The mineralogical properties of Asian dust have been examined in detail, and the results have been compared with the Chinese source soils using high-resolution scanning electron microscopy, transmission electron microscopy, X-ray diffraction, and chemical analysis. Mineralogical classification of Asian dust particles showed that the most common single particles were clay aggregates (48%) that were often mixed with nanosized calcite, followed by particles of quartz (22%), plagioclase (11%), coarse calcite (6%), K-feldspar (5%), muscovite, chlorite, kaolinite, amphibole, gypsum, Fe and Ti oxides which were either partly or entirely attached with clay-size mineral grains. The clay minerals in Asian dust were mostly illite and interstratified illite-smectite. The average mineral composition of the bulk dust samples by X-ray diffraction was quartz (28%), plagioclase (11%), K-feldspar (8%), calcite (8%), illite (19%), interstratified illite-smectite (22%), chlorite (2%), smectite (1%), and kaolinite (1%). In the silty soils from the source regions, the clay minerals and nanosized pedogenic calcite aggregated and covered silt-size minerals, while the sands were mostly composed of quartz and feldspars and were lined with clay minerals. The mineralogy of Asian dust was similar to that of the silty soil from the loess plateau, but the total phyllosilicate content increased from desert sands (7%), silty soil (23%), to Asian dust (45%), fining eastward. The optical properties of Asian dust and its interaction with atmospheric gases and cloud are probably affected by the clay-rich mineral composition, the aggregation and attachment of the clay minerals on the coarser minerals, and the nanosized calcite.
 Asian dust is an Earth material, and is mostly composed of minerals derived from regoliths that have disintegrated from the lithosphere. It continuously interacts with the atmosphere, hydrosphere, and biosphere during its journey. Recent studies have shown that mineralogical characterization of mineral dust is essential for modeling its climatic and atmospheric role, because the reactivity and optical properties of minerals vary widely depending on species, crystal size, shape, crystallinity, chemistry, and mixing state [Buseck and Pósfai, 1999; Claquin et al., 2003; Krueger et al., 2004; Lafon et al., 2006; Sokolik and Toon, 1999]. This characterization can be carried out by chemical and crystallographic analysis of single particles and bulk samples, followed by mineralogical and geological interpretation of the data. There is an abundant volume of work on the chemical properties of bulk dust [Arimoto et al., 2004; Kim et al., 2006; Sun et al., 2005; Zdanowicz et al., 2006], and single particles using a time-of-flight mass spectrometer [Matsumoto et al., 2006] and an X-ray spectrometer attached to a scanning electron microscope [Okada et al., 1990; Ro et al., 2005], a transmission electron microscope [Okada et al., 1990], and a synchrotron X-ray source [Ma et al., 2004]. Despite the large number of chemical studies on Asian dust, many of the previous studies have provided limited mineralogical data and interpretations. Bulk chemical analyses have often omitted some of the major elements of minerals (e.g., Si), or they have been rarely combined with X-ray diffraction (XRD) and electron microscopy. Single particles have been classified simply into chemical types, often without any mineralogical designation.
 Detailed and systematic mineralogical analyses have been carried out on Asian dust sampled in Korea, and soil samples from the Chinese source regions (Figure 1). This paper reports on the quantitative mineralogical properties of both bulk dust and single particles by comparing the data with that of soils from the source regions, and discusses the atmospheric implications.
2. Samples and Methods
 During the spring seasons of 2003, 2004, and 2005, eleven Asian dust events were recorded in Seoul by Korea Meteorological Administration with average duration time of 27 h. Eight events were selected to analyze the mineralogy of Asian dust. Backward trajectory analyses of the air masses performed using the NOAA/ARL HYSPLIT model [Draxler and Rolph, 2003] show that the dust was transported from the arid regions of Mongolia and north China, and arrived in Seoul within a period of 1–3 d (Figure 1b). By analyzing the 40-year records of spring dust storms in China, Sun et al.  showed that all the dust storms were associated with cold air outbreaks in Siberia, which resulted in Mongolian cyclonic depression and cold frontal systems. During the dust events, the peak PM10 concentration ranged from 194 to 683 μg/m3 (Figure 2a) with a dominant wind direction of west-northwest and a wind speed from 2.8 to 5.2 m/s (Korea Meteorological Administration, http://www.kma.go.kr/intro.html). The background PM10 concentration by averaging hourly data obtained during the non-Asian dust period was 53 ± 35 μg/m3 in 2003, 61 ± 38 μg/m3 in 2004, and 64 ± 38 μg/m3 in 2005. The number concentration data of aerosol particles in eight size intervals (0.3–0.5, 0.5–0.82, 0.82–1.35, 1.35–2.23, 2.23–3.67, 3.67–6.06, 6.06–10, 10–25 μm) measured every hour by an optical particle counter in the dates that the Asian dust event occurred were retrieved from the database of Korea Meteorological Administration and converted to volume size distributions, as shown in Figure 2b. The volume size distributions show that the aerosol particles of Asian dust are mostly distributed around 3 μm, being between 1 and 8 μm, which is consistent with the data of Kadowaki  and Chun et al. .
