Primary productivity in high-nitrate low-chlorophyll (HNLC) regions of the world's ocean has been an important topic because of the roles of this process in regulating atmospheric carbon dioxide levels over glacial–interglacial timescales (Boyd et al., 2000, 2004; Bopp et al., 2003; Jickells et al., 2005; Formenti et al., 2011). Iron (Fe) is a controlling micronutrient for phytoplankton growth in HNLC regions, with deep winter mixing (Tagliabue et al., 2010) and long-range transport of continental aerosols (including mineral dust and anthropogenic aerosols) being key Fe sources for surface water microbial communities. In addition, the dust-derived Fe supply to low-nitrate low-chlorophyll (LNLC) regions of the oceans has been shown to control dinitrogen fixation (Moore et al., 2009; Schlosser et al., 2014). Therefore, interest has grown in recent years regarding atmospheric aerosol transport, inputs to the surface ocean, and the subsequent dissolution of Fe from aerosols (Desboeufs et al., 2001; Jickells and Spokes, 2001; Hand et al., 2004; Guieu et al., 2005; Meskhidze et al., 2005; Baker and Jickells, 2006; Buck et al., 2006, 2010; Cwiertny et al., 2008; Journet et al., 2008; Mahowald et al., 2009; Shi et al., 2009, 2011; Aguilar-Islas et al., 2010; Baker and Croot, 2010; Fu et al., 2010; Johnson et al., 2010; Paris et al., 2010; Trapp et al., 2010; Formenti et al., 2011; Rubin et al., 2011; Takahashi et al., 2011; Sholkovitz et al., 2012).

Iron dissolution from aerosols has been represented by fractional Fe solubility ( %FeS), and varies strongly depending on the aerosol source (Mahowald et al., 2005; Sholkovitz et al., 2012). Sholkovitz et al. (2012) compiled total Fe loading (FeT) and %FeS for a global-scale set of aerosol samples, and found a hyperbolic trend in the %FeS as a function of FeT, which was explained by the mixing of mineral dusts of high FeT and low %FeS and anthropogenic aerosols of low FeT and high %FeS. However, mineral dust is an important supply of bioavailable Fe to the remote ocean, particularly during dust events originating from desert sources. Ito and Feng (2010) demonstrated using model simulations that, compared to Asian dust, soluble Fe from combustion sources contributed a relatively small amount to the soluble Fe supply to the North Pacific Ocean during spring periods. An enhancement of dust %FeS occurs during long-range transport. The factors responsible for this increase in %FeS are not yet fully understood, and include the type of Fe-bearing minerals of dust and their reactivity (Cwiertny et al., 2008; Journet et al., 2008), the photoreduction of Fe in dust particles (Siefert et al., 1994; Hand et al., 2004; Fu et al., 2010), reactions between dust particles and water during cloud processing (Desboeufs et al., 2001; Shi et al., 2009), reactions with acidic gases in the atmosphere (Zhuang et al., 1992; Meskhidze et al., 2003), and changes in particle size during long-range transport (Jickells et al., 2005; Baker and Croot, 2010).

Information regarding dust mineralogy enables the %FeS of mineral dust to be better understood, as highlighted by Cwiertny et al. (2008) after an extensive literature review. The mineralogical factors related to %FeS include solubility, reactivity with atmospheric acids, grain size, Fe content, and the Fe oxidation state of the minerals. Despite numerous articles on the measurements of %FeS (Sholkovitz et al., 2012, and references therein), the modeling of dust input (Mahowald et al., 2005, 2009; Johnson and Meskhidze, 2013), and the determinations of the aqueous geochemistry of Fe (Baker and Croot, 2010, and references therein), basic data are still lacking on the properties of Fe-bearing minerals in dust. Iron oxides/hydroxides are an important source of available Fe. However, the quantities of Fe (hydr)oxides in mineral dust are much lower than the quantity of Fe-bearing silicates. Mineralogical analyses have shown that clay minerals are the most abundant phases followed by quartz, feldspars, and calcite in the long-range transported dusts (Glaccum and Prospero, 1980; Avila et al., 1997; Jeong 2008; Jeong et al., 2014). The crystal structures of clay minerals can accommodate a significant quantity of Fe in their octahedral sites. Thus, both the clay minerals and Fe (hydr)oxides should be considered when investigating their roles in Fe availability (Raiswell and Canfield, 2012). Journet et al. (2008) reported a higher Fe solubility of clay minerals compared with Fe oxides, emphasizing the significant role of clay minerals in Fe availability. However, in the experiments performed by Journet et al. (2008), dissolution work was conducted for a limited set of clay minerals and Fe (hydr)oxides obtained from rocks. Clay minerals in atmospheric dust particles have diverse origins and a wide range of chemical compositions and particle sizes, depending upon the lithology, geological setting, and physical/chemical weathering process in their source regions. The contribution of clay minerals to Fe availability should be considered on the basis of the physical and chemical characteristics of the different clay mineral types in the natural dust. For example, the Fe content of clay minerals in dust for modeling and dissolution experiments is typically not known. The separate determination of the Fe content of each clay mineral species is almost impossible for bulk dust because of the agglomeration of many silicate mineral grains (Falkovich et al., 2001; Shi et al., 2005; Jeong, 2008; Jeong and Nousiainen, 2014; Jeong et al., 2014). This is in contrast to the exact determination of Fe content in the form of Fe (hydr)oxides using an established selective extraction procedure, such as the method of Mehra and Jackson (1960). However, the chemical composition of submicron grains of clay minerals can be determined by energy dispersive X-ray spectrometry (EDXS) of fluorescent X-ray induced by an electron microbeam. EDXS attached to a transmission electron microscope (TEM) is an excellent technique for the chemical and physical characterization of individual clay mineral grains.

