Compositions and mixing states of aerosol particles by aircraft observations in the Arctic springtime, 2018

Aerosol particles were collected at various altitudes in the Arctic during the Polar Airborne 15 Measurements and Arctic Regional Climate Model Simulation Project (PAMARCMiP 2018) conducted in the early spring of 2018. The composition, size, number fraction, and mixing state of individual aerosol particles were analyzed using transmission electron microscopy (TEM), and their sources and transport were evaluated by numerical model simulations. We found that sulfate, sea-salt, mineral-dust, K-bearing, and carbonaceous particles were the major aerosol constituents and were internally mixed. The number 20 fraction of mineral-dust and sea-salt particles decreased with increasing altitude. The K-bearing particles increased within a biomass burning (BB) plume at altitudes > 3900 m, which originated from Siberia. Chlorine in sea-salt particles was replaced with sulfate at high altitudes. These results suggest that the sources, transport, and aging of Arctic aerosols largely vary depending on the altitude and airmass history. We also provide the occurrences of solid-particle inclusions (soot, fly-ash, and Fe-aggregate particles), 25 some of which are light-absorbing and potential ice-nucleating particles. Our TEM measurements revealed, for the first time, the detailed mixing state of individual particles at various altitudes in the Arctic. This information facilitates the accurate evaluation of the aerosol influences on Arctic haze, radiation balance, cloud formation, and snow/ice albedo when deposited. 30 https://doi.org/10.5194/acp-2020-1114 Preprint. Discussion started: 2 November 2020 c © Author(s) 2020. CC BY 4.0 License.

only consists of its representative composition with the recognition that most particles are internally mixed and contain several components. Elemental mapping images of representative particles were also acquired using STEM-EDS. Details of the STEM-EDS analysis have been provided elsewhere (Adachi et al., 2019(Adachi et al., , 2020. https://doi.org/10.5194/acp-2020-1114 Preprint. Discussion started: 2 November 2020 c Author(s) 2020. CC BY 4.0 License. and the classification of the BC sources. In the current study, the simulation period was extended to 2018 (from January 2008 to December 2018), and the BC concentrations were further divided into those 165 originating from anthropogenic and BB sources. We used an atmospheric general circulation model (AGCM) with land processes (MRI-AGCM3.5) and the Model of Aerosol Species in the Global Atmosphere mark-2 (MASINGAR mk-2) as atmospheric and aerosol component models, respectively.
The model treats non-sea-salt sulfate, BC, organic carbon, sea salt, mineral dust, and aerosol precursor gases. The model uses a horizontal resolution with an approximately 120-km grid (TL159) and 80 170 vertical layers from the surface to the model top at 0.01 hPa in a hybrid sigma-pressure coordinate system. The horizontal wind fields were nudged toward the 6-hourly Japanese 55-year Reanalysis data (Kobayashi et al. 2015) to reproduce realistic meteorological fields in the simulations. We used the daily BB emissions from the global fire assimilation system dataset of Kaiser et al. (2012) and the monthly anthropogenic emissions dataset of Lamarque et al. (2010). In addition to the baseline 175 simulation, we performed a model sensitivity simulation without BC emissions from BB sources (i.e., anthropogenic BC). The BC concentration from BB sources (hereafter referred to as BB BC) was estimated by subtracting the anthropogenic BC concentration (the sensitivity simulation) from the total BC concentration (the baseline simulation). The emissions of mineral dust and sea salt were calculated based on the meteorological conditions in the simulations ( Tanaka and Chiba, 2005;Yumimoto et al., 180 2017).