 The dust samples were collected over a period of 9 h on a Whatman® No. 1441-866 cellulose filter using a Wedding & Associates PM10 high-volume sampler with a flow rate of 1.13 m3/min, which was installed at the Meteorological Research Institute, Seoul, Korea (37°29′37″N, 126°55′01″E). The mass of the dust samples collected on the filters ranged from 0.2 to 0.5 mg per cm3. The surface soils sampled in the arid to semiarid source regions were sandy soils from six localities in the northern deserts and silty soils from the loess plateau peripheral to the deserts (Figure 1a).
 The silty and sandy soils from the source regions were dispersed in water by ultrasonic agitation, and their particle size distribution in the range 0.02–2000 μm was measured using a Malvern Mastersizer 2000.
 For X-ray diffraction analysis, a portion of the filter paper (about 5 × 5 cm2) was cut into small strips and immersed into methanol in a 10 ml glass vial, which was subsequently treated in an ultrasonic bath. The liquid was evaporated to leave a few milligrams of sample behind. The sample was smeared on a glass slide with a drop of ethylene glycol. After drying, the glass slide was treated with ethylene glycol vapor at 60°C for a period of 24 h. XRD analysis was carried out between 3° and 40° (2θ) using a Rigaku RINT 2200 diffractometer equipped with a Cu target and a reflected beam monochromator. The count time was set to 50 s per 0.04° (2θ). Although the mass of the dust sample used in the XRD analysis was very small, an interpretable pattern was obtained on scanning for a period of 13 h. After the XRD analysis, the samples were subsequently heated to 300°C and 500°C, and rerun under the same conditions.
 The sands and silts were ground to a particle size of 5 μm by using a McCrone micromill and analyzed by a Bruker D8 Advance diffractometer located at the Korea Basic Science Institute, Daejeon, Korea. The mineral composition was quantified using the Bruker TOPAZ software package. Clay particles separated from the sands and silts were subjected to an ethylene glycol vapor treatment and subsequent heating at 300°C and 500°C.
 The chemical composition of bulk samples of Asian dust and source soils were analyzed at the Activation Laboratories, Ontario, Canada, using a Thermo Jarrell Ash ENVIRO II inductively coupled plasma emission spectrometer. The filters were ashed, fused with lithium metaborate/tetraborate, and subsequently digested.
 The morphology and chemistry of the dust particles were analyzed using a JEOL JSM 6700F field emission gun scanning electron microscope (FEGSEM) for high resolution and a JEOL JSM 6300 conventional scanning electron microscope (SEM) for lower resolution in the secondary electron (SE) image mode. Both electron microscopes were equipped with an Oxford energy-dispersive X-ray spectrometer (EDS). For morphological observations, the filter paper loaded with dust particles was attached to a glass slide using double-sided tape. The dust particles were allowed to freely fall on conducting carbon tape on a Cu stub by tapping the glass slide, and then were coated with gold. The transfer of dust particles from the filter on the carbon tape was inevitable in the high-resolution FEGSEM imaging to avoid the electron charging problem due to the loose and porous fabric of the filter. SEM images of the dust particles on the filter were compared to those on the carbon tape. The microscopic features of the individual dust particles on the filter were not different from those on the carbon tape after transfer. For a quantitative mineralogical analysis, a large set of single dust particles (>300 per sample) were analyzed from SEM images (×500) with the measurement of the long chord of the particles.
 Original blocks of silty soils and loose aggregates of sandy soils were impregnated with epoxy resin, and prepared to polished thin sections [Jeong and Kim, 1993]. The morphology and chemistry of the minerals in a thin section were analyzed using a JEOL JSM 6300 SEM in the backscattered electron (BSE) image mode after carbon coating.
 The morphology and chemistry of the submicrometer clay minerals of Asian dust and the source soils were analyzed using a JEOL 2010 transmission electron microscope (TEM) equipped with an Oxford EDS analyzer. The clay minerals on the filters and soils were dispersed in methanol using ultrasonic agitation, and loaded on a microgrid of lacey Formvar carbon. Quantification of the chemical composition of the clay minerals was carried out using the calculated k-factors under thin-film conditions [Cliff and Lorimer, 1975].