In this study, we report the mineral species, nanoscopic occurrence, and chemical compositions of the clay mineral grains in individual Asian and Saharan dust particles obtained by the combined application of TEM and EDXS. Analyses of clay minerals mixed in particles were conducted on cross-sectional slices of individual dust particles prepared by focused ion beam (FIB) milling. Clay mineral grains loaded on micro-grids by a conventional procedure were also analyzed by TEM and EDXS. We furthermore present the mineral compositions of bulk dusts obtained by X-ray diffraction (XRD) analysis.

Asian dust events were observed on 17 March 2009, 20 March 2010, 31 March 2012, and 18 March 2014 in the Republic of Korea. The dust outbreaks and subsequent migration of the Asian dusts were traced using dust index images derived from satellite remote sensing, which indicated the source of the four dust events in the Gobi desert of southern Mongolia and northern China (30∘–46∘ N, 90∘–110∘ E) and their migration to the east across the Republic of Korea. Details of the synoptic conditions during the dust outbreaks and migration of these events were provided in Jeong et al. (2014). The peak concentrations of particulate matter less than 10 µm in diameter (PM10) were 428, 1788, 220, and 378 µg m-3 in the 2009, 2010, 2012, and 2014 dust events, respectively (Korea Meteorological Administration, 2014).

The Asian dusts were sampled using a Thermo Scientific high-volume total suspended particulate (TSP) sampler fitted with Pallflex teflon-coated borosilicate glass-fiber filters (8 × 10 in.) or Whatman^® No. 1441-866 cellulose filters. The 2012 dust was sampled on a mountain peak at Deokjeok Island (190 m a.s.l., 37∘13′59′′ N, 126∘08′57′′ E) off the western coast of the Republic of Korea for 24 h (09:00, 31 March–08:00, 1 April) at a flow rate of 250 L min-1. The 2009, 2010, and 2014 Asian dusts were sampled using the same procedure on the roof of a four-story building at Andong National University (36∘32′34′′ N, 128∘47′56′′ E) over a 12 h period (09:00–21:00, 17 March 2009, 20:00–08:00, 19 March 2010, and 10:00–22:00, 18 March 2014).

Saharan dust samples were collected at the Cape Verde Atmospheric Observatory on the island of São Vicente, Cabo Verde (16∘51′50′′ N, 24∘52′03′′ W) in the eastern North Atlantic Ocean. Dust was sampled on Sterlitech polypropylene membrane filters (47 mm diameter, 0.4 µm pore size) using a low-volume aerosol sampler installed at the top of a 30 m tower in the period from 7 November 2007 to 14 March 2008 (Carpenter et al., 2010). The total volume of samples for the individual filters ranged from 50 to 100 m3 at flow rates of 20–30 L min-1 for a period of 3–5 days.

