) with EDS to examine the vertical distribution of OM and sulfate in individual particles. TEM grids were mounted onto an aluminum SEM stub and directly observed in secondary electron image mode. SEM analysis was operated at 10 kv of extra high tension (EHT) and 9.7 mm of work distance. Processes such as sample moving, analysis region selection and imaging were controlled by computer. The specimen stage in SEM was tilted at the range of 0–7∘, and then we vertically observed the thickness of OM coating and sulfate core on the substrate. Two typical samples that contain abundant sulfate particles were chosen for this analysis (Table 1).

Atomic force microscopy (AFM) with a digital nanoscope IIIa instrument operating in the tapping mode was used to observe surface morphology of individual aerosol particles and measure particle thickness. The tapping AFM has a cantilever and conical tip of 10 nm radius. By using AFM, a general image of the particles is taken at 10 µm full scan size, which generally includes 1–2 particles depending on the exact location. In this study, we are only interested in the sulfate-containing particles. AFM provides surface information and morphology of 17 particles but no composition. Samples were firstly quickly examined by the TEM under low magnification mode to find sulfate-containing particles. Because TEM grids have coordinate letters, we can find the same particles in the AFM. This analysis provides a 3-D image of the sulfate-containing particles and their volume. After we obtained AFM images of sulfate particles, the nanoscope analysis software can automatically obtain bearing area (A) and bearing volume (V). This can then be converted into equivalent circle diameter (d, ECD) and equivalent spherical diameter (D, ESD). 1A=43πr2=πd23→d=3Aπ2V=43πr3=43×πD38→D=6Vπ3 ESD and ECD of sulfate-containing particles are well corrected. Using the linear correlation equation (ESD = 0.38 ECD), we can then correct the ESD of individual particles from TEM analysis to obtain the ECD (Chi et al., 2015).

The refractive index used for the non-light-absorbing sulfate component was set to m=1.55 at 550 nm (Seinfeld and Pandis, 2006). The refractive index of OM (as BrC) is not known so we considered three scenarios: strongly absorbing (1.65-0.03i at 550 nm), moderately absorbing (1.65-0.003i at 550 nm), and non-absorbing OM (1.65 at 550 nm) (Feng et al., 2013). Here, we choose 550 nm as a case study to test how OM coating influences sulfate particles in Arctic air.

BHCOAT Mie code by Bohren and Huffman (1983) was used to calculate the optical properties, including scattering cross section (SCS), absorption cross section (ACS), and single scattering albedo (SSA), assuming a core–shell structure. We firstly calculated these parameters assuming a sulfate core and OM shell structure only (ignoring some of the particles that contain soot core). Because the Mie code can only calculate the core–shell structure or homogeneous model, we assume sulfate as a core and OM as a shell in individual particles to build the core–shell model. Based on the core–shell standard mode (Li et al., 2016), we can calculate optical properties of individual internally mixed particles.

Three-day (72 h) back trajectories of air masses were generated using a Hybrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT) model at the Chinese Arctic Yellow River Station during August 2012. Here we selected an altitude of 500 m as the end point in each back trajectory (Fig. 1). Back trajectories for 50 and 1000 m a.s.l. are similar.

A lagrangian particle dispersion model FLEXPART-WRF 3.1 (Brioude et al., 2013) was also used to examine the origin of particles. The FLEXPART-WRF model uses meteorological parameters from WRF dynamical simulation. The domain resolution is 5050 km with 64 vertical levels. The FLEXPART-WRF simulations were launched in a backward mode over 10 d, with the Chinese Arctic Yellow River Station as an origin. For each simulation (one per sample), 20 000 pseudo-particles were released in a small volume around the station position. Each single particle position evolution backward in time was determined by Lagrangian dispersion calculation.

Based on the TEM experiments and back trajectory of air masses (Fig. 1), we found that there were more S-rich with OM coating particles in the samples collected on 11, 12, 14, and 15 August 2012. Therefore, we further did the FLEXPART-WRF simulation of these four days (Fig. 2). The emission intensity in the Arctic area has been also shown in Fig. S2.