3.1. Source Soils
 The mineral compositions of the soils determined using XRD are given in Table 1. The sandy soils consisted mostly of quartz, plagioclase, and K-feldspar with minor phyllosilicates. Calcite was not detected in most of the sandy soils, except for Sand-B, which had a calcite content of 2 wt %. Compared to the sandy soils, the silty soils were rich in calcite and phyllosilicates. Small quantities of amphibole, dolomite, and gypsum were also detected in the silts. Compared to the sandy soils, the silty soils had higher Al2O3, Fe2O3, MgO, TiO2, and CaO content, and a lower SiO2 content (Table 2), due to the abundant clay minerals and carbonates. The average chemical composition of the silty soils was closer to that of the loess.
Table 1. Mineral Composition of Sandy and Silty Soils in Source Regionsa
Sample locations are as follows: Sand-A, Duolun, Inner Mongolia, China (N42°21′, E116°31′, average of two samples); Sand-B, Hobq, Inner Mongolia, China (N40°14′, E109°54′); Sand-C, Mu Us, Inner Mongolia, China (N39°06′, E109°39′, average of two samples); Sand-D, Mu Us, Inner Mongolia, China (N38°03′, E109°39′); Sand-E, Hustai, Mongolia (N47°41′, E105°54′); Sand-F, Zamin Ud, Mongolia (N43°42′, E111°54′); Silt-A, Yulin, Shaanxi, China (N38°17′, E109°50′, average of three samples); Silt-B, Zhangye, Xinjiang Uygur, China (N39°05′, E100°17′); Silt-C, Jiayuguan, Xinjiang Uygur, China (N39°48′, E98°13′, average of three samples).
 BSE images of thin sections showed that the sandy soils consisted of rounded sand grains with minor subangular silt grains of mineral and rock fragments (Figure 3a). Detrital calcite particles were not found in most of the sandy soils, except in Sand B, which is consistent with the XRD data. Calcite was not the dominant mineral in the sandy soils. Clay materials were difficult to detect in the BSE images of thin sections of the sandy soils. However, FEGSEM images showed thin layers of irregular plates of clay minerals encrusting the surface of the sand grains (Figures 3b and 3c), which were identified as smectite, illite, chlorite, and kaolinite from the XRD data.
 In the BSE images of thin sections, the silty soils consisted of mineral and rock fragments of silt to fine-sand sizes and clay minerals (Figure 4a). EDS analysis of the silt and fine-sand grains identified quartz, K-feldspar, plagioclase, calcite, amphibole, biotite, muscovite, chlorite, and epidote. The clay-mineral grains were aggregated, covering the surfaces of the silt and fine-sand grains or forming their own aggregates (insets in Figure 4b). SE images showed a thick cover of clay minerals having an irregular platy shape on the silt grains (Figures 4c and 4d) and nanosized calcite particles (Figure 4e). Illite, chlorite, kaolinite, and smectite were identified in the XRD and TEM analyses of the clays separated from the silty soils.
 The particle size distribution of the silty and sandy soils is given in Figure 2c. The sandy soils showed a relatively a narrow size range, around 200 μm, while the silty soils showed a wide size range, from 0.3 μm to 600 μm, with a maximum around 20 μm. However, even in the sandy soils there was a small subsidiary peak around 4 μm below 20 μm. The total volume fractions of the soils below 63 μm in Figure 2c were: 85% in Silt B, 76% in Silt C, 10% in Sand C, and 11% in Sand F.
3.2. Asian Dust
3.2.1. Mineralogy of Single Particles
 The combination of EDS (Figure 5) and morphological analysis (Figures 6 and 7) allowed for the identification of the mineralogical makeup of single dust particles. The most common dust particles were clay aggregates (Figures 6b–6e and 7b–7d) which were interpreted as being mixtures of submicrometer clay minerals of illite and interstratified illite-smectite with minor chlorite, kaolinite, and smectite, based on the extensive XRD, SEM-EDS (Figure 5), and TEM-EDS analysis of the single-grain clay minerals (see section 3.3). The uppermost EDS pattern shown in Figure 5c is typical of the common clay aggregates, where K and Ca are present in the interlayers of illite and smectite, respectively, while the following two patterns are representative of the aggregates of clay and nanosized calcite mixed in different proportions. Some of the clay aggregates (Figure 6e) could be distinguished from muscovite by the compact aggregation and from more common clay aggregates by the EDS pattern of illite (Figure 5b). Nanosized calcite particles were commonly associated with clay aggregates, mostly in the form of nanofibers (Figures 6b–6d, 6i, 7b, and 7h) [Jeong and Chun, 2006]. Some of the clay aggregates were associated with iron oxides (Figure 5c), titanium oxides (Figure 5c), gypsum (Figures 5c and 7c–7e), and halite (Figures 5c and 7d).