X-ray diffraction (XRD) analysis was conducted to obtain information on the mineral composition of bulk samples using a Rigaku Ultima IV diffractometer. A portion of the filter was cut and immersed in methanol in a 10 mL glass vial. The filter was agitated in an ultrasonic bath to remove dust particles from the filter. During the agitation, dust particles were disaggregated into clay minerals and other silicate grains. The suspension was sieved through 270-mesh sieve to remove cellulose fibers and dried on the glass plate. Then, dust was collected by razor blade. Several milligrams of dust samples (2–10 mg) were loaded on the 3 mm × 4 mm cavity of an aluminum plate. The analytical conditions were 60 s of counting per 0.03∘ step in the scan range of 3–65∘ 2θ, Cu Kα radiation, and 45 kV/35 mA. Because the quantity of the samples was small, the patterns obtained were not suitable for precise quantification. Thus, the compositional analysis was deemed semi-quantitative. Mineral compositions were derived using a SIROQUANT software package (version 4). Since intensity loss was significant in the lower angle region due to the small sample size irradiated with X-rays, the high angle region (24.5–65∘ 2θ) was used for the simulation of the observed XRD pattern. After the XRD analysis of the bulk dusts, the samples were subjected to ethylene glycol and heat treatments for detailed identification of clay minerals. The samples were wetted with water, smeared, and dried on a glass slide, and subsequently treated with ethylene glycol vapor at 60 ∘C in a desiccator for 2 days followed by heating at 350 ∘C for 30 min.

Electron-transparent thin slices of dust particles (2009, 2012, and 2014 Asian dusts; 28–31 December, 2007, 18–23 January, 23–26 February, 29 February–4 March, 12–14 March, 2008 Saharan dusts) were prepared for TEM analysis of clay minerals. Hereafter, the term “particle” refers to individual solid objects suspended in the atmosphere, while the term “grain” refers to the constituents of the particles. Thin slices (ca. 100 nm in thickness) of about ca. 6 µm × 6 µm size were cut from dust particles using a SII NanoTechnology SMI3050TB and a JEOL JIB4601F FIB instrument for the Asian and Saharan dusts. Prior to using the FIB, the dust particles were transferred onto adhesive carbon film and characterized using a JEOL JSM 6700F field emission gun scanning electron microscope (SEM) equipped with an Oxford EDXS system at 5 kV acceleration voltage and 8 mm working distance after being coated with platinum for electrical conductivity. Dust particles were selected for FIB work on the basis of mineralogical characteristics identified by SEM-EDXS analysis as reported in Jeong (2008) and Jeong et al. (2014). Individual Asian dust particles could be selected and prepared as thin slices by FIB milling because the particles were sufficiently separated from each other on the filters. However, the Saharan dusts considered in this study were highly concentrated and aggregated on the filters. Thus, the original atmospheric particles for FIB milling could not be identified with confidence. However, the purpose of the TEM analysis undertaken in this study was not to reveal the structures of individual original dust particles as reported by Jeong and Nousiainen (2014), but to analyze the chemistry and mineralogy of clay mineral grains. Thus, we prepared thin slices using FIB milling from the Saharan dust samples.

Clay-mineral grains loaded on micro-grids were also prepared for investigation of Asian (2009, 2010, 2012, and 2014) and Saharan (7–9 November, 13–14 December, 28–31 December 2007, 18–23 January, 29 February–4 March, 12–14 March 2008) dust. The clay mineral grains suspended in methanol were loaded on to 200-mesh Cu micro-grids covered with a carbon-coated lacey formvar support film by immersing the micro-grids in the suspension with tweezers and subsequent drying on filter paper. Every electron-transparent grain encountered during the movement of stage was analyzed by EDXS.

The TEM instruments used in this study were a JEOL JEM 2100F field emission gun STEM at 200 kV, a JEOL JEM 3010 TEM for high-resolution imaging, and a JEOL JEM 2010 TEM equipped with an Oxford ISIS EDXS system. Minerals in the FIB slices were identified on the basis of lattice fringe images, electron diffraction, and EDXS. General chemical formulas of minerals identified in this study are given in the Supplement Table 1. Digital images of the microstructures and lattice fringes were recorded using a Gatan digital camera and processed with a Gatan DigitalMicrograph^®. To obtain the elemental compositions of clay minerals, the X-ray counts of Si, Al, Fe, Mg, Ti, K, and Ca were converted to weight % (wt %) of the elements using the quantification procedures given by Cliff and Lorimer (1975). The k-factors of the elements for the conversion were obtained from an analysis of the FIB slices prepared from biotite and plagioclase of known composition occurring in the Palgongsan granite (Jeong, 2000).