TEM/EDS analysis shows that O, Na, S, and Cl are dominant elements in individual particles (Fig. S3). On the basis of the composition and morphology of individual particles, we classified the particles into four major groups: Na-rich (i.e., NaCl, Na2SO4, and NaNO3), S-rich (i.e., ammonium sulfate and sulfuric acid), carbonaceous (soot and OM), and mineral dust particles. The classification criteria of different particle types and their sources have been described in a separate study (Li et al., 2016). S-rich particles representing secondary inorganic particles (e.g., SO42-, NO3-, and NH4+) are converted from gaseous SO2, NOx, and NH3. OM can be divided into primary organic matter (POM) and secondary organic matter (SOM). SOM is produced from the chemical oxidation of volatile organic compounds (VOCs) and often exhibits OM coating on S-rich particles (Li et al., 2016; Moffet et al., 2013; Riemer et al., 2019). Na-rich particles in the marine air are from sea spray and have typical near-cubic shape. Mineral dust particles from natural ground soil contain Si, Al, or Ca and have irregular shape. Soot particles, which contain C with minor O, appear as a chain-like aggregate of carbon-bearing spheres. Chi et al. (2015) studied the aging mechanism of sea salt particles. Here, we focused on non-sea-salt (NSS) particles including S-rich, soot, and OM particles. NSS particles and sulfate-containing particles account for 39±5 % and 29±7 % by number of all the 2002 particles analyzed (Fig. 3).

TEM observations revealed a common core–shell mixing structure in fine sulfate-containing particles (Fig. 4a). Elemental mapping of such internally mixed sulfate particles shows C signals in the coating (C map, Fig. 4b) and S and O signals in the center (S and O map, Fig. 4c, d). The elemental line profile of a sulfate particle also shows sulfate core and C coating (Fig. S4). Furthermore, ion maps of individual particles from the NanoSIMS further show 12C14N- signals in the coating (red color in Fig. 4e, f) and 32S- signals in the core (green color in Fig. 4e, g). These results provide strong evidence that the coating is OM and the core is sulfate.

A majority of the 781 analyzed NSS particles (74 % by particle number) have a sulfate core and OM coating (Figs. 4 and 5). Approximately 20 % of them also contain small soot inclusions but they only appeared in organic coating, rather than as the core (Fig. 5b). The mixing structure is different from our previous findings in polluted air where soot is normally mixed with sulfate instead of OM coating (Li et al., 2016). Moreover, we observed some chain-like soot aggregates (1.3 % in all analyzed particles) (Fig. S5) but they only occurred in three samples during the whole sampling period (Table 1). Considering the remoteness of the sampling site, such fresh soot particles are likely to be of local origin, for example, shipping and flaring (Gilgen et al., 2018; Peters et al., 2011).

TEM observations showed that some sulfate-containing particles had the unique morphology of a sulfate particle being surrounded by some smaller particles (Fig. 5a). They are often called “satellite” particles as they were distributed from the central particles which impacted on the substrate during sample collection. Satellite particles were observed around 16 % of the analyzed sulfate-containing particles (Fig. 5a) in the samples (Table 1) collected during 9–15 August. NanoSIMS analysis further showed that the satellite particles (Table 1) have strong 32S- (Fig. 6a, c) and 16O- signals (Fig. 6d) as well as weak 12C14N- signals (Fig. 6a, b). Previous studies showed that the similar satellite particles are normally considered as acidic sulfate (Buseck and Posfai, 1999; Iwasaka et al., 1983). Our results show that these acidic satellites not only contain sulfuric acid but also some OM or organic acids. Indeed, Fu et al. (2008) found that polyacids are the most abundant organic compounds, followed by phthalates, aromatic acids, and fatty acids in Arctic aerosol particles. Based on the back trajectories of air masses and FLEXPART modeling, most air masses originate in the North America and Greenland during the sampling periods (Figs. 1 and 2). Figure 1 shows that these air masses brought abundant sulfate-containing particles into the sampling area in summertime.