Number fraction and size of the major aerosol types
The aerosol particles in our samples mainly consist of sulfate, sea-salt, mineral-dust, K-bearing, and carbonaceous particles. They are mixed at the single-particle scale and sometimes contain inclusions. 185 We measured the particle-size distribution and particle number fraction of each aerosol type (Figs. 2 and 3). We classified the samples based on the sampling altitude below 1000 m and above 1000 m. The former was mainly influenced by local emissions, and the latter included LRT aerosol particles.
The median diameter of all particles is 470 nm (σ = 410) in the AED (Fig. 2). The sulfate, K-bearing, and carbonaceous particles exhibit similar size distributions (the median AEDs are 470, 450, and 400 190 nm, respectively), whereas the carbonaceous particles demonstrate a sharper peak than do the others.
The mineral-dust and sea-salt particles attain larger and broader size distributions than do the others with median diameters of 680 and 740 nm, respectively.
Sulfate particles are the most abundant in all samples except the sample collected on 31 March, 14:40, which predominantly contains sea-salt particles. The sulfate number fraction ranges from 29% to 86% 195 (59% on average) (Table S1; Fig. 3). The second most dominant particle types are sea-salt and https://doi.org/10.5194/acp-2020-1114 Preprint. Discussion started: 2 November 2020 c Author(s) 2020. CC BY 4.0 License. carbonaceous particles for the samples collected at < 1000 m and > 1000 m, respectively. Potassiumbearing particles are relatively abundant (accounting for > 10% of the number fraction) in three samples collected at > 3900 m on 2, 3, and 4 April (Figs. 4 and 5). Based on the TEM observations and model simulations, we found that the above K-bearing particle-rich samples were collected from air masses 200 influenced by BB, and we denoted them as BB samples (please refer to section 3.2 for the transport of the BB plume).
The size-dependent number fractions (Fig. 3) indicate that sulfate particles dominate the < 1 µm particles in the samples collected below 1000 m. In the samples collected above 1000 m, sulfate particles dominate all bin sizes except the largest size bin (> 2 µm), which only contains 12 particles 205 and attains a low statistical significance ( Fig. 3 (b)). The fraction of mineral-dust particles increases with increasing particle size. The number fractions of the sea-salt and carbonaceous particles increase with increasing and decreasing particle size, respectively. In contrast to the negligible contributions of the K-bearing particles in the samples collected below 1000 m, their number fraction in the samples collected above 1000 m is ~10% in all size bins except the largest one. 210 The number fraction of the aerosol types varies depending on the sampling altitude. The mineral-dust and sea-salt number fractions decrease with increasing sampling altitude (Fig. 4). The K-bearing particle fraction is high in the BB samples collected above 3900 m.

K-bearing particles and biomass burning plume
K-bearing particles are the second dominant type (11-37% in number fraction) in the BB samples ( Fig.   5 and Table S1). The K-bearing particles mainly occur as potassium sulfate mixed with organic matter, soot, or both (Fig. 6). Other than in the K-bearing particles, K is detected as a minor component in 220 approximately half of all particles and 76% of the soot-bearing particles.
BB smoke contains a large amount of K in addition to other aerosols, such as organic matter and soot particles (Reid et al., 2005), and K can be used as a tracer element of BB. The occurrences of the Kbearing particles in the BB samples ( Fig. 6) are similar to those from BB in other areas (e.g., Li et al., 2003). 225 We used the modeled BB BC to identify the transport pathways of the BB samples (Fig. 7). The BB BC emissions in the model simulations show that the BB source is southeast Siberia (Fig. S2), where forest BB frequently occurs (Brock et al., 2011;Warneke et al., 2009;Schulz et a, 2019). During the campaign, BB started on approximately 20 March in the area and persisted for several months. On 28 March, the emitted BB BC plume was driven toward the northeast by a low-pressure system. 230 https://doi.org/10.5194/acp-2020-1114 Preprint. Discussion started: 2 November 2020 c Author(s) 2020. CC BY 4.0 License.
Thereafter, the BB plume was lifted and subsequently transported toward the North Pole until 1 April ( Fig. S2). On April 2, 3, and 4, a part of the BB plume approached the Fram Strait at ~600 hPa (~4000 m), where we conducted the sampling. Although the model simulations estimated the BB contributions only for the BC concentrations, we interpret that the K-bearing particles and other BB emissions were also transported along with BC as seen in the TEM results. In terms of the BB samples, the sampling 235 points were > 4600 km away from the BB sources, and the samples had aged for a week or more.