 Clay minerals, often together with nanofiber calcite particles, were attached to the coarser grains of quartz (Figures 6f and 6g), Na plagioclase (albite) (Figures 6i and 7i), Na-Ca plagioclase (Figures 7e and 7j), K-feldspar (Figure 7f), biotite (Figure 6a), chlorite (Figure 6a), muscovite (Figure 6a), amphibole (Figure 6j), and calcite (Figures 6h, 7g and 7h). In the dust sample 270303 (Figure 7), some of the large particles showed the original microstructure of the soils that consisted of coarse mineral grains and a fine matrix of clay minerals, nanosized calcite, or gypsum (Figures 7i and 7j).
 The long chords of almost all the particles in Figure 6a (total 450 particles) were measured, and their number and volume distributions were overlaid on Figure 6. The distributions with maxima around 3–4 μm were similar to that measured by optical particle counter in Figure 2b.
3.2.2. Mineral Composition of Bulk Dusts
 The XRD patterns obtained from a few milligrams of sample were not ideal for quantitative analysis of the mineral composition. Nevertheless, the mineral composition of Asian dust was quantified using XRD and chemical data. First, the ratios of the nonphyllosilicate minerals were evaluated from the intensities of the XRD peaks clustered between 26° and 30° (2θ), using the reference intensity ratio (RIR) method [Hillier, 2003]. Second, the ratios of the phyllosilicate minerals prone to preferred orientation were evaluated separately, using the RIR method from the intensities of the XRD peaks in the region where 2θ < 14°. Third, the two data sets were merged to provide the full mineral composition based on the total phyllosilicate content of Asian dust calculated from the chemical composition (Table 3), assuming that (1) the MgO content of Asian dust was proportional to the total phyllosilicate content and (2) the ratio of the total phyllosilicate content (in wt %) to MgO content (in wt % on a volatile-free basis) was equal to that of the source soils. These assumptions are reasonable, because Asian dust is composed of soil particles of fine-silt size blown from the source soils, and the clay particles in both Asian dust and soils have the same chemical characteristics, as shown in section 3.3. The silty soils on the loess plateau are mineralogically closer to Asian dust, and the average ratio of total phyllosilicates to MgO in the three silty soils (Tables 1 and 2) was estimated to be 8.5 ± 1.8. However, there is really no difference between the surface silty soil and the underlying loess, because the latter is a thick accumulation of the former. The chemical composition of Asian dust was similar to the average chemical composition of the loess (Table 3). A more rigorous and extensive analysis of the mineral composition and bulk chemistry of nineteen loess samples from the Chinese loess plateau estimated this ratio to be 10.6 ± 1.7 (G. Y. Jeong, unpublished data, 2007), which was used for calculating the mineral composition of Asian dust. In this quantification method, the grouping of XRD peaks of minerals used for quantification into those on low-angle side (mostly platy phyllosilicates, clay minerals) and on the high-angle side (nonphyllosilicates and carbonates) reduces the errors arising from the angle-dependent intensity loss due to the thinness of the samples on the glass slide, and this made it possible to determine easily oriented phyllosilicates separately from nonphyllosilicates. However, the differences in the relationship between the chemical and mineral composition of dust samples from that in loess would increase the error.
Sample number indicates the date of collection in day/month/year: for example, sample 130403 was collected on 13 April 2003. Data except for samples 230404 and 150405 were also presented by Jeong and Chun .
Average chemical composition of the 19 loess samples from the loess plateaus near Lanzhou and Tianshui, China (G. Y. Jeong, unpublished data, 2007).
 The mineral composition of Asian dust determined by XRD analysis is given in Table 4. The total phyllosilicate content ranged from 32% to 58%. The major clay minerals were interstratified illite-smectite (22%), and illite (19%) with minor chlorite (2%), kaolinite (1%), and smectite (1%). Quartz (20%–34%), plagioclase (7%–13%), K-feldspar (7%–10%), and calcite (2%–15%) were the major nonphyllosilicate minerals present.