AFM was used to obtain 3-D images of individual secondary particles impacting on the substrate. Figure 7a shows that the secondary particles normally have a smooth surface which is different from uneven surface of the Arctic fresh and aged NaCl particles (Chi et al., 2015). Furthermore, we observed particle thickness through tilting the specimen stage up to 75∘ in SEM. Figure 7a and b both show that the secondary particles look like thin pancake sticking on the substrate. Furthermore, the sections of two secondary particles in the AFM images show that the highest heights of particles are only 15 % (green line) and 26 % (red line) of the corresponding horizontal diameter (Fig. 7a). These results show that the shape of individual particles was modified when they impacted on the substrate following the airflow. Therefore, the measured ECDs of individual particles in TEM images are much larger than the real particle diameter. To calibrate the particle diameter, we obtained the volume of dry particles on the substrate and then calculated their ESD in the AFM images (Fig. 7c). ESD distribution of the secondary Arctic particles displayed a peak at 340 nm, ranging from 100 to 2000 nm (Fig. 7d). The core particles, as sulfate or soot, had a peak at 240 and 120 nm, respectively (Fig. 7d). It is estimated that OM on average accounted for 63±23 % of the dry-sulfate-containing particle volume. Our result shows that the OM volume increases following the particle size increase (Fig. S6).

Lagrangian particle dispersion modeling using the FLEXPART-WRF 3.1 showed that air masses arriving at the sampling site during our field measurement periods likely originated from the Greenland and North America (Fig. 2). Previous studies reported that air masses from North America or Greenland during the summer contain higher concentrations of black carbon, OM, and sulfate (Burkart et al., 2017; Chang et al., 2011; Fu et al., 2008; Moore et al., 2011; Park et al., 2013). Indeed, there is strong emission intensity of OC and SO2 around the Arctic area from the emission simulation as shown in Fig. S2. However, Weinbruch et al. (2012) observed soot particles when cruise ships were present in the area around Ny-Ålesund town. It is possible that minor soot particles are from the ship emissions and most of them are transported from outside the Arctic area in the free troposphere (Fig. S2).

The sulfate core–OM shell structure observed in the Arctic summer atmosphere is similar to those in the background or rural air in other places (Li et al., 2016; Moffet et al., 2013). Based on the images from electron microscopy, we can infer that OM coating thickness in the Arctic air was comparable with that in rural places but higher than that in urban places. During transport, organic coatings on sulfate were considered as the SOM and their masses increase following particle aging and growth (Li et al., 2016; Moffet et al., 2013; Sierau et al., 2014). Figures 1 and 2 show that most particles in the air masses are transported from North American. The sulfate or OM particles with soot inclusions are probably formed in a similar way to those found elsewhere (Li et al., 2016) – e.g., soot particles may have acted as nuclei for secondary sulfate or organic uptake during their transports (Riemer et al., 2009). Similarly, besides the OM coating in the Arctic particles, Leck and Svensson (2015) found biogenic aerosols like gel aggregate containing bacterium in ultrafine particles. However, we did not find any gel-like particles in the samples because our sampler had very low efficiency for ultrafine particles.

TEM images show that most of the internally mixed sulfate-containing particles display sulfate core and OM coating on the substrate (Figs. 4a and 5b, c). Knowledge of the phase separation in individual particles is important to understand particle hygroscopic properties, heterogeneous reactions of reactive gases on the particle surface, and organic aging (You et al., 2012). It is possible that the thick OM coatings were consistently built up during the long-range transport of sulfate-containing particles and part of the SOM in the coating likely formed in Arctic area. Indeed, there are various sources of organic precursors during the Arctic area, such as biogenic VOCs from ice melting and open water (Dall'Osto et al., 2017) and anthropogenic VOCs from shipping emissions in summertime (Gilgen et al., 2018). The dependence of OM volume on particle size (Fig. S6) suggests that the suspended sulfate particles are initially an important surface for SOM formation. Moreover, the presence of OM coating in 74 % sulfate particles indicates that SOM on the surfaces of fine particles may govern the possible heterogeneous reactions between reactive gases and sulfate-containing particles in the Arctic air.