Sea-salt particles
Sea-salt particles are globally abundant, especially over oceans, and considerably influence the climate as CCN and via sunlight scattering ( Lewi and Schwartz, 2004). They are formed from seawater film droplets over the open ocean and local leads in the Arctic. When the sea freezes, sea-salt particles form 240 on frost flowers over sea ice (Hara et al., 2017;Xu et al., 2016) and from blowing snow (Huang and Jaeglé, 2017). Our samples could originate from these sources as our research flights flew over both open water (near Svalbard islands) and sea ice (near Greenland) (Fig. 1). Original sea-salt particles mainly consist of sodium chloride as well as other inorganic salts (e.g., Mg, Ca, and K as chlorides or sulfates). In the atmosphere, the composition of sea-salt particles is altered through the reactions with 245 acidic gases, thus forming sodium sulfate or nitrate ( Adachi and Buseck, 2015;Gard et al., 1998;Yoshizue et al., 2019).
In our samples, the sea-salt particles are complex mixtures of, for example, sodium chloride, sodium sulfate, magnesium sulfate, Mg-C-O, calcium sulfate, and others ( Fig. 8 and Fig. S3). The sea-salt occurrences are similar to those found at a ground site in Svalbard (Chi et al., 2015). Although nitrate 250 can react with sea-salt particles, N was rarely detected in our sea-salt particles, possibly because the nitrate fraction relative to that of sulfate is limited in spring (Brock et al., 2011;Fenger et al., 2013). It is also possible that the measured particle sizes are too small for nitrate to retain as a particle phase and that nitrates are lost from the TEM samples after sampling because of their high volatility. Some Mg occurs around NaCl cores ( Fig. 8a-b). These mixtures may form either in the atmosphere or 255 on the substrate when liquid particles change into solid phases after sampling. Such Mg occurs with C and O as an amorphous phase, suggesting that they constitute organic matter. Similar organic Mg-C coatings on sea-salt particles have been observed in the Arctic (Chi et al., 2015). As Mg salts exhibit a high hygroscopicity, it is possible that they are liquid in the atmosphere and absorb organic matter either from anthropogenic or marine sources (Shaw et al., 2010). Some sea-salt particles demonstrate 260 homogeneous distributions of Mg, S, and Na ( Fig. 8c and Fig. S3e-f). These particles commonly attach soot particles, and Cl is replaced with sulfate, suggesting that they are well aged in the atmosphere.
Most Cl in the sea-salt particles was replaced with sulfate in the samples collected above 1000 m (Fig.   S4). Such Cl loss in NaCl at high altitudes in the Arctic has also been observed by Hara et al. (2002).
Mg and Ca are correlated with Na in all samples, although their weight percent is lower in the samples 265 collected above 1000 m than those collected below 1000 m (Fig. 9). This result suggests that most Mg, Ca, and Na originates from sea-salt particles and that their weight percent within the individual particles decreases with increasing sampling altitudes because of the condensation and coagulation of sulfate or other material on the sea-salt particles.
In the model simulations, the sea-salt concentrations near the sea surface are high (Fig. S5). On 30 and 270 31 March, the sea-salt particle concentrations are higher than those on the other sampling days based on the TEM analysis (Table S1), and the model simulation indicates that the sea-salt particles are transported from the north.

Mineral-dust particles
Certain types of mineral-dust particles (e.g., feldspar) act as INPs ( yields only a small amount of LRT mineral dust above ~700 hPa in the sampling area (Fig. S7). Longterm ground observations of the aerosol composition at station Nord have indicated small contributions from soil particles in the early spring, although they exhibit peaks in summer (Nguyen et al., 2013;Heidam et al., 1999). McNaughton et al. (2011) measured the dust mass concentration over the Western Arctic in spring during the ARCTAS/ARCPAC 2008 campaigns and demonstrated that the mineral-dust suggests that mineral-dust particles could be emitted from local ground surfaces, but the source of the mineral-dust particles in our samples cannot be confirmed. As the number fractions of the mineral-dust 300 particles are nonnegligible (4% on average), further observations are necessary to identify their sources.