Table 4. Mineral Composition of Asian Dust Samplesa
Unit is wt %. Minor minerals such as gypsum, amphibole, and dolomite were not included in the quantification because of their very weak XRD intensities.
 The mineral composition of bulk dust could also be evaluated in a different way using the mineralogical analysis of single dust particles. The dust particles were classified depending on the dominant mineral based on EDS analysis (Figure 5) and their morphological characteristics (Figures 6 and 7). The data in Table 5 were obtained from total 3125 particles and presented in both the number and volume %. Despite the limitations of this method due to the mixed mineralogy of the particles, the data matched those shown in Table 4 well, probably because of the similar specific gravity of the major minerals (quartz = 2.65, K-feldspar = 2.54–2.63, albite–andesine = 2.60–2.68, illite = 2.80, calcite = 2.71).
Table 5. Number and Volume Abundances of Asian Dust Particles Depending on Their Predominant Mineralsa
 The differences in the number % and volume % of major minerals in Table 5 were rather small because the number percentages of the minerals were little varied over different size fractions (Figure 8). The enhanced volume % of phyllosilicates partly resulted from the irregular shapes of the particles. In particular, the volume % of phyllosilicate particles tends to be overestimated. Muscovite and chlorite were mostly large plates (Figure 6a), and the clay aggregates were often plates (Figures 6a, 6b and 6e). In Figure 8, both the number of particles analyzed and the number % of clay decreased below 4 μm. Although most of the coarse dust particles in SEM images were analyzed, many of the very fine particles were not, resulting in the reduction of the population of fine particles. Particularly, fine platy clay particles were often missed out because of their low X-ray counts, emission from neighbor particles, and long analytical time in comparison to nonphyllosilicates such as quartz, K-feldspar, plagioclase, and calcite. Thus the number % of clays in fine fractions (< 4 μm) was lower than that in coarse fractions. On the other hand, the number % of minerals in the very coarse fractions (> 14 μm) was not as reliable because of the small population of particles. By limiting the size range of the dust particles to 4–14 μm for their mineralogical classification, the difference between number % and volume % was further reduced (Table 5). Some particles of the long chord exceeding 10 μm were included in PM10 sample, probably because of their irregular shapes, or the low density of the large porous clay aggregates. Although the volume % of the minerals should approximate their weight % in case of spherical particles, the true weight % is probably located somewhere between number % and volume % because of the inevitable overestimation of the volume % of the phyllosilicate-dominant particles. Therefore the mineralogy of Asian dust is best represented by the median of average number % and volume % of the single particles in the size range of 4–14 μm (Table 5): quartz 22%, plagioclase 11%, K-feldspar 5%, calcite 6%, amphibole 1%, Fe oxides 1%, Ti oxides 1%, gypsum 1%, clay 48%, chlorite 2%, kaolinite 1%, and muscovite 3%.
 The mineral quantification procedure employing XRD and chemical analysis was semiquantitative because of the small quantity of sample and the underlying assumptions made. Nevertheless, the mineral compositions shown in Table 4 are consistent with the mineralogical data determined from single particle analysis using SEM (Table 5). Comparing the results obtained by two methods, the ratios of major minerals to quartz are 0.39 (plagioclase), 0.29 (K-feldspar), 0.29 (calcite), and 1.61 (total phyllosilicates) from XRD method, while 0.50 (plagioclase), 0.23 (K-feldspar), 0.27 (calcite), and 2.50 (total phyllosilicates) from SEM method. The larger ratio of total phyllosilicate to quartz from SEM method could be attributed to the micropores within the clay aggregates. Part of the CaO content in Table 3 is assigned to calcite. The positive correlation coefficients between CaO and calcite contents were 0.59 (XRD method) and 0.55 (SEM method), supporting the reliability of the quantification methods employed in this study.
3.2.3. Chemical Composition of Bulk Dusts
 The chemical composition of the dust sample was highly uniform throughout the three continuous years studied. On comparing the average values of Asian dust and silt, SiO2 was depleted in the dust relative to the silt, but all the other elements were enriched (Table 3). However, the elemental composition varied slightly from sample to sample. Two dust samples (270303 and 130403) were different from the other samples in having higher Na2O and K2O contents.
3.3. Mineralogy of Submicrometer Clay Minerals in Dust and Soils
 XRD analysis of bulk samples of Asian dust showed that interstratified illite-smectite and illite were the major phyllosilicates, while smectite, chlorite, and kaolinite were relatively low in quantity. Since fine clay minerals in Asian dust have potential significance for atmospheric processes, their mineralogical characteristics were investigated further using TEM.