It should be noted that most SOM not only occurred on the surfaces of sulfate-containing particles but also its mass (mean mass at 63±23 %) dominated in individual particles (Fig. 7d). The OM dominating in individual particles can influence the IN and CCN activities of secondary sulfate-containing particles (Lathem et al., 2013; Martin et al., 2011). For example, some studies found that an increase in organic mass fraction in particles of a certain size would lead to a suppression of the Arctic CCN activity (Leck and Svensson, 2015; Martin et al., 2011). Moreover, OM on particle surfaces can significantly influence hygroscopicity and IN activity of sulfate-containing particles (Wang et al., 2012).

The internal mixing of soot, sulfate, and OM can change optical properties of individual particles in the atmosphere. Recent studies showed that BrC has been detected in the OM in the polluted and clean air and even in the upper troposphere (Laskin et al., 2015; Wang et al., 2018). Feng et al. (2013) further calculated the contribution to be up to 19 % of the optical absorption of the strongly absorbing BrC in global simulations, which is second to the absorbing BC aerosols. Various colored OM (e.g., nitrated/polycyclic aromatics and phenols), referred to as BrC, were detected in the Arctic atmosphere in different seasons (Fu et al., 2008; Wöhrnschimmel et al., 2013; Zangrando et al., 2013) and in surface ice or snowpacks (Browse et al., 2013; Doherty et al., 2013; Hegg et al., 2010). We also noticed that the 12C14N- signal occurred in all analyzed OM coating in sulfate-containing particles (Fig. 4e–f). 12C14N- from NanoSIMS is indicative of the presence of nitrogen-containing organics in the detected materials (Herrmann et al., 2007). Figure 4 showed that the nitrogen-containing OM was more or less homogeneously distributed in the OM coating in individual particles. This suggests that some of the OM in the coating has the potential to act as BrC (Jiang et al., 2019; Laskin et al., 2015).

To understand how OM coating influences optical properties of sulfate-containing particles, we make an assumption that OM coating is strongly absorbing (case 1), moderately absorbing (case 2) or non-absorbing BrC (case 3) with a refractive index of 1.65-0.03i, 1.65-0.003i, and 1.65 at 550 nm, as in Feng et al. (2013). Based on the size measurements shown in Fig. 7d, we can calculate the volume of sulfate and OM within each particle. We input the volume of each component and the corresponding refractive index into the Mie code and then calculated optical properties of individual sulfate-containing particles in the samples. Based on the optical data statistics of 575 particles, Fig. 8a shows that if the OM coating is strongly absorbing BrC (referred to case Abs1), the average absorption cross section (ACS) of individual particles is estimated to be 2.67×10-14 m2. This value is 8.30 times higher than the aerosol ACS (3.22×10-15 m2) when assuming that the BrC is moderately absorbing (referred to as case Abs2, Fig. 8a). However, the scattering cross section (SCS) of individual particles only shows a small change (Fig. 8b). Figure 8c also shows that the SSAs of individual particles are 0.92, 0.99, and 1 when assuming the OM to be strongly, moderately and non-absorbing BrC (cases SSA1 to SSA3). These results suggest that whether we consider organic coating as BrC may have a significant influence on the absorption properties of individual sulfate-containing particles.

In this study, we explored the relationship between the ACS of individual particles and particle diameters. Interestingly, Fig. 8d shows that the ACS of individual fine OM-coating sulfate particles increased with particle size. This suggests that the ACS of individual particles may increase as they grow and age in the atmosphere.

Current climate models estimated the radiative force of Arctic BC (Sand et al., 2013; Shindell, 2007; Winiger et al., 2017; Zanatta et al., 2018), but none specifically considered the optical properties of Arctic BrC. Our study revealed the OM coating on individual sulfate particles, which should be considered in aerosol radiation effect and cloud–aerosol interaction simulations in the models.