Sulfate particles
Sulfate particles are one of the most dominant aerosol species in the Arctic (Brock et al., 2011;Hara et al., 2003;Matsui et al., 2011). This study also showed that sulfate particles were the most dominant species (~60%) in their number fraction (Fig. 3). The sulfate particle number fractions are not well 305 correlated with the altitude (R 2 =0.16), suggesting that the detected sulfate particles and their precursors originate from both marine surface sources (Willis et al., 2017) and LRT at high altitudes. Sulfate commonly mixes with other species, and S is detected in almost all sampled particles. Sulfate particles generally exhibit various chemical forms, including ammonium sulfate, sodium sulfate, potassium sulfate, and calcium sulfate (Figs. 6 and 10; Fig. S3). The sulfate particles commonly have satellite 310 structures around their rims in the TEM images (Fig. 11), suggesting that sulfuric acid and ammonium bisulfate contribute to the sulfate particles (Kojima et al., 2004). This result is consistent with other measurements in the Arctic that show contributions of acidic sulfate (Brock et al., 2011;Hara et al., 2003;Fisher et al., 2011).

315
Carbonaceous particles primarily consist of C and O and include secondary and primary organic aerosols, soot particles, and tarballs. Secondary organic aerosols coat or embed other species and are mostly classified into the categories of their host species. Volatile organic compounds are likely lost before the analysis, and thus, their number fraction may be underestimated over the mass fraction measured using online instruments. Tarballs or tarball-like particles are spherical organic particles 320 originating from BB (Pósfai et al., 2004) (Fig. 12) and are a major aerosol type originating from BB as brown carbon (Chakrabarty et al., 2010;Sedlacek et al., 2018). As a result, they potentially influence the climate, although their detailed occurrences, including their removal processes, remain unknown. In the BB samples, we encountered a small number of tarball particles (<1 % in number fractions). These tarballs mainly consist of C and O and include some N and K (Fig. 12), and their composition is similar 325 to that of particles from young BB smoke plumes (e.g., Pósfai et al., 2004;Adachi and Buseck, 2011). Moroni et al. (2017Moroni et al. ( , 2020) also found tarballs with K-bearing particles using scanning electron microscopy in ground observations on the Svalbard Islands in the summer of 2015. In contrast to the smooth spherical shapes of fresh tarballs (e.g., Li et al., 2003), the surfaces of our tarballs are not smooth and contain sulfates (Fig. 12), suggesting that they reacted with other species and had aged Soot particles were also commonly found in our samples. However, only 13% of the carbonaceous particles consisted of soot with thin or no coatings, i.e., external mixtures, and they were mainly embedded within or attached to other particles (internal mixtures) and were classified as a part of their host species. 335

Soot, fly-ash, and Fe-aggregate inclusions
Inclusions are particles embedded within or attached to host particles. Inclusions are identified based on both their composition and shape determined from the TEM images after the removal of beam-sensitive materials (e.g., sulfate) by exposure to an electron beam during the STEM-EDS analysis ( Fig. 13 and Fig. S8). When they overlap with non-beam-sensitive materials (e.g., mineral dust and sea salt), it is 340 difficult to detect inclusions in the TEM images. Thus, the actual number fraction of the inclusions may be larger than that reported in this study. On the other hand, this technique has the advantage of identifying small inclusions (e.g., < 50 nm), which are difficult to detect using other methods. Although we observed several metal particles as inclusions (e.g., Zn and Pb), we focused on the occurrences of soot, fly-ash, and Fe-aggregate particles in this study. In our samples, soot particles were found in ~17% of all measured particles ( Fig. S9 and Table S1).
Most soot particles were internally mixed with other particles (Fig. 13 and Fig. S8). The number fraction of the externally mixed soot particles, which are those without any apparent coatings, was only ~1% of all soot particles. The relatively low number fractions of the externally mixed soot particles are 360 similar to the background conditions observed by Hara et al. (2003) and indicate that most LRT soot particles were internally mixed in the Arctic.
Soot particles were observed within all aerosol types, and the sampling altitude was not clearly correlated with the soot particle fraction (R 2 =0.13). On the other hand, more soot particles were found https://doi.org/10.5194/acp-2020-1114 Preprint. Discussion started: 2 November 2020 c Author(s) 2020. CC BY 4.0 License.
in large host particles than in small ones, i.e., the size distributions of the soot-bearing particles (host 365 particles) were larger than those of all particles (Fig. S10). This result is consistent with those in other areas ( Adachi and Buseck, 2008;Adachi et al., 2014). A possible explanation is that the coagulation process of soot particles with large host particles occurs more efficiently than that with small particles.
BB is one of the major sources of BC in the Arctic atmosphere and snow (Spackman et al., 2010;Hegg et al., 2010). In our BB samples, ~93% of the soot-bearing particles contained K, showing substantial 370 BC contributions from BB. In comparison, the ratio was 74% for the non-BB samples. However, there was no apparent enhancement of the soot particle number fraction in the BB samples (Table S1). This inconsistency may be explained by the fact that we measured the number fraction and that the other species, such as the K-bearing salts, sulfates, and organic matters, also increased within the BB plume, resulting in relatively lower soot number fractions in the BB samples.