 In Asian dust, the clay minerals have a shape of irregular plates, with clear outlines of illite (Figure 9a) and chlorite (Figure 9b), and diffuse outlines of interstratified illite-smectite (Figure 9c). Kaolinite grains were very rare (Figure 9d). As in the FEGSEM observations, the clay minerals adhered even to the submicrometer grains of quartz and plagioclase (Figures 9e and 9f). In the source soils, the clay minerals are interstratified illite-smectite (Figure 10a), illite (Figure 10b), chlorite (Figure 10b), and rare kaolinite (Figure 10c) in the form of irregular plates similar to those of Asian dust.
 The TEM-EDS data of the clay grains were converted to cation numbers based on a 44-anion charge, and allocated to the tetrahedral and octahedral sites assuming a 2:1 structure. The two data sets of Asian dust and source soils superimposed very well on the two diagrams (Figure 11), indicating that they had the same origin. In the Altotal–Fe-Mg ternary diagram, most of the clay particles occurred in the narrow Al-rich region, where the proportion of Altotal varied between 60% and 100%, with a uniform Fe to Mg ratio of about 1:1. The diagram of interlayer K versus sum of octahedral cations (AlVI + Fe + Mg + Ti) shows that most particles were located between illite and smectite, with some particles located in the area enclosed by chlorite, smectite, and illite.
 In 93% of the analyzed grains, the K ion per 44 anion charges occurred between the value of 0 and 1, particularly around a value of 0.5. Thus most grains belonged to the interstratified illite-smectite. The relative percentage of illite and smectite components in the submicrometer clay particles could be calculated, because the K ions are mostly allocated to the interlayer of illite [Jeong et al., 2004]. The average number of K ions per 44 anion charges was 0.47 from EDS data collected from 162 clay grains. Therefore the proportion of the illite component was estimated to be 24%, while the remaining 76% was the proportion of the smectite component.
4.1. Mineralogical Features
 The single dust particles ranged from coarser particles composed of nearly a single mineral to mixtures of submicrometer minerals. Even particles resembling a single mineral had a very thin lining of clay minerals on either part of or on the entire grain. Thus the dust particles were mostly aggregates of minerals. The formation of clay aggregates was attributed to the pedogenic process in the source regions.
 The submicrometer minerals of the aggregate particles were mostly clay minerals mixed with nanosized calcite particles. The XRD results showed that illite and interstratified illite-smectite minerals were the major phyllosilicate minerals present, and these were also the major minerals of the bulk dust. The common presence of interstratified illite-smectite could have been overlooked previously, because only a weak peak from ethylene-glycolated smectite around 17 Å is usually used for the XRD quantification of smectite in mineral dusts [Biscaye et al., 1997], neglecting the diffuse background peaks of interstratified illite-smectite. In comparison to the XRD results, the TEM-EDS analysis of the submicrometer clay minerals indicates a higher dominance of interstratified illite-smectite. The TEM-EDS data of submicrometer clay grains cannot represent the clay mineral composition of a bulk sample, because illite comminuted from coarse micas or eroded from hydrothermal and low-grade metamorphic rocks tends to be relatively coarse, while interstratified illite-smectites that are generally formed by diagenesis of smectites or weathering of illite tend to be fine. Nevertheless, the combination of both the XRD and TEM-EDS data confirms that the interstratified illite-smectite was the major type of clay mineral in Asian dust, and that abundant smectite layers are present in the form of interstratification with illite layers, although the content of discrete smectite was very low.
 There was no large variation in the mineral composition of Asian dust. However, the total phyllosilicate and calcite contents varied from sample to sample. In the samples 300304 and 200405, the calcite content was lower than others in both Tables 4 and 5 in consistent with their low CaO contents (Table 3). This variation may be related to the different sources of Asian dust, ranging from desert to clay- and calcite-rich loess plateau. The abnormally high Na2O and K2O contents in two samples collected in 2003 suggest a significant contribution of noncrustal materials, such as sea salt from the Yellow Sea, pollutants, or desert evaporites. It should be noted that the peak PM10 concentration values of the 2003 samples (194–218 μg/m3) were much lower than those collected in the 2004 and 2005 samples (390–683 μg/m3). This implies that the dust samples collected in 2003 had a higher contribution from noncrustal particles. The origin of the subtle variation in dust mineral composition is still uncertain, because there are many unknown factors, such as differences in source region, chemical reactions with atmospheric gases, sea salt, and pollutants, different transport routes, and possible mineralogical differentiation of the particles during transportation.