TEM-EDXS analysis often requires a small beam size down to ∼ 50 nm to determine the chemical composition of clay minerals that are mixed with other minerals. The X-ray generation volume is small due to the thin nature of the FIB slices (∼ 100 nm) and small analysis area. High-quality quantitative analysis requires at least 10 000 counts of each peak for the ideal specimens of large thin and resistant phases (Williams and Carter, 2009). However, such an ideal analytical condition was not obtained for the clay mineral grains, which were very sensitive to the electron beam because of their structural water, disorder, and nanocrystallinity. To minimize electron beam damage of the clay mineral grains, the electron dose was reduced by setting the spot size to four. X-rays of Si, Al, Fe, Mg, Ti, K, and Ca were counted for 100 s. For clay grains loaded on the micro-grids, 0.5–2 µm grains produced sufficient X-ray photons for analysis, while submicron thin clay grains (ca. < 0.5 µm) did not. The total X-ray counts of those elements in ISIS EDXS system were ca. 30 000 for the analysis of clay mineral grains loaded on the grid, and ca. 15 000 for the analysis of FIB specimens. The detection limits for these elements were ca. 0.1 wt %. For Asian dust samples, we conducted 206 analyses of clay minerals in the 50 FIB slices prepared from 50 dust particles, and 514 analyses of clay mineral grains loaded on the micro-grids. For Saharan dust, we conducted 116 analyses in the 10 FIB slices prepared from 10 particles, and 356 analyses of clay mineral grains loaded on micro-grids. In the calculation of elemental wt %, the total H2O content of clay minerals was assumed to be 14 wt %, which is the average H2O content of illite and smectite provided in Table II and Table XXVII of Weaver and Pollard (1975).

Global dust distributions and oceanic Fe deposition are modeled assuming the Fe content of bulk dust being equivalent to the average Fe content of the Earth crust (3.5 %) (Mahowald et al., 2005, 2009). A recent modeling study on Fe supply to the oceans progressed to include dust mineralogy and Fe content of clay minerals (Johnson and Meskhidze, 2013). Mineral compositions were predicted from available soil data (Claquin et al., 1999; Nikovic et al., 2012; Journet et al., 2014). However, mineralogical and chemical compositions of world soils are certainly variable and poorly constrained. In particular, the analytical data on the chemical composition (including Fe) of fine clay minerals in local soils are hard to find in literature because the analytical measurements are as difficult as the analysis of clay minerals in dust. The modeling study on the Fe supply to global oceans by Johnson and Meskhidze (2013) highlighted the importance of mineralogical data of dust in the deposition of dissolved Fe to the global oceans. They adopted mineral compositions of dust derived from a soil database of Nikovic et al. (2012) and Fe content of a commercial illite reported by Paris et al. (2011). However, their model sensitivity analysis showed large variations of dissolved Fe fluxes to the oceans, associated with the uncertainty in the dust mineralogy and Fe content of constituent minerals. They emphasized that realistic data of dust mineralogy and Fe content within individual Fe-containing minerals are essential for the improvement of the description of the Fe biogeochemical cycles in climate models. Although the chemical analysis of clay minerals in individual dust particles as carried out in this study is challenging, detailed analysis of representative global dust samples is likely an efficient approach for obtaining information on Fe mineralogy in dust. Long-range transported dust has the average mineralogical and chemical properties of fine soil particles lifted from the wide source regions of varying soil mineralogy. In addition, the mineral composition and chemistry in dust from the major source regions do not show significant inter-event and annual changes (Jeong, 2008). Further analytical works for dust samples transported over long distances from major source regions will allow the establishment of realistic mineralogical database for the evaluation of Fe inputs to the remote ocean.

Mineral composition of bulk dust samples, Fe content of constituent minerals, and their grain sizes form essential data for the evaluation of Fe supply to remote surface oceans. Particularly, the properties of clay minerals are important because of their abundance in mineral dust and high Fe contents, relative to other silicates. We presented for the first time mineralogical and chemical data of clay minerals in individual dust particles from several Asian and Saharan dust samples, following analyses by TEM and EDXS. The total clay content of Asian dusts determined by XRD analysis was lower than that of Saharan dusts. TEM analysis of thin cross-sectional slices of the dust particles revealed that nano-thin platelets of ISCMs (illite, smectite, and illite–smectite mixed layers) were most abundant in association with illite, chlorite, and kaolinite occurring as thicker plates. Asian dusts were enriched with chlorite relative to Saharan dust, while Saharan dusts were relatively enriched with kaolinite. Kaolinite in Saharan dust occurred as hexagonal plates that were better crystallized than in Asian dust. The average Fe content of the ISCMs in Asian dusts as determined by EDXS was 5.8 % assuming 14 % H2O, while the contents of illite and chlorite were 2.8 and 14.8 %, respectively. The average Fe content of the EDXS data of the clay mineral grains dispersed and loaded on the micro-grids was 6.7 and 5.4 % in Asian and Saharan dusts, respectively. The Fe dissolution from clay minerals is thought to be enhanced by the nanocrystallinity of ISCMs; furthermore, Fe-rich chlorite susceptible to acids may enhance dissolution of Fe. The establishment of realistic mineralogical data sets from global dust samples is important to reduce the uncertainty in the prediction of iron inputs to oceans using geochemical and meteorological models.