Fly-ash and Fe-aggregate inclusions
The fly-ash particles and primary particles of Fe aggregates commonly exhibited spherical shapes, which indicated that they formed through rapid cooling after melting or evaporation at high Thus, we suggest that our Fe-aggregate particles also exhibit light-absorbing properties.
The number fractions of the particles containing fly-ash or Fe-aggregate particles are 1.4% and 0.5%, respectively ( Fig. S9 and Table S1). Although these particles are commonly found in polluted areas 385 from anthropogenic sources such as stationary combustion sources and vehicles (Li et al. 2016), there are almost no such local anthropogenic sources in the Arctic area. Instead, the fly-ash and Fe-aggregate particles attain better relations with the soot number fraction (R 2 =0.35 and 0.24 for the fly-ash and Fe aggregates, respectively; Fig. S11), and soot, fly-ash, and Fe-aggregate particles are often found in the same particles ( Fig. 13 and Fig. S8). Our TEM measurements agree with the SP2 observations of 390 Yoshida et al. (2020), who also found correlations among BC and Fe oxide during this campaign. These results indicate that the soot, fly-ash, and Fe-aggregate particles originated from anthropogenic sources through LRT.

Implications for particle aging and the climate
In the Arctic, although there are few anthropogenic sources and vegetation, we found that many aerosol The soot and Fe-aggregate particles still exhibit fractal structures, especially the small particles (< several tens of primary particles; Fig. 13 and Fig. S8), similar to those found in source regions such as East Asia (Adachi et al., 2016). This observation is inconsistent with the result showing highly 410 compacted soot particles at a remote marine free-troposphere site (China et al., 2015). Whether soot particles become core-shell structures as they age in the atmosphere is of interest for the accurate evaluation of soot climate effects (Cappa et al., 2012;Adachi et al., 2010). Our observation implies that the soot particles in the Arctic cannot be assumed to be spherical particles but should be treated as fractal particles when calculating the optical properties of aged samples. One explanation is that soot 415 particles consisting of several tens or fewer primary particles hardly attain highly compact shapes because of the insufficient monomer numbers to turn them into spherical aggregates. On the other hand, most soot particles were coated with sulfate or other materials, suggesting that they possess enhanced light absorption and CCN activity. We also observed the aging of tarball particles in a BB plume, i.e., the surfaces of certain tarball particles had reacted with sulfate, making them hygroscopic. Aged tarballs 420 are removed more efficiently from the atmosphere by precipitation than the original ones. As the mass fraction of tarballs in fresh BB smoke may reach ~40% (Sedlaeck et al., 2018), which is much higher than that in our samples, most tarballs could have been removed from the atmosphere during LRT. This tarball removal process has been proposed by Pósfai et al. (2004) when they first characterized tarballs, and our study provides evidence of the above hypothesized processes.

Summary
This study reveals that the aerosol particles in the Arctic troposphere exhibit various composition, shape, and mixing state ranges depending on the sampling altitude and airmass history. Sulfate is the dominant aerosol type, and sea-salt, mineral-dust, K-bearing, and carbonaceous particles are also observed as major aerosol species. The aerosol particles are commonly mixtures of several components,

Data and code availability
The TEM and simulation data used in this publication are available upon request (adachik@mrijma.go.jp). Access to the MRI-ESM2 code is available under a collaboration framework with the MRI.

Author contributions
KA conducted the TEM analysis and data processing. SO, AY, and MK executed the TEM sampling 450 and field observations. KA and NM set up the TEM sampler. MK supervised the TEM sampling. NO performed the model simulations and analyses. KA prepared the manuscript with contributions from all coauthors.

Competing interests
The authors declare that they have no conflicts of interest.