 Currently, there is limited previous work on the bulk and single particle mineralogy of Asian dust. Lafon et al.  calculated the number abundance of the three assumed mineralogical types of dust particles using bulk-chemical data of a natural dust collected at Zhenbeitei area, China, and from dust generated from sandy soils in the Ulan Buh area in a wind tunnel, as follows: clay aggregates (66%), quartz (19%), and calcite (16%) in the natural dust, and clay aggregates (56%), quartz (18%), and calcite (26%) in the dust from the wind tunnel. The order is same as that measured in XRD quantification and direct counting in the SEM in this study (clay aggregates 66% > quartz 26% > calcite 7%, considering only three types). The clay content similar to that of Asian dust collected in Seoul could suggest the little variation of mineral composition over size fractions despite the decrease of particle size during the long-range transportation. However, in calculating the number abundance, Lafon et al.  assigned all the Al to clay, leading to an overestimation of clay aggregates, because a large quantity of the Al would be a structural component of plagioclase and K-feldspar. Similarly, the assignment of all the Ca to calcite resulted in the overestimation of calcite, because a portion of the Ca should have been assigned to plagioclase and amphibole. A proper estimation of iron oxide-clay aggregates should consider all aspects of the mineralogical characteristics of the dust.
 In comparison to the limited information on the mineralogy of the Asian dust, there have been several works on the bulk mineralogy of Saharan dust. However, the mineral composition of Saharan dust has been reported differently between investigators. Caquineau et al.  showed that the clay mineral composition of Saharan dust varies significantly, depending on the source region, from illite-rich dust to kaolinite-rich dust. Glaccum and Prospero  determined the bulk mineral composition at Sal Island, off the coast of Mauritania, as being illite = 53.8%, kaolinite = 6.6%, chlorite = 4.3%, quartz = 19.6%, plagioclase = 5.4%, K-feldspar = 2.2%, and calcite = 8.2%. Avila et al.  reported a significant content of smectite (11% on average), palygorskite (9%), and dolomite (5%) at the Montseny mountains near Barcelona, Spain, in addition to illite (39%), quartz (19%), feldspar (3%), calcite (9%), and kaolinite (6%). The distances to their dust sources from Sal Island and the Montseny mountains for the Saharan dust, and from Seoul for Asian dust are similar, ranging from 1000 to 2000 km. The ratio of minerals to quartz content showed that Asian dust has a higher feldspar content, a lower calcite content, and a lower total phyllosilicates content than Saharan dust, while kaolinite is enriched relative to chlorite in Saharan dust. The clay mineral composition of Asian dust is rather constant regardless of the trajectory, in contrast with Saharan dust, probably because the dust outbreak occurs in the relatively restricted regions of southern Mongolia and northern China. Illitic clay minerals always dominate the clay mineralogy of both the Saharan and Asian dust. Considering the different quantification methods adopted by the investigators, Saharan and Asian dusts are similar in their illite-dominating clay mineralogy, high total phyllosilicates and calcite contents.
4.2. Aeolian Processes Related to Sand, Silt, and Asian Dust
 Chemical and mineralogical analyses of sand, silt, and Asian dust provide data that has implications for the regional aeolian processes in the east Asia. Asian dust consists of fine-silt grains. Any regolith distributed in the arid to semiarid source regions could be a source of Asian dust. The sands in the deserts and silts on the loess plateau are known to be derived from the alluvial and fluvial deposits that originated in the glacial detritus or from the physical weathering products of high Asia [Smalley, 1995; Sun, 2002]. The parent materials supplied from the mountains surrounding the deserts were probably separated into desert sands and loess silts via aeolian processes under arid to semiarid climates dominated by westerly winds throughout the Quaternary Period. Although coarse silts deposited near the deserts to form the loess plateau, fine silts blown to high altitude could have been transported long distances. In addition, sand grains may have been further fragmented to form silt by aeolian abrasion, and have been continuously blown out of the desert to be added to the loess plateau [Smith et al., 2002], as shown in the silt-size particles of the desert sands (Figure 4a). The particle size distribution (Figure 2c) showed the presence of fine-silt particles in the sands, indicating that fine silty particles in the desert sands could be the source of Asian dust. The clay aggregates attached to the surface of sand grain could be detached by abrasion during the dust storm, and blown to high altitude. The source region of Asian dust includes the Chinese loess plateau, as well as the deserts of northwestern China and Mongolia [Chun et al., 2001]. Fine-silt particles could have been lifted from the loess soils and transported long distances, because the loess plateau near to the deserts is a partial source of Asian dust as well as being a depositional area of dust transported from the deserts. As shown in Figure 4, the silty soils included aggregates of the fine platy grains of clay minerals that were often mixed with nanosized calcite particles. It may be difficult to detach individual clay particles to form aerosol particles because of their cohesiveness, but the silt-size aggregates in the soils are able to be lifted to form aerosol particles. Therefore the particles of Asian dust are most likely to be formed by the erosion of the fine-silt fractions of the fluvial/alluvial deposits, the desert sands, and the reworking of the loess silts. On a regional scale, size separation of fluvial/alluvial parent materials into sand, loess silt, and Asian dust crossing the Korean peninsula could be regarded as an aeolian sorting eastward operated by the westerly wind system. The size separation of the parent materials results in their chemical and mineralogical fractionation (Figure 12). The most striking feature during the fractionation is the enhancement of the phyllosilicate content from the sand toward the Asian dust, and the lowering of the quartz, plagioclase, and K-feldspar contents.
4.3. Climatic and Atmospheric Implications
 The clay aggregates had abundant micropores, which provided a large surface area for chemical reactions to take place. Similarly, the particles of quartz and feldspar encrusted with clay minerals would be able to enhance the adsorption and reaction of atmospheric gases. The interstratified illite-smectite content was high in both the Asian dust and source soils, despite the very low content of pure smectite. Smectite is a highly hydrous clay mineral that adsorbs water, organic matter, and exchangeable cations into its interlayers and outer surfaces, and it mediates catalytic reactions. The hygroscopicity and other reactive characteristics of smectites in the form of interstratified illite-smectite warrant further investigation. Clay minerals in the dust particles are promising candidates for ice nuclei, influencing cloud properties and radiative forcing [Sassen, 2002].
 The optical properties of mineral dust are highly influenced by the mineral composition and aggregation state of the dust particles. Clay-rich mineral dust absorbs infrared radiation much more than quartz-rich dust does [Sokolik, 2002]. The output data of radiative forcing calculations would be more accurate using the most precise mineral composition data. The direct radiative forcing of Asian dust was recently estimated by Park et al. . The mineral composition data used in their model was similar to that of the sands observed in this study, which is significantly different from that of Asian dust. Since the dust particles are mostly internal mixtures of diverse minerals having a range of refractive indices, their single scattering albedo would be changed from that of the external mixtures. More realistic climatic models of Asian dust would be derived by inputting more reliable mineralogical data into the calculations of the optical properties.
5. Summary and Conclusions
 Since Asian dust is involved in the Earth's radiation budget, acid neutralization, and transport of chemical species, its mineralogy should be known in detail to construct a realistic climate model. In this study, the mineralogical properties of single dust particles and bulk dust were investigated from microscopic characterization and quantitative XRD and EDS analysis. Asian dust is mineralogically and chemically more similar to the silty soils of the Chinese loess plateau than to desert sands. The total phyllosilicate content of Asian dust (45% based on XRD method) is higher than that of the desert sands (7%) and loess silts (23%) in the source regions. Interstratified illite-smectite is an important clay mineral, in addition to illite. Fine illuvial clay minerals and secondary nanosized calcite particles of pedogenic origin were aggregated, and had coated coarser grains in the source soils. As a result, Asian dust consists of aggregates of minerals with different reactivity, size, shape, refraction index, and chemical composition. The most common single aerosol particles were clay aggregates that were often mixed with secondary nanosized calcite particles followed by quartz, plagioclase, coarser primary calcite, K-feldspar, chlorite, muscovite, gypsum, kaolinite, biotite, and amphibole in decreasing order of abundance, all of which were partly or entirely covered with clay minerals. The mineral and chemical compositions of Asian dust were similar, but there was a variation in the clay and calcite contents, which cannot be explained by any single factor. The abundant clay minerals and their aggregates that were attached to the coarser minerals, and their associated nanosized calcite, may promote chemical reactions between Asian dust and atmospheric gases and clouds, and may also affect its optical properties.
 The author gratefully thanks Y. Chun and the staff of the National Institute of Meteorological Research for providing aerosol size distribution data and soil and dust samples for this study. Kideok Kwon and Sae Jung Chang kindly helped this work by supplying many rare references. The author acknowledges the NOAA Air Resources Laboratory (ARL) for the provision of the HYSPLIT transport and dispersion model and READY website (http://www.arl.noaa.gov/ready.html) used in this publication. The manuscript was greatly improved by the comments of three anonymous reviewers. This work was supported by the Korea Polar Research Institute grant KOPRI (PE08010).